Datasets:

Dhruvil47

/

bioarxiv_papers

like 0

The ~1,000-fold higher affinities of the EO than the closed conformations for both α4β1 and α5β1 integrins are achieved by the ~25,000-fold slower off-rate of the EO conformation (Fig. 7C). Similar to the differences in on-rates, the differences in off-rates can be understood in terms of the structural details in the ligand-binding pocket and the much higher affinity of the EO state. The tighter Asp sidechain binding pocket and greater burial of the Asp provide a barrier to dissociation (Fig. 7D). The number of hydrogen bonds of the Asp sidechain to the β1-α1 loop backbone increases in the open state (Zhu et al., 2013). Furthermore, the greater burial of these polar bonds and increased network of hydrogen bonds around them increases their strength. The Arg sidechain also strengthens its hydrogen bonding in the open state. During opening, as the βI domain β1-α1 loop moves toward the Asp, the entire RGD moiety slides toward the α- subunit, which is to the left in the view of Fig. 7D. This movement is seen in Fig. 7D as the closer approach of the Asp to the α2-α3 loop in the open state. A hydrogen bond network with two waters between the RGD Arg sidechain and α5β1 residue Gln-221 in the closed state is exchanged for a direct Arg hydrogen bond to α5β1 Gln-221 as RGD slides toward α5β1 during opening (Xia & Springer, 2014). All these can contribute to the higher affinity and tens of thousands-fold slower off-rate of the EO conformation than the closed conformations. Our kinetic measurements were carried out at 22°C. At 37°C, both on- and off-rates will be higher. Increase in temperature generally has a much greater effect on dissociation rates than association rates (Johnstone et al, 1990). The amount of increase depends on the activation energy; i.e. the height of the energy barrier to dissociation. The intrinsic ligand-binding kinetics of integrin conformational states described here are consistent with previous kinetic observations. These studies showed that activating integrin ensembles with Mn2+ or activating IgG or Fab, using conditions that in retrospect would partially, but not completely, shift integrin ensembles to the EO state, decreased the ligand off-rates of integrins α4β1 and α5β1 (Chigaev et al, 2001; Takagi et al., 2003). The extremely long lifetime of the α5β1 complex with fibronectin in the EO state, around several hours, explains why the α5β1 complex with fibronectin in Mn2+ was much more rapidly reversed by mAb 13 IgG specific for the closed conformations than by competitive inhibitor (Mould et al, 2016; Mould et al., 2014). Integrin activation. A major impetus for these studies was to determine the pathway for activation of integrins in cells, i.e. the activation trajectory. Of key importance is how integrins on the cell surface first engage ligands. By dynamically linking the actin cytoskeleton to the extracellular environment, integrins transduce both external and internal mechanochemical cues and bi-directionally signal across the plasma membrane.

Integrin signaling is governed by 7 bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . cytoskeletal force and the force stabilized, high-affinity, extended-open conformation is the only state competent to mediate cell adhesion (Alon & Dustin, 2007; Astrof et al., 2006; Li & Springer, 2018; Li et al., 2017; Nordenfelt et al., 2016; Nordenfelt et al., 2017; Sun et al., 2019; Zhu et al., 2008). Mechanotransduction occurs when integrins bind to ligand anchored in the extracellular environment, the cytoplasmic domain simultaneously binds to a cytoskeletal adaptor and links to actin retrograde flow, and a tensile force is transmitted through the integrin that stabilizes the extended-open conformation over the bent-closed conformation. Thus the on- and off- rates of ligand binding to integrins are among the key parameters that determine the cytoskeletal force regulation efficiency. We found that the closed states, with loose ligand binding pockets, have higher on-rates for ligand binding, making them the most efficient state for encountering ligand. Because the BC state is >200-fold more populated than the EC state for both integrins α4β1 and α5β1 on the cell surface (Li & Springer, 2018; Li et al., 2017) (Fig. 7B), the BC state may have an important role in initial binding to ligand. There are few constraints on the orientation of integrins on cell surfaces until transmitted force orients them when they bridge extracellular ligands and the actin cytoskeleton (Nordenfelt et al., 2017; Swaminathan et al, 2017). In the absence of such engagement, linkers of largely disordered residues between the last module of integrin α- and β-subunit ectodomains and the beginning of their transmembrane α-helices allows large tilting motions of the ectodomain relative to the plasma membrane (Zhu et al, 2009). Thus, the common depiction of the leg domains of integrins and other receptors as oriented normal to the cell membrane (Fig. 1A) is only a conventional cartoon representation and has no experimental basis. Structures of integrin αIIBβ3 linker and transmembrane domains on cell surfaces, combined with the ectodomain, showed that large movements of the BC state relative to the membrane normal were possible. Nonetheless, none of these orientations have a ligand binding site with an orientation optimal for binding a ligand on the surface of another cell or in the extracellular matrix. In contrast, the ligand binding site is better exposed in the EC state (Fig. 1A) and the EC state also can bend at multiple domain-domain junctions and is less constrained in orientation relative to the plasma membrane. It is possible that either the BC state is the predominant ligand-binding state, or that the BC state provides a large reserve of integrins that, through conformational sampling of the EC state, allows the EC state to be the predominant ligand-binding state.

Once ligand is bound to the BC or EC state, the ~1000-fold higher ligand-binding affinity for the EO conformation strongly favors conformational change to the EO state (Li & Springer, 2018; Li et al., 2017). If an adaptor and the actin cytoskeleton are bound at the time when a ligand that is embedded in the extracellular environment is bound to the integrin, the ligand resists the force from actin retrograde flow, and tensile force is transmitted through the integrin, strongly stabilizing the ECL and EOL conformations (Li & Springer, 2017). Furthermore, the EOL state is ~2,000 and ~10,000 more populated than the ECL state for integrins α4β1 and α5β1, respectively (Fig. 7B). The off-rates of the EO states of α4β1 and α5β1 equate to lifetimes of about 0.4 hours and 3 hours, respectively, and serve to make integrin-ligand bonds highly resistant to detachment. In contrast, when engagement to the adaptor/actin cytoskeleton is reversed, BCL would become substantially populated in the basal ligand-bound conformations (Fig. 7B), and with a lifetime in millisecond to second range, would allow the integrin to dissociate from ligand. In summary, we have substantially advanced our understanding of how integrins on intact cells bind ligands by measuring the ligand binding and dissociation kinetics for the three conformational states of two integrins, α4β1 and α5β1. While it may seem surprising that the low affinity states bind more rapidly than the high affinity states, our findings concord with previous studies on selectins (Phan et al., 2006) and bacterial fimbriae adhesins (Yakovenko, 2015) that have two states, one flexed (bent) and the other extended, that are also subjected to regulation by force, in which the extended state is the higher affinity state. As there is no structural homology between the three classes of adhesins, convergent evolution appears to have selected a low affinity, flexed/bent state for rapid ligand binding that can be subsequently 8 bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . stabilized by force to a high affinity, extended state that can then better resist the tendency of force to accelerate receptor-ligand dissociation. In the two-state systems, force is applied externally by shear flow, while in the three-state integrin system, force is applied internally by engagement of actin retrograde flow. This empowers the actin cytoskeleton machinery to regulate integrin function, ensuring intimate coordination between the needs of adhering and migrating cells because the same signaling pathways that regulate actin polymerization and disassembly also regulate formation of cellular attachments through integrins to the extracellular environment.

While the kinetics of integrin conformational change remain to be measured, the excellent fit of kinetic measurements to the 1 vs. 1 Langmuir model found here and agreement between koff/konvalues and Kd measured at equilibrium suggest that integrin conformational change kinetics are also rapid. Rapid ligand binding, together with rapid cytoskeletal adaptor binding, would enable their coincidence to regulate integrin activation, thus providing a seamless method for activating integrins at cellular locations where actin is activated and at extracellular locations where ligand is available. MATERIALS AND METHODS Fabs. IgGs, 8E3 (Mould et al, 2005), 9EG7 (Bazzoni et al, 1995), 12G10 (Mould et al, 1995), HUTS4 (Luque et al, 1996), mAb13 (Akiyama et al, 1989) and SNAKA51 (Clark et al, 2005) were produced from hybridomas and purified by protein G affinity; Fabs were prepared with papain digestion in PBS (phosphate-buffered saline with 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4 and 1.8 mM KH2PO4, pH7.4) with 10 mM EDTA and 10 mM cysteine and papain: IgG mass ratio of 1:500 for 8 hrs at 37°C, followed by Hi-Trap Q chromatography in Tris-HCl pH 9 with a gradient in the same buffer to 0.5 M NaCl. Integrin α5β1 soluble preparations. Integrin α5β1 ectodomain (α5 F1 to Y954 and β1 Q1 to D708) with secretion peptide, purification tags, and C-terminal clasp (Takagi et al, 2001) were produced by co-transfecting the pcDNA3.1/Hygro(-) vector coding the α-subunit and pIRES vector coding the β-subunit into HEK 293S GnTI−/− (N-acetylglucosaminyl transferase I deficient) cells. Stable transfectants were selected with hygromycin (100 μg/ml) and G418 (1 mg/ml), and proteins were purified from culture supernatants by His tag affinity chromatography and Superdex S200 gel filtration after cleavage of C-terminal clasp and purification tags with Tev protease (Li et al., 2017). Peptidomimetic and macromolecule fragments. FITC-conjugated α4β1 specific probe, 4-((N′-2-methylphenyl)ureido)-phenylacetyl-L-leucyl-L-aspartyl-L-valyl-L-prolyl-L-alanyl-L- alanyl-L-lysine (FITC-LDVP) and its unlabeled version, LDVP, were from Tocris Bioscience (Avonmouth, Bristol, United Kingdom). Human VCAM D1D2 (mature residues F1 to T202) were expressed and purified from HEK 293S GnTI−/− cell line supernatants by affinity chromatography and gel filtration (Yu et al, 2013). VCAM D1D2 was fluorescently labeled with Alexa Fluor 488 NHS Ester (ThermoFisher Scientific). Human Fn39-10 S1417C mutant (mature residues G1326 to T1509) and its synergy and RGD sites (R1374A&P1376A&R1379A&S1417C&Δ1493-1496) mutated inactive version were expressed in E. coli and purified as described (Li et al., 2017; Takagi et al., 2001). Fn39-10 S1417C mutant was fluorescently labeled with Alexa Fluor 488 C5 maleimide (ThermoFisher Scientific) at residue Cys-1417. Both Fn39-10 S1417C mutant and its inactive version were biotinylated with Maleimide-PEG11-Biotin at residue 1417 (ThermoFisher Scientific) in PBS. Quantitative fluorescent flow cytometry.

Jurkat and K562 cells (106 cells/mL in RPMI- 1640 medium, 10% FBS) were washed twice with assay medium (Leibovitz’s L-15 medium, 10 mg/mL BSA) containing 5 mM EDTA, twice with assay medium alone, and suspended in assay medium. Cells at 2×106 cells/mL were incubated with indicated concentration of Fabs for 30min at 22°C. Addition of FITC-LDVP, Alexa488- VCAM D1D2 (1.6 labeling ratio) or Alexa488-Fn39-10 (1.0 labeling ratio) at indicated concentrations initiated association. Association was measured 9 bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . as mean fluorescence intensity (MFI) at successive time points after addition of the fluorescent ligands. Addition of 500-fold higher concentration of the unlabeled ligand at the end of the association phase initiated the dissociation phase. Background MFI for FITC-LDVP, Alexa488- VCAM D1D2 and Alexa488-Fn39-10 in presence of 10 mM EDTA was subtracted (Supplemental Fig. S1). Fitting flow cytometry and BLI kinetic binding traces with 1 vs. 1 Langmuir binding model Kinetic traces including both the association phase and the dissociation phase at different analyte concentrations were globally fitted to the following function. Rt=( +( 1 2 1 2 (tD-t) 2|tD-t| (tD-t) 2|tD-t| + ) ) Rmaxkon[A] koff+kon[A] (1-e-(koff + kon[A]) t) Rmaxkon[A] koff+kon[A] (1-e-(koff + kon[A]) tD) e-koff (t - tD), where t is time, Rt is response at time t, tD is the time that dissociation starts, [A] is the analyte concentration, and Rmax is the maximum response. The first term fits the data in the association phase and the second term fits the data in the dissociation phase. The prefactor of the first term is 1 prior to tD and becomes 0 after tD; whereas the prefactor for the second term is 0 prior to tD and becomes 1 after tD. Nonlinear least square fit of Rt, [A], and t to the above equation yields the on-rate, kon, off-rate, koff, and Rmax. Bio-Layer Interferometry (BLI). Binding kinetics of unclasped high-mannose α5β1 ectodomain and Fn39-10 was measured by BLI (Wallner et al., 2013) with streptavidin biosensors on an Octet RED384 System. The reaction was measured on 96 well plate (200 uL/well) in buffer with 20 mM Tris HCl (pH 7.4), 150mM NaCl, 1mM Ca2+, 1mM Mg2+ and 0.02% Tween20. Streptavidin biosensors were hydrated in reaction buffer for 10 min before starting the measurements. Each biosensor was sequentially moved through 5 wells with different components: (1) buffer for 3 minutes in baseline equilibration step; (2) 35 nM biotin-Fn39-10 for 1 minute for immobilization of ligand onto the biosensor; (3) indicated concentrations of Fabs for 5 minutes for another baseline equilibration; (4) indicated concentrations of α5β1 ectodomain and Fabs for the association phase measurement; (5) indicated concentrations of Fabs for the dissociation phase measurement.

Each biosensor has a corresponding reference sensor that went through the same 5 steps, except in step 2 the ligand was replaced with 35 nM inactive version of Fn39-10 with both the RGD binding site and the synergy site (PHSRN) mutated. Background subtracted response in both the association and dissociation phases, and at different α5β1 ectodomain concentrations, were globally fit to the 1 vs. 1 Langmuir binding model, with kon biosensor as individual fitting parameter. The equilibrium binding (response) was calculated from kon response curve to calculate Kd values as a check on koff response (Req), fitted kon, koff, and Rmax values at each α5β1 ectodomain concentration [A] were used to calculate Req at a time 1,000-folder longer than the "binding time", i.e., t = 1000* app as shared fitting parameters and maximum response (Rmax) for each app and koff app values at each each α5β1 ectodomain concentration and fit to a dose app and koff app app/k values. To calculate equilibrium on Req= 1 kon[A] Rmaxkon[A] koff+kon[A] , with the following equation: (1-e-(koff + kon[A]) t) Calculating ligand-binding and dissociation rates for the BC and EC states app) for each defined ensemble containing 2 or 3 app and koff The measured on- and off- rates (kon states shown in Figs. 2-4 was approximated by the on- and off- rates of each state weighted by its population in the ensemble (Fig. 7A, Eqs. 1-4). At steady state, the population of the free integrin states and the ligand-bound integrin states were calculated based on the previously determined population and intrinsic ligand-binding affinity of each state (Fig. S3B, Eqs. S5-S10) 10 bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . in the respective integrin α4β1 and α5β1 preparations (Li & Springer, 2018; Li et al., 2017) (Fig. 7B). Specifically, EO Ka EC and Ka EO Ka BC (in Fig.S3B, Eqs.S8-S10), are the intrinsic ligand-binding affinity Ka ratios of the EO state and the closed states. For integrin α4β1, the ratios were averaged to 745±237 from six α4β1 preparations, including α4β1 headpiece with high-mannose N-glycans, α4β1 ectodomain with high-mannose N-glycans, α4β1 ectodomain with complex N-glycans, and intact α4β1 on three different cell lines (Li & Springer, 2018); for integrin α5β1, the intrinsic ligand-binding affinity ratio of the EO state and the closed states were averaged to 3106 ±1689 from eight soluble α5β1 preparations that varied in presence or absence of the lower legs, of a loose clasp in place of the TM domain, and in whether the N-linked glycan was complex, high mannose, or shaved (Li et al., 2017). Using kon 2-4, kon ensembles, respectively (Fig. 7A, Eqs.1-2). By including the values for kon app (BC+EC+EO) and koff to kon measured in basal ensembles, respectively (Fig.

7A, Eqs. 3-4). EO and koff app(EC+EO) and koff EO rates experimentally measured in Figs. app(EC+EO) measured in extended EC were derived from the kon EC and koff EC in addition EC and koff app (BC+EC+EO) BC were then derived from kon BC and koff EO and koff EO, kon Acknowledgements. We thank Kelly L. Arnett in Center for Macromolecular Interactions of Harvard Medical school for training and consultation on BLI measurement. We thank Taekjip Ha for suggestions on our manuscript. This work was funded by NIH R01 HL131729 (“Activation trajectories of integrin a5b1”). Author Contributions. J.L. and T.A.S. designed research and wrote the paper. J.L. carried out the measurements and analyzed the data. J.B.Y. prepared the Fabs and purified soluble integrins and ligands. Conflict of Interest Statement. The authors declare no competing financial interests. 11 bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . References Akiyama SK, Yamada SS, Chen WT, Yamada KM (1989) Analysis of fibronectin receptor function with monoclonal antibodies: Roles in cell adhesion, migration, matrix assembly, and cytoskeletal organization. J Cell Biol 109: 863-875 Alon R, Dustin ML (2007) Force as a Facilitator of Integrin Conformational Changes during Leukocyte Arrest on Blood Vessels and Antigen-Presenting Cells. Immunity 26: 17-27 Alsallaq R, Zhou HX (2008) Electrostatic rate enhancement and transient complex of protein- protein association. Proteins 71: 320-335 Astrof NS, Salas A, Shimaoka M, Chen JF, Springer TA (2006) Importance of force linkage in mechanochemistry of adhesion receptors. Biochemistry 45: 15020-15028 Bazzoni G, Shih D-T, Buck CA, Hemler MA (1995) Monoclonal antibody 9EG7 defines a novel β1 integrin epitope induced by soluble ligand and manganese, but inhibited by calcium. J Biol Chem 270: 25570-25577 Bouvard D, Pouwels J, De Franceschi N, Ivaska J (2013) Integrin inactivators: balancing cellular functions in vitro and in vivo. Nat Rev Mol Cell Biol 14: 430-442 Chigaev A, Blenc AM, Braaten JV, Kumaraswamy N, Kepley CL, Andrews RP, Oliver JM, Edwards BS, Prossnitz ER, Larson RS et al (2001) Real-time analysis of the affinity regulation of α4-integrin: the physiologically activated receptor is intermediate in affinity between resting and Mn2+ or antibody activation. J Biol Chem 276: 48670-48678 Clark K, Pankov R, Travis MA, Askari JA, Mould AP, Craig SE, Newham P, Yamada KM, Humphries MJ (2005) A specific α5β1-integrin conformation promotes directional integrin translocation and fibronectin matrix formation. J Cell Sci 118: 291-300 Dong X, Hudson NE, Lu C, Springer TA (2014) Structural determinants of integrin β-subunit specificity for latent TGF-β. Nat Struct Mol Biol 21: 1091-1096 Dong X, Zhao B, Iacob RE, Zhu J, Koksal AC, Lu C, Engen JR, Springer TA (2017) Force interacts with macromolecular structure in activation of TGF-β.

Nature 542: 55-59 Dong X, Zhao B, Lin FY, Lu C, Rogers BN, Springer TA (2018) High integrin αVβ6 affinity reached by hybrid domain deletion slows ligand-binding on-rate. Proc Natl Acad Sci U S A 115: E1429-E1436 Iwamoto DV, Calderwood DA (2015) Regulation of integrin-mediated adhesions. Curr Opin Cell Biol 36: 41-47 Johnstone RW, Andrew SM, Hogarth MP, Pietersz GA, McKenzie IF (1990) The effect of temperature on the binding kinetics and equilibrium constants of monoclonal antibodies to cell surface antigens. Mol Immunol 27: 327-333 Kim C, Ye F, Ginsberg MH (2011) Regulation of integrin activation. Annu Rev Cell Dev Biol 27: 321-345 Kokkoli E, Ochsenhirt SE, Tirrell M (2004) Collective and single-molecule interactions of α5β1 integrins. Langmuir 20: 2397-2404 12 bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . Legate KR, Fassler R (2009) Mechanisms that regulate adaptor binding to β-integrin cytoplasmic tails. J Cell Sci 122: 187-198 Li J, Springer TA (2017) Integrin extension enables ultrasensitive regulation by cytoskeletal force. Proc Natl Acad Sci U S A 114: 4685-4690 Li J, Springer TA (2018) Energy landscape differences among integrins establish the framework for understanding activation. J Cell Biol 217: 397-412 Li J, Su Y, Xia W, Qin Y, Humphries MJ, Vestweber D, Cabanas C, Lu C, Springer TA (2017) Conformational equilibria and intrinsic affinities define integrin activation. EMBO J 36: 629-645 Luo BH, Carman CV, Springer TA (2007) Structural basis of integrin regulation and signaling. Annu Rev Immunol 25: 619-647 Luque A, Gomez M, Puzon W, Takada Y, Sanchez-Madrid F, Cabanas C (1996) Activated conformations of very late activation integrins detected by a group of antibodies (HUTS) specific for a novel regulatory region (355-425) of the common β1 chain. J Biol Chem 271: 11067-11075 Mould AP, Askari JA, Byron A, Takada Y, Jowitt TA, Humphries MJ (2016) Ligand-induced Epitope Masking: Dissociation of integrin α5β1-fibronectin complexes only by monoclonal antibodies with an allosteric mode of action. J Biol Chem 291: 20993-21007 Mould AP, Craig SE, Byron SK, Humphries MJ, Jowitt TA (2014) Disruption of integrin- fibronectin complexes by allosteric but not ligand-mimetic inhibitors. Biochem J 464: 301-313 Mould AP, Garratt AN, Askari JA, Akiyama SK, Humphries MJ (1995) Identification of a novel anti-integrin monoclonal antibody that recognises a ligand-induced binding site epitope on the β1 subunit. FEBS Lett 363: 118-122 Mould AP, Travis MA, Barton SJ, Hamilton JA, Askari JA, Craig SE, Macdonald PR, Kammerer RA, Buckley PA, Humphries MJ (2005) Evidence that monoclonal antibodies directed against the integrin β subunit plexin/semaphorin/integrin domain stimulate function by inducing receptor extension.

J Biol Chem 280: 4238-4246 Nagae M, Re S, Mihara E, Nogi T, Sugita Y, Takagi J (2012) Crystal structure of α5β1 integrin ectodomain: Atomic details of the fibronectin receptor. J Cell Biol 197: 131-140 Nordenfelt P, Elliott HL, Springer TA (2016) Coordinated integrin activation by actin-dependent force during T-cell migration. Nat Commun 7: 13119 Nordenfelt P, Moore TI, Mehta SB, Kalappurakkal JM, Swaminathan V, Koga N, Lambert TJ, Baker D, Waters JC, Oldenbourg R et al (2017) Direction of actin flow dictates integrin LFA-1 orientation during leukocyte migration. Nat Commun 8: 2047 Park YK, Goda Y (2016) Integrins in synapse regulation. Nat Rev Neurosci 17: 745-756 Phan UT, Waldron TT, Springer TA (2006) Remodeling of the lectin/EGF-like interface in P- and L-selectin increases adhesiveness and shear resistance under hydrodynamic force. Nat Immunol 7: 883-889 Schürpf T, Springer TA (2011) Regulation of integrin affinity on cell surfaces. EMBO J 30: 4712- 4727 13 bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . Su Y, Xia W, Li J, Walz T, Humphries MJ, Vestweber D, Cabañas C, Lu C, Springer TA (2016) Relating conformation to function in integrin α5β1. Proc Natl Acad Sci U S A 113: E3872-3881 Sun Z, Costell M, Fassler R (2019) Integrin activation by talin, kindlin and mechanical forces. Nat Cell Biol 21: 25-31 Sun Z, Guo SS, Fassler R (2016) Integrin-mediated mechanotransduction. J Cell Biol 215: 445- 456 Swaminathan V, Kalappurakkal JM, Mehta SB, Nordenfelt P, Moore TI, Koga N, Baker DA, Oldenbourg R, Tani T, Mayor S et al (2017) Actin retrograde flow actively aligns and orients ligand-engaged integrins in focal adhesions. Proc Natl Acad Sci U S A 114: 10648-10653 Takagi J, Erickson HP, Springer TA (2001) C-terminal opening mimics "inside-out" activation of integrin α5β1. Nat Struct Biol 8: 412-416 Takagi J, Strokovich K, Springer TA, Walz T (2003) Structure of integrin α5β1 in complex with fibronectin. EMBO J 22: 4607-4615 Wallner J, Lhota G, Jeschek D, Mader A, Vorauer-Uhl K (2013) Application of Bio-Layer Interferometry for the analysis of protein/liposome interactions. J Pharm Biomed Anal 72: 150- 154 Wei H, Mo J, Tao L, Russell RJ, Tymiak AA, Chen G, Iacob RE, Engen JR (2014) Hydrogen/deuterium exchange mass spectrometry for probing higher order structure of protein therapeutics: methodology and applications. Drug Discov Today 19: 95-102 Xia W, Springer TA (2014) Metal ion and ligand binding of integrin α5β1. Proc Natl Acad Sci U S A 111: 17863-17868 Xiao T, Takagi J, Wang J-H, Coller BS, Springer TA (2004) Structural basis for allostery in integrins and binding of fibrinogen-mimetic therapeutics. Nature 432: 59-67 Yakovenko OT, V.; Sokurenko, E.V. ; Thomas, W.E.. (2015) Inactive conformation enhances binding function in physiological conditions.

Proc Natl Acad Sci USA 112: 9884-9889 Yu Y, Schurpf T, Springer TA (2013) How natalizumab binds and antagonizes α4 integrins. J Biol Chem 288: 32314-32325 Zhu J, Luo BH, Barth P, Schonbrun J, Baker D, Springer TA (2009) The structure of a receptor with two associating transmembrane domains on the cell surface: integrin αIIbβ3. Mol Cell 34: 234-249 Zhu J, Luo BH, Xiao T, Zhang C, Nishida N, Springer TA (2008) Structure of a complete integrin ectodomain in a physiologic resting state and activation and deactivation by applied forces. Mol Cell 32: 849-861 Zhu J, Zhu J, Springer TA (2013) Complete integrin headpiece opening in eight steps. J Cell Biol 201: 1053-1068 14 bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . Figure Legends Figure 1. Ligand-interaction kinetics of integrin ensembles. (A) Three overall integrin conformational states (Luo et al, 2007). Individual domains are labeled next to the extended- open state. The structural motifs that move during opening (α1-helix, α7-helix and β6-α7 loop) are labeled in the βI domain of the EC and EO state. F represents tensile force exerted across ligand–integrin–adaptor complexes by the cytoskeleton and resisted by immobilized ligand. (B) Reaction scheme showing the apparent 1 vs. 1 kinetics of integrin and ligand binding (left), and the scheme required to correctly calculate ligand binding kinetics that takes into account the kinetics of conformational change (right). (C) Fabs utilized in this study, the integrin domains they bind, and their conformational specificity. Figure 2. Binding kinetics of ligands to intact α4β1 on Jurkat cells. (A-E) Binding and dissociation of FITC-LDVP (A-C) and Alexa488-VCAM D1D2 (D-E) to α4β1 on Jurkat cells measured by flow cytometry. Cartoons in panel A, B, and C show the schemes for measuring ligand binding and dissociation in the association phase and dissociation phase in basal ensemble (A), extended ensembles (EC+EO states) stabilized with Fab 9EG7 (4 μM) (B and D), and open ensemble (EO state) stabilized with Fabs 9EG7 (4 μM) and HUTS4 (2 μM) (C and E), respectively. Specific MFI with the MFI in EDTA (Fig. S1) subtracted is shown as open (association) or filled (dissociation) symbols; fits are shown as thin lines as indicated in keys. (F) app are from global fits of data at all ligand concentrations. Tabulation of results. Kon also shown and compared to previous equilibrium Kd measurements (Li & Springer, 2018), except for Alexa488-VCAM D1D2 binding to extended states, which was measured here (Fig. app values are s.e. from global nonlinear least square fitting; errors for S2). Errors for kon app koff app are propagated from errors of kon kon independent experiments. app and koff app and koff app app and koff ; errors for Kd values are s.d.

from three app Koff app is kon Figure 3. Binding kinetics of Alexa488-Fn39-10 to α5β1 on K562 cells. (A-B) Binding of Alexa488-Fn39-10 to α5β1 on K562 cells measured by flow cytometry. Measurements were on integrins in extended ensembles (EC+EO states) in presence of Fabs 9EG7 (6 μM) and SNAKA51 (2 μM) (A) or in the open (EO state) in presence of Fabs 9EG7 (6 μM) and HUTS4 (2 μM) (B), as illustrated in the cartoons. MFI with background in EDTA subtracted (Fig. S1) is shown as symbols and fits are shown as lines as explained in keys; the association phase has open symbols and solid lines, and the dissociation phase has filled symbols and dashed lines. app are from global fits and errors are from non-linear least (C) Tabulation of results. Kon app/kon app is also shown with propagated error and compared to previous equilibrium square fits. koff Kd measurements (Li et al., 2017). app and koff Figure 4. Binding kinetics of α5β1 ectodomain to Fn39-10. (A-D) Binding of unclasped high- mannose α5β1 ectodomain measured with BLI. Schemes for measuring ligand binding and dissociation in the association phase and dissociation phase are shown in each panels’ cartoon. α5β1 ectodomain (analyte) at the indicated concentrations in nM was bound to biotin- Fn39-10 immobilized on streptavidin biosensors without Fab (A) or with 2 μM Fab 8E3 (B), or with 2 μM 9EG7 and 5 μM HUTS4 Fabs, (C) or with 1 μM Fab 12G10 (D). Arrows mark the start of the dissociation phase. Response curves are in gray and fitting curves in black. (E) The equilibrium app and koff binding (response) was calculated from kon concentration and fit to a dose response curve to calculate Kd values. These values serve as a check on the koff analysis in Panel E, kon with 1 vs. 1 Langmuir binding model, and koff fitting errors from nonlinear least square fits; errors for koff app and koff kon app values at each α5β1 ectodomain app values in F. (F) Tabulation of Kd values from equilibrium response app/kon app and koff app values from nonlinear least square fit of data in Panel A-D app and Kd are app. Errors without * for kon app/kon app, koff app are propagated from errors of app/kon app. Errors with * are difference from the mean of two independent measurements. 15 bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . Figure 5. Dissociation of FITC-LDVP from α4β1 on Jurkat cells in presence of closure- stabilizing Fab. (A-B) FITC-LDVP dissociation from basal or extended ensembles of intact α4β1 on Jurkat cells measured using flow cytometry. FITC-LDVP (20nM) was incubated with Jurkat cells in absence (A) or in presence of extension-stabilizing Fab 9EG7 (4 μM) (B) for 10 minutes to reach steady state. Then, 10 μM unlabeled LDVP together with indicated concentrations of mAb13 Fab were added.

Observed MFI (MFIobs) values as a function of time at indicated mAb13 Fab concentrations were globally fitted to MFIobs=MFI0*e-koff with MFI at the start of dissociation (MFI0) and background MFI (MFIbackground) as shared parameters and koff Dependence of koff panels A and B were fitted to dose response curves to determine the maximum off-rate at saturating mAb13 Fab concentration, koff reaches half of the maximum, EC50 Figure 6. Dissociation of α5β1 ectodomain from biotin- Fn39-10 in presence of closure- stabilizing Fab. (A-B) Unclasped high-mannose α5β1 ectodomain dissociation from biotin-Fn39- 10 immobilized on streptavidin biosensors was monitored by BLI. Reaction schemes are illustrated in each panel’s cartoons. Specifically, 50 nM α5β1 ectodomain was incubated with biotin-Fn39-10 biosensors for 10 minutes to reach steady state binding in absence (A) or presence of 2μM 9EG7 Fab (B). Biosensors were then transferred into wells lacking the α5β1 ectodomain in presence or absence of 9EG7 Fab as before and also containing the indicated concentrations of mAb13 Fab for measurement of dissociation. The observed response (Robs) at each mAb13 Fab concentration as a function of time was individually fitted to the single exponential, Robs=R0*e-koff max at saturating mAb13 Fab concentration. koff (C) Determination of koff concentration using a dose response curve for the maximum off-rate at saturating mAb13 Fab max. The mAb13 Fab concentration when the off-rate reaches half concentration to determine koff of the maximum, EC50 Figure 7: Ligand-binding kinetics of each integrin state. (A) Defined integrin α4β1 and α5β1 ensembles utilized in this study to measure ligand-interaction kinetics, as well as equations to relate the apparent on- and off- rates with the on- and off-rates for each conformational state. (B) Conformational state populations (%) in absence and presence of ligand at steady state. Previously reported populations for integrins in the absence of ligand and their affinities for ligand (Li & Springer, 2018; Li et al., 2017) were used with Eqs.S5-S10 in Fig. S3B to calculate the populations in saturating ligand of ligand-bound integrin states in each type of ensemble studied here. (C) Values of kon and koff for conformational states of four integrin-ligand pairs. As discussed in the text and Methods, kinetic measurements on the EO state and the extended and basal ensembles were used with equations in panel A to calculate kinetics of the BC and EC states. The errors for directly measured values were fitting errors from non-linear least square fit; the errors for calculated BC and EC values were propagated. a: Intrinsic rates of EO state was from measurements in presence of HUTS4 & 9EG7 Fabs in Figs. 2-4, and intrinsic rates for BC and EC states were calculated with Eqs. 1-4 in panel A. b: From equilibrium measurements as specified in Fig.2 to Fig.4 legends. c: Calculated from the product of equilibrium Kd and kon. (D) Comparison of Asp-binding pocket in the open state (PDB: 3ze2 chains C+D) and closed state (PDB: 3zdy chains C+D) of integrin αIIBβ3 (Zhu et al., 2013).

The pocket in the β3 βI domain is shown with backbone and nearby sidechains in blue stick and blue dot surfaces and the MIDAS Mg2+ ion as a silver sphere. The ligand Asp sidechain and its backbone loop are shown in yellow, red sidechain carboxyl oxygens. The Asp sidechain Cβ carbon and carboxyl oxygens are shown as yellow and red dot surfaces, respectively. app t+MFIbackground, app as the individual fitting parameter at each mAb13 Fab concentration. (C) app on mAb13 Fab concentration. koff app at each mAb13 Fab concentration in max, and the mAb13 Fab concentration when the off-rate mAb13. All errors are from nonlinear least square fits. app t, for the initial response at the start of dissociation (R0) and the koff app was fit to mAb13 Fab app. t, for the initial response at the start of dissociation (R0) and the koff app was fit to mAb13 Fab app. 16 doi: A. Integrin conformational ensemble bioRxiv preprint https://doi.org/10.1101/2021.07.26.453735 this version posted July 26, 2021. ; The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made CC-BY-NC-ND 4.0 International license . available under a Ligand α1 α7 ECM or cell surface F β6-α7 α1 β6-α7 β-propeller β-propeller β-propeller α7 βI Thigh Genu Calf-1 PSI Hybrid I-EGF1 I-EGF2 I-EGF3 Calf-2 I-EGF4 β-tail e c e p d a e H i n a m o d o t c gE e i l r e w o L α β α β α Membrane Membrane β Bent-closed (BC) Extended-closed (EC) Extended-open (EO) F talin/kindlin/actin B. Integrin ligand binding Apparent 1vs1 binding: Integrin + L k app off app Ka app kon Binding coupled with conformational change: BC + L BC Ka kBC on kBC off KBC-EC EC + L EC Ka kEC on kEC off KEC-EO EO + L EO Ka kEO on kEO off Integrin•L BC•L K BC-EC L EC•L K EC-EO L EO•L C. Fabs against β1 subunit or α5 subunit used in this study Name Domain Conformational specificity 8E3 9EG7 SNAKA51 HUTS4 12G10 mAb13 PSI I-EGF2 Calf-1/Calf-2 hybrid βI βI BC + EC EC + EO EC + EO EO EO BC + EC Figure 1. Integrin conformational ensemble and conformational specific Fabs used to stabilize a defined ensemble . A Basal bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 this version posted July 26, 2021. The copyright holder for this preprint (which C CC-BY-NC-ND 4.0 International license . ; was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made B Extended available under a Open Association phase Association phase Association phase 9EG7 HUTS4 9EG7 BC membrane EC EO EC EO EO Dissociation phase Dissociation phase Dissociation phase Unlabeled ligand 9EG7 HUTS4 9EG7 BC EC EO EC EO EO 1400 ) I F M 1200 1000 ( P V D L - C T F 800 600 400 I FITC-LDVP 5 nM 10 nM 20 nM 1400 ) I F M 1200 1000 ( P V D L - C T F 800 600 400 I FITC-LDVP 5 nM 10 nM 20 nM 1400 ) I F M 1200 1000 ( P V D L - C T F 800 600 400 I FITC-LDVP 5 nM 10 nM 20 nM 200 200 200 0 0 5 10 15 Time (min) 20 0 0 5 15 Time (min) 10 20 0 0 5 10 15 Time (min) 20 25 D ) I F M ( 2 D 1 D M A C V - 8 8 4 a x e A Extended 800 600 400 200 Alexa488- VCAM D1D2 10 nM 20 nM 30 nM E Open 800 ) I F M ( 2 D 1 D M A C V - 8 8 4 a x e A 600 400 200 Alexa488- VCAM D1D2 10 nM 20 nM 30 nM l 0 l 0 0 5 10 Time (min) 15 20 0 5 10 Time (min) 15 20 25 F FITC-LDVP and Alexa488-VCAM D1D2 binding to intact α4β1 on Jurkat cells Condition Basal Extended Open FITC-LDVP app koff app kon app app koff /kon (nM) Kd (nM) 15.8±2.8 15.0±2.1 0.25±0.05 0.41±0.13 0.18±0.02 0.021±0.018 0.38±0.34 (104 M-1s-1) (10-3 s-1) 120±10 87±8 5.6±0.2 19±3 0.36±0.11 Alexa488-VCAM D1D2 app kon app koff app app koff /kon (nM) (104 M-1s-1) (10-3 s-1) K d (nM) 43±6 3.4±0.2 34±2 0.72±0.02 21±1 79±13 160±40 30±4 Figure 2.

Binding kinetics of ligands to intact α4β1 on Jurkat cells. A bioRxiv preprint doi: ; https://doi.org/10.1101/2021.07.26.453735 this version posted July 26, 2021. The copyright holder for this preprint (which Extended was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Association phase available under a CC-BY-NC-ND 4.0 International license . 9EG7 ) I F M 300 ( SNAKA51 membrane EO Dissociation phase EC 0 1 - 9 3 n F - 8 8 4 a x e A l 200 100 Alexa488-Fn39-10 10nM 20nM 30nM 9EG7 0 SNAKA51 0 10 30 Time (min) 20 40 EC EO B Open Association phase ) I F M 300 ( HUTS4 9EG7 EO Dissociation phase 0 1 - 9 3 n F - 8 8 4 a x e A l 200 100 Alexa488-Fn39-10 10nM 20nM 30nM 0 HUTS4 9EG7 0 10 20 Time (min) 30 40 EO C Alexa488-Fn39-10 binding to intact α5β1 on K562 cells app koff (10-3 s-1) 0.32±0.02 0.052±0.021 app kon (104 M-1s-1) 15.8±0.7 7.5±0.3 app appkoff /kon (nM) 2.0±0.2 0.7±0.3 Kd (nM) 1.9±0.1 1.3±0.1 Condition Extended Open Figure 3. Binding kinetics of Alexa488-Fn3 9-10 to α5β1 on K562 cells. A C Open (HUTS4 & 9EG7) The copyright holder for this preprint (which Association phase CC-BY-NC-ND 4.0 International license . bioRxiv preprint https://doi.org/10.1101/2021.07.26.453735 B Extended ; this version posted July 26, 2021. doi: Basal was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Association phase Association phase available under a Strep-biosensor Biotin-Fn39-10 8E3 9EG7 BC EC EO EC EO HUTS4 EO With free integrin in solution With free integrin in solution With free integrin in solution Dissociation phase Dissociation phase Dissociation phase 8E3 9EG7 BC EC EO EC EO EO HUTS4 Without free integrin in solution Without free integrin in solution Without free integrin in solution 0.6 0.7 0.8 ) m n ( e s n o p s e R 0.5 0.4 0.3 0.2 0.1 50 25 12.5 6.25 3.13 ) m n ( e s n o p s e R 0.6 0.5 0.4 0.3 0.2 0.1 50 25 12.5 6.25 3.13 1.56 ) m n ( e s n o p s e R 0.7 0.6 0.5 0.4 0.3 0.2 0.1 100 50 25 12.5 6.25 0 1.56 0 0.0 3.13 0 150 300 Time (second) 450 600 0 150 300 Time (second) 450 600 0 150 300 Time (second) 450 D Open (12G10) Association phase 0.5 E Equilibrium response analysis 0.8 e s n o p s e r 0.4 0.6 100 ) m n ( 12G10 0.3 e s n o p s e R Kd (nM) for Fn39-10 18.2±1.9 7.9±1.6 EO 0.4 m u i r b 50 With free integrin in solution 0.2 Basal Extended Open (HUTS4 & 9EG7) Open (12G10) Dissociation phase i l i u q E 25 12.5 6.25 3.13 0.2 0.1 1.1±0.6 12G10 5.0±0.1 33±1 18.2±1.9 14.9±0.1 0.0 1.6±0.7 0.0 0 20 5 1 ectodomain (nM) 40 60 80 100 600 Time (second) 300 0 900 1200 EO 2.3±0.1 7.0±0.1 7.9±1.6 33±1 Without free integrin in solution F α5β1 ectodomain binding to biotin-Fn39-10 app koff (10-3 s-1) app appk off /kon (nM) app Kd (nM) kon (104 M-1s-1 ) 1.2±0.2 1.1±0.6 Basal Extended 2.4±0.1 1.6±0.7 (HUTS4 &9EG7) Open (12G10) Open 7.9±0.2* 0.093±0.013* 0.047±0.008* 1.95±0.05* Figure 4. Binding kinetics of α5β1 ectodomain to Fn39-10 600 app off A Dissociation in presence of mAb13 Fab under basal condition doi: bioRxiv preprint this version posted July 26, 2021. ; The copyright holder for this preprint (which https://doi.org/10.1101/2021.07.26.453735 k 1000 µ was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

It is made available under a +mAb13 CC-BY-NC-ND 4.0 International license . EO•L EO k off kEO•L-C•L conf kC•L-EO•L conf EC BC C•L kC off k mAb13 on mAb13•C•L kC off mAb13 ) I F M ( P V D L - C T F I 900 800 700 600 500 mAb13 400 EO BC EC mAb13•BC mAb13•EC 0 60 120 180 Time (s) 240 B Dissociation in presence of mAb13 Fab under extended condition 1600 kEO•L-EC•L conf EC +mAb13 mAb13 1400 9EG7 kEC•L-EO•L conf 9EG7 kmAb13 on 9EG7 ) I F M 1200 µ EO•L k EO off EC•L kEC off mAb13•EC•L kEC off ( P V D L - C T F I 1000 800 mAb13 600 9EG7 9EG7 9EG7 400 EO EC mAb13•EC 0 60 120 180 Time (s) 240 C Dependence of k app off on mAb13 Fab 140 120 100 80 p p a f f o k 60 40 20 0 E C Basal Extended µ k 130±10 110±30 0 1 2 6 mAb13 Fab (µM) 3 4 5 7 8 Figure 5. Dissociation of FITC-LDVP from intact α4β1 in presence of closure-stabilizing Fab. 300 app koff 0.38±0.03 11.3±0.2 17.4±0.5 20.9±0.9 35.9±1.2 53.4±2.3 66.6±3.9 300 bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which A Basal Steady state was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license . kEO•L-C•L conf kC•L-EO•L Steady state 0.4 conf app off 5.8±0.1 350±14 543±22 757±31 940±38 1205±45 1445±69 s-1 mAb13 ( M) 0 1.25 2.5 5 10 20 40 k 0.3 ) m n ( e s n o p s e R EO•L BC•L With free integrin in solution EC•L 0.2 +mAb13 Dissociation phase Strep-biosensor Biotin-Fn39-10 0.1 k mAb13 on mAb13•BC•L mAb13 mAb13•EC•L 0.0 Without free integrin in solution 0 1 2 3 4 5 6 7 8 9 10 Time (s) B Extended Steady state k EO•L-EC•L conf Steady state kEC•L-EO•L conf 0.6 9EG7 EO•L EC•L With free integrin in solution +mAb13 Dissociation phase 9EG7 ) m n ( e s n o p s e R 0.5 0.4 0.3 0.2 mAb13 ( M) 0 1.25 2.5 5 10 20 40 app s-1 off 2.3±0.1 189±6 343±9 548±17 828±28 1125±44 1476±52 k k mAb13 on 0.1 mAb13 0.0 9EG7 mAb13•EC•L Without free integrin in solution 0 1 2 3 4 5 6 7 8 9 10 Time (s) app off C Dependence of k on mAb13 Fab 1600 1400 1200 1000 p p a f f o k 800 600 400 EC µ k s-1 200 0 Basal Extended 5.4±0.8 12.8±1.0 1600±100 1900±100 0 5 10 15 20 25 30 35 40 mAb13 Fab ( M) Figure 6. Dissociation of α5β1 ectodomain from biotin-Fn39-10 in presence of closure-stabilizing Fab. Condition States presentin the ensemble EOEO•L EC, EOEC•L, EO•L Basal BC, EC, EOBC•L, EC•L, EO•L A bioRxiv preprint doi: https://doi.org/10.1101/2021.07.26.453735 ; this version posted July 26, 2021. The copyright holder for this preprint (which Strategy for determining the intrinsic on- and off of each integrin state was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a for ligand CC-BY-NC-ND 4.0 International license . app koff app kon and EO kon EO koff app(EC+EO) Open PEC PEC + PEO P EC•L P EC•L+P EO•L BC + app(BC+EC+EO) PBCkon kon app(BC+EC+EO) PBC•Lkoff koff PEO PEC + PEO P EO•L + P EC•L+P EO•L EC kon EO kon (Eq.1) ≈ + kon Extended EC koff EO koff app(EC+EO) (Eq.2) ≈ koff EO EC + PEOkon PECkon PEC•L koff ≈ (Eq.3) EO EC + PEO•L koff Conformational state populations (%) in absence and presence of ligand at steady state for each characterized integrin preparation BC + ≈ (Eq.4) B intact α4β1-Jurkat BC EC EO 1.1 0.4 intact α5β1-K562 BC EC EO 0.05 α5β1-ectodomain BC EC EO 4.6 31.1 64.3 87.1 0 12.9 0 0 100 0.5 0.2 99.3 0.2 0 99.8 0 0 Ligand bound - - - + + 100 + Intrinsic on- and off- rates of each integrin state for ligand Ensemble 0.11 98.5 0 0 10.9 0 0 99.84 0 0 22.6 0 0 Basal Extended Open Basal Extended Open 28 0 0.04 89.0 0.05 0 72 100 30 0 0.01 77.4 0.01 0 70 100 99.95 99.99 100 100 C EO 5.6±0.2 0.021±0.018 0.18±0.02 0.010±0.001 3.4±0.2 0.72±0.02 30±4 BC 120±20 170±60 130±40 160±60 EC 300±170 640±590 130±40 390±250 150±90 57000±23000 22000±7000 33000±22000 1.0±0.1 7.5±0.3 0.052±0.021 1.3±0.1 0.098±0.001 7.9±0.2 0.093±0.013 1.1±0.6 0.085±0.046 a k on koff Kd koff k on k off Kd k off k on k off Kd k off k on k off Kd (nM) k off (104M-1s-1) a (10-3s-1) b (nM) =Kd*kon(10-3s-1) (104M-1s-1) (10-3s-1) (nM) =Kd*kon(10-3s-1) (104M-1s-1) a (10-3s-1) b (nM) =Kd*kon(10-3s-1) (104M-1s-1) (10-3s-1) b Intact α4β1 for LDVP c a a Intact α4β1 for VCAM D1D2 b c a 35±6 1900±1200 4000±2100 1400±800 37±2 1000±600 3400±2600 1300±1000 Intact α5β1 for Fn39-10 c a 34±2 740±400 3400±2600 1200±900 α5β1 ecto for Fn39-10 a c =Kd*kon(10-3s-1) D Asp-binding pocket in integrin αIIBβ3 Open state Closed state Asp Asp Mg2+ MIDAS Mg2+ MIDAS β1-α1 loop β1-α1 loop α2-α3 loop α2-α3 loop Figure 7.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . Activation by Cleavage of the Epithelial Na+ Channel a and g Subunits Independently Coevolved with the Vertebrate Terrestrial Migration Xue-Ping Wang1, Deidra M. Balchak1, Clayton Gentilcore1, Nathan L. Clark3, and Ossama B. Kashlan1,2 From the Departments of 1Medicine, Renal-Electrolyte Division and 2Computational and Systems Biology, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, and the 3Department of Human Genetics, University of Utah, Salt Lake City, Utah 84112 To whom correspondence should be addressed: Ossama Kashlan, Department of Medicine, Renal- Electrolyte Division, University of Pittsburgh, S828B Scaife Hall, 3550 Terrace St., Pittsburgh, Pennsylvania 15261, Telephone: (412)648-9275; E-mail: [email protected] 1 1 2 3 4 5 6 7 8 9 10 11 12 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . Abstract: Vertebrates evolved mechanisms for sodium conservation and gas exchange in conjunction with migration from aquatic to terrestrial habitats. Epithelial Na+ channel (ENaC) function is critical to systems responsible for extracellular fluid homeostasis and gas exchange. ENaC is activated by cleavage at multiple specific extracellular polybasic sites, releasing inhibitory tracts from the channel’s a and g subunits. We found that proximal and distal polybasic tracts in ENaC subunits coevolved, consistent with the dual cleavage requirement for activation observed in mammals. Polybasic tract pairs evolved with the terrestrial migration and the appearance of lungs, coincident with the ENaC activator aldosterone, and appeared independently in the a and g subunits. In summary, sites within ENaC for protease activation developed in vertebrates when renal Na+ conservation and alveolar gas exchange was required for terrestrial survival. 2 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . The migration of vertebrates to a terrestrial habitat began ~380 million years ago and was driven by competitive pressure and the relative availability of resources [1].

Survival in a terrestrial environment required several adaptations, including those for respiration, extracellular fluid volume balance and osmoregulation. Aldosterone signaling has a significant role in extracellular fluid homeostasis for tetrapods and depends on three proteins that appeared during the evolution of vertebrates: aldosterone synthase, the mineralocorticoid receptor (MR), and 11b-hydroxysteroid dehydrogenase (11b-HSD2) [2]. One of the major endpoints of aldosterone signaling through MR is increased activity of the epithelial Na+ channel (ENaC) in the kidney, which results in enhanced Na+ and fluid retention. Having originally evolved in a marine environment, these mechanisms to achieve electrolyte homeostasis may have required adaptive modifications as the ancestors of today’s terrestrial vertebrates moved to a relatively dry environment. ENaCs are heterotrimers comprising a, b, and g subunits that each have intracellular amino and carboxyl termini, two transmembrane domains, and a large extracellular region (Figs. 1A, B). An ENaC d subunit can substitute for the a subunit in assembled channels, resulting in channels with distinct functional properties [3]. However, the d subunit's absence in mice and rats has led to fewer reports of its function and physiology, as most of this work has been performed in these model systems. ENaC subunits belong to the ENaC/Degenerin family of proteins that form trimeric cation-selective channels with similar extracellular folds [4-6]. Aldosterone regulation of ENaC activity and abundance depends on essential sequences in both the extracellular and intracellular domains. These include polybasic tracts in extracellular GRIP (Gating Release of Inhibition by Proteolysis) domains unique to ENaC subunits; however, the polybasic tracts of ENaC are not present in all vertebrate lineages (Fig. 1B) [5, 7-9]. Aldosterone promotes double cleavage of the a and g subunits at polybasic tracts, which liberates embedded inhibitory tracts and converts dormant channels to constitutively open ones [10-13], thereby leading to increased fluid retention. ENaC subunits also feature PY motifs on their intracellular C-termini that are important for protein degradation (Fig. 1B) [14, 15]. These motifs recruit WW-domain containing Nedd4-2, whose binding results in the ubiquitylation of ENaC subunit N-termini and enhanced endocytosis and degradation [16, 17]. Nedd4-2 dependent reduction of ENaC abundance is inhibited by aldosterone [17]. These complementary regulatory mechanisms help aldosterone minimize Na+ wasting and fluid loss during volume contraction. Here we investigated the evolution of ENaC regulatory mechanisms to determine which features coevolved with the marine-terrestrial transition. We consistently found both activating cleavage sites in the ENaC a and g subunits of terrestrial vertebrates, while they appeared only sporadically in fishes. We confirmed that cleavage occurred at sites found in the g subunit from Australian lungfish, leading to channel activation.

Phylogenetic analysis and likelihood ratio tests showed a coevolutionary dependence of the polybasic tracts with each other. They also showed a coevolutionary dependence of tandem polybasic tracts with terrestrial status and with lungs. Analysis of ancestral reconstructions strongly suggests that the polybasic tracts appeared independently in the a and g subunits. Similar analyses of the PY motif showed no coevolutionary pattern and that the PY motif first arose in an ancient ancestral ENaC subunit. Our data suggest that changes associated with adaptation to terrestrial life provided selective pressure for the development of ENaC activation by cleavage. Results ENaC subunit sequence conservation and evolution The subphylum Vertebrata encompasses the vast majority of species in the phylum Chordata. We identified ENaC subunit sequences in each of the classes of Vertebrata, including all four ENaC subunits in each mammalian order. In order Rodentia, we found the d subunit in Spalacidae (blind mole rats; accession XP_008850903), but in neither Muridae (mice and rats) nor Cricetidae (hamsters and voles), suggesting that the d subunit was lost on the lineage to mice and rats after the divergence of Spalacidae ~34 million years ago [18, 19]. Notably, although we found ENaC subunits in Sarcopterygii (lobe-finned fishes) and the ray-finned Polypteriformes (ropefish), we did not find ENaC subunits in any other ray-finned fishes. Instead, BLAST searches using human ENaC subunits identified ENaC-related acid-sensing ion channels (ASICs) and uncharacterized ENaC-like proteins. This suggests ray-finned 3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . fishes lost all ENaC subunits after the early divergence of the Polypteriformes order and before the divergence of Chondrostei (sturgeons) and Neopterygii (gars and teleosts). Similarly, we only found ENaC-related proteins in non-vertebrate chordates (e.g. tunicates or cephalochordates). To examine sequence conservation within ENaC sequences, we aligned ENaC subunit sequences found in each of the classes and top teleost and non-vertebrate chordate hits identified in a BLAST search using human ENaC subunits (Supplementary Table 1). Sequence conservation was highest in the transmembrane helices (TM1 and TM2) and defined extracellular secondary structures in the ENaC and ASIC1 structures (Fig. 1C and Supplementary Fig. 1) [4, 5]. Sequence conservation was lowest in the intracellular N- and C-termini, extracellular loops, and the GRIP domain unique to ENaC subunits. Sequences associated with key functions demonstrated varied conservation. For example, G/S- X-S in TM2 contributing to ion selectivity [20] is well-conserved within a conserved region (Fig.

1C). The PY motif in the intracellular C-terminus is also well conserved, but in a poorly conserved region. The region in the GRIP domain containing the embedded inhibitory tracts and cleavage sites is poorly conserved. Figure 1. Sequence conservation in ENaC subunits. A, Space filling model of ENaC (pdb code: 6BQN) with plane indicating position of outer membrane border. The a and g subunits are white and grey, respectively. The b subunit is colored by domain as indicated in panel B. Intracellular structures are absent in this structural model. B, Linear model of human ENaC subunits showing domain organization and highlighting position of polybasic cleavage sites and PY motifs. C, Sequences (Supplementary Table 1) were aligned using MUSCLE [21]. Residue symbol sizes are proportional to frequency at a given position. Key features in the sequence are indicated, as are the approximate position of a helices (rounded rectangles) and b-strands (arrows). Colors correspond to protein domains, as indicated in panel B. The GRIP domain is unique to ENaC subunits in the ENaC/Deg family. The P1 and P2 b-strands in the GRIP domain are not indicated, but are likely near the inhibitory tract and distal site, respectively, if present. 4 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . To examine the evolution of ENaC subunits, we assessed the phylogenetic relationship of our 53 assembled sequences with maximum-likelihood methods (Fig. 2). In the resulting tree, we observed 6 clades: ENaC a, b, g, and d subunits, ASICs, and ENaC-like proteins. Each ENaC subunit clade largely recapitulated the expected phylogenetic relationships, with jawless fishes (lampreys) rooted closest to the base of the clade, followed by cartilaginous fishes (elephant shark), lobe-finned fishes (coelacanth and lungfishes), reptiles and birds, and mammals. Our evolutionary model suggests three gene duplication events for ENaC subunits, as previously reported [22]. The first duplication (node 1) gave ancestral b/g subunits (node 2) and a/d subunits (node 3). The second duplication (node 2) gave Figure 2. Phylogenetic tree of ENaC subunits. Maximum-likelihood tree calculated from ENaC subunit sequences of marine species and select terrestrial vertebrates, and ENaC-related proteins. Branch support bootstrap values are shown. Scale bar indicates the number of substitutions per site. Key ancestral nodes are indicated by circled numbers. A. Lungfish = Australian Lungfish, E. Shark = Elephant Shark, E. Lamprey = European River Lamprey, J. Lamprey = Japanese Lamprey, J. Medaka = Japanese Medaka, S. Lamprey = Sea Lamprey, W. Lungfish = West African Lungfish. 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . ancestral b and g subunits. Both of these duplication events occurred in ancestors of jawless fishes at least 550 million years ago. The third duplication (node 4) leading to the divergence of a and d subunits had more associated uncertainty. Given the presence of a and d subunits in coelacanth, the duplication likely occurred before the divergence of the coelacanth and tetrapods. Polybasic tracts in the ENaC GRIP domains varied over time We then examined the sequence conservation of the polybasic tracts in the ENaC subunit GRIP domains required for ENaC activation by cleavage. In mammals, the GRIP domains of the a and g subunits are each subject to double cleavage, leading to the release of embedded inhibitory tracts and channel activation. The b and d subunits are not similarly processed. The proprotein convertase furin cleaves both the proximal and distal sites of the a subunit, and the proximal site of the g subunit [7]. Cleavage distal to the g subunit inhibitory tract can be catalyzed by several proteases at the cell surface, including prostasin at a polybasic tract [23, 24]. Similar results were reported for ENaC from Xenopus laevis [25]. We inspected our multiple sequence alignment for polybasic tracts that aligned closely with the tracts in mammalian and frog a and g subunits (Supplementary Fig. 1). We found polybasic tracts aligning with the proximal and distal sites of the human a and g subunits for all tetrapod a and g subunits (Table 1). We also found both sites present in the g subunit from Erpetoichthys calabaricus (Ropefish) and Neoceratodus forsteri (Australian lungfish), but not in Protopterus annectens (West African lungfish). In coelacanth, we identified single polybasic tracts in the a, b, and g subunits. The elephant shark's g subunit also had a single distal polybasic tract. The human d subunit sequence also exhibits a polybasic tract in this region, but experimental evidence shows that it is not cleaved [26]. a d b g Animal S. Lamprey J. Lamprey E. Shark Ropefish Coelacanth W. Lungfish A. Lungfish Frog Salamander Turtle Chicken Cow Human Terr. Lungs X X X X X X X X X X X X X X X site 1 Ø Ø Ø Ø — Ø Ø RVKR RVSR — Ø RERR RVRR RSPR RHKR — RTSR RQKR — RSRR RGVR — RSRR RRAR — site 2 — — (⋯) — RSNR — — — site 1 — — (⋯) — — — site 2 Ø Ø Ø Ø — Ø Ø — Ø — — site 2 — — — — KRER — — — — Ø — — — — site 1 — — — — — — — Ø — — RLQR1 — RLQR1 — site 2 — — RQHR RKRR NRKR VKQR — site 1 — — — — — RKLR RQYR RSKR KRTR Ø Ø KVRR NKRK KVRR RKRK RKRR RKRK RKRR RKRK Table 1. Polybasic sequences aligning with human ENaC subunit proximal (site 1) and distal (site 2) cleavage sites. Terr., terrestrial. Dashes indicate absence of polybasic tract. 1The ENaC d subunit in mammals has a polybasic sequence in the aligned region, but is not cleaved in human channels [26]. Ø, no sequence was available. (⋯), GRIP domain sequence was missing in the available sequence.

Species abbreviations are as in Fig. 2. Australian lungfish ENaC g subunit is cleaved at GRIP domain polybasic tracts To determine whether apparent cleavage sites are functional in a subunit rooted before the emergence of tetrapods, we examined the ENaC g subunit from Australian lungfish (Ag) in Xenopus laevis oocytes. Ag has two furin cleavage motifs predicted to lead to activation (Table 1, Fig. 3A). To isolate functional effects to Ag, we coexpressed an a subunit from mouse lacking residues excised by furin (mouse a∆206- 231; maD), rendering it incapable of proteolytic activation [27]. We also coexpressed a mouse b subunit truncated before the C-terminal PY motif (mouse bR564X; mbT) to decrease channel turnover [28]. All g subunits contained a C-terminal hemagglutinin (HA) epitope tag to facilitate detection (Fig. 3A). Oocytes to expressing ENaC were conjugated with a membrane impermeant biotin reagent 6 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . Figure 3. Predicted cleavage sites in the ENaC g subunit from Australian lungfish are functional. A, Schematic of Ag topology. Ag has 2 predicted furin cleavage sites in its extracellular GRIP domain. All g subunits were labeled with C-terminal epitope tags to facilitate detection of full-length subunits and the larger of the cleaved fragments. B, Xenopus oocytes were injected with cRNAs encoding ma∆, mbT, and HA-tagged g subunits, as indicated. One day after injection, whole cell lysates and cell surface isolates were blotted and probed for HA and GAPDH. Full length and cleaved bands are indicated and band densities were quantified. An overexposed blot is shown to highlight cell surface bands. Over exposed areas are red. C, Cleavage %, calculated as cleaved/(cleaved +full length)*100 is shown. Data were analyzed by repeated measures two-way ANOVA with Šidák's multiple comparison test. P-values are shown for indicated comparisons. Cleavage was also greater for Ag than for mg (p=0.05). D, Normalized total expression was calculated by normalizing the sum of full length and cleaved bands to the mean of mg after normalizing each sample for loading based on GAPDH from the same blot. E, Surface expression % was calculated using band densities adjusted for the fraction of the respective sample loaded. F, Normalized surface expression was calculated by multiplying values form the same sample in D and E, and then normalizing to the mean of mg. Data in D–F were analyzed by one-way ANOVA with Tukey's multiple comparison test. No significant differences between groups were found. Note that due to the lack of GAPDH data for one blot, the number of replicates for D and F (n=6) are one fewer than for C and E (n=7).

G, Whole cell currents were measured in injected oocytes by two-electrode voltage clamp, with voltage clamped at -100 mV. Representative traces of indicated subunit combinations are shown. Currents were continuously recorded in a bath solution containing 110 mM Na+. The ENaC-blocking drug amiloride (100 µM) was added at the end of each experiment to determine the ENaC-mediated current. H, Log transformed amiloride-sensitive inward currents are plotted, and were analyzed by one-way ANOVA followed by Tukey's multiple comparison test. P-values for comparisons where p < 0.05 are shown. Bars indicate mean values; errors shown are SD. 7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . label surface proteins. We then lysed the oocytes, isolated biotin labeled proteins using NeutrAvidin beads, and analyzed whole cell and surface enriched samples by western blot using anti-HA antibodies. We detected full-length and cleaved forms of mγ and Ag (Fig. 3B). In both cases, the proportion of cleaved g subunit was higher in the surface pool than the total pool (Fig. 3C), consistent with trafficking-dependent processing reported for mammalian ENaC [29]. To confirm that cleavage occurred at the predicted sites, we mutated the terminal Arg in both sites to Ala (Ag2A). When we expressed Ag2A in oocytes, the higher molecular weight band remained readily apparent while the lower molecular weight band largely disappeared. Comparison of quantified band densities confirmed that mutation of predicted cleavage sites in Ag greatly diminished apparent cleavage (p = 0.007). The extent of any cleavage was similar in total and surface pools for Ag2A, in contrast to Ag. When we measured whole cell currents in oocytes expressing Ag, we found that mutating predicted furin sites decreased ENaC-mediated currents by 82% (p <0.0001), consistent with mutation precluding proteolytic activation (Figs. 3G–H). Notably, currents from oocytes expressing Ag2A were similar to currents from oocytes lacking g subunits altogether. Similar surface expression levels for Ag and Ag2A (Figs. 3D–F) suggest that differences in expression or surface delivery do not account for the differences we observed in ENaC-mediated currents. Together, these data provide evidence that Ag undergoes activating cleavage at the predicted furin cleavage sites. GRIP domain cleavage sites coevolved with each other Activation by cleavage requires two cleavage events within a single subunit and the release of embedded inhibitory tracts [8, 30]. The data in Table 1 suggest that sites 1 and 2 coevolved with each other.

To test this idea, we calculated the probability that the sites coevolved with each other using nested likelihood models based on our model of ENaC evolution in BayesTraits (see Supplementary Table 2) [31, 32]. Each sequence was assigned the traits of the presence or absence of "site 1" and "site 2", as indicated in Table 1. All models considered the evolutionary gains and losses of these traits over the full phylogenetic tree capturing the four ENaC subunits (Fig. 2). To test for dependence between the two sites, we contrasted two nested models, an independent model and a dependent model. The independent model contained two parameters: an appearance rate for both sites and a loss rate for both sites. The loss or gain of a site did not depend on the status of the other site, and therefore served as the null hypothesis. The dependent model added two free parameters by having separate gain and loss rates depending on the status of the other site. A likelihood ratio test between nested models showed a strong preference for the dependent model (p=0.01). Fits to the dependent model suggest little selection pressure in the change from 0 sites to 1 site, with loss rates 11- fold higher than gain rates. In contrast, in the change from 1 site to 2 sites, gain rates were 7-fold higher than loss rates. This is congruent with the requirement for two cleavage sites within a single subunit’s GRIP domain for channel activation, and supports the notion that selection pressure derives from the functional consequence of double cleavage [8, 30]. GRIP domain double cleavage coevolved with the terrestrial migration and with lungs The data in Table 1 also suggest that the terrestrial migration or the development of lungs may have provided the selection pressure for channel activation by cleavage. ENaC function is also important in aldosterone-insensitive tissues, including the airway, where control of airway surface liquids is essential for lung function. We first calculated the likelihood of site 1, site 2, or both sites coevolving with the terrestrial transition using nested likelihood models. Each sequence was assigned the traits of the presence or absence of "site 1", "site 2", or "tandem sites", and "terrestrial or marine" states, as indicated in Table 1. The independent model contained three parameters: a site appearance and loss rate, and a rate for transition from marine to terrestrial status (see Supplementary Table 2). The independent model does not allow a relationship between rates of site gain or loss with respect to marine or terrestrial state, and therefore served as the null hypothesis. The dependent model added two free parameters by having separate site gain and loss rates in the marine and terrestrial states. A likelihood ratio test between nested models showed a strong preference for the dependent model for site 1 (p=0.0053) and for tandem sites (p=0.0098), but not for site 2 (p=0.26). Fits to the dependent model suggest that in the marine state, site 8 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . appearance and disappearance was dynamic, with a disappearance rate being 5-fold higher than the appearance rate. However, once in the terrestrial state, both rates dropped to 0, supporting the notion of selection pressure favoring the presence of cleavage sites in the terrestrial state. We then calculated the likelihood of site 1, site 2, or both sites coevolving with the development of lungs using analogous procedures. A likelihood ratio test between nested models showed a preference for the dependent model for site 1 (p=0.031) and for tandem sites (p=0.018), but not for site 2 (p=0.94). ENaC expression in fishes with lungs and amphibians Lungs coevolved with the terrestrial migration of vertebrates (p=0.0004 in our dataset), and may have driven the coevolution of GRIP domain polybasic tracts with lungs that we observed. However, ENaC subunit transcripts were not detected in lung tissues from either of the lungfish species examined [33, 34]. To further investigate the role of ENaC in fishes with lungs, we examined the tissue distribution of ENaC transcripts in ropefish, which evolved lungs independently from tetrapods [35]. We observed clear bands for the transcripts of all three ENaC subunits in the gills and kidneys of the ropefish (Fig. 4A), which are important organs for ion homeostasis. In the lung, we observed a clear band for the b subunit, a faint band for the a subunit, and a barely perceptible band for the g subunit. In the other tissues examined, we observed broad expression of b subunit transcripts, faint bands for the a subunit in the liver and intestine, and no bands for the g subunit. Taken together with the lungfish data [33, 34], these data suggest that transcripts for ENaC subunits with GRIP domain polybasic tracts (Table 1) are readily detected at important sites of ion exchange, but are difficult to detect in lungs. Figure 4. Tissue distribution of Erpetoichthys calabaricus (ropefish) and Xenopus laevis ENaC subunit transcripts by RT-PCR. cDNA libraries were generated from tissue homogenates. PCR reactions were performed using primers indicated in Supplementary Table 3. 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . We also investigated the tissue distribution of ENaC subunit transcripts from the African clawed frog, in which both the a and g subunits have GRIP domain cleavage sites [25]. Similar to mammals, we detected bands for a, b, and g subunit transcripts in both kidneys and lungs (Fig.

4B). In contrast to humans where bands for the d subunit were relatively faint for both kidneys and lungs [36], we observed a strong band for the d subunit transcript in the kidney, and no band in the lung. The expression pattern we observed is consistent with a previous investigation of a and d subunit expression in Xenopus laevis tissues [25], and suggest that cleavage regulates ENaC function in both the lungs and kidneys of frogs, similar to mammals. a and g subunit GRIP domain cleavage sites evolved independently The absence of cleavage sites in species rooted closest to the early gene duplication events suggests that each cleavage site did not result from a common ancestor, but rather appeared independently in the a and g subunits. To test this idea, we performed ancestral reconstruction of nodes 1-4 in our phylogenetic tree (Fig. 2) using BayesTraits and MCMC methods. In the unconstrained evolutionary model, each of these nodes were free to adopt any status for the site of interest. In the divergent evolutionary model, nodes 1-4 were constrained to contain the site of interest. In the convergent evolutionary model, node 1 was constrained to exclude the site of interest. Models were compared by converting marginal likelihoods resulting from MCMC runs of each model to log Bayes Factors, where values of >2, 5-10, and >10 support positive, strong, and very strong preferences for the better fitting model, respectively. Within the framework of sites 1 and 2 coevolving with each other, comparison of the unconstrained evolutionary model with the convergent evolutionary model resulted in log Bayes Factors of 0.38 for site 1 and 0.34 for site 2, supporting no preference between models. Comparison of the divergent evolutionary model to the convergent model for each of the sites resulted in log Bayes Factors of 6.5 in both cases, supporting a strong preference for the convergent evolutionary model for each of the cleavage sites. This suggests that, despite analogous molecular mechanisms in the a and g subunits, the appearance of cleavage sites in the a and g subunits were independent events. Neither the proximal nor the distal cleavage site arose from a common pre-a/g ENaC subunit. ENaC PY motifs evolved through divergent evolution Our data suggest that cleavage of the a and g subunits evolved contemporaneously with aldosterone synthase. Indeed, aldosterone enhances activating cleavage of ENaC subunits [12, 13]. Due to the apparent connection to aldosterone, we examined another ENaC target of aldosterone-dependent regulation, the PY motifs in the C-termini. The PY motifs facilitate enhanced channel turnover via Nedd4- 2, which can be inhibited by aldosterone and other hormones through kinase signaling [37]. Mutation or deletion of the PY motifs increases ENaC function, leading to Liddle syndrome which is characterized by hypertension, hypokalemia and low aldosterone levels [38]. We found PY motifs in all a, b, and g subunits where sequences of the C-termini were available (Table 2, Supplemental Fig.

1), suggesting a divergent evolutionary model that did not depend on terrestrial status. Interestingly, d subunits apparently lost their PY motifs after their divergence from the a subunit in the coelacanth ancestor. We tested whether appearance of the PY motif depended on the terrestrial migration or the development of lungs using nested likelihood models, as above. A likelihood ratio test between the nested models showed no preference for the dependent model for either terrestrial status (p=0.54) or lungs (p=0.28), suggesting their evolution was affected by neither. To test whether the PY motifs evolved from a common ancestor, we tested models of PY motif evolution using ancestral reconstruction at nodes 1-4 in Fig. 2 and MCMC methods. The unrestricted model suggested probabilities of a PY motif present at 98.1%, 99.3%, 91.5%, and 75.9% for nodes 1, 2, 3, and 4, respectively. We then tested a series of restricted models to test competing evolutionary models. In the divergent evolutionary model, nodes 1-4 were restricted to contain PY motifs. Comparison of the unrestricted and divergent evolutionary models resulted in a log Bayes Factor of 0.19, suggesting these models were nearly equivalent. In convergent evolutionary models, one of the four early nodes was restricted to not contain a PY motif, resulting in 4 models. Comparison of the divergent model to each of the convergent models resulted in log Bayes Factors ranging from 4.7 to 8.9, suggesting a strong preference for the divergent model over each of the non-divergent models. These 10 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . data suggest that the PY motif in each subunit, in contrast to the polybasic tracts, arose from a common progenitor. Animal S. Lamprey J. Lamprey E. Shark Ropefish Coelacanth W. Lungfish A. Lungfish Frog Salamander Turtle Chicken Cow Human Terrestrial X X X X X X Lungs X X X X X X X X X a PPPSF PPDY (⋯) PPPAY PPAY PPPAY PPPAY PPPAY PPPAY LPSY LPSY PPPAY PPPAY d Ø Ø Ø Ø (⋯) Ø Ø — Ø — — — — b PPPHY PPPHY PPPRY PPPHY PPPNY PPPHY PPPKY PPPNY Ø PPPNY PPPNY PPPNY PPPNY g PPPQY PPPQY PPPNY PPPNY PPPTY PPPQY PPPQY PPPKY Ø PPPNY PPPNY PPPRY PPPKY Table 2. PY motifs (L/P-P-X-Y) in the C-terminal tails of ENaC subunits from various species. Ø, no sequence was available. (⋯), sequence for the C-terminal region was missing in the available sequence. Species abbreviations are as in Fig. 2. Discussion Our results suggest that the two ENaC GRIP domain cleavage sites in each of the a and g subunits co-evolved with each other, but appeared independently in each subunit. Our results also suggest that double cleavage coevolved with both the terrestrial migration of vertebrates and the development of lungs (Fig.

5). Co-evolution of site 1 with site 2 makes sense given that functional consequences of cleavage in ENaCs only result from double cleavage of a single subunit followed by liberation of the intervening inhibitory tract [8, 39]. We found that the polybasic tracts appeared sporadically in the GRIP domain of ENaC subunits from marine species, but appeared and remained consistently in terrestrial vertebrates. Terrestrial life required adaptive changes to overcome the stresses of electrolyte regulation and gas exchange. Co-evolution of the terrestrial migration with tandem GRIP domain polybasic tracts, and consequently higher ENaC activity, suggests that selection pressure could have arisen from either stressor. Air-breathing organs evolved independently several times as an adaptation to chronic or periodic environmental hypoxia [35]. Lungs appeared in lobe-finned fishes after their divergence from other bony fishes, and developmental and morphological evidence support homology between the lungs of lungfishes and tetrapods. Extant deep-water coelacanths possess vestigial lungs that correspond to the likely functional lungs of the Cretaceous shallow-water coelacanth Axelrodichthys [40, 41]. Air-breathing organs evolved several times in ray-finned fishes, giving Polypteridae (ropefish) morphologically distinct lungs, and other ray-finned fishes air-breathing capabilities through modified gas bladders and labyrinth organs (Fig. 5). Notably, we identified ENaC subunits in ropefish and not in any other ray-finned fish, suggesting ENaC subunits were lost soon after the divergence of Polypteriformes from ray-finned fishes in the Devonian period ~360 million years ago [42]. Their loss reflects a lack of selective pressure to maintain the ENaC subunits. Evidence supports electrogenic Na+ transport through an ENaC-like channel in the gills of ray-finned fishes [43]. Related proteins like ASIC4, proposed to play this role [44], or the uncharacterized ENaC-like paralogs may have left ENaC subunits dispensable. Airway surface liquids are critical to the function of mammalian lungs and are regulated by ENaC and other ion channels [45]. ENaC a subunit knockout mice die of asphyxiation shortly after birth due to an inability to clear fluid from the lung [46], and ENaC polymorphisms have been associated with lung dysfunction [47, 48]. ENaC regulation by cleavage is relevant in primary airway epithelial cell cultures [49, 50], where ENaC is regulated by glucocorticoids rather than by aldosterone [51]. However, transcripts of ENaC subunits with GRIP domain polybasic tracts were not detected in the lungs of lungfishes [33, 34] and were faint in the 11 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license .

lungs of the ropefish (Fig. 4). As lungs coevolved with the terrestrial migration of vertebrates, concurrent stresses may have provided selection pressure for ENaC activation by cleavage. Figure 5. Schematic view of evolution of aldosterone signaling, air-breathing organs, ENaC, and ENaC regulatory motifs. ENaC subunits were not found in non-vertebrate chordates or in teleosts. The ancestral ENaC subunit likely had a PY motif and was a substrate for Nedd4-2-dependent regulation. Like mammalian ENaC a subunits, the ancient ENaC subunit may have formed functional homotrimers, or may have formed channels with other ENaC paralogs. ENaC a, b and g subunits appeared before the emergence of jawless fishes, whereas ENaC d subunits first appeared in an ancestor of the lobe-finned coelacanth. Proteins required for aldosterone signaling (MR, 11b-HSD2, and aldosterone synthase) evolved before the emergence of tetrapods. Lungs appeared in lobe-finned fishes on the lineage to tetrapods. Air-breathing organs (e.g. respiratory gas bladders and labyrinth organs) evolved independently in ray-finned fishes, including morphologically distinct lungs in Polypteriformes. Individual GRIP domain ENaC cleavage sites first appeared sporadically in marine species. Dual cleavage sites appeared consistently in the ENaC a and g subunits in terrestrial vertebrates. CR, corticoid receptor; MR, mineralocorticoid receptor; GR, glucocorticoid receptor. Animal silhouettes courtesy of PhyloPic (http://www.phylopic.org). Emersion from an aquatic environment poses a risk of desiccation and a challenge to terrestrial life [52]. Aldosterone signaling through MR regulates electrolyte balance and total body volume in tetrapods. Key molecules required for aldosterone signaling appeared at various points during the evolution of vertebrates (Fig. 5) [2]. Interestingly, aldosterone synthase is absent in cartilaginous and ray- finned fishes, but appears in lobe-finned fishes on the lineage to terrestrial vertebrates [53]. The appearance of aldosterone synthase coincided with the appearance of the ENaC d subunit, and preceded the consistent appearance of tandem polybasic tracts in the ENaC a and g subunits. In ray-finned fishes, osmoregulation and euryhaline adaptation primarily occur through glucocorticoid receptor signaling [54]. Teleost MR signaling is implicated in central nervous system functions that include sodium appetite, but plays a secondary role in ion transporter regulation in osmoregulatory organs, in contrast to MR signaling in terrestrial vertebrates. Notably, aldosterone promotes cleavage of the ENaC a and g subunits in mammals through MR signaling [12, 13]. Analysis of transcripts from ropefish and lobe-finned lungfishes show expression of ENaC a, b, and g subunit transcripts in the gills and kidneys (Fig. 4), which are important sites of ion regulation in fishes [33, 34]. Furthermore, ENaC a subunit mRNA levels were positively correlated with plasma aldosterone in the West African lungfish Protopterus annectins, 12 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . suggesting an important role for aldosterone and ENaC in electrolyte homeostasis in lobe-finned fishes [33]. Why do ENaC subunits have polybasic tracts? Integral membrane proteins including ENaC are further processed and sorted in the Golgi apparatus for transport to their destinations. Furin and other proprotein convertases are present in various parts of the Golgi complex and cleave after polybasic motifs where Arg is preferred in the last position [55]. Selective double cleavage strongly increases ENaC currents, providing the functional change required for fitness-based selection. ENaC processing through the Golgi apparatus, where furin or other proprotein convertases are present, may have led to the emergence of polybasic tracts. Furin homologs are present in Drosophila, consistent with furin emerging well before ENaC subunits, and before the divergence of deuterostomes and protostomes [56, 57]. Prostasin, also known as PRSS8, cleaves the distal site of the g subunit in mammals. Prostasin may have emerged in amphibians [58] as the most prostasin-related proteins in non-tetrapods are more closely related to other mammalian proteases (PRSS22 and PRSS27). Notably, the last residue of the g subunit distal site switched from Arg in amphibians to Lys in other tetrapods (Table 1, g – site 2). Furin has a requirement Arg in the last position [59] while prostasin requires either Arg or Lys in the last position [60]. This Arg to Lys switch prevents furin cleavage in the trans-Golgi network, and allows aldosterone- inducible prostasin to cleave and activate the channel at the cell surface. The evolution of ENaC GRIP domain cleavage sites contrasts with the divergent evolution of PY motifs in the C-termini, which appear in all extant a, b, and g subunits, but was lost in d subunits that first appeared in a coelacanth ancestor. The presence of both ENaC subunit PY motifs and Nedd4-like proteins in ancestors of jawless fishes [61] suggests that aldosterone signaling co-opted Nedd4-2 dependent regulation. Differences in physiologic roles may underlie loss of the PY motif in ENaC d subunits. ENaC d subunit tissue distribution aligns poorly with aldosterone sensitivity, and is different than for the other ENaC subunits [3]. Despite the lack of PY motifs, d subunits can be ubiquitylated by Nedd4-2, presumably through binding b or g subunit PY motifs [62]. This is possible where subunit expression overlaps (e.g. lung and esophagus), but there are neuronal and reproductive tissues where only the d subunit has been detected. In summary, we found that two forms of ENaC regulation modulated by aldosterone signaling emerged through distinct evolutionary patterns. Channel activation by double cleavage of the a and g subunits coevolved with the terrestrial migration of vertebrates, closely paralleling that of aldosterone synthase and likely as a mechanism to enhance Na+ conservation and fluid retention.

Regulation of cell- surface stability through PY-motif interactions with Nedd4-2 was likely present in Cambrian period ancestral ENaC subunits and preceded the development of aldosterone signaling. Materials and Methods Sequence retrieval, alignment, phylogenetic tree calculation Human ENaC subunit protein sequences were used as query sequences in BLAST searches of databases (see Supplementary Table 1). Organism restrictions and organism specific databases were used to identify proteins in each of the major vertebrate classes. Alignment using MUSCLE 3.8.31 [63], curation using Gblocks 0.91b [64], and phylogenetic tree calculation using PhyML [65] were performed using Phylogeny.fr [66]. The final tree was calculated using the smaller final blocks option in Gblocks and the bootstrapping procedure option (100 bootstraps) in PhyML. Residue frequency in the aligned sequences were visualized using WebLogo [67]. Trees were visualized using FigTree (http://tree.bio.ed.ac.uk/software/figtree/). Motif analysis Furin cleaves human a ENaC after RSRR178 and RRAR204, and human g ENaC at RKRR138 [7]. Furin requires Arg at P1 and has a preferred P4-P3-P2-P1↓ = R-X-R/K-R↓ substrate sequence, although deviations at P2 and P4 have been observed, e.g. Ala-203 at P2 in the ENaC a subunit [68, 69]. Prostasin cleaves the human ENaC g subunit after RKRK181 [8]. Prostasin cleavage requires Arg or Lys at P1, and prefers basic residues at P2, P3, and P4 [60]. We inspected the region in the alignment that contains the 13 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . human ENaC cleavage sites (Supplementary Fig. 1) for ideal furin sites or alternatively, for polybasic tracts ending in Arg that aligned within five residues of either of the a subunit furin sites for a and d subunits, the g subunit furin site for b and g subunits, or for any of the furin sites for ASIC subunits. We also inspected the alignment for polybasic tracts ending in Arg or Lys that aligned within five residues of the human g subunit prostasin site for b and g subunits. For the PY motif (L/P-P-X-Y) required for Nedd4- 2 dependent regulation [70], we inspected the C-terminal region of the alignment. All sequences found aligned with human ENaC subunit PY motifs in the C-termini. Testing phylogenetic models of site gain and loss BayesTraits V3 [31, 32] was used to compare evolutionary models of two binary traits using maximum likelihood methods, as described above. Traits were assigned as indicated in Tables 1 and 2: absent, present, or ambiguous (in cases where the relevant region was missing).

ASICs in our dataset were all assigned "marine", and only coelacanth ASIC2 was assigned as having a polybasic tract at site 2 (Supplementary Fig. 1). BayesTraits run parameters are provided in Supplementary Table 2. Nested models were compared using a likelihood ratio test, with degrees of freedom equal to the difference in the number of parameters for each model. The likelihood ratio statistic is calculated as: likelihood ratio = 2×{log-likelihood(dependent model) – log-likelihood(independent model)}, and was converted to a p- value using the chisq.dist.rt function of Excel. BayesTraits was also used to determine whether traits likely appeared independently (convergent model), or derived from a common ancestor (divergent model). Program parameters are provided in Supplementary Table 2. Values of a given trait at all key ancestral nodes were determined while trait values at specific nodes were fixed to reflect each of the hypothetical evolutionary models. Maximum likelihood runs of each model were used to determine likely average values for model parameters. Uniform priors were selected for parameters with large expected values, and exponentially distributed priors with a mean of 0.001 were selected for parameters with small expected values. MCMC runs for each model were performed for the default number of iterations (1,010,000), and generated model parameters with means similar to those from maximum likelihood runs, consistent with convergence. The stepping stone sampler (100 stones with 10,000 iterations) was used to calculate log marginal likelihood values from MCMC runs. Log marginal likelihood values were converted to log Bayes Factors {log Bayes Factors = 2×(log marginal likelihood model 1 – log marginal likelihood model 2)} for model comparisons. Values of log Bayes Factors were used as evidence for model preference, where <2 was weak evidence, >2 was positive evidence, 5-10 was strong evidence, and >10 was very strong evidence. Tissue distribution of ENaC subunit transcripts All animals were handled according to approved institutional animal care and use committee (IACUC) protocols of the University of Pittsburgh. Total RNA was extracted from various tissues isolated from Erpetoichthys calabaricus and Xenopus laevis using TRIzol Reagent (Invitrogen) and denatured by heating at 70 ºC for 10 min. Using 2 µg of RNA isolated from each tissue, cDNA libraries were synthesized using RevertAid First Strand cDNA Synthesis Kit (ThermoFisher). Specific primers for the frog ENaC a, b and g subunits, each of the ropefish ENaC subunits, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as an internal standard (Supplementary Table 3) were designed using the NCBI Primer-Blast tool (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). Primers for frog d ENaC subunits and b-actin were previously described [25]. All primers were custom synthesized (Integrated DNA Technologies). PCR was run for 30 cycles using GoTaq® G2 Green master mix (Promega) and specific primers, with the annealing temperature set at 60 ºC.

PCR products were visualized after agarose gel electrophoresis using GelRed® stain (Sigma) and a GelDoc imaging system (Bio-Rad). Plasmids and site-direct mutagenesis cDNAs encoding mouse ENaC subunits were previously described [28, 71, 72]. Australian lungfish g subunit with a C-terminal epitope tag (Ag) was synthesized by Twist Bioscience (San Francisco, CA) and cloned into pcDNA 3.1 hygro (+). Site-directed mutagenesis of Ag was performed using the QuikChange 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . II XL Site-directed Mutagenesis Kit (Agilent, Santa Clara, CA). cRNAs were transcribed using mMessage mMachine Transcription Kits (Invitrogen), and purified using the RNeasy MiniElute Cleanup Kit (Qiagen). ENaC expression in Xenopus oocytes Oocytes were harvested from Xenopus laevis following a protocol approved by the University of Pittsburgh Institutional Animal Care and Use Committee, as previously described [73]. Stage V-VI oocytes were injected with 4 ng of cRNA per ENaC subunit, as indicated. Injected oocytes were maintained in modified Barth's saline (MBS: 88 mM NaCl, 1 mM KCl, 2.4 mM NaHCO3, 0.3 mm Ca(NO3)2, 0.41 mM CaCl2, 0.82 mM MgSO4, and 15 mM HEPES, pH 7.4) supplemented with 10 µg/ml sodium penicillin, 10 μg/ml streptomycin sulfate, and 100 μg/ml gentamycin sulfate at 18 °C for 24h. Current measurement by two-electrode voltage clamp Oocytes were mounted in a continuously perfused 20 µL recording chamber and impaled by two 0.1-1 MΩ recording pipettes filled with 3 M KCl. Voltage was clamped at -100 mV and currents were continuously recorded using an Axoclamp 900A voltage clamp amplifier (Molecular Devices, Sunnyvale, CA) and pClamp 10.5 software (Molecular Devices). Baseline currents were measured in Na-110 buffer (110 mM NaCl, 2 mM KCl, 2 mM CaCl2, and 10 mM HEPES, pH 7.4). Amiloride (Na-110 supplemented with 100 µM amiloride) was added at the end of each experiment to block ENaC and determine the ENaC-mediated portion of the current. Statistical analysis of current data was performed using Prism 9 (GraphPad Software). Surface biotinylation and western blotting One day after injection, oocytes were transferred to a 12-well dish and incubated in MBS on ice for 30 min. After washing twice, surface proteins were labeled with biotin by adding 1 mg/ml membrane impermeant EZ-LinkTM Sulfo-NHS-SS-Biotin (Thermo Fisher) in 137 mM NaCl and 15 mM sodium borate, pH 9.0 for 30 min. Excess reagent was quenched with MBS supplemented with 192 mM glycine. After two washes with MBS, approximately 40 oocytes were lysed in 500 µl detergent solution (100 mm NaCl, 40 mm KCl, 1 mm EDTA, 10% glycerol, 1% NP-40, 0.4% deoxycholate, 20 mm HEPES, pH 7.4, supplemented with protease inhibitor mixture III (Calbiochem)).

After reserving 2% of the whole cell lysate, the remainder was incubated with NeutrAvidin agarose (Thermo Fisher) overnight at 4 °C on a rocker to isolate biotinylated proteins. Biotinylated proteins were eluted by boiling in 2x Laemmli buffer (Bio-Rad). Biotinylated proteins representing the total and surface pools were separated by SDS-PAGE (4-15% Tris/glycine, Biorad) and blotted for the g subunit using anti-HA-Peroxidase antibodies (1:2000, Sigma) or GAPDH using anti-GAPDH-Peroxidase antibodies (1:10000, Proteintech). Band densities were measured using a ChemiDoc Imaging system (Bio-Rad). Statistical analyses were performed using Prism 9. Acknowledgements This work was supported by NIDDK, National Institutes of Health, Grant R01 DK125439 (to O.B.K). The Pittsburgh Center for Kidney Research was supported by P30DK079307 from NIDDK, and the Pittsburgh Liver Research Center was supported by P30DK120531 from NIDDK. Competing Interests The authors declare no conflicts of interest. References 1. fish-tetrapod transition. Physiol Biochem Zool. 2004;77(5):700-19. Epub 2004/11/18. doi: 10.1086/425183. PubMed PMID: 15547790. Long JA, Gordon MS. The greatest step in vertebrate history: a paleobiological review of the 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . 2. the story of our internal environment revisited. Physiol Rev. 2015;95(1):297-340. Epub 2014/12/30. doi: 10.1152/physrev.00011.2014. PubMed PMID: 25540145. 3. subunit: new notes for an old song. Am J Physiol Renal Physiol. 2012;303(3):F328-38. Epub 2012/05/11. doi: 10.1152/ajprenal.00116.2012. PubMed PMID: 22573384. 4. resolution and low pH. Nature. 2007;449(7160):316-23. Epub 2007/09/21. doi: nature06163 [pii] 10.1038/nature06163. PubMed PMID: 17882215. 5. sodium channel by cryo-electron microscopy. Elife. 2018;7. Epub 2018/09/27. doi: 10.7554/eLife.39340. PubMed PMID: 30251954. 6. based, homology model of the extracellular domain of the epithelial Na+ channel a subunit reveals a mechanism of channel activation by proteases. J Biol Chem. 2011;286(1):649-60. Epub 2010/10/27. doi: M110.167098 [pii] 10.1074/jbc.M110.167098. PubMed PMID: 20974852; PubMed Central PMCID: PMC3013024. 7. sodium channels are activated by furin-dependent proteolysis. J Biol Chem. 2004;279(18):18111-4. PubMed PMID: 15007080. 8. channels are fully activated by furin- and prostasin-dependent release of an inhibitory peptide from the g-subunit. J Biol Chem. 2007;282(9):6153-60. doi: 10.1074/jbc.M610636200. PubMed PMID: 17199078. 9. activation, and inhibition in epithelial sodium channel.

Elife. 2020;9. Epub 2020/07/31. doi: 10.7554/eLife.59038. PubMed PMID: 32729833; PubMed Central PMCID: PMCPMC7413742. 10. and Intracellular Factors. Annu Rev Physiol. 2018;80:263-81. Epub 2017/11/10. doi: 10.1146/annurev- physiol-021317-121143. PubMed PMID: 29120692; PubMed Central PMCID: PMCPMC5811403. 11. effects of dietary sodium. Am J Physiol Renal Physiol. 2009;297(5):F1249-55. Epub 2009/09/11. doi: 10.1152/ajprenal.00401.2009. PubMed PMID: 19741015; PubMed Central PMCID: PMCPMC2781327. Frindt G, Palmer LG. Acute effects of aldosterone on the epithelial Na channel in rat kidney. Am 12. J Physiol Renal Physiol. 2015;308(6):F572-8. Epub 2014/12/19. doi: 10.1152/ajprenal.00585.2014. PubMed PMID: 25520012; PubMed Central PMCID: PMCPMC4360037. 13. and Indirect Mineralocorticoid Effects Determine Distal Salt Transport. Journal of the American Society of Nephrology : JASN. 2016;27(8):2436-45. Epub 2015/12/30. doi: 10.1681/ASN.2015070815. PubMed PMID: 26712527; PubMed Central PMCID: PMCPMC4978056. 14. which Liddle's syndrome mutations increase activity of a human epithelial Na+ channel. Cell. 1995;83(6):969-78. PubMed PMID: 8521520. 15. Schild L, Lu Y, Gautschi I, Schneeberger E, Lifton RP, Rossier BC. Identification of a PY motif in the epithelial Na channel subunits as a target sequence for mutations causing channel activation found in Liddle syndrome. Embo J. 1996;15(10):2381-7. PubMed PMID: 8665845. 16. to the proline-rich PY motifs in the epithelial Na+ channel deleted in Liddle's syndrome. Embo J. 1996;15(10):2371-80. PubMed PMID: 8665844. 17. 2015;24(2):111-6. Epub 2015/01/21. doi: 10.1097/MNH.0000000000000097. PubMed PMID: 25602517. Staub O, Dho S, Henry P, Correa J, Ishikawa T, McGlade J, et al. WW domains of Nedd4 bind Terker AS, Yarbrough B, Ferdaus MZ, Lazelle RA, Erspamer KJ, Meermeier NP, et al. Direct Kleyman TR, Kashlan OB, Hughey RP. Epithelial Na(+) Channel Regulation by Extracellular Frindt G, Palmer LG. Surface expression of sodium channels and transporters in rat kidney: Rizzo F, Staub O. NEDD4-2 and salt-sensitive hypertension. Curr Opin Nephrol Hypertens. Snyder PM, Price MP, McDonald FJ, Adams CM, Volk KA, Zeiher BG, et al. Mechanism by 16 Rossier BC, Baker ME, Studer RA. Epithelial sodium transport and its control by aldosterone: Giraldez T, Rojas P, Jou J, Flores C, Alvarez de la Rosa D. The epithelial sodium channel delta- Jasti J, Furukawa H, Gonzales EB, Gouaux E. Structure of acid-sensing ion channel 1 at 1.9 Å Noreng S, Bharadwaj A, Posert R, Yoshioka C, Baconguis I. Structure of the human epithelial Kashlan OB, Adelman JL, Okumura S, Blobner BM, Zuzek Z, Hughey RP, et al. Constraint- Hughey RP, Bruns JB, Kinlough CL, Harkleroad KL, Tong Q, Carattino MD, et al. Epithelial Bruns JB, Carattino MD, Sheng S, Maarouf AB, Weisz OA, Pilewski JM, et al. Epithelial Na+ Noreng S, Posert R, Bharadwaj A, Houser A, Baconguis I. Molecular principles of assembly, 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . 18. revised: analysis of six nuclear genes from all major rodent clades. BMC Evol Biol. 2009;9:71. Epub 2009/04/04. doi: 10.1186/1471-2148-9-71. PubMed PMID: 19341461; PubMed Central PMCID: PMCPMC2674048. 19. diversification rates. PloS one. 2017;12(8):e0183070. Epub 2017/08/17. doi: 10.1371/journal.pone.0183070. PubMed PMID: 28813483; PubMed Central PMCID: PMCPMC5559066. 20. family member. Am J Physiol Renal Physiol. 2011;301(4):F684-96. Epub 2011/07/15. doi: ajprenal.00259.2011 [pii] 10.1152/ajprenal.00259.2011. PubMed PMID: 21753073; PubMed Central PMCID: PMC3191808. 21. Edgar RC. MUSCLE: a multiple sequence alignment method with reduced time and space complexity. BMC Bioinformatics. 2004;5:113. Epub 2004/08/21. doi: 10.1186/1471-2105-5-113. PubMed PMID: 15318951; PubMed Central PMCID: PMCPMC517706. 22. channel and the sodium pump as limiting factors of aldosterone action on sodium transport. Physiol Genomics. 2011;43(13):844-54. Epub 2011/05/12. doi: 10.1152/physiolgenomics.00002.2011. PubMed PMID: 21558422. 23. g subunit has a dominant role in channel activation. J Biol Chem. 2008;283(37):25290-5. Epub 2008/07/25. doi: M803931200 [pii] 10.1074/jbc.M803931200. PubMed PMID: 18650438; PubMed Central PMCID: PMC2533072. 24. Balchak DM, Thompson RN, Kashlan OB. The epithelial Na(+) channel gamma subunit autoinhibitory tract suppresses channel activity by binding the gamma subunit's finger-thumb domain interface. J Biol Chem. 2018;293(42):16217-25. Epub 2018/08/23. doi: 10.1074/jbc.RA118.004362. PubMed PMID: 30131333; PubMed Central PMCID: PMCPMC6200929. 25. Wichmann L, Vowinkel KS, Perniss A, Manzini I, Althaus M. Incorporation of the delta-subunit into the epithelial sodium channel (ENaC) generates protease-resistant ENaCs in Xenopus laevis. J Biol Chem. 2018;293(18):6647-58. Epub 2018/03/27. doi: 10.1074/jbc.RA118.002543. PubMed PMID: 29576549; PubMed Central PMCID: PMCPMC5936831. 26. channel (ENaC) enhances channel activity and alters proteolytic ENaC activation. J Biol Chem. 2009;284(42):29024-40. Epub 2009/09/01. doi: 10.1074/jbc.M109.018945. PubMed PMID: 19717556; PubMed Central PMCID: PMCPMC2781449. 27. an inhibitory domain in the a-subunit of the epithelial sodium channel. Am J Physiol Renal Physiol. 2008;294(1):F47-52. Epub 2007/11/23. doi: 00399.2007 [pii] 10.1152/ajprenal.00399.2007. PubMed PMID: 18032549. 28. by clathrin-mediated endocytosis. J Biol Chem. 1997;272(41):25537-41. PubMed PMID: 9325269. 29. are expressed at the plasma membrane. J Biol Chem. 2004;279(47):48491-4. PubMed PMID: 15466477. 30. epithelial Na+ channel by relieving Na+ self-inhibition. Am J Physiol Renal Physiol. 2006;290(6):F1488- 96. Epub 2006/02/02. doi: 00439.2005 [pii] 10.1152/ajprenal.00439.2005. PubMed PMID: 16449353. 31. 32. Syst Biol. 2004;53(5):673-84. Epub 2004/11/17. doi: 10.1080/10635150490522232. PubMed PMID: 15545248.

Pagel M, Meade A. BayesTraits V3. http://www.evolution.rdg.ac.uk2016. Pagel M, Meade A, Barker D. Bayesian estimation of ancestral character states on phylogenies. Shimkets RA, Lifton RP, Canessa CM. The activity of the epithelial sodium channel is regulated Carattino MD, Passero CJ, Steren CA, Maarouf AB, Pilewski JM, Myerburg MM, et al. Defining Hughey RP, Bruns JB, Kinlough CL, Kleyman TR. Distinct pools of epithelial sodium channels Sheng S, Carattino MD, Bruns JB, Hughey RP, Kleyman TR. Furin cleavage activates the Haerteis S, Krueger B, Korbmacher C, Rauh R. The delta-subunit of the epithelial sodium 17 Blanga-Kanfi S, Miranda H, Penn O, Pupko T, DeBry RW, Huchon D. Rodent phylogeny Steppan SJ, Schenk JJ. Muroid rodent phylogenetics: 900-species tree reveals increasing Kashlan OB, Kleyman TR. ENaC structure and function in the wake of a resolved structure of a Studer RA, Person E, Robinson-Rechavi M, Rossier BC. Evolution of the epithelial sodium Carattino MD, Hughey RP, Kleyman TR. Proteolytic processing of the epithelial sodium channel 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . 33. channel and its role in osmoregulation of aquatic and estivating African lungfish Protopterus annectens. Comp Biochem Physiol A Mol Integr Physiol. 2015;183:1-8. Epub 2014/12/30. doi: 10.1016/j.cbpa.2014.12.028. PubMed PMID: 25541184. 34. in the Australian lungfish, Neoceratodus forsteri (Osteichthyes: Dipnoi). Proc Biol Sci. 2012;279(1748):4795-802. Epub 2012/10/12. doi: 10.1098/rspb.2012.1945. PubMed PMID: 23055064; PubMed Central PMCID: PMCPMC3497098. 35. GE, editor. Respiratory Physiology of Vertebrates: Life With and Without Oxygen2010. p. 174-221. 36. Waldmann R, Champigny G, Bassilana F, Voilley N, Lazdunski M. Molecular cloning and functional expression of a novel amiloride-sensitive Na+ channel. J Biol Chem. 1995;270(46):27411-4. PubMed PMID: 7499195. 37. 2012;3:212. Epub 2012/06/28. doi: 10.3389/fphys.2012.00212. PubMed PMID: 22737130; PubMed Central PMCID: PMCPMC3380336. 38. Hypertens (Greenwich). 2017;19(5):524-9. Epub 2016/11/30. doi: 10.1111/jch.12949. PubMed PMID: 27896928. 39. channel is inhibited by a peptide derived from proteolytic processing of its a subunit. J Biol Chem. 2006;281(27):18901-7. doi: 10.1074/jbc.M604109200. PubMed PMID: 16690613. 40. extant coelacanth lung during ontogenetic development. Nat Commun. 2015;6:8222. Epub 2015/09/16. doi: 10.1038/ncomms9222. PubMed PMID: 26372119; PubMed Central PMCID: PMCPMC4647851. 41. lung of the fossil coelacanth Axelrodichthys araripensis (Actinistia: Mawsoniidae).

Palaeontology. 2010;53(6):1281-90. doi: 10.1111/j.1475-4983.2010.01015.x. 42. phylogeny of ray-finned fishes (Actinopterygii) based on transcriptomic and genomic data. Proc Natl Acad Sci U S A. 2018;115(24):6249-54. Epub 2018/05/16. doi: 10.1073/pnas.1719358115. PubMed PMID: 29760103. 43. Evans DH, Piermarini PM, Choe KP. The multifunctional fish gill: dominant site of gas exchange, osmoregulation, acid-base regulation, and excretion of nitrogenous waste. Physiol Rev. 2005;85(1):97-177. Epub 2004/12/25. doi: 10.1152/physrev.00050.2003. PubMed PMID: 15618479. 44. involved in epithelial Na+ uptake in the rainbow trout Oncorhynchus mykiss. Am J Physiol Cell Physiol. 2014;307(3):C255-65. Epub 2014/06/06. doi: 10.1152/ajpcell.00398.2013. PubMed PMID: 24898589; PubMed Central PMCID: PMCPMC4121585. 45. fibrosis: pathophysiology and therapeutic targets. Thorax. 2016;71(3):284-7. doi: 10.1136/thoraxjnl- 2015-207588. PubMed PMID: 26719229. 46. Hummler E, Barker P, Gatzy J, Beermann F, Verdumo C, Schmidt A, et al. Early death due to defective neonatal lung liquid clearance in alpha-ENaC-deficient mice. Nat Genet. 1996;12(3):325-8. PubMed PMID: 8589728. 47. epithelial sodium channel associated with atypical cystic fibrosis increases channel open probability and reduces Na+ self inhibition. The Journal of physiology. 2010;588(Pt 8):1211-25. Epub 2010/03/03. doi: jphysiol.2009.180224 [pii] 10.1113/jphysiol.2009.180224. PubMed PMID: 20194130; PubMed Central PMCID: PMC2872728. Rauh R, Diakov A, Tzschoppe A, Korbmacher J, Azad AK, Cuppens H, et al. A mutation of the Haq IJ, Gray MA, Garnett JP, Ward C, Brodlie M. Airway surface liquid homeostasis in cystic Dymowska AK, Schultz AG, Blair SD, Chamot D, Goss GG. Acid-sensing ion channels are Hughes LC, Orti G, Huang Y, Sun Y, Baldwin CC, Thompson AW, et al. Comprehensive Brito PM, Meunier FJ, Clément G, Geffard-Kuriyama D. The histological structure of the calcified 18 Uchiyama M, Konno N, Shibuya S, Nogami S. Cloning and expression of the epithelial sodium Uchiyama M, Maejima S, Yoshie S, Kubo Y, Konno N, Joss JM. The epithelial sodium channel Graham JB, Wegner NC. Breathing air in water and in air: The air-breathing fishes. In: Nilsson Rotin D, Staub O. Nedd4-2 and the regulation of epithelial sodium transport. Front Physiol. Cui Y, Tong A, Jiang J, Wang F, Li C. Liddle syndrome: clinical and genetic profiles. J Clin Carattino MD, Sheng S, Bruns JB, Pilewski JM, Hughey RP, Kleyman TR. The epithelial Na+ Cupello C, Brito PM, Herbin M, Meunier FJ, Janvier P, Dutel H, et al. Allometric growth in the 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license .

48. Schaedel C, Marthinsen L, Kristoffersson AC, Kornfalt R, Nilsson KO, Orlenius B, et al. Lung symptoms in pseudohypoaldosteronism type 1 are associated with deficiency of the alpha-subunit of the epithelial sodium channel. J Pediatr. 1999;135(6):739-45. PubMed PMID: 10586178. 49. Reihill JA, Walker B, Hamilton RA, Ferguson TE, Elborn JS, Stutts MJ, et al. Inhibition of Protease-Epithelial Sodium Channel Signaling Improves Mucociliary Function in Cystic Fibrosis Airways. Am J Respir Crit Care Med. 2016;194(6):701-10. Epub 2016/03/26. doi: 10.1164/rccm.201511-2216OC. PubMed PMID: 27014936. 50. Myerburg MM, Harvey PR, Heidrich EM, Pilewski JM, Butterworth MB. Acute regulation of the epithelial sodium channel in airway epithelia by proteases and trafficking. Am J Respir Cell Mol Biol. 2010;43(6):712-9. Epub 2010/01/26. doi: 10.1165/rcmb.2009-0348OC. PubMed PMID: 20097829; PubMed Central PMCID: PMCPMC2993091. 51. Stokes JB, Sigmund RD. Regulation of rENaC mRNA by dietary NaCl and steroids: organ, tissue, and steroid heterogeneity. Am J Physiol. 1998;274(6 Pt 1):C1699-707. PubMed PMID: 9611136. 52. Wright PA, Turko AJ. Amphibious fishes: evolution and phenotypic plasticity. J Exp Biol. 2016;219(Pt 15):2245-59. Epub 2016/08/05. doi: 10.1242/jeb.126649. PubMed PMID: 27489213. 53. Australian lungfish, Neoceratodus-Forsteri. J Comp Physiol B. 1994;164:378-82. 54. glucocorticoid signaling in the osmoregulation and 'central' actions of mineralocorticoid receptor. Gen Comp Endocrinol. 2013;181:223-8. Epub 2012/12/12. doi: 10.1016/j.ygcen.2012.11.016. PubMed PMID: 23220000. 55. Rev Drug Discov. 2012;11(5):367-83. Epub 2012/06/12. doi: 10.1038/nrd3699. PubMed PMID: 22679642. 56. Roebroek AJ, Pauli IG, Zhang Y, van de Ven WJ. cDNA sequence of a Drosophila melanogaster gene, Dfur1, encoding a protein structurally related to the subtilisin-like proprotein processing enzyme furin. FEBS Lett. 1991;289(2):133-7. Epub 1991/09/09. doi: 10.1016/0014- 5793(91)81052-a. PubMed PMID: 1915835. 57. genes of the proprotein convertase family, including aPC6C, a marker of epithelial fusions during embryology. Int J Biol Sci. 2006;2(3):125-32. Epub 2006/06/10. doi: 10.7150/ijbs.2.125. PubMed PMID: 16763672; PubMed Central PMCID: PMCPMC1474147. 58. Bhagwandin VJ, Hau LW, Mallen-St Clair J, Wolters PJ, Caughey GH. Structure and activity of human pancreasin, a novel tryptic serine peptidase expressed primarily by the pancreas. J Biol Chem. 2003;278(5):3363-71. Epub 2002/11/21. doi: 10.1074/jbc.M209353200. PubMed PMID: 12441343. 59. structure of the proprotein processing proteinase furin explains its stringent specificity. Nat Struct Biol. 2003;10(7):520-6. Epub 2003/06/10. doi: 10.1038/nsb941. PubMed PMID: 12794637. 60. characterization of prostasin, a channel activating protease. Biochem Biophys Res Commun. 2004;324(2):953-63. PubMed PMID: 15474520. 61. implicated in diverse cellular functions. Trends Cell Biol. 1999;9(5):166-9. Epub 1999/05/14. PubMed PMID: 10322449. 62. epithelial sodium channel (ENaC) by ubiquitination and Nedd8.

J Cell Physiol. 2013;228(11):2190-201. Epub 2013/04/17. doi: 10.1002/jcp.24390. PubMed PMID: 23589227. 63. 2010;38(7):2145-53. Epub 2010/01/06. doi: 10.1093/nar/gkp1196. PubMed PMID: 20047958; PubMed Central PMCID: PMCPMC2853116. Bertrand S, Camasses A, Paris M, Holland ND, Escriva H. Phylogenetic analysis of Amphioxus Henrich S, Cameron A, Bourenkov GP, Kiefersauer R, Huber R, Lindberg I, et al. The crystal Ly K, McIntosh CJ, Biasio W, Liu Y, Ke Y, Olson DR, et al. Regulation of the delta and alpha Harvey KF, Kumar S. Nedd4-like proteins: an emerging family of ubiquitin-protein ligases Edgar RC. Quality measures for protein alignment benchmarks. Nucleic Acids Res. Shipway A, Danahay H, Williams JA, Tully DC, Backes BJ, Harris JL. Biochemical 19 Joss JMP, Arnold-Reed DE, Balment RJ. The steroidogenic response to angiotensin-II in the Takahashi H, Sakamoto T. The role of 'mineralocorticoids' in teleost fish: relative importance of Seidah NG, Prat A. The biology and therapeutic targeting of the proprotein convertases. Nat 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. available under a CC-BY 4.0 International license . 64. Talavera G, Castresana J. Improvement of phylogenies after removing divergent and ambiguously aligned blocks from protein sequence alignments. Syst Biol. 2007;56(4):564-77. Epub 2007/07/27. doi: 10.1080/10635150701472164. PubMed PMID: 17654362. 65. methods to estimate maximum-likelihood phylogenies: assessing the performance of PhyML 3.0. Syst Biol. 2010;59(3):307-21. Epub 2010/06/09. doi: 10.1093/sysbio/syq010. PubMed PMID: 20525638. Dereeper A, Guignon V, Blanc G, Audic S, Buffet S, Chevenet F, et al. Phylogeny.fr: robust 66. phylogenetic analysis for the non-specialist. Nucleic Acids Res. 2008;36(Web Server issue):W465-9. Epub 2008/04/22. doi: 10.1093/nar/gkn180. PubMed PMID: 18424797; PubMed Central PMCID: PMCPMC2447785. Crooks GE, Hon G, Chandonia JM, Brenner SE. WebLogo: a sequence logo generator. 67. Genome Res. 2004;14(6):1188-90. Epub 2004/06/03. doi: 10.1101/gr.849004. PubMed PMID: 15173120; PubMed Central PMCID: PMCPMC419797. 68. Des Sel. 2004;17(1):107-12. Epub 2004/02/27. doi: 10.1093/protein/gzh013. PubMed PMID: 14985543. 69. substrates and their associated drugs. Int J Mol Sci. 2011;12(2):1060-5. Epub 2011/05/05. doi: 10.3390/ijms12021060. PubMed PMID: 21541042; PubMed Central PMCID: PMCPMC3083689. Persaud A, Alberts P, Amsen EM, Xiong X, Wasmuth J, Saadon Z, et al. Comparison of 70. substrate specificity of the ubiquitin ligases Nedd4 and Nedd4-2 using proteome arrays. Mol Syst Biol. 2009;5:333. Epub 2009/12/03. doi: 10.1038/msb.2009.85. PubMed PMID: 19953087; PubMed Central PMCID: PMCPMC2824488.

71. Passero CJ, Carattino MD, Kashlan OB, Myerburg MM, Hughey RP, Kleyman TR. Defining an inhibitory domain in the gamma subunit of the epithelial sodium channel. Am J Physiol Renal Physiol. 2010;299(4):F854-61. doi: 10.1152/ajprenal.00316.2010. PubMed PMID: 20630937; PubMed Central PMCID: PMC2957262. 72. the epithelial Na+ channel involves proteolytic processing of the α- and ɣ-subunits. J Biol Chem. 2003;278(39):37073-82. PubMed PMID: 12871941. 73. Wang XP, Im SJ, Balchak DM, Montalbetti N, Carattino MD, Ray EC, et al. Murine epithelial sodium (Na(+)) channel regulation by biliary factors. J Biol Chem. 2019;294(26):10182-93. Epub 2019/05/17. doi: 10.1074/jbc.RA119.007394. PubMed PMID: 31092599; PubMed Central PMCID: PMCPMC6664190. Hughey RP, Mueller GM, Bruns JB, Kinlough CL, Poland PA, Harkleroad KL, et al. Maturation of Guindon S, Dufayard JF, Lefort V, Anisimova M, Hordijk W, Gascuel O. New algorithms and Duckert P, Brunak S, Blom N. Prediction of proprotein convertase cleavage sites. Protein Eng Tian S, Huang Q, Fang Y, Wu J. FurinDB: A database of 20-residue furin cleavage site motifs, 20 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY 4.0 International license . bioRxiv preprint doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY 4.0 International license . bioRxiv preprint doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY 4.0 International license . bioRxiv preprint doi: https://doi.org/10.1101/2020.10.19.344341 ; this version posted December 14, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY 4.0 International license .

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Catestatin induces glycogenesis by stimulating phosphoinositide 3-kinase-AKT pathway Gautam Bandyopadhyay2*, Kechun Tang1*, Nicholas J.G. Webster1,2, Geert van den Bogaart3,4, and Sushil K. Mahata1,2¶ 1VA San Diego Healthcare System, 3350 La Jolla Village Drive, San Diego, CA, USA; 2Department of Medicine, University of California San Diego, La Jolla, CA, USA; 3Department of Tumor Immunology, Radboud Institute for Molecular Life Sciences, Radboud University Medical Center, Nijmegen, the Netherlands; 4Department of Molecular Immunology and Microbiology, Groningen Biomolecular Sciences and Biotechnology Institute, University of Groningen, Groningen, the Netherlands. Contributed equally to this work. Keywords: Catestatin; Phospho-AKT, Hyperglycemia; Glycogen; Glucose-6-phosphate; Glycogen synthase Short title: Catestatin diverts gluconeogenic substrates to the glycogenic pathway ¶Correspondence should be addressed to: Sushil K. Mahata, Ph.D. Metabolic Physiology & Ultrastructural Biology Laboratory Department of Medicine University of California San Diego 9500 Gilman Drive La Jolla, CA 92093-0732 Email: [email protected] (ORCID: 0000-0002-9154-0787) 1 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Abstract Aim: Defects in hepatic glycogen synthesis contribute to postprandial hyperglycemia in type 2 diabetic (T2D) patients. Chromogranin A (CgA) peptide Catestatin (CST: hCgA352-372) has been shown to improve glucose tolerance in insulin-resistant mice. Here, we seek to determine whether CST also reduces hyperglycemia by increasing hepatic glycogen synthesis. Methods: We determined liver glycogen, glucose-6-phosphate (G6P), uridine diphosphate glucose (UDPG), and glycogen synthase (GYS2) activities; plasma insulin, glucagon, norepinephrine (NE), and epinephrine (EPI) levels in fed and fasted liver of lean and obese mice as well as in CST knockout (CST-KO) mice after treatments with saline, CST, or insulin. We also determined glycogen synthesis and glycogenolysis in primary hepatocytes. In addition, we analyzed phosphorylation signals of Insulin receptor (IR), insulin receptor substrate-1 (IRS-1), phosphatidylinositol dependent kinase-1 (PDK-1), GYS2, glycogen synthase kinase-3β (GSK-3β), AKT (an enzyme in AKR mouse that produces Thymoma)/PKB (protein kinase B) and mTOR (mammalian/mechanistic target of rapamycin) by immunoblotting.

Results: CST stimulated glycogen accumulation in fed and fasted liver and in primary hepatocytes. CST reduced plasma NE and EPI levels, suggesting that CST promotes glycogenesis by inhibiting catecholamine-induced glycogenolysis. CST also directly stimulated glycogenesis and inhibited NE and EPI-induced glycogenolysis in hepatocytes. CST elevated the levels of UDPG and increased GYS2 activity, thus redirecting G6P to the glycogenic pathway. CST-KO mice had decreased liver glycogen that was restored by treatment with CST, reinforcing the crucial role of CST in hepatic glycogenesis. CST can improve insulin signals downstream of insulin receptor IR and IRS-1 by enhancing phospho-AKT signals through stimulation of PDK-1 and mTORC2 (mTOR complex 2) activities. Conclusions: We conclude that CST directly promotes the glycogenic pathway and reduces plasma glucose levels in insulin-resistant mice by (i) reducing glucose production, (ii) increasing glycogen synthesis from UDPG, and (iii) reducing glycogenolysis. This is achieved by enhancing downstream insulin signaling. 2 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Introduction Obesity, a major risk factor for type 2 diabetes (T2D) 1, results in the development of insulin resistance and post-prandial hyperglycemia 2-4. In the United States, >50% of the population have either diabetes or prediabetes 5. Obesity-induced T2D is thus emerging as a global health problem threatening to reach pandemic levels by 2030 6,7. Chronic hyperglycemia causes glucotoxicity 8,9, which results in decreased secretion of insulin and increased insulin resistance 8,10. The liver plays a central role in maintenance of blood glucose homeostasis. In the postprandial state, the liver responds to an increase in portal vein blood glucose and insulin levels by absorbing glucose via glucose transporter 2 (GLUT2) and converting it to glycogen. The stimulation of net glycogen synthesis is a major, direct physiological function of postprandial insulin on hepatocytes 11. In the postabsorptive state, the liver produces glucose by glycogenolysis and gluconeogenesis 12. In the liver, glucose metabolism is determined largely by the glucose concentration in the portal vein (substrate supply) and is regulated by feed-forward activation mechanisms 13,14. Liver glycogen synthesis is impaired in T2D patients 14,15. Glycogen synthesis is also impaired in other organs, as for instance the reduced glucose uptake in skeletal muscle in T2D subjects is partly accounted for by reduced glycogen synthesis 16. Defects in the hepatic glycogen synthesis have been shown to contribute to postprandial hyperglycemia of patients with poorly controlled insulin-dependent diabetes mellitus 15,17,18 and non-insulin-dependent diabetic subjects 19,20.

Therefore, targeting hepatic glucose storage or production is a potential therapeutic strategy for T2D. Glucose homeostasis is controlled by several hormones, with insulin being the most important anabolic hormone and glucagon and catecholamines being the most catabolic hormones. Glucagon and catecholamines act by the following mechanisms: (i) inhibiting secretion of insulin via an a- adrenergic mechanism 21, (ii) stimulating glucagon secretion 22, (iii) stimulating hepatic glucose production (HGP) via both b- and a-adrenergic regulation of hepatic glycogenolysis and gluconeogenesis 23-26, (iv) mobilizing gluconeogenic substrates such as lactate, alanine, and glycerol from extrasplanchnic tissues to the liver 27, and (v) decreasing glucose clearance by direct inhibition of tissue glucose uptake 28,29. In addition, both glucagon and the catecholamines counteract insulin- induced hypoglycemia by stimulating glycogenolysis and gluconeogenesis and augmenting HGP 30,31. Chromogranin A (CgA), a ~49 kDa proprotein, gives rise to peptides with opposing regulatory effects 32-36. While pancreastatin (PST: hCgA250-301) is an anti-insulin and proinflammatory peptide 37,38, catestatin (CST: CgA352-372) is a proinsulin and anti-inflammatory peptide 39,40 and is a risk factor for cardiovascular diseases in patients undergoing dialysis 41. The lack of all CgA peptides make the CgA knockout mice (CgA-KO) more sensitive to insulin despite being obese 42. In addition, opposite 3 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. to wild-type (WT) mice, CgA-KO mice can maintain insulin sensitivity even after 4 months of high fat diet (HFD) 43. Lack of both CgA and Chromogranin B (CgB) also caused increased sensitivity to insulin with aging 44. In contrast, CST knockout (CST-KO) mice are obese and display insulin resistance and hyperinsulinemia even on normal chow diet (NCD) 40. We have recently shown that supplementation of CST-KO mice with CST not only restored insulin sensitivity but also normalized plasma insulin levels in CST-KO mice 40. In HFD-induced obese (DIO) mice, CST improved insulin sensitivity by attenuating inflammation, inhibiting infiltration of proinflammatory macrophages into the liver, and inhibiting gluconeogenesis 40. We have also shown that CST improves insulin sensitivity in DIO mice by attenuating HFD-induced endoplasmic reticulum stress 45. Another mechanism by which CST could potentially control hyperglycemia in insulin-resistant mice is by inducing synthesis of glycogen in the postabsorptive state. Since CST inhibits both catecholamine secretion 39,46-49 and gluconeogenesis (by inhibiting expression of glucose 6-phosphatase gene (G6pc)) 40, we reasoned that glucose-6-phosphate (G6P), the substrate for G6pc, would accumulate after CST treatment, leading to stimulation of glycogenesis.

To address this, we measured plasma insulin, glucagon, NE and EPI in CST-KO mice and DIO mice treated with CST and determined their effects on glycogen levels. This enabled us to determine whether the improved glucose tolerance that we reported earlier in DIO mice by CST 40, was due to CST’s interaction with pancreatic and adrenomedullary hormones leading to mobilization of G6P from the gluconeogenic to the glycogenic pathway. Therefore, in the present communication, we have tested the glycogenic effects of CST on genetically obese and insulin-resistant CST-KO mice as well as in insulin-resistant DIO mice and compared the effects of CST with insulin. 4 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Results CST induces hepatic glycogen synthesis in an insulin-independent pathway. One of the major pathways by which liver removes glucose after a meal is through insulin-induced conversion of glucose into glycogen (glycogenesis) 10,11. This is reflected by the presence of >3.5-fold more glycogen in fed liver compared to fasted liver (Fig. 1A), likely because insulin levels are reduced in fasted mice (Fig. 1B). Supplementation with insulin (30 minutes) restored glycogen level in fasted mice (Fig. 1A) Interestingly, like insulin, both acute (1 hour) and chronic (10-15 days) treatment with CST (2 µg/g body weight) raised glycogen levels in fasted mice (Fig. 1C). Moreover, CST treatment caused greater increase in glycogen levels compared to insulin even in fed mice (Fig. 1A&C). The combined treatment of insulin and CST elicited an additive effect on glycogen in fasted mice (Fig. 1D), suggesting that there may exist parallel pathways for CST and insulin actions. Ultrastructural changes of glycogen granules (GG) strengthen and support the biochemical findings of glycogen level (Fig. S1). GG granules in the liver were increased both in fed and fasted mice after CST treatment (Fig. S1 A-E). It was revealed that CST caused dose-dependent (10 ng to 25 µg/g body weight) increase in glycogen levels with an EC50 0.013 µg (Fig. S1F). The half-life of CST in plasma in fed mice was approximately 6.7 hours (Fig. S1G). CST elevates muscle glycogen content. The bulk of postprandial glucose uptake takes place in the skeletal muscle 50. While ~80% of the glucose that enters muscle fibers in response to insulin is converted to glycogen 51, the rest is being oxidized to provide energy for muscle function 16. Being an insulin-sensitizing peptide 40, CST increased glycogen content in gastrocnemius muscle of NCD-fed WT mice (Fig. 1E). Electron microscopy revealed three distinct intracellular pools of glycogen 52: (i) subsarcolemmal glycogen, just beneath the sarcolemma, (ii) intermyofibrillar glycogen, located beneath the myofibrils (Fig.

S2 A & B); and (iii) intramyofibrillar glycogen, in the myofibril, mainly near the z-line. Exercise has been reported to cause marked depletion of intramyofibrillar glycogen, underscoring glycogen’s key roles in muscle function 53. CST increased the number of glycogen granules both in the subsarcolemmal and intramyofibrillar region (Fig. 1F, S2A&B). As previously shown 40, CST decreased lipid density in the liver of DIO mice (Fig. 1G, S2E&F), thereby reducing steatosis. Contribution of catecholamines and glucagon. 5 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Glycogenolysis, controlled by the counterregulatory hormones such as glucagon, NE and EPI 54, helps the liver to maintain blood glucose concentrations during the early stage of fasting. We measured the effect of CST on these hormones. While plasma NE and EPI levels were reduced upon CST treatment in both fed and fasted state, no changes in glucagon levels were detected in the fed state (Fig. 3A). These findings suggest that one of the mechanisms by which CST restores and increases hepatic glycogen content in the early stage of fasting, could be by inhibiting catecholamine- induced glycogenolysis. Interestingly, acute insulin treatment (for 30 minutes) raised plasma NE and EPI levels only in the fasted state but not in the fed state (Fig. 3B&C). Exogenously administered insulin therefore seems to play a dual role in the fasting state to maintain glucose levels, stimulating gluconeogenesis by catecholamines but also promoting glycogenesis (Fig. 1A&D). Since in the fasted state, insulin not only increased liver glycogen content, but also increased plasma catecholamines, insulin mediated inhibition of glycogenolysis dominates over catecholamine induced glycogenolysis. Since the glycogenic function of the liver is compromised in T2D 19,20, we tested the effects of CST on the hepatic glycogen content in insulin-resistant DIO mice. A schematic diagram shows the diet, peptide treatment and tissue harvesting (Fig. 3D). Chronic treatment of DIO mice with CST caused a >1.5-fold increase in liver glycogen content (Fig. 3E). When values in Figure 2E and F were compared, it appeared that in the fasted state, CST treatment increased liver glycogen content more than insulin in DIO mice. Thus, CST was more effective in inducing glycogenesis than insulin in DIO mice. Like in NCD mice, the effects of combined CST and insulin treatment in hepatic glycogen content were additive, indicating that these compounds both promote glycogenesis via divergent and synergistic pathways (Fig. 3F). We have previously shown that CST inhibits nicotine-induced secretion of NE from PC12 cells 39,46,48 as well as NE and EPI from mouse 47.

Unlike in NCD mice, CST caused a decrease only in NE levels but did not change EPI and glucagon levels in fed DIO mice (Fig. S3). In fasted mice, the effects of CST on counterregulatory hormones were comparable between NCD and DIO mice (Fig. 2A&S3). Importantly, insulin-induced glycogen synthesis is reduced in DIO mice, but CST is equally effective in NCD or DIO mice. (Figs. 1C&3E). Dose dependent improvement of glucose tolerance by CST in DIO mice. We have recently shown that CST reduces inflammation in DIO mice, resulting in an improvement in insulin sensitivity 40. Although in our previous work we demonstrated improvement of glucose tolerance by CST 40, here we have conducted a dose-response study of CST (0.1-1.0 µg/g body weight) on glucose tolerance by intraperitoneal glucose tolerance test (ip-GTT) (Fig. 3A&B), showing that CST caused dose-dependent improvement in glucose tolerance (Fig. 3A&B). We have also conducted a time-course of oral CST action by oral-GTT (Fig. 3C&D) demonstrating that in DIO 6 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. mice, revealing improved glucose tolerance by CST after 1 day of treatment (Fig. 3C&D). The absolute blood glucose values at 0 min are as follows: Pretreatment: 149.9±7.32; 1 day after CST: 142.7±8.836; 3 consecutive days of CST treatment: 138.6±8.16; 5 consecutive days of CST treatment: 134.0±11.66; 10 consecutive days of CST treatment: 128.9±11.9. Oral delivery of CST did not result in a significant reduction of the body weight (Fig. S4A). The intraperitoneal insulin tolerance test (ip-ITT) showed improvement in insulin tolerance in DIO mice by CST (Fig. 3F&G). The absolute blood glucose values at 0 min are as follows: saline: 155.5±10.62; CST: 121.2±9.63. Oral CST treatment also suppressed the expression of genes coding for enzymes involved in the gluconeogenesis: the G6Pase gene (G6pc) and phosphoenol-pyruvate kinase gene (Pck1) (Fig. S4 B & C), indicating that CST will limit de novo formation and hydrolysis of G6P. Together, these findings corroborated our previous work 40 by showing in further details that CST improves the GTT and ITT and inhibits the gluconeogenesis in DIO mice. Lower hepatic glycogen content in CST-KO compared to WT mice. To gain a better understanding of the role of CST in hepatic glucose homeostasis, we examined NCD-CST-KO mice, which are genetically obese and insulin-resistant 40. Hepatic glycogen content in fed NCD-CST-KO mice was lower than in WT mice (Fig. 4A), pointing to decreased insulin sensitivity. In the fed state, insulin failed to increase the hepatic glycogen content (Fig. 4B), supporting the presence of hepatic insulin resistance in CST-KO mice 40.

In the fasted state, insulin caused only a ~1.4-fold increase in liver glycogen content in CST-KO mice (Fig. 4B). Chronic treatment of CST-KO mice with CST resulted in ~3-fold increase in hepatic glycogen content in fed mice and >4.2-fold increase in hepatic glycogen content in fasted mice (Fig. 4C). Insulin alone had no effect on accumulation of hepatic glycogen content in fasted NCD-CST-KO mice, reinforcing insulin resistance in CST-KO mice (Fig. 4D). Insulin in combination with CST caused significant increase in hepatic glycogen content in fasted NCD-CST-KO mice (Fig. 4D). Plasma insulin level in fasting mice were as follows: WT: 0.66±0.04 ng/ml; CST-KO: 1.59±0.12 ng/ml. Thus, fasting plasma insulin is CST-KO mice was >2.4-fold higher than WT mice, as shown in our previous work 40. Despite the presence of higher plasma insulin levels in the fasted state, fasted CST-KO mice showed a lower hepatic glycogen content than fasted WT mice (Fig. 4A). These effects of CST correlated with decreased levels of plasma counterregulatory hormones (Fig. 4F). Plasma glucagon levels in NCD-CST-KO mice were higher than DIO-WT mice (Figs. 4F& S3) but in contrast to DIO-WT mice, the glucagon levels were lower in CST-KO mice after CST treatment (Figs. 4F& S3). Like in DIO-WT mice, CST treatment reduced plasma NE and EPI levels in both fed and fasted CST-KO mice (Fig. 4F). However, CST treatment of CST-KO mice also lowered glucagon levels (Fig. 7 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 4F), suggesting that CST might have aided glycogen accumulation by reducing glucagon-induced glycogenolysis. Acute insulin treatment only reduced NE (in contrast to WT mice) in fasted CST-KO mice but had no effects on NE or EPI in fed mice (Fig. 4G). Thus, CST-KO mice display lower hepatic glycogen content compared to WT mice, and this is elevated by CST supplementation. CST enhances glycogen synthesis and suppresses glycogenolysis. To confirm the in vivo studies and assess whether CST directly affects glycogenesis and glycogenolysis in hepatocytes, we measured glycogen synthesis from radiolabeled glucose and glycogenolysis from prelabeled stored glycogen in cultured primary hepatocytes. We found that CST and insulin alone caused a modest (>1.4-fold by CST; ~1.8-fold by insulin) increase in glucose incorporation into glycogen, while CST and insulin in combination resulted in a >2.6-fold increase in glycogen accumulation (Fig. 5A), consistent with a previous report 12. Analyzing the dose-dependent effects of CST on net glycogenesis, the EC50 turned out to be approximately 43.4 nM (Fig. 5B). We also found that combined treatments of counterregulatory hormones NE and EPI caused a ~2.3-fold increase in glucose release from prelabeled glycogen, which was inhibited by both CST and insulin (Fig.

5C), indicating that both CST and insulin can directly induce glycogen synthesis and inhibit catecholamine-induced glycogenolysis by hepatocytes. CST increases G6P and UDPG levels After entry of glucose into hepatocytes, glucose is immediately phosphorylated to G6P followed by conversion to glucose-1-phosphate (G1P) by phosphoglucomutase (PGM), and then to UDPG by UDP-glucose pyrophosphorylase to provide the substrate for glycogen synthesis. Glycogen synthase (GYS) or liver specific GYS2 catalyzes the transfer of glucosyl units from UDPG to glycogen by synthesis of a-1,4 bonds. A strong positive correlation exists between activation of GYS and intracellular G6P 55,56. Upon binding to GYS, G6P causes allosteric activation of the enzyme through a conformational rearrangement that simultaneously converts it into a better substrate for protein phosphatases 57-59. Dephosphorylation results in activation of GYS. CST increased G6P levels in both fed (>1.68-fold) and fasted (>1.88-fold) NCD-WT mice (Fig. 5D). In contrast, CST did not increase G6P levels in both the fed and fasted states in DIO mice (Fig. 5E). However, the basal G6P levels in fed DIO mice seemed to be already higher than the basal levels in fed NCD mice, almost close to the levels of CST-treated NCD mice (Fig. 5E). Since UDPG levels were increased by CST in fasted NCD and DIO mice (Fig. 5 F & G), these results underscore the critical role of UDPG-pyrophosphorylase in CST action in DIO mice where G6P levels are not changed. 8 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. CST augments GYS2 activity by inactivating GSK-3β through phosphorylation. GYS activity is controlled by covalent modification of the enzyme complex 60, allosteric activation by G6P 58,61, and enzymatic translocation by insulin 62. Therefore, we measured GYS2 activity by following incorporation of UDP-14C-glucose into glycogen in the presence or absence of saturating concentrations of G6P. CST stimulated GYS2 activity in fasted NCD-WT and DIO-WT mice (Fig. 5H &I). In addition to the allosteric activation of GYS2 by G6P, GYS2 is regulated by multi-site phosphorylation which causes inactivation. GYS2 is phosphorylated on nine serine residues by kinases such as glycogen synthase kinase (GSK)-3β 63,64. GSK-3β in turn is inactivated by phosphorylation of its Ser9 site by PKA, AKT and p90RSK and activated by autophosphorylation of Tyr216 65-68. Phosphorylation of GYS2 results in its progressive inactivation and decreased sensitivity to allosteric activators 69. GYS2 dephosphorylation is regulated by hormones like insulin and is mediated by phosphatases like protein phosphatase 1 58.

Immunoblotting revealed that in fasted NCD-WT and DIO-WT mice, CST and insulin alone increased GSK-3β Ser9 phosphorylation (Fig. 8A&B). The effects of CST in combination with insulin were additive in DIO-WT mice but not in NCD-WT mice (Fig. 8A&B). Since Ser641 is one of the GSK- 3b target sites in GYS2, we also measured Ser641 phosphorylation by immunoblotting. In fasted NCD- WT mice, CST or insulin alone caused dephosphorylation of Ser641, while their combination caused a small additive effect (Fig. 8A&C). However, in DIO-WT mice, neither CST nor insulin reduced Ser641 phosphorylation, but the combination did (Fig. 8A&C). In summary, CST increases glycogen production in liver by increasing UDPG levels, increasing phosphorylation of GSK-3b, and dephosphorylation of GYS2. CST stimulates AKT signaling without affecting insulin-induced tyrosine phosphorylation of insulin receptor and insulin receptor substrate-1. (i) Activation of PI-3 kinase and PDK-1 dependent phosphorylation of Akt at threonine-308 residue. Insulin binding to the insulin receptor (IR) stimulates tyrosine phosphorylation (pY) of IR (through auto phosphorylation by IR tyrosine kinase) and insulin receptor substrate-1 (IRS-1) (by the IR tyrosine kinase). These are the key initial steps of the insulin signaling pathway 70. We tested whether CST has any effect on these early steps. As expected, Insulin stimulated tyrosine phosphorylation of IR and IRS-1 in cultured hepatocytes. However, the levels of phosphorylation were not significantly changed upon CST treatment (Fig. 6A-D), suggesting that the proximal part of the insulin signaling pathway is separate from the pathway for CST action. Since CST had no effects on the phosphorylation of IR and IRS1, we tested whether CST exerts any post-IR signaling. Interestingly, 9 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. CST, like insulin, stimulated PI3-kinase activity (Fig. 6E). This enzyme primarily phosphorylates inositol lipids like phosphatidyl-inositol-phosphates such as PI-4,5 P (PIP2) at the 3 positions of the inositol moiety, to form PI-3,4,5 P (PIP3) which is a bioactive lipid. However, in this activity assay using anti-p85 immunoprecipitate (which pulls down catalytic subunit p110) as the source of the enzyme and phosphatidyl-inositol phosphate (PI), instead of PIP2, as the substrate, the product was PI3P (as shown in Fig. 6E). Thus, both insulin and CST treatment produced PI3P, and combined treatment with insulin and CST seems to have an additive effect. The endogenous products PIP2 and PIP3 binds to PH (Pleckstrin Homology) domain of AKT 71 as well as to mTORC2 72. PIP2/PIP3 binding to Akt allows phosphorylation of AKT by phosphatidyl-inositol phosphate dependent kinase-1 (PDK-1) at threonine 308 residue 71.

Downregulation of PDK-1 protein expression by si-RNA (Fig. 7A&B) led to the decreased pAKT (T308) signals (Fig. 7C&D). reenforcing the notion that the phosphorylation of Akt (T308), stimulated by insulin and CST, is PDK-1 dependent. (ii) Activation of mTORC2 and phosphorylation of AKT at the serine-473 residue. Another interesting aspect of the regulation of AKT phosphorylation is the existence of a positive feedback loop through which pT308-AKT phosphorylates SIN1 subunit of mTORC2 at the T86 residue enhancing mTORC2 kinase activity, which leads to phosphorylation of AKT at S473 (pS473-AKT) by mTORC2, thereby catalyzing full activation of AKT 73. Thus, mTORC2 is activated by both PIP3 as well as pT308-AKT. Furthermore, CST enhanced phosphorylation of mTOR at S2481 (pS2481) (Fig. 8A&B). This phosphorylation is known to increase functional integrity of the mTORC2 complex 74, causing increased AKT phosphorylation at S473 (Fig. 8C&D). Therefore, CST has a double acting role in AKT activation, one by providing PIP3 (via activation of PI3- kinase and PDK-1) and the other by strengthening the integrity of mTORC2 complex. CST augments GYS2 activity by inactivating GSK-3β through phosphorylation. Activation of Akt leads to phosphorylation of GSK-3β causing inactivation of GSK-3β which prevents subsequent phosphorylation and inactivation of GYS2. In addition to the allosteric activation, GYS2 is regulated by multi-site phosphorylation which causes inactivation. GYS2 is phosphorylated on nine serine residues by kinases such as glycogen synthase kinase (GSK)-3β 63,64. GSK-3β in turn is inactivated by phosphorylation of its Ser9 site by PKA, AKT and p90RSK and activated by autophosphorylation of Tyr216 65-68. Phosphorylation of GYS2 results in its progressive inactivation and decreased sensitivity to allosteric activators 69. GYS2 dephosphorylation is regulated by hormones like insulin and is mediated by phosphatases like protein phosphatase 1 58. Immunoblotting revealed that in fasted NCD-WT and DIO-WT mice, CST, and insulin alone increased GSK-3β Ser9 phosphorylation (Fig. 9A&B). The effects of CST in combination with insulin were additive in DIO-WT mice but not in NCD-WT mice (Fig. 9A&B). Since Ser641 is one of the GSK- 10 326 327 328 329 330 331 332 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 3b target sites in GYS2, we also measured Ser641 phosphorylation by immunoblotting. In fasted NCD- WT mice, CST or insulin alone caused dephosphorylation of Ser641, while their combination caused a small additive effect (Fig. 9A&C). However, in DIO-WT mice, neither CST nor insulin reduced Ser641 phosphorylation, but the combination did (Fig. 9A&C). In summary, CST increases glycogen production in liver by increasing UDPG levels, increasing phosphorylation of GSK-3b, and dephosphorylation of GYS2.

11 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Discussion In the postprandial state, a major pathway that contributes to the removal of glucose from the portal vein by the liver is conversion of glucose into glycogen 75,76 and defects in hepatic glycogen synthesis contribute to postprandial hyperglycemia in patients with poorly controlled insulin-dependent diabetes mellitus 15. The present study shows that CST induces hepatic glycogenesis in both the fed and fasted state in both lean (NCD) and obese (DIO) mice. CST caused an almost doubling of G6P levels in lean mice. In the liver of obese mice, basal G6P levels are already higher than in the liver of lean mice. Therefore, CST had no additional effects on G6P levels. In both lean and obese mice, GYS2 activity was stimulated by CST, resulting in the efficient conversion of G6P to glycogen. It is likely that the elevated G6P allosterically activates GYS to allow greater glycogen synthesis 55,57,58 after CST treatment. In fasted lean mice, CST or insulin alone induced >5-fold increase in hepatic glycogen content, while their combination caused an additive effect. This additive effect indicates that CST and insulin use divergent signaling pathways to induce hepatic glycogenesis in the postabsorptive state. Such an effect was limited in DIO mice because of insulin resistance. Nevertheless, CST also elevated glycogen content in the liver of DIO mice. In CST-KO mice, supplemented CST reaches a steady state level of 4-5 nM concentration in plasma in about 15 hours. Physiologically, such a concentration of CST could sustain a significant increase in glycogen formation. Together with insulin, this concentration of CST can boost glycogen loading even in obese mice, thus reducing circulating glucose level, implicating that delivery of circulatory glucose from G6P will be reduced by suppressing expression of G6pc with consequent diversion towards glycogenic pathway. Part of the CST effect in vivo could be indirect through changes in catecholamines. Indeed, CST decreased the levels of plasma catecholamines in the fed state, while insulin had no effect. However, CST did not change plasma glucagon levels. These findings indicate that CST restores and increases hepatic glycogen content by inhibiting catecholamine-induced glycogenolysis in the early stage of fasting. In contrast, insulin-induced hepatic glycogenesis is associated with increased plasma catecholamines in the fasted mice. The contribution of catecholamine-induced glycogenolysis to glucose production is well known 23,77. Suppression of catecholamine levels by CST can have dual indirect impacts as it could reduce both the gluconeogenic as well as the glycogenolytic effects of catecholamines 77.

Our results in hepatocytes showed that CST and insulin inhibit catecholamine-induced glycogenolysis. Although under stressed conditions (starvation) insulin elevates catecholamine levels, 12 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. its inhibitory effects on glycogenolysis may still dominate allowing accumulation of glycogen. In cultured hepatocytes, both glucose and insulin suppress glycogenolysis by inhibiting glycogen phosphorylase a, in a process possibly involving protein phosphatase 1 78. Protein phosphatase 1 inactivates glycogen phosphorylase via dephosphorylation of phosphorylase kinase (phosphorylated by PKA) as well as activates GYS2 by dephosphorylation 79. The net result of this is increased glycogen accumulation. Studies in humans show a dichotomy of EPI and NE responses during the oral glucose tolerance test: decreased plasma EPI and increased plasma NE 80, where the authors did not find a correlation between changed NE and EPI with glucose tolerance in mice. CST deficiency in CST-KO mice creates glucose intolerance and insulin resistance even in NCD-fed CST-KO mice. As would be predicted, CST-KO mice stored less glycogen in the liver than WT mice. In the absence of CST, elevated pancreatic glucagon and higher plasma NE may contribute to the glucose intolerance which could be corrected by insulin or CST. The glycogenic function of liver is compromised in both T2D subjects and DIO mice 19,20. CST increased the hepatic glycogen content in fed DIO mice and was more effective than insulin in inducing hepatic glycogenesis in fasted DIO mice. These findings indicate that the increased hepatic glycogenesis in both fed and fasted DIO mice by CST may be due to the increased insulin sensitivity that we have reported earlier 40. Besides the indirect effects through the catecholamines, CST seems to exert a direct effect on glycogenesis in hepatocytes. Both CST and insulin inhibit glycogenolysis in isolated hepatocytes but there is no additive effect, suggesting a common signaling pathway. In contrast, CST and insulin have an additive effect on glycogen accumulation that may indicate a divergence of CST and insulin effects at the level of glycogen synthesis. The fact that CST alone stimulates glycogen synthesis in insulin-resistant DIO mice and in isolated hepatocytes, supports a direct, insulin-independent regulatory role of CST in glycogenesis. CST treatment increased cellular G6P levels which is consistent with our previous finding that CST suppresses expression of the G6pc gene 40, thus reducing hydrolysis of G6P back to glucose. The present findings demonstrate that G6P is also directed by CST into the glycogenic pathway by stimulation of GYS and inhibition of glycogenolysis.

Consistent with glucose intolerance in CST-KO mice 40, we found here the reduced ability in CST-KO mice to convert glucose to glycogen owing to lack of CST. The nature of the relationship between insulin and CST-induced signaling pathways is incompletely understood. We show here that CST does not affect the proximal part of insulin signaling pathway, involving tyrosine phosphorylation of IR and IRS-1. On the other hand, under insulin resistant conditions (in DIO mice), insulin and CST cooperate to produce an additive effect on phosphorylation of GSK-3β and dephosphorylation of GYS2, suggesting that CST improves the weak insulin response 13 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. in obese mice. How might this happen? Our results suggest a two-prong effect of CST: (i) stimulation of PI3-kinase-mediated production of PIP3 leading to increased phosphorylation of AKT at T308 in a PDK-1 dependent manner, and (ii) increased autophosphorylation of mTOR at S2481 leading to improved integrity of the mTORC2 complex which allows increased phosphorylation of AKT at Ser473 74,81. A third mechanism could be the stimulation of a positive feedback loop, as suggested in the literature, where it was shown that activation of AKT (at T308) directly phosphorylate SIN1 subunit of mTORC2 complex and enhance phosphorylation of AKT at S473 73. That may mean CST, like insulin, could be an operator of an activation loop between PDK-1 and mTORC2 leading to phosphorylation of AKT at two sites, T308 and S473. Thus, the canonical insulin pathway and this alternative pathway for CST action may explain the additive effects seen with a combination of insulin and CST. The pathways of actions of insulin and CST are illustrated in the Figure 10. The mTORC2 pathway and its effects on AKT signaling has been extensively discussed in the literature 82. We have shown here three ways how CST causes glycogen accumulation. The first is to raise the substrate levels of glycogen synthesis (G6P and UDPG), one of which (G6P) is a known allosteric activator of GYS. The second is to enhance dephosphorylation of GYS by phosphorylating (and inactivating) GSK-3β. The third is to inhibit glycogenolysis. The insulin-independent stimulatory effect of CST on glycogenesis but not on glycogenolysis in hepatocytes, suggests that stimulation of glycogenesis by CST is the major pathway through which CST controls hepatic glucose production and improves glucose intolerance in obese mice. Given these characteristics, CST could be a new therapeutic peptide to treat both glucose intolerance and inflammation (in diabetes) as well as hypertension (due to its anti-adrenergic activity).

Conclusion CST improves glucose tolerance in obese and insulin-resistant mice by the following mechanisms: (i) reduced HGP from G6P, (ii) increased glycogen synthesis from G6P via formation of UDPG, (iii) phosphorylation of AKT at S473 and T308 by stimulating mTOR and PDK1, respectively, and (iv) reduced glycogen breakdown (due to suppression of plasma catecholamine levels). The net result is reduced free available glucose. These conclusions were verified in CST-KO mice and in hepatocyte cultures. 14 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Materials and Methods Animals, diets, and treatments. We used only male mice in this manuscript. Diet-induced obese (DIO) mice were created by feeding male WT mice (C57BL/6), starting at 8 weeks of age with HFD (Research diets D12492, 60% of calories from fat) for 112 days (16 weeks). Mice were kept in a 12:12 hour dark/light cycle; food and water were always available. Male CST knockout (CST-KO) mice with C57BL/6 background were also used. Both control WT mice as well as CST-KO mice were fed NCD (14% of calories from fat). Mice were injected intraperitoneally or orally with CST (2 µg/g body weight) for 1 hour (acute) on 168 days old mice (24 weeks) or 15 days (chronic) on day 153 days old mice. This dose of CST has been chosen from previous publications in rodents 40,83-88. Acute CST treatment (1 hour) was carried out in fasted mice 1 hour before sacrifice. A group of mice were fasted for 8 hours. Both fed and fasted mice were treated with saline or insulin (0.4 mU/g body weight) for 30 minutes before sacrificing for tissue collection that was done in the same way and at the same time of the day (8:00 to 9:30 AM). Tissues were rapidly frozen in liquid nitrogen and kept in -80oC freezer. Previously, we reported that CST treatment did not reduce body weight and did not change food intake in WT mice 40. In accordance with NIH animal care guidelines, all procedures and animals were housed and handled with the approval of The Institutional Animal Care and Utilization Committee. Glucose tolerance test (GTT) and insulin tolerance test (ITT). For GTT, glucose (1 mg/g) was injected intraperitoneally (ip-GTT) or gavaged orally (O-GTT) (at time zero) after an 8-hour fast. Food was removed from the cages between 12:00 PM to 0:30 AM. Glucose and insulin were injected on fasted (8 hours) between 8:00 to 8:30 AM, respectively for GTT and ITT. Tail-vein glucose levels were measured using a glucometer at 0, 15, 30, 60, 90, and 120 minutes. For ITT, insulin (0.4 mU/g) was injected intraperitoneally, and blood glucose levels were measured using a glucometer at the indicated time points.

Since plasma insulin level is low in fasting state, we injected insulin in fasted mice. GraphPad Prism software was used to determine the area under the curve (AUC) for each line curve. Hepatocyte isolation, culture, and assay for glycogen synthesis. Male mice (16-week-old) were fed NCD and used for perfusion of liver. Mice were perfused for 5 minutes with a calcium free buffer and followed by collagenase perfusion in a calcium containing buffer for another 5 minutes. Perfusion was carried out by inserting a catheter through inferior vena cava and passing buffer through a tube and allowing buffer to come out through portal vein which was cut for this purpose. The procedure was a modified version 40 of a published article 89. Livers, after 15 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. collagenase digestion, were excised out, hepatocytes were squeezed out in a petri dish inside a culture hood, filtered through 100-micron nylon filter, centrifuged at 50 x g for 5 minutes, and pellets were collected. The suspensions of cell pellets were then passed through 30% isotonic percoll by centrifuging at 100 x g for 10 minutes. Pellets were washed in buffer and suspended in culture medium (Williams E) containing glutamax, 10% FBS, 10 nM dexamethasone and antibiotics. Hepatocytes were seeded on collagen I coated plates. Cells were cultured in Williams E medium and then depleted of glycogen store by incubating in Hepes Krebs Ringer Bicarbonate buffer for 8 hours followed by switching to serum-free Williams E medium (25 mM glucose) with or without the presence of insulin (10 nM), CST (100 nM) or insulin+CST, incubated for 24 hours in presence of 14C-glucose (10 µCi/mL, 5 mM). At the end of incubation, cultures were washed, and cells were dissolved in 30% KOH, boiled for 20 minutes, glycogen was precipitated by 60% ethanol (in presence of 5 µg glycogen as carrier), pellets washed with 70% ethanol, dried free of traces of ethanol, dissolved in water and counted. Glycogen and enzyme analysis in liver. Fresh and frozen liver (25-30 mg) and muscle (90-100 mg) tissues from fed or fasted mice (collected between 8:00 to 9:30 AM under deep anesthesia) and glycogen was extracted by boiling with 30% KOH solution as described previously 90. Extracted glycogen was precipitated in cold by 66% ethanol and washed with 70% ethanol. After drying to remove traces of ethanol, the pellets were dissolved in water and then subjected to colorimetric determination of glycogen using Anthrone reagent (0.05% anthrone and 1% thiourea) in concentrated H2SO4. A group of lean and obese mice were fasted for 8 hours (12:00 PM to 0:30 AM) and treated with saline or insulin for 15 minutes.

Mice were sacrificed (8:00 to 8:30 AM) and livers were subjected to analysis of GYS2 activity and immunoblotting. GYS2 activity was measured by incorporation of UDP-14C-glucose into glycogen in presence of 10 mM and 0.1 mM G6P, an activator, and 14C- glycogen formed was spotted on GF/A filter paper, washed with 70% ethanol, dried and counted for radioactivity. Counts were subsequently normalized against protein. Results are presented as activation ratio (0.1 mM G6P/10 mM G6P). Measurement of catecholamines. Blood was drawn from the heart between 8:00 – 9:30 AM under deep isoflurane-induced anesthesia (1-2 min) and kept in potassium-EDTA tubes. Plasma catecholamines were measured by ACQUITY UPLC H-Class System fitted with Atlantis dC18 reversed-phase column (100A, 3 µm, 2.1 mm x 150 mm) and connected to an electrochemical detector (ECD model 2465) (Waters Corp, Milford, MA) as described previously 88. The mobile phase (isocratic: 0.3 mL/min) consisted of phosphate-citrate buffer 16 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. and acetonitrile at 95:5 (vol/vol). For determination of plasma catecholamines, DHBA (2 ng) was added to 150 µl plasma and adsorbed with ~15 mg of activated aluminum oxide for 10 min in a rotating shaker. After washing with 1 mL water adsorbed catecholamines were eluted with 100 µl of 0.1N HCl. 10 µl of the eluate was injected into UPLC where the ECD range was set at 500 pA. The data were analyzed using Empower software 3 (Waters Corp, Milford, MA). Catecholamine levels were normalized with the recovery of the internal standard. Plasma catecholamines are expressed as nM. Analysis of glycogenolysis in hepatocytes. Radiolabeled glycogen was synthesized in hepatocytes from 14C-glucose (10 µCi/mL) as described above in presence of insulin only (no CST). Cultures were washed (Krebs-Ringer bicarbonate buffer) to remove external radioactivity and then exposed to buffer (basal), NE (10 µM), EPI (10 µM), insulin (100 nM) and CST (100 nM) alone and in various combinations and incubated for 60 minutes. Incubation media were collected for radioactive counting. Cells were washed with cold buffer followed by cell lysis. An aliquot of each lysate was saved for protein assay and the rest was treated with 0.2 M barium hydroxide and 5% zinc sulphate (to precipitate protein) and centrifuged at 10,000 x g for 10 minutes. Supernatants were passed through 100 mg mixed bed ion-exchange resins (both cationic and anionic). Resins were washed twice with water and counted for bound radioactivity (glucose phosphates liberated from hydrolysis of radiolabeled glycogen will remain bound to the resins). For each incubation, total radioactivity was determined by combining the incubation medium, resin flow through and the radioactivity bound to the resin.

The total radioactivity was normalized with protein, and it represented 14C-glucose (phosphorylated and non-phosphorylated) released from 14C-glycogen due to glycogenolysis. Measurement of G6P and UDPG. G6P levels were measured in liver tissues by a commercial kit (Sigma-Aldrich, St. Louis, MO) and UDPG levels were measured by an UPLC-UV detection method in the UCSD core facility. Assay for PI-3-kinase activity. CST (100 nM for 1hour), insulin (10 nM for 10 minutes) and untreated hepatocytes were lysed, and proteins were subjected to immunoprecipitation with anti-p85 antibody. After extensive washing, immunoprecipitates were analyzed for kinase activity in presence of phosphatidylinositol (PI) and g- 32P-ATP (PerkinElmer, Santa Clara, CA) followed a published method 91. Extracted radiolabeled lipids, mixed with reference lipids, were analyzed by TLC using a developing solvent, CHCl3–MeOH–H2O– NH4OH (25%), 45:35:8.5:1.5 (v/v). Reference spots were identified by exposing to iodine vapor in a 17 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. closed chamber. TLC plates were exposed to X-ray film for 72 hours at -80°C. Reference spots for PI3P were counted for radioactivity. Real Time PCR. Total RNA from liver samples was isolated using RNeasy Mini Kit and reverse-transcribed using qScript cDNA synthesis kit. cDNA samples were amplified using PERFECTA SYBR FASTMIX L-ROX 1250 and analyzed on an Applied Biosystems 7500 Fast Real-Time PCR system. All PCRs were normalized to Rplp0, and relative expression levels were determined by the ΔΔCt method. Immunoblotting. Homogenization of livers were made in a lysis buffer supplemented with phosphatase and protease inhibitors. Homogenates were subjected to 10% SDS-PAGE and immunoblotted. The following primary antibodies were obtained from Cell Signaling Technology (Danvers, MA): GSK-3b (Catalog #9315) and pSer9-pGSK-3β (Catalog #5558), mouse monoclonal phospho-tyrosine 4G10 (Catalog # 96215), rabbit monoclonal GYS (Catalog #3886), and rabbit monoclonal, pSer641/640-GS (Catalog #3891). According to Cell Signaling Technology, this phospho-GYS antibody (Catalog #891) works to detect both muscle (pSer640) and liver (pSer641) signals. The antibodies against insulin receptor b (Catalog # sc-373975) and IRS-1 (Catalog # sc-8038) were purchased from Santa Cruz Biotechnology (Dallas, TX). Primary antibodies against PDK1, Heat shock protein-90 (HSP-900, p85 (a PI-3-kinase subunit), pT308-AKT, pS473-AKT were purchased from Cell Signaling Technology. Downregulation of PDK-1 by siRNA Four siRNAs targeting mouse PDK-1 and a non-targeting siRNA pool were purchased from Dharmacon (Lafayette, CO).

Hepatocytes were transfected with targeting and non-targeting siRNAs using Lipofectamine RNAiMax reagent (ThermoFisher Scientific) as the transfection reagent following manufacturer’s protocol. Transfection was continued for 48 hours before cells were homogenized. CST treatment (100 nM) was for 1 hour and insulin (10 nM) was for 10 minutes before cell homogenization. Statistical analysis. Statistics were performed with PRISM 8 (version 8.4.3) software (San Diego, CA). Normality was assessed by either D’Augustino-Pearson omnibus normality test or Shapio-Wilk normality test. After passing the normality test, data were analyzed using either unpaired two-tailed Student’s t-test for comparison of two groups or one-way or two-way analysis of variance (ANOVA) for comparison of 18 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. more than two groups followed by Tukey’s post hoc test if appropriate. All data are presented as mean ± SEM and significance were assumed when p<0.05. Author contributions. GB researched the data, contributed to the design and discussion, and wrote the manuscript. KT generated the data, contributed to the design and discussion. NJW and GvdB edited the MS and provided intellectual inputs. SKM conceived the idea, generated ultrastructural data, supervised the studies, analyzed the data and made the graphics, wrote the MS, and provided financial support. GB and SKM are the guarantor of this study. Acknowledgments. This work was supported by grants from the Veterans Affairs (I01 BX003934 to SKM). Conflicts of interest: The authors declare no competing interests. Availability of data and material: The datasets generated during this study are available from the corresponding author on reasonable request. 19 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure Legends Figure 1. Effects of chronic CST and acute insulin treatments on hepatic glycogen contents in both fed and fasted NCD-WT mice. Mice were treated with CST (2 µg/g body weight/day for 15 days or 1 hour; intraperitoneal) followed by fasting (8 hr) and treatment with saline or insulin (0.4 mU/g body weight for 30 min; intraperitoneal) before harvesting tissues under deep anesthesia. (A) Liver glycogen content in fed (n=12) and fasted (n=15) NCD-WT mice. Two-way ANOVA: Interaction: p<0.001; Food: p<0.001; Treatment: p<0.001. (B) Plasma insulin levels in fed and fasted WT-NCD mice (n=14).

Student’s t test. (C) Liver glycogen content in fed and fasted WT-NCD mice treated with saline (n=43), acute CST for 1 hour (n=12) or chronic CST for 15 days (n=43). Three-way ANOVA: Food: p<0.001; Acute vs Chronic: p<0.001; Saline vs CST: p<0.001; Food x Acute vs Chronic: ns; Food x Saline vs CST: p<0.05; Acute vs Chronic x Saline vs CST; p<0.001; Food x Acute vs Chronic x Saline vs CST: ns. (D) Liver glycogen content after saline (n=20), insulin for 30 min (n=9), acute CST plus insulin (n=9) or chronic CST for 15 days plus insulin for 30 min (n=11) treatment in fasted NCD-WT mice. One-way ANOVA. (E) Glycogen content in fed NCD-WT gastrocnemius muscle after saline or CST (2 µg/g body weight/day for 15 days; intraperitoneal) treatment (n=10). Student’s t test. (F) Morphometric analysis of 2-D TEM micrographs showing glycogen granules in saline and CST treated sub-sarcolemma and myofibril. Two-way ANOVA: Interaction: p<0.01; Treatment: p<0.001; Zone: p<0.001. (G) Morphometric assessment of lipid content in the TEM micrographs of steatotic liver of DIO mice after saline or chronic CST treatments. Student’s t test. *p<0.05; †, p<0.01; ‡, p<0.001. Figure 2. Effects of chronic CST and acute insulin treatments on plasma counterregulatory hormone levels in both fed and fasted NCD-WT mice. Fed or fasted NCD-WT mice were treated with CST (2 µg/g body weight/day for 15 days: intraperitoneal) before harvesting tissues under deep anesthesia after 24 hrs of last injection. (A) NE (n=16), EPI (n=16) and glucagon (n=6) levels in fed and fasted NCD-WT mice. Three-way ANOVA: Hormones: p<0.001; Fed vs Fast: p<0.001; Saline vs CST: p<0.001; Hormones x Fed vs Fast: p<0.05; Hormones x Saline vs CST: ns; Fed vs Fast x Saline vs CST: p<0.001; Hormones x Fed vs Fast x Saline vs CST: p<0.001. (B) NE levels in saline or insulin treated fed or fasted NCD-WT mice (n=7). Two-way ANOVA: Interaction: p<0.01; Treatment: p<0.05; Food: p<0.01. (C) EPI levels in saline or insulin treated fed or fasted NCD-WT mice (n=7). Two-way ANOVA: Interaction: p<0.01; Treatment: p<0.001; Food: p<0.001. (D) Schematic diagram showing age, diet, CST treatments, glucose and insulin tolerance tests, and tissue harvesting. (E) Liver glycogen content in fed and fasted DIO-WT mice after saline (n=27) or CST (n=16) treatments. Two- 20 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. way ANOVA: Interaction: p<0.01; Treatment: p<0.001; Food: p<0.001. (F) Effects of saline (n=47), insulin alone (n=9; 30 min) or in combination with chronic CST (n=12) on glycogen content in fasted DIO-WT mice. One-way ANOVA: p<0.001. †, p<0.01; ‡, p<0.001; ns, not significant. Figure 3.

Effects of intraperitoneal or oral CST treatment on glucose tolerance in DIO mice without causing weight loss. (A&B) Oral glucose tolerance test (O-GTT): (A) Dose-dependent effects of oral CST on blood glucose levels during O-GTT (n=6). Two-way ANOVA: Interaction: p<0.05; Time: p<0.001; Treatment: p<0.001. (B) The area under the curve (AUC) of the OGTT was used to determine EC50 of CST action. One-way ANOVA: p<0.001. (C): Blood glucose levels during O-GTT after time-course of CST treatment (2 µg/g body weight) (n=7). Two-way ANOVA: Interaction: p<0.001; Time: p<0.001; Treatment: p<0.001. (D) The area under the curve (AUC) of the O-GTT. One-way ANOVA: p<0.001. (E) Intraperitoneal insulin tolerance test (ip-ITT): Blood glucose levels during ip-ITT after chronic CST treatment (n=12). Two-way ANOVA: Interaction: p<0.001; Time: p<0.001; Treatment: p<0.001. (F) The area under the curve (AUC) of the ip-ITT. Student’s t test. p<0.05; †, p<0.01; ‡, p<0.001; ns, not significant. Figure 4. Effects of chronic CST and acute insulin treatments on hepatic glycogen content and plasma levels of counterregulatory hormones levels in NCD-CST-KO mice. NCD-WT and NCD- CST-KO mice were treated with CST (2 µg/g body weight/day for 15 days; intraperitoneal) followed by treatment with insulin (0.4 mU/g body weight for 30 minutes) before tissue harvesting. Tissues were harvested from fed or fasted (8 hr) mice under deep anesthesia. (A) Hepatic glycogen content in fed and fasted NCD-CST-KO mice compared to NCD-WT mice (n=12). Two-way ANOVA: Interaction: p<0.001; Treatment: p<0.001; Food: p<0.001. (B) Effects of insulin (30 min) on glycogen content in fed and fasted NCD-CST-KO mice (n=12). Two-way ANOVA: Interaction: ns; Time: Treatment: p<0.05; Food: p<0.001. (C) Liver glycogen content in fasted and fed NCD-CST-KO mice after saline (n=26) chronic CST treatment (n=14). Two-way ANOVA: Interaction: ns; Time: Treatment: p<0.05; Food: p<0.001. (D) Effects of insulin (30 min) alone or in combination with chronic CST on liver glycogen content in fasted CST-KO-NCD mice (n=12). One-way ANOVA: p<0.001. (E) Plasma insulin levels in fed or fasted NCD-CST-KO mice (n=14). Student’s t test. (F) NE (n=15), EPI (n=15), and glucagon (n=6) levels in fed or fasted NCD-CST-KO mice after chronic treatments with CST. Three-way ANOVA: Hormones: p<0.001; Fed vs Fast: p<0.01; Sal vs CST: p<0.001; Hormones x Fed vs Fast: ns; Hormones x Sal vs CST: p<0.001; Fed vs Fast x Sal vs CST: p<0.05; Hormones x Fed vs Fast x Sal vs CST: ns. (G) NE and EPI levels in saline or insulin (30 min) treated fed and fasted CST-KO-NCD mice (n=6). Three-way ANOVA: Catecholamines: p<0.001; Fed vs Fast: p<0.001; Sal 21 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder.

All rights reserved. No reuse allowed without permission. vs Insulin: p<0.01; Catecholamines x Fed vs Fast: p<0.001; Catecholamines x Sal vs Insulin: p<0.05; Fed vs Fast x Sal vs CST: ns; Catecholamines x Fed vs Fast x Sal vs CST: ns. *p<0.05; ‡, p<0.001; ns, not significant. Figure 5. Regulation of glycogenesis and glycogenolysis in hepatocytes by CST, and regulation of the levels of G6P and UDPG as well as GYS2 activity in mouse liver by CST. (A) Glycogenesis from 14C-glucose in cultured primary lean hepatocytes (n=4-5). One-way ANOVA: p<0.001. (B) Dose-dependent effects of CST on glycogenesis. EC50 was calculated using GraphPad PRISM software. (C) Glycogenolysis from preloaded radiolabeled glycogen (n=6). One-way ANOVA: p<0.001. (D) Effects of CST on liver G6P levels in fed or fasted NCD-WT mice (n=6). Two-way ANOVA: Interaction: ns; Treatment: p<0.001; Food: ns. (E) Effects of CST on liver G6P levels in fed or fasted DIO-WT mice (n=6). Two-way ANOVA: Interaction: ns; Treatment: ns; Food: p<0.01. (F&G) Effects of CST on UDPG in liver of fasted NCD-WT or DIO-WT mice (n=4). Student’s t test. (H & I)) Effects of CST on GYS2 activities in liver of fasted NCD-WT and DIO-WT mice (n=6). Student’s t test. For GYS2 activity, the activation ratios in presence of low (0.1 mM) and high (10 mM) G6P are shown (n=6). Tissues were from the same experiments shown in Figures 1&2. *p<0.05; †, p<0.01; ‡, p<0.001, ns, not significant. Figure 6. Unchanged tyrosine phosphorylation of IR and IRS-1 and increased PI3-kinase activity by CST in primary hepatocytes. (A) Immunoblots show tyrosine phosphorylation (pY) of IR in response to insulin and CST (n=3). (B) Immunoblots show tyrosine phosphorylation (pY) of IRS-1 in response to insulin and CST (n=3). (C) Corresponding density ratio of phospho-/total IR signals. One-way ANOVA: p<0.001. (D) Corresponding density ratio of phospho-/total IRS-1 signals. One-way ANOVA: p<0.01. (E) Autoradiograph of thin layer chromatography plate showing formation of PI3P from PI due to increased PI-3-kinase activity in the anti-p85-immunprecipitates after stimulation with insulin and CST. *p<0.05; †, p<0.01; ns, not significant. Figure 7. Increased phosphorylation of mTOR in the mTORC2 complex and AKT-S473 by insulin and CST in primary hepatocytes. (A) Immunoblots showing phosphorylation of mTOR at Ser2481 (n=3). (B) Corresponding density ratio of phospho-/total of mTOR signals. One-way ANOVA: p<0.001. (C) Immunoblots showing phosphorylation of AKT at Ser473 (n=3). (D) Corresponding density ratio of phospho-/total AKT signals are shown for AKT. One-way ANOVA: p<0.001. *p<0.05; †, p<0.01; ‡, p<0.001; ns, not significant. 22 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Figure. 8. Stimulation of phosphorylation of AKT at T308 residue by insulin and CST in a PDK- 1 dependent manner. (A) Immunoblots showing downregulation of PDK-1 protein expression by si- RNA. (B) The bar graphs showing signaling of the immunoblot. Student’s t test. (C) Immunoblots showing phospho-T308-AKT and total AKT signals after saline (Basal), insulin and CST treatments in presence and absence of si-RNA against PDK-1. (D) Bar graphs showing phospho-/total AKT signals. One-way ANOVA: p<0.001. ‡, p<0.001; ns, not significant. Figure 9. CST augments hepatic GYS2 activity (i.e., increased dephosphorylation) and decreases GSK-3b activity (i.e., increased phosphorylation) in fasted NCD-WT and DIO-WT mice. (A) Immunoblots showing total and phosphorylated signals for GSK-3β and GYS2 in NCD-WT and DIO-WT mice. (B) Densitometry values of GSK-3β in NCD-WT mice (n=4). One-way ANOVA: p<0.001. (C) Densitometry values of GSK-3β in DIO-WT mice (n=4). One-way ANOVA: p<0.001. (D) Densitometry values of GYS2 in NCD-WT mice (n=4). One-way ANOVA: p<0.01. (E) Densitometry values of GYS2 in DIO-WT mice (n=4). One-way ANOVA: p<0.001. Tissues were from the same experiments shown in Figures 1&2. *p<0.05; †, ‡, p<0.001; ns, not significant. Figure 10. Schematic diagram showing insulin and CST mediated signaling pathways involved in the regulation of hepatic glucose and glycogen metabolism. 23 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. References 1. 2. 3. 4. 5. 6. 7. 8. Martin BC, Warram JH, Krolewski AS, Bergman RN, Soeldner JS, Kahn CR. Role of glucose and insulin resistance in development of type 2 diabetes mellitus: results of a 25-year follow- up study. Lancet. 1992;340(8825):925-929. Kahn SE, Hull RL, Utzschneider KM. Mechanisms linking obesity to insulin resistance and type 2 diabetes. Nature. 2006;444(7121):840-846. Shoelson SE, Herrero L, Naaz A. Obesity, Gastroenterology. 2007;132(6):2169-2180. DeFronzo RA, Ferrannini E, Groop L, et al. Type 2 diabetes mellitus. Nat Rev Dis Primers. 2015;1:15019. Menke A, Casagrande S, Geiss L, Cowie CC. Prevalence of and Trends in Diabetes Among Adults in the United States, 1988-2012. JAMA. 2015;314(10):1021-1029. Zimmet P, Alberti KG, Shaw J. Global and societal implications of the diabetes epidemic. Nature. 2001;414(6865):782-787. Wild S, Roglic G, Green A, Sicree R, King H. Global prevalence of diabetes: estimates for the year 2000 and projections for 2030. Diabetes Care. 2004;27(5):1047-1053. Kawahito S, Kitahata H, Oshita S. Problems associated with glucose toxicity: role of hyperglycemia-induced oxidative stress.

World J Gastroenterol. 2009;15(33):4137-4142. Campos C. Chronic hyperglycemia and glucose toxicity: pathology and clinical sequelae. Postgrad Med. 2012;124(6):90-97. inflammation, and insulin resistance. 9. 10. Dubois M, Vacher P, Roger B, et al. Glucotoxicity inhibits late steps of insulin exocytosis. Endocrinology. 2007;148(4):1605-1614. 11. Roden M, Perseghin G, Petersen KF, et al. The roles of insulin and glucagon in the regulation of hepatic glycogen synthesis and turnover in humans. J Clin Invest. 1996;97(3):642-648. 12. Agius L. Physiological control of liver glycogen metabolism: lessons from novel glycogen phosphorylase inhibitors. Mini Rev Med Chem. 2010;10(12):1175-1187. 13. Cherrington AD. Banting Lecture 1997. Control of glucose uptake and release by the liver in vivo. Diabetes. 1999;48(5):1198-1214. 14. Moore MC, Coate KC, Winnick JJ, An Z, Cherrington AD. Regulation of hepatic glucose uptake and storage in vivo. Adv Nutr. 2012;3(3):286-294. 15. Hwang JH, Perseghin G, Rothman DL, et al. Impaired net hepatic glycogen synthesis in insulin- dependent diabetic subjects during mixed meal ingestion. A 13C nuclear magnetic resonance spectroscopy study. J Clin Invest. 1995;95(2):783-787. 16. Shulman GI, Rothman DL, Jue T, Stein P, DeFronzo RA, Shulman RG. Quantitation of muscle glycogen synthesis in normal subjects and subjects with non-insulin-dependent diabetes by 13C nuclear magnetic resonance spectroscopy. N Engl J Med. 1990;322(4):223-228. 17. Cline GW, Rothman DL, Magnusson I, Katz LD, Shulman GI. 13C-nuclear magnetic resonance spectroscopy studies of hepatic glucose metabolism in normal subjects and subjects with insulin-dependent diabetes mellitus. J Clin Invest. 1994;94(6):2369-2376. 18. Shulman GI. Cellular mechanisms of insulin resistance. J Clin Invest. 2000;106(2):171-176. 19. DeFronzo RA, Simonson D, Ferrannini E. Hepatic and peripheral insulin resistance: a common feature of type 2 (non-insulin-dependent) and type 1 (insulin-dependent) diabetes mellitus. Diabetologia. 1982;23(4):313-319. 20. Bogardus C, Lillioja S, Howard BV, Reaven G, Mott D. Relationships between insulin secretion, insulin action, and fasting plasma glucose concentration in nondiabetic and noninsulin- dependent diabetic subjects. J Clin Invest. 1984;74(4):1238-1246. 21. Porte D, Jr. Sympathetic regulation of insulin secretion. Its relation to diabetes mellitus. Arch Intern Med. 1969;123(3):252-260. 22. Gerich JE, Langlois M, Noacco C, Schneider V, Forsham PH. Adrenergic modulation of pancreatic glucagon secretion in man. J Clin Invest. 1974;53(5):1441-1446. 24 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder.

All rights reserved. No reuse allowed without permission. 23. Clutter WE, Rizza RA, Gerich JE, Cryer PE. Regulation of glucose metabolism by sympathochromaffin catecholamines. Diabetes Metab Rev. 1988;4(1):1-15. 24. Sacca L, Vigorito C, Cicala M, Corso G, Sherwin RS. Role of gluconeogenesis in epinephrine- stimulated hepatic glucose production in humans. Am J Physiol. 1983;245(3):E294-302. 25. Silverberg AB, Shah SD, Haymond MW, Cryer PE. Norepinephrine: hormone and neurotransmitter in man. Am J Physiol. 1978;234(3):E252-256. 26. Sacca L, Morrone G, Cicala M, Corso G, Ungaro B. Influence of epinephrine, norepinephrine, and isoproterenol on glucose homeostasis in normal man. J Clin Endocrinol Metab. 1980;50(4):680-684. 27. Christensen NJ, Alberti KG, Brandsborg O. Plasma catecholamines and blood substrate concentrations: studies in insulin induced hypoglycaemia and after adrenaline infusions. Eur J Clin Invest. 1975;5(5):415-423. 28. Abramson EA, Arky RA. Role of beta-adrenergic receptors in counterregulation to insulin- induced hypoglycemia. Diabetes. 1968;17(3):141-146. 29. Rizza RA, Cryer PE, Haymond MW, Gerich JE. Adrenergic mechanisms for the effects of epinephrine on glucose production and clearance in man. J Clin Invest. 1980;65(3):682-689. 30. Exton JH, Robison GA, Sutherland EW, Park CR. Studies on the role of adenosine 3',5'- monophosphate in the hepatic actions of glucagon and catecholamines. J Biol Chem. 1971;246(20):6166-6177. 31. Gerich J, Cryer P, Rizza R. Hormonal mechanisms in acute glucose counterregulation: the relative roles of glucagon, epinephrine, norepinephrine, growth hormone, and cortisol. Metabolism. 1980;29(11 Suppl 1):1164-1175. 32. Winkler H, Fischer-Colbrie R. The chromogranins A and B: the first 25 years and future perspectives. Neuroscience. 1992;49(3):497-528. 33. Bandyopadhyay GK, Mahata SK. Chromogranin A Regulation of Obesity and Peripheral Insulin Sensitivity. Front Endocrinol (Lausanne). 2017;8:20. 34. Muntjewerff EM, Dunkel G, Nicolasen MJT, Mahata SK, van den Bogaart G. Catestatin as a Target for Treatment of Inflammatory Diseases. Front Immunol. 2018;9:2199. 35. Mahata SK, Corti A. Chromogranin A and its fragments in cardiovascular, immunometabolic, and cancer regulation. Ann N Y Acad Sci. 2019;1455(1):34-58. Taupenot L, Harper KL, O'Connor DT. The chromogranin-secretogranin family. New England Journal of Medicine. 2003;348(12):1134-1149. Tatemoto K, Efendic S, Mutt V, Makk G, Feistner GJ, Barchas JD. Pancreastatin, a novel pancreatic peptide that inhibits insulin secretion. Nature. 1986;324(6096):476-478. 35. Mahata SK, Corti A. Chromogranin A and its fragments in cardiovascular, immunometabolic, and cancer regulation. Ann N Y Acad Sci. 2019;1455(1):34-58. Taupenot L, Harper KL, O'Connor DT. The chromogranin-secretogranin family. New England Journal of Medicine. 2003;348(12):1134-1149. Tatemoto K, Efendic S, Mutt V, Makk G, Feistner GJ, Barchas JD. Pancreastatin, a novel pancreatic peptide that inhibits insulin secretion.

Nature. 1986;324(6096):476-478. 36. 38. O'Connor DT, Cadman PE, Smiley C, et al. Pancreastatin: multiple actions on human intermediary metabolism in vivo, variation in disease, and naturally occurring functional genetic polymorphism. J Clin Endocrinol Metab. 2005;90(9):5414-5425. 39. Mahata SK, O'Connor DT, Mahata M, et al. Novel autocrine feedback control of catecholamine release. A discrete chromogranin A fragment is a noncompetitive nicotinic cholinergic antagonist. J Clin Invest. 1997;100(6):1623-1633. 40. Ying W, Mahata S, Bandyopadhyay GK, et al. Catestatin Inhibits Obesity-Induced Macrophage Infiltration and Inflammation in the Liver and Suppresses Hepatic Glucose Production, Leading to Improved Insulin Sensitivity. Diabetes. 2018;67(5):841-848. Luketin M, Mizdrak M, Boric-Skaro D, et al. Plasma Catestatin Levels and Advanced Glycation End Products in Patients on Hemodialysis. Biomolecules. 2021;11(3). 41. 42. Gayen JR, Saberi M, Schenk S, et al. A novel pathway of insulin sensitivity in chromogranin a null mice: A crucial role for pancreastatin in glucose homeostasis. J Biol Chem. 2009;284:28498-28509. 43. Bandyopadhyay GK, Lu M, Avolio E, et al. Pancreastatin-dependent inflammatory signaling mediates obesity-induced insulin resistance. Diabetes. 2015;64(1):104-116. 25 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 44. Pereda D, Pardo MR, Morales Y, Dominguez N, Arnau MR, Borges R. Mice lacking chromogranins exhibit increased aggressive and depression-like behaviour. Behavioural brain research. 2015;278:98-106. 45. Dasgupta A, Bandyopadhyay GK, Ray I, et al. Catestatin improves insulin sensitivity by attenuating endoplasmic reticulum stress: In vivo and in silico validation. Comput Struct Biotechnol J. 2020;18:464-481. 46. Mahata SK, Mahata M, Wakade AR, O'Connor DT. Primary structure and function of the catecholamine release inhibitory peptide catestatin (chromogranin A344-364): Identification of amino acid residues crucial for activity. Mol Endocrinol. 2000;14(10):1525-1535. 47. Mahata SK, Mahapatra NR, Mahata M, et al. Catecholamine secretory vesicle stimulus- transcription coupling in vivo. Demonstration by a novel transgenic promoter/photoprotein reporter and inhibition of secretion and transcription by the chromogranin A fragment catestatin. J Biol Chem. 2003;278:32058-32067. 48. Wen G, Mahata SK, Cadman P, et al. Both rare and common polymorphisms contribute functional variation at CHGA, a regulator of catecholamine physiology. Am J Hum Genet. 2004;74(2):197-207. 49. Mahata SK, Mahata M, Fung MM, O'Connor DT. Catestatin: a multifunctional peptide from chromogranin A. Regul Pept.

2010;162(1-3):33-43. 50. DeFronzo RA, Jacot E, Jequier E, Maeder E, Wahren J, Felber JP. The effect of insulin on the disposal of intravenous glucose. Results from indirect calorimetry and hepatic and femoral venous catheterization. Diabetes. 1981;30(12):1000-1007. Jue T, Rothman DL, Tavitian BA, Shulman RG. Natural-abundance 13C NMR study of glycogen repletion in human liver and muscle. Proc Natl Acad Sci U S A. 1989;86(5):1439- 1442. 51. 52. Marchand I, Chorneyko K, Tarnopolsky M, et al. Quantification of subcellular glycogen in resting human muscle: granule size, number, and location. J Appl Physiol (1985). 2002;93(5):1598-1607. 53. Hokken R, Laugesen S, Aagaard P, et al. Subcellular localization- and fibre type-dependent utilization of muscle glycogen during heavy resistance exercise in elite power and Olympic weightlifters. Acta Physiol (Oxf). 2021;231(2):e13561. 54. Rui L. Energy metabolism in the liver. Compr Physiol. 2014;4(1):177-197. 55. Ciudad CJ, Carabaza A, Guinovart JJ. Glucose 6-phosphate plays a central role in the activation of glycogen synthase by glucose in hepatocytes. Biochem Biophys Res Commun. 1986;141(3):1195-1200. 56. Carabaza A, Ciudad CJ, Baque S, Guinovart JJ. Glucose has to be phosphorylated to activate glycogen synthase, but not to inactivate glycogen phosphorylase in hepatocytes. FEBS Lett. 1992;296(2):211-214. 57. Villar-Palasi C, Guinovart JJ. The role of glucose 6-phosphate in the control of glycogen synthase. FASEB J. 1997;11(7):544-558. Ferrer JC, Favre C, Gomis RR, et al. Control of glycogen deposition. FEBS Lett. 2003;546(1):127-132. 57. Villar-Palasi C, Guinovart JJ. The role of glucose 6-phosphate in the control of glycogen synthase. FASEB J. 1997;11(7):544-558. Ferrer JC, Favre C, Gomis RR, et al. Control of glycogen deposition. FEBS Lett. 2003;546(1):127-132. 59. Aiston S, Green A, Mukhtar M, Agius L. Glucose 6-phosphate causes translocation of phosphorylase in hepatocytes and inactivates the enzyme synergistically with glucose. Biochem J. 2004;377(Pt 1):195-204. 60. Villar-Palasi C. Substrate specific activation by glucose 6-phosphate of the dephosphorylation of muscle glycogen synthase. Biochim Biophys Acta. 1991;1095(3):261-267. 61. Bouskila M, Hunter RW, Ibrahim AF, et al. Allosteric regulation of glycogen synthase controls glycogen synthesis in muscle. Cell Metab. 2010;12(5):456-466. 62. Prats C, Cadefau JA, Cusso R, et al. Phosphorylation-dependent translocation of glycogen synthase to a novel structure during glycogen resynthesis. J Biol Chem. 2005;280(24):23165- 23172. 26 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

63. Poulter L, Ang SG, Gibson BW, et al. Analysis of the in vivo phosphorylation state of rabbit skeletal muscle glycogen synthase by fast-atom-bombardment mass spectrometry. Eur J Biochem. 1988;175(3):497-510. 64. Roach PJ. Control of glycogen synthase by hierarchal protein phosphorylation. FASEB J. 1990;4(12):2961-2968. Jope RS, Johnson GV. The glamour and gloom of glycogen synthase kinase-3. Trends Biochem Sci. 2004;29(2):95-102. 64. Roach PJ. Control of glycogen synthase by hierarchal protein phosphorylation. FASEB J. 1990;4(12):2961-2968. Jope RS, Johnson GV. The glamour and gloom of glycogen synthase kinase-3. Trends Biochem Sci. 2004;29(2):95-102. 66. Hughes K, Nikolakaki E, Plyte SE, Totty NF, Woodgett JR. Modulation of the glycogen synthase kinase-3 family by tyrosine phosphorylation. EMBO J. 1993;12(2):803-808. 67. Cross DA, Alessi DR, Cohen P, Andjelkovich M, Hemmings BA. Inhibition of glycogen synthase kinase-3 by insulin mediated by protein kinase B. Nature. 1995;378(6559):785-789. Jensen J, Brennesvik EO, Lai YC, Shepherd PR. GSK-3beta regulation in skeletal muscles by adrenaline and insulin: evidence that PKA and PKB regulate different pools of GSK-3. Cell Signal. 2007;19(1):204-210. Lawrence JC, Jr., Roach PJ. New insights into the role and mechanism of glycogen synthase activation by insulin. Diabetes. 1997;46(4):541-547. Taniguchi CM, Emanuelli B, Kahn CR. Critical nodes in signalling pathways: insights into insulin action. Nat Rev Mol Cell Biol. 2006;7(2):85-96. 67. Cross DA, Alessi DR, Cohen P, Andjelkovich M, Hemmings BA. Inhibition of glycogen synthase kinase-3 by insulin mediated by protein kinase B. Nature. 1995;378(6559):785-789. Jensen J, Brennesvik EO, Lai YC, Shepherd PR. GSK-3beta regulation in skeletal muscles by adrenaline and insulin: evidence that PKA and PKB regulate different pools of GSK-3. Cell Signal. 2007;19(1):204-210. Lawrence JC, Jr., Roach PJ. New insights into the role and mechanism of glycogen synthase activation by insulin. Diabetes. 1997;46(4):541-547. Taniguchi CM, Emanuelli B, Kahn CR. Critical nodes in signalling pathways: insights into insulin action. Nat Rev Mol Cell Biol. 2006;7(2):85-96. 68. 69. 71. Storz P, Toker A. 3'-phosphoinositide-dependent kinase-1 (PDK-1) in PI 3-kinase signaling. Front Biosci. 2002;7:d886-902. 72. Gan X, Wang J, Su B, Wu D. Evidence for direct activation of mTORC2 kinase activity by phosphatidylinositol 3,4,5-trisphosphate. J Biol Chem. 2011;286(13):10998-11002. 73. Yang G, Murashige DS, Humphrey SJ, James DE. A Positive Feedback Loop between Akt and mTORC2 via SIN1 Phosphorylation. Cell Rep. 2015;12(6):937-943. 74. Copp J, Manning G, Hunter T. TORC-specific phosphorylation of mammalian target of rapamycin (mTOR): phospho-Ser2481 is a marker for intact mTOR signaling complex 2. Cancer Res. 2009;69(5):1821-1827. 75. Moore MC, Cherrington AD, Cline G, et al. Sources of carbon for hepatic glycogen synthesis in the conscious dog. J Clin Invest. 1991;88(2):578-587. 76. Agius L. Glucokinase and molecular aspects of liver glycogen metabolism.

Biochem J. 2008;414(1):1-18. 77. Chu CA, Sindelar DK, Neal DW, Cherrington AD. Direct effects of catecholamines on hepatic glucose production in conscious dog are due to glycogenolysis. Am J Physiol. 1996;271(1 Pt 1):E127-137. 78. Syed NA, Khandelwal RL. Reciprocal regulation of glycogen phosphorylase and glycogen synthase by insulin involving phosphatidylinositol-3 kinase and protein phosphatase-1 in HepG2 cells. Mol Cell Biochem. 2000;211(1-2):123-136. 79. Dent P, Lavoinne A, Nakielny S, Caudwell FB, Watt P, Cohen P. The molecular mechanism by which insulin stimulates glycogen synthesis in mammalian skeletal muscle. Nature. 1990;348(6299):302-308. 80. Chen M, Halter JB, Porte D, Jr. Plasma catecholamines, dietary carbohydrate, and glucose intolerance: a comparison between young and old men. J Clin Endocrinol Metab. 1986;62(6):1193-1198. 81. Sarbassov DD, Guertin DA, Ali SM, Sabatini DM. Phosphorylation and regulation of Akt/PKB by the rictor-mTOR complex. Science. 2005;307(5712):1098-1101. 82. Manning BD, Toker A. AKT/PKB Signaling: Navigating the Network. Cell. 2017;169(3):381- 405. 83. Gayen JR, Gu Y, O’Connor DT, Mahata SK. Global disturbances in autonomic function yield cardiovascular instability and hypertension in the chromogranin A null mouse. Endocrinology. 2009;150(11):5027-5035. 27 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 84. Bandyopadhyay GK, Vu CU, Gentile S, et al. Catestatin (chromogranin A(352-372)) and novel effects on mobilization of fat from adipose tissue through regulation of adrenergic and leptin signaling. J Biol Chem. 2012;287(27):23141-23151. 85. Rabbi MF, Munyaka PM, Eissa N, Metz-Boutigue MH, Khafipour E, Ghia JE. Human Catestatin Alters Gut Microbiota Composition in Mice. Front Microbiol. 2016;7:2151. 86. Rabbi MF, Eissa N, Munyaka PM, et al. Reactivation of Intestinal Inflammation Is Suppressed by Catestatin in a Murine Model of Colitis via M1 Macrophages and Not the Gut Microbiota. Front Immunol. 2017;8:985. 87. Muntjewerff EM, Tang K, Lutter L, et al. Chromogranin A regulates gut permeability via the antagonistic actions of its proteolytic peptides. Acta Physiol (Oxf). 2021;232(2):e13655. 88. Ying W, Tang K, Avolio E, et al. Immunosuppression of Macrophages Underlies the Cardioprotective Effects of CST (Catestatin). Hypertension. 2021;77(5):1670-1682. 89. Riopel M, Seo JB, Bandyopadhyay GK, et al. Chronic fractalkine administration improves glucose tolerance and pancreatic endocrine function. J Clin Invest. 2018;128(4):1458-1470. 90. Carroll NV, Longley RW, Roe JH. The determination of glycogen in liver and muscle by use of anthrone reagent. J Biol Chem. 1956;220(2):583-593. 91. Ciraolo E, Gulluni F, Hirsch E. Methods to measure the enzymatic activity of PI3Ks.

Methods Enzymol. 2014;543:115-140. 28 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. NCD-WT C A B Liver glycogen Plasma insulin Liver glycogen Sal: Acu CST: Acu Sal: Chr CST: Chr Fed Fast ) L m / g n ( n 60 2.0 120 ) r e v ) r e v i l ✱ 1.5 90 i l g m g µ ( n e g o c y G g m g µ ( n e g o c y G 40 ✱ ✱ / i l u s n / 1.0 60 i 20 a m s a P 0.5 30 l l 0 0.0 0 l S aline Insulin S aline Insulin F ast F ed Fed Fasting NCD-WT DIO-WT Liver lipid Muscle glycogen Fast: liver glycogen E F G D Sub-sarcolemma Myofibril DIO+Sal DIO+CST 150 ) r e v ) 2 m µ ( s e u n a r g n e g o c y G ) 100 15 150 ✱ S A G g m g µ ( n e g o c y G i l ) a e r a g m g µ ( n e g o c y G 80 100 y t i s n e d d p L 10 100 / l 60 / l a t o t 40 50 5 50 f o % i i 20 l ( l 0 0 0 0 l C hr-C S T +Insulin A cu-C S T +Insulin S aline Insulin DIO + C S T DIO + S al C S T S aline C S T C S T S aline S aline Figure 1 GG bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Hormones: NCD-WT B C A NE, EPI and glucagon NE EPI Fast:Sal Fast:CST Fed:Sal Fed:CST Fed Fast Fed Fast 200 120 150 n o i t a r t n e c n o c a m s a P ) M n ( ) M n ( ns ) l 150 m / g p r o M n ( 80 100 ns I P E a m s a P E N a m s a P ns ns ns 100 ns ✱ ✱ 40 50 50 l l l 0 0 0 Glucagon (pg/ml) N E (n M ) E PI (n M ) S aline Insulin S aline Insulin S aline Insulin S aline Insulin CST (3 days) D CST (1 day) CST (5 days) GTT/ITT/ Tissue harvest CST (10 days) CST (15 days) Normal chow Figh-fat Diet 1d 153d 56d 168d Age Glycogen: DIO-WT E F Fast Fast Fed 120 ) r e v 100 ) r e v i l i l 80 90 g m / g µ ( n e g o c y G g m / g µ ( n e g o c y G 60 60 40 30 20 l l 0 0 C S T C S T S aline S aline S aline C S T +Insulin Insulin Figure 2 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; DIO-WT The copyright holder for this preprint this version posted November 5, 2021. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. A Saline B CST (0.001 µg/g BW) CST (0.01 µg/g BW) DIO-WT: IC50~ 0.41 µg/g BW e s o c u g d o o B l l ) l o r t n o c f o t n e c r e p : L d g m / ( CST (0.1µg/g BW) CST (1µg/g BW) 400 CST (10µg/g BW) 300 200 100 0 15 min 30 min 0 min 60 min 180 min 150 min 120 min 90 min ) n m x L d g m i / ( C U A 2500 2000 1500 1000 500 C S T (0.001µg/g B W ) C S T (0.01µg/g B W ) C S T (10µg/g B W ) C S T (0.1µg/g B W ) C S T (1µg/g B W ) S aline C e s o c u g d o o B l l ) l o r t n o c f o t n e c r e p : L d / g m 500 400 300 200 100 0 Pretreatment O-CST (1 day) O-CST (3 day) O-CST (5 day) O-CST (10 day) D ) n m x L d / g m i ( C U A Pretreatment O-CST (1 day) O-CST (3 day) O-CST (5 day) O-CST (10 day) 3000 ns 2000 ns ns 1000 0 ( 0 min 180 min 150 min 120 min 90 min 60 min 30 min 15 min 5 min E e s o c u g d o o B l l ) l o r t n o c f o t n e c r e p : L d / g m 120 100 80 60 40 Saline CST F ) n m x L d / g m i ( C U A Saline 450 400 350 300 250 CST 200 ( 120 min 90 min 60 min 30 min 15 min 0 min S aline C S T Figure 3 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021.

NCD-CST-KO The copyright holder for this preprint NCD-WT and CST-KO A (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. B Glycogen Glycogen Fast Fed ns Fast Fed 20 40 ) r e v ) r e v 15 i l 30 i l g m / g µ ( n e g o c y G g m / g µ ( n e g o c y G 10 20 ✱ ✱✱ 5 10 0 0 l l S aline Insulin S aline Insulin C S T-K O W T C S T-K O W T NCD-CST-KO C D E Glycogen Glycogen Insulin Fast Fed Fast ) e u s s i t g m / g µ ( n e g o c y G ) e u s s i t g m / g µ ( n e g o c y G 80 20 ) L m g n ( n 5 / 60 4 15 3 i l 40 u s n 10 2 ns i 20 a m s a P 5 1 l 0 l l 0 0 C S T C S T S aline S aline C S T +Ins S aline Insulin F ast F ed F G NE, EPI and glucagon NE and EPI Fed:Insulin Fast:Insulin Fed:Sal Fast:Sal Fed+CST Fast+CST Fed+Sal Fast+Sal 200 300 n o i t a r t n e c n o c a m s a P n o i t a r t n e c n o c a m s a P ns ) l 150 m / g p r o M n ( 200 ns ns ns ) M n ( ns 100 ns ✱ ✱ ns 100 50 l l 0 0 Glucagon (pg/ml) N E (n M ) E PI (n M ) NE EPI Figure 4 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 Glycogenesis ; The copyright holder for this preprint this version posted November 5, 2021. Glycogenolysis (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. A Dose response B Hormone effect C Hormone effect d e a r o p r o c n t i e s o c u G l ) n e o r p g m i t / l o m n ( 5 4 3 2 1 0 EC50 ~7.3 nM 0 1 100 CST (nM) 10 30 1000 d e t a r o p r o c n i e s o c u G l ) n e t o r p g m i / l o m n ( 8 6 4 2 0 ✱ C S T C S T +Insulin Insulin B asal o h p s o h p - n o n & o h p s o h p l a t o T ) n e t o r p g m m p d ( e s o c u g - C 4 1 i / l 400 300 200 100 0 N E + E PI+Ins+ C S T N E + E PI+ C S T N E + E PI N E + E PI+Ins Insulin B asal C S T G6P level in liver D Normal chow E High fat Fed Fast Fed Fast ) r e v 150 ) r e v 200 ✱ ✱ i l g / s e o m n ( l P 6 G 100 50 i l g / s e o m n ( l P 6 G 150 100 50 ns ns 0 0 N C D-C S T N C D-S al N C D-C S T N C D-S al DIO-C S T DIO-S al DIO-C S T DIO-S al UDPG level in fast liver Gys2 activity in fast liver F Normal chow G High fat H Normal chow I High fat ) r e v i l g / s e o m n ( l G P D U 300 200 100 0 ) r e v i l g / s e o m n ( l G P D U 200 150 100 50 o i t a r y t i v i t c a S G ) P 6 G M m 0 1 / P 6 G M m 1 . 0 ( 0.4 0.3 0.2 0.1 0.0 o i t a r y t i v i t c a S G ) P 6 G M m 0 1 / P 6 G M m 1 . 0 ( 0.4 0.3 0.2 0.1 0.0 N C D + C S T N C D + S al DIO + C S T DIO + S al N C D + C S T N C D + S al DIO + C S T DIO + S al Figure 5 A B asal bioRxiv preprint IR blots doi: https://doi.org/10.1101/2020.10.31.363481 ; IRS-1 blots this version posted November 5, 2021. The copyright holder for this preprint B (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. C S T +Ins C S T +Ins Insulin Insulin B asal C S T C S T pY-IR pY-IRS1 IR IRS1 pY-IR pY-IRS1 IR IRS1 pY-IR pY-IRS1 IR IRS1 C IR densitometry 2.5 D IRS-1 densitometry 2.5 ✱ R I / R I - Y p 2.0 1.5 1.0 0.5 ns ns 1 S R I / 1 S R I - Y p 2.0 1.5 1.0 0.5 ns ns 0.0 0.0 B asal C S T C S T +Ins Insulin B asal C S T C S T +Ins Insulin E A nti-p85-IP Pre-IS PI3-P Insulin (10 nM) CST (100 nM) - + - + - + - + - Figure 6 - - bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021.

The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. B PDK1 densitometry A PDK1 blots C ontrol R N A C ontrol R N A si-P D K 1 si-P D K 1 1.2 0 9 P S H / 1 K D P 1.0 PDK1 HSP90 0.8 PDK1 0.6 HSP90 0.4 PDK1 C ontrol R N A siP D K 1 HSP90 C T308-AKT blot D T308-AKT densitometry Insulin Insulin B asal C S T C S T 4 si-PDK1 T K A T K A - 8 0 3 T p pT308-AKT 3 / AKT 2 pT308-AKT 1 AKT pT308-AKT 0 C S T Insulin Ins+si-P D K 1 C S T +si-P D K 1 B asal AKT pT308-AKT AKT Figure 7 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. A. B mTOR blot mTOR densitometry Ins+ C S T C S T Insulin 4 B asal ✱✱ R O T m R O T m p ✱✱✱ 3 pS2481-mTOR mTOR pS2481-mTOR ✱ / 2 1 mTOR pS2481-mTOR 0 Ins+ C S T Insulin B asal C S T mTOR C S473-AKT blot D S473-AKT densitometry Ins+ C S T C S T Insulin B asal 4 ✱ T K A T K A - 3 7 4 S p pS473-AKT ✱✱✱ 3 / AKT ✱ 2 pS473-AKT 1 AKT pS473-AKT 0 Ins+ C S T Insulin B asal C S T AKT Figure 8 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.31.363481 ; this version posted November 5, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. A GSK-3β and GYS2 blots WT-NCD WT-DIO C S T +Ins C S T +Ins C S T C S T k D a S al S al Ins Ins 50 pS9-GSK-3β 37 50 GSK-3β 37 100 pS641-GYS2 75 100 GYS2 75 WT-GSK-3β densitometry WT-GYS2 densitometry C D E B NCD DIO NCD DIO 2.5 3.0 2.0 ✱ 2.5 β 3 - K S G / β 3 - K S G - 9 S p 2 S Y G 2 S Y G - 1 4 6 S p ✱ β 3 - K S G / β 3 - K S G - 9 S p 2 S Y G / 2 S Y G - 1 4 6 S p ns 2.0 ✱ 2.5 1.5 2.0 ns ✱ ✱ / 1.5 2.0 ✱ 1.0 1.5 1.0 1.5 0.5 0.5 1.0 1.0 0.0 0.5 0.0 0.5 C S T C S T +Ins Insulin S aline C S T C S T +Ins Insulin S aline C S T C S T +Ins Insulin S aline C S T C S T +Ins Insulin S aline Figure 9 Glc Insulin bioRxiv preprint https://doi.org/10.1101/2020.10.31.363481 doi: this version posted November 5, 2021. ; The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Glc I I α α IR β β Glc P P P IRS-1 P P P GK Glc G6P mLST8 PRRS Rictor mTOR S2481 P SIN1 PGM mTORC2 UGP1 S473 T308 G1P UDPG P AKT P GSb Active Glycogen PI3K PIP2 CST/ Insulin PDK1 PIP3 S9 P GSK3β Inactive

bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 1 1 The role of genus and life span in predicting seed and vegetative trait variation and correlation in 2 Lathyrus, Phaseolus, and Vicia (Fabaceae) 3 4 Sterling A. Herron1,6, Matthew J. Rubin2, Matthew A. Albrecht3, Quinn G. Long4, Marissa C. Sandoval5, 5 Allison J. Miller1,6 6 7 1Saint Louis University, Department of Biology, 3507 Laclede Ave, St. Louis, MO 63103, USA 8 2Donald Danforth Plant Science Center, 975 North Warson Rd, St. Louis, MO 63132, USA 9 3Missouri Botanical Garden, Center for Conservation and Sustainable Development, 4344 Shaw Blvd, 10 St. Louis, MO 63110, USA 11 4Shaw Nature Reserve, 307 Pinetum Loop Rd, Gray Summit, MO 63039, USA 12 5University of California, Berkeley, Rausser College of Natural Resources, 319 Wellman Hall, Berkeley, 13 CA 94704, USA 14 6Authors for correspondence (email: [email protected]; [email protected]); phone: +1- 15 314-587-1232 16 17 Manuscript received _______; revision accepted _______. 18 19 Running head: Genus and life span trait variation and correlation in Fabaceae 20 21 ABSTRACT 22 23 PREMISE OF THE STUDY: Annual and perennial life history transitions are abundant among 24 angiosperms, and understanding the phenotypic variation underlying life span shifts is a key endeavor of 25 plant evolutionary biology. Comparative analyses of trait variation and correlation networks among 26 annual and perennial plants is increasingly important as new perennial crops are being developed in a bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 2 27 predominately annual-based agricultural setting. However, it remains unclear how seed to vegetative 28 growth trait relationships may correlate with life span. 29 30 METHODS: We measured 29 annual and perennial congeneric species of three herbaceous legume 31 genera (Lathyrus, Phaseolus, and Vicia) for seed size and shape, germination proportion, and early 32 vegetative height and leaf growth over three months in order to assess relative roles of genus and life span 33 in predicting phenotypic variation and correlation. 34 35 KEY RESULTS: Genus was the greatest predictor of seed size and shape variation, while life span 36 consistently predicted static vegetative growth traits. Correlation networks revealed that annual species 37 had significant associations between seed traits and vegetative traits, while perennials had no significant 38 seed-vegetative associations.

Each genus also differed in the extent of integration between seed and 39 vegetative traits, as well as within-vegetative trait correlation patterns. 40 41 CONCLUSIONS: Genus and life span were important for predicting aspects of early life stage phenotypic 42 variation and trait relationships. Differences in phenotypic correlation may indicate selection on seed size 43 traits will impact vegetative growth differently depending on life span, which has important implications 44 for nascent perennial breeding programs. 45 46 Keywords: annual; crop wild relative; herbaceous perennial; legume; life history; life history strategy; 47 perennial; perennial grain; phenotypic correlation; trait network. 48 49 Life history strategy in plants involves complex patterns of reproductive, growth, and survival trait 50 allocation and trade-offs, which can shed light on past adaptive drivers and influence future evolutionary 51 trajectories (Stearns, 1992). Describing ecological, genetic, and phenotypic life history patterns and 52 commonalities among diverse taxa is thus fundamental in advancing evolutionary biology (Díaz et al., bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 3 53 2016; Friedman and Rubin, 2015). Annuals and perennials are two broadly recognized life span 54 categories in plants (Friedman, 2020). While annual plants grow, reproduce, and senesce within a single 55 year, perennial plants vary from small, short-lived herbaceous individuals that live 3-5 years to expansive, 56 ancient clonal colonies (e.g., Populus) and woody individuals (e.g., Sequoia) that can live for hundreds or 57 thousands of years. Although traits associated with annual-perennial differences are well-defined in some 58 model systems, this comparative framework has garnered renewed attention with increasing focus on 59 understudied, wild species as sources of novel crop breeding material, particularly perennial species 60 closely related to annual crops, such as corn, rice, and wheat (Lundgren and Des Marais, 2020). 61 Additional analyses across diverse, non-model systems are needed to determine the phenotypic and 62 genetic factors underlying differences in annual and perennial species (Friedman, 2020). 63 64 Several broad life history frameworks have been proposed to interpret the large range of phenotypic 65 variation and correlation in plants. Classic plant life history theory balances an individual’s survival with 66 reproduction as modes of achieving fitness (Cole, 1954; Charnov and Schaffer, 1973; Gadgil and Solbrig, 67 1972). Derivations of this have described systems of interconnected phenotypes which span predictable 68 trait networks or spectra (e.g., Grime, 1977; Diaz et al., 2016).

One example, the plant economic 69 spectrum, categorizes plants by their resource-use strategy: resource-acquisitive plants use available 70 resources for immediate structural growth, and resource-conservative plants invest resources in expensive 71 storage organs to prioritize future growth and survival (Chapin, 1980; Chapin et al., 1990; Reich, 2014). 72 The plant economic spectrum has been primarily studied in leaves, with acquisitive plants having broad, 73 short-lived leaves and a high photosynthetic rate, and with conservative plants having thick, long-lived 74 leaves and a lower photosynthetic rate (Wright et al., 2004). This economic framework has also been 75 extended to stems (Chave et al., 2009) and roots (Roumet etl al., 2016). 76 77 Within the framework of the plant economic spectrum, annuals are generally categorized as acquisitive 78 species and perennials as conservative species. Evidence for differences in traits associated with life span bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 4 79 comes largely from leaf and shoot traits (e.g., specific leaf area, leaf water content, and relative growth 80 rate are greater in annuals than perennials; Garnier, 1992; Garnier and Laurent, 1994; Atkinson et al., 81 2016), reproductive traits (e.g., seed mass fraction and reproductive energy percent allocation are greater 82 in annuals; Pitelka, 1977; Vico et al., 2016) and root traits (e.g., specific root length and root nitrogen 83 concentration are greater in annuals, root tissue density and root diameter are greater in perennials; 84 Roumet et al., 2006). Nevertheless, it has long been recognized that plant species of different life histories 85 are capable of having diverse trait systems that are not always predicted by generalized models; these 86 traits are ultimately determined by the species’ distinct environmental stressors and evolutionary history 87 (Thompson and Hodkinson, 1998; Crews and DeHaan, 2015). Herbaceous perennials as a group have 88 been recognized as particularly variable in life history trait combinations (Grime, 1977; Kitchen, 1994). 89 Several studies have found evidence for wild herbaceous perennial species with acquisitive and ruderal 90 traits, including comparable reproductive biomass and relative growth rate to annuals (Verboom et al., 91 2004; González-Paleo and Ravetta, 2015; Vico et al., 2016). 92 93 In order to understand evolutionary patterns of life history strategy, it is helpful to study phenotypic 94 variation associated with life span within and among different phylogenetic lineages. Life span transitions 95 are common in many angiosperm systems (e.g., Castillejinae: Tank and Olmstead, 2008; core Pooideae: 96 Lindberg et al., 2020; Saxifragales: Soltis et al., 2013), which allows for investigation into phenotypic and 97 genetic commonalities in different life spans.

Previous studies in grasses and other systems have found 98 similar annual-perennial phenotypic differences across multiple genera, including in shoot, leaf, root, and 99 reproductive traits (as discussed above; Garnier, 1992; Garnier and Laurent, 1994; Garnier and 100 Vancaeyzeele, 1994; Vico et al., 2016), and accounting for phylogenetic relationship of the studied taxa 101 has shown to be key in discerning life span differences apart from other evolutionary influences (Garnier, 102 1992). However, the phenotypic and genetic factors underlying life history and life span variation remain 103 uncharacterized in many diverse plant groups. 104 bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 5 105 Networks of trait correlation are also fundamental to studies of life history diversity. Determining 106 relationships between plant organs, or phenotypic integration, helps to conclude whether key traits were 107 consistently selected together as a functional module, if trade-offs are occurring, or if the traits in question 108 remain relatively independent (Murren, 2002). These trait relationships can clarify fundamental 109 constraints on plant phenotypic diversity. Previous studies in herbaceous species have found negative 110 relationships between vegetative growth and some aspects of reproduction (e.g., earlier flowering; Geber, 111 1990) but generally positive relationships between vegetative size traits and size of seeds produced within 112 a species (Geber, 1990) and among species (Kleyer et al., 2019). Other studies have further found that life 113 span plays an important role in predicting functional reproductive-vegetative trait relationships, including 114 the association of flowering time with seed size and plant height, with seed size being derived from 115 separate studies (Bolmgren and Cowen, 2008; Du and Qi, 2010) or the size of seeds produced by the 116 plants on which vegetative traits were measured (Segrestin et al., 2020), and between leaf area and seed 117 size (Hodgson et al., 2017). However, to our knowledge few studies have focused in detail on congeneric 118 annual-perennial differences in seed size vs. vegetative plant size / growth rate correlations. Here we seek 119 to build on previous life history studies by investigating specifically how relationships between seed and 120 vegetative traits in multiple genera may shift with life span. 121 122 Due to its economic and ecological importance, as well as biological diversity, the legume family 123 (Fabaceae Lindl.) is an excellent system in which to study phenotypic patterns associated with life span. 124 Specifically, subfamily Papilionoideae DC. contains numerous genera of agricultural importance with 125 annual and herbaceous perennial congeners (Ciotir et al., 2019).

Similar to other systems, annual legume 126 species tend to have greater relative allocation to sexual reproduction than perennial congeners in terms of 127 percent energy allocation (Pitelka, 1977) and percent dry biomass allocation (Turkington and Cavers, 128 1978). Studies have also explored other functionally important traits in congeneric annual and perennial 129 legumes, including seed mass (Pitelka, 1977; Marshall et al., 1985; Kelly and Hanley, 2005; Ward et al., 130 2011; Herron et al., 2020); specific leaf area (den Dubbelden and Verburg, 1996; Roumet et al., 2000); bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 6 131 photosynthetic rate (Pitelka, 1977; Roumet et al., 2000); and growth rate (den Dubbelden and Verburg, 132 1996; Kelly and Hanley, 2005). In these studies, life span was commonly not the central focus and results 133 were mixed, or there were equivocal patterns in terms of whether annuals and perennials showed greater 134 trait values. These studies also tend to focus on two or three species per genus, while a much greater 135 range of diversity exists for both annuals and perennials within genera and across Fabaceae (Ciotir et al., 136 2019). 137 138 Exploring patterns of life span-associated differences in seeds and vegetative traits will be valuable not 139 only for better understanding their evolutionary background but also for informing crop breeding. 140 Although herbaceous perennial species were generally not domesticated by early farmers (Van Tassel et 141 al., 2010), attention is now focusing on these species as a potential means to slow or reverse soil erosion 142 associated with annual grain agriculture and to sustainably yield in marginal environments (Glover et al., 143 2010; Ryan et al., 2018). An increasing number of diverse crop wild relatives are being considered for 144 trait conferral via introgression or de novo domestication, including herbaceous perennial grains and 145 pulses. Capturing the range of genetic variation available in wild germplasm will be essential for guiding 146 breeding decisions (Schlautman et al., 2018; Smýkal et al., 2018). Characterizing phenotypic correlation 147 within plants may form a predictive foundation for early-stage selection based on seeds, and will 148 additionally inform whether breeding for certain phenotypes may lead to trade-offs with other desired 149 traits, e.g., resource competition between reproductive and vegetative organs such as roots (González- 150 Paleo et al., 2016; Pastor-Pastor et al., 2018). 151 152 Here we focus on a panel of annual and herbaceous perennial species from the agriculturally important 153 Fabaceae genera Lathyrus L., Phaseolus L., and Vicia L. in order to better characterize phenotypic 154 diversity associated with life span and add to our understanding of trait variation and correlation 155 associated with different life history strategies.

Starting with seeds, we measured phenotypic variation in 156 early life stages: seed size and shape, germination, and seedling vegetative growth. We ask the following bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 7 157 questions: 1) What is the relative importance of genus and life span in predicting seed and vegetative trait 158 variation? 2) How do seed and vegetative trait correlations differ across genera and life spans? We 159 address these questions by quantifying seed size, germination, and early life stage vegetative growth traits 160 in congeneric wild, annual and perennial species grown in a common environment. 161 162 MATERIALS AND METHODS 163 164 Plant material—We selected annual and herbaceous perennial species from three economically 165 important legume genera: Lathyrus (grass pea), Phaseolus (common bean), and Vicia (vetch). We 166 obtained 80 accessions (Appendix S1) from the United States Department of Agriculture’s National Plant 167 Germplasm System (Western Regional Plant Introduction Station, Pullman, Washington, USA) in spring 168 2017, which were stored in a desiccator at 3 to 4.5°C, 36% relative humidity prior to the start of the 169 experiment. All accessions used were designated as “wild” by the USDA (with the exception of one 170 accession of V. americana, see below), meaning that seed collection took place in a natural population 171 outside of cultivation, although this designation may include feral and/or naturalized escapees from 172 cultivation in the case of some species collected outside their native range (primarily Lathyrus collected 173 in the U.S.; Kenicer, 2008). All accessions were harvested from a seed increase effort at the germplasm 174 center, except for one accession of Lathyrus japonicus (W6 45319) and L. latifolius (PI 602368), which 175 were collections made directly from the original wild population. Time of seed increase / collection 176 ranged from two to 30 years prior to this study, the duration of which seeds were kept in frozen storage. 177 For this experiment, accessions were germinated and then grown from June to September 2017. For 178 Lathyrus, five annual and five perennial species were studied; for Phaseolus, four annual species and five 179 perennial species; and for Vicia, five annual species and five perennial species (Table 1). One to seven 180 accessions were obtained for each species, with origins from diverse geographic areas when possible. All 181 accession information is available in Appendix S1 (see Supplemental Data with this article). 182 bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 8 183 When possible, we obtained accessions of annual and perennial species from closely related phylogenetic 184 groups based on the most recently published, comprehensive molecular phylogenies of each genus. 185 Lathyrus and Vicia are both within the economically important tribe Fabeae, which also includes pea, 186 lentil, and grass pea (Schaefer et al., 2012). Both genera consist of c. 160 herbaceous annual and perennial 187 species distributed globally (Lewis et al., 2005). Within Lathyrus, clades sampled here include: 1) the 188 clade including sections Aphaca (L. aphaca), Pratensis (L. pratensis), and Orobus (L. japonicus); and 2) 189 the large section Lathyrus (L. annuus, L. cicera, L. hirsutus, L. odoratus, L. latifolius, L. sylvestris, and L. 190 tuberosus) (Asmussen and Liston, 1998; Schaefer et al., 2012; Table 1). Within Vicia, clades sampled 191 here included 1) section Pedunculatae (V. americana, V. dumetorum); 2) section Ervilia (V. ervilia, V. 192 hirsuta); 3) section Vicia (V. sepium and V. sativa); and 4) section Cracca (V. cracca, V. tenuifolia, V. 193 villosa, and V. benghalensis) (Endo et al., 2008; Castiglione et al., 2011; Schaefer et al., 2012; Jaaska, 194 2015; Table 1). Notably, section Ervilia is more distantly related to other sections of Vicia; it is sister to 195 all clades of the tribe Fabeae (including Vicia and Lathyrus, among other genera), and it has been 196 proposed that it be raised to genus level (Schaefer et al., 2012). These changes have yet to be finalized, so 197 we retain the current taxonomy of section Ervilia. Phaseolus, a monophyletic genus of c. 70-80 species, is 198 more distantly related to Lathyrus and Vicia and is in tribe Phaseoleae (Freytag and Debouck, 2002; 199 Delgado-Salinas et al., 2006). For Phaseolus, two broad clade groups were sampled here: the 1) the 200 Filiformis group (P. angustissimus and P. filiformis) and the Vulgaris group (P. acutifolius and P. 201 vulgaris); and 2) the Polystachios group (P. maculatus and P. polystachios), the Lunatus group (P. 202 lunatus), and the Leptostachyus group (P. leptostachyus) (Delgado-Salinas et al., 2006; Table 1). Also 203 included is the more distantly related P. parvulus in the Pauciflorus group (Delgado-Salinas et al., 2006; 204 Table 1). The USDA has since updated the taxonomic status of P. maculatus subsp. ritensis to P. ritensis, 205 but we retain its original taxonomic status here (Freytag and Debouck, 2002). The Lathyrus species in this 206 study are primarily native to temperate and subtropical regions of Eurasia, with L. japonicus having a 207 global temperate distribution (Wu et al. 2010; Schaefer et al., 2012; Global Biodiversity Information 208 Facility; http://www.gbif.org). Vicia species in this study are native to temperate regions of Eurasia bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 9 209 (though some also extend to subtropical regions), with the exception of the temperate North American 210 species V. americana (Wu et al. 2010; Schaefer et al., 2012; GBIF.org). Most Phaseolus species here are 211 from arid climates in the southwestern U.S. into northwest Mexico, or are native to dry or moist habitats 212 in Mexico, Central America and South America, with P. polystachios standing out as the only Phaseolus 213 species whose native range extends into temperate regions (Freytag and Debouck, 2002; Bitocchi et al. 214 2017; GBIF.org). 215 216 For each accession used in this study life span and cultivation status were derived from the online USDA 217 description in GRIN-Global (Germplasm Resources Information Network) and confirmed with literature 218 (Wu et al. 2010; Schaefer et al., 2012; Freytag & Debouck 2002). We use the annual and perennial life 219 span classifications only as a preliminary framework for understanding trait variation, acknowledging that 220 life span in the strict sense is only part of a broader, complex life history strategy. In the cases where the 221 USDA life span assignment (annual or perennial) and literature were contradictory, we conducted 222 extensive literature research and expert consultation (Daniel Debouck, pers. comm., 2020). This occurred 223 for Phaseolus filiformis, a Southwestern U.S. desert ephemeral (annual) (Buhrow, 1983; Nabhan and 224 Felger, 1985; Freytag and Debouck, 2002), and P. lunatus, a species with an expansive distribution from 225 Mexico to South America and which is variable in life span depending on the minimum temperature and 226 the severity of the dry season where it is growing (Freytag and Debouck, 2002; Bitocchi et al., 2017). 227 Although P. lunatus is capable of perenniality in mild, aseasonal climates, unlike other perennials in this 228 study its fibrous root system lacks substantial reserves and is not able to endure extended harsh conditions 229 (Freytag and Debouck, 2002); thus, in the absence of specific life span information for source populations 230 we deferred to the USDA’s original assignment of annual. Due to life span ambiguity, the linear models 231 were also tested with P. lunatus accessions removed to confirm that it was not driving any trends (see 232 below). If there was uncertainty regarding the cultivation status or origin of an accession, the original 233 collection information was consulted and collectors contacted when possible to confirm our information. 234 For example, the accession of V. americana from Canada (PI 452486) was noted as “cultivated” in the bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021.

The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 10 235 USDA system; while it may sometimes be seeded in restoration efforts, V. americana is not known to be 236 domesticated and this accession’s likely wild status was confirmed by the Plant Gene Resources of 237 Canada, which also curates the accession. 238 239 Seed size and shape measurements— Seed mass and two-dimensional seed shape parameters 240 were characterized for each accession prior to sowing. A pool of seeds for each accession was first 241 weighed on an Ohaus® Adventurer® Pro precision balance (Parsippany, New Jersey, USA) to the nearest 242 0.1 mg and divided by the total number of seeds for that accession to estimate mean single seed mass. 243 Small funicles (rare) were not removed before weighing, except in the case of Vicia sepium, where 244 uniquely conspicuous funicles were present on a few seeds. Seeds were then scanned at a resolution of 245 1200 dpi using an EPSON DS-50000 scanner (Nagano, Japan). When seeds were bilaterally symmetric 246 (in Lathyrus cicera and all Phaseolus), they were oriented with the flat (lateral) side facing down and the 247 hilum parallel lengthwise to the scanner surface. After scanning, images were converted from a CMYK 248 jpg to png (400dpi) and analyzed in ImageJ (Schneider et al., 2012). Severely damaged seeds and seeds in 249 a non-standardized orientation (only occurring for a few bilaterally symmetric seeds of Lathyrus cicera) 250 were removed from the analysis. In ImageJ seed images were cropped and converted to binary (“Make 251 Binary” function), or if the contrast was not well defined, the image was converted to 8-bit grayscale and 252 a binary threshold (“Threshold” function) was applied and the threshold value adjusted for the highest 253 seed contrast and lowest noise (shadows). Remaining pixel holes within seeds were removed using the 254 “Fill Holes” function, and erroneous gaps on the perimeter of seeds were filled using the “Convex Hull” 255 function (gift wrapping algorithm) or traced in manually when this was not adequate. Accessory 256 structures still attached to the seeds (primarily funicles) were manually removed from the image. From 257 this final image we extracted four size parameters per seed: length (Feret’s diameter, or the maximum 258 distance between any two points along the perimeter of the seed), width (minimum distance between any 259 two points along the perimeter of the seed), perimeter, and area, and two shape parameters, circularity and 260 roundness. Circularity is calculated as 4 π × (Area)/(Perimeter)2, and represents the extent to which the bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license .

11 261 seed shape approximates a circle, ranging from 0 (infinitely elongated shape) to 1 (perfect circle). 262 Roundness is calculated as 4 × (Area)/( π × (Major axis length)2) and is the inverse of the seed’s aspect 263 ratio (length to width ratio of the best fitting ellipse); as this value approaches 1 it represents a width that 264 is closer to the length of the seed. A majority of the seeds measured for size and shape were randomly 265 chosen for use in germination experiments. In this study, seed traits are the initial juvenile stage in plants 266 later measured for vegetative growth. 267 268 Germination measurements— All seeds were sterilized in 6% sodium hypochlorite aqueous 269 solution for 5 to 6.5 minutes, then rinsed with water purified via reverse osmosis (RO water) and patted 270 dry (Frehner and Conn, 1987; Galasso et al., 1997). Due to the prevalence of physical dormancy induced 271 by a water-impermeable seed coat in legumes (Baskin and Baskin, 2014), all seeds were scarified 272 collectively by accession with sandpaper, pressing firmly until a breach of the seed coat was visible on the 273 majority of seeds; this was completed in short (~10 sec) bursts, usually 3-5 times. This was done without 274 specificity to any part of the seed scarified, to simulate heterogeneous, natural scarification. P100 or P60 275 grade sandpaper was used depending on the observed hardness of the seed coat. A subset of seeds was 276 taken randomly from each accession for germination, excluding malformed and broken seeds. Minimal 277 damage occurred from the germination protocol, except for one accession of Vicia hirsuta (PI 219631) 278 where about half of the seeds were damaged from bleach sterilization, which were not used. Germination 279 models were tested with this accession removed to determine any impact (see sect. 2.3). All scarified 280 seeds were surface-sown in unsterilized quartz sand (Fairmount Santrol Handy Sand (Chesterland, Ohio, 281 USA); approximately 34 mL) in 20 mm deep plastic petri dishes; seeds were oriented horizontally 282 (morphology allowing, hilum parallel lengthwise to substrate surface) in a grid pattern. Dishes were 283 initially watered to saturation (~11-12 mL) and were remoistened after one week and subsequently when 284 dry. Excess water and condensation were blotted off. All seeds were germinated in 12:12 h light:dark 285 conditions inside a temperature-controlled incubator, with 20:10°C and 25:15°C light:dark temperature 286 settings for temperate (“temperate settings”) and subtropical / tropical species (“tropical settings”), bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 12 287 respectively, in order to expose the species to temperatures similar to their native range.

Vicia seeds were 288 incubated at temperate settings, Phaseolus seeds were incubated at tropical settings , and Lathyrus seeds 289 were divided into both settings depending on the species (Lathyrus annuals included both temperate and 290 subtropical / Mediterranean species, perennials only temperate) (Appendix S1). Each accession had 2-3 291 replicate petri dishes and 5-26 seeds per replicate. Replicate dishes were randomized when placed in the 292 incubator and re-randomized after each germination check. Germination was defined as an extension of 293 the radicle past the seed coat, or in rare cases backwards emergence of the seedling due to the radicle 294 pushing against the seed coat. 295 296 Germinated and imbibed seeds were counted beginning one day following placement on the substrate, 297 then every one to four days up to 10-12 days, and then at three weeks and/or four weeks, or until all 298 imbibed, viable seeds germinated (Baskin and Baskin, 2014). A subset of four to 38 (mean 15) vigorous 299 seedlings were planted soon after germination, in order to have at least three separate replicates of five 300 seedlings for each accession in most cases (see vegetative growth section). The remaining seedlings were 301 returned to that accession’s petri dish; after this point, germination counts were based on the subtraction 302 of ungerminated seeds from the original total seeds in the petri dish. Seeds which decayed and developed 303 severe fungal infections were considered nonviable and removed but still included in the total count since 304 their nonviability was a pre-existing property of the seedstock. From the germination counts, days to 50% 305 germination (T50) was calculated as a measure of germination time, using the “PROBIT” procedure in 306 SAS 9.4 software, which calculates a maximum likelihood estimate of germination timing with a default 307 maximum iteration of 50 (University Edition; SAS Institute, 2017). In the case of germination increasing 308 from 0 to 90-100% between only two time points, model convergence was not attained, and thus the T50 309 value estimated was that of the last maximum likelihood iteration, which was approximately a linear 310 midpoint between the two time points (this occurred for 37 / 225 data points). Negative T50 values were 311 excluded (1 / 225). At least two replicate dishes had to have at least five viable, imbibed seeds and at least 312 one germinant in the allotted time in order to be included for germination T50 analyses; most accessions bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 13 313 exceeded this. Seeds remaining un-imbibed after four weeks were re-scarified precisely using a scalpel 314 and placed with the scar-side down onto the water-saturated sand substrate to determine if their non- 315 germination was due to insufficient scarification; these seeds were then checked weekly for two to three 316 weeks.

The final total of germinated seeds was used to calculate germination proportion by accession. 317 Seeds which were damaged from scarification but still germinated were included in germination 318 proportion but not germination T50 due to potential changes in germination rate; seeds which were 319 severely damaged from scarification were excluded for both traits. Accession age (years in frozen 320 storage) was the only covariate for germination T50 and proportion. Any ungerminated, imbibed seeds by 321 the end of the experiment that were not severely decayed were tested for viability by bisecting the seeds 322 and treating them with tetrazolium chloride (TZ) overnight (Gosling, 2009; Patil and Dadlani, 2009). 323 Dormancy (if present) was broken by physical scarification for most accessions; only two accessions had 324 >5 imbibed, still viable, ungerminated seeds after all germination tests: W6 2427 (Lathyrus aphaca: 84% 325 viable ungerminated; 0% germinated) and PI 494749 (Vicia dumetorum: 22% viable ungerminated; 25% 326 germinated). This viability was considered evidence that some physiological requirements for 327 germination were not completely met in these accessions; thus, their data for germination T50 and 328 proportion was dropped from further analyses. 329 330 Vegetative growth measurements—A subset of seedlings from at least two petri dish replicates 331 of each accession was planted in Ball Professional Growing Mix (no peat; West Chicago, Illinois, USA) 332 within 38-cell trays (cell diameter: 1.95”, cell depth: 4.98”) as soon as possible following germination. 333 For each accession, 4 - 38 seedlings (mean: 15.3 +/- 5.6; after removal of any problematic data) were 334 transplanted and measured and were arranged in replicate groups across trays (1-7 replicates per 335 accession, mean: 3.6); since a subset of perennial seedlings were retained for year two growth, usually 336 more replicates were planted for perennial than annual accessions. From June 19, 2017 - August 25, 2017, 337 plants grew in hoophouse A-6 at the Missouri Botanical Garden (St. Louis, Missouri, USA), which was 338 covered in a light shade cloth; day-night temperatures ranged from 14 - 40°C. Plants were watered bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 14 339 typically daily, and 150 ppm 15-5-15 NPK aqueous fertilizer was applied approximately weekly, 340 beginning the earliest week that any plants were measured (for day 21 growth; see below for details); 341 days from most recent fertilization ranged from one to nine days, with the majority measured within five 342 days. Exceptions included some replicates of nine accessions, where their first measurements were taken 343 after less than a full day had elapsed from the most recent fertilization, and one accession that had to be 344 measured after transport to a different greenhouse (see below), which was measured 13 days after the 345 most recent fertilization (at 35 days after planting).

We tested our models with these plants dropped 346 respectively to determine if this influenced the results (statistical analysis section). Beginning at the first 347 measurement (~21 days after planting), climbing plants were trained around thin bamboo poles for taller 348 plants (usually Phaseolus) or 18-inch hyacinth sticks for shorter plants. Growth trays were spatially 349 randomized in the hoophouse on August 1, 2017, to avoid potential microclimate heterogeneity. 350 Pesticides were applied on August 15, 2017 (after the majority of plants had been measured) in the 351 hoophouse to control thrips, consisting of a liquid cocktail of Pylon (3mL/gal), Mavrik (3mL/gal), and ⅛ 352 Tristar ( teaspoon/quart). A subset of plants had to be transferred on August 25, 2017 to the Saint Louis 353 University (SLU) Biology Department greenhouse (see growth measurements below), where 13 354 accessions of a mix of annual and perennial Lathyrus and Vicia had their second measurement taken over 355 the course of two weeks after transfer (up to September 5); here plants were watered every few days, 356 temperatures were similar to the previous hoophouse, but there was somewhat heterogeneous lighting. 357 Due to any impact of transfer or environmental differences, linear models were also tested with these 358 accessions dropped (statistical analysis section). 359 360 Vegetative growth was assessed non-destructively for each individual plant over a two-week period. 361 Planting date in the hoophouse was used as a baseline, since seedlings began to show most growth 362 development after this date. All seedlings were first assessed for height, leaf number, vigor, and 363 reproductive status beginning 20-22 days after planting (DAP-21; time point one) in the hoophouse, with 364 the same measurements taken at 34-37 days after planting (DAP-35; time point two) to determine growth bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 15 365 rate on an individual plant level, as well as the singular height and leaf measurements at the respective 366 time points (“static” growth measurements). Stem height was measured from ground level to the base of 367 the shoot apex on the tallest stem. Leaf number was counted for each node on the same stem measured for 368 height, up to the most recently developed countable node, at least ~2 mm from the next node near the 369 shoot apex. If a potential node was leafless, a leaf was only counted if there was a clear remaining 370 structure indicating a leaf was present, such as stipules. For all Phaseolus species in this study, two 371 unifoliate eophylls are present at the first true node and thus two leaves were counted for that node for all 372 individuals, even if they had dehisced.

Otherwise, there were only single, alternate leaves present along 373 the stem for all genera, with some variation in the number of leaflets per leaf. Lastly, the leaves of 374 Lathyrus aphaca are reduced to tendrils, and the stipules are enlarged, foliaceous, and are functionally the 375 main photosynthetic organ of the plant; thus each pair of stipules was counted as one leaf (Sharma and 376 Kumar, 2012). Absolute growth rate (AGR) was calculated on a per-day basis by dividing the difference 377 in height and leaf number by the number of days elapsed between the two time points (as defined in Rees 378 et al., 2010; Pommerening and Muszta, 2016). Relative growth rate (RGR) was calculated for height and 379 leaf number by taking the difference of the natural log of the trait at both dates and dividing by the days 380 elapsed: (ln(trait DAP-35) – ln(trait DAP-21)/(time point 2 – time point 1) (as in Perez-Harguindeguy et 381 al., 2013). Leaf number RGR is also known as relative leaf production rate (RLPR; Garnier, 1992; 382 Grotkopp et al., 2002). 383 384 Plants with severe damage were removed from analysis; this included large cuts, bends, or extensive dead 385 tissue in the stem being measured which could have compromised growth, or otherwise irreparable 386 damage to the whole plant. If the damage occurred by DAP-21, all vegetative data was excluded; if the 387 damage occurred by DAP-35, only DAP-35, AGR, and RGR data were removed. Any accessions were 388 dropped for a trait if fewer than three total plants remained following removal of problematic data. A 389 vigor covariate was assessed for each plant qualitatively at both static time points, which considered the 390 impacts of environmental and endogenous factors; this was scored categorically as “low” (high damage bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 16 391 and/or was unhealthy, with a severe impact on growth), “medium” (average growth, with moderate to 392 high damage or generally unhealthy, but with a nonlethal impact on growth), or “high” (little to no 393 damage and an overall robust form, with effectively unimpeded growth). The lowest vigor observed 394 between the two dates was assigned as the covariate for AGR and RGR analyses. Due to the possible 395 preferential allocation of energy to reproductive structures, we scored the reproductive status of the plant 396 (“reproductive” or “nonreproductive”) at both dates measured (if either date was “reproductive”, that was 397 assigned to AGR and RGR as well). Lastly, for AGR and RGR, height at DAP-21 was included as a 398 covariate to help control for the effect of different developmental sizes at the start of the growth period. 399 Each of these covariates were tested as random effects in our linear models (statistical analysis section).

400 401 Statistical analyses—Statistical analyses were performed in R v. 3.6.1 (R Core Team, 2019). All 402 analyses were performed on the dataset following removal of any problematic data. Principal component 403 analyses (PCAs) were computed on data scaled to unit variance using the “prcomp” function (base R). 404 PCA was first implemented for accession-level data for all traits in the full dataset (Fig. 1) and then for 405 individual-level seed and vegetative datasets for each genus separately (Figs. S1-S3). In the case of 406 individual-level vegetative data, individuals were dropped if measurement data was missing from an 407 entire date (DAP-21 or DAP-35; and thus growth rate was missing as well). For both accession and 408 individual-level PCAs, any remaining missing data values were imputed using a regularized iterative PCA 409 algorithm (“imputePCA” function in the “missMDA” package; Josse and Husson, 2016), for which the 410 number of components were estimated using generalized cross-validation ( “estim_ncpPCA” function). 411 PCs were retained which had an eigenvalue of 1 or greater; in each case these PCs cumulatively explained 412 at least >80% of the total variation in the dataset. Eigenvalues were derived using the “get_eigenvalue” 413 function from the “factoextra” package (Kassambara and Mundt, 2020). PCA figures were created using 414 the package “ggplot2” (Wickham, 2016), and the variable PCA plot was created using the package 415 “ggbiplot” (Vu, 2011). For individual-level genus seed and vegetative PCAs (Fig. S1-S3), PC axes were 416 rotated so they could all be visualized in the same orientation. Accessions and individuals with much bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 17 417 reduced data (about half of the traits missing) or which had removed outliers (see below) were excluded 418 from PCAs. Based on these criteria, six accessions were dropped from the accession-level PCA, one each 419 of Lathyrus annuus, L. aphaca, L. cicera, Phaseolus acutifolius, Vicia dumetorum, and V. hirsuta, and 420 imputation was necessary for 7 out of 1,258 data values. For individual-level Lathyrus vegetative data, 44 421 out of 370 individuals were dropped and six out of 2608 data values were imputed. For Phaseolus, 50 out 422 of 474 individuals were dropped and no imputation was necessary. For Vicia, 36 out of 353 individuals 423 were dropped and one out of 2536 data values was imputed. There were no missing values for individual 424 seed data of any genus. 425 426 Linear mixed models were employed on a trait-by-trait basis to assess the contribution of genus, life span, 427 species, and covariates on phenotypic variation using the “lmer” function in the R package “lme4” (Bates 428 et al., 2015).

For all ImageJ seed traits (all seed traits excluding seed mass) and all vegetative traits, 429 models were tested on individual seed- and plant-level data respectively. For germination T50, the model 430 was tested on individual replicate-level data. The remaining trait models used accession-level data. Each 431 mixed model included genus, life span, genus × life span, and species (nested within genus × life span) as 432 the main fixed effects and accession nested within species as a random effect (except for PC1, PC2, seed 433 mass, and germination proportion, since they measured at the accession level). For growth traits, replicate 434 nested within accession was additionally included in the model as a random effect. Trait models were 435 assessed using a type III ANOVA with Satterthwaite’s method for degrees of freedom, with the exception 436 of PC, seed mass, and germination proportion models, which were assessed using a type I ANOVA (and 437 the base R “lm” function) due to the data consisting of only accession-level means with no significant 438 random effects. Type III ANOVA significance was assessed using the package “afex” (Singmann et al., 439 2020) along with “lme4.” Using the R package “emmeans,” adjusted means for each genus-life span 440 combination were derived from our models (“emmeans” function) and post hoc custom contrasts were 441 conducted using a Bonferroni correction (“contrast” function; Lenth, 2020). The mentioned covariates 442 under germination and vegetative measurements were added as random effects for the appropriate trait bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 18 443 model. The significance of random effects in our models was assessed by taking the likelihood ratio test 444 (LRT) of the model following sequential removal of each random term, using the “rand” function in the 445 package “lmerTest” (Kuznetsova et al., 2017); nonsignificant random effects were dropped from the final 446 model (the “reduced model”). In some cases, certain accessions or species were dropped from models 447 (mentioned above) in order to assess their influence where potential confounding factors were present. 448 449 Trait-by-trait correlations were calculated in order to identify phenotypic relationships in the full dataset 450 and within each genus and life span group. Correlation matrices were generated from accession-level data 451 using Pearson correlations (“cor” function in base R) on scaled and centered data, with p values being 452 assessed at a 95% confidence level (“cor.mtest” function in the “corrplot” package; Wei and Simko, 453 2017). Correlation networks were created from this correlation data using the package “igraph”, with 454 nodes representing traits (organized with “tkplot” function; Csardi and Nepusz, 2006).

Nonsignificant 455 correlations (Pearson; P < 0.05) were excluded from correlation networks. Full correlation matrix plots 456 for each data group were also generated from Pearson correlation coefficient values using the “corrplot” 457 function (“corrplot” package). 458 459 Accessions with trait values which were statistical outliers in their genus, defined as being greater than 460 1.5 × the interquartile range below or above the 25% and 75% quartiles, respectively, were removed from 461 all analyses of that trait (models, PCA, and correlations) if their sampling replication was also reduced. 462 This was done to reduce the impact of outliers in statistical modelling and estimation of 463 correlations/networks. This resulted in the removal of one accession of Phaseolus acutifolius (PI 535200) 464 and Vicia dumetorum (PI 494749) for germination T50 (also removed due to incomplete germination), and 465 one accession of Lathyrus annuus (PI 358829) for height DAP-21 and height and leaf number AGR and 466 RGR (which were dependent on height DAP-21 as a covariate). 467 468 RESULTS bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 19 469 470 We investigated variation and correlation in seed, germination, and vegetative growth traits of plants 471 derived from germinated seeds in 80 accessions of annual and perennial Lathyrus, Phaseolus, and Vicia 472 species. Principal component analyses highlighted distinct phenotypic space occupied by each genus and 473 life span group, as well as important species-level variation within each genus. Linear models identified 474 genus as the primary predictor of seed size variation and life span as the primary predictor of static 475 vegetative growth variation (the singular height and leaf measurements at DAP-21 and DAP-35). 476 Correlation networks further demonstrate genus- and life span-specific trait relationships, including 477 correlations between and within seed and vegetative traits. 478 479 Genus and life span predict trait variation—Principal component analyses identified key 480 phenotypic differences in the dataset among genera and between life spans, and showed correlated groups 481 of traits (Fig. 1). The first two principal components explained 42.2 and 24.7% of the variance 482 respectively. Seed size traits clustered tightly (Fig. 1A) and loaded positively on both PC1 and PC2. Seed 483 shape (circularity and roundness) loaded negatively onto PC1 (Fig. 1A; Appendix S2), suggesting that 484 seed size and seed shape are negatively correlated. Vegetative traits were less tightly grouped: height at 485 DAP-21 and DAP-35 and leaf number at DAP-21 loaded positively onto PC1, while leaf number at DAP- 486 35 and both growth rate metrics (AGR and RGR) loaded negatively onto PC2 (Fig.

1A; Appendix S2). 487 Germination T50 loaded positively onto PC2, opposite of vegetative traits, whereas germination 488 proportion loaded positively onto PC1 in conjunction with vegetative traits (Fig. 1A; Appendix S2). 489 Phaseolus occupied the greatest area in the first two PCs, followed by Vicia then Lathyrus (Fig. 1B). 490 Annuals occupied a greater PC area than perennials, with much of this area occupied by Phaseolus 491 annuals (Fig. 1C). Among genera, Phaseolus accessions tended to occupy space showing greater seed size 492 and greater static plant height and leaf number, as well as higher growth rate; Lathyrus and Vicia tended 493 to have smaller, more circular seeds with a delayed T50 and smaller vegetative growth. Likewise, annual bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 20 494 accessions showed greater seed size and vegetative traits (static and growth rate), while perennials 495 showed the opposite pattern (Fig. 1B,C). 496 497 Principal component analyses of seed and vegetative data highlighted similar seed trait correlations in 498 each genus (Appendix S3, S4, S5); however, differences among genera existed in vegetative trait 499 correlations and in how annual and perennial species separated in phenotypic space. In each of the three 500 genera, seed size traits were tightly correlated and loaded positively onto PC1, while seed shape 501 (circularity and roundness) loaded positively onto PC2, approximately orthogonal to seed size (Appendix 502 S3, S4, S5, S6). These data consistently supported a negative relationship between seed size and shape. In 503 general, annual and perennial species did not separate consistently in PC space across genera with respect 504 to seed size; however, Phaseolus annuals had more distinctly less circular / round seeds than perennials 505 (Appendix S3, S4, S5). In contrast, vegetative traits showed less consistent and less closely grouped 506 loading patterns in PCAs across genera (Appendix S3, S4, S5, S7). Similar to seed PCAs, Lathyrus and 507 Vicia had more overlap in annual-perennial vegetative variation than Phaseolus. Annual Phaseolus 508 species had consistently higher static vegetative growth than perennial species (Appendix S3, S4, S5). 509 Nevertheless, two annual Vicia species, V. villosa and V. benghalensis, showed distinctly greater static 510 vegetative growth than other Vicia species (Appendix S5). These data revealed important species-level 511 variation underlying the broad lifespan patterns within each genus. 512 513 In our linear mixed models, genus was the largest predictor of seed trait variation (Table 2). Genus was 514 significant for all seed traits except seed mass, whereas life span was not significant for any seed trait 515 (Table 2).

Despite life span nonsignificance, post hoc tests revealed that for Phaseolus all seed size traits 516 except seed mass were significantly larger in annuals (Appendix S8). For seed circularity and roundness 517 there was a significant genus × life span interaction where both traits were significantly greater in 518 perennials than annuals for Phaseolus, but roundness was significantly lower in perennials than annuals 519 for Lathyrus (Table 2 and Appendix S8). Linear models revealed a significant genus and life span effect bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 21 520 for germination T50, while only genus was significant for germination proportion (Table 2). Perennials in 521 all genera had delayed T50 compared to annuals, but this was only significant for Lathyrus (Appendix S8). 522 523 Life span was the most consistent predictor of vegetative trait variation (Table 2). Life span was a 524 significant predictor for all static height and leaf number measurements at DAP-21 and DAP-35, with 525 more variation attributable to life span than genus (nonsignificant) for static height traits (Table 2). For 526 static leaf traits, life span explained a similar amount of variation to genus, and both effects were 527 significant (Table 2). For all genera, perennials had lower mean static vegetative height and leaf number 528 than congeneric annuals; however, the difference was only significant for Phaseolus (all static vegetative 529 traits) and for Vicia (static leaf number only; Appendix S8). Vegetative growth rate patterns (AGR and 530 RGR) were more variable, and neither genus nor life span was significant for these traits in the linear 531 models (Table 2). However, Lathyrus showed a significant difference in leaf number RGR (perennials 532 greater than annuals) and Phaseolus in height AGR and RGR (annuals greater than perennials; Appendix 533 S8). Although nonsignificant, in Lathyrus both height and leaf AGR and height RGR were also higher in 534 perennials; in Vicia height and leaf number AGR were higher in annuals than perennials, but height and 535 leaf number RGR were higher in perennials (nonsignificant; Appendix S8). In addition, species was a 536 significant predictor for most traits, with the exception of germination T50, germination proportion, height 537 at DAP-21, and leaf number RGR (Table 2). Several random effects controlled for in the models were 538 significant (Appendix S9), while dropping data with potentially confounding factors from the models 539 resulted in minimal change in significance (Appendix S10). 540 541 Genus and life span patterns in trait correlations—Correlation networks revealed dynamic seed 542 and vegetative trait correlations which reflected unique genus and life span-specific patterns, and which 543 were generally consistent with PCA variable loadings (Fig.

2). Considering the full dataset network and 544 commonalities among each subgroup, there were always strong, significant positive correlations among 545 all seed size traits, which were always significantly negatively correlated with seed circularity and/or bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 22 546 roundness (Fig. 2; Appendix S11). Germination T50 was usually negatively correlated with at least some 547 vegetative traits for all subgroups, while germination proportion was generally positively correlated with 548 vegetative traits; both germination traits generally positively correlated with seed traits when significant 549 (Fig. 2). All significant correlations among vegetative growth traits were positive, with the exception of 550 height and leaf number RGR, but significance and magnitude of correlation among vegetative traits 551 varied for each subgroup (Fig. 2). Seed size traits were more commonly significantly correlated to static 552 vegetative growth traits (positive) than AGR or RGR (Fig. 2). Where significant correlations did occur, 553 height and leaf number RGR usually negatively correlated with seed size (Fig. 2). Overall, static height 554 measurements had a consistently high degree (number of significant connections to other traits) in each 555 subgroup (Fig. 2). Despite commonalities, considerable variation in magnitude, direction, and 556 significance in trait integration also existed among the subgroups. 557 558 The three genera differed in the extent of integration between seed and vegetative traits, as well as the 559 strength of correlations among vegetative traits (Fig. 2D-E). Lathyrus showed the least amount of 560 significant seed to vegetative trait correlation, with only static height positively correlated to seed size 561 (Fig. 2D; Appendix S12). Lathyrus also showed the least connectivity among vegetative traits, with 562 height and leaf trait groups having few significant correlations (Fig. 2D, Appendix S12). Phaseolus 563 showed significant positive correlations between seed size and height at DAP-21 and significant negative 564 correlations between seed size and height/leaf RGR, with the opposite pattern for seed shape traits with 565 vegetative traits (Fig. 2E, Appendix S13). Phaseolus also showed predominantly positive correlations 566 between vegetative traits, and it had more significant seed and vegetative trait connections to RGR than 567 the other genera (Fig. 2E, Appendix S13). Vicia showed the greatest amount of connectivity between seed 568 and vegetative traits, with all static DAP-21 and DAP-35 height and leaf measurements significantly 569 positively correlated with all seed size traits, although there was minimal correlation between AGR/RGR 570 and seed traits (Fig.

2F, Appendix S14). Vicia also showed the most robust cluster of significant 571 positively correlated vegetative traits, with the exception of RGR (Fig. 2F, Appendix S14). Lastly, seed bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 23 572 size was significantly negatively correlated to seed circularity for Lathyrus and Vicia, while for Phaseolus 573 seed size traits were significantly negatively correlated with seed roundness but not circularity (Fig. 2D-F, 574 Appendix S12, S13, S14). For all three genera, height at DAP-21 had a consistently high degree (Fig. 2D- 575 F). 576 577 Annual and perennial species differed in their seed trait to vegetative growth correlations and their overall 578 connectivity among traits (Fig. 2B-C). While annuals showed significant positive relationships between 579 all seed size and static height measurements, as well as significant negative correlations between seed 580 shape and all height traits, perennials lacked any significant correlation between seed and vegetative traits 581 (Fig. 2B-C; Appendix S15, S16). Annuals also had more significant correlations between seed size and 582 seed shape traits (negative) than perennials, and annuals showed significant positive correlations between 583 seed shape and germination T50, while seed and germination traits were not connected for perennials (Fig. 584 2B-C; Appendix S15, S16). However, both annual and perennial subgroups showed significant positive 585 correlations among most vegetative traits, with the only gaps in correlation occurring for height and leaf 586 AGR/RGR (Fig. 2B-C; Appendix S15, S16). The degree of seed and static vegetative traits was low in 587 perennials compared to annuals, but perennials generally had greater positive correlations between height 588 and leaf traits, as well as between static height and height AGR/RGR (Fig. 2B-C; Appendix S15, S16). 589 Similar to the genera subgroups, static vegetative traits showed some of the highest degrees within both 590 the annual and perennial networks (Fig. 2B-C). 591 592 DISCUSSION 593 594 Life history strategy and the associated life span classification (annual, perennial) is associated with 595 various aspects of species reproductive and vegetative biology, but the role of life span in predicting seed 596 to vegetative trait correlation is not well characterized. Here we examined trait variation and correlation 597 of seed size and shape, and the germination and vegetative growth characters of seedlings derived from bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021.

The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 24 598 those seeds, in annual and perennial congeners from three herbaceous legume genera. We found that 599 while genus was a stronger predictor of seed size and shape, life span most consistently predicted static 600 vegetative growth. Patterns of trait correlation between and among seed size and vegetative growth traits 601 differed by life span and genus; specifically, annual species had more significant seed to vegetative trait 602 correlations than perennial species. 603 604 Evolutionary basis of genus and life span differences— In this study genus was consistently the 605 greatest predictor of seed size and shape variation, which illustrates the importance of phylogenetic 606 context in understanding some aspects of life history variation (Silvertown and Dodd, 1996). Annual- 607 perennial overlap in species-level variation presented here in PCAs further demonstrates that seed size 608 and shape variation does not consistently separate by life span group (Appendix S3, S4, S5). Mazer 609 (1989) similarly found a greater amount of variation in seed mass explained by phylogenetic family 610 (30%) than life history type (22%) in a large study of ten families of Indiana Dune angiosperms. 611 However, they still found life history to be significant, which was attributed to the distinctly large seeds 612 of tree species (Mazer, 1989). Our study builds on this previous work in that it focuses on herbaceous 613 species of three legume genera and quantifies multiple dimensions of seed size and shape on an individual 614 seed level (instead of seed mass class), without specificity to a particular habitat (Mazer, 1989). 615 Importantly, Mazer (1989) also noted that life history was usually not a significant predictor of seed mass 616 looking at the within-family level; also, there was not a significant difference in seed mass among annuals 617 and herbaceous perennials in their study. Our lack of clear life span signal in seed size (but see post hoc 618 tests for Phaseolus) is also consistent with a meta-analysis of c. 3000 congeneric comparisons of annual 619 and perennial species (Vico et al., 2016). 620 621 While genus predicted seed traits, our data demonstrated that life span consistently predicted static 622 vegetative traits. In this study, annual species displayed greater mean height and leaf number measured at 623 21 and 35 days from planting than congeneric perennial species in the first year of growth; this was bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 25 624 significant for both traits in Phaseolus and for leaf number in Vicia. Higher growth in annuals is 625 consistent with the predicted pattern for annuals and perennials according to the acquisitive-conservative 626 resource economics spectrum (e.g., Roumet et al., 2006, González-Paleo and Ravetta, 2015).

Vegetative 627 growth is commonly measured in terms of relative growth rate (RGR), for which life span is usually a 628 significant predictor (e.g., Garnier, 1992; Atkinson et al., 2016); however, in this study neither life span 629 nor genus was a significant predictor of either absolute growth rate (AGR) or RGR in terms of plant 630 height and leaf number. This could reflect the timing of growth measurements (made at the seedling 631 stage), since species have different trajectories of RGR over their life span, with RGR typically 632 decreasing with time as the plant becomes larger (Turnbull et al., 2008). Also, while RGR is typically 633 measured using successive destructive biomass harvests on different plants (Perez-Harguindeguy et al., 634 2013), we nondestructively measured height and leaf number growth on the same individual plants, which 635 may result in different patterns. Ideally, vegetative measurements will span a much larger developmental 636 window, with multiple time points allowing for more precise modeling of growth through the life cycle of 637 each plant. 638 639 Annual species exhibited significant correlations between seed and vegetative traits (plant height and leaf 640 number), but the same correlations were not significant in perennial species, suggesting that some 641 phenotypic relationships are at least partially dependent on life span (Fig. 2). Other studies of herbaceous 642 perennial species have focused on relationships between vegetative traits and traits of seeds produced by 643 that individual or from individuals in separate studies, a slightly different approach than our study, where 644 we examined correlations between seed traits and the vegetative traits expressed by the individuals 645 germinated from those seeds. Kleyer et al. (2019), studying predominantly herbaceous perennials in a 646 NW European flora, also found no significant correlation between plant height and mass of seeds 647 produced by that plant, although seed mass was somewhat correlated to plant biomass. Similarly, a study 648 of 526 species in Sweden found no significant relationship between plant height and seed mass in 649 herbaceous perennials, while that relationship did exist for herbaceous annuals, woody perennials, and the bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 26 650 full dataset (Bolmgren and Cowen, 2008). In this case seed and plant height traits were combined from 651 separate studies in a meta-analysis (Bolmgren and Cowen, 2008). Our study extends these results and 652 demonstrates similar correlation patterns between source seeds and the plants derived from them. 653 Significant seed to vegetative trait correlations in annual species could be related to annuals’ complete 654 reliance on seeds to perpetuate their population; thus many aspects of annual plant form should be 655 optimized for producing a greater reproductive output.

Plant size and seed size are often significantly 656 positively correlated in broad interspecific meta-analyses (Leishman et al., 1995; Moles et al., 2004; Díaz 657 et al., 2016); greater vegetative size may allow greater total reproductive output through increased seed 658 size and/or number. 659 660 Unique properties of this dataset must also be taken into consideration for their influence on correlation 661 trends. For example, life span seed vs. vegetative trends were heavily influenced by Phaseolus accessions, 662 which tended to have some of the highest trait values but were nonetheless supported by multiple 663 replicated accessions. Further research with greater replication will be necessary to determine if this 664 intergeneric life span trend is consistent within each genus. Phenotypic integration is also sensitive to 665 growth environment and developmental stage (Murren, 2002); thus it remains to be seen if these 666 phenotypic patterns are robust in each species’ native environment and across a longer period of growth, 667 particularly in the case of perennial species. 668 669 This study examined a broad diversity of annual and perennial interspecific variation, but variation at the 670 intraspecific level was not as well represented. The fact that species was highly significant for almost all 671 traits reflects the plethora of phenotypic diversity that exists within and among annual and perennial 672 species. For simplicity, we adopted the broad titles of annual and perennial in this study, but life span is 673 not a simple binary or categorical trait, and life span alone cannot be expected to encompass the 674 biological complexity of all underlying traits. There is substantial intra- and interspecific variation in 675 traits associated with life span: reproductive patterns (e.g., semelparous and iteroparous), growth bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 27 676 determinacy vs. indeterminacy, clonality, length of the juvenile phase, and mating system (autogamous, 677 allogamous, and mixed), among many others (Friedman, 2020). Life span is also inherently context 678 dependent and potentially variable at the intraspecific level: both annual and perennial populations can 679 occur within a single species across its range, often due to variation in environmental disturbance, e.g., 680 Mimulus guttatus (Friedman et al., 2015), Oryza perennis (Morishima et al., 1984), and Zostera marina 681 (Reynolds et al., 2017). We are also limited in our ability to know and control for the maternal growth 682 environment and past environmental drivers for these accessions, which may include either direct or 683 indirect artificial selection for those collected in feral populations and as agricultural weeds.

In order to 684 gain a more precise understanding of life span variation in these species, a thorough assessment of 685 variation within and across many diverse populations is needed. 686 687 Relevance to perennial breeding goals—With increasing interest in de novo domestication of 688 wild species and breeding of hybrid herbaceous perennial crops, broadening our understanding of trait 689 networks in herbaceous perennials is essential. Much of our understanding of phenotypic change and 690 correlation in domestication targeting reproductive structures comes from annual and woody perennial 691 cultivars (Miller and Gross, 2011; Meyer et al., 2012), and a gap exists in the case of herbaceous 692 perennials, which have not been broadly domesticated for human grain consumption (Van Tassel et al., 693 2010). There is evidence that annual crops exhibit a decrease in phenotypic integration during crop 694 improvement, potentially releasing them from ecological trait trade-offs present in more resource-limited 695 wild conditions (Milla et al., 2014). Further characterization of functional traits such as those in this study 696 can advance understanding of evolution under domestication and can also help in identifying promising 697 wild candidates for domestication. Evidence to date highlights potential seed yield - vegetative trade-offs 698 in emergent perennial crops (e.g., González-Paleo et al., 2016; Pastor-Pastor et al., 2018); however, the 699 challenge of co-selecting for negatively correlated traits has been accomplished in modern plant breeding, 700 and breeders have successfully selected for both high seed yield and sustained perenniality in rice 701 (DeHaan et al., 2005; Huang et al., 2018). bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 28 702 703 Seed size is a key trait in perennial grain breeding programs (Kantar et al., 2016), and the effect of seed 704 size selection on other plant phenotypes is relevant to understanding constraints to evolution under 705 artificial selection. Our results suggest that different phenotypic ties exist between seed and early 706 vegetative growth traits in annuals and perennials measured under controlled conditions, with seed traits 707 ostensibly being less integrated with growth traits in perennials under these circumstances. In these 708 genera, this could mean that selection on seed size may not greatly impact early vegetative growth in 709 perennials. This could be particularly important for grain crops which are dual-purposed as forage, where 710 vigorous vegetative growth is also favorable, e.g., Thinopyrum intermedium (Pugliese, 2017) and 711 Silphium integrifolium (Vilela et al., 2020). It remains to be seen if the seed-vegetative disconnect extends 712 to later lifetime vegetative characteristics such as biomass allocation and plant size at reproductive 713 maturity, and if these relationships hold when mature plants are measured under field conditions.

714 715 The next steps in investigating phenotypic integration in herbaceous perennials should extend to traits 716 important to perennial agriculture and total perennial lifetime fitness, measured in the field over multiple 717 years. Specifically, total seed yield and root allocation (among other belowground and perennating 718 structures) are dual targets of current perennial grain breeding, and understanding potential trade-offs 719 between these traits is fundamental to the advancement of perennial breeding (Van Tassel et al., 2017). 720 Our results highlight phenotypic patterns only in the first few weeks of perennial growth, whereas 721 multiyear patterns in yield and vegetative allocation in perennial species will more comprehensively 722 reflect their total life history strategy. While logistically difficult, understanding yearly fluctuations in 723 reproductive output, above-ground biomass, and root growth over the total life span of perennials is 724 paramount, as will be connecting this variation to environmental variables. By gathering this lifetime 725 phenotypic information, early seed and growth traits may be used to predict hard-to-measure later life 726 span aspects of perennial crops and thus expedite the breeding process. 727 bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 29 728 CONCLUSIONS 729 730 Despite the prevalence of annual-perennial transitions in angiosperms, we know relatively little about 731 how life history classification is associated with phenotypic correlations across ecologically and 732 agriculturally important traits. Here we show that in three genera of legumes seed variation was primarily 733 explained by genus and static vegetative variation by life span. Further, annual species showed stronger 734 seed to vegetative trait correlation in early growth than perennial congeners. These findings call for 735 further investigation into how these trait correlations differ throughout the life span of perennial plants, as 736 well as into correlations among other important life history traits, particularly allocation to reproductive 737 and perennating organs. This study highlights that both life span and phylogenetic context are important 738 in predicting phenotypic variation, and there remains numerous underexplored systems in which to 739 expand life history studies. 740 741 ACKNOWLEDGEMENTS 742 743 This research was funded by the Perennial Agriculture Project (Malone Family Land Preservation 744 Foundation and The Land Institute). S.A.H. was supported by a graduate assistantship from Saint Louis 745 University. M.J.R. is supported by the Donald Danforth Plant Science Center and the Perennial 746 Agriculture Project.

The National Science Foundation Research Experiences for Undergraduates (REU) 747 program at the Missouri Botanical Garden supported student involvement in this research, including 748 M.C.S. A special thanks also goes to REU student Summer Sherrod for her assistance in this work and 749 enthusiastic engagement in this project as a part of her internship. Seeds were provided by the United 750 States Department of Agriculture Western Regional PI Station (Pullman, WA); from the USDA facility in 751 Pullman, we specifically thank curators Clarice Coyne and Barbara Hellier for their assistance in 752 clarifying improvement status and other provenance details for each accession. We are grateful to Daniel 753 Debouck (former curator of Phaseolus at the International Center for Tropical Agriculture) for invaluable bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 30 754 discussions regarding appropriate species life span assignment. We are grateful to the greenhouse and 755 horticulture staff at the Missouri Botanical Garden for providing space and assistance in plant care, 756 especially Joshua Higgins, Justin Lee, and Derek Lyle. We would also like to thank the Saint Louis 757 University Department of Biology and Donald Danforth Plant Science Center, and specifically Kasey 758 Fowler-Finn, Kristina Haines, and Kevin Reilly, for assistance with plant growth facilities. We 759 acknowledge the following researchers who helped provide important information regarding accession 760 origin: Alexandr Afonin, Clarice Coyne, Ken Friesen, Stephanie Greene, Richard Hannan, Barbara 761 Hellier, Douglas Johnson, Francis Kilkenny, Nigel Maxted, Luis Guillermo Santos Meléndez, Sergey 762 Shuvalov, and Filip Vandelook. We are thankful to the Miller Lab and Elizabeth Kellogg for helpful 763 comments on previous versions of this manuscript, and specifically for Zachary Harris and Julia Pratt for 764 assistance in figure generation. Lastly, we are especially grateful to all who assisted in plant measurement 765 and caretaking: Niyati Bhakta, Emily Boeckenstedt, Leah Brand, Claudia Ciotir, Emma Frawley, Jordan 766 Hathaway, Danielle Hopkins, Tanvi Kadiyala, Aidan Leckie-Harre, Alex Linan, Kazi Maharun Nessa, 767 Brittany Pace, Joshua Reinl, Heather Schier, and William Shoenberger. 768 769 AUTHOR CONTRIBUTIONS 770 771 S.A.H. and A.J.M. designed the study, and S.A.H. implemented the research and wrote the manuscript. 772 M.A.A., Q.G.L., A.J.M., M.J.R., and M.C.S. assisted in critical review and writing of the manuscript. 773 M.J.R. assisted in germination calculations and statistical methods and interpretation. M.A.A. and Q.G.L. 774 assisted in crafting germination experiments and provided the necessary resources.

M.C.S. assisted in data 775 acquisition. 776 777 DATA AVAILABILITY 778 bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 31 779 All phenotypic data (individual seed, germination, vegetative growth, and accession-level data) are 780 available in the Figshare database. Data can be accessed at: doi: [data is currently in preparation for 781 submission to this database]. 782 783 SUPPORTING INFORMATION 784 785 Additional supporting information may be found online in the Supporting Information section at the end 786 of the article. 787 788 APPENDIX S1. Accession descriptive metadata and sample size for each trait. 789 790 APPENDIX S2. Variable loadings for principal components 1-5 of the full accession-level dataset 791 principal component analysis (Fig. 1). 792 793 APPENDIX S3. Principal component analyses of the full seed and vegetative trait datasets for Lathyrus. 794 795 APPENDIX S4. Principal component analyses of the full seed and vegetative trait datasets for Phaseolus. 796 797 APPENDIX S5. Principal component analyses of the full seed and vegetative trait datasets for Vicia. 798 799 APPENDIX S6. Variable loadings for principal components 1-2 of the full seed dataset principal 800 component analysis, parsed by genus. 801 802 APPENDIX S7. Variable loadings for the full vegetative dataset principal component analysis, parsed by 803 genus (principal components 1-3 for Lathyrus, principal components 1-2 for Phaseolus and Vicia). 804 bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 32 805 APPENDIX S8. Table of adjusted means of each trait for annuals and perennials of each genus and post 806 hoc significance tests. 807 808 APPENDIX S9. Table of all significant random effects in addition to accession for each trait model. 809 810 APPENDIX S10. Analysis of variance table of each trait model with subsets of the data removed to test 811 for influence on significance. 812 813 APPENDIX S11. Correlation matrix for the full dataset, showing Pearson correlation coefficients 814 between every combination of traits, using accession-level data. 815 816 APPENDIX S12. Correlation matrix for the Lathyrus data subgroup, showing Pearson correlation 817 coefficients between every combination of traits, using accession-level data. 818 819 APPENDIX S13. Correlation matrix for the Phaseolus data subgroup, showing Pearson correlation 820 coefficients between every combination of traits, using accession-level data.

821 822 APPENDIX S14. Correlation matrix for the Vicia data subgroup, showing Pearson correlation 823 coefficients between every combination of traits, using accession-level data. 824 825 APPENDIX S15. Correlation matrix for the annual data subgroup, showing Pearson correlation 826 coefficients between every combination of traits, using accession-level data. 827 828 APPENDIX S16. Correlation matrix for the perennial data subgroup, showing Pearson correlation 829 coefficients between every combination of traits, using accession-level data. 830 bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 33 831 LITERATURE CITED 832 833 Asmussen, C.B., and A. Liston. 1998. Chloroplast DNA characters, phylogeny, and classification of 834 Lathyrus (Fabaceae). American Journal of Botany 85: 387–401. 835 836 Atkinson, R.R.L., E.J. Mockford, C. Bennett, P.A. Christin, E.L. Spriggs, R.P. Freckleton, K. Thompson, 837 et al. 2016. C4 photosynthesis boosts growth by altering physiology, allocation and size. Nature Plants 2: 838 1–5. 839 840 Baskin, C.C., and J.M. Baskin. 2014. Seeds: Ecology, biogeography, and evolution of dormancy and 841 germination, 2nd ed. Academic Press, San Diego, California, USA. 842 843 Bates, D., M. Mächler, B. Bolker, S. Walker. 2015. Fitting linear mixed-effects models using lme4. 844 Journal of Statistical Software 67: 1–48. 845 846 Bitocchi, E., D. Rau, E. Bellucci, M. Rodriguez, M.L. Murgia, T. Gioia, D. Santo, et al. 2017. Beans 847 (Phaseolus ssp.) as a model for understanding crop evolution. Frontiers in Plant Science 8: 1–21. 848 849 Bolmgren, K., and P.D. Cowan. 2008. Time - size tradeoffs: A phylogenetic comparative study of 850 flowering time, plant height and seed mass in a north-temperate flora. Oikos 117: 424–429. 851 852 Buhrow, R. 1983. The wild beans of southwestern North America. Desert Plants 5: 67–88. 853 854 Castiglione, M.R., M. Frediani, and M.T. Gelati. 2011. Cytology of Vicia species. X. Karyotype evolution 855 and phylogenetic implication in Vicia species of the sections Atossa, Microcarinae, Wiggersia and Vicia. 856 Protoplasma 248: 707–716. bioRxiv preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made doi: https://doi.org/10.1101/2021.02.17.431656 ; this version posted February 17, 2021. The copyright holder for this preprint available under a CC-BY-NC-ND 4.0 International license . 34 857 858 Chapin, III, F.S. 1980. The mineral nutrition of wild plants. Annual Review of Ecology and Systematics 859 11: 233–260. 860 861 Chapin, III, F.S., E.D. Schulze, and H.A. Mooney. 1990. The ecology and economics of storage in plants.