https://orcid.org/0000-0002-3060-7749
Figures
Abstract
Background
Tuberculosis (TB) remains a major public health problem globally, even compared to COVID-19. Genome-wide studies have failed to discover genes that explain a large proportion of genetic risk for adult pulmonary TB, and even fewer have examined genetic factors underlying TB severity, an intermediate trait impacting disease experience, quality of life, and risk of mortality. No prior severity analyses used a genome-wide approach.
Methods and findings
As part of our ongoing household contact study in Kampala, Uganda, we conducted a genome-wide association study (GWAS) of TB severity measured by TBScore, in two independent cohorts of culture-confirmed adult TB cases (n = 149 and n = 179). We identified 3 SNPs (P<1.0 x 10–7) including one on chromosome 5, rs1848553, that was GWAS significant (meta-analysis p = 2.97x10-8). All three SNPs are in introns of RGS7BP and have effect sizes corresponding to clinically meaningful reductions in disease severity. RGS7BP is highly expressed in blood vessels and plays a role in infectious disease pathogenesis. Other genes with suggestive associations defined gene sets involved in platelet homeostasis and transport of organic anions. To explore functional implications of the TB severity-associated variants, we conducted eQTL analyses using expression data from Mtb-stimulated monocyte-derived macrophages. A single variant (rs2976562) associated with monocyte SLA expression (p = 0.03) and subsequent analyses indicated that SLA downregulation following MTB stimulation associated with increased TB severity. Src Like Adaptor (SLAP-1), encoded by SLA, is highly expressed in immune cells and negatively regulates T cell receptor signaling, providing a potential mechanistic link to TB severity.
Conclusions
These analyses reveal new insights into the genetics of TB severity with regulation of platelet homeostasis and vascular biology being central to consequences for active TB patients. This analysis also reveals genes that regulate inflammation can lead to differences in severity. Our findings provide an important step in improving TB patient outcomes.
Author summary
Severity of tuberculosis (TB) is an important determinant of transmission, morbidity, mortality, disease experience and quality of life; it can be affected by human genetic variation. In this study, we performed a genome wide study (GWAS) to examine the role of human genetic variation on TB severity, as measured by the Bandim TBscore, a metric developed for resource-limited environments. Three single nucleotide polymorphisms (SNPs) in one gene, RGS7BP, associated with severity. This gene is expressed in blood vessels and plays a role in infectious disease pathology. In addition, we used the GWAS data to identify pathways linked to TBscore, and found that the platelet homeostasis and transport of organic anion pathways both associated with severity. Functional analyses were also performed, examining the effects of stimulation with Mycobacterium tuberculosis, the causative agent of TB. Down regulation of the gene encoding Src Like Adaptor (SLAP-1) that is highly expressed in immune cells and negatively regulates T cell receptor signaling associated with MTB exposure, providing a potential mechanistic link to TB severity. Together these findings indicate that host factors involved in regulation of platelet homeostasis and vascular biology affect TB severity. These findings may lead to new ways to decrease disease severity.
Citation: McHenry ML, Simmons J, Hong H, Malone LL, Mayanja-Kizza H, Bush WS, et al. (2023) Tuberculosis severity associates with variants and eQTLs related to vascular biology and infection-induced inflammation. PLoS Genet 19(3): e1010387. https://doi.org/10.1371/journal.pgen.1010387
Editor: Gregory S. Barsh, HudsonAlpha Institute for Biotechnology, UNITED STATES
Received: August 21, 2022; Accepted: February 13, 2023; Published: March 27, 2023
Copyright: © 2023 McHenry et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All summary data is within the manuscript or supporting material. Because of the IRB restriction on the data from Uganda, individual level data are only available upon request from the Uganda Genetics of TB Data Access Committee by contacting the committee: chair Dr. Sudha Iyengar (ski@case.edu). The other members of the committee are: Alberto Santana of CWRU, and Dr. Alphonse Okware and Ms Dorcas Lumunu from Uganda. None of these people are involved with the research presented in the paper.
Funding: This work was supported by the following grants from the National institutes of Health: N01-AI70022 to WHB, R01AI124348 to WHB, HL007567 to MLM, N01-AI95383 to WHB, T32 GM007250, U19AI16258 to TRH, T32 HL007567, R01HL096811 to CMS, TL1 TR002549 and OPP1151836 from the Bill and Melinda Gates Foundation to TRH. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Pulmonary tuberculosis (TB) caused more deaths per year than any other pathogen prior to the COVID-19 pandemic [1]. It is the leading cause of death among people infected with human immunodeficiency virus (HIV) [2]. Although incidence is decreasing globally, TB is re-emerging in Sub-Saharan Africa and Southeast Asia [3]. The bacterium, Mycobacterium tuberculosis (MTB), causes most TB and is transmitted via airborne droplets from coughing and sneezing by people with active disease. Therefore, it can be a very mobile pathogen in the age of frequent global travel, making global exposure high; between one fourth and one third of the entire global population is latently (asymptomatically) infected. However, far fewer people develop active disease than are infected. In 2020, only 10 million people had active disease with 1.5 million people dying of it [1].
As compared to studies of disease susceptibility or resistance, few studies have focused on TB severity, an important determinant of transmission, morbidity, mortality, as well as disease experience and quality of life [4,5]. Determinants of TB severity remain uncertain due to heterogeneous definitions. Measures of severity include the Bandim TBscore (a validated outcome based on 11 clinically relevant symptoms, S1 Table), bacillary load, and radiologic findings (e.g. enumeration of pulmonary lesions or area, presence of cavitation) [6–17]. In this study, we chose to use TBscore as it is: 1) based on simple measures of relevant and meaningful clinical parameters that can be ascertained in the resource-limited environments where TB is most prevalent; 2) has previously been validated through comparisons to other measures of TB disease progression and severity [18–20]; and 3) at presentation, it is predictive of mortality in TB patients receiving treatment, associated with quality of life, and even a one point increase is clinically meaningful in some contexts [18–22]. As such, the TBscore presents several advantages in terms of ease of measurement and prediction of clinical outcomes. Earlier genetic studies of severity phenotypes focused on candidate genes and not genome-wide analyses, making it likely additional genetic variants elsewhere in the genome associate with TB severity. Notably, some of the severity associating genes (i.e., IFNG, SLC11A1, MCP1, TLR variants, and HLA variants) are similar or identical to those implicated in previous studies of susceptibility or resistance to TB [23]. However, the overlap in genes for risk and severity may simply be due to the fact that candidate gene studies tended to choose the same genes for the two phenotypes. In contrast, genome-wide analyses may better define how the genetics of TB risk or progression do or do not correlate across severity definitions and may inform distinct functional aspects of TB host genetics.
Understanding the link between DNA variants and RNA expression can inform an understanding of immunological responses and TB pathogenesis. Directly linking genetic variation to immunological function is critical in identifying valid targets for new therapeutics and vaccines [24–28]. This understanding can be partially achieved using bioinformatic databases that annotate function of specific genotypes by connecting tissue specific gene expression or aggregating SNPs from analyses into functional pathways (e.g., gene set enrichment analysis, GSEA) and can also be achieved by expression quantitative trait locus (eQTL) studies [27,29–32]. We identified nine studies that have tied gene expression to tuberculosis phenotypes, but only four of them assessed the role of DNA variants in the regulation of gene expression, and none examined the regulation of gene expression in the context of clinical TB severity (S2 Table). While bioinformatic databases can provide valuable information, they are limited because the cells and tissues available are not always the most relevant to a given phenotype (e.g. TB severity) and they are not under conditions that recapitulate disease related exposures (e.g. active TB) [32]. Further, some eQTL’s are only associated with gene expression in specific contexts, such as an active MTB infection. In the case of MTB infection, resident alveolar macrophages and recruited monocyte-derived macrophages are early targets of MTB in the lung, making these cells informative with respect to the regulation of gene expression in the context of active infection [33]. This situation has, however, yet to be explored [34,35]. Thus, datasets containing MTB stimulated macrophages can provide functionally relevant information on how variants associated with severity affect RNA expression during MTB infection in vitro.
In the present study, we first conducted a case-only genome-wide association study (GWAS) of TBscore to identify variants associated with TB severity. We then followed-up these associated loci in a different set of subjects without active TB disease, utilizing data from monocyte-derived macrophages before and after in vitro stimulation with MTB to observe how severity associated variants from our GWAS analysis associated with changes in gene expression in the context of active infection. This study addresses the aforementioned gaps in the current literature on active TB severity by 1) studying the genomic underpinnings of TB severity using a meaningful, replicable, and validated clinical phenotype; and 2) bridging the gap between genetic variants and immunological function by studying gene expression in the macrophage response, as well as that of other immune cells. Our underlying hypothesis is that genomic variation in humans affects the immunological response to active TB disease, as measured by gene expression, and that this correlates with clinical severity.
Results
Study population
To discover the major genetic variants associated with TB severity, we examined 328 subjects with pulmonary TB (Table 1, 149 subjects in Cohort 1 and 179 in Cohort 2). There were statistically significant differences between the cohorts with respect to HIV status and TBscore, with Cohort 2 having more HIV+ subjects and a lower average TBscore. Both cohorts included more males than females, most subjects were HIV-, and the average age was just under 29 in both cohorts.
Genome-wide association results
Individually, the two cohorts showed no sign of genome-wide inflation (S1 and S2 Figs) in their Q-Q plots or genomic control statistics (λ<1.0). PCA analysis revealed that none of the top 10 PC’s were associated with TBscore in either cohort. Based on this and the low amount of variation explained by the PC’s, we decided not to include them in the initial regression equation (S3 and S4 Figs). The Q-Q plot for the full range of meta-analytic P-values appears to deviate from the line but the genomic control parameter (λ) was 0.98, indicating little to no genome-wide inflation (Fig 1). Neither cohort showed any SNPs that were GWAS significant when considered individually (S5 and S6 Figs). There were a total of 11,323 SNPs showing an association with P<0.05 in both cohorts and beta values with the same direction of effect (i.e. both negative or both positive) (S3 Table). Of these, 10,750 SNPs passed the I2 threshold for heterogeneity (S3 Table). Out of the 10,750, there was one SNP on chromosome 5 (rs1848553) that was GWAS significant with a meta-analytic P-value of 2.97x10-8 and a beta of -0.97 (Table 2 and Fig 2). Two other SNPs in this gene are close to genome wide significant as well (p = 7.78x10-8 for both, Table 2 and Fig 3).
The quantile-quantile (Q-Q) plot shows the inverse log(10) of the observed p-values on the Y-axis relative to what is expected if there was no association on the x-axis. Deviations above the line indicate an association with the outcome. If the line deviates at the low quantiles, then this is considered evidence to suggest genome-wide inflation of the test statistics, which typically indicates unmeasured confounding (λ = 0.98).
The Manhattan plot shows the inverse log(10) of the p-values for the association between each SNP and TBscore on the y-axis and the x-axis represent the physical location of each SNP on the chromosomes, which are in order from 1–22.
The LocusZoom plot shows the region surrounding rs1848553 on chromosome 5, using an LD panel and reference genome from the AFR super-population in the 1000G project. Yellow and orange indicate higher levels of LD.
We conducted sensitivity analysis to evaluate the impact of PCs, and this analysis demonstrated that the association of rs1848553 at a GWAS significant level is not sensitive to the PC’s being in the regression equation in the analyses (S4 Table and S1 Methods). In addition, to assess whether the top to PCs had any impact on association with TBscore, we adjusted each of the 10,750 SNPs that passed heterogeneity threshold and there were no major changes in results (S4 Table).
Lastly, the role of HIV status was assessed stratifying analyses by HIV status. As the number of HIV+ individuals was small (n = 58), we compared HIV- to the entire data set. The effect sizes (betas) were similar in HIV- and all individuals, but as expected p values were less significant due to reduced sample sizes and an extra degree of freedom (S5 Table).
Annotation for GWAS Significant SNP: rs1848553
SNP rs1848553 is a C to T allele change, the MAF in the two cohorts combined was 24.7%, and did not vary much between cohorts (S6 Table). The T allele is generally very rare in 1000G reference populations. However, in African populations, the T allele is much more common (21%) than in other continental populations, where the T allele ranged from 0–2.6% (S7 Table). The beta value indicates each copy of the minor allele is associated with just under a 1-point decrease in the TBscore in the additive model. This represents a clinically meaningful reduction in severity [18], as well as an important reduction in the risk of mortality. Considering each T allele is associated with nearly a 1-point reduction in TBscore, T/T homozygotes have nearly a 2 point reduction in TBscore relative to C/C homozygotes (Fig 4). The beta and P-values did not vary across cohorts and the I2 for this SNP was <1%, indicating little to no heterogeneity. The GWAS significant SNP, rs1848553, is located within an intron of the protein-coding RGS7BP gene [36]. There are 18 other SNPs in the same area (63.6Mb to 64.0Mb) on chromosome 5 that show an association with TBscore with P<1x10-5, including the 2 SNPs that had a P<1x10-7 (Fig 3 and Table 2). RGS7BP codes for a protein that regulates trafficking of G-proteins between the nucleus and the plasma membrane [37].
Distribution of TBscore by genotype.
In previous work we included lineage x SNP interaction terms in analyses of data that overlapped with the present study and demonstrated that lineage itself did not associate with TBscore [38]. Further, both this prior paper as well as an earlier candidate gene study demonstrate that the lack of first-order association between lineage and TBscore (i.e. lack of significance in the absence of interaction) is a hallmark of likely co-evolution [39]. Therefore, we included lineage as a first order covariate in sensitivity analysis for all SNPs with P<1e-05 in the original analysis for association between SNP and TBscore (without adjustment for lineage) (S8 Table). This sensitivity analysis showed that there was a slight increase in P-value due to an increase in degrees of freedom, but the lineage variable itself was not associated with TBscore even when included in the regression model using linear adjustment.
Chest X-Ray analysis
Of the 169 SNPs showing an association with TBscore at P<1 x10-5, 165 of these showed the same direction of effect in the chest X-ray (CXR) analyses: i.e., the beta value for association with TBscore was <0 and the OR for association with CXR extent was <1, or the TBscore association beta was >0 and the CXR OR was >1 (S7, S8 and S9 Figs, and S4 and S9 Tables). While the direction of effect was consistent in 97.6% of the SNPs that showed an association with TBscore at the P<1 x10-5 threshold, 20 of the 165 SNPs showed an association with CXR extent with P<1 x10-5. Thus, the consistency in the direction of effects provides evidence that the TBscore is a similar and valid measure for quantifying clinical severity.
Decile regression analyses
The decile regression for rs1848553 and TBscore was consistent with a negative or zero beta value for the whole range of the TBscore (S10 Fig). This analysis shows that for effect of rs1848553 genotype, there is a stronger effect at the highest end of the TBscore deciles than elsewhere, but the effect never reverses direction. Thus, it appears that genetics has the greatest influence on severity among the subset of active TB patients who have most severe or most mild disease, and it makes the least impact on those with TBscores close to the mean.
Examination of prior published results related to susceptibility and severity
We examined our genome-wide TBscore association results for 12 previously identified TB susceptibility SNPs and 3 SNPs in our data showed association with TBscore gene regions associated with disease (P<0.05) (S10 Table)[40]. IFNG, TLR4, and VDR all showed P<0.05 for the association with TBscore, and SLC11A1 had P = 0.053. Most of the gene regions examined did not show a statistically significant association, indicating that there are likely unique sets of genes that drive severity versus susceptibility. To follow-up our previously identified association with IL12B variants [17], we examined the association between any SNPs +/- 50kb from IL12B with P<0.05 for association with TBscore in Cohort 2 only, because Cohort 1 was included in our previous publication [17]. We identified three SNPs with P<0.05 in this region (S11 Table) in this analysis. These findings add further evidence that there are likely distinct sets of genes associated with susceptibility to and severity of TB disease.
Enrichment analyses
A total of 169 SNP associations had P<1 x10-5 and were used for annotation and enrichment analysis [41–43]. The analyses showed enrichment for two Reactome pathways: platelet homeostasis and organic anion transport (Table 3). The genes driving the significance of these two pathways did not overlap. Nor do any genes overlap between the two pathways (S12 and S13 Tables). The MAGMA analysis of gene-level associations with TBscore did not show any significant single gene results at the genome-wide level; the most significant association (P = 2.1x10-3) was in the PSORS1C2 gene, also known as the psoriasis susceptibility 1 candidate 2 gene that is thought to confer susceptibility to psoriasis. There were 19,220 genes represented in the summary statistics and the resulting p-value threshold for genome-wide significance at the gene-level was P<2.6x10-6. While no individual genes or gene sets were found to be significant, the MAGMA gene level analysis of tissue specificity showed that the genes represented in the data are significantly differentially expressed, and specifically up-regulated, in blood vessels based on tissue specificity data from GTEx v8 (Fig 5).
FUMA GWAS uses MAGMA gene-level analyses and differential gene expression data from GTEx v8 to determine if the genes to which the SNPs are mapped are significantly differentially expressed in any tissues. This analysis showed that the mapped genes were significantly upregulated in blood vessels (indicated by the red color).
Using FUMA to determine if any of the 169 SNPs from our GWAS analysis were eQTL’s, we found that 25 of the SNPs were eQTL’s that associate with expression of 28 genes; some of the eQTL’s showed an association with more than one gene. There were a total of 150 eQTL effects across different tissues and databases, as many of the 25 eQTL’s associated with expression in multiple tissues and/or databases (S11 Fig shows the 28 genes). Interestingly, 5 of the SNPs with P<1x10-5 were cis eQTL’s for RGS7BP in the same region on chromosome 5 as the top GWAS hit. This is consistent with the annotations that showed it is a known regulatory variant. The STRING database analysis of protein-protein interactions (PPIs) did not show significant enrichment for interactions or for any pathways (S11 Fig).
eQTL analysis of severity associated variants
We next examined whether these suggestive TB severity SNPs were eQTLs in MTB-stimulated monocytes (cohort characteristics in S14 and S15 Tables). There were four SNPs that showed significant SNP by stimulation interactions that were significant with FDR corrected P<0.1, and all of these were cis-eQTL’s (Tables 4 and S15). Upon examining the marginal effects (no interaction term in model) for these four significant SNPs, only rs2976562 at the Src-like adaptor (SLA) gene was significantly associated with expression in the MTB-stimulated stratum (Table 4). This eQTL did not have a significant effect in media-only. Thus, rs2976562 meets the definition of a stimulation dependent eQTL, as it is active only within the context of MTB infection. In the analyses that were stratified by RSTR or LTBI status, this SNP showed a similar Beta (-0.71), but was not statistically significant (FDR = 0.15), likely due to the decreased sample size. This indicates that the relationship between rs2976562 and SLA is similar between RSTR and LTBI, and thus, robust to differences in patients’ clinical characteristics, despite only being significant when the data are combined. Notably, rs2976562 showed interaction with MTB stimulation that associated with both SLA and NDRG1 expression, but the association with NDRG1 was not significant after FDR correction (Table 4). However, rs2976562 was also found to be an eQTL for NDRG1 in the eQTLGen cis eQTL’s database during our interrogation of publicly available eQTL databases described above (S11 Fig and S16 Table).
MTB stimulation increases the expression of SLA for individuals with CC and CT genotypes (Fig 6). The T allele for rs2976562 is associated with lower expression of SLA after stimulation with MTB but is associated with a relative increase in SLA expression prior to MTB stimulation (p = 0.03). Hence, the effect of rs2976562 is dependent on MTB infection status of the cells. Comparing this to the boxplot of the relationship between rs2976562 and severity shows that this same allele is associated with increased severity, especially among homozygotes, who appear to have very severe disease on average (Fig 7). Thus, there appears to be a relationship between downregulation of SLA upon MTB stimulation and more severe TB disease among T/T homozygotes.
rs2976562 Genotype Distribution: 96 subjects CC, 44 CT, and 4 TT. P value in media = 0.41, β in media = 0.6, P-value in MTB stimulated = 0.03, β in MTB stimulated = -0.32.
Distribution of TBscore by genotype.
The data show that a statistically significant relationship between SLA expression and rs2976562 genotype only becomes apparent in the context of MTB stimulation, though the T/T homozygotes do appear to have higher expression of SLA prior to stimulation but the relationship was not statistically significant. An examination of population genetic data shows that the T allele is found at lowest frequency (11%) in African populations, but it is actually the major allele in Asian and American populations (60–70%) and common in European populations [44] (S17 Table).
Discussion
Overall, our results show that human genetic variation is associated with TB severity. We found a single SNP that was GWAS significant, but enrichment analyses and tissue specificity both revealed variation in several processes related to vascular biology that have been previously implicated in the inflammatory response to infectious disease. Our eQTL analysis results indicated that human genetic variation is an important aspect of antigen presentation may affect the host response to infection with MTB. A better understanding of these mechanisms, and how they relate to severity and mortality in active TB patients, may lead to greater therapeutic insight that can reduce the negative impact of active TB disease on human health.
While the MAGMA analysis did not show any statistically significant single-gene effects, the tissue expression analysis of our MAGMA results showed that, collectively, the genes represented in our severity-associated results are significantly up-regulated in blood vessels in response to MTB infection. The vascular endothelium plays an important role in thrombosis and inflammation, and the vasculature is responsible for enabling the extravasation of immune effector cells in the response to infection[45,46]. Acute changes in blood pressure during active infection can lead to organ failure and death in COVID-19 (hypertension) and acute sepsis (hypotension) [47,48]. Further, vasculitis and stroke (particularly in the context of TB meningitis) have been posited as complications of TB disease as a result of inflammation and dysregulation of vascular function [49–51]. If differences in inflammation, coagulation, and regulation of vascular function lead to more severe outcomes and/or death, then an understanding of this phenomenon, and how to address it, could drive improvement in outcomes and reduction in mortality for active TB patients.
Previous studies of TB susceptibility identified genes that are primarily involved in the host immune response. However, the genetics of severity appears to incorporate a different biological process, namely platelet homeostasis. Specifically, platelet homeostasis was found to be enriched in the GSEA analyses using FUMA, and platelets are an important part of the response to both inflammation and infection [52,53]. Platelets are involved in crosstalk between immune effector cells and aid in the body’s ability to sense pathogens and enable infection-induced inflammation [52–55]. This inflammation often leads to a state that boosts coagulation in humans and a previous study showed that TB patients are in a pro-coagulatory state [56,57]. The damage that dysregulation of platelet homeostasis and coagulation can cause has also been demonstrated in acute sepsis and septic shock [58]. Notably, septic shock has been`previously reported as a common cause of death in pulmonary TB patients [59].That most genes previously associated with susceptibility did not significantly associate with TB severity emphasizes that although the two phenotypes have some overlap they are likely genetically and biologically distinct.
Our eQTL results help clarify the effect that some of the severity associated DNA variants exert on gene expression. While the GWAS significant SNP was not shown to be an eQTL, multiple SNPs within the same region as the GWAS significant SNP (rs1848553) appear to be eQTL’s for RGS7BP, implying that there may be a functional role for this gene in active TB. The stimulation-dependent eQTL, rs2976562, was an important regulator of SLA in the context of in vitro stimulation in monocytes and resides in a flanking promoter region [60]. While this region does not contain the promoter for the SLA gene, it has been associated with expression of other genes and may be an enhancer for SLA.
The SLA (Src-like adaptor) gene codes for the SLAP-1 protein. The DICE database indicates that SLA is expressed in a number of immune cells, including monocytes, but that its expression is highest in T-cells (S12 Fig) [61]. SLAP-1 is an adapter protein that negatively regulates T-cell receptor signaling, inhibits T-cell antigen-receptor induced activation of nuclear factor of activated T-cells, and is involved in the negative regulation of positive selection and mitosis of T-cells [37,62,63]. SLAP has a role in activation and maturation of monocyte and dendritic cells through downregulation of granulocyte macrophage colony-stimulating factor receptor (GM-CSFR) [35]. SLAP deficient bone marrow-derived dendritic cells produce less TNF-α and IL12 in response to LPS, fail to stimulate T-cells in mixed lymphocyte reactions, and are less effective at inducing IFN-γ secretion from T cells [64]. Thus, a deficiency of SLAP-1 likely impairs a robust immune response, and reduces ability to generate cytokines (IL12, IFN-γ, and TNF-α) that are important drivers of the host immune response to active TB infection [65–67]. Further, variants that code for these proteins have previously been implicated in TB severity and/or susceptibility. Our results in conjunction with prior literature on the role of SLA indicate that decreased SLA expression may be associated with more severe disease and this may be explained by dysregulation in the maturation of monocytes and dendritic cells.
This study is not without limitations: 1) our sample size is small by the standards of many modern GWAS studies; and 2) there were differences between the two cohorts, namely in TBscore and in the proportion of HIV+ individuals. The analyses controlled for the latter factor by adjusting for HIV status and then assessing HIV- individuals only. That said, the genomic control parameter was below 1, so it does not appear that there was significant genome-wide inflation. Furthermore, we used a stringent I2 threshold to exclude SNPs showing heterogeneity of effects. Thus, it is unlikely that differences between the two cohorts substantially affected the associations. For our eQTL analyses, a more directly related phenotype would have been cells harvested from patients with active TB, as this was the patient population from our GWAS analysis. However, the stratified analysis indicated that the association we observed was similar in latent tuberculosis infected (LTBI) patients, and thus unlikely to be sensitive to the inclusion of RSTR subjects in the data. The small sample size also affected our eQTL analysis; the strongest association between rs2976562 genotype and SLA expression was among TT homozygotes, of which there were only 4.
In conclusion, this study demonstrates that TB severity associates with genetic variation and indicates that variation in both the regulation of platelet homeostasis and vascular function may be driving outcomes for active TB patients. Our eQTL results indicated that regulation of SLA may be an important driver of variation in active TB severity due to an impact on a fundamental aspect of the host immune response to TB disease. Further study of the impact of SLA may yield insight into a more effective host immune response that is associated with less severity and mortality. Future studies should consider the role that infection-induced inflammation plays in active TB severity, in the hope that mortality and other severe outcomes might be mitigated or avoided through a better understanding of these processes.
Methods
Ethics statement
The study protocol was approved by the National HIV/AIDS Research Committee of Uganda and the institutional review board at University Hospitals Cleveland Medical Center. Final clearance was given by the Uganda National Council for Science and Technology. All participants provided written informed consent.
Study participants
TB cases and extensive clinical data were ascertained as part of the Kawempe Community Health Study (KC Study) [68], a longitudinal household contact study conducted in Kampala, Uganda from 2002–2017 through the Uganda-CWRU Research Collaboration. The KC Study enrolled 3,818 total participants, of which there were 872 adult active TB index cases. In the present case-only analysis of severity, a subset, for whom available genotype data and the information necessary for assigning the TBscore exists, were identified from these 872 index cases (S1 Table). The TBscore, developed for adults, may not appropriate for individuals under 15. Thus, our sample was limited to TB cases 15 years old and older. We examined two subsets, as described in our past publications[38,39], that will be referred to as Cohort 1 and Cohort 2 (N = 149 and N = 179, respectively) according to different genotyping platforms available at different times during the study (described below). The Bandim TBscore was constructed as described previously [19,69]; further details are provided in the S1 Methods.
All TB cases were culture-confirmed based on isolation of MTB from sputum and clinical characteristics were assessed during the visit at which subjects were diagnosed with active TB. X-rays were performed at the Uganda Cancer Institute. Additional details about the original study protocol are described elsewhere [68]. The two subsets differed in percentage of HIV positive individuals (Table 1); therefore, HIV status was used as a covariate in all regression models. Previous analyses of microsatellite data from these cohorts indicated no substantial population substructure, as confirmed by previous principal components (PC) analyses [17,70].
Monocyte transcriptional profiles
A follow-up cohort including 72 Ugandan subjects without active TB infection was included in a transcriptome-wide study to assess the association between gene expression and variants we identified in our GWAS. These subjects were ascertained from the KC Study but included only subjects without active TB because this cohort had both genotype and RNA expression data available. The follow-up cohort, without active TB infection, was part of an analysis that examined whole blood and monocytes isolated from highly MTB-exposed, HIV-negative donors in Uganda [71,72]. Some of these subjects (n = 34) showed resistance (classified using TST and IGRA) to infection by MTB after repeated exposure, and are referred to as resisters (RSTR), as previously described [73,74]. In addition to the RSTR subjects, there were donors with latent tuberculosis (LTBI) defined by concordant positive TST/IGRA. For this study, monocytes were isolated from PBMCs and then stimulated with MTB for 6 hours. RNA expression levels were assessed in these samples using RNA-seq with and without stimulation by MTB [75].
Genotyping and QC
Cohort 1 was genotyped on the Illumina Infinium MegaEX chip, comprising 2.1M markers genome-wide. For Cohort 2, we used the Illumina HumanOmni5 microarray comprising 4.3M markers genome-wide, offering high genome wide coverage of common genetic variation within African populations [76]. Prior to imputation, only SNPs that had a call rate greater than 0.98, minor allele frequency (MAF) > 0.05, and did not show deviation from Hardy-Weinberg equilibrium (p<10−6) before and after imputation in both samples were used in the analysis. The total number of SNPs that overlapped between the two cohorts after imputation quality control (see S1 Methods) was 6,421,278 (S13 Fig). Principal components were computed using Plink v1.9 (S3 and S4 Figs).
Our follow-up cohort without active TB used for the analysis of monocyte derived macrophages was genotyped separately from TB cases in Cohorts 1 and 2. The genotyping for this cohort was done on the Illumina MegaEx Chip and had 1,042,921 SNPs prior to imputation. To maximize the overlap between the list of SNPs associated with TBscore and those included in the genotype data for the 72 subjects with RNA-seq data, we imputed SNP calls on the follow-up cohort with the Michigan Imputation Server (Minimac4) with a Minimac R2 >0.3 on the whole MegaEx chip [77]. The 1000G Phase 3 v5 AFR panel was used as the reference population and Eagle2 was used for phasing [78]. After imputation, we restricted the analysis to SNPs with MAF > 0.05, genotyping rate > 98%, and those that did not violate Hardy-Weinberg equilibrium (at p<10−6). Gene expression was measured using RNA-seq in monocyte-derived macrophages, as previously described (S1 Methods) [72,75].
Genome-wide association analysis
To assess the association between genetic variants and TBscore, we utilized a linear regression model with sex and HIV status as covariates in Plink v1.9 software. We then combined the summary statistics from the two cohorts to generate meta-analysis derived p-values. To determine meta-analysis p-values and beta coefficients across the two cohorts, we utilized random effects meta-analysis with inverse variance weighting. Based on the Cochrane handbook recommendations, all variants with an I2 > 40% were excluded from the analysis to reduce heterogeneity between the cohorts [79]. To be considered GWAS significant, the association between a variant and TBscore had to have a p<0.05 in both cohorts, the sign of the beta value had to be the same in both cohorts, and the meta-analysis needed to meet the canonical GWAS threshold (P <10−8). A power calculation showed that, with our sample size and mean TBscore, we had 48% power to detect a difference of 1 point on the TBscore at p = 5x10-8 and at a minor allele frequency of 0.25, using an additive model. To be included in further enrichment and annotation analyses, the meta-analytic P-value had to be below 1x10-5. We chose this latter threshold because previous studies have shown that some variants that do not meet the GWAS threshold may still have important regulatory or biological functions, and may be worthy of further study and follow-up, especially in the context of gene regulation [41–43]. We used FUMA GWAS to annotate and enrich our SNPs below this threshold. Analyses performed with FUMA included gene mapping, regulatory annotation, tissue specificity, MAGMA analysis (gene-based analysis), gene set enrichment, and pathway analyses (S1 Methods) [80]. In addition to FUMA, we utilized GeneCards, Ensembl, DICE, and STRING DB to annotate and enrich our results with respect to function, expression, and downstream protein interactions [36,37,61,81]. MAGMA is a computational biology tool that utilizes the significance and direction of effect of multiple SNPs aggregated together in the same genes [82]. It is similar to a multiple linear principal components regression model, using an F-test to compute the gene-level p-value [82].
To look more closely at SNPs that showed GWAS significance, we performed decile regression on all significant SNPs to determine if and how their relationship with TBscore varied across the distribution. Decile regression, performed using QuantReg Software in R 3.6.3, shows how the beta values for the SNPs in the linear regression differ by deciles of the TBscore. The decile regression utilized the same covariates in the regression equation as the GWAS analysis (HIV status and sex).
To validate our GWAS summary statistics showing association with TBscore, we also examined the association between the genetic variants of interest and radiological severity, as determined by extent of disease using chest X-ray (CXR) data. Extent of lung involvement was measured using the US National Tuberculosis and Respiratory Disease Association (US NTRDA) grading system from radiographs taken at the Uganda Cancer Institute and performed on the same patients as the analysis of TBscore (i.e. they were identical to Cohorts 1 and 2) [83]. The US NTRDA grading system includes categories 0–3 based on extent of disease upon radiological examination (S7 Fig). The categories ranged from 0 as the least extensive disease (i.e. least severe) to 3 with the most extensive. For these analyses, we combined categories 0 and 1 and categories 2 and 3 from the US NTRDA grading system. This was operationalized as a binary variable and the analysis was done using a logistic regression model that adjusted for sex and HIV status. We tested all SNPs with P<1x10-5 for the association with TBscore for association with CXR severity.
eQTL analysis
For our eQTL analysis in the follow-up cohort of 72 subjects without active TB, we first generated a list of genes that were differentially expressed in response to MTB infection, so we could narrow the list to genes that are regulated in response to active infection. We first analyzed eQTL’s using the cross linear model that tests for interactions between genotype (coded additively) and MTB stimulation status and its association with expression in a linear regression model. For our cross-linear model, if Y = expression, X1 = SNP(additive coding), X2 = Age, X3 = Sex, and X4 = MTB stimulation status, then the regression equation for this analysis was: Y = β0 +β1X1 + β2X2 + β3X3 + β4 (X4) + β5 (X1X4) + ε. We followed up eQTL’s that showed an FDR < 0.1 for the interaction term by testing them for association with gene expression in an analysis that stratified by MTB stimulation. All models were adjusted for age and sex. To meet the definition of a stimulation dependent eQTL, a SNP had to show a statistically significant effect (an FDR < 0.1) in the MTB stimulated samples and a non-significant effect in the unstimulated samples (i.e., it had to be active specifically within the context of active MTB infection). We examined our list of severity-associated SNPs (i.e. those with P<1x10-5 for association with TBscore) within this list of genes that were differentially expressed in response to MTB stimulation (e.g., genes with significant interaction terms in the model above). All eQTL analysis was performed using the Matrix eQTL package in R v3.6.3 [84]. Additional details about the eQTL analysis are found in the S1 Methods. Thus, the primary goal was to identify eQTL’s solely active after MTB stimulation, so we could narrow our results down to variants that play a role within the context of active TB that influenced severity. To determine if the eQTL effects observed in the whole cohort differ between those who are RSTR or LTBI, we performed a stratified analysis among only LTBI subjects and observed how this affected the associations we identified (eg. first order term).
In addition to the eQTL analysis using the data from our monocyte derived macrophage samples, we used FUMA GWAS to query a number of publicly available eQTL databases with the severity associated SNPs [80]. FUMA simultaneously queried several databases to see if these SNPs are eQTL’s for any of the genes and in any of the tissues included in the eQTL catalogue, eQTLgen, BIOSQTL, Blood eQTL Browser, DICE, xQTL Server, and GTEx v8 databases [61,85–90]. In order to enrich these results and look for common biological functions, we uploaded the list of genes for which these SNPs were eQTL’s into the STRING database [91]. This database can implicate protein-protein interactions (PPIs) between the proteins downstream of the genes of interest as well as look for pathway enrichment such as KEGG or Gene Ontologies. PPIs among genes for which the SNPs from our GWAS analysis showed evidence of regulation can link their changes in expression to potential functional roles for their associated proteins and help explain why SNPS associated with severity are regulating these genes.
Supporting information
S1 Table. Components of the TBscore and Points Contributed to Final Score.
https://doi.org/10.1371/journal.pgen.1010387.s002
(DOCX)
S2 Table. Studies of eQTL’s and Gene Expression.
https://doi.org/10.1371/journal.pgen.1010387.s003
(DOCX)
S3 Table. Table of all SNPs showing P<0.05 and Same Direction of Effect for Association with TBscore in Both Cohorts, accompanied by Meta-Analytic Statistics.
(excel file)
https://doi.org/10.1371/journal.pgen.1010387.s004
(XLSX)
S4 Table. Table of all SNPs showing P<0.05 and Same Direction of Effect for Association adjusted for PCs.
(Excel file)
https://doi.org/10.1371/journal.pgen.1010387.s005
(XLSX)
S5 Table. Table of all SNPs showing P<0.05 and Same Direction of Effect for Association with TBscore in Both Cohorts, adjusted for HIV- status.
(Excel file)
https://doi.org/10.1371/journal.pgen.1010387.s006
(XLSX)
S6 Table. MAF, P-value, and β value for rs1848553 Across Cohorts 1 and 2.
https://doi.org/10.1371/journal.pgen.1010387.s007
(DOCX)
S7 Table. Allele Frequencies for rs1848553 in 1000G Project in African Populations and Non-African Super-Populations.
https://doi.org/10.1371/journal.pgen.1010387.s008
(DOCX)
S8 Table. Table of all SNPs showing P<0.05 and Same Direction of Effect for Association with TBscore in Both Cohorts, adjusted for MTB lineage.
(Excel file)
https://doi.org/10.1371/journal.pgen.1010387.s009
(XLSX)
S9 Table. Table of all SNPs showing significant association with TBscore, accompanied by Summary Statistics from CXR GWAS Analysis.
(excel file)
https://doi.org/10.1371/journal.pgen.1010387.s010
(XLSX)
S10 Table. Association with TBscore for SNPs near Genes Previously Associated with TB Susceptibility.
The genes analyzed are from a prior review paper by Stein et al. A +/- 5kb window around each gene was interrogated for potential associations with TBscore and the SNP with the lowest P-value in each window is shown in the table.
https://doi.org/10.1371/journal.pgen.1010387.s011
(DOCX)
S11 Table. Association with TBscore for SNPs within +/- 50Kb of IL12B.
https://doi.org/10.1371/journal.pgen.1010387.s012
(DOCX)
S12 Table. List of genes in REACTOME platelet homeostasis pathway.
(Excel)
https://doi.org/10.1371/journal.pgen.1010387.s013
(XLSX)
S13 Table. List of genes in REACTOME organic anion transport pathway.
(Excel)
https://doi.org/10.1371/journal.pgen.1010387.s014
(XLSX)
S14 Table. Cohort Characteristics for Ugandan Subjects in Matrix eQTL Analysis.
https://doi.org/10.1371/journal.pgen.1010387.s015
(DOCX)
S15 Table. Table of all Differentially Expressed Genes.
(excel file)
https://doi.org/10.1371/journal.pgen.1010387.s016
(XLSX)
S16 Table. Table of all eQTL’s from SNPs Associated with TBscore at P<1e-05.
(excel file)
https://doi.org/10.1371/journal.pgen.1010387.s017
(XLSX)
S17 Table. rs2976562 Allele Frequencies in 1000G Project.
https://doi.org/10.1371/journal.pgen.1010387.s018
(DOCX)
S1 Fig. Quantile-Quantile Plot for P-values for Association between SNPs and TBscore in Cohort 1.
The quantile-quantile (Q-Q) plot shows the inverse log(10) of the observed p-values on the Y-axis relative to what is expected if there was no association on the x-axis. Deviations above the line indicate an association with the outcome. If the line deviates at the low quantiles, then this is considered evidence to suggest genome-wide inflation of the test statistics, which typically indicates unmeasured confounding.
https://doi.org/10.1371/journal.pgen.1010387.s019
(DOCX)
S2 Fig. Quantile-Quantile Plot for P-Values for Association Between SNPs and TBscore in Cohort 2.
The quantile-quantile (Q-Q) plot shows the inverse log(10) of the observed p-values on the Y-axis relative to what is expected if there was no association on the x-axis. Deviations above the line indicate an association with the outcome. If the line deviates at the low quantiles, then this is considered evidence to suggest genome-wide inflation of the test statistics, which typically indicates unmeasured confounding.
https://doi.org/10.1371/journal.pgen.1010387.s020
(DOCX)
S3 Fig. Plot of PC1 vs. PC2 in Cohort 1.
PC1 vs. PC2 plots show the first two principal components of the genotype data, i.e. the ones that represent the greatest amount of variation in the data. Multiple distinct clusters would suggest the existence of sub-groups within the population in question. As we do not see any here, it does not indicate that there any genetically distinct sub-groups within our data. PC1 explains 2.8% of the variation and PC2 explains 2.6% of the variation. The fact that both of these percentages (especially for PC1) are so low also suggests that the principal components are not meaningful representations of sub-groups within the population.
https://doi.org/10.1371/journal.pgen.1010387.s021
(DOCX)
S4 Fig. Plot of PC1 vs. PC2 in Cohort 2.
PC1 vs. PC2 plots show the first two principal components of the genotype data, i.e. the ones that represent the greatest amount of variation in the data. Multiple distinct clusters would suggest the existence of sub-groups within the population in question. As we do not see any here, it does not suggest that there are genetically distinct sub-groups within our data. PC1 explains 4.3% of the variation and PC2 explains 3.5% of the variation. The fact that both of these percentages (especially for PC1) are so low also suggests that the principal components are not meaningful representations of sub-groups within the population.
https://doi.org/10.1371/journal.pgen.1010387.s022
(DOCX)
S5 Fig. Manhattan Plot of P-values for association between SNPs and TBscore in Cohort 1.
The Manhattan plot shows the inverse log(10) of the p-values for the association between each SNP and TBscore on the y-axis and the x-axis represent the physical location of each SNP on the chromosomes, which are in order from 1–22.
https://doi.org/10.1371/journal.pgen.1010387.s023
(DOCX)
S6 Fig. Manhattan Plot of P-values for association between SNPs and TBscore in Cohort 2.
The Manhattan plot shows the inverse log(10) of the p-values for the association between each SNP and TBscore on the y-axis and the x-axis represent the physical location of each SNP on the chromosomes, which are in order from 1–22.
https://doi.org/10.1371/journal.pgen.1010387.s024
(DOCX)
S8 Fig. Manhattan Plot for Meta-Analytic P-values for association between SNPs and CXR Extent.
The Manhattan plot shows the inverse log(10) of the p-values for the association between each SNP and CXR extent on the y-axis and the x-axis represent the physical location of each SNP on the chromosomes, which are in order from 1–22.
https://doi.org/10.1371/journal.pgen.1010387.s026
(DOCX)
S9 Fig. Q-Q Plot for Meta-Analytic P-Values for association between SNPs and CXR Extent.
The quantile-quantile (Q-Q) plot shows the inverse log(10) of the observed p-values on the Y-axis relative to what is expected if there was no association on the x-axis. Deviations above the line indicate an association with the outcome. If the line deviates at the low quantiles, then this is considered evidence to suggest genome-wide inflation of the test statistics, which typically indicates unmeasured confounding.
https://doi.org/10.1371/journal.pgen.1010387.s027
(DOCX)
S10 Fig. Decile Regression between rs184553 and TBscore, adjusted for Sex and HIV.
The Y-axis shows the β value for the association between rs184553 and TBscore while the X-axis shows the quantiles of TBscore. Each dot on the line shows the β value at each decile of TBscore.
https://doi.org/10.1371/journal.pgen.1010387.s028
(DOCX)
S11 Fig. STRING Network for PPI’s from eQTL Response Genes.
This figure is a string diagram showing protein-protein interactions among the genes identified in my analysis. Lines represent an experimentally determined protein-protein interaction and multiple lines between the same two proteins indicates multiple interactions that have been identified, but multiple interactions between the same two proteins are still considered to be one edge.
https://doi.org/10.1371/journal.pgen.1010387.s029
(DOCX)
S12 Fig. Expression of SLA in Immune Cells from DICE Database.
https://doi.org/10.1371/journal.pgen.1010387.s030
(DOCX)
S13 Fig. Imputation and QC of SNPs in Cohorts 1 and 2.
https://doi.org/10.1371/journal.pgen.1010387.s031
(DOCX)
Acknowledgments
We would like to acknowledge the invaluable contribution made by the study medical officers, health visitors, laboratory and data personnel: Dr. Lorna Nshuti, Dr. Roy Mugerwa, Dr. Sarah Zalwango, Dr. Christopher Whalen, Dr. Deo Mulindwa, Dr. Christina Lancioni, Allan Chiunda, Denise Johnson, Bonnie Thiel, Mark Breda, Dennis Dobbs, Hussein Kisingo, Mary Rutaro, Albert Muganda, Richard Bamuhimbisa, Yusuf Mulumba, Deborah Nsamba, Barbara Kyeyune, Faith Kintu, Mary Nsereko, Gladys Mpalanyi, Janet Mukose, Grace Tumusiime, Pierre Peters, Annet Kawuma, Saidah Menya, Joan Nassuna, Alphonse Okwera, Keith Chervenak, Karen Morgan, Alfred Etwom, Micheal Angel Mugerwa, and Lisa Kucharski. We would like to acknowledge and thank Dr. Francis Adatu Engwau, Head of the Uganda National Tuberculosis and Leprosy Program, for his support of this project. We would like to acknowledge the medical officers, nurses and counselors at the National Tuberculosis Treatment Centre, Mulago Hospital, the Ugandan National Tuberculosis and Leprosy Program and the Uganda Tuberculosis Investigation Bacteriological Unit, Wandegeya, for their contributions to this study. This study would not be possible without the generous participation of the Ugandan patients and families.
References
- 1.
World Health Organization. Tuberculosis https://www.who.int/news-room/fact-sheets/detail/tuberculosis2020.
- 2. Bell LCK, Noursadeghi M. Pathogenesis of HIV-1 and Mycobacterium tuberculosis co-infection. Nature Reviews Microbiology. 2017;16:80. pmid:29109555
- 3. World Health Organization. Global Tuberculosis Report 2018 Factsheet. 2018.
- 4. Takarinda KC, Sandy C, Masuka N, Hazangwe P, Choto RC, Mutasa-Apollo T, et al. Factors Associated with Mortality among Patients on TB Treatment in the Southern Region of Zimbabwe, 2013. Tuberc Res Treat. 2017;2017:6232071-. Epub 2017/03/02. pmid:28352474.
- 5. Adamu AL, Gadanya MA, Abubakar IS, Jibo AM, Bello MM, Gajida AU, et al. High mortality among tuberculosis patients on treatment in Nigeria: a retrospective cohort study. BMC infectious diseases. 2017;17(1):170. pmid:28231851
- 6. Bukhari M, Aslam MA, Khan A, Iram Q, Akbar A, Naz AG, et al. TLR8 gene polymorphism and association in bacterial load in southern Punjab of Pakistan: an association study with pulmonary tuberculosis. International journal of immunogenetics. 2015;42(1):46–51. Epub 2015/01/13. pmid:25572425.
- 7. Ganachari M, Guio H, Zhao N, Flores-Villanueva PO. Host gene-encoded severe lung TB: from genes to the potential pathways. Genes and immunity. 2012;13(8):605–20. Epub 2012/09/21. pmid:22992722; PubMed Central PMCID: PMC3518758.
- 8. Jiang D, Wubuli A, Hu X, Ikramullah S, Maimaiti A, Zhang W, et al. The variations of IL-23R are associated with susceptibility and severe clinical forms of pulmonary tuberculosis in Chinese Uygurs. BMC infectious diseases. 2015;15:550. Epub 2015/12/03. pmid:26626589; PubMed Central PMCID: PMC4665827.
- 9. Jurado JO, Pasquinelli V, Alvarez IB, Pena D, Rovetta AI, Tateosian NL, et al. IL-17 and IFN-gamma expression in lymphocytes from patients with active tuberculosis correlates with the severity of the disease. Journal of leukocyte biology. 2012;91(6):991–1002. Epub 2012/03/15. pmid:22416258; PubMed Central PMCID: PMC3360475.
- 10. Kim HS, Park MH, Song EY, Park H, Kwon SY, Han SK, et al. Association of HLA-DR and HLA-DQ genes with susceptibility to pulmonary tuberculosis in Koreans: preliminary evidence of associations with drug resistance, disease severity, and disease recurrence. Human immunology. 2005;66(10):1074–81. Epub 2006/01/03. pmid:16386650.
- 11. Magee MJ, Sun YV, Brust JCM, Shah NS, Ning Y, Allana S, et al. Polymorphisms in the vitamin D receptor gene are associated with reduced rate of sputum culture conversion in multidrug-resistant tuberculosis patients in South Africa. PloS one. 2017;12(7):e0180916. Epub 2017/07/13. pmid:28700743; PubMed Central PMCID: PMC5507304.
- 12. Masood KI, Rottenberg ME, Carow B, Rao N, Ashraf M, Hussain R, et al. SOCS1 gene expression is increased in severe pulmonary tuberculosis. Scandinavian journal of immunology. 2012;76(4):398–404. Epub 2012/06/08. pmid:22670716.
- 13. Najmi N, Kaur G, Sharma SK, Mehra NK. Human Toll-like receptor 4 polymorphisms TLR4 Asp299Gly and Thr399Ile influence susceptibility and severity of pulmonary tuberculosis in the Asian Indian population. Tissue antigens. 2010;76(2):102–9. Epub 2010/04/21. pmid:20403143.
- 14. Singh A, Gaughan JP, Kashyap VK. SLC11A1 and VDR gene variants and susceptibility to tuberculosis and disease progression in East India. The international journal of tuberculosis and lung disease: the official journal of the International Union against Tuberculosis and Lung Disease. 2011;15(11):1468–74, i. Epub 2011/10/20. pmid:22008758.
- 15. Streata I, Weiner J 3rd, Iannaconne M, McEwen G, Ciontea MS, Olaru M, et al. The CARD9 Polymorphisms rs4077515, rs10870077 and rs10781499 Are Uncoupled from Susceptibility to and Severity of Pulmonary Tuberculosis. PloS one. 2016;11(9):e0163662. Epub 2016/09/30. pmid:27684065; PubMed Central PMCID: PMC5042433.
- 16. Thye T, Scarisbrick G, Browne EN, Chinbuah MA, Gyapong J, Osei I, et al. CTLA4 autoimmunity-associated genotype contributes to severe pulmonary tuberculosis in an African population. PloS one. 2009;4(7):e6307. Epub 2009/07/18. pmid:19609446; PubMed Central PMCID: PMC2707621.
- 17. McHenry ML, Bartlett J, Igo RP Jr., Wampande EM, Benchek P, Mayanja-Kizza H, et al. Interaction between host genes and Mycobacterium tuberculosis lineage can affect tuberculosis severity: Evidence for coevolution? PLOS Genetics. 2020;16(4):e1008728. pmid:32352966
- 18. Wejse C, Gustafson P, Nielsen J, Gomes VF, Aaby P, Andersen PL, et al. TBscore: Signs and symptoms from tuberculosis patients in a low-resource setting have predictive value and may be used to assess clinical course. Scand J Infect Dis. 2008;40(2):111–20. Epub 2007/09/14. pmid:17852907.
- 19. Rudolf F. The Bandim TBscore—reliability, further development, and evaluation of potential uses. Global health action. 2014;7:24303. Epub 2014/05/27. pmid:24857613; PubMed Central PMCID: PMC4032506.
- 20. Theron G, Zijenah L, Chanda D, Clowes P, Rachow A, Lesosky M, et al. Feasibility, accuracy, and clinical effect of point-of-care Xpert MTB/RIF testing for tuberculosis in primary-care settings in Africa: a multicentre, randomised, controlled trial. Lancet (London, England). 2014;383(9915):424–35. Epub 2013/11/02. pmid:24176144.
- 21. Rasmussen TA, Sogaard OS, Camara C, Andersen PL, Wejse C. Serum procalcitonin in pulmonary tuberculosis. The international journal of tuberculosis and lung disease: the official journal of the International Union against Tuberculosis and Lung Disease. 2011;15(2):251–6, i. Epub 2011/01/12. pmid:21219690.
- 22. Rudolf F, Joaquim LC, Vieira C, Bjerregaard-Andersen M, Andersen A, Erlandsen M, et al. The Bandim tuberculosis score: reliability and comparison with the Karnofsky performance score. Scand J Infect Dis. 2013;45(4):256–64. Epub 2012/11/02. pmid:23113626.
- 23. McHenry ML, Williams SM, Stein CM. Genetics and evolution of tuberculosis pathogenesis: New perspectives and approaches. Infection, Genetics and Evolution. 2020:104204. pmid:31981609
- 24. Seshadri C, Sedaghat N, Campo M, Peterson G, Wells RD, Olson GS, et al. Transcriptional networks are associated with resistance to Mycobacterium tuberculosis infection. PLoS One. 2017;12(4):e0175844. pmid:28414762; PubMed Central PMCID: PMC5393882.
- 25. Seshadri C, Shenoy M, Wells RD, Hensley-McBain T, Andersen-Nissen E, McElrath MJ, et al. Human CD1a deficiency is common and genetically regulated. Journal of immunology (Baltimore, Md: 1950). 2013;191(4):1586–93. Epub 2013/07/17. pmid:23858036; PubMed Central PMCID: PMC3748949.
- 26. Shah JA, Vary JC, Chau TT, Bang ND, Yen NT, Farrar JJ, et al. Human TOLLIP regulates TLR2 and TLR4 signaling and its polymorphisms are associated with susceptibility to tuberculosis. J Immunol. 2012;189(4):1737–46. Epub 2012/07/11. pmid:22778396; PubMed Central PMCID: PMC3428135.
- 27. Barreiro LB, Tailleux L, Pai AA, Gicquel B, Marioni JC, Gilad Y. Deciphering the genetic architecture of variation in the immune response to Mycobacterium tuberculosis infection. Proc Natl Acad Sci U S A. 2012;109(4):1204–9. Epub 2012/01/12. pmid:22233810; PubMed Central PMCID: PMC3268270.
- 28. Pacis A, Nedelec Y, Barreiro LB. When genetics meets epigenetics: deciphering the mechanisms controlling inter-individual variation in immune responses to infection. Current opinion in immunology. 2014;29:119–26. Epub 2014/07/02. pmid:24981784.
- 29. Wainberg M, Sinnott-Armstrong N, Mancuso N, Barbeira AN, Knowles DA, Golan D, et al. Opportunities and challenges for transcriptome-wide association studies. Nature genetics. 2019;51(4):592–9. Epub 2019/03/31. pmid:30926968; PubMed Central PMCID: PMC6777347.
- 30. Jabot-Hanin F, Cobat A, Feinberg J, Orlova M, Niay J, Deswarte C, et al. An eQTL variant of ZXDC is associated with IFN-γ production following Mycobacterium tuberculosis antigen-specific stimulation. Sci Rep. 2017;7(1):12800. pmid:28993696
- 31. Ai JW, Zhang H, Zhou Z, Weng S, Huang H, Wang S, et al. Gene expression pattern analysis using dual-color RT-MLPA and integrative genome-wide association studies of eQTL for tuberculosis suscepitibility. Respir Res. 2021;22(1):23. Epub 2021/01/22. pmid:33472618; PubMed Central PMCID: PMC7816316.
- 32. Gamazon ER, Segrè AV, van de Bunt M, Wen X, Xi HS, Hormozdiari F, et al. Using an atlas of gene regulation across 44 human tissues to inform complex disease-and trait-associated variation. Nature genetics. 2018;50(7):956–67. pmid:29955180
- 33. Cohen SB, Gern BH, Delahaye JL, Adams KN, Plumlee CR, Winkler JK, et al. Alveolar Macrophages Provide an Early Mycobacterium tuberculosis Niche and Initiate Dissemination. Cell host & microbe. 2018;24(3):439–46 e4. Epub 2018/08/28. pmid:30146391; PubMed Central PMCID: PMC6152889.
- 34. Stein CM, Nsereko M, Malone LL, Okware B, Kisingo H, Nalukwago S, et al. Long-term Stability of Resistance to Latent Mycobacterium tuberculosis Infection in Highly Exposed Tuberculosis Household Contacts in Kampala, Uganda. Clinical infectious diseases: an official publication of the Infectious Diseases Society of America. 2019;68(10):1705–12. Epub 2018/08/31. pmid:30165605; PubMed Central PMCID: PMC6495009.
- 35. Simmons JD, Stein CM, Seshadri C, Campo M, Alter G, Fortune S, et al. Immunological mechanisms of human resistance to persistent Mycobacterium tuberculosis infection. Nat Rev Immunol. 2018;18(9):575–89. Epub 2018/06/14. pmid:29895826.
- 36. Flicek P, Amode MR, Barrell D, Beal K, Brent S, Chen Y, et al. Ensembl 2011. Nucleic acids research. 2011;39(Database issue):D800–D6. Epub 2010/11/02. pmid:21045057.
- 37. Stelzer G, Rosen N, Plaschkes I, Zimmerman S, Twik M, Fishilevich S, et al. The GeneCards Suite: From Gene Data Mining to Disease Genome Sequence Analyses. Current Protocols in Bioinformatics. 2016;54(1):1.30.1–1..3. pmid:27322403
- 38. McHenry ML, Wampande EM, Joloba ML, Malone LL, Mayanja-Kizza H, Bush WS, et al. Interaction between M. tuberculosis Lineage and Human Genetic Variants Reveals Novel Pathway Associations with Severity of TB. Pathogens. 2021;10(11). Epub 2021/11/28. pmid:34832643; PubMed Central PMCID: PMC8617877.
- 39. McHenry ML, Bartlett J, Igo RP Jr, Wampande EM, Benchek P, Mayanja-Kizza H, et al. Interaction between host genes and Mycobacterium tuberculosis lineage can affect tuberculosis severity: Evidence for coevolution? PLoS genetics. 2020;16(4):e1008728. Epub 2020/05/01. pmid:32352966; PubMed Central PMCID: PMC7217476.
- 40. Stein CM, Sausville L, Wejse C, Sobota RS, Zetola NM, Hill PC, et al. Genomics of human pulmonary tuberculosis: from genes to pathways. Current genetic medicine reports. 2017;5(4):149–66. Epub 2017/10/12. pmid:29805915.
- 41. Williams SM, Haines JL. Correcting away the hidden heritability. Annals of human genetics. 2011;75(3):348–50. Epub 2011/04/15. pmid:21488852.
- 42. Wang X, Tucker NR, Rizki G, Mills R, Krijger PHL, de Wit E, et al. Discovery and validation of sub-threshold genome-wide association study loci using epigenomic signatures. eLife. 2016;5:e10557. pmid:27162171
- 43. Hammond RK, Pahl MC, Su C, Cousminer DL, Leonard ME, Lu S, et al. Biological constraints on GWAS SNPs at suggestive significance thresholds reveal additional BMI loci. eLife. 2021;10:e62206. pmid:33459256
- 44. Consortium GP. A global reference for human genetic variation. Nature. 2015;526(7571):68. pmid:26432245
- 45. Pober JS, Sessa WC. Inflammation and the blood microvascular system. Cold Spring Harb Perspect Biol. 2014;7(1):a016345–a. pmid:25384307.
- 46. Pober JS, Tellides G. Participation of blood vessel cells in human adaptive immune responses. Trends in immunology. 2012;33(1):49–57. Epub 2011/10/24. pmid:22030237.
- 47. Wentowski C, Mewada N, Nielsen ND. Sepsis in 2018: a review. Anaesthesia & Intensive Care Medicine. 2019;20(1):6–13.
- 48. Pranata R, Lim MA, Huang I, Raharjo SB, Lukito AA. Hypertension is associated with increased mortality and severity of disease in COVID-19 pneumonia: A systematic review, meta-analysis and meta-regression. Journal of the renin-angiotensin-aldosterone system: JRAAS. 2020;21(2):1470320320926899. Epub 2020/05/16. pmid:32408793; PubMed Central PMCID: PMC7231906.
- 49. Sheu J-J, Chiou H-Y, Kang J-H, Chen Y-H, Lin H-C. Tuberculosis and the risk of ischemic stroke: a 3-year follow-up study. Stroke. 2010;41(2):244–9. pmid:20035070
- 50. Misra UK, Kalita J, Maurya PK. Stroke in tuberculous meningitis. Journal of the neurological sciences. 2011;303(1–2):22–30. Epub 2011/01/29. pmid:21272895.
- 51. Chen S, Chen J, Chen L, Zhang Q, Luo X, Zhang W. Mycobacterium tuberculosis infection is associated with the development of erythema nodosum and nodular vasculitis. PloS one. 2013;8(5):e62653. Epub 2013/05/08. pmid:23650522; PubMed Central PMCID: PMC3641092.
- 52. Portier I, Campbell RA. Role of Platelets in Detection and Regulation of Infection. Arteriosclerosis, Thrombosis, and Vascular Biology. 2021;41(1):70–8. pmid:33115274
- 53. Speth C, Löffler J, Krappmann S, Lass-Flörl C, Rambach G. Platelets as immune cells in infectious diseases. Future microbiology. 2013;8(11):1431–51. Epub 2013/11/10. pmid:24199802.
- 54. Akira S, Uematsu S, Takeuchi O. Pathogen recognition and innate immunity. Cell. 2006;124(4):783–801. Epub 2006/02/25. pmid:16497588.
- 55. Chen L, Deng H, Cui H, Fang J, Zuo Z, Deng J, et al. Inflammatory responses and inflammation-associated diseases in organs. Oncotarget. 2017;9(6):7204–18. pmid:29467962.
- 56. Kothari H, Keshava S, Vatsyayan R, Mackman N, Rao LVM, Pendurthi UR. Role of tissue factor in Mycobacterium tuberculosis-induced inflammation and disease pathogenesis. PloS one. 2014;9(12):e114141–e. pmid:25462128.
- 57. Caccamo N, Dieli F. Inflammation and the coagulation system in tuberculosis: Tissue Factor leads the dance. European journal of immunology. 2016;46(2):303–6. Epub 2016/01/15. pmid:26763085.
- 58. McDonald B, Dunbar M. Platelets and Intravascular Immunity: Guardians of the Vascular Space During Bloodstream Infections and Sepsis. Frontiers in immunology. 2019;10:2400–. pmid:31681291.
- 59. Lin C-H, Lin C-J, Kuo Y-W, Wang J-Y, Hsu C-L, Chen J-M, et al. Tuberculosis mortality: patient characteristics and causes. BMC infectious diseases. 2014;14:5–. pmid:24387757.
- 60. Cunningham F, Achuthan P, Akanni W, Allen J, Amode MR, Armean IM, et al. Ensembl 2019. Nucleic acids research. 2019;47(D1):D745–D51. pmid:30407521
- 61. Ha B, Greenbaum JA, Shmiedel BJ, Singh D, Madrigal A, Valdovino-Gonzalez AG, et al. Database of Immune Cell EQTLs, Expression, Epigenomics. The Journal of Immunology. 2019;202(1 Supplement):131.18–.18.
- 62. Sosinowski T, Pandey A, Dixit VM, Weiss A. Src-like adaptor protein (SLAP) is a negative regulator of T cell receptor signaling. J Exp Med. 2000;191(3):463–74. Epub 2000/02/09. pmid:10662792; PubMed Central PMCID: PMC2195826.
- 63. Tang J, Sawasdikosol S, Chang JH, Burakoff SJ. SLAP, a dimeric adapter protein, plays a functional role in T cell receptor signaling. Proceedings of the National Academy of Sciences of the United States of America. 1999;96(17):9775–80. pmid:10449770.
- 64. Liontos LM, Dissanayake D, Ohashi PS, Weiss A, Dragone LL, McGlade CJ. The Src-Like Adaptor Protein Regulates GM-CSFR Signaling and Monocytic Dendritic Cell Maturation. The Journal of Immunology. 2011;186(4):1923–33. pmid:21220694
- 65. Cooper AM, Magram J, Ferrante J, Orme IM. Interleukin 12 (IL-12) is crucial to the development of protective immunity in mice intravenously infected with mycobacterium tuberculosis. The Journal of experimental medicine. 1997;186(1):39–45. Epub 1997/07/07. pmid:9206995; PubMed Central PMCID: PMC2198958.
- 66. Harris J, Keane J. How tumour necrosis factor blockers interfere with tuberculosis immunity. Clinical and experimental immunology. 2010;161(1):1–9. Epub 2010/05/25. pmid:20491796; PubMed Central PMCID: PMC2940142.
- 67. Nunes-Alves C, Booty MG, Carpenter SM, Jayaraman P, Rothchild AC, Behar SM. In search of a new paradigm for protective immunity to TB. Nature reviews Microbiology. 2014;12(4):289–99. Epub 2014/03/05. pmid:24590243; PubMed Central PMCID: PMC4085047.
- 68. Stein CM, Zalwango S, Malone LL, Thiel B, Mupere E, Nsereko M, et al. Resistance and susceptibility to Mycobacterium tuberculosis infection and disease in tuberculosis households in Kampala, Uganda. Am J Epidemiol. 2018;187:1477–89. pmid:29304247
- 69. Fluegge K, Malone LL, Nsereko M, Okware B, Wejse C, Kisingo H, et al. Impact of geographic distance on appraisal delay for active TB treatment seeking in Uganda: a network analysis of the Kawempe Community Health Cohort Study. BMC Public Health. 2018;18(1):798. Epub 2018/06/27. pmid:29940918; PubMed Central PMCID: PMC6019214.
- 70. Stein CM, Zalwango S, Malone LL, Won S, Mayanja-Kizza H, Mugerwa RD, et al. Genome scan of M. tuberculosis infection and disease in Ugandans. PloS one. 2008;3(12):e4094. pmid:19116662
- 71. Stein CM, Nsereko M, Malone LL, Okware B, Kisingo H, Nalukwago S, et al. Long-term Stability of Resistance to Latent Mycobacterium tuberculosis Infection in Highly Exposed Tuberculosis Household Contacts in Kampala, Uganda. Clinical Infectious Diseases. 2019;68(10):1705–12. pmid:30165605
- 72. Simmons JD, Dill-McFarland KA, Stein CM, Van PT, Chihota V, Ntshiqa T, et al. Monocyte Transcriptional Responses to Mycobacterium tuberculosis Associate with Resistance to Tuberculin Skin Test and Interferon Gamma Release Assay Conversion. Msphere. 2022;7(3):e00159–22. pmid:35695527
- 73. Stein CM, Nsereko M, Malone LL, Okware B, Kisingo H, Nalukwago S, et al. Long-term Stability of Resistance to Latent Mycobacterium tuberculosis Infection in Highly Exposed Tuberculosis Household Contacts in Kampala, Uganda. Clinical Infectious Diseases. 2018;68(10):1705–12. pmid:30165605
- 74. Stein CM, Zalwango S, Malone LL, Thiel B, Mupere E, Nsereko M, et al. Resistance and Susceptibility to Mycobacterium tuberculosis Infection and Disease in Tuberculosis Households in Kampala, Uganda. Am J Epidemiol. 2018;187(7):1477–89. Epub 2018/01/06. pmid:29304247; PubMed Central PMCID: PMC6031055.
- 75. Simmons JD, Van PT, Stein CM, Chihota V, Ntshiqa T, Maenetje P, et al. Monocyte metabolic transcriptional programs associate with resistance to tuberculin skin test/interferon-γ release assay conversion. The Journal of clinical investigation. 2021;131(14). Epub 2021/06/11. pmid:34111032; PubMed Central PMCID: PMC8279582.
- 76. Igo RP Jr, Hall NB, Malone LL, Hall JB, Truitt B, Qiu F, et al. Fine-mapping analysis of a chromosome 2 region linked to resistance to Mycobacterium tuberculosis infection in Uganda reveals potential regulatory variants. Genes Immun. 2018. Epub 2018/08/14. pmid:30100616; PubMed Central PMCID: PMC6374218.
- 77. Das S, Forer L, Schönherr S, Sidore C, Locke AE, Kwong A, et al. Next-generation genotype imputation service and methods. Nat Genet. 2016;48(10):1284–7. Epub 2016/08/30. pmid:27571263; PubMed Central PMCID: PMC5157836.
- 78. McCarthy S, Das S, Kretzschmar W, Delaneau O, Wood AR, Teumer A, et al. A reference panel of 64,976 haplotypes for genotype imputation. Nature genetics. 2016;48(10):1279–83. Epub 2016/08/23. pmid:27548312; PubMed Central PMCID: PMC5388176.
- 79. Deeks JJ, Higgins JP, Altman DG. Analysing Data and Undertaking Meta-Analyses. Cochrane Handbook for Systematic Reviews of Interventions2008. p. 243–96.
- 80. Watanabe K, Taskesen E, van Bochoven A, Posthuma D. Functional mapping and annotation of genetic associations with FUMA. Nature communications. 2017;8(1):1826–. pmid:29184056.
- 81. von Mering C, Jensen LJ, Snel B, Hooper SD, Krupp M, Foglierini M, et al. STRING: known and predicted protein-protein associations, integrated and transferred across organisms. Nucleic acids research. 2005;33(Database issue):D433–7. Epub 2004/12/21. pmid:15608232; PubMed Central PMCID: PMC539959.
- 82. de Leeuw CA, Mooij JM, Heskes T, Posthuma D. MAGMA: generalized gene-set analysis of GWAS data. PLoS computational biology. 2015;11(4):e1004219. Epub 2015/04/18. pmid:25885710; PubMed Central PMCID: PMC4401657.
- 83. Diagnostic Standards and Classification of Tuberculosis in Adults and Children. This official statement of the American Thoracic Society and the Centers for Disease Control and Prevention was adopted by the ATS Board of Directors, July 1999. This statement was endorsed by the Council of the Infectious Disease Society of America, September 1999. Am J Respir Crit Care Med. 2000;161(4 Pt 1):1376–95.
- 84. Shabalin AA. Matrix eQTL: ultra fast eQTL analysis via large matrix operations. Bioinformatics. 2012;28(10):1353–8. pmid:22492648
- 85. Zhernakova DV, Deelen P, Vermaat M, van Iterson M, van Galen M, Arindrarto W, et al. Identification of context-dependent expression quantitative trait loci in whole blood. Nature genetics. 2017;49(1):139–45. Epub 2016/12/06. pmid:27918533.
- 86. Consortium G. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science (New York, NY). 2020;369(6509):1318–30.
- 87. Westra HJ, Peters MJ, Esko T, Yaghootkar H, Schurmann C, Kettunen J, et al. Systematic identification of trans eQTLs as putative drivers of known disease associations. Nature genetics. 2013;45(10):1238–43. Epub 2013/09/10. pmid:24013639; PubMed Central PMCID: PMC3991562.
- 88. Kerimov N, Hayhurst JD, Manning JR, Walter P, Kolberg L, Peikova K, et al. eQTL Catalogue: a compendium of uniformly processed human gene expression and splicing QTLs. BioRxiv. 2020.
- 89. Arends D, van der Velde KJ, Prins P, Broman KW, Möller S, Jansen RC, et al. xQTL workbench: a scalable web environment for multi-level QTL analysis. Bioinformatics. 2012;28(7):1042–4. pmid:22308096
- 90. van der Wijst MG, de Vries DH, Groot HE, Trynka G, Hon C-C, Bonder M-J, et al. Science Forum: The single-cell eQTLGen consortium. eLife. 2020;9:e52155.
- 91. Szklarczyk D, Gable AL, Lyon D, Junge A, Wyder S, Huerta-Cepas J, et al. STRING v11: protein–protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic acids research. 2019;47(D1):D607–D13. pmid:30476243