Monocyte Phenotype and IFN-γ-Inducible Cytokine Responses Are Associated with Cryptococcal Immune Reconstitution Inflammatory Syndrome

A third of adults with AIDS and cryptococcal meningitis (CM) develop immune reconstitution inflammatory syndrome (IRIS) after initiating antiretroviral therapy (ART), which is thought to result from exaggerated inflammatory antigen-specific T cell responses. The contribution of monocytes to the immunopathogenesis of cryptococcal IRIS remains unclear. We compared monocyte subset frequencies and immune responses in HIV-infected Ugandans at time of CM diagnosis (IRIS-Baseline) for those who later developed CM-IRIS, controls who did not develop CM-IRIS (Control-Baseline) at CM-IRIS (IRIS-Event), and for controls at a time point matched for ART duration (Control-Event) to understand the association of monocyte distribution and immune responses with cryptococcal IRIS. At baseline, stimulation with IFN-γ ex vivo induced a higher frequency of TNF-α- and IL-6-producing monocytes among those who later developed IRIS. Among participants who developed IRIS, ex vivo IFN-γ stimulation induced higher frequencies of activated monocytes, IL-6+, TNF-α+ classical, and IL-6+ intermediate monocytes compared with controls. In conclusion, we have demonstrated that monocyte subset phenotype and cytokine responses prior to ART are associated with and may be predictive of CM-IRIS. Larger studies to further delineate innate immunological responses and the efficacy of immunomodulatory therapies during cryptococcal IRIS are warranted.


Intracellular Cytokine Staining
Monocyte cytokine responses were determined by intracellular cytokine staining. Following 2 h of stimulation at 37 • C, Brefeldin A (100 µL/mL) (BD) was added and cells were incubated for another 4 h, then overnight at 4 • C, and then washed, fixed, and permeabilized with successive washes in Fluorescence Activated Cell Sorter (FACS) permeabilizing solution (BD Biosciences). The cells were stained with surface antibodies to define phenotype as above and intracellularly with monoclonal antibodies reactive with IL-6 FITC (clone AS12, BD Horizon) and TNF-α APC (clone MAb11, BioLegend). One million events were collected on a FACSCanto II and analyzed using FlowJo version 10.0.5 (TreeStar, Ashland, OR, USA).

Statistical Analysis
Data were analyzed using GraphPad Prism, version 6.0b (GraphPad Software Inc., La Jolla, CA, USA) and Spice, version 5.35 (NIAID, Bethesda, MD, USA). Paired samples at diagnosis and CM-IRIS were compared using the Wilcoxon signed-rank test. Cell phenotype and activation variables were compared between CM-IRIS and controls using the Mann-Whitney rank sum test. Significance was defined as p-value ≤ 0.05. Cytokine expression profiles were compared by permutation analysis, using a 10,000-iteration Monte Carlo simulation model, as described elsewhere [23].

Results
We enrolled 17 HIV-infected adults with CM, 11 of whom developed CM-IRIS and 6 of whom did not develop IRIS at Mulago Hospital in Kampala, Uganda between November 2010 and April 2012. Blood samples were obtained at CM-diagnosis (n = 15), IRIS-Baseline for participants who went on to develop CM-IRIS and Control-Baseline for controls who did not develop IRIS; at CM-IRIS (n = 11), (IRIS-Event), and controls without CM-IRIS at comparable ART duration (n = 6), (Control-Event).
Age, baseline CD4 + and CD8 + T cell counts, cryptococcal antigen (CrAg) titer, plasma HIV RNA (viral load), cerebrospinal fluid (CSF) protein, and white blood cells were similar among participants who later developed CM-IRIS and control participants (Table 1). At meningitis diagnosis, participants who later developed CM-IRIS showed a trend to higher median cryptococcal colony counts in CSF compared with those who did not develop IRIS (5.33 vs. 3.97 log10 colony forming units/mL (p = 0.078). Among the entire clinical cohort (n = 172), the median quantitative CSF culture was 5.1 log10 colony forming units/mL [19].
Among IRIS cases, the median duration from ART initiation to IRIS was 78 days (IQR, 43-144 days). The six control participants' samples were obtained at diagnosis (Control-baseline) and at matched time points of ART duration (Control-event) with CM-IRIS subjects (±14 days), a median of 71 days (IQR, 63-84 days) after ART initiation.
At CM-IRIS (IRIS-event), the median serum CrAg titer had decreased three-fold from CM diagnosis (p = 0.078) (IRIS-baseline) ( Table 1). Among controls, serum CrAg titer did not change significantly 81,920 (Interquartile range (IQR), 5,122-245,760) at CM diagnosis to 46,080 (IQR, 3-163,840; p = 0.681) at the later time point. CD4 + T cell numbers at baseline were low, particularly among those who developed IRIS (Table 1). By three months following initiation of ART, the proportion of CD4 + and CD8 + T cells increased in both populations (Table 1). However, at the time of CM-IRIS, the percent CD4 + T cells was lower among patients experiencing CM-IRIS (p = 0.039), respectively ( Figure S1). In addition to smaller numbers of CD4 + T cells, we also considered that antigen presenting cell numbers and/or function might be contributing to CM-IRIS.

Baseline Monocyte Subsets
To evaluate a role for monocytic cells in the evolution of CM-IRIS, we measured the numbers of total monocytes and of the three phenotypic categories of monocyte subsets. Classical monocytes were the predominant subset at IRIS-Baseline and Control-Baseline (Figure 1d, Figure 2 panel A). Whereas no differences were detected in the intermediate subset, the frequencies of non-classical monocytes, the minority subset, were significantly lower at IRIS-Baseline vs. Control-Baseline (0% vs. 4.1%, p < 0.001) (Figure 2 panel A). The frequency of activated non-classical monocytes, as determined by co-expression of PD-L1 and CD25 (Figure 1e), was significantly lower in CM-IRIS compared with controls, 0% vs. 11% (6-16%), p < 0.001 (Figure 2 panel B).
In contrast to baseline values, at the time of CM-IRIS, monocyte activation was significantly elevated in the total monocyte population from those with CM-IRIS compared with cells from time-matched controls, median frequency, (62% (IQR, 44-76%) vs. 12% (IQR, 5-25%); p < 0.001) ( Figure 2 panel C), but these differences were not confined to one subset. These differences persisted from baseline to IRIS, and did not change significantly in either group. These data suggest that activation of monocytes both precedes and accompanies the development of CM-IRIS and may play a role in the immunopathogenesis of this clinical syndrome.

Cytokine Expression and Responses at Baseline
To characterize functional differences between monocyte populations isolated during IRIS-event and Control-event, we examined constitutive and induced expression of two relevant cytokines, TNF-α and IL-6. At IRIS-baseline, unstimulated monocytes consistently expressed a substantial and higher median frequency of TNF-α (16% (IQR, 2-31%) than Control-baseline, 0.8% (IQR, 0.3-4%); p = 0.039) (Figure 3 panel A). Results with IL-6 expression showed no differences. The TLR4 agonist, LPS, elicited a significant and comparable increase in expression of both cytokines in both groups ( Figure 3 panel A), and significantly increased IL-6 expression compared to unstimulated monocytes among controls ( Figure S2), highlighting the differences in cellular responses between the two groups. In contrast, stimulation with IFN-γ elicited increased expression of TNF-α and IL-6 by monocytes only at IRIS-Baseline. Indeed, following IFN-γ stimulation, the median frequency of activated monocytes expressing TNF-α or IL-6 was higher at IRIS-Baseline compared with Control-Baseline (Table 2). We found IFN-γ-induced IL-6 and TNF-α expression by activated (PD-L1 + CD25 + ) monocytes was higher at IRIS-Baseline compared with Control-Baseline. Thus, spontaneous expression of TNF-α and, to a lesser extent, IL-6 were increased at IRIS-Baseline, as was the response of both cytokines to IFN-γ stimulation. ; (Panel C) displays activated monocytes and subsets at IRIS-baseline and Control-baseline. Activation was measured by surface co-expression of programmed death ligand-1 (PD-L1) and CD25 (Interleukin 2 receptor α chain). Bars represent medians; * represents p < 0.001. Median values were compared using the Mann-Whitney test. Non-classical monocytes were absent at IRIS-baseline. Data is presented as percentage of monocytes but not percentage of total white blood cells, as the laboratory did not perform complete blood counts with every peripheral blood mononuclear cell (PBMC) isolation.

Cytokine Expression and Responses at Baseline
To characterize functional differences between monocyte populations isolated during IRIS-event and Control-event, we examined constitutive and induced expression of two relevant cytokines, TNF-α and IL-6. At IRIS-baseline, unstimulated monocytes consistently expressed a substantial and higher median frequency of TNF-α (16% (IQR, 2-31%) than Control-baseline, 0.8% (IQR, 0.3-4%); p = 0.039) (Figure 3 panel A). Results with IL-6 expression showed no differences. The TLR4 agonist, LPS, elicited a significant and comparable increase in expression of both cytokines in both groups ( Figure  3 panel A), and significantly increased IL-6 expression compared to unstimulated monocytes among controls ( Figure S2), highlighting the differences in cellular responses between the two groups. In ; (Panel C) displays activated monocytes and subsets at IRIS-baseline and Control-baseline. Activation was measured by surface co-expression of programmed death ligand-1 (PD-L1) and CD25 (Interleukin 2 receptor α chain). Bars represent medians; * represents p < 0.001. Median values were compared using the Mann-Whitney test. Non-classical monocytes were absent at IRIS-baseline. Data is presented as percentage of monocytes but not percentage of total white blood cells, as the laboratory did not perform complete blood counts with every peripheral blood mononuclear cell (PBMC) isolation.
We determined the differential expression of these cytokines in the three monocyte subsets, including activated populations. Without stimulation, the frequency of TNF-α + classical monocytes was higher at IRIS-Baseline compared with Control-Baseline (19% (IQR, 12-37%) vs. 5% (IQR, 3-8%); p = 0.039), whereas IL-6 expression did not differ (Figure 3 panel B). IL-6 expression by activated non-classical monocytes did not differ between the two groups, whereas TNF-α expression by the
Consistent with these paired results over time, the frequencies of IFN-γ-induced IL-6 + monocytes at IRIS-Event were also higher compared with Control-Event (Table 3). These differences extended to all monocyte subsets (Table 3, Figure S3). About a quarter to a third of each monocyte subset from IRIS-Event expressed IL-6 as well as TNF-α. Even in the absence of IFN-γ stimulation, the frequencies of TNF-α + and IL-6 + monocytes were significantly elevated among IRIS-Event compared to Control-Event ( Figure S3), further illustrating that monocyte activation and inflammatory mediators are linked to emergence of CM-IRIS. During IRIS-Event, constitutive and/or IFN-γ-induced cytokine responses were greater at IRIS-Event compared with Control-Event. A clear demonstration of IL-6 expression following different experimental conditions is shown in Figure S4.
inflammatory mediators are linked to emergence of CM-IRIS. During IRIS-Event, constitutive and/or IFN-γ-induced cytokine responses were greater at IRIS-Event compared with Control-Event. A clear demonstration of IL-6 expression following different experimental conditions is shown in Figure S4.

Monocyte Population with Intracellular Cytokine Production
Control-Event (n = 6) We assessed the monocyte functional phenotype following IFN-γ stimulation at IRIS-Event compared with Control-Event by measuring the possible combinations of 1-or 2-cytokine responses. At IRIS-Event, the immune responses were characterized by significantly elevated dual-function monocytes producing IL-6-and TNF-α, with a frequency of 39% (IQR, 20-61%) compared to 8% (IQR, 3-10%) at Control-Event, p = 0.045 ( Figure 5). In summary, significantly different monocyte subset distribution, particularly non-classical monocytes, and function were detected at IRIS-Event and IRIS-Baseline compared to Control-Event and Control-Baseline. These differences extended to the distribution of monocyte subsets, particularly non-classical monocytes at baseline, to the degree of monocyte activation (CD25 + PD-L1 + ) at IRIS-Event, and, in particular, to the expression of intracellular proinflammatory cytokines (IL-6 and, less consistently, TNF-α) prior to ART and subsequently at IRIS-Event with and without IFN-γ stimulation.

IRIS-Event (n = 11) p-Value
p < 0.045 Figure 5. Functional Monocyte Cytokine responses to IFN-γ Stimulation. Peripheral blood mononuclear cells collected at IRIS-Event and Control-Event were stimulated with interferon-γ (IFN-γ). The bar chart shows each of the three possible combination responses on the x-axis. The percentage of the total cytokine response is shown on the y-axis, with the filled bar representing the interquartile range and a black line at the median. Responses at Control-Event are shown as blue bars, responses at IRIS-Event are shown as red bars on the graph. Statistically significant differences (p < 0.05) by rank-sum testing are indicated by the plus sign. The pie charts show the representative fractions according to the pie-slice colors shown at the bottom of the bar chart, with color-coded slices indicating the contributions of IL-6 and TNF-α (red), only IL-6 (green), and only TNF-α (blue) to the 2-and 1-function responses. Statistical comparisons of the overall responses by permutation testing are shown in the pie category test result chart where the red arc represents IL-6 expression and the green arc represents TNF-a producing cells. Patients with CM-IRIS had immune responses characterized by significantly elevated dual-function monocytes producing IL-6-and TNF-α compared to controls.
In summary, significantly different monocyte subset distribution, particularly non-classical monocytes, and function were detected at IRIS-Event and IRIS-Baseline compared to Control-Event and Control-Baseline. These differences extended to the distribution of monocyte subsets, particularly non-classical monocytes at baseline, to the degree of monocyte activation (CD25 + PD-L1 + ) at IRIS-Event, and, in particular, to the expression of intracellular proinflammatory cytokines (IL-6 and, less consistently, TNF-α) prior to ART and subsequently at IRIS-Event with and without IFN-γ stimulation.

Discussion
There is a notable paucity of data on innate immune responses occurring during cryptococcal IRIS, and, in particular, the role of monocytes [24,25]. We have shown that monocyte phenotype and function may predict the incidence of cryptococcal IRIS, with the absence of non-classical monocytes at CM diagnosis and upregulation of IFN-γ induced IL-6 expression that precedes and accompanies the development of CM-IRIS.
Monocytes circulate in peripheral blood and migrate into tissues, maturing into phenotypically and functionally distinct macrophages at different anatomical locations. The main functions of monocytes include phagocytosis, antigen presentation, and cytokine production. The three subsets of circulating monocytes are characterized by the differential expression of surface CD14 and CD16: classical (CD14 hi++ CD16 − ), intermediate (CD14 ++ CD16 + ), and non-classical monocytes (CD14 lo+ CD16 ++ ), each with certain unique phenotypic and functional parameters [21]. Classical monocytes can develop into intermediate monocytes and then to non-classical CD14 + CD16 ++ monocytes, also described as patrolling monocytes [21]. The absence of non-classical monocytes prior to ART in patients with cryptococcal meningitis was associated with the incidence of IRIS, similar to paradoxical tuberculosis (TB)-IRIS, where the non-classical monocyte subset was also significantly decreased prior to ART initiation [26]. Non-classical monocytes have been demonstrated in CSF during the immune response to primary cryptococcal meningitis [25], however, it is unclear whether the absence of this monocyte subset was associated with poor cryptococcal antigen clearance with resulting CM-IRIS, as demonstrated in TB-IRIS [26].
During TB-IRIS, dysregulation of monocyte gene expression is functionally related to infection, inflammation, and inappropriate control of complement activation prior to initiating ART, during ART, and during the IRIS event [27,28]. Whether this scenario with an intracellular bacterial infection also occurs during the development of IRIS with an encapsulated fungal pathogen, such as Cryptococcus, is unknown. Barber and colleagues hypothesize that when patients with low CD4 + T cells acquire an opportunistic infection, subsequent paradoxical IRIS results from accumulated monocytes/macrophages harboring abundant organisms producing cytokines for cell signaling (e.g., IL-6), and then undergoing en masse activation upon an effective T cell response to produce excessive innate inflammatory cytokines. This activation coincides with the reconstitution of antigen-specific CD4 + T cells, which deliver previously diminished IFN-γ signaling as ART is initiated [18]. In this study, we provided a surrogate for restored T-cell signals to the monocyte population by in vitro stimulation of PBMCs with IFN-γ. By measuring the intracellular expression of the proinflammatory cytokines IL-6 and TNF-α in monocytes from patients with and without CM-IRIS, we simultaneously characterized measurement and phenotype of cytokine-producing monocytes at the single cell level [29].
We found elevated expression of TNF-α and IL-6 by monocytes at baseline among participants with future CM-IRIS, suggesting that elevation of these innate cytokines may predict future CM-IRIS, as demonstrated for IL-6 in TB or herpes virus-associated IRIS [30][31][32][33][34]. IL-6 induced during microbial infection influences macrophage differentiation and shifts the T helper 1(Th1)/Th2 balance toward the Th2 axis by promoting IL-4 production and Th2 differentiation to inhibit IFN-γ production and inhibition of regulatory T cells [9,35]. Another important pathway in the immune response against fungal infections is Th17, which is associated with increased IFN-γ production by T cells and influenced by IL-6 kinetics [36]. Cryptococcus similarly promotes an imbalance toward a Th2 response by suppressing Th1 responses via IL-12 produced by macrophages [37][38][39]. Together, the influence of IL-6 and cryptococcal infection could promote poor antigen clearance, persistence of residual antigen, and subsequent risk of CM-IRIS, consistent with the elevated serum CrAg titer among those pre-disposed to CM-IRIS.
We found elevated monocyte IL-6 expression among participants with CM-IRIS compared to controls without IRIS. Barber and colleagues have demonstrated in a murine model of experimentally-inducible Mycobacterium avium IRIS that by neutralizing IL-6, C-reactive protein levels diminished, wasting disease was alleviated, and host survival was prolonged [40]. Further, combined blockade of IL-6 and IFN-γ further reduced IRIS pathology even after onset of wasting disease, pointing to the possible role of the IL-6 pathway in the immunopathology of IRIS. Our results are consistent with a few other human studies that have demonstrated elevated IL-6 during CM-IRIS, which show different cytokine responses in CSF and blood [9,13]. We posit that an anti-IL-6 monoclonal antibody (e.g., siltuximab) could be further evaluated in understanding the efficacy of alleviating the pathology and mortality associated with cryptococcal IRIS in the enclosed central nervous system space cognizant of its pleotropic nature. Despite the increased risk of developing cryptococcosis among patients using anti-TNF antagonists [41,42], a therapeutic option to evaluate in larger studies would be to target inhibition of TNF-α for the alleviation of the neurological symptoms associated with CM-IRIS, as suggested in two case reports [43,44]. However, the clinical response may be variable by person.
The frequency of activated monocytes co-expressing PD-L1 and CD25 was increased at CM-IRIS compared with that in matched controls. PD-L1, expressed by various effector cytokines, has the ability to co-stimulate naïve T cells and co-inhibit effector T cells through the programmed cell death protein 1 (PD-1) receptor, and is expressed on most activated immune cells [45,46]. IFN-γ induces monocytes to express TNF-α and IL-6, both of which were elevated during CM-IRIS compared to CM diagnosis and to controls without CM-IRIS in response to IFN-γ stimulation. Thus, increased monocyte activation and restored cytokine responses to IFN-γ signaling from recovering CD4 + T cells appear to coincide with the timing of CM-IRIS, consistent with the prior proposed paradigm [9,18].
Taken together, elevated frequencies of activated monocytes, our previously published data showing the presence of proinflammatory intermediate monocytes in CSF [47], the elevation of IL-6 during CM-IRIS [9,13], and our data showing increased monocyte cytokines at baseline suggest monocyte dysregulation associated with augmented disease. Moreover, data showing increased CCL2 and CCL3 chemokines in the CSF compared with blood [48], and the association between baseline CCL2, CCL3, and GM-CSF expression in CSF and subsequent IRIS [49], suggest enhanced trafficking of activated monocytes from blood into the CSF during CM-IRIS. In addition, within the overall COAT trial, we found that earlier ART initiation was associated with CSF microglial/macrophage activation and increased cellular infiltrate in persons with cryptococcal meningitis [50].
Higher fungal burden or cryptococcal antigen titer at CM diagnosis is known to be a risk factor for poor outcomes and CM-IRIS [9][10][11]. Participants with CM-IRIS in our study had a tendency to have higher cryptococcal quantitative colony counts compared to control participants at baseline.
Despite the small sample size, our findings provide a basis for further examining correlations between monocyte subsets, antigen burden, cytokine responses, and clinical outcomes in larger studies. We did not have a sufficient sample size to evaluate the correlation of our findings with ART timing, as there were only six participants with IRIS in the deferred arm and four participants in the early arm of the COAT trial [19]. We studied monocyte function in peripheral blood and not directly in the CNS, the site of CM-IRIS pathology, thus limiting extrapolation to CSF. Although these data may not mirror compartment-specific responses, they may be reflective of monocytes already exposed and exhausted following IFN-γ signaling in vivo during CM-IRIS, thus we interpret the cytokine responses with some caution. Due to the overlap and a degree of heterogeneity in IL-6 responses among those who developed CM-IRIS and controls who did not develop CM-IRIS, we can only surmise that IL-6 could be a co-predictor of increased risk for CM-IRIS. Nevertheless, building on our previous studies of CM-IRIS using cytokine levels in serum [9] and CSF [13], an instructive feature of this study is the detection of augmented monocyte cytokine responses at the single cell level prior to the development of and during cryptococcal IRIS. There was likely a gender bias in the selection of controls, since we did not match on gender, however, we do not envisage any correlation between gender and cytokine expression during IRIS.
In conclusion, we have demonstrated that monocyte subset phenotype and cytokine responses prior to ART are associated with and may be predictive of CM-IRIS. Larger studies to further delineate innate immunological responses and the efficacy of immunomodulatory therapies during cryptococcal IRIS are warranted.