In this large, prospective cohort of HIV-infected persons with AIDS and CM in Sub-Saharan Africa, a paucity of CSF inflammation at the time of initial CM diagnosis was associated with subsequent development of IRIS. This finding that persons who later develop IRIS had less initial CSF inflammation compared to those without IRIS suggests that persons at risk for IRIS had ineffectual protective immune responses, despite similar CD4+ T-cell counts and cryptococcal burdens.
Specifically, compared with non-IRIS patients, persons with future IRIS had lower CSF levels of the Th1
-associated cytokine IFN-γ and of the pro-inflammatory cytokines IL-6, IL-8, and TNF-α at the time of initial CM diagnosis. These attenuated cytokine responses were consistent with decreased local inflammation in the CSF and were associated with decreased levels of CSF protein and WBCs. Normally, an effective immune response to cryptococcus requires a Th1
T-cell response for cryptococcal clearance directly by T-cell cytotoxicity or cytokine-enhanced antibody-dependent killing by macrophages [16
]. Thus, persons with diminished Th1
responses would be expected to exhibit greater cryptococcal loads for longer durations. Because cryptococcal antigen persists for months [39
], the initially ineffectual inflammatory response in CSF from patients at risk for IRIS appears to transform in the setting of immune restoration into an exaggerated inflammatory response directed at the persisting antigen burden.
These data advance our understanding of CM-IRIS pathophysiology by characterizing the inflammation in the CSF, at the site of the exaggerated response. A robust Th1 T-cell response was evident in the CSF at the time of IRIS, with a 3-fold increase in IFN-γ compared to initial CM. Also present at elevated levels at the time of IRIS were pro-inflammatory cytokines, such as IL-6 and TNF-α, which are typically produced by macrophages and antigen presenting cells, as well at T-cells. In contrast, IL-17 levels at the time of IRIS were similar to initial levels (P=.099), suggesting that CM-IRIS pathogenesis is not driven by a Th17 response. Also, levels of Th2 cytokines such as IL-4, IL-5, and IL-10 were negligible at the time of IRIS. Overall, our results suggest that pro-inflammatory cytokine responses, including Th1 cytokines, are involved in IRIS pathogenesis.
Chemokines present in CSF at the time of IRIS included CCL2, CCL11, and VEGF. CCL2 (MCP-1), a chemotactic factor for dendritic cells, monocytes, and T-cells, did not appear to contribute to IRIS pathogenesis. Levels of CCL2 (MCP-1) decreased at the time of IRIS compared to initial levels at the time of CM diagnosis and were comparable to HIV-infected persons without IRIS receiving ART [41
]. In contrast, levels of TNF-α were elevated at the time of IRIS. TNF-α induces VEGF-A which increases vascular permeability, stimulates chemotaxis of macrophages and CD4+
memory T-cells [42
], and VEGF has co-stimulatory activity for IFN-γ-secreting Th1
memory T-cells [42
]. Lastly, granulocyte-colony stimulating factor (G-CSF) was elevated in CSF at the time of CM-IRIS. In cryptococcosis, G-CSF levels may correlate with CSF clearance of cryptococcus [38
]. Whereas G-CSF is normally considered a hematopoietic growth factor, G-CSF is also produced by macrophages and increases the innate antifungal activity of neutrophils and macrophages [43
]. Overall, our results revealed increasing CSF inflammation (cytokines, WBCs, and protein) coincident with the development of IRIS. These results differ from those of a pioneering but smaller study from Cape Town in which levels of CSF cytokines did not differ between subjects with CM and CM-IRIS [1
Whether CSF cytokine profiles can be used to clinically risk-stratify HIV-infected patients starting ART remains an open question. Our results suggest that widely available CSF parameters such as protein and WBC are generally informative of IRIS risk, yet we were unable to identify a particular cut point as an ideal diagnostic threshold. However, those with more prominent CSF inflammation (WBC >25 cells/μL, protein >50 mg/dL) at time of initial CM diagnosis only infrequently (25.5%) developed IRIS on ART whereas 71% of those below both thresholds developed IRIS. Thus, high risk persons may be identifiable before starting ART. Potential theoretical strategies to reduce IRIS risk could include more aggressive management of increased intracranial pressure, more potent antifungal therapy, altering the timing of ART, or prophylactic use of anti-inflammatory medications in patients at higher risk of IRIS. However, the association between decreased inflammation and the development of IRIS may create problems in preventative anti-inflammatory therapy. For instance, although TNF-α levels were increased at time of IRIS, low levels of TNF-α at initial CM were associated with increased IRIS risk. Thus, prophylaxis with accessible anti-TNF agents, such as chloroquine or thalidomide, could slow cryptococcal clearance and inadvertently increase the risk of IRIS and death.
We also identified a distinct difference in CSF cytokine profiles between cases of CM-relapse and CM-IRIS at the time of these events. CM-relapse was associated with persistent viable organisms and a lack of inflammation in CSF whereas CM-IRIS showed no organisms but a more robust inflammatory profile. CM-relapse was characterized by the lack of pro-inflammatory cytokines in the CSF similar to the cytokine profile observed with initial CM. However, relapse and IRIS are not always distinct or completely separate. Two of the five cases of CM-relapse later developed culture-negative paradoxical-IRIS. IRIS is an immunologic event that can occur in the presence of live or dead organisms. While the presence of live organisms, as seen in relapse, influences clinical management, in contrast, whether the cryptococcal antigens driving the inflammatory response in IRIS derive from live organisms, dead intact organisms or cellular debris, may likely be inconsequential from an immunologic perspective. Thus, some persons could have IRIS with low-level positive cultures.
The diagnostic value of positive CSF cultures within the first few weeks of ART and the development of IRIS remains unsettled. Within the expected treatment course, 50% of patients treated with amphotericin B in Uganda show positive CSF cultures at 2 weeks [30
], and a minority of CSF cultures can remain positive up through 8–10 weeks [39
]. Clinical considerations in suspected IRIS events early in ART would suggest anti-fungal regimens should be intensified while awaiting CSF culture results to exclude relapse, especially if anti-inflammatory therapies are given. The degree of inflammation in the CSF, evident by increasing CSF WBCs coupled with decreasing CRAG titers at the time of recurrent symptoms, is a useful initial diagnostic tool to differentiate IRIS from relapse. However, culture is essential. Cryptococcal cultures were more informative than CRAG titers alone for distinguishing IRIS from relapse, as 25% of persons with IRIS had negative CSF cultures but a <4-fold decrease in CRAG titer.
Overall, our results indicate that a paucity of initial inflammation at the time of CM diagnosis was associated with increased risk for subsequent development of exaggerated inflammatory responses associated with CM IRIS. Future studies are needed to determine whether our findings can be validated in other populations and to determine predictors of the response to medical treatment of CM-IRIS. In characterizing the pathophysiology of CM-IRIS at the site of inflammation, this study is a first step toward selecting rational therapeutic strategies for CM-IRIS.