A daisy chain of inferences: the role of single-cell and single-genome proviral sequencing in characterizing HIV-1 reservoirs

Abstract

Purpose of review

Understanding and targeting the HIV reservoir requires navigating a hierarchy of inferential assays. Quantitative viral outgrowth assays, FLIP-seq, and intact proviral DNA assays (IPDA) – though methodologically distinct - are all fundamentally single-cell technologies. Each relies on limiting dilution to isolate and interrogate individual proviral genomes derived from single infected cells, offering high-resolution proxies for the outcome of greatest interest: replication competence and the risk of viral rebound. Rather than providing direct measurements, these assays infer one another in a nested framework. This review highlights the importance of critically interpreting assay outputs within this chain of inference to guide cure-directed strategies and reservoir quantification.

Recent findings

Recent studies emphasize the complexity and limitations of current assays measuring HIV-1 reservoirs. Key themes include reliance on bioinformatics definitions of genome-intactness to infer replication competence, alongside significant limitations due to viral diversity, PCR amplification length biases, and definitional inconsistencies. Modified assays like subtype-specific IPDA aim to address these issues.

Summary

Standardized, subtype-specific single-cell methodologies are crucial for accurate HIV reservoir characterization. Future research should integrate large-scale sequencing with replication competence validation, and should refine bioinformatics approaches to enhance predictive accuracy. Enhanced assay precision is essential to inform effective HIV cure strategies.

INTRODUCTION

The search for a cure for HIV hinges on measuring and characterizing the proviral reservoir – the pool of infected cells that persists despite effective antiretroviral therapy. These reservoirs are established early during infection [1–3], and serve as the source of virologic rebound in the absence of treatment. Characterizing viral reservoirs is critical not only for understanding viral persistence but also for assessing the efficacy of cure strategies, including latency-reversing agents, immune therapies, and gene editing technologies. Multiple single-infected-cell/single-genome laboratory techniques have been developed to either measure and/or characterize viral reservoirs. They include but are not limited to quantitative viral outgrowth assays (QVOA) [4], near-full-length intact proviral sequencing (FLIP-seq and similar assays) [5–7], and the intact proviral DNA assay (IPDA) [8]. Although differing in methodology, QVOA, FLIP-seq, and IPDA are fundamentally single-cell approaches in that they rely on limiting dilution to isolate and interrogate individual proviruses – each presumably derived from a single infected cell – thus achieving single-cell resolution. In the broader field, the term “single-cell technologies” is often associated with high-throughput transcriptomic platforms such as 10× Genomics paired with Illumina short-read sequencing. However, such methods are currently poorly suited for the study of HIV proviruses. First, they lack the sequencing length needed to identify internal deletions or defects in proviral genomes. Second, given the extremely low frequency of infected cells ex vivo – on the order of 1 to 1000 per million CD4+ T cells [5,9–16] – profiling infected cells without a reliable universal surface marker remains infeasible, making it impractical to apply droplet-based single-cell platforms to larger scale HIV proviral characterization research. Instead, limiting-dilution-based methods remain the standard for single-cell interrogation of HIV-infected cells, which will be the focus of this review. Each of QVOA, FLIP-seq and IPDA has unique strengths and weaknesses and should be applied thoughtfully in HIV persistence research. Importantly, all of these techniques are predictors: IPDA is a technique that predicts proviral genome-intactness in FLIP-seq data, while FLIP-seq proviral genome-intactness predicts proviral replication competence, and finally QVOA reveals proviruses that may reactivate and lead to subsequent viral rebound in the absence of therapy. Sensitivity/specificity of inferences varies depending on multiple factors such as assay design, genome-intactness definition, HIV-1 subtypes and intrahost viral diversity. This review article intends to provide an overview of these technologies in the context of their target of inference, with an emphasis on the role of proviral sequencing as a link across the various techniques. 

Box 1.

no caption available

INDUCIBLE AND REPLICATION-COMPETENT RESERVOIRS: QUANTITATIVE VIRAL OUTGROWTH ASSAYS

QVOA interrogates single infected cells to quantify inducible HIV reservoirs. It is considered the gold standard of HIV reservoir quantification. In this assay, CD4⁺ T cells from HIV-infected individuals on suppressive antiretroviral therapy are limiting diluted and activated ex vivo – typically using mitogens and irradiated feeder cells – to induce the reactivation of latent proviruses. Any replication-competent virus that is produced is then amplified by co-culture with uninfected CD4⁺ cells and detected using methods such as p24 ELISA or PCR-based approaches [4]. At limiting dilution, each p24 or HIV-PCR-positive well corresponds to a single infected cell. However, not all proviruses are reactivatable after a single round of stimulation using a given antigen [17]. Therefore, depending on the assay setup, this technique could be associated with underestimation of reservoir sizes of an unknown magnitude. This magnitude is likely mitogen-dependent. It may also vary across individuals depending on viral integration sites, host cell status, and epigenetic profiles. In other words, inducible viruses derived from QVOA likely only partially predict the probable source of rebound virus in case of treatment interruption. Viral genomes obtained from limiting-diluted and positive QVOA wells can be individually sequenced [5,17]. These QVOA-derived proviral genomes are replication competent, and therefore by definition are genome-intact.

GENOME-INTACT PROVIRUSES VIA SEQUENCING: FLIP-SEQ AND RELATED SEQUENCING ASSAYS

Proviral genomes can also be directly sequenced from ex vivo samples to interrogate genome-intactness. FLIP-seq [5,6] and similar assays [6,7,18] are high-resolution single-genome sequencing methods developed to characterize HIV-1 reservoirs by sequencing near full-length individual proviral genomes. Like QVOA, FLIP-seq relies on limiting dilution to isolate and analyze single proviruses derived from individual infected cells, enabling single-cell resolution of the latent reservoir. Proviral sequencing has gained popularity since the canonical study by Ho et al. [19] in which the study team revealed that over 90% of HIV-1 proviral DNA genomes in individuals under long-term suppressive therapy are highly mutated and/or truncated, and do not lead to further rounds of infection. In other words, typically less than 10% of the intrahost proviral pool are “genome-intact” and may contribute to virologic rebound in case of treatment interruption. FLIP-seq and similar technologies combines limiting-dilution with nested long-range PCR amplification to recover single proviruses from genomic DNA isolated from infected CD4⁺ T cells [20]. These proviruses are then sequenced using next-generation sequencing and assembled to determine proviral genome integrity. Each proviral genome obtained from FLIP-seq and its related assays are classified into “intact” or various classes of defectiveness including those that contains, for example, large deleterious truncations, signatures of APOBEC-3G/F G-to-A mutations, premature stop codons, and/or defects in the 5’ packaging signal (psi) region [5,20,21]. However, the definition of genome-intactness varies among research groups depending on the bioinformatics algorithms used (Table 1) [5,6,18–22]. Researchers should be actively aware of how genome-intactness is defined when interpreting data. Importantly, when combined with multiple displacement amplification or similar approaches, limiting-dilution sequencing strategies such as FLIP-seq can also be extended to interrogate viral integration sites at the single-cell level [10,23], enabling paired analysis of proviral genomes and chromosomal context.

Table 1.

Examples of definitions for proviral genome intactness used by various research groups

Citation	Method/tool	Detailed criteria for defining genome-intactness
Ho et al. [33]	Experimental validation	Genome-intact proviruses were rigorously defined through both sequencing and direct functional validation: (1) Full-length open reading frames (ORFs) for all viral genes; (2) intact cis-regulatory elements including major splice donor (MSD) site, packaging signal (psi), and Rev-response element (RRE); and (3) experimentally verified capability to produce replication-competent virus via near-full-genome cloning.
Lee et al. [34–36]	FLIP-seq and HIVSeqinR (Bioinformatics algorithm)	Genome-intact proviruses must lack: (1) large internal deletions; (2) internal inversions; (3) APOBEC3G/F-induced hypermutations; (4) premature stop codons in essential viral genes (gag, pol, env); and (5) deletions ≥15 base pairs (bp) in the 5’ packaging signal (psi) region compared to reference strain NL4-3. Validated using QVOA-derived sequences.
Hiener et al. [37]	FLIPS	Genome-intact proviruses must satisfy these bioinformatics criteria: (1) Absence of inversion sequences; (2) no large internal deletions >100 bp; (3) absence of APOBEC3G/F-induced hypermutations causing premature stop codons; (4) no frameshift mutations within any essential viral ORFs; (5) intact 5’ packaging signal (psi) region; and (6) intact MSD site.
Wright et al. [38]	HIVIntact (Bioinformatics algorithm)	Genome-intact proviruses must satisfy these bioinformatics criteria: (1) All essential ORFs (gag, pro-pol, env) must be complete without premature stop codons or frameshift mutations; (2) minimal internal deletions permitted (≤30 bp in gag/pol, ≤100 bp in env); (3) packaging signal (psi) must be largely intact (allowing ≤10 bp deletions); (4) major splice donor (MSD) site must be fully conserved without mutations; and (5) Rev-response element (RRE) must be largely intact.
Barton et al. [39]	HIVε-seq	Genome-intact proviruses must satisfy these bioinformatics criteria: (1) Complete absence of large deletions; (2) no internal genomic inversions; (3) no premature stop codons in any viral genes; (4) no frameshift mutations affecting viral protein production; (4) no mutations in MSD; and (5) Deletions of >10 bp in psi or 20 bp in the RRE were considered to be defective.

Although all QVOA-derived replication competent proviruses are by definition genome-intact, not all bioinformatically-inferred genome-intact proviruses are proven to be replication competent. Furthermore, replication competence is also a spectrum – proviruses can have a range of replication fitness. Therefore, the word “genome-intactness” is at best an inference of replication competence defined by the lack of known deleterious mutations and/or insertions/deletions. Of note, the type of deleterious mutations in the 5’ packaging signal (psi) region of the proviral genome that would inhibit viral replication is in particular poorly understood in relation to replication competence, especially outside the context of subtype B HIV-1 [20,24^▪]. During the development of FLIP-seq and its associated bioinformatics pipeline HIVSeqinR [5,21], we validated our bioinformatic definition of “genome-intactness” to maximize its predictability for replication competence by ensuring that 14 donor-matched QVOA-derived subtype B HIV-1 sequences in the study were bioinformatically classified as intact. However, in our recent study on Ugandan proviruses using FLIP-seq, we found that 3 out of 18 QVOA-derived replication-competent subtype A1 and D viral genomes were misclassified as having 5’ psi defect if definitions for subtype B HIV-1 were to be used [24^▪]. The three genomes did not share a consistent deletion pattern in the 5’ psi region, and did not belong to the same subtype: one was A1, one was D, and one was a recombinant A1/D virus. These observations and our bioinformatics validation is limited by the extremely small number of QVOA-derived near-full-length viral genomes in our studies, but nonetheless highlight the need to further refine our definition of genome-intactness in the context of viral subtypes to infer replication competence. So far, no large-scale studies have systematically compared viral genotypes of replication-competent and defective proviruses, likely due to associated costs. To improve the predictiveness of bioinformatically-defined “genome-intactness,” future studies will need to collect thousands of viral DNA sequences derived from both QVOA-positive and -negative wells and/or from plasma circulating viruses, followed by machine-learning approaches to train bioinformatics algorithms on the association between viral genotype and replication competence.

Despite imperfect bioinformatics inferences, proviral sequencing remains a powerful tool for the study of HIV persistence. High-resolution viral sequence data provides a wealth of additional information on the persistence reservoir. Proviral sequences can be used to study the phylogenetic relationships between archived genomes and can be used to monitor the longitudinal genotypic changes in the proviral landscape [25–27]. Other information such as archived drug resistance, archived immune escape mutations, and viral tropism can also be inferred from such datasets. Relative to QVOA, proviral sequencing also requires fewer input cell materials and does not require prolonged viral culture in BSL2+ facilities. However, since these sequencing techniques are PCR-dependent, and PCR amplification efficiency is higher in shorter relative to longer templates, these long-range PCR-based methods has been shown to bias towards capturing shorter truncated viral DNA genomes resulting in underestimation of the levels of genome-intact viruses [20,28]. DNA shearing during sample processing can also result in underestimation. Using 8E5/LAV cells, we compared the sensitivity of FLIP-seq against a short-range quantification method (single-target LTRgag Droplet Digital PCR), and have shown that FLIP-seq's capture rate of intact genome is approximately 20% [5]. Due to these considerations, proviral sequencing is currently not considered an accurate method for the quantification of genome-intact proviruses. Short-range PCR approaches such as IPDA would remove PCR amplification bias due to template lengths. For further discussion on the biochemical and methodological considerations underlying FLIP-seq, including amplification bias and assay design constraints, please refer to our commentary in Viruses[20].

GENOME-INTACT PROVIRUSES VIA SHORT-RANGE EMULSION PCR: INTACT PROVIRAL DNA ASSAYS AND RELATED ASSAYS

IPDA is a droplet digital PCR-based single-genome detection method developed to estimate the number of intact and defective HIV-1 proviruses in infected individuals [8]. IPDA achieves single-cell resolution by encapsulating individual proviral genomes into separate emulsion droplets, allowing parallel interrogation of single proviral templates. It is scalable and high-throughput. Due to its short target fragment length (approximately 100 bp) relative to near-full-length proviral sequencing (up to 10 000 bp), this method circumvents PCR amplification template length bias and has been shown to more accurately measure the copy number of genome-intact proviruses in clinical samples [28]. The assay uses two pairs of probes targeting relatively conserved regions of the HIV-1 genome – specifically the packaging signal (psi) and the rev response element (RRE) within the env gene. These two regions are strategically chosen and could jointly identify over 90% of truncated proviruses as defective in the original study [8]. This value was derived from an analysis of a near-full-length subtype B HIV-1 proviral genome dataset (FLIP-seq-type data), consisting of a training set of 431 HIV-1 DNA sequences obtained from 28 adults living with HIV and a validation set of an additional 474 sequences. A competitive hypermutation probe eliminates signals from genomes that are hypermutated due to APOBEC3G/F activity. A provirus is classified as “intact” if it tests positive for both probes, while signals from only one or neither region indicate deletions or hypermutation.

In other words, IPDA was designed to predict proviral genome-intactness as defined by proviral near-full-genome sequencing data. In its original design, IPDA was able to successfully rule out over 90% of HIV-1 DNA genomes that are defective in silico[8]. Due to its high-throughput nature and low-cost relative to QVOA and proviral sequencing, this assay has been widely adopted in clinical trials that aim to reduce viral reservoir sizes. However, recent studies, especially in studies on nonsubtype B HIV-1, has shown viral sequence diversity results in IPDA assay failures. In the eCLEAR trial in which approximately 50% of study donors had nonsubtype B HIV-1, a modified version of the Intact Proviral DNA Assay (IPDA) was utilized to quantify the HIV-1 reservoir in participants [29]. Among the 23 individuals with HIV-1 subtype B in eCLEAR, IPDA detection failure occurred in six donors, accounting for approximately 26% of this subgroup. These failures were attributed to viral polymorphisms in the IPDA probe and/or primer regions [29,30]. For each of the other study donors living with nonsubtype B HIV-1 and for the six donors who failed, donor-specific viral sequencing of the IPDA primer/probe binding regions was performed, which allowed for donor-specific IPDA primer/probe modifications. In a separate study, our group analyzed a dataset of 603 FLIP-seq–derived, near-full-length proviral genome sequences from subtypes A1 and D, obtained from a Ugandan cohort, to develop a modified IPDA assay called IPDA-A1D, taking account of subtype A1 and D polymorphisms using degenerated primer/probes [24^▪]. IPDA-A1D is tailored to capture specific polymorphisms of subtype A1 and D HIV-1 circulating in that specific geographical region. This modified assay successfully rescued samples that failed to produce signals by the original IPDA. Other groups have developed alternative versions of IPDA to address viral polymorphism such as Cross-Subtype IPDA (CS-IPDA) [31], which targets subtypes A, B, C, D and CRF01_AE, while the IPDA-BC [32^▪] is designed specifically for subtype B and C HIV-1.

It is important to note that an IPDA failure, as mentioned above, is defined as the absence of signal in one or both of the psi and env targets despite evidence of the presence of proviruses using alternative approaches such as sequencing and/or QVOA. However, there could be cases in which IPDA signals are detected, but at reduced levels both in terms of signal intensity and the absolute positive signal counts. This is because IPDA detection relies on the strength of the fluorescent signal emitted by the probes: A greater number of mismatches between the primer/probe and the proviral template translates to less efficient PCR amplification and/or weaker fluorescent signals, resulting in a reduction or abscence of positive signals. In other words, detection may occur (i.e. IPDA did not fail), but the sensitivity of the assay against bioinformatically-defined genome-intactness may vary depending on both the count and location of nucleotide mismatches between the primers/probes and the template DNA. Consequently, the extent of underestimation of intact reservoir sizes may differ between individuals and/or samples. In light of these considerations, IPDA is perhaps best suited for intrahost longitudinal or pre/postintervention studies aimed at measuring changes in intact reservoir size, as intrahost viral diversity can be presumed to remain relatively stable over time.

CONCLUSION. THE ULTIMATE HOLY GRAIL: SOURCE OF REBOUND VIRUS

In summary, QVOA, proviral sequencing, and IPDA form a continuum of assays linked by a hierarchy of inference. Each method offers valuable insights into different facets of the HIV reservoir, yet none provides a complete picture on its own. Despite their methodological differences, all three are inherently single-cell approaches – using limiting dilution or droplet partitioning to isolate and interrogate individual proviral genomes – making them uniquely suited to resolve the reservoir at the level of single infected cells. In the context of HIV cure research, the ultimate goal is to eliminate or permanently silence all proviruses capable of reactivation and causing subsequent virology rebound following treatment interruption. As such, it is critical for researchers to apply these tools with a clear understanding of their respective strengths, limitations, and predictive capacities. When used thoughtfully and in combination, these assays serve as complementary strategies to approach the central challenge in HIV persistence: identifying and eliminating the true sources of rebound virus.

Acknowledgements

None.

Financial support and sponsorship

This work was supported in part by NIH grants R21AI150398 (GQL), R01AI162221 (GQL), and UM1AI164565 (RBJ).

Conflicts of interest

There are no conflicts of interest.

REFERENCES AND RECOMMENDED READING

Papers of particular interest, published within the annual period of review, have been highlighted as:

• ▪ of special interest

• ▪▪ of outstanding interest