Leu8 and Pro8 oxytocin agonism differs across human, macaque, and marmoset vasopressin 1a receptors

Oxytocin (OXT) is an important neuromodulator of social behaviors via activation of both oxytocin receptors (OXTR) and vasopressin (AVP) 1a receptors (AVPR1a). Marmosets are neotropical primates with a modified OXT ligand (Pro8-OXT), and this ligand shows significant coevolution with traits including social monogamy and litter size. Pro8-OXT produces more potent and efficacious responses at primate OXTR and stronger behavioral effects than the consensus mammalian OXT ligand (Leu8-OXT). Here, we tested whether OXT/AVP ligands show differential levels of crosstalk at primate AVPR1a. We measured binding affinities and Ca2+ signaling responses of AVP, Pro8-OXT and Leu8-OXT at human, macaque, and marmoset AVPR1a. We found that AVP binds with higher affinity than OXT across AVPR1a, and marmoset AVPR1a show a 10-fold lower OXT binding affinity compared to human and macaque AVPR1a. Both Leu8-OXT and Pro8-OXT produce a less efficacious response than AVP at human AVPR1a and higher efficacious response than AVP at marmoset AVPR1a. These data suggest that OXT might partially antagonize endogenous human AVPR1a signaling and enhance marmoset AVPR1a signaling. These findings aid in further understanding inconsistencies observed following systemic intranasal administration of OXT and provide important insights into taxon-specific differences in nonapeptide ligand/receptor coevolution and behavior.

were performed to explore species-level differences in binding affinities for the antagonist radioligand 125 I-OVTA to AVPR1a (SI Fig. 1). B max values (i.e., maximal binding) for human, macaque, and marmoset ranged from ~6. 21-12.71 fmol/well with marmosets AVPR1a CHO clone showing the highest B max values (Table 1). Overall, human, marmoset, and macaque AVPR1a had relatively similar K d values, ranging from 331-1165 pM ( Table 1). The affinities for AVPR1a were also similar to those for 125 I-OVTA at human, macaque, and marmoset OXTR (161-481 pM) published previously . These findings confirm that there are only relatively small differences in binding affinities for 125 I-OVTA among AVPR1a from these three primate species.
OXT partial antagonism of AVP Ca 2+ mobilization at human and marmoset AVPR1a. Partial agonists should also exhibit partial antagonism. To confirm this, Leu 8 -OXT and Pro 8 -OXT were coadministered with 10 nM AVP (a concentration of AVP that alone produces approximately 80% of the maximal Ca 2+ response) for both human and marmoset AVPR1a. For human AVPR1a, both Leu 8 -OXT and Pro 8 -OXT caused a concentration-dependent decrease in AVP-stimulated Ca 2+ mobilization. Both Leu 8 -OXT and Pro 8 -OXT at concentrations < 1 μM with and without coadministration of 10 nM AVP produced only 50% of the maximal Ca 2+ mobilization response to AVP, confirming that both OXT ligands act as both partial agonists and partial antagonists at human AVPR1a (Fig. 3). For marmoset AVPR1a, OXT did not function as a partial antagonist of AVP, (i.e., OXT did not induce inhibition of the Ca 2+ signaling response when co-administered with AVP) (Fig. 3). Interestingly, in both human and marmoset AVPR1a, Pro 8 -OXT coadministered with AVP always produced a greater maximal response than Leu 8 -OXT when coadministered with AVP. This occurred even at concentrations of OXT that did not produce Ca 2+ responses on their own. OXT partial antagonism experiments were not performed in macaque AVPR1a because OXT did not partially agonize or superagonize Ca 2+ responses at macaque AVPR1a.
Coupling efficiency at primate AVPR1a. We measured a simplified form of coupling efficiency as a ratio of the concentration of ligand needed to mobilize Ca 2+ responses in primate AVPR1a (potency/EC50) relative to the ligand binding affinity (K i ) at AVPR1a (Table 3). This metric provides insight into whether OXT/AVP ligands produce equal signaling responses across different receptors in the presence or absence of unbound/spare receptors, i.e., a extra receptors than what is required to produce a maximal response 32    www.nature.com/scientificreports www.nature.com/scientificreports/ both Leu 8 -OXT and Pro 8 -OXT, respectively. Conversely, human AVPR1a showed negative coupling efficiency values, suggesting that human AVPR1a require relatively more AVPR1a to produce Ca 2+ responses from OXT ligands compared to marmoset AVPR1a. Macaque AVPR1a coupling efficiencies for AVP, Leu 8 -OXT, and Pro 8 -OXT showed minimal difference (less than one log unit) compared to those for human and marmoset AVPR1a.

Discussion
This study is the first to evaluate potential differences in Leu 8 -OXT and Pro 8 -OXT binding affinity, Ca 2+ signaling potency and efficacy, OXT partial antagonism, and receptor coupling efficiency across a variety of primate AVPR1a. Previous studies have evaluated whether the documented coevolutionary changes in OXT ligands and OXTR in Platyrrhini primates would produce demonstrable and unique properties for OXT-OXTR signaling in these species [23][24][25] . These studies found that changes to the OXT molecule, namely the Leu 8 to Pro 8 AA substitution in OXT, produced only modest changes in binding and signaling across primate OXTR. OXT exhibits a significant degree of 'cross-talk' with AVP receptors (primarily AVPR1a), and OXT and AVP exhibit potential overlap in behavioral outcomes via OXT and AVP signaling in the brain. It is therefore plausible that OXT modifications would lead to a functional selective advantage through differences in OXT interactions with AVPR1a. The data from this study support three key conclusions: (1) AVP binds with significantly higher affinity than OXT at human, marmoset, and macaque AVPR1a, and marmoset receptor AVPR1a show a 10-fold lower OXT binding affinity compared to human and macaque AVPR1a. (2) There are no significant differences in binding affinity or Ca 2+ signaling potency between Leu 8 -OXT and Pro 8 -OXT at primate AVPR1a. (3) Both OXT isoforms exhibit differential levels of agonism/antagonism across primate AVPR1a, acting as partial agonists and partial antagonists at human AVPR1a and as superagonists at marmoset AVPR1a.
The idea that differences in the OXT ligand structure would result in functional differences in primate AVPR1a binding and/or signaling properties was only partially supported. While Leu 8 -OXT and Pro 8 -OXT showed no differences in binding affinity or Ca 2+ mobilization potencies at any of the primate AVPR1a, there was a significant difference in levels of OXT agonism both across primate AVPR1a and between OXT variants. Pro 8 -OXT produced a significantly higher maximal response compared to Leu 8 -OXT at marmoset AVPR1a. These pharmacological findings also partially align with previous work examining effects of Leu 8 -OXT and Pro 8 -OXT signaling at OXTR. Pro 8 -OXT exhibited modestly higher potencies than Leu 8 -OXT at primate OTRs 25 ; Pro 8 -OXT produced more efficacious Ca 2+ responses at marmoset OXTR but not at human OXTR 24 ; and Pro 8 -OXT produced lower recruitment of β-arrestin and less receptor desensitization and internalization at both human OXTR and AVPR1a, where only human receptors were tested 23 . Perhaps the most compelling finding from this study was that OXT exhibits differential agonism at human and marmoset AVPR1a. Similar to OXT signaling at OXTR, Pro 8 -OXT was more efficacious than Leu 8 -OXT at marmoset but not human AVPR1a. Previous pharmacological studies of marmoset OXTR did not explicitly test if different OXT ligands were partial agonists at marmoset and human OXTR 24 , and that study did not make direct comparisons of OXT agonism to AVP agonism at OXTR. However, AVP appears to be a full agonist relative to both Leu 8 -OXT and Pro 8 -OXT at primate OXTR based on data reported across human, marmoset, macaque, and titi monkey OXTR 25 .
The observation that Leu 8 -OXT and Pro 8 -OXT act as a partial agonists/antagonists at human AVPR1a is a novel finding. The first reported study evaluating the pharmacological profile of Pro 8 -OXT at human AVPR1a showed that Pro 8 -OXT is a full agonist at producing Ca 2+ responses compared to AVP, while Pro 8 -OXT was only a partial agonist for β-arrestin recruitment at both human AVPR1a or OXTR 23 . It is unclear what underlies the difference in Ca 2+ responses from the human AVPR1a tested in this study and the human AVPR1a tested previously 23 . Based on the clear partial agonism at human AVPR1a, we tested whether, as expected, OXT also functioned as a partial antagonist of the AVP Ca 2+ response at both human and marmoset AVPR1a. Adding either OXT isoform along with AVP reduced the AVP Ca 2+ response, but only for human AVPR1a. This confirms that OXT is a partial antagonist at human AVPR1a but not at marmoset AVPR1a. We further corroborated this finding by testing a marmoset AVPR1a clone with lower receptor expression (as indicated by saturation binding with 125 I-OVTA), and again both OXT ligands functioned as full agonists with slightly lower potency, with the Pro 8 -OXT response greater than for Leu 8 -OXT, eliminating concerns that species-differences in OXT agonism at AVPR1a were due to different expression levels of AVPR1a across species (SI Fig. 2) and/or differences in AVP signaling efficacy at 10 nM doses across primate AVPR1a. We also observed that all non-maximal OXT doses (<100 nM) of Pro 8 -OXT coadministered with 10 nM AVP at each primate AVPR1a produced a more efficacious agonism than comparable Leu 8 -OXT doses coadministered with 10 nM AVP, even in the presence of OXT doses (<1 nM) that would produce no measurable Ca 2+ signaling response on their own. The mechanism underlying this finding is currently unclear.
The conclusion that OXT functions as a partial agonist and a partial antagonist for AVP activation of Ca 2+ signaling responses in human AVPR1a has important implications. Though evidence for endogenously released OXT producing functionally important responses at AVP receptors is limited, some studies have shown that stimulating endogenous OXT release can induce social behavioral responses in rodents via AVPR1a 33,34 . However, a  www.nature.com/scientificreports www.nature.com/scientificreports/ majority of studies that examine the effects of OXT on behavior use exogenous intranasal OXT administration, causing systemic distribution and leading to supraphysiological increases in circulating OXT throughout the periphery. OXT is known to exert dose-dependent behavioral effects 35,36 ; thus further studies are warranted to evaluate whether high doses of OXT, in addition to activating OXTR, might also partially antagonize endogenous human AVPR1a signaling, which could aid in further understanding of the inconsistencies observed in behavioral responses following systemic administration of OXT 37 and the reported "inverted -U-shaped" relationship between OXT dose and behavior 38,39 .
Moreover, differential OXT agonism at AVPR1a could have important implications for understanding the therapeutic potential of nonapeptide treatments in alleviating symptoms associated with neurodevelopmental disorders such as autism spectrum disorders (ASD). For instance, animal models of nonapeptide signaling may not generalize in a simple way to human clinical trials. The impact of intranasal OXT on behaviorally relevant clinical outcomes has shown mixed support in the literature 40,41 , but recent evidence has shown that peripheral use of both a highly selective AVPR1a antagonist and intranasal AVP administration has markedly improved behavioral outcomes for individuals with ASD 30,31 . These findings are important given that similar OXT treatment strategies for ASD have shown mixed efficacy 42,43 , and high doses of OXT could even mitigate potential therapeutic benefits of AVPR1a activation in ASD and surely other behavioral contexts as well. Whether the partial agonism/antagonism at human AVPR1a explains these anomalies merits further study.
It is also noteworthy that differences in OXTR and AVPR1a functioning are important for our broader understanding of the coevolution of nonapeptide signaling system in Platyrrhini primates. While OT acted as a partial agonist at human AVPR1a, OXT (both Leu 8 -OXT and Pro 8 -OXT) instead acted as a superagonist at marmoset AVPR1a. This is especially important from an evolutionary context because the Callitrichid clade has evolved widespread Pro 8 -OXT expression of the OXT ligand, and the Pro 8 ligand produces stronger behavioral effects 44,45 , potency and efficacy effects at marmoset OXTR 24,25 , and efficacy effects at marmoset AVPR1a (this study). It is unclear whether the higher agonism and coupling efficiency of OXT at marmoset AVPR1a is an important or conserved mechanism underlying the potential coevolution between OXTR and AVPR1a variability with socially monogamous phenotypes in primates 15,46 . More pharmacological and behavioral work utilizing Pro 8 -OXT and other OXT ligands is needed across a broader sampling of primates. Such examples include Leu 8 -OXT expressing titi monkeys that are viewed as socially monogamous and biparental and Pro 8 -OXT expressing primates such as capuchins or squirrel monkeys that are highly social but non-monogamous/biparental. These data combined with the important data published on OXTR and AVPR1a central expression profiles in marmoset, titi, macaque, and humans 7 would serve as a powerful tool to begin utilizing and targeting diverse non-human primate models of nonapeptide regulation of social behavioral phenotypes.
Clearly, there are many contributing factors to the ways in which OXT and AVP regulate physiological and behavioral outcomes across primates, the relative roles of OXTR and AVPR1a activation, and how well these and other in vitro findings translate directly to neural transmission and ultimately behavioral modulation. These relationships are difficult to ascertain, especially in light of currently limited access to primate neural tissue and primate gene-editing techniques. An important first step is to evaluate whether the pharmacological and physiological findings and principles already established for nonapeptide biology in rodents are divergent or conserved across diverse nonhuman primate species. Our findings will serve as a roadmap to target specific pharmacological and physiological properties that may underlie species-or individual-level differences in behavioral and social phenotypes. Behavioral studies have been at the forefront of this effort and have elucidated many key findings about how OXT regulates social behavior in nonhuman primates 14,[47][48][49] , but many of these studies have yet to identify specific neural mechanisms underlying these behavioral effects. Overall, the findings from this study provide important molecular insights into species-level differences in nonapeptide ligand/receptor coevolution and 'cross-talk' between OXT and AVP.

Methods
Primate AVPR1a transfection and cell culture. Chinese hamster ovary (CHO; Female origin) cells were purchased from American Type Culture Collecton (Manassas, VA) and cultured at 37 • C with 5% CO 2 using Ham's F12 medium supplemented with 10% fetal bovine serum and 100 units/mL penicillin and 100 µg/ml streptomycin. Human, marmoset, and macaque AVPR1a plasmids were purchased from Genscript (Piscataway, NJ) in a pcDNA3.1+ vector based on confirmed genetic sequences. CHO cells were transfected using Turbofect according to the manufacturer's instructions and were kept under selective pressure using 400 µg/mL G418 antibiotic. Individual clonal lines were generated from batch-transfected cells by plating approximately 10 cells/mL (1 cell/100 µL) into 96-well plates. Clonal lines that originated from a single colony were screened using an intact cell 125 I-ornithine vasotocin analog ( 125 I-OVTA) binding assay and selected for similar receptor expression across species, defined as specific radioligand binding. CHO cells showed no endogenous OXTR and AVPR1a binding or signaling activity in response to OXT and/or AVP ligands (SI Fig. 3).
Intact cell saturation binding assays. CHO cells expressing primate AVPR1a were plated at 150,000 cells/mL (15,000 cells per well/100 µL) into 96-well plates and incubated at 37 °C for 48 hours to achieve 80-90% confluence. On the day of assay, growth medium was aspirated and cells were quickly washed once with 100 µL ice-cold high glucose HEPES-buffered Dulbecco's Modified Eagle's Medium containing 0.1% bovine serum albumin (HGH-BSA) and then placed on ice. 50 µL of ice-cold HGH-BSA containing 125 I-OVTA (PerkinElmer) in doubling concentrations (~15 to 2000 pM) was added in triplicate (technical replicates) and incubated for 3 hours on ice. 3 hours is the minimum incubation time on ice for 125 I-OVTA and 125 I-OVTA + AVP/OXT to reach equilibrium (SI Fig. 4) in CHO cells transfected with human and marmoset AVPR1a. Cells were washed four times with 100 µL ice-cold HGH-BSA, solubilized with 100 µL 0.2 N NaOH, and radioactivity quantified with a gamma counter. We also counted aliquots of the used binding medium (i.e., free 125 I-OVTA) to quantify free radioligand www.nature.com/scientificreports www.nature.com/scientificreports/ CPM directly, eliminating any concerns about differential depletion of ligand due to differential receptor expression levels across species or time. Non-specific binding was defined as 125 I-OVTA binding occurring in the presence of excess competitor (10 −4 M AVP). Binding affinity for 125 I-OVTA was determined after correcting for non-specific binding by plotting specific bound/free vs. bound using a single-site binding equation (Graphpad Software Inc., La Jolla, CA). Assays were done at least three times on three different days using fresh aliquots of 125 I-OVTA and competitor, and K d values were averaged across at least three biological replicates.
Intact cell competition binding assays. CHO cells expressing primate AVPR1a were plated at 150,000 cells/mL (15,000 cells per well/100 ul) into 96-well plates and incubated at 37 °C for 48 hours/grown to 80-90% confluence. On the day of assay, growth medium was aspirated and cells were quickly washed once with 100 µL ice-cold HGH-BSA and then placed on ice. Then 50 µL of ice-cold HGH-BSA containing ~50,000 CPM 125 I-OVTA were added in triplicate (technical replicates) to all wells in the presence or absence of 10 −11 to 10 −5 M Pro 8 -OXT (CYIQNCPPG-NH2; Anaspec), Leu 8 -OXT (CYIQNCPLG-NH2; Anaspec) or AVP (CYFQNCPRG-NH2; Anaspec), and incubated for three hours on ice. Cells were washed four times with 100 µL ice-cold HGH-BSA, solubilized with 100 µL 0.2 N NaOH, and radioactivity was quantified with a gamma counter. Half-maximal inhibitory concentrations (IC 50 ) were determined by plotting bound 125 I-OVTA vs. competitor concentration. IC 50 values were then corrected using the Cheng-Prusoff equation with each receptor's K D for 125 I-OVTA to produce K i values for the competing ligands. Assays were done at least three times on three different days using fresh aliquots of 125 I-OVTA and Leu 8 -OXT, Pro 8 -OXT, and AVP with at least three biological replicates per clone. ca 2+ mobilization assays. CHO cells expressing primate AVPR1a were plated at 150,000 cells/mL (15,000 cells per well/100 µL) into 96-well plates and incubated at 37 °C for 48 hours/grown to 80-90% confluence. On the day of assay, growth medium was aspirated and cells were incubated at 37 °C with 100 µL Fluo-4 Direct dye mixed in Fluo-4 Direct Ca 2+ Assay Buffer with 5 mM probenecid for ~45 minutes. Using a FlexStation 2 (Molecular Devices), baseline fluorescence was measured at 37 °C followed by stimulated fluorescence in the presence or absence of 10 −12 to 10 −5 M Pro 8 -OXT, Leu 8 -OXT, or AVP (3 × technical replicates). Peak fluorescence minus baseline fluorescence was plotted as a function of ligand concentration to determine EC 50 values. Assays were done at least three times on three different days using fresh aliquots of Leu 8 -OXT, Pro 8 -OXT, and AVP for three biological replicates per clone. We determined the degree of OXT Ca 2+ agonism/antagonism at AVPR1a by repeating the same procedures for Leu 8 -OXT alone, Pro 8 -OXT alone, and OXT ligands coadministered with 10 −8 M AVP (10 nM) (3 × technical and biological replicates) with concentrations of OXT from 10 −10 to 10 −5 M compared to coadministration of OXT concentrations from 10 −10 to 10 −4 M together with 10 −8 AVP.

Data analyses.
Binding affinities for 125 I-OVTA at each primate AVPR1a were calculated by subtracting nonspecific binding and then plotting bound 125 I-OVTA vs. free 125 I-OVTA. Because concentrations of 125 I-OVTA were not identical from experiment to experiment, technical replicates within each experiment (n = 3) were normalized and then corrected using the Cheng-Prusoff equation. Technical replicates were averaged and used as biological replicates (n = 3 per clone) to determine and compare K i values for each ligand within species. Differences in Ca 2+ mobilization potency (EC 50 ) and maximal response to OXT were determined by normalizing OXT-induced (Log M) Ca 2+ responses as a percentage of maximal (100%) AVP-induced Ca 2+ response. We averaged across technical replicates (n = 3) within each biological replicate and then averaged across the biological replicates (n = 3), normalized the data, and tested for significant differences of best-fit LogEC 50 using one-way ANOVA analyses. Post hoc analyses to assess ligand comparisons (Pro 8 -OXT vs. Leu 8 -OXT, Pro 8 -OXT vs. AVP, Leu 8 -OXT vs. AVP) were performed using Tukey's posthoc test with a Bonferroni-corrected cutoff to determine statistically significant differences in best-fit LogEC 50 . All best-fit data (K i , EC50, and Ca 2+ maximal responses) were analyzed using the nonlinear least squares curve-fitting capabilities of GraphPad Prism.

Data availability
Raw data and clonal cell lines (CHO clonal cell lines expressing either human, macaque, or marmoset AVPR1a) are available upon request.