Archives of Toxicology

, Volume 91, Issue 5, pp 2067–2078 | Cite as

Transcriptomics and methylomics of CD4-positive T cells in arsenic-exposed women

  • Karin Engström
  • Tomasz K. WojdaczEmail author
  • Francesco Marabita
  • Philip Ewels
  • Max Käller
  • Francesco Vezzi
  • Nicola Prezza
  • Joel Gruselius
  • Marie Vahter
  • Karin BrobergEmail author
Open Access
Inorganic Compounds


Arsenic, a carcinogen with immunotoxic effects, is a common contaminant of drinking water and certain food worldwide. We hypothesized that chronic arsenic exposure alters gene expression, potentially by altering DNA methylation of genes encoding central components of the immune system. We therefore analyzed the transcriptomes (by RNA sequencing) and methylomes (by target-enrichment next-generation sequencing) of primary CD4-positive T cells from matched groups of four women each in the Argentinean Andes, with fivefold differences in urinary arsenic concentrations (median concentrations of urinary arsenic in the lower- and high-arsenic groups: 65 and 276 μg/l, respectively). Arsenic exposure was associated with genome-wide alterations of gene expression; principal component analysis indicated that the exposure explained 53% of the variance in gene expression among the top variable genes and 19% of 28,351 genes were differentially expressed (false discovery rate <0.05) between the exposure groups. Key genes regulating the immune system, such as tumor necrosis factor alpha and interferon gamma, as well as genes related to the NF-kappa-beta complex, were significantly downregulated in the high-arsenic group. Arsenic exposure was associated with genome-wide DNA methylation; the high-arsenic group had 3% points higher genome-wide full methylation (>80% methylation) than the lower-arsenic group. Differentially methylated regions that were hyper-methylated in the high-arsenic group showed enrichment for immune-related gene ontologies that constitute the basic functions of CD4-positive T cells, such as isotype switching and lymphocyte activation and differentiation. In conclusion, chronic arsenic exposure from drinking water was related to changes in the transcriptome and methylome of CD4-positive T cells, both genome wide and in specific genes, supporting the hypothesis that arsenic causes immunotoxicity by interfering with gene expression and regulation.


Arsenic Transcriptomics Methylomics CD4 cells Immune system Immunotoxic 


Inorganic arsenic occurs in drinking water worldwide and in certain foods, mainly rice-based products. Long-term exposure to inorganic arsenic is a strong risk factor for different types of cancers as well as numerous non-cancer conditions, such as cardiovascular and lung diseases (IARC 2012). More recently, studies have identified immunotoxic effects of arsenic. For example, studies on newborns and children found that exposure to arsenic during pregnancy was associated with immunosuppressive effects, such as decreased thymus size and impaired thymic function, altered cytokine expression, and increased susceptibility to infections (Moore et al. 2009; Raqib et al. 2009; Ahmed et al. 2011, 2012, 2014; Rahman et al. 2011; Farzan et al. 2016). Arsenic-exposed adults and children had reduced proliferation of lymphocytes, fewer CD4-positive T cells, altered CD4 + :CD8 + T cell ratio, and fewer T-regulatory cells (Biswas et al. 2008; Hernandez-Castro et al. 2009; Kile et al. 2014; Nadeau et al. 2014). A few studies have reported that arsenic exposure may also increase the risk of allergies and autoimmune diseases (Tseng 2004; Soto-Pena et al. 2006). These findings suggest that arsenic exposure could result in both immune suppression and immune stimulation.

The mechanisms behind the immunotoxic effects of arsenic remain unclear (Ahmed et al. 2014; Ferrario et al. 2016). One potential mechanism is that arsenic could interfere with the expression of immune-related genes. This could result in failure to maintain homeostasis in the immune response to foreign antigens, potentially leading to immune dysfunction and immune-related diseases. In support of such a mechanism, microarray-based epidemiological studies have shown that arsenic exposure is associated with altered expression of immune-related genes in whole blood (Argos et al. 2006; Andrew et al. 2008; Salgado-Bustamante et al. 2010). The mode of action by which arsenic changes gene expression could, as recently indicated, occur by disturbing epigenetic processes that regulate gene expression, such as DNA methylation (Pilsner et al. 2012; Koestler et al. 2013; Broberg et al. 2014; Kile et al. 2014; Cardenas et al. 2015; Rojas et al. 2015).

We hypothesized that chronic arsenic exposure affects gene expression via alteration of DNA methylation of central components of the immune system. To evaluate this, we performed next-generation sequencing (NGS) of the transcriptome (by RNA sequencing) and the methylome (by target-enrichment NGS) of primary CD4-positive T cells, key immune cells in the defense against infections, in individuals chronically exposed to inorganic arsenic via drinking water.


Study participants, CD4-positive cell enrichment, and measurement of arsenic

This study involved women living in San Antonio de los Cobres and surrounding villages at around 4000 m above sea level in the Andes Mountains, Salta Province, Argentina. This area has varying concentrations of arsenic in the drinking water, but little industrial or traffic pollution (Concha et al. 2010). This population has been found to have an efficient metabolism of arsenic, i.e., they efficiently methylate inorganic arsenic to dimethylarsinic acid (DMA) with little excretion of the monomethylated metabolite, MMA, in urine (Vahter et al. 1995).

The study had 34 women enrolled in 2011. A detailed description of the study group has been published previously (Harari et al. 2013). Blood samples from the first 20 women recruited were sorted for CD4-positive T cells immediately after blood collection (K2EDTA tubes (Vacuette, Greiner Bio-One GmbH, Greiner, Germany), using the Dynabeads CD4 kit (Life Technologies, USA) and following the manufacturer’s protocol. Subsequently, RNA and DNA were extracted with the AllPrep DNA/RNA/Protein Mini kit (Qiagen, Hilden, Germany). Exposure to arsenic was assessed based on the sum concentration of metabolites of inorganic arsenic in urine (inorganic arsenic, MMA, and DMA), as measured by high-performance liquid chromatography (Agilent 1100 Series System, Agilent Technologies, Waldbronn, Germany) coupled with hydride generation and inductively coupled plasma mass spectrometry (Agilent 7500ce; Agilent Technologies, Tokyo, Japan) (HPLC-HG-ICPMS), as described previously (Concha et al. 2010). To correct for variation in urine dilution, we adjusted the obtained concentrations for the average specific gravity mean specific gravity of urine (1.020 g/mL), determined by a digital refractometer (EUROMEX RD 712 clinical refractometer; EUROMEX, Arnhem, the Netherlands). To obtain a large contrast in exposure, we selected samples from the eight women who showed the greatest differences in arsenic exposure (four with high exposure and four with lower exposure) and that had similar age and BMI. However, there was an overlap between the samples used for DNA and RNA analyses for only three of the women, due to limitations in sample amount. For the transcriptome analysis all samples were analyzed individually, and for the methylome analysis the samples in each exposure group were pooled. None of the women reported any disease or other health problems, and none reported drinking alcohol or smoking. The high- and low-exposure groups showed a fourfold to fivefold difference in urinary arsenic concentrations (Table S1).

Transcriptome characterization: RNA sequencing (RNA-seq)

RNA concentration and purity were evaluated on a NanoDrop spectrophotometer, and RNA integrity (RIN) was evaluated on a Bioanalyzer 2100; the results showed good RNA quality (RIN > 7.5). Total RNA was used to prepare libraries, which were sequenced as paired-end, 100-bp reads on an Illumina HIseq 2500 (HiSeq Control Software 2.2.38/RTA 1.18.61). Library construction and sequencing were performed according to the manufacturer’s protocols by researchers at The Swedish National Center for Molecular Biosciences, Science for Life Laboratory (SciLifeLab), Stockholm, Sweden.

Methylome characterization: target-enrichment NGS

Equal quantities of DNA from all four women (250 ng) for each exposure group were pooled and sequenced in two technical replicates. Sample processing included DNA quality control (using Qubit assays, from Life Technologies, USA, and Bioanalyzer, from Agilent, USA), target enrichment, bisulfite modification, library preparation, and sequencing. Target enrichment was performed with SureSelectXT Human Methyl-Seq kit (Agilent, USA), including the following genomic sequences likely to undergo changes in methylation: CpG islands, shores, and shelves, DNase I hypersensitive sites, Refseq genes, and Ensembl regulatory features. The enriched products were bisulfite treated with the Zymo EZ DNA Methylation Lightning kit (Zymo Research, USA) and sequenced by HiSeq 2500, in High Output mode 2 × 100 bp with the TruSeq SBS and the TruSeq PE Cluster kits, both v3 (Illumina, USA). The sample preparation was performed according to the manufacturers’ protocols by SciLifeLab.

MS-HRM analysis

Methylation-sensitive high-resolution melting (MS-HRM) assays were designed and optimized as described previously (Wojdacz and Dobrovic 2007). LightCycler 480 (Roche, Germany) and LightCycler 480 High Resolution Melting Master (Roche) PCR chemistry were used.

RNA-seq data analysis

RNA-seq reads were processed with TrimGalore (v0.3.7) for adapter removal and adaptive trimming. Alignment was performed with TopHat (v2.0.12), and reads were counted with HT-seq (v0.6.1) with GENCODE V19 GTF annotation. The features with zero counts were removed, and the size factor was estimated to normalize the libraries. The cumulative distribution of the counts before and after normalization was examined to check for proper normalization. Regularized log2 transformation (due to skewed data) was applied for further visualization and data exploration. Pairwise correlation and distance between samples were calculated and visualized, and principal component analysis (PCA) was used to explore sample clustering. The variability of gene expression was measured by coefficient of variation (standard deviation divided by the mean), and the variance explained was calculated on the top 20% variable genes ranked using the coefficient of variation. Differential expression analysis was performed using DESeq 2, contrasting high- vs. low-arsenic exposure groups. Differentially expressed genes (DEGs) were defined as having a log-twofold change in expression (log2FC) between arsenic exposure groups and with a false discovery rate (FDR) adjusted p value (q) <0.05. DEGs with a positive log-twofold change in the group with higher exposure, compared to the group with lower exposure, were defined as upregulated, while DEGs with a negative log-twofold change in the group with higher exposure compared to the group with lower exposure were defined as downregulated. Independent filtering was employed to calculate cutoff (<2.7) for the number of genes with low expression. There were 28,351 input genes, of which 11,326 (40%) showed low expression (<2.7) and 69 were defined as outliers (genes whose observed counts might not fit to a negative binominal distribution). Heatmaps were obtained for the DEGs. Enrichment for gene ontology was analyzed using TopGo with a Fisher test and the algorithm weight01, to take into account the structure of the gene ontology tree and to eliminate redundancy.

Target-enrichment NGS data analysis

Adapter removal, adaptive trimming (quality score <28), and 5′ clipping (4 nucleotides) were performed using Trim Galore (v0.3.3). Trimmed sequences were mapped to the human genome (build hg19), de-duplication was performed, and methylation calls were extracted using Bismark (v0.10.0, with Bowtie2 v2.0.6) (Krueger and Andrews 2011). Downstream analysis was performed using bsseq (Hansen et al. 2012). CpGs with ≥10 × coverage in all samples were retained, and 2,705,455 CpGs were included in subsequent analysis. Genomic clusters of CpGs were identified: Regions covered by the capture probes were extended 100 bp on either side, and regions separated by <300 bp were merged into single clusters. To identify differentially methylated regions (DMRs), we first calculated a difference in methylation (∆Meth) for each CpG position between high- and low-exposure groups. The function “regionFinder” was used in the bumphunter package version 1.2.0 [modified from a previously published method (Jaffe et al. 2012)], providing the locations of the clusters and using a cutoff of ∆Meth = 10%. The DMRs were then filtered for those with at least four CpGs. DMRs with higher methylation in the high-exposure group compared to the lower-exposure group were defined as hypermethylated; DMRs with lower methylation in the high-exposure group compared to the lower-exposure group were defined as hypomethylated. For evaluation of technical reproducibility, SeqMonk (Zhao et al. 2014) was used to generate cumulative distribution plots that describe the methylation level at each CpG site versus a number of CpG sites with a given methylation level. The GREAT platform was used for analysis of gene ontology (McLean et al. 2010).

Alignment RNA-seq and target-enrichment NGS

Gene overlap for DMRs and DEGs that were both statistically significantly associated with arsenic exposure group was further evaluated. DMRs included in these analyses were restricted to those in promoter regions, defined as within 500-base-pair downstream and 1500-base-pair upstream of the transcription start site.


Descriptive data

We compared the transcriptomes and methylomes of CD4-positive T cells from four women with high-arsenic exposure (~300 µg/L in urine) to those from four women with lower-arsenic exposure (~60 µg/L; Table S1). The two exposure groups and the women evaluated for transcriptomes and methylomes showed no statistically significant differences in age, body mass index (BMI), or coca chewing (Table S1).

Transcriptomics of CD4-positive T cells

We did not observe any bias in the distribution of number of transcripts, i.e., count data, with respect to exposure group, or any sample outliers when performing pairwise scatterplots of the samples (data not shown).

We performed PCA to evaluate the genome-wide effects of arsenic on gene expression in CD4-positive T cells. The first principal component clearly separated samples based on their exposure group and explained 53% of the variance in gene expression among the top 20% variable genes (based on coefficient of variation) (Fig. 1, left panel). We identified multiple DEGs between arsenic exposure groups (Fig. 1, right panel): out of 28,351 input genes, the total fraction of DEGs, up- (9.7%, 2744 genes) or downregulated (9%, 2553 genes) in the high-arsenic group was 18.7%. We found the largest log2FC for downregulated genes: 79 genes showed a log2FC below −3, with a maximum of −4.7. For upregulated DEGs, 20 genes showed a log2FC above 3 with a maximum of 4.0.
Fig. 1

Principal components analysis (left panel) and heatmap (right panel) for differently expressed genes between high- and lower-arsenic exposure groups. In the principal components analysis, the first component separates samples based on their exposure group and explains 53% of the variance in gene expression for the top 20% variable genes. In the heatmap, the row Z-score of differentially expressed genes is plotted, which shows the relative change for each row (gene). Samples are correctly clustered according to the exposure group

Among the top 10 DEGs (based on p value), four have been linked to immune system function, and all were downregulated in the high-arsenic exposure group (Table S2): these genes encode tumor necrosis factor, alpha-induced protein 3 (TNFAIP3), CD69 molecule (CD69), SBDS ribosome assembly guanine nucleotide exchange factor (SBDS), and DNA-damage-inducible transcript 3 (DDIT3). Other significantly downregulated immune-related DEGs included TNIP1, TNIP2, TNF, IFNG, IL10, IL12A, as well as several genes related to MHC class 2 (HLA-DRA, HLA-DQA2, HLA-DPB1, HLA-DMA), and the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-kappa-beta complex: REL, RELA, and RELB) (Table 1). Upregulated immune-related genes included IL16 and IL24 (encoding cytokines), TLR1, TLR4, TLR5, TLR6, TLR7, TLR8, and TLR10 (encoding Toll-like receptors), and genes coding signal transducer and activator of transcription (STAT) proteins (Table 1). Also, some important apoptosis-related genes were upregulated, such as BCL2, FASLG, MAF, and CASP2 as well as CASP3, CASP6, CASP8, and CASP10. We note that DNA (cytosine-5-)-methyltransferase 1 (DNMT1) was upregulated (log2FC = 0.59, q = 0.043), but none of the de novo DNA methyltransferases (DNMT3A, DNMT3B, and DNMT3L) was differently expressed.
Table 1

Selection of immune-related differentially expressed genes (DEGs) between high-arsenic (n = 4) and lower-arsenic (n = 4) exposure groups

Gene name

Gene description


Q a


v-rel avian reticuloendotheliosis viral oncogene homolog B


4.1 × 10−19


Tumor necrosis factor


5.2 × 10−16


Caspase 3, apoptosis-related cysteine peptidase


2.8 × 10−12


Interferon, gamma


5.8 × 10−10


Toll-like receptor 8


1.0 × 10−08


Caspase 10, apoptosis-related cysteine peptidase


2.6 × 10−08


Major histocompatibility complex, class II, DQ alpha 2


8.4 × 10−08


Toll-like receptor 4


1.8 × 10−06


Toll-like receptor 10


5.3 × 10−06


B cell CLL/lymphoma 2


7.6 × 10−06


Caspase 6, apoptosis-related cysteine peptidase


1.2 × 10−05


Toll-like receptor 7


1.4 × 10−05


v-maf avian musculoaponeurotic fibrosarcoma oncogene homolog


4.4 × 10−05


v-rel avian reticuloendotheliosis viral oncogene homolog A




Interleukin 12A




Signal transducer and activator of transcription 2




Toll-like receptor 5




Signal transducer and activator of transcription 5A




v-rel avian reticuloendotheliosis viral oncogene homolog




Interleukin 24




Interleukin 16




TNFAIP3 interacting protein 1




Caspase 8, apoptosis-related cysteine peptidase




Interleukin 10




Signal transducer and activator of transcription 1




Signal transducer and activator of transcription 33response factor)




Toll-like receptor 1




Fas ligand (TNF superfamily, member 6)




TNFAIP3 interacting protein 2




Major histocompatibility complex, class II, DM alpha




Major histocompatibility complex, class II, DP beta 1




Toll-like receptor 6




Caspase 2, apoptosis-related cysteine peptidase




Major histocompatibility complex, class II, DR alpha



The DEGs are sorted based on p value (immune genes presented in Table S2 are not included)

a Log2FC = log-twofold change, q = FDR-adjusted p value

We then performed a gene ontology analysis with the topGO algorithm, considering the upregulated and downregulated DEGs separately and together. The top gene ontologies for the downregulated DEGs were related to RNA metabolism, including one category for viral transcription. The top gene ontology category for upregulated DEGs was regulation of transcription, DNA-templated. In general, categories related to RNA metabolism were significantly enriched among both the up- and downregulated DEGs, including general transcriptional and translational processes (Table S3).

Methylomics of CD4-positive cells

For each exposure group, we used bisulfite sequencing to examine two technical replicates of pooled DNA from all individuals in each exposure group and evaluated the technical variability of the data by plotting the cumulative distribution of methylation (Figure S1). The two replicates were almost identical, indicating very little technical variability in the data. A clear shift between the plots of the high- and low-exposure groups suggested that a large number of CpG sites were differentially methylated between the exposure groups. To evaluate the epigenome-wide effect of arsenic exposure, we computed methylation levels for each site and compared the number of CpGs showing full methylation (defined as >80% methylation) between exposure groups. We found that in the two technical replicates of the lower-exposure group, 41.3 and 41.2% of the CpG sites showed full methylation, compared with 44.2% in both technical replicates of the high-exposure group, i.e., 3 percentage points higher full methylation, genome wide. DNA methylation increased for the high-exposure group at certain functional elements, particularly in CpG shelves, gene bodies, CpG shores, and gene promoters (Figure S2), but not in CpG islands.

We identified 1168 differentially methylated regions (DMRs) of which 391 (33.5%) showed hypermethylation and 777 (66.5%) hypomethylation in the high-exposure group. The top 10 DMRs are presented in Table S4, and examples of statistically significant methylation changes at a single-base resolution are shown in Fig. 2. Among the top 50 DMRs, two immune-related genes have previously been associated with arsenic exposure (Ramsey et al. 2013; Ding et al. 2016), encoding secretoglobin family 3A member 1 (SCGB3A1) and nuclear factor of activated T cells 1 (NFATC1). Both genes were hypomethylated in the high-arsenic group. The 1168 DMRs included several immune-related genes, such as interleukins IL9, IL10, IL13, IL32 (hypermethylated), IL36A (hypomethylated), members of the interferon regulatory factor (IRF) family; IRF1, IRF4, IRF7, and IRF8 (all hypermethylated), members of the cluster of differentiation (CD) family CD2, CD58 (hypermethylated), CD6, CD27, CD40 (hypomethylated), two genes related to MHC class (HLA-DQAI, which was hypermethylated, and HLA-DRB5, for which one DMR was hypermethylated and one DMR hypomethylated), as well as IFNG and TLR10.
Fig. 2

Single CpG resolution of differences in DNA methylation between arsenic exposure groups. ac Differentially methylated regions (DMRs) hypermethylated in the high-arsenic exposure group; eg DMRs hypomethylated in the high-arsenic exposure group. The chromosome number and genomic position are shown above the figure. Exposure groups are shown in blue (lower-arsenic exposure) and red (high exposure); vertical bars represent CpGs covered by targeted sequencing in specific regions. The methylation on the y-axis is shown as beta values. Green boxes are regions of the genome where DMRs (defined here as four consecutive CpG sites with a difference in methylation above 10% between high- and low-arsenic groups) were identified. Bottom bars indicate distances of the DMR to the closest gene (G) and CpG island (CGI)

The gene ontology analyses of hypermethylated DMRs (performed by the GREAT platform) showed an enrichment of cellular processes that relate to immune function of CD4-positive T cells, such as processes related to immunoglobulins and isotype switching (positive regulation of isotype switching, positive regulation of isotype switching to IgE isotypes, positive regulation of isotype switching to IgG isotypes, regulation of immunoglobulin production) and lymphocyte activation and differentiation (Table 2). By contrast, hypomethylated DMRs showed an enrichment of developmental processes, such as embryonic organ morphogenesis (Table 2). Hypermethylated DMRs also showed enrichment of genes related to V(D)J recombination.
Table 2

Top gene ontologies (GOs) related to differently methylated regions between high-arsenic (n = 4) and lower-arsenic (n = 4) exposure groups



P value

DMRs hypermethylated in high-arsenic exposure group



Positive regulation of isotype switching

5.25 × 10−6


Positive regulation of DNA recombination

8.91 × 10−6


Lymphocyte activation

1.45 × 10−5


Positive regulation of isotype switching to IgE isotypes

2.95 × 10−5


Regulation of immunoglobulin production

3.09 × 10−5


Lymphocyte differentiation

3.89 × 10−5


Translational elongation

4.07 × 10−5


Regulation of leukocyte-mediated immunity

5.5 × 10−5


Innate immune response

6.17 × 10−5


Regulation of isotype switching

6.31 × 10−5


Positive regulation of isotype switching to IgG isotypes

7.24 × 10−5

DMRs hypomethylated in high-arsenic exposure group



Embryonic organ morphogenesis

1.26 × 10−9


Prostate gland stromal morphogenesis

1.78 × 10−6


Histone H2A monoubiquitination

3.31 × 10−6


Negative regulation of osteoblast differentiation

8.13 × 10−6


Epithelial–mesenchymal cell signaling

9.55 × 10−6


Histone H2A ubiquitination

1.29 × 10−5


Regulation of ERK1 and ERK2 cascade

1.48 × 10−5


V(D)J recombination

1.58 × 10−5


Embryonic cranial skeleton morphogenesis

2.19 × 10−5


Negative regulation of sequence-specific DNA binding transcription factor activity

2.19 × 10−5


Regulation of mesenchymal cell proliferation

2.29 × 10−5


Positive regulation of ERK1 and ERK2 cascade

2.82 × 10−5


Regulation of myotube differentiation

3.16 × 10−5

To validate the accuracy of the methylome analysis, we analyzed four statistically significant DMRs using MS-HRM assays, and all confirmed a methylation difference between samples from high- and lower-exposure groups (data not shown).

Alignment of DEGs and promoter DMRs associated with arsenic exposure group

In total, 32 genes had DMRs and DEGs that both statistically significantly associated with arsenic exposure group (top 10 genes in Table S5). Two of the 32 genes are involved in immune responses (Miyamoto et al. 1988; Dorn et al. 2007) (Figure S3): (1) interferon regulatory factor 1 (IRF1), for which the gene expression was downregulated and the DMR hypermethylated, and (2) ras homolog family member H (RHOH) for which the gene expression was downregulated and the DMR hypomethylated.


Our study suggests that chronic arsenic exposure through drinking water is related to changes in the transcriptome and methylome, both genome wide and for specific genes, of CD4-positive T cells in Andean women. Several key genes in the immune system were differently expressed at higher-arsenic exposure, and DMRs that were hypermethylated in the high-arsenic exposure group showed enrichment of immune-related processes that constitute the basic functions of CD4-positive T cells. These findings support the hypothesis that elevated arsenic exposure may cause immunotoxicity by interfering with gene expression and gene regulation. Still, we could not link many DMRs in promoter regions to changes in gene expression, which may be due to limited sequencing coverage, few samples analyzed and only partial overlap, or the arsenic-related changes in gene expression may be partly mediated by mechanisms other than DNA methylation.

Arsenic exposure was associated with changes in genome-wide gene expression and expression of specific immune-related genes

Arsenic exposure was associated with genome-wide gene expression. In spite of the few samples analyzed, the first principal component, which separated samples based on their exposure group, explained more than half of the observed variance in gene expression for the top 20% variable genes and more than 5000 genes were differently expressed between the arsenic exposure groups. Among the top ten DEGs, four (all downregulated) have been linked to immune system function, and for three of these (TNFAIP3, CD69, and DDIT3), changes in expression have previously been associated with different types of arsenic exposure (Wang et al. 2006; Andrew et al. 2008; Li et al. 2013; Nadeau et al. 2014). The deubiquitinating enzyme TNFAIP3 (also called A20) is an important negative regulator of innate and adaptive immunity. TNFAIP3 participates in immune and inflammatory responses signaled by cytokines, such as TNF, and TNFAIP3 is rapidly induced by TNF (Dixit et al. 1990). TNFAIP3 also responds to signaling by Toll-like receptors (Boone et al. 2004) and affects immune responses by strongly inhibiting NF-kappa-B activity (Heyninck and Beyaert 2005). TNFAIP3 is most highly expressed in T cells and, in particular, in CD4 positive T cells ( The downregulation of TNFAIP3 is consistent with the downregulation of TNF with increasing arsenic exposure identified in this study, as well as with the downregulation of TNFAIP3-interacting proteins 1 and 2 (TNIP1 and TNIP2). Also, TNFAIP3 was upregulated in a multiple myeloma cell line after treatment with realgar, an arsenic sulfide mineral (Wang et al. 2006).

CD69 codes for a surface marker expressed on CD4 positive T cells. The activation of T lymphocytes induces expression of CD69. CD69 appears to be the earliest inducible cell surface glycoprotein acquired during lymphoid activation and is involved in lymphocyte proliferation. CD69 functions as a signal-transmitting receptor in lymphocytes, including NK cells, and platelets (Cambiaggi et al. 1992). Downregulation of CD69, as observed here, could thus be linked to decreased activation of T lymphocytes. The downregulation of CD69 is in line with the results from a population exposed to low levels of arsenic in New Hampshire (N = 21), where increasing arsenic exposure from drinking water was associated with decreased CD69 expression in peripheral blood lymphocytes (Andrew et al. 2008). Another study in New Hampshire showed that in utero environmental arsenic exposure was inversely related to absolute total cord blood CD69 + T cell counts (Nadeau et al. 2014).

DNA-damage-inducible transcript 3 (DDIT3), also known as C/EBP-homologous protein (CHOP), encodes a transcription factor. DDIT3 has previously been shown to be upregulated in CD4-positive T cells in mice in response to arsenic trioxide treatment (Li et al. 2013). The upregulation of TNFAIP3 and DDIT3 found in previous studies is in contrast to the downregulation of these genes observed in our study. However, the expression of these genes could be dose dependent, since the doses of arsenic trioxide or realgar were much higher than the doses from drinking water in the current study.

Furthermore, we found that several key players in the immune system were downregulated, such as TNF, IFNG, and IL10, as well as several genes related to MHC class 2 and the NF-kappa-B complex, while several key genes related to apoptosis were upregulated. Most of these findings are in accordance with previous studies investigating the association between gene expression (measured by microarray) and arsenic exposure in peripheral blood (Dangleben et al. 2013). Apoptosis is important for the homeostasis of the immune system, and abnormal immune cell apoptosis can contribute to dysregulation of immune function (Boise et al. 1995). The decrease in T cells with increasing arsenic exposure could be mediated by the upregulation of apoptotic genes. The regulation of mRNA levels of these genes may hypothetically be due to an effect of arsenic exposure on RNA metabolism, since several gene ontologies related to RNA metabolism were strongly enriched in the high-arsenic group.

Exposure to arsenic influenced genome-wide and gene-specific 5-methylcytosine methylation

Our analysis of the methylome indicated that 3% of the genome-wide CpGs gained full methylation in the high-arsenic samples. By contrast, earlier in vitro and in vivo experiments associated arsenic exposure with global hypomethylation, measured as methylation of repetitive elements (Intarasunanont et al. 2012; Bailey and Fry 2014; Kile et al. 2014). Here, we measured methylation at CpG sites in regulatory sequences, which showed that these regions were hypermethylated in women in the high-arsenic exposure group. This suggests that in response to arsenic exposure, transcriptionally active parts of the genome gain methylation while repetitive elements lose methylation. Arsenic exposure did not appear to affect the methylation of CpG islands, but previous work showed that those elements are generally protected from changes in methylation (Deaton and Bird 2011). The increase in genome-wide DNA methylation may partly be explained by the upregulation of DNMT1 expression in the high-arsenic exposure group, as DNMT1 is the major enzyme responsible for maintaining methylation patterns following DNA replication in somatic cells.

In addition, arsenic exposure also seemed to be associated with locus-specific DNA methylation. Two of the top 50 DMRs in the arsenic exposure group (both hypomethylated) covered immune-related genes that have previously been associated with gene expression in relation to arsenic. The secretoglobin gene SCGB3A1 was upregulated in the lung in C57BL/6 mice exposed to arsenic in drinking water (Ramsey et al. 2013). Secretoglobins have anti-inflammatory and immunomodulatory functions (Jackson et al. 2011). NFATC1 plays a central role in inducible transcription of cytokine genes in T cells (Chuvpilo et al. 1999) and was downregulated by arsenic in HCT116, a colon cancer cell line (Ding et al. 2016).

Gene ontology analysis of the DMRs that were hypermethylated in the high-arsenic group showed enrichment of immune-related processes that constitute the basic functions of CD4-positive T cells, such as isotype switching, especially the switch to IgE or IgG. Hypomethylated DMRs, by contrast, were linked to embryonic organ development, in particular. This is of interest, as a previous study found genome-wide hypomethylation in blood from newborns with elevated prenatal exposure to arsenic (Broberg et al. 2014).

Differentially methylated regions in gene promoters correlated partly with gene expression

Thirty-two genes were significantly differentially expressed in the transcriptome analyses and also harbored a DMR. Two of the top genes, IRF1 and RHOH, are involved in immune responses, and both were downregulated in the high-arsenic group. IRF1 is a member of the interferon regulatory transcription factor (IRF) family and activates transcription of interferons alpha and beta, as well as of genes induced by interferons alpha, beta, and gamma (Miyamoto et al. 1988). Thus, the downregulation of IRF1 may be linked to the downregulation of IFNG seen in this study. IRF1 plays an important role in Th1 development and maturation of T helper cells (Lohoff and Mak 2005) and is involved in the suppression of regulatory T cell development (Fragale et al. 2008). RHOH is a critical regulator of thymocyte development and T cell antigen receptor signaling (Dorn et al. 2007).

Strengths and limitations of the current study

The main strength of this study was that we used large-scale NGS on sorted CD4-positive T cells, which allowed us to conduct a genome-wide investigation of how this specific immune cell may respond to arsenic exposure. The technical variability in our sequencing data was minor, and the contrast in arsenic exposure was large; this enabled us to reveal group differences in the methylome and the transcriptome, despite the limited number of samples. However, further studies are needed for evaluation of causality as well as the dose–effect relationships. To note, the group with lower exposure was also somewhat exposed to arsenic; thus, there is no unexposed group in this study. A limitation of the study is the small number of samples and that the samples for the two different sequencing analyses were not completely overlapping, due to sample limitations. Also, only women were included; thus, we do not know whether these results also extend to men. Moreover, this is a unique population for studies on the effects of arsenic exposure, as they have an efficient metabolism of arsenic (Vahter et al. 1995), known to be related to lower risk of toxicity This population has been exposed to arsenic in drinking water for thousands of years and likely has adapted to tolerate arsenic (Schlebusch et al. 2013, 2015). Thus, the indicated changes in gene expression and DNA methylation may not necessarily be considered toxic responses, since they may occur following an adaptive response, e.g., acquired resistance over time. Thus, the effects in a population without this background may differ from the effects found in this study. Further and larger studies, in other populations with wide ranges in chronic arsenic exposure, and also including men, will be needed to validate the results.


We showed that chronic arsenic exposure from drinking water is related to changes in the transcriptome and methylome of CD4-positive T cells, both genome wide and for specific genes, supporting the hypothesis that arsenic causes immunotoxicity by interfering with gene expression and gene regulation.



We thank the women for participating in this study, as well as Dr. Gabriela Concha and Dr. Florencia Harari for help with recruiting the study participants. This research was supported by grants from the Swedish Council for Health, Working Life and Welfare; Kungliga fysiografiska sällskapet; Eric Philip Sörensen’s Foundation; and the Karolinska Institutet. The authors declare they have no actual or potential competing financial interests. The authors would like to acknowledge support from Science for Life Laboratory, the National Genomics Infrastructure, NGI, and Uppmax for providing assistance in massive parallel sequencing and computational infrastructure.

Compliance with ethical standards

Ethical standards

All study participants gave their informed consent prior to their inclusion in the study. The study has been approved by the Regional Ethics Committee, Stockholm, and has been performed in accordance with the Declaration of Helsinki and its later amendments.

Conflict of interest

The authors declare that they have no conflict of interest.

Supplementary material

204_2016_1879_MOESM1_ESM.pdf (2.3 mb)
Supplementary material 1 (PDF 2338 kb)


  1. Ahmed S, Mahabbat-e Khoda S, Rekha RS, Gardner RM, Ameer SS, Moore S et al (2011) Arsenic-associated oxidative stress, inflammation, and immune disruption in human placenta and cord blood. Environ Health Perspect 119:258–264CrossRefPubMedGoogle Scholar
  2. Ahmed S, Ahsan KB, Kippler M, Mily A, Wagatsuma Y, Hoque AM et al (2012) In utero arsenic exposure is associated with impaired thymic function in newborns possibly via oxidative stress and apoptosis. Toxicol Sci 129:305–314CrossRefPubMedGoogle Scholar
  3. Ahmed S, Moore SE, Kippler M, Gardner R, Hawlader MD, Wagatsuma Y et al (2014) Arsenic exposure and cell-mediated immunity in pre-school children in rural Bangladesh. Toxicol Sci 141:166–175CrossRefPubMedPubMedCentralGoogle Scholar
  4. Andrew AS, Jewell DA, Mason RA, Whitfield ML, Moore JH, Karagas MR (2008) Drinking-water arsenic exposure modulates gene expression in human lymphocytes from a U.S. Population. Environ Health Perspect 116:524–531CrossRefPubMedPubMedCentralGoogle Scholar
  5. Argos M, Kibriya MG, Parvez F, Jasmine F, Rakibuz-Zaman M, Ahsan H (2006) Gene expression profiles in peripheral lymphocytes by arsenic exposure and skin lesion status in a Bangladeshi population. Cancer Epidemiol Biomark Prev 15:1367–1375CrossRefGoogle Scholar
  6. Bailey KA, Fry RC (2014) Arsenic-associated changes to the epigenome: what are the functional consequences? Curr Environ Health Rep 1:22–34CrossRefPubMedPubMedCentralGoogle Scholar
  7. Biswas R, Ghosh P, Banerjee N, Das JK, Sau T, Banerjee A et al (2008) Analysis of T-cell proliferation and cytokine secretion in the individuals exposed to arsenic. Hum Exp Toxicol 27:381–386CrossRefPubMedGoogle Scholar
  8. Boise LH, Minn AJ, Thompson CB (1995) Receptors that regulate T-cell susceptibility to apoptotic cell death. Ann N Y Acad Sci 766:70–80CrossRefPubMedGoogle Scholar
  9. Boone DL, Turer EE, Lee EG, Ahmad RC, Wheeler MT, Tsui C et al (2004) The ubiquitin-modifying enzyme a20 is required for termination of toll-like receptor responses. Nat Immun 5:1052–1060CrossRefGoogle Scholar
  10. Broberg K, Ahmed S, Engstrom K, Hossain MB, Jurkovic Mlakar S, Bottai M et al (2014) Arsenic exposure in early pregnancy alters genome-wide DNA methylation in cord blood, particularly in boys. J Dev Orig Health Dis 5:288–298CrossRefPubMedPubMedCentralGoogle Scholar
  11. Cambiaggi C, Scupoli MT, Cestari T, Gerosa F, Carra G, Tridente G et al (1992) Constitutive expression of CD69 in interspecies T-cell hybrids and locus assignment to human chromosome 12. Immunogenetics 36:117–120CrossRefPubMedGoogle Scholar
  12. Cardenas A, Houseman EA, Baccarelli AA, Quamruzzaman Q, Rahman M, Mostofa G et al (2015) In utero arsenic exposure and epigenome-wide associations in placenta, umbilical artery, and human umbilical vein endothelial cells. Epigenetics 10:1054–1063CrossRefPubMedPubMedCentralGoogle Scholar
  13. Chuvpilo S, Avots A, Berberich-Siebelt F, Glockner J, Fischer C, Kerstan A et al (1999) Multiple NF-aATc isoforms with individual transcriptional properties are synthesized in T lymphocytes. J Immunol 162:7294–7301PubMedGoogle Scholar
  14. Concha G, Broberg K, Grander M, Cardozo A, Palm B, Vahter M (2010) High-level exposure to lithium, boron, cesium, and arsenic via drinking water in the Andes of northern Argentina. Environ Sci Technol 44:6875–6880CrossRefPubMedGoogle Scholar
  15. Dangleben NL, Skibola CF, Smith MT (2013) Arsenic immunotoxicity: a review. Environ Health 12:73CrossRefPubMedPubMedCentralGoogle Scholar
  16. Deaton AM, Bird A (2011) CpG islands and the regulation of transcription. Genes Dev 25:1010–1022CrossRefPubMedPubMedCentralGoogle Scholar
  17. Ding W, Tong Y, Zhang X, Pan M, Chen S (2016) Study of arsenic sulfide in solid tumor cells reveals regulation of nuclear factors of activated T-cells by PML and p53. Sci Rep 6:19793CrossRefPubMedPubMedCentralGoogle Scholar
  18. Dixit VM, Green S, Sarma V, Holzman LB, Wolf FW, O’Rourke K et al (1990) Tumor necrosis factor-alpha induction of novel gene products in human endothelial cells including a macrophage-specific chemotaxin. J Biol Chem 265:2973–2978PubMedGoogle Scholar
  19. Dorn T, Kuhn U, Bungartz G, Stiller S, Bauer M, Ellwart J et al (2007) RhoH is important for positive thymocyte selection and T-cell receptor signaling. Blood 109:2346–2355CrossRefPubMedGoogle Scholar
  20. Farzan SF, Li Z, Korrick SA, Spiegelman D, Enelow R, Nadeau K et al (2016) Infant infections and respiratory symptoms in relation to in utero arsenic exposure in a U.S. cohort. Environ Health Perspect 124:840–847PubMedGoogle Scholar
  21. Ferrario D, Gribaldo L, Hartung T (2016) Arsenic exposure and immunotoxicity: a review including the possible influence of age and sex. Curr Environ Health Rep 3:1–12CrossRefPubMedGoogle Scholar
  22. Fragale A, Gabriele L, Stellacci E, Borghi P, Perrotti E, Ilari R et al (2008) Ifn regulatory factor-1 negatively regulates CD4 + CD25 + regulatory T cell differentiation by repressing foxp3 expression. J Immunol 181:1673–1682CrossRefPubMedGoogle Scholar
  23. Hansen KD, Langmead B, Irizarry RA (2012) Bsmooth: from whole genome bisulfite sequencing reads to differentially methylated regions. Genome Biol 13:R83CrossRefPubMedPubMedCentralGoogle Scholar
  24. Harari F, Engstrom K, Concha G, Colque G, Vahter M, Broberg K (2013) N-6-adenine-specific DNA methyltransferase 1 (N6AMT1) polymorphisms and arsenic methylation in andean women. Environ Health Perspect 121:797–803CrossRefPubMedPubMedCentralGoogle Scholar
  25. Hernandez-Castro B, Doniz-Padilla LM, Salgado-Bustamante M, Rocha D, Ortiz-Perez MD, Jimenez-Capdeville ME et al (2009) Effect of arsenic on regulatory T cells. J Clin Immunol 29:461–469CrossRefPubMedGoogle Scholar
  26. Heyninck K, Beyaert R (2005) A20 inhibits NF-kappab activation by dual ubiquitin-editing functions. Trends Biochem Sci 30:1–4CrossRefPubMedGoogle Scholar
  27. IARC (2012) Arsenic, metals, fibres, and dusts. IARC Monogr Eval Carcinog Risks Hum World Health Organ Int Agency Res Cancer 100:11–465Google Scholar
  28. Intarasunanont P, Navasumrit P, Waraprasit S, Chaisatra K, Suk WA, Mahidol C et al (2012) Effects of arsenic exposure on DNA methylation in cord blood samples from newborn babies and in a human lymphoblast cell line. Environ Health 11:31CrossRefPubMedPubMedCentralGoogle Scholar
  29. Jackson BC, Thompson DC, Wright MW, McAndrews M, Bernard A, Nebert DW et al (2011) Update of the human secretoglobin (SCGB) gene superfamily and an example of ‘evolutionary bloom’ of androgen-binding protein genes within the mouse scgb gene superfamily. Hum Genom 5:691–702CrossRefGoogle Scholar
  30. Jaffe AE, Murakami P, Lee H, Leek JT, Fallin MD, Feinberg AP et al (2012) Bump hunting to identify differentially methylated regions in epigenetic epidemiology studies. Int J Epidemiol 41:200–209CrossRefPubMedPubMedCentralGoogle Scholar
  31. Kile ML, Houseman EA, Baccarelli AA, Quamruzzaman Q, Rahman M, Mostofa G et al (2014) Effect of prenatal arsenic exposure on DNA methylation and leukocyte subpopulations in cord blood. Epigenetics 9:774–782CrossRefPubMedPubMedCentralGoogle Scholar
  32. Koestler DC, Avissar-Whiting M, Houseman EA, Karagas MR, Marsit CJ (2013) Differential DNA methylation in umbilical cord blood of infants exposed to low levels of arsenic in utero. Environ Health Perspect 121:971–977CrossRefPubMedPubMedCentralGoogle Scholar
  33. Krueger F, Andrews SR (2011) Bismark: a flexible aligner and methylation caller for bisulfite-seq applications. Bioinformatics 27:1571–1572CrossRefPubMedPubMedCentralGoogle Scholar
  34. Li K, Zhang L, Xiang X, Gong S, Ma L, Xu L et al (2013) Arsenic trioxide alleviates airway hyperresponsiveness and promotes apoptosis of CD4 + T lymphocytes: evidence for involvement of the ER stress-CHOP pathway. Ir J Med Sci 182:573–583CrossRefPubMedGoogle Scholar
  35. Lohoff M, Mak TW (2005) Roles of interferon-regulatory factors in T-helper-cell differentiation. Nat Rev Immunol 5:125–135CrossRefPubMedGoogle Scholar
  36. McLean CY, Bristor D, Hiller M, Clarke SL, Schaar BT, Lowe CB et al (2010) GREAT improves functional interpretation of cis-regulatory regions. Nat Biotechnol 28:495–501CrossRefPubMedPubMedCentralGoogle Scholar
  37. Miyamoto M, Fujita T, Kimura Y, Maruyama M, Harada H, Sudo Y et al (1988) Regulated expression of a gene encoding a nuclear factor, IRF-1, that specifically binds to IFN-beta gene regulatory elements. Cell 54:903–913CrossRefPubMedGoogle Scholar
  38. Moore SE, Prentice AM, Wagatsuma Y, Fulford AJ, Collinson AC, Raqib R et al (2009) Early-life nutritional and environmental determinants of thymic size in infants born in rural Bangladesh. Acta Paediatr 98(7):1168–1175CrossRefPubMedPubMedCentralGoogle Scholar
  39. Nadeau KC, Li Z, Farzan S, Koestler D, Robbins D, Fei DL et al (2014) In utero arsenic exposure and fetal immune repertoire in a us pregnancy cohort. Clin Immun 155:188–197CrossRefGoogle Scholar
  40. Pilsner JR, Hall MN, Liu X, Ilievski V, Slavkovich V, Levy D et al (2012) Influence of prenatal arsenic exposure and newborn sex on global methylation of cord blood DNA. PLoS ONE 7:e37147CrossRefPubMedPubMedCentralGoogle Scholar
  41. Rahman A, Vahter M, Ekstrom EC, Persson LA (2011) Arsenic exposure in pregnancy increases the risk of lower respiratory tract infection and diarrhea during infancy in Bangladesh. Environ Health Perspect 119:719–724CrossRefPubMedGoogle Scholar
  42. Ramsey KA, Bosco A, McKenna KL, Carter KW, Elliot JG, Berry LJ et al (2013) In utero exposure to arsenic alters lung development and genes related to immune and mucociliary function in mice. Environ Health Perspect 121:244–250PubMedGoogle Scholar
  43. Raqib R, Ahmed S, Sultana R, Wagatsuma Y, Mondal D, Hoque AM et al (2009) Effects of in utero arsenic exposure on child immunity and morbidity in rural Bangladesh. Toxicol Lett 185:197–202CrossRefPubMedGoogle Scholar
  44. Rojas D, Rager JE, Smeester L, Bailey KA, Drobna Z, Rubio-Andrade M et al (2015) Prenatal arsenic exposure and the epigenome: identifying sites of 5-methylcytosine alterations that predict functional changes in gene expression in newborn cord blood and subsequent birth outcomes. Toxicol Sci 143:97–106CrossRefPubMedGoogle Scholar
  45. Salgado-Bustamante M, Ortiz-Pérez MD, Calderón-Aranda E, Estrada-Capetillo L, Niño-Moreno P, González-Amaro R et al (2010) Pattern of expression of apoptosis and inflammatory genes in humans exposed to arsenic and/or fluoride. Sci Total Environ 408:760–767CrossRefPubMedGoogle Scholar
  46. Schlebusch CM, Lewis CM Jr, Vahter M, Engström K, Tito RY, Obregón-Tito AJ et al (2013) Possible positive selection for an arsenic-protective haplotype in humans. Environ Health Perspect 121:53–58PubMedGoogle Scholar
  47. Schlebusch CM, Gattepaille LM, Engström K, Vahter M, Jakobsson M, Broberg K (2015) Human adaptation to arsenic-rich environments. Mol Biol Evol 32:1544–1555CrossRefPubMedGoogle Scholar
  48. Soto-Pena GA, Luna AL, Acosta-Saavedra L, Conde P, Lopez-Carrillo L, Cebrian ME et al (2006) Assessment of lymphocyte subpopulations and cytokine secretion in children exposed to arsenic. FASEB J 20:779–781PubMedGoogle Scholar
  49. Tseng CH (2004) The potential biological mechanisms of arsenic-induced diabetes mellitus. Toxicol Appl Pharmacol 197:67–83CrossRefPubMedGoogle Scholar
  50. Vahter M, Couch R, Nermell B, Nilsson R (1995) Lack of methylation of inorganic arsenic in the chimpanzee. Toxicol Appl Pharmacol 133:262–268CrossRefPubMedGoogle Scholar
  51. Wang MC, Liu SX, Liu PB (2006) Gene expression profile of multiple myeloma cell line treated by realgar. JECCR CR 25:243–249Google Scholar
  52. Wojdacz TK, Dobrovic A (2007) Methylation-sensitive high resolution melting (ms-hrm): a new approach for sensitive and high-throughput assessment of methylation. Nucleic Acids Res 35:e41CrossRefPubMedPubMedCentralGoogle Scholar
  53. Zhao MT, Whyte JJ, Hopkins GM, Kirk MD, Prather RS (2014) Methylated DNA immunoprecipitation and high-throughput sequencing (medip-seq) using low amounts of genomic DNA. Cell Reprogr 16:175–184CrossRefGoogle Scholar

Copyright information

© The Author(s) 2016

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors and Affiliations

  • Karin Engström
    • 1
    • 2
  • Tomasz K. Wojdacz
    • 2
    • 8
    Email author
  • Francesco Marabita
    • 3
    • 4
  • Philip Ewels
    • 5
  • Max Käller
    • 6
  • Francesco Vezzi
    • 5
  • Nicola Prezza
    • 6
  • Joel Gruselius
    • 7
  • Marie Vahter
    • 1
  • Karin Broberg
    • 2
    Email author
  1. 1.Department of Laboratory Medicine, Division of Occupational and Environmental MedicineLund UniversityLundSweden
  2. 2.Unit of Metals and Health, Institute of Environmental MedicineKarolinska InstitutetStockholmSweden
  3. 3.Unit of Computational Medicine, Center for Molecular MedicineKarolinska University HospitalStockholmSweden
  4. 4.Bioinformatics Infrastructure for Life SciencesKarolinska InstitutetSolnaSweden
  5. 5.Science for Life Laboratory (SciLifeLab), Department of Biochemistry and BiophysicsStockholm UniversityStockholmSweden
  6. 6.Science for Life Laboratory, School of Biotechnology, Division of Gene TechnologyRoyal Institute of TechnologySolnaSweden
  7. 7.Department of Mathematical and Computer ScienceUniversity of UdineUdineItaly
  8. 8.Aarhus Institute of Advanced StudiesAarhus UniversityAarhus CDenmark

Personalised recommendations