Research ArticleBiochemistry

Structure-function guided modeling of chemokine-GPCR specificity for the chemokine XCL1 and its receptor XCR1

See allHide authors and affiliations

Science Signaling  03 Sep 2019:
Vol. 12, Issue 597, eaat4128
DOI: 10.1126/scisignal.aat4128

Modeling ligand-receptor specificity

Ligand-receptor interactions trigger many cellular behaviors and changes that underlie various physiologic processes. Understanding what determines the specificity of these interactions, therefore, can explain disease and inform the development of targeted therapeutics. Using a hybrid modeling approach, Fox et al. generated a model for the chemokine XCL1 in complex with its G protein–coupled receptor XCR1 and identified key interaction sites. Changes within a region of XCL1 that determined its binding energy to XCR1 changed the activity of the receptor and, consequently, cell migration. In addition to showing the power of such hybrid approaches, these findings specifically on this chemokine signaling axis may have implications for diseases associated with altered dendritic and T cell immune responses.


Chemokines interact with their G protein–coupled receptors (GPCRs) through a two-step, two-site mechanism and, through this interaction, mediate various homeostatic and immune response mechanisms. Upon initial recognition of the chemokine by the receptor, the amino terminus of the chemokine inserts into the orthosteric pocket of the GPCR, causing conformational changes that trigger intracellular signaling. There is considerable structural and functional evidence to suggest that the amino acid composition and length of the chemokine amino terminus is critical for GPCR activation, complementing the size and amino acid composition of the orthosteric pocket. However, very few structures of a native chemokine-receptor complex have been solved. Here, we used a hybrid approach that combines structure-function data with Rosetta modeling to describe key contacts within a chemokine-GPCR interface. We found that the extreme amino-terminal residues of the chemokine XCL1 (Val1, Gly2, Ser3, and Glu4) contribute a large fraction of the binding energy to its receptor XCR1, whereas residues near the disulfide bond–forming residue Cys11 modulate XCR1 activation. Alterations in the XCL1 amino terminus changed XCR1 activation, as determined by assessing inositol triphosphate accumulation, intracellular calcium release, and directed cell migration. Computational analysis of XCL1-XCR1 interactions revealed functional contacts involving Glu4 of XCL1 and Tyr117 and Arg273 of XCR1. Subsequent mutation of Tyr117 and Arg273 led to diminished binding and activation of XCR1 by XCL1. These findings demonstrate the utility of a hybrid approach, using biological data and homology modeling, to study chemokine-GPCR interactions.


The chemokine family consists of ~50 ligands and ~20 corresponding G protein–coupled receptors (GPCRs). Promiscuity, the ability of one chemokine to interact with multiple GPCRs and vice versa, is abundant in the chemokine network. Despite this promiscuity, chemokines mediate an extraordinarily specific and diverse array of homeostatic, inflammatory, and pathologic functions. This study focuses on understanding the interactions between the X-C motif chemokine ligand 1 (XCL1) and its cognate GPCR, XCR1. The XCL1-XCR1 axis is a major facilitator of dendritic cell and T cell immune responses throughout the body. For example, mice lacking either XCL1 or XCR1 have a diminished CD8+ T cell response and lack the ability to generate T regulatory cells. This not only affects the T cell population in the thymus (1) but also influences T cell populations in the intestines of mouse models of colitis (2). The XCL1-XCR1 axis has been explored as a venue to enhance the efficacy of various cancer treatments, including antitumor vaccines (310), cancer gene therapies (1113), and combined gene and adoptive T cell therapies for cancers (1416). These treatments have been proposed for a wide variety of cancers including myeloma (9, 16), B16 melanoma (3, 7, 8, 13), 3LL lung carcinoma (5, 8), breast cancer (11, 17), neuroblastoma (4, 10), lymphoma (14, 15), hepatocellular carcinoma (6), and colon carcinoma (12). As the immunological role and therapeutic potential of the XCL1-XCR1 axis continues to emerge, a detailed understanding of the key components that mediate XCL1-XCR1 binding and activation will enhance our ability to design XCL1-based therapeutics ranging from prophylactic vaccines to cancer treatments.

Chemokines engage their cognate GPCRs through a two-step, two-site binding mechanism of recognition and response (1820). Like a nametag, the core of the chemokine bears an epitope that is recognized by the receptor N terminus (site 1). Once identity has been confirmed, information is passed from the ligand to the receptor through insertion of the flexible N-terminal domain of the chemokine ligand into the orthosteric pocket of the GPCR (site 2). The orthosteric pocket is defined as the main ligand-binding pocket, and for chemokine GPCRs, this binding site is typically located within the transmembrane (TM) domain of the receptor (21). Residues of the chemokine N terminus make specific contacts within the orthosteric site that induce conformational changes and ultimately lead to a cellular response (step 2) (Fig. 1) (20). In addition to cytoskeletal remodeling and cellular migration, these pathways also signal for receptor desensitization and GPCR internalization (22). Under normal circumstances, the process of receptor binding, activation, and desensitization is highly controlled. When this control is lost, aberrant chemokine signaling can contribute to a number of pathological conditions including autoimmune disease (23) and cancer progression (24).

Fig. 1 Working model of XCL1 CC3-XCR1 binding and receptor activation.

This figure illustrates the typical chemokine-receptor activation through a two-site mechanism. Once activated, the GPCR XCR1 activates a G protein that initiates several downstream signaling cascades. For brevity, only a few pathways are illustrated here. To pinpoint amino acids that are crucial for XCR1 activation, we completed alanine-scanning mutagenesis on the first 10 amino acids of XCL1-CC3 (V1 to T10). Each variant was then analyzed for its ability to bind and activate XCR1 and induce downstream signaling events. These signaling events, indicated by shaded gray boxes, were measured through various cellular assays.

In this study, we conducted a detailed structure-function and computational analysis of the site 2 interactions between XCL1 and XCR1. Unlike other chemokine family members, XCL1 exists in a unique, dynamic equilibrium between two native state conformations under near physiological conditions (2527). One state is monomeric and retains the canonical chemokine fold. This state binds and activates XCR1 and does not bind to glycosaminoglycans (GAGs) (28). The second state is an atypical β-sandwich dimer that does interact with GAGs and establishes chemotactic gradients while not activating XCR1 (27). Descriptions of these extreme dynamic fluctuations are uncommon in structural biology, and proteins like XCL1 that exhibit similar behaviors are classified as metamorphic (29). Several studies demonstrate that the metamorphic behavior of XCL1 plays an important role in its biological activity. For example, the alternative XCL1 dimer conformation inhibits HIV infection in peripheral blood mononuclear cells, but the monomeric chemokine fold does not (3032).

Using an engineered XCL1 monomer that retains full XCR1 agonist activity but is locked into the monomeric chemokine fold (28), we conducted alanine-scanning site-directed mutagenesis on the 10 N-terminal amino acids of XCL1 to identify key residues important for binding and activation of XCR1 (site 2). After measuring the receptor binding of each variant, we then conducted several assays to capture discrete signaling events and measured the ability of each XCL1 variant to activate XCR1. Last, in conjunction with the aforementioned experimental approaches, we used computational modeling, docking, and molecular dynamics (MD) simulations to generate a model of the XCL1-XCR1 site 2 interactions. We then validated a subset of these computationally defined interactions in vitro through binding and signaling assays. Together with our previous work mapping its GAG binding site (33), our findings define the functional roles for a number of XCL1 residues. This analysis also demonstrates the utility of using a combined in vitro and in silico approach to study chemokine-GPCR interactions.


Amino acid substitutions in the XCL1 N terminus result in diminished XCR1 binding

Our initial attempts to develop a radioligand-binding assay using 125I-labeled wild-type (WT) XCL1 were unsuccessful due to high levels of nonspecific binding. We speculated that the metamorphic native state of XCL1 was the cause and overcame this limitation by using an engineered variant of XCL1 that prevents metamorphic interconversion. This variant, XCL1 CC3, contains two substitutions (V21C and V59C) that form a second disulfide bond, which locks the protein into the monomeric chemokine fold (28). Our previous work demonstrated that XCL1 CC3 was slightly more potent than XCL1 WT as XCR1 agonist in an intracellular Ca2+ flux assay (28). A recent study demonstrated that a murine version of XCL1 CC3 is a more potent XCR1 agonist in vitro than is WT mXCL1 and is more effective in vivo as an adjuvant for the induction of antigen-specific effector CD8+ T cells (34).

First, radioligand displacement assays were used to measure the ability of each XCL1 N-terminal variant to displace 125I-XCL1 CC3 having the native N terminus. XCL1 WT was included in all analyses for comparison. COS-7 cells transfected with human XCR1 were treated with 125I-XCL1 CC3 and various concentrations of XCL1 variants. XCL1 CC3 displaces 125I-XCL1 CC3 with an EC50 (median effective concentration) of 1 nM (Fig. 2A and table S1). XCL1 WT had a sevenfold decreased affinity (EC50 = 7.7 nM), which may be due to its metamorphic conversion to the dimer form, thus lowering agonist concentrations and leading to weaker potency. The XCL1 CC3 variants, V1A, G2A, S3A, and E4A had decreased affinities ranging from 30- to 375-fold. G2A and E4A variants had the lowest affinity. Mutations that remove the N-terminal valine (∆Val1) or add an additional glycine on the extreme N terminus (+GlyN-term) also showed decreased affinities of >100-fold. With the exception of V5A and K8A (EC50 = 8.0 and 5.2 nM, respectively), mutant residues S6A and R9A had increased affinity for XCR1 binding (EC50 < 1.0 nM). The EC50 values for D7A and T10A were not significantly different when compared to XCL1 CC3 (table S1). Overall, alterations to residues 1 to 5 showed larger changes in binding affinity than did residues 6 to 10 (Fig. 2A).

Fig. 2 125I-XCL1 CC3 displacement and 3H-IP3 accumulation by XCL1 N-terminal variants.

(A) XCR1 transfected COS-7 cells were incubated with 125I-XCL1 CC3 along with unlabeled XCL1 ligands at concentrations indicated above for 3 hours, and γ radiation was measured (n ≥ 4). Corresponding EC50 values and other fitting parameters for each ligand can be found in table S1. The dotted line indicates the EC50 for XCL1 CC3 = 1 nM and is added to all graphs as a point of reference. (B) COS-7 cells transfected with XCR1 and Gqi4myr were incubated with [3H]myo-inositol overnight, washed, incubated with XCL1 ligands for 1.5 hours, and lysed, and then scintillation was measured (n = 3 experiments). The solid lines represent 3H-IP3 measurement for each XCL1 ligand. The dotted curves represent 3H-IP3 measurement for XCL1 CC3 and are added to all other XCL1 ligand plots as a point of reference. Corresponding EC50 values and other fitting parameters for each ligand can be found in table S1.

Alteration of XCL1 N-terminal residues 1 to 5 disrupt inositol triphosphate accumulation

Inositol triphosphate (IP3) accumulation assays are a common approach used to study chemokine receptor and G protein activation (35, 36). After chemokine binding and receptor activation, the Gα and βγ subunits dissociate from the intracellular face of the receptor. Depending on the type of GPCR, either the Gα or the βγ subunit then activates phospholipase C (PLC), leading to cleavage of the membrane-bound phosphatidylinositol-4,5-bisphosphate (PIP2) and the formation of IP3 and diacylglycerol. For chemokine receptors, the βγ subunit is responsible for activation of PLC and formation of IP3 (37). IP3 functions as a secondary messenger by diffusing to the endoplasmic reticulum (ER) and mobilizing intracellular calcium stores that lead to a variety of intracellular signaling cascades (Fig. 1) (38).

XCL1 N-terminal variants were incubated with COS-7 cells transfected with XCR1 and Gqi4myr. Gqi4myr is an engineered chimeric Gαq protein containing several modifications that allow for Gαi receptors (i.e., chemokine receptors) to signal through the Gαq pathway and activate PLC. Note that most of the PLC is activated through the Gαq pathway when using the Gqi4myr system. However, it is likely that the endogenous βγ subunit, as mentioned above, may also activate PLC but to a lesser extent. XCL1 N-terminal variants were assayed for their ability to stimulate tritium (3H)–IP3 accumulation within the cell (Fig. 2B and table S1). XCL1 CC3 and WT had comparable IP3 accumulation (EC50 = 14.1 and 17.6 nM, respectively), likely indicating that the metamorphic behavior of WT has a negligible effect on XCR1 activation in the context of the IP3 accumulation assay. Consistent with radioligand-binding results, variant residues within the extreme N terminus of XCL1 (V1A, G2A, S3A, E4A, V5A, ∆Val1, and +GlyN-term) displayed large defects in IP3 accumulation with potency decreases of 2.8- to 90-fold. IP3 was below detectable levels in cells treated with G2A, and an EC50 could not be determined. E4A and ∆Val1 had the largest decrease in potency (~90-fold). With the exception of S6A, all remaining variants proximal to Cys11 (D7A, K8A, R9A, and T10A) displayed IP3 accumulation similar to CC3. It is likely that a correlation exists between receptor affinity of variants 1 to 5 and their ability to promote IP3 accumulation. Conversely, mutations of residues 7, 8, 9, and 10 had a negligible effect on XCR1 activation, likely indicating that Asp7, Lys8, Arg9, and Thr10 in the native XCL1 sequence play a minimal role in XCL1-mediated XCR1 signaling response. Similar to the binding assay, S6A displayed a marked increase in potency (threefold) for IP3 accumulation when compared to XCL1 CC3. It is possible that Ser6 may function to divide the extreme N-terminal residues (Val1-Val5) from the residues near the Cys11 (Asp7-Thr10) and dampen XCL1-mediated XCR1 signaling response (table S1).

Several XCL1 N-terminal variants have decreased potency for intracellular calcium flux

Accumulated IP3 triggers the release of calcium from the ER, causing a pulse in intracellular calcium release, which can be measured by a calcium flux assay (Fig. 1). Like IP3 accumulation, intracellular calcium flux is another second messenger and reporter of GPCR activation, allowing for measurement of ligand potency (39). Potency describes the concentration of XCL1 ligand needed to produce a calcium flux response and is quantified by EC50, whereas efficacy describes the maximal calcium flux response achievable by an XCL1 ligand.

XCR1-expressing human embryonic kidney (HEK) 293 cells were treated with various concentrations of XCL1 variants (Fig. 3A and table S1). XCL1 WT and S6A have mildly decreased potencies (1.4- and 2.1-fold, respectively) and shared efficacies of ~80 relative fluorescent units (RFUs) for calcium flux when compared to CC3. The V5A, R9A, and T10A variants also had similar decreased potencies (1.6- to 2.6-fold); however, these variants display decreased efficacies of about twofold (RFUs <60). All other N-terminal variants (V1A, G2A, S3A, E4A, D7A, K8A, ∆Val1, and +GlyN-term) displayed reduced potencies of 6- to 93-fold (EC50 > 268 nM), with E4A and ∆Val1 having the most reduced potencies of 93- and 69-fold, respectively. The reduction in calcium release by D7A, K8A, R9A, and T10A does not correlate with the observed lack of effect of these variants in the IP3 accumulation, indicating that responses to changes in the XCL1 agonist can vary between assay types (table S1).

Fig. 3 Calcium flux and chemotaxis response in XCR1-expressing cells treated with XCL1 N-terminal variants.

(A) HEK293 cells were transfected with human XCR1 and incubated with XCL1 N-terminal variants at various concentrations. Corresponding EC50 values are provided in table S1. The dotted curves represent calcium flux measurement for XCL1 CC3 and are added to all other XCL1 ligand plots as a point of reference. The data are the mean ± SEM (n ≥ 3 experiments). (B) Murine L1.2 cells were transfected with human XCR1 and incubated with various concentrations of XCL1 N-terminal variants in a transwell assay. The dotted box represents the maximal chemotaxis of XCL1 CC3 (1 nM) and is shown on each graph to highlight deviations in maximal chemotaxis relative to CC3. Data are mean ± SEM (n = 3 experiments). *P < 0.05 compared with CC3, bottom right, by using the Holm-Sidak test. Maximal chemotaxis concentrations for each variant can be found in table S2. (C) Correlation plot comparing log(EC50) for each variant with the percent of migrated cells at 1 nM. Residues shaded in purple indicate variants with largest defect; residues shaded in cyan have XCR1 activation properties similar to those of XCL1 CC3.

Alterations of Val1-Glu4 cause diminished cellular migration

The ability of a cell to respond and move toward a chemical stimulus is known as chemotaxis, and measurement of this phenomenon is a foundational tool for assessing chemokine function (40). Typically, cells exhibit a biphasic (bell-shaped) chemotactic response, dependent on chemokine concentration. This means that cellular migration increases with increasing chemokine concentration until a concentration threshold is reached. Upon reaching this threshold, the cell is unable to respond to the additional increases in chemokine concentration and migration is decreased (41).

L1.2 murine cells were transfected with human XCR1 and incubated in transwell plates with various concentrations of XCL1 variants (Fig. 3B and table S2). Maximal chemotaxis occurs at 1 nM for XCL1 CC3, S6A, D7A, K8A, R9A, and T10A. XCL1 WT stimulated maximal chemotaxis at 10 nM, which was similar to V5A. The XCL1 variants that showed the largest defects in chemotactic migration were V1A, G2A, S3A, E4A, ∆Val1, and +GlyN-term, all having ~1000-fold decrease in potency (maximal chemotaxis ≥1000 nM). Of these, G2A displayed the largest defect in chemotaxis stimulation (Fig. 3B). When correlating calcium flux EC50 for each variant with its maximal chemotaxis at 1 nM, variants S6A-T10A retained levels of XCR1 activation similar to XCL1 CC3, whereas extreme N-terminal variants had the largest disruption in XCR1 activation (Fig. 3C).

Modeling of XCR1-XCL1 N-terminal interactions

To better interpret the experimental data for XCR1 activation, we constructed a structural model of the site 2 interactions between the N terminus of XCL1 and the orthosteric pocket (i.e., the main ligand-binding pocket) of XCR1. For chemokine receptors, this site is composed of molecular contributions from TM regions III to VII (major binding pocket) and TM regions I to III and VII (minor binding pocket) (42). Because an experimentally determined structure of XCR1 has not yet been published, we began our efforts by generating a homology model of XCR1 using the comparative modeling protocol of Rosetta (RosettaCM) (43, 44). The benefit of this technique for creating a homology model of XCR1 is that it uses multiple template structures instead of a single homolog, leading to the generation of more accurate models than other currently available methods (44). The generated XCR1 model has a favorable energy score and was found to be stable over the course of a 300-ns all-atom MD simulation [average Cα root mean square deviation (RMSD) = 3.2 Å; fig. S1, A and B], indicating stable intramolecular interactions. Because we wanted to dock the N-terminal peptide of the chemokine into the orthosteric pocket, we chose three models of XCR1, which were selected by visual inspection from frames across the MD trajectory in which the orthosteric pocket was allowed to relax from a relatively closed conformation to become more accessible to a chemokine ligand. The N-terminal peptide of XCL1 (Val1-Gly2-Ser3-Glu4-Val5-Ser6-Asp7-Lys8-Arg9-Thr10) was docked using Rosetta’s FlexPepDock ab initio protocol (45). For each XCR1 starting pose, 100,000 models were generated in the first round of unbiased docking. Three percent to 5% of these models docked to the orthosteric pocket of XCR1, as chosen by a cutoff distance of <10 Å between Val1 and Trp2.60 (Ballesteros-Weinstein nomenclature, used here and throughout for numbering GPCR residues, is a scheme in which the first number denotes the TM helix, and the second number denotes relative position with respect to the most conserved residue in each helix, which is assigned number 50). As mutation of Glu4 to alanine resulted in a severe disruption of XCL1-XCR1 binding and activation in vitro (Fig. 4, A to C, and tables S1 and S2), the energy contribution of Glu4 was used to select models in which Glu4 was making a favorable interaction (meaning, a negative energy contribution). Models that adhered to the aforementioned distance and energy requirements were clustered, and a single representative model was selected for each XCR1 starting pose. These models were each used to seed the generation of an additional 100,000 docked models. Of these 100,000 redocked models, ~90% docked to the orthosteric pocket of XCR1. Models were again filtered and clustered as above, and a single structure was selected for further analysis. The final model selection was based on agreement of in vitro data (Fig. 4, A to C), with the results of a computational alanine scan performed in the Robetta online server (46, 47). During the computational alanine scan, each nonglycine residue of the N terminus of XCL1 was individually mutated to alanine, and the change in binding energy (ddG) was calculated (Fig. 4D and table S2). When compared to the experimental EC50 values obtained from radioligand displacement, calcium mobilization, and IP3 accumulation assays (Fig. 4, A to C, and table S1), the ddG values calculated by Robetta for the selected pose showed agreement at critical residues, including Glu4 (Fig. 4D). The ddG value for the G2A substitution was not calculated because the energy function used by the Robetta server requires alanine to be smaller than the residue it is replacing (46). Thus, our computational alanine scan could not predict the large loss of binding affinity seen experimentally with the G2A substitution. In addition, the computational alanine scan incorrectly predicted a large loss for the S6A variant (Fig. 4, A to D). These data contributed to a selected XCR1-XCL1N-terminal peptide model (Fig. 4E).

Fig. 4 Modeling of XCR1-XCL1 N-terminal interactions.

(A to C) Graphical representations of radioligand displacement (A), IP3 accumulation (B), and calcium release (C) data, shown as the fold change in EC50 value as compared to XCL1 CC3 for each N-terminal alanine variant (labeled according to residue number, 1 to 10). Color scheme is consistent throughout the figure. Data are means of more than four experiments for radioligand displacement, three experiments for IP3 release, and more than three experiments for calcium flux. (D) Changes in binding energies (ddG) upon mutation of each residue in the XCL1 N-terminal peptide to an alanine using the Robetta online server. ND, no data. (E) Docked XCR1-XCL1N-terminal Peptide structure chosen for further analysis. (F) Cα position of each residue of the N-terminal peptide of XCL1, colored according to residue number, over the course of a representative 1-μs MD simulation. Residues with low Cα RMSD distribution (more contacts with XCR1) are indicated by tightly grouped sets of spheres; residues with high Cα RMSD (less contacts with XCR1) have dispersed grouping of spheres. (G) Histogram representing the average Cα RMSD of each residue of the N-terminal peptide of XCL1 over the course of the representative 1-μs MD simulation as shown in (F). (H) Contact network map showing contacts between the top 100 docked poses between the XCL1 N-terminal peptide (residues 1 to 10, y axis) and XCR1 (residues numbered according to Ballesteros-Weinstein nomenclature, x axis) for each of the three XCR1 structures docked. The size of each graphed circle represents the number of poses with the contact. ECL, extracellular loop. Black squares beneath XCR1 residues indicate known contacts for other chemokine–chemokine receptor complexes. (I) Contacts between Val1 of the XCL1 N-terminal peptide (shown in purple) and XCR1 (shown in gray). XCR1 residues are identified with their Ballesteros-Weinstein nomenclature. Purple box indicates that the residue makes contact with Val1 of XCL1. (J) Contacts between Glu4 of the XCL1 N-terminal peptide (green) and XCR1 (gray). Residues making a direct contact are shown in green.

To further analyze the contact stability of the selected pose, we performed four 1-μs all-atom MD simulations on the XCR1-XCL1N-terminal peptide complex (fig. S2). When calculating RMSDs for our simulations, superpositioning was performed using the R package Bio3D, with the first frame of the MD simulation used as the reference coordinate set and the coordinates from the rest of the MD simulation used as the mobile dataset (a single representative simulation is shown in Fig. 4 for clarity). Over the course of the four 1-μs simulations, the peptide remained in the orthosteric pocket of the receptor, with an average Cα RMSD of <5 Å for the first four residues of the peptide (Fig. 4, F and G, and fig. S2A). Whereas the first five residues appear to make substantial contacts that enable low RMSD of the backbone over the course of the simulation, the last five residues make substantially fewer contacts and have a much larger Cα RMSD distribution (Fig. 4, F and G, and fig. S2A).

To gain a more global sense of the molecular contacts made in our modeled XCR1-XCL1N-terminal peptide complexes, we analyzed the top scoring 100 docked models from each XCR1 starting pose for their molecular contacts (Fig. 4H). Despite that each of the three initial XCR1 poses was randomly selected from the MD simulation trajectory, the complexes generated from all three starting poses share contacts with XCR1 residues known to make contacts in other chemokine–chemokine receptor complexes, including contacts with Trp2.60 and Asn3.29 (Fig. 4H).

Returning to our single selected model, visual inspection of the docked pose was performed with a focus on residues that caused large changes in the signaling ability of XCL1 in vitro. This revealed key XCR1-Val1 and XCR1-Glu4 intermolecular contacts. Specifically, important contacts were seen between the side-chain Val1 and the side chains of Tyr45.52, Trp5.35, and Phe3.32, as well as potential hydrogen bond interactions between the N terminus of Val1 and His4.64 and Asn3.29 (Fig. 4I), implicating these residues as key mediators of the XCL1-XCR1 interaction. Visual inspection also revealed three important contacts for Glu4. A potential hydrogen bond interaction between Glu4 and Arg6.62 and Tyr45.52 (Fig. 4J) provides a possible explanation for the large binding and activation defect seen experimentally for the XCL1 E4A variant (Fig. 4, A to C). Although not within a hydrogen bonding distance of Arg7.39 in the selected model, Glu4 is within 5 Å of Arg7.39, and thus, we identified Arg7.39 as a third residue that may also be important for Glu4-mediated XCL1-XCR1 interactions. Furthermore, position 7.39 is highly conserved among chemokine receptors as a negatively charged glutamate (48), suggesting the positively charged Arg7.39 may be mediating the interaction specific between XCR1 and XCL1.

In the context of our experimental data, we find agreement between the in silico and in vitro results in that both indicate the importance of the Val1, Gly2, Ser3, and Glu4 in the XCL1-XCR1 site 2 interaction. In addition, the experimental and computational data suggest that Ser6, Asp7, Lys8, Arg9, and Thr10 act as a linker between the core of XCL1 and the site 2 interaction, providing the length necessary for XCL1 to fully bind to and activate XCR1.

MD simulations predict key interactions of the N terminus of XCL1 with Asp4.60 and Asn3.29

Because truncation (ΔVal1) or extension (+GlyN-term) of the N terminus of XCL1 drastically alters XCL1 binding to and activation of XCR1 (table S1), we further investigated which XCR1 residues were making important contacts with Val1 by analyzing the relative position of Val1 over the course of four 1-μs MD simulations. We analyzed the interactions made between the N terminus of the XCL1N-terminal peptide and XCR1 over the course of four 1-μs all-atom MD simulations (Fig. 5 and fig. S2B). The MD simulations allowed for capture of additional binding contacts that were not previously identified in our initial Rosetta-generated XCR1-XCL1N-terminal peptide complexes due in part to differences in the Rosetta energy function and the Chemistry at Harvard Macromolecular Mechanics (CHARMM) force field. We found that the N terminus of the XCL1N-terminal peptide primarily interacts with three residues: Asn3.29, His4.64, and Asp4.60 (Fig. 5A). In the selected Rosetta model, the N terminus of the XCL1N-terminal peptide is making hydrogen bond contacts with His4.64 and Asn3.29, but the MD simulations revealed an additional possible interaction as the N terminus of the XCL1N-terminal peptide moved within hydrogen bonding distance of Asn3.29 and Asp4.60 (Fig. 5, B to D, and fig. S2B). Although positions 3.29 and 4.60 have been identified as contacts in previously crystalized ligand–chemokine receptor complexes (49), neither position is highly conserved, though a negatively charged residue is common at position 4.60 (Fig. 5E). Because it is already known that the distal N termini of chemokines commonly interact with the orthosteric pockets of their respective chemokine receptors (49), we did not focus our efforts on in vitro validation of XCL1 Val1 receptor contacts, but rather on XCL1 Glu4 receptor contacts, which we expect to be more important for enhancing XCL1-XCR1 specificity of binding.

Fig. 5 MD simulation predicts key interactions between the N terminus of XCL1 and Asn3.29 and Asp4.60.

(A) Positions of the XCL1 N-terminal peptide N terminus (N Val1; nitrogen represented as spheres, purple) and its potential interacting partners His4.64, Asn3.29, and Asp4.60 over the course of a representative 1-μs MD simulation. (B) The N terminus primarily interacts with Asn3.29 and Asp4.60. (C) Distance between the nitrogen atom of the N terminus of the XCL1 N-terminal peptide (N Val1) and Asn3.29 (OD1, orange), His4.64 (NE2, blue), and Asp4.60 (OD2, green) over the course of the representative 1-μs MD simulation. Distance in angstroms. (D) Average frequency of interactions between the N terminus of the XCL1 N-terminal peptide and either Asn3.29 (OD1, orange), His4.64 (NE2, blue), and Asp4.60 (OD2, green) over the course of the representative 1-μs MD simulation. Distance in angstroms. (E) Amino acids at positions 3.29 and 4.60 in other chemokine receptors.

Modeling dynamics simulation predicts key interaction of XCL1 Glu4 with Arg6.62, Arg7.39, and Tyr45.52 of XCR1

Because Glu4 of XCL1 was shown to be critical for binding and activation of XCR1 by 125I-displacement, IP3 accumulation, calcium flux, and chemotaxis assays, we sought to determine the XCR1 contacts that might be mediating this critical interaction. Analysis of the intermolecular contacts (i.e., atom-atom distances <3 Å) between Glu4 and XCR1 in the top 100 models for each XCR1 pose (Fig. 4H) revealed that Glu4 may interact with residues in TM domains VI and VII as well as residues within ECL2 (extracellular loop 2). However, the most frequent Glu4 contact observed in our modeling was with Arg6.62 (Fig. 4H). Thus, further investigation of Glu4-XCR1 interactions was performed through analysis of the relative position of the Glu4 side chain over the course of the aforementioned four 1-μs MD simulations. During the simulations, Glu4 interacts with Arg6.62, Arg7.39, and Tyr45.52 (Fig. 6, A to C, and fig. S3, A to C). Glu4-Arg7.39 is particularly interesting, given the importance of position 7.39 for chemokine receptors (50). In nearly all other chemokine receptors, residue position 7.39 is a glutamate and is thought to make important contacts with the chemokine ligand (Fig. 6D) (51). In the three currently published crystal structures of chemokine-receptor complexes, the chemokine N terminus makes important contacts with residue 7.39 (5254). Of all chemokine receptors, only XCR1 and CCRL2 have positively charged residues at position 7.39. Given the significance of Glu4 in the N terminus of XCL1 and the potential for a Glu4-Arg7.39 interaction shown via MD simulations (Fig. 6, E and F, and fig. S2C), we propose that transient interactions between Glu4 (XCL1) and Arg7.39 (XCR1) are important for binding and stabilization of the N terminus of XCL1 in the orthosteric pocket of XCR1. In addition, the MD simulation further suggested a possible interaction between Glu4 in the N terminus of XCL1 and Tyr45.52 in ECL2 of XCR1 (Fig. 6, E and F, and figs. S2C and S3, A to C). However, this position is not highly conserved among chemokine receptors (fig. S3D).

Fig. 6 Modeling and in vitro validation of key interactions between glutamate 4 of XCL1 and two arginine residues and a tyrosine residue of XCR1.

(A) Positions of Glu4 of the XCL1 N-terminal peptide and XCR1 residues Arg6.62, Arg7.39, and Tyr45.52 (Ballesteros-Weinstein nomenclature) over the course of a representative 1-μs MD simulation. Cα represented as spheres. (B) and (C) Glu4 of the XCL1 N-terminal peptide interacts not only with Arg6.62 primarily (B) but also with Arg7.39 briefly (C). (D) Amino acids at positions 6.62 and 7.39 in other chemokine receptors. Red, negatively charged; blue, positively charged; black, not charged. (E) Distances between Glu4 (CD) and Arg6.62 (CZ, dark blue), Arg7.39 (CZ, light blue), and Y45.52 (CZ, green) over the course of a representative 1-μs MD simulation. Distance in angstroms. (F) Average frequency of interaction between Glu4 and Arg6.62 (dark blue), Arg7.39 (light blue), or Y45.52 (green) over the course of the representative 1-μs MD simulation shown in (E). Distance in angstroms. (G to L) 125I-XCL1 CC3 displacement and 3H-IP3 accumulation by XCR1 variants. XCR1 variants were designed based on computational modeling results and tested for binding (G to I) and signaling (J to L), respectively, in response to CC3. WT XCR1 (G, J) shown for reference. Data are mean ± SD from three experiments.

Alteration of XCR1 residues Arg7.39 and Tyr45.52 result in diminished XCL1 binding and accumulation of IP3

To test the importance of XCR1 residues Arg6.62, Arg7.39, and Tyr45.52 for XCL1 binding and signaling, we mutated each of these residues to alanine and performed binding and signaling assays. The ability of XCL1 CC3 to displace 125I-XCL1 CC3 from XCR1 variants was measured using radioligand displacement assays. XCR1 WT was used as a positive control. These assays were performed as described previously. As compared with XCR1 WT (Fig. 6G), XCR1 R7.39A and XCR1 Y45.52A showed major defects in their ability to bind XCL1 CC3 (Fig. 6, H and I). XCR1 R6.62A was able to bind CC3 with an affinity similar to that of WT XCR1 (fig. S4). This demonstrates that XCR1 residues Arg7.39 and Tyr45.52 are critical in facilitating XCR1 binding to XCL1. IP3 accumulation in response to treatment of cells expressing XCR1 variants with XCL1 CC3 was measured. XCR1 R7.39A and XCR1 Y45.52A, but not XCR1 R6.62A, display large defects in IP3 accumulation (Fig. 6, J to L, and fig. S4), consistent with radioligand-binding assays. In all, these results support the conclusion drawn from our computational modeling that binding and signaling at the XCL1-XCR1 axis critically depends on XCR1 residues Arg7.39 and Tyr45.52. Broadly, this demonstrates the utility of a hybrid approach including both in vitro and in silico components to better understand ligand-receptor interfaces.


Mutagenic studies of chemokines including CCL5 (55), CXCL8 (56), and CXCL12 (57) have shown that amino acid composition in the N terminus is important for receptor binding and activation. With the exception of some CXC chemokines that harbor an N-terminal Glu-Leu-Arg (ELR) motif (58), there is little consensus among chemokine N termini. Because chemokines orchestrate a large variety of homeostatic and pathologic functions, knowledge of the structural features that encode chemokine-receptor specificity is important for understanding immune function and developing new therapeutic strategies. In this study, we combined structure-function studies of XCL1 with computational modeling to identify key residues within the chemokine N terminus that mediate binding and activation of XCR1. A panel of N-terminal XCL1 variants (V1A, G2A, S3A, E4A, V5A, S6A, D7A, K8A, R9A, and T10A) including an N-terminal truncation (∆Val1) and extension (+GlyN-term) was analyzed using cell-based assays (radioligand displacement, IP3 accumulation, calcium flux, and chemotaxis) and computational analysis (modeling, docking, and MD simulations). The results define several key interactions between the N terminus of XCL1 and XCR1.

N-terminal substitutions (Val1, Gly2, Ser3, and Glu4), as well as N-terminal modifications (∆Val1 and +GlyN-term), introduce substantial defects in XCR1 binding, suggesting that both the length and amino acid composition of the XCL1 N terminus are crucial. These results are supported by computational modeling analysis. Docking and MD simulations show that alterations of Val1 through mutation (V1A), deletion (∆Val1), or extension (+GN-term) likely prohibit interactions between the XCL1 N terminus and residues lining the bottom of the orthosteric pocket of XCR1. These XCR1 residues, found within the TM helices, include Trp2.60 (TM2), His4.64 (TM4), and Trp5.35 (TM5) and most likely serve as an anchoring point for the N terminus of XCL1 in the orthosteric pocket. Removal or alternation of Val1 prohibits anchoring and diminishes XCL1-XCR1 binding. This finding is consistent with previous studies that have shown that changes or truncations of the extreme N terminus of chemokines result in altered receptor binding and activation (19, 59).

Two additional residues in the N terminus of XCL1, Gly2 and Glu4, mediate specific contacts with residues of XCR1. The G2A and E4A mutations lead to >100-fold decrease in XCR1 binding affinity when compared to XCL1 CC3 in radioligand displacement assays. E4A displayed the largest change in binding energies (ddG > 1) according to docking analysis, whereas the ddG calculation was not able to predict the large loss of binding affinity caused by the G2A variant. MD simulations further support contributions of Glu4 in XCR1 binding, illustrating a primary interaction with Arg6.62 and Tyr45.52 and a secondary interaction with Arg7.39. Inspection of chemokine receptor homologs at the 6.62, 7.39, and 45.52 positions revealed that most of the chemokine receptors have a Glu7.39 (48), whereas there is no consensus for 6.62 or 45.52. Previous studies have shown that position 7.39 in other chemokine receptors makes important contacts with chemokine N termini, illustrating the importance of this position in chemokine-receptor interactions (49, 5254). Previous works have also shown that in several GPCRs, both in the chemokine family and in other receptor families, position 45.52 makes key contacts with ligands (60). For example, receptor position 45.52 has been shown to make key contacts in the chemokine receptor–ligand complexes US28-CXC3CL1, US28–vMIP-II, and CXCR4-CVX15 (49). In addition, position 45.52 in rhodopsin interacts with rhodopsin ligand retinal (61), and position 45.52 in the 5-HT2A receptor interacts with the antipsychotics zotepine and pimavanserin (62). In all, this demonstrates the importance of position 45.52 in GPCR-ligand interactions including but not limited to interaction between chemokine receptors and their ligands. The Glu4 carboxylate of XCL1 likely interacts with Arg6.62, Arg7.39, and Tyr45.52 due to their close proximity in the XCR1 orthosteric pocket. Upon mutation of Glu4 to alanine, binding energy analysis reveals a large variation in ddG (>1). Ser3 and Val5 may also contribute to XCR1 binding according to our analysis, but their impact is minimal in comparison to Gly2 and Glu4.

As expected, N-terminal variants that displayed defective XCR1 binding (Val1, Gly2, Ser3, Glu4, Val5, ∆Val1, and +GN-term) also exhibit downstream defects in receptor activation as measured through IP3 accumulation, calcium flux, and cell migration assays. However, not all variants displayed defects in XCL1-XCR1 binding, such as those closest to Cys11 and distal from the extreme N terminus: S6A, D7A, K8A, R9A, and T10A. According to 125I-XCL1-CC3 displacement assays, these variants bound to XCR1 with affinities equal to or greater than XCL1 CC3 (EC50 ≤ 1.0 nM). Both 125I-XCL1-CC3 displacement assays and the MD simulation demonstrate that Ser6, Asp7, Lys8, Arg9, and Thr10 residues do not contribute to XCL1-XCR1 affinity. Specifically, the MD simulation illustrates that these residues are more dynamic, displaying more flexibility than the extreme N-terminal residues, suggesting that they serve as a link to the folded XCL1 chemokine domain.

Mutation of Val5, Ser6, Asp7 Lys8, Arg9, and Thr10 led to inconsistent effects on XCL1-mediated XCR1 responses in vitro (minimal effects on IP3 accumulation and cellular migration and larger effects in calcium response). The MD simulation demonstrated enhanced flexibility of residues Ser6-Thr10 within the orthosteric binding pocket of XCR1. It is possible that dynamics within this region of the N terminus may influence the ability of residues Ser6-Thr10 to activate XCR1. These computational and biological findings warrant further examination.

Note that no single mutation displayed antagonistic properties, meaning displayed tight binding without activation. This is consistent with the findings in a recent study by Kroczek et al. (63), showing that deletion of the seven N-terminal amino acids of murine diminished XCR1 binding by about 50-fold and led to reduced chemotactic activity. These results indicate that multiple residues contribute to both receptor binding and activation. Previous studies have demonstrated that site 1 interactions, between the chemokine and the N terminus of the cognate receptor, are important for initial chemokine receptor binding (42). These contacts may also be major contributors to XCL1-XCR1 binding, and contributions from both sites 1 and 2 are needed for high-affinity binding. Studies targeting the site 1 interaction between XCL1-XCR1 were not examined in this article. In addition to XCL1, humans and several other species have a closely related paralog XCL2. XCL2 varies form XCL1 by two N-terminal amino acid changes (D7H and K8R). We have previously examined the ability of XCL2 to signal through XCR1, finding that XCL1 and XCL2 displayed similar activation profiles for calcium release and cellular migration (64). The structure-function analysis of the XCL1 N terminus presented here is consistent with our previous findings that XCR1 activation is not particularly sensitive to the amino acids at positions 7 and 8.

In the absence of a solved structure of XCL1-XCR1, we have developed an in silico and in vitro approach to study both the functional and structural effects of XCL1-XCR1 site 2 binding. Homology modeling and MD simulations complemented the functional data and allowed us to visualize important contacts for XCR1 binding and activation including Glu4 interactions with Arg6.62, Arg7.39, and Tyr45.52 in XCR1. Because a sequence comparison across the chemokine family revealed that most chemokine receptors have a glutamate at position 7.39, and multiple studies have identified position 7.39 as an important mediator of ligand-receptor interactions (48, 49, 5254), we have identified Arg7.39 as being a unique residue that may be playing an important role in mediating XCL1-XCR1 interaction specificity. To probe the importance of XCR1 residues Arg6.62, Arg7.39, and Tyr45.52, we performed binding and signaling assays with XCR1 variants that altered these residues. XCR1 R7.39A and XCR1 Y45.52A showed substantially diminished binding to XCL1 CC3. Cells expressing these XCR1 variants also showed a major decrease in IP3 accumulation in response to treatment with XCL1 CC3. XCR1 R6.62A exhibited binding and signaling capacity similar to that of WT XCR1. These data demonstrate that XCR1 residues Arg7.39 and Tyr45.52 are critical for XCL1-XCR1 binding and signaling. Moreover, these results support the utility of the computational modeling pipeline developed here, inspiring confidence in future application of these techniques to other chemokine receptor pairs.

Additional studies measuring GPCR kinase (GRK) phosphorylation, β-arrestin recruitment, and receptor internalization are needed to better understand XCL1-XCR1–mediated signal transduction that drives dendritic cell chemotaxis. Here, we have described a hybrid approach that combines experimental and in silico methods to characterize the chemokine-receptor interface of XCL1 and its GPCR XCR1. We propose that this strategy can be readily adapted for other members of the chemokine family to better interpret existing data or guide future functional studies.


Mutagenesis of XCL1 and XCR1 and purification of recombinant XCL1

The QuikChange Site-Directed Mutagenesis Kit (Agilent) was used to complete alanine-scanning mutagenesis of the N-terminal residues (1 to 10) of the locked-monomer conformation of XCL1 (CC3) that was previously described by Tuinstra et al. (28). Recombinant protein expression and purification was carried out as previously described (64). The molecular weights of all XCL1 purified proteins were verified by matrix-assisted laser desorption/ionization–time-of-flight (MALDI-TOF) mass spectrometry. XCR1 variants were commercially provided by GenScript in the pcDNA3.1(+) vector, using the company’s gene synthesis and mutagenesis services.

Molecular biology

The following methods were based on previously published methods (65). XCR1 complementary DNA (cDNA) was cloned into the pcDNA3.1(+) vector (Invitrogen) using sticky end ligation. XCR1 was amplified from this vector using polymerase chain reaction (PCR) with end primers without a stop codon before insertion into the pCMV-ProLink 1 (PK1) Vector (DiscoveRx, Birmingham, UK) directly upstream of the PK1 tag needed for the β-arrestin2 recruitment assays. The product was transformed into XL1-Blue cells for monoclonal vector selection, and a purified vector product was sequenced before use.

In vitro mammalian cell culture

As previously described (65), COS-7 cells were grown in Dulbecco’s modified Eagle’s medium 1885 [10% fetal bovine serum, 2 mM glutamine, penicillin (180 U/ml), and streptomycin (45 μg/ml)] at 37°C and 10% CO2. Cells were transiently transfected with the pcDNA3.1(+) XCR1 vector using the calcium phosphate precipitation method and incubated for ~40 hours before performing assays.

Radioligand competition binding assay

Assay was performed as previously described (65). An XCL1 tracer was produced by iodine labeling of tyrosine residues of the XCL1 protein by using an oxidative iodination procedure [using chloramine-T to incorporate a 125I isotope (PerkinElmer)], and the product was purified and verified by reversed-phase high-performance liquid chromatography (HPLC) (66). COS-7 cells transfected with XCR1 were seeded in 96-well plates at 35,000 cells per well (in duplicates) for growth 1 day before the assay. On the day of the assay, the cells were washed and changed to a 50 mM Hepes buffer supplemented with bovine serum albumin (BSA) (5 g/liter) and chilled to 5°C. Ligands were added shortly before the labeled XCL1 CC3 tracer (calibrated to result in ≈10% tracer binding), and the cells were incubated at 4°C for 3 hours before being washed in a 50 mM Hepes buffer containing BSA (5 g/liter) and NaCl (29.22 g/liter). The cells were lysed, and γ radiation of the lysate was measured.

IP3 accumulation assay

Assay was performed as previously described (65). COS-7 cells were transfected with human XCR1 and Gqi4myr, a large G protein chimera with the Gαi recognition interface, and a Gαq output. Cells were seeded into 96-well plates at 35,000 cells per well (in duplicate) 1 day before the assay and incubated for growth with [3H]myo-inositol (5 μl/ml, 2 μCi/ml) overnight. On the day of the assay, cells were washed in Hanks’ balanced salt solution (HBSS) buffer and changed to a solution of 10 mM LiCl in HBSS before ligands were added. After incubating for 90 min at 37°C, the cells were lysed in 10 mM formic acid, and 35-μl lysis solution (≈90%, v/v) was transferred to white and opaque 96-well plates and mixed with a solution of agitated SPA-Ysi beads (80 μl per well, 12.5 mg/ml, PerkinElmer). The plates were shaken for 30 min before being left for an 8-hour equilibration period, and last, scintillation was measured using a Packard TopCount NXT counter (PerkinElmer). Assays were conducted in triplicate.

Calcium flux assay

Calcium flux was measured in stable transfected XCR1-expressing HEK293 cells, provided by J. Hedrick (Schering-Plough Research Institute) (67), using a previously described method (64). Briefly, cell cultures were grown to ~90% confluency, lifted from the culture plate, washed twice in warmed phosphate-buffered saline (PBS), and suspended in assay buffer [1× HBSS, 20 mM Hepes (pH 7.4), and 0.1% BSA]. Cells were plated at 2.0 × 105 cells per well in 100 μl in a 96-well plate and incubated with 100 μl of FLIPR Calcium 4 Assay Dye (Molecular Devices) for 1 hour at 37°C and 5% CO2. A FlexStation 3 (Molecular Devices) was used to treat cells with various concentrations of purified proteins and monitor calcium flux. All experiments were conducted in triplicate.

Chemotaxis assay

Chemotaxis assays were carried out as previously described (64) using murine L1.2 cells stably expressing human XCR1 (68). In brief, 2.0 × 105 cells were suspended in 25 μl of assay media [phenol red–free RPMI 1640, 0.5% BSA, and 10 mM Hepes (pH 7.4)] and added to the upper chambers of a transwell assay plate. XCL1 proteins were diluted to various concentrations in 30 μl of the same medium and added to the lower chambers. The plates were incubated at 37°C and 5% CO2 for 1.5 hours. After incubation, cells that migrated into the lower wells were lysed with 0.1% Triton X-100 and measured by PicoGreen doubled-stranded DNA quantitation reagent (Molecular Probes). Assays were conducted in duplicate, and the numbers of cells in the lower wells were expressed as a percentage of input cells. Results are shown as mean ± SEM for three separate experiments.

Homology modeling

Homology modeling of XCR1 was performed using the RosettaCM protocol (43, 44). The human XCR1 sequence was obtained from UniProt (69) (UniProtKB accession number P46094) and subsequently edited to remove both the N-terminal (1-MESSGNPESTTFFYYDLQS-19) and C-terminal domains (304-QFWFCRLQAPSPASIPHSPGAFAYEGASFY-333). Seven template structures were chosen on the basis of availability of structural information and sequence similarity to human XCR1: human CCR5 [Protein Data Bank (PDB) ID: 4MBS], human CXCR4 (PDB ID: 4RWS), human CCR2 (PDB ID: 5T1A), human CCR9 (PDB ID: 5LWE), human angiotensin II receptor type I (AT1R; PDB ID: 4YAY), human δ-opioid receptor (δOR; PDB ID: 4RWA), and murine μ-opioid receptor (μOR; PDB ID: 4DKL). The template structure PDBs were edited to remove any crystallographic inserts (e.g., T4 lysozyme and cytochrome b) and to remove the N- and C-terminal domains. The edited template PDBs were cleaned to remove extraneous information using the Rosetta script Sequences were aligned to the edited XCR1 sequence using Clustal Omega (70), and the sequence alignment was manually adjusted to eliminate gaps in TM regions and to align the conserved cysteine residue in ECL2. XCR1 fragments were generated using the Robetta server (71), and the TM topology of XCR1 was predicted using OCTOPUS (72). The aligned sequence of XCR1 was threaded onto each of the seven template structures, and the resultant threaded models were used to generate 5000 hybridized models using the hybridize mover of Rosetta scripts (43, 44). Each of the 5000 generated models was subjected to two all-atom sampling relaxation runs using Rosetta’s relax application (73). The top 20 scoring models were visually inspected, and one model was selected for further analysis. To examine the stability of the selected models, we performed a 300-ns all-atom MD simulation via Desmond (74). The Cα RMSD of the final selected model plateaued under ~3.5 Å, supporting a stable fold and feasible intramolecular interactions (fig. S1).

Peptide docking

The N-terminal peptide of XCL1 (residues 1 to 10; N-[Val-Gly-Ser-Glu-Val-Ser-Asp-Lys-Arg-Thr]-C) was docked into the homology model of XCR1 generated above using Rosetta’s FlexPepDock ab initio protocol (45, 75, 76), similar to what has been previously published (77). Because the orthosteric pocket of the Rosetta-generated homology model of XCR1 was largely inaccessible to ligand, three unique “open-pocket” poses of XCR1 were randomly selected by visual inspection for docking from the 300-ns MD simulation trajectory (i.e., three frames from the 1000-frame trajectory; specifically frame 99, frame 475, and frame 957). The peptide was manually built in PyMOL and placed away from the receptor in an extended conformation. A total of 100,000 docked models were generated with FlexPepDock ab initio and subsequently filtered according to their relative position to the orthosteric pocket (i.e., distance from CH2 of Trp2.60 to CA of Val1 of XCL1 < 10 Å) and the energy score contributed by residue Glu4 of the N-terminal peptide from XCL1 (i.e., negative energy contribution). The resulting models were clustered using Calibur (78), and a representative model from the largest cluster was selected for further refinement. The selected model for each initial XCR1 pose (three models in total) was used to seed the generation of an additional 100,000 models with FlexPepDock ab initio. The 100,000 newly generated models were subjected to the filtering procedure described above, and top scoring 100 models were selected for visual inspection. The top 100 scoring poses were also analyzed for their molecular contacts using an in-house R script, using the cmap function in Bio3D (79) and a 3-Å cutoff to define intermolecular distances. Last, 10 models were chosen by a visual inspection process comprising the selection of models in which the XCL1 N-terminal peptide sampled a variety of conformations within the orthosteric pocket. The chosen 10 models were subjected to a computational alanine scan (47) using the Robetta Computational Interface Alanine Scanning Server (46, 47), and the in silico results were compared to the experimental alanine scan results presented within to select a single final model.

MD simulations

Selected XCR1-XCL1N-terminal peptide models were prepared for simulation by first positioning the membrane using the PPM Server of the Orientations of Proteins in Membranes (OPM) database (80) and subsequently using the Protein Preparation Wizard of Maestro (Schrödinger) to cap the N and C termini of the receptor and to perform H-bond optimization and protein minimization of the complex (81). Histidine residues were simulated in the neutral state (Nε tautomer, HIE), and all glutamate and aspartate residues were simulated in the charged state, after several recent studies of GPCR systems (8284). The System Builder in Maestro (Schrödinger) was then used to insert the prepared system into an equilibrated POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) bilayer according to the PPM positioning described above. Sodium and chloride ions were added to neutralize the system and reach a final NaCl concentration of 150 mM. The viparr and build_constraints utilities of Desmond were used to adjust the force field parameters to use the CHARMM 36 force field and the TIP3P water model (85). The system was equilibrated at 310 K using the utility of Desmond. Four 1-μs MD simulations were subsequently performed with randomized starting velocities, each with a 300-ps recording interval and a 2-fs time step, at 310 K and 1 bar in the isothermal-isobaric (NTP) ensemble using a Nose-Hoover thermostat and a Martyna-Tobias-Klein barostat with a 2.0-ps relaxation time. The resulting trajectory was analyzed using Visual Molecular Dynamics (VMD) (86) and the R package Bio3D (79). Note that in Figs. 4 to 6 and fig. S3, a single representative simulation is shown for clarity. The representative simulation corresponds to “simulation 2” as represented in fig. S2.

Nonlinear regression and statistical analysis

All data are expressed as mean ± SEM. Nonlinear regression was performed in Prism version 7.0c and version 8 (GraphPad Software) for 125I-XCL1 CC3 radioligand displacement, IP3 accumulation, and calcium flux assays. A sigmoidal dose-response (variable slope) equation with a least squares regression was used to fit all the data. Values for the bottom and upper asymptote, LogEC50, and Hill slope were all fit, except when indicated as constrained. Values were constrained under certain instances to fit the available data points. To account for statistical differences in the data, we used one-way analysis of variance (ANOVA) testing to compare the mean LogEC50 and SEM for each XCL1 construct. Multiple comparisons were made using Dunnett’s post hoc testing. Mean chemotaxis data for each XCL1 construct were statistically analyzed using multiple t tests, and multiple comparisons were made using the Holm-Sidak method. A P value of ≤0.05 is considered statistically significant.


Fig. S1. Homology modeling of XCR1.

Fig. S2. One-microsecond MD simulations of the XCR1-XCL1N-terminal peptide complex.

Fig. S3. MD simulations reveal Glu4 interactions with Tyr45.52.

Fig. S4. 125I-XCL1 CC3 displacement and 3H-IP3 accumulation by XCR1 R6.62A.

Table S1. Summary of quantitative values from nonlinear regression for each XCL1 N-terminal variant.

Table S2. Summary of quantitative values for cellular migration and docking analysis.


Acknowledgments: We thank M. Sigvardt Baggesen (University of Copenhagen) for assistance with XCR1 binding and IP3 accumulation assays, J. Hedrick (Schering-Plough Research Institute) for providing XCR1-expressing HEK293 cells, and A. G. Getschman (Medical College of Wisconsin) for fruitful discussion and assistance with manuscript editing. Funding: This work was supported, in part, by NIH grants R01 AI058072 (to B.F.V.), F30 HL134253 (to M.A.T.), and F30 CA236182 (to A.F.D.). M.A.T. and A.F.D. are members of the NIH-supported (T32 GM080202) Medical Scientist Training Program at the Medical College of Wisconsin (MCW). Author contributions: J.C.F. and B.F.V. conceived and planned experiments. J.C.F. produced and purified proteins, performed cell-based assays, analyzed data, and wrote the manuscript. M.A.T. performed the computational modeling, analyzed modeling data, and assisted in manuscript writing and revision. O.L. performed cell-based assays, analyzed results, and assisted in manuscript revision. T.N. performed cell-based assays, analyzed results, and assisted in manuscript revision. A.F.D. produced and purified protein, assisted M.A.T. with computational modeling, and assisted with preparation of manuscript figures and manuscript revision. O.Y., M.M.R., and B.F.V. provided financial assistance and support for all experiments represented in this manuscript and assisted in manuscript revision. Competing interests: B.F.V. has ownership interests in Protein Foundry, LLC. All other authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article