key: cord-0272838-161fbfz8
authors: Lerm, M.; Das, J.; Paues, J.; Idh, N.; Pehrson, I.
title: A DNA methylome biosignature in alveolar macrophages from TB-exposed individuals predicts exposure to mycobacteria
date: 2021-03-20
journal: nan
DOI: 10.1101/2021.03.16.21253732
sha: 6c2a530d9ae41bea13ddd3bbbe1dcbb7f7d3a9a1
doc_id: 272838
cord_uid: 161fbfz8

Several studies have identified biomarkers for tuberculosis (TB) diagnosis based on blood cell transcriptomics. Here, we instead studied epigenomics in the lung compartment by obtaining induced sputum from subjects included in a TB contact tracing. CD3- and HLA-DR-positive cells were isolated from the collected sputum and DNA methylome analyses performed. Unsupervised cluster analysis revealed that DNA methylomes of cells from TB-exposed individuals and controls appeared as separate clusters and the numerous genes that were differentially methylated were functionally connected. The enriched pathways were strongly correlated to data from published work on protective heterologous immunity to TB. Taken together, our results demonstrate that epigenetic changes related to trained immunity occurs in the pulmonary immune cells of TB-exposed individuals and that a DNA methylation signature can be derived from the DNA methylome. Such a signature can be further developed for clinical use as a marker of TB exposure.

Tuberculosis (TB) is caused by Mycobacterium tuberculosis, which is transmitted between individuals through the inhalation of aerosols generated by coughing 1 . The disease still claims more than one million human lives annually and an expansion of the current toolkit for diagnosis, prevention and treatment is critical for reaching the United Nations' Sustainable Development Goals for 2030 of ending the TB epidemic 2 . In response to the urgent need for new approaches to diagnose TB, recent studies identified TB-specific biosignatures based on RNA transcription profiles in peripheral blood of TB-infected individuals 3, 4 . Similar TB biosignatures have been identified that could predict disease progression 5, 6 and reflect treatment monitoring 6, 7 . Biosignatures for the detection of TB-exposure in easily accessible clinical samples could provide a basis for the development of novel effective point-of-care diagnostic tools.

The only available TB vaccine is Bacillus Calmette Guérin (BCG), which is based on live attenuated M. bovis 8 . In a recent study, we showed that administration of the BCG vaccine to healthy subjects induced profound epigenetic alterations in immune cells, which correlated with enhanced anti-mycobacterial activity in macrophages isolated from the vaccinees 9 . The changes were reflected in the DNA methylome, with the strongest response being recorded within weeks after vaccination 9 . Our observation that BCG induces alterations of the DNA methylome of immune cells has later been confirmed by others 10,11 . Since BCG vaccination reflects an in vivo interaction between immune cells and viable mycobacteria, we hypothesized that natural exposure to M. tuberculosis would induce similar changes not only in TB patients, but also in individuals who have been exposed to TB. To enable recruitment of control subjects with very low likelihood of previous TB exposure, we performed the study in a low-endemic setting. We All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 20, 2021. ; https://doi.org/10.1101/2021.03.16.21253732 doi: medRxiv preprint established a protocol for the isolation of distinct cell populations from induced sputum 12 , from which DNA could be isolated. Analyses of DNA methylomes of immune cells isolated from lungs and peripheral blood allowed us to identify distinct DNA methylation signatures in TBexposed individuals. The signature was most prominent in the lung-derived cell populations.

Pathway analyses revealed strong overlaps with previous studies on BCG-induced epigenetic signatures that could be correlated with protection against M. tuberculosis. We also identified pathway overlap with previous work on trained immunity induced by -glucan from Candida albicans. In conclusion, we found a distinct pattern of DNA methylome changes in immune cells isolated from lungs in individuals with documented exposure to TB. The alterations strongly overlap with pathways described as reflecting trained immunity and an enhanced antimycobacterial response. The identified signature has potential to be used as a tool to identify TB exposure in low-endemic settings.

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 20, 2021. ; https://doi.org/10.1101/2021.03.16.21253732 doi: medRxiv preprint

To determine epigenetic changes in the immune cells in TB-exposed individuals, we recruited subjects enrolled in a routinely performed TB contact tracing at Linköping University Hospital, Sweden. Age-matched individuals were included as controls ( Table 1 ). The index case was diagnosed with drug-sensitive pulmonary TB and had completed two out of six months of standard treatment at the time of sample collection. All included subjects except one (a TB contact) were BCG-vaccinated (Table 1) . Interferon-Gamma Release Assay (IGRA) status was determined and among the exposed individuals, two were positive (including the index case) and among the controls, one individual (C2) was classified as 'borderline'positive 13 (Table   1) . From induced sputum, HLA-DR-positive (antigen-presenting cells, dominated by macrophages 14 and CD3-positive (T cell) populations were purified, whereas the PBMC fraction extracted from blood were kept as a mixed population (Fig. 1) .

DNA methylome data from TB-exposed individuals form a separate cluster DNA isolation from the studied cell populations was followed by global DNA methylation analysis using the Illumina 450K protocol. After curation of the data 15 , the datasets were subjected to unsupervised hierarchical cluster analysis based on DNA CpG methylation values (see project work-flow, Fig. 1 ). This approach accurately clustered the participants into TB-exposed and controls based on the DNA methylome data derived from both HLA-DR-and CD3-positive cell populations. (Fig. 2a, b) . On the other hand, in the PBMC-derived dataset, the TB index case appeared outside the clusters and two of the controls clustered with the other exposed individuals, one of them ("Con_2") being the individual identified as border linepositive in the IGRA test (Fig. 2c ).

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Next, we identified the differentially methylated CpG sites (DMCs) and DMGs by comparing the TB-exposed and controls groups for each cell population. To filter out the most significantly altered DMGs in the dataset, the stringency criteria of log2 0.3 fold increased or decreased values and Benjamini-Hochberg (BH)-corrected p-value < 0.05 (HLA-DR), <0.1 (CD3) and <0.2 (PBMC) were applied. The results are depicted as volcano plots, which show that the DNA methylomes of TB-exposed most strongly differ in the HLA-DR cells as compared to control subjects, followed by the CD3 population, whereas PBMC datasets reveled fewer DMGs (Fig.   2d , e, f, Table 2 ). To highlight the locus position of the DMGs, chromosome maps were constructed (Suppl. Fig. S1 ). Using the same stringency criteria as for the HLA-DR analysis, we tested whether DMGs would emerge when the datasets were arranged in other possible groups as derived from the demographics (>/< median age, sex, IGRA status), Neither age nor IGRA status generated any significant DMGs with these settings, and gender rendered only three ( Table 2) .

Using the Panther Database, we investigated whether the identified DMGs were enriched in known pathways (Fig. 3a,b,c) . The analysis revealed pathways with relevance for TB infection, including hypoxia-inducible factor (HIF)1- activation, Vitamin D metabolism and p38, Wnt, Notch, interleukin, chemokine and cytokine signaling pathways [16] [17] [18] [19] [20] [21] [22] [23] . Common pathways shared between at least two of the cell populations included B cell activation, glycolysis, angiotensin II signaling, and cholecystokinin signaling. Notably, several pathways being named after their known functions in the nervous system were enriched in the studied cell populations, including pathways involved in axon guidance and adrenaline, acetylcholine and glutamate All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 20, 2021. ; https://doi.org/10.1101/2021.03.16.21253732 doi: medRxiv preprint signaling. In the PBMC population but not in the lung cell populations, the interferon- signaling pathway was identified among the enriched pathways.

Comparisons across cell populations and species reveals the existence of a common DNA methylome-based biosignature in mycobacteria-exposed immune cells Given the fact that the interaction between mycobacteria and eukaryotes is evolutionary ancient, we predicted that highly conserved pathways exist that are common among the studied cell populations. Combining the identified DMGs from the HLA-DR, CD3 and PBMCs in a Venn analysis, we discovered 185 common DMGs (Fig. 4a) . We expanded the Venn analyses to include data from our previous work on BCG vaccine-induced DMGs that correlated with enhanced mycobacterial control 9 , the rationale being that natural exposure to TB and BCG vaccination both represent in vivo encounters between mycobacteria and host immune cells.

Even though the routes of mycobacterial exposure differ profoundly in these settings, a set of 151 DMGs could be identified that overlapped between our previous BCG study 9 and all cell populations studied here (Fig. 4b) , suggesting that a highly conserved epigenetic response to mycobacterial challenge exists.

In 2018, Hasso-Agopsowicz et al. described alterations in DNA methylation patterns in PBMCs from BCG-vaccinated individuals, with concomitant enrichment in many immune-related pathways 10 . In order to compare that study with ours, we performed Panther analysis with the 185 common DMGs and matched the identified enriched pathways with those from that study, revealing that 75% of those pathways were the same as in the present study ( Fig. 5a and Suppl. Fig. S2a-c) . Figure 5b and Suppl. Table 1b demonstrates that for our PBMC data, the GO terms "biological processes" overlapped to 100% with the mouse study (same cell population) and to 31% and 65% for HLA-DR and CD3 cells respectively. In 2014, Saeed et al 24 demonstrated the induction of trained immunity pathways by another immune-training agent, -glucan. We assessed possible pathway overlap with that study and although there were fewer overlaps as compared to the BCG-induced pathways described above, again the strongest correlation was found in the PBMC fraction, in this case in the GO terms "cellular components" (Fig. 5c and Table 1c ).

Finally, we assessed how well the 284 CpG sites corresponding to the 185 overlapping DMGs performed in an unsupervised cluster analysis. To this end, we included one additional TB patient and two contacts, and collected HLA-DR cells from induced sputum, since the DNA methylome data this cell type was clearly outperforming the others with respect to accurate separation of the groups. Fig. 6 shows a k means-based dendrogram with a heatmap of the β values of the 284 CpG sites from the previous and the new subjects' samples.

Together the results demonstrate that the identified biosignature with the strongly enriched pathways can be linked to modifications of immune cell functions that result in improved antimycobacterial defense, possibly through trained immunity.

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in 

In this study, we present data suggesting that exposure to TB generates a very distinct DNA methylation signature in pulmonary immune cells. The signature was found not only in those with active or latent TB infection but also in individuals who are exposed but IGRA-negative.

The finding that also healthy, TB-exposed carry the signature opens up the possibility that the epigenetic alterations reflect a host-beneficial reprogramming of the immune mechanisms rather than being induced by M. tuberculosis as a step to evade the immune defense. This notion is supported by our observation that the DMGs identified in the present study strongly overlapped with the previously reported DNA methylation changes induced during BCG vaccination, which correlated with increased anti-mycobacterial capacity of macrophages 9 . In addition, we demonstrate that the GO data derived from our dataset display a strong overlap with data from a study on protective BCG vaccination in mice 11 . BCG vaccination has convincingly been shown to induce heterologous immunity protecting against childhood mortality from other causes than TB 25,26 . Based on our finding that natural TB exposure and BCG vaccination trigger similar epigenetic changes we propose the hypothesis that a "beneficial exposure" to TB exists, which protects against other infections through heterologous immunity. Along the same line, it has been shown that a substantial fraction of individuals exposed to TB can be defined as 'early clearers', since they do not test positively in the tuberculin skin test (TST) or IGRA 27 , suggesting effective eradication of the infection 27 . Identifying these early clearers and understanding the biology behind their resistance to TB infection could move the field forward towards novel strategies of TB prevention.

The rationale of comparing overlaps in mycobacteria-induced DNA methylation changes between different immune cells is based on the accumulated evidence of co-evolution of mycobacteria and amoeba 28,29 and the origin of phagocytic immune defense in metazoans from All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 20, 2021. ; https://doi.org/10.1101/2021.03.16.21253732 doi: medRxiv preprint amoeboid cells 30,31 , predicting evolutionary conserved pathways to be engaged in antimycobacterial defense. In line with this prediction, we identified the ubiquitously expressed and evolutionary conserved Wnt signaling pathway, which is found in all metazoans 32 and with homologs in amoeba 33 to be strongly enriched across all cell populations, settings and species.

The role for Wnt signaling in mycobacterial defense remains elusive, but many studies have Although macrophages and lymphocytes are not generally viewed as having many similarities, we found 34 of the identified pathways to overlap between HLA-DR and CD3. In data derived from the CD3 and PBMC populations, both of which represent lymphocytes, overlaps were All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 20, 2021. ; https://doi.org/10.1101/2021.03.16.21253732 doi: medRxiv preprint identified for glycolysis, glutamate receptor and angiotensin II pathways. Interestingly, a metabolic shift towards increased glycolysis, representative of the Warburg effect, has been strongly associated with trained immunity 34 . However, the literature is dominated by the view that this event takes place in trained myeloid cells, while we identified this circuit in CD3 cells (lymphocytes) and not in the HLA-DR cells (dominated by macrophages). The glutamate receptor is widely expressed on immune cells and have been described as having an important regulatory role in T cells, which can also produce and release glutamate 41 . The role for angiotensin II pathway in TB remains elusive, while Angiotensin II Converting Enzyme 2 is currently in the spotlight due to fact that the SARS-CoV2 virus utilizes it as a receptor for entry into host cells 42 . In the PBMC population, which over all showed a weaker epigenetic response, we found the interferon- signaling pathway, which has a well-established role in antimycobacterial defense (reviewed in 43 ), to be among the reprogramed pathways.

A weakness of our study is the small cohort, which warrants testing in larger cohorts performed in different clinical settings such as areas high-and low endemic for TB. Taken together, we present data supportive of DNA methylation changes that are induced through exposure to TB.

The changes correlate with findings from studies on BCG vaccination including TB protection, heterologous and trained immunity.

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in 

Patients with pulmonary TB, participants with occupational-or household-related TB-exposure and healthy controls, with an age ranging from 18 to 53 years, were enrolled at Linköping University Hospital and Linköping University, respectively. Included subjects (please see Table   1 for demographics) donated peripheral blood and induced sputum samples 12 following oral and written informed consent (ethical approval obtained from the regional ethical review board in Linköping, #2016/237-31). The study protocol included questionnaires on respiratory and overall health, the evaluation of IGRA-status and sputum samples for DNA extraction.

Induced sputum is a well-tolerated, non-invasive method to collect cells from the surface of the bronchial airways after inhalation of a hypertonic saline solution. The procedure of sputum induction takes approximately 30 minutes and is both cost effective and safe with minimal clinical risks 44 . Sputum specimens were collected as described by Alexis et al 45 , with the following modifications: premedication with an adrenergic β2-agonist, salbutamol (Ventoline, 1ml 1mg/ml) was administrated before the inhalation of hypertonic saline, using a nebulizer (eFlow, PARI). The subsequent steps of sputum processing were adopted from Alexis et al. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in 49,50 packages. The type 1 and type 2 probes were normalized using the quantile normalization method. Using the default setup of the ChAMP package, following probes were filtered out: i) probes below the detection p-value (>0.01), ii) non-CpG probes, iii) multi-hit probes, and iv) all probes of X and Y chromosomes. Cell type heterogeneity was corrected for the PBMC cell types using the Houseman algorithm 51 and batch effects were fixed using ComBat from the SVA package (v3.38.0) 52 Differential methylation analysis were performed with the linear modeling (lmFit) using the limma package (v3.46.0) 53 in a contrast matrix of the TB-exposed and TB-non-exposed (Control) individuals. All Differentially methylated CpGs (DMCs) were considered significant at the Bonferroni-Hochberg (BH) corrected p-value < 0.05 (for HLA-DR cell types), <0.1 (for CD3 cell types) and <0.2 (for PBMC cell types).

Hierarchical clustering of the all TB-exposed and control individuals was performed with the normalized -values obtained after the data filtration in each cell type individually. The distance was calculated using the Euclidean distance matrix. The dendextend (v1.14.0) 54 and ape (v5.4-1) 55 packages in R were used to construct the horizontal hierarchical plots from the three different cell populations using the hclust and dendrogram functions.

The EnhancedVolcano 56 package (v1.8.0) was used to generate the individual volcano plots from all cell populations. The ChromoMap 57 package (v0.3) was used to annotate and visualize the genome-wide chromosomal distribution of the DMGs. The interactive plots were generated using the plotly (v4.9.3) package 58 .

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The heatmaps were generated from the filtered DMGs with their respective CpGs for each cell type using the ComplexHeatmap (v2.6.2) package. The clustering dendrogram in heatmaps were plotted using the Euclidean distance matrix.

We also used Panther database (PantherDB v15) 59 to identify the enriched pathways related to our identified DMGs. In addition, to assess functional enrichment, we used the ReactomePA 

Venn analyses were performed in order to detect the DMGs overlapping between cell populations and between studies. We constructed the Venn diagrams by using matplotlib-venn package (https://github.com/konstantint/matplotlib-venn) using in-house python script. The overlap analyses were calculated and plotted using the go.Sunburst function from plotly using an in-house python script.

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in 

All differences with a p-value < 0.05 were considered significant if not otherwise stated. We calculated family-wise error rate (FWER) using the BH correction method. All analyses were performed in R (v4.0.2) with the mentioned packages.

This study was funded through generous grants from Forskningsrådet Sydöstra Sverige perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Table 3 : Demographic data of the participants in the second recruitment ("test dataset") † The standard deviation of the mean values of age, height, weight and BMI.

All rights reserved. No reuse allowed without permission.

perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in initial subjects (Exp_1-4 and Con_1-6) and the three additional exposed subjects (Exp_6-8).

Purple=exposed, red=TB index case, green=controls. shows the logFC values ranging from -1 (blue) to 1 (red).

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 20, 2021. ;

Droplets, dust and Guinea pigs: An historical review of tuberculosis transmission research, 1878-1940

WHO. The End TB strategy

An interferon-inducible neutrophil-driven blood transcriptional signature in human tuberculosis

Transcriptional Blood Signatures Distinguish Pulmonary Tuberculosis, Pulmonary Sarcoidosis, Pneumonias and Lung Cancers

A blood RNA signature for tuberculosis disease risk: a prospective cohort study

Four-gene pan-African blood signature predicts progression to tuberculosis

Distinct phases of blood gene expression pattern through tuberculosis treatment reflect modulation of the humoral immune response

Trained immunity: A new avenue for tuberculosis vaccine

ChAMP: 450k Chip Analysis Methylation Pipeline

A Network-guided Association Mapping Approach from DNA Methylation to Disease

Surrogate Variable Analysis

limma powers differential expression analyses for RNA-sequencing and microarray studies | Nucleic Acids Research | Oxford Academic. limma powers Differ. Expr. Anal. RNA-sequencing microarray Stud

dendextend: An R package for visualizing, adjusting and comparing trees of hierarchical clustering

Ape 5.0: An environment for modern phylogenetics and evolutionary analyses in R

Publication-ready volcano plots with enhanced colouring and labeling

chromoMap: An R package for Interactive Visualization and Annotation of Chromosomes

Interactive Web-Based Data Visualization with R, plotly, and shiny. Interactive Web-Based Data Visualization with R, plotly, and shiny

PANTHER: A library of protein families and subfamilies indexed by function

ReactomePA: an R/Bioconductor package for reactome pathway analysis and visualization

clusterProfiler: an R Package for Comparing Biological Themes Among Gene Clusters

GOplot: An R package for visually combining expression data with functional analysis

The R Onto-Tools suite

Generally applicable gene set enrichment for pathway analysis

A novel signaling pathway impact analysis

Pathview: an R/Bioconductor package for pathway-based data integration and visualization

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 20, 2021. ; https://doi.org/10.1101/2021.03.16.21253732 doi: medRxiv preprint All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in