A multi-residue electrostatic clamp in HIV-1 CRF01_AE Tat underlies high-affinity engagement with NMDA and hDAT receptors

HIV-1 CRF01_AE is an epidemic in Southeast Asia associated with rapid HIV disease progression. Despite effective antiretroviral therapy, serious non-AIDS-defining complications, including HIV-associated neurocognitive disorder, remain a clinical burden. HIV transactivator of transcription (Tat) protein is a mediator of neurotoxicity known to vary by subtype. Subtype B Tat-mediated neurovirulence has been linked to an arginine residue at position 57 (R57) as a critical structural anchor. However, the CRF01_AE variant lacks the motif, presenting a structural paradox regarding its neuropathogenicity. To elucidate the functional mechanisms of the CRF01_AE Tat exon 1 domain, comprehensive in silico biophysical modeling and protein-protein docking were conducted. Thermodynamic profiling, residue-level mutational scanning, and normal mode analysis (NMA) were utilized to evaluate the epistatic architecture and kinetic stability of the protein when bound to human N-methyl-d-aspartate (NMDA) and dopamine transporter (hDAT) receptors against subtypes B and C. Our computational profiling revealed that the Philippine CRF01_AE Tat exon 1, lacking the R57 motif, evolved a stabilizing, multi-residue lateral electrostatic clamp (K53 and H54), alongside other electropositive substitution mutations. CRF01_AE Tat exhibited high nanomolar affinity for the NMDA (ΔG = -9.6 kcal/mol; K d = 1.7 × 10 –7 M) and hDAT (ΔG = -10.6 kcal/mol; K d = 3.2 x 10 –8 M) neuroreceptors. NMA predicted that the epistatic adaptations specific to CRF01_AE would mechanically lock the central basic domain (residues 40–55) into the neuroreceptor cleft. This was in contrast with subtype B, which relied on the R57 insertion anchor, allowing greater localized conformational freedom at the docking interface. The predicted CRF01_AE Tat structure suggests a strong binding affinity and enhanced kinetic stability with both NMDA and hDAT neuroreceptors. These computational findings indicate that structural adaptations in CRF01_AE Tat may alter neuroreceptor engagement and could contribute to its neurovirulence in HAND.

In contrast to the Global North’s decline in HIV cases, the incidence of HIV infections and AIDS-related deaths is steadily increasing in the Asia-Pacific region from the 2010s to the present (1). Local HIV epidemiology in both the Philippines and the broader Southeast Asian (SEA) region is predominated by the CRF01_AE variant, deviating from the globally dominant subtype B (2–8). The tropical variant is clinically associated with a more aggressive disease progression, marked by higher viral loads and accelerated decline in CD4 + T-cell counts (9–11).

HIV-associated neurocognitive disorder (HAND), a non-AIDS defining event, persists among people living with HIV (PLHIV) even post-active viremia (12–15). While global HAND prevalence has not changed after clinical implementation of antiretroviral therapy (ART), presentations were in milder phenotypes (16, 17). Nevertheless, HAND poses a significant burden in the daily activities of PLHIV. In the Philippines, virally suppressed PLHIV were shown to have significant neurocognitive impairment, specifically in verbal memory and fine motor performances (18). Similar neurocognitive burden was observed in other SEA countries (6–8). Clinically, while subtype B-associated HAND manifests as gradual neurocognitive decline, the CRF01_AE epidemic in SEA is characterized by rapid-onset neurological deficits and profound synergistic toxicity with methamphetamine use (19–21).

The HIV-1 transactivator of transcription (Tat) mediates a role in the pathogenesis of HAND (22–24). Although the virus does not directly infect neurons, infected peripheral immune cells and central nervous system (CNS) reservoirs can actively secrete Tat, exerting neurotoxicity (22, 25, 26). Even among virologically suppressed PLHIV, persisting extracellular Tat, transcribed during HIV-1 latency, drives chronic neuroinflammation, neuronal apoptosis, and synaptic injury (27–31). Mechanistically, this neurovirulence can be attributed to direct structural interactions between Tat and neuronal executioner receptors. Potentiation of the N-methyl-D-aspartate (NMDA) receptor function triggers an uncontrolled calcium influx, leading to severe glutamate excitotoxicity (32, 33). HIV-1 Tat can also bind to the human dopamine transporter (hDAT), functionally impairing dopamine reuptake, exacerbating cognitive and motor dysfunctions frequently observed in HAND, often worsened by concurrent psychostimulant use (34, 35).

Crucially, the neurotoxic potential of HIV-1 Tat is not uniform, rather being markedly subtype-dependent through distinct genetic and structural signatures (36). Majority of foundational neurovirology research is centered on Subtype B, predominant in Global North countries, possessing a highly neurovirulent Tat protein characterized by specific functional motifs, including the canonical arginine 57 (R57) anchor (26, 31). Contrastingly, Subtype C, more globally distributed, presents with an attenuated neurovirulent profile attributed to key structural mutations (e.g., C31S substitution, localized alterations at R57), compromising its ability to inflict neurotoxicity (26, 30). However, there is a paucity on the structural biophysical rationale of the Philippine-dominant CRF01_AE Tat, presenting a critical gap in understanding neurocognitive deficits currently observed among PLHIV in SEA cohorts.

To address this, a comprehensive in silico pipeline, leveraging advances in AI-driven structural predictions, was employed to elucidate the structural and thermodynamic mechanisms that structurally predispose the HIV-1 CRF01_AE lineage to heightened neurovirulence. Through mapping the mutational, thermodynamic, and kinetic binding profiles of different Tat protein variants against neuronal executioner receptors, a molecular rationale can be provided for the observed aggressive neuropathological clinical course associated with the local subtype. Here, we predicted that HIV-CRF01_AE Tat exon 1 utilizes an acquired K53/H54 electrostatic clamp, mounting high-affinity binding with NMDA and hDAT neuroreceptors, structurally predisposing CRF01_AE for neurocognitive disorders associated with the Philippine and the broader Southeast Asian epidemics (7, 14, 37).

Collection and phylogenetic reconstruction of HIV-1 Tat sequences

To assemble the study cohort and ensure high fidelity of subtype classification, HIV-1 Tat exon 1 sequences representing the Philippine CRF01_AE epidemic, alongside global subtype B and subtype C references, were retrieved from https://www.hiv.lanl.gov/content/index and NCBI GenBank ® (38, 39). Inclusion criteria were strictly limited to isolates sampled between 2016 and 2025, with a minimum sequence length of 200 bp. Off-target recombinants and sequences flagged as problematic were excluded from the dataset. Intra-host sampling bias was mitigated by computational retention of a single randomly selected sequence per patient identifier. Sequences from the Los Alamos National Laboratory (LANL) HIV Sequence Database were aligned with the reference HIV-1 HXB2 (NC_001802) Tat nucleotide sequence using HIValign, then clipped to the Tat Exon 1 through Gene Cutter, both built-in the database (46–48). Hence, the final dataset comprised 60 sequences (Supplementary Table 1), distributed evenly across three target lineages (n = 20 per group), allowing the construction of localized phylogenetic quality control without over-sampling bias from localized clinical outbreaks.

Multiple sequence alignment (MSA) was performed using the MUSCLE algorithm in MEGA12, leaving only nucleotides that encode for Tat exon 1 (49). A maximum likelihood (ML) phylogenetic tree (Supplementary Figure 1) was subsequently generated, implementing the General Time Reversible model with a discrete Gamma distribution and estimated proportion of invariable sites (GTR + G + I) (50). The optimal nucleotide substitution model was mathematically determined based on the lowest Bayesian Information Criterion (BIC) score (Supplementary Table 2) (51). Branch support was evaluated by utilizing 1,000 bootstrap replicates, rooted to the HXB2 Tat sequence as the outgroup. To filter out hyper-variable, patient-specific polymorphisms that could skew docking results, the individual isolate exhibiting the shortest phylogenetic branch length from the ancestral root of its respective subtype clade was selected as the clinically representative isolate, which were used for downstream biophysical simulations (Supplementary Table 3). The selection of a representative clinical sequence was done to serve as a high-confidence structural baseline for each subtype.

Mutational profiling and information theory

To identify conserved mutational signatures within Tat protein variants, individual nucleotide alignment files for each subtype (n = 20) were curated, translating them into amino acid residues. Specifically, residues flanking the neurotoxic cysteine-rich (residues 20–35) and basic domains (residues (40–55)) were compared against the HXB2 reference sequence. Focus on these regions were rooted on previous evidence of Tat directly binding to the NMDA receptor through cysteine-residue interactions, and the high affinity of Arg/Lys residues to negatively charged surface targets on CNS cells (32, 46–48). Identified mutation profiles were validated using information theory by calculating the residue-level Shannon entropy and Kullback-Leibler (KL) divergence through the ggseqlogo module in R 4.6.0, quantifying regional conservation in bits (49, 50).

In silico structural prediction and validation of Tat exon 1 variants

Due to the absence of complete crystallographic templates for intrinsically disordered proteins (IDPs) such as HIV-1 Tat, three-dimensional spatial coordinates were generated de novo using AlphaFold3 via Google DeepMind (51–53). Globally predicted Template Modeling scores (pTM) and predicted aligned error (PAE) values were noted for the three variants (54, 55). Peak pLDDT values in the cysteine-rich, and basic domains flanking the C-terminal tail were separately listed. Minimizing ensemble noise and representing the most thermodynamically probable native fold, the highest-confidence prediction (Rank 1/Model 0) was selected for each variant.

To ensure the geometric viability and thermodynamic stability of the most thermodynamically probable model prior to targeted molecular docking, an in silico validation pipeline was employed. Stereochemical quality of the models were assessed via Ramachandran plot analysis using PROCHECK algorithm, through the Structure Validation Server (SAVES) 6.1 (56). Moreover, overall energetic folding quality of the models were evaluated using Protein Structure Analysis (PROSA)-web, verifying alignment of computational coordinates with the energetic profiles of experimentally determined native protein structures (57, 58). Lastly, global model reliability was quantified using QMEANDisCo scoring function via the SWISS-MODEL workspace (59).

Selection and processing of neuronal executioner receptors

The human NMDA receptor (PDB: 6IRA) was retrieved from the RCSB Protein Data Bank, while the hDAT target (AlphaFold DB: Q01959) was obtained from the AlphaFold Protein Structure Database (60–64). Solvent artifacts and stabilizing antibodies were removed from the structure via PyMOL 3.1 Molecular Graphics System (65). To preserve the tetrameric biological structure of NMDA, alternate conformations were systematically removed, followed by a sequential numerical offset (+2000, +4000, +6000) was applied to chains B, C, and D to prevent spatial indexing collisions. All ligand models were standardized to Chain T formatting prior to simulation.

Targeted molecular docking simulations

A 3 x 2 docking matrix was executed using the HADDOCK 2.4 web server, simulating rigid-body, energy-minimized interactions between the Tat variant and each neurocognitive receptor with water as the solvent (66). Ambiguous interaction restraints (AIRs) were explicitly defined based on valid literature (67, 68). Active residue parameters were identified as: residues 31 and 48–57 for Tat, T chain; residues 743–745, 797–799 (and structurally shifted equivalents 4743–4745, 4797–4799) for human NMDA; and residues 87–89, 91–93, 469–471, and 546–548 for hDAT receptor (32, 33, 46–48). Passive residues were automatically defined around the active boundaries. The top-ranked structural cluster, representing the most energetically favorable docking interaction, was extracted for further simulations.

Thermodynamic calculation and electrostatic mapping

To map structural mechanisms driving aforementioned thermodynamic profiles, surface maps were generated using PyMOL’s built-in protein contact potential function, calculating Coulombic vacuum electrostatics and mapping them onto solvent-accessible surface area (69). In order to predict the binding affinity between the proteins in each docking model, the PRODIGY web server was utilized, calculating for the predicted binding affinity value (ΔG), dissociation constant (K d) at both 25 °C and 37 °C, and the number of intermolecular contacts within the 5.5 Å threshold distance were noted for each model (70).

Biophysical validation of key mutations

To validate the thermodynamic influence of specific CRF01_AE mutations independent of the global fold, in silico single-mutation scanning was performed on representative HADDOCK complexes, utilizing the mCSM-PPI2 server (71). The resulting change in ΔG (ΔΔG) was calculated. Lastly, the dynamic impact of mutations on the flexibility of the basic domain was determined through normal mode analysis (NMA) via the iMODS server, comparing main-chain deformability internal atomic mobility profiles B-factors that account for the intrinsically disordered nature of the Tat protein (72–74). The C-alpha (Cα) setting was used as the coarse-grain model for the analysis, simulating oscillating motions of the nodes between atoms through computationally efficient preservation of the carbon backbone topology (75). All downstream 1-D kinetic profiles mapping the exon 1 domain were visualized using the ggplot2 module in R 4.6.0 (76, 77).

Philippine CRF01_AE Tat exhibits mutations within neurovirulent domains

To evaluate the structural and functional divergences across HIV-1 lineages, comparative sequence alignment and corresponding sequence logos were generated for the Tat exon 1 domain of subtypes B, C, and Philippine CRF01_AE (Figure 1). Canonical subtype B (Figure 1A) predicted robust conservation across the basic domain (residues 49–57), highlighted by strict preservation of the arginine 57 (R57) residue, identified as the primary structural anchor for classical neuroreceptor docking (26, 31).

Evaluation of subtype C (Figure 1B) revealed the known C31S substitution within the cysteine-rich region. In previous studies, C31S substitution contributes to a characteristic lower neurovirulence compared to subtype B (78, 79). Most strikingly, the Philippine CRF01_AE alignment (Figure 1C) highlighted seven key mutations. This included the basic R57 residue, replaced by polar serine 57 (S57) located at the basic domain directly involved in neuroreceptor docking (36, 80). The Tat sequence specific to CRF01_AE also had upstream conservation of basic and bulky residues, such as T40K, R53K, and Q54H substitution mutations. The presence of subtype-specific mutations in CRF01_AE suggests potential structural conformation and binding interaction changes of HIV-1 Tat with neuroreceptors.

Structural modeling revealed conformational plasticity among HIV-1 Tat exon 1 variants

We also predicted and evaluated three-dimensional conformations of HIV-1 Tat exon 1 of subtypes CRF01_AE, C, and B. Overall, there was good model confidence, as indicated by the pLDDT scores in all subtypes (Figure 2). Distinct key mutations identified in the primary structure translated into visible 3-D structural changes among proteins. While the Tat proteins of subtypes B and CRF01_AE exhibited similar unbound conformations, subtype C variant showed a more clumped structure, with confident pLDDT values across central residues. While the Tat proteins of subtypes B and CRF01_AE similarly exhibit highly extended, natively unfolded conformations, the subtype C variant adopts a distinctly compacted spatial arrangement. Localized structural collapse observed in subtype C is accompanied by elevated pLDDT confidence scores across central residues, indicating altered structural integrity consistent with C31S induced disruption (79). Conversely, the CRF01_AE protein maintains higher predictive confidence within N-terminal and central domains, yet with drastically lower pLDDT values across C-terminal tail, relative to the subtype B model. These observations are supportive of the intrinsically disordered nature of HIV-1 Tat protein, maintaining conformational plasticity in the solvent-exposed tail (53).

The geometric and thermodynamic integrity of the most thermodynamically probable native fold models were evaluated (Table 1). Stereochemical validation via Ramachandran analysis predicted that 100% of the amino acid residues across all three variants were situated within physically permissible geometries (Supplementary Figure 2). The distribution of residues within the favored and allowed regions are consistent with the biophysical profile of HIV-1 Tat as an IDP (47, 81). Global energetic quality verification yielded negative Z-scores indicative of native-like folding profiles. Lastly, QMEANDiSco evaluations generated global structural scores ranging from 0.38 to 0.44, penalizing structures lacking a buried hydrophobic core, anticipated for highly solvent-exposed IDPs (82, 83).

Electrostatic topography of unbound HIV-1 Tat exon 1 revealed a substantially broadened, highly electropositive surface architecture in the CRF01_AE variant

To evaluate intrinsic biochemical properties of the viral protein prior to neuroreceptor engagement, vacuum electrostatic surface potential maps were generated for the unbound HIV-1 Tat exon 1 variants (Figure 3). Subtype B exhibited a distinct, highly localized region of dense electropositive charges projecting from the protein core, corresponding directly to the conserved arginine 57 (R57) residue, the primary structural probe for classical receptor docking. Conversely, the subtype C variants showed a severely disrupted electrostatic profile, eliminating the electropositive motif required for secure receptor anchoring. However, the CRF01_AE variant revealed profound topological and electrostatic remodeling. Evolutionary loss of the canonical R57 has been compensated by epistatic substitutions (e.g., T40K, R53K, Q54H) that generate a laterally extended electropositive ridge across the solvent-accessible surface which suggest an evolutionary transition from a localized insertion anchor toward a generalized electrostatic steering mechanism, allowing high-affinity engagement with electronegative vestibules of neuroreceptors (84, 85).

To validate the predictive accuracy of the distinct unbound topologies, predicted aligned error (PAE) matrices were evaluated (Supplementary Figure 3). Across all three variants, residues 1–55 suggested exceedingly high positional confidence (<5 Å). Notably, the structural integrity of the electropositive ridge in the CRF01_AE variant is supported by robust off-diagonal PAE confidence scores in residues 40–55. Furthermore, PAE matrices show transition to high positional error (>25 Å) after residue 55 across all variants, physically validating the presence of the solvent-exposed C-terminal tail. The expanded basic topology of the CRF01_AE Tat structure predisposes the unbound protein to engage with neuroreceptors, using broader electronegative vestibules compared to other subtypes.

Epistatic remodeling drives a lateral shift in the docking architecture of Philippine CRF01_AE

Protein–protein molecular docking simulations were executed to model the interface between the Tat exon 1 variants and human neuroreceptors. The canonical subtype B predicted highly localized binding with the human NMDA receptor (Figure 4A), with the conserved R57 residue acting as a singular structural key, inserting vertically into the electronegative binding pocket. However, the interaction between NMDA and CRF01_AE Tat (Figure 4B) predicted a laterally extended binding interface, horizontally spanning the receptor vestibule to maximize spatial contact. The magnified inset reveals the realization of the K53/H54 electrostatic clamp, forming a lateral anchor that secures the Tat protein across a significantly broader surface area of the receptor cleft.

Docking simulations were also executed to the hDAT receptor, visualizing the structural basis of dopaminergic disruption (Figure 5). The canonical subtype B (Figure 5A) captures the conserved R57 residue, descending vertically into the core of the hDAT binding cleft, functionally mirroring a classic lock-and-key mechanism. Contrastingly, the CRF01_AE variant (Figure 5B) suggested structural plasticity in the absence of the R57 residue, leveraging the epistatic electropositive ridge to adopt a highly lateralized docking pose across the upper hDAT vestibule. The multi-domain K53/H54 epistatic clamp engages a broader receptor surface area. Given these observations, CRF01_AE Tat exon 1 presented compensation that allows it to function as a high-affinity ligand for both neuroreceptors, compared to subtype B and C models.

CRF01_AE Tat exon 1 supported advantage in thermodynamic binding affinity with human neuroreceptors attributed to mutations

To quantify the binding efficiency and thermodynamic stability of the HIV-1 Tat variants, the predicted Gibbs free energy (ΔG) and dissociation constants (K d) were calculated for each Tat-neuroreceptor complex (Table 2). Due to binding affinities being inherently temperature-dependent, K d was evaluated at both standard (25˚C) and simulated physiological (37˚C) conditions, representing the kinetic environment of the human CNS (86). All complexes exhibited slight kinetic relaxation (increase in K d) when subjected to physiological thermal stress, but relative hierarchical affinities between viral clades remained conserved (Supplementary Table 4).

Within the NMDA receptor docking interface, the CRF01_AE Tat supported the most favorable binding free energy and the lowest dissociation constant under physiological conditions, superseding canonical subtype B. The subtype C variant yielded the weakest binding energetics under physiological conditions, consistent with its role as the negative control for this study (78, 79). These suggest the CRF01_AE Tat is predicted to achieve greater glutamatergic excitotoxicity compared to other subtypes.

Contrastingly, subtype B predicted the strongest overall binding affinity and physiological dissociation constant among the three models. While the CRF01_AE model did not surpass subtype B thermodynamics at the hDAT interface, it still maintains a competitive binding profile. Subtype C, consistent with its phenotype for reduced neurovirulence, showed the weakest affinity for this receptor. Collectively, the evaluation of both receptor interfaces provides a structural hypothesis for the multi-pathway neural disruption driven by the Philippine CRF01_AE lineage. The local subtype successfully maintains high nanomolar affinities across both distinct neurological targets (3.1 nM for NMDA, 34.0 nM for hDAT).

To isolate the specific thermodynamic impact of individual amino acid substitutions acquired by the non-B HXB2 lineages, in silico mutational scanning was performed (Table 3). A positive ΔΔG indicates a stabilizing effect on the protein-protein interaction, whereas a negative value change signifies thermodynamic destabilization. Overall, the mutational landscape reveals several non-synonymous mutations in the CRF01_AE Tat exon 1 exert a net stabilizing effect across both neuroreceptor interfaces. However, localized destabilization has been introduced through specific structural ablations, such as Y47A and the evolutionary loss of the R57 residue (R57G in CRF01_AE; R57S in subtype C). These findings suggest the maintenance of viability in the CRF01_AE variant through compensatory epistatic networks.

At the NMDA receptor docking interface, most profound thermodynamic enhancements are directly driven by acquisition of positively charged residues within the central basic domain, such as R53K and Q54H. These mutations operate at close intermolecular proximity, physically suggesting an energetically favorable electrostatic clamp domain. Secondary structural stabilization is supplied by mutations localized to the flexible C-terminal tail, such as S70P and T64D.

A distinctly different residue-level functional hierarchy governs hDAT docking with HIV-1 Tat. The T40K mutation emerges as the primary thermodynamic driver, engaging the transporter at a close intermolecular distance. This is closely followed by the C-terminal substitution mutations Q54H and T64D. Critically, severe thermodynamic penalties have been incurred by the loss of the R57 anchor (R57G, R57S) and the Y47A substitution. However, epistatic compensation in CRF01_AE occurs, sustaining a high-affinity profile against the dopaminergic neuroreceptor.

The R53K/Q54H epistatic clamp restricts local deformability and thermodynamically locks the CRF01_AE docking interface

To assess the sustained mechanical stability of the bound Tat-neuroreceptor complexes, kinetic mobility and overall complex rigidity were evaluated using NMA (Figure 6). Global dynamic properties, quantified through motion eigenvalues, revealed that the CRF01_AE variant harbors a rigid and mechanically sound complex across both neurological targets (Supplementary Table 5). The overall thermodynamic stability of subtypes B (1.55 x 10 -5) and C (1.71 x 10 -5) was lower than that of CRF01_AE (1.50 x 10 -5) when bound to NMDA (Supplementary Figure 4). However, subtype B Tat bound to the hDAT receptor required the least amount of energy to excite (1.34 x 10 -6), followed by CRF01_AE (1.87 x 10 -6) and subtype C (2.69 x 10 -6).

Residue-specific main-chain deformability plots show the precise mechanical rationale for global HIV-1 Tat stability at the bound state. The CRF01_AE variant within the NMDA binding complex exhibited a dramatic suppression of kinetic mobility across the central basic domain, with localized deformability lowest at the exact region harboring the R53K/Q54H mutations, approaching zero. These strongly predict that the electrostatic clamp of the CRF01_AE Tat anchors the domain deeply and tightly into the neuroreceptor cleft, suggesting a structural constraint at the interacting complex.

Critically, the CRF01_AE deformability profile exhibits a sudden, sharp rise after residue 60, supporting that the C-terminal tail is highly mobile, unbound, and solvent-exposed. When the CRF01_AE is bound to the hDAT receptor, deformability fluctuates more broadly, maintaining sufficient structural rigidity. Conversely, the NMA profile of subtype C displays massive, anomalous deformability peaks between residues 20–35, corresponding directly to the localized C31S mutation and distinct from the CRF01_AE profile. Meanwhile, subtype B presents high localized mobility at the N-terminus, with stability moderating throughout the middle regions.

Understanding the neurovirulence potential of the locally circulating CRF01_AE variant is a clinical imperative in the landscape of HAND in the Asia-Pacific region (2, 3, 5, 7). Our study provides biophysical baseline data on the genetic signatures of the CRF01_AE Tat exon 1 interacting with neuroreceptors NMDA and hDAT, affecting neuropathogenicity (32, 35). Principally, our results indicate that the Philippine HIV-1 CRF01_AE Tat exon 1 utilizes a distinct structural mechanism maintaining high-affinity interactions with both neuroreceptors. Thermodynamic profiling revealed CRF01_AE has evolutionary divergence when compared to the globally dominant subtype B, which showed greater binding affinity to the hDAT receptor (ΔG = -12.8 kcal/mol) reliant on the deep-pocket insertion of the canonical R57 residue.

While the CRF01_AE lacks the R57 anchor, it actively compensates for it through an acquired K53/H54 motif, generating a broad electrostatic clamp along with other key mutations. These motifs are highly prevalent evolutionary signatures in non-B lentiviral lineages (87–89). In vitro nuclear magnetic resonance (NMR) studies experimentally validated the alteration of topological folding and binding mechanics in Tat-receptor models with presence of the H54 residue (90) CRF01_AE, a circulating recombinant between parent subtypes A and E, is a robust, ancestral evolutionary strategy inherent to A-derived Tat proteins, revealing that the R53K and Q54H motifs are also conserved in European variant subtype A6, also prevalent of severe neurological complications (87, 91–93). While slightly less specialized for hDAT (ΔG = -10.6 kcal/mol), the adaptation predicts stronger thermodynamic binding affinity to the NMDA receptor (ΔG = -9.6 kcal/mol) than subtype B (ΔG = -7.9 kcal/mol).

The functional significance of predicting a stable docking interface is further underscored by our subtype C comparative model. Consistent with established literature, signature C31S substitution in subtype C induces localized collapse of the hydrophobic core, attenuating neuroreceptor docking capacity (NMDA ΔG = -5.5 kcal/mol; hdDAT ΔG = -9.1 kcal/mol) (26, 78, 79). In contrast, the CRF01_AE lateral clamp avoids the structural collapse by distributing an electropositive network across a broader buried surface area, supporting potent nanomolar dissociation constants for both neuroreceptors (NMDA K d = 1.7 x 10– 7 M; hDAT K d = 1.0 x 10– 9 M). Furthermore, NMA findings suggest that upon engagement with the neuroreceptor cleft, the clamp induces structural rigidity across the central basic domain while preserving the flexibility of the solvent exposed C-terminal tail, a kinematic profile aligned with the intrinsically disordered nature of the HIV-1 Tat protein (51, 53, 94).

Computational predictions in our study offered potential structural rationale for the neurocognitive impairment observed in the Southeast Asian HIV cohort (7, 8, 18). First, the predicted high-affinity engagement of CRF01_AE with the NMDA receptor suggested a biomolecular rationale for chronic, calcium-mediated glutamatergic excitotoxicity (32, 33). This hypothesis aligns strongly with the early-stage motor and memory dysfunctions frequently reported in CRF01_AE-dominant cohorts across Southeast Asia (7, 17). Concurrently, sustained capacity of the CRF01_AE variant in disrupting the hDAT receptor offers implications for patients with comorbid substance abuse disorders (95, 96). Management of the HIV epidemic in the Philippines is frequently complicated by the concurrent use of psychostimulants (e.g., methamphetamine), which also disrupt dopaminergic dysregulation through hDAT (97–99).

Consequently, both dysregulated neural pathways are directly attributed to the observed epistatic substitutions that generated a wider electrostatic ridge. The predicted CRF01_AE Tat binding to multiple receptors may predispose PLHIV to exacerbated synergistic neurotoxicity, which has been an independent risk factor for death in HIV infection (34, 35, 100). Thus, recognizing neurovirulence of this recombinant variant is clinically relevant. However, receptor affinity alone cannot definitely establish neuronal toxicity. HAND pathogenesis is multifactorial syndrome, which can be influenced by host-related factors such as immune activation, immune exhaustion, and host genetics (101). It is also further influenced by viral replication dynamics such as secretion rates of the Tat protein, localized receptor expression levels (102). While viral suppression in the CNS is achieved with ART with optimal penetration effectiveness, certain regimens may also contribute to neurotoxicity (103). Lastly, aging and co-morbidities (e.g., cardiovascular disease, coinfection) also contribute to hypoxia and inflammation, which can worsen cognitive impairment (104).

However, evaluating protein-protein interactions and their kinematics requires critical methodological consideration, specifically regarding algorithmic limitations. The use of AlphaFold3 to model HIV-1 Tat, an IDP, only presents a functional predicted docking, which underrepresents the conformational landscape and structural plasticity that characterize IDPs in vivo. Nevertheless, the lowered pLDDT scores localized to the C-terminal tail represent highly accurate biophysical reflections of the natively unfolded, hyper-flexible nature of the domain. Furthermore, while the HADDOCK algorithm successfully isolated the most energetically favorable conformational clusters, protein-protein docking inherently provides robust structural hypotheses, which remain model-dependent (66–68). We acknowledge that transient docking orientations may also be energetically viable in a dynamic physiological environment. Moreover, PRODIGY derives robust predictive power from structural interfacial contacts (ICs) and non-interacting surface (NIS) properties, making them reliable for modeling interactions involving highly flexible domains.

While targeted NMA successfully captured large-scale, collective, harmonic motions characterizing viral interfaces, we acknowledge that it does not provide the high-resolution temporal data regarding kinetic stability afforded by molecular dynamics (MD) simulations, which are recommended for future simulations. However, HIV-1 Tat being an IDP, short-timescale MD simulations frequently result in computational noise and non-converging trajectories. Consequently, extensive microsecond-timescale MD utilizing specialized IDP-centric force fields represents a critical future directive to deeply validate the kinetic stability of the K53/H54 clamp in this study. Ultimately, by systematically addressing and compensating for the aforementioned inherent limitations of individual computational tools, our framework provides highly robust and confident predictive models of the viral interface (105, 106).

Our structural predictions necessitate extensive experimental testing. We recommend that future in vitro studies should quantify the binding kinetics of Tat variants and determine their functional effects, utilizing surface plasmon resonance (SPR), isothermal titration calorimetry (ITC), microscale thermophoresis (MST), and cryo-EM structural resolution. To confirm hypothesized excitotoxic and dopaminergic profiles, future studies should evaluate these variants in primary cortical and striatal neuronal cell lines (107). To provide direct evidence of heightened NMDA-mediated excitotoxicity, Fura-2 ratiometric imaging could be performed to measure intracellular Ca 2+ influx following neuronal exposure to variant-specific pseudovirions. Furthermore, mechanical efficacy of the dopaminergic lateral camp can be quantified by monitoring rapid internalization rates of surface-expressed hDAT through multi-parametric flow cytometry, electrophysiology, and confocal microscopy. Elucidating the neurovirulence profile of CRF01_AE may support the implementation of earlier, genotype-informed neurocognitive screening protocols for at-risk cohorts in the Asia-Pacific region (17, 108).

HIV-1 CRF01_AE Tat exon 1 has evolved a distinct receptor-binding mechanism that compensates for the absence of the canonical R57 motif. Through epistatic electropositive substitutions, particularly the formation of a stabilizing lateral electrostatic clamp involving K53 and H54, CRF01_AE Tat supports strong binding affinity and enhanced kinetic stability with both NMDA and hDAT neuroreceptors. Normal mode analysis further predicts that these adaptations mechanically constrain the central basic domain within the receptor-binding cleft, which may support a more rigid and stable interaction than that observed in subtype B. Collectively, these results highlight a predicted unique structural and functional evolutionary pathway in CRF01_AE Tat that may contribute to altered neuroreceptor engagement and potentially influence HIV-associated neuropathogenesis. Future research must prioritize the validation of docking energetics in vitro, alongside the confirmation of heightened functional excitotoxicity and receptor internalization using primary neuronal cell cultures. Ultimately, characterizing lineage-specific docking architecture provides foundational information necessary for developing targeted neuroprotective therapeutics, minimizing the global burden of HAND.