Subclinical Cytomegalovirus and Epstein-Barr Virus Shedding Is Associated with Increasing HIV DNA Molecular Diversity in Peripheral Blood during Suppressive Antiretroviral Therapy

As part of this study, we evaluated the molecular characteristics of the HIV DNA reservoir over time during antiretroviral treatment (ART) in relation to those of other chronic viral infections (i.e., cytomegalovirus [CMV] and Epstein-Barr virus [EBV]). We demonstrated that the presence of CMV and high-level EBV DNA in peripheral blood cells was associated with changes in HIV DNA molecular diversity. Specifically, HIV DNA molecular diversity increased over time among participants with detectable CMV and high-level EBV DNA, while it significantly declined among participants with no/low viral shedding. Although the current study design does not allow causality to be inferred, it does support the theory that persistent CMV and EBV shedding could contribute to the dynamics of the HIV DNA reservoir during suppressive ART, even when ART is initiated during the earliest phases of HIV infection.

ABSTRACT

Cytomegalovirus (CMV) almost universally infects persons with HIV (PWH), and it is a driver of persistent inflammation and HIV persistence. The mechanisms underlying the association between CMV (and possibly other herpesviruses) and HIV persistence are unclear. Serially collected blood samples were obtained from men who have sex with men (MSM) who started antiretroviral therapy (ART) within 1 year of their estimated date of HIV infection (EDI). Total CMV and Epstein-Barr virus (EBV) DNA were quantified in peripheral blood mononuclear cells by droplet digital PCR (ddPCR). Deep sequencing of the HIV DNA partial env gene was performed, and the dynamics of viral diversity over time were analyzed in relation to CMV and EBV shedding status. In total, 37 MSM PWH were included and followed for a median of 23 months (IQR, 22 to 28). Participants started ART within a median of 3.1 months (IQR, 1.5 to 6.5) after EDI and remained virally suppressed thereafter. A total of 18 participants (48.6%) were classified as high EBV shedders, while 19 (51.4%) were classified as CMV shedders. In longitudinal analyses, normalized molecular diversity levels tended to increase over time among participants with detectable CMV and high EBV DNA (0.03 ± 0.02, P = 0.08), while they significantly declined among participants with no/low viral shedding (−0.04 ± 0.02, P = 0.047, interaction P < 0.01). Subclinical CMV and EBV shedding could contribute to the dynamics of the HIV DNA reservoir during suppressive ART. Whether persistent CMV/EBV replication could be targeted as a strategy to reduce the size of the latent HIV reservoir is an avenue that should be explored.

IMPORTANCE As part of this study, we evaluated the molecular characteristics of the HIV DNA reservoir over time during antiretroviral treatment (ART) in relation to those of other chronic viral infections (i.e., cytomegalovirus [CMV] and Epstein-Barr virus [EBV]). We demonstrated that the presence of CMV and high-level EBV DNA in peripheral blood cells was associated with changes in HIV DNA molecular diversity. Specifically, HIV DNA molecular diversity increased over time among participants with detectable CMV and high-level EBV DNA, while it significantly declined among participants with no/low viral shedding. Although the current study design does not allow causality to be inferred, it does support the theory that persistent CMV and EBV shedding could contribute to the dynamics of the HIV DNA reservoir during suppressive ART, even when ART is initiated during the earliest phases of HIV infection.

INTRODUCTION

A common driver of inflammation in the setting of HIV is cytomegalovirus (CMV), which almost universally infects persons with HIV (PWH) (1). CMV establishes a lifelong infection characterized by persistent antigen exposure and complex interactions with the host immune system (2). Even PWH who have intact immune systems have higher levels of subclinical CMV shedding in mucosal sites and tissues than persons without HIV (3–8). This asymptomatic CMV shedding profoundly affects circulating T-cell subset dynamics and is associated with extensive proliferative expansion of T cells (“inflation”), immune dysfunction, and impaired immune recovery (2, 3, 9–12). Data from our group and others have consistently demonstrated that CMV drives several mechanisms promoting persistence of HIV DNA (13). For example, in our initial cross-sectional studies, we found that the presence of subclinical CMV replication was associated with higher levels of HIV DNA in both antiretroviral therapy (ART)-naive and ART-suppressed PWH (14, 15). In a longitudinal study of 107 men starting ART within 4 months after primary HIV infection, the presence of detectable CMV and Epstein-Barr virus (EBV) DNA in peripheral blood mononuclear cells (PBMCs) was associated with a slower decay rate of HIV DNA over approximatively 19 months that was not accounted for by low-level HIV RNA transcription (16). Importantly, a recent study found that CMV/EBV-specific CD4⁺ T cells were enriched for HIV DNA in PWH who underwent myeloablative chemotherapy and then had immune reconstitution (17).

While causality cannot be inferred in these observational studies, they collectively support our scientific premise that persistent CMV, and possibly also EBV replication and associated inflammation, promotes the maintenance of HIV reservoirs. The mechanisms underlying this association are still unclear (13).

As part of this study, we evaluated the dynamics of HIV DNA molecular diversity over time during ART in relation to CMV and EBV shedding status, time from ART initiation, and their interaction effects.

RESULTS

Study cohort and samples.

Study participants (n = 37) were all men who have sex with men (MSM) diagnosed with recent HIV infection and followed for up to 53 months (median, 23; IQR, 22 to 28) (Table 1). Participants were started on ART after a median of 3.1 months (IQR, 1.5 to 6.5) from EDI and achieved viral suppression within 5.5 months (IQR, 2.7 to 7.1) of starting ART, without any documented HIV RNA blips in blood plasma. At study entry, participants were a median of 35 years old (IQR, 29 to 44) and had a median baseline CD4⁺ T-cell count of 548 cells/μl (IQR, 394 to 710), median CD4:CD8 ratio of 0.5 (IQR, 0.4 to 0.7), and median peak HIV RNA of 5.7 log₁₀ copies/ml (IQR, 4.9 to 6.4).

TABLE 1.

Demographics and clinical metrics of study participants

Participant characteristicc	Description of participantsb
CMV shedding status (no. [%])
Nonshedder	18 (48.6)
Shedder	19 (51.4)
EBV shedding status (no. [%])
Low shedder	19 (51.4)
High shedder	18 (48.6)
MSM (no. [%])	37 (100.0)
Age at enrollment (yrs)	35 (29–44)
Length of follow-up (mo)	23 (22–28)
No. of mo from EDI to ART initiation	3.1 (1.5–6.5)
Peak HIV RNA, log10 copies/ml	5.7 (4.9–6.4)
Baseline CD4+ cell count (cells/μl)a	548 (394–710)
CD4/CD8 ratioa	0.5 (0.4–0.7)
HIV DNA (log10 copies/106 CD4+)a	2.7 (1.9–3.4)
HIV RNA level (log10 copies/ml)a	5.2 (4.7–5.6)
Normalized entropy level at the ENV region [entropy/ln(nhaplotype)]	3.0 (2.6–3.8)

^aAt antiretroviral therapy (ART) initiation.

^bN = 37. Values given are median and interquartile range (IQR) unless otherwise stated.

^cCMV, cytomegalovirus; EBV, Epstein-Barr virus; MSM, men who have sex with men; EDI, estimated date of infection; HIV, human immunodeficiency virus, n_Haplotype, no. of haplotypes recovered.

Among the 37 participants, 18 (48.6%) did not shed CMV at any time point, while 19 (51.4%) were classified as CMV shedders (i.e., defined as at least one positive measure with droplet digital PCR [ddPCR]). For EBV, 18 were classified as high shedders (i.e., defined as average EBV DNA above the median value of 64.1 copies/10⁶ cells). Participants who had detectable CMV DNA in PBMC, were more likely to have high EBV DNA (P = 0.03). Because CMV and EBV shedding patterns were correlated, we decided to combine these variables to classify participants into the following groups, depending on their shedding patterns: both CMV-negative (CMV_NEG) and low EBV (EBV_LOW) (n = 13), either CMV-positive (CMV_POS) or high EBV (EBV_HIGH) (n = 11), and both CMV_POS and EBV_HIGH (n = 13).

Sequencing recovery.

Deep sequencing of the HIV DNA env C2V3 region (HXB2 coordinates 6962 to 7315) was performed using the Illumina MiSeq platform. A median of 3 samples (IQR, 2 to 3) per individual were investigated. After quality filtering through our in-house pipeline (18) adapted from Zanini et al. (19), a median of 320,452 (IQR, 198,053 to 1,494,661) reads/sample were obtained. From these, a median of 5 (IQR 3 to 10) HIV DNA haplotypes/sample with a minimal frequency threshold of 0.01 were obtained and further analyzed.

Impact of CMV and EBV shedding on the dynamics of HIV diversity during ART.

Changes in within-host HIV molecular diversity over time were assessed by evaluating the normalized Shannon entropy measure of the partial HIV envelope. Univariable models showed no association with either viremic shedding (P = 0.41 in the univariable models in Table 2), time (P = 0.97), or any of the covariates examined (P > 0.11). In the multivariable model, we evaluated the dynamics of HIV DNA diversity over time during ART in relation to CMV and EBV shedding status, time from ART initiation, and their interaction effects. We showed a significant interaction between HIV DNA molecular diversity and shedding status over time during ART (P = 0.03; Table 2). In fact, HIV DNA molecular diversity (i.e., normalized entropy levels) tended to increase over time among participants with both CMV_POS and EBV_HIGH (0.03 ± 0.02, P = 0.08), while it significantly declined among participants with both CMV_NEG and EBV_LOW (−0.04 ± 0.02, P = 0.047). The difference in slope between these two groups was significant (P = 0.009; Fig. 1). The slope among participants who had either CMV_POS or EBV_HIGH did not differ from either group (P > 0.28). Including other covariates (listed in Table 2) in the multivariable model did not significantly improve model fit or alter the regression coefficient for the interaction term (time by viremic shedding) by more than 3%.

TABLE 2.

Univariable and multivariable models of the relation of CMV and EBV shedding with normalized entropy levels (sequences recovered from the ENV region)

Predictor	Univariable models			Multivariable model
	Coefficient	95% CId	Pe	Coefficient	95% CI	P
Intercept	3.41	(2.86, 3.96)	<0.001	3.84	(3.16, 4.52)	<0.001
Time (no. of mo since ART initiation)	0.00	(−0.02, 0.03)	0.969	−0.04	(−0.09, 0)	0.047
Viremic sheddingc			0.412b
Either positive CMV or high EBV	−0.43	(−1.24, 0.39)	0.294	−0.87	(−1.87, 0.13)	0.089
Both positive CMV and high EBV	0.06	(−0.73, 0.84)	0.886	−0.74	(−1.71, 0.23)	0.132
Viremic shedding by timec						0.028b
Either positive CMV or high EBV				0.05	(−0.02, 0.11)	0.156
Both positive CMV and high EBV				0.08	(0.02, 0.14)	0.009
Covariates
Age at enrollment (yrs)	0.01	(−0.02, 0.05)	0.400
Peak viral load (log10 copies /ml)	0.11	(−0.21, 0.44)	0.482
CD4 ($\sqrt{\text{cell counts}}$/μl)a	−0.05	(−0.12, 0.01)	0.107
CD4/CD8 ratioa	−0.61	(−1.7, 0.48)	0.260
Early ART initiation (≤90 days)	0.32	(−0.35, 0.99)	0.335

^aAt antiretroviral therapy (ART) initiation.

^bResult of a likelihood-ratio test against the model without the referenced factor.

^cThe negative-CMV and low-EBV group was used as the reference category.

^dCI, confidence interval.

^dBoldface indicates significant results (P < 0.05).

FIG 1.

Dynamics of HIV DNA molecular diversity (env) over time since ART initiation according to CMV or EBV shedding status. Molecular diversity of the partial env region was estimated with Shannon entropy (normalized). See Materials and Methods for details. Participants were classified into one of the following groups depending on their shedding patterns: CMV-negative (CMV_NEG) and low EBV (EBV_LOW), either CMV-positive (CMV_POS) or high EBV (EBV_HIGH), and both positive CMV and high EBV. Gray dots and line indicate normalized entropy values and participants, respectively. Blue lines and shaded areas indicate model-predicted values and corresponding 95% confidence intervals, respectively.

Of note, two samples in the both CMV_POS and EBV_HIGH group had higher values than any other samples at the last time point. However, even after removing these two individuals, the reported difference in the slopes between the both CMV_POS and EBV_HIGH and both CMV_NEG and EBV_LOW groups remained significant (interaction P = 0.045).

DISCUSSION

The complex connections between CMV and HIV are becoming increasingly defined (1). To better understand how asymptomatic CMV and EBV replication in blood cells is associated with HIV reservoir composition over time, the present study evaluated the dynamics of HIV DNA molecular diversity over time during early ART in relation to chronic CMV/EBV shedding, time from ART initiation, and their interaction effects. Our study demonstrated that the presence of asymptomatic CMV and EBV DNA in peripheral blood cells was associated with changes in HIV DNA molecular diversity. Specifically, HIV DNA molecular diversity tended to increase over time among participants with detectable CMV DNA and high EBV shedding, while it significantly declined among participants with no detectable CMV and low EBV shedding. Although the current study design does not allow causality to be inferred, it does support the theory that persistent CMV and EBV replication could contribute to the dynamics of the HIV DNA reservoir during suppressive ART, even when ART is initiated during the earliest phases of HIV infection. This could suggest ongoing HIV replication during early HIV infection, although a previous study from our group did not find any correlation between CMV shedding and cellular HIV RNA transcription in this same data set (16). However, this previous study evaluated levels of cellular HIV RNA transcription only in PBMCs, and we cannot exclude the possibility that a difference in cellular transcription exists in other compartments (e.g., lymphoid or gut tissue). Another possible explanation is that the reservoir is more dynamic as a result of clonal expansions and contractions of memory CD4⁺ T cells as a consequence of antigen stimulation (20–22). Specifically, HIV-infected CMV/EBV-reactive CD4⁺ T cells might be responsible for the changes in diversity over time as a result of their expansions/contractions when CMV and EBV DNA are detectable, but this needs to be evaluated as part of future studies.

This study had several limitations. Most importantly, because this was an observational study, we cannot establish a definitive causal relationship between CMV/EBV shedding and HIV DNA molecular diversity. The extent of immune activation during treated HIV infection could be a determinant of both chronic viral shedding and increased in molecular diversity. We also did not determine if the observed increase in HIV DNA CD4⁺ T cells represents replication competent latent provirus, and which cellular subset carries the integrated HIV DNA in the CMV/EBV-shedding group. Also, further studies will be needed to determine if presence of detectable CMV and EBV DNA in PBMCs is a surrogate of low-level viral replication or rather just latent virus present in circulating mononuclear cells.

Despite these limitations, this study provides important insights regarding connections between asymptomatic CMV/EBV replication and the HIV DNA reservoir and complements our previous findings by providing longitudinal data on a well-characterized cohort of individuals with early HIV infection. Specifically, it demonstrated that even among individuals who started ART early during HIV infection, asymptomatic CMV and EBV reactivation is associated with increasing molecular diversity of HIV DNA during effective ART. Since almost all PWH and especially MSM are infected with CMV and EBV (6), future studies are needed to determine if persistent viral replication could be targeted as a strategy to reduce the size of the latent HIV reservoir.

MATERIALS AND METHODS

Ethics statement.

The study was reviewed and approved by the University of California San Diego Human Research Protections Program. All adult participants (age, >18 years) provided written informed consent.

Study participants.

Men who have sex with men (MSM) starting ART after recent HIV infection were recruited from the San Diego Primary Infection Consortium. A total of 107 people enrolled in a previous reservoir study, providing a total of 515 longitudinal blood samples (16). Of the 107 participants, we selected 37 who started ART within 1 year of the estimated date of infection (EDI), had the longest follow-up, and had and enough frozen PBMCs available at baseline and after ART initiation for further analysis by deep sequencing. A summary of participant characteristics and sample time points is shown in Table 1, as well as in Fig. 2.

FIG 2.

Diagram displaying the study population and details about each sampled time point. ART, antiretroviral therapy; IQR, interquartile range.

Sample storage and data available from previous study.

PBMCs were isolated from whole blood using a density gradient medium (Lymphoprep, Stemcell Technologies, Cambridge, MA) per manufacturer’s instructions, and cryopreserved at –150°C in 90% fetal bovine serum plus 10% dimethyl sulfoxide (DMSO) within 24 h of blood collection.

Detection of EBV and CMV.

DNA was extracted from 5 million PBMCs for each time point using the AllPrep DNA/RNA minikit (Qiagen, Carlsbad, CA). Total CMV and EBV DNA was quantified by droplet digital PCR (ddPCR) (16, 23). Copy numbers were calculated as the mean of replicate PCR measurements and normalized to cell input, determined by RPP30 (24). Based on the dichotomized measures of detectable CMV or within-participant mean EBV copies above the median level (>64 per 10⁶ cells), we characterized each participant’s CMV and EBV status as either “shedder” (detectable CMV or high EBV) or “nonshedder” (no detectable CMV or low EBV measure).

Sequencing and molecular analyses.

(i) Library preparation. Deep sequencing of the HIV DNA env C2V3 region (HXB2 coordinates 6962 to 7315) was performed using the Illumina MiSeq platform. DNA library preparations were created per manufacturer specifications using the Nextera XT index kit (Illumina, San Diego, CA). Briefly, a minimum of 100 ng of DNA template was combined with 25 μl of 2× Kapa HiFi HotStart ReadyMix (Kapa Biosystems, Wilmington, Massachusetts) and 5 μl of each of the two Nextera index primers provided with the Nextera XT index kit v2 set A (Illumina, San Diego, CA) The total volume was adjusted to 50 μl with the addition of PCR-grade water. The preparations were then thermocycled, and >100 ng DNA was purified using AMPure XP beads (ratio of 1.2 μl beads:1μl DNA, catalog no. A63881; Beckman) and eluted with 10 mM Tris (pH 8.5).

(ii) Sequencing. The Illumina MiSeq instrument and MiSeq reagent kit v3 600-cycle paired-end sequencing kits (catalog no. MS-102-2003 and MS-102-3003) were used to sequence the DNA libraries.

(iii) Read mapping and filtering. The reads were analyzed using a custom pipeline adapted from Zanini et al. (19). Briefly, (i) reads were first mapped onto the HIV-1 reference HxB2, (ii) reads mapped to partial env region were kept (ambiguous reads were discarded) and trimmed for a Phred quality score greater than or equal to 30, (iii) a consensus sequence was computed in each sample from a subset of the reads, using a chain of overlapping local multiple sequence alignments, (iv) reads were remapped against their own consensus, (v) reads were trimmed for mapping errors at the edges (small indels), (vi) filtered reads were mapped a third time against a patient-specific consensus sequence from the initial time point, and (vii) reads were refiltered and checked again for cross-contamination. Reads with a distance larger than a sample- and fragment-specific threshold were discarded. Each threshold, calculated to exclude even traces of cross-contamination that might have happened during RNA extraction, PCR amplification, or library preparation, was established by plotting the distribution of Hamming distances of reads from the sample consensus and by excluding reads that were further away than the tail of the main peak (19).

HIV DNA partial env haplotypes representative of distinct variants present in at least 1% or more in the viral population were generated and further analyzed (25, 26). For each sample, the HIV diversity was estimated by measuring the Shannon entropy index (S) and accounting for both the number of cleaned mapped reads and the number of haplotypes (25, 26) (i.e., normalized entropy). Specifically, we normalized the outcome by dividing it by the log of the number of haplotypes. This normalized version represents an “averaged measure of entropy per sequence haplotype,” taking the variability associated with the number of haplotypes across participants and time points into consideration.

Statistical analyses.

All statistical analyses were performed using R statistical software (27). The primary outcome was the measure of normalized HIV DNA env entropy. To examine the dynamics of the normalized HIV DNA env entropy measures over time by CMV and EBV DNA shedding status, we used the continuous time predictor, defined as the number of months elapsed from the onset of ART to the date of sample collection. We also controlled for the following covariates: age, peak HIV viral load, time from the EDI to ART, CD4 count, CD4/CD8 ratio, and early ART initiation (≤90 days from EDI).

First, we built a single-predictor model to test the association of the outcome (HIV DNA molecular diversity) with each of the predictors (i.e., CMV and EBV shedding status) and covariates. Second, we built a multiple-predictor (i.e., multivariable) model to evaluate HIV diversity dynamics over time in relation to CMV and EBV shedding status. In this model, we included the CMV and EBV shedding status, time, and their interaction, as well as any of the covariates that were associated with the outcome at a 10% significance level in the single-predictor models.

Data availability.

All read files were uploaded to the NCBI BioSample database under accession numbers SAMN15361593 to SAMN15361644.