Five PBMC cohorts were assayed in this study. The groups included normal anonymous blood bank donors, and anonymously labeled patient samples from four classes of HIV-1 seropositive individuals [i.e. patients with high CD4+ T cell count and low viral load (class I), high CD4+ T cell count and high viral load (class II), low CD4+ T cell count and low viral load (class III), and low CD4+ T cell count and high viral load (class IV) (Figure 1A)]. These four classifications of HIV-1 individuals are operationally defined; other ways to stratify patients are possible and merit additional consideration. Nevertheless, small RNAs were extracted from the phlebotomized PBMC samples, and the expression of 327 well-characterized human cellular miRNAs was analyzed using miRNA microarrays as previously described 21. En toto, 12 normal and 36 discrete patient PBMCs were characterized by microarray miRNA profiling.

Primary PBMC samples are expected to show some degree of individual-to-individual variability. To analyze the raw miRNA readouts, we applied two levels of filtering. First, we considered only those miRNAs that were at least two fold or more changed (either up or down) when compared to normal controls. Second, we discarded miRNA changes that were not replicated in more than 50% of the patients in any of the four different classes. When these two filters were applied to the 327 miRNA readouts, 62 miRNAs satisfied both criteria (Figure 1B). The frequencies of these 62 miRNA changes were then compared between class I, II, III, and IV patients using JMP software and BRB array tools (see Materials and Methods). The resulting in silico clustering patterns indicated a closer relatedness in the frequencies of miRNA changes between class II and class III patients; and between class I and class IV patients (Figure 1B). It is unclear at this juncture what these relationships mean biologically.

Of the 62 frequently-changed miRNAs in the four classes of patients, 59 were down regulated while 3 were up regulated when compared to normal PBMCs (Figure 2). As expected, some polycistronic miRNA clusters such as miR-451 and miR-144; and miR-23a, miR-27a, and miR-24 were down regulated simultaneously. In figure 2, we show an example of the typical data points graphed from the microarray results for the 62 miRNAs from the 9 class IV patients. Similar patterns of mostly down regulated miRNAs were also observed for the other three patient classes (data not shown).

Since the vast majority of miRNAs were down regulated, we next asked whether these 59 miRNAs segregated into specific patterns (Figure 3). In parsing the results, we noted certain "signatures". For example, the down regulation of 14 mRNAs was specific to class IV, but was absent from class I, II or III; and the changes in 4 other miRNAs (hsa-miR-143, hsa-miR-199a, hsa-miR30e-3p, hsa-miR335) were unique to class I, but not observed in class II, III, or IV (Figure 3A). 8 other miRNAs were changed in both class I and IV patients, but not in class II or III patients (Figure 3B); while a further 8 miRNAs (hsa-let-7a, hsa-miR-1, hsa-miR-106b, hsa-miR-20a, hsa-miR-25, hsa-miR-29a, hsa-34b, and hsa-miR-520b) were changed in class I, II, and IV patients, but were absent from class III patients (Figure 3C). Lastly, 12 miRNA changes were present in all four classes of patients (Figure 3D). These patterns suggest class-specific "signatures" that plausibly correlate stage-specific miRNA alterations with the in vivo course of HIV-1 infection.

In seropositive individuals, HIV-1 infects only a very small fraction of the circulating CD4+ T – cells. Thus, most of the PBMCs from our 36 patients (Figure 1A) are not infected by virus. The observed miRNA changes are likely indirect bystander results from systemic changes in activation status or cytokine levels in the infected individuals. To ask how the changes in patient miRNAs correlate with those seen from direct viral infection, we compared our 62 frequently changed miRNAs to those observed from cultured primary PBMCs that were infected ex vivo with HIV-1 pNL4-3. While there was some overlap between "Patients" miRNAs and "Infected PBMC" miRNAs, 50% or greater of the miRNAs in the two sets were discordant (Figure 4), indicating that a significant portion of the "Patients" changes could not be accounted by direct viral infection.

We next queried how "Patients" changes might resemble primary PBMCs treated with an activating stimulus (anti-CD3; Figure 4) or an inactivating cytokine (IL-10; Figure 5). In PBMCs treated with anti-CD3, 48 miRNAs changes were seen. Amongst these 48 miRNAs, 31 (64%) overlapped with the miRNAs frequently changed in "Patients" (Figure 4). Interestingly, all of the down regulated miRNAs shared between "Infected PBMC" and "anti-CD3" treated PBMCs were also down regulated in the "Patients" (Figure 4). By comparison, IL-10 treated PBMCs showed only 18 miRNA changes (Figure 5), and only a single (6%) miRNA overlapped with the "Patients" (Figure 5). These results suggest that the state of in vivo HIV-1 patient PBMCs, as profiled by miRNAs, is more closely modeled by anti-CD3 activation, rather than IL-10 inactivation.

MiRNA expression is cell-type specific 23. HIV-1 infection in vivo is expected to exert physiologic effects on T-cell function which could be reflected in significant miRNA changes. Elsewhere, miR-223, mR-150, miR-146b, miR-16, and miR-191 have been described to be highly expressed in human T-cells 2425. We wondered next whether these abundant T-cell miRNAs could be dysregulated in our patient samples. In our data set, the five T-cell abundant miRNAs showed class specific presentations; and, on average, each was down regulated by 3 to 9 fold (figure 6). Thus in vivo HIV-1 infection, in all classes of patients, has sufficient impact to affect significantly the levels of even highly expressed miRNAs. These abundantly expressed T-cell miRNAs are anticipated to provide important biological functions which would be altered accordingly in infected versus uninfected individuals.

Normal and human immunodeficiency virus-infected patient PBMCs were obtained from the NIH blood bank and Saint Michael's Medical Center. The study protocol was approved by the St. Michael's Medical Center's Institutional Review Board. Written informed consent was obtained from each subject. The IRB approval letter and the signed, informed consents are available for review. Plasma viral loads were quantified by the Bayer SIV bDNA assay (Bayer Reference Testing Laboratory, Emeryville, CA) 39. Peripheral blood CD4⁺ T-cell concentrations were quantified using standard techniques, as previously described 40. PBMCs were isolated using standard Ficoll separation procedure. Ficoll-purified PBMCs were directly lysed for RNA isolation or stored in liquid nitrogen.

RNAs with a cutoff size < 200 nts were hybridized on a microarray printed with 327 probes complementary to mature miRNAs. The probe design and the experimental procedures are the same as previously described 41. After hybridization, excessive RNA was removed by washing in 0.1 × SSC. Unhybridized probes were removed using exonuclease I (New England Biolabs) for 3 hours. Since the probe design contains a stretch of thymidine, polyadenylation from the 3'end of the hybridized miRNAs was achieved by addition of biotin-label dATP (Enzo Life Sciences). Detection of the labeled miRNA under the 532 nm wavelength was facilitated by addition of streptavidin-conjugated Alexa-flur-555. Data points collected from GenePix 4000B (Molecular Devices) were exported into BRBarray tools (developed by Richard Simon and Amy Peng Lam; http://linus.nci.nih.gov/BRB-ArrayTools.html) and JMP software (SAS) for further analysis. Microarray signal normalization was performed using the "median-normalization" procedure. This method is applicable for normalizing arrays in which the majority of data points do not change significantly in values. Essentially, the log-intensities of an array and the reference array are normalized to a median value such that the unchanged gene-by-gene difference between the normalized array and the reference array is 0. The linearity of the microarray readouts has been previously validated using quantitative RT-PCR assays.