https://orcid.org/0000-0003-0667-9926
Figures
Abstract
Mycobacterium tuberculosis infects two billion people across the globe, and results in 8–9 million new tuberculosis (TB) cases and 1–1.5 million deaths each year. Most patients have no known genetic basis that predisposes them to disease. Here, we investigate the complex genetic basis of pulmonary TB by modelling human genetic diversity with the Diversity Outbred mouse population. When infected with M. tuberculosis, one-third develop early onset, rapidly progressive, necrotizing granulomas and succumb within 60 days. The remaining develop non-necrotizing granulomas and survive longer than 60 days. Genetic mapping using immune and inflammatory mediators; and clinical, microbiological, and granuloma correlates of disease identified five new loci on mouse chromosomes 1, 2, 4, 16; and three known loci on chromosomes 3 and 17. Further, multiple positively correlated traits shared loci on chromosomes 1, 16, and 17 and had similar patterns of allele effects, suggesting these loci contain critical genetic regulators of inflammatory responses to M. tuberculosis. To narrow the list of candidate genes, we used a machine learning strategy that integrated gene expression signatures from lungs of M. tuberculosis-infected Diversity Outbred mice with gene interaction networks to generate scores representing functional relationships. The scores were used to rank candidates for each mapped trait, resulting in 11 candidate genes: Ncf2, Fam20b, S100a8, S100a9, Itgb5, Fstl1, Zbtb20, Ddr1, Ier3, Vegfa, and Zfp318. Although all candidates have roles in infection, inflammation, cell migration, extracellular matrix remodeling, or intracellular signaling, and all contain single nucleotide polymorphisms (SNPs), SNPs in only four genes (S100a8, Itgb5, Fstl1, Zfp318) are predicted to have deleterious effects on protein functions. We performed methodological and candidate validations to (i) assess biological relevance of predicted allele effects by showing that Diversity Outbred mice carrying PWK/PhJ alleles at the H-2 locus on chromosome 17 QTL have shorter survival; (ii) confirm accuracy of predicted allele effects by quantifying S100A8 protein in inbred founder strains; and (iii) infection of C57BL/6 mice deficient for the S100a8 gene. Overall, this body of work demonstrates that systems genetics using Diversity Outbred mice can identify new (and known) QTLs and functionally relevant gene candidates that may be major regulators of complex host-pathogens interactions contributing to granuloma necrosis and acute inflammation in pulmonary TB.
Author summary
We investigated the genetic basis of susceptibility to Mycobacterium tuberculosis using Diversity Outbred mice, a mouse population suited for studying complex genotype-phenotype relationships. We found five new genetic loci and three known loci. Three loci associated with multiple correlated disease traits likely contain genes that are major regulators of host inflammatory responses which favor M. tuberculosis growth. Further, these three loci contain four gene candidates with single nucleotide polymorphisms that are predicted to have deleterious effects upon protein functions.
Citation: Gatti DM, Tyler AL, Mahoney JM, Churchill GA, Yener B, Koyuncu D, et al. (2024) Systems genetics uncover new loci containing functional gene candidates in Mycobacterium tuberculosis-infected Diversity Outbred mice. PLoS Pathog 20(6): e1011915. https://doi.org/10.1371/journal.ppat.1011915
Editor: Padmini Salgame, New Jersey Medical School, UNITED STATES
Received: December 22, 2023; Accepted: April 17, 2024; Published: June 11, 2024
Copyright: © 2024 Gatti 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: Data has been deposited in Gene Expression Omnibus (GEO), and assigned Series ID GSE179417.
Funding: • The work was supported by funding from the National Institutes of Health R21 AI115038 (GB); the National Institutes of Health R01 HL145411 (GB); the American Lung Association Biomedical Research Grant RG-349504 (GB); Tufts University’s Cummings School of Veterinary Medicine Summer Research Program (AS), and the National Institutes of Health HL126066 (IK). 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
The number of humans who develop active pulmonary tuberculosis (TB) is minor compared to those who eliminate or control Mycobacterium tuberculosis (5–10% vs 90–95%), yet morbidity and mortality from TB remain high. COVID-19 mortality temporarily surpassed global TB mortality, but TB has remained in the top two leading causes of death due to an infectious disease for decades, killing more people than HIV/AIDS and malaria. Pulmonary TB is the most common and most contagious form of TB, with mortality rates >40% if untreated and higher if caused by antibiotic resistant M. tuberculosis [1–8]. Human responses to M. tuberculosis range from fulminant pulmonary TB that develops within weeks to lifelong control or complete clearance of bacilli [9–11]. Further, a body of evidence shows an interesting paradox: Immune competence is necessary to restrict M. tuberculosis growth [12], but is not sufficient to prevent disease [13].
The variable responses to M. tuberculosis and lack of single gene defects in most patients indicate a complex genetic basis for pulmonary TB, and this has been investigated by linkage association mapping, genome wide association studies, and other methods, recently reviewed [14–17]. Reviews identify knowledge gaps attributable to using laboratory mouse strains that do not replicate key disease traits of human pulmonary TB, e.g., granuloma necrosis [18–23]. To address these gaps, we and others investigate genetic contribution, molecular pathogenesis, and vaccine efficacy in the Diversity Outbred mouse population, panels of Collaborative Cross inbred mice and their progeny, some of which do develop human-like pulmonary TB [24–29]. These mice are valuable resources to model complex genotype-phenotype associations; tools to dissect the genetic basis of disease; and a means to assess effects of candidate genetic polymorphisms in vivo.
The Diversity Outbred mouse population originated by breeding eight inbred founder strains together, resulting in a reproducible experimental population with balanced allele frequencies of one-eighth across the genome [30]. This is important for genetic mapping studies because low allele frequencies in natural populations can diminish power and increase false positive associations [31]. Further, Diversity Outbred mice carry over 40 million variants [32], some of which alter regulatory elements, splice sites, and protein-coding sequences. This defined genetic architecture allows rigorous investigation of genotype-phenotype association in context of M. tuberculosis infection. Here, we selected phenotypes of pulmonary TB (i.e., quantified traits) for biological relevance, many of which are shared by humans, non-human primates, and Diversity Outbred mice [33]. Traits were selected to span clinical, granuloma, microbiological, and immune/inflammatory mediators that are known to positively correlate with disease, or are investigated as diagnostic biomarkers, or have been shown to activate or negatively regulate immune/inflammatory responses to M. tuberculosis. Briefly, the clinical trait of disease was weight loss. The granuloma trait was necrosis. The microbiological trait was M. tuberculosis lung burden. Chemokine correlates of disease were CXCL1, CXCL2, CXCL5 [24,34]. TH1 proinflammatory cytokines and acquired immune mediators were tumor necrosis factor (TNF) [35], interferon-gamma (IFN-gamma) [36], and interleukin (IL)-12 [37]. Biomarker candidates were S100A8, matrix metalloproteinase (MMP)8, and vascular endothelial growth factor (VEGF) [34] and the immune regulatory cytokine was IL-10 [38].
To find genetic loci associated with pulmonary TB traits, we used quantitative trait locus (QTL) mapping. Next we ranked candidate genes within the Diversity Outbred tuberculosis susceptibility (Dots) loci that were associated with correlated, colocalized traits by using a machine learning algorithm [39,40] to find genes functionally related to the mapped traits and the fit models scored each candidate [41]. All candidates contain a variety of SNPs as annotated in Mouse Variation Registry (MVAR). Seven of the eleven candidates contain missense SNPs in protein coding regions, and of those, the SNPs in four candidates (S100a8, Itgb5, Fstl1, and Zfp318) are predicted to have deleterious consequences on protein functions. Published in vitro and in vivo studies show roles for three candidates (Itgb5, Fstl1, S100a8) in bacterial lung infections including M. tuberculosis [34,42–47], although questions remain. To our knowledge, the other eight candidates have no published roles in M. tuberculosis infection but have been shown in other systems to contribute to cell stress responses, signaling pathways, adhesion and migration; extracellular matrix synthesis, tissue remodeling and angiogenesis; immune cell metabolism; macrophage inflammatory responses; and viral hepatitis [48–60]. Overall, ten candidate genes have roles in innate immune responses suggesting that genetically controlled responses of epithelial and endothelial cells, neutrophils, and monocytes, macrophages to M. tuberculosis bacilli are the primary drivers of susceptibility to M. tuberculosis and to disease progression in pulmonary TB. Only one candidate (Zfp318) has a direct role in acquired, antigen-specific immunity.
Methods
Ethics statement
Tufts University’s Institutional Animal Care and Use Committee (IACUC) approved this work under protocols G2012-53; G2015-33; G2018-33; and G2020-121. Tufts University’s Institutional Biosafety Committee (IBC) approved this work under registrations: GRIA04; GRIA10; GRIA17, and 2020-G61.
Mice
Female Diversity Outbred mice from generations 15 16, 21, 22, 34, 35, 37 and 42 and the inbred founder strains: A/J, C57BL/6J, 129S1/SvlmJ), NOD/LtJ, NZO/HILtJ, CAST/EiJ, PWK/PhJ, and WSB/EiJ mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and group housed (n = 5–7 mice per cage) on Innovive (San Diego, CA) or Allentown Inc (Allentown, NJ) ventilated, HEPA-filtered racks in the New England Regional Biosafety Laboratory (Tufts University, Cummings School of Veterinary Medicine, North Grafton, MA) or at The Ohio State University, Columbus, OH. The light cycle was 12 hours of light; 12 hours of dark. Two breeding pairs of female and male C57BL/6 inbred mice carrying null mutation for S100a8 gene were a gift from Dr. Philippe Tessier, Department of Microbiology and Immunology, Faculty of Medicine, Université Laval. After quarantine, breeders were used to establish a colony of S100a8 homozygous knock out (KO), heterozygous (HET) and wild-type (WT) C57BL/6 inbred mice, and genotypes confirmed (Transnetyx, Cordova, TN). Mice were housed in disposable sterile caging or re-usable autoclaved caging containing sterile corn-cob bedding, with sterile paper nestlets (Scotts Pharma Solutions, Marlborough, MA), and/or sterile enrichment paperboard or plastic “houses.” Cages were changed every other week or sooner if soiled. Mice were provided with sterile mouse chow (Envigo, Indianapolis, IA) and sterile, acidified water ad libidum.
M. tuberculosis Aerosol Infection
Female Diversity Outbred mice and inbred founder strains were infected with aerosolized M. tuberculosis strain Erdman bacilli using a custom-built CH Technologies system [24,34,61] or a Glas-col (Terre Haute, ID) system [62,63] between eight and ten weeks of age. Male and female C57BL/6 S100a8 KO, HET, and WT mice were infected between eight and sixteen weeks of age. For each aerosol infection, the retained lung dose was determined by euthanizing a cohort of four to twelve mice 24 hours after exposure, homogenizing the entire lungs in 5mL sterile phosphate buffered saline, and plating the entire homogenate onto OADC-supplemented 7H11 agar. After 3–4 weeks at 37°C, M. tuberculosis colony forming units were counted. Mice were infected with ~100 colony forming units in the first two experiments, and ~25 colony forming units in the subsequent eight experiments.
Quantification of TB-related Traits (Phenotyping)
Survival. IACUC protocols disallowed natural death as an endpoint. Therefore, as a proxy of survival, we used the day of euthanasia due to any single criterion: Severe weakness/lethargy; respiratory distress; or body condition score < 2 [64]. We confirmed morbidity was due to pulmonary TB by finding: (i) Large nodular, or severe diffuse lung lesions; (ii) histopathology confirmation of severe granulomatous lung infiltrates; (iii) growth of viable M. tuberculosis colonies from lung tissue; and (iv) absence of other diseases. Twenty-one M. tuberculosis infected Diversity Outbred mice were excluded due to co-morbidity that developed during the in-life portions.
Weight loss. Mice were weighed 1 to 3 days prior to M. tuberculosis aerosol infection, at least once per week during infection, and immediately before euthanasia. For each mouse, weight loss was calculated as the percent loss from peak body weight.
Lung granuloma necrosis. Immediately after euthanasia, lung lobes were removed and inflated and fixed in 10% neutral buffered formalin (5–10 mL per lobe), processed, and embedded in paraffin, sectioned at 5μm, and stained with hematoxylin and eosin with or without carbol fuschin for acid-fast bacilli at Tufts University, Cummings School of Veterinary Medicine, Core Histology Laboratory (North Grafton, MA). Hematoxylin and eosin-stained glass slides were magnified 400 times and digitally scanned by Aperio, LLC (Sausalito, CA) ScanScope scanners at 0.23 microns per pixel at The Ohio State University’s Comparative Pathology and Mouse Phenotyping Shared Resources Core resource (Columbus, OH) or by Aperio, LLC (Sausalito, CA) AT2 scanners at 0.23 microns per pixel at Vanderbilt University Medical Center’s Digital Histology Shared Resource (Nashville, TN). Lung granuloma necrosis was quantified in one lung lobe per mouse by our previously validated, deep learning image analysis method [65] and reported here as a ratio of granuloma necrosis per lung tissue section area.
M. tuberculosis lung burden. Immediately after euthanasia, 2 or 3 lung lobes were removed from each mouse and homogenized in sterile phosphate buffered saline (1mL per lobe), serially diluted, plated onto OADC-supplemented 7H11 agar, incubated at 37°C for 3–4 weeks, after which colonies were counted, and M. tuberculosis lung burden in the lungs was calculated as described [66].
Lung cytokines and chemokines. Lung homogenates were stored at -80°C until the experiment ended. Lung homogenates were then thawed overnight at 4°C, serially diluted and tested for CXCL5, CXCL2, CXCL1, tumor necrosis factor (TNF), matrix metalloproteinase 8 (MMP8), S100A8, interferon-gamma (IFN-γ), interleukin (IL)-12p40, I-L12p70, IL-10, and vascular endothelial growth factor (VEGF) by sandwich ELISA using antibody pairs and standards from R&D Systems (Minneapolis, MN), Invitrogen (Carlsbad, CA), eBioscience (San Diego, CA), or BD Biosciences (San Jose, CA, USA), per kit instructions. Lung homogenate ELISA results from five of the experiments using Diversity Outbred mice have been published and analyzed for biomarkers previously [34].
Phenotype correlation
We took the log of each phenotype after adding one (to ensure that zero was not converted to negative infinity) and regressed out the effect of the experimental batch. We then standardized the residuals and estimated the Pearson correlation between all pairs of phenotypes.
Gene expression
Gene expression profiling was performed on a subset of 117 Diversity Outbred mice representing the spectrum of observed phenotypes. Detailed selection description, additional analyses, and released datasets are described in [34,67]. Briefly, one lung lobe was homogenized in TRIzol, stored at -80°C, and RNA was extracted using Pure Link mini-kits (Life Technologies, Carlsbad, CA). Boston University’s Microarray and Sequencing Resource Core Facility (Boston, MA) confirmed quality and quantity were sufficient for microarray analyses. Mouse Gene 2.0 ST CEL files were normalized to produce gene-level expression values using the implementation of the Robust Multiarray Average (RMA) in the affy package (version 1.62.0) included in the Bioconductor software suite and an Entrez Gene-specific probeset mapping (24.0.0) from the Molecular and Behavioral Neuroscience Institute (Brainarray) at the University of Michigan. Array quality was assessed by computing Relative Log Expression (RLE) and Normalized Unscaled Standard Error (NUSE) using the affyPLM package (version 1.59.0). The CEL files were also normalized using Expression Console (build 1.4.1.46) and the default probesets defined by Affymetrix to assess array quality using an AUC metric computed from sets of negative and positive control probes; all samples used in this analysis had an AUC > 0.8. To remove microarray probes that intersected with Diversity Outbred SNPs, we intersected the Diversity Outbred founder strain SNPs [68] with the vendor-provided probes and removed probes containing SNPs. All microarray analyses were performed using the R environment for statistical computing (version 3.6.0). A related microarray dataset and other secondary analyses have been published elsewhere [34,67] and deposited in Gene Expression Omnibus (GEO), and assigned Series ID GSE179417.
Genotyping
We collected tail tips from each Diversity Outbred mouse and sent them to Neogen (Lincoln, NE) for genomic DNA isolation and genotyping. Neogen genotyped the mice on the Illumina GigaMUGA platform, which contains 143,259 markers [69]. Genotypes of S100a8 KO, HET, and WT C57BL/6 inbred mice were confirmed by polymerase chain reaction (TransnetYX, Cordova, TN).
Haplotype reconstruction and SNP imputation
We used 137,302 GigaMUGA marker positions located on the autosomes and chromosome X found at https://github.com/kbroman/MUGAarrays/blob/main/UWisc/gm_uwisc_v1.csv and the R package qtl2 to reconstruct the Diversity Outbred haplotypes using the founder and Diversity Outbred allele calls, and used the haplotype reconstructions to impute the founder SNPs onto the Diversity Outbred genomes [70].
Quantitative Trait Locus (QTL) mapping
We performed multiple rounds of genetic mapping to identify the least noisy dataset, ensure appropriate sample size for statistical power, and generate robust results. The final QTL mapping results include all available genotype and phenotype data from Diversity Outbred mice that survived M. tuberculosis infection for up to 250 days because this dataset produced the most statistically rigorous results, and minimized confounding effects of age-related comorbidities (e.g., lung tumors, lymphoma, uterine lesions). We used qtl2 [70] to perform linkage mapping using the founder haplotypes and association mapping using the imputed SNPs. We calculated the kinship between mice using the leave-one-chromosome-out method, which excludes the current chromosome in kinship calculations [71]. We standardized each phenotype and mapped with the Diversity Outbred outbreeding generation as an additive covariate and used the linear mixed-effects model with one kinship matrix per chromosome. We estimated the genome-wide significance thresholds by permuting the samples 1,000 times and performed a genome scan with each permutation. We retained the maximum log10 of the odds ratio (LOD) score from each permutation and estimated the genome-wide significance threshold of 7.6 from the 95th percentile of the empirical distribution of maximum LOD scores under permutation. We estimated the support interval around each peak using the 95% Bayesian Credible Interval.
For each peak with a LOD score above the genome-wide threshold > 7.6, we then searched for peaks associated with other traits that had LOD scores > 6 and confidence intervals that overlapped [72]. Our rationale was that the probability that a peak is biologically relevant, given that another trait has a co-located peak, is greater than the probability that a peak is significant with no prior evidence.
Candidate gene selection
Within each QTL interval, we imputed the founder SNPs onto the Diversity Outbred mouse genomes using qtl2 and performed association mapping. We selected the SNPs that were within a 1 LOD drop of the peak SNP in the QTL interval and filtered them to retain ones with missense, splice, or stop codon effects as annotated by the Sanger Mouse Genome Project [68]. We considered the genes in which these polymorphisms occurred as candidate causal genes for the associated trait(s).
Trait-related gene sets
Because causal variants within a QTL may exert their influence through mechanisms other than gene expression, identifying differentially expressed genes within the QTL may be insufficient for ranking causal genes. Here we took an alternative approach and ranked candidates in each QTL based on their predicted association with the mapped traits. To do this, we trained a Support Vector Machine (SVM) classifier to classify trait-related genes, and then used the trained SVM to score each positional candidate gene as trait-related or not-trait-related. We defined the training set of trait-related genes for the SVM as those genes that were highly correlated to the measured trait using the gene expression data described above. We calculated the Pearson correlation between the abundance of each transcript, and each physiological trait using rank Z normalized gene expression and traits. For each trait, we defined the training set of trait-related genes as the 500 genes with the largest magnitude Pearson correlation to the trait. We have made these gene lists available as a set of zipped text files in S1 File.
Support Vector Machine classifier training
We trained SVMs to classify genes in each gene list as trait-related using features derived from the Functional Network of Tissues in Mouse [41]. The nodes in this network are genes, and the edges between them are weights between 0 and 1 that predict the likelihood that each pair of genes is annotated to the same Gene Ontology (GO) term or KEGG pathway [39]. Values closer to one indicate more certainty that the genes are more likely to be annotated to the same GO term or KEGG pathway and thus functionally related. The weights were derived using Bayesian integration of data sets from numerous sources of data, including gene expression, protein-protein interaction data, and phenotype annotations [41]. We used the top edges of the mouse lung network downloaded on March 31, 2021, from http://fntm.princeton.edu.
Application of Support Vector Machine classifiers to identify genes functionally related to traits
We used SVMs to classify each positional candidate as trait-related or not-trait-related, as described previously [39,40]. Briefly, the expression-derived gene sets for each lung trait served as the positive labeled set of genes. We used the R package e1071 [73] to train SVMs to distinguish this set of genes from a balanced set of genes drawn randomly from the remaining genes in the lung network. The randomly selected genes were the unlabeled set. We performed this training 100 times, each time with a new set of random unlabeled genes. The SVMs were trained to distinguish positive labeled genes from unlabeled genes using the connection weights to the positive labeled genes. It is expected that positively labeled genes have strong connections to each other because they are functionally related. It is further expected that randomly drawn genes will be unrelated to the trait and to the positive labeled set and will thus have lower connection weights to the positive labeled genes. The SVM learns to distinguish these two groups of genes, and the resulting model can be used to classify genes that have not been seen before based on their connection weights to the positively labeled genes. We initialized each run by tuning the SVM over a series of cost parameters, starting with the sequence 1025 to 102 by factors of ten, and iteratively narrowing the range until we found a series of eight cost parameters that maximized accuracy. In running each SVM, we used a linear kernel and 10-fold cross-validation.
We calculated the area under the receiver operating characteristic curves (AUC) for each set of trait-related genes as follows. We defined labeled positives (LP) as positive labeled genes that were classified by the SVM as trait related. Unlabeled negatives (UN) were unlabeled genes that were classified by the SVM as not trait related. Unlabeled positives (UP) were unlabeled genes that were classified as trait-related, and labeled negatives (LN) were positive labeled genes that were classified as not trait-related. These terms are conceptually like true/false positive and true/false negative scores. However, because unlabeled genes may not be truly unrelated to the trait, we cannot call them true negatives. Instead, we call them unlabeled. We generated ROC curves using the Unlabeled Predicted Positive Rate (UPPR = UP/(UP+UN)), which is akin to the false positive rate, and the labeled positive rate (LPR = LP/(LP+UN)), which is akin to the false negative rate, along a series of SVM scores from the minimum to the maximum. We then calculated the average AUC across all 100 SVMs.
Positional candidate scoring
After training SVMs for each trait, we scored all positional candidate genes in each QTL, defined as the minimum to the maximum position across a set of overlapping QTLs. Each candidate gene received one score for each trait that mapped to that location. To compare scores across traits, we used the UPPR for each gene at its calculated SVM score. The UPPR varies between 0 and 1, allowing us to compare scores for candidate genes across models. To visually compare across models, we used the -log10(UPPR) such that genes with small UPPR (highnconfidence) got large positive values. In contrast, the SVM scores cannot be used to compare across models because they are unbounded and vary from model to model. Within each pleiotropic QTL, each gene received a score from each trait that mapped to the QTL.
Mouse genome build and database versions
We used mouse genome build GRCm38 and SNPs and Indels from the Sanger Mouse Genomes Project, version 7, which uses Ensembl version 97 gene models. We also used and cross-referenced candidates with the Mouse Phenome Database GenomeMUSter, the Mouse Genome Informatics databases [74], and Ensembl’s Variant Effect Predictor tool.
Results
Survival and body weight changes
We infected Diversity Outbred mice by aerosol with ~100 M. tuberculosis colony forming units in the first two experiments (N = 167), and ~25 colony forming units in the subsequent eight experiments (N = 683). Infection reduced survival of Diversity Outbred mice compared to identically housed, age-, gender-, and generation-matched uninfected Diversity Outbred mice and to identically housed age- and gender-matched infected C57BL/6J inbred mice (Fig 1A). Approximately one-third of infected Diversity Outbred mice succumbed prior to 60 days post infection (Fig 1A) reflecting early mortality between 20–56 days that peaked at 30 days (Fig 1B). This supersusceptible fraction of the Diversity Outbred population has been named Progressors [24,33,65,67]. Morbidity in all Progressors was due to pulmonary TB, confirmed by histology, recovery of viable M. tuberculosis bacilli from the lungs and absence of other diseases. After the first mortality wave subsided, cumulative survival declined slowly to nearly 600 days with no discernable mortality waves (Fig 1A and 1B). This resistant fraction of Diversity Outbred mice has been named Controllers [24,33,65,67]. The eight founder strains survived at least 40 days of M. tuberculosis infection, without early mortality (Fig 1A).
Diversity Outbred (DO) mice (n = 680, brown solid line), and the eight inbred founder strains (n = 15 to 78, colored lines) were infected with M. tuberculosis strain Erdman bacilli by aerosol. Panel A: Cumulative survival. Approximately 30% of the DO population succumbed to pulmonary TB by 60 days post infection and approximately 70% succumbed to between 60- and 600-days post infection. Of the eight inbred founder strains, survival studies were completed for the C57BL/6J inbred strain; the other seven inbred founder strains were euthanized 40 days after M. tuberculosis infection. No inbred founder or non-infected (NI) DO mice (n = 53, dashed line) showed mortality in that period. Panel B: Daily mortality of M. tuberculosis-infected DO mice, highlighting the early wave of mortality that peaked between 25- and 35-days post infection.
All mice were weighed prior to infection, during infection, and immediately before euthanasia. Non-infected Diversity Outbred mice gained weight until they developed other diseases or were euthanized at the experiment end (S1A Fig). Progressors gained weight for 2–3 weeks, and then lost weight rapidly (S1B Fig). Controllers and C57BL/6J inbred mice gained weight for long and variable durations through about 250 days of infection, and then most but not all slowly lost weight (S1C and S1D Fig). We questioned whether pre-infection body weight influenced differential susceptibility. Retrospective analysis identified no significant differences in pre-infection body weights of non-infected Diversity Outbred mice compared to Progressors; and a significant difference of on average 1.75 gm less in mean body weights of non-infected Diversity Outbred mice and Progressors compared to Controllers (S2A Fig). Whether this is spurious or biologically relevant (i.e., heavier pre-infection body weight is partially protective) remains to be determined. S2B, S2C and S2D Fig show correlations between survival and eight clinical indicators. Seven indicators positively correlated with survival, including pre-infection body weight which was weakly positive. Only one indicator, the rate of body weight loss, negatively correlated with survival.
Lung histology and automated image analysis of granuloma necrosis
By eight weeks of M. tuberculosis infection, Diversity Outbred mice showed a spectrum of lung lesions visible at low magnification (Fig 2) with variation in severity (minimal to marked); distribution of cellular infiltrates (focal, multifocal, and diffuse); and granuloma content (e.g., necrotizing, and non-necrotizing, shown in S3 Fig at higher magnification). Additional lesions included fibrin thrombosis with alveolar septal necrosis; cavities with peripheral fibrosis; foamy and multinucleated macrophages with cholesterol clefts; formation of secondary lymphoid follicles; alveolar septal fibrosis; and intra- and extracellular M. tuberculosis bacilli described elsewhere [19,24,28,67,75–77]. Since granuloma necrosis is a key feature of pulmonary TB in humans, we focused on this feature of disease and used our automated image analysis method [65] to quantify granuloma necrosis in lung tissue sections. The ratio of necrotic to non-necrotic lung tissue was then input as a trait for genetic mapping.
Lung lobes were formalin-fixed, paraffin-embedded, sectioned, and stained with hematoxylin & eosin. Panel A: Lung section from a non-infected DO mouse. Panels B through I: Lung sections from M. tuberculosis-infected DO mice euthanized eight weeks after infection show a spectrum of lesions from mild to severe (upper left to bottom right); focal lesions (e.g., Panel B) to diffuse infiltration (Panel H); and include necrotizing (Panels C, E, F, I) and non-necrotizing granulomas (Panels B, D, G, H). Low magnification (15X).
Quantification of lung traits: M. tuberculosis burden and immune responses
We quantified M. tuberculosis lung burden by counting colonies from lung tissue homogenate and used the remaining lung homogenate to quantify cytokines and chemokines by ELISA [24,34]. Most lung traits were significantly greater in mice infected with M. tuberculosis compared to non-infected mice (Fig 3A), including neutrophil and monocyte/macrophage chemokines (CXCL1, CXCL2, CXCL5); mediators of innate immunity (S100A8, Tumor Necrosis Factor (TNF), interleukin (IL)-10, matrix metalloproteinase-8 (MMP8); mediators of acquired immunity (interferon-gamma (IFN-γ)) and M. tuberculosis burden. Pairwise Pearson correlation of all traits in infected mice showed that all the lung traits except IL10 and VEGF positively correlated with each other (Fig 3B). Like previous findings in a small study of Diversity Outbred mice [24], correlations were strongest between M. tuberculosis lung burden and mediators of acute neutrophilic inflammation, innate immunity, and extracellular matrix degradation: CXCL1, CXCL2, TNF, and MMP8 with a mean correlation of 0.75.
(A) Boxplots of phenotypes in non-infected (n = 77) and infected (n = 930) mice at euthanasia, which ranged from 5 to 553 days post-infection. Each box shows the median (central line) and the inter-quartile range in the box. The whiskers show 90% of the data and the dots beyond the whiskers show data points outside of the 90th percentile of the distribution. Significant differences at a Bonferroni corrected p-value of 0.004 are shown with an asterisk. (B) Pearson correlation of selected phenotypes with n = 930 infected mice. Each cell in the upper triangle contains an ellipse in which the color and shape show the strength of the correlation. Positive correlations are shown in blue and negative correlations are shown in red. Strong correlations are shown by narrow ellipses, and weak correlations are shown by round ellipses. The more circular the ellipse, the closer the correlation is to zero. Numerical values of the Pearson correlation are shown in the lower triangle and range from -0.14 to 0.84. Phenotypes are ordered by hierarchical clustering of the similarity in correlations. Absolute Pearson correlations greater than 0.105 are significant at a Bonferroni-corrected p-value of 0.0007 and are shown with an asterisk.
Overview of genetic mapping and gene prioritization within QTLs
We performed linkage mapping on each trait by regressing each on the additive founder allele dosage at each locus using the R package qtl2 [70]. We selected peaks with a permutation-derived significance threshold of 7.62 (pGW ≤ 0.05) and found eight peaks associated with multiple traits on chromosomes 1, 2, 3, 4, 16, and 17 (Table 1 and Fig 4). We observed that correlated traits colocalized to shared QTLs, and had similar patterns of allele effects, so we used a two-step procedure in which we recorded the confidence interval for peaks with a LOD ≥ 7.62 and then looked for peaks of colocalized traits with a LOD ≥ 6.0 (pGW ≤ 0.6). We reasoned that once we had found the first significant peak for one trait, the threshold for colocalized peaks with the same pattern of founder allele effects should be less, allowing refinement of the loci. Fig 5 shows a flow diagram of the types of input data for genetic mapping to identify QTLs, and the subsequent methods of gene prioritization.
Quantitative trait locus (QTL) mapping results from M. tuberculosis infected Diversity Outbred mice that were euthanized on or prior to 250 days post infection (if morbidity developed), and complete genotype and phenotypic trait data which also passed QA/QC (n = 853). The mouse genome, from chromosome 1 through X, is shown on the horizontal axis. Phenotypes are shown on the vertical axis. Each cell shows the logarithmic of the odds (LOD) score on one chromosome for the phenotype listed on the left, colored by the color scale. Higher LOD scores (oranges and yellows) meet or exceed statistical thresholds and show the significant genotype-phenotype associations. The phenotypes are hierarchically clustered based on the correlation between LOD curves, i.e., phenotypes with similar LOD curves are clustered next to each other.
Traits were mapped to identify positional candidate genes in QTLs. Gene expression data were analyzed for differential gene expression. The top 500 differentially expressed genes (DEG) were used to train SVMs to distinguish these trait-related genes from other genes in the genome using the FNTM mouse lung network. The fitted models were used to score positional candidates in each trait QTL. Positional candidates were then ranked as trait-related based on their functional scores.
Six QTLs with LOD scores > 7.62 (pGW < 0.05) and two with lower thresholds were identified by linkage mapping. Correlated traits with lower LOD thresholds (≥ 6) mapped to the same positions. The columns from left to right show: Lung trait(s), chromosome, LOD score of the peak, the position of the peak in megabases (Mb), lower and upper confidence intervals (C.I.) in Mb, and the width of the confidence interval in Mb. Five QTLs are new, and two QTLs are not new. Dots3 overlaps with the QTL tbs1, previously identified by crossing A/Sn and I/St inbred mouse strains. Dots7 and Dots8 overlap with sst5 and sst6, previously identified by crossing C3HeB/FeJ and C57BL/6J inbred mouse strains.
Chromosome 1: Diversity Outbred Tuberculosis Susceptibility locus 1 (Dots1)
Dots1 is a new QTL on chromosome 1 with a peak LOD >7.6 (pGW < 0.05) at 155.36 M and interval 154.25–156.71 Mb shared by two correlated traits, M. tuberculosis burden and CXCL1 (Table 1). Two correlated traits (CXCL2 and MMP8) that mapped to the same position had LOD thresholds ≥ 6.0 (pGW ≤ 0.6) (Fig 4). Notably, these four correlated traits shared patterns of founder allele effects (S5 Fig), suggesting this QTL contains an important mechanism of genetic regulation for neutrophil-mediated activities, extracellular matrix remodeling, and M. tuberculosis growth. To refine the locus, we calculated the first principal component of those four traits and plotted the LOD curve, which also peaked between 154–156 Mb (Fig 6A) and plotted the founder allele effects. A/J, C57BL/6J and WSB/EiJ alleles contributed to greater values of principal component 1 and CAST/EiJ alleles contributed to lesser values (Fig 6B). We next imputed the founder SNPs onto the Diversity Outbred genomes and performed association mapping in a 10 Mb region around the peak (Fig 6C). Interestingly, the SNPs with highest LOD scores were outside of the peak, and none of the SNPs with the highest LOD scores were missense, stop, or splice site SNPs. This suggested the SNPs in the confidence interval could regulate expression of nearby genes, including some of the 47 protein-coding genes in the interval (Fig 6D and S1 File). To find and prioritize trait-related gene candidates within Dots1, we used the trained SVM model to rank gene candidates based on the strength of their functional relationship in gene expression network modules (S2 File). Fam20b and Ncf2 ranked highest by functional scoring (Fig 6E). Table 2 shows known protein functions, MVAR genome annotations, allele effects, founder alleles containing SNPs, and predicted effects of missense SNPs in Fam20b and Ncf2 genes on protein functions.
Panel A: LOD curve for PC1 between 150 and 160 Mb on chromosome 1 with peak near 155.36 Mb. Panel B: Founder allele effects for PC1 in the same genomic interval. Each colored line is the best linear unbiased predictor for one of the founder alleles. Founder colors are shown in the upper left. Panel C: LOD score of the imputed SNPs in the same genomic interval. Each point represents the LOD score of one imputed SNP. Panel D: Genes in the confidence interval. Panel E: Functional scores for genes in the chromosome 1 QTL. Each dot represents a single gene. Its position on the x axis is its position within the QTL. Its position on the y axis is the functional -log10(UPPR) derived from the SVM. Points are colored based on is correspondence with the trait—green with CXCL1, orange with MMP8, and blue with M. tuberculosis burden. Fam20b and Ncf2 genes had the highest functional scores. QTL mapping results from n = 853 M. tuberculosis infected Diversity Outbred mice up to 250 days post infection.
Predicted effects of missense SNPs on proteins were identified using Ensembl’s Variant Effect Predictor and Sorting Intolerant from Tolerant (SIFT) databases. SIFT produces scores based on sequence homology and physicochemical similarity between amino acids encoded by SNPs. Normalized scores are based on the probability that the amino acid change is tolerated, with scores < 0.05 are predicted to be deleterious and all other scores are predicted ’tolerated’.
Chromosome 2: Diversity Outbred Tuberculosis Susceptibility locus 2 (Dots2)
Dots2 is a new QTL, not shared by correlated traits (Table 1), and has a peak LOD of 7.69 (pGW < 0.05) at 22.43 Mb that was associated with lung CXCL2 protein levels (Fig 4). Dots2 contains 19 protein coding genes (S1 File). Because this QTL was associated with only one trait, gene prioritization by functional scoring was not pursued.
Chromosome 3: Diversity Outbred Tuberculosis Susceptibility locus 3 (Dots3)
Dots3 is not a new QTL (Table 1) and overlaps with tbs1, a QTL previously identified by crossing A/Sn and I/St inbred mouse strains [78]. These strains are not founder strains of the Diversity Outbred population. Dots3 was identified by a single peak with a high LOD of 16.57 at 90.69 Mb and interval 90.52–92.02 (pGW < 108) associated with lung S100A8 (calgranulin A) protein levels (Figs 4 and 7A). CAST/EiJ alleles effects were high and PWK/PhJ allele effects were low (Fig 7B). SNPs with the highest LOD scores within the peak are shown (Fig 7C). The interval contains 12 protein coding genes (S1 File) including the S100a8 gene (Fig 7D), suggesting that genetic variants which affect S100a8 transcription regulate S100A8 (calgranulin A) protein levels in M. tuberculosis infection. Further, based on based on the strength of the functional relationship in gene expression network modules (S2 File) the trained SVMs identified S100a8 as the gene with the highest functional score in Dots3 (Fig 7E). Dots3 also contains the gene S100a9, which encodes S100A9 (calgranulin B), a protein binding partner of S100A8 (calgranulin A) required to form the heterodimer, calprotectin. Table 2 shows known protein functions, MVAR genome annotations, allele effects, founder alleles containing SNPs, and predicted effects of missense SNPs in S100a8 and S100a9 genes on protein functions.
Panel A: LOD score in the confidence interval from 90 to 92 Mb on chromosome 3. Panel B: Founder allele effects within the confidence interval. Panel C: LOD score of the imputed SNPs with the highest SNPs colored in red. Panel D: Genes in the same interval. The gene S100a8 is directly under the SNPs with the highest LOD scores. Panel E: Functional scores for genes in the chromosome 3 QTL. Each dot represents a single gene. Its position on the x axis is its position within the QTL. Its position on the y axis is the functional -log10(UPPR) derived from the SVM. The gene S100a8 had the highest functional score. QTL mapping results from n = 853 M. tuberculosis infected Diversity Outbred mice up to 250 days post infection.
Chromosome 4: Diversity Outbred Tuberculosis Susceptibility locus (Dots4)
Dots4 is a new QTL, not shared by correlated traits (Table 1), and has a peak LOD of 7.64 (pGW < 0.05) at 22.43 Mb and interval 22.18–23.79 Mb associated with lung S100A8 (calgranulin A) protein levels (Fig 4). Dots4 contains two protein coding genes (S1 File). Because this QTL was associated with only one trait and contained few protein coding genes, prioritization by functional scoring was not pursued.
Chromosome 16: Diversity Outbred Tuberculosis Susceptibility locus (Dots5)
Dots5 is a new QTL on chromosome 16, shared by three correlated traits: lung M. tuberculosis burden (LOD = 8.45, pGW ≤ 0.01), weight loss, and granuloma necrosis with a peak at 38.3 Mb and interval 33.28–43.28 Mb (Table 1 and Fig 4). We calculated the first principal component of these traits and plotted the LOD curve showing its peak (Fig 8A). The founder allele effects indicate that C57BL/6J alleles contribute to greater values of principal component 1 and that PWK/PhJ and NZO/HILtJ alleles contribute to lesser (Fig 8B). We imputed the founder SNPs onto the Diversity Outbred genomes and performed association mapping around the peak, showing the SNPs with the highest LOD scores (Fig 8C). The SNPs with high LOD scores (Fig 8D) were not missense, stop, or splice site SNPs in the 75 protein coding genes within the interval (S1 File). By prioritizing genes based on functional relationships in network modules, we identified Fstl1 and Itgb5 as functional candidates associated with weight loss, and Zbtb20 as a functional candidate associated with M. tuberculosis burden (Fig 8E and S2 File). Table 2 shows known protein functions, MVAR genome annotations, allele effects, founder alleles containing SNPs, and predicted effects of missense SNPs in Fstl1, Itgb5, and Zbtb20 genes on protein functions.
M. tuberculosis lung burden, weight loss, and granuloma necrosis map to a region on chromosome 16 near 38 Mb. Panel A: LOD profile for M. tuberculosis in the confidence interval. The genomic position on chromosome 16 is on the horizontal axis and the LOD score is on the vertical axis. Panel B: Founder allele effect in the confidence interval. The vertical axis shows the estimates effect of gaining one founder allele. Panel C shows the SNP LOD score for association mapping using imputed SNPs. Panel D: Genes in the confidence interval. Panel E: Functional scores for genes in the chromosome 16 QTL. Each dot represents a single gene. Its position on the x axis is its position within the QTL. Its position on the y axis is the functional -log10(UPPR) derived from the SVM. Points are colored based corresponds to the trait—green with weight loss and blue with M. tuberculosis burden. Fstl1 was the top ranked gene overall followed by Itgb5 and Zbtb20. QTL mapping results from n = 853 M. tuberculosis infected Diversity Outbred mice up to 250 days post infection.
Chromosome 16: Diversity Outbred Tuberculosis Susceptibility locus (Dots6)
Dots6 is a new QTL on chromosome 16 and shared by two correlated traits: M. tuberculosis burden and CXCL5 with a peak at 52.23 Mb and interval 37.97–57.67 Mb (Table 1 and Fig 4). The interval contains 101 protein coding genes (S1 File). Because the LOD score was less than LOD threshold 7.64 for significance (pGW < 0.05), prioritization by functional scoring was not pursued.
Chromosome 17: Diversity Outbred Tuberculosis Susceptibility locus (Dots7)
Dots7 on chromosome 17 is not new, and overlaps with sst5 and sst6, QTLs that were previously identified by crossing C3HeB/FeJ and C57BL/6J inbred mouse strains [79]. Dots7 is associated with M. tuberculosis burden and the LOD peaks ~20 Mb (Table 1 and Fig 4). The interval contains 198 protein coding genes (S1 File). Because the LOD score was less than the LOD threshold 7.64 for significance (pGW < 0.05), prioritization by functional scoring was not pursued.
Chromosome 17: Diversity Outbred Tuberculosis Susceptibility locus (Dots8)
Dots8 is not a new QTL and also overlaps with sst5 and sst6, two QTLs that were previously identified by crossing C3HeB/FeJ and C57BL/6J inbred mouse strains [79]. Five traits: lung granuloma necrosis (“Necr Ratio”), weight loss, MMP8, CXCL1, and IL-10 mapped to Dots8. Lung granuloma necrosis had the highest LOD score (LOD = 8.12, pGW ≤ 0.02) at 35.02 Mb and interval 33.94–41.06 Mb Of those five traits, four positively correlated with each other (Fig 3) and had similar patterns of allele effects, and one, IL-10 had weak correlation and different pattern of founder allele effects.
We calculated the first principal component of the correlated traits and again performed QTL mapping. Principal component 1 mapped to a wide interval 30–50 Mb with a peak near 38 Mb (Fig 9A). The founder allele effects indicate PWK/PhJ alleles contribute to greater trait values, and NZO/HILtJ and NOD/ShiLtJ alleles contribute to lesser values (Fig 9B). We expected to find polymorphisms in the proximal peak of Dots8 at 34–38 Mb because it contains the mouse histocompatibility-2 (H-2; or Major Histocompatibility Complex-II MHCII). This locus contains genes known to regulate innate and adaptive immunity and is known to be highly polymorphic. Indeed, the highest SNP association mapping LOD scores were over the mouse H-2 locus, located 36–38 Mb (Fig 9C), and there were 27 SNPs with protein-coding or splice site variation which occurred in 15 genes (S3 File). Among these were several histocompatibility genes (H2-M1, H2-M5, H2-M9, H2-M11) and several tripartite motif (TRIM) family genes (Trim10, Trim26, Trim31, Trim40).
Panel A: LOD score for PC1 of lung granuloma necrosis ratio, M. tuberculosis burden, MMP8, CXCL1, and weight loss in the interval where the phenotypes map. Panel B: Founder allele effects for the two peaks. Panel C: LOD score of the imputed SNPs in the interval, with the highest scoring SNPs colored in red. Panel D: Functional scores for genes in the chromosome 17 QTL. Each dot represents a single gene. Its position on the x axis is its position within the QTL. Its position on the y axis is the functional -log10(UPPR) derived from the SVM. Points are colored based on correspondence to the traits—blue with CXCL1 and green with granuloma necrosis. Vegfa was the top ranked gene overall followed by Ier3, Ddr1, and Zpf318. QTL mapping results from n = 853 M. tuberculosis infected Diversity Outbred mice up to 250 days post infection.
The genes within the broad interval of Dots8 are difficult to show in the figure and are provided in S1 File. There were 361 protein-coding genes within the 30–50 Mb locus. We prioritized positional candidate genes based on their functional relationships in network modules to the correlated traits (S2 File). This identified candidates Ddr1, Ier3, and Vegfa associated with CXCL1; and Zfp318 associated with granuloma necrosis (Fig 9D). Table 2 shows known protein functions, MVAR genome annotations, allele effects, founder alleles containing SNPs, and predicted effects of missense SNPs in Ddr1, Ier3, Vegfa, and Zfp318 genes on protein functions.
Methodological and gene candidate validation
We performed three validation types shown in S6 Fig. This included (i) survival analysis of Diversity Outbred mice carrying PWK/PhJ alleles at the H-2 locus in Dots8 on chromosome 17; (ii) quantification of S100A8 protein levels in lungs of M. tuberculosis infected PWK/PhJ and CAST/EiJ inbred founder strains; and (iii) infection of S100a8 C57BL/6 gene deficient mice. Notably, infected Diversity Outbred mice carrying at least one copy of the PWK/PhJ allele at the mouse H-2 locus had shorter survival than mice carrying other alleles at the H-2 locus (S6A Fig).
The Dots3 QTL on chromosome 3, associated with S100A8, had the highest LOD score, narrowest interval, distinct founder allele effects, and the gene candidate was the S100a8 gene itself. For these reasons, the locus was selected for additional study. To help validate the method and ensure the founder allele effects generated by QTL mapping were biologically relevant, S100A8 was measured in lungs of the two founder inbred strains with the highest and lowest allele effects at the locus (Fig 7). As predicted by the allele effects, lungs from M. tuberculosis infected CAST/EiJ inbred mice contained significantly more S100A8 protein than lungs from M. tuberculosis infected PWK/PhJ inbred mice (S6B Fig). To evaluate in vivo effects of S100a8 genetic deficiency, we obtained C57BL/6 breeding pairs to generate knockout, heterozygous, and wild-type mice. Littermates with null mutation (knockout), heterozygous, and wild-type C57BL/6 S100a8 alleles were infected with M. tuberculosis for up to 50 days. The absence of one or both copies of the C57BL/6 S100a8 allele had minimal impact on M. tuberculosis lung burden (S6C Fig). The lung granulomas of knockout, heterozygous, and wild type C57BL/6 mice were similar and non-necrotizing, composed of small dense aggregates of lymphocytes and macrophages, without neutrophils and lacking interalveolar pyknotic debris (S6D, S6E and S6F Fig). This indicates that a lack or reduction of S100A8 has no major impact on early M. tuberculosis growth or granuloma formation in C57BL/6 inbred mice. These findings are unlikely to generalize to all genetic backgrounds, however, especially those genetic backgrounds predisposed to early accumulation of neutrophils and necrotizing granulomas with relatively high S100A8 levels, like the Diversity Outbred mice that are highly susceptible to M. tuberculosis [24,34].
Discussion
TB remains a major public health concern in the United States and across the globe, with an estimated 2 billion people infected with M. tuberculosis; 8–9 million patients diagnosed each year, and 1–1.5 million deaths annually [2]. Fortunately, most humans (~90%) are highly resistant to M. tuberculosis and clear or control infection [11,80]. In susceptible adults, active pulmonary TB develops a few years following exposure and tends to occur in young to middle-aged adults in the prime years of their lives [81]. The disease is usually restricted to the lungs and is characterized by granuloma necrosis and cavitation, neutrophilic infiltration, and cachexia [82,83]. The genetic basis of pulmonary TB is complex and not attributable to single-gene defects that cause severe immune deficiency (i.e., Mendelian susceptibility to mycobacterial disease does not explain pulmonary TB) [15,84–89]. Although genome-wide association studies have identified loci, gene candidates, and SNPs associated with increased or decreased odds ratios for pulmonary TB, only a few (e.g., Ipr1/SP110b and HLA variants/I-A Major Histocompatibility genes) have been validated [16,90–96]. This has led investigators to seek alternative experimental mouse models such as Diversity Outbred mice and Collaborative Cross recombinant inbred strains to examine effects of genetics on host responses to M. tuberculosis [24–26,47].
An advantage of the Diversity Outbred mouse population is that infection with M. tuberculosis induces phenotypes that are rare in common laboratory inbred strains of mice [15–17]. Further, a growing body of evidence shows similarities in M. tuberculosis-infected Diversity Outbred mice and humans in biomarkers, gene expression signatures, and BCG vaccination [28,33,34,47,97]. The phenotypic similarities suggest that humans and Diversity Outbred mice may share underlying genetic pathways of immunity and disease. And, because SNP variants in the Diversity Outbred mouse genomes are dense, with balanced allele frequencies, any gene that plays a role in disease is theoretically detectable [98]. This eliminates a problem common to human genetic studies where under-represented alleles cannot be confidently associated with disease phenotypes because they are low-frequency genetic events.
We performed genetic mapping in M. tuberculosis-infected Diversity Outbred mice and used orthogonal methods to rationally select candidate genes. We first used DOQTL mapping, which relies entirely on phenotypic and genetic variation, to find eight QTLs on six different chromosomes named Dots1 through Dots8. To refine loci, we then subjected the QTLs on chromosomes 1, 16, and 17 (Dots1, Dots5, and Dots 8) by mapping the first principal component of the correlated traits with similar patterns of allele effects that colocalized to the same interval. Finally, we applied a gene-based machine-learning SVM to identify and rank gene candidates based on functional scores. The sequential methods narrowed the candidate gene list to eleven polymorphic, protein coding genes. Finally, the SNPs were critically examined using publicly available databases to find four candidates (S100a8, Itgb5, Fstl1, Zfp318) where missense SNPs are predicted to have deleterious effects on protein function.
All eleven candidates have roles in infection, inflammation, cell migration, extracellular matrix remodeling, or intracellular signaling. Of those, only one (mouse Ncf2 in Dots1 on chromosome 1) has a human homologue where a single SNP (G nucleotide in human NCF2 rs10911362) may provide a protective effect because this SNP was associated with a lower odds ratio for pulmonary TB in humans [99]. Absence of Ncf2 in C57BL/6 inbred mice temporarily impairs resistance to M. tuberculosis infection by abrogating superoxide production, but the defect does not affect overall survival due to compensation by T cell mediated immunity [100]. Experimental validation of mouse Ncf2 and human NCF2 polymorphisms remains to be confirmed.
We identified Fam20b in Dots1 QTL on chromosome 1. The gene encodes a xylosylkinase that functions in glycosaminoglycan synthesis to produce extracellular matrix components in tissues. Deficiencies are embryologically lethal or cause cranioskeletal malformations [52,53,101,102]. A functionally homologous enzyme phosphorylates cadherins [103] which regulate immune cell migration [104] by interacting with extracellular matrix components, and since cell migration is required to form mycobacterial granulomas, Fam20b polymorphisms may alter host susceptibility to M. tuberculosis by changing extracellular matrix.
We identified S100a8 and S100a9 genes in Dots3 QTL on chromosome 3, which encode S100A8 (calgranulin A) and S100A9 (calgranulin B). The proteins form monomers, homodimers, heterodimers, and multimers in inflammation, host defense, and nociception [105–108]. Some forms activate Toll-Like receptor 4; some activate the receptor for advanced glycation end-products (RAGE); and some sequester calcium, zinc, and manganese metal ions [105,109,110]. In pulmonary TB, S100A9 contributes to neutrophil localization to granulomas, and both S100A8 and S100A9 are biomarkers of TB-related lung damage [34,47,97,110–112]. Our results showing S100a8 deficiency does not impact on M. tuberculosis lung burden, or granulomas, during early infection of C57BL/6 inbred mouse strain aligns with findings by Scott et. al. [97]. More specifically, the authors showed that S100A9 was elevated during chronic M. tuberculosis infection of C57BL/6 mice where it promotes neutrophil recruitment by CD11b, and C57BL/6 mice lacking S100A9 had lower M. tuberculosis CFU by 100 days post infection [97]. Whether these findings and mechanisms generalize to all genetic backgrounds and to all forms of S100A8 and S100A9 remain open questions, as S100A8 has anti-inflammatory functions [113] that appear independent from S100A9.
We identified Itgb5 in Dots5 QTL on chromosome 16 as a gene candidate. Itgb5 encodes the beta 5 (β5) integrin subunit which dimerizes with the alpha v subunit to mediate cell adhesion and signaling by binding to fibronectin and vitronectin [49]. Notably, the β5 subunit is on the surface of cancer cells, and normal epithelial cells and activated endothelial cells but not on lymphoid or myeloid cells [49,114–120]. To our knowledge, neither the mouse nor human gene, nor subunit β5, nor the αvβ5 integrin heterodimer have been investigated in pulmonary TB.
We identified Fstl1 in Dots5 QTL on chromosome 16. The primary transcript encodes microRNA (miR)-198. The product is a secreted glycoprotein, FSTL1 with activities in angiogenesis, cell proliferation, differentiation, embryogenesis, metastasis, and wound healing; specifically reducing inflammation and fibrosis in cardiovascular disease [51,121–124]. Notably, Fstl1 affects survival of M. tuberculosis-infected macrophages [43,44,125]. Given the central roles of macrophages, inflammatory mediators, and fibrosis in M. tuberculosis infection, understanding how Fstl1 polymorphisms and FSTL1 function in vivo may inform TB pathogenesis, and host-directed therapy.
We identified Zbtb20 in Dots5 QTL on chromosome 16. The gene encodes a transcriptional repressor involved in glucose homeostasis; growth; hematopoiesis; innate immunity; neurogenesis; and B cell development and long-term survival of plasma cells [54–57,126–130]. Natural mutations occur in humans with Primrose Syndrome, although immune deficiencies are not reported [131]. To our knowledge, there are no studies on Zbtb20 and M. tuberculosis infection or pulmonary TB. However, in Listeria monocytogenes infection, Zbtb20-deficiency improved CD8 T cell memory functions due to efficient use of diverse fuel sources [57]. Whether the same is true in pulmonary TB is unknown.
We identified Ddr1 in Dots8 QTL on chromosome 17. Ddr1 encodes for the discoidin domain receptor 1 (DDR1), which interacts with collagen [50]. Initial studies suggested DDR1 function was restricted to epithelial cells; however, recent work shows expression on solid tumors, metastatic cells, and mouse histiocytic cancer cell lines, J774 and Raw264.7 [132–138]. DDR1 has additional roles in demyelination, fibrosis, vitiligo, and wound healing, and it is also a promising target for anti-fibrotic therapy [139–143]. Whether Ddr1 (mouse) or DDR1 (human) gene polymorphisms contribute to pulmonary TB, or whether it could be a target for anti-fibrotic therapy in TB are areas open for investigation.
We identified the immediate early response gene, Ier3 in Dots8 QTL on chromosome 17. Ier3 transcription is triggered by cytokines, hormones, DNA damage, and infections. The protein, IER3, regulates apoptosis, DNA repair, differentiation, and proliferation by interfering with NF-κB, MAPK/ERK and PI3K/Akt signaling pathways [58–60,144–146]. Mice lacking Ier3 are more susceptible to Leishmania [147], an intracellular pathogen that shares some similar immune responses profiles with those induced by M. tuberculosis but we did not find studies showing that mutated Ier3 also increases susceptibility to M. tuberculosis. One in vitro study of human macrophages, however, had high levels of IER3 mRNA following infection with a hypervirulent strain of M. tuberculosis [148] indicating the transcriptional pathway is triggered.
We identified Vegfa in Dots8 QTL on chromosome 17. Mouse Vegfa and human VEGFA, encode for a heparin-binding protein and essential growth factor that induces proliferation, migration, and permeability changes in vascular endothelial cells by binding VEGFR1 and VEGFR2 [149–152]. Roles for VEGF in pathogenesis and diagnostics for extrapulmonary TB, cavitary TB, and active TB have been published [153–155]. Myeloid-specific gene deletion of Vegfa extended survival of C57BL/6J inbred mice infected with M. tuberculosis [156], highly noteworthy because very few gene deletions improve survival. Whether Vegfa or VEGFA polymorphisms have the same effect is unknown.
Lastly, we identified Zfp318 in Dots8 QTL on chromosome 17. The gene encodes the transcription factor, zinc finger protein 318 and it is expressed in testes, hematopoietic, and lymph nodes [157]. In B cells, the protein represses transcription required for class switching, helping to maintain B cell anergy and prevent autoimmunity [158–160]. Database and literature searches identified no publications on mouse Zpf318 or human ZPF318 in infectious diseases.
When we compared genetic mapping results from Diversity Outbred mice to results from colleagues using Collaborative Cross inbred strains [26], QTLs and gene candidates did not overlap although we measured a few of the same traits by standard laboratory methods (e.g., body weight changes, lung M. tuberculosis burden, and lung CXCL1 by immunoassays). This suggests that phenotype-genotype relationships in the Collaborative Cross strains may be fundamentally different than in the Diversity Outbred mouse population, possibly due to heterozygosity in the Diversity Outbred population which is absent from Collaborative Cross inbred strains. Other reasons could be differences in routes of infection that change the host cell types first encountering M. tuberculosis bacilli which alters antigen presentation, T-cell, and B-cell priming. Here, we modeled natural aerosol exposure by delivering a low dose of M. tuberculosis bacilli to the lungs of Diversity Outbred mice in nebulizer-delivered aerosol mist, and then focused on quantification of lung disease. In contrast, Smith et al [26] took a different approach by using intravenous infection with 1x106 bacilli to take advantage of their rich library of transposon mutants, allowing detailed assessment of pathogen-associated QTLs. As the intravenous route of infection favors rapid induction of acquired immunity by delivering bacilli directly to lymphoid organs (i.e., spleen, thoracic, and abdominal lymph nodes by portal and systemic circulation), this approach maximized identification of unique Host-Interacting-with Pathogen QTLs and resulted in a prioritized list of candidate genes involved in immunity.
Overall, by using a systems genetics approach focused on the lungs, we identified multiple new and existing QTLs, and 11 candidate genes. Of those, gene products for five (Ncf2, Fstl1, Zbtb20 Vegfa, Zfp318) have known roles in recruitment, activation, or regulation of effector functions of immune cells (e.g., neutrophils, monocytes, macrophages and CD8 T cells). The gene products for three candidates (Fam20b, Itgb5, Ddr1) have known roles in epithelial cell, endothelial cell, and (possibly) macrophage adhesion to extracellular matrix glycoproteins or participate in remodeling of extracellular matrix. The gene products for two candidates (S100a8 and S100a9) have complex and context-dependent roles in innate immune response signaling and in host defenses. Finally, the gene product of one candidate (Ier3) controls early stress responses of cells, including cell survival and death pathways. Ten of the eleven candidates have annotated polymorphisms; six have missense SNPs in protein coding regions; and the SNPs in four candidates (S100a8, Itgb5, Fstl1, and Zfp318) are predicted to have deleterious consequences on protein functions. Together, these results yield a concise list of candidates that may be major regulators of host necrotizing and inflammatory responses during M. tuberculosis infection and pulmonary TB disease progression. Future studies will focus on testing effects of these gene candidates and polymorphisms in vivo and identification of pathogenic molecular and cellular mechanisms.
Supporting information
S1 Fig. Mouse body weight following a low dose of aerosolized M. tuberculosis strain Erdman.
Mice were infected with a low dose of M. tuberculosis strain Erdman by aerosol and infection progressed unmanipulated until mice were euthanized due to IACUC-approved morbidity criteria. Panel A: Body weight changes of identically housed, age-, gender-, and generation-matched non-infected Diversity Outbred (DO) mice (n = 49) compared to baseline. Panels B, C, and D: Body weight changes of Progressor DO mice (n = 195); Controller DO mice (n = 145); and C57BL/6J inbred founder strain mice that succumbed to pulmonary TB (n = 39), are shown over time compared to pre-infection baseline. All mice were weighed 1 to 3 days prior to M. tuberculosis infection, at least twice per week during infection, and immediately before euthanasia. Each line is the body weight expressed as a percent change compared to initial pre-infection body weight.
https://doi.org/10.1371/journal.ppat.1011915.s001
(TIF)
S2 Fig. Clinical correlates of survival due to pulmonary TB in Diversity Outbred (DO) mice following exposure to a low dose of aerosolized M. tuberculosis strain Erdman.
Age-, gender-, and generation-matched DO mice were assigned to cages at random, and infected (or not infected) with a low dose of M. tuberculosis strain Erdman by aerosol exposure. All mice were initially weighed 1–3 days prior to infection, at least twice per week during infection, and immediately before euthanasia. Panel A: shows retrospective analysis of pre-infection body weights of Non-infected mice (n = 76) compared to pre-infection body weights of Progressors (n = 298) and pre-infection body weights of Controllers (n = 195), shown as box-and-whisker plots with the line at the mean for each group, and whiskers at the minimum and maximum. Data were analyzed by 1-way ANOVA with Tukey’s multiple comparisons test ***p<0.001; ****p<0.0001. Panels B, C, D: Infection progressed unmanipulated until mice were euthanized due to IACUC-approved morbidity criteria. Panel B: The rate of weight loss (gm/day) and duration of body weight (BW) loss in days negatively correlate with each other. Panel C: Duration of BW loss was strongly, positively, and linearly correlated with survival by Spearman correlation analysis (r = 0.848 with dashed lines indicating the 95% confidence interval, 0.8204 to 0.8717, p<0.0001). Panel D: Correlation matrix to show how survival and eight clinical indicators correlate with each other. Only correlations with p-values <0.00001 are shown on the matrix. Cells marked by an “X” were not significantly correlated.
https://doi.org/10.1371/journal.ppat.1011915.s002
(TIF)
S3 Fig. Examples of necrotizing and non-necrotizing lesions in lungs of M. tuberculosis infected Diversity Outbred (DO) mice by light microscopy at high magnification.
Lung lobes were formalin-fixed, paraffin-embedded, sectioned, and stained with carbol fuschin and counterstained with hematoxylin & eosin. Panels A and B: High magnification images of necrotizing lung lesions. One example contains abundant pyknotic nuclear debris (A) and one example contains abundant fibrin, eosinophilic cellular debris, and less nuclear debris (B). Panels C and D: High magnification images of non-necrotizing lung lesions. Both examples contain mostly viable cells, including macrophages, foamy macrophages, and foci of lymphocytes (400X).
https://doi.org/10.1371/journal.ppat.1011915.s003
(TIF)
S4 Fig. Similar patterns of founder allele effects for four traits that map to distal chromosome 1 QTL.
The pattern of founder allele effects of four correlated traits all with LOD > 6.0 on chromosome 1 at 155.36 Mb is similar. Panel A: Founder allele effects for CXCL1. Panel B: Founder allele effects for CXCL2. Panel C: Founder allele effects for MMP8. Panel D: Founder allele effects for M. tuberculosis burden. Founder strain names are on the horizontal axis and the standardized allele effect are on the vertical axis. Allele effects from QTL mapping results from n = 853 M. tuberculosis infected Diversity Outbred mice up to 250 days post infection.
https://doi.org/10.1371/journal.ppat.1011915.s004
(TIF)
S5 Fig. Receiver operator characteristic (ROC) curves for SVM training on traits used in gene prioritization.
Each panel shows the true positive rate of the trained SVM as a function of the false positive rate for each trait. The area under the curve (AUC) is noted for each panel.
https://doi.org/10.1371/journal.ppat.1011915.s005
(TIF)
S6 Fig. Validation of QTL mapping results.
Panel A: M. tuberculosis infected Diversity Outbred (DO) mice with one or more PWK/PhJ alleles (red) at the H-2 locus on chromosome 17 in Dots8 (near 36 Mb) have significantly reduced survival compared to DO mice carrying other alleles (grey) (p = 0.00075, Cox-PH test). Days of survival are shown on the horizontal axis and the proportion of mice surviving is shown on the vertical axis. Panel B: Lungs from M. tuberculosis infected CAST/EiJ inbred mice contain significantly more S100A8 protein (calgranulin A) than lungs of PWK/PhJ inbred mice at the time points indicated on the horizontal axis. PWK/PhJ (red) and CAST/EiJ (green) inbred founder strains with 4–6 mice per strain per time point, analyzed by Mann-Whitney t-tests within each time point, *p<0.05. Panel C: M. tuberculosis infected S100a8 knockout (KO), heterozygotes (HET), wild-type (WT) C57BL/6 inbred mice were euthanized at the time points indicated, and M. tuberculosis lung burden assessed by CFUs with total combined 15–22 mice per genotype per time point from 2 independent experiments, shown as average and standard error of the mean. No significant (ns) differences were identified within each time point by mixed effects ANOVA with Tukey’s post-test (p<0.05). Panel D, E, F: Representative examples of lymphohistiocytic non-necrotizing granulomas day 50 post infection from S100a8 knockout (D), heterozygous (E), and wild type (F) C57BL/6 inbred mice magnified 400x normal.
https://doi.org/10.1371/journal.ppat.1011915.s006
(TIF)
S1 File. This file is an Excel workbook containing worksheets that list all protein-coding genes in each QTL with the functional candidates highlighted.
https://doi.org/10.1371/journal.ppat.1011915.s007
(XLSX)
S2 File. This is an Excel file that lists the top ten functional candidates for each trait in each QTL.
https://doi.org/10.1371/journal.ppat.1011915.s008
(XLSX)
S3 File. This is an Excel file that lists genes within chromosome 17 QTL.
https://doi.org/10.1371/journal.ppat.1011915.s009
(XLSX)
Acknowledgments
We thank Ms. Julie Tzipori, Mr. Curtis Rich, Mr. Donald Girouard, and Dr. Sam Telford III for study support at the New England Regional Biosafety Laboratory at Tufts University Cummings School of Veterinary Medicine, North Grafton, MA. Ms. Frances Brown, Ms. Linda Wrijil, Ms. Sarah Ducat, Ms. Gina Scarglia, Ms. Dian Taylor provided histology services at Tufts University’s Cummings School of Veterinary Medicine Comparative Pathology and Genomics Shared Resource, directed by Dr. Amanda Martinot and Dr. Heather Gardner. The Digital Histology Shared Resource at Vanderbilt University Medical Center performed whole slide imaging. The following reagents were obtained through BEI Resources: ESAT-6, Recombinant Protein Reference Standard, NR-49424; CFP-10, Recombinant Protein Reference Standard, NR49425; Plasmid pMRLB.7 Containing Gene Rv3875 (Protein ESAT-6) from Mycobacterium tuberculosis, NR-50170; Plasmid pMRLB.46 Containing Gene Rv3874 (Protein Cfp10) from Mycobacterium tuberculosis, NR-13297; Mycobacterium tuberculosis, Strain H37Rv, Culture Filtrate Proteins, NR-14825; and Mycobacterium tuberculosis, Strain H37Rv, Cell Wall Fraction, NR-14828. All microarray protocols were conducted by the Boston University Microarray and Sequencing Resource (BUMSR), and we thank Mr. Eduard Drizik of the BUMSR for initial microarray analyses. The following individuals are thanked for their technical expertise: Ms. Victoria Mello at Tufts Cummings School of Veterinary Medicine, North Grafton, MA; Mr. Austin Hossfeld at The Ohio State University, Columbus, OH; and Dr. Joanne Turner at Texas Biomedical Research Institute, San Antonio, TX.
Boston University Microarray and Sequencing Resource received support from the National Institutes of Health UL1TR001430. The New England Regional Biosafety Laboratory at Tufts University’s Cummings School of Veterinary Medicine received support from the National Institutes of Health UC6A1066843. Tufts University’s Cummings School of Veterinary Medicine Comparative Pathology and Genomics Shared Resource received support from the Massachusetts Life Sciences Center.
References
- 1. WHO. 14.9 million excess deaths associated with the COVID-19 pandemic in 2020 and 2021 2022 [updated 5/5/2022. Available from: https://www.who.int/news/item/05-05-2022-14.9-million-excess-deaths-were-associated-with-the-covid-19-pandemic-in-2020-and-2021.
- 2.
Organization WH. Global tuberculosis report 2021. Geneva, Switzerland: World Health Organization; 2021.
- 3. WHO. Global Tuberculosis Report: Multidrug-/ rifampicin-resistant TB (MDR/RR-TB):Update 2017. 2017.
- 4. WHO. Global Tuberculosis Report. 2013.
- 5. Soleimanpour S, Yaghoubi A, Sadat Seddighinia F, Rezaee SAR. A century of attempts to develop an effective tuberculosis vaccine: Why they failed? International immunopharmacology. 2022;109:108791. pmid:35487086
- 6. Heidary M, Shirani M, Moradi M, Goudarzi M, Pouriran R, Rezaeian T, et al. Tuberculosis challenges: Resistance, co-infection, diagnosis, and treatment. Eur J Microbiol Immunol (Bp). 2022. pmid:35420996
- 7. Kassa GM, Tadesse A, Gelaw YA, Alemayehu TT, Tsegaye AT, Tamirat KS, et al. Predictors of mortality among multidrug-resistant tuberculosis patients in central Ethiopia: a retrospective follow-up study. Epidemiol Infect. 2020;148:e258. pmid:33054897
- 8. Tiemersma EW, van der Werf MJ, Borgdorff MW, Williams BG, Nagelkerke NJ. Natural history of tuberculosis: duration and fatality of untreated pulmonary tuberculosis in HIV negative patients: a systematic review. PLoS One. 2011;6(4):e17601. pmid:21483732
- 9. Vilaplana C, Cardona PJ. The lack of a big picture in tuberculosis: the clinical point of view, the problems of experimental modeling and immunomodulation. The factors we should consider when designing novel treatment strategies. Frontiers in microbiology. 2014;5:55. pmid:24592258
- 10. Achkar JM, Jenny-Avital ER. Incipient and subclinical tuberculosis: defining early disease states in the context of host immune response. J Infect Dis. 2011;204 Suppl 4:S1179–86. pmid:21996700
- 11. Verrall AJ, Netea MG, Alisjahbana B, Hill PC, van Crevel R. Early clearance of Mycobacterium tuberculosis: a new frontier in prevention. Immunology. 2014;141(4):506–13. pmid:24754048
- 12. Flynn JL, Chan J. Immunology of tuberculosis. Annu Rev Immunol. 2001;19:93–129. pmid:11244032
- 13. Hunter RL, Jagannath C, Actor JK. Pathology of postprimary tuberculosis in humans and mice: contradiction of long-held beliefs. Tuberculosis (Edinb). 2007;87(4):267–78. pmid:17369095
- 14. Azad AK, Sadee W, Schlesinger LS. Innate immune gene polymorphisms in tuberculosis. Infect Immun. 2012;80(10):3343–59. pmid:22825450
- 15. Gibbs KD, Schott BH, Ko DC. The Awesome Power of Human Genetics of Infectious Disease. Annu Rev Genet. 2022. pmid:35697043
- 16. Medhasi S, Chantratita N. Human Leukocyte Antigen (HLA) System: Genetics and Association with Bacterial and Viral Infections. J Immunol Res. 2022;2022:9710376. pmid:35664353
- 17. Varshney D, Singh S, Sinha E, Mohanty KK, Kumar S, Kumar Barik S, et al. Systematic review and meta-analysis of human Toll-like receptors genetic polymorphisms for susceptibility to tuberculosis infection. Cytokine. 2022;152:155791. pmid:35158257
- 18. Williams A, Orme IM. Animal Models of Tuberculosis: An Overview. Microbiol Spectr. 2016;4(4). pmid:27726810
- 19. Kramnik I, Beamer G. Mouse models of human TB pathology: roles in the analysis of necrosis and the development of host-directed therapies. Semin Immunopathol. 2016;38(2):221–37. pmid:26542392
- 20. Clark S, Hall Y, Williams A. Animal models of tuberculosis: Guinea pigs. Cold Spring Harb Perspect Med. 2015;5(5):a018572.
- 21. Scanga CA, Flynn JL. Modeling tuberculosis in nonhuman primates. Cold Spring Harb Perspect Med. 2014;4(12):a018564. pmid:25213189
- 22. Dharmadhikari AS, Nardell EA. What animal models teach humans about tuberculosis. Am J Respir Cell Mol Biol. 2008;39(5):503–8. pmid:18556589
- 23. Helke KL, Mankowski JL, Manabe YC. Animal models of cavitation in pulmonary tuberculosis. Tuberculosis (Edinb). 2006;86(5):337–48. pmid:16359922
- 24. Niazi MK, Dhulekar N, Schmidt D, Major S, Cooper R, Abeijon C, et al. Lung necrosis and neutrophils reflect common pathways of susceptibility to Mycobacterium tuberculosis in genetically diverse, immune-competent mice. Dis Model Mech. 2015;8(9):1141–53. pmid:26204894
- 25. Smith C, Proulx M, Olive A, Laddy D, Mishra B, Moss C, et al. Tuberculosis Susceptibility and Vaccine Protection Are Independently Controlled by Host Genotype. mBio. 2016;7(5):e01516. pmid:27651361
- 26. Smith CM, Baker RE, Proulx MK, Mishra BB, Long JE, Park SW, et al. Host-pathogen genetic interactions underlie tuberculosis susceptibility in genetically diverse mice. Elife. 2022;11. pmid:35112666
- 27. Kurtz SL, Rydén P, Elkins KL. Transcriptional signatures measured in whole blood correlate with protection against tuberculosis in inbred and outbred mice. PLoS One. 2023;18(8):e0289358. pmid:37535648
- 28. Kurtz SL, Rossi AP, Beamer GL, Gatti DM, Kramnik I, Elkins KL. The Diversity Outbred Mouse Population Is an Improved Animal Model of Vaccination against Tuberculosis That Reflects Heterogeneity of Protection. mSphere. 2020;5(2).
- 29. Lai R, Gong DN, Williams T, Ogunsola AF, Cavallo K, Lindestam Arlehamn CS, et al. Host genetic background is a barrier to broadly effective vaccine-mediated protection against tuberculosis. J Clin Invest. 2023;133(13). pmid:37200108
- 30. Churchill GA, Gatti DM, Munger SC, Svenson KL. The Diversity Outbred mouse population. Mammalian genome: official journal of the International Mammalian Genome Society. 2012;23(9–10):713–8. pmid:22892839
- 31. Tabangin ME, Woo JG, Martin LJ. The effect of minor allele frequency on the likelihood of obtaining false positives. BMC Proc. 2009;3 Suppl 7(Suppl 7):S41. pmid:20018033
- 32. Gatti DM, Svenson KL, Shabalin A, Wu LY, Valdar W, Simecek P, et al. Quantitative trait locus mapping methods for diversity outbred mice. G3. 2014;4(9):1623–33. pmid:25237114
- 33. Ahmed M, Thirunavukkarasu S, Rosa BA, Thomas KA, Das S, Rangel-Moreno J, et al. Immune correlates of tuberculosis disease and risk translate across species. Science translational medicine. 2020;12(528). pmid:31996462
- 34. Koyuncu D, Niazi MKK, Tavolara T, Abeijon C, Ginese ML, Liao Y, et al. CXCL1: A new diagnostic biomarker for human tuberculosis discovered using Diversity Outbred mice. PLoS Pathog. 2021;17(8):e1009773. pmid:34403447
- 35. Flynn JL, Goldstein MM, Chan J, Triebold KJ, Pfeffer K, Lowenstein CJ, et al. Tumor necrosis factor-alpha is required in the protective immune response against Mycobacterium tuberculosis in mice. Immunity. 1995;2(6):561–72. pmid:7540941
- 36. Flynn JL, Chan J, Triebold KJ, Dalton DK, Stewart TA, Bloom BR. An essential role for interferon gamma in resistance to Mycobacterium tuberculosis infection. J Exp Med. 1993;178(6):2249–54. pmid:7504064
- 37. 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. J Exp Med. 1997;186(1):39–45. pmid:9206995
- 38. Wong EA, Evans S, Kraus CR, Engelman KD, Maiello P, Flores WJ, et al. IL-10 Impairs Local Immune Response in Lung Granulomas and Lymph Nodes during Early Mycobacterium tuberculosis Infection. J Immunol. 2020;204(3):644–59. pmid:31862711
- 39. Guan Y, Ackert-Bicknell CL, Kell B, Troyanskaya OG, Hibbs MA. Functional genomics complements quantitative genetics in identifying disease-gene associations. PLoS Comput Biol. 2010;6(11):e1000991. pmid:21085640
- 40. Tyler AL, Raza A, Krementsov DN, Case LK, Huang R, Ma RZ, et al. Network-Based Functional Prediction Augments Genetic Association To Predict Candidate Genes for Histamine Hypersensitivity in Mice. G3. 2019;9(12):4223–33. pmid:31645420
- 41. Goya J, Wong AK, Yao V, Krishnan A, Homilius M, Troyanskaya OG. FNTM: a server for predicting functional networks of tissues in mouse. Nucleic acids research. 2015;43(W1):W182–7. pmid:25940632
- 42. Crouser ED, White P, Caceres EG, Julian MW, Papp AC, Locke LW, et al. A Novel In Vitro Human Granuloma Model of Sarcoidosis and Latent Tuberculosis Infection. Am J Respir Cell Mol Biol. 2017;57(4):487–98. pmid:28598206
- 43. Zhang ZM, Zhang AR, Xu M, Lou J, Qiu WQ. TLR-4/miRNA-32-5p/FSTL1 signaling regulates mycobacterial survival and inflammatory responses in Mycobacterium tuberculosis-infected macrophages. Experimental cell research. 2017;352(2):313–21. pmid:28215633
- 44. Deng Q, Huang J, Yan J, Mao E, Chen H, Wang C. Circ_0001490/miR-579-3p/FSTL1 axis modulates the survival of mycobacteria and the viability, apoptosis and inflammatory response in Mycobacterium tuberculosis-infected macrophages. Tuberculosis (Edinb). 2021;131:102123. pmid:34555658
- 45. Duan XY, Sun Y, Zhao ZF, Shi YQ, Ma XY, Tao L, et al. Baicalin attenuates LPS-induced alveolar type II epithelial cell A549 injury by attenuation of the FSTL1 signaling pathway via increasing miR-200b-3p expression. Innate Immun. 2021;27(4):294–312. pmid:34000873
- 46. Henkel M, Dutta JA, Partyka J, Eddens T, Hirsch R, Kolls JK, et al. Regulation of Pulmonary Bacterial Immunity by Follistatin-Like Protein 1. Infect Immun. 2020;89(1). pmid:33077624
- 47. Gopal R, Monin L, Torres D, Slight S, Mehra S, McKenna KC, et al. S100A8/A9 proteins mediate neutrophilic inflammation and lung pathology during tuberculosis. Am J Respir Crit Care Med. 2013;188(9):1137–46. pmid:24047412
- 48. Qiu J, Peng P, Xin M, Wen Z, Chen Z, Lin S, et al. ZBTB20-mediated titanium particle-induced peri-implant osteolysis by promoting macrophage inflammatory responses. Biomater Sci. 2020;8(11):3147–63. pmid:32363359
- 49. Pasqualini R, Bodorova J, Ye S, Hemler ME. A study of the structure, function and distribution of beta 5 integrins using novel anti-beta 5 monoclonal antibodies. Journal of cell science. 1993;105 (Pt 1):101–11. pmid:7689573
- 50. Vogel W. Discoidin domain receptors: structural relations and functional implications. The FASEB journal. 1999;13(9001):S77–S82. pmid:10352148
- 51. Mattiotti A, Prakash S, Barnett P, van den Hoff MJB. Follistatin-like 1 in development and human diseases. Cell Mol Life Sci. 2018;75(13):2339–54. pmid:29594389
- 52. Kuroda Y, Murakami H, Enomoto Y, Tsurusaki Y, Takahashi K, Mitsuzuka K, et al. A novel gene (FAM20B encoding glycosaminoglycan xylosylkinase) for neonatal short limb dysplasia resembling Desbuquois dysplasia. Clinical genetics. 2019;95(6):713–7. pmid:30847897
- 53. Koike T, Izumikawa T, Tamura J, Kitagawa H. FAM20B is a kinase that phosphorylates xylose in the glycosaminoglycan-protein linkage region. The Biochemical journal. 2009;421(2):157–62. pmid:19473117
- 54. Zhu C, Chen G, Zhao Y, Gao XM, Wang J. Regulation of the Development and Function of B Cells by ZBTB Transcription Factors. Frontiers in immunology. 2018;9:580.
- 55. Sutherland AP, Zhang H, Zhang Y, Michaud M, Xie Z, Patti M-E, et al. Zinc finger protein Zbtb20 is essential for postnatal survival and glucose homeostasis. Molecular and cellular biology. 2009;29(10):2804–15. pmid:19273596
- 56. Lightman SM, Utley A, Lee KP. Survival of Long-Lived Plasma Cells (LLPC): Piecing Together the Puzzle. Frontiers in immunology. 2019;10:965. pmid:31130955
- 57. Sun Y, Preiss NK, Valenteros KB, Kamal Y, Usherwood YK, Frost HR, et al. Zbtb20 Restrains CD8 T Cell Immunometabolism and Restricts Memory Differentiation and Antitumor Immunity. J Immunol. 2020;205(10):2649–66. pmid:32998985
- 58. Schilling D, Pittelkow MR, Kumar R. IEX-1, an immediate early gene, increases the rate of apoptosis in keratinocytes. Oncogene. 2001;20(55):7992–7. pmid:11753682
- 59. Im H-J, Pittelkow MR, Kumar R. Divergent Regulation of the Growth-promoting GeneIEX-1 by the p53 Tumor Suppressor and Sp1. Journal of Biological Chemistry. 2002;277(17):14612–21. pmid:11844788
- 60. Arlt A, Schäfer H. Role of the immediate early response 3 (IER3) gene in cellular stress response, inflammation and tumorigenesis. Eur J Cell Biol. 2011;90(6–7):545–52. pmid:21112119
- 61. Harrison DE, Astle CM, Niazi MK, Major S, Beamer GL. Genetically diverse mice are novel and valuable models of age-associated susceptibility to Mycobacterium tuberculosis. Immun Ageing. 2014;11(1):24. pmid:25606048
- 62. Cyktor JC, Carruthers B, Kominsky RA, Beamer GL, Stromberg P, Turner J. IL-10 inhibits mature fibrotic granuloma formation during Mycobacterium tuberculosis infection. J Immunol. 2013;190(6):2778–90.
- 63. Vesosky B, Rottinghaus EK, Stromberg P, Turner J, Beamer G. CCL5 participates in early protection against Mycobacterium tuberculosis. J Leukoc Biol. 2010;87(6):1153–65.
- 64. Ullman-Cullere MH, Foltz CJ. Body condition scoring: a rapid and accurate method for assessing health status in mice. Lab Anim Sci. 1999;49(3):319–23. pmid:10403450
- 65. Tavolara TE, Niazi MKK, Ginese M, Piedra-Mora C, Gatti DM, Beamer G, et al. Automatic discovery of clinically interpretable imaging biomarkers for Mycobacterium tuberculosis supersusceptibility using deep learning. EBioMedicine. 2020;62:103094. pmid:33166789
- 66. Beamer GL, Flaherty DK, Vesosky B, Turner J. Peripheral blood gamma interferon release assays predict lung responses and Mycobacterium tuberculosis disease outcome in mice. Clin Vaccine Immunol. 2008;15(3):474–83.
- 67. Tavolara TE, Niazi MKK, Gower AC, Ginese M, Beamer G, Gurcan MN. Deep learning predicts gene expression as an intermediate data modality to identify susceptibility patterns in Mycobacterium tuberculosis infected Diversity Outbred mice. EBioMedicine. 2021;67:103388. pmid:34000621
- 68. Keane TM, Goodstadt L, Danecek P, White MA, Wong K, Yalcin B, et al. Mouse genomic variation and its effect on phenotypes and gene regulation. Nature. 2011;477(7364):289–94. pmid:21921910
- 69. Morgan AP, Fu CP, Kao CY, Welsh CE, Didion JP, Yadgary L, et al. The Mouse Universal Genotyping Array: From Substrains to Subspecies. G3. 2015;6(2):263–79. pmid:26684931
- 70. Broman KW, Gatti DM, Simecek P, Furlotte NA, Prins P, Sen S, et al. R/qtl2: Software for Mapping Quantitative Trait Loci with High-Dimensional Data and Multiparent Populations. Genetics. 2019;211(2):495–502. pmid:30591514
- 71. Cheng R, Abney M, Palmer AA, Skol AD. QTLRel: an R package for genome-wide association studies in which relatedness is a concern. BMC Genet. 2011;12:66. pmid:21794153
- 72. Van Ooijen JW. LOD significance thresholds for QTL analysis in experimental populations of diploid species. Heredity (Edinb). 1999;83 (Pt 5):613–24.
- 73. David Meyer ED, Kurt Hornik, Andreas Weingessel and Friedrich Leisch. e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. 1.7–7 ed2021. p. R package.
- 74. Blake JA, Baldarelli R, Kadin JA, Richardson JE, Smith CL, Bult CJ, et al. Mouse Genome Database (MGD): Knowledgebase for mouse-human comparative biology. Nucleic acids research. 2021;49(D1):D981–D7. pmid:33231642
- 75. Kus P, Gurcan MN, Beamer G. Automatic Detection of Granuloma Necrosis in Pulmonary Tuberculosis Using a Two-Phase Algorithm: 2D-TB. Microorganisms. 2019;7(12). pmid:31817882
- 76. Niazi MKK, Beamer G, Gurcan MN. An application of transfer learning to neutrophil cluster detection for tuberculosis: efficient implementation with nonmetric multidimensional scaling and sampling: SPIE; 2018.
- 77. Niazi MKK, Beamer G, Gurcan MN. A computational framework to detect normal and tuberculosis infected lung from H and E-stained whole slide images: SPIE; 2017.
- 78. Sanchez F, Radaeva TV, Nikonenko BV, Persson AS, Sengul S, Schalling M, et al. Multigenic control of disease severity after virulent Mycobacterium tuberculosis infection in mice. Infect Immun. 2003;71(1):126–31. pmid:12496157
- 79. Yan BS, Kirby A, Shebzukhov YV, Daly MJ, Kramnik I. Genetic architecture of tuberculosis resistance in a mouse model of infection. Genes Immun. 2006;7(3):201–10. pmid:16452998
- 80. Dodd CE, Schlesinger LS. New concepts in understanding latent tuberculosis. Curr Opin Infect Dis. 2017;30(3):316–21. pmid:28177961
- 81. Behr MA, Edelstein PH, Ramakrishnan L. Revisiting the timetable of tuberculosis. Bmj. 2018;362:k2738. pmid:30139910
- 82. Loddenkemper R, Lipman M, Zumla A. Clinical Aspects of Adult Tuberculosis. Cold Spring Harb Perspect Med. 2015;6(1):a017848.
- 83. Hunter RL, Actor JK, Hwang SA, Karev V, Jagannath C. Pathogenesis of post primary tuberculosis: immunity and hypersensitivity in the development of cavities. Annals of clinical and laboratory science. 2014;44(4):365–87. pmid:25361920
- 84. Naranbhai V. The Role of Host Genetics (and Genomics) in Tuberculosis. Microbiol Spectr. 2016;4(5). pmid:27787193
- 85. Abel L, El-Baghdadi J, Bousfiha AA, Casanova JL, Schurr E. Human genetics of tuberculosis: a long and winding road. Philos Trans R Soc Lond B Biol Sci. 2014;369(1645):20130428. pmid:24821915
- 86. Kondratieva E, Logunova N, Majorov K, Averbakh M, Apt A. Host genetics in granuloma formation: human-like lung pathology in mice with reciprocal genetic susceptibility to M. tuberculosis and M. avium. PLoS One. 2010;5(5):e10515. pmid:20463893
- 87. Bustamante J, Boisson-Dupuis S, Abel L, Casanova JL. Mendelian susceptibility to mycobacterial disease: genetic, immunological, and clinical features of inborn errors of IFN-gamma immunity. Semin Immunol. 2014;26(6):454–70.
- 88. Qu HQ, Fisher-Hoch SP, McCormick JB. Molecular immunity to mycobacteria: knowledge from the mutation and phenotype spectrum analysis of Mendelian susceptibility to mycobacterial diseases. International journal of infectious diseases: IJID: official publication of the International Society for Infectious Diseases. 2011;15(5):e305–13. pmid:21330176
- 89. Cottle LE. Mendelian susceptibility to mycobacterial disease. Clinical genetics. 2011;79(1):17–22. pmid:20718793
- 90. Kramnik I. Genetic dissection of host resistance to Mycobacterium tuberculosis: the sst1 locus and the Ipr1 gene. Current topics in Microbiology and Immunology. 2008;321:123–48. pmid:18727490
- 91. Tosh K, Campbell SJ, Fielding K, Sillah J, Bah B, Gustafson P, et al. Variants in the SP110 gene are associated with genetic susceptibility to tuberculosis in West Africa. Proc Natl Acad Sci U S A. 2006;103(27):10364–8. pmid:16803959
- 92. Pan H, Yan BS, Rojas M, Shebzukhov YV, Zhou H, Kobzik L, et al. Ipr1 gene mediates innate immunity to tuberculosis. Nature. 2005;434(7034):767–72. pmid:15815631
- 93. Kramnik I, Demant P, Bloom BB. Susceptibility to tuberculosis as a complex genetic trait: analysis using recombinant congenic strains of mice. Novartis Found Symp. 1998;217:120–31; discussion 32–7. pmid:9949805
- 94. Apt AS, Avdienko VG, Nikonenko BV, Kramnik IB, Moroz AM, Skamene E. Distinct H-2 complex control of mortality, and immune responses to tuberculosis infection in virgin and BCG-vaccinated mice. Clin Exp Immunol. 1993;94(2):322–9. pmid:8222323
- 95. Barrera LF, Kramnik I, Skamene E, Radzioch D. I-A beta gene expression regulation in macrophages derived from mice susceptible or resistant to infection with M. bovis BCG. Mol Immunol. 1997;34(4):343–55. pmid:9244347
- 96. Logunova N, Korotetskaya M, Polshakov V, Apt A. The QTL within the H2 Complex Involved in the Control of Tuberculosis Infection in Mice Is the Classical Class II H2-Ab1 Gene. PLoS genetics. 2015;11(11):e1005672. pmid:26618355
- 97. Scott NR, Swanson RV, Al-Hammadi N, Domingo-Gonzalez R, Rangel-Moreno J, Kriel BA, et al. S100A8/A9 regulates CD11b expression and neutrophil recruitment during chronic tuberculosis. J Clin Invest. 2020;130(6):3098–112. pmid:32134742
- 98. Yang H, Bell TA, Churchill GA, Pardo-Manuel de Villena F. On the subspecific origin of the laboratory mouse. Nat Genet. 2007;39(9):1100–7. pmid:17660819
- 99. Jiao L, Song J, Ding L, Liu T, Wu T, Zhang J, et al. A Novel Genetic Variation in NCF2, the Core Component of NADPH Oxidase, Contributes to the Susceptibility of Tuberculosis in Western Chinese Han Population. DNA and cell biology. 2020;39(1):57–62. pmid:31794672
- 100. Cooper AM, Segal BH, Frank AA, Holland SM, Orme IM. Transient loss of resistance to pulmonary tuberculosis in p47(phox-/-) mice. Infect Immun. 2000;68(3):1231–4. pmid:10678931
- 101. Wu J, Tian Y, Han L, Liu C, Sun T, Li L, et al. FAM20B-catalyzed glycosaminoglycans control murine tooth number by restricting FGFR2b signaling. BMC Biol. 2020;18(1):87. pmid:32664967
- 102. Saiyin W, Li L, Zhang H, Lu Y, Qin C. Inactivation of FAM20B causes cell fate changes in annulus fibrosus of mouse intervertebral disc and disc defects via the alterations of TGF-beta and MAPK signaling pathways. Biochim Biophys Acta Mol Basis Dis. 2019;1865(12):165555.
- 103. Ishikawa HO, Takeuchi H, Haltiwanger RS, Irvine KD. Four-jointed is a Golgi kinase that phosphorylates a subset of cadherin domains. Science. 2008;321(5887):401–4. pmid:18635802
- 104. Perez TD, Nelson WJ. Cadherin adhesion: mechanisms and molecular interactions. Handb Exp Pharmacol. 2004(165):3–21. pmid:20455088
- 105. Zygiel EM, Nolan EM. Transition Metal Sequestration by the Host-Defense Protein Calprotectin. Annual review of biochemistry. 2018;87:621–43. pmid:29925260
- 106. Wang S, Song R, Wang Z, Jing Z, Wang S, Ma J. S100A8/A9 in Inflammation. Frontiers in immunology. 2018;9:1298. pmid:29942307
- 107. Dale CS, Altier C, Cenac N, Giorgi R, Juliano MA, Juliano L, et al. Analgesic properties of S100A9 C-terminal domain: a mechanism dependent on calcium channel inhibition. Fundam Clin Pharmacol. 2009;23(4):427–38. pmid:19515068
- 108. Dale CS, Pagano Rde L, Paccola CC, Pinotti-Guirao T, Juliano MA, Juliano L, et al. Effect of the C-terminus of murine S100A9 protein on experimental nociception. Peptides. 2006;27(11):2794–802. pmid:16920226
- 109. Hudson BI, Lippman ME. Targeting RAGE Signaling in Inflammatory Disease. Annu Rev Med. 2018;69:349–64. pmid:29106804
- 110. Garcia V, Perera YR, Chazin WJ. A Structural Perspective on Calprotectin as a Ligand of Receptors Mediating Inflammation and Potential Drug Target. Biomolecules. 2022;12(4).
- 111. Ryckman C, Vandal K, Rouleau P, Talbot M, Tessier PA. Proinflammatory activities of S100: proteins S100A8, S100A9, and S100A8/A9 induce neutrophil chemotaxis and adhesion. The Journal of Immunology. 2003;170(6):3233–42. pmid:12626582
- 112. Liu Q, Li R, Li Q, Luo B, Lin J, Lyu L. High levels of plasma S100A9 at admission indicate an increased risk of death in severe tuberculosis patients. Journal of Clinical Tuberculosis and Other Mycobacterial Diseases. 2021;25:100270. pmid:34849408
- 113. Cesaro A, Defrene J, Lachhab A, Page N, Tardif MR, Al-Shami A, et al. Enhanced myelopoiesis and aggravated arthritis in S100a8-deficient mice. PLoS One. 2019;14(8):e0221528. pmid:31437241
- 114. Li HM, Huang Q, Tang F, Zhang TP. Altered NCF2, NOX2 mRNA Expression Levels in Peripheral Blood Mononuclear Cells of Pulmonary Tuberculosis Patients. Int J Gen Med. 2021;14:9203–9.
- 115.
Leung K. (99m)Tc-Diamine dioxime-Lys-Cys-Arg-Gly-Asp-Cyc-Phe-Cys-polyethylene glycol. Molecular Imaging and Contrast Agent Database (MICAD). Bethesda (MD): National Center for Biotechnology Information (US); 2004.
- 116. Lin Z, He R, Luo H, Lu C, Ning Z, Wu Y, et al. Integrin-β5, a miR-185-targeted gene, promotes hepatocellular carcinoma tumorigenesis by regulating β-catenin stability. Journal of Experimental & Clinical Cancer Research. 2018;37:1–13.
- 117. Shi W, He J, Huang Y, Zeng Z, Feng Z, Xu H, et al. Integrin beta5 enhances the malignancy of human colorectal cancer by increasing the TGF-beta signaling. Anticancer Drugs. 2021;32(7):717–26.
- 118. Zhuang H, Zhou Z, Ma Z, Li Z, Liu C, Huang S, et al. Characterization of the prognostic and oncologic values of ITGB superfamily members in pancreatic cancer. J Cell Mol Med. 2020;24(22):13481–93. pmid:33073486
- 119. Nurzat Y, Su W, Min P, Li K, Xu H, Zhang Y. Identification of Therapeutic Targets and Prognostic Biomarkers Among Integrin Subunits in the Skin Cutaneous Melanoma Microenvironment. Front Oncol. 2021;11:751875. pmid:34660316
- 120. Zhang LY, Guo Q, Guan GF, Cheng W, Cheng P, Wu AH. Integrin Beta 5 Is a Prognostic Biomarker and Potential Therapeutic Target in Glioblastoma. Front Oncol. 2019;9:904. pmid:31616629
- 121. Ouchi N, Asaumi Y, Ohashi K, Higuchi A, Sono-Romanelli S, Oshima Y, et al. DIP2A functions as a FSTL1 receptor. J Biol Chem. 2010;285(10):7127–34. pmid:20054002
- 122. Jin X, Nie E, Zhou X, Zeng A, Yu T, Zhi T, et al. Fstl1 promotes glioma growth through the BMP4/Smad1/5/8 signaling pathway. Cellular Physiology and Biochemistry. 2017;44(4):1616–28. pmid:29212066
- 123. Murakami K, Tanaka M, Usui T, Kawabata D, Shiomi A, Iguchi-Hashimoto M, et al. Follistatin-related protein/follistatin-like 1 evokes an innate immune response via CD14 and toll-like receptor 4. FEBS letters. 2012;586(4):319–24. pmid:22265692
- 124. Horak M, Fairweather D, Kokkonen P, Bednar D, Bienertova-Vasku J. Follistatin-like 1 and its paralogs in heart development and cardiovascular disease. Heart Fail Rev. 2022;27(6):2251–65. pmid:35867287
- 125. Zhang Y, Luo D, Tang MI, Jiang DI, Yi H. Circ-WDR27 regulates mycobacterial vitality and secretion of inflammatory cytokines in Mycobacterium tuberculosis-infected macrophages via the miR-370-3p/FSTL1 signal network. J Biosci. 2022;47.
- 126. Zhang W, Mi J, Li N, Sui L, Wan T, Zhang J, et al. Identification and characterization of DPZF, a novel human BTB/POZ zinc finger protein sharing homology to BCL-6. Biochem Biophys Res Commun. 2001;282(4):1067–73. pmid:11352661
- 127. Xie Z, Zhang H, Tsai W, Zhang Y, Du Y, Zhong J, et al. Zinc finger protein ZBTB20 is a key repressor of alpha-fetoprotein gene transcription in liver. Proceedings of the National Academy of Sciences. 2008;105(31):10859–64. pmid:18669658
- 128. Xie Z, Ma X, Ji W, Zhou G, Lu Y, Xiang Z, et al. Zbtb20 is essential for the specification of CA1 field identity in the developing hippocampus. Proceedings of the National Academy of Sciences. 2010;107(14):6510–5. pmid:20308569
- 129. Zhang Y, Xie Z, Zhou L, Li L, Zhang H, Zhou G, et al. The zinc finger protein ZBTB20 regulates transcription of fructose-1, 6-bisphosphatase 1 and β cell function in mice. Gastroenterology. 2012;142(7):1571–80. e6.
- 130. Liu X, Zhang P, Bao Y, Han Y, Wang Y, Zhang Q, et al. Zinc finger protein ZBTB20 promotes Toll-like receptor-triggered innate immune responses by repressing IκBα gene transcription. Proc Natl Acad Sci U S A. 2013;110(27):11097–102.
- 131.
Arora V, Ferreira CR, Dua Puri R, Verma IC. Primrose Syndrome. In: Adam MP, Mirzaa GM, Pagon RA, Wallace SE, Bean LJH, Gripp KW, et al., editors. GeneReviews. Seattle (WA): University of Washington, Seattle; 1993.
- 132. Kim SH, Lee S, Suk K, Bark H, Jun CD, Kim DK, et al. Discoidin domain receptor 1 mediates collagen-induced nitric oxide production in J774A.1 murine macrophages. Free radical biology & medicine. 2007;42(3):343–52. pmid:17210447
- 133. Wang H, Wen Y, Wang L, Wang J, Chen H, Chen J, et al. DDR1 activation in macrophage promotes IPF by regulating NLRP3 inflammasome and macrophage reaction. International immunopharmacology. 2022;113(Pt A):109294. pmid:36257259
- 134. Yeh YC, Lin HH, Tang MJ. Dichotomy of the function of DDR1 in cells and disease progression. Biochim Biophys Acta Mol Cell Res. 2019;1866(11):118473. pmid:30954568
- 135. Taciak B, Białasek M, Braniewska A, Sas Z, Sawicka P, Kiraga Ł, et al. Evaluation of phenotypic and functional stability of RAW 264.7 cell line through serial passages. PLoS One. 2018;13(6):e0198943. pmid:29889899
- 136. Ralph P, Prichard J, Cohn M. Reticulum cell sarcoma: an effector cell in antibody-dependent cell-mediated immunity. J Immunol. 1975;114(2 pt 2):898–905. pmid:1089721
- 137. Zhang X, Hu Y, Pan Y, Xiong Y, Zhang Y, Han M, et al. DDR1 promotes hepatocellular carcinoma metastasis through recruiting PSD4 to ARF6. Oncogene. 2022;41(12):1821–34. pmid:35140331
- 138. Su H, Yang F, Fu R, Trinh B, Sun N, Liu J, et al. Collagenolysis-dependent DDR1 signalling dictates pancreatic cancer outcome. Nature. 2022;610(7931):366–72. pmid:36198801
- 139. Ma R, Xie X, Zhao L, Wu Y, Wang J. Discoidin domain receptors (DDRs): Potential implications in periodontitis. J Cell Physiol. 2022;237(1):189–98. pmid:34431091
- 140. Moll S, Desmoulière A, Moeller MJ, Pache JC, Badi L, Arcadu F, et al. DDR1 role in fibrosis and its pharmacological targeting. Biochim Biophys Acta Mol Cell Res. 2019;1866(11):118474. pmid:30954571
- 141. Sirvent A, Espie K, Papadopoulou E, Naim D, Roche S. New functions of DDR1 collagen receptor in tumor dormancy, immune exclusion and therapeutic resistance. Front Oncol. 2022;12:956926.
- 142. Marchioro HZ, Silva de Castro CC, Fava VM, Sakiyama PH, Dellatorre G, Miot HA. Update on the pathogenesis of vitiligo. An Bras Dermatol. 2022;97(4):478–90. pmid:35643735
- 143. Vilella E, Gas C, Garcia-Ruiz B, Rivera FJ. Expression of DDR1 in the CNS and in myelinating oligodendrocytes. Biochim Biophys Acta Mol Cell Res. 2019;1866(11):118483. pmid:31108116
- 144. Wu M. Roles of the stress-induced gene IEX-1 in regulation of cell death and oncogenesis. Apoptosis. 2003;8:11–8. pmid:12510147
- 145. Arlt A, Grobe O, Sieke A, Kruse M-L, FoÈlsch UR, Schmidt WE, et al. Expression of the NF-κB target gene IEX-1 (p22/PRG1) does not prevent cell death but instead triggers apoptosis in HeLa cells. Oncogene. 2001;20(1):69–76.
- 146. Sebens Müerköster S, Rausch A, Isberner A, Minkenberg J, Blaszczuk E, Witt M, et al. The apoptosis-inducing effect of gastrin on colorectal cancer cells relates to an increased IEX-1 expression mediating NF-κB inhibition. Oncogene. 2008;27(8):1122–34.
- 147. Akilov OE, Ustyugova IV, Zhi L, Hasan T, Wu MX. Enhanced susceptibility to Leishmania infection in resistant mice in the absence of immediate early response gene X-1. J Immunol. 2009;183(12):7994–8003. pmid:19923449
- 148. Leisching G, Pietersen RD, van Heerden C, van Helden P, Wiid I, Baker B. RNAseq reveals hypervirulence-specific host responses to M. tuberculosis infection. Virulence. 2017;8(6):848–58. pmid:27763806
- 149. Coultas L, Chawengsaksophak K, Rossant J. Endothelial cells and VEGF in vascular development. Nature. 2005;438(7070):937–45. pmid:16355211
- 150. Ferrara N, Gerber HP. The role of vascular endothelial growth factor in angiogenesis. Acta Haematol. 2001;106(4):148–56. pmid:11815711
- 151. Ferrara N, Alitalo K. Clinical applications of angiogenic growth factors and their inhibitors. Nature medicine. 1999;5(12):1359–64. pmid:10581076
- 152. Shibuya M. Vascular Endothelial Growth Factor (VEGF) and Its Receptor (VEGFR) Signaling in Angiogenesis: A Crucial Target for Anti- and Pro-Angiogenic Therapies. Genes Cancer. 2011;2(12):1097–105. pmid:22866201
- 153. Davuluri KS, Singh AK, Kumar V, Singh SV, Singh AV, Kumar S, et al. Stimulated expression of ELR+ chemokines, VEGFA and TNF-AIP3 promote mycobacterial dissemination in extrapulmonary tuberculosis patients and Cavia porcellus model of tuberculosis. Tuberculosis (Edinb). 2022;135:102224. pmid:35763913
- 154. Delemarre EM, van Hoorn L, Bossink AWJ, Drylewicz J, Joosten SA, Ottenhoff THM, et al. Serum Biomarker Profile Including CCL1, CXCL10, VEGF, and Adenosine Deaminase Activity Distinguishes Active From Remotely Acquired Latent Tuberculosis. Frontiers in immunology. 2021;12:725447. pmid:34691031
- 155. Golubinskaya EP, Filonenko TG, Kramar TV, Yermola YA, Kubyshkin AV, Gerashenko AV, et al. Dysregulation of VEGF-dependent angiogenesis in cavernous lung tuberculosis. Pathophysiology. 2019;26(3–4):381–7. pmid:31791834
- 156. Harding JS, Herbath M, Chen Y, Rayasam A, Ritter A, Csoka B, et al. VEGF-A from Granuloma Macrophages Regulates Granulomatous Inflammation by a Non-angiogenic Pathway during Mycobacterial Infection. Cell Rep. 2019;27(7):2119–31.e6. pmid:31091450
- 157. Tao R-H, Kawate H, Ohnaka K, Ishizuka M, Hagiwara H, Takayanagi R. Opposite effects of alternative TZF spliced variants on androgen receptor. Biochemical and biophysical research communications. 2006;341(2):515–21. pmid:16446156
- 158. Pioli PD, Debnath I, Weis JJ, Weis JH. Zfp318 regulates IgD expression by abrogating transcription termination within the Ighm/Ighd locus. J Immunol. 2014;193(5):2546–53. pmid:25057009
- 159. Enders A, Short A, Miosge LA, Bergmann H, Sontani Y, Bertram EM, et al. Zinc-finger protein ZFP318 is essential for expression of IgD, the alternatively spliced Igh product made by mature B lymphocytes. Proceedings of the National Academy of Sciences. 2014;111(12):4513–8. pmid:24616512
- 160. Schwickert TA, Tagoh H, Schindler K, Fischer M, Jaritz M, Busslinger M. Ikaros prevents autoimmunity by controlling anergy and Toll-like receptor signaling in B cells. Nat Immunol. 2019;20(11):1517–29. pmid:31591571