Cerebrospinal fluid cytokine profiles predict risk of early mortality and immune reconstitution inflammatory syndrome in HIV-associated cryptococcal meningitis

Abstract

Understanding the host immune response during cryptococcal meningitis (CM) is of critical importance for the development of immunomodulatory therapies. We profiled the cerebrospinal fluid (CSF) immune-response in ninety patients with HIV-associated CM, and examined associations between immune phenotype and clinical outcome. CSF cytokine, chemokine, and macrophage activation marker concentrations were assayed at disease presentation, and associations between these parameters and microbiological and clinical outcomes were examined using principal component analysis (PCA). PCA demonstrated a co-correlated CSF cytokine and chemokine response consisting primarily of Th1, Th2, and Th17-type cytokines. The presence of this CSF cytokine response was associated with evidence of increased macrophage activation, more rapid clearance of Cryptococci from CSF, and survival at 2 weeks. The key components of this protective immune-response were interleukin (IL)-6 and interferon-γ, IL-4, IL-10 and IL-17 levels also made a modest positive contribution to the PC1 score. A second component of co-correlated chemokines was identified by PCA, consisting primarily of monocyte chemotactic protein-1 (MCP-1) and macrophage inflammatory protein-1α (MIP-1α). High CSF chemokine concentrations were associated with low peripheral CD4 cell counts and CSF lymphocyte counts and were predictive of immune reconstitution inflammatory syndrome (IRIS). In conclusion CSF cytokine and chemokine profiles predict risk of early mortality and IRIS in HIV-associated CM. We speculate that the presence of even minimal Cryptococcus-specific Th1-type CD4+ T-cell responses lead to increased recruitment of circulating lymphocytes and monocytes into the central nervous system (CNS), more effective activation of CNS macrophages and microglial cells, and faster organism clearance; while high CNS chemokine levels may predispose to over recruitment or inappropriate recruitment of immune cells to the CNS and IRIS following peripheral immune reconstitution with ART. These results provide a rational basis for future studies of immune modulation in CM, and demonstrate the potential of baseline immune profiling to identify CM patients most at risk of mortality and subsequent IRIS.

Fig 1. Baseline cerebrospinal fluid cytokine and chemokine concentrations in a cohort of ninety patients with HIV-associated cryptococcal meningitis.

Baseline concentrations of the thirteen cytokines and chemokines measured and detected in the CSF of patients with HIV-associated cryptococcal meningitis are shown. Cytokine and chemokine concentrations were log transformed to approximate normal distributions. Filled bars are the inter-quartile range with a line at the median, with error bars to the 10^th and 90^th centiles. IL-12, IL-21, IL-22 and IL-23 concentrations were below the limit of detection in the majority of cases.

Fig 2. Principal component analysis scores and weightings.

Figure (A) shows the variance of each of the principal component scores amongst the 90 patients studied. Figure (B) shows the proportion of variance in the dataset explained by each of the principal components. The majority of the variance in the dataset was explained by the first two principal components, which were used to explore associations between CNS immune responses and outcome. Contribution of each variable to the variance between patients in PC1 and PC2 is shown in the heat map. PC1 was composed primarily of pro-inflammatory cytokines IL-6, IFNγ and the chemokine IL-8. PC2 contained a heavy negative weighting of chemokines MCP-1, MIP-1α and the cytokine GM-CSF. Figure (C) shows the associations between PC1 and baseline CSF IL-6 concentrations, and PC2 and baseline CSF MCP-1 concentrations. See S2 Fig for more detailed PC1 and PC2 associations.

Fig 3. Relationships between cerebrospinal fluid interferon-γ, IL-4, IL-10 and IL-17 concentrations, and their associations with disease severity and outcome.

Associations between baseline IL-4 (top row), IL-10 (middle row) and IL-17 concentrations (bottom row), and baseline IFNγ concentrations, baseline fungal burden (quantitative cryptococcal culture, QCC), rate of clearance of infection (early fungicidal activity, EFA) and 2 week mortality. Best-fit regression lines are shown with 95% confidence intervals. Box plots show the median and extend to the inter-quartile range, with whiskers denoting minimum and maximum values. P-values derived from Student’s t-tests are shown for the mortality associations (both unadjusted and adjusted for family wise error rate using permutation testing). All EFA and mortality associations were adjusted for treatment group. Pearson’s correlation coefficients and significance levels were: IL-4/IFNγ r = 0.57 p<0.001, IL-4/QCC r = -0.23 p = 0.03, IL-4/EFA r = -0.21 p = 0.1. IL-10/IFNγ r = 0.64 p<0.001, IL-10/QCC r = -0.30 p = 0.005, IL-10/EFA r = -0.15 p = 0.2. IL-17/IFNγ r = 0.47 p<0.001, IL-17/QCC r = -0.25 p = 0.02, IL-17/EFA r = -0.29 p = 0.03. These associations all remained significant when controlling for a family wise error rate of 0.05.

Fig 4. Associations between baseline cerebrospinal fluid immune response profiles and clinical outcome.

Associations between baseline PC1 and PC2 scores and 2-week mortality and IRIS. The points represent the mean values, with standard errors denoted by the error bars. The adjusted p value are shown, derived from (in the case of mortality) a linear regression model adjusting for treatment group, CD4⁺-cell count, and the previously described risk factors for mortality, baseline fungal burden and altered mental status; and (in the case of IRIS) a linear regression model adjusting for treatment group, CD4⁺-cell count, and the previously described risk factors for IRIS, baseline fungal burden and CSF white cell count.

Fig 5. Relationship between CSF MCP-1 concentrations and CD4 cell counts, cerebrospinal fluid lymphocyte counts, and immune reconstitution syndrome.

Associations between baseline MCP-1 concentrations and baseline CD4 cell count, concentrations, cerebrospinal fluid (CSF) lymphocyte count and immune reconstitution syndrome (IRIS). The IRIS association was adjusted for treatment group. Best-fit regression lines are shown with 95% confidence intervals. Box plots show the median and extend to the inter-quartile range, with whiskers denoting minimum and maximum values. These associations all remained significant when controlling for a family wise error rate of 0.05.There was no significant difference in CSF MCP-1 concentrations between those who survived (749.1 pg/ml) and those who died (858.4 pg/ml), p = 0.76.