Abstract
Methodological advances in X-ray crystallography have made possible the recent solution of X-ray structures of pharmaceutically important G protein-coupled receptors (GPCRs), including receptors for biogenic amines, peptides, a nucleoside, and a sphingolipid. These high-resolution structures have greatly increased our understanding of ligand recognition and receptor activation. Conformational changes associated with activation common to several receptors entail outward movements of the intracellular side of transmembrane helix 6 (TM6) and movements of TM5 toward TM6. Movements associated with specific agonists or receptors have also been described [e.g., extracellular loop (EL) 3 in the A2A adenosine receptor]. The binding sites of different receptors partly overlap but differ significantly in ligand orientation, depth, and breadth of contact areas in TM regions and the involvement of the ELs. A current challenge is how to use this structural information for the rational design of novel potent and selective ligands. For example, new chemotypes were discovered as antagonists of various GPCRs by subjecting chemical libraries to in silico docking in the X-ray structures. The vast majority of GPCR structures and their ligand complexes are still unsolved, and no structures are known outside of family A GPCRs. Molecular modeling, informed by supporting information from site-directed mutagenesis and structure-activity relationships, has been validated as a useful tool to extend structural insights to related GPCRs and to analyze docking of other ligands in already crystallized GPCRs.
Introduction
In the past 5 years, progress in the structure-based design of ligands for G protein-coupled receptors (GPCRs) has greatly accelerated. The major contributing factor has been the elucidation of X-ray crystallographic structures of high resolution for various drug-relevant GPCRs, initially in the inactive antagonist-bound forms and more recently in agonist-bound forms. The initial breakthrough were the reports in 2007 by the groups of Kobilka (Stanford Univ.), Stevens (Scripps Research Inst.), Schertler and Tate (Medical Research Council Laboratory of Molecular Biology in Cambridge, UK), and colleagues of the first nonrhodopsin GPCR structure (e.g., the inactive human β2-adrenergic receptor in complex with the inverse agonist carazolol) (Cherezov et al., 2007; Rasmussen et al., 2007; Rosenbaum et al., 2007). These landmark studies were followed by the determination of other GPCRs (Table 1), and the rapid pace of these reports is continuing. Biogenic amine receptor complexes (epinephrine, dopamine, histamine, muscarinic), nucleoside (adenosine) receptor complexes, sphingolipid (S1P1) receptor complexes, and peptide (CXCR4, opioid) receptor complexes have been reported. All of the crystallized receptors belong to the GPCR family known as class A, family 1, or rhodopsin family, which in humans accounts for more than 80% of all GPCRs (Costanzi, 2012; Krishnan et al., 2012). As shown in Fig. 1, most of these receptors belong to a branch of class A that comprises receptors for biogenic amines melanocortin, endothelial differentiation sphingolipids, cannabinoid and adenosine receptors. The only exceptions are the chemokine CXCR4 and opioid receptors (κ and μ), which are found in a branch of class A predominantly populated by peptide receptors, and rhodopsin, which is found in a small branch of class A populated by opsins. Moreover, the solution of the nociceptin/orphanin FQ peptide receptor has recently been announced. A large portion of the dendrogram of class A GPCRs, including a branch that comprises mostly receptors for nucleotides and lipids, is still unexplored. There is reason to expect that many other structures will be solved in the near future to shed light onto a still-uncharted region of the GPCR phylogenetic dendrogram. Furthermore, the solution of receptors belonging to families beyond class A is expected.
The technical advances that led to this dramatic progress include the following:
Fusion of the receptor with the T4-lysozyme, which increases the tendency to form crystals—the T4-lysozyme is usually inserted in lieu of intracellular loop 3 (Cherezov et al., 2007; Rosenbaum et al., 2007) but has also been fused with the N terminus, to facilitate the cocrystallization of a receptor-G protein complex (Rasmussen et al., 2011b);
Structurally stabilizing point mutations, either in the ligand-binding regions or more remotely—this approach has been introduced by the groups in Cambridge and the associated company Heptares (Warne et al., 2008). These thermostabilizing mutations can be made to favor either an antagonist-bound conformation or agonist-bound conformation (although poorly activatable because of the energetic stabilization) of the same GPCR. The stabilization is so effective that the GPCR protein can be captured on a Biacore chip to allow characterization of small-molecule binding by measuring surface plasmon resonance (Zhukov et al., 2011);
Stabilization of the receptors through antibodies (Rasmussen et al., 2007)—recently, nanobodies generated in llamas inoculated with receptors as well as receptors cross-linked with G protein heterotrimers have been used to solve crystal structures of the activates state of the β2-adrenergic receptor (Rasmussen et al., 2011a,b);
Specialized agonists, such as irreversibly binding agonists (Rosenbaum et al., 2011) or an agonist that has multiple arms extending from the core pharmacophore structure (Xu et al., 2011); and
Specialized techniques for producing crystals of membrane-bound proteins, including adaptation of the lipidic cubic phase, which forms a single lipid-aqueous bilayer that allows ordered protein molecules to make contacts in their hydrophobic as well as hydrophilic portions (Cherezov, 2011).
The Ligand Binding Cavity
All GPCRs are constituted by a single polypeptide chain that spans the plasma membrane seven times with seven α-helical structures (Costanzi et al., 2009). As the crystal structures revealed, the helical bundle of most class A GPCRs hosts a ligand-binding cavity opened toward the extracellular milieu, which provides access to diffusible ligands. Alternately, the cavities of rhodopsin and S1P1 receptor are sealed from the extracellular space by the second extracellular loop (EL2) and the N terminus, respectively. It is likely that the hydrophobic ligands of these receptors make their way into the binding cavity through the transmembrane domains. For most GPCRs, this ligand-binding cavity is lined by transmembrane domains (TMs) 2, 3, 5, 6, and 7 and is deeper in proximity to TMs 5 and 6 but shallower in proximity to TMs 2 and 7 (Fig. 2). Nevertheless, in some cases, TMs 1 and 4 also form the pocket and can affect ligand binding. The various ligands cocrystallized with their receptors in the currently solved GPCR structures variously occupy different regions of this cavity. This is evident from Fig. 3, which shows a side-by-side comparison of six representative receptors featuring the bound ligands and the residues that surround them, as well as Fig. 4, which provides an overlay of the same ligands of these receptors resulting from a structural superposition of the receptors. It is noteworthy that different ligands that bind to the same receptor may occupy different regions of the binding cavity. This situation is particularly evident from Fig. 5A, which compares the binding, to the CXCR4 receptor, of a small molecule antagonist and a cyclic peptide antagonist (Wu et al., 2010): there is virtually no overlap between the two molecules. Less extreme is the case illustrated in Fig. 5B, which compares the binding to the adenosine receptor of a nonpurine antagonist and an adenosine-based agonist 2-(3-[1-(pyridin-2-yl)piperidin-4-yl]ureido)ethyl-6-N-(2,2-diphenylethyl)-5′-N-ethylcarboxamidoadenosine-2-carboxamide (UK-432097) bearing large substituents at the C2 and N6 positions of the purine ring (Xu et al., 2011): although there is substantial overlap between the two molecules, the larger agonist occupies regions of the receptor unexploited by the antagonist. The ribose moiety, which is an essential component of nearly all adenosine receptor agonists, confers to the ligand its agonistic properties. This moiety is accommodated in a space of the binding cavity close to TM3 that in the antagonist-bound state is unoccupied by a ligand and is partially filled by water molecules. The hydroxyl and other H-bonding groups of agonists displace these water molecules, thereby gaining an entropic advantage in the binding process. This is consistent with the typically nanomolar affinities of agonists at three of the four subtypes of adenosine receptors. The other large substituents present on the agonist UK-432097 cocrystallized with the A2A receptor further enhance the binding affinity of the agonist by establishing contacts with the receptor.
The outer regions of a GPCR can serve as meta-binding sites for a ligand on its path to the principle orthosteric binding site and can also contribute to the lining of the orthosteric binding site. The X-ray structures have confirmed the hypothesis, based on mutagenesis as well as molecular modeling (Olah et al., 1994; Moro et al., 1999; Peeters et al., 2011), that parts of the ELs are intimately involved in recognition of ligands for both agonists and antagonists. In particular, we now know that the C-terminal part of EL2 tends to drop more or less deeply, depending on the receptor, into the ligand-binding region to establish contacts with the ligands (Fig. 3), as predicted using site-directed mutagenesis. Peculiar are the cases of rhodopsin and the S1P1 receptor, in which the ligand-binding cavity is substantially more enclosed than in other receptors, thanks to a singular conformation of EL2 that occludes the entrance of the cleft (Palczewski et al., 2000). Conversely, the conformation that EL2 assumes in the chemokine CXCR4 receptor renders the binding cavity particularly open to the extracellular space, facilitating the binding of large peptides (Wu et al., 2010). However, the role of ELs in recognition of and activation by large peptide ligands will require further studies (Dong et al., 2011).
In general, the extracellular loops are typically more varied between subtypes than the TMs, which could facilitate the rational design of more selective orthosteric and allosteric ligands, as guided by the three-dimensional structures of the receptor complexes. For instance, the M2 (Haga et al., 2012) and M3 (Kruse et al., 2012) muscarinic acetylcholine receptors contain a vestibule in their outer portions that corresponds to the regions of the receptor associated with the binding of allosteric modulators. The fact that the orthosteric site of the muscarinic receptors and the specific residues involved in ligand recognition are identical across the family has impeded the medicinal chemical effort to design selective competitive ligands. The possibility of targeting the vestibule, which is more divergent in sequence, now offers further opportunities for drug design. However, for those receptors for which a crystallographically solved structure is not yet available, the high variability of the extracellular regions makes their modeling subject to greater uncertainty compared with the modeling of the seven TMs (Goldfeld et al., 2011).
Inactive and Activated Structures.
The determination of both agonist- and antagonist-bound states of the same receptor was accomplished for two classes (i.e., β-adrenergic and adenosine receptors, in addition to visual pigment receptors with the structures of opsin and rhodopsin) (Park et al., 2008; Scheerer et al., 2008; Choe et al., 2011; Rasmussen et al., 2011a,b; Standfuss et al., 2011; Xu et al., 2011). This major step forward allowed a deeper understanding of the activation processes operating in GPCRs. Thus, the details of agonist binding and activation (i.e., characteristic agonist-induced movement of helices and key residues) are beginning to be understood. The comparison of the structures of inactive rhodopsin with those of the meta II state of rhodopsin as well as the unliganded opsin revealed several conformational changes associated with the activation of the receptor. The most notable of these conformational changes entail outward movements of the intracellular side of TM6 and movements of TM5 toward TM6 (Fig. 6, A and B). These conformational changes were also found in the agonist-bound β2-adrenergic and A2A adenosine receptors, although the displacement of TM6 is substantially less pronounced in the A2A receptor because of the T4-lysozyme fused between the cytosolic ends of TM5 and TM6 (Fig. 6, C and D). In addition, the distance between oppositely charged amino acid side chains near the cytosolic side (e.g., an Arg and an Asp residue, respectively, found in TMs 3 and 6 and predicted to be involved in a putative ionic lock characteristic of the inactive state, increased upon binding of the agonist in each case. However, it should be noted that this ionic lock was not fully closed in the inactive states of the β2-adrenergic and A2A adenosine receptors. Furthermore, some movements that were not predicted in the opsin structure were seen in other receptors, such as a seesaw movement of TM7 in the A2A adenosine receptor by which the intracellular end moves inward (Fig. 5, C and D). Movements in the EL regions were also noted. These movements seem to be more ligand-specific than the movements in the TM regions. For example, the outward displacement of EL3 of the A2A adenosine receptor was considerably greater for an agonist having bulky substitutions of the adenine ring than in the unsubstituted cases. It is noteworthy that increasing evidence demonstrates that the stimulation of one GPCR can trigger different signaling cascades in a ligand-dependent manner through a phenomenon known as biased agonism (Kahsai et al., 2011). As a result, the same receptor can be selectively induced to activate a variety of pathways mediated by G proteins as well as β-arrestins. These different signaling states are probably due to distinct conformations of the same receptor that individual ligands can induce or stabilize. Further light will be shed on the correlation between conformation and signaling state of GPCRs once multiple structures of the same receptor with a variety of biased ligands are solved. In this respect, some progress has been made for the β1- and β2-adrenergic receptors through crystallographic (Warne et al., 2012) and NMR studies (Liu et al., 2012), respectively. For the β1-adrenergic receptor, the crystal structures revealed that the biased agonists bucindolol and carvedilol, which stimulate β-arrestin-mediated signaling but act as inverse agonists or partial agonists of G protein-dependent pathways, interact with additional residues located in EL2 as well as TM7 compared with unbiased β-adrenergic receptor blockers. Moreover, the above-mentioned NMR spectroscopic analyses of the β2-adrenergic receptor suggest that ligands, including biased ligands, do not “induce” states but shift equilibrium between pre-existing states. Specifically, the study indicates that unbiased ligands affect mostly the conformational state of TM6, whereas biased agonists shift primarily the conformational state of TM7 (Liu et al., 2012).
Molecular Docking at GPCR Homology Models to Predict the Structure of Receptor Ligand Complexes
Until recently, the three-dimensional study of the interactions of GPCRs with their ligands was limited to molecular modeling. The modeling efforts began more than 2 decades ago, using as a crude template the structure of bacteriorhodopsin, and progressed with the major advance of the high-resolution structure of bovine rhodopsin more than a decade ago (Ballesteros et al., 2001; Costanzi et al., 2007, 2009). Although there has been much variability in the confidence in the modeling results (articles were published with diametrically opposed modes of ligand binding for the same receptor-ligand complex), there were examples of well supported modeling that was later validated crystallographically (among others, see Ivanov et al., 2009; Michino et al., 2009; Kufareva et al., 2011). Comparisons of crystal structures and homology models revealed that, for some GPCRs, reasonably accurate receptor-ligand complexes can be constructed through homology modeling followed by molecular docking (Costanzi, 2008, 2010, 2012; Reynolds et al., 2009). However, accurate complexes cannot always be obtained through the sole use of specialized software. Purely computational docking of ligands at GPCR homology models could lead to substantially inaccurate results if the contributions of specific residues to ligand biding or, even worse, the location of the binding cavity were incorrectly recognized. A community-wide challenge to molecular modelers to predict the docking mode of antagonist 4-(2-(7-amino-2-(furan-2-yl)-[1,2,4]triazolo[1,5-a][1,3,5]triazin-5-ylamino)ethyl)phenol (ZM241385) to the A2A adenosine receptor before the release of the X-ray structure showed how the ligand could assume almost a random placement around the receptor (Michino et al., 2009). However, when informed with accessory information about the molecular recognition elements, the placement of the ligand in various docking models was shown to be within a reasonable range of accuracy (Michino et al., 2009). In many cases, the identification of the correct receptor-ligand interactions is strictly dependent on an expert selection of the docking poses based on insights derived from site-directed mutagenesis, comparisons of SAR within the same chemical series, and bioinformatics studies within a receptor family.
A further controlled assessment clarified that, when the target receptor shares a significant sequence similarity with one of the available templates, the models are particularly accurate. Conversely, predictions are more challenging for receptors that are more distant from the available templates, in which case the modeling strategies need to be more closely guided through the incorporation of the above-mentioned external information (Kufareva et al., 2011). Because, as we pointed out, there are not yet X-ray structures representative of all major branches of the GPCR dendrogram (Fig. 1), the modeling of receptors that are more distant from the available templates is still challenging. Moreover, the modeling of GPCRs belonging to classes B and C (Hu et al., 2006; Wheatley et al., 2012) remains more uncertain, because none of the structures of the members of these families have been solved crystallographically.
As mentioned, particularly important is the use of data gathered from site-directed mutagenesis. In this regard, one variation of the use of site-directed mutagenesis that has been particularly informative with respect to probing molecular recognition among GPCRs has been that of re-engineering the binding site to accept agonists that have been chemically modified. These approaches, with some important differences, are known by various terms introduced by different research groups: neoceptors, “receptors activated solely by synthetic ligands,” and “designer receptors exclusively activated by designer drugs” (Conklin et al., 2008). Different degrees of design insight versus empirical screening have been used to match a given mutant receptor with an orthogonally activating agonist analog. The neoceptor approach, in particular, has focused on the adenosine receptor family to accurately predict the placement of the ribose moiety of nucleoside agonists, using the three-way integrated combination of mutagenesis, modeling, and chemical modification. Complementary changes in the structures of the ligand and receptor that lead to enhanced affinity can help establish the orientation of a ligand within the binding site.
Receptor-ligand interactions can also be studied through a systematic approach that was recently introduced, termed biophysical mapping (Zhukov et al., 2011). This method is based on the characterization of the functional contour of the binding pocket of a given GPCR using a thermostabilized form of the receptor. The effects of site-directed mutagenesis within the binding site are correlated with binding data for diverse ligands obtained through SPR measurements. Then, molecular modeling and docking are used to map the small molecule-binding site with respect to each chemical class of ligands. This approach was used to identify novel chemotypes (later to be optimized by chemical modification), such as chromones and triazines, binding to the A2A adenosine receptor (Congreve et al., 2012; Langmead et al., 2012).
Structure-Based Discovery of GPCR Ligands Is Increasingly More Practical
Looking ahead, it will be increasingly feasible to tap the potential of structure-based design for GPCRs, initially for class A (Congreve et al., 2011; Salon et al., 2011), and then, hopefully, for the other classes as well. For now, the newly revealed detailed knowledge of GPCR structures has already facilitated a recent flurry of studies directed toward ligand discovery.
The careful stepwise modification of known classes of agonist or antagonist for a given receptor has been for years a major successful approach of medicinal chemists when applied empirically. This study can now be expedited using accurate three-dimensional knowledge of receptor-ligand recognition. For instance, it is now feasible to target specific amino acid residues near the bound pharmacophore that might confer enhanced affinity or receptor subtype selectivity to the modified ligands. For example, recent reports have shown that A2A adenosine receptor agonists and antagonists can be modified in this manner (Congreve et al., 2012; Deflorian et al., 2012). Moreover, the interaction of small fragments with regions of the binding cavity proximal to the ligand, to be considered as candidate substituents to enhance the receptor-ligand interactions, can also be studied. In this context, in recent studies, the structure of a known GPCR agonist was systematically varied using the ICM software (Internal Coordinate Mechanics; Molsoft LLC, San Diego, CA). A library of 2000 small fragments was screened in silico for fit within a small pocket, to successfully predict those favoring adenosine receptor affinity when linked to the 5′-carbonyl group of modified adenosine (Tosh et al., 2012).
For the identification of ligands based on novel chemotypes, a technique that has proven its value is the virtual screening through molecular docking of chemically diverse libraries for the discovery of novel chemotypes that bind to various GPCRs (Costanzi, 2011). A number of controlled experiments targeting the β-adrenergic and adenosine receptors, conducted by subjecting to molecular docking known agonists and blockers together with a larger number of nonbinders, clearly illustrated that such virtual screening campaigns are most effective when applied to X-ray structures (Reynolds et al., 2009; Vilar et al., 2010, 2011a; Costanzi and Vilar, 2012). This observation is consistent with the successful identification of novel structurally diverse ligands on the basis of virtual screenings conducted by targeting the crystal structures of β-adrenergic, adenosine, dopamine, and histamine receptors (Sabio et al., 2008; Kolb et al., 2009; Carlsson et al., 2010, 2011; Katritch et al., 2010a; de Graaf et al., 2011; van der Horst et al., 2011; Langmead et al., 2012). The above-mentioned controlled experiments also demonstrated that virtual screening campaigns, although not as effective as when applied to a crystal structure, are useful when applied to accurate homology models (Cavasotto et al., 2008; Katritch et al., 2010b; Phatak et al., 2010; Vilar et al., 2010, 2011a; Cavasotto, 2011). This observation is in line with the results of virtual screening campaigns through which, before the explosion of GPCR crystallography, novel GPCR ligands were identified using rhodopsin-based homology models (Engel et al., 2008; Tikhonova et al., 2008). More recently, Carlsson et al. (2011) conducted a virtual screening campaign by targeting the crystal structure of the dopamine D3 receptor as well as a model of the same receptor based on the β2-adrenergic homolog, with which it shares a relatively high sequence identity (38% within the TMs and 61% within the binding cavity, defined as the residues found within a 4 Å radius from the bound ligands). It is noteworthy that each of these two parallel campaigns yielded two overlapping sets of novel ligands, suggesting that models based on a relatively close template are indeed useful to ligand discovery.
Controlled experiments demonstrated also that not only blockers but also agonists are substantially prioritized over nonbinders in docking experiments (Costanzi and Vilar, 2012). In particular, such controlled virtual screening campaigns revealed that the activated structure of the β2-adrenergic receptor is significantly biased toward the preferential recognition of agonists over blockers. Moreover, they also showed that structures of receptors crystallized in the inactive state can be modified in silico by modeling the shape of the binding cavity around a docked agonist, thus being turned into structures that preferentially recognize agonists rather than blockers (Vilar et al., 2011b; Costanzi and Vilar, 2012). However, the identification of agonists on the basis of novel chemotypes through the screening of diverse libraries may prove particularly challenging, in light of the likely stricter structural requirements for agonism than for blockade. Moreover, as mentioned, it is increasingly evident that the same GPCR may trigger a variety of different signaling cascades (Kahsai et al., 2011). The basis for distinguishing ligands needed for selective effector pathway activation (i.e., biased ligands) is an important area for future investigation. Several examples of ligand-specific interactions of the same receptor have been reported, but the implications of these differences for signaling are still largely unknown.
Undoubtedly, docking-based virtual screening campaigns will become increasingly more feasible with the experimental determination of new GPCR structures. Moreover, they will also benefit from the fast pace at which supercomputing is progressing as well as the continuous improvement of computational algorithms. In particular, the scoring functions that are currently used to estimate the likelihood of binding of large sets of screened compounds represent a compromise between accuracy and rapidity. Large computer clusters as well as specialized purpose-built supercomputers will increasingly allow the development and application of more complex methods for the calculation of free binding energies (Mitchell and Matsumoto, 2011). Moreover, an increasingly higher number of alternative receptor conformations, either solved experimentally or generated computationally from a single experimental structure, will be applicable in parallel to the screening campaigns. This practice, known as receptor-ensemble docking, was already demonstrated to significantly improve virtual screening yields by providing a means to account for receptor flexibility (Bottegoni et al., 2011; Costanzi, 2011; Vilar et al., 2011a; Costanzi and Vilar, 2012).
Conclusions
On the basis of methodological advances in X-ray crystallography, the structural elucidation of GPCRs has begun a revolution in the medicinal chemical approaches applied to the discovery of new GPCR ligands. The binding sites of different receptors partly overlap but differ significantly in ligand orientation, depth and breadth of contact areas in TM regions, and the involvement of the ELs. Conformational changes associated with activation have been analyzed for several receptors. However, there are still large areas in which knowledge is lacking. For example, there still is an uncharacterized large portion of the GPCR phylogenetic dendrogram, the interaction of large peptide ligands with their receptors are unclear, and the structural basis for functional selectivity (i.e., why agonists display different spectra of activation properties through the same GPCR) are poorly understood. Molecular modeling, informed by supporting information from site-directed mutagenesis and structure activity relationships, has been validated as a useful tool to extend structural insights to related GPCRs and to analyze docking of other ligands in already crystallized GPCRs. Further exploration of the interactions of GPCRs with their G protein and non–G-protein intracellular targets, through medicinal chemistry as well as techniques of structural biology, will undoubtedly be required.
Authorship Contributions
Wrote or contributed to the writing of the manuscript: Jacobson and Costanzi.
Footnotes
This work was supported by the National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases [Grants Z01-DK031126-08, Z01-DK013025-05].
Article, publication date, and citation information can be found at http://molpharm.aspetjournals.org.
ABBREVIATIONS:
- GPCR
- G protein-coupled receptor
- S1P
- sphingosine-1-phosphate
- EL
- extracellular loop
- FAUC50
- (R)-5-(2-((4-(3-((2-aminoethyl)disulfanyl)propoxy)-3-methoxyphenethyl)amino)-1-hydroxyethyl)-8-hyfroxyquinolin-2(1H)-one
- UK-432097
- 2-(3-[1-(pyridin-2-yl)piperidin-4-yl]ureido)ethyl-6-N-(2,2-diphenylethyl)-5′-N-ethylcarboxamidoadenosine-2-carboxamide
- ZM241385
- 4-(2-(7-amino-2-(furan-2-yl)-[1,2,4]triazolo[1,5-a][1,3,5]triazin-5-ylamino)ethyl)phenol
- IT1t
- (Z)-6,6-dimethyl-5,6-dihydroimidazo[2,1-b]thiazole-3-yl-N,N′-dicycylohexylcarbamimidothioate
- ML056
- (R)-3-amino-(3-hexylphenylamino)-4-oxobutylphosphonic acid
- CVX15
- cyclic disulfide of H-Arg-Arg-Nal-Cys-Tye-Gln-Lys-d-Pro-Pro-Tyr-Arg-Cit-Cys-Arg-Gly-d-Pro-OH.
- Received April 16, 2012.
- Accepted June 13, 2012.
- U.S. Government work not protected by U.S. copyright
References
MolPharm articles become freely available 12 months after publication, and remain freely available for 5 years.Non-open access articles that fall outside this five year window are available only to institutional subscribers and current ASPET members, or through the article purchase feature at the bottom of the page.
|