Atypical activation and molecular glue-like dimerization mechanism of an intrinsically-biased chemokine receptor

GPCRs, biased -agonism, chemokines, chemokine receptors, CXCR7, ACKR3, 15 drug discovery, cellular signaling, arrestins, opioid receptors 16 17 18 19 20 21 22 23 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 2

24 CXCR7, also known as atypical chemokine receptor 3 (ACKR3), is a naturally -biased, β -25 arrestin-coupled seven transmembrane receptor (7TMR) that lacks productive coupling with 26 heterotrimeric G-proteins. Despite a critical involvement in cancer metastasis, cardiovascular 27 pathophysiology, and inflammatory disorders, the molecular basis of non-canonical activation 28 and functional divergence of CXCR7 remains elusive. Here, we present a complete landscape 29 of CXCR7 activation using a series of cryo-EM structures, and discover an atypical activation 30 mechanism that is distinct from prototypical GPCRs. CXCR7 is maintained in a basal 31 conformation by a unique tripartite ionic -lock involving TM5 -TM6, in contrast to a broadly 32 conserved TM3 -TM6 ionic -lock in GPCRs, which is disrupted upon receptor activation. 33 Importantly, activation of CXCR7 results in a constricted pocket and distinct surface topology 34 on the intracellular side compared to prototypical GPCRs. Serendipitously, we capture novel 35 dimeric arrangements of CXCR7 with an inter -protomer stitching by a native phospholipid 36 serving as a molecular glue, and identify previously unanticipated extrahelical allosteric sites 37 on the receptor. Surprisingly, in an intermediate state structure of CXCR7, the second 38 extracellular loop (ECL2) displays a self -blocking conformation, in stark contrast to ECL2 -39 mediated self-activating mechanism reported recently for some orphan GPCRs. Finally, we 40 unequivocally establish CXCR7 as an atypical opioid receptor via a large peptide library 41 screening and structure elucidation in complex with distinct opioid peptides imparting full 42 receptor activation. In summary, our study elucidates an atypical mechanism of CXCR7 43 activation, and establishes it as an alternative, non -canonical opioid receptor target with 44 potential for novel pain therapeutics. 45 46 47 48 49 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 3

50 G protein-coupled receptors (GPCRs) constitute a large superfamily of integral membrane 51 proteins, and they are critically involved in nearly all aspects of human physiology1,2. Aberrant 52 GPCR signaling is linked to a broad spectrum of human disorders, and as a result, they are 53 targeted by approximately one third of the currently prescribed medicines 3-5. The broadly 54 conserved paradigm of GPCR activation and signaling involves agonist-induced interaction of 55 the receptors with heterotrimeric G-proteins, GRKs, and β-arrestins (βarrs). Structural biology 56 efforts using crystallography and cryo -EM have illuminated the intricate details of receptor 57 activation and transducer -coupling mechanisms over the past two decades 6,7. Interestingly, 58 the classical paradigm of GPCR activation has been expanded further with the discovery and 59 integration of functional selectivity framework, also referred to as signaling -bias, and it has 60 profoundly influenced the ongoing drug discovery efforts 8-11. A series of recent studies 61 combining biophysical and structural methodologies have started to provide generalizable 62 insights into functional selectivity at the level of receptor and/or transducer conformations12-14. 63 However, these studies are primarily limited to synthetic ligands and mutant receptors, and 64 relatively little is known about the molecular basis and underlying principles of naturally -65 encoded signaling-bias at GPCRs. 66 Several members of the GPCR superfamily are incapable of activating heterotrimeric 67 G-proteins, despite harboring a conserved seven transmembrane (7TM) architecture, but they 68 still maintain robust GRK and βarr recruitment 15-18. These arrestin-coupled receptors (ACRs) 69 represent naturally -encoded GRK/βarr -biased 7TMRs, and offer a unique opportunity to 70 decode the fundamental mechanisms directing distinct modalities of receptor activation and 71 ensuing signaling-bias19,20. Most of these receptors belong to the subfamily of chemokine 72 receptors, and also referred to as Atypical Chemokine Receptors (ACKRs) 16,17, although 73 additional examples from other receptor subfamilies have also emerged19,21. Previous studies 74 on these receptors have focused predominantly on establishing their ligand -binding 75 properties, transducer-coupling patterns, and non-canonical downstream signaling19,22, but the 76 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 4 structural mechanism of their functional divergence remains mostly elusive. For example, 77 these receptors appear to preferentially engage GRK5/6 for their phosphorylation and 78 subsequent βarr recruitment, which aligns with the lack of G -protein activation and hence, 79 inefficient translocation of GRK2/3 to the plasma membrane19.23,24 Unlike prototypical GPCRs, 80 ACRs do not seem to elicit canonical downstream signaling cascades, potentially suggesting 81 distinct modalities of transducer-coupling and receptor activation compared to GPCRs 19. For 82 instance, agonist -stimulation does not result in a measurable ERK1/2 phosphorylation 83 downstream of ACRs, which has been used as a quintessential readout to define the 84 spatiotemporal aspects of βarr signaling at prototypical GPCRs 19. These unique functional 85 manifestations distinguish these receptors from the functional selectivity scenario operational 86 at prototypical GPCRs imparted by synthetic biased ligands, and establish them as unique 87 system to explore the molecular mechanisms of naturally-encoded signaling-bias. 88 One of the naturally-biased 7TMRs, CXCR7, also referred to as ACKR3, is expressed 89 widely in the endothelial and immune cells, and plays a central role in cellular migration, 90 embryonic development, immune regulation, tissue homeostasis, and vascular functions 25. 91 Aberrant CXCR7 signaling is implicated in numerous disease contexts including tumour 92 growth, metastasis and angiogenesis, vascular remodelling, multiple sclerosis, 93 neuroinflammation, and autoimmune disorders 26. CXCR7 recognizes two distinct C -X-C 94 chemokines namely, CXCL11 and CXCL12, which also activate prototypical GPCRs i.e. 95 CXCR3 and CXCR4, respectively, and couples efficiently to βarrs27. Similar to class A GPCRs, 96 CXCR7 harbors the conserved sequence motifs such as the DRY in TM3 and NPXXY in TM7, 97 which are implicated in G -protein-coupling and activation 28. Therefore, its inability to 98 functionally engage and activate G-proteins is even more perplexing, and potentially suggests 99 an atypical activation mechanism driving the functional divergence. Two recent studies have 100 provided the first glimpse of CXCL12 binding29 and βarr-coupling to CXCR7 using cryo-EM30. 101 In these structural snapshots, CXCL12 appears to adopt a significantly different binding pose, 102 compared to that by other chemokines on their cognate receptors, as visualized in the 103 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 5 structures reported previously 29,30. Still however, considering that the inactive structure of 104 CXCR7, and that of any other ACRs, still remains unresolved, deciphering the activation 105 mechanism, in particular, the distinguishing features compared to prototypical GPCRs, 106 remains rather speculative. These knowledge gaps represent major lacunae in our current 107 understanding of activation, divergent signaling, and non -canonical functions mediated by 108 ACRs including CXCR7. Interestingly, CXCR7 is also proposed to recognize and scavenge 109 opioid peptides, and potentially modulate the endogenous opioid system implicated in 110 nociception31-33. However, the molecular basis of opioid -recognition by CXCR7 and ensuing 111 receptor activation has not yet been studied at molecular level, which in turn, has limited the 112 efforts to explore the potential of modulating CXCR7 for developing novel pain therapeutics. 113 Here, we elucidate an atypical activation mechanism of CXCR7 involving a significant 114 rearrangement of unique ionic-locks distinct from prototypical GPCRs, and discover a pivotal 115 switch that is also broadly conserved in chemoattractant receptors. We identify previously 116 unanticipated extrahelical allosteric sites on the receptor for agonists and phospholipids, and 117 surprisingly, also trap receptor dimers using a native lipid as molecular glue with an 118 architecture that is reminiscent of distinct conformational states of membrane transporters. 119 Moreover, using a combination of peptide library screening and cryo -EM-based structural 120 elucidation, we unequivocally establish CXCR7 as an atypical opioid receptor, thereby, 121 expanding the subfamily of opioid receptors beyond the conventional members. Our findings 122 offer important insights into CXCR7 activation and signaling with broad and generalizable 123 implications for the framework of naturally-encoded signaling-bias and development of novel 124 therapeutics. 125

126 A rich tapestry of CXCR7 ligand pharmacology 127 CXCR7 is one of the most well characterized chemokine receptors with respect to a wide 128 spectrum of ligands having distinct chemical structures and molecular efficacies34. Therefore, 129 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 6 we selected a combination of small molecules, peptide, nanobody, and natural chemokine 130 ligands, each having a distinct pharmacology, to capture the complete landscape of CXCR7 131 activation. Of these, VUF16840 is a small molecule inverse agonist in terms of βarr-coupling, 132 VUN701 is a nanobody -based antagonist, VUF11207 is a CXCR7 -selective agonist, 133 VUF10661 is a small molecule dual agonist at CXCR3 and CXCR7, TC14012 is a cyclic 134 peptide agonist at CXCR7 but an inverse agonist at CXCR4, and CXCL11 is a natural 135 chemokine agonist 34,35 (Figure 1a-b). We first validated these ligands in extensive cellular 136 assays in terms of βarr recruitment, which confirmed their pharmacology and efficacy profile, 137 and also established TC14012 as CXCR7 -selective cyclic-peptide agonist ( Figure 1c and 138 Supplementary Figure S1A-D). We also observed that CXCL11 is nearly as efficacious 139 agonist as CXCL12 albeit with slightly lower potency, and VUN701 and VUF16840 are able to 140 antagonize both, CXCL11 and CXCL12, although only VUF16840 exhibits an inverse agonism 141 for βarr recruitment. Taken together, these ligands provide an opportunity to visualize distinct 142 conformations of CXCR7 and fully understand receptor activation. 143 Structure determination of CXCR7 using cryo-EM 144 In order to determine the structures of agonist -bound CXCR7, we leveraged a previously 145 described antibody fragment, CID24, which binds on the intracellular side of the receptor, and 146 presumably recognizes an active conformation 29. We observed that CID24 is not able to 147 recognize antagonist or inverse -agonist-bound CXCR7 ( Supplementary Figure S2 A), and 148 therefore, we anticipated that capturing the inverse agonist/antagonist-bound CXCR7 will be 149 technically challenging due to the small size of the receptor. Serendipitously however, we 150 observed during cryo-EM data processing that VUF16840-CXCR7 predominantly exhibited a 151 dimeric population yielding a structure at 3.2Å resolution (Figure 1D and Supplementary 152 Figure S3). For the VUN701-CXCR7 complex, both, monomeric and dimeric populations were 153 apparent during cryo -EM data processing, and we determined their structures at 3.6Å and 154 3.7Å, respectively (Figure 1 D and Supplementary Figure S3 ). On t he other hand, the 155 agonist-bound CXCR7 complexes, stabilized by CID24, yielded structures ranging from 2.9Å 156 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 7 to 3.2Å, where the receptor was present mostly as a monomer except VUF10661 (Figure 1D 157 and Supplementary Figure S 5). We observed both, monomer and dimer population for 158 VUF10661-bound CXCR7, and determined their structures at 3.2Å and 3.3Å, respectively 159 (Figure 1D and Supplementary Figure S 4 and S 5). Interestingly, the monomer structure 160 contains CID24 on the intracellular side of the receptor, but the dimer structure did not show 161 discernible evidence of CID24 binding (Figure 1D). All the structures showed clear densities 162 for the bound ligands, majority of the receptor segment, and also distinct lipid molecules bound 163 to the receptor, leading to the discovery of several fundamental insights as described 164 subsequently. The technical details of cryo-EM data collection and processing are outlined in 165 Table S1 , and the cryo-EM densities for the TMs and ligands are presented in the 166 Supplementary Figure S6-S8. Notably, we observed a significantly larger portion of the N -167 terminus of CXCR7 in the CXCL11 -bound structure due to an engagement with the ligand, 168 which was not resolved in other structures as they do not extensively engage the N-terminus 169 of the receptor (Supplementary Figure S9A -B). In addition, we also observed a clear 170 disulfide bond between Cys34N-term and Cys2877.25 in the active structures of CXCR7, while it 171 was absent in the inactive state structures due to a linear shift of the receptor N -terminus 172 (Supplementary Figure S 9C-D). Interestingly, this disulfide bond was not resolved in the 173 previously determined CXCR7 structure in complex with CXCL1229, although it appears to be 174 a critical determinant of receptor stabilization in an active state based on our structural analysis 175 and extensive site-directed mutagenesis studies36. 176 Overall ligand binding to CXCR7 177 Most ligands expectedly engage primarily in the orthosteric binding pocket of the receptor with 178 the exception of VUF10661, which shows a dual binding mode. For example, VUF16840 179 occupies the base of the orthosteric pocket with a distance of ~4Å from the conserved 180 Trp2656.48 and encompasses a buried surface area of ~1600Å 2 (Figure 2a). The orthosteric 181 binding pocket is characterized by the presence of several aromatic amino acid residues such 182 as Tyr51 1.39, Trp100 2.60, His121 3.29, Phe124 3.32, Phe129 3.37, Trp265 6.48, and Tyr268 6.51, and 183 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 8 hydrophobic residues such as Leu128 3.36, Ile132 3.40, Leu305 7.43, and Leu183 4.64. Together, 184 these residues create an optimal environment for the binding of VUF16840 (Supplementary 185 Figure S 10A). Small molecule agonists VUF10661 and VUF11207 also occupy a similar 186 space in the orthosteric binding pocket although they penetrate a little deeper compared to 187 VUF16840 in the pocket ( Figure 2a and Supplementary Figure S 10B-C). As mentioned 188 earlier, VUF10661 is a dual agonist for CXCR3 and CXCR7, and it occupies a similar position 189 in the orthosteric of both, CXCR3 and CXCR7, and engages mostly a similar set of residues 190 (Supplementary Figure S 10D-E). This is in excellent agreement with our site -directed 191 mutagenesis data showing a comparable impact of analogous residues in both receptors on 192 VUF10661 binding37. On the other hand, CXCL11 exhibits a two -site binding mechanism on 193 CXCR7, wherein the core domain engages the N -terminus and ECL2 of the receptor, while 194 the N -terminus penetrates deep into the orthosteric binding pocket ( Figure 2 A and 195 Supplementary Figure S11A). Overall, CXCL11 binding encompasses more than 3,000Ų 196 buried surface area, and is stabilized by an array of hydrogen bonds, hydrophobic interactions, 197 salt bridges, and polar contacts (Supplementary Figure S11B). A partially negatively charged 198 segment in the N -terminus of CXCR7 spanning residues Ser 23-Met37 docks into a positively 199 charged groove formed by helix-1 and β3 strand of the CXCL11 core domain as a part of the 200 first binding site (Supplementary Figure S11B). On the other hand, the proximal N-terminus 201 of CXCL11 makes a series of contacts with distinct residues in TM1-7 in the orthosteric binding 202 pocket as the second binding site on the receptor. 203 Structural alignment of CXCL11-bound CXCR7 complex with the previously reported 204 CXCL12-CXCR7 structure indicates that despite adopting broadly similar binding orientations, 205 the chemokine core domains are laterally displaced relative to each another (Supplementary 206 Figure S11 C). Structural superimposition of currently available C -X-C type chemokine 207 receptors in complex with different C-X-C chemokines indicates that the chemokines adopt an 208 overall similar binding pose with minor localized variations. In contrast, in the CXCL11–CXCR7 209 complex, the chemokine core domain is oriented at ~90°, while the N -terminus adopts an 210 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 9 open, slender conformation that penetrates more deeply into the orthosteric pocket of CXCR7 211 (Supplementary Figure S 11D). This is also reflected in a comparison of CXCL11 -CXCR7 212 with previously determined structure of CXCL11 -CXCR338, which reveals a large shift in the 213 positioning of the same ligand i.e. CXCL11 on the two receptors ( Supplementary Figure 214 S11E). Specifically, the core domain of CXCL11 on CXCR7 adopts a torsional conformation 215 rotated by approximately 90°, which facilitates more extensive interactions with the receptor 216 N-terminus, compared to CXCR3 ( Supplementary Figure S 11E). Moreover, while N -217 terminus of CXCL11 aligns closely within the orthosteric pocket of both receptors, the N-loop 218 residues in CXCL11 diverge by nearly 180° between the two receptors (Supplementary 219 Figure S11E). These pronounced differences in the binding of CXCL11 to CXCR3 vs. CXCR7 220 likely contribute to their functional divergence in terms of transducer -coupling and functional 221 outcomes. 222 In case of VUN701, the extended CDR3 penetrates into the orthosteric pocket of the 223 receptor, and surprisingly, the N-terminus of CXCR7 wraps around the CDR3 to stabilize the 224 binding further (Supplementary Figure S11F). This is in stark contrast with CXCL11-CXCR7 225 structure, wherein the N-terminus of the receptor adopts a vertical orientation to interact with, 226 and stabilize, the core domain of CXCL11 ( Supplementary Figure S 11F). In addition, 227 CXCL11 penetrates approximately 8Å deeper in the orthosteric binding pocket compared to 228 the CDR3 of VUN701 (Figure 2A ). This observation underscores a relatively superficial 229 binding of VUN701 and likely suggests that it antagonizes CXCR7 by sterically obstructing the 230 entry and docking of agonists including chemokines. Interestingly, TC14012 adopts a slender 231 conformation despite being a cyclic peptide, and inserts deep into the orthosteric pocket of the 232 receptor, almost to a comparable depth with CXCL11, and makes similar interactions as that 233 of the N-terminus of CXCL11 (Figure 2A and Supplementary Figure S11G). 234 Distinct ionic-locks stabilize inactive CXCR7 conformation 235 Class A GPCRs typically harbor a TM3 -TM6 ionic-lock where Arg 3.50 of the conserved DRY 236 motif in TM3 forms an interhelical salt-bridge with a conserved Asp/Glu6.30 in TM6, and helps 237 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 10 maintain the receptors in their basal conformation 39,40. This salt -bridge is disrupted upon 238 receptor activation leading to large TM movements, especially TM6, leading to subsequent 239 coupling and activation of heterotrimeric G-proteins41,42. Interestingly, in CXCR7, the position 240 6.30 in TM6 is occupied by a lysine residue, which makes the formation of a typical salt-bridge 241 with Arg3.50 unfeasible, and VUF16840-bound structure indeed corroborates the same (Figure 242 2B). A previous study has proposed that Glu 6.29 in CXCR7 may serve as a surrogate for 243 Asp/Glu6.30 present in other class A GPCRs, and thereby, still allow the formation of a slightly 244 altered TM3-TM6 ionic-lock43. In contrast, VUF16840-bound CXCR7 structure elucidates that 245 not only Arg3.50 is oriented towards the receptor core and away from Glu6.29 but the side-chain 246 of Glu6.29 is also oriented in an opposite direction, which in turn makes any contact with Arg3.50 247 implausible (Figure 2B). Instead, we observed that Glu 6.29 forms a distinct tripartite lock by 248 making a salt -bridge with Arg 6.34, which in turn interacts with Tyr 5.58 via an ionic -interaction 249 (Figure 2C). In addition, we also observed that Tyr 7.53 in TM7, which is also a part of the 250 conserved NPxxY motif, interacts with Asn8.47 in helix 8, and this engagement is likely to restrict 251 the mobility of TM7 and helix8 in the basal conformation of CXCR7 (Figure 2C). Interestingly, 252 these two distinct ionic-locks are also conserved in both, monomeric and dimeric structures of 253 VUN701-bound CXCR7 (Supplementary Figure S12A-B), and therefore, establish them as 254 a bonafide feature that maintains CXCR7 in its basal conformation. 255 Atypical activation of CXCR7 256 Structural comparison of the inactive and active CXCR7 structures reveals large 257 conformational changes in the receptor, especially, an outward movement of cytoplasmic side 258 of TM6 (~7Å) accompanied by a significant rotation of helix 8 (~50°) towards the cytoplasmic 259 core of the receptor, and a relatively smaller movement of TM7 (~4Å) (Supplementary Figure 260 S12C). Unlike other class A GPCRs, the movement of TM5 upon CXCR7 activation appears 261 to be restricted (<1Å). The movement of cytoplasmic end of TM6 upon receptor activation 262 appears to be essential for CID24 binding, and the position of TM6 in the inactive state is likely 263 to clash with CID24 engagement. Interestingly, upon receptor activation, Tyr5.58 undergoes an 264 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 11 upward rotation (~75°) towards the receptor core, while Arg 6.34 and Glu 6.29 shift by 265 approximately 5 -6Å away from the core, and these structural changes result in complete 266 disruption of the tripartite ionic-lock (Figure 2B). Concomitantly, Tyr7.53 in TM7 and Asn8.47 in 267 helix8 also undergo a large upward and downward rotation, respectively, which disrupts their 268 interaction by positioning their side -chains in the opposite directions ( Figures 2 B). 269 Collectively, these structural changes in CXCR7 disrupt the non -canonical ionic-locks and 270 thereby, relieve the restraints on TM6, 7 and helix8, and consequently, result in receptor 271 activation. Importantly, these activation -dependent structural changes in CXCR7 are 272 consistent across all the active state structures stabilized by different agonists, and therefore, 273 unequivocally establish the mechanistic basis of receptor activation. Interestingly, the 274 cytoplasmic surface of CXCR7 upon activation also reveals two distinctive features compared 275 to prototypical GPCRs. First, the overall dimension of the cytoplasmic cavity generated upon 276 receptor activation is significantly smaller than that observed for prototypical GPCRs including 277 other chemokine receptors such as CXCR3, which may not be sufficient to support a stable 278 interaction with the α5 -helix of the G -proteins ( Figures 2 E and Supplementary Figure 279 S12D). This constricted intracellular pocket likely results from a relatively smaller outward 280 movement of TM5 and TM6 in CXCR7, compared to prototypical GPCRs. Second, the 281 cytoplasmic surface of CXCR7 in active conformation lacks a positively charged patch of 282 residues that is apparent in other chemokine receptors and accommodates the αN helix of Gβ 283 subunit (Figure 2E and Supplementary Figure S12E). Taken together, these two features 284 provide a plausible explanation for the lack of G-protein-coupling and activation by CXCR7. 285 Distinct dimerization of CXCR7 286 Several class A GPCRs have a propensity to homo - and hetero-dimerize as assessed using 287 cellular assays, and this impacts their functional outcomes in terms of trafficking and 288 downstream signaling44-46. CXCR7 is reported to form homodimers as well as heterodimers 289 with other chemokine receptors such as CXCR4 with distinct functional consequences47,48. As 290 mentioned earlier, we have captured CXCR7 dimeric arrangement in three different structures 291 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 12 in complex with three distinct ligands (Supplementary Figure S13A-B). The dimeric structure 292 in complex with VUF16840 encompasses a dimension of ~80Å along the longer axis and ~65Å 293 along the broader axis, and harbors a buried surface area and interface area of ~2200Å2 and 294 ~1100Å2, respectively. The dimer interface is mediated primarily by TM1 and helix8 from each 295 protomer, wherein TM1 is positioned diagonally at an angle of ~90º while helix8 is aligned 296 almost parallel to each other ( Figure 3A-B). Three hydrogen bonds and a large number of 297 non-bonded contacts stabilize the dimer interface suggesting a robust interaction. These 298 include, for example, hydrophobic interactions of Met59 1.47 and Trp671.55 from one protomer 299 with Leu912.51 and Ile601.48/Val641.52 in the other protomer (Figure 3C) and hydrogen binding 300 between Ser631.51 from each protomer across the interface (Figure 3C). In addition, the helix8 301 interface involves a hydrogen bond between Arg3238.51 of one protomer with Met3278.55 of the 302 other as well as a cation -π interaction between Arg320 8.48 and Phe330 8.58 of the two 303 protomers, respectively ( Figure 3C). Interestingly, the dimeric assembly of VUN701 -bound 304 CXCR7 closely resembles that observed in VUF16840-CXCR7 complex, with the dimerization 305 interface mediated primarily by TM1 and helix8, and a comparable cross -angle between the 306 protomers (Figure 3A-B). 307 On the other hand, the dimeric assembly of VUF10661-bound CXCR7 is distinct from 308 that observed in the VUF16840/VUN701-bound CXCR7 dimers. Here, the two protomers align 309 parallel to each other with TM4 and TM5 constituting the primary dimer interface with minor 310 contribution from the cytoplasmic region of TM3 (Figure 3A-B). The absence of CID24 in this 311 structure suggested an inactive-like receptor conformation, and further structural analysis as 312 discussed subsequently indeed corroborates this. The overall buried surface area at the dimer 313 interface is approximately 1000Ų, and Tyr181 4.62 of each protomer forms a hydrogen bond 314 with Glu2075.33 of the opposing protomer, and vice-versa. A prominent hydrophobic cluster at 315 the interface further contributes to the stability of the oligomeric complex, which includes 316 Leu1674.48, Val1704.51, Pro1784.59, Trp2085.34, and Val2185.44 from both protomers (Figure 3C). 317 The overall conformation of each protomers in the dimeric structure are nearly identical with a 318 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 13 rmsd of <1Å. Interestingly however, while the receptor conformations between VUN701 -319 CXCR7 monomer vs. dimer are mostly similar except a slight shift in CDR3 positioning 320 (Supplementary Figure S 13C), there are stark differences between the receptor 321 conformation between the monomer vs. dimer of VUF10661 -bound CXCR7 as discussed 322 subsequently. 323 Interaction of native lipids to CXCR7 and a molecular glue-like mechanism 324 Several cryo -EM structures of GPCRs determined previously have shown a presence of 325 bound cholesterol molecules, and for some, specific functional impact has also been 326 demonstrated using cellular and biophysical assays49,50. Surprisingly, the cryo-EM structures 327 of CXCR7 determined here exhibit clear densities consistent with several lipid molecules, not 328 only at the dimeric interface but also bound to the receptor monomers (Figure 4A-D). As there 329 is no prior information on the interaction of non -cholesterol lipids to CXCR7, their modelling 330 and placement are guided by the observed densities, assignment in other membrane protein 331 structures in the Protein Data Bank, and experimentally confirmed abundance in cultured 332 insect cells. In the active state CXCR7 structures, there are two distinct lipid densities, which 333 correspond to dipalmitoyl phosphatidic acids (LPP) molecules . The first LPP molecule 334 occupies a hydrophobic sub-pocket formed with contributions of residues from TM2, TM3, and 335 TM4 (Figure 4A-B). The phosphate head group of this LPP molecule forms hydrogen bonds 336 with Lys73 ICL1 and Arg1624.43, while the aliphatic chains engage with several residues from 337 TM2, TM3, and TM4 via hydrophobic interactions (Figure 4A and 4D ). The second LPP 338 molecule occupies a similar hydrophobic sub -pocket composed of residues from TM3, TM4, 339 and TM5, and it is stabilized by a series of non-bonded contacts with the receptor. Importantly, 340 in the previously published CXCL12 -CXCR7 structure 29, two cholesterol molecules were 341 modelled in the densities corresponding to the lipid moieties ( Supplementary Figure S14A-342 B). However, the structures presented here at improved resolution clearly suggest that these 343 densities correspond to phospholipids with two aliphatic chains. In order to corroborate these 344 findings, we carried out mass spectrometry (MS) -based lipidomics analysis to identify the 345 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 14 specific lipid moieties that are bound to CXCR7. Interestingly, we identified several 346 phospholipids that are abundantly present in the Sf9 membrane, and some of these are 347 enriched in purified receptor sample (Supplementary Figure S14C). Importantly, we noticed 348 a robust enrichment of phosphatidic acid (16:0/16:1) in the purified receptor sample, despite 349 relatively lower abundance in the membranes. This is in an excellent agreement with the 350 placement of LPP in the cryo -EM density, which is repeatedly observed in every CXCR7 351 structure determined here including different batches of purified receptor. 352 In addition, we also observed two putative phosphatidylcholine (PC) molecules at the 353 dimer interface in the inactive state structures, each coming from the individual protomers 354 (Figure 4B and Supplementary Figure S15A-B). The lipid-binding microenvironment at the 355 dimeric interface is predominantly hydrophobic, formed by several residues contributed 356 primarily by TM2 and TM3 (Supplementary Figure S 15C) These two phosphatidylcholine 357 molecules located at this interface form extensive hydrophobic interactions with residues from 358 both protomers, thereby, effectively stitching the two protomers together by serving as a 359 molecular glue and stabilizing the dimeric arrangement. In addition to these lipids, we also 360 observed additional density corresponding to a lipid moiety in some of the structures such as 361 VUF11207-bound CXCR7 and VUF16840/10661-bound dimers, which we infer to represent 362 myristic acid ( Figure 4 F). For example, in the VUF10661 -CXCR7 structure, myristic acid 363 molecule is present at the dimeric interface, and it is positioned between the cytoplasmic 364 segment of TM3 of one protomer and the ICL2 of the adjacent protomer. It is further stabilized 365 via non -bonded interactions involving Tyr143 3.51 in the DRY motif of one protomer and 366 Arg156ICL2 of the other protomer. We did not model lipid density in the VUN701-bound CXCR7 367 structures due to relatively lower resolution , although, at lower contour levels, the 368 corresponding density appears to emerge. We also note that a precise identification of these 369 lipids, and their functional roles receptor activation and signaling remains to be determined in 370 future studies. 371 A novel allosteric site and an intermediate CXCR7 conformation 372 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 15 As mentioned earlier, the monomeric structure of VUF10661-bound CXCR7 exhibits an active 373 conformation of the receptor as reflected by the large conformational changes compared to 374 the inactive state, and VUF10661 occupies the orthosteric binding pocket ( Supplementary 375 Figure S10B and S10D). Interestingly however, in the dimeric structure, VUF10661 binds to 376 an extra -helical allosteric site located at the dimer interface formed by the extracellular 377 segments of TM3, TM4, and part of ECL2 (Figure 5A-B). Specifically, each copy of the ligand 378 interacts with Glu1143.22, Lys1183.26, Val1193.27, and Thr1804.61 of one protomer and Glu2075.33 379 of the other (Figure 5C). Additionally, we also observed that the phosphatidylcholine molecule 380 is positioned adjacent to VUF10661, and engages with Lys206 5.32, Ile2105.36 and Leu2145.40, 381 thereby, providing further stabilizing the binding of the ligand to the allosteric site and serving 382 as a molecular glue -like (Figure 5C). Interestingly, both CXCR7 protomers in this structure 383 are present in an inactive-like conformation, and they align well with the inverse-agonist-bound 384 conformation with an overall rmsd of <1Å. Still however, some notable conformational 385 differences are observed in the extracellular loop regions. For example, ECL1 exhibits an 386 outward linear displacement of approximately 11Å (measured at the Cα atom of Gln109) while 387 the β-hairpin of ECL2 shifts laterally away from the orthosteric pocket by approximately 6Å. 388 Surprisingly, a part of ECL2, spanning the residues from Tyr 195 to His 203, folds back and 389 penetrate the orthosteric pocket ( Figure 5 D). Interestingly, ECL2 fold -back structure and 390 positioning in the orthosteric binding pocket is reported as a self -activating mechanism for a 391 couple of orphan receptors such as GPR161 51 and GPR52 52. Contrary to this, ECL2 in 392 VUF10661-bound CXCR7 appears to block the orthosteric binding pocket of the receptor, and 393 thereby, pushing the ligand to an allosteric site, stabilizing an inactive -like conformation 394 (Supplementary Figure S 16A). To our knowledge, this is the first example of an 395 autoinhibitory mechanism mediated by ECL2 in any GPCR structure reported till date, and it 396 underscores the versatility and ligand-specific conformational adaptation encoded by the 7TM 397 scaffold. Moreover, the newly identified ionic -locks in CXCR7 are present in an intermediate 398 state in this structure. Specifically, the Arg2516.34-Glu2466.29 and Tyr315 7.53-Asn3198.47 399 interactions are maintained, while Arg2516.34 is positioned more than 4Å away from Tyr2325.58, 400 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 16 thereby, weakening the tripartite lock compared to the fully inactive structure (Supplementary 401 Figure S1 6B). We acknowledge that future studies are required to further explore the 402 functional consequences of this allosteric site in CXCR7 and the role of ECL2 in receptor 403 activation. 404 CXCR7 as an atypical opioid receptor 405 In addition to chemokines, CXCR7 is also proposed to recognize opioid peptides and 406 scavenge them to potentially modulate their physiological availability for the canonical opioid 407 receptors31-33. In order to assess the specificity of CXCR7 for opioid peptides, we screened a 408 library of >1000 biologically active peptides and analogues, covering a wide variety of ligand 409 families including >150 orphan peptides, on cells expressing CXCR7 using a Nano -BRET-410 based ligand competition assay (Figure 6A). We observed that only about a dozen different 411 peptides were able to displace a significant level of CXCL12 from CXCR7, and most of these 412 were opioid peptides from the enkephalin and dynorphin family ( Figure 6A). We selected a 413 set of these peptides and measured their ability to induce βarr1/2 recruitment in a dose 414 dependent manner, which further confirmed their robust agonism at CXCR7 (Supplementary 415 Figure S17A-B). 416 Next, we reconstituted CXCR7 complexes with dynorphin A (dynorphin), BAM22, and 417 an adrenorphin-based CXCR7-selective synthetic peptide (LIH383), and stabilized them using 418 CID24 for structural analysis (Supplementary Figure S17C-E). We determined the structures 419 using cryo-EM at resolutions ranging from 3.2Å to 3.7Å (Figure 6B-D). The technical details 420 of cryo-EM data collection and processing are presented in Table S1 and Supplementary 421 Figure S18A-C, and the cryo -EM densities for the TMs and ligands are presented in the 422 Supplementary Figure S1 9A-C. Overall, these structures reveal a clear binding of the 423 ligands, and the lipid moieties in a similar position as described earlier. As dynorphin-bound 424 CXCR7 structure was resolved at a higher resolution, we primarily used this for subsequent 425 analysis as described below. In this structure, the first ten residues of dynorphin were clearly 426 discernible, which adopted a “Z -shaped” conformation with residues Gly3 -Phe4-Leu5-Arg6 427 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 17 taking a helical turn, which penetrates deep into the orthosteric pocket ( Figure 6 E). In 428 particular, the N -terminal tyrosine residue of dynorphin (Tyr1 Dyn) is oriented vertically 429 downwards, forming an extensive receptor -peptide interface with an interface area of more 430 than 800Å2 and buried surface area of about 2,200Å2. We observed that three hydrogen bonds 431 and multiple non-bonded contacts stabilize the positioning of dynorphin within the amphipathic 432 orthosteric pocket. Tyr1Dyn is nestled within a hydrophobic patch formed by Phe124 3.32, 433 Leu1283.36, and Phe129 3.37 of CXCR7 ( Figure 6 F). Molecular interactions that anchor 434 dynorphin include the cationic hydroxyl of Tyr1 Dyn forming ionic bonds with His121 3.29 and 435 Gln3017.39, while Tyr2686.51 engages with the backbone nitrogen of Tyr1 Dyn (Figure 6E). The 436 Phe4Dyn residue participates in hydrophobic interactions with Leu297 7.35, while the guanidino 437 group of Arg6Dyn forms a cation-π interaction with Trp110ECL1 (Figure 6E). In addition, Arg7Dyn 438 and Arg9Dyn establish hydrogen bonds with Asp2756.58 in CXCR7 Figure 6E). 439 A direct comparison of dynorphin -bound CXCR7 with the inverse -agonist-bound 440 CXCR7 displays similar conformational changes in the receptor as that observed for CXCL11, 441 indicating that dynorphin is able to fully activate the receptor ( Figure 6G). In line with this 442 interpretation, the structures of CXCR7 in complex with dynorphin and CXCL11 look nearly -443 identical with an overall rmsd of <1Å (Supplementary Figure S 20A). Moreover, dynorphin 444 and CXCL11 are positioned in the orthosteric binding pocket at comparable depths with similar 445 spatial placement of the terminal ligand residues (Tyr1 in dynorphin vs. Phe1 in CXCL11), 446 leading to similar interaction patterns ( Supplementary Figure S 20A). However, the three 447 residues following the two terminal residues i.e. Gly3-Phe4-Leu5 in dynorphin -CXCR7 and 448 Met3-Phe4-Lys5 in CXCL11, deviate by ~5.5Å although the subsequent stretches of both 449 peptides realign again within the orthosteric pocket (Supplementary Figure S20A). While the 450 backbone atoms of Arg 7 and Arg9 in dynorphin occupy similar sites as Arg6 and Arg8 in 451 CXCL11, their side chains are oriented in opposite direction s, thereby, engaging different 452 receptor residues. Specifically, Arg6 and Arg9 in dynorphin interact with Asp2756.58 in CXCR7 453 via hydrogen bonds, whereas Arg6 and Arg8 in CXCL11 engage with Asp1794.60 and Ile2796.62 454 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 18 in the receptor via a hydrogen bond and hydrophobic interaction , respectively 455 (Supplementary Figure S20B). These structural observations underline an overall conserved 456 mode of interaction of dynorphin and CXCL11 with CXCR7 with respect to the orthosteric site, 457 and observed differences in their interaction networks are likely to fine -tune their respective 458 potencies for βarr recruitment. 459 Structural comparison of dynorphin-bound CXCR7 with that of kappa -opioid receptor 460 (kOR)53 reveals some interesting patterns of ligand -interaction and receptor conformation. 461 Overall, the two receptors are similar to each other in dynorphin-bound conformation with an 462 overall rmsd of <1Å ( Supplementary Figure S 20C), and dynorphin also penetrates to a 463 similar depth in the orthosteric binding pockets (Figure 6F). However, the side-chain rotamer 464 of Tyr1 in dynorphin differs significantly. In kOR, Tyr1 points in an upward direction while in 465 CXCR7, it points downward, which in turn, results in a divergent interaction network (Figure 466 6F). Despite this deviation at the N-terminus, the other end of dynorphin appears to converge 467 in a similar fashion on the two receptors as reflected by an identical interaction of Arg9 in 468 dynorphin with Asp297 6.58 and Asp275 6.58 in the kOR and CXCR7, respectively 469 (Supplementary Figure S 20C-D). Importantly, distinct structural arrangements are also 470 evident in the intracellular loops (ICLs) of the two receptors in complex with dynorphin. For 471 example, in the dynorphin-bound kOR structure, the ICL1, ICL2, and ICL3 extend outward 472 from the receptor core, creating a wide cytoplasmic pocket that accommodates heterotrimeric 473 G-proteins (Supplementary Figure S 20E). On the other hand , in the dynorphin -bound 474 CXCR7 structure, these ICLs shift linearly towards the receptor core, resulting in a constricted 475 cytoplasmic pocket as observed in the other agonist -bound structures (Supplementary 476 Figure S2E). 477 The other two opioid peptides namely BAM22 and LIH383 also exhibit an overall 478 similar binding to CXCR7. For example, similar to dynorphin, residues “Gly3-Phe4-Met5-Arg6” 479 of LIH383 form a helical turn, and the overall conformation of LIH383 within the orthosteric 480 pocket closely mirrors that of dynorphin and engages a comparable buried surface area 481 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 19 (Supplementary Figure S 20F). Specifically, Phe1 in LIH383 penetrates deepest into the 482 orthosteric pocket, with its phenyl ring occupying a spatial position analogous to that of Tyr1 483 in dynorphin, and the following residues also interact in analogous manner to that of dynorphin, 484 leading to receptor activation as assessed by outward movement of TM6 and TM7 485 (Supplementary Figure S 20G). Taken together, these findings unequivocally establish 486 CXCR7 as an atypical opioid receptor and thereby, expand the opioid receptor subfamily 487 beyond the conventional receptors. 488 A pivotal ligand-recognition switch in the chemoattractant receptors 489 The analysis of CXCR7 structures expectedly revealed several residues that are involved in 490 ligand interactions, and some of these appear to be mostly conserved across different ligands 491 while others seem to show a ligand -specific pattern of interaction ( Figure 7 A). Of these, 492 Tyr2686.51 stands out in particular as it appears to make a contact with nearly every ligand 493 used here for structural analysis. We generated a series of alanine mutants of CXCR7 by 494 replacing the key residues involved in ligand-interaction, and measured agonist-induced βarr 495 recruitment (Figure 7B and Supplementary Figure S21A-B). We observed that the mutation 496 of some of the residues such as Tyr2686.51 affected βarr recruitment across all ligands while 497 the others impacted only selected ligands (Figure 7B and Supplementary Figure S21C-H). 498 For example, the mutation of Leu1283.36 resulted in a significant attenuation of βarr recruitment 499 for CXCL11 and dynorphin but not for CXCL12 and VUF11207. These findings corroborate 500 the structural observations, and underscore ligand-specific receptor contacts potentially fine-501 tuning downstream functional responses. Considering the robust impact of Tyr2686.51 502 mutation, we explored if this residue is conserved across other chemokine and non-chemokine 503 peptide receptors. We observed that the majority of the chemokine receptors and all three 504 complement anaphylatoxin receptors harbor Tyr6.51, and it also makes a robust interaction with 505 the corresponding natural agonists of these receptors (Figure 7C). On other hand, only very 506 few of the non -chemokine peptide receptors harbor the conserved Tyr 6.51 in TM6. Taken 507 together with site-directed mutagenesis data and structural analysis, we propose that Tyr 6.51 508 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 20 serves as a pivotal switch for ligand -binding and activation of a broad repertoire of chemo -509 attractant receptors, underscoring a previously unknown conserved structure feature of these 510 receptors. 511

512 The conceptual framework of biased agonism has redefined GPCR pharmacology, signaling, 513 and therapeutic design considering the inherent potential to minimize the side -effects of 514 GPCR-targeting drugs51. Biophysical and structural studies using synthetic biased ligands and 515 their cognate receptors such as the angiotensin II subtype 1 receptor (AT1R) have elucidated 516 distinct conformational changes in the receptor as an underlying principle to impart signaling-517 bias12-14. On the other hand, the molecular basis of distinct activation of naturally -encoded, 518 signaling-biased 7TMRs, also referred to as Arrestin -Coupled Receptors (ACRs), which lack 519 measurable G -protein activation, remains mostly speculative. Earlier studies have 520 unequivocally demonstrated a lack of functional engagement of heterotrimeric G -proteins to 521 ACRs, and have suggested distinct conformational signatures in βarrs upon their interaction 522 with ACRs vs. GPCRs19. A previous study on CXCR7 has proposed distinct conformation and 523 structural features of the 2 nd and 3 rd intracellular loops (ICL2 and ICL3) as a plausible 524 mechanism to preclude G -protein-coupling29. However, replacing these loops either 525 individually or in combination with that of a prototypical chemokine receptor, CXCR2, did not 526 allow any measurable gain in G-protein-coupling. Therefore, it is likely that additional structural 527 features may contribute to the inherent defect in G-protein activation. In fact, our study reveals 528 a constricted intracellular pocket in CXCR7, resulting particularly from a relatively smaller 529 movements of TM5 and 6 upon activation, compared to other GPCRs. In addition, unlike 530 prototypical GPCRs, the lack of positively charged patch on the intracellular surface in CXCR7 531 is also readily apparent. Taken together, these distinctive features in activated CXCR7 may 532 not be suitable for efficient docking and alignment of the α5 and αN helices of the Gα subunit, 533 respectively, resulting in the lack of G -protein activation. These observations provide a 534 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 21 testable hypothesis for future experiments, and may also guide the investigation on other 535 ACRs going forward (Figure 8). 536 The functional specialization of ACRs in terms of selective βarr -coupling potentially 537 argues for a distinct molecular mechanism of activation compared to prototypical GPCRs, 538 possibly at the level of receptor conformation and micro-switches. This is further substantiated 539 for CXCR7 considering that broadly conserved GPCR sequence motifs such as DRY and 540 NPxxY are conserved in this receptor. Due to the lack of CXCR7 structure in an inactive state, 541 previous inferences about receptor activation have been drawn using the inactive state 542 structures of other chemokine receptors as a reference. However, this is likely to provide only 543 a close approximation at best, and therefore, should be interpreted with caution. For example, 544 a previous study using HDX -MS and site-directed mutagenesis has proposed that Arg 3.50 in 545 TM3 of CXCR7 may form an ionic lock with Glu 6.29 in TM6, which serves as an alternative to 546 broadly conserved TM3 -TM6 ionic in GPCRs formed between Arg 3.50 in TM3 and Glu 6.30 in 547 TM643. Our study now demonstrates that the formation of TM3 -TM6 ionic lock in CXCR7 is 548 structurally not plausible, at least based on the receptor conformations captured here using 549 cryo-EM. This, in turn, raises the possibility of alternative inter -helical restraints to maintain 550 the receptor in its basal state. Indeed, CXCR7 structures highlight two novel inter-helical ionic-551 locks spanning TM5-TM6 and TM7-H8, which are present in the antagonist/inverse -agonist-552 bound states of the receptor, but are disrupted upon agonist-binding and receptor activation. 553 This observation also aligns with the previous experimental data where the mutation of Glu6.29 554 to alanine imparts higher constitutive activity of CXCR7 compared to the wild-type receptor43. 555 We also note that unlike CXCR7, other ACRs lack the conserved sequence motifs such as 556 DRY and NPxxY, and thus, may utilize an analogous mechanism as described here. However, 557 direct structural visualization and biochemical analysis of these additional ACRs is essential 558 to probe this hypothesis. 559 Dimerization of class A GPCRs are well documented in cellular context using 560 proximity-based assays, and in some cases, also appear to have a direct functional 561 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 22 relevance44-46. There is a growing list of GPCR cryo -EM structures that also capture GPCR -562 G-proteins and GPCR-βarr complexes in dimeric arrangements 6. While some of these may 563 simply represent concentration-dependent non-physiological dimers, others appear to use a 564 specific interface and may have direct functional consequences. On the other hand, direct 565 visualization of distinct dimeric interface and/or arrangements adopted by the same receptor, 566 wherein the protomers are stitched together by a native phospholipid originating from the 567 membrane bilayer, is rather scarce. Previous studies using cell -based experiments have 568 reported homo-dimerization of CXCR7 as well as hetero-dimerization with other GPCRs, and 569 some of these also appear to be sensitive to ligand stimulation 47,48. Moreover, in case of 570 CXCR7-CXCR4 heterodimer, CXCR7 is also proposed to modulate the activation and 571 signaling of CXCR4, presumably by scavenging the ligand and/or βarrs47,48. CXCR7 structures 572 resolved here show three distinct dimeric arrangements, each stabilized via native 573 phospholipid molecules positioned at the inter -protomer interface. While two of them utilize 574 primarily TM1-TM1 interface, the inter -protomer rotation are significantly different between 575 these two dimers, and the other dimeric assembly is stabilized by TM4/5-TM4/5 interface. Still 576 however, the overall positioning of the native phospholipid molecule, designated as 577 phosphatidylcholine here based on the resolved EM density, is nearly identical in all three 578 dimers. A striking observation is that all three dimers exhibit an inactive conformation of the 579 receptor although in one of them, the receptor is bound to an agonist albeit at a non-orthosteric 580 binding pocket. On the other hand, all the active state structures are predominantly present 581 as a monomer, which was also reflected in the 2D class averages during data processing. It 582 would be interesting to probe in future studies, especially in cellular context, if the propensity 583 of receptor dimerization correlates with the transition from inactive to active state (Figure 8). 584 Allosteric sites on GPCRs are emerging as a hot-spot for therapeutic targeting as they 585 can be leveraged to modulate receptor signaling, presumably with a better precision than 586 orthosteric binding pocket54-56. Binding of VUF10661, a dual agonist of CXCR3 and CXCR7, 587 to an extra-helical allosteric site is intriguing, especially considering that it can also bind to the 588 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 23 orthosteric pocket of CXCR7 and CXCR337. Previous studies have also revealed dual binding 589 mode of ligands for the muscarinic M5 receptor 57 and the bitter taste receptor TASR2 58. 590 However, the ability of VUF10661 to stabilize an active receptor conformation when occupying 591 the orthosteric pocket vs. an intermediate, inactive-like conformation from the allosteric site is 592 unique and first -of-its-kind efficacy switch to the best of our knowledge. Future studies are 593 required to probe the unanticipated ECL2 conformation in this structure that blocks the 594 orthosteric site, and thereby, likely pushes VUF10661 to occupy the allosteric site on the 595 receptor. Nonetheless, this self-blocking mechanism is in stark contrast with ECL2-mediated 596 self-activating mechanism observed for a couple of class A GPCRs reported recently 58, and 597 in turn, underscores a remarkable versatility in structural mechanisms inherently encoded by 598 7TM scaffold. 599 The first evidence of opioid peptide recognition by CXCR7 and an ensuring modulation 600 of emotional behaviour circuitry via adrenal -opioid signaling cascade surface more than a 601 decade ago 31. Subsequently, a comprehensive study reaffirmed the ability of CXCR7 to 602 recognize multiple opioid peptides, which in turn results in βarr recruitment, potentially leading 603 to opioid peptide scavenging 32,33. This is analogous to chemokine scavenging activity 604 proposed for ACKRs, and may in turn be a natural mechanism to regulate opioid peptide 605 availability in physiological context. Still however, follow-up studies to exploit CXCR7 as a non-606 canonical opioid receptor from therapeutic perspective have remained mostly non -existent, 607 possibly due to lack of a direct visualization of the binding to opioid peptides to CXCR7. We 608 now demonstrate the CXCR7 not only selectively recognizes primarily opioid peptides from a 609 large library of biologically active peptides, but also that the binding of opioid peptides such as 610 dynorphin and BAM22 fully activates the receptor. These structural observations should 611 alleviate any doubts about CXCR7 being a bonafide receptor for opioid peptides, and thereby, 612 expand the opioid receptor subfamily beyond conventional opioid receptors namely the μ-OR, 613 κ-OR, and δ -OR53. More importantly, the structural template of LIH383 -CXCR7 should also 614 facilitate the design and optimization of the new generation of synthetic molecules that are 615 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 24 more suitable from therapeutic perspective by modulating the opioid concentration under in-616 vivo condition for their cognate receptors. 617 In summary, our study presents a series of novel insights into atypical CXCR7 618 activation and functional divergence, and also unravels novel allosteric sites for ligands and 619 native phospholipids. Importantly, our findings also firmly establish CXCR7 as a non-canonical 620 opioid receptor, thereby, expanding the endogenous repertoire of opioid receptors beyond the 621 conventional members. These findings have direct and immediate implications to better 622 understand the framework of biased GPCR signaling and designing novel therapeutics. 623

624 Research in A.K.S.’s laboratory is supported by the Senior Fellowship of the DBT Wellcome 625 Trust India Alliance (IA/S/20/1/504916) awarded to A.K.S., the Science and Engineering 626 Research Board (SPR/2020/000408 and IPA/2020/000405), the Indian Council of Medical 627 research (F.NO.52/15/2020/BIO/BMS), and IIT Kanpur. A.K.S. is the Sonu Agrawal Memorial 628 Chair Professor. All the cryo -EM data collection described here was carried at the ANRF -629 sponsored National cryo -EM Facility at IIT Kanpur (IPA/2020/000405). This study was also 630 supported in part by the Luxembourg Institute of Health (LIH) through the NanoLux Platform, 631 the Cancer Foundation Luxembourg, Luxembourg National Research Fund (INTER/FNRS 632 CXCL12 20/15084569, CORE IMPACTT C23/BM/18068832, INTER/AUDACE/25/19352183 633 and AFR Hope-ioid 17191672) to AC, Natural Science Foundation of China (grant #22077012) 634 to XC, and a Wellcome Trust grant (number 221795/Z/20/Z) to CVR. We thank Dr. Martin 635 Gustavsson for kindly providing the CID24 expression plasmid and Calvin D’Souza for help 636 with CID24 purification. 637 Authors’ contributions 638 NB and AR purified the receptor, prepared, and characterized the complexes for cryo -EM 639 analysis with help from MK and AD; MG and RB processed the cryo-EM data, determined the 640 structures, and carried out structural analysis with help from NR; SM and AD characterized 641 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 25 the ligands and receptor mutants in cellular assays with help from SS and DT; XG synthesized 642 and characterized VUF16840 with help from HG under the supervision of XC; MM and CBP 643 carried out the peptide library screening under the supervision of AC; all authors contributed 644 to manuscript writing and editing; AKS supervised and coordinated the overall study. 645 Conflict of interest 646 The authors declare no conflict of interest. 647 Data availability 648 The cryo-EM maps and structures have been deposited in the PDB and EMDB with accession 649 numbers 9WLD and EMD -66055 for CXL11 -CXCR7-CID24, 9WLE and EMD -66056 for 650 VUF11207-CXCR7-CID24, 9WLF and EMD-66057 for VUF10661-CXCR7-CID24, 9WLL and 651 EMD-66065 for Dynorphin-CXCR7-CID24, 9WLJ and EMD-66061 for LIH383-CXCR7-CID24, 652 9WLH and EMD -66059 for TC14012 -CXCR7-CID24, 9WLK and EMD -66062 for BAM22 -653 CXCR7-CID24, 9WLM and EMD-66066 for VUF10661-CXCR7-dimer, 9WLG and EMD-66058 654 for VUF16840-CXCR7, 9WLI and EMD-66060 for VUN701-CXCR7-monomer, and 9WLO and 655 EMD-66067 for VUN701 -CXCR7-dimer respectively. The accession codes of other PDB 656 coordinate files referenced in this study are 7SK3 (CXCL12 -CXCR7-Fab), 7SK5 (CXCL12 -657 CXCR7-Fab), 7SK9 (CCX662 -CXCR7-Fab), 8HNK (CXCL11 -CXCR3), 8F7W (Dynorphin -658 KOR-Gi), 8XYI (VUF10661 -CXCR3), 8IC0 (CXCL8-CXCR1-Gi), 6LFM (CXCL8-CXCR2-Gi), 659 8XWV (CXCL1-CXCR2-Go), 8XVU (CXCL2 -CXCR2-Go), 8XWA (CXCL1 -CXCR2 receptor-660 ligand focused map), 8XWF (CXCL3 -CXCR2-Go), 8XWM (CXCL6 -CXCR2-Go), 8XWN 661 (CXCL8-CXCR2), and 8XWS (CXCL5-CXCR2). 662

663 Human cell line 664 HEK293T cells used in this study were purchased from ATCC (Cat. No. CRL -3216). They 665 were cultured in DMEM, supplemented with 10% FBS, at 37°C under 5% CO ₂. The cell line 666 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 26 was routinely monitored under the microscope for appropriate morphology, but they were not 667 authenticated. In this study no knockout or knockdown cell lines were generated. 668 Insect cell line 669 Sf9 cell line was obtained from Expression Systems (Cat. No: 94 -001F), maintained in a 670 shaker incubator at 27°C with 135 rpm shaking speed, and sub-cultured in protein-free insect 671 cell media purchased from Gibco (Cat. No: 10902 -088). The cells were regularly inspected 672 under microscopy for appropriate morphology, but they were not authenticated. 673 General reagents and chemicals 674 Most standard molecular biology reagents were purchased from Sigma -Aldrich, unless 675 otherwise specified. Dulbecco’s Modified Eagle Medium (DMEM), trypsin -EDTA, Penicillin-676 Streptomycin solution, Phosphate -Buffered Saline (PBS), Fetal Bovine Serum (FBS), and 677 Hanks’ Balanced Salt Solution (HBSS) were purchased from ThermoFisher Scientific. HEK -678 293 cells were purchased from ATCC (Cat. no: CRL -3216) and maintained in 10 cm dishes 679 (Corning, Cat. no: 430167) in DMEM (Gibco, Cat. no: 12800 -017) supplemented with 10% 680 (v/v) FBS (Gibco, Cat. no: 10270 -106) and 100 U/mL penicillin and 100 µg/mL streptomycin 681 (Gibco, Cat. no: 15140122) at 37°C and 5% CO 2. Plasmid constructs for NanoBiT -based 682 assays have been described previously 59. VUF11207 was synthesized as described 683 previously60 while VUF10661 was purchased from Sigma -Aldrich (Cat. No: SML0803) and 684 TC14012 was purchased from MedChemExpress (Cat. no: HY -P1102). The opioid peptides 685 used in βarr recruitment assays and structural studies were obtained from Genscript. 686 NanoBiT-based G-protein-coupling and βarr recruitment 687 The standard methods for plasmid transfection and NanoBiT assays were carried out as 688 described previously59. Briefly, HEK -293T cells were seeded in the 10 cm dishes (Corning, 689 Cat. no: 430167) at a density of 3 million, 24 h prior to the transfection. For preparing the 690 transfection mixture, plasmid DNA and polyethyleneimine, linear (PEI) (Polysciences, Cat. no: 691 23966) were mixed in 1:3 ratio in 500 µL serum-free DMEM (Cellclone, Cat. no: CC3004) in a 692 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 27 microcentrifuge tube. This mix was incubated for 10 min. Prior to the addition of the 693 transfection mixture, the FBS-supplemented DMEM of the cell culture plate was replaced with 694 serum-free DMEM. Subsequently, the transfection mix was added to the cells, and serum-free 695 DMEM was replaced with FBS-supplemented DMEM 6 h post-transfection. 696 To measure ligand-induced ꞵarr recruitment, cells were transfected with the indicated 697 receptors tagged at the C-terminus with SmBiT and N-terminally LgBiT-tagged ꞵarr1/2. After 698 16-18 h of transfection, cells were washed with 1X PBS, trypsinized, and resuspended in the 699 assay buffer containing 1X HBSS, 5 mM HEPES, pH 7.4, 0.01% BSA, and 10 µL 700 coelenterazine (GoldBio, Cat. no: CZ05). The cells were seeded in a white, flat-bottom 96 well 701 plate at the density of 0.1 million cells per well, and incubated at 37°C for 90 min, followed by 702 a 30 min incubation at room temperature. Afterwards, basal luminescence values were 703 recorded for three cycles, followed by addition of ligands at indicated concentrations, and 704 luminescence readings were taken for additional ten cycles using a multimode plate reader 705 (LUMIstar/FLUOstar microplate reader, BMG Labtech). The analysis was carried out using 706 GraphPad Prism v 10.4.0 software, taking an average of luminescence values for cycles 5 -707 10. Normalization was carried out by taking luminescence values of the lowest dose as 1. 708 CXCR7 mutants were generated on wild-type template using a site-directed mutagenesis kit 709 (NEB, Cat no. E0554S) as indicated, and all constructs were verified by sequencing 710 (Macrogen). Subsequently, βarr recruitment was measured as described above. 711 To measure G-protein dissociation, cells were transfected using the indicated receptor 712 constructs and heterotrimeric G-proteins, and 16-18 h post-transfection, cells were harvested 713 by washing with 1X PBS followed by trypsinization. Subsequently, cells were resuspended in 714 the assay buffer containing 1X HBSS, 5 mM HEPES, pH 7.4, 0.01% BSA and 10 µL 715 coelenterazine (GoldBio, Cat. no: CZ05), and seeded in a white, flat-bottom 96-well plate at a 716 density of 0.1 million cells per well. The plate was incubated at 37°C for 90 min followed by 717 incubation at room temperature for an additional 30 min. Post-incubation, basal luminescence 718 reading was taken for three cycles, followed by ligand stimulation at indicated concentrations 719 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 28 and luminescence was recorded for ten additional cycles using a multimode plate reader 720 (LUMIstar/FLUOstar microplate reader, BMG Labtech). Data analysis was carried out using 721 GraphPad Prism v 10.4.0 software, and the luminescence value at 10 min post -ligand 722 stimulation was corrected for basal signal, followed by percentage normalization in terms of 723 the decrease in luminescence value. 724 In cell-based assays, the surface expression of the receptors was measured using a 725 previously described whole cell-based surface ELISA assay61. Briefly, cells were seeded in a 726 poly-D-lysine-coated 24 well plate at a density of 0.2 million cells per well, and allowed to 727 adhere for 24 h. Afterwards, growth media was removed, cells were washed with ice-cold 1X 728 Tris-buffered saline (TBS), and then fixed using 4% paraformaldehyde for 20 min on ice. 729 Subsequently, cells were washed three times with 1X TBS, incubated with 1% BSA (w/v, 730 prepared in 1X TBS) for 1 h at room temperature, followed by 1 h incubation with anti -FLAG 731 M2-HRP (Sigma-Aldrich, Cat. no: A8592; RRID:AB_439702) at 1:10,000 dilution (prepared in 732 1% BSA). Then, cells were washed three times with 1% BSA, followed by addition of TM Ultra 733 TMB-ELISA (Thermo Fisher Scientific, Cat. no: 34028), the reaction was quenched using 1 M 734 H2SO4. Signal was measured at 450 nm using the PerkinElmer VictorTM X4 multimode plate 735 reader, and to normalize the signal, Janus Green staining was performed (Sigma-Aldrich, Cat. 736 no: 201677) with absorbance measured at 595 nm. Surface expression was calculated by 737 taking a ratio of absorbance at 450 nm and 595 nm, and normalizing with respect to mock -738 transfected cells. 739 Expression and purification of CXCR7 740 CXCR7 was expressed and purified using baculovirus-mediated infection of Sf9 cells following 741 the protocols published previously62-65. Briefly, a full-length CXCR7 construct with N-terminal 742 FLAG-tag was generated using baculovirus DNA (Cat. No: 91-002). Cells were harvested 72 743 h post-infection, homogenized in a hypotonic buffer (20 mM HEPES pH 7.4, 20 mM KCl, 10 744 mM MgCl₂, 1 mM PMSF, 2 mM benzamidine), followed by a hypertonic buffer (20 mM HEPES 745 pH 7.4, 20 mM KCl, 10 mM MgCl₂, 1 M NaCl, 1 mM PMSF, 2 mM benzamidine). Subsequently, 746 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 29 the receptor was solubilized in lysis buffer (20 mM HEPES pH 7.4, 450 mM NaCl, 1 mM PMSF, 747 2 mM benzamidine, 0.1% cholesteryl hemisuccinate, 2 mM iodoacetamide) containing 1% L-748 MNG (Anatrace, Cat. no: NG310) for 2 h at 4°C. Cell lysate containing the solubilized receptor 749 was centrifuged (~48,000 X g for 30 min), filtered, and loaded onto pre-equilibrated M1-FLAG 750 column. The column was washed alternately with low salt buffer (20 mM HEPES pH 7.4, 150 751 mM NaCl, 2 mM CaCl2, 0.01% cholesteryl hemisuccinate, 0.01% L-MNG) and high salt buffer 752 (HSB; 20 mM HEPES pH 7.4, 350 mM NaCl, 2 mM CaCl2, 0.01% L-MNG). Subsequently, the 753 receptor was eluted using elution buffer (20 mM HEPES pH 7.4, 150 mM NaCl, 0.01% L-MNG, 754 2 mM EDTA, and 250 ug/mL FLAG peptide), free cysteines were alkylated by incubation with 755 2 mM iodoacetamide, and any remaining iodoacetamide was quenched with 2 mM L-cysteine. 756 Purified receptor was analyzed by SDS-PAGE, flash-frozen with 10% glycerol, and stored at 757 -80°C until further use. 758 Expression and purification of CXCL11 759 CXCL11 was purified following a previously published protocol66. Briefly, E. coli (Shuffle) cells 760 were transformed with CXCL11 expression plasmid, and transformed cells were first grown 761 overnight at 30°C in 50 mL of TB media containing 100 μg/mL ampicillin. The primary culture 762 was then used to inoculate 1 L of TB media and grown at 30°C until the OD 600 reached 1.2, 763 and subsequently, protein expression was induced with IPTG (1mM). After an additional 764 growth at 20°C for 48 h, cells were harvested and lysed in lysis buffer (50 mM Tris-HCl pH 8, 765 150 mM NaCl, 6 M Guanidine hydrochloride, pH 8) for 1 h at 4°C followed by sonication at 766 40% amplitude for 20 min. Cell lysate was then centrifuged at ~48,000 x g at 4°C for 1 h, 767 supernatant was filtered, and then loaded onto a pre-equilibrated Ni-NTA column. The beads 768 were washed with the same buffer, and the bound protein was eluted with elution buffer (50 769 mM Tris-HCl pH 8, 150 mM NaCl, 500 mM imidazole pH 8). Eluted protein was incubated with 770 20 mM DTT for 1 h at room temperature and then added dropwise in renaturation buffer (20 771 mM Tris -HCl pH 8, 200 mM NaCl, 550 mM Arginine pH 8, 1 mM EDTA, 1 mM reduced 772 glutathione, 0.1 mM oxidized glutathione) followed by an incubation for 48 h at 4°C. After 773 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 30 renaturation, the protein solution was dialyzed against 20 mM Tris -HCl pH 8, 150 mM NaCl, 774 and hexa-histidine tag was cleaved using enterokinase in presence of 10 mM CaCl ₂ for 16 h 775 at 4°C. The Cleaved CXCL11 was purified on Resource S cation exchange chromatography 776 with elution using a linear gradient of NaCl, and the elution fractions were analyzed by SDS -777 PAGE. Subsequently, fractions containing purified CXCL11 were pooled, dialyzed overnight 778 against 20 mM HEPES pH 7.4, 150 mM NaCl buffer at 4°C, flash -frozen with 10% glycerol, 779 and stored at -80°C until further use. 780 Expression and purification of CID24 781 CID24 was expressed and purified following previously published protocol 67. Briefly, CID24 782 expression plasmid was transformed in E. coli BL21 (DE3) cells, and the primary culture was 783 grown in 2XYT media containing 100 μg/mL ampicillin at 37°C and expression was induced 784 with 1 mM IPTG at OD600 of 1.0. The culture was grown for an additional 4 h at 37°C followed 785 by harvesting and lysis of cells in lysis buffer (20 mM HEPES pH 7.4, 500 mM NaCl, 0.5 mM 786 MgCl2, 1 mM PMSF, 2 mM Benzamidine, and 0.5% (v/v) Triton X-100). Cell lysis was 787 performed using sonication for 20 min at 40% amplitude, and the lysate was heated at 65°C 788 for 30 min. Subsequently, lysate was centrifuged at ~48,000 x g for 40 min at 4°C, supernatant 789 was clarified by filtration, and then loaded onto a pre -equilibrated Capto-L (Cytiva) column. 790 The beads were washed with washing buffer (20 mM HEPES pH 7.4, 500 mM NaCl), Fab was 791 eluted with 0.1 M acetic acid, and immediately neutralized with 10% (v/v) 1M HEPES pH 7.4. 792 Purified protein was dialyzed against 20 mM HEPES pH 7.4, 150 mM NaCl buffer, analyzed 793 using SDS-PAGE, and flash-frozen with 10% glycerol and stored at -80°C until further use. 794 Expression and purification of VUN701 795 The gene encoding VUN701 was cloned in pET-22b (+) vector with a hexa-histidine tag at the 796 N-terminus, and transformed in E. coli (Rosetta) cells. Freshly transformed cells were used to 797 inoculate primary culture at 37°C in 50 mL of 2XYT media supplemented with 100 μg/mL 798 ampicillin. The culture was grown overnight and further expanded in 1L 2XYT media and 799 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 31 grown in 37°C until OD600 reached 1.1. Afterwards, the culture was induced with 100 μM IPTG 800 and incubated for 16 h at 18°C. Subsequently, cells were harvested and resuspended in lysis 801 buffer (20 mM HEPES pH 8, 150 mM NaCl, 1 mM PMSF, 2 mM Benzamidine and 1 mg/mL 802 Lysozyme) and homogenized for 1 h at 4°C. Cells were lysed by ultrasonication at 40% 803 amplitude for 20 min, and the lysate was centrifuged at ~48,000 x g for 40 min at 4°C. The 804 supernatant was filtered and loaded onto a pre -equilibrated Ni-NTA column, and the column 805 was washed with 20 mM HEPES pH 8, 150 mM NaCl, and 30 mM imidazole pH 8 to remove 806 unbound protein. Subsequently, bound protein was eluted with 20 mM HEPES pH 8, 150 mM 807 NaCl, and 500 mM imidazole pH 8, and dialyzed against 20 mM HEPES pH 8, 150 mM NaCl 808 at 4°C for 16 h. After dialysis, protein was loaded onto a pre-equilibrated HiTrap Q FF column, 809 fractions corresponding to VUN701 were pooled, and dialyzed against 20 mM HEPES pH 7.4 810 and 150 mM NaCl. Finally, the purified protein was flash-frozen with 10% glycerol and stored 811 at -80°C until further use. 812 Chemical synthesis of VUF16840 813 Chemical reagents and instruments 814 All chemical reagents were purchased from Energy Chemicals and Aladdin Chemicals 815 (Shanghai, China) and used as received. All moisture -sensitive reactions were carried out 816 using anhydrous solvents and under nitrogen atmosphere. 1H and 13C NMR spectra were 817 recorded on a Bruker Avance III spectrometer (400 MHz) using TMS as the internal standard. 818 High resolution mass spectrometry (HRMS) was performed on Agilent Technologies 6540 819 UHD Accurate-Mass Q-TOF. Flash column chromatography was performed with silica gel 820 (300-400 mesh). 821 1-(tert-Butyl) 3 -methyl (S) -4-((1-phenylethyl)amino)-5,6-dihydropyridine-1,3(2H)-822 dicarboxylate（2）：To a solution of 1-(tert-butyl) 3-methyl 4-oxopiperidine-1,3-dicarboxylate 823 (5 g, 19.4 mmol) and Na 2SO4 (5.53 g, 38.8 mmol) in 2 -methyltetrahydrofuran (2-MeTHF, 25 824 mL) was added (S)-1-phenylethan-1-amine (2.6 g, 21.4 mmol) at rt. After being stirred at rt for 825 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 32 18h, the reaction mixture was filtered via a celite pad, and the filter cake was washed with 2 -826 MeTHF (2×5 mL). The combined filtrates were collected and kept under N₂, used for next step 827 without further purification. 828 1-(tert-Butyl) 3 -methyl (3R,4S) -4-(((S)-1-phenylethyl)amino)piperidine-1,3-829 dicarboxylate TFA salt（3）：NaBH4 (1.5 g, 23.3 mmol) in 2-MeTHF (15 mL) was cooled to -830 5°C, followed by dropwise addition of TFA (3.1 g, 27.2 mmol) over 5 min. The freshly prepared 831 compound 2 2 -MeTHF solution described above was added dropwise at -15°C. After being 832 stirred at -15°C for 15 min, the reaction mixture was quenched by adding 4 M NaOH (18 mL). 833 The resulting mixture was allowed to warm to ambient temperature, and then extracted with 834 2-MeTHF (3 × 30 mL). The combined organic layers were washed with brine, dried over 835 Na2SO4, and concentrated. The residue was dissolved in TBME (40 mL) at 65°C, then cooled 836 to ambient temperature, and was treated with TFA (2.67 g, 23.3 mmol) to induce 837 crystallization. The precipitate was filtered, washed with 2-MeTHF/TBME (1:1, 2 × 30 mL), and 838 dried under vacuum at 55°C to afford 3 as a white solid (4.53 g, 63% yield for two steps). 1H 839 NMR (400 MHz, MeOD): δ 7.56-7.48 (m, 5H), 4.69-4.41 (m, 2H), 3.80 (s, 3H), 3.81-3.24 (m, 840 5H), 2.90-1.60 (m, 4H), 1.69 (d, J = 6.8 Hz, 3H), 1.43 (s, 9H). 13C NMR (101 MHz, MeOD): δ 841 173.00, 155.10, 137.20, 130.93, 130.68, 128.81, 81.71, 58.50, 56.31, 42.15, 28.49, 26.90, 842 20.24. 843 1-(tert-Butyl) 3-methyl (3R,4S)-4-aminopiperidine-1,3-dicarboxylate TFA Salt （4）：844 A mixture of 3 (1.9 g, 5.2 mmol) and 20% Pd/C (0.42 g, 0.78 mmol) in MeOH (20 mL) was 845 evacuated and purged with hydrogen gas (3 times). After being stirred for 3 h under hydrogen, 846 the reaction mixture was filtered through a celite pad, and the filter cake was washed with 847 MeOH (2×15 mL). The combined filtrates were concentrated and recrystallized from TBME 848 (20 mL), giving 4 (1.09 g, 81% yield) as a white solid. 1H NMR (400 MHz, DMSO-d₆): δ 4.00 849 (m, 2H), 3.89 (s, 3H), 3.65 (m, 3H), 3.49 (dt, J = 10.5, 4.1 Hz, 1H), 3.23 (dd, J = 14.0, 3.8 Hz, 850 1H), 3.11–2.75 (m, 2H), 1.90 (m, 1H), 1.76–1.72 (m, 1H), 1.37 (s, 9H). 851 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 33 1-(tert-Butyl) 3 -methyl (3R,4S) -4-(5-(2,4-difluorophenyl)isoxazole-3-852 carboxamido)piperidine-1,3-dicarboxylate（5）：A mixture of 5 -(2,4-853 difluorophenyl)isoxazole-3-carboxylic acid (455 mg, 2 mmol) and HOAt (330 mg, 2.4 mmol) in 854 DMF (8 mL) were stirred at 0°C for 10 min. Amine 4 (720 mg, 2 mmol) and NMM (142 mg, 1.4 855 mmol) were then added at 0°C, followed by EDCI (465 mg, 2.4 mmol). The ice bath was 856 removed, and the reaction mixture was then stirred at ambient temperature for 12 h. After the 857 solvent and volatiles were evaporated under the reduced pressure, H2O (10 mL) was added, 858 and the mixture was extracted with EtOAc (3×15 mL). The combined organic phases were 859 washed with brine (20 mL) and dried (anhydrous Na 2SO4). The crude products were purified 860 by flash column chromatography (eluting with petroleum ether/EtOAc, 6:1) to afford 5 (580 861 mg, 62% yield) as a white solid. 1H NMR (400 MHz, MeOD): δ 8.08 -7.97 (m, 1H), 7.21 (m, 862 2H), 7.05 (s, 1H), 4.52-4.27 (m, 2H), 4.04 (brs, 1H), 3.71 (s, 3H), 3.37-3.26 (m, 4H), 2.25-2.05 863 (m, 1H), 1.78 (m, 1H), 1.46 (s, 9H), 1.36-1.25 (m, 1H). 13C NMR (101 MHz, MeOD): δ 174.09, 864 161.00, 160.77, 130.90, 114.20 (d, J = 22.1 Hz), 113.50, 106.57, 106.30 (d, J = 24.5 Hz), 865 103.66, 81.89, 52.95, 44.97, 29.12. 866 (3S,4S)-1-(tert-Butoxycarbonyl)-4-(5-(2,4-difluorophenyl)isoxazole-3-carbox-867 amido)piperidine-3-carboxylic acid（7）A 22% KOⁱPr solution (1.11 g, 2.5 mmol) in iPrOH (25 868 mL) was azeotropically dried via co -evaporation at 70°C. iPrOAc (127 mg, 1.2 mmol) was 869 added, followed by 5 (580 mg, 1.2 mmol) at rt. The reaction mixture was stirred at rt for 2 h, 870 and the solution containing the resulting ester 6 was then cooled to -5°C with an ice-bath. 2 M 871 KOH (1.8 mL, 3.6 mmol) was added to the cooled mixture, and the whole reaction mixture was 872 stirred at the same temperature for 24 h. The pH value of the reaction solution was adjusted 873 to 3.5 by adding 1 M HCl, resulting in some precipitates. The mixture was heated to 75°C until 874 a clear solution formed, and some solid crystallized from the solution upon cooling. The solid 875 was filtered, washed with H ₂O/iPrOH (3:1, 2 × 15 mL), and dried at 65°C under vacuum to 876 give 7 (410 mg, 73% yield). 1H NMR (400 MHz, MeOD): δ 8.41 -7.86 (m, 1H), 7.36 -7.14 (m, 877 2H), 7.12-6.99 (m, 1H), 4.39 (m, 1H), 4.12 (m, 1H), 2.95 (m, 2H), 2.69 (m, 1H), 1.98 (m, 1H), 878 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 34 1.61 (m, 1H), 1.48 (s, 9H). 13C NMR (101 MHz, MeOD): δ 174.68, 165.92, 160.53 (d, J = 26.4 879 Hz), 156.13, 130.27 (d, J = 13.6 Hz), 113.66 (d, J = 22.3 Hz), 106.03 (t, J = 26.2 Hz), 103.09, 880 81.63, 50.20, 49.90, 28.60. 881 tert-Butyl(3S,4S)-4-(5-(2,4-difluorophenyl)isoxazole-3-carboxamido)-3-((2-882 methoxyphenethyl)carbamoyl)piperidine-1-carboxylate（8）：A solution of 7 (410 mg, 0.9 883 mmol) and 2 -(2-methoxyphenyl)ethan-1-amine (165 mg, 1.1 mmol) in MeCN (8 mL) was 884 treated with NMI (298 mg, 3.6 mmol) at rt. TCFH (383 mg, 1.4 mmol) in MeCN (3 mL) was 885 added dropwise. After being stirred at rt for 30 min, the reaction mixture was diluted with H₂O, 886 heated at 75°C for 45 min, and then allowed to cool to ambient temperature, resulting in 887 formation of some precipitate. The solid was filtered, washed with H ₂O/MeCN (4:1, 2 × 10 888 mL), and dried at 65°C under vacuum to afford 8 (376 mg, 71% yield) as a white solid. 1H 889 NMR (400 MHz, DMSO-d6): δ 8.65 (d, J = 8.6 Hz, 1H), 8.09-7.84 (m, 2H), 7.46 (m, 1H), 7.25 890 (d, J = 2.5 Hz, 1H), 7.13-7.03 (m, 2H), 6.99 (d, J = 7.4 Hz, 1H), 6.85 (d, J = 8.2 Hz, 1H), 6.74 891 (t, J = 7.4 Hz, 1H), 4.23 (d, J = 11.2 Hz, 1H), 3.69 (s, 3H), 3.46 (m, 4H), 3.20 (t, J = 6.9 Hz, 892 2H), 2.67 (d, J = 1.3 Hz, 2H), 2.60 (q, J = 7.5 Hz, 1H), 1.88 -1.80 (m, 1H), 1.40 (s, 9H). 13C 893 NMR (101 MHz, DMSO-d6): δ170.49, 165.42, 164.15, 162.53 (J = 12.5 Hz), 160.38 (J = 12.8 894 Hz), 159.59, 157.90, 157.77, 157.25, 153.91, 129.86, 127.34 (J = 49.2 Hz), 120.22, 113.03 (J 895 = 20.8 Hz), 112.93, 105.47 (t, J = 26.1 Hz), 102.56, 79.20, 59.90, 55.18, 48.03, 38.65, 38.29, 896 29.91, 28.10, 20.78, 14.12. 897 5-(2,4-Difluorophenyl)-N-((3S,4S)-3-((2-methoxyphenethyl)carbamoyl)piperidin-4-898 yl)isoxazole-3-carboxamide（9）：To a solution of compound 8 (170 mg, 0.3 mmol) in iPrOH 899 (4 mL) was added 4 M HCl/iPrOH (0.4 mL, 1.2 mmol), and the mixture was stirred at 50°C for 900 5 h. After the solvent and volatiles were evaporated under the reduced pressure, the residue 901 was dissolved in EtOAc (30 mL), washed with saturated NaHCO 3 solution (10 mL) and brine 902 (10 mL), and dried (anhydrous Na 2SO4). The crude product was purified by flash column 903 chromatography (eluting with petroleum ether/EtOAc) to afford 9 (61 mg, 43%) as light off -904 white solid. 1H NMR (400 MHz, DMSO-d6): δ 8.96 (d, J = 8.5 Hz, 1H), 8.19 (s, 1H), 7.98 (q, J 905 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 35 = 8.0 Hz, 1H), 7.49 (m, 1H), 7.35 -7.19 (m, 2H), 7.08 (m, 1H), 6.99 (d, J = 7.3 Hz, 1H), 6.83 906 (m, 1H), 6.73 (m, 1H), 4.31-4.14 (m, 4H), 3.64 (s, 3H), 3.42-3.12 (m, 2H), 3.11-2.93 (m, 2H), 907 2.55 (d, J = 9.1 Hz, 2H), 2.01 (m, 1H), 1.82 (m, 1H). 13C NMR (101 MHz, DMSO-d6): δ169.22, 908 164.94, 164.12, 162.50 (d, J = 12.4 Hz), 160.33 (d, J = 12.8 Hz), 159.40, 157.85, 157.71, 909 157.19, 129.65 (d, J = 20.1 Hz), 127.30 (d, J = 71.8 Hz), 120.28, 113.19, 111.59, 110.63, 910 105.79, 102.63, 102.54, 55.24, 47.21, 44.54, 44.13, 42.41, 29.76, 28.42, 24.16. 911 N-((3S,4S)-1-Cyclohexyl-3-((2-methoxyphenethyl)carbamoyl)piperidin-4-yl)-5-(2,4-912 difluorophenyl)isoxazole-3-carboxamide（VUF16840）A mixture of 9 (316 mg, 0.65 mmol), 913 cyclohexanone (128 mg, 1.3 mmol), MeOH (5 mL) and catalytic amount of AcOH (1 drop) was 914 stirred at rt for 3 h. NaBH4 (49 mg, 1.3 mmol) was then added, and the stirring was continued 915 for an additional 12 h. The reaction was quenched by adding H₂O (3 mL). The resulting mixture 916 was concentrated to dryness, and the residue was extracted with EtOAc (3×10 mL). The 917 combined organic phases were washed with brine (10 mL) and dried (anhydrous Na ₂SO₄). 918 The crude product was purified by flash column chromatography (eluting with DCM/MeOH) to 919 afford VUF16840 (116 mg, 32% yield) as a grey solid. 1H NMR (400 MHz, DMSO-d6): δ 8.85 920 (brs, 1H), 8.06-8.01 (m, 2H), 7.58 (ddd, J = 11.5, 9.2, 2.5 Hz, 1H), 7.32 (td, J = 8.6, 2.5 Hz, 921 1H), 7.19 (d, J = 2.8 Hz, 1H), 7.17 -7.09 (m, 1H), 7.05-7.00 (m, 1H), 6.88 (d, J = 8.1 Hz, 1H), 922 6.77 (m, 1H), 4.05 (m, 1H), 3.69 (s, 3H), 3.27-3.12 (m, 3H), 2.64-2.55 (m, 3H), 2.08-1.95 (m, 923 3H), 1.81-1.78 (m, 2H), 1.61 -1.57 (m, 2H), 1.34 -1.28 (m, 5H), 1.24 -1.10 (m, 2H). 13C NMR 924 (101 MHz, DMSO -d6): δ172.19, 164.23, 160.42, 159.45, 157.68, 157.21, 130.14, 129.88, 925 129.72, 128.54, 127.65, 126.93, 124.61, 120.21, 113.09 (J = 19.6 Hz), 110.80 (J = 31.0 Hz), 926 105.56 (J = 25.7 Hz), 102.56, 68.65, 60.83, 55.24, 47.43, 43.41, 38.59, 32.62, 29.84, 28.71, 927 26.90, 25.57. HRMS (ESI, positive): Calcd. for C 31H37F2N2O4 [M+1]+ 567.2783; observed: 928 567.2787. 929 Reconstitution of CXCR7 complexes for structural analysis 930 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 36 For preparing CXCR7 complexes, purified CXCR7 was incubated with 3-5 fold molar excess 931 of the indicated ligands, and approximately 1.5 fold molar excess of CID24, when applicable, 932 for 1.5 hours at room temperature. Following incubation, the protein complex was 933 concentrated using a 100 MWCO concentrator (Cytiva, Cat. no: 28932363) and injected into 934 SuperoseTM 6 Increase 10/300 GL column. Peak fractions corresponding to the target 935 complexes were analyzed by SDS -PAGE, pooled, concentrated to 15 -18 mg/mL, and used 936 for freezing cryo -EM grids. In those cases where the cryo -EM grids were not frozen 937 immediately after sample preparation, the samples were flash -frozen in liquid N 2 and stored 938 at -80°C until further use. 939 Cryo-EM grid preparation and data collection 940 Purified complex samples were applied onto glow -discharged Quantifoil R1.2/1.3 gold grids 941 (300 mesh, holey carbon) at a protein concentration of 12 -18 mg/ml (~3 μl per grid). Grids 942 were blotted for 4 s at 4°C and 100% relative humidity (blot force 0) using a Vitrobot Mark IV 943 (Thermo Fisher Scientific) and immediately plunge-frozen in liquid ethane. Cryo-EM data were 944 collected on a Titan Krios G4 electron microscope (Thermo Fisher Scientific) operated at 300 945 kV, equipped with a Gatan K3 direct electron detector and BioQuantum K3 energy filter. 946 Automated data acquisition was performed using EPU software, with movies recorded in 947 counting mode at a nominal magnification corresponding to a pixel size of 0.86 Å. Dose rates 948 were maintained at ~15.6 e⁻/Ų/s, with a total dose of 55 e ⁻/Ų distributed over 50 frames for 949 each 1 s exposure. For the VUN701 -CXCR7 monomer complex, movies were recorded at a 950 nominal magnification of 165,000x corresponding to a pixel size of 0.53 Å with a total dose of 951 75 e⁻/Ų. Images were collected across a defocus range of −0.8 to −1.8 μm. The total number 952 of micrographs for each complex are indicated in the Figure S3-S5 and S18. 953 Cryo-EM data processing 954 All datasets for CXCR7 complexes were processed following a standardized workflow in 955 cryoSPARC v4.5 -4.6, unless otherwise specified. Dose -fractionated movie stacks were 956 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 37 corrected for beam -induced motion using the Patch motion correction (multi) sub -program, 957 and contrast transfer function parameters were determined with Patch CTF (multi)68. Particles 958 were automatically picked using the blob-picker sub-program, applying a circular mask with a 959 diameter range of 80-160 Å. The auto-picked particles were extracted with a box size of 320 960 px (fourier cropped to 72 px) and subjected to reference -free 2D classification and 961 heterogeneous refinement to eliminate ice contamination and distorted particles. Particles 962 corresponding to the best 3D class were selected and re-extracted using a box size of 320 px 963 (fourier cropped to 256 px) and subjected to non -uniform (NU) refinement and local 964 refinement. A complete data processing pipeline including the number of particles selected at 965 different stages are presented in Figure S3 -S5 and S13. Local resolution estimations for all 966 cryo-EM maps were performed using the LocRes module in cryoSPARC, with half -maps 967 provided as input. Protein-protein interfaces were analyzed using PDBsum69. 968 Model building and refinement 969 Coordinates for CXCR7 were generated using SWISS-MODEL and the atomic coordinates of 970 CID24 were derived from the previously determined structure of the CCX662 -CXCR7-CID24 971 complex (PDB: 7SK9) 29. Initial models for the chemokines were generated with SWISS -972 MODEL, employing the previously determined structure of CXCL11 (PDB: 1RJT) as a 973 template. All initial models were docked into the respective cryo -EM maps using UCSF 974 Chimera70,71, followed by flexible fitting using the “all -atom refine” function in COOT 72. 975 Subsequent model refinement was performed with phenix.real_space_refine, applying 976 secondary structure restraints, combined with iterative manual adjustments in COOT 72. The 977 quality of the final models was assessed using MolProbity as implemented in Phenix73,74. Data 978 collection, processing, and model refinement statistics are summarized in Supplemental 979 Table 1 . All structural figures were prepared with UCSF Chimera 70 and ChimeraX 69. 980 Chemokine residues are numbered from the first residue following signal sequence cleavage 981 site. 982 Lipidomics analysis 983 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 38 The lipidomics analysis was performed based on the method described previously 75. Briefly, 984 the purified CXCR7 sample (40 µL) and Sf9 cell lysate with overexpressed CXCR7 (40 µL) 985 were mixed with 300 µL methanol and 1mL MTBE and incubated for 1 h on a roller at room 986 temperature. 250 µl of HPLC H 2O was added. The sample was vortex for 2 -5 min to induce 987 phase separation and centrifuged at 400g, at room temperature for 10 min. The supernatant 988 was transferred to a new tube and further dried by a SpeedVac vacuum concentrator (Thermo 989 Fisher Scientific). The lipids mixture was dissolved in 80% lipidomics mobile phase A/20% 990 lipidomics mobile phase B (100 µL) and sonicated for 10 min. Mobile phase A consists of 991 acetonitrile/H2O (60/40, vol/vol) supplemented with 10 mM ammonium formate and 0.1% 992 (vol/vol) formic acid. Mobile phase B contains isopropanol/acetonitrile (90/10, vol/vol) 993 supplemented with 10 mM ammonium formate and 0.1% (vol/vol) formic acid. For LC-MS/MS 994 analysis, lipids were analyzed using a Dionex UltiMate 3000 RSLC nano system (Thermo 995 Scientific) coupled to an Eclipse Tribrid Orbitrap mass spectrometer (Thermo Scientific). The 996 samples were loaded onto a C18 column (Acclaim PepMap 100, C18, 75 µm × 150 mm, 3 µm 997 particle size). A binary solvent system was applied. Lipid separation was performed at 40 °C 998 using a gradient from 30% to 99% buffer B over 30 minutes, at a flow rate of 300 nL/min. The 999 electrospray voltage was set to 2.2 kV, with a funnel RF level of 40 and a heated capillary 1000 temperature of 320 °C. For data-dependent acquisition (DDA), the full MS scan range was set 1001 from m/z 300 to 2,000, with a resolution of 120,000 and an AGC target of 100%. Fragmentation 1002 spectra were collected in the Orbitrap at a resolution of 15,000 using higher-energy collisional 1003 dissociation (HCD) with stepped collision energies of 25%, 30%, and 35%. Phospholipids were 1004 detected in the negative ion mode. The raw data were processed using MSDIAL for 1005 phospholipid identification and quantification76. 1006 Peptide library screening on CXCR7 using NanoBRET 1007 Peptide library screening on ACKR3 was monitored by NanoBRET on living cells. HEK293T 1008 cells, stably expressing ACKR3, N -terminally fused to Nanoluciferase were distributed into 1009 white 384-well plates (1.5 × 104 cells per well). Peptides from the Chop-Suey/Variety Peptide 1010 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 39 Library (L-001, 238 peptides), Bioactive Secretory Peptide Library (L-009, 1016 peptides) and 1011 Orphan Receptor Peptide Ligand Library (L -005, 150 peptides) from Phoenix Peptides were 1012 added to the cells at a final concentration of 1 µM and incubated for 5 min on ice. Cells were 1013 subsequently incubated with CXCL12 -AZ568 (2 nM) and incubated for 2 h on ice. NanoGlo 1014 Live Cell Assay System substrate was then added and donor emission (450/8 nm BP filter) 1015 and acceptor emission (600 nm LP filter) were immediately measured on a GloMax Discover 1016 plate reader (Promega). BRET binding signal was defined as acceptor/donor ratio, and cells 1017 not treated with CXCL12 -AZ568 were used to define 0% BRET binding, whereas cells that 1018 were treated with CXCL12-AZ568 alone were used to define 100% BRET binding. 1019

1020 1 Pierce, K. L., Premont, R. T. & Lefkowitz, R. J. Seven-transmembrane receptors. Nat Rev 1021 Mol Cell Biol 3, 639-650 (2002). https://doi.org:10.1038/nrm908 1022 2 Rosenbaum, D. M., Rasmussen, S. G. & Kobilka, B. K. The structure and function of G-1023 protein-coupled receptors. Nature 459, 356-363 (2009). 1024 https://doi.org:10.1038/nature08144 1025 3 Lorente, J. S. et al. GPCR drug discovery: new agents, targets and indications. Nat Rev 1026 Drug Discov 24, 458-479 (2025). https://doi.org:10.1038/s41573-025-01139-y 1027 4 Hauser, A. S., Attwood, M. M., Rask-Andersen, M., Schioth, H. B. & Gloriam, D. E. Trends 1028 in GPCR drug discovery: new agents, targets and indications. Nat Rev Drug Discov 16, 1029 829-842 (2017). https://doi.org:10.1038/nrd.2017.178 1030 5 Hauser, A. S. et al. Pharmacogenomics of GPCR Drug Targets. Cell 172, 41-54 e19 1031 (2018). https://doi.org:10.1016/j.cell.2017.11.033 1032 6 Gusach, A., Garcia-Nafria, J. & Tate, C. G. New insights into GPCR coupling and 1033 dimerisation from cryo-EM structures. Curr Opin Struct Biol 80, 102574 (2023). 1034 https://doi.org:10.1016/j.sbi.2023.102574 1035 7 Congreve, M., de Graaf, C., Swain, N. A. & Tate, C. G. Impact of GPCR Structures on 1036 Drug Discovery. Cell 181, 81-91 (2020). https://doi.org:10.1016/j.cell.2020.03.003 1037 8 Smith, J. S., Lefkowitz, R. J. & Rajagopal, S. Biased signalling: from simple switches to 1038 allosteric microprocessors. Nature Reviews Drug Discovery 17, 243-260 (2018). 1039 https://doi.org:10.1038/nrd.2017.229 1040 9 Whalen, E. J., Rajagopal, S. & Lefkowitz, R. J. Therapeutic potential of β-arrestin- and G 1041 protein-biased agonists. Trends Mol Med 17, 126-139 (2011). 1042 https://doi.org:10.1016/j.molmed.2010.11.004 1043 10 Rajagopal, S., Rajagopal, K. & Lefkowitz, R. J. Teaching old receptors new tricks: biasing 1044 seven-transmembrane receptors. Nature Reviews Drug Discovery 9, 373-386 (2010). 1045 https://doi.org:10.1038/nrd3024 1046 11 Reiter, E., Ahn, S., Shukla, A. K. & Lefkowitz, R. J. Molecular Mechanism of β-Arrestin-1047 Biased Agonism at Seven-Transmembrane Receptors. Annu Rev Pharmacol 52, 179-197 1048 (2012). https://doi.org:10.1146/annurev.pharmtox.010909.105800 1049 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 40 12 Suomivuori, C. M. et al. Molecular mechanism of biased signaling in a prototypical G 1050 protein-coupled receptor. Science 367, 881-+ (2020). 1051 https://doi.org:10.1126/science.aaz0326 1052 13 Wingler, L. M. et al. Angiotensin Analogs with Divergent Bias Stabilize Distinct Receptor 1053 Conformations. Cell 176, 468-+ (2019). https://doi.org:10.1016/j.cell.2018.12.005 1054 14 Wingler, L. M. et al. Angiotensin and biased analogs induce structurally distinct active 1055 conformations within a GPCR. Science 367, 888-+ (2020). 1056 https://doi.org:10.1126/science.aay9813 1057 15 Rajagopal, S. et al. β-arrestin- but not G protein-mediated signaling by the "decoy" 1058 receptor CXCR7. P Natl Acad Sci USA 107, 628-632 (2010). 1059 https://doi.org:10.1073/pnas.0912852107 1060 16 Szpakowska, M. et al. Inclusion of ACKR5 in the systematic nomenclature of atypical 1061 chemokine receptors. Nat Rev Immunol 25, 225-226 (2025). 1062 https://doi.org:10.1038/s41577-025-01135-8 1063 17 Comerford, I. & McColl, S. R. Atypical chemokine receptors in the immune system. Nat 1064 Rev Immunol 24, 753-769 (2024). https://doi.org:10.1038/s41577-024-01025-5 1065 18 Nibbs, R. J. & Graham, G. J. Immune regulation by atypical chemokine receptors. Nat 1066 Rev Immunol 13, 815-829 (2013). https://doi.org:10.1038/nri3544 1067 19 Pandey, S. et al. Intrinsic bias at non-canonical, β-arrestin-coupled seven 1068 transmembrane receptors. Molecular Cell 81, 4605-+ (2021). 1069 https://doi.org:10.1016/j.molcel.2021.09.007 1070 20 Kleist, A. B. et al. Conformational selection guides β-arrestin recruitment at a biased G 1071 protein-coupled receptor. Science 377, 222-+ (2022). 1072 https://doi.org:10.1126/science.abj4922 1073 21 Li, X. X., Lee, J. D., Kemper, C. & Woodruff, T. M. The Complement Receptor C5aR2: A 1074 Powerful Modulator of Innate and Adaptive Immunity. J Immunol 202, 3339-3348 (2019). 1075 https://doi.org:10.4049/jimmunol.1900371 1076 22 Borroni, E. M. et al. β-Arrestin-Dependent Activation of the Cofilin Pathway Is Required 1077 for the Scavenging Activity of the Atypical Chemokine Receptor D6 (vol 6, ra30, 2013). 1078 Sci Signal 6 (2013). 1079 23 Nguyen, H. T. et al. CXCR7: a β-arrestin-biased receptor that potentiates cell migration 1080 and recruits β-arrestin2 exclusively through Gβγ subunits and GRK2. Cell Biosci 10 1081 (2020). https://doi.org:ARTN 134 1082 10.1186/s13578-020-00497-x 1083 24 Sarma, P. et al. Molecular insights into intrinsic transducer-coupling bias in the CXCR4-1084 CXCR7 system. Nat Commun 14, 4808 (2023). https://doi.org:10.1038/s41467-023-1085 40482-9 1086 25 Koenen, J., Bachelerie, F., Balabanian, K., Schlecht-Louf, G. & Gallego, C. Atypical 1087 Chemokine Receptor 3 (ACKR3): A Comprehensive Overview of its Expression and 1088 Potential Roles in the Immune System. Mol Pharmacol 96, 809-818 (2019). 1089 https://doi.org:10.1124/mol.118.115329 1090 26 Wang, C., Chen, W. L. & Shen, J. Z. CXCR7 Targeting and Its Major Disease Relevance. 1091 Front Pharmacol 9 (2018). https://doi.org:ARTN 641 1092 10.3389/fphar.2018.00641 1093 27 Smit, M. J. & van Muijlwijk-Koezen, J. E. From Insight to Modulation of CXCR4 and ACKR3 1094 (CXCR7) Function. Mol Pharmacol 96, 735-736 (2019). 1095 https://doi.org:10.1124/mol.119.118364 1096 28 Koch, C. & Engele, J. Functions of the CXCL12 Receptor ACKR3/CXCR7-What Has Been 1097 Perceived and What Has Been Overlooked. Mol Pharmacol 98, 577-585 (2020). 1098 https://doi.org:10.1124/molpharm.120.000056 1099 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 41 29 Yen, Y. C. et al. Structures of atypical chemokine receptor 3 reveal the basis for its 1100 promiscuity and signaling bias. Sci Adv 8 (2022). https://doi.org:ARTN eabn8063 1101 10.1126/sciadv.abn8063 1102 30 Chen, Q. Y. et al. Effect of phosphorylation barcodes on arrestin binding to a chemokine 1103 receptor. Nature (2025). https://doi.org:10.1038/s41586-025-09024-9 1104 31 Ikeda, Y., Kumagai, H., Skach, A., Sato, M. & Yanagisawa, M. Modulation of Circadian 1105 Glucocorticoid Oscillation via Adrenal Opioid-CXCR7 Signaling Alters Emotional 1106 Behavior. Cell 155, 1323-1336 (2013). https://doi.org:10.1016/j.cell.2013.10.052 1107 32 Meyrath, M. et al. The atypical chemokine receptor ACKR3/CXCR7 is a broad-spectrum 1108 scavenger for opioid peptides. Nature Communications 11 (2020). 1109 https://doi.org:10.1038/s41467-020-16664-0 1110 33 Szpakowska, M. et al. The natural analgesic conolidine targets the newly identified 1111 opioid scavenger ACKR3/CXCR7. Signal Transduct Tar 6 (2021). https://doi.org:ARTN 1112 209 1113 10.1038/s41392-021-00548-w 1114 34 Pillaiyar, T. & Laufer, S. A patent review of CXCR7 modulators (2019-present). Expert 1115 Opin Ther Pat 35, 543-569 (2025). https://doi.org:10.1080/13543776.2025.2477475 1116 35 Adlere, I. et al. Modulators of CXCR4 and CXCR7/ACKR3 Function. Mol Pharmacol 96, 1117 737-752 (2019). https://doi.org:10.1124/mol.119.117663 1118 36 Szpakowska, M. et al. Mutational analysis of the extracellular disulphide bridges of the 1119 atypical chemokine receptor ACKR3/CXCR7 uncovers multiple binding and activation 1120 modes for its chemokine and endogenous non-chemokine agonists. Biochem 1121 Pharmacol 153, 299-309 (2018). https://doi.org:10.1016/j.bcp.2018.03.007 1122 37 Saha, S. et al. Structural visualization of small molecule recognition by CXCR3 uncovers 1123 dual-agonism in the CXCR3-CXCR7 system. Nat Commun 16, 3047 (2025). 1124 https://doi.org:10.1038/s41467-025-58264-w 1125 38 Jiao, H. Z. et al. Structural insights into the activation and inhibition of CXC chemokine 1126 receptor 3. Nature Structural & Molecular Biology 31 (2024). 1127 https://doi.org:10.1038/s41594-023-01175-5 1128 39 Hauser, A. S. et al. GPCR activation mechanisms across classes and 1129 macro/microscales. Nature Structural & Molecular Biology 28, 879-+ (2021). 1130 https://doi.org:10.1038/s41594-021-00674-7 1131 40 Kooistra, A. J., Munk, C., Hauser, A. S. & Gloriam, D. E. An online GPCR structure 1132 analysis platform. Nature Structural & Molecular Biology 28, 875-+ (2021). 1133 https://doi.org:10.1038/s41594-021-00675-6 1134 41 Latorraca, N. R., Venkatakrishnan, A. J. & Dror, R. O. GPCR Dynamics: Structures in 1135 Motion. Chem Rev 117, 139-155 (2017). https://doi.org:10.1021/acs.chemrev.6b00177 1136 42 Kobilka, B. K. G protein coupled receptor structure and activation. Biochim Biophys 1137 Acta 1768, 794-807 (2007). https://doi.org:10.1016/j.bbamem.2006.10.021 1138 43 Otun, O. et al. Conformational dynamics underlying atypical chemokine receptor 3 1139 activation. P Natl Acad Sci USA 121 (2024). https://doi.org:ARTN e2404000121 1140 10.1073/pnas.2404000121 1141 44 Milligan, G., Ward, R. J. & Marsango, S. GPCR homo-oligomerization. Curr Opin Cell Biol 1142 57, 40-47 (2019). https://doi.org:10.1016/j.ceb.2018.10.007 1143 45 Ferré, S. et al. G Protein-Coupled Receptor Oligomerization Revisited: Functional and 1144 Pharmacological Perspectives. Pharmacol Rev 66, 413-434 (2014). 1145 https://doi.org:10.1124/pr.113.008052 1146 46 Terrillon, S. & Bouvier, M. Roles of G-protein-coupled receptor dimerization - From 1147 ontogeny to signalling regulation. Embo Rep 5, 30-34 (2004). 1148 https://doi.org:10.1038/sj.embor.7400052 1149 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 42 47 Levoye, A., Balabanian, K., Baleux, F., Bachelerie, F. & Lagane, B. CXCR7 1150 heterodimerizes with CXCR4 and regulates CXCL12-mediated G protein signaling. 1151 Blood 113, 6085-6093 (2009). https://doi.org:10.1182/blood-2008-12-196618 1152 48 Decaillot, F. M. et al. CXCR7/CXCR4 heterodimer constitutively recruits beta-arrestin to 1153 enhance cell migration. J Biol Chem 286, 32188-32197 (2011). 1154 https://doi.org:10.1074/jbc.M111.277038 1155 49 Sarkar, P. & Chattopadhyay, A. Cholesterol in GPCR Structures: Prevalence and 1156 Relevance. J Membrane Biol 255, 99-106 (2022). https://doi.org:10.1007/s00232-021-1157 00197-8 1158 50 Taghon, G. J., Rowe, J. B., Kapolka, N. J. & Isom, D. G. Predictable cholesterol binding 1159 sites in GPCRs lack consensus motifs. Structure 29, 499-+ (2021). 1160 https://doi.org:10.1016/j.str.2021.01.004 1161 51 Hoppe, N. et al. GPR161 structure uncovers the redundant role of sterol-regulated 1162 ciliary cAMP signaling in the Hedgehog pathway. Nature Structural & Molecular Biology 1163 31 (2024). https://doi.org:10.1038/s41594-024-01223-8 1164 52 Lin, X. et al. Structural basis of ligand recognition and self-activation of orphan GPR52. 1165 Nature 579, 152-+ (2020). https://doi.org:10.1038/s41586-020-2019-0 1166 53 Wang, Y. et al. Structures of the entire human opioid receptor family. Cell 186, 413-+ 1167 (2023). https://doi.org:10.1016/j.cell.2022.12.026 1168 54 Zhang, X. et al. Allosteric modulation and biased signalling at free fatty acid receptor 2. 1169 Nature 643, 1428-1438 (2025). https://doi.org:10.1038/s41586-025-09186-6 1170 55 Zhang, M., Lan, X., Li, X. & Lu, S. Pharmacologically targeting intracellular allosteric 1171 sites of GPCRs for drug discovery. Drug Discov Today 28, 103803 (2023). 1172 https://doi.org:10.1016/j.drudis.2023.103803 1173 56 Shen, S. et al. Allosteric modulation of G protein-coupled receptor signaling. Front 1174 Endocrinol (Lausanne) 14, 1137604 (2023). 1175 https://doi.org:10.3389/fendo.2023.1137604 1176 57 Burger, W. A. C. et al. The Clinical Candidate Xanomeline Displays a Dual Orthosteric 1177 and Allosteric Binding Profile at the M4 mAChR. Neuropsychopharmacol 47, 339-340 1178 (2022). 1179 58 Peri, L. et al. A bitter anti-inflammatory drug binds at two distinct sites of a human bitter 1180 taste GPCR. Nature Communications 15 (2024). https://doi.org:ARTN 9991 1181 10.1038/s41467-024-54157-6 1182 59 Inoue, A. et al. Illuminating G-Protein-Coupling Selectivity of GPCRs. Cell 177, 1933-1183 1947 e1925 (2019). https://doi.org:10.1016/j.cell.2019.04.044 1184 60 Wijtmans, M. et al. Synthesis, modeling and functional activity of substituted styrene-1185 amides as small-molecule CXCR7 agonists. Eur J Med Chem 51, 184-192 (2012). 1186 https://doi.org:10.1016/j.ejmech.2012.02.041 1187 61 Pandey, S., Roy, D. & Shukla, A. K. Measuring surface expression and endocytosis of 1188 GPCRs using whole-cell ELISA. Methods Cell Biol 149, 131-140 (2019). 1189 https://doi.org:10.1016/bs.mcb.2018.09.014 1190 62 Yadav, M. K. et al. Molecular basis of anaphylatoxin binding, activation, and signaling 1191 bias at complement receptors. Cell 186, 4956-4973 e4921 (2023). 1192 https://doi.org:10.1016/j.cell.2023.09.020 1193 63 Yadav, M. K. et al. Structure-guided engineering of biased-agonism in the human niacin 1194 receptor via single amino acid substitution. Nat Commun 15, 1939 (2024). 1195 https://doi.org:10.1038/s41467-024-46239-2 1196 64 Saha, S. et al. Molecular mechanism of distinct chemokine engagement and functional 1197 divergence of the human Duffy antigen receptor. Cell 187, 4751-4769 e4725 (2024). 1198 https://doi.org:10.1016/j.cell.2024.07.005 1199 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 43 65 Saha, S. et al. Molecular basis of promiscuous chemokine binding and structural 1200 mimicry at the C-X-C chemokine receptor, CXCR2. Mol Cell 85, 976-988 e979 (2025). 1201 https://doi.org:10.1016/j.molcel.2025.01.024 1202 66 Goncharuk, M. V. et al. Purification of native CCL7 and its functional interaction with 1203 selected chemokine receptors. Protein Expr Purif 171, 105617 (2020). 1204 https://doi.org:10.1016/j.pep.2020.105617 1205 67 Eberle, S. A. & Gustavsson, M. Bilayer lipids modulate ligand binding to atypical 1206 chemokine receptor 3. Structure 32, 1174-1183 e1175 (2024). 1207 https://doi.org:10.1016/j.str.2024.04.018 1208 68 Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for 1209 rapid unsupervised cryo-EM structure determination. Nat Methods 14, 290-296 (2017). 1210 https://doi.org:10.1038/nmeth.4169 1211 69 Laskowski, R. A., Jablonska, J., Pravda, L., Varekova, R. S. & Thornton, J. M. PDBsum: 1212 Structural summaries of PDB entries. Protein Sci 27, 129-134 (2018). 1213 https://doi.org:10.1002/pro.3289 1214 70 Pettersen, E. F. et al. UCSF Chimera--a visualization system for exploratory research 1215 and analysis. J Comput Chem 25, 1605-1612 (2004). https://doi.org:10.1002/jcc.20084 1216 71 Pettersen, E. F. et al. UCSF ChimeraX: Structure visualization for researchers, 1217 educators, and developers. Protein Sci 30, 70-82 (2021). 1218 https://doi.org:10.1002/pro.3943 1219 72 Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta 1220 Crystallogr D Biol Crystallogr 60, 2126-2132 (2004). 1221 https://doi.org:10.1107/S0907444904019158 1222 73 Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular 1223 structure solution. Acta Crystallogr D Biol Crystallogr 66, 213-221 (2010). 1224 https://doi.org:10.1107/S0907444909052925 1225 74 Liebschner, D. et al. Macromolecular structure determination using X-rays, neutrons 1226 and electrons: recent developments in Phenix. Acta Crystallogr D Struct Biol 75, 861-1227 877 (2019). https://doi.org:10.1107/S2059798319011471 1228 75 Wu, D. et al. Native MS-guided lipidomics to define endogenous lipid 1229 microenvironments of eukaryotic receptors and transporters. Nat Protoc 20, 1-25 1230 (2025). https://doi.org:10.1038/s41596-024-01037-4 1231 76 Tsugawa, H. et al. MS-DIAL: data-independent MS/MS deconvolution for comprehensive 1232 metabolome analysis. Nat Methods 12, 523-526 (2015). 1233 https://doi.org:10.1038/nmeth.3393 1234 1235 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint A B C CXCR7 D β-arrestin1 recruitment -12 -10 -8 -6 -4 0 5 10 CXCL12 CXCL11 -12 -10 -8 -6 -4 0 5 10 VUF11207 VUF10661 TC14012 log [ligand] (M) -12 -10 -8 -6 0 2 4 6 CXCL11 VUN701 +CXCL11 -12 -10 -8 -6 -4 0.6 0.8 1.0 2 4 6 VUF16840 VUF16840+CXCL11 CXCL11 Fold normalized CXCR7 CXCR3 CXCR4 VUF16840 VUN701 VUF10661 VUF11207 TC14012 CXCL11 CXCL12 Does not bind Inverse agonist Antagonist Agonist βarrs CXCR4 CXCL10 CXCL11 CXCL12 βarrsG-protein βarrsG-protein G-protein CXCR3 CXCR7 CID24Intracellular VUF16840 VUN701 VUN701 VUF10661 VUF10661 VUF11207 TC14012 CXCL11 3.2 Å 3.7 Å 3.6 Å 3.3 Å 3.2 Å 3.1 Å 3.2 Å 2.9 Å Figure 1. Ligand pharmacology and structure determination of CXCR7. (A) A schematic illustration showing chemokine-binding and transducer-coupling to CXCR3, CXCR4, and CXCR7. (B) Distinct pharmacology of selected CXCR7 ligands presented based on published literature. (C) Ligand-induced βarr1 recruitment at CXCR7 measured using NanoBiT-based assay in HEK-293T cells (mean±sem; n=3; fold normalized with basal signal). (D) Structures of CXCR7 in complex with indicated ligands determined using cryo-EM, stabilized using CID24 (an antibody fragment) as indicated. Figure 1. .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint Cys26 Asp30 K86 I48 S47 A46 W2.60 Y1.39 Y6.51 H3.29 F3.32 W6.48 F3.37 A VUF16840 VUF10661 VUF11207 TC14012 CXCL11 VUN701 N108ECL2 W2.60 Y1.39 Q7.39 Y6.51 L4.64 S5.42 Y6.51 W6.48 Y6.51 T31N-term H6.64 R97 R104 S31 4Å <1Å <1Å 3.3Å 2.5Å E 6.30 R 3.50 Inactive β 2 AR R 3.50 D 3.49 Y 3.51 E 6.30 H8 TM6 TM3 R 3.50 K 6.30 E 6.29 K 6.30 R 3.50 D 3.49 Y 3.51 E 6.29 H8 TM6 TM3 Inactive CXCR7 Y 5.58 R 6.34 E 6.29 Y 7.53 N 8.47 N 8.47 Y 5.58 R 6.34 E 6.29 Y 7.53 N 8.47 Active CXCR7 TM6 TM5 TM7 H8 Y 5.58 R 6.34 E 6.29 Y 7.53 N 8.47 Y 5.58 R 6.34 E 6.29 Y 7.53 TM6 TM5 H8 TM7 Inactive CXCR7 H8 ICL2 CXCR3 Pocket dimension 31Å High abundance of positively charged residues ICL3 H8 ICL1 ICL2 ICL3 CXCR7 Pocket dimension 21Å Less abundance of positively charged residues Cytosolic cavity in CXCR7 vs. CXCR3 B E C Y6.51 D6.58 Q7.39 R1 10.8Å CXCR7 activation 0 1.5 3 RMSD (Å) VUF11207 vs VUF16840 D Figure 2. .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint Figure 2. Ligand-recognition and atypical activation of CXCR7. (A) Overall binding pose (upper panel) and key interactions for the indicated ligands in the orthosteric binding pocket. The distance between the deepest point of the ligands from the conserved W6.48 is indicated in the upper panel. (B) A schematic representation and ribbon diagram showing the presence and absence of a broadly conserved TM3-TM6 ionic-lock in β2AR and CXCR7, respectively. The orientations of R3.50 (TM3) and E6.30/6.29 (TM6) are presented based on the structural snapshots (PDB IDs: 2RH1 for β2AR; 9WLD for CXCL11-CXCR7). (C) A schematic representation and ribbon diagram showing two distinct ionic locks i.e. TM5-TM6 and TM7-H8 in the inactive state of CXCR7 limiting the TM movement (left panel; red), and their disruption upon receptor activation (right panel; blue). (D) Activation dependent conformational changes in CXCR7 depicted in terms of RMSD between inverse-agonist vs. agonist-bound structures (VUF16840 vs. VUF11207-bound structures) mapped onto a ribbon representation. (E) Distinct cytoplasmic surface of the active CXCR7 compared to CXCR3 indicating a relatively constricted dimension of the cytoplasmic pocket and lesser abundance of positively charged residues on a patch involved in αN-helix engagement (PDB IDs: 9WLD for CXCL11-CXCR; 8HNK for CXCR3). .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint B M1.47 W1.55 S1.51 L2.51 S1.51 I1.48 V1.52 R8.48 R8.51M8.55 F8.58 Protomer 1 Protomer 2 Protomer 1 Protomer 2 W 1.55 W 1.55 S 1.51 S 1.51 I 1.48 I 1.48 F 7.55 F 7.55 R 8.51 R 8.51 F 8.58 F 8.58 K 8.61 K 8.61 Protomer 1 Protomer 1 Protomer 2 Protomer 2 Y 4.62 Y 4.62 E 5.33 E 5.33 L 4.48 L 4.48 P 4.59 V 4.51 V 4.51 P 4.59 V 5.44 W 5.34 W 5.34 V 5.44 Protomer 1 Protomer 1 Protomer 2 Protomer 2 A C TM1 TM1 H8 H8 CXCR7 CXCR7 H8 H8 TM1 TM1 TM4 TM5 TM4 TM5 H8 H8 TM1 TM1 TM4/5 TM4/5 CXCR7 CXCR7 H8 H8 CXCR7 CXCR7 H8 H8 TM1 TM1 VUF16840 - CXCR7 VUN701 - CXCR7 VUF10661 - CXCR7 Protomer 1 Protomer 2 Protomer 1 Protomer 2 Protomer 1 Protomer 2 Figure 3. Dimerization of CXCR7 via distinct interfaces. (A) Structural representation of CXCR7 dimers as captured by cryo-EM in complex with VUF16840, VUN701, and VUF10661. Ribbon diagram at two different orientations are presented for visualization (PDB IDs: 9WLG for VUF16840-CXCR7; 9WLO for VUN701-CXCR7; 9WLM for VUF10661-CXCR7). (B) Schematic illustration and ribbon diagram of CXCR7 dimers showing the dimerization interface. Although the dimerization interface between VUF16840-CXCR7 and VUN701-CXCR7 are similar (i.e., TM1-TM1 and H8-H8), the two protomers are oriented differently with respect to each other. (C) The key interacting residues in CXCR7 dimers, contributed by each protomer, are presented based on the corresponding structural snapshots. .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint PC8 PC8 PC8 Myristic acid PC8 Myristic acid Myristic acid PC8 PC8 Myristic acid VUF11207 CXCR7 VUF10661 CXCR7 TC14012 CXCR7 CXCL11 CXCR7 VUF16840 - CXCR7 (dimer) VUF10661 - CXCR7 (dimer) TM2 TM4 TM3 TM2 TM4 TM3 TM4 TM2 TM3 Active CXCR7 Inactive CXCR7 Lipid engagement with TM2/TM3/TM4 (active state) Lipid engagement with TM2/TM4 (inactive state) Inactive CXCR7 Active CXCR7 CXCR7 Inter - protomer stitching Protomer bound lipid CXCR7 CXCR7 Native lipid as molecular glue in CXCR7 dimer Myristic acid PC8 Inter - protomer stitching Protomer bound lipid LPP LPP A B D C E F LPP Figure 4. Interaction of native lipids with CXCR7. (A) Ribbon diagram of CXCR7 structures depicting the binding of native lipids to the receptor monomer in the agonist-bound state. The lipid-binding interface in terms of the TM helices are highlighted. (B-C) Ribbon diagram of CXCR7 dimeric structures depicting the binding of native lipids to each protomer and at the dimer interface. The lipid-binding interface in terms of the TM helices contributed by each protomer is highlighted. (D) Ribbon diagram and schematic illustration depicting the rearrangement of lipid-binding interface and the aliphatic chains due to conformational change in CXCR7 upon activation. (E-F) Lipid-mediated inter-protomer stitching as observed in VUF16840-CXCR7 and VUF10661-CXCR7 structures. The cross-interaction of the protomers mediated by the lipid moieties is indicated. CXCl11 .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint VUF10661 Orthosteric site VUF10661 - CXCR7 (monomer) VUF10661 - CXCR7 (dimer) VUF10661 VUF10661 allosteric site interaction VUF10661 - CXCR7 Protomer 1 Protomer 2 E5.33 E3.22 K3.26 V3.27 T4.61 VUF10661 Allosteric site Allosteric site CXCR7 CXCR7 A B Y195ECL2 H203ECL2 Y195ECL2 H203ECL2 VUF10661 ECL2 in VUF10061 - CXCR7 (monomer) ECL2 in VUF10061 - CXCR7 (dimer) Dimer Monomer ECL2 occupying the orthosteric pocket 0 1.5 3 RMSD (Å) VUF16840 - CXCR7 vs VUF10661 - CXCR7 CXCL11 - CXCR7 vs VUF10661 - CXCR7 Conformational changes in CXCR7 D E E5.33 E3.22 K3.26 V3.27 T4.61 VUF10661 K5.32 I5.36 L5.40 VUF10661 Lipid (PC8) VUF10661 Lipid (PC8) CXCR7 CXCR7 C VUF10661 and lipid as molecular glue in CXCR7 dimer Figure 5. .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint Figure 5. An allosteric site as a molecular glue and an intermediate conformation of CXCR7. (A) Ribbon diagram of the VUF10661-bound monomeric and dimeric structures showing the orthosteric and extra-helical allosteric binding sites (side-view) (PDB ID: 9WLF for VUF10661-CXCR7 monomer; 9WLM for VUF10661-CXCR7 dimer). (B) Key interactions of VUF10661 with CXCR7 in the extra-helical allosteric binding site as visualized in the cryo-EM structure (PDB ID: 9WLM ) and stabilization of VUF10661 in the extra-helical binding pocket by the positioning and interaction of PC at the dimer interface. (C) Top-view of VUF10661-bound CXCR7 showing the allosteric sites occupied by VUF101661 on each protomer and the key residues interacting with VUF101661. (D) Structural differences between the VUF101661-bound CXCR7 and inactive/active conformations of CXCR7 depicted in terms of RMSD (VUF10661-CXCR7 vs. VUF16840/CXCL11-CXCR7 structures) mapped onto a ribbon representation. (E) Distinct ECL2 conformations in the VUF10661-bound monomeric and dimeric CXCR7 structures. A stretch of ECL2 protrudes into the orthosteric binding pocket of the receptor protomers in the dimeric (intermediate) state, thereby, blocks of the pocket for VUF10661 entry. .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint 0 50 100 150 200 Normalized response (%) BAM22 Dyn A BAM22 Dyn A 1-13 Adrenorphin Dyn A Peptide library screening on CXCR7 Dynorphin A - CXCR7 3.2 Å BAM22 - CXCR7 3.7 Å LIH383 - CXCR7 3.3 Å S198ECL2 D6.58 Y6.48 Q7.39 F4 R6 R7 R9 Y1 Q7.39 Y6.51 H3.29 F4 R6 L7.35 W110ECL1 Dynorphin A - CXCR7 Dynorphin A - κ OR Dynorphin A - CXCR7 0 1.5 3 RMSD (Å) CXCL11 - CXCR7 Dynorphin A - CXCR7 VUF16840 - CXCR7 CXCL11 - CXCR7 Y1 D3.32 Y7.43 Q2.60 T2.56 Y1 F3.32 L3.36 F3.39 κOR CXCR7 Y1 Y1 CXCR7 Dynorphin A CXCR7 Dynorphin A CXCR7 Dynorphin A CXCR7 Dynorphin A A A CID24 CID24 CID24 C D B E F G Fig. 6: CXCR7 as an atypical opioid receptor. (A) Screening of peptide libraries on CXCR7-expressing cells using the displacement of a fluorescently-labeled CXCL12 as a readout. A displacement of 60% of CXCL12 used as a cut-off and the clusters indicated in different colors correspond to three distinct libraries. (B-D) Structures of CXCR7 in complex with indicated ligands determined using cryo-EM, stabilized using CID24 as indicated. (E) Dynorphin A stabilizing interactions in the CXCR7 orthosteric pocket. (F), Relative binding depth and key interactions of Dynorphin A with respect to Dynoprhin A in the orthosteric pocket of KOR and CXCR7 respectively. (G), Ribbon diagram showing overall structural changes in CXCR7 upon activation via endogenous ligand CXCL11 and opioid Dynorphin A. .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint Y 1.39 I2.57 W 2.60 S 2.63 T 3.28 H 3.29 F 3.32 S 3.33 L 3.36 F 3.37 I3.40 D 4.60 L 4.64 I5.31 L 5.35 S 5.42 W 6.48 Y 6.51 H 6.52 V 6.55 D 6.58 S 6.61 I6.62 H 6.64 T 7.24 E 7.28 L 7.31 L 7.35 H 7.36 Q 7.39 S 7.42 L 7.43 CXCL11 VUF11207 VUF10661 TC14012 VUF16840 VUN701 Dynorphin A BAM22 LIH383 0 1 2 3 4 5 6 7 8 9 Unstim CXCL11 CXCL12 Dynorphin A VUF11207 WT C26N-termA D30N-termA Y511.39A W1002.60A S1032.63A F1243.32A S1253.33A L1283.36A D1794.60A R197ECL2A Y200ECL2A S2165.42A Y2686.51A D2756.58A H2816.64A Q3017.39A S3047.42A L3057.43A 1.0 1.5 2.0 NPFF1/2R NPSR NPBW1R NPBW2R Y1/2/4/5/6R NTS1/2R δ/κ/μ/NOPR OX1/2R QRFPR PrRPR RXFP1/2/3/4R SST1/2/3/4R NK1R NK2/3R TRH1R V1A/1B/2R BB1/2/3R B1R B2R CCK1/2R ETA/BR FPR1/3R FPR2R GAL1/2/3R GhrelinR GnRH1R KisspeptinR MCH1/2R MC1/2/3/4/5R MotilinR NMU1/2R CXCR1 CXCR2 CXCR3 CXCR4 CXCR5 CXCR6 CXCR7 CXC3CR1 XCR1 ACKR1 ACKR2 ACKR4 ACKR5 CCRL2 ApelinR C3aR C5aR1 C5aR2 CCR1 CCR2 CCR3 CCR4 CCR5 CCR6 CCR7 CCR8 CCR9 CCR10 AT1/2R UTR OTR Tyr 6.51 as pivotal switch in chemoattractant receptors Y6.51 present Y6.51 absent ECL2N-term H8 Y 6.51 TM4 ICL2ICL3 TM1 Ligand - receptor interaction (orthosteric binding pocket) β arr interaction (CXCR7 mutants) Fig. 7: A pivotal tyrosine switch in CXCR7 and chemoattract receptors. (A) A heatmap representation depicting the ligand-interacting residues in the TM domain of CXCR7 based on structural snapshots determined here. Color-coding indicates the relative frequencies of the interaction. (B) Ligand-induced βarr- recruitment for CXCR7 mutants measured using NanoBiT-based bystander assay in transfected HEK-293T cells presented as a heatmap (mean; n=3; fold-normalized with basal; CXCL11, 1µM; CXCL12, CXCL11, 1µM; VUF11207, 10µM; dynorphin, 31.6µM .(C) Sequence analysis of peptide GPCRs uncovers a broadly conserved occurrence of Tyr6.51 in TM6 of the chemoattractant receptors. A B C .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint Constricted cytoplasmic pocket Extra-helical allosteric site Native lipid as a molecular glue Figure 8. Atypical activation and functional divergence of CXCR7. A schematic illustration depicting a unique activation mechanism, constricted cytoplasmic pocket, allosteric site for a small molecule agonist, dimerization, and recognition of endogenous opioid peptides. TM5 Unique TM5-TM6 ionic-lock TM6 Extracellular Intracellular CXCR7 CXCR7 CXCR7 Opioid peptides Scavenging/modulation .CC-BY-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 31, 2025. ; https://doi.org/10.1101/2025.10.30.685530doi: bioRxiv preprint