HIV Diversity Considerations in the Application of the Intact Proviral DNA Assay (IPDA)

The Intact Proviral DNA Assay (IPDA) was developed to address the critical need for a precise and scalable method for intact HIV reservoir quantification1. This duplexed droplet digital PCR (ddPCR) assay simultaneously targets the HIV Packaging Signal (Ψ) and the Rev Responsive Element (RRE) within Envelope (env) to distinguish genomically intact proviruses against a large background of defective ones2. The IPDA requires less time, resources, and biological material than the gold standard for replication-competent HIV reservoir measurement, the Quantitative Viral Outgrowth Assay (QVOA)3, and is being adopted in research and clinical studies4–7. In our cohort of HIV-1 subtype B-infected individuals from North America however, the IPDA yielded a 28% failure rate due to HIV polymorphism. We further demonstrate that within-host HIV diversity can lead the IPDA to underestimate intact HIV reservoir size, which could negatively impact clinical trial results interpretation. While the IPDA represents an important methodological advance, HIV diversity should be addressed before its widespread adoption.


Opening Paragraph (serves as abstract for submission) and Body 1 2
The Intact Proviral DNA Assay (IPDA) was developed to address the critical need for a 3 precise and scalable method for intact HIV reservoir quantification 1 . This duplexed droplet 4 digital PCR (ddPCR) assay simultaneously targets the HIV Packaging Signal (Ψ) and the Rev 5 Responsive Element (RRE) within Envelope (env) to distinguish genomically intact proviruses 6 against a large background of defective ones 2 . The IPDA requires less time, resources, and 7 biological material than the gold standard for replication-competent HIV reservoir measurement, 8 the Quantitative Viral Outgrowth Assay (QVOA) 3 , and is being adopted in research and clinical 9 studies 4-7 . In our cohort of HIV-1 subtype B-infected individuals from North America however, 10 the IPDA yielded a 28% failure rate due to HIV polymorphism. We further demonstrate that 11 within-host HIV diversity can lead the IPDA to underestimate intact HIV reservoir size, which 12 could negatively impact clinical trial results interpretation. While the IPDA represents an 13 important methodological advance, HIV diversity should be addressed before its widespread 14

adoption. 15
We applied the IPDA to 46 HIV subtype B-infected, virally-suppressed individuals from 16 North America, yielding a median of 29 (interquartile range [IQR] 0-93) intact 17 proviruses/million CD4+ T-cells (Extended Data Figure 1). Of note, the IPDA did not detect any 18 intact (i.e. Ψ and env double-positive) proviruses in 17 participants (37%). In four of these 19 individuals, both Ψ-and env-single-positive proviruses were detected, suggesting a true negative 20 result (Extended Data Figure 3) In the remaining 13 individuals however, the IPDA did not 21 detect Ψ-and/or env-single-positive proviruses above background levels despite recovery of 22 replication competent HIV in 11/11 cases where QVOA was performed, suggesting that the 23 assay failed to detect autologous proviruses. Specifically, in eight of these participants only Ψ-24 will occur, and not infrequently. The first step towards mitigation is awareness: samples that 1 yield no Ψ or env single-positive proviruses should be flagged as 'unreportable' until HIV 2 polymorphism has been addressed (e.g. using autologous primers/probes 1,5 ). Analysis of HIV 3 subtype B sequences from unique individuals in the Los Alamos HIV database revealed that 23% 4 of 9,360 env sequences harbored at least one env probe mismatch (which is similar to our env 5 detection failure rate of 9/46 or 20%), while 50% of 1,489 sequences harbored at least one Ψ 6 probe mismatch (which is substantially greater than our Ψ detection failure rate of 5/46 or 11%). 7 This suggests that the env reaction may be particularly sensitive to polymorphism. 8 Though laborious to correct, complete detection failures due to HIV polymorphism are 9 nevertheless easy to flag. In contrast, within-host HIV diversity (see Figure 1C for an example) 10 could lead the IPDA to underestimate intact reservoir size if the within-host variants were 11 differentially detectable by the assay. Such partial detection failures would not be easy to 12 identify. Moreover, if IPDA-detectable and non-detectable reservoir subpopulations were 13 differentially susceptible to HIV cure interventions, this could lead to erroneous conclusions 14 regarding intervention efficacy. 15 We illustrate this using HIV-specific broadly neutralizing antibodies (bNAbs). bNAbs 16 can facilitate elimination of HIV-infected cells 9 in part by targeting them for antibody-dependent 17 cellular cytotoxicity (ADCC) 9,10 , and are being evaluated in clinical trials, some of which use the 18 IPDA as a readout (e.g: ACTG A5386 6 ). Participant 91C33 from a published trial 11 provides a 19 hypothetical example. This individual did not respond to (off-ART) infusions of the bNAbs 20 3BNC117 and 10-1074 because they harbored a plasma HIV subpopulation that was resistant to 21 both bNAbs 11 (Figure 2A). Of note, this plasma HIV subpopulation also harbored a mismatch to 22 the IPDA env probe. We confirmed that the published IPDA could not detect templates 23 harboring this mismatch, while those representing the bNAb-sensitive strains were detected 1 readily ( Figure 2A; Extended Data Figure 5). If a person harboring such diversity in their 2 reservoir were to be successfully treated with one or both bNAbs, the IPDA could over-estimate 3 the intervention's effect ( Figure 2C). 4 ARV-suppressed participant OM5346 provides another example. Pre-ART HIV drug 5 resistance genotyping identified subtype B infection, however single-genome sequencing of pre-6 ART plasma and proviruses sampled during long-term ART revealed co-infection with a non-B 7 strain (Extended Data Figure 6) that harbored a G-to-A mismatch at position 13 (G13A) of the 8 env probe but no other critical mismatches to the IPDA primers. While the IPDA is only 9 designed for subtype B, G13A is the most frequent env probe polymorphism [~5%] in subtype B 10 (and was observed in 2/10 study participants for whom env was sequenced). Moreover, the 11 original report indicated that the IPDA could detect G13A, at least when present on a plasmid 12 template 1 . Using QVOA, we isolated replication-competent subtype B ("virus 3") and co-13 infecting ("virus 4") viruses from OM5346's reservoir, confirming both of these as true intended 14 IPDA targets ( Figure 2B). We further observed that these viruses were differentially susceptible 15 to 3BNC117-and 10-1074-mediated ADCC: while virus 3-infected cells could be eliminated by 16 10-1074-, but not 3BNC117-, mediated ADCC (25% and 0% relative reduction, respectively), 17 the opposite was true for virus 4 (0% and 32% relative reduction, respectively) ( Figure 2B Figure 8B). Importantly, a secondary 23 env primer/probe set (see below and methods) readily detected both viruses at all concentrations, 1 regardless of template type. If this individual were to be successfully treated with 10-1074 the 2 IPDA would overestimate the intervention's effect on the reservoir, whereas if the individual 3 were to be successfully treated with 3BNC117, the IPDA would erroneously conclude that the 4 intervention had no effect ( Figure 2C). Our findings, taken together with Gaebler et al's 5 observation that 4 of 9 (44%) studied reservoirs were heterogeneous in an IPDA probe region 8 , 6 suggests that the impact of within-host HIV diversity on IPDA accuracy may be non-negligible. 7 During our investigation we identified another source of potential error, resulting from 8 variable spillover of fluorescence from the Ψ channel into the env channel, which has 9 implications for the appropriate placement of thresholds defining negative and positive 10 populations. Importantly, the extent of spillover is HIV sequence-specific. We demonstrate this 11 by applying the IPDA to synthetic templates encoding the Ψ regions of OM5346 viruses 3 and 4, 12 without a corresponding env template present (Extended Data Figure 9). Virus 3, which harbors 13 mismatches to the Ψ probe, produced modest yet clearly discernible Ψ signal that did not spill 14 over into the env channel. By contrast, virus 4 yielded high-amplitude Ψ signal that spilled over 15 noticeably. Thus, drawing a tight threshold based on a template-negative (or virus 3-containing) 16 sample and applying this threshold to virus 4 creates env signal when no env template was 17 present. This Ψ channel spillover for virus 4 thus likely contributed to the small amount of 'env 18  The IPDA offers major scalability advantages over existing molecular 8 or culture-based 22 approaches 3 , and Gaebler et al. confirmed that, of any two probes, those of the IPDA offered the 23 greatest selectivity for intact proviruses 8 . Any molecular assay targeting a genetically variable 1 pathogen however must address polymorphism. While the requirement to discriminate proviral 2 defects in the Ψ region limits the placement and sequence of this probe 2 , we developed a 3 secondary primer/probe set in the intact-discriminating RRE region 1 , approximately 50 bases 4 downstream of the original location, as a first step towards addressing diversity. This 5 primer/probe set rescued detection of env-positive proviruses in 9/9 participants with IPDA env 6 detection failure (Extended Data Figure 10A). When we applied it to 36 participants for whom 7 the IPDA detected env-positive proviruses (excluding OM5346 who harbored within-host 8 diversity), it failed to detect env-positive proviruses in 3 (8%) individuals, indicating that it is not 9 a universal solution. For the remaining 33 (92%) however it yielded measurements that were not 10 significantly different from the IPDA (Extended Data Figure 10B). The secondary primer/probe 11 set could therefore be used to identify instances of detection failure, though users must note that 12 it cannot discriminate hypermutated sequences. were isolated by density gradient separation and cryopreserved (-150°C, 90% Fetal Bovine 7 Serum+ 10% DMSO). Participants recruited at other sites provided a leukapheresis sample from 8 which PBMC were isolated and cryopreserved as above. All participants were on long-term, 9 virally suppressive combination antiretroviral therapy (cART) at time of sampling, with the 10 exception of one elite controller. Participants' duration of untreated HIV infection ranged from 1 11 month to >10 years, though the exact duration was unknown for most participants. Ethical 12 approval to conduct this study was obtained from the Institutional Review Boards of Simon 13 Fraser University, Providence Health Care/University of British Columbia, Weill Cornell 14 Medicine, and the George Washington University. All participants provided written informed 15 consent. 16

Quantitative Viral Outgrowth Assay (QVOA) 17
For participants for whom sufficient biological material was available, the Quantitative CCTTGGGTTTCTGGGA-MGBNFQ). Droplets were prepared using either the Automated or 5 QX200 Droplet Generator (BioRad) and cycled at 95°C for 10 minutes; 45 cycles of (94°C for 6 30 seconds, 59°C for 1 minute) and 98°C for 10 minutes, as previously described 1 . Droplets were 7 analyzed on a QX200 Droplet Reader (BioRad) using QuantaSoft software (BioRad, version For participant OM5346, HIV RNA was extracted from QVOA outgrowth viruses and 23 pre-ART plasma, after which gp160 (QVOA) and gp120/Pol (plasma) were amplified from 24 endpoint-diluted templates by nested RT-PCR using HIV-specific primers and high fidelity 1 enzymes. For participants with suspected IPDA detection failure, and for whom HIV sequences 2 were not already available, the IPDA Ψ and/or env amplicon regions were bulk-amplified from 3 extracted proviral DNA using HIV-specific primers. Amplicons were sequenced using either 4 Sanger (3730xl, Applied BioSystems) or Next Generation (Illumina MiSeq) technologies. Sanger 5 chromatograms were analyzed using Sequencher (version 5.0.1, Gene Codes), while Illumina 6 MiSeq reads were de novo assembled as described above. 7

Antibody-dependent cellular cytotoxicity (ADCC) assays 8
Total CD4+ T-cells were isolated, as described above, from HIV-negative donors and CCCCAGACTGTGAGTTGCA (HXB2 nucleotides 7,939-7,921, 88% conservation). Reaction 2 composition and cycling conditions were same as those used for the IPDA as described above. 3 Accurate quantification using the Secondary env primer/probe set was verified using DNA 4 extracted from the J-Lat 9.2 cell line (obtained from the NIH AIDS Reagent Program, Division 5 of AIDS, NIAID, NIH, contributed by Dr. Eric Verdin) 21 (Extended Data Figure 11). 6

Statistical Analyses 7
All statistical analyses were performed using GraphPad Prism (version 8). 8

Data and Sequence Availability 9 10
All sequences collected for the present study have been deposited in GenBank (Accession 11 Numbers pending). Sequences for participant 91C33 were previously deposited under Accession 12 Numbers: MH632930-MH632955 11 . All other data are available from the corresponding authors 13 upon request, in compliance with institutional and REB requirements.   H  S  H   ------------------------------   (right) An intervention's impact on the reservoir could be underestimated when a subset of within-host HIV sequences are undetectable by IPDA and sensitive to treatment (e.g. OM5346 virus 4 during treatment with 3BNC117). In the worst-case scenario shown here, one could erroneously conclude that the treatment had no effect.