|Home | About | Journals | Submit | Contact Us | Français|
The gut is a major reservoir for HIV in patients receiving antiretroviral therapy (ART). We hypothesized that distinct immune environments within the gut may support varying levels of HIV.
In 8 HIV-1+ adults on ART with CD4>200 and plasma VL<40, levels of HIV and T-cell activation were measured in blood and endoscopic biopsies from the duodenum, ileum, right colon, and rectum.
HIV DNA and RNA per CD4+T-cell were higher in all four gut sites compared to blood. HIV DNA increased from the duodenum to the rectum, while the median HIV RNA peaked in the ileum. HIV DNA correlated positively with T-cell activation in the PBMC but negatively with T-cell activation in the gut. Multiply-spliced RNA was infrequently detected in gut, and unspliced RNA/DNA ratios were lower in the colon and rectum relative to PBMC, reflecting paradoxically low HIV transcription given the higher T-cell activation in the gut.
HIV DNA and RNA are both concentrated in the gut, but the inverse relationship between HIV DNA and T-cell activation in the gut and the paradoxically low levels of HIV expression in the large bowel suggest that different processes drive HIV persistence in the blood and gut.
The gut plays critical roles in HIV transmission, pathogenesis, and persistence on combined antiretroviral therapy (ART). In patients who are receiving ART and whose viral load (VL) is undetectable (<40 copies/ml), HIV DNA [1–6], RNA [1, 3, 4, 7–10], and even replication-competent virus  have been recovered from the gut. While most studies have reported detection of HIV DNA in the gut, RNA detection has been more variable [1, 3, 4, 7–10], and relatively few studies have attempted to quantify the levels of HIV RNA [3, 7, 10] or DNA [3, 5, 6] in ART-suppressed patients. These studies differed in many important respects, including the stage of HIV infection, gut site and method of sampling, tissue processing, and assay for HIV.
A recent study by Chun et al, using colonoscopic-guided biopsies from the terminal ileum, showed that HIV DNA levels per million CD4+ cells were on average five times higher in the gut than in the peripheral blood . It is not known whether there is a similarly disproportionate concentration of HIV RNA in the gut. No study has sought to systematically measure both HIV DNA and RNA throughout the gut of ART-suppressed patients or to compare HIV RNA/DNA ratios in gut to those in peripheral blood mononuclear cells (PBMC).
To address the gap in the understanding of HIV burden across the spectrum of gut tissue, we systematically measured levels of HIV DNA, HIV RNA, and T-cell activation throughout the small and large bowel of chronically HIV-infected, ART-suppressed adults. We hypothesized that distinct immune environments in the gut would support varying levels of HIV persistence.
HIV+ adults meeting entry criteria were recruited from two hospital-based HIV clinics. Inclusion criteria included: 1) age 18–65 years; 2) infection with HIV-1; 3) ART for ≥ 12 months; 4) no change in ART for ≥ 3 months; 5) CD4+T-cell count>200 cells/μl; and 6) VL<40 copies/ml for >6 months prior to study entry. Exclusion criteria included factors that would increase the risk from sedation, endoscopy, or biopsy. The study was approved by the Institutional Review Board of the University of California, San Francisco.
Blood was sampled at screening and at study entry. At study entry, all participants had phlebotomy followed by esophagogastroduodenoscopy and colonoscopy with 7–10 biopsies each from the duodenum, terminal ileum, right colon, and rectum. Duodenal biopsies were taken with a Radial Jaw 4, 2.8mm forceps (Boston Scientific). Colonoscopic biopsies were taken using the Radial Jaw 3, 3.7mm forceps or the Radial Jaw 4, 2.8mm forceps (Boston Scientific).
68ml of blood was collected for plasma and PBMC in acid-citrate-dextrose tubes (BD). Plasma was obtained by centrifuging twice at 1000g for 10minutes without braking. The buffy coat from the first spin was used to isolate PBMC by centrifugation on Ficoll according to the procedure recommended for Dynal Invitrogen Bead Separations. PBMC were resuspended in PBS+0.1% BSA+2mM EDTA (buffer A) and aliquots were saved for flow cytometry, HIV DNA, and HIV RNA. 6×107 PBMC were used to isolate CD4+ cells by negative selection using the Dynabeads Untouched Human CD4+T-cell Kit (Invitrogen). 5×105 CD4+T-cells were saved for flow cytometry, while the remainder was frozen as cell pellets for HIV DNA and RNA.
Six to nine biopsies from each site (initially placed in RPMI with L-Glu, penicillin/streptomycin, and 15% fetal calf serum) were separated into single cells using a modification of a published method [11, 12]. Briefly, biopsies were subjected to 3 rounds of collagenase digestion, mechanical disruption (by passing through a blunt 16gauge needle), clarification (by passing through a 70μm cell strainer), and washing. The three aliquots of strained and washed gut cells were then combined, counted, and resuspended in buffer A. 5×105 cells were set aside for flow cytometry, while the remainder was divided and frozen at −80C for subsequent DNA or RNA extraction.
Plasma HIV RNA was measured using a modification of the Abbott Real Time HIV Viral Load assay that involves pelleting virions from up to 30ml of plasma (detection limit<0.5 copies/ml; S. Yukl, et al, submitted). Plasma was diluted 1:1 with PBS, divided in two portions, layered onto 10ml of 6% iodixanol (OptiPrep Density Gradient Medium (Sigma) diluted 1:10 in PBS) in 50ml polypropylene tubes (Beckton), and centrifuged at 47,810g for 3hrs at 4C without braking. Viral pellets were resuspended in a total of 1000μl of PBS and the HIV RNA was measured according to the Abbott protocol. Copy values were extrapolated from the Ct values of the standards and then adjusted for the concentration factor.
DNA was extracted using the DNA Blood Mini Kit (Qiagen) and measured using a NanoDrop 1000 Spectrophotometer. Three replicates of 500ng of DNA from each sample were then used in a real time Taqman PCR assay for HIV DNA that uses primers and a Locked Nucleic Acid probe from the Gag region. This assay detected a 10 copy standard 19/20 times (95%) with a mean value of 9.5+/−1.7 copies (P. Li et al, submitted). Primers were G-19–2-F (5’-AGCAGCYATGCAAATGTTA-3’;1374–1392) and G-20-R (5’-AGAGAACCAAGGGGAAGTGA-3’;1474–1493). The probe, 5’-CCATCAATGAGGA-3’ (1400–1412; underlined bases are locked), was dual-labeled with 6-FAM(5’) and Black quencher BHQ-1(3’). Reaction volume was 50μl with primer and probe concentrations of 200nM, 1× Taqman master mix, and 500ng of sample DNA. Cycling conditions were: 50C for 2 min, 95C for 5 min, then 50 cycles of 95C for 15s and 59C for 1 min. External standards (105 to 1) were prepared from DNA extracted from serial dilutions of known numbers of 8E5 cells (NIH AIDS Reagent Program), each of which contains one integrated HIV genome per cell.
HIV DNA copy numbers were extrapolated from the Ct values of the samples and expressed as copies/106 cells (assuming 1 μg total DNA corresponds to 160,000 cells). To account for variation in the number of CD4+T-cells in different samples, results were also normalized by the percent of all cells that were CD45+CD3+CD4+ (by flow cytometry) and expressed as copies/106 CD4+T-cells. To verify the DNA concentrations and assess for PCR inhibitors, samples from four patients were assayed using a separate real time PCR for β-actin.
RNA was extracted using the Rneasy Kit (Qiagen) with on-column digestion using RNase-free DNase (Qiagen). To maximize the sensitivity of the qRT-PCR assays, which approaches 1 copy/reaction [13, 14], primers and probes were matched to each subject based on the sequence of HIV DNA in peripheral CD4+T-cells. Unspliced HIV RNA (UsRNA) was measured using primers from the pol region (2536–2562 and 2634–2662). Multiply-spliced HIV RNA encoding for Tat and Rev (MsRNA tat-rev) was measured using primers from tat exon 1 (5956–5979) and tat/rev exon 2 (8433–8459). Total multiply-spliced HIV RNA (MsRNA Tot) was measured using primers from rev exon 1 (6012–6045) and tat/rev exon 2 (8433–8459). qRT-PCR was done under conditions described previously . HIV copy numbers (the mean of replicate measurements) were extrapolated using the formula Ct= I+Slog10 (copy number), where the slopes(S) were derived from external standard curves and the intercepts(I) were calculated for each patient using the mean Ct value from all sites where 50% of replicates were detectable, reflecting 1 copy [13, 14, 16].
HIV RNA copy numbers were normalized to cellular input into the PCR, as determined both by total RNA concentration (measured by NanoDrop 1000), assuming that 1ng RNA correspond to 1000 cells , and by levels of glyceraldehyde phosphate dehydrogenase (GAPDH) RNA, as determined by a separate qRT-PCR. Results (copies/106 cells) from the two different methods of normalization correlated well. To account for variation in the number of CD4+T-cells in different samples, copy numbers were further normalized by the percent of all cells that were CD45+CD3+CD4+.
Gut biopsy specimens were fixed in 4% paraformaldehyde over night and then transferred to 80% ethanol. Fixed gut biopsies were assayed for cell-associated HIV RNA by ISH with 35S-labeled HIV-1 antisense riboprobe (Lofstrand Labs Limited) and quantitative image analysis, as described previously .
CD45, CD3, CD4, CD8, CD38, and HLA-DR were measured on blood and gut cells using flow cytometry adapted from previously published methods . Antibodies included the following: CD45-APC, CD3-Pacific Blue (BD Bioscience), CD4-ECD (Beckman Coulter), CD8-Q-Dot605 (Invitrogen), CD38-PE (BD Bioscience), and HLA-DR-FITC (BD Bioscience). Cells were washed with PBS+1% BSA (Wash Buffer), resuspended in Wash Buffer+ 1% human gamma globulin, incubated for 15min at 4C, stained with 25μl of antibody mix or controls for 30min at 4C, washed twice, fixed with 0.5% formaldehyde, and stored at 4C overnight.
Data were acquired on a customized BD LSR II Flow cytometer and analyzed using Flowjo Software (Treestar Inc). Cells were gated on a scatter plot to remove debris, then sequentially gated for CD45 (to define total leukocytes), CD3 (to define T-cells), and CD4 or CD8. CD38 and HLA-DR gates were set using Fluorescent-Minus-One controls for each marker on a PBMC sample then applied to PBMC and gut samples from the same subject.
For each outcome measure, the results from any two given sites were compared across all study participants using the paired Wilcoxon signed rank test. For each site, pairs of outcome measures were correlated using the Pearson correlation test (data meeting the D’Agostino and Pearson omnibus normality test) or Spearman correlation test (all other data). Statistics were calculated using GraphPad Prism 5.0.
Of 14 patients who were screened, 13 met study criteria and 8 consented to enter the study. The eight participants (all men) had a median age of 51 years (range: 33–63), median duration of HIV infection of 15 years (11–22), and median CD4 nadir of 219 cells/μl (49–469). At study entry, they had maintained VL<40 copies/ml for 2.8–12 years (median 6.7) and had CD4 counts of 289–1552 cells/μl (median 478). ART regimens included 2 nucleoside reverse transcriptase inhibitors (typically emtricitabine and tenofovir) plus either efavirenz (3 subjects), atazanavir+/-ritonavir (3 subjects), nevirapine+lopinavir/ritonavir (1 subject), or nevirapine+atazanavir/ritonavir+maraviroc (1 subject).
Plasma HIV RNA was undetectable in all study participants using the standard Abbott assay, but was detectable in all participants using our modified high volume assay. Mean plasma viral loads ranged from 0.5–6 copies/ml (median 2.3).
CD4 content was measured by flow cytometry in PBMC and cell suspensions from the four gut sites. The percent of all cells that were CD4+T-cells was lower in all four gut sites (range: 0.2–6.4%, Figure 1A) compared to PBMC (range: 14.3–44.4%, not shown). As a percent of all cells and of T-cells, CD4 content increased from the small to the large bowel (Figure 1A,B).
HIV DNA was measured in PBMC, peripheral CD4+T-cells, and cell suspensions from the four gut sites. To account for significant differences between the PBMC and gut sites in the percent of all cells that were CD4+T-cells, the HIV DNA levels were normalized to CD4+T-cells, as measured by flow cytometry. HIV DNA levels per 106 CD4+T-cells (Figure 2A) were higher in all 4 gut sites compared to blood (p=NS, 0.016, 0.008, 0.008 for duodenum, ileum, colon, and rectum, respectively). The median HIV DNA per 106 CD4+T-cells in the duodenum was 2.8 times higher than that of the blood; fold differences for other sites were 6.5, 6.3, and 9.1, respectively. HIV DNA levels per 106 CD4+T-cells in the blood correlated positively with the mean plasma HIV RNA (Pearson r=0.82, p=0.012; Figure 3A).
HIV RNA was undetectable by ISH in any of the gut sites at 14 days of film exposure. In contrast, unspliced HIV RNA [UsRNA] was detectable by qRT-PCR using samples from PBMC, peripheral CD4+T-cells, and total cells isolated from the four gut sites. For participants A185 and A186, the yield of cells from the colon and rectum was insufficient for testing for HIV RNA. UsRNA was usually detectable in the gut, whereas MsRNAs were rarely detectable (Table 1).
When normalized for CD4+T-cell content (Figure 2B), the mean and median UsRNA levels were highest in the ileum and were greater in all 4 gut sites compared to blood (median fold difference=1.8, 10.2, 1.6, 2.4 for duodenum, ileum, colon, rectum; mean fold difference=5.7, 12.5, 4.6, 3.8), though results were significant only for comparison of ileum to blood (p=0.016).
To approximate the transcriptional activity per HIV-infected cell, we divided the HIV UsRNA per 106 cells (normalized by μg RNA) by the HIV DNA per 106 cells (normalized by μg DNA) to determine the ratio of UsRNA to DNA. Within the gut, the median HIV RNA/DNA ratio was highest in the terminal ileum (Figure 4). There was a trend towards a higher HIV RNA/DNA ratio in PBMC compared to either site of large bowel (p=0.062 for comparison of PBMC to right colon; p=0.094 for PBMC to rectum) and a trend towards a higher HIV RNA/DNA ratio in the duodenum compared to either site of large bowel (p=0.094 for comparison of duodenum to right colon; p=0.062 for duodenum to rectum).
Though immune “activation” markers have not been clearly defined for the gut, we measured CD38 and HLA-DR because these markers have been used to distinguish activated T cells in the peripheral blood and have been extensively correlated with poor prognosis [20–31]. T-cell activation was higher in all four gut sites compared to blood (Figure 5).
The mean plasma HIV RNA level correlated positively with the percentage of peripheral CD4+T-cells that were CD38+ (Pearson r=0.73, p=0.039; Figure 3B) and tended to correlate with the percent that were CD38+HLA-DR+ (Spearman r=0.71, p=0.058; Figure 3C). For the PBMC, comparison of HIV DNA (per total cells or CD4+T-cells) to any measure of immune activation (CD38+HLA-DR+, total CD38+, or total HLA-DR+, on CD4+ or CD8+T cells) yielded a correlation coefficient (r) that was invariably positive (Table 2), though p values were not always significant. In contrast, for all gut sites, comparison of either measure of HIV DNA to any measure of immune activation yielded an r value that was always negative or neutral (Table 2).
In HIV-infected adults with a history of sustained viral suppression on ART, we measured levels of HIV RNA, DNA, RNA/DNA ratios, CD4+T-cells, and T-cell activation in blood and four different regions of gut. HIV DNA levels per CD4+T-cell were on average 5–10 fold higher in the gut compared to the peripheral blood, in agreement with Chun, et al . If the gut CD4+T-cells have on average five times more HIV DNA than CD4+T-cells in the blood (and possibly the rest of the body), and the gut contains 50–80% of total body CD4+T-cells, then the gut harbors 83% to >95% of all infected cells in the body. Based on the average (across sites and participants) HIV DNA level of 20,000 DNA copies per 106 CD4+T-cells, an estimate of 1.2×1011 total body CD4+T-cells, and an estimate of 50% of total body CD4+T-cells residing in the gut, the gut contains 1.2×109 infected CD4+T-cells after a median of 7 years of suppressive ART. If 1 in 100 infected cells carries replication-competent HIV, the gut alone contains 1.2×107 latently-infected HIV genomes. This estimate exceeds by an order of magnitude earlier estimates of the latent reservoir size in PBMC and central lymphoid tissues [32, 33].
It is unclear why the gut harbors such a disproportionate concentration of infected CD4+T-cells. Possible explanations include greater rates of initial infection (especially in the rectum, which may be a site of transmission), differences in the content of memory CD4+ T cell subpopulations (such as transitional memory cells) that may harbor more HIV DNA , higher rates of replication of integrated HIV DNA by cell division, increased establishment of latent infection, reduced reactivation from latency, slower clearance, trafficking from the blood, or ongoing replication (the latter was suggested by Chun et al ).
While ISH for HIV RNA was negative in all gut sites, UsRNA was detectable in the majority of gut samples using qRT-PCR. ISH can detect as few as 2–5 genomic HIV equivalents per cell but has a detection limit of 104 copies/g tissue for dispersed HIV RNA, whereas PCR provides very sensitive detection of HIV RNA pooled from many cells but cannot discriminate RNA produced from few or many cells. By qRT-PCR, UsRNA levels per CD4+T-cell were higher in all four gut sites compared to blood. It is unclear whether this RNA represents reactivation of latently-infected cells, stable chronically-infected cells, or cells newly-infected as a result of ongoing replication. Based on the absence of HIV RNA+ cells by ISH, we suspect that productively-infected cells in the gut, if present, must be very infrequent or exhibit very attenuated production, and that the HIV RNA detected by qRT-PCR represents modest viral transcription distributed across many HIV DNA+ cells. This assumption is in agreement with the observation that MsRNAs, which may be found in latently-infected cells but are expressed at high level in productive infection , were rarely detected in the gut.
The RNA/DNA ratio peaked in the ileum, where the median ratio tended to be higher than that of the PBMC, suggesting that this site may have a greater ratio of productive to latent infection and should be sampled in studies aimed at detecting ongoing replication. In contrast, in the large bowel, the median HIV RNA/DNA ratio tended to be lower than in the PBMC, suggesting that more of these cells behave as if they are “latently” infected. However, since most gut lymphocytes display markers of T-cell activation, “latent” infection of these cells may differ from the classic latent infection in the blood, which was originally described in resting CD4+T-cells. Differences between gut sites could reflect differences in T-cell activation, memory CD4+T-cell subsets, or the proportion of lymphocytes from lymphoid aggregates.
The infrequent detection of MsRNA and the trend towards lower HIV RNA/DNA ratio in most gut sites suggest lower levels of HIV transcription. Given the high degree of T-cell activation in the gut, it is very surprising that gut cells have such low levels of HIV transcription, suggesting that they are hyporesponsive to activating stimuli, or that T-cell activation has different consequences (or activation markers have different meanings) in the gut compared to the blood. Gut lymphocytes may have a reduced ability to respond to antigens and may resemble more immunotolerant or “anergic” T-cells. Previous reports have shown that CD4+T-cell tolerance and anergy can be caused by epigenetic modification [35–37]. Given that epigenetic modification of the LTR has also been implicated as a feature of latent infection with HIV [38–45], it is tempting to hypothesize that the unique environment of the gut favors both induction of CD4+T-cell tolerance and HIV latency through epigenetic modification. It is not clear whether these cells would respond to therapies that may reduce latently-infected cells in the blood.
Whereas HIV DNA levels in PBMC tended to correlate positively with T-cell activation, in the gut, we found a surprising trend towards a negative correlation between HIV DNA and T-cell activation. Immune activation could have divergent effects on HIV infection. Systemic immune activation may increase the susceptibility of CD4+T-cells to infection, cause replication of proviral DNA by cell division, or serve as a marker for spread of infection, thus explaining the positive correlation between activation and HIV DNA seen in the blood. On the other hand, activation of HIV-specific T-cells can lead to death of virally-infected cells, and HIV-nonspecific activation can reduce the number of susceptible target cells (by apoptosis) or lead to reactivation and clearance of latently infected cells, thus explaining the negative correlation seen in the gut. If verified, the opposing directions for the correlations seen in the gut and the peripheral blood further suggest that activation may have different consequences for HIV persistence in these two sites.
Potential limitations of the study should be noted. First, the number of participants was relatively small, thus limiting generalizability and the power to detect small differences. Second, even with multiple biopsies, there remains the possibility of insufficient sampling. Previous studies have shown that HIV DNA and RNA can be reproducibly quantified from a single endoscopic biopsy [5, 46], and we pooled 6 to 9 biopsies from each site. While in situ studies confirmed the presence of lymphoid aggregates (in 50% of ileal biopsies), additional sampling error may have been introduced by the tissue digestion, which resulted in some cell loss and death (AARD staining showed that 75–80% of gut cells were viable). Third, the normalization per CD4 cell assumes that all of the HIV is in CD4 cells. Finally, the PCR detection methods, while sensitive, do not overcome confounding effects of sampling that occur when target nucleic acids are present at low copy numbers, so that Poissonian effects have a greater influence on results.
Nevertheless, the findings here confirm and extend the important role for GALT as a reservoir for HIV in patients on suppressive ART. Additional studies are needed to better define and distinguish the modes of viral persistence in blood and different regions of gut, and to investigate whether site-specific differences result in different responses to therapies designed to “re-activate” HIV from latently-infected cells.
We thank the following: 1) the study participants; 2) PLUS staff members Michele Downing and Marc Gould; 3) VA study nurses Sandra Charles and Linda Adams; 4) members of the SFVAMC GI Endoscopy Unit, the UCSF Core Immunology Lab, and the San Francisco Department of Public Health; and 5) the NIH AIDS Research Reagent Program. This work was supported in part by the U.S. Department of Veterans Affairs (VA Merit Award [JW/SY]), the National Institute of Health (NIH grants P30-AI027763 [SY], NS051145 [JW/SY] and T32 AI60530 [DH/SY]), and the Swiss National Science Foundation (3100A0-112670 [MF] and 324730-130865 [HFG]).
Sources of financial support include:
No author has a commercial or other association that may pose a conflict of interest.
Since completion of the study, Sara Gianella’s affiliation has changed to the University of California, San Diego (UCSD).
The data in this paper has not been presented at any prior meeting, though it will be presented at the 2010 CROI in February 2010 in San Francisco, CA, USA.
Registry: ClinicalTrials.gov, PLUS1