Study Population

Participants consisted of confirmed IRIS cases and controls selected from an HIV-infected ART-naïve adult (>18 years) cohort who participated in a prospective surveillance study to examine the incidence of IRIS in Sub-Saharan Africa between January 6, 2006 and July 7, 2007 [5]. ART initiation criteria were in accordance with the South Africa National ART Program and included a CD4 count ≤200 cells/mm³ or WHO stage IV AIDS-defining illness irrespective of CD4 count. Patients received scheduled clinical assessments, including at least one pre-treatment assessment, one treatment commencement assessment, and regularly scheduled assessments at weeks 2, 4, 8 and every 3 months thereafter [13]. Enrollment into the nested case-control study required participation in the prospective surveillance cohort, a pre-treatment CD4 count and HIV RNA level, and willingness to provide written informed consent for an additional blood draw and sample storage.

In general, IRIS is clinically defined as a paradoxical clinical worsening due to a subclinical opportunistic pathogen ("unmasking" IRIS) or previously known treated (completed or ongoing) opportunistic pathogen ("paradoxical" IRIS) in the setting of an adequate response to ART [14–16]. For the "unmasking" form of IRIS, a new localized infection was required from a focal inflammatory process (suppurative lymph node, pulmonary infiltrate, positive CSF culture, etc.) in a patient who, prior to ART, exhibited no signs or symptoms of disease and in whom adequate OI screening and clinical assessment had been performed (i.e. negative pre-ART sputum AFBs in the case of "unmasking" pulmonary TB). For organisms for which cultures or diagnostic studies were available (i.e. TB, Cryptococcus) demonstration of the organism or a pathological process characteristic of the organism (i.e. caseous necrosis, granulomatous inflammation) were required. For the "paradoxical" form of IRIS, a patient required the diagnosis and treatment initiation of an OI prior to ART initiation with a positive clinical response. Following ART, the patient experienced a new inflammatory process (worsening lymphadenopathy or suppuration, expansion of Kaposi's lesions, recurrence of meningeal signs and symptoms) at the original or new site of infection accompanied by systemic symptoms (fever, loss of weight, elevated white blood cell count). For all cases, no other identifiable pathogen could be present after thorough diagnostic evaluation.

For this immunological analysis, only confirmed IRIS cases were examined. These consisted of subjects who: 1) exhibited symptoms consistent with an infectious or inflammatory condition while on ART which could not be explained by the expected clinical course of a previously recognized infectious agent or by side effects of therapy, and 2) whose treatment led to a ≥1 log₁₀ drop in HIV RNA at time of IRIS diagnosis, and 3) whose treatment resulted in adequate immune reconstitution 6 months post ART, defined as a ≥1 log₁₀ drop in HIV RNA or a CD4 count equal to or above the pre-treatment baseline value. Peripheral blood sampling was performed at the time of IRIS diagnosis for IRIS cases and at study enrollment for matched controls. Eligible control subjects were matched on ART duration within ± 2 weeks in a 1:1 ratio. The study protocol was reviewed and approved by all participating institutional review boards.

CD4 cell count and HIV RNA level measurements

Longitudinal CD4 and CD8 cell counts were measured as previously described [17]. Plasma samples were assayed for HIV-1 RNA levels using the Amplicor polymerase-chain reaction test (Roche Diagnostics, Switzerland).

Cell isolation, storage, and thawing

Peripheral blood mononuclear cells (PBMCs) were isolated from sodium heparin anticoagulated human blood by density centrifugation (Histopaque, Sigma, Germany). PBMCs were harvested, washed using Hanks Balanced Salt Solution (H9394, Sigma, Germany), and centrifuged at 400 g for 10 minutes. PBMCs were then resuspended in Hanks Balanced Salt Solution and counted using an automated cell counter (Guava Technologies, USA). After counting, 10 million PBMCs/ml were cryopreserved using 1 ml of 10% Dimethyl Sulfoxide (Sigma, Germany) and 90% Fecal Calf Serum (Sigma, Germany) cryopreservation solution. Samples were rate-controlled frozen to -80°C, stored overnight, and then transferred into liquid nitrogen at -180°C until analysis.

Culture medium (R10) was prepared by adding HEPES buffer (1 M), penicillin-streptomycin (1%), sodium pyruvate (100 mM), L-glutamine (200 mM) and heat-inactivated fetal bovine serum (10%) (Sigma, USA) to RPMI 1640 medium (Sigma, Germany). PBMC vials were thawed in a 37°C water bath and 1 ml of R10 with 50 U/ml benzonase (Novagen, Denmark) was added, after which cells were resuspended to 8 ml using R10. Cells were centrifuged for 10 minutes at 250 g, resuspended and washed again using R10. Cells were then counted (Guava Technologies, USA) and rested overnight in a 37°C 5% CO₂ atmosphere.

Cell surface staining and measurements

Following overnight rest, 1.0–2.0 × 10⁶ viable PBMC were stained as previously described[18] using the following optimally titrated monoclonal antibody conjugates in a single tube 9-color panel: HLA-DR APC-Cy7 (L243); CD38 PE Quantibrite (HB7); CD8 Alexa700 (RPA-T8); CD27 APC (L128); CD57 FITC (HNK-1); CD3 AmCyan (SK7); CD4 PerCP Cy5.5 (SK3) (BD Biosciences, CA, USA); CD45RO ECD (UCHL1, Beckman Coulter, France) and vAmine (LIVE/DEAD^® Fixable Violet Dead Cell Stain, Invitrogen, USA). An additional 9-color tube consisted of isotype controls for HLA-DR and CD38 using mouse IgG1 APC-Cy7 isotype (X40) and mouse IgG1 PE isotype (X40) (BD Biosciences, USA).

All data were immediately acquired using an LSRII flow cytometer (BD Biosciences, USA) within 4 hours of staining. Between 500,000 and 2 million events were acquired per sample. Compensation was performed daily digitally using FACSDiva software (version 5.0, BD Biosciences, USA) by staining compensation beads (anti-mouse Igκ, BD Biosciences, USA) with the antibodies described with the exception of vAmine which was substituted with CD14 Pacific Blue (BD Biosciences, USA). PMT voltages were checked daily through gating of peak 8 beads (Spherotech, Il, USA). CD38 PE quantitation was performed using CD38 PE Quantibrite and Quantibrite PE Beads (BD Biosciences, USA) according to manufacturer recommended procedures [19].

List mode data were analyzed with FlowJo v8.5 (Treestar). Initial singlet gating used a forward scatter width (FSC-W) versus height (FSC-H) plot. Events were then gated through a CD3⁺ versus vAmine to identify a minimum of 10,000 viable CD3⁺ lymphocytes and sequentially gated on CD4⁺ and CD8⁺ populations. Maturational subsets were defined for CD4⁺ and CD8⁺ populations as follows: naïve (N) (CD45RO^-CD27⁺), central memory (CM) (CD45RO⁺CD27⁺), effector memory (EM) (Effector Memory; CD45RO⁺CD27^-), E (Effector; CD45RO⁺CD57⁺), & TE (Terminal Effector; CD45RO^-CD57⁺) [20, 21]. Focusing on cells co-expressing CD38 and HLA-DR antigens, cellular activation was measured quantitatively by the number of CD38 PE antibodies bound per cell (ABC), which reflects the density of CD38 antigen on the surface of activated T cells [19]. Cellular activation was measured for each respective CD4⁺ & CD8⁺ maturational subset. Figure 1 is an example of gating methods employed for all samples.

Statistical analyses

Group comparison of baseline clinical and immunological data were performed using the Wilcoxon rank sum and Fisher's exact tests (Stata v10, College Station, USA), where appropriate, with an a priori definition of <0.05 considered significant. Group comparisons (IRIS vs. non-IRIS controls) of CD4⁺ and CD8⁺ T population subset percentages were performed using the Wilcoxon-Mann-Whitney test. The activation markers CD38 and HLA-DR were gated from CD4⁺ and CD8⁺ cells their subsets in which quadrant gates, using an isotype control, were used to define positive and negative populations. For the measurement of CD38 PE ABCs, a minimum of 50 CD38⁺ HLA-DR⁺ events were required within each T cell population maturational subset for statistical analysis.

Cohort characteristics and clinical outcomes

Immunovirologic outcomes

Of the 44 cases, 22 IRIS cases and 22 matched controls met enrollment criteria for the immunological nested case-control study. Reasons for failure to enroll in the immunological study were prisoner status (n = 2), patient refusal (n = 3), missing baseline or follow-up CD4 count or viral load (n = 4), and failure to enroll patient within two weeks of IRIS identification (n = 13). The median interval of time between ART initiation and the development of IRIS was 38 days (Interquartile range (IQR): 24, 56), with immunological sampling occurring one week later after IRIS diagnosis and clinical evaluations were complete (45 days, IQR: 27–59). Cases demonstrated significantly lower baseline CD4 cell counts compared to controls (79 versus 142 cells/mm³, respectively, p = 0.02) with similar baseline HIV RNA levels (Table 1). This observation persisted at IRIS diagnosis and control enrollment (183 versus 263 cells/mm³, respectively, p = 0.05) with a continued, but nonsignificant trend by after 6 months of ART (161 versus 277, respectively, p = 0.10). Absolute and percentage changes in CD4 cell count at IRIS diagnosis (for cases) or control enrollment (for matched controls) and at 6 months follow-up were not significant between cases or controls (data not shown). At IRIS diagnosis or control enrollment, controls demonstrated higher absolute CD8 cell counts than cases (875 versus 578 cells/mm³, p = 0.02), but no difference in CD8 cell percentage (57 versus 55%, p = 0.15) or CD4:CD8 cell ratio (0.32 versus 0.28, p = 0.64). Response to ART was similar between groups, with no differences in HIV RNA levels at enrollment or at 6 months.

CD4 and CD8 T cell subset phenotyping

Within CD8⁺ T cells, the overall percentage of CD8⁺ T cells was similar between groups, but subset maturational profiles were heterogeneous. Controls again exhibited significantly higher percentages of central memory (CM) compared to cases (13.9 versus 7.81%, p = 0.01) and exhibited more of an effector (E) phenotype compared to cases (13.2 versus 8.8%, p = 0.04). On the other hand, cases demonstrated an increase in terminal effectors (TE) compared to controls (28.8 versus 15.1%, p = 0.05).

CD4 and CD8 T cell activation profiles

T cell activation, as measured by the number of CD38 PE ABCs on respective CD38⁺ HLA-DR⁺ T cell subsets, demonstrated significant differences between IRIS cases and matched controls (Figure 2). Within CD8⁺ subsets, but not within CD4 subsets, cases tended to have more activated CD8⁺ T cells, demonstrating significantly higher levels of activated effector memory (CD27^-CD45RO⁺), terminal effectors (CD57⁺CD45RO^-), and effector (CD57⁺CD45RO⁺) subsets (p = 0.04, 0.02, and 0.02, respectively) and tended to have higher activation levels of central memory subsets (CD27⁺CD45RO⁺) (p = 0.10).