molecules

Article Computational Study of C-X-C Chemokine Receptor (CXCR)3 Binding with Its Natural Agonists Chemokine (C-X-C Motif) Ligand (CXCL)9, 10 and 11 and with Synthetic Antagonists: Insights of Receptor Activation towards Drug Design for Vitiligo

Giovanny Aguilera-Durán 1,2 and Antonio Romo-Mancillas 2,*

1 Posgrado en Ciencias Químico Biológicas, Facultad de Química, Universidad Autónoma de Querétaro, Cerro de las Campanas S/N, Querétaro 76010, Mexico; [email protected] 2 Laboratorio de Diseño Asistido por Computadora y Síntesis de Fármacos, Facultad de Química, Universidad Autónoma de Querétaro, Centro Universitario, Querétaro 76010, Mexico * Correspondence: [email protected]



Received: 31 August 2020; Accepted: 23 September 2020; Published: 25 September 2020


Abstract: Vitiligo is a hypopigmentary skin pathology resulting from the death of melanocytes due to the activity of CD8+ cytotoxic lymphocytes and overexpression of chemokines. These include CXCL9, CXCL10, and CXCL11 and its receptor CXCR3, both in peripheral cells of the immune system and in the skin of patients diagnosed with vitiligo. The three-dimensional structure of CXCR3 and CXCL9 has not been reported experimentally; thus, homology modeling and molecular dynamics could be useful for the study of this chemotaxis-promoter axis. In this work, a homology model of CXCR3 and CXCL9 and the structure of the CXCR3/Gαi/0βγ complex with post-translational modifications of CXCR3 are reported for the study of the interaction of chemokines with CXCR3 through all-atom (AA-MD) and coarse-grained molecular dynamics (CG-MD) simulations. AA-MD and CG-MD simulations showed the first activation step of the CXCR3 receptor with all chemokines and the second activation step in the CXCR3-CXCL10 complex through a decrease in the distance between the chemokine and the transmembrane region of CXCR3 and the separation of the βγ complex from the α subunit in the G-protein. Additionally, a general protein–ligand interaction model was calculated, based on known antagonists binding to CXCR3. These results contribute to understanding the activation mechanism of CXCR3 and the design of new molecules that inhibit chemokine binding or antagonize the receptor, provoking a decrease of chemotaxis caused by the CXCR3/chemokines axis.

Keywords: CXCR3; CXCL9; CXCL10; CXCL11; protein homology modeling; molecular dynamics

1. Introduction Vitiligo is a hypopigmentary skin autoimmune pathology characterized by the development of lesions lacking melanin, which is the natural pigment of the skin produced by melanocytes. Vitiligo affects 0.5–2% of the global population, with the highest incidence rates occurring in India, Japan, and Mexico [1–3]. This pathology can be classified as either nonsegmental, due to the symmetrical arrangement of the lesions, or segmental, due to the nonsymmetrical disposition of the lesions [4]. Vitiligo is not a fatal pathology; however, the patient’s quality of life is drastically diminished due to social reasons such as discrimination, social isolation, and depression [5–9]. Current treatments for vitiligo are focused on the promotion of repigmentation through the stimulation of melanogenesis and decreased immune system activity and oxidative stress. Unfortunately, these treatments are

Molecules 2020, 25, 4413; doi:10.3390/molecules25194413 www.mdpi.com/journal/molecules Molecules 2020, 25, 4413 2 of 30 not totally effective and have limitations, such as the time of treatment (phototherapy), risk of side effects (corticosteroids), and high cost (topical calcineurin inhibitors such as pimecrolimus and tacrolimus) [10,11]. The etiology of vitiligo is not completely understood; nevertheless, different theories could explain the development and progression of this pathology. Amongst these, oxidative stress and the involvement of the immune system are positioned as the most accepted theories, in addition to genetic factors that predispose individuals to vitiligo [3,12,13]. An increase in reactive oxygen species that cause DNA damage, the release of exosomes with intracellular peptides and misfolded proteins, and apoptosis have been reported. Thus, released peptides and misfolded proteins can be recognized by the immune system and stimulate the development of autoreactive lymphocytes responsive to melanocyte epitopes [3,14,15]. Moreover, CD8+ T lymphocytes were identified as being responsible for the progress and intensity of vitiligo; once melanocyte antigens are recognized, these lymphocytes release granzyme and perforin, promoting apoptosis in melanocyte cells. Additionally, the proinflammatory process triggers structural damage in the epidermis. All of these factors contribute to the development of new vitiligo lesions [16]. Furthermore, an increase in the concentration of Interferon-γ-dependent chemokines and their respective receptors, which are related to the progression of the pathology, have been observed in patients with vitiligo [11]. When the CD8+ T lymphocytes are activated by antigen recognition, the chemotaxis of new cells of the immune system is promoted to the site of damage [17]. Among the factors involved in the chemotaxis process, the most important are the chemokines, which are low molecular weight proteins (between 8–12 kDa) that promote immune cell migration to the site of injury when interacting with their receptor [18]. In patients with vitiligo, the overexpression of chemokine receptors of the CXC family has been reported, namely CXCR3 and CXCR6, as well as the chemokines CXCL9, 10, and 11, natural ligands of CXCR3, and CXCL16, the only known natural ligand of CXCR6 [11,19]. In the case of the CXCR3 axis, the overexpression of the CXCR3 receptor in CD8+ and memory T lymphocytes in the skin and peripheral blood of patients with vitiligo has been reported. Moreover, the reduction in CXCL10 and CXCR3 concentrations decreases the lymphocyte infiltration in areas of injury in animal models. Therefore, CXCL10 is considered an important marker in the progression of vitiligo due to the lower overexpression in patients with stable vitiligo than the overexpression in patients with active vitiligo. Additionally, it is considered that stressed keratinocytes are responsible for the production and release of CXCL10 to recruit lymphocytes into the epidermis and promote the melanocytotoxic activity of CD8+; in comparison, the inhibition of CXCR3 by antibodies decreases the presence of CD8+ lymphocytes in lesions in animal vitiligo models [20–22]. The CXCR3 chemokine receptor is a transmembrane G-protein-coupled receptor (GPCR) with three isoforms, CXCR3A (342 aminoacids, a.a.), CXCR3B (451 a.a.), and CXCR3-alt (267 a.a.), by alternative splicing [23–25]. CXCR3A is the most expressed isoform, followed by CXCR3B; it has been reported in the selective expression of these isoforms in different pathologies such as cancer, endometriosis, and Crohn’s disease [24,25]. The natural ligands for CXCR3 are chemokines CXCL9 (MIG), CXCL10 (IP-10), and CXCL11 (I-TAC) [26]. These chemokines have different affinities with the receptor; CXCL11 has the highest affinity, followed by CXCL10 and CXCL9. CXCR3A is a GPCR coupled with a G-protein (GP) Gαi/0 subunit and its natural ligands, and this isoform is overexpressed in different pathologies of an inflammatory nature, in addition to several types of cancer [27,28], and its activation promotes the mobilization of calcium in the cell, resulting in the chemotaxis of immune system cells [28]. In contrast, CXCR3B is coupled to the Gαs subunit and β-arrestin signaling [23–25], and the expression of CXCR3B in the melanocytes of patients with vitiligo is related to the apoptosis of melanocytes in early stages of the pathology by the stimulus of CXCL10 excreted by innated lymphoid cells [29]. In the chemokine–receptor interaction, a “two-site” model for complex formation has been proposed: the receptor first recognizes the binding site located in the N-loop region (“docking domain”) after two cysteines in chemokines. This initial contact facilitates the subsequent binding and proper positioning of the flexible N-terminal region (“triggering domain”), which activates the receptor by Molecules 2020,, 25,, 4413x FOR PEER REVIEW 3 of 30

and proper positioning of the flexible N-terminal region (“triggering domain”), which activates the interactingreceptor by presumablyinteracting presumably with multiple with receptor multiple helical receptor sites helical and inducingsites and ainducing change a in change the receptor in the conformationreceptor conformation [30]. Furthermore, [30]. Furthermore, recent reports recent propose reports that propose there that is an there alternative is an alternative binding site binding to the two-sitesite to the model, two-site reported model, in reported an extensive in an study extensive of the study interaction of the betweeninteraction CXCL12 between and CXCL12 CXCR4 thatand confirmsCXCR4 that the confirms relevance the of sulfationrelevance on of tyrosinessulfation inon CXC tyrosines receptors, in CXC thus receptor expandings, thus the expanding knowledge the of chemokine–CXCRknowledge of chemokine–CXCR interactions [31 interactions]. [31]. As forfor activation,activation, itit has has been been proposed proposed that that chemokine chemokine receptors receptors are are activated activated in twoin two steps: steps: (1) the(1) interactionthe interaction of chemokine of chemokine with with the the extracellular extracellula regionr region and and the the movement movement into into thethe transmembranetransmembrane region, promoting aa conformational changechange inin thethe receptor,receptor, andand (2)(2) separationseparation ofof thethe βγβγ complex from the αii/0/0 subunitsubunit (Figure (Figure 11))[ [32–34].32–34]. Moreover, Moreover, GPCR GPCR acti activationvation is related to changes in the rotation angle and polar interactions between the transmembranetransmembrane segments (TMs). In the rhodopsin receptor (Rho),(Rho), activation has been reported by the rotation of TMs 1, 3, 5, 6, and 7,7, wherewhere relevantrelevant residuesresidues such as W265 allow the rotation of TM6 for Rho Rho acti activation.vation. Similarly, Similarly, the highly conserved DRY motif is relevantrelevant toto thethe stabilizationstabilization ofof activeactive andand inactiveinactive states.states. This motif contains an ionicionic interactioninteraction between D and R, known as the “arginine cage” [[35,36],35,36], which is related to an inactive state of the receptors; whenwhen thethe receptor receptor is is activated, activated, this this interaction interaction breaks breaks down down and and promotes promotes interactions interactions with otherwith other TMs andTMs with and the with alpha the subunit alpha subunit of the G-protein. of the G-protein. The DRY The motif DRY is present motif inis mostpresent chemokine in most receptors,chemokine with receptors, the exception with the of exception CXCR6, whichof CXCR has6, an which analogous has an DRF analogous motif [ 36DRF–41 motif]. [36–41].

Figure 1. CXCR3 receptor activation. The first first activation step is the arrival of the chemokine and the subsequent orientation towards the transmembrane region. The The second step is the release of thethe βγβγ complex caused by the conformational changechange ofof thethe receptor.receptor.

Synthetic antagonists antagonists for for CXCR3 CXCR3 have have been been report reporteded in recent in recent years, years, mostly mostly found found by high- by high-throughputthroughput screening screening (HTS) (HTS)efforts e andfforts subsequent and subsequent optimization optimization methods methods [42–51]; [however,42–51]; however, studies studiesto understand to understand the mechanism the mechanism of antagonist of antagonist and natural and natural ligand ligand binding binding are limited, are limited, because because the thecrystallographic crystallographic structure structure of ofthis this receptor receptor has has not not been been resolved resolved experimentally [[26].26]. Protein homology modelingmodeling is ais computational a computational technique technique that allows that anallows approximation an approximation to the three-dimensional to the three- structuredimensional of a structure protein that of doesa protein not have that an does experimentally not have an resolved experimentally tridimensional resolved structure, tridimensional based on thestructure, principle based of on “similar the principle primary of sequences“similar prim willary have sequences a similar will three-dimensional have a similar three-dimensional conformation”. Thus,conformation”. the accuracy Thus, of thethe models accuracy depends of the on models the similarity depends of on the the amino similarity acid sequences of the amino to proteins acid whosesequences crystallographic to proteins whose structures crystallographic are reported struct inures databases are reported such in as databases the Protein such Data as the Bank Protein [52]. CXCR3Data Bank homology [52]. CXCR3 models homology have been models reported have been to elucidate reported theto elucidate mechanism the mechanism of receptor of activation receptor andactivation interactions and interactions with their with chemokines their chemokines and antagonists; and antagonists; however, however, these models these did models not consider did not post-translationalconsider post-translational modifications modifications contained contained in the receptor in the recept or wereor or modeled were modeled using a using template a template with a lowwith percentage a low percentage of identity of identity [26,53]. [26,53]. Due to the lack of efficient efficient therapeutic options, the search for new active compounds is necessary to improve the quality of life of patients with vi vitiligo.tiligo. The The antagonism antagonism of CXCR3 to inhibit chemotaxis is a target of interest in the search for new bioactive molecules due to a critical role in the recruitment + of CD8+ melanocytotoxicmelanocytotoxic lymphocytes lymphocytes in in the the skin; skin; howeve however,r, the lack of a three-dimensional three-dimensional structure limits thethe development development of of new new specific specific drugs drugs targeted targeted to CXCR3. to CXCR3. In this work,In this the work, homology the homology modeling ofmodeling the chemokine of the chemokine receptor CXCR3 receptor and CXCR3 the chemokine and the chemokine CXCL9 is reported;CXCL9 is additionally,reported; additionally, the molecular the Molecules 2020, 25, 4413 4 of 30 dynamics study of interactions between CXCR3 and their chemokines CXCL9, CXCL10, and CXCL11 are presented. The aim of the study was to observe the conformational changes of CXCR3 in the presence of its natural agonists, in addition to the protein–protein interactions present in these complexes, to better understand the two-step activation of this receptor and the interactions of the known antagonists with those conformations. The results could contribute to the future design of new specific antagonists in the therapy of vitiligo and other pathologies in which the CXCR3 axis is involved.

2. Results and Discussion

2.1. Protein Structures, Homology Modeling, and Relaxation by Molecular Dynamics (MD) Simulations

2.1.1. Protein Structures and Homology Modeling All Ramachandran plots are shown in the Supplementary Information (Figures S1–S3). The I-TASSER server built five models of CXCR3 using CCR5 as a template (PDB code: 4MBS, chain A), with a sequence identity of 36.49%, considering that a percentage greater than 30% facilitates the prediction of structures at the time of modeling [54]. The best model by I-TASSER was the model1, which was selected by I-TASSER through the evaluation of the C-score and TM-score: model1 had a C-score of -0.65 (with a range of 5 to 2, higher values represent greater confidence in the − model) and a TM-score of 0.63 0.14 (greater than 0.5 is considered a model with correct topology). ± Additionally, model1 was evaluated using the MolProbity public server [55] to obtain a Ramachandran plot to validate this model. The percentage of residues in favorable areas was 85.52% (Figure S1A); the higher the percentage of residues in the favored regions in the Ramachandran plot, the better the stereochemical quality of the homology models [56]. According to this metric, the CXCR3 model by I-TASSER was a suitable model, even without a relaxation process by MD. In the case of CXCL9, the I-TASSER server built five models of CXCL9 using CXCL2 as a template (PDB code: 5OB5, chain A), with a sequence identity of 47.14%. The best model was the model1 and had a C-score of -1.11 and a TM-score of 0.58 0.14. The percentage of residues in favorable areas in ± the Ramachandran plot was 75.25% (Figure S1C), indicating a model that needed relaxation by MD to improve its quality. Similarly, the αi/0 subunit was modeled using the structure of chimeric G-alpha proteins (Rattus norvegicus and Bos taurus) as a template (PDB code: 3V00), with the sequence identities of 87.89% and 65.92%, respectively. The best model had a C-score of 1.00 and TM-score of 0.85 0.08. ± The percentage of residues in favorable areas in the Ramachandran plot was 91.19% (Figure S1E) [57].

2.1.2. Relaxation of Homology Models In the short production MD simulation for equilibration of the homology models, the most representative conformation (cluster1) of the CXCL9 five-ns all-atom molecular dynamics (AA-MD) simulation was aligned with the homology model, with a root mean standard deviation (RMSD) of the α-carbon of 2.675 Å (Figure S1), and the percentage of favored residues of the Ramachandran plot was 89.11%, showing that cluster1 conformation was an appropriate conformation for subsequent studies (Figure S1D). RSMD was used as a measure of change in the system with respect to a starting structure. Likewise, the alignment of cluster1 from the CXCR3 50-ns AA-MD simulation presented an RMSD of 3.234 Å compared to the homology model obtained by I-TASSER (Figure2), corresponding to the adjustment of residues that were in unfavorable conformations to improve the protein stability. The percentage of favored residues in the Ramachandran plot increased to 90.44% (Figure S1B), relaxing the conformation to a state of lower energy, thus obtaining a viable model for subsequent studies. MoleculesMolecules 20202020,, 2525,, 4413x FOR PEER REVIEW 55 ofof 3030 Molecules 2020, 25, x FOR PEER REVIEW 5 of 30

Figure 2. All-atom molecular dynamics (AA-MD) simulation of CXCR3 50 ns. (A) CXCR3 model Figure 2. All-atomAll-atom molecular molecular dynamics dynamics (AA-MD) (AA-MD) simulation simulation of of CXCR3 50 ns. ( (AA)) CXCR3 CXCR3 model obtainedobtained fromfrom I-TASSERI-TASSER (initial(initial confirmationsconfirmations oror TT0),), ((B)) cluster1cluster1 ofof thethe simulation,simulation, andand ((CC)) thethe obtained from I-TASSER (initial confirmations or T0), (B) cluster1 of the simulation, and (C) the alignmentalignment ofof TT0 and cluster1. alignment of T0 and cluster1.

2.2.2.2. CXCR3-Gαii/0/0βγ Complex Building and Relaxation 2.2. CXCR3-Gαi/0βγ Complex Building and Relaxation AlignmentAlignment betweenbetween thethe 5HT1B5HT1B (PDB(PDB Code:Code: 6G79) and cluster1cluster1 fromfrom thethe CXCR3CXCR3 50-ns50-ns AA-MDAA-MD Alignment between the 5HT1B (PDB Code: 6G79) and cluster1 from the CXCR3 50-ns AA-MD simulationsimulation was made for for the the addition addition of of G Gααi/0iβγ/0βγ toto the the latter. latter. Once Once the the subunits subunits were were combined, combined, the simulation was made for the addition of Gαi/0βγ to the latter. Once the subunits were combined, the theresulting resulting complex complex was was evaluated evaluated in inMolProbity MolProbity to toobserve observe the the residues residues in in the the favored favored areas, resulting complex was evaluated in MolProbity to observe the residues in the favored areas, obtainingobtaining 91.91%91.91% (Figure(Figure S2A).S2A). Then,Then, thethe 100-ns100-ns MDMD simulationsimulation waswas performedperformed toto relaxrelax thethe system.system. obtaining 91.91% (Figure S2A). Then, the 100-ns MD simulation was performed to relax the system. Cluster1Cluster1 ofof the the MD MD simulation simulation of the of CXCR3-G the CXCR3-Gαi/0βγαcomplexi/0βγ complex had an had RMSD an of RMSD 5.834 Åof compared 5.834 Å Cluster1 of the MD simulation of the CXCR3-Gαi/0βγ complex had an RMSD of 5.834 Å tocompared the initial to confirmations the initial confirmations (T0) (Figure (T30),) (Figure possibly 3), because possibly the because conformation the conformation presented presented by the GP byi/0 compared to the initial confirmations (T0) (Figure 3), possibly because the conformation presented by belongsthe GPi/0 to belongs a stable to interaction a stable interaction with the serotonin with the 5HT1B serotonin receptor. 5HT1B The receptor. simulation The allowed simulation the relaxation allowed the GPi/0 belongs to a stable interaction with the serotonin 5HT1B receptor. The simulation allowed ofthe residues relaxation to aof favorable residues to interaction a favorable between interaction the GPbetweeni/0 and the the GP CXCR3i/0 and receptor.the CXCR3 After receptor. alignment After the relaxation of residues to a favorable interaction between the GPi/0 and the CXCR3 receptor. After betweenalignment the between GPi/0 T 0theand GP GPi/0 iT/00 cluster1and GPi/0 of cluster1 the 100-ns of the dynamics, 100-ns dynamics, a RMSD of a 5.022RMSD Å of was 5.022 observed, Å was alignment between the GPi/0 T0 and GPi/0 cluster1 of the 100-ns dynamics, a RMSD of 5.022 Å was suggestingobserved, suggesting a conformational a conformati changeonal that change promoted that thepromoted stabilization the stabilization of the complex. of the complex. observed, suggesting a conformational change that promoted the stabilization of the complex.

FigureFigure 3.3. AA-MD simulationsimulation ofof CXCR3-GCXCR3-Gααi/i/00βγ 100 ns. ( A) Structure of thethe complexcomplex atat TT00,(, (B) cluster1 Figure 3. AA-MD simulation of CXCR3-Gαi/0βγ 100 ns. (A) Structure of the complex at T0, (B) cluster1 ofof thethe simulation,simulation, andand ((CC)) thethe alignmentalignment ofof TT00 andand cluster1.cluster1. of the simulation, and (C) the alignment of T0 and cluster1. TheThe alignmentalignment betweenbetween only only the the CXCR3 CXCR3 receptor receptor was was 2.553 2.553 Å, Å, showing showing significant significant changes changes in the in The alignment between only the CXCR3 receptor was 2.553 Å, showing significant changes in regionsthe regions corresponding corresponding to the extracellularto the extracellular region, loops, region, and loops, the intracellular and the region.intracellular Additionally, region. the regions corresponding to the extracellular region, loops, and the intracellular region. theAdditionally, Ramachandran the Ramachandran plot indicated plot 92.72% indicated of favored 92.72% residues of favored (Figure residues S2B) after (Figure the MDS2B) simulation after the MD of Additionally, the Ramachandran plot indicated 92.72% of favored residues (Figure S2B) after the MD thissimulation system. of Thus, this thissystem. structure Thus of, this CXCR3 structure was used of CXCR3 for building was theused complexes for building for subsequentthe complexes studies for simulation of this system. Thus, this structure of CXCR3 was used for building the complexes for ofsubsequent the interaction studies with of the CXCL9, interaction 10, and with 11. CXCL9, 10, and 11. subsequent studies of the interaction with CXCL9, 10, and 11.

Molecules 2020, 25, 4413 6 of 30

2.3. CXCR3/Chemokines Complex Building and 50-ns AA-MD Simulation A summary of the clustering of the AA-MD simulations is shown in Table1.

Table 1. Timestep and clustering of the AA-MD simulations.

Most Representative Conformation Timestep (ns) Complex 5 ns 50 ns 100 ns Cutoff (nm) Clusters CXCR3 32.1 86.9 0.4 */0.2 ** 3 */10 ** CXCL9 3.8 0.2 14 CXCR3-CXCL9 31.1 0.4 9 CXCR3-CXCL10 33.3 0.51 3 CXCR3-CXCL11 26.6 0.4 10 * 50-ns simulation and ** 100-ns simulation. AA-MD: all-atom molecular dynamics.

Using the CXCR3 conformation in cluster1 of the CXCR3-Gαi/0βγ complex, the protein–protein dockings were made in the ClusPro server to observe the protein–protein interactions preliminarily between CXCR3 and the different chemokines and positioning the CXCR3/chemokines complex for the MD simulations. The docking was guided by the selected constrains in ClusPro (first 16 residues, Y27, and Y29 in CXCR3) derived by experimental mutagenesis data (see Materials and Methods). The conformation representative of the highest number of conformations in the cluster based on the combination of hydrophobic and electrostatic interactions given by ClusPro was chosen for the AA-MD and coarse-grained molecular dynamics (CG-MD) simulations. These complexes presented a good percentage of residues in favored areas of the Ramachandran plot: 91.60%, 90.87%, and 90.46%, respectively (Figure S3A–C). Once the complexes were built, a 50-ns AA-MD simulation was performed; the interactions between CXCR3 and the chemokines in T0 and cluster1 are presented in Figures4–6. All observed salt bridges of the CXCR3–chemokine interactions are included in the Supplementary Information as tables (Tables S1–S5). The CXCR3-CXCL9 complex has two salt bridges and the CXCR3-CXCL10 complex has one salt bridge; in contrast, the CXCR3-CXCL11 complex does not have salt bridges (Table S1). No hydrogen bonds were made with the post-translational modifications included in CXCR3 in any complex. Additionally, for CXCR3-CXCL9, cluster1 had a hydrogen bond between Q99 and glycosylation in the N32 residue of CXCR3 (Figure4B). By comparison, CXCL10 and CXCL11 did not show interactions with glycosylations, showing the importance of post-translational modifications for the interaction between CXCL9 and CXCR3 and disregarding their relevance in the initial interaction of CXCL10 and CXCL11; however, in cluster1 of CXCR3-CXCL11, the residue N22 participates in the hydrophobic environment, promoting the interactions between CXCR3 and CXCL11 (Figure6B). The salt bridges between CXCR3 and the chemokines had an interesting change: in the CXCR3-CXCL9 complex, they increased to a total of nine, while there were none in CXCR3-CXCL10 and two in CXCR3-CXCL11 (Table S2). Residues with post-translational modifications were not involved in any of the cases. Additionally, the binding of chemokines with CXCR3 caused changes in the receptor, with an RMSD of 2.841 Å for CXCL9, 2.646 Å for CXCL10, and 2.538 Å for CXCL11, mainly in the transmembrane region but, also, in Gαi/0βγ (Figures7A,8A and9A). The comparison of the initial conformations (T0) and the representative clusters shows a considerable change, which indicates that the system conformationally evolves to find a more stable interaction between the different chemokines and CXCR3. CXCL11 presents a considerable change, with an RMSD of 6.339 Å; however, interactions with CXCR3 are decreased compared to CXCL9 and CXCL10 (Figures4B,5B and6B). Molecules 2020, 25, 4413 7 of 30 MoleculesMolecules 2020 2020, 25, 25, x, xFOR FOR PEER PEER REVIEW REVIEW 7 7of of 30 30

FigureFigureFigure 4. Interactions 4. 4. Interactions Interactions between between between CXCL9 CXCL9 CXCL9 and and CXCR3.and CXCR3. CXCR3. (A )( A( CXCR3-CXCL9_TA) )CXCR3-CXCL9_T CXCR3-CXCL9_T0.(0.0 .(B B()B) Cluster1_50-ns)Cluster1_50-ns Cluster1_50-ns AA- AA-MD.AA- The RMD.MD. chain The The corresponds R R chain chain corresponds corresponds to CXCR3 to to CXCR3 CXCR3 and the and and Q the the chain Q Q ch ch toainain CXCL9. to to CXCL9. CXCL9. An An An increase increase increase ofof of hydrogen hydrogen bonds bonds in cluster1inin cluster1 ofcluster1 the 50-nsof of the the simulation50-ns 50-ns simulation simulation can becan can observed, be be observed observed as, wellas, as well well as as anas an an interaction interaction interaction betweenbetween between the thethe Q99 Q99 residue residue residue of CXCL9ofof CXCL9 and CXCL9 the and and glycosylation the the glycosylation glycosylation present present present in thein in the N32the N32 N32 residue re residuesidueof of of CXCR3.CXCR3. CXCR3. The The residues residues residues in in red in red redand and andpurple purple purple represent hydrophobic interactions between proteins. representrepresent hydrophobic hydrophobic interactions interactions between between proteins. proteins.

Figure 5. Interactions between CXCR3 and CXCL10. (A) CXCR3-CXCL10_T0.(B) Cluster1_50 ns. The R chain corresponds to CXCR3 and the Q chain to CXCL10. An increase of hydrogen bonds can be observed, along with a decrease of the hydrophobic environment involved and an increase in CXCL10 residues that interact with CXCR3 in cluster1 of the 50-ns simulation. Molecules 2020, 25, x FOR PEER REVIEW 8 of 30 Molecules 2020, 25, x FOR PEER REVIEW 8 of 30

Figure 5. Interactions between CXCR3 and CXCL10. (A) CXCR3-CXCL10_T0. (B) Cluster1_50 ns. The Figure 5. Interactions between CXCR3 and CXCL10. (A) CXCR3-CXCL10_T0. (B) Cluster1_50 ns. The RR chain chain corresponds corresponds to to CXCR3 CXCR3 and and the the Q Q chain chain to to CXCL10. CXCL10. An An increase increase of of hydrogen hydrogen bonds bonds can can be be observed, along with a decrease of the hydrophobic environment involved and an increase in CXCL10 Moleculesobserved,2020, 25 ,along 4413 with a decrease of the hydrophobic environment involved and an increase in CXCL108 of 30 residuesresidues that that interact interact with with CXCR3 CXCR3 in in cluster1 cluster1 of of the the 50-ns 50-ns simulation. simulation.

Figure 6. Interactions of CXCR3 and CXCL11. (A) CXCR3-CXCL11_T0. (B) Cluster1_50 ns. The R chain Figure 6. InteractionsInteractions of of CXCR3 CXCR3 and and CXCL11. CXCL11. (A) ( ACXCR3-CXCL11_T) CXCR3-CXCL11_T0. (B0).( Cluster1_50B) Cluster1_50 ns. The ns. R The chain R correspondschaincorresponds corresponds to to CXCR3 CXCR3 to CXCR3 and and the the and Q Q chain the chain Q to chain to CXCL11. CXCL11. to CXCL11. A A decrease decrease A decrease of of interactions interactions of interactions between between between CXCR3 CXCR3 CXCR3 and and CXCL11andCXCL11 CXCL11 is is observed. observed. is observed. Glycosylation Glycosylation Glycosylation in in the the in N22 N22 the part N22participates participatesicipates in in the the in hydropho hydropho the hydrophobicbicbic environment environment environment for for the the for interaction of CXCL11 T0 and cluster1. theinteraction interaction of CXCL11 of CXCL11 T0 and T0 and cluster1. cluster1.

FigureFigure 7. 7. AA-MDAA-MD simulation simulation of of CXCR3-CXCL9. CXCR3-CXCL9. (A (A) )Alignment Alignment of of CXCR3-CXCL9_T CXCR3-CXCL9_T0 (blue)(blue) and and Figure 7. AA-MD simulation of CXCR3-CXCL9. (A) Alignment of CXCR3-CXCL9_T0 (blue) and CXCR3-CXCL9_50CXCR3-CXCL9_50 ns ns (red). (red). (B (B) )Alignment Alignment of of CXCL9_T CXCL9_T0 (blue)(blue) and and CXCL9_50 CXCL9_50 ns ns (magenta). (magenta). There There is is CXCR3-CXCL9_50 ns (red). (B) Alignment of CXCL9_T00 (blue) and CXCL9_50 ns (magenta). There is aa root root mean mean standard standard deviation deviation (RMSD) (RMSD) of of 2.841 2.841 Å Å between between T T0 0andand cluster1 cluster1 of of the the 50-ns 50-ns simulation, simulation, a root mean standard deviation (RMSD) of 2.841 Å between T0 and cluster1 of the 50-ns simulation, andand changes changes in in the the G Gααi/0i/βγ0βγ areare observed; observed; in in addition, addition, there there is is a aconformational conformational change change of of CXCL9 CXCL9 with with and changes in the Gαi/0βγ are observed; in addition, there is a conformational change of CXCL9 with an RMSD of 2.972 Å, which indicates that the system evolves into a more favorable state for interactions anan RMSD RMSD of of 2.972 2.972 Å, Å, which which indicates indicates that that the the sy systemstem evolves evolves into into a amore more favorable favorable state state for for between CXCL9 and CXCR3. interactionsinteractions between between CXCL9 CXCL9 and and CXCR3. CXCR3. Molecules 2020, 25, x FOR PEER REVIEW 9 of 30

Molecules 2020, 25, 4413 9 of 30 Molecules 2020, 25, x FOR PEER REVIEW 9 of 30

Figure 8. AA-MD simulation of CXCR3_CXCL10. (A) Alignment of CXCR3-CXCL10_T0 (blue) and

CXCR3-CXCL10_50 ns (green). (B) Alignment of CXCL10_T0 (blue) and CXCL10_50 ns (cyan). There Figure 8. AA-MD simulation of CXCR3_CXCL10. (A) Alignment of CXCR3-CXCL10_T (blue) and isFigure an RMSD 8. AA-MD of 2.646 simulation Å between of T CXCR3_CXCL10.0 and cluster1 of the(A) 50-nsAlignment simulation; of CXCR3-CXCL10_T changes in the G00 α(blue)i/0βγ areand CXCR3-CXCL10_50 ns (green). (B) Alignment of CXCL10_T (blue) and CXCL10_50 ns (cyan). There is observed;CXCR3-CXCL10_50 additionally, ns a (green). conformati (B) Alignmentonal change of of CXCL10_T CXCL100 exists0 (blue) wi andth an CXCL10_50 RMSD of 4.745ns (cyan). Å, which There an RMSD of 2.646 Å between T and cluster1 of the 50-ns simulation; changes in the Gα βγ are showsis an RMSDthat the of system2.646 Å evolves between into0 T0 anda more cluster1 favorable of the state 50-ns for simulation; interactions changes between in theCXCL10 Gαi/0i/0βγ and are observed; additionally, a conformational change of CXCL10 exists with an RMSD of 4.745 Å, which CXCR3.observed; additionally, a conformational change of CXCL10 exists with an RMSD of 4.745 Å, which showsshows thatthat the the system system evolves evolves into into a more a more favorable favorable state forstate interactions for interactions between between CXCL10 CXCL10 and CXCR3. and CXCR3.

FigureFigure 9. 9. AA-MDAA-MD simulation simulation of of CXCR3_CXCL11. CXCR3_CXCL11. ( (AA) )Alignment Alignment of of CXCR3-CXCL11_T CXCR3-CXCL11_T0 0(blue)(blue) and and

CXCR3-CXCL11_50CXCR3-CXCL11_50 ns ns (yellow). (yellow). (B (B) )Alignment Alignment of of CXCL11_T CXCL11_T0 0(blue)(blue) and and CXCL11_50 CXCL11_50 ns ns (orange). (orange). α βγ ThereThereFigure is is an9. an AA-MDRMSD RMSD of ofsimulation 2.538 2.538 Å Å between between of CXCR3_CXCL11. T T00 andand cluster1 cluster1 (A of of) Alignmentthe the 50-ns 50-ns simulation; simulation; of CXCR3-CXCL11_T changes in the 0 (blue) G αi/0i/ 0βγand areareCXCR3-CXCL11_50 observed observed in in addition addition ns to(yellow). to a a co conformationalnformational (B) Alignment change change of ofCXCL11_T of CXCL10 CXCL10 0with (blue) with an an andRMSD RMSD CXCL11_50 of of 6.339 6.339 Å; Å;ns however, however,(orange). unlikeunlikeThere CXCL9 is CXCL9 an RMSD and and 10, 10,of the2.538 the hydrog hydrogen Å betweenen bonds bonds T0 and and and cluster1 hydrophobic-type hydrophobic-type of the 50-ns interactionsinteractions simulation; decreased. decreased.changes in the Gαi/0βγ are observed in addition to a conformational change of CXCL10 with an RMSD of 6.339 Å; however, Once the interactions and evolution of the CXCR3-chemokine systems were observed, it was Onceunlike the CXCL9 interactions and 10, the and hydrog evolutionen bonds of andthe hydrophobic-typeCXCR3-chemokine interactions systems decreased. were observed, it was expectedexpected that that the the two-sites two-sites model model interactions interactions were were present present in in the the complexes. complexes. In In the the case case of of the the CXCL9-CXCR3 complex, the chemokine had interactions at T in the N-loop region; after 50 ns of CXCL9-CXCR3Once the interactionscomplex, the and chemokine evolution had of theinteractions CXCR3-chemokine at T0 0in the systems N-loop wereregion; observed, after 50 itns was of simulation,simulation,expected that the the interactionsinteractions two-sites increased increasedmodel interactions to to the the other othe residueswerer residues present of CXCL9, of in CXCL9, the similarly complexes. sim toilarly the In casesto the the case of cases CXCL10 of theof and CXCL11. Nevertheless, this simulation time was insufficient, since the first activation step could CXCL10CXCL9-CXCR3 and CXCL11. complex, Nevertheless, the chemokine this simulation had interactions time was at T insufficient,0 in the N-loop sinc eregion; the first after activation 50 ns of stepnotsimulation, becould observed, not the be interactions whichobserved, consists which increased of theconsists approach to the of thothe ofe approach chemokinesr residues of ofchemokines to CXCL9, the transmembrane sim to ilarlythe transmembrane to regionthe cases of the of receptorCXCL10 andand theCXCL11. subsequent Nevertheless, conformational this simulation change of time the receptor.was insufficient, Within thissinc timee the frame, first activation only the firststep interactionscould not be between observed, the chemokineswhich consists and of CXCR3 the approach could be of observed. chemokines to the transmembrane Molecules 2020, 25, x FOR PEER REVIEW 10 of 30 Molecules 2020, 25, 4413 10 of 30 region of the receptor and the subsequent conformational change of the receptor. Within this time frame, only the first interactions between the chemokines and CXCR3 could be observed. 2.4. Complex Stabilization in CG-MD Simulations 2.4. Complex Stabilization in CG-MD Simulations To verify the two-step activation of the CXCR3 receptor, CG-MD simulations were performed. Currently,To verify the the Martini two-step 2.2p forceactivation field implementedof the CXCR3 on receptor, CHARMM-GUI CG-MD simula does nottions have were parameters performed. for Currently,small molecules the Martini and modified 2.2p force residues; field implement thus, theed addition on CHARMM-GUI of post-translational does not CXCR3 have parameters modifications for smallwas not molecules possible. and However, modified CG-MDs residues; allowed thus, the larger addition simulation of post-translational times with shorter CXCR3 calculation modifications times. was notThe possible. complexes However, of CXCR3-G CG-MDsαi/0βγ allowed/chemokines larger (obtainedsimulation from times docking with shorter in ClusPro) calculation were used times. as a startingThe complexes structure. of According CXCR3-G toαi/0 theβγ/chemokines results of the (obtained CG-MD simulations,from docking the in representativeClusPro) were used clusters as awere starting around structure. 500–800 According ns, as seen to in the Table results2. of the CG-MD simulations, the representative clusters were around 500–800 ns, as seen in Table 2. Table 2. Timestep and clustering of the coarse-grained molecular dynamics (CG-MD) simulations. Table 2. Timestep and clustering of the coarse-grained molecular dynamics (CG-MD) simulations. Most Representative Conformation Timestep (ns) Complex MostCluster1 Representative Cluster2 Conformation Timestep Cluster3 (ns) Cutoff (nm) Clusters Complex Cluster1 Cluster2 Cluster3 Cutoff (nm) Clusters CXCR3-CXCL9 607.0 131.0 39.0 0.6 10 CXCR3-CXCL9 607.0 131.0 39.0 0.6 10 CXCR3-CXCL10 787.9 159.6 51.2 0.6 10 CXCR3-CXCL11CXCR3-CXCL10 532.7787.9 171.3 159.6 979.3 51.2 0.6 0.6 10 10 CXCR3-CXCL11 532.7 171.3 979.3 0.6 10

InIn these these simulations, simulations, the the first first step step of receptor of receptor activation activation could could be inferred be inferred by the by decrease the decrease of the distanceof the distance between between the chemokine the chemokine and the and transmembrane the transmembrane region region of CXCR3; of CXCR3; additionally, additionally, a conformationala conformational changechange was was observed observed in in the the complex, complex, expressed expressed as as an an RMSD RMSD of of 9.397 9.397 Å Å for for CXCL9, CXCL9, 10.72410.724 Å for CXCL10, andand 9.8859.885 ÅÅ forfor CXCL11, CXCL11, with with respect respect to to the the T 0Tof0 of each each simulation. simulation. The The changes changes in α βγ inthe the extracellular extracellular region region and and G i /0Gαi/0inβγ the in representativethe representative clusters clusters of CG-MD of CG-MD could becould detected be detected visually. visually.Moreover, Moreover, there was there a decrease was ina decrease the distance in th betweene distance the chemokinesbetween the and chemokines CXCR3 proportional and CXCR3 to proportionalthe simulation to timethe simulation (Figures 10 time–12). (Figures 10–12).

FigureFigure 10.10. Coarse-grainedCoarse-grained molecular molecular dynamics dynamics (CG-MD) (CG-MD) simulation simulation of CXCR3-CXCL9. of CXCR3-CXCL9. (A) Alignment (A)

Alignmentbetween CXCR3-CXCL9_T between CXCR3-CXCL9_T0 (blue) and0 cluster1_CG-MD(blue) and cluster1_CG-MD (red), with an (red), RMSD with of an 9.409 RMSD Å. (B of) Distances 9.409 Å. (amongB) Distances the CXCR3-CXCL9 among the CXCR3-CXCL9 complex chains. complex The distancechains. The decreases distance during decreases the simulation during the by simulationapproximately by approximately 2 to 3 Å. 2 to 3 Å. Molecules 2020, 25, 4413 11 of 30 Molecules 2020, 25, x FOR PEER REVIEW 11 of 30 Molecules 2020, 25, x FOR PEER REVIEW 11 of 30

Figure 11. CG-MD simulation of CXCR3-CXCL10. (A) Alignment of CXCR3-CXCL10_T0 (blue) and Figure 11. CG-MD simulation ofof CXCR3-CXCL10.CXCR3-CXCL10. (A) AlignmentAlignment ofof CXCR3-CXCL10_TCXCR3-CXCL10_T00 (blue) and cluster1_CG-MD (green), with an RMSD of 9.409 Å. (B) Distances among the CXCR3-CXCL10 cluster1_CG-MD (green),(green), with with an an RMSD RMSD of 9.409of 9.409 Å. (B )Å. Distances (B) Distances among among the CXCR3-CXCL10 the CXCR3-CXCL10 complex complex chains. The distance decreases during the simulation by approximately 2 to 3 Å. chains.complex The chains. distance The decreasesdistance decreases during the during simulation the simulation by approximately by approximately 2 to 3 Å. 2 to 3 Å.

Figure 12. CG-MDCG-MD simulation simulation of ofCXCR3-CXCL11. CXCR3-CXCL11. (A) (AlignmentA) Alignment of CXCR3-CXCL11_T of CXCR3-CXCL11_T0 (blue)0 (blue) and Figure 12. CG-MD simulation of CXCR3-CXCL11. (A) Alignment of CXCR3-CXCL11_T0 (blue) and cluster1_CG-MDand cluster1_CG-MD (yellow), (yellow), with with an RMSD an RMSD of of9.853 9.853 Å. Å. (B ()B )Distances Distances among among the the CXCR3-CXCL11 cluster1_CG-MD (yellow), with an RMSD of 9.853 Å. (B) Distances among the CXCR3-CXCL11 complex chains. The The distance distance decreases decreases during during the simulation simulation by approximately 1 Å less than CXCL9 complex chains. The distance decreases during the simulation by approximately 1 Å less than CXCL9 and CXCL10. and CXCL10. Furthermore, the interactions between CXCR3 and th thee chemokines in the re representativepresentative clusters Furthermore, the interactions between CXCR3 and the chemokines in the representative clusters of CG-MD simulations were calculated. For For both both CXCL9 CXCL9 and and CXCL11, the hydrogen bonds and of CG-MD simulations were calculated. For both CXCL9 and CXCL11, the hydrogen bonds and hydrophobic interactionsinteractions increasedincreased significantly, significantly, whereas whereas CXCL10 CXCL10 did did not not show show a large a large increase increase in the in hydrophobic interactions increased significantly, whereas CXCL10 did not show a large increase in thenumber number of polar of polar interactions; interactions; however, however, the number the number of hydrophobic of hydrophobic interactions interactions increased increased with respect with tothe T numberor cluster1 of polar of the interactions; 50-ns AA-MD however, simulation the nu (Figuresmber of 13 hydrophobic–15). interactions increased with respect0 to T0 or cluster1 of the 50-ns AA-MD simulation (Figures 13–15). respect to T0 or cluster1 of the 50-ns AA-MD simulation (Figures 13–15). Interestingly, CXCR3 residues involved in the interaction with the chemokines were in the Interestingly, CXCR3 residues involved in the interaction with the chemokines were in the extracellular loops and near the transmembrane re region.gion. For For CXCL9, there we werere no coincidences of extracellular loops and near the transmembrane region. For CXCL9, there were no coincidences of direct interactionsinteractions withwith the the relevant relevant residues residues considered considered for for protein–protein protein–protein docking docking (see (see Materials Materials and direct interactions with the relevant residues considered for protein–protein docking (see Materials andMethods). Methods). However, However, they werethey close were to close residues to residues that form that the CXCR3-CXCL9form the CXCR3-CXCL9 interaction observedinteraction in and Methods). However, they were close to residues that form the CXCR3-CXCL9 interaction observedthe CG-MD in simulation;the CG-MD specifically,simulation; thesespecifically, were the these residues were N22the residues and Y27, N22 which and form Y27, hydrophobic which form observed in the CG-MD simulation; specifically, these were the residues N22 and Y27, which form hydrophobicinteractions, andinteractions, those that and should those have that post-translational should have post-translational modifications modifications (glycosylation (glycosylation and sulfation). hydrophobic interactions, and those that should have post-translational modifications (glycosylation andTherefore, sulfation). it cannot Therefore, be discarded it cannot that be thesediscarded modifications that these could modifications have a significant could have influence a significant on the and sulfation). Therefore, it cannot be discarded that these modifications could have a significant influenceinteraction on with the interaction CXCL9. By with comparison, CXCL9. By saltcomparis bridge-typeon, salt interactionsbridge-type inte increasedractions to increased a total of to 15 a influence on the interaction with CXCL9. By comparison, salt bridge-type interactions increased to a total(Table of S3). 15 (Table S3). total of 15 (Table S3). Molecules 2020, 25, 4413 12 of 30 MoleculesMolecules 2020 2020, 25, 25, x, FORx FOR PEER PEER REVIEW REVIEW 12 of12 30 of 30

Figure 13. Interactions between CXCR3-CXCL9 from the CG-MD simulation. There is an increase in Figure 13.13. Interactions between CXCR3-CXCL9CXCR3-CXCL9 from the CG-MD simulation. There is an increase in hydrogen bonds and hydrophobic interactions of CXCR3 with CXCL9 after the CG-MD simulation. hydrogen bondsbonds andand hydrophobichydrophobic interactionsinteractions ofof CXCR3CXCR3 withwith CXCL9CXCL9 afterafter thethe CG-MDCG-MD simulation.simulation. InIn the the case case of of CXCL10, CXCL10, there there was was also also no no coincidence coincidence between between thethe representativerepresentative conformationsconformations of In the case of CXCL10, there was also no coincidence between the representative conformations theof CG-MD the CG-MD simulation simulation and the and previously the previously reported re residues.ported residues. Additionally, Additionally, the hydrophobic the hydrophobic interactions of the CG-MD simulation and the previously reported residues. Additionally, the hydrophobic increased;interactions these increased; interactions these are interactions relevant inare the rele activationvant in the of activation CXCR3 byof CXCR3 CXCL10, by andCXCL10, the interaction and the interactionsinteraction increased; with residue these Y27 interactions of CXCR3 are was rele detected,vant in thesuggesting activation that of thisCXCR3 could by be CXCL10, a significant and the with residue Y27 of CXCR3 was detected, suggesting that this could be a significant influence of interactioninfluence withof post-translational residue Y27 of modificationsCXCR3 was detected,in the interaction suggesting between that thisCXCL10 could and be CXCR3.a significant In post-translational modifications in the interaction between CXCL10 and CXCR3. In parallel, there was influenceparallel, ofthere post-translational was an increase ofmodifications 12 salt bridges in (Table the interaction S4). between CXCL10 and CXCR3. In an increase of 12 salt bridges (Table S4). parallel, there was an increase of 12 salt bridges (Table S4).

Figure 14. Interactions between CXCR3 and CXCL10 from the CG-MD simulation. An increase in hydrogen bonds and hydrophobic interactions is observed, although, in comparison with CXCL9, the Figurenumber 14.14. ofInteractions polar interactions between is CXCR3lower. The and intera CXCL10CXCL10ctions from are thethemaintained CG-MDCG-MD simulation.withsimulation. the first An residues increase of in CXCR3. hydrogen bonds and and hydrophobic hydrophobic interactions interactions is isob observed,served, although, although, in comparison in comparison with with CXCL9, CXCL9, the thenumber number of polar of polar interactions interactions is lower. is lower. The Theintera interactionsctions are aremaintained maintained with with the thefirst first residues residues of In CXCL11, hydrogen bonds are observed with R197 and E293 of CXCR3, which is reported as ofCXCR3. CXCR3. being important for CXCR3-CXCL11 binding to stabilize the complex. By comparison, the salt bridges increasedInIn CXCL11,CXCL11, to a total hydrogenhydrogen of 15 (Table bondsbonds S5). are observedobserved with R197 and E293E293 ofof CXCR3,CXCR3, whichwhich isis reportedreported asas being important for CXCR3-CXCL11 bindingbinding toto stabilizestabilize thethe complex.complex. By comparison, the salt bridges increasedincreased toto aa totaltotal ofof 1515 (Table(Table S5).S5). Molecules 2020, 25, 4413 13 of 30

Molecules 2020, 25, x FOR PEER REVIEW 13 of 30

FigureFigure 15. Interactions 15. Interactions between between CXCR3-CXCL11 CXCR3-CXCL from11 from the CG-MDthe CG-MD simulation. simulation. An increaseAn increase of hydrogen of bondshydrogen and hydrophobic bonds and hydrophobic interactions interactions is observed. is observed. The diagram The diagram shows shows the interaction the interaction with with the R197 the R197 residue of CXCR3, a residue reported to be important for the binding of CXCL11, and residue of CXCR3, a residue reported to be important for the binding of CXCL11, and interactions with interactions with E293, which is reported to be necessary for the stabilization of the CXCL11-CXCR3 E293, which is reported to be necessary for the stabilization of the CXCL11-CXCR3 complex. complex.

Finally,Finally, due due to theto the interactions interactions observed observed betweenbetween CXCR3 CXCR3 and and the the different different chemokines, chemokines, it can it can be suggestedbe suggested that that more more residues residues could be be responsi responsibleble for the for conformational the conformational change changetowards towardsthe first the firstactivation step step of of the the CXCR3 CXCR3 receptor. receptor. However, However, the βγ the complexβγ complex did not did separate not separate from thefrom αi/0 the αi/0 subunit;subunit; on the contrary, they they approached approached each each other, other, possibly possibly because because this thissimulation simulation did not did not containcontain the nucleotidethe nucleotide necessary necessary for for the the activation activation of G Gααi/0i/βγ0βγ. In. Ingeneral, general, GPs GPs are arein the in presence the presence of of guanosine-diphosphateguanosine-diphosphate (GDP), (GDP), which which promotespromotes an an inactive inactive state state and and changes changes to an to activated an activated state state whenwhen the nucleotidethe nucleotide exchanges exchanges from from GDP GDP toto guanosineguanosine triphosphate triphosphate (GTP) (GTP) with with the thesubsequent subsequent separationseparation of the of theβγ βγcomplex complex with with the theα αsubunit, subunit, triggering triggering the the signaling signaling process process [41]. [ 41Nevertheless,]. Nevertheless, the conformational changes present in the transmembrane region in CXCR3 indicated an activation the conformational changes present in the transmembrane region in CXCR3 indicated an activation process that corresponded to that expected when the different chemokines interact; in the first processtimesteps, that corresponded there was an to interact that expectedion with whenthe N-terminal the different region chemokines of CXCR3; interact;then, the inchemokine the first timesteps, was thereorientated was an interaction to the transmembrane with the N-terminal region of CXCR3 region as of the CXCR3; simulation then, progressed the chemokine towards was the orientated active to the transmembraneconformation of regionthe receptor. of CXCR3 as the simulation progressed towards the active conformation of the receptor.To verify if the conformational changes of CXCR3 induced by the presence of the chemokines contributedTo verify if athe functional conformational state of the changesreceptor, ofanalyses CXCR3 of the induced rotation by and the interactions presence of the the TMs, chemokines the contributedpresence aor functional absence of the state arginine of the cage, receptor, and in analysesteractions between of the rotation each chemokine and interactions and CXCR3 of every the TMs, the presence100 ns were or absenceperformed. of theThe arginineconformation cage, at and0 ns interactionswas considered between as a reference; each chemokine this structure and was CXCR3 obtained after the minimization and equilibrium process of the system in the CG-MD simulation. The every 100 ns were performed. The conformation at 0 ns was considered as a reference; this structure presence or absence of the interaction between D148-R149 forming the arginine cage, which is related was obtained after the minimization and equilibrium process of the system in the CG-MD simulation. to the activation of the GPCRs, was present in different conformations of the CG-MD simulations The presence(Table 3). or absence of the interaction between D148-R149 forming the arginine cage, which is related to the activation of the GPCRs, was present in different conformations of the CG-MD simulations (Table3). Table 3. Presence of the so-called arginine cage in the CG-MD simulation.

Time Frame (ns) CXCR3_CXCL9 CXCR3_CXCL10 CXCR3_CXCL11 Table 3. Presence0 of the so-calledX arginine cageX in the CG-MD X simulation. 100 X X Time Frame200 (ns) CXCR3_CXCL9X CXCR3_CXCL10 CXCR3_CXCL11 X 0300 X X X X 100400 X X 200500 X X 300600 X X 700 X X 400 X 800 X 500 X 900 600 1000 X 700 X X 800 X 900 1000 X Molecules 2020, 25, 4413 14 of 30 Molecules 2020, 25, x FOR PEER REVIEW 14 of 30

2.4.1. Conformational Changes of the Transmembrane SegmentsSegments (TMs)(TMs) The rotation of of TM1, TM1, TM3, TM3, TM5, TM5, TM6, TM6, and and TM7 TM7 was was important important to confirm to confirm the theactivation activation of the of theGPCRs; GPCRs; in addition, in addition, the the hydrogen hydrogen bonds bonds among among th themem changed changed in in the the presence presence of of agonists and antagonists [[39–41].39–41]. The CXCR3 transmembrane segm segments,ents, identified identified by UniProt and considered in this work, are presentedpresented in TableTable4 4[ [58].58]. AllAll polarpolar andand hydrophobichydrophobic interactionsinteractions are providedprovided in thethe Supplementary Information.

Table 4.4. Transmembrane segmentssegments inin CXCR3.CXCR3. TM1 TM2 TM3 * TM4 TM5 * TM6 TM7 TM1 TM2 TM3 * TM4 TM5 * TM6 TM7 54–80 90–110 126–147 170–189 213–233 256–277 299–321 * Some residues54–80 were 90–110 included in 126–147 the analysis; 170–189 in TM3, the 213–233 residues 148–150 256–277 are the 299–321 DRY motif, and *the Some residue residues 234 were and included 235 are inrelevant the analysis; in the in TM5 TM3, rotation. the residues 148–150 are the DRY motif, and the residue 234 and 235 are relevant in the TM5 rotation. In the CXCR3-CXCL9 CG-MD simulation, changes in the rotation angles of the TMs were observed.In the TM2, CXCR3-CXCL9 TM3, TM4, CG-MD and TM7 simulation, were maintained changes inwith the smaller rotation rotations angles of of the TM1, TMs TM5, were and observed. TM6. TM2,Drastic TM3, changes TM4, in and the TM7 TM wererotations maintained occurred with in the smaller 600–1000-ns rotations range. of TM1, The TM5, rotations and TM6. of TM5 Drastic and changesTM6 at 600 in theand TM 700 rotations ns agreed occurred with the in loss the 600–1000-nsand regeneration range. of The the rotations arginine ofcage, TM5 as and seen TM6 in Figure at 600 and16. TM1, 700 ns TM5, agreed and with TM6 the presented loss and regenerationintense rotations of the from arginine 600 ns, cage, which as seen corresponded in Figure 16 to. TM1,the loss TM5, of andthe arginine TM6 presented cage in intense the simulation; rotations fromhowever, 600 ns, this which interaction corresponded was regenerated to the loss at of the700 arginineand 1000 cage ns, incorresponding the simulation; to however,the rotations this of interaction TM5 and was TM6 regenerated at 700 ns and at 700 TM1 and and 1000 TM5 ns, at corresponding 1000 ns. The topolar the rotationsinteractions of TM5between and CXCL9 TM6 at and 700 CXCR3 ns and TM1(Table and S6) TM5 increased at 1000 as ns.the Thesimulati polaron interactions progressed, between and the CXCL9most frequent and CXCR3 residues (Table were S6) increased lysine types, as the agreeing simulation with progressed, previous and reports the most on frequentthe relevance residues of werepositively lysine charged types, agreeing residues with in previousthe chemokines reports onfor the the relevance interaction of positively with CXCR3. charged The residues interactions in the chemokinesbetween the forTMs the increased interaction during with CXCR3. the simulation The interactions. The residue between R216 the in TMs TM5 increased was reported during as the a simulation.relevant residue The residue for CXCR3 R216 inactivity TM5 was[59]; reported in this simulation, as a relevant R216 residue was for kept CXCR3 in constant activity [interaction59]; in this simulation,between N132 R216 (TM3) was and kept D186 in constant (TM4) interaction(Table S7). between N132 (TM3) and D186 (TM4) (Table S7).

Figure 16.16. RotationRotation of of the the transmembrane transmembrane segments segments (TMs) (TMs) from from the CG-MD simulation of CXCR3_CXCL9. TheThe rotationsrotations ofof thethe TMs TMs throughout throughout the the simulation simulation are are seen, seen, with with intense intense rotations rotations of TM1,of TM1, TM5, TM5, and and TM6. TM6. In In blue blue is shownis shown the the initial initial conformation; conformation; in red,in red, the the conformation conformation at 800at 800 ns. ns.

Regarding thethe CXCR3-CXCL10 CXCR3-CXCL10 complex, complex, the receptorthe rece exhibitedptor exhibited different differ behaviorsent behaviors in the presence in the ofpresence CXCL10 of comparedCXCL10 compared to the binding to the of binding CXCL9, of and CXCL9, all the and TMs all had the higher TMs rotationshad higher throughout rotations thethroughout simulation, the assimulation, seen in Figure as seen 17 .in The Figure rotation 17. The of TM2rotation underwent of TM2 underwent a drastic change a drastic at 500 change ns and at continued500 ns and with continued relevant with changes relevant during changes the subsequent during the nanoseconds; subsequent nanoseconds; TM1, TM3, and TM1, TM7 TM3, showed and similarTM7 showed behaviors similar with behaviors intense with rotations intense between rotations 300 between and 600 300 ns, and and 600 from ns, 500 and ns, from TM5 500 and ns, TM6TM5 underwentand TM6 underwent relevant changes.relevant changes. The greatest The changesgreatest changes in the rotation in the rotation of the TMs of the occurred TMs occurred after 300 after ns, 300 ns, matching with the absence of the arginine cage. Furthermore, after 300 ns, CXCL10 reached an equilibrium in terms of the distance from the center of the mass of CXCR3. The polar and Molecules 2020, 25, 4413 15 of 30

matchingMolecules 2020 with, 25, xthe FOR absence PEER REVIEW of the arginine cage. Furthermore, after 300 ns, CXCL10 reached15 of an 30 equilibrium in terms of the distance from the center of the mass of CXCR3. The polar and hydrophobic interactionshydrophobic between interactions CXCL10 between and CXCR3 CXCL10 (Table and S8) CXCR3 increased (Table from S8) 300 increased ns, and interactions from 300 betweenns, and theinteractions R197 residue between of CXCR3 the R197 and residue A64 and of E71CXCR3 of CXCL10 and A64 were and formedE71 of CXCL10 at 600 ns. were R197 formed is an important at 600 ns. residueR197 is an for important binding CXCL10 residue andfor binding CXCR3 activity.CXCL10 and CXCR3 activity.

Figure 17. Rotation of the TMs from the CG-MD simulationsimulation of CXCR3-CXCL10.CXCR3-CXCL10. The rotations of the TMs throughout the simulation are seen, with intense rotations of TM1,TM1, TM2, TM5, and TM6. In blue, the initial conformationconformation isis shown;shown; inin redred isis thethe conformationconformation atat 700700 ns.ns.

The interactions between the transmembrane segments were lower compared to the presence of CXCL9; additionally, no interactions of R216 with TM3 and TM4 were present in any of the frames analyzed, butbut thethe interactions interactions between between N132 N132 and and Q219 Q219 and and D186 D186 and and Q219 Q219 were were maintained maintained (TM3 (TM3 and TM4and TM4 with TM5,with respectively).TM5, respectively). The interaction The intera betweenction between D186 and D186 Q219 and was Q219 maintained was maintained practically throughoutpractically thethroughout simulation; the on simulation; the contrary, on the the interaction contrary, between the interaction N132 and between Q219 occurred N132 constantlyand Q219 fromoccurred 300 nsconstantly and was from lost at 300 700 ns ns, and coinciding was lost with at 700 the ns, regeneration coinciding of with the argininethe regeneration cage. This of may the suggestarginine thatcage. the This TM4-TM5 may suggest interaction that the between TM4-TM5 D186 interaction and Q219 between is necessary D186 and for CXCR3Q219 is necessary to have a stablefor CXCR3 conformation to have a of stable the transmembrane conformation of segments, the transmembrane and this interaction segments, acts and as athis stabilization interaction point acts betweenas a stabilization the active point and inactivebetween states. the active By comparison, and inactive TM2 states. constantly By comparison, interacted withTM2 TM3,constantly TM4, andinteracted TM7 from with 300TM3, ns, TM4, and theand interactions TM7 from 300 with ns, TM3 and andthe interactions TM4 were lost with at TM3 700 ns,and coinciding TM4 were with lost theat 700 regeneration ns, coinciding of the with arginine the regeneration cage. Therefore, of th TM2e arginine is relevant cage. for Therefore, stabilization TM2 of is the relevant functional for conformationstabilization of of the CXCR3 functional (Table conformation S9). of CXCR3 (Table S9). In the simulation of the CXCR3-CXCL11 complex,complex, the CXCR3 transmembrane segments, by the stimulation of CXCL11, rotated with greater intensityintensity compared to the stimulation of CXCL9 and CXCL10 (Figure 1818).). At At 200 ns, the TMs (with the exceptionexception of TM3) rotated intensely, concurringconcurring with the loss of the arginine cage, and TM3, TM4, and TM7 maintained a similar behavior throughout the simulation without without major major changes, changes, with with a relevant a relevant peak peak at 600 at 600 ns. ns.By comparison, By comparison, TM5 TM5and TM6 and TM6presented presented drastic drastic changes changes from 200 from ns, 200 in ns,agreement in agreement with the with loss the of lossthe arginine of the arginine cage from cage 200 from ns 200and nsregeneration and regeneration at 400 at and 400 and800 800ns. TM2 ns. TM2 underwent underwent a relevant a relevant rotation rotation during during the the simulation, simulation, changing before the regeneration of the arginine cage at 300 andand 700700 ns.ns. A change in the rotation of the transmembrane segmentssegments occurredoccurred atat 600600 ns, wherewhere TM1, TM5, and TM6 had greater magnitudes concerning TM2, TM4, TM3, and TM7. Similarly, the polar interactionsinteractions between CXCL11 and CXCR3 increased from 200 ns (Table(Table S10)S10) butbut decreaseddecreased beforebefore andand duringduring thethe regenerationregeneration of thethe argininearginine cage (300, 400, 700, and 800800 ns).ns). From 100 ns, interactionsinteractions existed between R197 and E293 of CXCR3 with residues ofof CXCL11.CXCL11. These residuesresidues are relevant for the bindingbinding of CXCL11, and the interaction of R197 waswas maintainedmaintained throughoutthroughout the the simulation. simulation. In In contrast, contrast, the the interactions interactions with with E293 E293 were were lost lost at 500at 500 and and 900 900 ns, ns, matching matching with with the the 100-ns 100-ns conformation conformation after after the the regeneration regeneration of of the the arginine arginine cage cage at at 400 and 800 ns. This indicates that the behavior of the DRY segment possibly influences the generation of interactions between the chemokine and E293. Interactions of the transmembrane segments increased from 200 ns. In the case of the stimulation of CXCL11 in CXCR3, the interactions between TM3 and TM5 through N132-Q219 were not present in any of the analyzed time frames. However, interactions between TM3 and TM6 through G138- Molecules 2020, 25, x FOR PEER REVIEW 16 of 30

Molecules 2020, 25, 4413 16 of 30 W268 were observed after 200 ns, and the TM4-TM5 interactions between D186 and Q219 were maintained during the simulation. TM2 showed interactions with TM3, TM4, and TM7 after 200 ns, and400 andthere 800 was ns. an This interaction indicates between that the behaviorthe DRY ofsegment the DRY with segment Y233 possiblyof TM5 and influences interaction the generation between C234of interactions of intracellular between loop the 3 (Table chemokine S11). and E293.

FigureFigure 18. 18. RotationRotation of of the the TMs TMs from from the the CG-MD CG-MD simulati simulationon of of CXCR3_CXCL11. CXCR3_CXCL11. Rotations Rotations of of the the TMs TMs throughoutthroughout the simulation areare present,present, withwith relevant relevant rotations rotations of of TM1, TM1, TM2, TM2, TM5, TM5, and and TM6, TM6, starting starting at at200 200 ns. ns. In In blue, blue, the the initial initial conformation conformation is is shown; shown; in in red red is is the the conformation conformation at at 600 600 ns. ns.

TransmembraneInteractions of the segment transmembrane 2 is relevant segments for increased the conformational from 200 ns. chan In theges case of of CXCR3 the stimulation by the stimulationof CXCL11 inof CXCR3, CXCL10 the and interactions CXCL11 betweenbut not TM3of CX andCL9. TM5 In throughthe presence N132-Q219 of CXCL9, were R216 not present in TM5 in continuedany of the to analyzed interact time with frames. N132 (TM3) However, and D186 interactions (TM4); betweenin contrast, TM3 with and the TM6 stimulation through G138-W268of CXCL10 andwere CXCL11, observed the after interaction 200 ns, and between the TM4-TM5 N132 and interactions R216 was between not seen D186 during and Q219 the weresimulation. maintained The networkduring the of interactions simulation. between TM2 showed TM3, interactionsTM4, and TM5 with appears TM3, TM4,to be necessary and TM7 afterfor the 200 stabilization ns, and there of thewas functional an interaction conformation between theof CXCR3, DRY segment specifically with Y233N132-D186-R216-Q of TM5 and interaction219 for CXCL9 between and C234 N132- of D186-Q219intracellular for loop CXCL10 3 (Table and S11). CXCL11. Interactions between TM3 and TM6 were present due to the stimulationTransmembrane of CXCL11 segment through 2 is G138-W268, relevant for theconcurri conformationalng with the changes loss of of the CXCR3 arginine by the cage stimulation and the rotationof CXCL10 of the and transmembrane CXCL11 but notsegments. of CXCL9. These In resu thelts presence reinforced of CXCL9,the fact that R216 the in CXCR3 TM5 continued receptor underwentto interact withconformational N132 (TM3) changes and D186 that promoted (TM4); in acti contrast,vation within the the presence stimulation of the of chemokines CXCL10 and in ourCXCL11, simulations, the interaction proving between the first N132step of and receptor R216 was activation not seen of duringthe chemokines. the simulation. The network of interactionsTo support between the results TM3, TM4,of the and CXCR3/chemokine TM5 appears to be CG-MD necessary simulations, for the stabilization a 1-µs CG-MD of the simulation functional ofconformation the CXCR3/GP of CXCR3,complex specifically was performed, N132-D186-R216-Q219 and the rotation angles for CXCL9 and polar and interactions N132-D186-Q219 between for theCXCL10 transmembrane and CXCL11. segments Interactions were between analyzed. TM3 Al andl polar TM6 interactions were present between due to thethe stimulation TMs in the of CXCR3/GPCXCL11 through complex G138-W268, simulation concurring are listed in with Table the S12. loss Unexpectedly, of the arginine the cage transmembrane and the rotation segments of the remainedtransmembrane in constant segments. movement These (Figure results S5) reinforced compared theto when fact thata chemokine the CXCR3 was receptorpresent, where underwent only someconformational of the TMs changes had relevant that promoted rotations. activation Additiona inlly, the a presenceconstant interaction of the chemokines between in TM3, our simulations, TM4, TM5 andproving TM6 the with first TM7 step was of receptor present activation through ofthe the interact chemokines.ions of residues A127-S304, L130-Y308, and N134-Y308To support between the resultsTM3 and of theTM7, CXCR3 D186-K300/chemokine between CG-MD TM4 simulations,and TM7, Q219-K300 a 1-µs CG-MD between simulation TM5 and of TM7the CXCR3 in the /firstGP complex400 ns, and was Y271-K300 performed, between and the TM6 rotation and anglesTM7. These and polar results interactions may suggest between that the CXCR3transmembrane receptor segmentswithout stimulation were analyzed. of any All of polarthe chemokines interactions is between in constant the TMsmovement, in the CXCR3and when/GP thecomplex chemokine simulation is present, are listed these in movements Table S12. Unexpectedly, are diminished, the stabilizing transmembrane the TM segments rotations remained towards ina meta-activeconstant movement conformation. (Figure The S5) TM7 compared transmembrane to when a segment chemokine could was act present, as a central where axis only of somean inactive of the state,TMshad since relevant the main rotations. interactions Additionally, with TM3, a constant TM4, interactionTM5 and TM6 between with TM3, Y308 TM4, and TM5K300 and are TM6between with relevantTM7 was residues present throughfor the thestabilization interactions conformati of residuesons A127-S304, of the meta-active L130-Y308, and and active N134-Y308 states between of the receptorTM3 and as TM7, the D186-K300N132-D186-Q219 between interaction TM4 and TM7,network. Q219-K300 between TM5 and TM7 in the first 400 ns, and Y271-K300To assess if betweenthese conformational TM6 and TM7. changes These resultspromot maye the suggest second thatactivation the CXCR3 step of receptor CXCR3, without a 250- nsstimulation AA-MD simulation of any of the was chemokines performed, is taking in constant cluster1 movement, of the CXCR3-CXCL10 and when the CG-MD chemokine simulation is present, as thethese starting movements structure. are diminished, stabilizing the TM rotations towards a meta-active conformation. The TM7 transmembrane segment could act as a central axis of an inactive state, since the main interactions with TM3, TM4, TM5 and TM6 with Y308 and K300 are between relevant residues for the Molecules 2020, 25, 4413 17 of 30 stabilization conformations of the meta-active and active states of the receptor as the N132-D186-Q219 interaction network. To assess if these conformational changes promote the second activation step of CXCR3, a 250-ns AA-MDMolecules 2020 simulation, 25, x FORwas PEER performed, REVIEW taking cluster1 of the CXCR3-CXCL10 CG-MD simulation17 as of the 30 starting structure. 2.4.2. Second Activation Step of the CXCR3-CXCL10 Complex 2.4.2.Chemokine Second Activation CXCL10 Step is relevant of the CXCR3-CXCL10 in vitiligo due to Complex an increased concentration in the serum and skin ofChemokine patients. Thus, CXCL10 this ischemokine relevant inis vitiligoan import dueant to marker an increased to assess concentration the progression in the of vitiligo. serum and To verifyskin of the patients. second Thus, activation this chemokine step of CXCR3, is an important a 250-ns markersimulation to assesswas perf theormed, progression preserving of vitiligo. the disulfideTo verify bridges the second and activationadding the step post-translational of CXCR3, a 250-ns modifications simulation present was performed, in CXCR3. preservingUsing cluster1 the ofdisulfide the CG-MD bridges simulation and adding at 787.9 the post-translational ns, which was converted modifications to the presentall-atom in system CXCR3. with Using the cluster1Martini protocol,of the CG-MD as a simulationstarting structure, at 787.9 ns,which which is wasa state converted before tothe the active all-atom conformation system with of theCXCR3, Martini it representsprotocol, as the a starting interaction structure, between which D148 is aan stated R149, before forming the active the conformationarginine cage, of and CXCR3, Gi/0 was it represents added in +2 the interactioncomplex with between GTP D148and Mg and. R149, The conformations forming the arginine present cage, every and 25 G ins/0 was were added analyzed, in the with complex the withconformation GTP and at Mg 0 +ns2. obtained The conformations after the minimizati present everyon and 25 nsequilibrium were analyzed, proces withses of the the conformation system taken at as0 ns a reference. obtained after the minimization and equilibrium processes of the system taken as a reference. The rotation of TM5 and TM6 was observed after 50 ns, and there was a change in the rotation at 175 ns, where the TMs converged, with a greater intensityintensity of TM2,TM2, TM3,TM3, andand TM5.TM5. TM7 showed an intense rotation at 50 ns, after maintaining a beha behaviorvior with discrete rotations that coincided with the rotation rotation of of TM2, TM2, TM3, TM3, and and TM5 TM5 at at 175 175 ns, ns, as as seen seen in inFigure Figure 19. 19 Compared. Compared with with the therotation rotation in the in CG-MDthe CG-MD simulation, simulation, the thetransmembrane transmembrane segments segments presented presented a adiscrete discrete rotation, rotation, although although with similar behaviors of TM2, TM3, TM4, TM5, and TM6. The The rotation rotation of TM2 TM2 was was necessary necessary for the active active conformation of the receptor.

Figure 19. Figure 19. RotationRotation of of TMs TMs from from the the 250-ns 250-ns AA-MD AA-MD simulation simulation of of CXCR3-CXCL10-G CXCR3-CXCL10-Gi/0.i /0Rotations. Rotations of theof the TMs TMs throughout throughout the the simulation simulation are are presented, presented, with with the the relevant relevant rotations rotations of of TM1, TM1, TM2, TM2, TM3, TM3, TM5, and TM6. TM7 TM7 has has an intense rotation at 50 ns ns.. In In blue, the initial conformation; in in red is the conformation at 600 ns. The polar interactions between CXCL10 and CXCR3 decreased compared to the 0 ns between the The polar interactions between CXCL10 and CXCR3 decreased compared to the 0 ns between different time frames analyzed; however, the hydrophobic interactions increased as the simulation the different time frames analyzed; however, the hydrophobic interactions increased as the progressed. Post-translational modifications are relevant for the interaction of CXCL10 with CXCR3, simulation progressed. Post-translational modifications are relevant for the interaction of CXCL10 and the interaction between glycosylation in N32 and K87 of CXCL10 occurred at 25 ns. Additionally, with CXCR3, and the interaction between glycosylation in N32 and K87 of CXCL10 occurred at 25 ns. interactions with phosphorylation in Y27 and K87-K88 occurred from 50 ns to 175 ns (Table S12). Additionally, interactions with phosphorylation in Y27 and K87-K88 occurred from 50 ns to 175 ns The interactions of the transmembrane segments changed in the presence of a functional GP. (Table S12). The most relevant were the loss of the interaction of TM3-TM5 between N132 and Q219, which may be The interactions of the transmembrane segments changed in the presence of a functional GP. due to the influence of G . The interactions of TM3-TM4 between N132 and D186 were maintained at The most relevant were thei/0 loss of the interaction of TM3-TM5 between N132 and Q219, which may different points of the simulation; at the same time, an interaction between TM3 and TM6 was formed be due to the influence of Gi/0. The interactions of TM3-TM4 between N132 and D186 were maintained at different points of the simulation; at the same time, an interaction between TM3 and TM6 was formed through N134-W268. From 50 to 150 ns, interactions of TM5 and TM6 between R216 and Q219 with Y271 were present. These changes promoted the stabilization of the CXCR3-CXCL10-Gi/0 complex in the first 100 ns of the simulation (Table S14). Once the conformation of CXCR3 and GP stabilized, separation of the βγ complex from the αi/0 subunit of GP began at 175 ns, coinciding with the constant changes in the rotation of the Molecules 2020, 25, 4413 18 of 30 through N134-W268. From 50 to 150 ns, interactions of TM5 and TM6 between R216 and Q219 with Y271 were present. These changes promoted the stabilization of the CXCR3-CXCL10-Gi/0 complex in theMolecules first 1002020, ns 25, of x FOR the PEER simulation REVIEW (Table S14). 18 of 30 Once the conformation of CXCR3 and GP stabilized, separation of the βγ complex from the αi/0 subunittransmembrane of GP began segments at 175 ns, until coinciding the end with of the the simula constanttion changes (Figure in 20). the Moreover, rotation of thethetransmembrane presence of the segmentsN132-D186 until and the N134-W268 end of the interactions simulation (Figureand the 20 in).teraction Moreover, between the presence R149 of of CXCR3 the N132-D186 with R193 and of N134-W268the alpha subunit interactions was present and the in interaction the simulation between. Additionally, R149 of CXCR3 the withinteraction R193 of between the alpha N132 subunit and wasD186 present and N134 in the and simulation. W268 stabilized Additionally, the active the co interactionnformation betweenof CXCR3 N132 and andpromoted D186 andthe interaction N134 and W268between stabilized R149 of the CXCR3 active with conformation R193 of the of α CXCR3 subunit and necessary promoted for the GP interaction activation between(Table S15). R149 This of CXCR3result confirmed with R193 that of theCXCR3α subunit presented necessary a meta-act for GPive activationstate with (Tablethe orientation S15). This of resultCXCL10 confirmed towards thatthe extracellular CXCR3 presented loops aand meta-active the rotation state of withthe transmembrane the orientation segments of CXCL10 and, towards when thecoupled extracellular with the loopsfunctional and the Gi/0 rotation, promoted of the the transmembrane conformational segmentschanges in and, the when α subunit coupled that with were the promoted functional for G thei/0, promotedseparation the of the conformational βγ-complex, changesstabilizing in the activeα subunit state that through were the promoted interactions for the between separation N132-D186 of the βγand-complex, N134-W268. stabilizing Additionally, the active it was state reported through that the interactionsthe N132 residue between is fundamental N132-D186 and for N134-W268.activity and Additionally,chemokine binding. it was reported When this that resi thedue N132 is residuemutated, is fundamentalthe expression for in activity the membrane and chemokine and activity binding. is Whendiminished this residue [60]. is mutated, the expression in the membrane and activity is diminished [60].

FigureFigure 20.20. TheThe distance distance of of complex complex proteins proteins in inthe the AA-MD AA-MD 250-ns 250-ns simulation simulation of CXCR3-CXCL10. of CXCR3-CXCL10. The βγ α Theseparation separation of the of theβγ complexcomplex to the to theα subunitsubunit occurs occurs at 175 at 175ns. ns. 2.5. General Protein–Ligand Interaction Model of CXCR3 Antagonists 2.5. General Protein–Ligand Interaction Model of CXCR3 Antagonists To provide information to support the design of new antagonists to CXCR3, a general To provide information to support the design of new antagonists to CXCR3, a general protein– protein–ligand interaction model was constructed based on 894 ligands reported as antagonists in the ligand interaction model was constructed based on 894 ligands reported as antagonists in the ZINC ZINC public database. This interaction model is similar to a pharmacophore, which is a representation public database. This interaction model is similar to a pharmacophore, which is a representation of of the chemical features necessary for the recognition of a ligand by macromolecules; in this case, the chemical features necessary for the recognition of a ligand by macromolecules; in this case, the the general protein–ligand interaction model provides these chemical features [61]. Three different general protein–ligand interaction model provides these chemical features [61]. Three different conformations of CXCR3 were used—the initial structure (F0) of CXCR3-CXCL10_CG-MD, the most conformations of CXCR3 were used—the initial structure (F0) of CXCR3-CXCL10_CG-MD, the most representative conformation (C1) of CXCR3-CXCL10_CG-MD, and the conformation at 175 ns (F7) of representative conformation (C1) of CXCR3-CXCL10_CG-MD, and the conformation at 175 ns (F7) of CXCR3-CXCL10_AA-MD—to observe if the antagonists could bind before and after the two receptor CXCR3-CXCL10_AA-MD—to observe if the antagonists could bind before and after the two receptor activation steps occurred; thus, CXCL10 was removed from each complex. Once the molecular docking activation steps occurred; thus, CXCL10 was removed from each complex. Once the molecular complexes were obtained, the interactions present in the complexes were grouped for the construction docking complexes were obtained, the interactions present in the complexes were grouped for the of the general protein–ligand interaction model to observe the chemical features and three-dimensional construction of the general protein–ligand interaction model to observe the chemical features and three-dimensional position, which are most relevant for the interaction of antagonists with the three conformations of CXCR3. Three general protein–ligand interaction models were built, one for each CXCR3 conformation.

Molecules 2020, 25, 4413 19 of 30 position, which are most relevant for the interaction of antagonists with the three conformations of MoleculesCXCR3. 2020 Three, 25, generalx FOR PEER protein–ligand REVIEW interaction models were built, one for each CXCR3 conformation.19 of 30

2.5.1.2.5.1. General General Protein–Ligand Protein–Ligand Interaction Interaction Mo Modeldel of of Conformation Conformation F0 F0 of of CXCR3-CXCL10_CG-MD CXCR3-CXCL10_CG-MD InIn this this structure, structure, CXCR3 CXCR3 presents presents an an inactive inactive conformation conformation and and CXCL10 CXCL10 interacts interacts through through the the firstfirst residues ofof thethe extracellular extracellula regionr region of CXCR3.of CXCR3. The The position position of the of hydrogenthe hydrogen bond acceptorsbond acceptors (HAc), (HAc),hydrogen hydrogen bond donors bond (HDn),donors and(HDn), positive-charged and positive-charged (Pin)-type (Pin)-type pharmacophoric pharmacophoric elements inelements the region in theof theregion extracellular of the extracellular loops, and a hydrophobicloops, and a (Hph)-type hydrophobic pharmacophoric (Hph)-type elementpharmacophoric between TM5element and betweenTM6, are TM5 shown and in FigureTM6, are 21. shown Residues in aroundFigure 521. Å Residues for each pharmacophoric around 5 Å for elementeach pharmacophoric are relevant for elementthe interaction are relevant and antagonisticfor the interact activityion and of antagonistic the ligands withactivity CXCR3 of the (Table ligands5). with CXCR3 (Table 5).

FigureFigure 21. 21. GeneralGeneral protein–ligand protein–ligand interaction interaction model model of of F0 F0 CXCR3-CXCL10_CG-MD. CXCR3-CXCL10_CG-MD. In In magenta: magenta: hydrogenhydrogen bond bond acceptors acceptors (HAc), (HAc), gray: gray: hydrogen hydrogen bo bondnd donors donors (HDn), (HDn), yello yellow:w: hydrophobic hydrophobic groups groups (Hph),(Hph), and and blue: blue: positive-charged positive-charged groups groups (Pin). (Pin).

TableTable 5. 5. PharmacophoricPharmacophoric elements elements of of F0 F0 CXCR3-CXCL10_CG-MD. CXCR3-CXCL10_CG-MD.

PharmacophoricPharmacophoric Element Element Residues (5 Å) A192, H194, D195, R212, M281, D282, G284, L286, A287, HAcHAc A192, H194, D195, R212, M281, D282, G284, L286, A287, R288 HDn D195R288, E196 , R197, N289, C290 HphHDn Y271,D195 H272,, E196 L273,, R197 V275,, N289, L276, C290 I279 PInHph H194 Y271,, D195 H272,, E196 L273,, R197 V275,, R288 L276, I279 Bold is used to showPIn the residues that areH194, repeated D195, in the E196, different R197, pharmacophoric R288 elements. HAc: hydrogen bond Boldacceptors, is used HDn: to hydrogenshow the bondresidues donors, that Hph: are hydrophobicrepeated in groups,the different and Pin: pharmacophoric positive-charged groups.elements. HAc: hydrogen bond acceptors, HDn: hydrogen bond donors, Hph: hydrophobic groups, and Pin: positive- chargedThe residues groups. R197 and R212 are found around the pharmacophoric elements and are relevant for the binding and activity of all chemokines; D282 is relevant for the binding and activity of CXCL9 andThe CXCL10. residues Moreover, R197 and residues R212 aroundare found the around CWTP sequence,the pharmacophoric a conserved elements sequence and in theare GPCRs relevant of forclass the A binding (CWxP), and which activity are residues of all chemokines; related to theD282 rotation is relevant of TM6 for [ 40the] necessarybinding and for activity CXCR3 of activation, CXCL9 andare closeCXCL10. to the Moreover, Hph-type residues pharmacophoric around the element. CWTP sequence, This general a cons protein–liganderved sequence interaction in the GPCRs model, of class A (CWxP), which are residues related to the rotation of TM6 [40] necessary for CXCR3 activation, are close to the Hph-type pharmacophoric element. This general protein–ligand interaction model, calculated from antagonists docking to an inactive CXCR3 conformation, suggest that these antagonists could hinder the interactions among the chemokines and the residues of the extracellular loops and, subsequently, inhibit the rotation of TM6. Molecules 2020, 25, 4413 20 of 30 calculated from antagonists docking to an inactive CXCR3 conformation, suggest that these antagonists could hinder the interactions among the chemokines and the residues of the extracellular loops and, subsequently,Molecules 2020, 25 inhibit, x FOR PEER the rotation REVIEW of TM6. 20 of 30

2.5.2.2.5.2. General Protein–Ligand InteractionInteraction ModelModel ofof ConformationConformation C1C1 CXCR3-CXCL10_CG-MDCXCR3-CXCL10_CG-MD InIn thisthis structure,structure, CXCR3CXCR3 isis inin aa pre-activepre-active conformationconformation wherewhere thethe firstfirst stepstep ofof activationactivation ofof thethe receptorsreceptors toto thethe chemokineschemokines hashas alreadyalready occurred.occurred. LikeLike thethe F0F0 generalgeneral protein–ligandprotein–ligand interactioninteraction model,model, pharmacophoricpharmacophoric elementselements areare concentratedconcentrated inin thethe extracellularextracellular loopsloops andand thethe regionregion betweenbetween TM5TM5 andand TM6TM6 (Figure(Figure 2222).). The Hph-type pharmacophoric element is positioned close to residue T269, thethe residueresidue secondsecond toto W268W268 inin importanceimportance forfor TM6TM6 rotation;rotation; similarly,similarly, residuesresidues relevantrelevant forfor CXCR3CXCR3 activation,activation, suchsuch asas R197R197 andand R212R212 (Table(Table6 ),6), are are found found around around the the elements elements of of HAc, HAc, HDn, HDn, and and Hph. Hph.

Figure 22. General protein–ligand interaction model of C1 CXCR3-CXCL10_CG-MD. In magenta: HAc, Figure 22. General protein–ligand interaction model of C1 CXCR3-CXCL10_CG-MD. In magenta: gray: HDn, yellow: Hph, and blue: PIn. HAc, gray: HDn, yellow: Hph, and blue: PIn. Table 6. Pharmacophoric elements of C1 CXCR3-CXCL10_CG-MD. Table 6. Pharmacophoric elements of C1 CXCR3-CXCL10_CG-MD. Pharmacophoric Element Residues (5 Å) PharmacophoricHAc Element R197, L198, P208, Q209, ResiduesF224, T269 (5, Å)H272 , L273, L276, D282 HAcHDn R197 D195,, L198E196,, P208R197, ,Q209 N206,, F224 F207,, T269R212, H272, L273, L276, D282 HDnHph D195,E196, R197E196,, R197L198, P208N206,, Q209 F207,, R212, F224, T269, H272, L273 Hph E196, R197, L198, P208, Q209, R212, F224, T269, H272, L273 Bold is used to show the residues that are repeated in the different pharmacophoric elements. Bold is used to show the residues that are repeated in the different pharmacophoric elements. 2.5.3. General Protein–Ligand Interaction Model of Conformation F7 CXCR3-CXCL10_AA-MD 2.5.3. General Protein–Ligand Interaction Model of Conformation F7 CXCR3-CXCL10_AA-MD This structure corresponds to the active state of the receptor once the chemokine is oriented towardsThis the structure TMs, causing corresponds the conformational to the active changesstate of inthe the receptor TMs for once the couplingthe chemokine of the GPis oriented and the subsequenttowards the separation TMs, causing of the theβγ conformationalsubunit. In this change conformation,s in the TMs the generalfor the coupling protein–ligand of the GP interaction and the modelsubsequent can be separation divided into of the two: βγ first, subunit. the pharmacophoric In this conformation, elements the general are positioned protein–ligand in the region interaction of the model can be divided into two: first, the pharmacophoric elements are positioned in the region of the extracellular loops, and second, the pharmacophoric elements are in TM2, TM3, and TM4. Additionally, the latter presented an aromatic rings (Arm) and negative-charged (Nin)-type pharmacophoric element that was absent in the previous general interaction models (Figure 23). Molecules 2020, 25, 4413 21 of 30 extracellular loops, and second, the pharmacophoric elements are in TM2, TM3, and TM4. Additionally, the latter presented an aromatic rings (Arm) and negative-charged (Nin)-type pharmacophoric element that was absent in the previous general interaction models (Figure 23). Molecules 2020, 25, x FOR PEER REVIEW 21 of 30

FigureFigure 23. 23.General General protein–ligand protein–ligand interaction interaction model model of F7 of CXCR3-CXCL10_CG-MD. F7 CXCR3-CXCL10_CG-MD. In magenta: In magenta: HAc, gray:HAc, HDn, gray: yellow: HDn, yellow: Hph, green: Hph, aromaticgreen: aromatic rings (Arm), rings (Arm), and red: and negative-charged red: negative-charged groups groups (Nin). (Nin).

TheThe residues residues around around the pharmacophoricthe pharmacophoric elements elemen werets diwerefferent different from the from residues the ofresidues the previous of the generalprevious interaction general interaction models. However, models. theHowever, relevant the residues relevant for residues the activation for the activation of CXCR3 of are CXCR3 listed inare Tablelisted7, suchin Table as L198 7, such and as N199, L198 and residues N199, close residues to R197, close which to R197, is essential which is for essential the binding for the and binding activity and ofactivity chemokines. of chemokines. Thus, antagonist Thus, antagonist binding inbinding the zone in the between zone between TM2, TM3, TM2, and TM3, TM4, and shown TM4, by shown the calculatedby the calculated general general protein–ligand protein–ligand interaction interaction model, model, could becould related be related to the to destabilization the destabilization of the of interactionsthe interactions of TM3 of withTM3 thewithα thesubunit α subunit of GP. of GP.

Table 7. Pharmacophoric elements of F7 CXCR3-CXCL10_AA-MD. Table 7. Pharmacophoric elements of F7 CXCR3-CXCL10_AA-MD.

PharmacophoricPharmacophoric Element Element Residues Residues (5 (5 Å) Å) ArmArm A129A129, I133, I133, L180,, L180,A183 A183, L184,, L184,F187 F187 HAcHAc V126V126, A129, A129, L130, L130, I133, I133, A183,, A183,F187 F187, L198, L198, N199, N199, L286, L286, A287,, A287,R288 R288 HDnHDn S36,S36,L198 L198, N199, N199, L286, L286, R288, R288, N289, N289 HphHph V136V136, A129, A129, L130, L130, I133, I133, A183, A183, F187, F187 NInNIn E31,E31L286, L286,, A287, A287R288, R288 BoldBold is usedis used to show to show the residues the residues that are that repeated are repeat in the diedff erentin the pharmacophoric different pharmacophoric elements. Arm: elements. aromatic rings Arm: and Nin: negative-charged groups. aromatic rings and Nin: negative-charged groups.

ComparingComparing the the three three general general interaction interaction models, models, the the pharmacophoric pharmacophoric elements elements HAc, HAc, HDn, HDn, and and HphHph are are the the most most relevant relevant for thefor interactionthe interaction and activityand activity with CXCR3.with CXCR3. The three The general three general protein–ligand protein– interactionligand interaction models dimodelsffer to differ some to extent, some principallyextent, principally between between F0 and F0 F175 and ns, F175 suggesting ns, suggesting that the that the conformation that occurs in CXCR3 is important in antagonist binding. The general interaction models of F0 and C1 are more relevant, because they can avoid the coupling of the GP, stopping the changes in the first activation step of CXCR3 and inhibiting the chemotaxis derived by receptor stimulation. Finally, since the antagonists used for general interaction model building were obtained mostly through the HTS campaigns and subsequent optimization [42–51], the library had considerable Molecules 2020, 25, 4413 22 of 30 conformation that occurs in CXCR3 is important in antagonist binding. The general interaction models of F0 and C1 are more relevant, because they can avoid the coupling of the GP, stopping the changes in the first activation step of CXCR3 and inhibiting the chemotaxis derived by receptor stimulation. Finally, since the antagonists used for general interaction model building were obtained mostly through the HTS campaigns and subsequent optimization [42–51], the library had considerable structural diversity, with at least eight different structural scaffolds. Therefore, although the calculated general interaction models consisted of 6–8 pharmacophoric elements, they provided information on the important characteristics of the compounds, regardless of the structural scaffold, in addition to the type of major interactions and the residues relevant to the binding of the compounds. All of the above could help in the future rational designs of CXCR3 antagonists.

3. Materials and Methods

3.1. Protein Structures and Homology Modeling

The primary sequence of the full legth-CXCR3A receptor, the chemokine CXCL9, and the αi/0 subunit of the G-protein were obtained from the UniProt database [58] (codes P49682, Q07325, and P04899, respectively). The three-dimensional structures of CXCL10, CXCL11, and the G-protein βγ subunits complex were obtained from the Protein Data Bank (PDB) database with codes 1LV9 [62], 1RJT [32], and 6G79 [63], respectively. The amino acid sequences of CXCR3, CXCL9, and Gαi/0 were uploaded to the GPCR-I-TASSER server for CXCR3 and the general module I-TASSER server for CXCL9 and Gαi/0 [64] for homology model building. The CXCR3 and CXCL9 models were calculated with the best sequence identity found by the BLAST server (CCR5 and CXCL2, respectively) [65,66].

3.2. Protein-Protein Docking To build the CXCR3-chemokine complex, molecular docking was performed using the ClusPro server [67,68] with the structures from the PDB codes stated above for CXCL10 and CXCL11 and the most representative conformation of the molecular dynamics simulations (cluster1) for CXCR3 and CXCL9. Relevant residues that were reported for the interaction of chemokines with the receptor and their subsequent activation are shown in Table8. Therefore, the proximity to the first 16 residues and Y27 and Y29 of CXCR3 was considered as a restriction for molecular docking. In the case of the chemokines, no restriction was used [59,69,70].

Table 8. Relevant amino acids for the interactions between CXCR3 and its chemokines.

Residues Observation Residues 1–16 Relevant for interactions between the CXCR3 receptor and its chemokines [28]. Fundamental for chemotaxis and the mobilization of Ca+2 but not in the union of R216 chemokines [59]. Y27, Y28 Sulfation in these residues is necessary for the interactions of the chemokines [28,31,59]. N197, N212 Essential for the union of CXCL10 and CXCL11 [59]. Relevant in the interactions of the three chemokines, but only CXL10 and CXCL11 have D112, D278 an influence on the activity [59]. D282, E293 Influence on the activity of CXCL9 and CXCL10 and the interaction of CXCL11 [59].

3.3. Molecular Dynamics (MD) Simulations The preparation of all the systems for MD simulations, including the addition of post-translational modifications (disulfide bridges, glycosylations, and lipidations) present in the proteins, was undertaken on the CHARMM-GUI web server [71,72]. In the case of sulfation in CXCR3, phosphorylation was used to generate a similar electrostatic environment to the original, because the parameters of sulfation are not readily available in CHARMM-GUI [71–73]. MD simulations were performed with Gromacs 2018.7 [74,75] using an isothermic-isobaric ensemble (NPT) at 1 atm and 310.15 K, employing a temperature coupling and velocity rescaling with a stochastic term [76] and a Parrinello-Rahman Molecules 2020, 25, 4413 23 of 30 barostat [77]. The systems were submitted to 5000 steps steepest-descent minimization, followed by the 6-equilibration steps scheme by reducing the force constants suggested by the CHARMM-GUI server prior to performing the production simulations. The TIP3 water model and POPC lipids were used. The modules “distance” and “cluster” of Gromacs 2018.7 were used to measure the distance between each chemokine and CXCR3 and to find the relevant conformations of the simulation using the “gromos” algorithm [78], respectively. All graphs were built with R [79] using ggplot2 libraries [80]. All the MD simulations were performed in the ADA cluster of the National Laboratory of Advanced Scientific Visualization at Campus Juriquilla of the National Autonomous University of Mexico (LAVIS-UNAM).

3.3.1. All-Atom Molecular Dynamics (AA-MD) Simulations

Relaxation of the Homology Models Once the homology models were calculated, all systems for the molecular dynamics simulations of CXCR3 were built using the Membrane Builder module in CHARMM-GUI; CXCL9 was built using the Solution Builder module in CHARMM-GUI. In the all-atom systems, the CHARMM36 force field was used in an NPT ensemble, as stated previously. The production simulations for model relaxation were 50-ns-long for CXCR3 and 5 ns for CXCL9.

Relaxation of the Chemokine-CXCR3/GP Complexes

Once the homology models were relaxed, the Gi/0 protein was added to the conformation representative of CXCR3 simulation, and a 100-ns molecular dynamics simulation system was produced for relaxation of the CXCR3/GP complex. Next, the system was prepared for the CXCR3-chemokine-GP complexes with a 50-ns simulation to observe the evolution of the system. Interactions between CXCR3 and the chemokines were observed by LigPlot+ [81,82] and the ESBRI Server for salt bridges [83–85] using representative structures from the simulations. The RMSD of this simulation is shown in Figure S6.

3.3.2. Coarse-Grained Molecular Dynamics (CG-MD) Simulations To assess if the receptor reached an activated conformation, the systems for the molecular dynamics simulations of CXCR3/chemokines were built using the Martini Maker [86] module in CHARMM-GUI; the Martini2.2p force field was used in an NPT assemble at 1 atm and 310.15 K, and the coarse-grained simulations of 1 µs were produced. Representative structures of each simulation were analyzed in LigPlot+ and ESBRI to observe the interactions between CXCR3 and the chemokines. Finally, a 250-ns AA-MD simulation was performed; the system using the representative structure of the biggest conformation cluster of CXCR3 and CXCL10 (cluster1) of the CG simulation was built, and the GP in the complex with guanosine triphosphate (GTP) and Mg2+ was added. This system was built as previously described for the AA-MD simulations. The RMSD of these simulations are shown in Figures S6 and S7.

3.4. Ligand Docking and the General Protein–Ligand Interaction Model

3.4.1. Ligand Docking The 3D structures of 894 unique structures of reported antagonists of CXCR3 [42–51] were obtained from the ZINC public database [87], with pKi values (minus logarithm of the equilibrium dissociation constant of an antagonist) [88] ranging from 5 to 9.7. A grid of 40 32 40 Å with × × spacing of 0.375 Å, centered on the residues identified as relevant for both receptor activation and agonist binding, was calculated with AutoGrid 4.2.6 [89]. Finally, molecular docking was performed with AutoDock 4.2.6 optimized for graphics-processing units, using a total of 25 runs and 25,000,000 evaluations, with a Lamarckian genetic algorithm and Solis–Wets local search [90] Molecules 2020, 25, 4413 24 of 30 using three different conformations of CXCR3: the initial frame (F0) of the CXCR3-CXCL10 complex coarse-grained molecular dynamics simulation (CXCR3-CXCL10_CG-MD), the most representative conformation (C1) of the CXCR3-CXCL10_CG-MD simulation, and frame 7 (F7) corresponding to the 175-ns timestep of the 250-ns all-atom molecular dynamics (AA-MD) simulation of CXCR3-CXCL10.

3.4.2. Pharmacophoric Elements Identification and General Protein–Ligand Interaction Model Calculation The software Pharmer (2015, Pittsburg, PA, USA) [91] was used to identify the key pharmacophoric elements of the docked ligands in their most frequent binding conformations, obtained by clustering the docking results of each ligand with a root mean standard deviation (RMSD) cutoff of 2 Å on their protein–ligand complexes: hydrogen bond acceptors (HAc), hydrogen bond donors (HDn), aromatic rings (Arm), hydrophobic groups (Hph), positive-charged groups (PIn), and negative-charged groups (NIn) in the three-dimensional conformations of the docked ligands. These characteristics were clustered according to their nature, position, and size using a partition-around-medoids statistical method [92], implemented in the cluster package [93], available in the statistical software R (3.5.2, Vienna, Austria) [79].

3.5. Manipulation of Complexes and Figures All of the protein figures presented in this article and the alignments were made in the PyMOL program [94]. The manipulation of the protein complexes was performed using the academic version of the Schrödinger-Maestro program [95].

4. Conclusions

A homology model of CXCR3 was calculated to build the CXCR3/Gαi/0βγ complex and study the interactions with its chemokines through AA-MD and CG-MD simulations. The early interaction of CXCL9 with a glycosylation of the receptor was observed, which indicates the importance in the first interactions of CXCL9 with CXCR3. These interactions enable the conformational change of the system towards a favorable state for hydrogen bonds and hydrophobic and salt bridge interactions. CG-MD simulations showed the first activation step of CXCR3, which corresponds to the binding and orientation of the chemokines towards the transmembrane region and the conformational change of the receptor towards a functional state for the coupling of the GP. Later, the 250-ns AA-MD simulation confirmed the importance of post-translational modifications for the binding and activity of CXCL10, showing interactions with phosphorylation in Y27 throughout the simulation, and glycosylation in N32 occurred in the first 25 ns, indicating that glycosylations form a relevant interaction between the chemokine and receptor. Next the changes in the rotations of the transmembrane segments, the interaction between R149 and D193 of the α subunit, and the separation of the βγ complex confirmed the second activation step. Thus, a simulation of the process by which CXCR3 is activated by CXCL10 was obtained, in which N132-D186 and N134-W268 were the relevant interactions to stabilize the active state of the receptor in the presence of GP complexed with GTP and Mg+2. Moreover, the pharmacophoric elements HAc, HDn, Hph, and PIn are the most relevant for the binding and activity of the antagonists listed in the ZINC database: most of those are in two regions of the receptor: the extracellular loops and the TM5-TM6 region. Antagonist binding in the first region could prevent the orientation of the chemokine towards the TMs and the interaction of the chemokine with the extracellular loops, and binding in the second region could hinder the rotation of TM6 to activate the receptor. These general protein–ligand interaction models can be used to design new molecules to inhibit the first step of activation in CXCR3. Furthermore, these results allow a better understanding of the activation of two steps of CXCR3, the involved residues, and the rotation of transmembrane segments, in agreement with the reported experimental data. This information is relevant for the design of molecules that can prevent the binding of chemokines to CXCR3, the orientation towards the transmembrane region, and the activation of GP. As a consequence of antagonizing CXCR3, a reduction of the chemotaxis of the cells of the immune Molecules 2020, 25, 4413 25 of 30 system could prevent the appearance of new lesions and the progression of existing lesions in vitiligo patients. Additionally, the antagonism of CXCR3 can inhibit other pathological processes in which CXCR3 is involved, such as cancer metastasis. In contrast, CXCR3B is a receptor with 100 residues in the N-terminus, compared to the 53 residues on CXCR3A, and is related with melanocytic apoptosis in the early stages of vitiligo; therefore, the general protein–ligand interaction model reported here could be relevant also in CXCR3B, a question that stands as a perspective for future works. Still, it is particularly important that these results be experimentally validated in collaboration with experts in this area.

Supplementary Materials: The following are available online. Figure S1: Ramachandran plot. Figure S2: Ramachandran plot. Figure S3: Ramachandran plot. Figure S4: AA-MD simulation of CXCL9 5 ns. Figure S5: Rotation of the TMs from the CG-MD simulation of the CXCR3/GP complex. Figure S6: RMSD of the AA-MD simulations. Figure S7: RMSD of the CG-MD simulations. Table S1: Salt bridge interactions of CXCR3 with CXCL9, CXCL10, and CXCL11 in molecular docking. Table S2: Salt bridge interactions of CXCR3 with CXCL9, CXCL10, and CXCL11 in AA-MD 50 ns. Table S3: Salt bridge interactions of CXCR3 with CXCL9 in the CG-MD simulation. Table S4: Salt bridge interactions of CXCR3 with CXCL10 in the CG-MD simulation. Table S5: Salt bridge interactions of CXCR3 with CXCL11 in the CG-MD simulation. Table S6: Polar interactions between CXCR3-CXCL9 CG-MD. Table S7: Polar interactions between the TMs in CXCR3-CXCL9 CG-MD. Table S8: Polar interactions between CXCR3-CXCL10 CG-MD. Table S9: Polar interactions between the TMs in CXCR3-CXCL10 CG-MD. Table S10: Polar interactions between CXCR3-CXCL11 CG-MD. Table S11: Polar interactions between the TMs in CXCR3-CXCL11 CG-MD. Table S12: Polar interactions between the TMs in the CXCR3/GP complex 1µ. Table S13: Polar interactions between CXCR3-CXCL10 250-ns AA-MD. Table S14. Polar interactions between the TMs in CXCR3-CXCL10 250-ns AA-MD. Table S15: Polar interactions between CXCR3_alpha subunit 250-ns AA-MD. Author Contributions: Conceptualization: G.A.-D. and A.R.-M.; calculations: G.A.-D.; data analysis: G.A.-D. and A.R.-M.; writing—original draft preparation: G.A.-D.; writing—review and editing: G.A.-D. and A.R.-M.; project administration and funding acquisition: A.R.-M. All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by the University of Queretaro (UAQ) through the FOFI-UAQ research grant FCQ201825 and by the National Council of Research and Technology (CONACYT) for doctoral scholarship number 335496 to G.A-D. Acknowledgments: The authors gratefully acknowledge the technical support from Luis Aguilar, Alejandro de León, Carlos Flores, and Jair García of the National Laboratory of Advanced Scientific Visualization (LAVIS-UNAM). Conflicts of Interest: The authors declare no conflict of interests.

References

1. Ezzedine, K.; Lim, H.W.; Suzuki, T.; Katayama, I.; Hamzavi, I.; Lan, C.C.E.; Goh, B.K.; Anbar, T.; Silva de Castro, C.; Lee, A.Y.; et al. Revised classification/nomenclature of vitiligo and related issues: The Vitiligo Global Issues Consensus Conference. Pigment. Cell Melanoma Res. 2012, 25, E1–E13. [CrossRef] [PubMed] 2. Krüger, C.; Schallreuter, K.U. A review of the worldwide prevalence of vitiligo in children/adolescents and adults. Int. J. Dermatol. 2012, 51, 1206–1212. [CrossRef][PubMed] 3. Qiao, Z.; Wang, X.; Xiang, L.; Zhang, C. Dysfunction of Autophagy: A Possible Mechanism Involved in the Pathogenesis of Vitiligo by Breaking the Redox Balance of Melanocytes. Oxid. Med. Cell. Longev. 2016, 2016, 3401570. [CrossRef][PubMed] 4. Ezzedine, K.; Eleftheriadou, V.; Whitton, M.; Van Geel, N. Vitiligo. Lancet 2015, 386, 74–84. [CrossRef] 5. Hedayat, K.; Karbakhsh, M.; Ghiasi, M.; Goodarzi, A.; Fakour, Y.; Akbari, Z.; Ghayoumi, A.; Ghandi, N. Quality of life in patients with vitiligo: A cross-sectional study based on Vitiligo Quality of Life index (VitiQoL). Health Qual. Life Outcomes 2016, 14, 86. [CrossRef] 6. Nguyen, C.; Beroukhim, K.; Danesh, M.; Babikian, A.; Koo, J.; Leon, A. The psychosocial impact of acne, vitiligo, and psoriasis: A review. Clin. Cosmet. Investig. Dermatol. 2016, 9, 383–392. [CrossRef] 7. Hagen, S.L.; Grey, K.R.; Korta, D.Z.; Kelly, K.M. Quality of life in adults with facial port-wine stains. J. Am. Acad. Dermatol. 2017, 76, 695–702. [CrossRef] 8. Morales-Sánchez, M.A.; Vargas-Salinas, M.; Peralta-Pedrero, M.L.; Olguín-García, M.G.; Jurado-Santa Cruz, F. Impact of Vitiligo on Quality of Life. Actas Dermo-Sifiliogr. Engl. Ed. 2017, 108, 637–642. [CrossRef] Molecules 2020, 25, 4413 26 of 30

9. Lai, Y.C.; Yew, Y.W.; Kennedy, C.; Schwartz, R.A. Vitiligo and depression: A systematic review and meta-analysis of observational studies. Br. J. Dermatol. 2017, 177, 708–718. [CrossRef] 10. Huggins, R.H.; Schwartz, R.A.; Janniger, C.K. Vitiligo. Acta Dermatovenerol. Alp Pannonica Adriat. 2005, 14, 137–144. 11. Rashighi, M.; Agarwal, P.; Richmond, J.M.; Harris, T.H.; Dresser, K.; Su, M.-W.; Zhou, Y.; Deng, A.; Hunter, C.A.; Luster, A.D.; et al. CXCL10 Is Critical for the Progression and Maintenance of Depigmentation in a Mouse Model of Vitiligo. Sci. Transl. Med. 2014, 6, 223ra23. [CrossRef][PubMed] 12. Laddha, N.C.; Dwivedi, M.; Mansuri, M.S.; Gani, A.R.; Ansarullah, M.; Ramachandran, A.V.; Dalai, S.; Begum, R. Vitiligo: Interplay between oxidative stress and immune system. Exp. Dermatol. 2013, 22, 245–250. [CrossRef][PubMed] 13. Picardo, M.; Dell’Anna, M.L.; Ezzedine, K.; Hamzavi, I.; Harris, J.E.; Parsad, D.; Taieb, A. Vitiligo. Nat. Rev. Dis. Prim. 2015, 1, 15011. [CrossRef][PubMed] 14. Jimbow, K.; Chen, H.; Park, J.-S.; Thomas, P.D. Increased sensitivity of melanocytes to oxidative stress and abnormal expression of tyrosinase-related protein in vitiligo. Br. J. Dermatol. 2001, 144, 55–65. [CrossRef] 15. Xie, H.; Zhou, F.; Liu, L.; Zhu, G.; Li, Q.; Li, C.; Gao, T. Vitiligo: How do oxidative stress-induced autoantigens trigger autoimmunity? J. Dermatol. Sci. 2016, 81, 3–9. [CrossRef] 16. van den Boorn, J.G.; Konijnenberg, D.; Dellemijn, T.A.M.; Wietze van der Veen, J.P.; Bos, J.D.; Melief, C.J.M.; Vyth-Dreese, F.A.; Luiten, R.M. Autoimmune Destruction of Skin Melanocytes by Perilesional T Cells from Vitiligo Patients. J. Investig. Dermatol. 2009, 129, 2220–2232. [CrossRef] 17. Brinkman, C.C.; Peske, J.D.; Engelhard, V.H. Peripheral Tissue Homing Receptor Control of Naïve, Effector, and Memory CD8 T Cell Localization in Lymphoid and Non-Lymphoid Tissues. Front. Immunol. 2013, 4, 1–8. [CrossRef] 18. Müller, M.; Carter, S.; Hofer, M.J.; Campbell, I.L. Review: The chemokine receptor CXCR3 and its ligands CXCL9, CXCL10 and CXCL11 in neuroimmunity—A tale of conflict and conundrum. Neuropathol. Appl. Neurobiol. 2010, 36, 368–387. [CrossRef] 19. Li, S.; Zhu, G.; Yang, Y.; Jian, Z.; Guo, S.; Dai, W.; Shi, Q.; Ge, R.; Ma, J.; Liu, L.; et al. Oxidative stress drives CD8 + T-cell skin trafficking in patients with vitiligo through CXCL16 upregulation by activating the unfolded protein response in keratinocytes. J. Allergy Clin. Immunol. 2017, 140, 177–189.e9. [CrossRef] 20. Zhu, Y.; Zhong, L.; Peng, J.; Yuan, Q.; Xu, A. The Therapeutic Effects of Baicalin on Vitiligo Mice. Biol. Pharm. Bull. 2019, 42, 1450–1455. [CrossRef] 21. Yang, L.; Yang, S.; Lei, J.; Hu, W.; Chen, R.; Lin, F.; Xu, A. Role of chemokines and the corresponding receptors in vitiligo: A pilot study. J. Dermatol. 2018, 45, 31–38. [CrossRef][PubMed] 22. Boniface, K.; Jacquemin, C.; Darrigade, A.-S.; Dessarthe, B.; Martins, C.; Boukhedouni, N.; Vernisse, C.; Grasseau, A.; Thiolat, D.; Rambert, J.; et al. Vitiligo Skin Is Imprinted with Resident Memory CD8 T Cells Expressing CXCR3. J. Investig. Dermatol. 2018, 138, 355–364. [CrossRef] 23. Van Raemdonck, K.; Van den Steen, P.E.; Liekens, S.; Van Damme, J.; Struyf, S. CXCR3 ligands in disease and therapy. Cytokine Growth Factor Rev. 2015, 26, 311–327. [CrossRef] 24. Smith, J.S.; Alagesan, P.; Desai, N.K.; Pack, T.F.; Wu, J.-H.; Inoue, A.; Freedman, N.J.; Rajagopal, S. C-X-C Motif Chemokine Receptor 3 Splice Variants Differentially Activate Beta-Arrestins to Regulate Downstream Signaling Pathways. Mol. Pharmacol. 2017, 92, 136–150. [CrossRef][PubMed] 25. Metzemaekers, M.; Vanheule, V.; Janssens, R.; Struyf, S.; Proost, P. Overview of the Mechanisms that May Contribute to the Non-Redundant Activities of Interferon-Inducible CXC Chemokine Receptor 3 Ligands. Front. Immunol. 2018, 8.[CrossRef] 26. Huang, D.; Gu, Q.; Ge, H.; Ye, J.; Salam, N.K.; Hagler, A.; Chen, H.; Xu, J. On the Value of Homology Models for Virtual Screening: Discovering hCXCR3 Antagonists by Pharmacophore-Based and Structure-Based Approaches. J. Chem. Inf. Model. 2012, 52, 1356–1366. [CrossRef][PubMed] 27. Palladino, P.; Portella, L.; Colonna, G.; Raucci, R.; Saviano, G.; Rossi, F.; Napolitano, M.; Scala, S.; Castello, G.; Costantini, S. The N-terminal Region of CXCL11 as Structural Template for CXCR3 Molecular Recognition: Synthesis, Conformational Analysis, and Binding Studies. Chem. Biol. Drug Des. 2012, 80, 254–265. [CrossRef] 28. Groom, J.R.; Luster, A.D. CXCR3 ligands: Redundant, collaborative and antagonistic functions. Immunol. Cell Biol. 2011, 89, 207–215. [CrossRef] Molecules 2020, 25, 4413 27 of 30

29. Tulic, M.K.; Cavazza, E.; Cheli, Y.; Jacquel, A.; Luci, C.; Cardot-Leccia, N.; Hadhiri-Bzioueche, H.; Abbe, P.; Gesson, M.; Sormani, L.; et al. Innate lymphocyte-induced CXCR3B-mediated melanocyte apoptosis is a potential initiator of T-cell autoreactivity in vitiligo. Nat. Commun. 2019, 10, 1–13. [CrossRef] 30. Clark-Lewis, I.; Mattioli, I.; Gong, J.-H.; Loetscher, P. Structure-Function Relationship between the Human Chemokine Receptor CXCR3 and Its Ligands. J. Biol. Chem. 2003, 278, 289–295. [CrossRef] 31. Ngo, T.; Stephens, B.S.; Gustavsson, M.; Holden, L.G.; Abagyan, R.; Handel, T.M.; Kufareva, I. Crosslinking-guided geometry of a complete CXC receptor-chemokine complex and the basis of chemokine subfamily selectivity. PLoS Biol. 2020, 18, e3000656. [CrossRef][PubMed] 32. Booth, V.; Keizer, D.W.; Kamphuis, M.B.; Clark-Lewis, I.; Sykes, B.D. The CXCR3 Binding Chemokine IP-10/CXCL10: Structure and Receptor Interactions. Biochemistry 2002, 41, 10418–10425. [CrossRef][PubMed] 33. Benredjem, B.; Girard, M.; Rhainds, D.; St.-Onge, G.; Heveker, N. Mutational Analysis of Atypical Chemokine Receptor 3 (ACKR3/CXCR7) Interaction with Its Chemokine Ligands CXCL11 and CXCL12. J. Biol. Chem. 2017, 292, 31–42. [CrossRef][PubMed] 34. Sanchez, J.; e Huma, Z.; Lane, J.R.; Liu, X.; Bridgford, J.L.; Payne, R.J.; Canals, M.; Stone, M.J. Evaluation and extension of the two-site, two-step model for binding and activation of the chemokine receptor CCR1. J. Biol. Chem. 2019, 294, 3464–3475. [CrossRef] 35. Neumann, A.; Engel, V.; Mahardhika, A.B.; Schoeder, C.T.; Namasivayam, V.; Kie´c-Kononowicz,K.; Müller, C.E. Computational Investigations on the Binding Mode of Ligands for the Cannabinoid-Activated G Protein-Coupled Receptor GPR18. Biomolecules 2020, 10, 686. [CrossRef] 36. Singh, S.P.; Foley, J.F.; Zhang, H.H.; Hurt, D.E.; Richards, J.L.; Smith, C.S.; Liao, F.; Farber, J.M. Selectivity in the Use of G i/o Proteins Is Determined by the DRF Motif in CXCR6 and Is Cell-Type Specific. Mol. Pharmacol. 2015, 88, 894–910. [CrossRef] 37. Okada, T.; Ernst, O.P.; Palczewski, K.; Hofmann, K.P. Activation of rhodopsin: New insights from structural and biochemical studies. Trends Biochem. Sci. 2001, 26, 318–324. [CrossRef] 38. Hofmann, K.P.; Scheerer, P.; Hildebrand, P.W.; Choe, H.-W.; Park, J.H.; Heck, M.; Ernst, O.P. A G protein-coupled receptor at work: The rhodopsin model. Trends Biochem. Sci. 2009, 34, 540–552. [CrossRef] 39. Deupi, X. Relevance of rhodopsin studies for GPCR activation. Biochim. Biophys. Acta Bioenerg. 2014, 1837, 674–682. [CrossRef] 40. Woods, K.N.; Pfeffer, J.; Dutta, A.; Klein-Seetharaman, J. Vibrational resonance, allostery, and activation in rhodopsin-like G protein-coupled receptors. Sci. Rep. 2016, 6, 37290. [CrossRef] 41. Gao, Y.; Hu, H.; Ramachandran, S.; Erickson, J.W.; Cerione, R.A.; Skiniotis, G. Structures of the Rhodopsin- Transducin Complex: Insights into G-Protein Activation. Mol. Cell 2019, 75, 781–790.e3. [CrossRef][PubMed] 42. Cole, A.G.; Stroke, I.L.; Brescia, M.-R.; Simhadri, S.; Zhang, J.J.; Hussain, Z.; Snider, M.; Haskell, C.; Ribeiro, S.; Appell, K.C.; et al. Identification and initial evaluation of 4-N-aryl-[1,4]diazepane ureas as potent CXCR3 antagonists. Bioorg. Med. Chem. Lett. 2006, 16, 200–203. [CrossRef] 43. Watson, R.J.; Allen, D.R.; Birch, H.L.; Chapman, G.A.; Galvin, F.C.; Jopling, L.A.; Knight, R.L.; Meier, D.; Oliver, K.; Meissner, J.W.G.; et al. Development of CXCR3 antagonists. Part 3: Tropenyl and homotropenyl- piperidine urea derivatives. Bioorg. Med. Chem. Lett. 2008, 18, 147–151. [CrossRef][PubMed] 44. Bongartz, J.-P.; Buntinx, M.; Coesemans, E.; Hermans, B.; Van Lommen, G.; Wauwe, J. Synthesis and structure–activity relationship of benzetimide derivatives as human CXCR3 antagonists. Bioorg. Med. Chem. Lett. 2008, 18, 5819–5823. [CrossRef][PubMed] 45. Allen, D.R.; Bolt, A.; Chapman, G.A.; Knight, R.L.; Meissner, J.W.G.; Owen, D.A.; Watson, R.J. Identification and structure–activity relationships of 1-aryl-3-piperidin-4-yl-urea derivatives as CXCR3 receptor antagonists. Bioorg. Med. Chem. Lett. 2007, 17, 697–701. [CrossRef] 46. Du, X.; Gustin, D.J.; Chen, X.; Duquette, J.; McGee, L.R.; Wang, Z.; Ebsworth, K.; Henne, K.; Lemon, B.; Ma, J.; et al. Imidazo-pyrazine derivatives as potent CXCR3 antagonists. Bioorg. Med. Chem. Lett. 2009, 19, 5200–5204. [CrossRef] 47. McGuinness, B.F.; Carroll, C.D.; Zawacki, L.G.; Dong, G.; Yang, C.; Hobbs, D.W.; Jacob-Samuel, B.; Hall, J.W.; Jenh, C.-H.; Kozlowski, J.A.; et al. Novel CXCR3 antagonists with a piperazinyl-piperidine core. Bioorg. Med. Chem. Lett. 2009, 19, 5205–5208. [CrossRef] 48. Crosignani, S.; Missotten, M.; Cleva, C.; Dondi, R.; Ratinaud, Y.; Humbert, Y.; Mandal, A.B.; Bombrun, A.; Power, C.; Chollet, A.; et al. Discovery of a novel series of CXCR3 antagonists. Bioorg. Med. Chem. Lett. 2010, 20, 3614–3617. [CrossRef] Molecules 2020, 25, 4413 28 of 30

49. Shao, Y.; Anilkumar, G.N.; Carroll, C.D.; Dong, G.; Hall, J.W.; Hobbs, D.W.; Jiang, Y.; Jenh, C.-H.; Kim, S.H.; Kozlowski, J.A.; et al. II. SAR studies of pyridyl–piperazinyl-piperidine derivatives as CXCR3 chemokine antagonists. Bioorg. Med. Chem. Lett. 2011, 21, 1527–1531. [CrossRef] 50. Wijtmans, M.; Verzijl, D.; Bergmans, S.; Lai, M.; Bosch, L.; Smit, M.J.; de Esch, I.J.P.; Leurs, R. CXCR3 antagonists: Quaternary ammonium salts equipped with biphenyl- and polycycloaliphatic-anchors. Bioorg. Med. Chem. 2011, 19, 3384–3393. [CrossRef] 51. Thoma, G.; Baenteli, R.; Lewis, I.; Jones, D.; Kovarik, J.; Streiff, M.B.; Zerwes, H.-G. Special ergolines efficiently inhibit the chemokine receptor CXCR3 in blood. Bioorg. Med. Chem. Lett. 2011, 21, 4745–4749. [CrossRef] [PubMed] 52. Floudas, C.A.; Fung, H.K.; McAllister, S.R.; Mönnigmann, M.; Rajgaria, R. Advances in protein structure prediction and de novo protein design: A review. Chem. Eng. Sci. 2006, 61, 966–988. [CrossRef] 53. Trotta, T.; Costantini, S.; Colonna, G. Modelling of the membrane receptor CXCR3 and its complexes with CXCL9, CXCL10 and CXCL11 chemokines: Putative target for new drug design. Mol. Immunol. 2009, 47, 332–339. [CrossRef][PubMed] 54. Hillisch, A.; Pineda, L.F.; Hilgenfeld, R. Utility of homology models in the drug discovery process. Drug Discov. Today 2004, 9, 659–669. [CrossRef] 55. Chen, V.B.; Arendall, W.B.; Headd, J.J.; Keedy, D.A.; Immormino, R.M.; Kapral, G.J.; Murray, L.W.; Richardson, J.S.; Richardson, D.C. MolProbity: All-atom structure validation for macromolecular crystallography. Acta Crystallogr. Sect. D Biol. Crystallogr. 2010, 66, 12–21. [CrossRef][PubMed] 56. Jha, P.; Chaturvedi, S.; Swastika; Pal, S.; Jain, N.; Mishra, A.K. Improvising 5-HT 7 R homology model for design of high affinity ligands: Model validation with docking, embrace minimization, MM-GBSA, and molecular dynamic simulations. J. Biomol. Struct. Dyn. 2018, 36, 2475–2494. [CrossRef] 57. Singh, G.; Ramachandran, S.; Cerione, R.A. A Constitutively Active Gα Subunit Provides Insights into the Mechanism of G Protein Activation. Biochemistry 2012, 51, 3232–3240. [CrossRef] 58. The Uniprot Consortium. Activities at the Universal Protein Resource (UniProt). Nucleic Acids Res. 2014, 42, D191–D198. [CrossRef] 59. Colvin, R.A.; Campanella, G.S.V.; Manice, L.A.; Luster, A.D. CXCR3 Requires Tyrosine Sulfation for Ligand Binding and a Second Extracellular Loop Arginine Residue for Ligand-Induced Chemotaxis. Mol. Cell. Biol. 2006, 26, 5838–5849. [CrossRef] 60. Nedjai, B.; Viney, J.M.; Li, H.; Hull, C.; Anderson, C.A.; Horie, T.; Horuk, R.; Vaidehi, N.; Pease, J.E. CXCR3 antagonist VUF10085 binds to an intrahelical site distinct from that of the broad spectrum antagonist TAK-779. Br. J. Pharmacol. 2015, 172, 1822–1833. [CrossRef] 61. Sivashanmugam, M.; Sulochana, K.N.; Umashankar, V. Virtual screening of natural inhibitors targeting ornithine decarboxylase with pharmacophore scaffolding of DFMO and validation by molecular dynamics simulation studies. J. Biomol. Struct. Dyn. 2019, 37, 766–780. [CrossRef][PubMed] 62. Booth, V.; Clark-Lewis, I.; Sykes, B.D. NMR structure of CXCR3 binding chemokine CXCL11 (ITAC). Protein Sci. 2004, 13, 2022–2028. [CrossRef][PubMed] 63. García-Nafría, J.; Nehmé, R.; Edwards, P.C.; Tate, C.G. Cryo-EM structure of the serotonin 5-HT1Breceptor coupled to heterotrimeric Go. Nature 2018, 558, 620–623. [CrossRef][PubMed] 64. Zhang, J.; Yang, J.; Jang, R.; Zhang, Y. GPCR-I-TASSER: A Hybrid Approach to G Protein-Coupled Receptor Structure Modeling and the Application to the Human Genome. Structure 2015, 23, 1538–1549. [CrossRef] [PubMed] 65. Altschul, S.F.; Gish, W.; Miller, W.; Myers, E.W.; Lipman, D.J. Basic local alignment search tool. J. Mol. Biol. 1990, 215, 403–410. [CrossRef] 66. States, D.J.; Gish, W. QGB: Combined Use of Sequence Similarity and Codon Bias for Coding Region Identification. J. Comput. Biol. 1994, 1, 39–50. [CrossRef][PubMed] 67. Kozakov, D.; Hall, D.R.; Xia, B.; Porter, K.A.; Padhorny, D.; Yueh, C.; Beglov, D.; Vajda, S. The ClusPro web server for protein–protein docking. Nat. Protoc. 2017, 12, 255–278. [CrossRef] 68. Vajda, S.; Yueh, C.; Beglov, D.; Bohnuud, T.; Mottarella, S.E.; Xia, B.; Hall, D.R.; Kozakov, D. New additions to the ClusPro server motivated by CAPRI. Proteins Struct. Funct. Bioinf. 2017, 85, 435–444. [CrossRef] 69. Colvin, R.A.; Campanella, G.S.V.; Sun, J.; Luster, A.D. Intracellular Domains of CXCR3 That Mediate CXCL9, CXCL10, and CXCL11 Function. J. Biol. Chem. 2004, 279, 30219–30227. [CrossRef] Molecules 2020, 25, 4413 29 of 30

70. Nedjai, B.; Li, H.; Stroke, I.L.; Wise, E.L.; Webb, M.L.; Merritt, J.R.; Henderson, I.; Klon, A.E.; Cole, A.G.; Horuk, R.; et al. Small molecule chemokine mimetics suggest a molecular basis for the observation that CXCL10 and CXCL11 are allosteric ligands of CXCR3. Br. J. Pharmacol. 2012, 166, 912–923. [CrossRef] 71. Jo, S.; Kim, T.; Iyer, V.G.; Im, W. CHARMM-GUI: A web-based graphical user interface for CHARMM. J. Comput. Chem. 2008, 29, 1859–1865. [CrossRef][PubMed] 72. Lee, J.; Cheng, X.; Swails, J.M.; Yeom, M.S.; Eastman, P.K.; Lemkul, J.A.; Wei, S.; Buckner, J.; Jeong, J.C.; Qi, Y.; et al. CHARMM-GUI Input Generator for NAMD, GROMACS, AMBER, OpenMM, and CHARMM/OpenMM Simulations Using the CHARMM36 Additive Force Field. J. Chem. Theory Comput. 2016, 12, 405–413. [CrossRef][PubMed] 73. Brooks, B.R.; Brooks, C.L.; Mackerell, A.D.; Nilsson, L.; Petrella, R.J.; Roux, B.; Won, Y.; Archontis, G.; Bartels, C.; Boresch, S.; et al. CHARMM: The biomolecular simulation program. J. Comput. Chem. 2009, 30, 1545–1614. [CrossRef][PubMed] 74. Berendsen, H.J.C.; van der Spoel, D.; van Drunen, R. GROMACS: A message-passing parallel molecular dynamics implementation. Comput. Phys. Commun. 1995, 91, 43–56. [CrossRef] 75. Van Der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A.E.; Berendsen, H.J.C. GROMACS: Fast, flexible, and free. J. Comput. Chem. 2005, 26, 1701–1718. [CrossRef] 76. Bussi, G.; Donadio, D.; Parrinello, M. Canonical sampling through velocity rescaling. J. Chem. Phys. 2007, 126, 014101. [CrossRef] 77. Parrinello, M.; Rahman, A. Polymorphic transitions in single crystals: A new molecular dynamics method. J. Appl. Phys. 1981, 52, 7182–7190. [CrossRef] 78. Daura, X.; Gademann, K.; Jaun, B.; Seebach, D.; van Gunsteren, W.F.; Mark, A.E. Peptide Folding: When Simulation Meets Experiment. Angew. Chem. Int. Ed. 1999, 38, 236–240. [CrossRef] 79. Team, R.C. R: A Language and Environment for Statistical Computing. Available online: https://www.r- project.org (accessed on 12 August 2020). 80. Wickham, H. Ggplot2: Elegant Graphics for Data Analysis; Springer: New York, NY, USA, 2009; ISBN 978-0-387-98140-6. 81. Wallace, A.C.; Laskowski, R.A.; Thornton, J.M. Ligplot: A program to generate schematic diagrams of protein-ligand interactions. Protein Eng. Des. Sel. 1995, 8, 127–134. [CrossRef] 82. Laskowski, R.A.; Swindells, M.B. LigPlot+: Multiple Ligand–Protein Interaction Diagrams for Drug Discovery. J. Chem. Inf. Model. 2011, 51, 2778–2786. [CrossRef] 83. Kumar, S.; Nussinov, R. Salt bridge stability in monomeric proteins 1 1Edited by J. M. Thornton. J. Mol. Biol. 1999, 293, 1241–1255. [CrossRef][PubMed] 84. Kumar, S.; Tsai, C.J.; Ma, B.; Nussinov, R. Contribution of salt bridges toward protein thermostability. J. Biomol. Struct. Dyn. 2000, 17, 79–85. [CrossRef][PubMed] 85. Kumar, S.; Nussinov, R. Relationship between Ion Pair Geometries and Electrostatic Strengths in Proteins. Biophys. J. 2002, 83, 1595–1612. [CrossRef] 86. Qi, Y.; Ingólfsson, H.I.; Cheng, X.; Lee, J.; Marrink, S.J.; Im, W. CHARMM-GUI Martini Maker for Coarse-Grained Simulations with the Martini Force Field. J. Chem. Theory Comput. 2015, 11, 4486–4494. [CrossRef][PubMed] 87. Sterling, T.; Irwin, J.J. ZINC 15—Ligand Discovery for Everyone. J. Chem. Inf. Model. 2015, 55, 2324–2337. [CrossRef][PubMed] 88. Kenakin, T.P. A pharmacological method to estimate the pK I of competitive inhibitors of agonist uptake processes in isolated tissues. Naunyn. Schmiedebergs. Arch. Pharmacol. 1981, 316, 89–95. [CrossRef][PubMed] 89. Morris, G.M.; Huey, R.; Lindstrom, W.; Sanner, M.F.; Belew, R.K.; Goodsell, D.S.; Olson, A.J. AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility. J. Comput. Chem. 2009, 30, 2785–2791. [CrossRef] 90. Santos-martins, D.; Solis-vasquez, L.; Koch, A.; Forli, S. Accelerating AUTODOCK 4 with GPUs and Gradient-Based Local Search. ChemRxiv 2019.[CrossRef] 91. Koes, D.R.; Camacho, C.J. Pharmer: Efficient and Exact Pharmacophore Search. J. Chem. Inf. Model. 2011, 51, 1307–1314. [CrossRef] 92. Reynolds, A.P.; Richards, G.; de la Iglesia, B.; Rayward-Smith, V.J. Clustering Rules: A Comparison of Partitioning and Hierarchical Clustering Algorithms. J. Math. Model. Algorithms 2006, 5, 475–504. [CrossRef] Molecules 2020, 25, 4413 30 of 30

93. Maechler, M.; Rousseeuw, P.; Struyf, A.; Hubert, M.; Hornik, K. Cluster: Cluster Analysis Basics and Extensions. R Package Version 2.1.0. Available online: https://cran.r-project.org/package=cluster (accessed on 12 August 2020). 94. Schrödinger Inc. The PyMOL Molecular Graphics System; Version 2; Schrödinger Inc.: New York, NY, USA, 2015. 95. Schrödinger. Schrödinger Release 2019-3: Maestro; Schrödinger Inc.: New York, NY, USA, 2019.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).