Mycobacterium tuberculosis infection, immune activation, and risk of HIV acquisition.

BACKGROUND: Although immune activation is associated with HIV acquisition, the nature of inflammatory profiles that increase HIV risk, which may include responses to M. tuberculosis (Mtb) infection, are not well characterized.
METHODS: We conducted a nested case-control study using cryopreserved samples from persons who did and did not acquire HIV during the multinational Step clinical trial of the MRKAd5 HIV-1 vaccine. PBMCs from the last HIV-negative sample from incident HIV cases and controls were stimulated with Mtb-specific antigens (ESAT-6/CFP-10) and analyzed by flow cytometry with intracellular cytokine staining and scored with COMPASS. We measured inflammatory profiles with five Correlates of TB Risk (CoR) transcriptomic signatures. Our primary analysis examined the association of latent Mtb infection (LTBI; IFNγ+CD4+ T cell frequency) or RISK6 CoR signature with HIV acquisition. Conditional logistic regression analyses, adjusted for known predictors of HIV acquisition, were employed to assess whether TB-associated immune markers were associated with HIV acquisition.
RESULTS: Among 465 participants, LTBI prevalence (21.5% controls vs 19.1% cases, p = 0.51) and the RISK6 signature were not higher in those who acquired HIV. In exploratory analyses, Mtb antigen-specific polyfunctional CD4+ T cell COMPASS scores (aOR 0.96, 95% CI 0.77, 1.20) were not higher in those who acquired HIV. Two CoR signatures, Sweeney3 (aOR 1.38 (1.07, 1.78) per SD change) and RESPONSE5 (0.78 (0.61, 0.98)), were associated with HIV acquisition. The transcriptomic pattern used to differentiate active vs latent TB (Sweeney3) was most strongly associated with acquiring HIV.
CONCLUSIONS: LTBI, Mtb polyfunctional antigen-specific CD4+ T cell activation, and RISK6 were not identified as risks for HIV acquisition. In exploratory transcriptomic analyses, two CoR signatures were associated with HIV risk after adjustment for known behavioral and clinical risk factors. We identified host gene expression signatures associated with HIV acquisition, but the observed effects are likely not mediated through Mtb infection.

Background

Several lines of evidence suggest that systemic inflammation increases risk of HIV acquisition. For example, in the HIV Vaccine Trials Network Step Study (V520-023/HVTN 502), an efficacy trial of an Adenovirus 5 (Ad5)-vectored HIV-1 vaccine, some vaccinees experienced a transient increased risk of HIV acquisition for several months after vaccination [1, 2]. In Step, ELISpot mock responses (interferon-gamma [(IFN-ɣ] secretion in the absence of antigen), but not HIV-antigen stimulated responses, directly correlated with a 61% increase in HIV acquisition risk [3]. Furthermore, CAPRISA-004 and Partners in Prevention studies each identified different cytokine profiles associated with HIV acquisition [4, 5]. Other studies suggest associations between herpes simplex virus-2 and chronic filarial infections with HIV acquisition [6-8]. Although there is some understanding of the local genital inflammation and microbiome changes associated with HIV susceptibility [9-11], the biologic underpinnings of peripheral blood immune profiles potentially associated with HIV risk are not well understood. It may be that other infections or exposures induce immune phenotypes that increase susceptibility of CD4 T cells and macrophages to HIV infection.

Mycobacterium tuberculosis (Mtb) latently infects approximately 24% of the world’s population, with an estimated 10 million new cases of clinical disease (TB) in 2019 alone [12-14], Latent Mtb infection (LTBI), is defined by a positive response to Mtb antigens (Purified Protein Derivative (PPD) skin test or an IFN-ɣ release assay (IGRA)) in the absence of clinical symptoms. Recent data indicates that many asymptomatic individuals have “incipient TB” with evidence of systemic inflammation [15], which is associated with an increased risk of progression to clinical TB disease [16, 17]. While the effects of HIV on risk of LTBI progression have been intensively investigated, the effect of Mtb infection on risk of acquiring HIV has never been formally assessed, other than in ex-vivo experiments [18-21]. In a recent study, infant macaques were vaccinated with TB vaccines and then challenged with simian immunodeficiency virus (SIV) orally. Animals vaccinated with either an Mtb-SIV auxotroph vaccine or BCG vaccine demonstrated significantly increased SIV risk compared to placebo recipients. Both vaccines resulted in immune activation, which correlated with SIV acquisition [22].

Because the global burden of Mtb infection is high and the lifetime risk of progression to active disease is only ~10% in HIV-negative persons, no countries with co-endemic TB and HIV routinely treat LTBI in HIV-negative persons [23]. If treating LTBI were to reduce the risk of acquiring HIV, then targeting LTBI treatment to the subset of individuals most at-risk for HIV could prevent development of TB disease as well as decrease risk of HIV. We therefore designed a study of LTBI- and Mtb-associated immune activation using prospectively collected samples from an HIV vaccine trial to address this question. We included several Correlates of Risk (CoR) transcriptomic signatures that have been validated to predict TB-disease states, including incipient active TB and TB treatment success, based on differential human gene activation.

Methods

Study population

We conducted a case control ancillary study nested within the completed Step study (V520-023/HVTN502). The Step study enrolled 3,000 high-risk HIV-negative males and females aged 18–45 who received the MRKAd5 HIV-1 gag/pol/nef vaccine at 34 study sites in the Americas, Caribbean, and Australia from 2004–2007. Participants received vaccine/placebo injection at 0, 4, and 26 weeks. HIV testing was done at 6-monthly visits for up to 4 years of in the Step study and the follow-on HVTN 504; stored blood from the prior visit was tested for HIV RNA to more precisely time HIV acquisition [2]. A total of 172 incident infections were diagnosed in men (3.3/100 person-years) and 15 infections in women (0.45/100 PY) [1]. We identified as cases all participants within Step and HVTN 504 with incident HIV who had available peripheral blood mononuclear cells (PBMC) at the last study visit prior to HIV acquisition. Cases who acquired HIV before week 8 (first available PBMC) were excluded. Controls were selected by HVTN staff as persons who did not acquire HIV during study follow-up, matched to cases 2:1 on study site and treatment arm, and with a pooled gender distribution similar to cases. We acquired PBMC from the study visit corresponding to that of their matched case; if no sample from a matched visit was available a new control was selected. An unblinded team member not involved in the laboratory work constructed a batching scheme that allowed for a balanced selection of cases and controls throughout the laboratory work, while allowing our laboratory team to remain blinded to participant characteristics. The following baseline variables from the Step study were assessed: age, gender, self-reported race/ethnic identity, and study site; study treatment (vaccine/placebo); baseline Ad5 antibody titer, self-reported behavioral HIV risks: drug use, number and type of sexual partners, condomless sex, substance use; circumcision status and HSV-2 serostatus for males. Step participants were generally not eligible for TB prophylaxis and no prophylactic or therapeutic TB drugs were documented as given to any participant during study follow up. Because this ancillary study used only deidentified data and samples from the Step trial, the work described in this manuscript was determined not to be engaged in human subjects research by the University of Washington Institutional Review Board. All participants who participated in the Step study and whose samples and data were used here were verified to have given specific informed consent for future use of stored samples at time of participation in the vaccine trial.

Cell activation and flow cytometry

Cryopreserved PBMCs were thawed, washed and rested in RPMI 1640 media containing 10% heat-inactivated fetal bovine serum (FBS) overnight prior to antigen stimulation with a concentration of 2 x 106 cells /mL. PBMCs were counted with Guava easyCyte (Millipore) using Guava ViaCount reagent (Luminex) and GuavaSoft v.2.6 software. Samples with less than 66% viability were discarded. Cells were stimulated with a pool of early secretory Mtb antigen target-6 (ESAT-6) and culture filtrate protein (CFP-10) (BEI Resources). PMA (25ng/mL)/ionomycin (1ug/mL) was used as a positive control and DMSO (0.5%) was used as a negative control. In addition, costimulatory antibody anti-CD28/49d, cytokine secretion inhibitor Brefeldin A and Monensin were added to each stimulation cocktail. Cells were lysed and permeabilized with FACS Lyse and FACS Perm-II buffer and stained on a BD LSRFortessa with an antibody panel developed for analyzing CD4 and CD8 T-cell responses (ICS) including IFN-ɣ, TNF, IL-2, and IL-17a (Table A in [24]. Flow cytometry data were analyzed in FlowJo™ v10.7. Each sample was compensated and gated manually.

RNA isolation and RT-PCR

Approximately 2 x 106 PBMCs from each sample were re-suspended in RNAlater (Invitrogen) preservative and stored at -20°C for processing [25]. Total RNA was isolated using RNeasy spin columns (Qiagen) and cDNA templates for qRT-PCR were generated using Applied Biosystems high-capacity cDNA reverse transcriptase kits. qRT-PCR was performed with 44 primer-probe sets (Table B in ) from Taqman using the Fluidigm 96.96 dynamic array platform. cDNA was preamplified using a pool of specific TaqMan primer-probe sets for 16 cycles with a 15 second denaturation step at 95°C and 4 minutes at 60°C. The pre-amplified reaction was added to the Fluidigm 96.96 dynamic array platform with TaqMan gene expression assays and amplified for an additional 40 cycles. A positive control sample was run on every chip to monitor consistency. Primer-probe sets were included to encompass the components of 5 previously described CoR signatures from peripheral blood associated with TB disease outcomes. Signature scores were calculated from raw Ct values in R v3.6.1 as previously described [25-30] (Table B in ; additional primer probes not pertaining to selected CoRs were included for future analysis).

Statistical analysis

The outcome of interest in all analyses was HIV acquisition. To evaluate the association of Mtb antigen-specific T-cell activation and HIV acquisition, we used three measures:

Individuals with LTBI were defined using a flow cytometry-based method that is strongly correlated with the tuberculin skin test and IGRA. Samples were classified as positive if the frequency of IFN-ɣ+ CD4+ cells stimulated with ESAT-6/CFP-10 pooled peptides doubled compared to negative control [31-35]. LTBI was prespecified as the primary Mtb exposure variable.

We used Combinatorial polyfunctionality analysis of antigen-specific T-cell subsets (COMPASS) to determine overall Mtb-antigen-specific T cell activation. COMPASS is an analytic tool that uses a Bayesian hierarchical framework to model all observed cell subsets. COMPASS outputs functional scores (FS) and polyfunctional scores (PFS) that define the posterior probabilities of antigen-specific response across cell subsets, described by a single numerical score that ranges from 0 to 1 [36]. We have previously used COMPASS to investigate polyfunctionality in CD4+ T cells in an epidemiologic study of South African adolescents screened for LTBI, as well as to predict HIV-specific responses in a prior HIV vaccine trial [37, 38]. FS and PFS derived from COMPASS output were included as untransformed variables of secondary interest.

To determine whether transcriptomic evidence of Mtb-associated immune activation also predicts risk of HIV acquisition, we used Correlates of Risk (CoR) scores previously validated to detect different TB-associated states. The primary signature of interest was RISK6, a 6-gene transcriptomic CoR signature that predicts progression to TB disease (incipient TB) [26]. Additional CoR scores assessed in exploratory analysis include Suliman4 and Maertzdorf4, both of which also predict incipient TB; Sweeny3, which accurately differentiates between current LTBI vs active TB; and RESPONSE5, which predicts cure at time of TB treatment initiation among those with active TB (Table B in ) [27-30]. CoR scores that were not normally distributed were log10 transformed (RISK6, and Suliman4). To account for score values of zero (n = 3; all RISK6), zeros were replaced by dividing the next lowest value by two.

For all measures, the association between exposures and HIV acquisition was estimated using univariate and multivariate conditional logistic regression accounting for the variables used to match controls to cases (country of residence and treatment arm within the Step study). Variables assessed for inclusion in multivariate regression included known predictors of HIV risk in the Step study (i.e. gender, validated behavioral risk score (scale 0–7) [39], baseline Ad5 titer, and self-reported race/ethnicity). Each variable of interest was assessed for association with HIV acquisition using conditional logistic regression, and associations that were significant at p<0.1 were included in the multivariate model. For sub-analyses restricted to males, we included HSV-2 serostatus and circumcision, as these were documented in the parent trial for males only. Analyses of the association of PFS, FS, and CoR scores with HIV acquisition were repeated after stratification by LTBI status. Stratified analyses were performed as standard logistic regression with matching variables included, as logistic regression conditioned on LTBI would have excluded all case/control sets with discordant LTBI status. PFS, FS, and CoR scores were reported as the Odds Ratio (OR) for 1 standard deviation change to allow for comparability across outcomes. As a sensitivity analysis, we repeated the primary analyses restricted to case/control sets with samples within 6 and 12 months prior to HIV diagnosis. All analyses were conducted in R v3.6.1 and Stata v15.1 (College Station, TX, USA, 2017).

Results

We selected 155 cases who acquired HIV and 310 controls (n = 465), with baseline characteristics presented in Among cases, the median interval between sample collection date and HIV diagnosis was 287 days (IQR 161, 483). The majority of participants (88%) were from Peru or the US. By design, there was no difference in distribution of vaccine treatment arm or country in cases vs. controls, although there was a trend toward more males among cases. As expected, the behavioral risk score previously found to be associated with HIV acquisition was higher in cases than controls (3.3 vs 2.8, SD 1.2, p<0.0001), and HSV-2 seropositivity was more common in male cases than controls (40.1% vs 30.7%, p = 0.049).

To assess one of our primary hypotheses, whether Mtb infection was associated with HIV acquisition, we examined 454 participants, including 152 cases and 302 controls (n = 11 missing samples or gating failures) for LTBI status via flow cytometry. Overall, 94 (20.7%) study participants were considered LTBI positive via detection of ESAT-6-CFP-10-specific CD4+ T cells: 65 (21.5%) controls compared to 29 (19.1%) cases. LTBI positivity was not associated with HIV acquisition in unadjusted or adjusted conditional logistic regression analyses (aOR 0.85 (0.51, 1.43) per 1 SD change, p = 0.73) (). When stratified by gender, the inference did not change. Among men, when adjusting for all covariates in the primary models as well as HSV-2, LTBI status was not associated with HIV acquisition (aOR: 0.99; 95% CI 0.58, 1.69). Likewise, there was no association between LTBI status and HSV-2 serostatus (31.8% seropositive in LTBI negative vs 30.0% in LTBI positive, p = 0.77).

We also performed two exploratory analyses with ESAT-6/CFP-10-specific functional and polyfunctional cytokine expression, as measured through the COMPASS FS and PFS, respectively, which were both significantly higher in LTBI-positive versus LTBI-negative samples (p<0.0001; ). Among all 152 cases the median FS was 0.026 (IQR 0.009, 0.077) and of 302 controls the median FS was 0.029 (0.009, 0.077). In controls, the median PFS was 0.015 vs 0.015 among cases and controls. shows box plots and a heatmap of COMPASS posterior probabilities by case/control status. The distributions of mono- and poly-functional Mtb-specific subsets across persons who did or did not acquire HIV were similar. When assessing the association using conditional logistic regression, neither FS (OR 0.94 per 1 SD change, 95% CI 0.76, 1.16) nor PFS (OR 0.96, 95% CI 0.78, 1.18) were associated with HIV acquisition. we found no evidence of an association of FS or PFS with HIV acquisition when evaluating only LTBI positive participants, (Table C in ).

Combinatorial polyfunctionality analysis of antigen-specific T-cell subsets (COMPASS) describes Mtb-antigen-specific T cell activation.

Box and whisker plots depict median and interquartile ranges for FS and PFS scores as well as the range. A. COMPASS Polyfunctional Scores (PFS) demonstrate expected higher levels of Mtb-specific cellular immune activation in persons with LTBI than those without (p<0.0001). B. COMPASS Mtb-specific Functional Scores (FS) were no different between persons who acquired HIV and controls. PFS was similarly indistinguishable between cases and controls. C. Heatmap of COMPASS posterior probabilities depicts combinations of intracellular cytokine staining composing the combinatorial polyfunctional subsets of Mtb-specific CD4+ T cells. Columns show the subsets with detectable antigen-specific responses color-coded by the cytokines they express and ordered by the degree of functionality from one function on the left to 3 of 4 functions on the right; the combinations of cytokines are found in colored boxes at the bottom. Horizontal rows depict one participant, with lime shaded rows representing control samples stacked above violet rows representing persons who acquired HIV. The intensity of purple shading of each box shows the probability that a given participant sample that expresses the subset has a Mtb-antigen specific response ranging from white (zero) to purple (one). This heatmap depicts relatively similar distributions of mono- and polyfunctional Mtb-specific subsets across persons who did or did not acquire HIV.

RISK6 score was calculated for 439 participants (n = 141 cases, n = 298 controls) in whom RNA extraction was successful. The mean of untransformed RISK6 scores among cases was 0.119 (SD 0.147) compared to 0.107 (SD 0.141) among controls (). In both unadjusted and adjusted analyses, RISK6 was not associated with HIV acquisition (aOR 1.15 per 1 SD change, 95% CI 0.92, 1.44) (). shows the distribution of CoR scores in cases who acquired HIV and controls for RISK6, as well as the other four CoR signatures (Suliman4, Sweeney3, Maertzdorf4, and RESPONSE5) examined in exploratory analyses. In multivariate conditional logistic regression, two of the four CoR measures demonstrated associations with HIV acquisition. A one SD increase of the Sweeney3 score was associated with a 38% increased odds of HIV acquisition (aOR 1.38, 95% CI 1.07, 1.78). In contrast, the RESPONSE5 CoR score was associated with a moderate decrease in odds of HIV acquisition (aOR 0.78, 95% CI 0.61, 0.98). There were no important trends in CoR scores between those with lower baseline Ad5 titer, gender, or behavioral risk score. HSV-2 serostatus (in males only) was associated with score of one signature and vaccine vs placebo treatment was associated with a different signature. The mean of three CoR scores differed significantly by country of residence. Bivariate associations between predictors of HIV acquisition and each CoR are shown in Table D in . In adjusted analyses, Ad5 titer was not influential in evaluating associations between any CoR score and HIV acquisition.

Distribution of 5 TB Correlates of Risk (CoR) scores among participants who subsequently acquired HIV vs controls.

Box plots demonstrate median and interquartile range of values, with dots completing the range. RISK6 and Suliman4 were log10 transformed as this allowed for the data to be normally distributed and for all scores to be evaluable on a similar scale.

Due to the long interval between PBMC collection and HIV diagnosis in some cases, we performed sensitivity analyses including the 282 participants (n = 94 cases, n = 188 controls) with samples within 12 months, and then the 159 participants (n = 53 cases, n = 106 controls) with 6 months or less between sampling and HIV diagnosis. The aOR for RISK6 increased to 1.28 (0.94, 1.75) but neither this or any of the other primary exposures (LTBI, FS, or PFS), became statistically significant for participants sampled within a year or 6 months of diagnosis. As in the full dataset, RESPONSE5 remained associated with decreased risk of HIV and Sweeney3 with the association strengthened by >10% (OR 1.56 (1.10, 2.21)). Due to loss of power, the point estimates for several exploratory analyses changed by >10% for the participants sampled within 6 months, but no findings were significant (Table E in ).

In adjusted analyses stratified by LTBI status, a one SD increase in logRISK6 score was associated with a trend toward decreased odds of HIV acquisition among LTBI-positive participants (aOR 0.58, 95% CI 0.34, 0.99) and a trend toward increase in odds of HIV acquisition among LTBI negative participants (aOR 1.24, 95% CI 0.97, 1.59) (Table F in ). Stratified by LTBI status, some associations between CoR scores and HIV remained in the LTBI negative participants, but not in the smaller subset of LTBI positive participants. In univariate analyses however, none of the CoR scores were associated with LTBI status (Table G in ).

Discussion

In this study, we accessed rare, cryopreserved specimens from the completed Step HIV vaccine trial to evaluate the influence of sub-clinical Mtb infection on HIV acquisition risk, employing a case-control design. Despite the fact that many of the biomarkers that have previously been associated with increased risk of acquiring HIV are also those elevated in subclinical/incipient Mtb infection or TB disease, we found little evidence that prior infection with Mycobacterium tuberculosis increases risk of acquiring HIV. To our knowledge, there are no cohorts in which LTBI status has been determined using clinical methods (TST or IGRA) among comparable persons with subsequent HIV endpoints. Our use of comprehensive flow cytometry and COMPASS analysis allowed us to retrospectively determine LTBI status and to characterize Mtb antigen-specific polyfunctional T-cell responses within a homogenous trial population with subsequent evaluation of HIV status. Neither LTBI status, nor polyfunctional cellular activation scoring was associated with risk of HIV acquisition, either bivariately, or after controlling for the previously described HIV risk factors. Transcriptional TB CoR signatures do not measure Mtb-specific responses, but rather the differential expression of genes associated with incipient or active TB disease; we hypothesized that these same signatures, many of which include interferon stimulated genes (ISGs), could also reveal HIV susceptibility states. However, the primary CoR signature of interest, the RISK6 signature, comprising several ISGs, was not associated with HIV acquisition in the overall study. Although aORs differed by LTBI status, as RISK6 was associated with lower HIV risk in those with LTBI and showed a trend towards increased HIV risk in those without LTBI, we could not rule out that this difference was due to chance. It is possible that pathways other than the specific ISG signaling characterized by RISK6 are associated with HIV risk and are also differentially regulated by co-infections. In addition, evaluation of other CoR profiles suggested associations with HIV acquisition, suggesting that there may be an unmeasured source of immune activation predictive of HIV acquisition. The different gene sets in each signature describe different activation patterns, and possibly various underlying exposures. Because LTBI status was not independently associated with any of the 5 CoRs in this cohort, especially RISK6, Suliman4, or Maetzdorf4, it is likely that there was little incipient TB in the Step study. It is therefore possible that these CoRs were detecting ISGs provoked by stimuli other than Mtb infection. The Mtb-specific cellular responses that define LTBI are well-known to persist despite LTBI treatment or spontaneous clearance of Mtb infection [40, 41], and therefore persons identified as LTBI in this study may not have had differential ISG or other gene expression measured concurrently in that sample or at the time of HIV exposure.

The results of the exploratory analyses provide some evidence that a transcriptomic biomarker could evaluate HIV risk, regardless of Mtb infection. The Sweeney3 signature was most strongly associated with HIV risk in unadjusted and adjusted analyses, and in the LTBI-negative, but not the LTBI-positive group. This three-gene set predicts the presence of active TB disease versus LTBI with better accuracy (86% sensitivity, 86% specificity) than most combinations of clinical prediction tools, and with similar sensitivity to GeneXpert-MTB RIF sputum tests [42, 43]. Because the three genes, GBP5, DUSP3, and KLF2, are associated with macrophage regulation and other immune pathways [44-46], this gene set may detect a high-risk systemic immune status, provoked by genital or gastrointestinal infection or altered microbiome composition, which could also be differentially associated with LTBI status. Alternatively, the transcription of these genes could relate to other stable host factors that impact both Mtb and HIV susceptibility.

In contrast, the RESPONSE5 profile evaluates differential gene expression that predicts cure after TB treatment among those with active TB, wherein a high score represents higher innate immune activation [28], and in this study, higher scores were associated with lower odds of HIV acquisition. These findings suggest that Mtb infection and polyfunctional Mtb antigen-specific T-cell activation are not associated directly with HIV risk, but that other unmeasured exposures or intrinsic factors may ultimately activate these same host transcriptional pathways and explain the observed association between some transcriptional signatures and HIV acquisition. For example, some risk scores were modestly associated with receipt of the Ad5 HIV DNA vaccine in the parent trial, with HSV-2 serostatus, or with country of residence.

Our study has several limitations, the foremost of which was lack of availability of stored PBMC close to the time of HIV acquisition in some cases, which we tried to address through sensitivity analyses restricted to cases with proximal samples. Some samples were collected more than one year prior to documented HIV diagnosis. Due to wide intervals between study visits (up to 6 months between visits), many persons likely acquired HIV much closer to the sampling timepoint than indicated by their diagnosis visit, but the timing of HIV acquisition is not clearly pinpointed. That said, CoR signatures are able to predict progression to incident TB up to 24 months prior to clinical disease and show accurate prediction within 1 year [17, 26]. As the parent Step study occurred within some countries with moderate or high TB prevalence, there is a possible risk that interval Mtb infection or sub-clinical reactivation occurred between sampling and HIV outcome. Additionally, LTBI status was not determined in the Step study; thus, we were unable to use clinically accepted measures of LTBI status, such as IGRA or TST. Having access to a single blood sample limited our ability to assess interval LTBI status conversion or reversion prior to HIV acquisition, and there is a possibility of other unevaluated infections or other confounders. Nonetheless, the Step study represents a rare opportunity to address this issue as there are few, if any, cohorts in which Mtb infection status, future HIV outcome, and longitudinal cryopreserved specimens are available. Instead, we used a rigorous Mtb-antigen-specific flow cytometry assay that provided richer descriptive data, not only about LTBI status, but about the degree and quality of Mtb responses. Other strengths of this study include rigorous characterization of study participants as part of the parent Step study, and novel application of TB CoR signatures.

Based on this translational study we would not expect that treating LTBI in persons at risk for HIV would decrease risk of HIV acquisition substantially, especially in the context of increasingly effective and available HIV pre-exposure prophylaxis (PrEP) options. Treating LTBI has other intrinsic benefits, including prevention of TB disease and Mtb transmission, especially in those who are living with HIV or otherwise at high risk of developing active TB disease.

To our knowledge there is little published data on differential gene regulation being associated with HIV susceptibility or protection. While we found variable associations with some purpose-designed TB CoR signatures, global transcriptomic analysis could assist in discovery of similar, simpler HIV CoR signatures. Although such HIV CoR signatures would be unlikely to replace patient-reported indications for PrEP and other prevention measures, such signatures could be beneficial in risk-stratifying participants within HIV prevention and vaccine studies. Transcriptomic signatures could also be employed to detect high-risk post-vaccine immune responses in early-phase trial volunteers. Designing purpose-built signatures could help identify vaccine candidates that might induce risky immune states for HIV acquisition, such as occurred in the Step study, before moving the vaccine candidate to larger trials in HIV-exposed populations.

Conclusions

In this study, we found no evidence that LTBI or polyfunctional Mtb-antigen-specific T cell responses were associated with risk of HIV acquisition in a population at high risk for HIV. However, the exploratory analyses provided a proof of concept that transcriptomic signatures could provide additional information about immunologic profiles of HIV susceptibility beyond known behavioral and viral co-infection risk factors identified in the Step study and other prior studies.

Supplemental Tables A and B provide materials used in flow cytometry and transcriptional signature analysis, respectively.

Tables C, E, and F provide sensitivity analyses stratified or restricted by LTBI status (C, F) and time restricted analyses (E). Table D shows univariate analyses between predictors of HIV acquisition and CoR scores. Table G shows univerate associate.

(DOCX)

Click here for additional data file.

6 Mar 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Manish Sagar, MD

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

Reviewer #3: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This nested case-control study used prospectively collected data to investigate whether Tuberculosis-associated immunological variables measured in PBMC samples could predict HIV acquisition. The methods are in general appropriate and clearly reported.

The Benjamini-Hochberg methods using false discovery rates (FDR) was developed for the case where very large numbers of significance tests are performed, in the absence of structure or hierarchy in the set of tests. I am not sure whether it is so appropriate in this study: for the main analysis there seem to have been 2 tests (univariate, multivariate) for measure (1) =LTBI, 4 for measure (2) =FS, PFS and 10 tests for measure (3) =CoR) – 16 in total. Sensitivity analyses on subsets defined by sample collection date resulted in a further 2 x 16 tests, while stratification by LBTI status produced 10 further tests. Other significance tests (supplement) were used to investigate correlations between measures, but not association with HIV, so these should presumably not be counts in the BH procedure?

A suggestion would be to restrict the multiple testing adjustment to the main analyses, defining either univariate or multivariate regression as primary (i.e. 8 primary tests), and to describe all other analyses as sensitivity analyses and the corresponding tests as purely descriptive.

In any case, for clarity, the number of tested hypotheses and the preset FDR (=0.05?) should be stated. In order to allow the reader to apply alternative multiple testing approaches, it would be helpful to report the nominal p-values as well as the adjusted ones.

Minor points

1. Please describe more precisely how matched controls were selected for cases with more than two potential controls – using random numbers? How was the balance in pooled gender distribution maintained?

2. P.7 missing word(s) in ‘which predicts cure of at time of TB treatment initiation’?

3. What was the reasoning behind stratification by LTBI status for analyses of the association of PFS, FS, and CoR scores with HIV acquisition (p.7).

4. Were confidence intervals also adjusted using FDR methods as described by Benjamini et al. (reference 40)?

5. On p.10 the mean and standard deviation are given for FS and PFS, although Fig 1 boxplots show very skew distributions. Therefore, median and interquartile range would be more informative.

6. The legend to Fig. 1 seems to state the wrong labels 1C and 1D (heatmap is 1C, there is no 1D).

7. In Fig. 2, the log transformation does not seem to have resulted in a normal (symmetrical) distribution for RESPONSE5 (many extreme negative log values).

Reviewer #2: The authors leverage a unique cohort sample bank to investigate association of tuberculosis (TB) infection and incipient disease with HIV acquisition. Strengths of the manuscript and project are the longitudinal data and access to samples prior to HIV infection. The stronger finding is that TB infection, as defined by the team, was not associated with higher HIV infection risk. The team though makes leaps between signature patterns and interpreting their meaning that seems beyond the associations they can more accurately describe. Other weakness includes lack of TB infection testing in host (which the team acknowledges) and long intervals between pre-HIV blood sample and HIV infection. I believe the manuscript needs to address the following issues.

• In the abstract, describe the cohort at high risk for HIV acquisition (not assume knowledge of what the Step MRKAd5 HIV-1 study included). Last sentence in the conclusion that host gene expression is associated with HIV acquisition feels the wrong focus, instead that pattern is associated with higher likelihood of infection is what I think is captured.

• Throughout the manuscript, concerned saying that a signature predicts HIV acquisition, given there are many mediating steps including need for exposure which will vary by participant. Instead would stick with associated with, with I think main hypothesis which the signature indicates a potential increased vulnerability to successful infection.

• Multiple transcript signatures were tested, with 1 associated with a significant increased odds and another with a lower odds. Would benefit from a clearer indication of what was considered a cutoff with each score that is meaningful and what number of participants were in the “incipient” TB group by each score. Were there enough in that category to have captured their particular risk?

• The discussion appropriately focuses on lack of association with TB infection and HIV infection in this cohort. But the leap that transcriptional signatures might be revealing HIV susceptibility is not well supported, as that is not what these signatures were developed for. Can consider future development but seems inappropriate to interpret from this data.

• Would keep focus on paper as this is exploratory, not making major decisions such as role of TB preventive therapy.

Reviewer #3: Bender Ignacio et al takes advantage of the Step MRKAd5 HIV-1 vaccine study to use samples from a cohort in which Mtb status and approximate time of HIV infection are known to determine if several risk factors associated with latent TB infection or immune activation correlated with HIV infection. In general, no differences were observed and latent TB infection did not appear to be a predictive risk for HIV-1. Even though the data are “negative,” this study, despite the potential caveats pointed out by the authors, is one of the few studies using human samples providing a useful reference for others examining infections in other cohorts. Some minor criticisms include:

1. Page 10, there is a reference to data not shown. These data should be included if not in the text, then in the supplemental data.

2. This is not an easy paper to read; there is an "alphabet soup" of clinical measurements that at times detract from the overall conclusions. Its clinical focus on the outcomes and statistical analysis rather than what the assays are actually measuring limits the appeal to what may be a more general audience. Defining and discussing the immune and cellular signatures in the context of the results, would put into perspective of why these are relevant correlates for TB and HIV and provide a foundation for those with broader interests in immunology, TB, HIV and coinfections. This is also an opportunity to expand the discussion and put into the context with the body of work that has suggested cytokine and functional immune changes.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Jeremy Franklin

Reviewer #2: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

5 Apr 2022

Point by point response to reviewers:

Reviewer #1: This nested case-control study used prospectively collected data to investigate whether Tuberculosis-associated immunological variables measured in PBMC samples could predict HIV acquisition. The methods are in general appropriate and clearly reported.

The Benjamini-Hochberg methods using false discovery rates (FDR) was developed for the case where very large numbers of significance tests are performed, in the absence of structure or hierarchy in the set of tests. I am not sure whether it is so appropriate in this study: for the main analysis there seem to have been 2 tests (univariate, multivariate) for measure (1) =LTBI, 4 for measure (2) =FS, PFS and 10 tests for measure (3) =CoR) – 16 in total. Sensitivity analyses on subsets defined by sample collection date resulted in a further 2 x 16 tests, while stratification by LBTI status produced 10 further tests. Other significance tests (supplement) were used to investigate correlations between measures, but not association with HIV, so these should presumably not be counts in the BH procedure?

A suggestion would be to restrict the multiple testing adjustment to the main analyses, defining either univariate or multivariate regression as primary (i.e. 8 primary tests), and to describe all other analyses as sensitivity analyses and the corresponding tests as purely descriptive.

In any case, for clarity, the number of tested hypotheses and the preset FDR (=0.05?) should be stated. In order to allow the reader to apply alternative multiple testing approaches, it would be helpful to report the nominal p-values as well as the adjusted ones.

Thank you for this discussion. We had previously erred conservatively by adjusting all presented analyses, even though the primary hypotheses are few, and the majority of the paper, as pointed out by this reviewer, was done as exploratory testing to evaluate further associations after the primary hypotheses were found null. We have two primary hypotheses in this paper: 1. whether LTBI, measured by the frequency of IFN-ɣ+ CD4+ cells stimulated with ESAT-6/CFP-10, is associated with HIV acquisition, or 2. Whether the RISK6 CoR transcriptional signature is associated with HIV acquisition. We also examine whether several exploratory variables including flow cytometry (COMPASS derived FS and PFS scores) and transcriptional (Suliman4, Maertzdorf4, Sweeny3, RESPONSE5) measurements are associated with HIV acquisition. We revised the manuscript to more clearly distinguish these primary from secondary hypotheses and analyses. For example, we added a horizontal cut to Tables 2 and 3 to distinguish the primary analyses on top from exploratory below the line in both cases.

We had previously only adjusted for the number of tests within a given hypothesis set but had done this separately for each set of data presented (results from flow cytometry as a separate set from transcriptomic signatures, which resulted in correction for only a few comparisons, except in the supplementary tables). We did not count univariate and multivariate analyses as separate tests since they provide different types of information about the same hypothesis test. For a similar reason, we feel that providing both the nominal and BH adjusted p values puts disproportionate emphasis on the significance test, rather than the effect size and direction, and the inference that can be gained by comparing adjusted and unadjusted tests of the primary hypotheses.

Therefore, given that there are only two primary hypotheses tested (LTBI or RISK6 association with HIV risk), we are no longer presenting unadjusted p values for the flow analyses in Table 2, which includes only 3 tests (1 primary and 2 secondary), with no change in inference with either strategy (none approached significance with nominal p-values). For the CoR scores, we had prespecified RISK6 as the primary exposure, and therefore now present RISK6 as the single primary signature of interest and more clearly specify that the other 4 scores were done as exploratory analyses. Again, the nominal p value for RISK6 did not approach 0.05, so we are embracing the null hypothesis with either method.

At the suggestion of the reviewer, we have reverted to nominal p values for the exploratory analyses in the supplement or else do not display p values, as the overarching goal of these additional analyses was to identify plausible explanations for which known clinical or demographic risk factors for HIV, could be associated with immune responses that could render a person more susceptible to HIV if Mtb itself was not a risk factor.

Minor points

1. Please describe more precisely how matched controls were selected for cases with more than two potential controls – using random numbers? How was the balance in pooled gender distribution maintained?

The HVTN laboratory team offered controls based on availability of PBMCs with the matched visit to the cases selected, of those who still had valid consent for future use. The list was generated by HVTN statisticians based on matching requirements, and since they were not able to be directly matched on gender, they selected roughly the same number of female and male controls as in cases. The HVTN laboratory did this before providing the entire transfer of samples through their ancillary studies mechanism, which arranges receipt of samples and a matching dataset for external or collaborating investigators.

An unblinded team member not involved in the laboratory work constructed a batching scheme that maintained a rough 2:1 proportion of cases and controls for each planned plate but then stripped these details from the sample manifest. These steps allowed for balanced selection of controls to cases throughout the lab work, and for our laboratory team to remain blinded to case/control status and the other clinical/demographic details as they ran the samples.

We have added more detail in the methods to this point.

2. P.7 missing word(s) in ‘which predicts cure of at time of TB treatment initiation’?

Addressed, thank you.

3. What was the reasoning behind stratification by LTBI status for analyses of the association of PFS, FS, and CoR scores with HIV acquisition (p.7).

The stratified analysis was initially planned as a mediation analysis to evaluate what proportion of effect of LTBI on HIV incidence was attributable to the pathway of transcriptional activation through RISK6 (or another score). However, LTBI was not associated with incident HIV, but some of the CoRs were. So, we then performed this analysis to explore additional hypotheses as to why some CoRs could be associated with HIV in the absence of LTBI signal:

1) Whether CoRs were associated with HIV, but only among people with LTBI. If the effect size was strengthened or only present in the LTBI subset, compared to the overall group, that could have indicated that LTBI alone was not sufficient to increase risk of HIV, but that Mtb-associated immune activation was. This was one of our primary hypotheses- that being infected with Mtb wasn’t sufficient, and incipient TB was needed to see the association with HIV acquisition. We did not find this.

2) If the CoR-> HIV pathway was only seen in people without LTBI, then maybe there was a different unmeasured exposure that was associated with HIV acquisition among those without LTBI. This would require something else to be different among people without LTBI related to social status, exposures, sexually transmitted infections, which is possible. We therefore looked at whether HSV status, Ad5 titer, or treatment arm, for example were associated with CoRs in univariate analyses (Supplementary Table 4) in case we could identify a risk driver that was not LTBI. We found some hints of this, and therefore one conclusion is that while LTBI is not a unique risk, one might consider making a purpose-built CoR for HIV risk, and then further exploring associations with other non-Mtb exposures that drive them.

We have tried to add more of this into the Discussion, as we de-emphasized the interpretation of those CoR findings, as suggested by Reviewer 2.

4. Were confidence intervals also adjusted using FDR methods as described by Benjamini et al. (reference 40)?

No. We present the original confidence intervals. See longer discussion above, but we now present unadjusted p values also for the 3 flow results.

5. On p.10 the mean and standard deviation are given for FS and PFS, although Fig 1 boxplots show very skew distributions. Therefore, median and interquartile range would be more informative.

Thank you for this suggestion. We have made this change.

6. The legend to Fig. 1 seems to state the wrong labels 1C and 1D (heatmap is 1C, there is no 1D).

This has been corrected.

7. In Fig. 2, the log transformation does not seem to have resulted in a normal (symmetrical) distribution for RESPONSE5 (many extreme negative log values).

We had previously chosen the log transformation so as to bring the range of scores closer to other CoRs and to create a better spread in the data as the range is very narrow. However, we also appreciate this point, and now present RESPONSE5 as neat rather than log transformed throughout.

Reviewer #2: The authors leverage a unique cohort sample bank to investigate association of tuberculosis (TB) infection and incipient disease with HIV acquisition. Strengths of the manuscript and project are the longitudinal data and access to samples prior to HIV infection. The stronger finding is that TB infection, as defined by the team, was not associated with higher HIV infection risk. The team though makes leaps between signature patterns and interpreting their meaning that seems beyond the associations they can more accurately describe. Other weakness includes lack of TB infection testing in host (which the team acknowledges) and long intervals between pre-HIV blood sample and HIV infection. I believe the manuscript needs to address the following issues.

We appreciate this assessment. We also note that while the parent trial did not include either PPD or clinically used Interferon Gamma Release Assays (IGRAs), the flow cytometry assessment in this study using ESAT-6 and CFP-10 uses these same antigens found in authorized IGRA tests and has been shown to perform similarly to IGRAs in the hands of the authors and as published by several others (Refs 31-36). There are no tests for current TB infection other than sputum analysis for M. tuberculosis DNA or growth, which only identifies active TB; the IGRA tests or this analogous flow cytometry version are the most specific available tests for Mtb exposure.

• In the abstract, describe the cohort at high risk for HIV acquisition (not assume knowledge of what the Step MRKAd5 HIV-1 study included).

We clarified that the Step study was a previously completed multinational HIV vaccine study.

Last sentence in the conclusion that host gene expression is associated with HIV acquisition feels the wrong focus, instead that pattern is associated with higher likelihood of infection is what I think is captured.

Thank you, we’ve tried to make the wording clearer throughout the discussion, including moving some of the sentences from the conclusion to the part of the discussion that expounds on exploratory analyses and next steps. We have followed this reviewers’ suggestion and now end with a more succinct summary of the study, as follows:

In this study, we found no evidence that LTBI or polyfunctional Mtb-antigen-specific T cell responses were associated with risk of HIV acquisition in a population at high risk for HIV. However, the exploratory analyses provided a proof of concept that transcriptomic signatures could provide additional information about immunologic profiles of HIV susceptibility beyond known behavioral and viral co-infection risk factors identified in the Step study and other prior studies.

• Throughout the manuscript, concerned saying that a signature predicts HIV acquisition, given there are many mediating steps including need for exposure which will vary by participant. Instead would stick with associated with, with I think main hypothesis which the signature indicates a potential increased vulnerability to successful infection.

We have made edits throughout to use language that does not imply causality. In some places, we left this nomenclature where the word “predict” was appropriate, as in discussing a future event in longitudinal analysis.

• Multiple transcript signatures were tested, with 1 associated with a significant increased odds and another with a lower odds. Would benefit from a clearer indication of what was considered a cutoff with each score that is meaningful and what number of participants were in the “incipient” TB group by each score. Were there enough in that category to have captured their particular risk?

We used these TB CoRs as continuous variables in all cases, whether log-transformed or neat. Because these CoRs are not tuned as HIV risk scores, we used them without assumptions of cutoffs, which also gave us the full power of continuous rather than binary predictors. For example, even if there is a published cutoff for the TB status of interest (which is incipient TB in only 3 scores, LTBI vs active TB in 1 score (Sweeney3), and prediction of TB cure at treatment start in the last score (RESPONSE5)), we did not make assumptions as to what level of immune activation would be associated with HIV acquisition.

We present OR throughout for 1 SD change of each score to use these scores in an unbiased way.

The fact that the 5 scores predict different TB statuses is likely why some scores are positively associated and other negatively associated with HIV. We have tried to clarify this more with what the scores represent in the discussion. For example, RESPONSE5 is likely negatively associated with HIV risk because this score predicts successful TB cure at the start of treatment, rather than associated with incipient TB.

We also have tried to more clearly identify that RISK6 was the primary transcriptomic signature of interest, with the rest as exploratory exposure variables.

• The discussion appropriately focuses on lack of association with TB infection and HIV infection in this cohort. But the leap that transcriptional signatures might be revealing HIV susceptibility is not well supported, as that is not what these signatures were developed for. Can consider future development but seems inappropriate to interpret from this data.

We have tried to now state this more carefully. What we concluded was not that we should use any of these signatures to predict HIV acquisition, but rather that the exploratory work we did gives credence to the idea that an HIV purpose-built CoR could be designed. We would encourage colleagues to use unbiased approaches (eg high-throughput sequencing) to identify transcriptomic signatures that could be similarly used for HIV.

To address this point, we added the following text to the discussion to call for purpose-built signatures for HIV risk developed via unbiased approaches: “While we found variable associations with some purpose-designed TB CoR signatures, global transcriptomic analysis could assist in discovery of similar, simpler HIV CoR signatures.”

And have revised the discussion as follows: “However, the exploratory analyses provided a proof of concept that transcriptomic signatures could provide additional information about immunologic profiles of HIV susceptibility”

• Would keep focus on paper as this is exploratory, not making major decisions such as role of TB preventive therapy.

We agree and are not advocating for TB preventative therapy in our conclusion. We discuss the idea that TB prevention could have been considered to avert HIV infections in populations with HIV risk in TB-endemic areas worldwide, as this was an actionable public health intervention that motivated us to pursue this project. Because we did not find an association with LTBI or flow or transcriptomic markers of TB-associated immune activation, we cannot justify pursuing a clinical trial with this aim, which would have been the next step had our analyses shown a strong association. Because our primary hypotheses were negative, we did perform exploratory analyses evaluating whether LTBI or CoRs were associated with other known HIV risk factors in the parent trial.

Reviewer #3: Bender Ignacio et al takes advantage of the Step MRKAd5 HIV-1 vaccine study to use samples from a cohort in which Mtb status and approximate time of HIV infection are known to determine if several risk factors associated with latent TB infection or immune activation correlated with HIV infection. In general, no differences were observed and latent TB infection did not appear to be a predictive risk for HIV-1. Even though the data are “negative,” this study, despite the potential caveats pointed out by the authors, is one of the few studies using human samples providing a useful reference for others examining infections in other cohorts. Some minor criticisms include:

1. Page 10, there is a reference to data not shown. These data should be included if not in the text, then in the supplemental data.

Thank you for this suggestion. We added this data as a new Supplementary Table 3 and renumbered the subsequent Supplemental tables.

2. This is not an easy paper to read; there is an "alphabet soup" of clinical measurements that at times detract from the overall conclusions. Its clinical focus on the outcomes and statistical analysis rather than what the assays are actually measuring limits the appeal to what may be a more general audience. Defining and discussing the immune and cellular signatures in the context of the results, would put into perspective of why these are relevant correlates for TB and HIV and provide a foundation for those with broader interests in immunology, TB, HIV and coinfections. This is also an opportunity to expand the discussion and put into the context with the body of work that has suggested cytokine and functional immune changes.

We have tried to improve the language used in the discussion to make it more easily interpretable.

For example, we added more detail to how the 5 CoRs are used for TB prediction, so that they are not simply acronyms. We also completely rewrote a large section of the discussion and conclusion (see response to Reviewer #2) in an attempt to make it easier to follow.

Thank you for the opportunity to respond to these comments.

Submitted filename: Response_Review_05Apr22.docx

Click here for additional data file.

14 Apr 2022

Mycobacterium tuberculosis infection, immune activation, and risk of HIV acquisition

PONE-D-21-33120R1

Dear Dr. Bender Ignacio,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Manish Sagar, MD

Academic Editor

PLOS ONE

21 Apr 2022

PONE-D-21-33120R1

Mycobacterium tuberculosis infection, immune activation, and risk of HIV acquisition

Dear Dr. Bender Ignacio:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Manish Sagar

Academic Editor

PLOS ONE

Table 1

Participant characteristics.

Characteristic	Control (n = 310)	Case (n = 155)
Age (mean, SD)	31.9 (7.8)	30.6 (7.8)
Gender
Male	264 (85.2%)	142 (91.6%)
Female	46 (14.8%)	13 (8.4%)
Country
Australia	4 (1.3%)	2 (1.3%)
Brazil	12 (3.9%)	6 (3.9%)
Canada	14 (4.5%)	7 (4.5%)
Dominican Republic	2 (0.6%)	1 (0.6%)
Haiti	4 (1.3%)	2 (1.3%)
Peru	68 (21.9%)	34 (21.9%)
USA	206 (66.5%)	103 (66.5%)
Race and Ethnicity *
White	152 (49.0%)	79 (51.0%)
Black	43 (13.9%)	15 (9.7%)
Multiracial or Other Race	12 (3.9%)	11 (7.1%)
Mestizo/a	68 (21.9%)	34 (21.9%)
Hispanic	35 (11.3%)	16 (10.3%)
Vaccine trial treatment arm
Intervention	177 (57.1%)	86 (55.5%)
Comparison	133 (42.9%)	69 (44.5%)
AD5 Titer stratum
>18	176 (56.8%)	84 (54.2%)
< = 18	134 (43.2%)	71 (45.8%)
Behavioral Risk score (mean, SD) **	2.8 (1.2)	3.3 (1.2)
Circumcised (male participants)	157 (59.5%)	84 (59.2%)
HSV-2 seropositive (male participants)	81 (30.7%)	57 (40.1%)

*Race and ethnicity were asked in a singular question about self-identity in the parent study without option of multiselect; options including race or ethnicity therefore sum to 100%.

**Combined risk score for male and female includes theoretical and observed range 0–7 points.

Table 2

LTBI status, Mtb-specific CD4 T-cell activation, and HIV acquisition.

CoR Score	Univariate regression estimates		Multivariate regression estimates
	ORa (95% CI)	p-value	aORa,b (95% CI)	p-value
LTBI Positivec	0.84 (0.51, 1.39)	0.51	0.85 (0.51, 1.43)	0.54
PFSd	0.96 (0.78, 1.18)	0.68	0.96 (0.77, 1.20)	0.73
FSd	0.94 (0.76, 1.16)	0.57	0.95 (0.76, 1.18)	0.62

Unadjusted and adjusted odds ratios for LTBI status, COMPASS polyfunctional scores (PFS) and functional scores (FS) in those who acquired HIV versus controls.

aAll models used conditional logistic regression accounting for matching variables (treatment arm and country).

bModels were fully fit with all variables that were associated with case control status: gender, risk score, and age.

cPrespecified as the primary exposure of interested.

dORs are reported for 1 standard deviation change.

Table 3

Associations between each Correlates of Risk (COR) signature and HIV acquisition, univariate and adjusted analyses.

CoR Scorea	Univariate regression estimates	Multivariate regression estimates
	ORc (95% CI)	aORb,c (95% CI)
RISK6d,e	1.06 (0.87, 1.31)	1.15 (0.92, 1.44)
Suliman4d	0.89 (0.72, 1.09)	0.89 (0.72, 1.11)
Sweeney3	1.34 (1.05, 1.70)	1.38 (1.07, 1.78)
Maertzdorf4	0.90 (0.73, 1.12)	0.89 (0.71, 1.13)
RESPONSE5	0.84 (0.67, 1.05)	0.78 (0.61, 0.98)

aAll CoR scores in were modeled using conditional logistic regression accounting for matching variables (treatment arm, country).

bOR and adjusted OR (aOR) are reported for 1 standard deviation change.

cModels were fully fit with all variables that were associated with case control status: gender, risk score, and age.

dVariables log transformed.

ePrespecified as the primary transcriptomic signature of interest. The other four signatures were considered exploratory predictors.