Cerebrospinal Fluid Cytokines and Chemokines Involved in Cytotoxic Cell Function and Risk of Acute 14-Day Mortality in Persons with Advanced HIV and Cryptococcal Meningitis

Abstract

Background

The role of the immune response in acute mortality of cryptococcal meningitis remains unclear.

Methods

Cerebrospinal fluid (CSF) from 337 Ugandans with first-episode cryptococcal meningitis was collected. CSF cytokines and chemokines were quantified and compared by 14-day survival, stratification by quartiles, and logistical regression to determine association with acute mortality.

Results

Eighty-four (24.9%) participants died by day 14. Persons who survived to day 14 had higher levels of proinflammatory macrophage inflammatory protein (MIP)-3β and interferon (IFN)-β and cytotoxicity-associated granzyme B and inteferon gamma-induced protein (IP)-10 compared to those who died (P < .05 for each). Logistic regression analysis revealed that per 2-fold increase in proinflammatory interleukin (IL)-6, IL-1α, MIP-1β, MIP-3β, and IFN-β and cytotoxicity-associated IL-12, tumor necrosis factor–α, granzyme-B, and IP-10 CSF concentrations, the risk of acute 14-day mortality decreased. Similar biomarkers were implicated when stratified by quartiles and further identified that lower concentrations of anti-inflammatory IL-10 and IL-13 were associated with 14-day mortality (P < .05 for each).

Conclusions

Proinflammatory and cytotoxicity-associated cytokine and chemokine responses in the CSF decrease the risk of acute 14-day mortality. These data suggest that a cytotoxic immune environment in the CSF could potentially improve acute survival. Further research on cytotoxic cells is crucial to improve understanding of innate and adaptive immune responses in cryptococcal meningitis.

Persons living with advanced human immunodeficiency virus (HIV) are at risk of developing disease from the environmental fungal pathogen Cryptococcus neoformans [1, 2]. Cryptococcosis begins as a localized pulmonary infection, from which the fungus disseminates into the central nervous system (CNS) causing cryptococcal meningitis [3–5]. Cryptococcal meningitis is the most common cause of adult meningitis in sub-Saharan Africa [6]. Despite antifungal and antiretroviral therapies, 1-year mortality remains high at >40% [7]. Therefore, the need to improve morbidity and mortality persists.

Due to underlying advanced HIV, dysregulated CD4⁺ T-cell responses contribute to uncontrolled fungal growth in the lung parenchyma, resulting in fungal dissemination to the CNS [8]. During C. neoformans infection, CD4⁺ T cells provide crucial T-helper 1 (Th1) inflammatory cytokine and chemokine signals, such as interferon (IFN)-γ, tumor necrosis factor (TNF)-α, and macrophage inflammatory protein (MIP)-1α, to surrounding innate immune cells and stromal cells [9–11]. This in turn polarizes macrophages toward an inflammatory M1 activation status and recruits neutrophils to promote inflammation, fungal phagocytosis, and killing [12–15].

Conversely, T-helper 2 (Th2) CD4⁺ T cells provide anti-inflammatory and allergy-mediated stimuli, through the secretion of cytokines such as interleukin (IL)-4, IL-5, and IL-13, which promote eosinophil recruitment and alternatively activated M2 macrophages poised for wound healing [15]. Unfortunately, Th2 responses are detrimental in C. neoformans infection and are associated with poor clinical outcomes in murine models [16–19]. A balance between Th1 and Th2 cytokine signaling is necessary for an optimal antifungal immune response to clear the infection with minimal damage to host tissues and neurological sequela, as postulated in the damage response framework of cryptococcal meningitis [20, 21]. Lastly, studies on the T-helper 17 (Th17) CD4⁺ T-cell signaling pathways have yielded conflicting results of protection from disease and enhanced fungal dissemination in murine models and human cytokine studies [10, 17, 22, 23]. Therefore, the role of Th17-mediated immunity in cryptococcal meningitis remains to be clarified.

Studies investigating cytokine and chemokine milieu in the cerebrospinal fluid (CSF) can provide crucial information regarding the immune environment in the CNS during infection and association with acute clinical outcome. Acute 14-day survival is a particularly relevant measure in cryptococcal meningitis. First, during the first 14 days of hospitalization, patients receive an intravenous and oral antifungal induction therapy regimen with penetration into the CNS and therapeutic lumbar punctures to remove excess CSF to manage raised intracranial pressure [1, 24–26]. Other supportive interventions, including electrolyte supplementation and nutritional management, are provided while in hospital [1, 26]. Elevated CSF fungal burden, altered mental status, and early fungicidal activity are independently associated with 14-day mortality [24, 27]. Therefore, investigating acute mortality is highly clinically relevant to assess. Lastly, during the first 14 days of hospitalization, mortality is most often meningitis-related, whereas longer-term, other advanced HIV disease–related causes of death increasingly can occur [24].

Early studies examining acute survival have shown that acute phase reactant molecules (eg, IL-6, IL-8), proinflammatory Th1 and Th17 cytokines (eg, IFN-γ, TNF-α, IL-17A), and anti-inflammatory cytokines (eg, IL-10) in the CSF are important in acute survival [23, 28, 29]. However, previous studies were restricted by a limited cohort size and the number of biomarkers investigated. There is a need to perform additional studies investigating cytokine and chemokine levels in the CSF in relation to acute survival in a large cohort of participants to allow for statistical power to perform a risk assessment. These data could provide insight into the utility of CSF biomarker quantification during treatment and their impact on clinical prognosis, and further our understanding of the CSF immune environment.

In the present study, we analyze baseline CSF in a cohort of 337 participants with first-episode cryptococcal meningitis to determine the association between cytokine and chemokine concentrations and acute 14-day mortality. We postulate that elevated levels of adaptive Th1 inflammatory cytokines and chemokines could be protective from mortality. However, we identify novel findings implicating innate inflammatory and cytotoxicity-associated cytokines and chemokines in acute 14-day mortality.

METHODS

Study Design and Participant Enrollment

CSF samples were collected at time of hospitalization from participants with first-episode cryptococcal meningitis enrolled into the parent clinical trial, which aimed to determine the impact of adjunctive sertraline on the development of neurological sequalae and mortality from cryptococcal meningitis (ASTRO-CM; NCT01802385) [30–34]. Trial sites were located in Uganda at Mulago National Referral Hospital, Kiruddu National Referral Hospital, and Mbarara Regional Referral Hospital. The Mulago Hospital Research Ethics Committee approved the study protocols. All participants or legal guardians completed written informed consent. A detailed description of the eligibility criteria for the parent clinical trial is published elsewhere [30, 33].

All demographic data, clinical data, and laboratory specimens were de-identified at time of enrollment. Enrolled participants were randomized to receive standard of care of amphotericin B deoxycholate 0.7–1.0 mg/kg with fluconazole 800 mg/day alone (control arm) or with 100–400 mg/day of sertraline (experimental arm) and were followed for 18 weeks. As it was determined that adjunctive sertraline did not reduce mortality from cryptococcal meningitis, all participants with available sample volume, irrespective of clinical trial arm randomization, were included in this study [33].

Sample Processing and Storage

Lumbar punctures were performed at time of hospitalization and study enrollment to collect CSF for clinical diagnostics and immunological studies. CSF was centrifuged and supernatant was stored at −80°C. The supernatant was transported to the University of Minnesota for analysis.

Cytokine and Chemokine Quantification

CSF supernatant was thawed, centrifuged, and run on the R&D Human XL Cytokines Discovery Premixed Kit 45-Plex (catalog number FCSTM18–45). Samples were run in singles with provided standards and controls following manufacturer's instructions. Approximately 10% of samples were run in duplicates to measure coefficient of variance. Details regarding the analytes included in the assay can be found in Supplementary Method Table 1.

The plate was read on a Luminex MAGPIX microplate instrument with xPONENT analysis software (R&D Systems, Minneapolis, Minnesota). A 6-point serial dilution of standards and control was used to generate a standard curve for each biomarker and experimental samples were plotted on the standard curve. In cases where computed values were below the level of detection, samples were set halfway between the minimum level of detection and zero.

Statistical Methods

Baseline demographic and clinical data were analyzed by χ² test for categorical variables and Kruskal-Wallis test for continuous variables comparing participants by 14-day survival, reporting the absolute number (percentages) and median (interquartile range). Adjustments to the cohort size are due to variations in clinical data availability. Significance was set at P < .05.

CSF cytokine concentrations were log₂ transformed. Mean cytokine and chemokine levels were compared with a linear regression to ascertain differences based on 14-day survival. To determine the association of each biomarker with 14-day mortality, a logistical regression analysis was completed reporting the odds ratio and 95% confidence interval, adjusting for baseline Glasgow Coma Scale (GCS) score and CSF quantitative Cryptococcus culture colony-forming units (CFU)/mL of CSF. For the analysis by quartiles, cytokine and chemokine levels were stratified into bottom 25% (Low [Q1]), middle 50% (reference group [Q2 + Q3]), and top 25% (High [Q4]). Subsequently, the Low-Q1 and High-Q4 groups were compared to the reference group (Q2 + Q3) and the risk of 14-day mortality was assessed, adjusting for baseline GCS score and CSF CFU. Significance was set at P < .05. All the statistical analyses were completed using SAS (SAS Institute Inc, Cary, North Carolina).

RESULTS

Evaluation of Baseline Demographic and Clinical Data

A total of 337 participants were included in the analysis, among which 253 (75.1%) survived and 84 (24.9%) died by day 14 after diagnosis (Table 1). There was no significant difference in sex, age, weight, current use of tuberculosis medication, or history of seizures across the 2 groups. There was no significant difference in peripheral blood CD4 count (P = .16) or number of participants with CD4 count >100 cells/μL (P = .32). Current use of antiretrovirals was comparable among participants who survived (47.8%) and those who died (45.2%) (P = .68). However, participants who died by day 14 had significantly lower hemoglobin (P = .01) and GCS (P < .01) on bedside assessment, indicating anemia and declining mental status and cognitive function, respectively.

Analysis of baseline CSF samples demonstrated no significant difference in CSF opening pressure (P = .18), percentage of participants with sterile culture (P = .58), protein level (P = .14), or total leukocyte count (P = .12) when comparing 14-day survival, though participants who died trended toward higher opening pressures, higher protein levels, and lower leukocyte counts (Table 1). Participants who died within 14 days of hospitalization had significantly higher CSF Cryptococcus burden (P = .02). Together, these data suggest that participants with advanced HIV and cryptococcal meningitis who died acutely have altered mental status and higher CSF fungal burden compared to those who survive, despite similar rates of antiretroviral utilization and peripheral blood CD4 counts.

CSF Cytokine and Chemokine Analysis

First, we compared the mean levels of CSF cytokines and chemokines between those who survived (n = 253) and those who died (n = 84) within 14 days of hospitalization. We identified 4 statistically significant biomarkers (Supplementary Table 1 and Supplementary Figure 1). Levels of innate immune response inflammatory cytokines MIP-3β (P < .01) and IFN-β (P = .03) were significantly lower in participants who died by day 14 compared to those who survived. Among biomarkers functioning primarily in the adaptive immune response, we identified 2 that are commonly involved in cytotoxic cell function. Participants who died within 14 days of hospitalization had significantly lower levels of chemoattractant interferon gamma-induced protein 10 (IP-10) (P < .01) and cellular apoptosis inducing protein granzyme B (P = .02) compared to those who survived. There was no significant difference in biomarkers functioning primarily in hemostasis and proliferation pathways.

Subsequently, to identify biomarkers predictive of 14-day mortality as a continuous risk factor, we performed a logistical regression analysis. Eleven biomarkers in total were statistically significant, indicating that per 2-fold increase in CSF concentration, the risk of acute 14-day mortality decreased (Figure 1 and Supplementary Table 2). We did not identify biomarkers where per 2-fold increase in CSF concentration was associated with increased risk of mortality. Five biomarkers including IL-6 (P = .015), MIP-1β (P = .003), MIP-3β (P = .009), IFN-β (P = .02), and IL-1a (P = .017) are involved in innate immune cell inflammatory pathways and cell activation, specifically dendritic cells and macrophage chemotaxis. Additionally, the Th2 cytokine IL-5 (P = .022) and Th17 cytokine IL-17A (P = .026) were also associated with decreased risk of acute mortality. Importantly, we identified biomarkers related to cytotoxic cell function, target cell killing, and chemotaxis, respectively, that were associated with 14-day mortality including IL-12 (P = .012), TNF-α (P = .03), granzyme B (P = .007), and IP-10 (P = .017).

Next, we examined how well our data align with the damage response framework of microbial pathogenesis, recognizing that the extremes of either too little or too much inflammation (Th2 or Th1 skewed cytokine environments, respectively) may convey risk of mortality, which is not detected by statistical analysis methods of means or continuous variables. Participants were grouped into quartiles based on biomarker concentration and the percentage of participants who died within 14 days of diagnosis by quartile was assessed. We identified 4 biomarkers in which there was a significant difference across the 4 quartiles including innate inflammatory molecules MIP-3β (P < .001) and IFN-β (P < .01), checkpoint marker PDL1 (P = .02), and cytotoxicity-associated chemokine IP-10 (P = .03) (Figure 2A). None of the later biomarkers that were significant mirrored the damage response framework parabola. There was no common trend among these cytokines; for example, the percentage of participants who died by day 14 steadily decreased as the concentration of MIP-3β increased; however, a similar trend was not observed for biomarkers IFN-β, IP-10, and PDL1. Additionally, the well-studied Th1 cytokine IFN-γ and TNF-α were not significant in our acute 14-day mortality quartile analysis. However, several biomarkers were trending toward significance including MIP-1β, growth-regulated oncogene-β (GRO-β), IL-10, and granzyme B (Figure 2B). Several biomarkers identified as significant in the logistical regression analysis including IL-6, IL-1a, IL-12, TNF-α, IL-17A, and IL-15, which were associated with decreased risk of mortality, were not significant in the quartile analysis (Supplementary Figure 2).

Last, to determine the degree of risk of low and high cytokine concentrations on 14-day mortality, we combined the middle 25%–75% (Q2 and Q3) into a “reference group” and compared Low-Q1 or High-Q4 to the reference group (Q2 + Q3) (Supplementary Table 3). When comparing Low-Q1 to the reference group (Q2 + Q3), we determined that low levels of inflammatory cytokines and chemokines MIP-3β (P = .03), IL-1a (P = .02), and GRO-β (P = .007); anti-inflammatory IL-10 (P = .03) and IL-13 (P = .02); cytotoxicity-associated cytokines and chemokine TRAIL (P = .04) and IP-10 (P = .01) increased the risk of acute mortality (Figure 3). In contrast, when comparing High-Q4 to the reference group (Q2 + Q3), we identified that high levels of the innate inflammatory biomarkers MIP-1β (P = .04), MIP-3β (P = .008), and IFN-β (P = .02) and the hemostasis marker PDGF-aa (P = .04) decreased the risk of acute mortality. MIP-3β was implicated in both Low-Q1 and High-Q4 comparisons, indicating that low concentrations of MIP-3β increase the risk of mortality whereas high concentrations are protective.

DISCUSSION

Our results demonstrate that an immune environment in the CSF at the time of cryptococcal meningitis diagnosis with higher concentrations of innate inflammatory and cytotoxicity-associated cytokine and chemokine signaling molecules are associated with decreased risk of acute 14-day mortality. Prior studies investigating acute mortality have shown a role for Th1 CD4⁺ T cells producing inflammatory cytokines and chemokines. In a cohort of 90 participants with cryptococcal meningitis, principal component analysis demonstrated that participants who died within 14 days of diagnosis had lower levels of IL-6, IFN-γ, IL-8, IL-10, IL-17A, and RANTES in the CSF compared to their counterparts [28]. A similar study of 30 participants identified that persons who died acutely from cryptococcal meningitis had lower levels of IFN-γ, IL-12, IL-8, and TNF-α [23].

However, previous studies were limited by small cohort size and range of biomarkers. The present analysis overcomes earlier limitations by using a larger cohort size and >40 biomarkers quantified. Our larger analysis with more cytokines allowed for (1) the identification of additional biomarkers that provide better resolution of immune responses in the CSF critical for 14-day survival and (2) revealed a novel role for proinflammatory and cytotoxicity-associated cytokines and chemokines in the CSF immune response.

By utilizing several analysis methodologies including direct comparison of mean cytokine concentration to acute mortality, logistical regression to assess for risk as log₂-transformed continuous variables, and quartile assessment to investigate patterns within our cohort, we identified that higher levels of MIP-1β, MIP-3β, IFN-β, IL-12, IP-10, granzyme-B, and TNF-α are associated with a decreased risk in acute 14-day mortality from cryptococcal meningitis. Additionally, CNS concentrations of granzyme B and IP-10 were significantly lower in participants who died within 14 days of diagnosis compared to those who survived. Similar to previously published studies, we also identified IL-6, IL-17A, and TNF-α associated with acute mortality [23, 28]. Together these novel data may implicate cytotoxic cells in acute mortality from cryptococcal meningitis as these cytokines and chemokines function within crucial cytotoxic cell processes. Such a hypothesis requires future confirmation.

Cytotoxic cells, primarily CD8⁺ T cells and natural killer (NK) cells, though other cytotoxic cell populations of lesser frequency exist, function to target pathogens and distressed host cells via granules containing cytolytic and proapoptotic proteins (granulysin, perforin, granzymes, and others) to trigger apoptosis [35, 36]. This process of degranulation occurs through a combination of activating receptor–ligand interactions and cytokine signals from IL-12, IL-15, IL-18, IL-2, and type 1 interferons [35, 37]. Activated cytotoxic cells produce inflammatory chemoattractant signals such as MIP-3β, MIP-3α, MIP-1β, MIP-1α, and granulocyte macrophage colony-stimulating factor, which are important recruitment signals to attract innate inflammatory macrophages and dendritic cells [35]. Last, activated cytotoxic cells contribute to the production of several inflammatory cytokine populations, notably TNF-α and IFN-γ, to enhance inflammation in target tissues and CD4⁺ T-cell polarization toward the Th1 phenotype, crucial for fighting intracellular and extracellular pathogens [35, 37].

Previous work on human NK cells has shown that NK cells from healthy HIV-negative donors in vitro target and kill Cryptococcus deneoformans (B3501) via engagement of cytotoxicity activating receptor NKp30 and ligand β-1,3 glucan on C deneoformans [38, 39]. Conversely, NK cells from virally suppressed HIV-positive donors have impaired ability to prevent the growth of C deneoformans [39]. Furthermore, stimulation with IL-12 can enhance cytotoxic function and improve perforin polarization in vitro against C deneoformans [40]. Collectively these data demonstrate the ability of NK cells to target C deneoformans in vitro. However, it is unclear if NK cells have similar functional abilities against C neoformans clinical isolates acquired from patients with cryptococcal meningitis.

CD8⁺ T cells in the CSF of participants with cryptococcal meningitis make up a predominant proportion of infiltrating leukocytes at baseline and day 14 of hospitalization [41]. These CD8⁺ T cells express activating markers HLA-DR and PD-L1, indicating that they are recently activated, though the pathogen (eg, HIV, C neoformans, or others) producing the antigenic stimuli is unclear [41]. In work performed using CD4⁻ mouse models, to mimic the state of CD4⁺ T-cell depletion in human advanced HIV, depleting CD8⁺ T cells during Cryptococcus pulmonary infection led to increased fungal burden in the lung and decreased leukocyte recruitment [42]. In vitro studies have shown that CD8⁺ T cells from healthy HIV-negative donors can prevent the growth of C deneoformans (Cap67 mutant) in an IL-15–dependent manner through the utilization of cytolytic protein granulysin [43]. However, it is unclear if these CD8⁺ T cells maintain their ability to produce inflammatory cytokines and chemokines post–C deneoformans co-culture.

Considering our findings and the current body of literature regarding cytotoxic cells in Cryptococcus infection, it is possible that these immune cell subsets play a crucial role in cryptococcal meningitis infection that has been previously overlooked. We postulate that a combination of low levels of activation and chemotaxis signaling in the CNS may be impairing infiltrating cytotoxic cell populations, CD8⁺ T cells and NK cells, degranulation against extracellular Cryptococcus and Cryptococcus-infected macrophages for destruction, and inflammatory cytokine secretion for downstream polarization of Th1 CD4⁺ T cells (Figure 4). Extensive testing of said hypothesis is essential to determine the role of cytotoxic cells in the CNS during Cryptococcus infection and their role in acute mortality.

While a majority of the biomarkers identified in our analyses participate in cytotoxic cell processes, several overlap with classic Th1 inflammatory cytokines, specifically IL-12 and TNF-α. These cytokines are also produced by Th1 CD4⁺ T cells to trigger downstream polarization of inflammatory macrophage populations. Both the CD4⁺ T-cell count (cells/μL) and the proportion of subjects with >100 cells/μL CD4⁺ T-cell count did not significantly differ across 14-day survival status (P = .16 and P = .32, respectively). Therefore, it is unlikely that the decrease in acute mortality per 2-fold increase in cytokine and chemokine concentrations is due primarily to CD4⁺ T cells. When considering all the cytokines and chemokines collectively that were significant in our analyses, the predominant pattern is more consistent with cytotoxic cells.

Likewise, several biomarkers do not fit well into our cytotoxic cell model, including acute phase reactant cytokine IL-6, inflammatory mediator cytokine IL-5, and Th17 cytokine IL-17A. Proinflammatory cytokine IL-6 has been implicated in protective immune responses at 14 days post–cryptococcal meningitis diagnosis [28]. The Th2 cytokine IL-5 has not been documented in relation to acute mortality in previous human cytokine studies, though similar biomarkers such as IL-4 and IL-10 have been negatively correlated with baseline fungal burden and rate of infection clearance [28]. Potentially deficient IL-5 signaling may also reflect a dysfunctional anergic immune response.

Th17-polarized CD4⁺ T cells produce cytokine IL-17A, among other cell types, and are involved in autoimmune disease and infection [44]. IL-17A levels in the CSF are significantly lower in persons who die within 14 days of cryptococcal meningitis diagnosis compared to those who survive [23]. Prior mouse studies using C deneoformans (52D) has shown that IL-17A^−/− mice have higher fungal burdens at late-stage infection and reduced counts of myeloid cells in the lung, potentially impacted by IL-10 signaling [17, 22]. Together these data demonstrate that IL-17A may perhaps be important in acute survival from cryptococcal meningitis; however, the mechanism remains to be elucidated.

Though we postulate the interactions between cytokines, chemokines, and cellular mechanisms, we were limited in our study by our inability to firmly identify the cellular origin of these biomarkers. In addition, it is unclear if the cytokines and chemokines present in the CSF travel to the CNS via lymphatics or vasculature or were secreted by infiltrating immune cells. CSF cellular studies of cell subsets and cytotoxic cell phenotypes are warranted. We recognize that the results presented here are exploratory to formulate testable hypotheses for prospective clinical and laboratory-based research studies.

In summary, our findings are consistent with previous studies demonstrating an important role for inflammation in the CNS during cryptococcal meningitis infection in participants with advanced HIV. Moreover, our data describe a novel and potentially crucial role for cytotoxic cells during the early stage of cryptococcal meningitis disease that appears to be important for preventing acute mortality. Future targeted research investigating cytotoxic NK cells and CD8⁺ T cells that have infiltrated the CNS is warranted. Such immunological studies could provide information regarding the characterization, functionality, and gene expression profiles of cytotoxic cells. With the accumulation of supportive data, we may be able to identify candidate targets for immunomodulatory therapy, which could enhance proinflammatory and or cytotoxic cell capabilities to improve acute mortality from cryptococcal meningitis in patients with advanced HIV.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

ASTRO-CM Team. John Kasibante, Jane Gakuru, Jane Francis Ndyetukira, Cynthia Ahimbisibwe, Florence Kugonza, Carolyne Namuju, Alisat Sadiq, Alice Namudde, James Mwesigye, Kiiza K. Tadeo, Paul Kirumira, Michael Okirwoth, Tonny Luggya, Julian Kaboggoza, Eva Laker, Leo Atwine, Davis Muganzi, Stewart Walukaga, Emily E. Evans, Bilal Jawed, Matthew Merry, Anna Stadelman, Nicole Stephens, Ayako W. Fujita, Richard Kwizera, Sarah M. Lofgren, Fiona V. Cresswell, Bozena M. Morawski, Caleb P. Skipper, Nathan C. Bahr, Melanie R. Nicol, Ananta S. Bangdiwala, and Katelyn A. Pastick.

Author contributions. E. C. O. conceptualized the study, framed the hypothesis, designed statistical models, compiled the results and data interpretation, drafted the manuscript, edited the manuscript, managed the publication review process, and sourced and contributed research funding. L. M. performed data generation, compiled the results, and reviewed the manuscript. K. H. H. and N. E. designed statistical models, performed statistical analyses, compiled the results, and reviewed the manuscript. L. T., K. S., A. M., E. N., E. M., D. A. W., and C. M. managed participant enrollment, patient treatment, patient diagnostics, and sample acquisition and reviewed the manuscript. K. N. conceptualized the study, framed the hypothesis, guided data analysis, edited the manuscript, and sourced and contributed research funding. J. R., D. B. M., and D. R. B. conceptualized the study, framed the hypothesis, designed statistical models, managed participant enrollment, patient treatment, and patient diagnostics, guided data analysis, edited the manuscript, and sourced and contributed research funding.

Acknowledgments. We express our deepest appreciation and gratitude to the patients and their families for participating in the parent clinical trial. We are also thankful for the dedicated work of the medical officers, nurses, and laboratory staff on our teams and in the hospitals across Uganda who provide medical care and support for study participants and their families. Figure 4 was generated using Biorender.com.

Data sharing. All relevant data are included in the article and supplemental data.

Financial support. This research was supported by the National Institute of Allergy and Infectious Diseases (National Research Service Award F31AI62230 to E. C. O., grant numbers T32AI055433 to J. R., R01AI176922 to K. N., and R01NS110519); the National Institute of Neurological Disorders and Stroke (grant numbers R01NS086312 to D. R. B. and D. B. M. and R01NS118538 to K. N. and D. B. M.); and the Fogarty International Center (award number K01TW010268 to J. R. and L. M.). This work was also supported in part by the Doris Duke International Clinical Research Fellows Program at the University of Minnesota Medical School, United Kingdom Medical Research Council, and the Wellcome Trust (MR/M007413/1).

References

Author notes