Shared and unique antibody and B cell profiles in HIV-positive and HIV-negative individuals with cryptococcal meningoencephalitis

Abstract

Host non-T cell markers to aid in the diagnosis of cryptococcal meningoencephalitis (CM) have not been identified. In this case-control study, we characterized antibody and B cell profiles in HIV-negative and HIV-positive Vietnamese individuals of the Kinh ethnicity recently diagnosed with CM and controls. The study included 60 HIV-negative with no known immunocompromising condition and 60 HIV-positive individuals, with 30 CM cases and 30 controls in each group. Participants were matched by age, sex, HIV serostatus, and CD4 count in the HIV-positive group. Plasma immunoglobulin (Ig) levels, including IgG1, IgG2, IgM, and IgA, Cryptococcus spp. glucuronoxylomannan (GXM)- and laminarin (branched -[1-3]-glucan)-binding IgG, IgM, IgA levels, and peripheral blood B cell subsets were measured. Logistic regression, principal component, and mediation analyses were conducted to assess associations between antibody, B cell levels, and CM. The results showed that GXM-IgG levels were higher and IgG1 and IgG2 were lower in CM cases than controls, regardless of HIV status. In HIV-negative individuals, IgG2 mediated an inverse association between CD19⁺CD27⁺CD43⁺CD5⁻ (B-1b-like) cells and CM. In HIV-positive individuals, lower levels of IgA, laminarin-IgA, and CD19⁺CD27⁺IgM⁺IgD⁻ (IgM⁺ memory B) cells were each associated with CM. The shared and distinct antibody and B cell profiles identified in HIV-negative and HIV-positive CM cases may inform the identification of non-T-cell markers of CM risk or unsuspected disease, particularly in HIV-negative individuals.

Introduction

Cryptococcal meningoencephalitis (CM) is a life-threatening fungal infection caused by encapsulated Cryptococcus spp, most commonly in the setting of human immunodeficiency virus (HIV)-associated immunosuppression.¹, ² However, an increasing number of cases are being reported in HIV-negative individuals, most of whom have no identified immunosuppressive condition in resource-rich countries,³, ⁴ as well as in less-resourced regions, such as Vietnam.⁵, ⁶

The diagnosis of CM in HIV-negative individuals who do not have CD4 deficiency is often clinically unsuspected and delayed³ due to the absence of known host biomarkers of CM risk to prompt cryptococcal antigen (CrAg) testing.⁷, ⁸ In contrast, in HIV-positive individuals, decreased CD4 T cell counts, typically below 100 cells/mm³, prompt CrAg testing.⁹ In resource-rich settings, the median time from symptom onset to cryptococcosis diagnosis in HIV-negative individuals was 4 weeks¹⁰ and 78% of patients required three or more visits to a physician before a diagnosis was made due to an absence of clinical findings that would prompt CrAg testing.¹¹ Missed opportunities for earlier testing¹² have been associated with higher mortality rates¹³ and worse neurological outcomes.¹⁴ In resource-poor settings, limited access to care and a lack of subspecialty experts may further delay diagnosis.¹⁵

Distinct profiles of cryptococcal capsular (glucuronoxylomannan [GXM])- and laminarin (branched -[1-3]-glucan) or curdlan (linear -[1-3]-glucan)-binding antibodies have been identified across the spectrum of human cryptococcal infection from latent infection¹⁶ to HIV-associated asymptomatic cryptococcal antigenemia,¹⁷, ¹⁸ symptomatic cryptococcosis in HIV-positive¹⁹ and HIV-negative individuals,²⁰, ²¹ and cryptococcus-associated immune reconstitution inflammatory syndrome.²²In vitro studies and experimental models in mice demonstrate that defined antibodies that bind GXM and β-glucans promote cryptococcal phagocytosis²³ and protect against lethal Cryptococcus neoformans infection.²⁴, ²⁵ However, data on antibody and B cell profiles in HIV-negative and HIV-positive individuals with CM diagnosis is limited, particularly in HIV-negative individuals.

The primary objective of this study was to identify non-T cell factors, including plasma antibody and B cell subset profiles that associate with early CM in HIV-positive and HIV-negative individuals compared to controls.

Materials and Methods

Study design

Case-control study of 120 participants is comprised of four groups: HIV-positive individuals with CM (n = 30) and without CM (n = 30) and HIV-negative individuals with CM (n = 30) and without CM (n = 30). CM-negative individuals (controls) were matched with CM cases (cases) 1:1 by HIV-serostatus, age, and sex. Age was matched 5 years. For HIV-positive individuals, CD4 count was matched by nadir counts if lower and known than when measured in this study within the categories, <25, 25–50, 50–100, and >100 cells/mm³. Baseline demographic and clinical data including symptom duration, antiretroviral therapy (ART) use, infecting cryptococcal molecular type (identified in 18 HIV-negative and all HIV-positive cases using isolates from sub-cultures of the blood or cerebrospinal fluid [CSF]), and mortality data up to day 70 in HIV-negative and month 6 in HIV-positive individuals were collected. CSF studies, including opening pressure, white cell count (WCC), and CSF quantitative fungal counts, were available in HIV-positive cases.

Study population

Adults of Vietnamese Kinh ethnicity 18 years of age were enrolled between 2013 and 2017 from the following groups: HIV-positive cases from the 04CN CryptoDex trial randomized to the placebo arm²⁶; HIV-negative otherwise immunocompetent cases from the study “A prospective Descriptive Study of CNS Infections at the Hospital for Tropical Diseases (HTD)”²⁷ conducted between 2007 and 2013 where participants were contacted by telephone based on records of previous participation and invited to re-participate (n = 2), and the remaining from patients prospectively admitted to HTD or Cho Ray Hospital (CRH) (n = 28); HIV-positive controls (as evidenced by negative blood or CSF lateral flow assay [LFA]) were enrolled from outpatient departments of HTD or CRH; and HIV-negative controls amongst employees of Oxford University Clinical Research Unit (OUCRU) or the HTD, or the family members of either.

CM was defined as a clinical presentation consistent with CM and one or more of positive CSF India ink, Cryptococcus spp. cultured from CSF or blood, or positive CrAg LFA in CSF or blood. Individuals were excluded if they had unknown HIV status (except for healthy controls who were not tested); hemoglobulin <7.5g/dl (except for healthy controls who were not tested); a history of cryptococcal disease (except in HIV-positive individuals); or were pregnant or receiving immunosuppressive treatment. Individuals diagnosed with CM were initiated on anti-fungal therapy. None of the participants were receiving corticosteroids at baseline.

Sample collection and processing

Plasma and whole blood were collected following informed consent, the time that we defined as a baseline. Peripheral blood mononuclear cells (PBMCs) were separated from whole blood, stored at -80°C, and shipped to Einstein. Before analysis, plasma, which was frozen at -80°C was thawed and heat-inactivated at 56°C for 30 min.

Antibody level measurements

Plasma immunoglobulin (Ig) A, IgM, IgG1, and IgG2 concentrations were measured using a Luminex platform (Austin, TX) and quantified in units of µg/ml as previously described.¹⁸, ²²Cryptococcus spp. GXM (generously provided by Dr. Arturo Casadevall, Johns Hopkins School of Medicine)- and laminarin-binding IgA, IgM, and IgG were measured and reported as inverse titers as previously described.¹⁸, ²²

Flow cytometry

Flow cytometry was performed with PBMCs prepared from anticoagulated whole blood by density gradient centrifugation in Ficoll paque (GE Healthcare Biosciences AB, Uppsala, SE). B and T cell subsets were identified by immunostaining and flow cytometry, as described previously,²⁰, ²⁸ using mouse anti-human Alexa Fluor-700 CD19, PerCP-Cy5.5 CD27, APC IgM, PE IgD, PE-Cy7 CD43, and V450 CD5 for B cells, and mouse anti-human PE-CF594 CD3, APC-Cy7 CD4, and FITC CD8 (BD Biosciences, Franklin Lakes, NJ) for T cells. Fluorescence Minus One (FMO) and CompBeads (BD Bioscience) compensation controls were used for each reagent. Among living cells (Live/Dead Blue, Life technology), percentages of the following cell types (closest mouse homologs in brackets) were identified with FlowJo^TM software (Version 10. Ashland, OR; Becton, Dickinson and Company; 2023): CD19⁺ (total B cells), CD19⁺CD27⁺ (memory B cells), CD19⁺CD27⁺IgM⁺IgD⁻ (IgM⁺ memory B cells), CD19⁺CD27⁺IgM⁺IgD⁺ (unswitched memory B cells), CD19⁺CD27⁺IgM⁻IgD⁻ (class-switched memory B cells), CD19⁺CD27⁺CD43⁺CD5⁺ (B-1a cells), CD19⁺CD27⁺CD43⁺CD5⁻ (B-1b cells), CD3⁺ (T cells), CD3⁺CD4⁺ (CD4 T cells), and CD3⁺CD8⁺ (CD8 T cells) (Supplementary Figure 1). While not definitely established, B cells with the phenotypes CD20⁺CD27⁺CD43⁺CD5⁺ and CD20⁺CD27⁺CD43⁺CD5⁻ have been referred to as human counterparts of mouse B-1a and B-1b cells, or B-1a-like and B-1b-like, respectively.^29–32 CD19⁺ and CD20⁺ cells in peripheral blood largely overlaps.²⁹

Ethics

The study was approved by the Vietnamese Ministry of Health, the HTD and CRH, and the Oxford Tropical Ethics Committee (OxTREC 25-12, HTD reference number CS/ND/12/17). All participants provided written informed consent, and de-identified samples were studied under an Albert Einstein College of Medicine Institutional Review Board-approved protocol (1989–228).

Statistical analysis

Bivariate analysis

Clinical markers and levels of antibodies and percentages of B and T cell subsets in cases and controls, stratified by HIV status, were compared using the Wilcoxon rank-sum test, Fisher's exact test, or χ² test, as appropriate. The balance in covariables between the comparison groups in Table 1 was assessed using the standardized mean difference, with an absolute value of 0.10 or more indicating covariable imbalances.³³ Correlations were assessed using Spearman’s correlation coefficient.

Table 1.

Baseline and clinical characteristics of participants by CM and HIV status.

Covariables	HIV-negative		HIV-positive		SMD1 (HIV-negative cases vs. HIV-positive cases)
Baseline characteristics	Controls (n = 30)	Cases (n = 30)	Controls (n = 30)	Cases (n = 30)
Age, median (IQR)	50 (40–62)	49.5 (43–62)	34.5 (29.5–40)	35 (30–38)	1.39
Sex, no. (%)
Male	17 (56.7)	17 (56.7)	28 (93.3)	28 (93.3)	0.92
Female	13 (43.3)	13 (43.3)	2 (6.7)	2 (6.7)
Co-morbid conditions, no. (%)
History of tuberculosis, pulmonary, and extra-pulmonary	1 (3.3)	0	5 (16.7)	4/29 (13.8)
Chronic liver disease (hepatitis B, hepatitis C, and/or cirrhosis)	1 (3.3)	1 (3.3)	5 (16.7)	2/29 (6.9)
Chronic renal disease	0	0	0	1/29 (3.4)
Diabetes mellitus	2 (6.7)	3 (10)	0	1/29 (3.4)
History of corticosteroids use, no. (%)	0	0	0	0
Has health insurance, no. (%)	n/a	n/a	n/a	13 (43.3)
Factors related to HIV status
Duration of HIV, median days (IQR)	n/a	n/a	12.5 (6–46.5), n = 24	252.5 (85–2,804), n = 10
On ART at baseline, no. (%)	n/a	n/a	10 (34.5)	7 (23.3)
Duration of ART at baseline, median days (IQR)	n/a	n/a	40.5 (15.5–1032.5), n = 4	31 (26–74), n = 7
Absolute CD4 counts at baseline (cells/mm3), median (IQR)	n/a	495 (340–1197), n = 7	12 (6–38)	15 (3–32)	2.19
On ART	n/a	n/a	15.5 (6–50), n = 10	31 (18–96), n = 7
Not on ART	n/a	n/a	9 (5–38), n = 19	11 (2–25), n = 23
Absolute CD8 counts at baseline (cells/mm3), median (IQR)	n/a	533.5 (260–797), n = 6	n/a	247 (121–517), n = 7	0.99
Absolute CD4/CD8	n/a	1.43, n = 6	n/a	0.03, n = 7	12.10
Primary fluconazole prophylaxis, no. (%)	n/a	n/a	n/a	2 (6.7)
Secondary fluconazole prophylaxis, no. (%)	n/a	n/a	n/a	0
At study baseline
Symptom duration, median days (IQR)	n/a	23 (14–30)	n/a	14 (8–20), n = 28	0.44
Time between CM diagnosis and enrollment, median days (IQR)	n/a	2.5 (1–5)	n/a	1 (0–2)	0.69
Inclusion criteria, no. (%)
Positive serum CrAg LFA	n/a	30 (100)	0	29 (96.7)
Positive CSF CrAg LFA	n/a	20 (66.7)	n/a	18 (60)
Positive CSF or blood cultures2	n/a	16 (53.3)	n/a	18 (60)
Positive CSF India Ink	n/a	n/a	n/a	28 (93.3)
Baseline CSF parameters3
Opening pressure (cmH2O), median (IQR)	n/a	n/a	n/a	28 (20–35), n = 25
CSF fungal load (log10 CFU/ml), median (IQR)	n/a	n/a	n/a	5 (4.5–6), n = 24
CSF white-cell count (cells/mm3), median (IQR)	n/a	n/a	n/a	53 (16–92), n = 29
Infecting Cryptococcus species4, no. (%)
Cryptococcus gattii	n/a	9/18 (50)	n/a	0
Cryptococcus neoformans var grubii	n/a	9/18 (50)	n/a	30 (100)
Mortality
Within 70 days, no. (%)	n/a	9/29 (31.0)	n/a	6 (20.0)	0.25
On ART at baseline	n/a	n/a	n/a	1 (3.3)
Not on ART at baseline	n/a	n/a	n/a	5 (16.7)
Within 6 months, no. (%)	n/a	n/a	n/a	7 (23.3)

Note: Percentage calculated using n = 30 as the denominator, unless indicated when complete data are not available.

Abbreviations: ART, anti-retroviral therapy; CFU, colony forming unit; CM, cryptococcal meningoencephalitis; CrAg, cryptococcal antigen; CSF, cerebrospinal fluid; HIV, human immunodeficiency virus; IQR, interquartile range; LFA, lateral flow assay; n/v, not available; SMD, standardized mean differences.

¹Standardized mean difference indicates the difference in means or proportions divided by standard error. An absolute standardized difference of 0.10 or more indicates that covariates are imbalanced between groups.

²Data for the HIV-positive group includes CSF culture results; and for the HIV-negative group, either CSF or blood culture data are available.

³If a baseline sample was not available, the earliest available CSF samples were used for baseline measurements.

⁴Identified via molecular typing of isolates from blood or CSF.

Regression analysis

Association between log-transformed antibody levels and B and T cell subset percentages and CM status was estimated using conditional logistic regression stratified by HIV-serostatus. In the HIV-positive group, we performed a sensitivity analysis using multivariable logistic regression to adjust for ART use, a factor not included in the matched analysis, and to investigate interactions between ART use, sex, and the markers in relation to their association with CM status. Previous research has indicated that matched and unmatched analyses of a matched case-control study are valid methodologies,³⁴ particularly when loose matching criteria are employed.³⁵

Principal component analysis

Antibody and B cell subset markers were each analyzed using principal component analysis (PCA) in each stratum of HIV-serostatus to reduce complex correlated datasets into a series of linear, non-correlated principal components (PC).¹⁷, ²¹ Log-transformed markers were normalized to the mean and centered. Difference in the mean PC scores by CM status was estimated using Student t test.

Mediation analysis

Based on the regression findings and biological plausibility, we conducted a mediation analysis using the Zhao et al. approach to testing mediation.³⁶ We used the Stata medsem command with zlc, which allowed us to estimate the direct and indirect effects of the exposure (B cell subsets) on the outcome variable (CM) via the mediator (antibody levels). We estimated the indirect effect using the product of coefficients method and tested the significance using bootstrapping with 1000 replications.

Subgroup analysis

We assessed potential factors that could modify the relationship between host immune response and disease status. These factors included cryptococcal molecular types, and baseline CSF culture sterility in HIV-negative individuals, baseline ART status in HIV-positive individuals, as well as sex. Limited data prevented us from analyzing CSF culture sterility in the HIV-positive group, as was performed in a previous study.³⁷ For detailed statistical analysis methods, see Supplementary Material.

All tests were two sided (α = 0.05). All analysis was exploratory and correction for multiple testing was not done. Analyses were performed in Stata/IC 16.1 (College Station, TX: StataCorp LLC).

Results

Study cohort

The study cohort consisted of Vietnamese individuals, all of whom belonged to the Asian race and Kinh ethnicity. Compared to HIV-negative cases, HIV-positive cases were younger (median age, interquartile range [IQR]; 35 [30–38] vs. 50 [43–62] years; P < .0001), predominantly male (93% vs. 57%; P = .001) and had lower absolute CD4 counts (15 [3–32] vs. 495 [340–1197] cells/mm³; P < .0001). Among HIV-positive individuals, 7 (23.3%) of cases were receiving ART at baseline, compared to 10 (34.5%) of controls. Among the cases, the seven individuals on ART at baseline had higher CD4 counts than those not on ART (31 [18–96] vs. 11 [2–25] cells/mm³; P = .02), and six out of these seven individuals had initiated ART within the previous three months. Cases for whom data was available (10 out of 30) had a longer duration of HIV infection than controls (median days, IQR; 252.5 [85–2804] vs. 12.5 [6–46.5]; P = .003) (Table 1).

The enrolled HIV-negative individuals did not have any known immunosuppressive conditions and were relatively healthy, with few co-morbid conditions such as chronic liver disease and diabetes mellitus, which were balanced between the case and control groups. HIV-negative cases had a longer symptom duration than HIV-positive cases (median days, IQR; 23 [14-–30] vs. 14 [8–20]; P = .02). Mortality by day 70 was higher in the HIV-negative group than the HIV-positive group, but this was not statistically significant (9 [31%] vs. 6 [20%]; P = .38). Among HIV-positive cases, those not on ART had higher mortality than those on ART (5 [16.7] vs. 1 [3.3]; P = 1.00), but this difference was not statistically significant (Table 1).

Antibody levels among HIV-negative individuals

Compared to controls, cases had significantly lower IgG1 (1933 [1566–2386] vs. 2948 [2654–3104] µg/ml; P < .0001) and IgG2 (2081 [1784–2585] vs. 5674 [3784–7263] µg/ml; P < .0001), and higher GXM-IgG (inverse titer, 894 [316–1392] vs. 135 [98–175]; P < .0001) and GXM-IgM (39 [20–48] vs. 18 [13–37]; P = .01) (Figs. 1, 2, Supplementary Table 1).

Figure 1.

Total Ig levels, μg/ml of HIV-negative and HIV-positive individuals by CM status, are depicted as medians and interquartile ranges as shown on the y-axis for each group of 30 individuals shown on the x-axis. HIV and CM status is represented by (+) or (-) to indicate positive or negative HIV or CM status. Wilcoxon rank-sum test in controls vs. cases by HIV-serostatus. CM, cryptococcal meningoencephalitis; HIV, human immunodeficiency virus; Ig, immunoglobulin. *P < .05, ** P ≤ .01, and ***P ≤ .001.

Figure 2.

Plasma GXM- and laminarin-binding antibody levels in HIV-negative and HIV-positive individuals by CM status. Inverse ELISA titers, depicted as medians and interquartile ranges, are shown on the y-axis for each group of 30 individuals shown on the x-axis. HIV and CM status is represented by (+) or (-) to indicate positive or negative HIV or CM status. Wilcoxon rank-sum test in controls vs. cases by HIV-serostatus. CM, cryptococcal meningoencephalitis; ELISA, enzyme-linked immunosorbent assay; GXM, glucuronoxylomannan; HIV, human immunodeficiency virus; Ig, immunoglobulin. *P < .05, ** P ≤ .01, and ***P ≤ .001.

Among HIV-negative individuals with CM for whom molecular typing of their infecting strain was available (n = 18), those with C. gattii (n = 9) had higher GXM-IgG levels than those with Cryptococcus neoformans var grubii (n = 9) (inverse titer, 1496.3 [1072.0–2744.7] vs. 426.3 [308.9–900.0]; P = .03). No significant differences were identified in the other measured antibody markers (Supplementary Table 2).

Among the HIV-negative cases, those with sterile CSF at baseline (n = 14) had lower GXM-IgM than those with positive CSF culture (n = 16) (inverse titer, 34.9 [15.9–40.7] vs. 46.1 [34.7–121.2]; P = .03), a finding confirmed in a multivariable logistic regression analysis adjusting for age and sex (adjusted OR, 4.00; 95% CI, 1.11–14.40). No significant differences in symptom duration, absolute CD4 count, infecting cryptococcal species, or mortality were observed between groups stratified by CSF sterility (Supplementary Table 3).

Antibody levels among HIV-positive individuals

Compared to controls, cases had significantly lower IgG1 (3103 [1962–3829] vs. 4299 [3455–5347] µg/ml; P < .001), IgG2 (1836 [1099–2749] vs. 3237 [2505–4765] µg/ml; P < .001), IgA (2131 [1229–3180] vs. 2560 [1728–3915] µg/ml; P = .04), and laminarin-IgA (16 [13–36] vs. 48 [16–116]; P < .01) and higher GXM-IgG (374 [165-541] vs. 171 [121–375]; P = .02) (Figs. 1, 2, Supplementary Table 1).

GXM-IgG was lower in individuals who were on ART at baseline (n = 17) than those who were not on ART (n = 42) (134.6 [51.6–383.9] vs. 362.4 [150.9–522.0]; P = .02), a finding confirmed in a multivariable linear regression analysis adjusting for age, sex, and absolute CD4 count (coefficient, -1.01; 95% CI, -1.81, -0.22; being on ART was associated with a 1.01 log titer decrease in GXM-IgG). When stratified by CM status, the significant inverse association between ART use and GXM-IgG levels was observed only in controls (on ART vs. not on ART; 75.4 [41.1–134.6] vs. 344.6 [146.1–480.3]; coefficient, -1.37; 95% CI, -2.38, -0.35, P = .01).

Clinically, those on ART had a longer duration of HIV (median days, IQR; 685 (85–2507) vs. 11 (6–34); P = .0003) and higher absolute CD4 counts (median, IQR; 20 [13–51] vs. 10 [3–20] cells/mm³; P = .02) than those not on ART. No significant differences were found in CSF findings, mortality, or other antibody markers based on ART status at baseline (Supplementary Table 4).

In terms of sex differences, females had higher levels of IgM than males in the HIV-negative group (776.6 [574.6–1181.1] vs. 624.4 [451.3–835.2]; P = .03) and higher IgG2 in the HIV-positive group (4843.6 [4111.0–5670.1] vs. 2505.3 [1400.4–3465.3], P = .01), with the caveat that there were 4 females in the HIV-positive group (Supplementary Table 5).

B cell subset levels

Among CD19⁺CD27⁺ B cells, compared to their respective controls, HIV-negative cases had a lower percent of CD43⁺CD5⁻ cells (12 [8–21] vs. 25 [16–40] %; P < .01), and HIV-positive cases had a lower percent of IgM⁺IgD⁻ cells (3 [2–9] vs. 8 [5–13] %; P = .001) (Fig. 3). These findings were similar when comparing the proportion of B cell subsets among CD19⁺ cells (Supplementary Table 1).

Figure 3.

B cell subset levels as a proportion of CD19⁺CD27⁺ cells in HIV-negative and HIV-positive individuals by CM status. Medians with interquartile range are shown on the y-axis for each group of 30 individuals shown on the x-axis. HIV and CM status is represented by (+) or (-) to indicate positive or negative HIV or CM status. Wilcoxon rank-sum test in controls vs. cases by HIV-serostatus. Closest mouse homologs of the listed human B cell phenotypes (in the order) are, unswitched, IgM⁺-only, switched memory B cells, and B-1a and B-1b cells. CM, cryptococcal meningoencephalitis; HIV, human immunodeficiency virus. *P < .05, ** P ≤ .01, and ***P ≤ .001.

T cell subset levels

Among CD3⁺ cells, no significant differences were observed in CD4⁺ or CD8⁺ percent between cases and controls in both HIV-negative and HIV-positive groups, although CD4/CD8 ratios were >1 and <1, respectively, for HIV-negative and HIV-positive individuals (Supplementary Table 1). Absolute CD4 count was inversely correlated with GXM-IgG in the HIV-negative cases ( = -0.93, P = .003, n = 7) and with CD5⁺CD43⁺ (B-1a-like) cells in the HIV-positive cases (= -0.40, P = .03, n = 30) (Supplementary Table 6).

Associations of antibody and B cell subset levels with CM status by conditional regression analysis

For HIV-negative individuals, lower IgG1 (odds ratio, 95% confidence interval; 0.01, 0.0006–0.28), IgG2 (0.01, 0.0006–0.31), and CD43⁺CD5⁻ B cell (0.36, 0.15–0.88) levels and higher GXM-IgM (4.27, 1.57–11.6) and GXM-IgG (4.54, 1.51–13.6) levels were associated with CM status.

For HIV-positive individuals, lower IgA (0.17, 0.04–0.69), laminarin-IgA (0.32, 0.13–0.80), IgG1 (0.37, 0.14–0.99), and IgM⁺IgD⁻ B cell (0.48, 0.24–0.95) and higher GXM-IgG (2.35, 1.12–4.94) levels were associated with CM status (Fig. 4).

Figure 4.

Forest plots of the conditional logistic regression analysis estimating the association between antibody and B cell subsets and CM status in (A) 60 HIV-negative and (B) 60 HIV-positive individuals. Antibody and B cell subset variables were log-transformed. CM cases and controls in each group were matched 1:1 by age and sex, and CD4 count in HIV-positive individuals. P-values <.05 are bolded. CI, confidence interval; CM, cryptococcal meningoencephalitis; GXM, glucuronoxylomannan; HIV, human immunodeficiency virus; Ig, immunoglobulin; OR, odds ratio.

Sensitivity analysis

An unmatched multivariable logistic regression model was used to estimate the association between antibody and B cell subset levels and CM status adjusted for age, sex, absolute CD4 count, and ART status within the HIV-positive group. While the overall association profile largely aligned with the findings from the conditional regression analysis, there was an interaction between ART use and both IgG2 and laminarin-IgA (with interaction term P-values of .03 and .02, respectively); notably, a significant inverse association between these markers and CM status was found only in individuals not on ART in a stratified analysis. No significant interaction with sex was found in either the HIV-negative or HIV-positive cohorts (Supplementary Figure 2).

PCA

Measured antibody and B cell subset levels were correlated to varying degrees (Supplementary Table 7). The second principal component (PC2) of the antibody markers distinguished individuals by CM status in each stratum of HIV-serostatus. In the HIV-negative group, PC2, composed of IgG2, IgG1, and negative loading of GXM-IgG, explained 21% of the variance and was significantly lower in cases compared to controls; mean PC2 score, -1.21 vs. 1.21; 95% CI, 1.99–2.84, P < .0001. In the HIV-positive group, PC2, composed of IgA, laminarin-IgA, and GXM-IgA, explained 19% of the variance and was significantly lower in cases compared to controls; mean PC2 score, -0.70 vs. 0.70; 95% CI, 0.78–1.99, P < .0001 (Fig. 5, Supplementary Table 8, Supplementary Figure 3).

Figure 5.

Plots of individual patient values of the first and second principal components derived using the log of antibody levels in (A) 60 HIV-negative and (B) 60 HIV-positive individuals. HIV, human immunodeficiency virus.

PC2 of the B cell subsets also distinguished HIV-negative and HIV-positive individuals by CM status, driven by CD43⁺CD5⁻ B cells and IgM⁺IgD⁻ B cells, respectively (Supplementary Table 9).

Mediation analysis

We performed mediation analysis to examine if the influence of CD43⁺CD5⁻ B cells (the exposure) on IgG2 levels (the potential mediator) explained some or all the association between CD43⁺CD5⁻ B cell levels and CM status (the outcome) in the HIV-negative group. We found a significant indirect effect of the exposure on the outcome through the mediator variable (coefficient, -0.15; standard error, 0.05; P < .01), confirmed by both the Sobel test and Monte Carlo test, indicating that IgG2 completely mediated the relationship between CD43⁺CD5⁻ B cells and CM (Supplementary Figure 4). The relative indirect effect size was 0.65, meaning that about 65% of the effect of log-transformed CD43⁺CD5⁻ B cells on CM status was mediated by IgG2. Mediation was sought but not found in the HIV-positive group.

Discussion

In this matched case-control study, we analyzed plasma antibody and B cell subset levels in HIV-negative and HIV-positive individuals recently diagnosed with CM and compared them to controls. Irrespective of HIV status, cases had higher GXM-IgG and lower IgG2 and IgG1 levels than controls, and GXM-IgM levels were higher in HIV-negative cases than controls. Of note, GXM-IgG levels were higher in HIV-negative cases with Cryptococcus gattii than C. neoformans, a finding similar to that of a 1996 Australian study that suggested C. gattii was more immunogenic.³⁸ We also found that lower CD43⁺CD5⁻ (B-1b-like) B cells and IgG2 levels were associated with CM status in HIV-negative individuals, whereas lower IgM⁺IgD⁻ (IgM⁺ memory) B cells and IgA levels were associated with CM status in HIV-positive individuals.

Overall, our results are consistent with previous studies in which GXM-IgG levels were higher in those with cryptococcal antigenemia¹⁷, ¹⁸ or cryptococcosis in HIV-positive¹⁹, ³⁹ and HIV-negative²⁰, ⁴⁰ individuals than controls. Together, these findings suggest that GXM-IgG levels, irrespective of HIV infection status, and perhaps GXM-IgM in HIV-negative individuals, may reflect a response to the fungal burden. Consistent with this concept, GXM-IgM levels were higher in HIV-negative cases with positive CSF cultures than in those with sterile cultures. Although antibody levels and quantitative CSF yeast counts were not correlated, the analysis was limited by the small sample size. Notably, lower GXM-IgG levels were associated with 6-month mortality in HIV-positive Africans with asymptomatic cryptococcal antigenemia.¹⁷ While we did not identify an association between GXM-IgG and mortality in CM cases in this study, this question requires longitudinal studies of antibody levels across the continuum of cryptococcal infection from latency to cryptococcal antigenemia to cryptococcosis.

Levels of IgG1 and IgG2 were lower in CM cases than controls. In contrast, levels of these Igs were higher in HIV-positive African individuals with than those without asymptomatic cryptococcal antigenemia.¹⁷, ¹⁸ This difference suggests the hypothesis that cryptococcal antigenemia stimulates an increase in Igs, but fungal burdens associated with CM may be less effective in doing so due to dysfunction or depletion of the responding B cell populations. Consistent with this idea, for HIV-negative individuals, lower CD43⁺CD5⁻ (B-1b-like) B cell and IgG2 levels were each associated with CM status and directly correlated. When we used a mediation model to test the relationship between this B cell subset, IgG2, and CM status statistically, we found that IgG2 mediated the inverse association between CD43⁺CD5⁻ B cell levels and CM in HIV-negative individuals. Lending mechanistic support to these associations in another encapsulated pathogen, mouse B-1b cells⁴¹ and human CD43⁺CD5⁻ (B-1b-like) cells³⁰, ³¹ each produce pneumococcal capsular polysaccharide-specific antibodies and a lack of B-1b cells was associated with pneumococcal susceptibility in mice.⁴¹

Regarding the possible importance of IgG2 in cryptococcosis, HIV-negative organ transplant recipients (OTRs) with cryptococcosis²¹ and other fungal infections had lower IgG2 than controls.⁴² IgG2 deficiency, which is a long-recognized feature of HIV infection,⁴³ is the predominant IgG subclass of human capsular polysaccharide, including GXM antibodies.⁴⁴, ⁴⁵ Notably, human GXM-IgG2 enhanced C. neoformans phagocytosis²³ and mediated protection in mice, whereas IgG1 did not.⁴⁶ In addition, a case of CM caused by C. gattii was associated with IgG2 deficiency,⁴⁷ although the study did not measure anti-granulocyte macrophage-colony stimulating factor (GM-CSF) auto-antibodies, which were later described in association with C. gattii cases.⁴⁸ Nonetheless, just as CD4 deficiency alone does not explain CM occurrence in HIV-positive individuals, IgG2 deficiency alone is also unlikely to do so as a multi-hit model to explain CM occurrence has been proposed.⁴⁹ CM is rare in individuals with common variable immunodeficiency, possibly due to residual B or T cell function,⁵⁰, ⁵¹ but it does occur in association with anti-GM-CSF autoantibodies (C. gattii)⁴⁸ and antibody defects, such as those in X-linked agammaglobulinemia,⁵² X-linked Hyper IgM syndrome,⁵³ and Fc receptor polymorphism.⁵⁴

In HIV-positive individuals, lower IgM⁺ memory B cell and IgA levels were each associated with CM status. IgM⁺ memory B cells in humans, like B-1a cells in mice, produce naturally occurring antibodies, including laminarin-binding IgM⁵⁵ and natural IgA in mice.⁵⁶ An inability to produce natural antibodies has been linked to host susceptibility to fungi including C. neoformans²⁸, ⁵⁵, ⁵⁷ and Aspergillus fumigatus⁵⁸ in mice and humans. IgA, like IgG2, can dimerize, which increases antibody avidity and allows more effective binding and triggering of effector functions such as Fc receptor binding, complement activation, and phagocytosis.⁵⁹ We found lower IgA, laminarin-IgA, and GXM-IgA levels in HIV-positive individuals with CM than controls, suggesting a possible role for mucosal immunity in resistance to CM. Notably, cryptococcal nasal colonization has been documented in koalas,⁶⁰, ⁶¹ and C. neoformans has been identified in the sinus microbiomes of healthy volunteers and those with chronic rhinosinusitis.⁶² These observations lend biological plausibility to the concept that cryptococcal nasal colonization could induce mucosal immunity. Given that memory B cells and IgA are important components of lung mucosal immunity,⁶³ further research is needed to investigate this question.

HIV-negative cases in this study did not have known underlying cellular immunodeficiency. Since T cells can amplify B cell responses and are required for adaptive antibody responses,²⁹ intact anti-fungal T cell-mediated responses⁶⁴ might delay cryptococcal dissemination, symptoms, and clinical presentation in HIV-negative individuals which could paradoxically result in poorer clinical outcomes. The inverse relationship between GXM-IgG and CD4 count in HIV-negative cases and the lower levels of GXM-binding antibodies observed in those with sterile CSF may suggest a reduced fungal burden, potentially influenced by T cell activity. Along these lines, Skipper et al. showed that HIV-positive individuals with CM and negative CSF cultures had higher CD4 counts and mortality rates,³⁷ and in mice, Neal et al. found that, despite achieving microbiological control, central nervous system Th1-biased CD4 cells contributed to detrimental pathology.⁶⁵ The ability of CD4 cells to both augment fungal clearance and promote inflammation and damage is consistent with a foundational concept of the Damage response framework, namely, that host damage can occur in the setting of a strong or a weak immune response.⁶⁶ While characterization of T cell subsets and/or measurement of cytokines, such as interleukin-17³⁷ or interferon-⁶⁷ was beyond the scope of this study, our findings underscore the need to better understand the relationships between antibody and B cell subsets, fungal clearance, and inflammation.

In this study, the shorter duration of symptoms prior to presentation in the HIV-positive than HIV-negative cases may be attributable to the absence of CD4-mediated functions that may temporarily control the fungal burden. Therefore, in HIV-positive individuals, early control of fungal growth may depend on residual CD4 immunity, non-T cell-dependent innate immune mechanisms, or prior antibody responses.⁶⁸ Consistent with this idea, GXM-IgG levels were lower among HIV-positive individuals on ART than those who were not, and B-1a-like cell levels inversely correlated with CD4 count. B-1a cells can promote differentiation of CD4 cells into Th1 and Th17 cells,⁶⁹ both of which can mediate fungal clearance.⁶⁷, ⁷⁰ It is noteworthy that multiple B cell populations, including B-1a, B-1b, B-2, and marginal zone B cells,⁷¹ produce laminarin-binding antibodies and/or GXM-IgG in mice,⁵⁵ and B-1a cells limited fungal dissemination to the brain in mice.⁵⁵ This redundancy may partially explain the absence of a mediation effect or a direct correlation between a specific B cell subset and antibody population among the HIV-positive individuals in this study.

Strengths of our study include the use of an individual matching strategy to control potential confounders. We enhanced rigor by the use of PCA and mediation analysis and used antibody assays well-validated in prior studies. Unlike antibodies binding to cryptococcal protein antigens, which have been shown to cross-react with other fungi,⁷² antibodies to cryptococcal GXM only cross-react with Trichosporon polysaccharides⁷³ and are otherwise specific. Also, we compared the association profiles of antibody and B cell subsets between HIV-negative individuals without known immunocompromising conditions and HIV-positive individuals. To our knowledge, this has not been done previously.

Limitations of our study are that there may have been latent or unaccounted variables that we were not able to control by matching and may have confounded the associations we identified, e.g., concurrent or subclinical infections,¹⁷ which may partially explain the difference in GXM-IgG levels by ART status in the HIV-positive controls; ART or anti-fungal use; and in the HIV-negative group, undiagnosed immunodeficiency. Although only 23% of the HIV-positive cases in our study were on ART, we identified an interaction between ART use and IgG2 and laminarin-IgA, which may be in line with older studies that suggest ART use is associated with normalization of memory B cell levels.⁷⁴ As CM is increasingly being diagnosed in the ART-experienced individuals,⁷⁵ further investigation of antibody and B cell profiles in a larger cohort is warranted. Sex and race may affect antibody responses, limiting generalizability, e.g., IgG2 levels in HIV-negative controls were higher than published norms, likely reflecting reportedly lower IgG2 levels in Whites than Asians,⁷⁶ the only race represented in this study. The HIV-positive cases sourced from the CryptoDex study²⁶ for which matching controls could be found were predominantly male—a demographic that a previous study suggested to have a less effective immune response at controlling C. neoformans infections.⁷⁷

The study of antigen-specific or antibody-producing B cells and functional assays with patient samples were beyond the scope of this study. We only examined B cell and antibody levels at a single time point, hence we were not able to determine if our measured antibody and/or B cell levels preceded or were a response to CM diagnosis or whether our findings were transient and limited to the point of measurement, similar to the temporary decrease in IgG2 levels in severe illness caused by bacterial community-acquired pneumonia.⁷⁸ The mediation analysis was exploratory as it relies on assumptions, albeit biologically plausible, of the causal order between the independent (B-1b-like cells), mediator (IgG2), and dependent (CM status) variables. Nevertheless, since cryptococcosis is a subacute disease that spans a continuum from latency to reactivation and dissemination, the profiles we identified are likely to represent a snapshot along a continuum from rising fungal burden to CM diagnosis.

Despite the limitations of our study, our data suggest that changes in GXM-IgG and/or IgG2 levels may hold promise for earlier diagnosis of high-risk HIV-negative groups, such as OTRs or individuals on chronic immunosuppressants. This is important because HIV-negative patients without known CD4 deficiency are often under-tested using CrAg LFA⁷⁹ as physicians primarily rely on classical symptoms to consider CrAg testing. Our findings suggest that studies of the trajectory of antibody and B cell subset levels over time as a function of cellular immune status could potentially identify antibody-based biomarkers for earlier diagnosis of CM in high-risk HIV-negative individuals.

Notes

Previous presentation: Mycoses Study Group Education and Research Consortium 2022 Biennial meeting, Albuquerque, New Mexico, USA, September 9, 2022, poster presentation; 11^th International Conference on Cryptococcus & Cryptococcosis, Kampala, Uganda, January 12, 2023, Plenary Session 5, oral presentation.

Author contributions

Hyunah Yoon (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review & editing), Antonio S. Nakouzi (Formal analysis, Investigation, Methodology, Project administration, Validation, Visualization, Writing – review & editing), Van Anh Duong (Conceptualization, Data curation, Project administration, Resources, Writing – review & editing), Le Quoc Hung (Conceptualization, Data curation, Project administration, Resources, Writing – review & editing), Tran Quang Binh (Conceptualization, Data curation, Project administration, Resources, Writing – review & editing), Nguyen Le Nhu Tung (Conceptualization, Data curation, Project administration, Resources, Writing – review & editing), Jeremy N. Day (Conceptualization, Data curation, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review & editing), Liise-anne Pirofski (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing – original draft, Writing – review & editing)

Data availability

The data for this project may be obtained with Data Use Agreements with Albert Einstein College of Medicine. All code for data cleaning and analysis associated with the current submission will be available upon request.

Conflict of interest

None.