Overall structure of human CXCR4 tetramer

To obtain the human CXCR4 structure, we fused a Flag-tag at the C terminus of the full-length optimized coding DNA sequence. The purification process began with affinity chromatography, followed by gel filtration after expression in the FreeStyle^TM 293-F cells. The elution volume of gel filtration suggested that CXCR4 primarily exists as an oligomer. Additionally, we observed a minor peak following the main peak. SDS-PAGE analysis confirmed that both peaks correspond to CXCR4 protein (Supplementary Fig. 1a–c). Native page showed that the minor peak has an apparent molecular weight of approximately 120 kDa, while the main peak is around 480 kDa—roughly four times larger. Based on these observations and consistent with a recent publication²⁵, we assigned these two peaks to monomeric and tetrameric forms of CXCR4, respectively (Supplementary Fig. 1d, e). The rising molecular weights observed on the native gel are likely influenced by detergent micelles surrounding the CXCR4 proteins.

On the SDS page, we observed diffuse CXCR4 bands, which shifted to a low molecular weight upon treatment with PNGase F—an enzyme for removing almost all N-linked oligosaccharides from glycoproteins, suggesting that CXCR4 is highly glycosylated (Supplementary Fig. 1f). Preliminary cryo-EM analysis revealed that the CXCR4 particles exhibited severe preferred orientation on holey-carbon gold grids (Quantifoil Au 300 mesh R1.2/1.3). To address this issue, we added 0.5 mM detergent octyl maltoside fluorinated (FOM) to the solution prior to vitrification of cryo-grids. Through a cascade of two-dimensional (2D) and 3D classification procedures, a non-uniform refinement of 734,989 particles yielded a 2.9-Å density map (Supplementary Fig. 2a–c). This map enabled us to reconstruct nearly the full-length of the CXCR4 structure, except for one region: the N-terminus of CXCR4 (residues 1–24), which remains unresolved due to its intrinsic flexibility, likely related to its role in CXCL12 binding³². 2D classification identified a minor population (~3%) of monomeric CXCR4 (Supplementary Fig. 2d). However, the limited particle number and size constraints of cryo-EM prevented us from reconstructing a high-resolution structure.

The CXCR4 structure adopts a homotetramer topology with C4 symmetry (Fig. 1a). The tetrameric assembly is primarily mediated by interactions between transmembrane helices 5, 6, 7 (TM5, TM6, TM7) from one protomer and TM1 of an adjacent protomer (Fig. 1b). While hydrophobic interactions (Fig. 1c), typically considered the main driving force of transmembrane assembly, play an important role, a few polar residues near the end of helices form stable hydrogen bonds, further strengthening the organization of CXCR4 tetramer (Fig. 1c). On the extracellular side, Q272 from TM6 and E275 from TM7 establish interactions with K38 and N35 of TM1, respectively (Fig. 1c). On the intracellular side, two polar interactions are observed between TM1 (Q66, K68) and TM5 (S224, S227) (Fig. 1c). Additionally, π–π interactions also contribute to the stability of the tetramer architecture, with F36 and F40 of TM1, along with W283 of TM7, forming key interfaces between CXCR4 monomers (Fig. 1c). The interface between two monomers in the cryo-EM-based tetrameric alignment differs notably from that in the crystal structure-based dimeric alignment of CXCR4 (Fig. 1b)²⁴, with the latter assembled through interactions between two TM5 helices. The structure of CXCR4 tetramer is nearly identical to previously reported tetrameric CXCR4 structures (PDB: 8YU7 and 8U4T)^25,33, with root mean square deviation (RMSD) values of 0.840 Å and 0.871 Å, respectively. Only minor differences are observed in the N-terminus and the ECL2 region of CXCR4, which are responsible for the interactions with the Fab (Supplementary Fig. 2e). Alignment of one CXCR4 monomer of the tetrameric and dimeric structures reveal a significant movement in TM6 and TM7, with a rotation of 10° in TM6 and 5° in TM7 (Fig. 1b). In summary, these structural differences highlight an alternative assembly mode for CXCR4 and provide a plausible explanation for oligomer switching, as CXCR4 has been shown to exist as a monomer when in complex with the downstream factor Gi proteins^25,34.

a Different views of CXCR4 tetramer, with the four promoters are shown in blue, magenta, green, and yellow. b Comparison of the CXCR4 dimer interface in the CXCR4 tetramer (in color) and the CXCR4 crystal structure (PDB:3OE9) (in silver). c Detailed interactions between CXCR4 monomers at the tetramer interface.

Two distinct assembly mechanisms of CXCL12-CXCR4 complex formation

To understand how the cognate ligand CXCL12 binds to its receptor CXCR4 and the mechanism by which it blocks HIV entry, we coexpressed CXCL12 and CXCR4 that yielded a single peak on a size-exclusion column (Supplementary Fig. 3a) and determined the structure of the complex. We obtained two distinct structures of the CXCL12-CXCR4 complex (Supplementary Fig. 3b–d), namely with 8:4 stoichiometry at 3.4 Å (Fig. 2a) and 8:8 stoichiometry at 3.3 Å (Fig. 2b). Both states were confirmed by native PAGE analysis (Supplementary Fig. 3e). CXCL12 has been reported to exist in monomeric and dimeric states, each leading to distinct signaling events³⁵. Its binding to CXCR4 mirrors interactions observed for other C-X-C chemokines and receptors in the presence of Gi proteins during activation^25,34,36,37. In the CXCL12-CXCR4 structures reflecting the inactive state, CXCR4 adopts a tetrameric state (Fig. 2a), with each tetramer associated with four CXCL12 dimers (Fig. 2a). Interestingly, the four CXCL12 dimers can further attach to another CXCR4 tetramer, forming a “head-to-head” assembly configuration (Fig. 2b). Like the CXCR4 tetramer alone, the CXCL12-CXCR4 complexes also adopt C4 symmetry, with a RMSD of 0.658 Å in CXCR4 tetramer regions (Supplementary Fig. 4a).

a Different views of CXCL12-CXCR4 complex with 8:4 stoichiometry. b Different views of CXCL12-CXCR4 complex with 8:8 stoichiometry, with the two head-to-head CXCR4 tetramers labeled as tetramerᵃ and tetramerᵇ. c Detailed interactions between CXCL12 and CXCR4 with residues labeled in the boxed panels. d The CXCL12 mutants exhibited either similar or impaired binding to CXCR4. The Co-immunoprecipitation (Co-IP) experiments were repeated at least two times with similar results. Source data are provided as a Source data file. e Structural comparison of CXCL12-CXCR4 complex between inactive (in color) and active (in silver) (PDB:8K3Z) states.

The interaction of CXCL12 and CXCR4 is in line with a “two-site” model of chemokine binding. The N-terminus of CXCR4 acts as the chemokine recognition site 1 (CRS1), engaging with the globular core of CXCL12, while the seven transmembrane helices of CXCR4 form chemokine recognition site 2 (CRS2), which accommodates the insertion of the N-terminus of CXCL12 into a pocket^38,39,40. In our structure, similar to the CXCR4 tetramer alone, the N-terminal residues (1–25) of CXCR4 in the CXCL12-CXCR4 complex are not visible, which is consistent with all reported CXCR4 ligand-bound and apo states, including complexes with chemokines (PDBs: 8U4O and 8K3Z)^25,34, viral peptides (e.g., vMIP-II, PDB: 4RWS)⁴¹, small-molecule antagonists (PDBs: 8ZPL, 8ZPM, 8ZPN)⁴². At the CRS2 recognition site, K1 of CXCL12 forms a salt bridge with E288 of CXCR4, along with possible cation-π interactions involving Y116_CXCR4 and W94_CXCR4 (Fig. 2c, boxed segment). Additionally, V3_CXCL12 and L5_CXCL12 engage in hydrophobic interactions with the I259_CXCR4 and I284_CXCR4 (Fig. 2c, boxed segment). Within the ECL2 of CXCR4, two key interaction sites are observed between CXCL12 and the ECL2 of CXCR4, which is designated CRS3. The first involves Y7_CXCL12, which stacks against F189_CXCR4 and Y190_CXCR4 through π-π interactions (Fig. 2c, boxed segment). The second site involves the curved loop between β1 and β2 of CXCL12, where the main chain probably contacts with D181_CXCR4 (Fig. 2c, boxed segment). On the opposite side, in the ECL3 region, E268_CXCR4 forms a single salt bridge with R12_CXCL12, further supported by interactions with the main chains of I265_CXCR4 and L266_CXCR4 (Fig. 2c, boxed segment), consistent with the mutagenesis studies highlighting the importance of this region for CXCL12 binding^23,43. In terms of the CSR1, E26_CXCR4 is sandwiched between the N-loop and β3 regions of CXCL12, reinforcing the binding between CXCL12 and its receptor (Fig. 2c). Mutagenesis experiments reveal that single mutation at K1, S6 or Y7 in CXCL12 have minimal effects on CXCL12-CXCR4 binding (Fig. 2d). In contrast, a mutation at R12, or simultaneous mutations in all interaction residues, completely abolish interactions between CXCL12 and CXCR4 (Fig. 2d). The R12 _CXCL12 mutation was also reported to cause a dramatic loss of potency in CXCR4-mediated intracellular calcium signaling and chemotactic responses^44,45,46. These findings highlight the critical role of the ECL3 region of CXCR4 in mediating its interaction with CXCL12.

Comparison of CXCL12-CXCR4 complexes in the inactive and active (PDB: 8K3Z) states reveals two striking divergences. One, as expected from GPCR activation⁴⁷, TM5-7 undergo notable shifts (approximately 12 Å in TM5, 13 Å in TM6 and 10 Å in TM7) between these two states (Fig. 2e). The other, the relative orientation of CXCL12 and CXCR4 is rotated, with a displacement of 7 Å in the active state compared to the inactive state (Fig. 2e). This shift alters the conformation of CXCL12, particularly the N-terminal loop (Fig. 2e), which corresponds to different binding interfaces between these two states³⁴. The CXCL12 dimer interface in our structures is primarily mediated by main-chain interactions between the β1 strands of the two monomers (Supplementary Fig. 4b, c), closely resembling both the crystal and NMR structures^48,49,50 (PDB IDs: 1QG7, 1A15, and 2K01, corresponding to a mammalian crystal form, a synthetic variant, and an NMR model, respectively). The CXCL12 dimer in our structure closely resembles, except for two regions: the N-terminal segment, which is critical for receptor interaction, and the C-terminal helix (Supplementary Fig. 4d).

Architecture of CXCR4-gp120^HIV-2 complex

The first step in reconstituting the CXCR4-gp120 complex involved purifying the CXCR4 and gp120 independently (see “Methods”). We selected gp120 from HIV-2 instead of HIV-1 due to its higher binding affinity for CXCR4²³. CXCR4 and gp120 were mixed at a 1:2 molar ratio before being subjected to size-exclusion chromatography (SEC). The peak corresponding to CXCR4-gp120^HIV-2 complex (Supplementary Fig. 5a) was collected and concentrated for preparation of cryo-grids. Structural data processing revealed CXCR4-gp120^HIV-2 complex exhibits multiple conformations (Fig. 3a and Supplementary Fig. 6), with the relative position between CXCR4 and gp120^HIV-2 being dynamic. This flexibility results in various alignments of the CXCR4-gp120^HIV-2 complex. Local refinement of the gp120 ^HIV-2 region enabled the reconstruction of the entire V3 loop and docking of most of the gp120 ^HIV-2 structure (Fig. 3b, c). Notably, the V3 loop shows minimal changes, amongst conformers, indicating a conserved mechanism of gp120^HIV-2 binding to its co-receptor CXCR4. The CXCR4 and gp120^HIV-2 complex exhibits a 4:1 stoichiometry (Fig. 3a, b). A small number of particles containing two gp120^HIV-2 proteins were also observed in the 2D classifications (Supplementary Fig. 6b, boxed segment), but these particles could not be resolved to high resolution. Structural modeling suggests that the CXCR4 tetramer can accommodate four gp120^HIV-2 proteins without any apparent steric clashes (Supplementary Fig. 7a), indicating that the complex with a single gp120^HIV-2 bound may be more stable than other configurations, at least when only gp120^HIV-2 is present.

a Structural diversity of the CXCR4-gp120^HIV-2 complex, superscripts^a–d indicate distinct conformations with slight structural variations. b Different views of CXCR4-gp120^HIV-2 complex, highlighting the V3 loop of gp120. c Highlighted key residues within the V3 loop that are critical for CXCR4 recognition, including the labeled GFKF motif. d Detailed interactions between HIV-2 gp120 and CXCR4. Interfacial residues are labeled in the boxed segment. e The gp120^HIV-2 mutants exhibited either similar or impaired binding to CXCR4. The pull-down assays were repeated at least two times with similar results, and source data are provided as a Source data file. f Structural comparison of CXCR4-gp120^HIV-2 (in pink and cyan) and CCR5-gp120^HIV-1 (PDB:6MEO) (in silver and green) complexes.

For these structures, residues 305–313 of gp120^HIV-2 are buried into the CXCR4 CRS2 pocket, adopting a reverse cross conformation (Fig. 3d). In the turn region of the V3 loop, residues F309, K310, and F311 form the GFKF motif, alongside G308, which is essential for HIV-2 infection (Fig. 3d). F309 is the most deeply embedded, engaging in both a cation-π interaction with R188_CXCR4 and a π-π interaction with H203_CXCR4. K310 forms a salt bridge with E288_CXCR4, while F311 establishes local hydrophobic interactions with I185 and F189 in the ECL2 region of CXCR4 (Fig. 3d). At CRS2, H312 plays a critical role by making extensive contacts with neighboring CXCR4 residues. It forms a bifurcated hydrogen bond with both E277_CXCR4 and D262_CXCR4, and a π–π interaction with H281_CXCR4 (Fig. 3d, left boxed segment). In particular, the loop formed by residues M306-F309 of gp120 comes into close proximity with the ECL2 region of CXCR4, producing main-chain contacts. In the ECL3 region, K314 interacts with E268_CXCR4 through a salt bridge (Fig. 3d, right boxed segment), which further stabilizes the V3 loop. K300, situated above the inserted loop, is positioned near the N-terminus of CXCR4, suggesting potential interactions due to their proximity (Fig. 3d, boxed segment). However, mutating lysine to alanine does not affect the gp120^HIV-2-CXCR4 binding (Fig. 3e). Further mutational analyses highlight the functional significance of key residues. Single mutations at K310 or H312 disrupt gp120^HIV-2-CXCR4 interactions, while mutations at F309 and F311 do not, indicating the critical roles of K310 and H312 in assembling the gp120^HIV-2-CXCR4 complex, which closely corresponds to the key hydrogen bonds formed between K310, H312, and CXCR4. Combined mutations of residues F309-H312 or an expanded set including K300, F309-H312, and K314 effectively block the interaction between gp120 and CXCR4 (Fig. 3e). Our structural model also reveals that the N-terminus of CXCR4 runs parallel and anti-parallel to the entry and exit loops of the V3 region of gp120^HIV-2 (Fig. 3d). These three loops are tightly packed, making significant connections with one another (Fig. 3d). Furthermore, we noticed three proline residues-P304, P315 in gp120, and P27 in CXCR4-closely positioned within these loops, potentially altering the loop direction (Fig. 3d, boxed segments). This structural arrangement could explain the dynamics of the gp120^HIV-2 binding, consistent with the multiple observed structural conformations. The CXCR4 residues involved in binding HIV-2 gp120 are similar to those involved in HIV-1 gp120 binding^23,51, suggesting a conserved mechanism for both HIV-1 and HIV-2 infection.

To determine whether the CXCR4 pocket in this configuration could accommodate the V3 loop of HIV-1, we compared our structure to the CCR5-gp120^HIV-1 complex by aligning the CXCR4 and CCR5 protomers (Fig. 3f). As both CXCR4 and CCR5 serve as co-receptors for HIV, they adopt a similar configuration, with an RMSD of 1.131 Å. This close alignment suggests that the V3 loop of HIV-1 is also well-suited for binding to the CXCR4 pocket. Structural and sequence alignments reveal that the key residues involved in HIV-1 binding to CCR5 are largely conserved in CXCR4 (Fig. 3f and Supplementary Fig. 8a). This conservation suggests that the combination of HIV-1 gp120 and CXCR4 may closely resemble the structure of the CXCR4-gp120^HIV-1 complex. The binding pocket of chemokine receptors is typically divided into two sub-pockets: the minor sub-pocket, formed by TM1-3 and TM7, and the major sub-pocket, formed by TM3-7⁵². Interestingly, the V3 loop of HIV-1 primarily occupies the minor sub-pocket, while the V3 loop of HIV-2 predominantly fills the major sub-pocket (Fig. 3d). This distinction may arise from the absence of two residues in the HIV-2 V3 loop or from the sequence diversity in this region (Supplementary Fig. 8b). Another notable feature is the relative position of gp120 in HIV-1 compared to HIV-2. In HIV-1, gp120 exhibits a significant shift of 12 Å and a clockwise rotation of 150°, making its position closer to the center of the CXCR4 tetramer, particularly around the bridging sheet, where it contacts the N-terminus of CCR5. This movement also results in steric clashes in the bridging sheet region, making it unlikely that four, or even two, HIV-1 gp120 proteins could bind the CXCR4 tetramer simultaneously (Supplementary Fig. 7b). However, we cannot entirely rule out the possibility of four HIV-1 gp120 proteins binding, as structural rearrangements in HIV-1 gp120 could potentially accommodate such recognition (Supplementary Fig. 7b).

Reconstitution of CXCR4-gp120^HIV-2-CD4 complex

The mature viral spike contains three copies of gp120 (gp120₃), which potentially engage three copies of CD4. However, the extent of CD4 binding leads to distinct conformational outcomes. When no or only one CD4 molecule is bound, the trimeric Envs remain in a closed, prefusion state. By contrast, binding of two or three CD4 molecules induces an open state in the Env trimer^16,53,54. This asymmetrical engagement of CD4 molecules has also been observed in native membranes⁵⁵. In the CCR5-gp120^HIV-1-CD4 structure, no significant differences were observed in the core region of gp120 in the presence of CCR5²¹. It is essential to determine whether this holds true for the CXCR4-gp120^HIV-2-CD4 structure, as HIV-1 and HIV-2 exhibit notable differences in gp120 configuration when bound to CXCR4 (Fig. 3f). To obtain the CXCR4-gp120^HIV-2-CD4 complex, we first expressed the extracellular D1-D4 domains of CD4 and incubated them with gp120. The mixture was then combined with CXCR4 to form the CXCR4-gp120-CD4 complex (Supplementary Fig. 5b). The resulting structure, at a modest 5.6 Å resolution, enabled us to dock the CXCR4-gp120 complex and the D1-D2 domain of CD4 (Fig. 4a, Supplementary Fig. 9). Unlike the CXCR4-gp120 complex, most of the particles of CXCR4-gp120-CD4 complex displayed two gp120 molecules bound, as indicated by the 2D classification (Supplementary Fig. 9b). However, the density map revealed that only one half of the gp120-CD4 complex was well-resolved, while the other half showed only part of the gp120 molecule. This contrasts with the CXCR4-gp120 complex alone, where most particles had only one bound gp120. Native page gel results showed a broad distribution of CXCR4-gp120^HIV-2 and CXCR4-gp120^HIV-2-CD4 complexes (Supplementary Fig. 5c, d), making it difficult to determine the exact stoichiometry of gp120 and CD4 binding to CXCR4 tetramer. To further clarify this, we employed mass photometry to assess their molecular weights (Supplementary Fig. 5e–g). Despite the heterogeneity, the main peak suggested that one or two gp120 molecules, along with one or two CD4 molecules, bind to CXCR4 tetramer, consistent with cryo-EM observations. One possible explanation for this structural variability is that CD4 binding may increase the stability of gp120, favoring the presence of two gp120 molecules in the complex.

a Different views of CXCR4-gp120^HIV-2-CD4 complex. b Structural comparison of CXCR4-gp120^HIV-2 (in silver) and CXCR4-gp120^HIV-2-CD4 (in color) complexes. c Structural modeling of high-order CXCR4-gp120₃^HIV-2-CD4₃ complex. d Structural modeling of high-order CXCR4-gp120₃^HIV-1-CD4₃ complex. Steric clashes are circled with a red dashed rectangle.

Alignment of CXCR4-gp120 and CXCR4-gp120-CD4 complexes shows no obvious conformational changes in gp120 upon CD4 binding (Fig. 4b). Comparison of our structure to CD4-bound SOSIP Env trimer reveals no significant differences in the core region of gp120, consistent with the CCR5-gp120^HIV-1-CD4 structure. The HIV-1 Env trimer could align with three gp120 molecules from the CXCR4_tetramer-gp120^HIV-2-CD4 complexes without experiencing any spatial clashes (Fig. 4c). To investigate whether this compatibility extends to HIV-1 gp120, we modeled the Env trimer with three generated CXCR4-gp120^HIV-1-CD4 complexes. Surprisingly, the model revealed dominant clashes within adjacent CXCR4 tetramers, preventing the trimeric Env from simultaneously binding three CXCR4-gp120^HIV-1-CD4 complexes (Fig. 4d). This structural incompatibility highlights the notable differences between HIV-1 and HIV-2 gp120 when bound to CXCR4, suggesting that HIV-2 may adopt a distinct assembly architecture during host cell infection. This difference may contribute to the divergent infection mechanisms of HIV-1 and HIV-2, reflecting the unique ways these viruses engage their co-receptors for efficient entry.

Potential mechanism of co-receptor switch

It is well established that the amino acid sequence of V3 determines HIV-1 co-receptor usage-whether the virus binds to CCR5 (“R5 viruses”), predominantly infecting macrophages, or to CXCR4 (“X4 viruses”), primarily infecting T cells^29,30,31. We revisited the modeled CXCR4-gp120^HIV-1 complex to investigate how a specific mutation could drive the transition from an R5 to an X4 phenotype. Previous studies have reported that mutations such as D320R or S306R in the V3 loop are associated with enhanced CXCR4 binding. In our structural model, this aspartate-to-arginine substitution enables a salt bridge with E31, located in the N-terminus of CXCR4, which strengthens the gp120-CXCR4 interaction (Fig. 5a, b). Similarly, the S306R introduces a positively charged residue that can form a cation-π with F189_CXCR4, further stabilizing the interaction (Fig. 5c). Residues 323 and 324 in the V3 loop, which are typically negatively charged or uncharged and associated with CCR5 usage, may from hydrogen bonds or salt bridges with D181 in the ECL2 region of CXCR4 when mutated to positively charged residues (Fig. 5d). Notably, these residues are not conserved in CCR5, highlighting their roles in co-receptor specificity and V3 loop recognition. Our structural model offers a mechanistic explanation for how these site-specific mutations influence co-receptor preference, consistent with prior experimental observations^27,28,30. It also provides fundamental structural evidence for how a single mutation can enable the switch from CCR5 to CXCR4 usage, contributing to our understanding of viral tropism and pathogenesis.

a Structural model of HIV-1 gp120 (green) bound to CXCR4 (magenta), overlaid with CCR5-bound gp120^HIV-1 from PDB: 6MEO (gray), highlighting the V3 loop and positions of key interface residues. b–d Close-up comparisons of wild-type (left panels) and mutant (right panels) V3 loop residues and their interactions with CXCR4. b The D320R mutation introduces a bifurcated salt bridge with CXCR4 residues E31 (N-terminus) and E277 (ECL3), enhancing electrostatic complementarity. c The S306R mutation enables a cation-π interaction with CXCR4 F189. d Double mutation G323R/D324R allows formation of hydrogen bonds or salt bridges with CXCR4 D181 and D182 in ECL2.

Insight into CXCL12 and HIV antagonists’ inhibition

Our structures reveal that CXCR4, whether alone, in complex with CXCL12, or bound to HIV-2 gp120, undergoes minimal rearrangement at its four key binding sites: CRS1, CRS2, ECL2, and ECL3. Notably, the N-terminus of CXCR4 exhibits an outward movement when comparing the CXCL12-bound or gp120-bound states to CXCR4 alone (Fig. 6a). This shift likely accommodates the ligand binding or HIV recognition, providing additional space for their loop insertion. Interestingly, the inserted loops of CXCL12 and HIV-1 penetrate roughly twice as deep as that of HIV-2 (Fig. 6b). Furthermore, the N-terminal loop of CXCL12 passes through the major sub-pocket to reach the minor sub-pocket, with its terminal residue K1 extending into the minor sub-pocket to form contacts with surrounding residues. The N-terminus of CXCL12 also exhibits extensive spatial clashes with the V3 loop of both HIV-1 and HIV-2 (Fig. 6b), explaining why CXCL12 is capable of inhibiting HIV infection by disrupting its binding.

a Comparison between the CXCR4 tetramer, the CXCL12-CXCR4 complex and the CXCR4-gp120^HIV-2 complex. Segments are labeled in the boxed panels. b Comparison between CXCL12 and the V3 loop of HIV-1 and HIV-2. Segments are labeled in the boxed panels. Structural modeling of the V3 loops of HIV-1, HIV-2, and other known CXCR4 and small molecule bound structures: c IT1t (PDB:3OE9); d AMD3100 (PDB:8U4P); e CVX15 (PDB:3OE0), and f vMIP-II (PDB:4RWS), the minor and major pockets of CXCR4 are highlighted with green and blue circular shading, respectively.

To explore whether HIV antagonists inhibit both HIV-1 and HIV-2 entry, we analyzed the relative spatial positioning of the V3 loops of HIV-1 and HIV-2 in the presence of various HIV antagonists, using available structural data^24,25,41. For the small molecule IT1t, it is located in the minor sub-pocket, which causes a clear clash with the V3 loop of HIV-1 but not HIV-2 (Fig. 6c), indicating strong inhibition of HIV-1 infection with a weaker effect on HIV-2. By contrast, the small molecule AMD3100 possesses two “wings” that occupy both the major and minor sub-pockets (Fig. 6d), suggesting its ability to inhibit both HIV-1 and HIV-2 infections. Additionally, a cyclic peptide CVX15 blocks the entry of HIV-1 gp120 by stacking primarily into the major sub-pocket, with its N- and C-termini stretching into the minor sub-pocket, effectively blocking the insertion of the V3 loops of both HIV-1 and HIV-2 (Fig. 6e). Similar to the cognate ligand CXCL12, the viral chemokine vMIP-II also obstructs HIV-1 and HIV-2 entry by creating substantial clashes with the V3 loop (Fig. 6f). This comparative analysis offers valuable insights into how the binding of these antagonists can interfere with viral entry by targeting conserved or distinct regions within the V3 loops of both HIV-1 and HIV-2, providing a foundation for the rational design of therapeutic agents that enhance inhibition efficiency against both viral strains.