Expression quantitative trait loci influence DNA damage-induced apoptosis in cancer Authors: Jessica Bigge, Laura L Koebbe, Ann-Sophie Giel, Dorothea Bornholdt, Benedikt Buerfent, Pouria Dasmeh, Alexander M Zink, Carlo Maj, Johannes Schumacher Published: 2024 ABSTRACT AbstractBackgroundGenomic instability and evading apoptosis are two fundamental hallmarks of cancer and closely linked to DNA damage response (DDR). By analyzing expression quantitative trait loci (eQTL) upon cell stimulation (called exposure eQTL (e2QTL)) it is possible to identify context specific gene regulatory variants and connect them to oncological diseases based on genome-wide association studies (GWAS).ResultsWe isolate CD8+ T cells from 461 healthy donors and stimulate them with high doses of 5 different carcinogens to identify regulatory mechanisms of DNA damage-induced apoptosis. Across all stimuli, we find 5,373 genes to be differentially expressed, with 85% to 99% of these genes being suppressed. While upregulated genes are specific to distinct stimuli, downregulated genes are shared across conditions but exhibit enrichment in biological processes depending on the DNA damage type. Analysis of eQTL reveals 654 regulated genes across conditions. Among them, 47 genes are significant e2QTL, representing a fraction of 4% to 5% per stimulus. To unveil disease relevant genetic variants, we compare eQTL and e2QTL with GWAS risk variants. We identify gene regulatory variants for KLF2, PIP4K2A, GPR160, RPS18, ARL17B and XBP1 that represent risk variants for oncological diseases.ConclusionOur study highlights the relevance of gene regulatory variants influencing DNA damage-induced apoptosis in cancer. The results provide new insights in cellular mechanisms and corresponding genes contributing to inter-individual effects in cancer development.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12864-024-11068-6.Keywords: DNA damage, Apoptosis, Cancer, GWAS, eQTL Background The DNA damage response (DDR) is a complex cellular network that allows an appropriate reaction to different DNA lesions. Depending on the type and severity of DNA lesions, the cellular response ranges from successful repair to the elimination of damaged cells by apoptosis. While repair pathways aim to remove or tolerate damage and preserve cellular functionality, DNA damage-induced apoptosis utilizes cell death as a protective mechanism to prevent cells becoming dysfunctional or malignant [1–3]. This mechanism is particularly relevant in oncological diseases, as cancer cells are characterized by genomic instability and evasion of apoptosis [4]. Dysregulation of DNA damage-induced apoptosis affects carcinogenesis, cancer maintenance, therapy and resistance [5–7]. Thus, genes involved in this pathway are frequently mutated in cancer cells and alterations in apoptosis genes particularly correlate with a lower overall survival compared to alterations in repair genes [8, 9]. Since the majority of oncological diseases are multifactorial, their genetic architecture is defined by various common single nucleotide polymorphisms (SNPs), each with a small effect size [10, 11]. Genome-wide association studies (GWAS) have identified numerous SNPs associated with different cancer entities [12, 13]. As these SNPs are predominantly located in non-coding regions, they affect the regulation of gene expression rather than altering the gene itself [14, 15]. By integrating genotype with expression data, expression quantitative trait loci (eQTL) analyses facilitate the functional characterization of regulatory variants [16–19]. To account for exogenous factors that influence regulatory processes under pathophysiological conditions, the analysis of exposure eQTL (e2QTL) allows for the identification of context-specific effects and enables the explanation of previously undetected mechanisms [20–23]. With regard to the ability of damaged cells to evade apoptosis, previous studies have shown that DNA-damaging agents provoke apoptosis by transcriptionally distinct responses and are accompanied by a selective downregulation of genes, including DNA repair genes [24–26]. In addition, many DDR proteins play a dual role in repair and apoptosis, which is mediated by posttranslational modifications [27, 28]. However, it is still unclear how the decision to repair lesions or commit apoptosis is made and how these processes are regulated. Thus, the analysis of e2QTL in the context of DNA damage-induced apoptosis is fundamental for understanding pathomechanisms and developing new prevention and treatment strategies. In this study, we analyze the cellular mechanisms of DNA damage-induced apoptosis with specific interest in how inter-individual genetic variability influences their regulation. Cytotoxic CD8+ T cells are chosen because monocytes as well as resting T cells are impaired in DNA repair [29, 30]. We stimulate these cells with high doses of 5 different DNA-damaging agents to cover the broad spectrum of responses to DNA damage-induced apoptosis. Chemical alterations of single bases are induced using the alkylating agent Methyl-methanesulfonate (MMS) [31] and the reactive oxygen species tert-butyl-hydroperoxide (TBOOH) [32]. Furthermore, we utilize benzo(a)pyrene-7,8-diol-9,10-epoxide (BPDE) [33] and ultraviolet (UVC) radiation [34] to cause bulky lesions like adducts or dimers. The preferred repair pathway for the former chemical lesions (MMS, TBOOH) is base excision repair (BER), whereas the latter bulky lesions (BPDE, UVC) predominantly address nucleotide excision repair (NER) [35]. Additionally, we induce a third kind of DNA damage by stimulating cells with 4-hydroxycyclophosphamide (HC) [36]. The resulting DNA crosslinks are repaired by a complex interplay of different pathways, including excision-based repair mechanisms, homologous recombination (HR) and the Fanconi anaemia (FA) pathway [37]. Our results exhibit regulatory mechanisms in the context of DNA damage-induced apoptosis ranging from a general inhibition of transcription to eQTL and e2QTL that act in a stimulus specific manner. Finally, we test whether eQTL and e2QTL represent risk variants from GWAS for multifactorial diseases and, thus, provide further insights into the pathomechanisms of corresponding phenotypes. Methods Sample collection, isolation and stimulation of CD8+ T cells The study includes blood samples from 461 healthy European donors (non-smoker, no infection or vaccination 4 weeks prior to blood withdrawal, white blood count (WBC) < 10 × 103/ μl). We isolated peripheral blood mononuclear cells (PBMCs) of all participants by density gradient centrifugation using Pancoll (PAN biotech). CD8+ T cells were isolated by magnetic cell separation using CD8 MicroBeads (Milteny Biotec) according to manufacturer’s instructions. Cell purity > 95% was evaluated based on FACS analysis (CytoFLEX, Beckmann Coulter Life Science) with an antibody against CD8 (Milteny Biotec). Cells were cultured in TexMACS GMP medium (Milteny Biotec) supplemented with 50 µg/ml Gentamycin (Gibco) in a density of 2,000,000 cells in 400 μl overnight. Stimulation of cells was performed the next day by dividing cells of each participant into six subsets that were either left untreated or treated with 5 µM benzo(a)pyrene-7,8-diol-9,10-epoxide (Toronto Research Chemicals), 0.5 mM 4-hydroxycyclophosphamide (Cayman Chemical), 0.5 mM tert-butyl-hydroperoxide (Sigma-Aldrich), 1 nM Methyl-methanesulfonate (Sigma-Aldrich) for 6 h or irradiated with 50 mJ/ cm3 UVC (Stratalinker 2400, Stratagene). Dose and duration of stimulation were determined based on the highest number of differentially expressed genes (DEGs) (Additional file 1: Table S1). Cell viability and induced apoptosis were assessed by Annexin V/ PI assay (Biozol) and Trypan Blue Exclusion Test (Invitrogen). After treatment cells were lysed using RLT reagent (Qiagen) and stored at -80 °C. DNA and RNA isolation DNA of untreated CD8+ T cells and RNA of untreated and all stimulated CD8+ T cells were isolated using the AllPrep 96 DNA/RNA Kit from Qiagen. Quantity of nucleic acids was determined using a microplate reader (Varioskan LUX, Thermo Fisher Scientific) and appropriate Quant-iT kits (Invitrogen). Quality was determined by electrophoresis (TapeStation, Agilent Technologies) for a subset of samples. Genotyping and imputation Participants were genotyped using DNA of untreated CD8+ T cells. Genotyping was conducted on Infinium Global Screening Array v3.0 (Illumina) and raw data was processed using the GenomeStudio Software (Illumina). Quality control (QC) of generated PLINK [38] files included minor allele frequency (MAF) > 5%, call rate > 97% and Hardy–Weinberg equilibrium (HWE) (PHWE > 10–4) as well as statistics for sex, kinship, heterozygosity and population stratification based on 1000 Genomes reference data as recommended [39–41]. A total of 482,620 SNPs were used as input for the TOPMed Imputation Server [42–44]. Gene expression analysis RNA of untreated and stimulated CD8+ T cells was used for library preparation. Libraries were generated using QuantSeq FWD (Lexogen) with unique dual indices according to manufacturer’s instructions and sequenced on NovaSeq6000 (Illumina) with a read length of 100 bp and 10 million reads per sample. Raw data was processed using BBDuk [45] for adapter trimming and STAR [46] for alignment. QC included FastQC [47], RSeQC [48] and MultiQC [49]. An additional quality control step was added based on expression data by performing principal component (PC) analysis of control samples together with a single stimulus. We used the Gaussian mixture model of the scikit-learn library [50] to predict clusters based on the first two PCs using k-means clustering algorithm for initialization. After removing identified outliers, the final dataset consisted of 2,728 samples and 60,617 genes for further analysis. Differential expression analysis was performed for each stimulus separately as described by Law et al. [51] using edgeR [52] and limma [53]. Multiple testing correction was performed using the Benjamini–Hochberg method [54]. eQTL and e2QTL analysis We used FastQTL [55] to analyze cis-eQTL within a 1 megabase window of a gene’s transcription start site. Filtering and normalization of expression data was performed as described by the Genotype Tissue Expression (GTEx) project [56] including 60 PEER factors, top 3 genotype PCs and sex as covariates. Genotypes were filtered by PHWE > 10–6 and MAF > 5%. Adjusted p-values were generated using 1,000 to 10,000 permutations. Based on our eQTL data, we determined the most significant variant for each analyzed gene (referred to as top eQTL) to calculate e2QTL in a z-test and corrected for multiple testing using Bonferroni correction as described by Kim-Hellmuth et al. [23]. Enrichment Enrichment of Gene Ontology (GO) terms [57, 58] as well as Gene Set Enrichment Analysis (GSEA) [59, 60] were performed using clusterProfiler [61]. GO terms only included biological processes and GSEA was performed for hallmark gene sets of the Molecular Signatures Database (MSigDB) [62] with all parameters set to default. Categories with a q-value < 0.05 were reported as significantly enriched. Trait association and colocalization Associations of eQTL and genome-wide significant associated GWAS variants were assessed using the Open Targets Platform [63, 64]. GWAS variants had to be a significant eQTL (qval < 0.05) and in strong linkage disequilibrium (LD) (r2 > 0.7) to the top eQTL variant to be reported. LD was calculated using LDlink [65]. If GWAS summary statistics were available, colocalization analysis was performed using coloc [23] with all parameters set to default. Results Study setup and expression profiling of CD8+ T cells upon carcinogen treatment To examine regulatory processes of DNA damage-induced apoptosis, we isolated CD8+ T cells from blood samples obtained from 461 healthy donors. We stimulated cells for 6 hours (h) using high doses of 5 different DNA-damaging agents to induce early apoptosis. As the DDR is specific to different types of lesions, we selected MMS, TBOOH, UVC, BPDE, and HC to address a broad spectrum of responses. Transcriptional analysis was performed of all stimulated cells as well as untreated controls by RNA-Sequencing. In addition, all donors were genotyped genome-wide by analyzing the untreated controls on Global Screening Array v3.0 (Fig. 1a). Study setup and expression profiling of CD8+ T cells upon carcinogen treatment. a Overview of the study design. Blood samples from 461 healthy donors were used to isolate CD8+ T cells. Cells were either left untreated or stimulated with different carcinogens for 6 h. Expression analysis and genotype profiling were performed to identify regulatory variants. b Number of significantly downregulated (red) and upregulated (blue) genes (adj. p < 0.05 and abs(logFC) > 1.5) identified for each stimulus. c Distribution of the logFC for significant DEG (adj. p < 0.05 and abs(logFC) > 1.5) in each stimulus. Genes were classified as specific (significant only in one stimulus) and common genes (significant in two or more stimuli). d Clustered heatmap of significantly enriched GO terms for biological processes of downregulated genes. Hierarchical clustering was performed using complete-linkage clustering and Euclidean distances. BPDE—benzo(a)pyrene-7,8-diol-9,10-epoxide, HC—4-hydroxycyclophosphamide, MMS—Methyl-methanesulfonate, TBOOH—tert-butyl-hydroperoxide, UVC—ultraviolet radiation Expression profiling revealed a total of 5,373 differentially expressed genes (DEG) across all stimuli (adj. p < 0.05 and abs(logFC) > 1.5) (Additional file 1: Table S2). We found a high number of repressed genes in all conditions ranging from 85% in HC to 99% in UVC (Fig. 1b, Additional File 2: Figure S1a). To exclude the observed repression is due to RNA degradation, we performed extensive quality control steps (see Methods). In addition, we evaluated the count distribution for CD8A, CD8B and the reference genes GAPDH and UBE2D2 and did not find a general reduction in the expression of these genes upon stimulation (Additional file 2: Figure S1b). These results were in line with differential expression analysis showing no significant downregulation of the genes (logFC > -1.5). Our results were consistent with prior research indicating an inhibition of transcription and induced apoptosis upon exposure to high dose DNA-damaging agents [26]. Notably, data of a previous study by Lu et al. [25] analyzing expression response of epithelial cells to high dose 0.5 μM BPDE matched our results regarding the direction of fold change (FC). Based on our significant DEG we found 414 common genes in both datasets with 99% indicating the same FC direction (Additional File 1: Table S3). To analyze expression patterns with respect to regulatory differences across stimuli, we categorized DEG into common genes, that were significant across all stimuli, and specific genes, that were significant in only one stimulus. Common DEG were predominantly repressed, while specific DEG exhibited induction or repression (Fig. 1c). Given the general inhibition of transcription, we conducted a grouped enrichment analysis of Gene Ontology (GO) terms for biological processes [57, 58] using all significantly downregulated genes (Additional file 1: Tab S4). To evaluate stimulus-specific differences, we determined the 10 most significant terms per stimulus and performed hierarchical cluster analysis. Subclusters were identified for stimuli inducing chemical alterations (TBOOH and MMS) or bulky lesions (BPDE and UVC). While genes downregulated upon treatment with TBOOH or MMS showed enrichment for terms associated with autophagy, organelle organization or transport, genes downregulated upon treatment with BPDE or UVC predominantly showed enrichment for terms associated with protein modifications. Of note, genes downregulated upon treatment with HC only exhibited significant enrichment for GO terms related to RNA metabolism (Fig. 1d, Additional file 1: Table S5). Due to the small number of upregulated genes and their tendency to be stimulus specific, only 8 common induced genes were identified, including the spliceosomal RNA U2 and the snRNAs RNU4-2 and SNORD3A (Additional file 2: Figure S1c). Gene set enrichment analysis (GSEA) [59, 60] for all induced genes determined apoptosis (q = 1.22e−4), p53 signaling (q = 1.22e−4), oxidative phosphorylation (q = 2.17e−2) and E2F targets (q = 4.6e−2) as significant hallmarks. However, it should be noted that these results mainly rely on genes induced by BPDE and HC, because stimulation with all other exposures exhibited no significant enrichment due to the smaller number of upregulated genes (Additional file 1: Table S6, Additional file 2: Figure S1d). Identification of eQTL and e2QTL in the context of DNA damage-induced apoptosis In order to assess how genetic variability influences DNA damage-induced apoptosis based on gene expression, we performed eQTL and e2QTL analyses. Here, we focused on cis-eQTL that were defined as eQTL within a 1 MB window of a gene’s transcription start site. All eQTL were calculated using FastQTL [55] and e2QTL were defined by comparing the regression slope of the most significant variant for each analyzed gene (referred to as top eQTL) in unstimulated and stimulated cells in a z-test as described in Kim-Hellmuth et al. [23]. Across all conditions, we found 654 genes (called eGenes) that were significantly regulated by eQTL (qval < 0.05) (Additional file 1: Table S7). Of these, 49% were specific for one stimulus and 20% were shared across all stimuli. Especially cells treated with HC showed a high fraction of stimulus specific eGenes (Additional file 2: Figure S2a). For 47 of all identified eGenes, we found additional e2QTL effects due to differences in genetic regulation upon stimulation (Additional file 1: Table S8), which represented a fraction of 4% to 5% of eGenes per stimulus (Fig. 2a). In order to evaluate our eQTL results with regard to DEG, 61% to 69% of eGenes were differentially expressed including the identified e2QTL. Noteworthy, 550 of 654 identified eGenes were found in unstimulated CD8+ T cells, with smaller numbers reaching from 409 to 271 eGenes for stimulated T cells. To understand why the number of eQTL dropped upon stimulation, we compared effect sizes of eQTL before and after stimulation and found a reduction of slopes in > 60% of eQTL in all stimuli (Additional file 2: Figure S2b). In addition, we observed an inverse correlation of eGenes and DEG (Fig. 2b) indicating a reduced genetic effect on gene regulation upon stimulation. This tendency was previously described in a single cell eQTL study of immune cells [66]. Identification of eQTL and e2QTL in the context of DNA damage-induced apoptosis. a Number of eGenes with at least one significant eQTL separated by stimulus. The fraction of eGenes that are significant DEGs (blue) or significant e2QTL (red) is indicated. eQTL under unstimulated condition are referred to as control (CTRL). b Comparison of the number of significant DEGs (blue) and eGenes (green) for each stimulus. c Scatter plot of eQTL effect size beta for eQTL present in both GTEx whole blood and our untreated CD8+ T cell data (referred to as CTRL). eQTL with identical direction of effect size are shown in grey and eQTL with opposite effect size are shown in black. eQTL of SMPD4 are highlighted in red. d Enrichment analysis of eGenes for GO terms associated with biological processes separated by stimulus. The gene ratio is indicated by the size and the q-value by the color of dots. Numbers under each stimulus indicate the number of eGenes included in the analysis. eGenes under unstimulated condition are referred to as control (CTRL). BPDE—benzo(a)pyrene-7,8-diol-9,10-epoxide, HC—4-hydroxycyclophosphamide, MMS—Methyl-methanesulfonate, TBOOH—tert-butyl-hydroperoxide, UVC—ultraviolet radiation As we used CD8+ T cells for our study, we were interested in the cell type specificity of the identified eGenes. Therefore, we compared the effect sizes of our eQTL under basis condition to whole blood eQTL data of the Genotype Tissue Expression project (GTEx) [56]. By analyzing all eQTL present in both datasets the Spearman coefficient for the correlation of effect sizes was 0.73. While the direction of the regression slopes was identical for most of the eQTL, we found 38 eGenes with a consistently opposite slope in GTEx and our data (Fig. 2c, Additional file 1: Table S9). Of these, SMPD4 (NSMASE3) was reported as e2QTL in our study with an absolute difference of slopes to GTEx reaching from 0.44 to 1.12. Furthermore, we conducted an enrichment analysis of GO terms for biological processes using all eGenes. We found leukocyte and T cell specific processes to be enriched but also observed stimulus specific differences. While cells treated with HC exhibited the highest number of eGenes upon stimulation, the enrichment of GO terms was more similar to untreated cells compared to other conditions. Moreover, stimuli addressing BER (MMS, TBOOH) likewise lacked enrichment of single leukocyte specific processes that were still significant enriched for stimuli addressing NER (BPDE, UVC) (Fig. 2d, Additional file 1: Table S10). Trait-association of eQTL and e2QTL based on GWAS data Since our results have already illustrated regulatory characteristics of DNA damage-induced apoptosis, we completed our analyses by evaluating the role of eQTL and e2QTL variants for multifactorial diseases. We utilized the Open Targets Platform [63, 64] and tested whether risk variants that have been identified by GWAS for multifactorial diseases represent top eQTL or e2QTL. Here, we filtered for genome-wide significant associated GWAS variants that represent significant eQTL or e2QTL in our study (qval < 0.05) and are in strong linkage disequilibrium (LD, r2 > 0.7) to the top eQTL or e2QTL. In addition to all significant e2QTL, we only included eQTL with a strong differential expression (adj. p < 0.05 and abs(logFC) > 2) to ensure that the eQTL effects are in context with the stimulation and induced DDR. Overall, we identified 3 eQTL and 8 e2QTL referring to 7 different eGenes that represent GWAS risk variants for 5 oncological and 2 neuropsychiatric diseases (Table 1). Table 1eQTL and e2QTL that represent GWAS risk variants for multifactorial diseasesGeneStimulusTraitGWAS varianteQTL / e2QTL variantbetaq-valueLD r2XBP1TBOOHBreast Cancerrs5997389 (G/A)rs3788408 (A/G)0.514.77E-040.802TBOOHOvarian Cancerrs6005807 (T/C)rs3788408 (A/G)0.514.77E-040.73HCOvarian Cancerrs6005807 (T/C)chr22:28639216:A:ATAAG*0.541.37E-060.746PLECBPDEBipolar Disorderrs6992333 (A/G)rs7822511 (T/C)-0.351.77E-040.852Multiple Sclerosisrs3923387 (C/T)rs7822511 (T/C)-0.351.77E-040.804KLF2MMSMultiple Myelomars11086029 (T/A)rs3745318 (T/C)-0.426.94E-150.833BPDEMultiple Myelomars11086029 (T/A)rs2362475 (A/C)-0.435.53E-180.983UVCMultiple Myelomars11086029 (T/A)rs7246538 (A/G)-0.439.00E-150.907TBOOHMultiple Myelomars11086029 (T/A)rs4808046 (G/A)-0.397.94E-190.983PIP4K2AMMSB cell ALLrs2296624 (C/T)rs10430590 (A/T)-0.321.36E-030.761GPR160TBOOHTesticular germ cell tumorrs3755605 (C/T)rs6778888 (C/T)0.393.47E-020.824RPS18HCBreast Cancerrs17215231 (C/T)rs17215231 (C/T)-0.806.86E-101ARL17BUVCBreast Cancerrs199533 (G/A)rs141783865 (A/C)0.842.60E-270.91BPDE benzo(a)pyrene-7,8-diol-9,10-epoxide, HC – 4 hydroxycyclophosphamide, MMS Methyl-methanesulfonate, TBOOH tert-butyl-hydroperoxide, UVC ultraviolet radiation, ALL acute lymphoblastic leukaemiaEffects of XBP1 and PLEC refer to eQTL (upper part of the table) and effects of KLF2, PIP4K2A, GPR160, RPS18 and ARL17B refer to e2QTL (lower part of the table). Effect alleles are highlighted in bold. Variants are indicated by an asterisk (*) if rs ID is not available. Effect size beta and q-value refer to the eQTL effect upon stimulation or e2QTL effect, respectively eQTL and e2QTL that represent GWAS risk variants for multifactorial diseases BPDE benzo(a)pyrene-7,8-diol-9,10-epoxide, HC – 4 hydroxycyclophosphamide, MMS Methyl-methanesulfonate, TBOOH tert-butyl-hydroperoxide, UVC ultraviolet radiation, ALL acute lymphoblastic leukaemia Effects of XBP1 and PLEC refer to eQTL (upper part of the table) and effects of KLF2, PIP4K2A, GPR160, RPS18 and ARL17B refer to e2QTL (lower part of the table). Effect alleles are highlighted in bold. Variants are indicated by an asterisk (*) if rs ID is not available. Effect size beta and q-value refer to the eQTL effect upon stimulation or e2QTL effect, respectively Of the disease related eQTL (Additional file 1: Table S11), both eGenes represent significant downregulated DEG and their disease association is specific to distinct stimuli. The top eQTL for PLEC upon BPDE exposure is in strong LD with GWAS risk variants for bipolar disorder [67] (r2 = 0.852) and multiple sclerosis [68] (r2 = 0.804). Carriers of the GWAS risk alleles exhibit decreased expression of PLEC in both traits (Table 1). The top eQTL for XBP1 represents a GWAS risk variant for breast cancer [69] upon TBOOH treatment (r2 = 0.802) and for ovarian cancer [70] upon TBOOH treatment (r2 = 0.73) and HC treatment (r2 = 0.746). The GWAS risk allele associated with breast cancer leads to an increased XBP1 expression, while the opposite effect is observed for the GWAS risk alleles associated with ovarian cancer (Table 1, Fig. 3a). Trait association of eQTL and e2QTL based on GWAS data. a Violin plots of the eQTL for XBP1 in breast cancer upon TBOOH treatment (1), ovarian cancer upon TBOOH treatment (2) and ovarian cancer upon HC treatment (3). Genomic variants on the x-axis represent GWAS lead variants for the corresponding trait. Individuals are indicated by small ticks and horizontal lines represent the median as well as the 25% and 75% quartiles in each violin. The dashed lines indicate the linear regression for the effect of the genotype on gene expression. b Violin plots of the e2QTL for KLF2 in untreated cells (CTRL, blue) and upon BPDE treatment (red). The genomic variant on the x-axis represents the GWAS lead variant for MM. c Regional association plots for common variants in our eQTL data upon BPDE treatment and the GWAS for MM. The lead variant of the MM GWAS is highlighted in black and annotated, respectively. BPDE—benzo(a)pyrene-7,8-diol-9,10-epoxide, HC—4-hydroxycyclophosphamide, MM – Multiple Myeloma, TBOOH—tert-butyl-hydroperoxide Of all GWAS related e2QTL (Additional file 1: Table S12), only KLF2 represents a significant downregulated DEG and the disease associations are present across all BER and NER associated stimuli, whereas all other eGenes show not differential expression but are stimulus specific. We found that the e2QTL for ARL17B upon UVC and RPS18 upon HC exposure represent GWAS risk variants for breast cancer [71, 72] (r2 = 0.91, r2 = 1), both exhibiting reduced expression in carriers of the risk alleles (Table 1). Of note, RPS18 is the only eGene exhibiting a stronger eQTL effect upon stimulation and the same variant for the GWAS and the top e2QTL (Additional file 2: Figure S3a). Furthermore, the top e2QTL for GPR160 upon TBOOH exposure represents a GWAS risk variant for testicular germ cell tumors [73] (r2 = 0.824) and the e2QTL for PIP4K2A upon MMS exposure represents a GWAS risk variant for B cell acute lymphoblastic leukaemia (ALL) [74] (r2 = 0.761). Both findings exhibit increased gene expression for carriers of the GWAS risk alleles and reduced eQTL effect sizes upon stimulation (Table 1). We were particularly interested in the e2QTL for KLF2 that represent a GWAS risk variant for multiple myeloma (MM) [75] and are present among all stimuli associated to BER and NER (r2 = 0.833 – 0.983). The e2QTL for KLF2 exhibit the strongest difference of slopes before and after treatment in our analysis (abs(∆slope) = 0.7 – 0.75). For all conditions we observed an increased expression of KLF2 in carriers of the GWAS risk alleles and reduced eQTL effect sizes upon stimulation that matches the observed DEG effect (Table 1, Fig. 3b, Additional file 2: Figure S3b). As summary statistics for the MM GWAS were available, we additionally tested whether the GWAS and our e2QTL data are likely to share a causal variant at the KLF2 locus using coloc [76]. We found PP.H4 > 0.993 and PP.H4/PP.H3 > 143 in all conditions confirming strong evidence of sharing the same causal variant in our e2QTL analysis and the MM GWAS (Fig. 3c, Additional file 2: Figure S3b). Altogether, our results demonstrate the relevance of regulatory variants in the context of DNA damage-induced apoptosis and highlight their role in oncological diseases. Discussion In this study, we analyzed regulatory mechanisms in the context of DNA damage-induced apoptosis with a particular interest in eQTL and e2QTL effects and their role for multifactorial diseases. Even though genomic instability and evading apoptosis are known hallmarks for various diseases, less is known about the inter-individual genetic influence on the underlying biological processes. Our results offer new insights into the regulation of DNA damage-induced apoptosis as well as provide genes and cellular mechanisms that might be relevant for future research. Consistent with previous studies, differential expression analysis revealed a general inhibition of transcription with no upregulation of specific repair pathways. Our results further support the assumption that apoptosis induced by high dose DNA-damaging agents does not lead to a general shutdown of cellular processes but is accompanied by a selective downregulation of genes [24, 25]. Due to the high number of suppressed genes, we had a particular interest in the upregulated genes. Except for the spliceosomal RNA U2, the majority of commonly induced genes is poorly characterized, but at least for the snRNAs RNU4-2 and SNORD3A, cancer-related functionalities were reported previously [77–79]. Especially SNORD3A differs from other upregulated genes by a conspicuous strong induction (logFC > 6) across all conditions, emphasizing the role of regulatory RNAs in the context of DNA damage-induced apoptosis. Moreover, other genes upregulated in distinct but also specific stimuli like PPA1 [80, 81], GADD45B [82], C1D [83], TRPV2 [84] and RPS14 [85, 86] were associated with DDR, cancer and apoptosis. Therefore, our results reflect known characteristics of DNA damage-induced apoptosis and underline their relevance in oncological disease but also highlight the importance of still unexplained genes and regulatory networks. To evaluate how inter-individual genetic variability influences DNA damage-induced apoptosis, we analyzed eQTL and e2QTL. On the one hand, smaller effect sizes of eQTL upon stimulation as well as the inverse correlation of eGenes and DEG support earlier hypotheses of a restricted influence of gene regulatory variants in the cellular response to external stimuli [66]. On the other hand, cell type and context specificity enable to find new and divergent gene regulatory mechanisms, as seen for SMPD4, that exhibited opposite eQTL compared to the results of the GTEx project [56] and showed an additional e2QTL effect in our study. SMPD4 is involved in DNA damage-induced apoptosis and dysregulated in several tumors [87], but neither cell type nor context specific expression effects have been described before. Our findings, thus, highlight the specificity of gene regulatory variants in certain biological processes and the need for context specific analyses to explain gene regulatory mechanisms with regard to exogenous factors. Another goal of our study was to identify eGenes that are associated to multifactorial diseases in order to gain new insights into cellular pathomechanisms. As expected, the majority of associated diseases were oncological and only one eGene was found for neuropsychiatric diseases. Although DDR and apoptosis have already been described as risk factors for multiple sclerosis and bipolar disorder [88, 89], the associated gene PLEC has not been described in the context of these diseases before. PLEC encodes a cytolinker protein that is relevant for the structural scaffold as well as signaling of cells, including astrocytes, that are known to play an important role in both traits [90–92]. Of the cancer related eGenes, XBP1 is relevant for different processes in tumors due to its role in unfolded protein response (UPR). Moreover, its responsiveness to estrogen highlights the importance in estrogen-dependent cancers like breast and ovarian cancer [93]. In line with our results, carriers of the GWAS risk alleles exhibit an increased expression in triple-negative breast cancer (TNBC) cells [94], while ovarian cancer cells show a decreased XBP1 expression [95]. In addition, context specific regulatory effects for ARL17B and RPS18 were found to be associated with breast cancer. The cellular function of both genes is poorly understood. RPS18 was not differentially expressed (p.adj = 0.99), but its strong genotype dependent regulation upon stimulation should be considered regarding its usage as housekeeper under certain conditions [96]. Moreover, we found an increased expression of GPR160 and PIP4K2A among carriers of GWAS risk variants for testicular germ cell tumors and B cell ALL. These results further support previous studies reporting high expression of GPR160 and PIP4K2A in corresponding cancer cells that undergo apoptosis after knockdown of the corresponding genes in both cases [97, 98]. Together, these data indicate that DNA damaged cells from carriers of GPR160 and PIP4K2A risk alleles might be at higher risk to evade apoptosis. Thus, both context specific targets may help to improve therapy and overcome resistances of both cancer entities in future. At least, we found that regulatory effects of KLF2 are related to MM across BER and NER associated stimuli. KLF2 is a transcription factor involved in various biological processes and is discussed as tumor suppressor for several cancer entities [99, 100]. MM cells are characterized by an increased KLF2 expression and knockdown of KLF2 leads to apoptosis, which further highlights the relevance of KLF2 in MM maintenance [101]. The findings are in line with our e2QTL effects which show that carriers of the MM risk allele exhibit a stronger KLF2 downregulation than carriers of the opposite allele. Furthermore, the fact that all conditions except from HC showed e2QTL effects, indicates that single base modifications that induce BER and NER are more likely to contribute to KLF2 related MM pathophysiology. Despite the promising results of this analysis, our study has limitations. We used CD8+ T cells as model cell type to study the influence of genetic variability on DNA damage-induced apoptosis. However, we cannot determine to what extent our data on apoptosis can be transferred to other cell types. In addition, all studies were performed in European individuals. It is, thus, unclear to what extent the results can be transferred to non-European populations. Conclusions This analysis provides new insights into the context specific regulation of gene expression by genetic variants. As evading apoptosis is a key hallmark of cancer, particularly the regulation of DNA damage-induced apoptosis is crucial to understand the pathophysiology of oncological diseases. Our study reveals individual regulatory mechanisms and corresponding genes that promote apoptosis. The findings might represent promising targets for future research in the field of apoptosis related diseases. Supplementary Information Additional file 1: Master file for Tables S1-S12. (Excel https://www.microsoft.com/de-de/microsoft-365/excel?market=de). Additional file 2: Supplementary figures. (pdf, Acrobat Reader https://www.adobe.com/de/acrobat.html). Acknowledgements Not applicable. Abbreviations Acute lymphoblastic leukaemia Base excision repair Benzo(a)pyrene-7,8-diol-9,10-epoxide DNA damage response Differentially expressed genes Exposure expression quantitative trait loci Expression quantitative trait loci Fanconi anaemia Fold change Gene Ontology Gene Set Enrichment Analysis Genotype Tissue Expression Genome-wide association study 4-Hydroxycyclophosphamide Homologous recombination Hardy-Weinberg equilibrium Linkage disequilibrium Minor allele frequency Multiple Myeloma Methyl-methanesulfonate Molecular Signature DataBase Nucleotide excision repair Peripheral blood mononuclear cells Principal component Quality control Single nucleotide polymorphism Tert-butyl-hydroperoxide Triple negative breast cancer Unfolded protein response Ultraviolet radiation White blood count Authors’ contributions JS and BB conceived the project. DB collected samples and stimulated them. JB prepared sequencing libraries, performed bioinformatic analysis with the help of LLK, AG, PD, AMZ and CM, and wrote the manuscript with contribution of JS. All authors read and approved the final manuscript. Funding Open Access funding enabled and organized by Projekt DEAL. This work was funded by intramural funding of the University of Marburg and Inventum Genetics GmbH. Data availability The raw data of this study are available from Inventum Genetics GmbH but restrictions apply to the availability of these data, which were used under license for the current study. Data are however available from the authors upon reasonable request and with permission of Inventum Genetics GmbH. Data has been deposited at the European Genome-phenome Archive (EGA) under accession number EGAS50000000666. Processed data that support the findings is either included in the published article and its supplementary information files or publicly available at the Gene Ontology knowledgebase [57, 58], the Genotype-Tissue Expression project [56], the Molecular Signatures Database [62] or the Open Targets Platform [63, 64]. Declarations Ethics approval and consent to participate The study was approved by the ethics committee of the University of Marburg and complied with the Helsinki Declaration. All participants gave written informed consent. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Additional file 1: Master file for Tables S1-S12. (Excel https://www.microsoft.com/de-de/microsoft-365/excel?market=de). Additional file 2: Supplementary figures. (pdf, Acrobat Reader https://www.adobe.com/de/acrobat.html). Data Availability Statement The raw data of this study are available from Inventum Genetics GmbH but restrictions apply to the availability of these data, which were used under license for the current study. Data are however available from the authors upon reasonable request and with permission of Inventum Genetics GmbH. Data has been deposited at the European Genome-phenome Archive (EGA) under accession number EGAS50000000666. Processed data that support the findings is either included in the published article and its supplementary information files or publicly available at the Gene Ontology knowledgebase [57, 58], the Genotype-Tissue Expression project [56], the Molecular Signatures Database [62] or the Open Targets Platform [63, 64].