Title Structure-Activity Relationship Studies of Nand C-TerminallyModified Secretin Analogs for the Human Secretin Receptor

The pleiotropic role of human secretin (hSCT) validates its potential use as a therapeutic agent. Nevertheless, the structure of secretin in complex with its receptor is necessary to develop a suitable therapeutic agent. Therefore, in an effort to design a three-dimensional virtual homology model and identify a peptide agonist and/or antagonist for the human secretin receptor (hSR), the significance of the primary sequence of secretin peptides in allosteric binding and activation was elucidated using virtual docking, FRET competitive binding and assessment of the cAMP response. Secretin analogs containing various Nor C-terminal modifications were prepared based on previous findings of the role of these domains in receptor binding and activation. These analogs exhibited very low or no binding affinity in a virtual model, and were found to neither exhibit in vitro binding nor agonistic or antagonistic properties. A parallel analysis of the analogs in the virtual model and in vitro studies revealed instability of these peptide analogs to bind and activate the receptor.


Introduction
GPCRs are one of the largest receptor families [1]; these receptors share features in their molecular structure and signaling mechanisms and are regulated by a wide range of ligands such as hormones, peptides, neurotransmitters, chemokines, etc. GPCRs serve as the most important link between extracellular conditions and intracellular responses and are involved in most aspects of physiological processes [2]. Among GPCRs, the class B secretin receptor family has been found to mediate a broad array of homeostatic functions hence, represents putative drug target. Class B ligands such as calcitonin, glucagon and parathyroid hormone are currently used as therapeutic agents [3]. Secretin is a 27-residue linear peptide that is widely expressed throughout the body [4]. Secretin receptors are present in central and peripheral tissues [5], with an established role in the gastro-intestinal tract in which it regulates intraduodenal pH [6], acid release, stomach motility [7] and insulin secretion [8]. Recently, the roles of secretin in water balance [9], motor function [10], lipid homeostasis [11,12], and appetite regulation [13] have also been demonstrated. This integrated role in physiology makes secretin a potential target for the treatment of metabolic disorders. The lack of structural insights into the interaction of hSR and secretin peptide remains the primary obstacle in the development of secretin agonist/antagonist and using the natural ligand as a therapeutic agent has not been feasible because of the short half-life of the peptide. Previously, most studies have investigated the physiological role of secretin, without much knowledge about the structure of its receptor [14]. The structures of some class A [15,16] (http://tools.gpcr.org/crystalstructure/table) and class C [17,18] GPCRs have been determined, and the entire structures of class B receptors [19,20] have only been recently elucidated. These structural studies have led to an understanding of the active and inactive receptor conformations. Instead of distinct pharmacophores which are generally seen in ligands, the class B GPCR receptor ligands have a distributed interaction interface with its receptor [21]. It is also established that the extracellular C-terminal (Ct) region of the ligand is necessary for initial binding to the receptor and is also responsible for the specificity and allosteric activity of the receptor. In contrast, the extracellular N-terminal (Nt) region of the ligand is involved in secondary binding with the extracellular loop region and is responsible for downstream signaling [22][23][24]. Based on this information new molecules were designed in this study with modifications in the Nt region, whereas secretin from various non-mammalian vertebrates that contain variations in the Ct region were used as Ct-modified analogs to assess their effect on the human secretin receptor (hSR). In the absence of an experimentally determined structure for hSR, a homology-modeled 3D receptor structure was developed to provide additional details on the receptor-ligand interaction. These secretin analogs were studied in parallel with virtual docking, in vitro binding and functional assays to investigate their interaction with the hSR.
The following secretin peptide analogs were evaluated in this study:

Homology modeling
The 3D model of hSR was prepared by modeling the Nt and the trans membrane (TM) region separately. The Nt model was prepared using multiple templates like Nt region of the PACAP receptor (PAC1) (PDB ID: 2JOD) [26] (Table 1), Nt region of the VPAC-2 receptor (PDB ID: 2X57) at 2.10 Å resolution [27], the Nt region of the GLP-1 receptor (PDB ID: 3C5T) at 2.10 Å [28] etc. All the initial models were then evaluated by SAVES server (http://services.mbi.ucla. edu/SAVES/) [29] and the model based on the template of Nt extracellular domain of human pituitary adenylate cyclase 1 receptor (PDB ID: 3N94) (Figure A in S1 File) was found to be the best structure [30] with sequence identity of 46.43 percent ( Figure A in S1 File). The TM backbone model was generated using the crystal structures of the glucagon receptor (PDB ID: 4L6R) at 3.30 Å resolution [20] and corticotropin-releasing factor receptor 1 (PDB ID: 4K5Y) at 2.98 Å resolution [19] ( Figure B in S1 File). The structure of the glucagon receptor (GCGR) with sequence similarity of 49.63 percent was used as the primary template for the TM region. The alignments were performed using the details available from Uniprot C [31] by fixing the residues and regions according to sequence and structural similarity. The developed model was validated for structural orientation and disulfide bonds. After validation, the Nt and TM were fused at the over-hanged region ( Figure C in S1 File) with the help of rigid docking by Pydock [32] to produce 102 models which were later screened for correct orientation of EC domain and TM region by structural validation, the tertiary structure of the receptor was compared with other class B templates [24] reducing the suitable models to 28 and on further screening 5 models were observed to have the best ERRAT score (crystallographic errors) [33]. The best model was selected based on the docking score between the receptor and ligands (hSCTF, human vasoactive peptide (hVIP), human gastric inhibitory polypeptide (hGIP) and human pituitary adenylate cyclase-activating polypeptide (hPACAP). The TM region was validated by the help of ProQM which is the only model quality assessment algorithm for membrane protein [34]. The 3-D models of hGIP, hVIP and hPACAP were retrieved from the RCSB protein data bank 2OBU [35], 2RRH [36] and 2D2P (yet to be published), respectively. The homology model of secretin and its analogs were also generated from solution NMR structures using hPACAP (PDB ID: 2D2P) and, crystal structure of glucagon (PDB ID: 3IOL) [37]. The 3D structures of analog number 5, 7, 9, 10 and 12 were not modeled due to unavailability of a template with non-standard amino acids.

Virtual docking
The docking stimulation was performed using Schrödinger biological suite [38] and patchdock/firedock was used to estimate the docking score [39]. The refined and validated receptor and ligand models were used for virtual docking. The receptor was analyzed for possible binding sites using Schrödinger biological suite for docking between hSCT and hSR models. Using hSCT, hVIP as positive control and GIP and others as negative control, ligand model was first docked with the five receptor models with the best ERRAT score in order to find the best docking site between hSCT and hSR EC domain [40]. The best model was used to calculate the docking score with all the secretin analogs by patchdock/firedock. The peptide and the protein were docked by patchdock and refinement was performed by firedock, the top 10 models after firedock refinement were analyzed. The binding energy of the best-docked model (among the top ten) at the correct binding site was used for affinity analysis. If the top ten structure with the ligand at correct binding site were not available, structure with the highest delta G score was used.

FRET binding assay
A non-radioactive binding assay was used to evaluate the binding affinity of the analogs [41]. SNAPtag-hSR-transfected CHO-K1 cells expressed the hSR at the plasma membrane with a SNAP-tag at the Nt region of the receptor. The cells were washed, and the SNAPtag-hSR was labeled with 100 nM Lumi4Tb at 37°C for 1 h. The h[Gly 28 , Alexa-Cys 29 ]SCT analog was used as an acceptor. The labeled cells were detached with Versene and counted. In 384-well black plates, 10,000 cells in 5 μL of labeling media, 5 μL of labeling media, 5 μL of labeled hSCT (final concentration of 500 nM), and 5 μL of the peptide analogs (final concentration of 10 μM) were added to each well. The microplate was centrifuged and incubated at 4-8°C for 1 h. Unlabeled hSCT and hGLU were used as positive and negative controls, respectively. Peptide analogs that exhibited binding at a concentration of 10 μM were evaluated for a dose response ranging from 10 −12 M to 10 −4 M to determine the IC 50 value. The nonspecific FRET signals were measured with 100 μM unlabeled hSCT at respective concentrations of peptide analogs and was corrected from the total binding. The absorbance was measured at 615 and 520 nm in a VICTOR™ X4 spectrofluorometer (PerkinElmer), and the FRET signals were analyzed.

Functional assays
Agonist-cAMP response. The cAMP responses in hSR-transfected CHO-K1 cells were detected using the HTRF-LANCE 1 cAMP assay kit. After 48 h of transfection, the cells were detached by nonenzymatic treatment using Versene. The cells were counted and processed according to the manufacturer's instructions to measure the cAMP response following cell treatment with the secretin analogs. hSCT was used as a positive control. The TRF signals (340 nm excitation/665 nm emission) were detected using a VICTOR X4 spectrofluorometer (Perki-nElmer), and the cAMP concentration was determined using a standard curve.
Antagonist-cAMP response. The antagonistic properties of the peptide analogs were analyzed by coincubation of hSCT with increasing concentrations of the analogs, and the changes in the cAMP responses were measured using the HTRF-LANCE 1 cAMP assay kit. The detached cells were diluted in the stimulation buffer with Alexa Fluor 1 647-conjugated anti-cAMP antibody, and 10,000 cells per well (10 μL) were seeded in 384-well microplates and incubated with 5 μL of 5 nM hSCT plus 5 μL of the peptide analogs (10 −12 to 10 −6 M) for 30 min. hSCT was used as a positive control for agonist activity. Detection was subsequently performed according to the manufacturer's instructions.

Statistical analysis
All in vitro results were analyzed in quadruplicate, and the data are expressed as the means ± SEM. Significant differences were identified using Student's t-test at p<0.05. The dose response was analyzed using GraphPad Prism 5.0 software with a variable slope (four parameters).

Model building
The complete 3 dimensional model of hSR was created by in silico homology modeling approach. As there is no direct template available for 3D modeling of full hSR, the Nt domain was modeled with a multiple template approach separately from TM. First the Nt domain was modeled with multiple templates and the best model was selected with the help of SAVES (server analysis for model validation). The Ramachandran plot of the final model shows zero residues in the outlier region by rampage [42] (Fig 1A). Verify 3D [43] results showed 81.18% to have 3D-1D score > = 0.2. The TM region was then modeled using GCGR crystal structure as the template. The template query alignment was performed without gaps in the helices and gaps reduced to one in the loop region. These gaps were separately modeled using comparative loop modeling (Fig 2). The structural geometry was validated with Ramachandran plot and through confirmation of the beta sheet and helical conformation for the predicted backbone [42,[44][45][46] (Fig 1B). The position of each outlier was analyzed and was found to be negligible ( Table 2). The model of the Nt and TM were fused by rigid body docking using Pydock [47] (Fig 3). The best five models from a total of 102 were screened for structural validity. The tertiary structure of the receptor was screened by comparison with other class B templates [24] Table 2. and by considering the best ERRAT value. This models were screened by ProQM which is the only model quality assessment algorithm for membrane protein (Fig 4). The binding site of SCT on to the receptor was determined by Schrodinger protein-protein docking algorithm ( Fig  3B). The docking site was confirmed using binding energy estimation by docking the receptor with hVIP, hPACAP and hGIP (Table 3).

Model verification by in vitro assay
The model was validated by verifying the docking score and by in vitro cAMP (Fig 5) and FRET binding assays (Fig 6). The assay results were consistent with the docking results where hSCT was used as a positive control. Peptides like hVIP, hPACAP were found to exhibit some binding affinities in docking analyses could also activate hSR to produce cAMP, and their interactions with hSR were also confirmed by the FRET assay. The binding energy of hSCT, hVIP and hPACAP were -11.53, -9.51 and -6.31, respectively, while hGIP has a binding energy of +10.63, clearly indicates its inability to interact with hSR.

Virtual docking and cAMP studies on hSCT analogs
Structures of hSCT analogs were prepared by homology modeling and were used for virtual docking with hSR. Functional cAMP assays were performed in parallel to confirm the docking  result and to test the ability of any of these analogs acting as agonist or antagonist. The best model was used to calculate the binding energy of these secretin analogs. All the analogs, with the exception of analog 15 (rat secretin) that showed very high delta G, exhibited low binding potential at the binding site ( Table 4). The binding affinity of these SCT analogs were further evaluated, and only rat SCT was found to exhibit binding affinity for hSR (Fig 7) with an IC 50 value of 0.40 ± 0.35 nM, which is higher than that of hSCT (1.6 ± 1.1 nM). The virtual docking and in vitro binding assay show positive correlation for class B ligands (Table 4). Through cAMP assays, only rat SCT was identified as an agonist of hSR, whereas all other SCT analogs exhibited neither agonistic nor antagonistic properties ( Figure D in S1 File).

Discussion
The structural elucidation of membrane proteins, particularly GPCRs, is challenging. Therefore, homology modeling is a tempting alternative to generate a virtual 3D model of these receptors that can be used as a primary tool in virtual screening to understand the binding affinity against a library of analogs. GPCRs possess a similar topology and activation mechanism [48,49]. In this study, we have used a multiple-template approach to generate a 3D model of the hSR. Both the Nt and the TM regions have been individually modeled using respective templates for the active state receptor conformations. The secondary structure of the final Nt-TM fused model was studied/analyzed for structural validity. The tertiary structure of the receptor was compared with other class B templates [24]. The structural geometry was assessed using Ramachandran plot (Fig 3), all of the five outliers were found to lie in the nonbinding regions and the total outliers are below 2%. The model was further virtually confirmed by docking with hSCT and class B ligands and experimentally confirmed using biological assays. There has been an allosteric model proposed on receptor activation in class B GPCRs  where the C terminal of the peptide hormone first interacts with the Nt domain of the receptor while the N-terminal of the ligand subsequently interacts with the extracellular TM loops [50][51][52][53][54][55][56]. It is also worth mentioning here that there is a different model for hSR activation. It was found that minor modification at the N-terminal of the secretin peptide resulted in no direct interaction of the ligand with the receptor TM loops, but binding of the ligand only with the Nt of the secretin receptor was sufficient to activate the receptor [57]. This observation suggests the presence of an agonist epitope hidden within the receptor Nt [58], and it was hypothesized that the binding of natural ligands to the conserved disulfide-bonded in the N-terminal domain of the receptor may lead to a conformational change in the N-terminal for receptor activation [59]. In virtual docking, hSCT, hPACAP and hVIP were found to bind hSR with high binding energy in respective order. In vitro experiments revealed that hPACAP and hVIP also bind to hSR and activate downstream signaling, but only at high concentrations. These results confirm the weak affinity of hPACAP and hVIP for hSR and support the specificity of our active receptor model, hence the model may serve as a tool to identify agonists or antagonists by targeting the active binding site. We have therefore used this hSR model for virtual docking of the peptide analogs that we have synthesized. We found those analogs with modifications in either the Nt or the Ct region exhibited very low or no binding affinity for the receptor. However, human and rat SCT exhibited strong ligand interactions with good docking scores. Frog, chicken and coelacanth SCT exhibited docking scores lower than the one observed for hSCT. To verify the in silico data, all these analogs were studied by in vitro FRET-based competitive binding assay. In this assay, FRET signals were detected only when the labeled Nt of the receptor and the Ct of the labeled ligand were in close proximity. In this study, only human and rat SCTs could bind to hSR (IC 50 1.6 ± 1.1 and 0.40 ± 0.35 nM, respectively). Frog, chicken, coelacanth SCT and all other SCT analogs did not exhibit binding affinity even at a high concentration (Fig 6). As GPCRs are known to possess more than one binding site, we have also tested the agonistic (Table 5) or antagonistic properties of these analogs by cAMP assays (Figure D in S1 File), but none of these analogs exhibit activities excluding the possibility of allosteric binding for these modified peptide analogs, which is consistent with the virtual docking data. GPCRs possess a similar topology and activation mechanism [48,49]. Receptor-ligand interactions give rise to structural changes that result in multiple conformations, which is evident from the active and the inactive states [60,61]. Recent studies have shown both interesting and complex processes of GPCR ligand responses, with different signaling outcomes upon activation   [62]. The pharmacophores of hSCT are spread throughout the native ligand, as a result, the binding between the receptor and the SCT analogs could not be established due to instability/ inefficiency. For class B GPCRs, it was hypothesized that the Ct region of the ligand is involved in initial receptor binding, and the Nt region is involved in receptor activation [23]. In summary, both in vitro and in silico studies indicate that small modifications in either the Ct or Nt of SCT result in complete abolishment of their activity in activating the receptor. Analog 15 (rSCT) has modification in the center of the peptide, and hence remains active. In analyzing the binding energy score, the main deviation was in the repulsive van der waals forces, and these forces were low in hSCT, rSCT, hVIP, hPACAP while all other peptide analogs have very high values ( Table 6). Van der waals force is the sum of the attractive and repulsive force between molecules caused by fluctuating polarization of nearby particles [63]. The net total of van der waals forces can be attractive or repulsive [64]. The repulsive van der waals force in the peptide analogs (excluding rat secretin) were significantly elevated which is the main reason for the decrease in their binding affinities. Analyses of the docked structures by ligPlot+v.1.4.5 [41] reveal changes in the ligand-receptor interactions for all the analogs ( Figure E in S1 File) due to changes in the secondary structures arising from amino acid substitutions. It may be augmented that due to the small size of the ligand, minor modification can result in dramatic changes to its structure, resulting in loss of function. It was believed that the C-terminal of secretin is crucial to its initial interaction with the receptor while the N-terminal for subsequent activation of the receptor. Nt (activation region) modified analogs with the intact Ct should theoretically be able to interact with the receptor and may function as antagonists, but none of these Nt-modified analogs could bind or affect the cAMP response of hSR. In our study, modifications of either the N-or the Cterminal resulted in a loss of function as well as loss of affinity for the receptor, shown though Table 5. Pair wise sequence alignment of all analogs with hSCT.
Peptide sequence alignment of all the peptide analogs of hSCT with modification highlighted in yellow. Supporting Information S1 File. Fig A of S1. Primary sequence alignment of HSR. Depicts the primary sequence alignment of HSR fasta sequence with Class B N-terminal templates PACAP N-terminal (PDB 2JOD), VIPR N-terminal (PDB 2X57) and GLPR-1 N-terminal (PDB 3C5T). The NT constraints were preserved and aligned for HSR NT modeling. Fig B of S1. The primary sequence is alignment hSR. The primary sequence is aligned and highlighted in seven different colors depicting the TM1 to TM7.