CROI 2025: Summary of Basic Science Research in HIV

Abstract

The 2025 Conference on Retroviruses and Opportunistic Infections (CROI) in San Francisco maintained its existing format with a combination of plenary lectures, workshops, oral and poster abstract sessions, themed discussions, and interactive symposia to deliver the latest advances in HIV/AIDS research to the approximately 4000 delegates in attendance. The conference featured a comprehensive collection of presentations addressing the molecular biology of HIV-1, with the basic virology track offering mechanistic insights into viral replication, immune evasion, and host-pathogen interactions. CROI showcased a range of innovative approaches to decipher and target the latent reservoir. From high-resolution lineage tracking in nonhuman primates to dissection of chromatin landscapes and latency regulatory circuits, studies presented at this year's meeting underscored the complexity of HIV persistence and the need for multidimensional intervention strategies. Selected abstracts are highlighted, emphasizing mechanistic insights, methodologic innovations, and therapeutic implications. As with prior renditions of the conference, CROI continues to set the standard for engagement of early career investigators. Sessions such as the Scott M. Hammer Workshop for New Investigators and Trainees provide an effective forum for orientation to the various thematic areas covered at CROI.

Introduction

The 2025 Conference on Retroviruses and Opportunistic Infections (CROI) presented an extensive array of talks focused on the molecular biology of HIV-1. The basic virology track provided in-depth insights into the mechanisms of viral replication, immune evasion, and host-pathogen interactions. The conference also highlighted a variety of innovative strategies aimed at understanding and targeting the latent HIV reservoir.

Virology

HIV-1 integration into the host genome is not random but is influenced by host chromatin organization and transcriptional activity. Integration into transcriptionally active regions can facilitate viral gene expression but may also expose proviruses to immune surveillance. Elite controllers — rare individuals who control viral replication without antiretroviral therapy (ART) — provide a unique model to understand how immune pressure may select for latent integrations. The mechanisms guiding integration to specific chromatin contexts remain a crucial focus for understanding reservoir establishment and persistence.

Bushman (Abstract 13) presented studies involving high-throughput integration site sequencing on peripheral blood mononuclear cells (PBMCs) from ART-treated individuals and elite controllers. Genomic DNA was extracted and subjected to ligation-mediated polymerase chain reaction (PCR) followed by next-generation sequencing (NGS). The resulting integration site data were mapped to the human genome (hg38) and compared across individuals using custom bioinformatic pipelines. Integration-site enrichment was assessed relative to chromatin state maps, lamina-associated domains (LADs), and nuclear lamina boundaries identified by DNA adenine methyltransferase sequencing (DamIDseq). In vitro models using pseudotyped lentiviral vectors were used to assess transcriptional outcomes of integration in defined chromatin contexts. Functional assays included luciferase reporter constructs with engineered integration sites, and single-cell RNA sequencing (scRNA-seq) to quantify expression variability. Elite controllers exhibited a highly skewed integration pattern, with a marked enrichment in LADs and repressive chromatin regions defined by histone 3 lysine 9 trimethylation (H3K9me3) and H3K27me3 histone modifications. Integration in ART-treated individuals showed a broader distribution with substantial overlap in transcriptionally active genes marked by H3K4me3 and open chromatin (assay for transposase-accessible chromatin with high-throughput sequencing [ATAC-seq] peaks). Reporter assays revealed that integrations within active promoters or enhancers led to robust viral gene expression, whereas those in repressive chromatin were transcriptionally silent. The 4 mutagenic mechanisms identified included (1) enhancer hijacking, (2) promoter interference, (3) gene truncation via exon disruption, and (4) 3’ untranslated region (UTR) interference. These modes were confirmed through functional readouts and integration site mapping near key host oncogenes and immune regulatory genes. These findings suggest that in elite controllers, the immune system preferentially clears cells harboring actively transcribed proviruses, leading to the persistence of deeply latent clones. This supports a model of immune editing that shapes the integration landscape.

Lentiviruses such as HIV-1 uniquely retain the ability to infect nondividing cells by actively transporting their replication complexes through the nuclear pore complex (NPC). This has facilitated the development of lentiviral vectors that can be used to transduce terminally differentiated cells (such as muscle cells or neurons) for gene therapy applications. The HIV-1 capsid (CA) plays an essential role in mediating this nuclear import. The CA hexamer lattice presents a hydrophobic pocket that binds to host factors involved in nuclear trafficking, such as cleavage and polyadenylation specificity factor 6 (CPSF6), and now, newly identified, phenylalanine-glycine (FG)-nucleoporins (Nups). This mimicry of karyopherin-like interactions permits viral complexes to bypass canonical nuclear import receptors and directly access the nuclear interior.

Jacques (Abstract 29) characterized the interaction between HIV-1 CA and FG-repeat-containing Nups (Nup153 and Nup358) using various orthogonal approaches. CA tubes were generated in vitro and subjected to binding assays with purified FG-repeat peptides and recombinant Nups. Surface plasmon resonance (SPR) and isothermal titration calorimetry (ITC) were used to determine affinity and binding kinetics. Point mutations (eg, N74D, A105T) were introduced into the CA to disrupt the hydrophobic FG-binding pocket. The impact of these mutations on nuclear import was assessed in synchronized cell cycle-arrested HeLa cells using subcellular fractionation and quantification of 2-long terminal repeat (LTR) circles. Cryo-electron tomography (cryo-ET) visualized CA-NPC engagement in situ. Lenacapavir was tested for its ability to inhibit binding in vitro and nuclear import in cell culture. The CA hexamer specifically bound to FG-repeat domains of Nup153 and Nup358 with low-nanomolar affinity, mimicking canonical karyopherin-FG interactions. Structural analysis showed that the CA engages these motifs through a conserved hydrophobic cavity formed between hexamer interfaces. Mutations in the FG-binding pocket abolished FG binding and blocked nuclear import, resulting in a 90% reduction in 2-LTR circles in nondividing cells. Cryo-ET imaging revealed partial docking of CA lattices at the NPC face, with structural deformation of the NPC membrane consistent with pore dilation. Lenacapavir abrogated FG binding and phenocopied the N74D mutant, confirming that its antiviral activity is mediated through disruption of CANup interactions. This study demonstrates that HIV-1 CA functions as a viral karyopherin, enabling nuclear import through direct mimicry of host transport receptors. This represents an evolutionary adaptation that allows lentiviruses to replicate in terminally differentiated, nondividing cells such as macrophages and microglia. The CA FG-binding pocket is not only structurally conserved but also pharmacologically targetable, making it an attractive candidate for next-generation CA inhibitors.

HIV-1 integration exhibits strong preferences for gene-dense and transcriptionally active regions of the host genome. This site selection is not random but is rather facilitated by host proteins that interact with the preintegration complex (PIC). Two major cofactors, CPSF6 and lens epithelium-derived growth factor p75 (LEDGF/p75), have been shown to guide HIV-1 to nuclear speckle-associated domains (SPADs) and transcriptionally active gene bodies, respectively. The biophysical mechanisms underlying their targeting behavior remain incompletely understood, but evidence suggests that both proteins form liquid-liquid phase-separated (LLPS) condensates that may concentrate the integration machinery at specific nuclear compartments.

Engelman (Abstract 30) presented a comprehensive analysis of the process of HIV integration through a combination of chromatin immunoprecipitation sequencing (ChIP-seq), super-resolution microscopy, and optogenetic perturbation to investigate the roles of CPSF6 and LEDGF/p75 in integration targeting. Wildtype and LLPS-deficient mutants of both proteins were expressed in primary CD4+ T cells and Jurkat cells. Time-lapse imaging tracked HIV-1 PIC localization in real time, using fluorescently tagged integrase and RNA aptamer-tagged viral genomes. Integration site mapping was performed using linear amplification-mediated (LAM) PCR and deep sequencing. Disruption of phase separation was achieved using 1,6-hexanediol and targeted mutagenesis of intrinsically disordered regions (IDRs). Additional validation was conducted using a clustered regularly interspaced short palindromic repeats (CRISPR)-associated protein 9 (Cas9)-based recruitment system to reposition LEDGF domains. LEDGF/p75 was enriched at transcriptionally active mid-gene regions marked by H3K36me3 and bound directly to chromatin through its PWWP (proline-tryptophan-tryptophan-proline) domain. CPSF6 localized near nuclear speckles and mediated PIC trafficking to SPADs, as visualized by structured illumination microscopy. Integration sites shifted toward lamina-proximal and heterochromatic regions in cells expressing mutant cofactors. Pharmacologic disruption of phase separation using 1,6-hexanediol led to similar mislocalization of PICs and reduced integration efficiency. Artificial tethering of LEDGF IDRs to heterochromatic loci partially restored targeting specificity. By forming condensates at distinct nuclear compartments, CPSF6 and LEDGF/p75 enrich viral PICs in favorable genomic regions for transcriptional competence. This spatial organization likely enhances the probability of successful proviral expression and persistence.

The HIV-1 protease (PR) is a key enzyme responsible for group-specific antigen (Gag) and Gag-polymerase (Pol) polyprotein processing during viral maturation. Although PR activation is tightly temporally regulated, premature activation can expose viral components to host immune sensors. Caspase recruitment domain family member 8 (CARD8) is a cytosolic innate immune sensor that is activated by HIV-1 PR activity, leading to inflammasome formation, caspase-1 activation, interleukin (IL)-1β secretion, and pyroptotic cell death. Understanding the molecular regulation of PR and its implications for immune evasion and viral cytopathicity is important to the design of reservoir-elimination strategies that promote infected cell death.

Hughes and colleagues (Abstract 134) engineered a series of HIV-1 Gag mutants targeting the matrix (MA), CA, nucleocapsid (NC), and p6 domains to disrupt PR activation timing and viral assembly. HEK293T cells and THP-1-derived macrophages were transfected or infected with wild-type and mutant HIV constructs. CARD8 activation was measured via IL-1β enzyme-linked immunosorbent assay (ELISA), lactate dehydrogenase release assays, and Western blotting of caspase-1 cleavage. In addition, CRISPR-mediated CARD8 knockout cell lines were used to validate specificity. Primary CD4+ T cells from healthy donors were also infected with mutant viruses to assess inflammasome activation in a more physiologic context. The team used confocal microscopy to evaluate Gag localization and virus assembly defects and measured infectivity using TZM-bl reporter assays.

Gag mutants that disrupted membrane targeting (MA deletion), multimerization (CA mutations), or budding (p6 ΔPTAP) led to premature PR activation, as evidenced by increased PR autocleavage and early processing of Gag. This correlated with robust CARD8 activation, including increased IL-1β secretion and pyroptotic cell death. In primary CD4+ T cells, premature PR activity induced pronounced CD4+ T-cell death and inflammasome activation, which was absent in CARD8-deficient cells.

These data demonstrate that Gag structure and assembly tightly regulate PR activation to prevent premature sensing by CARD8. HIV-1 evades innate immune detection by coupling PR activation to late stages of viral assembly. Disruption of this process exposes infected cells to inflammasome-mediated cell death. These findings suggest a novel therapeutic approach: pharmacologic agents that destabilize Gag multi-merization or uncouple PR activation from budding may unmask infected cells to immune clearance via CARD8 inflammasomes. This approach could complement latency reversal strategies and enhance reservoir clearance.

The mechanism of HIV-1 nuclear entry has long been a subject of debate, particularly in nondividing cells where the nuclear envelope remains intact. Although biochemical evidence has supported the role of the CA in nuclear import, direct structural visualization of intact HIV-1 cores engaging the NPC in a near-native state has been limited. Advances in cryo-focused ion beam (FIB) milling and cryo-ET have made it possible to observe viral particles in cellular contexts at nanometer resolution.

Hou and colleagues (Abstract 135) developed a correlative light and cryo-electron microscopy pipeline to directly observe nuclear translocation of viral PICs. Primary CD4+ T cells infected with HIV-1 were high-pressure frozen and prepared into thin lamellae using cryo-FIB. Fluorescently labeled HIV-1 cores were tracked with live-cell imaging and correlated to cryo-ET fields. Structural classification of CAs was performed using subtomogram averaging. Morphologic features were correlated with nuclear envelope docking and pore interaction events. Additional immunogold labeling was performed for nuclear pore markers (Nup153, Nup358). Out of more than 500 virions analyzed, approximately 87% showed nuclear association. HIV-1 CAs approaching the NPC were predominantly in tubular or irregular morphologies rather than the classic conical form, suggesting a remodeling process during nuclear engagement. NPCs interacting with CAs showed dilation and distortion of the nuclear envelope. Cryo-ET slices revealed electron-dense material consistent with FG-Nup meshwork along the CA surface. CA structures were partially intact during nuclear pore docking, with some displaying visible uncoating intermediates. These observations were consistent across independent biologic replicates and aligned with CA mimicry of karyopherins, as previously shown biochemically. This study offers the most direct structural visualization to date of HIV-1 nuclear import and supports a model in which the intact or partially uncoated CA interacts directly with the NPC. The observed deformation of CA and pore architecture suggests a dynamic translocation mechanism. The data support the notion of CA flexibility as a functional determinant of successful nuclear entry.

HIV-1 negative regulatory factor (Nef) is a multifunctional accessory protein known to manipulate host signaling pathways to enhance viral replication and immune evasion. Recent work has implicated Nef in the suppression of innate immune detection via cytosolic pattern recognition receptors, including retinoic acid-inducible gene-I (RIG-I) and melanoma differentiation-associated protein 5 (MDA5), which signal through the mitochondrial antiviral signaling protein (MAVS). The actin cytoskeleton plays a key role in regulating RIG-I/MDA5 trafficking and activation, yet the role of Nef in modulating cytoskeletal dynamics to subvert immune recognition remains poorly understood.

Laliberté and colleagues (Abstract 141) investigated how Nef interferes with actin remodeling and RIG-I/MDA5 signaling using a combination of site-directed mutagenesis, phospho-cofilin Western blotting, and reporter assays. THP-1 macrophage-like cells and primary monocyte-derived dendritic cells were transduced with wild-type or mutant HIV-1 constructs lacking functional Nef. Nef mutants included R191A and F191A substitutions, designed to disrupt Pak2-dependent cofilin inactivation. Cofilin phosphorylation was assessed by Western blot; RIG-I/MDA5 activation was quantified using luciferase reporters for interferon (IFN)-stimulated response elements (ISRE) and quantitative (q) PCR of IFN-stimulated gene (ISG) transcripts (MX1, IFNB1, C-X-C motif chemokine ligand 10 [CXCL10]). CRISPR/Cas9 was used to knock out MAVS to confirm pathway specificity. Imaging of actin structures was performed by confocal microscopy and phalloidin staining. Cells infected with Nef-deficient or Nef-mutant HIV-1 showed elevated levels of phospho-cofilin, indicative of restored actin depolymerization. This correlated with increased RIG-I/MDA5 activation and upregulation of ISGs, including a more than 10-fold induction of MX1 and CXCL10 compared with wild-type virus. Wild-type Nef suppressed cofilin phosphorylation, limiting cytoskeletal remodeling and dampening pattern-recognition receptor activation. MAVS knockout abrogated these responses, confirming the reliance on the RIG-I/MDA5 axis. Imaging revealed altered cortical actin networks in the presence of functional Nef, consistent with disruption of pattern-recognition receptor (PRR) scaffold formation. These findings suggest that HIV-1 Nef suppresses innate immune signaling by modulating the actin cytoskeleton, thereby inhibiting RIG-I/MDA5 trafficking and MAVS-dependent signaling. This represents a novel mechanism of immune evasion that operates at the interface of cytoskeletal and innate immune regulation.

The early stages of HIV-1 infection are shaped by the selective transmission of CCR5-tropic (R5) viruses, and CXCR4-tropic (X4) variants tend to emerge later in the disease. This pattern suggests that host restriction mechanisms may differentially affect viral strains depending on coreceptor usage. Solute carrier family 35 member A2 (SLC35A2) is a nucleotide-sugar transporter involved in Golgi-mediated glycosylation of cell surface proteins, including the HIV-1 envelope glycoprotein (Env) receptor complex. Disruption of glycosylation has been shown to alter virus-cell interactions and may serve as a barrier to X4-tropic virus fusion and entry.

Guenthoer and colleagues (Abstract 323) used CRISPR-Cas9 to generate SLC35A2-knockout (KO) primary CD4+ T cells from healthy donors. They infected wild-type and SLC35A2-KO cells with CXCR4-tropic HIV-1 strain NL4-3 and CCR5-tropic strain JRFL. Viral fusion was assessed using a β-lactamase (BlaM)-Vpr assay, and early reverse transcription products were quantified using qPCR. Flow cytometry was used to assess CD4 and CXCR4 surface expression and Env binding. Western blot and mass spectrometry were used to profile glycosylation changes. To determine the stage of restriction, the team also evaluated viral attachment, fusion, reverse transcription, and nuclear import.

SLC35A2 KO enhanced fusion and entry of CXCR4-tropic HIV-1 by about 5-fold without substantially affecting CCR5-tropic infection. The effect was specific to viral fusion, as prefusion binding and postfusion reverse transcription were elevated in the KO condition for X4 virus. BlaM-Vpr assays confirmed a stage-specific enhancement of fusion in SLC35A2-KO cells. Glyco-proteomic profiling revealed altered sialylation and fucosylation patterns in the KO cells, affecting CXCR4 surface glycosylation and Env-coreceptor interaction. These changes appeared to reduce steric hindrance at the virus-cell interface, facilitating X4 Env-mediated fusion.

This study identifies SLC35A2 as a novel host restriction factor that selectively inhibits CXCR4-tropic HIV-1 by altering glycosylation of the viral entry machinery. Although these findings suggest the involvement of a host restriction that selectively impacts CXCR4-tropic virus infection, more work is needed to assess whether SLC35A2 is part of the puzzle as to why CCR5-tropic and not CXCR4-tropic viruses are selectively involved in sexual transmission of HIV-1 virus or whether it plays a role in the emergence of CXCR4-tropic virus later in disease progression.

HIV-1 Vpr is a multifunctional accessory protein implicated in viral replication enhancement, immune modulation, and DNA damage response. Recent studies have highlighted its role in manipulating host restriction pathways, particularly those linked to IFN signaling. Ras homologue family member A (RhoA), a small guanosine triphosphatase (GTPase) known for its role in cytoskeletal regulation, has emerged as a positive regulator of IFN signaling. Its degradation during HIV infection suggests a possible immune evasion strategy by Vpr.

Prelli Bozzo and colleagues (Abstract 327) performed a genome-wide CRISPR-Cas9 knockout screen in primary human CD4+ T cells to identify host factors restricting HIV replication. RhoA emerged as a hit, and its functional role was validated through targeted KO, overexpression, and pharmacologic inhibition experiments. The interaction between Vpr and RhoA was assessed by co-immunoprecipitation, and RhoA ubiquitination and degradation were tracked using Western blotting and proteasome inhibition assays. Downstream signaling was evaluated by quantifying signal transducer and activator of transcription 1 (STAT1) phosphorylation and IFN-stimulated gene (ISG) expression using qPCR and phospho-flow cytometry. In vivo relevance was tested using simian immunodeficiency virus (SIV) mac239- infected rhesus macaques, including wild-type and Vpr-deficient viral strains. RhoA KO increased HIV-1, confirming its role as a restriction factor. Vpr directly bound to RhoA and promoted its ubiquitination and proteasomal degradation. This degradation resulted in reduced phosphorylation of STAT1 and decreased ISG expression, including ISG15 and MX1. Treatment with proteasome inhibitors rescued RhoA levels and partially restored antiviral signaling. In SIV-infected macaques, animals receiving Vpr-deficient virus displayed higher RhoA expression and more robust ISG induction in lymphoid tissues, accompanied by lower plasma viral loads than wild-type SIV. This study defines RhoA as an IFN-inducing restriction factor targeted by HIV-1 Vpr. Vpr-mediated RhoA degradation allows HIV to subvert type I IFN signaling, promoting viral replication and persistence. These findings underscore the importance of innate immune modulators in shaping viral pathogenesis.

The gastrointestinal mucosa is a primary site of early HIV replication and CD4+ T-cell depletion. The viral accessory protein Vpr contributes to immune dysregulation and pathogenesis, yet its function in the gut microenvironment remains poorly understood. Given the extensive microbial exposure at mucosal surfaces, the interplay between Vpr and microbial signals may critically influence immune cell fate and barrier integrity during acute infection.

Mickens and colleagues (Abstract 345) used the lamina propria aggregate culture (LPAC) model of human jejunal tissue to study Vpr function in ex vivo gut mucosa. Lamina propria mononuclear cells (LPMCs) were isolated from jejunal biopsies of healthy donors and infected with transmitted/founder (T/F) HIV-1 strains, either wild-type or Vpr-deficient. Some cultures were co-treated with Escherichia coli lysate to simulate bacterial translocation. Viral replication was monitored via p24 ELISA, and cell death was assessed by annexin V/propidium iodide staining. Granzyme B expression in CD8+ and CD4+ T cells was measured using intracellular flow cytometry. CD4+ T-cell subset depletion and cytotoxic differentiation were quantified by flow cytometry and single-cell RNA-seq.

Wild-type HIV-1, particularly in the presence of E coli lysate, induced elevated granzyme B expression in CD8+ and CD4+ T cells, leading to substantially more CD4+ T-cell death than Vpr-deficient viruses. Granzyme B+ CD4+ T cells exhibited features of cytotoxic effector differentiation, including perforin expression and IFN-γ production. scRNA-seq revealed an inflammatory transcriptional program enriched in Vpr-expressing infections, including upregulation of tumor necrosis factor (TNF), IL32, and cytotoxic effector genes. Viral replication was markedly enhanced in the wild-type conditions, correlating with increased CD4+ T-cell depletion and microbial responsiveness. This study demonstrates that Vpr enhances the sensitivity of gutresident CD4+ T cells to microbial stimuli, promoting cytotoxic differentiation and accelerating mucosal immune damage. The synergy between Vpr and bacterial products may drive early CD4+ T-cell loss and compromise gut barrier integrity, facilitating systemic immune activation and viral dissemination. Therapeutic strategies that limit microbial translocation or Vpr activity may preserve mucosal immunity and reduce the severity of early infection.

Viral Reservoirs and Cure

Combination ART effectively suppresses HIV-1 replication, transforming what was once a fatal illness into a chronic, manageable condition. However, ART is noncurative due to the persistence of a latent viral reservoir, a population of long-lived, replication-competent proviruses integrated into the host genome, primarily within resting memory CD4+ T cells. These proviruses are transcriptionally silent, shielded from immune surveillance and unaffected by ART, which only targets actively replicating virus. Upon ART interruption, reactivation of even a small number of these proviruses is sufficient to reignite systemic infection, underscoring the resilience of this latent pool.

Mechanistically, HIV latency is reinforced by a combination of factors: integration into transcriptionally repressive chromatin environments, restricted availability of transcriptional cofactors (eg, nuclear factor [NF]-κB, NF of activated T cells [NFATs]), absence or sequestration of the viral transactivator Tat, and epigenetic silencing via histone deacetylation and methylation. In addition, anatomic sanctuary sites, such as the gut-associated lymphoid tissue (GALT), lymph nodes, and central nervous system (CNS), provide anatomic and immunologic barriers that limit drug penetration and immune effector function.

Current cure strategies are broadly divided into 2 paradigms: (1) “shock and kill,” where latency-reversing agents (LRAs) are used to induce proviral expression followed by immune-mediated or cytopathic clearance, and (2) “block and lock,” aimed at enforcing durable silencing of the provirus to prevent reactivation. These strategies are challenged by the intrinsic heterogeneity of the reservoir, including proviral genetic integrity, integration site selection, cellular phenotype, and tissue localization.

The Latent Reservoir

The 2025 CROI showcased a range of innovative approaches to decipher and target the latent reservoir. From high-resolution lineage tracking in nonhuman primates to dissection of chromatin landscapes and latency regulatory circuits, studies presented at this year's meeting underscored the complexity of HIV persistence and the need for multidimensional intervention strategies.

Despite profound suppression of plasma viremia by ART, viral rebound occurs universally upon cessation of therapy. This rebound is thought to originate from rare cells harboring replication-competent proviruses, yet the specific anatomic origin, clonal composition, and dynamics of early recrudescent events remain poorly understood. Classical phylogenetic studies using unbarcoded virus have limited capacity to resolve the initial seeding events due to low sequence diversity and sampling limitations. Furthermore, previous analyses have largely been restricted to blood or accessible tissues, underestimating the potential contribution of deep lymphoid or mucosal compartments.

Keele and colleagues (Abstract 18) developed a barcoded version of the pathogenic SIV_mac239 clone, SIV_mac239m, engineered with a 34-nucleotide degenerate tag inserted into a noncoding region of the genome. This enables tracking of more than 10,000 unique viral lineages. Thirty rhesus macaques were infected and initiated on ART during early chronic infection to achieve suppression. During ART, and at defined timepoints following analytic treatment interruption (ATI), a broad set of tissues was sampled, including PBMCs, mesenteric and peripheral lymph nodes, spleen, gut lamina propria, and rectal mucosa. Barcode sequencing of viral RNA and DNA was performed using NGS, allowing high-resolution mapping of the temporal and spatial dynamics of viral rebound.

Analysis revealed that rebound viremia was seeded by a remarkably restricted number of barcoded lineages, consistent with oligoclonal reactivation. Importantly, these rebound-initiating barcodes were most frequently found in gut-associated tissues and mesenteric lymph nodes prior to ART cessation, with minimal representation in peripheral blood or distal lymphoid organs. Early rebound was highly localized — lineages initially detected in gut tissue expanded locally before disseminating systemically. Moreover, some rebound lineages were undetectable during ART despite intensive sampling, suggesting that latency reversal may occur from deep, cryptic reservoirs. Kinetic modeling of barcode frequencies suggested exponential expansion of a small number of founder clones within days of ART withdrawal. This indicates that early rebound originates in the GALT, a tissue compartment with a high burden of memory CD4+ T cells, elevated immune activation, and, perhaps, suboptimal ART penetration. The restricted number of reactivated lineages supports the concept of a reservoir “bottleneck,” which may offer a narrow window for targeted intervention post ART cessation. These findings also highlight the inadequacy of blood as a surrogate for reservoir monitoring and the need to develop minimally invasive techniques for sampling mucosal sites.

Ott and colleagues (Abstract 31) presented an account of the molecular events that govern HIV transcriptional latency as well as its potential for therapeutic disruption. Although ART drugs suppress active replication to below detection limits, they do little to address the silent, chromatin-bound proviruses embedded deep within the host genome. Ott's group has pursued the hypothesis that, under ART, HIV can occupy a range of transcriptional configurations, from deep, nuclease-protected silence to partial, abortive transcription. Ott described the reservoir as stratified by its integration landscape: proviruses inserted into transcriptionally active euchromatin may remain poised for reactivation, and those within heterochromatic or LADs are locked behind numerous layers of repression. It is the latter that resist all current LRAs, and the former that might yield under pharmacologic or immunologic pressure.

To dissect these configurations, Ott's team employed an integrative multiomic platform, analyzing HIV-infected primary CD4+ T cells from ART-suppressed individuals alongside in vitro models. They mapped chromatin accessibility using ATAC-seq, profiled histone modifications and factor occupancy through ChIP-seq, and tracked RNA output via transcriptomics. This revealed that transcriptionally silent proviruses were typically embedded in genomic loci marked by high levels of H3K9me3 and dense nucleosome occupancy at Nuc-0 and Nuc-1 around the LTR, inaccessible to RNA polymerase II and transcription factors. In contrast, pro-viruses with permissive marks — H3K27ac, H3K4me3 — were positioned in open chromatin and responded robustly to exogenous stimulation.

Ott also described “in-between” proviruses: those transcribing at low levels under ART, producing full-length or near-full-length viral RNAs, but failing to complete splicing or generate viral proteins. Ott presented evidence that these transcripts were often stalled at the elongation stage or subjected to incomplete processing. Notably, even in the absence of external stimulation, low-level expression from these proviruses might sustain chronic immune activation, an important consideration for non-AIDS comorbidities linked to persistent inflammation. In exploring the mechanics of latency reversal, Ott turned to bromodomain and extraterminal domain (BET) inhibitors like JQ1, which displace bromodomain-containing protein 4 from the HIV LTR and free positive transcription elongation factor b (P-TEFb) for recruitment by the viral Tat protein. She described how the synergy between Tat and BET inhibition dramatically improved the efficiency of transcriptional elongation, promoting not only unspliced genomic RNA but also fully processed, multiply spliced transcripts necessary for viral protein production. Ott made the case for targeting epigenetic repression and transcriptional elongation simultaneously, an approach that may distinguish future LRAs from their more limited predecessors.

In the search for an HIV cure, one of the most elusive questions is not merely how to extinguish the virus, but where to find it. Vandekerckhove (Abstract 19) outlined efforts in his group to identify cellular and tissue origins of HIV rebound. Vandekerckhove presented evidence that HIV persists in a wide array of cellular niches, and in anatomic compartments that current diagnostics are unable to track. His team leveraged fine-resolution phenotyping, tissue sampling, and integrated multiomics to portray the reservoir not as a uniform population but as a fragmented and compartmentalized entity, with some cells merely carrying intact proviral DNA and others actively transcribing or even translating viral genes, in ways that escape detection in peripheral blood. One of his key messages was that despite the blood's accessibility, it is a poor surrogate for the reservoir's true dynamics. Lymphoid tissues, especially GALT, mesenteric lymph nodes, and spleen, harbor large populations of memory CD4+ T cells, many of which exhibit a prosurvival phenotype (eg, Bcl-2^high), express latency-favoring transcriptional programs (eg, BTB domain And CNC homologue 2 [BACH2], forkhead box protein 01 [FOXO1]), and are located in cytokine-rich niches that promote survival and restrain activation. These cells may not merely house latent virus, they may actively support its long-term maintenance. Vandekerckhove underscored the role of myeloid cells, macrophages and microglia, in harboring persistent virus. Although controversial in the field, evidence is accumulating that these cells, particularly in the CNS and gut lamina propria, may be infected in vivo and contribute to viral rebound, especially under conditions of immune suppression or ART interruption. Unlike T cells, these cells are long-lived and largely resistant to cytopathic effects, making them ideal candidates for sustaining the virus during periods of host immune quiescence.

To support these assertions, his team employed tissue-resident cell isolation combined with qPCR for proviral DNA and viral outgrowth assays to test replication competence. They complemented this with single-cell RNA-seq to assess transcriptional activity and chromatin accessibility, and flow cytometry to define activation and differentiation states. Across these approaches, one consistent theme emerged: rebound-competent cells often reside in tissues that are poorly sampled, poorly penetrated by ART, and structurally resistant to immune cell infiltration.

Vandekerckhove concluded that although reservoir quantification has grown more sophisticated, reservoir targeting remains elusive. If Keele showed us how rebound begins and Ott explained why proviruses stay silent, then Vandekerckhove showed us where to look next: in the nodes, the gut, the brain, and the macrophage-rich niches that have long evaded scrutiny.

Kearney (Abstract 32) presented studies to assess the relationship between the proviral landscape and latency maintenance and reactivation. She began by revisiting a long-standing paradox in HIV cure research: why do some proviruses persist for decades in a transcriptionally silent state, but others remain transcriptionally active and potentially vulnerable to immune surveillance or pharmacologic clearance? The answer, she argued, lies in part in where HIV chooses to embed itself within the host genome — a process once thought to be random, but now known to be highly biased toward actively transcribed genes and open chromatin. Yet despite this general preference, the precise context of integration, including epigenetic environment, gene function, orientation, and proximity to regulatory elements, can make all the difference. Drawing from a broad dataset of integration sites derived from in vitro infected primary CD4+ T cells and ex vivo samples from ART-suppressed individuals, Kearney offered a layered picture of integration fate. Proviruses integrated into actively transcribed, intragenic regions, especially those within transcription units that support high levels of Pol II elongation, were more likely to be transcriptionally competent. These loci also conferred greater sensitivity to LRAs, suggesting a “permissive” integration profile that could be exploited therapeutically.

In contrast, proviruses inserted near transcriptional pause sites, enhancers, or repressive chromatin domains, often marked by high levels of H3K9me3 and DNA methylation, exhibited deep latency and were refractory to even aggressive LRA combinations. Interestingly, integration into repetitive elements or noncoding regions, once dismissed as transcriptionally inert, was found to harbor a large fraction of defective proviruses. Yet even among these, rare intact proviruses could be found that resisted reactivation, hinting at a chromatin-based silencing mechanism akin to position effect variegation. Her team also leveraged matched integration site and proviral sequencing data to track the clonality of infected cells over time. Clonally expanded cells, those harboring identical provirus-host junctions across numerous timepoints, were predominantly enriched in specific genomic “hotspots,” many of which were associated with genes involved in cell proliferation, immune regulation, or survival pathways, such as MKL2, BACH2, STAT5B, and ZNF genes. This raised the intriguing possibility that HIV integration not only passively reflects chromatin openness but may in some cases confer a proliferative advantage, contributing to clonal expansion and long-term persistence. Kearney also linked these integration patterns to differential responses to LRA stimuli. Using primary cell models of latency and ex vivo stimulation assays, her team showed that proviruses integrated into genic, euchromatic regions (eg, within introns of highly expressed genes) could be efficiently reactivated with combinations of BET inhibitors, histone deacetylase (HDAC) inhibitors, and T-cell receptor (TCR) mimetics. In contrast, deeply latent proviruses near repressive elements required very strong global activation or remained completely uninducible, even under maximal stimulation.

Kearney emphasized that understanding the integration site landscape is not merely of academic interest: it is crucial to the design of cure strategies. Whether one aims to “shock” the reservoir into visibility or “lock” it into permanent silence, knowing which proviruses are accessible, which are inducible, and which are off-limits is essential.

In the evolving story of HIV latency, most maps have been drawn from bulk assays, population averages, and static snapshots of viral transcription. Karn and colleagues (Abstract 136) offered a dynamic and almost cinematic glimpse into the spatial choreography of HIV proviral DNA as cells transition between active and latent states, shedding light on one of the most elusive dimensions of latency: its 3-dimensional nuclear positioning.

Karn's group utilized refined latency models using primary human CD4+ T cells driven into quiescence following effector activation. These models better mimic the physiologic milieu in which HIV latency is established and maintained. Central to his group's approach is an advanced imaging platform that combines CRISPR-based fluorescence in situ hybridization (Cas-FISH) with multiplex immunofluorescence and high-resolution confocal microscopy. This allows for simultaneous visualization of HIV proviral DNA, key host chromatin regulators, and nuclear substructures. By applying these tools, Karn's group was able to watch in exquisite detail how the spatial organization of the nucleus shapes the transcriptional fate of the integrated provirus.

Karn demonstrated that in actively infected cells, HIV proviral DNA localized within the intermediate euchromatic compartment (IEC), a region of the nucleus enriched in transcriptionally active domains and accessible to RNA polymerase II and chromatin remodelers. Upon transition into cellular quiescence, however, the provirus migrated to the perinucleolar compartment (PNC), a transcriptionally repressive subnuclear zone associated with nucleolar periphery and enriched for heterochromatin protein 1 (HP1), LADs, and histone marks such as H3K9me3.

Karn showed that treatment with raltegravir, an integrase strand transfer inhibitor, prevented this spatial rearrangement, implying that postintegration nuclear positioning is an active process influenced by integration machinery and chromatin remodeling complexes. The relocation of proviral DNA to the PNC was accompanied by profound chromatin condensation, increased nucleosome density at the LTR, and reduced accessibility to transcriptional regulators, effectively locking the provirus into latency.

Perhaps most revealing were the events surrounding reactivation. Upon stimulation via the TCR or pharmacologic LRAs, proviruses relocalized away from the PNC and toward regions of higher transcriptional activity. This repositioning was accompanied by a burst of nuclear reorganization, including the assembly of interchromatin granule clusters, membraneless nuclear foci enriched for transcriptional coactivators, splicing factors, and the P-TEFb complex. Proviral reactivation was temporally coupled with the accumulation of Tat, 7SK snRNA, and cyclin T1/CDK9, all recruited to these foci. Not all proviruses, however, participated equally in this process. Those integrated near LADs or buried in deeply repressive chromatin failed to reposition or initiate transcription, even in the presence of strong stimuli. This heterogeneity underscores what many in the field have long suspected: latency is not simply a function of proviral sequence or integration site, but of nuclear geography, a physical, spatial state that defines whether proviruses can respond to activation signals. These studies point to a need to reconfigure nuclear structure, or design therapies that selectively target the subnuclear niches where inducible proviruses reside.

Eshetu and colleagues (Abstract 137) presented technically innovative studies aimed at mapping, with single-cell precision, the physical location, transcriptional activity, and immunologic context of intact HIV-infected cells in lymphoid tissues. Their approach used spatial transcriptomics to aid in identification and localization of HIV-positive cells within tissue architecture. Lymph nodes, dynamic immunologic hubs where T cells, B cells, dendritic cells, and macrophages converge, have long been known to harbor large fractions of the latent reservoir. Yet until recently, technical limitations made it challenging to directly visualize which cells harbored intact HIV genomes, or how they were embedded within the immunologic microenvironment. Eshetu's group used the 10x Genomics Xenium platform, which combines spatially barcoded oligonucleotide probes with single-molecule fluorescence imaging and highresolution tissue segmentation. Their study focused on frozen lymph node sections from 4 participants in the “Last Gift” cohort: 2 viremic individuals with untreated HIV and 2 virally suppressed individuals on long-term ART, with matched lymph nodes from HIV-negative donors as controls. To distinguish intact vs defective proviruses, they designed 5 highly specific probe sets targeting: (1) the packaging signal (ψ), (2) the Rev response element (RRE), (3) unspliced gag-pol RNA, (4) viral splice junctions, and (5) a pan-HIV probe for total viral transcripts.

The simultaneous detection of ψ and RRE within the same single cell was used as a surrogate for intact, potentially replication-competent proviruses, based on previous validation with full-length sequencing. The resulting dataset comprised more than 790,000 individual cells profiled across 11 tissue sections. In viremic donors, they detected more than 1900 HIV RNA-positive cells per million total cells, including 17 cells per million that carried transcript signatures consistent with intact proviruses. The transcriptional profiles of HIV-positive cells revealed key distinctions. Cells with intact HIV RNA displayed elevated expression of genes involved in type I IFN signaling (IFN-stimulated gene 20 kDa protein [ISG20], IFN regulatory factor [IRF] 8), cytotoxicity (perforin [PRF] 1, granzyme [GZMK] K, natural killer cell G protein [NKG] 7), and T-cell memory and activation (selectin [SEL] L, transcription factor [TCF] 7, TNF receptor superfamily member [RSF]9). These cells also upregulated markers of immune exhaustion and survival, including thymocyte selection-associated HMG-box protein (TOX), programmed cell death protein (PDCD) PDCD1, and Bcl-2-related protein A1 (BCL2A1), suggesting that they were not only intact and transcriptionally active, but poised for persistence.

Eshetu's spatial data also revealed the immunologic climate surrounding these reservoir cells. In viremic individuals, HIV-negative bystander cells in proximity to infected cells also exhibited elevated expression of ISGs, including apolipoprotein B mRNA editing enzyme, catalytic subunit 3G (APOBEC3G) and IFN-induced transmembrane protein (IFITM) 3, along with heightened expression of cytotoxic and T helper cell 1 markers. This suggested that even in regions where infected cells were rare, there existed a tissue-level immune activation gradient, potentially contributing to the inflammation and immune dysregulation seen in chronic infection. For the first time, HIV researchers can move beyond bulk DNA quantification or indirect surrogate markers and begin to literally see the reservoir: where it resides, what it expresses, and how it coexists, or collides, with the host immune system. As Eshetu put it, “We're no longer just sequencing the virus. We're mapping it.”

Wei (Abstract 101) continued with the theme of the importance of tissue localization and cellular phenotype in defining the HIV reservoir. Wei's study demonstrated not just where the virus persists, but why certain cells are so exquisitely suited to maintain it. His focus was on a gut-resident population of CD4+ T cells sculpted by the transcription factor BACH2, whose survival, quiescence, and epigenetic configuration enable them to serve as stable sanctuaries for HIV. Wei's group focused on the question of why the gut mucosa, more than any other site, remains such a robust reservoir, even in individuals on long-term ART. Although previous studies have implicated immune activation, microbial translocation, and ART penetration issues, Wei's team hypothesized that cell-intrinsic programs within gut-resident memory T cells may be central to their capacity to harbor HIV. They paired bulk RNA-seq and scRNA-seq on CD4+ T cells isolated from gut biopsies of ART-suppressed individuals. These analyses revealed an enrichment of CCR6+ CD4+ T cells with a tissue-resident memory (TRM) phenotype, defined by high expression of CD69, CD103, and CD49a, along with low expression of sphingosine-1-phosphate receptor 1 (S1PR1) and Krüppel-like factor 2 (KLF2), consistent with tissue retention. These cells co-expressed IL-7R (CD127) and Bcl2, indicating a survival advantage, and were notably enriched for HIV DNA and RNA, suggesting that they were not only infected, but transcriptionally active.

At the center of their transcriptional identity was BACH2, a transcription factor known to regulate quiescence, antiinflammatory states, and memory cell differentiation. Using ATAC-seq and ChIP-seq, Wei's team showed that BACH2 bound to regulatory elements near key effector genes (eg, IFNG, TNF, GZMB) and suppressed their expression, effectively restraining cytotoxic and proinflammatory outputs. Simultaneously, BACH2 enhanced the expression of survival-associated genes (eg, IL7R, TCF7, TOX), promoting a homeostatic, nonexhausted phenotype that supports longevity and immune evasion. To test whether this phenotype was functionally linked to HIV persistence, the team performed in vitro infection assays with sorted gut TRMs and their circulating counterparts. The gut TRMs were significantly more permissive to HIV infection and sustained higher levels of proviral DNA and integrated HIV-1 copies over time. When exposed to LRAs, these cells showed attenuated reactivation responses, despite the presence of intact proviruses, consistent with a chromatin landscape primed for deep latency but not for cytolytic clearance. Wei related these findings to in vivo data. In ART-suppressed individuals, BACH2+ TRMs were spatially enriched in gut lamina propria and near epithelial crypts, areas with persistent inflammation and antigen exposure. These cells displayed low major histocompatibility complex class I (MHC-I) expression, low metabolic activity, and limited transcription of cytopathic viral proteins, allowing them to persist in a stealth mode that evades ART and immune detection.

From studying longitudinal gut biopsies in a small cohort of ART-treated individuals over year's, BACH2-expressing TRMs remained stable in frequency and phenotype, even as other gut T-cell populations declined or shifted. This long-lived stability supports the notion that TRMs are foundational to reservoir durability, not simply residual vestiges of early infection. As such, therapies that target the BACH2 axis, by disrupting its transcriptional network or reprogramming TRM epigenetics, could form the basis for novel reservoir-directed strategies.

Sun's presentation examined the evolution of the viral reservoir and its shaping by immune responses. Their longitudinal study reframed the reservoir as a dynamic, co-evolving population, shaped not only by intrinsic viral characteristics or drug pressure, but also by the gradual sculpting effects of host immune responses. Examining 20 years of ART, they showed that the virus and the immune system adapt, sometimes in lockstep, sometimes in conflict. Sun's study leveraged rare, longitudinal samples from 5 ART-suppressed individuals enrolled in prospective observational studies, all of whom began ART in chronic infection and maintained undetectable plasma viremia for 2 decades. Using paired peripheral blood samples collected at about 1, 10, and 20 years of suppression (designated T1, T2, and T3), the team profiled how the phenotypic, transcriptional, and genomic characteristics of reservoir-harboring cells changed over time, and how these changes might reflect evolving immune selection. To do this, the team employed a powerful combination of techniques. First, they used phenotypic proviral sequencing (PheP-seq) to identify and isolate memory CD4+ T cells harboring integrated HIV proviruses. This method integrates full-length proviral sequencing with surface marker phenotyping, allowing each proviral genome to be linked to its host cell's immunologic profile. In parallel, cellular indexing of transcriptomes and epitopes (CITE) sequencing and scATAC-seq were used to characterize the landscape of immune effector populations, revealing how cytotoxic, regulatory, and helper T cells evolved in relation to the reservoir.

From more than 1 million individual CD4+ memory T cells, the team identified 2433 HIV-infected cells, including 265 harboring genome-intact proviruses and 356 parts of clonally expanded lineages. Across timepoints, a clear trajectory emerged: although the overall frequency of HIV-infected cells declined slightly, the proportion of cells harboring intact proviruses in hetero-chromatin-rich integration sites increased dramatically, from just 2.5% at T1 to 25% at T3. These heterochromatin-associated integrations were strongly associated with a phenotypically quiescent, survival-biased profile: high expression of Bcl2, low activation markers, and increased expression of latency-associated transcription factors such as BACH2, TCF7, and KLF2. These reservoir cells also expressed lower levels of MHC-I and costimulatory molecules, suggesting increasing resistance to immune detection and clearance. This parallel evolution hinted at a coevolutionary dynamic: the reservoir, under long-term immune surveillance, appears to adapt by shifting toward harder-to-kill phenotypes: those with epigenetically repressed integration sites, immune-inert profiles, and durable survival programs. At the same time, the immune system narrows its focus, expanding cytotoxic clones capable of targeting antigen-expressing cells, but potentially missing deeply latent ones. These findings echoed a theme that had emerged across the conference: the idea of a “survivor bias” within the reservoir, a selective process in which only the most resistant cells genetically and immunologically persist over years of therapy. The study suggests that cure strategies effective against early reservoirs, such as TLR agonists or broadly neutralizing antibodies, may be less effective in long-term suppressors whose proviruses reside in immune-privileged, transcriptionally silenced compartments.

For decades, HIV reservoir research has focused on integration site, latency depth, and immune escape. Nicodemos (Abstract 133) brought attention to a deeper level: the chemical composition of the provirus itself. His team's findings revealed that a large fraction of replication-competent HIV proviruses contain deoxyuracil (dU), a noncanonical DNA base, and that this unusual biochemical feature may be a defining mark of infection established in resting CD4+ T cells. This insight adds an entirely new dimension to our understanding of reservoir formation and maintenance.

The premise of the study was simple: what happens when HIV infects cells that lack robust deoxynucleotide biosynthesis, specifically, resting CD4+ T cells? These cells, which form the backbone of the long-lived latent reservoir, are metabolically quiescent and maintain low intracellular levels of deoxythymidine triphosphate (dTTP), the canonical substrate used during reverse transcription to incorporate thymidine into proviral DNA. In such a context, the RT enzyme may substitute deoxyuracil triphosphate (dUTP) for dTTP, incorporating dU into the nascent proviral DNA.

Using a CCR5-tropic, GFP-expressing replication-competent HIV-1 reporter virus, Nicodemos and colleagues infected highly purified, nonactivated primary CD4+ T cells, carefully validated to be negative for activation (CD25, CD69), proliferation (Ki-67, phospho-H3.3), and metabolic markers. Remarkably, within 3 to 5 days, they observed the emergence of GFP+ cells that retained a resting phenotype, demonstrating that HIV can indeed infect and establish latency in the absence of overt T-cell activation. These GFP+ resting cells underwent detailed transcriptomic and biochemical analysis. RNA-seq revealed a transcriptional program indicative of cell-cycle-independent nucleotide salvage, and critically, a deficiency in dUTP-to-dTTP conversion enzymes, including dUTPase and thymidylate synthase. This biochemical profile suggested that the cells were permissive to the incorporation of deoxyuracil during reverse transcription. To confirm this, the team designed a qPCR-based assay capable of distinguishing proviruses containing dU vs dT at defined loci. Applying this method to in vitro infected cells and ex vivo samples from ART-treated individuals, they found that dU-containing proviruses were not only present but enriched in cell populations harboring replication-competent virus, as validated by quantitative viral outgrowth assays (QVOAs). These dU+ proviruses were detectable in distinct CD4+ T-cell subsets, with unique transcriptional and epigenetic profiles. Why does this matter? The incorporation of dU into DNA is generally considered mutagenic and potentially destabilizing, but Nicodemos proposed that it may actually facilitate long-term latency. Deoxyuracil is poorly recognized by host DNA repair machinery in resting T cells and may reduce detection by cytoplasmic DNA sensors like cyclic GMP-AMP synthase (cGAS), limiting innate immune activation. Moreover, it may affect the recruitment of chromatin modifiers to the LTR, influencing the transcriptional fate of the provirus. Importantly, the presence of dU was not synonymous with replication defectiveness. Nicodemos presented data from limiting dilution outgrowth cultures showing that dU+ proviruses could produce infectious virus upon stimulation, satisfying the definition of replication competence. This challenges the long-held assumption that chemically abnormal proviruses are necessarily defective and underscores the need to revise our molecular definitions of reservoir integrity.

The CNS, long known as a sanctuary for viral persistence, remains among the least accessible and most poorly understood compartments of the reservoir. Churchill (Abstract 44) presented an overview of the CNS reservoir, reframing it not only as a virologic challenge but as a neuropathologic dilemma, one with consequences that extend beyond viral latency to affect neurocognitive function, inflammation, and therapeutic safety. Churchill pointed out that the CNS is not immunologically silent, nor is it virologically inert. HIV can cross the blood-brain barrier (BBB) within days of systemic infection, seeding a long-lived population of infected cells. Although ART substantially suppresses systemic replication, HIV genomes, and, importantly, viral transcripts and proteins, persist in the CNS for years. Even in individuals with undetectable plasma viremia, cerebrospinal fluid (CSF) often contains cell-free HIV RNA, virally infected cells, and biomarkers of ongoing immune activation. But what exactly constitutes the CNS reservoir? Churchill laid out the cellular architecture in detail. The dominant reservoir cell type appears to be microglia, the brain's resident macrophages, which are long-lived, self-renewing, and capable of supporting HIV replication. Perivascular macrophages and, potentially, astrocytes also contribute, although the latter's role remains controversial due to conflicting evidence around productive infection. Resident and infiltrating CD4 T cells may also harbor HIV, particularly in inflamed regions where immune cell trafficking into the CNS is enhanced. However, these populations are sparse compared with the periphery, making quantification and phenotyping technically demanding.

Using a combination of postmortem brain tissue analysis, CSF cell sorting, and in situ hybridization, Churchill's group showed that even under suppressive ART, many of these CNS cells express HIV RNA and proteins, particularly Nef and Tat, which have been implicated in neurotoxicity and chronic immune activation. Most of the proviruses identified in CNS-resident cells were defective, but as in the periphery, a subset of intact, replication-competent proviruses persists — capable of rekindling infection upon ART withdrawal or immune perturbation.

One of the most pressing issues Churchill highlighted was the pharmacologic inaccessibility of the CNS. The BBB, although essential for protecting neural tissue, also restricts the penetration of many ART drugs and virtually all latency-reversing agents and gene therapy vectors. Although some LRAs, such as certain histone deacetylase inhibitors or TLR agonists, have shown CNS penetration in vitro or in animal models, their ability to reverse latency in situ remains unproven. Furthermore, the activation of latent virus within the CNS carries significant risk: even modest increases in viral transcription can provoke local inflammation, excitotoxicity, and damage to the delicate architecture of neural circuits. Churchill also explored the immunologic dynamics within the CNS. Microglia and astrocytes mount a local innate immune response to HIV RNA and proteins, producing interferons, cytokines (eg, IL-1β, TNF-α), and neurotoxic mediators (eg, glutamate, reactive oxygen species). These responses can persist even in the absence of productive infection, contributing to HIV-associated neurocognitive disorders (HAND), a spectrum of cognitive, motor, and affective disturbances observed in a substantial subset of ART-treated individuals. CSF analysis reveals elevated markers of neuroinflammation (eg, neopterin, NFL), often correlating with impaired cognitive function and brain imaging abnormalities. Churchill noted that most interventional studies, including those testing LRAs, immune checkpoint inhibitors, or cell-based therapies, do not monitor CNS involvement, nor do they assess potential neurotoxicity or viral rebound in the brain. This omission may leave patients vulnerable to CNS viral escape or unintended neuropathology, especially in settings where ART is interrupted, or immune pressure is modulated.

Despite these challenges, Churchill offered grounds for cautious optimism. Several LRAs, including bromodomain inhibitors and TLR7 agonists, have shown the ability to activate latent virus in ex vivo CNS models without inducing widespread inflammation. Some delivery systems, including lipid nanoparticles and adeno-associated virus (AAV) vectors, show promise for crossing the BBB and targeting CNS-resident cells. Furthermore, emerging technologies such as CSF cell-free DNA sequencing, neuroimaging-based reservoir tracking, and spatial transcriptomics of brain tissue may allow more precise monitoring of CNS reservoirs in future cure trials.

Even in the era of highly effective ART, where plasma HIV RNA is typically suppressed below the limits of detection, a subset of individuals continues to exhibit low-level persistent or intermittent viremia. This phenomenon, often referred to as nonsuppressible viremia (NSV), occurs in the absence of adherence issues, drug resistance, or overt virologic failure.

Box (Abstract 100) presented an exploration of this clinical enigma, uncovering a key role for defective but transcriptionally active proviruses bearing mutations in the 5’ leader (5’L) region of the HIV genome. The 5’L, located at the very start of the viral genome, contains essential regulatory elements required for proper RNA splicing, packaging, dimerization, and translation, including the major splice donor (MSD) site, the packaging signal (ψ), and the transcriptional start site. Mutations in this region can profoundly disrupt the fate of viral transcripts, potentially generating RNA that is abundant, stable, and released into plasma, but incapable of supporting replication or virion production. To investigate whether such defective clones could explain NSV, her team enrolled 30 ART-suppressed individuals with persistent detectable plasma HIV RNA (20-1500 copies/mL) in the absence of treatment interruptions or drug resistance. From these individuals, they isolated cell-free plasma RNA, purified it via high-speed centrifugation, and performed single genome sequencing (SGS) of 2 key regions: U5-gag (spanning the 5’ leader and Gag start codon) and protease-reverse transcriptase (p6-RT). Across 29 participants, the viral sequences were highly clonal, with 90% of sequences per individual being genetically identical, suggesting plasma virus derived from 1 or a few long-lived, transcriptionally active infected cell clones.

A large fraction of these sequences harbored 5’L mutations, especially within the MSD site, which is necessary for proper splicing of HIV transcripts. The team performed site-directed mutagenesis of the reference NL4-3 strain, engineering the specific MSD mutations observed in clinical isolates. These mutant viruses were then tested for infectivity, packaging, and replication competence. Although the mutant proviruses were able to produce RNA that was stable and released into the extracellular space, they were incapable of producing infectious virus, due to impaired splicing and virion assembly. Importantly, the lack of drug resistance mutations across all samples suggested that this persistent viremia was not driven by ART failure, but rather by chronically activated or transcriptionally permissive clones of infected cells that continued to produce defective but detectable HIV RNA. This RNA may not pose a direct risk of transmission or disease progression, but it complicates clinical management by triggering concerns about adherence, failure, or treatment resistance and it raises key questions about the true meaning of “undetectable” in clinical monitoring.

By correlating the frequency of 5’L-defective clones with residual viremia levels, Box and colleagues found that in individuals with persistent viremia over years, the same defective clone persisted longitudinally, suggesting that these cells are long lived, clonally expanded, and transcriptionally stable. These characteristics align them closely with other well-defined subsets of the HIV reservoir, but with a key difference: they are visible in plasma, blurring the line between “latent” and “active” infection. The team's findings help explain the clinical phenomenon of residual or intermittent low-level viremia without ART failure, offering a mechanistic reassurance to providers and patients. Second, they highlight a new dimension of the HIV reservoir — one that is not replication competent, but not silent either. This hybrid state of defective, transcriptionally active proviruses represents a source of immune activation, a barrier to true virologic silence, and a potential confounder in cure-directed interventions.

Most latently infected cells do not express enough viral antigen to be recognized by the immune system, and those that do often escape due to immune exhaustion, restricted MHC presentation, or lack of T-cell priming. But what if these limitations could be bypassed? Dorrell and colleagues (Abstract 104) examined whether T cells could be redirected with precision, and even against cells harboring trace amounts of viral protein. Her group assessed immune mobilizing monoclonal TCRs against virus (ImmTAV), a novel class of bispecific fusion proteins that combines high-affinity TCRs targeting conserved HIV epitopes with anti-CD3 single-chain variable fragments (scFvs) to redirect polyclonal T cells toward infected targets. Dorrell presented the first-in-human trial of ImmTAV GAGxCD3, a molecule engineered to recognize an HLA-A*02:01- restricted Gag epitope with picomolar affinity and recruit CD3+ T cells to mediate killing. Gag is abundantly expressed in early viral gene expression, even in cells not producing full virions, and is relatively conserved across clades. By using affinity-matured TCRs that exceed the sensitivity of endogenous CD8+ T cells by several orders of magnitude, the ImmTAV platform circumvents low antigen density, one of the primary bottlenecks in HIV reservoir clearance. Dorrell's trial, conducted in collaboration with Immunocore, enrolled 5 ART-suppressed individuals with HIV who were HLA-A*02:01 positive and had plasma HIV RNA under 50 copies/mL. Participants received a single intravenous dose of ImmTAV GAGxCD3, with close monitoring of pharmacokinetics, immune activation, and viral reservoir metrics.

No ART interruption was performed; the intervention was tested in the setting of fully suppressed viremia, where any impact on reservoir markers would be especially meaningful. The most immediate observation was safety: no serious adverse events were reported. Cytokine levels rose transiently within hours of infusion (eg, IFN-γ, IL-6, TNF-α), peaking at 24 hours and resolving by day 3. Mild flu-like symptoms were observed in 2 participants, consistent with transient immune activation. No significant changes in CD4+ or CD8+ T-cell counts were noted. Within 7 days of dosing, all 5 participants showed a modest but consistent increase in plasma HIV RNA (typically from <20 copies/mL to 50-200 copies/mL) suggesting that ImmTAV may have triggered low-level reactivation or clearance of HIV-expressing cells. By day 14, 3 participants exhibited a reduction of more than 1 log in cell-associated unspliced HIV RNA in CD4+ T cells, and 2 had a reduction in total HIV DNA, as measured by droplet digital PCR. These findings suggest that ImmTAV can engage infected cells in vivo and initiate clearance mechanisms.

Dorrell's group further investigated immune correlates of activity. Flow cytometry and single-cell transcriptomics revealed expansion of activated, cytotoxic CD8+ T cells post dosing, including upregulation of perforin, granzyme B, and NKG7. Notably, these cells were not HIV specific by TCR, confirming that ImmTAV redirection had mobilized non-HIV-specific T cells toward viral targets. Preliminary follow-up at 28 days post infusion suggested that reservoir metrics had stabilized, with no evidence of viral rebound or immune exhaustion. Importantly, Dorrell emphasized that this was a first-in-human pharmacodynamic signal, not yet a cure, but it was proof of principle that latently infected cells can be identified and targeted under ART suppression.