The HIV reservoir is constituted of cells that contain integrated HIV DNA (proviruses) capable of producing new virions, and is the impediment to HIV cure. The majority of replication-competent HIV proviruses reside in a population of resting CD4+ T cells that can be either central memory, transitional memory, effector memory, or other subtypes . In persons on antiretroviral therapy (ART), the reservoir can be maintained by several mechanisms; long-lived cells contribute to the reservoir by persisting for years whereas cells with shorter life spans can either die or proliferate leading to contraction or expansion of the reservoir, respectively . The reservoir could also be maintained by new infection of uninfected target cells with rounds of complete viral replication .
The size of the reservoir varies between individuals [4,5] and is difficult to quantify as all currently available methods of measurement are imperfect. Quantitative viral outgrowth (QVOA) from highly purified CD4+ T cells is the gold-standard method, but is expensive, time-consuming, and likely underestimates the reservoir size as not all intact and full-length integrated proviruses can be induced to make new infectious virions . Quantitative PCR (qPCR)-based methods detect integrated proviral sequences, but qPCR methods cannot distinguish between replication competent proviruses (the true reservoir) and other incomplete HIV DNA fragments or mutated proviruses that are replication-incompetent and, thus, likely overestimate the reservoir size by orders of magnitude. Nevertheless, studies that used both techniques (QVOA and qPCR) on the same sample have demonstrated that the measurements are proportionate to one another , correlate with the actual reservoir size, and can be used to measure the change in reservoir over time.
Studies using QVOA estimated the HIV reservoir half-life to be as short as 6.3 months to as long as 96 months [1,3,4,6,8]. The broad range of observed half-lives is likely explained by differences in assay methods, patient characteristics, degree of viral suppression, duration of observation, as well as the small sample size. In some studies, the reservoir half-life was longer (the decay rate slower) for participants who experienced intermittent ‘blips’ of plasma viremia as compared with those without blips [3,4].
To address limitations of prior studies of HIV reservoir decay (low sample sizes, the unclear effect of long-term viral suppression, and relatively short periods of observation), we investigated the clinical correlates with PBMC HIV DNA levels and decay in a large cohort of 111 HIV patients who had contributed serial peripheral blood mononuclear cell (PBMC) specimens to the University of Washington/Fred Hutch Center for AIDS Research (CFAR) Specimen Repository after at least 5 years of viral suppression before the first measurement of reservoir size.
Participants, specimens, and clinical data
University of Washington/Fred Hutch CFAR maintains a cohort of HIV-infected individuals who receive their care at the University of Washington HIV outpatient clinics and have agreed to contribute their clinical data to the University of Washington HIV Information System (UWHIS) . Many of these patients have also joined the University of Washington HIV Specimen Repository and serially donated blood for HIV research. For this study, we identified chronically HIV-infected participants in the cohort who had continuously taken suppressive antiretroviral therapy (ART) for at least 5 years with evidence for continuous viral suppression, defined as having HIV plasma viral loads (HIV pVLs) that were not detectable or consistently less than 40 copies/ml; patients with viral load ‘blips’ in this 5-year, premeasurement period were excluded from the study. Figure S1 (https://links.lww.com/QAD/B322) shows the frequency of HIV VL measurement during this 5-year period; demonstrating a median of 99 days between observations with nearly all individuals having at least annual testing. Clinical and demographic data were obtained from the UWHIS including ART history and laboratory tests (HIV RNA pVL, CD4+ and CD8+ T-cell counts). All patients were chronically infected. All patients provided informed consent before participating.
Measurement of reservoir size via quantitative polymerase chain reaction
One aliquot of five million PBMCs from each of 477 specimen donation times from all 111 patients was tested for HIV proviral DNA by qPCR. Cryopreserved PBMC specimens were thawed viably and extra chromosomal nucleic acid was removed using the Qiagen QIAprep Miniprep Kit. The remaining genomic DNA was extracted using the Qiagen QIAamp DNA mini kit. HIV DNA was amplified by PCR using primers targeting both gag (SK431 and HXB2) and pol (Abbott proprietary) regions, quantified using gag mix probes (HXB2 and AR-8) on the ABI 7900 and Abbott proprietary pol probe on the Abbott m2000, normalized for genomic DNA content based on the DNA concentration (NanoDrop) and corrected for the closest CD4+ cell count, CD4+ and reservoir measurements were on the same day for 198 of the 477 samples (41.5%); the CD4+ measurements were within 30 days of reservoir measurements for 320 of the samples (67.1%); and the measurements were within 90 days of each other for 444 of the reservoir measurements (93.1%).
The HIV DNA level assessed by qPCR was first normalized to the total PBMC count of the sample, and then to the nearest absolute CD4+ cell count (cells/μl) for the patient. The log10-transformed normalized qPCR value with zero values replaced with half of the minimum detected value before log-transformation defined the log-HIV DNA level, the dependent variable for subsequent analysis. We used linear mixed effects modeling, accounting for repeated measures on the same individual by including a random effects term to test for associations between clinical factors and HIV DNA level at baseline and the rate of change in level over time.
Using the estimated coefficient for time in years in the linear mixed effects model, we calculated the half-life of the HIV DNA level by the formula:
The statsmodel  package (version 0.6.1) of Python was used to implement the strategy, with the pandas  package (version 0.19.2) of Python used to normalize the data.
We examined whether there was a difference in the HIV DNA decay rate (half-life) among patients for whom all pVLs remained undetectable versus those with detectable-but-not-quantifiable pVLs (i.e. a positive pVL that was below the level of assay quantification, that is less than 40 copies/ml for most of the assays), as well as those with detectable and quantifiable pVLs (referred to as ‘viral blips’). We used linear mixed effects modeling with a per-participant intercept as a random effect to correct for multiple comparisons.
We identified 111 individuals meeting criteria for analysis; their demographic and clinical features are shown in Table 1. All study participants received combinations of a nucleotide or nucleoside reverse transcriptase inhibitors. A majority (86%) of participants also received a protease inhibitor and approximately one-half, (47; 42.3%) received an integrase inhibitor.
Quantitative polymerase chain reaction of gag and pol for reservoir size and decay assessment
Study participants contributed a median of three (range 1–23) blood donations to the repository, over a median of 1.4 years [range 0–8.5 (Fig S2, https://links.lww.com/QAD/B322)], resulting in 477 specimens that were tested for HIV proviral DNA by qPCR.
The HIV DNA levels as determined by qPCR for the HIV pol gene were consistent with the estimates for the HIV gag gene (Spearman's R2 of 0.60), as depicted in Fig. 1. However, there were 68 (14.1%) samples for which only the pol gene amplified but the gag gene did not and 3 (0.63%) samples for which the reverse was true. For all subsequent analysis, we estimated HIV DNA levels based on the pol gene qPCR normalized to both total PBMC genomic DNA in the sample, and to the most recent absolute CD4+ cell count (cell/μl), which were then log (base 10) transformed.
Peripheral blood mononuclear cell HIV DNA level correlates
We used linear mixed/random effects modeling to test the associations between the HIV DNA levels and demographic or clinical characteristics of a patient, and to calculate an HIV DNA level half-life while accounting for repeated sampling of individuals.
Table 2 shows the results of the multivariate regression. HIV DNA levels (at all observed time points, correcting for time) were significantly associated with the age (as a continuous variable) of the patient when clinical suppression was achieved. Age did not correlate with the pre-ART peak observed viral load (Fig S3, https://links.lww.com/QAD/B322; Spearman R2 = 0.001) or CD4+ nadir (Fig S4, https://links.lww.com/QAD/B322; Spearman R2 = 0.003). We did not find an association between the HIV DNA levels and the risk factor for HIV acquisition, specific antiretroviral class used (analyzed by ‘ever exposed’ and ‘months on particular antiretroviral classes’), gender, or race.
Time was included as a covariate in this model, corresponding to time in years since the first observation of reservoir size for each individual T0. We found that HIV DNA levels were significantly associated with time since study entry. The regression coefficient for time estimated a decay rate of −0.26 (pol/gDNA/CD4+ log 10 over time in years).
Peripheral blood mononuclear cell HIV DNA decay
Figure 2 shows the HIV DNA levels and change over time for the entire study cohort with the median and 95% confidence interval (CI) adjusted decay rate as determined by the linear mixed effects model. By this method, the half-life of the HIV DNA content for all study participants was 11.7 years (95% CI of 6.3–86 years).
Results of the subgroup analysis among patients for whom all pVLs remained undetectable versus those with detectable-but-not-quantifiable pVLs (i.e. a positive pVL that was below the level of assay quantification, i.e. <40 copies/ml for most of the assays) and those with detectable and quantifiable pVLs (referred to as ‘viral blips’) are shown in Table 3. We discovered a (not statistically significant) trend towards shorter HIV DNA level half-lives in 21 patients (19.6%) who remained consistently undetectable at 87 months, compared with the 82 patients (76.6%) with at least one episode of detectable-but-not-quantifiable of 145 months, or four people (3.7%) with blips of quantifiable viral RNA in the plasma at a half-life of 264 months.
The subset of patients without blips had the same HIV DNA level (1.3 copies of pol/CD4+) as those with detectable-but-not-quantifiable plasma RNA (1.3 copies of pol/CD4+). Surprisingly, for those with blips (quantifiable plasma RNA), there was a trend towards higher HIV DNA level (0.6 copies of pol/CD4).
This is one of the largest observational cohort studies of PBMC HIV DNA level kinetics to date, with 477 separate measurements from 111 participants on suppressive ART for a median of 8 years. Among patients with consistently undetectable pVLs throughout the observation period, the half-life was estimated to be 87 months, which is consistent with other studies of chronically virally suppressed HIV patients. Most patients in our study (77%) had at least brief stretches of detectable-but-not-quantifiable serum HIV virus, for whom the half-life (144 months) was similar to the cohort as a whole (140 months), and longer than the typical HIV reservoir half-life previously reported [1,3,4,6,8].
Our study used quantitative DNA PCR for the pol gene as a surrogate for HIV persistence. We paired quantitative PCR measurements of the gag gene, but found in a subset of samples gag did not amplify when pol was detectable (often at high levels). We hypothesize that mutations in the gag gene (perhaps better tolerated than mutations in pol) prevented proviral amplification with gag-specific primers in these instances. Sequence analyses of the gag and pol from discordant samples are needed to test this hypothesis.
Only age at the start of study correlated with reservoir size. This may be because of age at entry being confounded by or being a surrogate for a longer time before HIV diagnosis and longer time without virologic control that, in turn, led to higher PBMC HIV DNA levels. Unfortunately, we do not know the dates of HIV infection, and therefore, the timing of ART relative to infection, for these patients. However, we found no clear relationship between age at study entry and peak pVL or CD4 nadir prior to the study observation period, suggesting that the relationship of age to PBMC HIV DNA level is not simply mediated by plasma virus zenith or disease stage (see supplemental Figures S3 and S4, https://links.lww.com/QAD/B322).
Our results demonstrate a longer HIV persistence than previously reported, which may be because of: differences in the reservoir composition as estimated by qPCR as compared with QVOA; nonlog-linear kinetics of PBMC HIV DNA decay resulting in a slowing of the half-life for participants with a longer time interval of viral suppression; the larger size of our cohort better captured the HIV DNA decay kinetics compared with other studies with fewer participants; and the possibility that limited viral replication because of incomplete adherence led to reseeding of the HIV DNA pool.
We have used quantitative PCR to measure integrated HIV proviral DNA as a surrogate for HIV reservoir size, or HIV proviral persistence. The QVOA is the gold standard for measuring the inducible, infectious HIV reservoir. We note prior studies [7,8] found a linear relationship between reservoir size as estimated by qPCR and QVOA. Estimates of HIV persistence by quantitative PCR are orders of magnitude greater than that estimated by QVOA, as qPCR can detect defective provirus that is not replication competent. However, changes in the PBMC HIV DNA level over time (the decay rate or half-life) should be similar whether measured by qPCR or QVOA, provided the death and proliferation rates of cells with replication-competent or defective provirus are similar. However, the relative persistence of cells with defective or replication-competent proviruses remains controversial; one study reported an inverse correlation between proviral DNA clone size and the probability of reactivation suggesting cells with replication-competent provirus are targeted for destruction over proliferation . In contrast, a more recent publication using ex-vivo CD4+ T cells from HIV-infected, ART-suppressed persons, demonstrated CD8+ T-cell killing of cells with defective provirus with no effect on cells harboring replication-competent provirus . Our results are similar to that of Siliciano et al. and showed a trend toward a longer HIV persistence in those patients with detectable viral RNA in plasma, which likely represents expansion of the PBMC HIV DNA pool because of cell proliferation , given that all patients were maintained on suppressive ART throughout the period of observation. Additional studies that examine proviral integration site analysis are needed to prove the concept of clonal expansion and to determine whether the ratio of cells with defective and replication competent provirus changes over time.
Funding: ACTG Laboratory Center (UM1-AI-106701); CFAR (P30-AI-027757); HVTN HIV Diagnostic Laboratory (UM-AI-068618).
A portion of this work was previously presented as an oral abstract (#953) at ID Week 2016 (New Orleans, Los Angeles, USA).
Author contributions: J.L.G. was primarily responsible for preparing figures, writing of the manuscript, and the data analysis. J.S. and S.H. contributed to the statistical analysis plan. M.M.K., H.M.K., and A.E.W. contributed to the study design and metadata. R.W.C. contributed to the study design and HIV Proviral DNA measurement strategy. E.K. contributed the HIV Proviral DNA measurements. R.D.H. contributed to the study design, organization, and writing.
Conflicts of interest
There are no conflicts of interest.
1. Finzi D, Blankson J, Siliciano JD, Margolick JB, Chadwick K, Pierson T, et al. Latent infection of CD4+ T cells provides a mechanism for lifelong persistence of HIV-1, even in patients on effective combination therapy
. Nat Med
2. Chomont N, El-Far M, Ancuta P, Trautmann L, Procopio FA, Yassine-Diab B, et al. HIV reservoir size and persistence are driven by T cell survival and homeostatic proliferation
. Nat Med
3. Ramratnam B, Mittler JE, Zhang L, Boden D, Hurley A, Fang F, et al. The decay of the latent reservoir of replication-competent HIV-1 is inversely correlated with the extent of residual viral replication during prolonged antiretroviral therapy
. Nat Med
4. Siliciano JD, Kajdas J, Finzi D, Quinn TC, Chadwick K, Margolick JB, et al. Long-term follow-up studies confirm the stability of the latent reservoir for HIV-1 in resting CD4+ T cells
. Nat Med
5. Cockerham LR, Siliciano JD, Sinclair E, O’Doherty U, Palmer S, Yukl SA, et al. CD4+ and CD8+ T cell activation are associated with HIV DNA in resting CD4+ T cells
. PLoS One
6. Eriksson S, Graf EH, Dahl V, Strain MC, Yukl SA, Lysenko ES, et al. Comparative analysis of measures of viral reservoirs in HIV-1 eradication studies
. PLoS Pathog
7. Kiselinova M, De Spiegelaere W, Buzon MJ, Malatinkova E, Lichterfeld M, Vandekerckhove L. Integrated and total HIV-1 DNA predict ex vivo viral outgrowth
. PLOS Pathog
8. Crooks AM, Bateson R, Cope AB, Dahl NP, Griggs MK, Kuruc JD< ET-AL>. Precise quantitation of the latent HIV-1 reservoir: implications for eradication strategies
. J Infect Dis
9. Kitahata MM, Dillingham PW, Chaiyakunapruk N, Buskin SE, Jones JL, Harrington RD, et al. University of Washington HIV Study Cohort. Electronic human immunodeficiency virus (HIV) clinical reminder system improves adherence to practice guidelines among the University of Washington HIV Study Cohort
. Clin Infect Dis Off Publ Infect Dis Soc Am
10. Seabold S, Perktold J. Statsmodels: Econometric and statistical modeling with Python. Proceedings of the 9th Python in Science Conference
11. McKinney W. Data structures for statistical computing in Python. In van der Walt S, Millman J, editors. Proceedings of the 9th Python in Science Conference
, 2010; 51–56.
12. Lorenzi JCC, Cohen YZ, Cohn LB, Kreider EF, Barton JP, Learn GH, et al. Paired quantitative and qualitative assessment of the replication-competent HIV-1 reservoir and comparison with integrated proviral DNA
. Proc Natl Acad Sci USA
13. Huang S-H, Ren Y, Thomas AS, Chan D, Mueller S, Ward AR, et al. Latent HIV reservoirs exhibit inherent resistance to elimination by CD8+ T cells
. J Clin Invest