Defining Discriminatory Antibody Fingerprints in Active and Latent Tuberculosis

Abstract

Tuberculosis (TB) is among the leading causes of death worldwide from a single infectious agent, second only to COVID-19 in 2020. TB is caused by infection with Mycobacterium tuberculosis (Mtb), that results either in a latent or active form of disease, the latter associated with Mtb spread. In the absence of an effective vaccine, epidemiologic modeling suggests that aggressive treatment of individuals with active TB (ATB) may curb spread. Yet, clinical discrimination between latent (LTB) and ATB remains a challenge. While antibodies are widely used to diagnose many infections, the utility of antibody-based tests to diagnose ATB has only regained significant traction recently. Specifically, recent interest in the humoral immune response to TB has pointed to potential differences in both targeted antigens and antibody features that can discriminate latent and active TB. Here we aimed to integrate these observations and broadly profile the humoral immune response across individuals with LTB or ATB, with and without HIV co-infection, to define the most discriminatory humoral properties and diagnose TB disease more easily. Using 209 Mtb antigens, striking differences in antigen-recognition were observed across latently and actively infected individuals that was modulated by HIV serostatus. However, ATB and LTB could be discriminated, irrespective of HIV-status, based on a combination of both antibody levels and Fc receptor-binding characteristics targeting both well characterized (like lipoarabinomannan, 38 kDa or antigen 85) but also novel Mtb antigens (including Rv1792, Rv1528, Rv2435C or Rv1508). These data reveal new Mtb-specific immunologic markers that can improve the classification of ATB versus LTB.

Keywords: HIV; active and latent tuberculosis; antibodies; biomarkers; diagnostics.

Figure 1

Antibody levels against common Mtb antigens differ between ATB and LTB patients. Relative levels of IgG1, IgG3, IgA1 and IgM against 8 common Mtb antigens (LAM, PPD, ESAT6/CFP10, Ag85A and 85B, groES and PSTS3) were quantified via Luminex in the plasma of ATB (n=21) and LTB (n=22) patients in the HIV negative group and ATB (n=12) and LTB (n=22) patients in the HIV positive group. (A) For the heatmap, Z-score transformation of the MFI values was performed, and the median of each group was graphed. Antibody levels that are significantly different between ATB and LTB are graphed separately for all individuals in the HIV negative (B) and the HIV positive (C) groups. A Mann-Whitney U-test test followed by a Benjamini-Hochberg (BH) correction for multiple comparisons was used to test for statistically significant differences ATB and LTB. Multivariate analysis using LASSO and PLS-DA model shows antibody levels comparisons between ATB and LTB in HIV negative (D) and positive (E) populations. Cross-validation accuracy for (D, E) was 62.8% and 67.6%, respectively.

Figure 2

Mtb-specific antibody levels differ between ATB and LTB in HIV negative and HIV positive populations. Relative levels of IgG1, IgG3, IgA1 and IgM against 209 Mtb antigens were quantified via Luminex in the plasma of ATB (n=21) and LTB (n=22) patients in the HIV negative group and ATB (n=12) and LTB (n=22) patients in the HIV positive group. (A) Heatmaps illustrate the median of Z-scored MFI data for each group indicated. (B) The volcano plots characterize the magnitude (log2 fold change of ATB/LTB) and the significance (p values) of antibody levels between ATB and LTB. Values above black dashed lines are statistically different between ATB and LTB (p < 0.05). For adjusted p values, significant data are shown in red, non-significant differences are in black.

Figure 3

Mtb-specific FcR binding distinguishes ATB from LTB among HIV negative and positive individuals. The ability of antibodies to bind to FcgR2AR, FcgR2B, FcgR3AV and FcgR3B was measured in the plasma of ATB (n=21) and LTB (n=22) patients without HIV infection as well as ATB (n=12) and LTB (n=22) patients with HIV. (A) Heatmaps show median values of FcR binding after Z-score transformation to each of the 209 Mtb antigens. (B) Fold change between FcR binding in ATB and LTB as well as the significance (p values) of differences are plotted the volcano plots. Values above black dashed lines are statistically different between ATB and LTB (p < 0.05). For adjusted p values, significant data are shown in red, non-significant differences are in black.

Figure 4

Individuals with ATB and LTB exhibit distinct humoral profile against Mtb antigens regardless of HIV infection. Multivariate analysis comparing antibody response in ATB and LTB among HIV negative (A, B) and HIV positive (E, F) populations. (A, E) PLS-DA models showing dot plots of nearly non-overlapping antibody features in ATB (blue) and LTB (yellow) from HIV negative (A) and positive (E) groups. Ellipses show 95% confidence intervals. PLDS-DA models were trained using LASSO-selected features that are plotted on VIP plots for HIV negative (B) and positive (F) patients. Cross-validation accuracy for (A, E) was 74.2% and 79.1%, respectively. Antibody features showing significant differences between ATB and LTB are graphed for the HIV negative (C) and HIV positive (G) cohorts. A Mann-Whitney U-test test followed by a Benjamini-Hochberg (BH) correction for multiple comparisons was used to test for statistically significant differences ATB and LTB. Network correlations depicting the additional non-LASSO-selected features that were correlated with the LASSO-selected parameters statistically different between ATB and LTB in the HIV negative (D) and positive (H) groups. Data on antibody levels are in purple, features related to FcγR binding are in green.

Figure 5

Discriminatory features between ATB and LTB. ROC curves showing the sensitivity and the specificity of potential biomarkers discriminating ATB and LTB in HIV negative (A) and positive (B) populations. (C) Loading plot showing LASSO-selected features enriched in ATB, both among HIV negative and HIV positive individuals (upper right quadrant) versus those enriched in LTB among HIV negative and HIV positive groups (low-left quadrant). Features in the upper left quadrant are increased in ATB compared to LTB among HIV+ but are enriched in LTB among HIV-. Features in the low right quadrant are enriched in LTB compared to ATB among HIV+ but are higher in ATB among HIV-.