If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Institute of Biomedicine and Translational Medicine, University of Tartu, Tartu, EstoniaDepartment of Genetics, Tartu University Hospital, Tartu, Estonia
Institute of Biomedicine and Translational Medicine, University of Tartu, Tartu, EstoniaDepartment of Reproductive Biology, Estonian University of Life Sciences, Tartu, Estonia
Address correspondence to Sulev Kõks, M.D., Ph.D., Institute of Biomedicine and Translational Medicine, University of Tartu, 19 Ravila St., Tartu 50411, Estonia.
Institute of Biomedicine and Translational Medicine, University of Tartu, Tartu, EstoniaDepartment of Reproductive Biology, Estonian University of Life Sciences, Tartu, Estonia
Despite the described clear epigenetic effects of smoking, the effect of smoking on genome-wide gene expression in the blood is obscure. We therefore studied the smoking-induced changes in the gene-expression profile of the peripheral blood. RNA was extracted from the whole blood of 48 individuals with a detailed smoking history (24 never-smokers, 16 smokers, and 8 ex-smokers). Gene-expression profiles were evaluated with RNA sequencing, and results were analyzed separately in 24 men and 24 women. In the male smokers, 13 genes were statistically significantly (false-discovery rate <0.1) differentially expressed; in female smokers, 5 genes. Although most of the differentially expressed genes were different between the male and female smokers, the G-protein–coupled receptor 15 gene (GPR15) was differentially expressed in both male and female smokers compared with never-smokers. Analysis of GPR15 methylation identified significantly greater hypomethylation in smokers compared with that in never-smokers. GPR15 is the chemoattractant receptor that regulates T-cell migration and immunity. Up-regulation of GPR15 could explain to some extent the health hazards of smoking with regard to chronic inflammatory diseases.
Tobacco smoking is considered the leading preventable cause of morbidity and mortality. Smoking affects >1 billion individuals worldwide and accounts for an estimated 3 million deaths per year.
The association of tobacco use with various chronic diseases is smoking-duration dependent, and the effects of tobacco depend on the amount (pack-years) of smoking.
These findings suggest roles for epigenetic reprogramming in the modulation of the biological effects of smoking and in the development of smoking-induced signature. Indeed, previous studies have identified an association between global DNA methylation and tobacco smoking in cancer-related tissues.
Focus on the methylation of particular gene loci has identified several regions as differentially methylated between smokers and never-smokers. The CpG site in the COMT gene at position -193 is methylated in 22.2% of smokers and 18.3% of never-smokers.
Determination of Methylated CpG Sites in the Promoter Region of Catechol-O-Methyltransferase (COMT) and their Involvement in the Etiology of Tobacco Smoking.
In addition, a clear correlation between methylation and smoking status (smokers, ex-smokers, and never-smokers) has been observed; smokers had a significantly lesser amount of methylation in the MAOA locus compared with that in never-smokers, and the pattern in ex-smokers was in between.
In one study, 27,000 sites in the peripheral blood DNA were analyzed for methylation. Factor II receptor–like 3 gene (F2RL3) expression was robustly associated with smoking status, and this finding was replicated in two independent samples of European ancestry.
In another study with the same methylation array (27K BeadChip; Illumina, San Diego, CA), the F2RL3 finding was replicated, and additionally a novel association at the G-protein–coupled receptor 15 gene (GPR15) locus was identified.
Recently, higher-density arrays have been used. An epigenome-wide association study in 374 Europeans replicated the smoking-related hypomethylation of F2RL3 and identified three additional loci, including the aryl hydrocarbon receptor repressor gene (AHRR).
Epigenome-wide association study in the European Prospective Investigation into Cancer and Nutrition (EPIC-Turin) identifies novel genetic loci associated with smoking.
In addition to the confirmation of the AHRR methylation, at least eight additional loci were found. Methylation was found to depend on the cessation time and pack-years of smoking.
All of these studies indicate the broad effects of smoking on the genome and cell physiology.
Although the clear effects of smoking on epigenetics have been described, the effects of smoking on genome-wide gene expression in blood is not often studied. Our goals were to describe the smoking-induced changes in the gene-expression profile of the blood and to identify methylated regions that match the altered RNA levels. We analyzed RNA and DNA extracted from the whole blood of donors at the Estonian Genome Center, University of Tartu (Tartu, Estonia).
Materials and Methods
Study Cohort
The study cohort was derived from the Estonian Biobank of the Estonian Genome Center.
The Ethics Review Committee on Human Research of the University of Tartu approved the protocols and informed-consent forms used in this study. All of the participants signed a written informed-consent form.
The Estonian Biobank cohort is a volunteer-based sample of the Estonian resident adult population (aged over 18 years). The current number of participants (52,000) represents 5% of the adult population of Estonia, making it ideally suited to population-based studies. The Biobank stores DNA and peripheral blood mononuclear cells from donors, along with data on lifestyle (eg, smoking and alcohol-intake habits, physical activity). Forty-eight samples (from 24 men and 24 women) were used for gene-expression profiling in the present study.
RNA Extraction
For collecting whole-blood samples, Tempus blood RNA tubes (Thermo Fisher Scientific Inc., Waltham, MA) were used. The samples were stored at −20°C until RNA extraction. For total RNA extraction, the combination of TRIzol reagent (Thermo Fisher Scientific) and the RNeasy Mini Kit (Qiagen, Hilden, Germany) was used. The protocol was as follows—after samples were thawed and mixed the in Tempus tube, blood was transferred to an empty 50-mL tube. The Tempus tube was additionally washed with 3 mL of phosphate-buffered saline. The sample in the 50-mL tube was mixed and centrifuged at 4°C for 60 minutes at 3000 × g. The supernatant was discarded (the invisible precipitation was at the bottom of the tube), and the tube without the cap was placed onto the clean tissue, upside down, for 2 minutes. TRIzol reagent (1 mL) was added into the tube, mixed, and incubated for 5 minutes at room temperature. The sample was lifted to a new 1.5-mL tube, and 200 μL of chloroform was added. The sample was mixed with vortexing for 15 seconds, incubated at room temperature for 2 to 3 minutes, and centrifuged at 4°C for 15 minutes at 12,000 × g. Five hundred microliters of the upper (clear) phase was lifted to a new 1.5-mL tube, and an equal amount of isopropanol was added to the sample. The sample was mixed with vortexing for 15 seconds, incubated at room temperature for 10 minutes, and centrifuged at 4°C for 10 minutes at 12,000 × g. The invisible RNA sediment was at the bottom. The supernatant was discarded, and 1 mL of freshly prepared 75% ethanol was added and centrifuged at 4°C for 5 minutes at 7500 × g. The supernatant was discarded and the ethanol wash step was repeated once more.
After the second wash, the tube was dried with open cap at room temperature for 5 minutes. The RNA was eluted in 50 μL of nuclease-free water and incubated at 55°C for 5 minutes. The DNase treatment was conducted with an Ambion Turbo DNA-free kit (Thermo Fisher Scientific) according to the manufacturer's protocol. The final volume of DNase-treated total RNA was 50 μL. The total RNA was then cleaned with an RNeasy Mini Kit. Three hundred fifty microliters of RLT buffer was added and the sample was mixed, followed by the addition of 1225 μL of 100% ethanol and mixing. The sample was centrifuged through RNeasy Mini Kit columns for 15 seconds at 8000 × g and washed twice with 500 μL of buffer RPE. The RNA was eluted in 50 μL of nuclease-free water. The quality of total RNA was evaluated with an Agilent 2100 Bioanalyzer and the RNA 6000 Nano kit (Agilent Technologies Inc., Santa Clara, CA); the RNA Integrity Number of all of the samples was >5.
Total RNA extraction was conducted at the Estonian Genome Center. From there, 2 μg of total RNA per individual was transported to the Core Facility of Clinical Genomics, University of Tartu, where all of the following procedures were conducted.
Globin Clear Treatment
Total RNA from whole blood consists of up to 70% of Ig mRNA; a Globin Clear Human kit (Thermo Fisher Scientific) was applied to purify the samples from globin mRNA. After Globin Clear treatment, nearly 1.5 μg of RNA was left. The RNA Integrity Number remained >5. The RNA quality was assessed using the Agilent 2100 Bioanalyzer and the RNA 6000 Nano kit (Agilent Technologies).
SOLiD WT RNAseq Library Preparation and Sequencing
Whole transcriptome RNAseq libraries for 48 RNA samples were prepared. Fifty nanograms of each Globin Clear kit–treated total RNA sample was taken as library input. The rest of the RNA material was stored at −80°C. For library preparation, the Ovation RNAseq V2 Kit (NuGen, Emeryville, CA) together with the 5500 Series Fragment Library Core Kit (Thermo Fisher Scientific) were used according to the manufacturers' protocols.
An automated SOLiD EZ Bead System and EZ Bead E80 System Consumables (Thermo Fisher Scientific) were applied for emulsion PCR. With each template preparation, the pool of 12 libraries was used (marked with barcoding sequences to distinguish the samples on data analysis). All together, four template preparations were achieved.
The samples were sequenced with SOLiD 5500 xl platform (Thermo Fisher Scientific) on two flowchips. For each sample, at least 40 million mappable reads were received, which is enough for gene-expression analysis and also fusion and exon junction analysis. Paired-end chemistry for barcoded libraries was used, which provides up to 110 Bp (75 Bp forward and 35 Bp reverse) per one paired-end read.
DNA Methylation Analysis
We also compared methylation patterns between the smokers and never-smokers using MassArray EpiTyper DNA methylation technology (Agena Bioscience, San Diego, CA). Samples were prepared using an EpiTyper T Complete Reagent Set according to the manufacturer's instructions (Agena Bioscience). The bisulfite-treated DNA (25 ng) was amplified with Hot FirePol DNA Polymerase (Solis BioDyne, Tartu, Estonia), and CpG methylation was determined by the MassArray Analyzer 4 (Agena Bioscience). The specific primers were designed with the EpiDesigner software beta version (Agena Bioscience), and the primer sequences for GPR15 were 5′-AGGAAGAGAGTATTGTTTTTTTGGGTGGATAAAGA-3′ and 5′-CAGTAATACGACTCACTATAGGGAGAAGGCTCAATAACAAATCACAATACTCAACAAAA-3′.
Bioinformatics and Statistical Analysis
Raw reads were color-space mapped to the human genome hg19 reference using a Maxmapper algorithm implemented in Lifescope software version 2.5.1 (Life Technologies Corporation, Carlsbad, California). Mapping to multiple locations was permitted. The quality threshold was set at 10, providing a mapping confidence of >90. Reads with a Phred score <10 were filtered out. The mean mapping quality was 30. RNA content and gene-based annotation were analyzed with whole-transcriptome workflow. Raw sequencing data with appropriate experimental information are available from the Gene Expression Omnibus repository (www.ncbi.nlm.nih.gov/geo; accession number GSE68549).
We analyzed RNA samples of blood from 24 smokers and 24 never-smokers with SOLiD 5500xl RNA sequencing technology. There were three main groups: smokers (n = 16), ex-smokers (n = 8) and individuals who had never smoked (n = 14) (Table 1). We analyzed men and women separately and left the data from the ex-smokers out of the initial statistical comparison. The data from ex-smokers were used only for illustrative purposes.
Table 1Characteristics of the Subjects Enrolled in the Present Study
Non-normalized raw counts were used for the edgeR package version 3.2.0 (Bioconductor, http://bioconductor.org) to perform differential gene-expression analysis after quality control of samples. edgeR performs model-based scale normalization, estimates dispersions, and applies negative binomial modeling. edgeR is a flexible tool for RNAseq data analysis to identify differentially expressed genes.
It implements negative binomial model fitting, followed by testing procedures for determining differential expression.
To detect differentially expressed genes, we used negative binomial fitting followed by Fisher exact testing. False-discovery rate adjustment was used for multiple-testing correction.
A false-discovery rate threshold of 0.1 for statistical significance was applied. Genes with greater differential expression were defined with a threshold of log fold-change 0.5 (ie, 50% change between experimental conditions). We analyzed men and women as separate data sets (24 individuals each) (Table 1).
Results
RNA Sequencing
The basic characteristics of the study groups are listed in the Table 1. For each sample, at least 40 million mappable reads were received, which is enough for gene-expression analysis and also fusion and exon junction analysis. Paired-end chemistry for barcoded libraries were used, which gives up to 110 Bp (75 Bp forward and 35 Bp reverse) per one paired-end read. RNA sequencing provided high-quality reads with good similarity between different samples. Multidimensional scaling analysis of fold-change differences in gene expression indicated good separation of study groups by sex (Figure 1). Multidimensional scaling analysis showed the sample distances and illustrated the similarity of samples based on the biological coefficient of variation. As we saw strong separation of samples by sex, to reduce confounding effects, it was better to analyze samples separately by sex. Therefore, in further statistical analysis we used data from men and women separately.
Figure 1Multidimensional scaling (MDS) plot of expression data illustrates a significant effect of sex on variations in global gene expression. Normalized RNAseq counts were used in the MDS analysis to find the similarities of individual cases of a data set. Based on the results from MDS, we decided to analyze data from men (triangles) and women (circles) separately. logFC, log fold-change.
In women, a comparison of gene-expression profiles between smokers and never-smokers revealed differential expression (false-discovery rate <0.1) of five genes: FKBP10, GPR128, GPR15, L1TD1, and SMOC1 (Table 2). Of these, the only gene found to be related to smoking in earlier studies was GPR15. Other genes were not identified in earlier studies as related to smoking.
Table 2Gene Expression of Smokers and Never-Smokers among Women
Gene name
Symbol
logFC
logCPM
P
FDR
G-protein–coupled receptor 128
GPR128
4.36
1.27
1.31 × 10−7
0.003
G-protein–coupled receptor 15
GPR15
1.75
2.35
3.36 × 10−7
0.004
LINE-1 type transposase domain-containing protein 1
L1TD1
1.63
1.36
8.88 × 10−6
0.063
SPARC-related modular calcium binding protein 1
SMOC1
−1.97
−0.37
1.08 × 10−5
0.063
FK506 binding protein 10, 65 kDa
FKBP10
−2.04
−0.43
2.11 × 10−5
0.097
Long intergenic nonprotein–coding RNA 518
C6orf218
−3.15
−1.09
4.29 × 10−5
0.142
Prostaglandin D2 synthase, 21 kDa (brain)
PTGDS
−1.46
1.87
3.81 × 10−5
0.142
Uncharacterized locus MGC21881
MGC21881
−2.16
−0.25
6.56 × 10−5
0.189
RNAseq data were analyzed with edgeR software package version 3.2.0.
FDR, false-discovery rate (P value corrected for multiple testing); LINE, long interspersed nuclear element; logFC, log fold-change (smokers – never-smokers); logCPM, gene expression level in log of counts per million; SPARC, secreted protein, acidic, cysteine-rich protein.
In men, a comparison of the gene-expression profiles between smokers and never-smokers identified differential expression (false-discovery rate <0.1) of 15 genes: C4BPA, C8orf42, CCDC3, CCR8, CD177, CREG1, FRMD4B, FSTL1, GPR15, MYCT1, NFIA, PLCH1, and TPM1 (Table 3). Again, the only gene found in earlier studies to be related to smoking was GPR15. Other genes were not identified in earlier studies as related to smoking. The gene-expression levels of four genes related to smoking status (smokers, ex-smokers, and never-smokers) are illustrated in Figure 2. The lymphocyte antigen 6 complex, locus G6C (LY6G6C) was significantly down-regulated in smokers when we analyzed the entire study sample (men and women together). GPR15 expression correlated well with smoking status. The highest expression was in smokers, and the lowest was in never-smokers. The expression level was in between in ex-smokers (Figure 2). Similarly, expression of the CCDC3 gene followed the smoking status with the lowest level in smokers, intermediate level in ex-smokers, and the greatest level in never-smokers (Figure 2). Interestingly, the expression levels of MYCT1 and LY6G6C were reduced only in smokers and not in ex-smokers.
Table 3Differential Gene Expression of Smokers and Never-Smokers among Men
Gene name
Gene
logFC
logCPM
P
FDR
G-protein–coupled receptor 15
GPR15
2.61
2.96
3.21 × 10−13
7.42 × 10−9
Myc target 1
MYCT1
−1.06
4.38
1.59 × 10−6
0.02
Coiled-coil domain-containing protein 3
CCDC3
−3.59
1.00
5.19 × 10−6
0.03
Cellular repressor of E1A-stimulated genes 1
CREG1
−0.74
8.30
4.52 × 10−6
0.03
Chemokine (C-C motif) receptor 8
CCR8
1.87
1.74
7.26 × 10−6
0.03
Phospholipase C, η1
PLCH1
−1.04
2.98
1.04 × 10−5
0.04
Complement component 4 binding protein α
C4BPA
−3.22
2.60
2.01 × 10−5
0.07
Testis development–related protein
C8orf42
−1.24
1.41
2.48 × 10−5
0.07
CD177 molecule
CD177
−2.60
2.77
4.10 × 10−5
0.09
FERM domain-containing protein 4B
FRMD4B
−0.68
5.76
3.43 × 10−5
0.09
Follistatin-like 1
FSTL1
−1.08
3.83
4.83 × 10−5
0.09
Nuclear factor I/A
NFIA
−0.74
6.20
4.49 × 10−5
0.09
Tropomyosin 1 (α)
TPM1
−0.79
6.69
3.97 × 10−5
0.09
RNAseq data were analyzed with edgeR package.
FDR, false-discovery rate (P value corrected for multiple testing); FERM, 4.1, ezrin, radixin, moesin protein; logFC, log fold-change (smokers – never-smokers); logCPM, gene expression level in log of counts per million.
Figure 2Boxplot illustrating the relationship between the normalized expression of the most significantly differentially expressed genes [GPR15 (A), MYCT1 (B), CCDC3 (C), and LY6G6C (D)] and smoking status in men. The levels of expression of the genes in this figure showed the greatest statistical differences between smokers and never-smokers (Never) on comparison of the RNAseq data (Fisher exact test). For illustrative purposes, the group of ex-smokers was added.
The heatmap of gene-expression data based on 50 genes with the lowest P values illustrated a clear smoking-related expressional pattern (Figure 3). Male smokers were clustered into two groups with characteristic gene-expression profiles. GPR15, RTKN2, USP46, CCR4, and CCR8 formed a cluster of genes up-regulated in smokers (Figure 3). In addition, FOXP3 and GPR15 showed similar expression patterns, suggesting potential correlation. Indeed, analysis found a correlation coefficient of 0.59 between GPR15 and FOXP3, with a P value of 1.15E–05.
Figure 3Heatmap illustrating RNAseq expression (normalized counts) of the 50 genes with the greatest statistical differences between smokers (green) and never-smokers (Never, red) (Fisher exact test). Deep blue indicates greater expression; light green indicates lesser expression.
DNA methylation analysis in the GPR15 locus was performed, and two CpG sites were analyzed. Hypomethylation of CpG1 was significantly greater in smokers compared with that in never-smokers (Figure 4). This finding correlated with the gene-expression results, in which significant overexpression of GPR15 was observed. We did not find differential methylation with another CpG site in the GPR15 gene.
Figure 4Methylation of the GPR15 locus CpG1 is dependent on smoking status. Smokers have significantly greater hypomethylation compared with never-smokers (Never). Pairwise (eg, smoker versus never-smoker) approach and t-test were used for determining statistical significance. ∗P < 0.05 never-smokers versus smokers.
Our study showed that smoking induces the overexpression of the GPR15, and this change was statistically significant in both men and women. The level of GPR15 expression followed smoking status. GPR15 expression was greatest in smokers, in the intermediate range in ex-smokers, and least in never-smokers. Moreover, we found that increased GPR15 gene expression was related to the hypomethylation of the CpG1 locus in the GPR15 gene, and we conclude that hypomethylation of this locus is involved in the up-regulation of GPR15. Although smoking-induced hypomethylation of the GPR15 locus was reported in earlier studies, differential expression of GPR15 RNA caused by smoking has not been described.
Therefore, GPR15 seems to be an interesting target for the biological effects of smoking. The function of GPR15 gives additional impact and significance that this finding can have in the smoking-related pathologies.
GPR15 was discovered as a novel G-protein–coupled receptor in chromosome 3.
It is involved in lymphocyte homing in the large intestine, which the inflammatory bowel disease most commonly affects. Indeed, recent studies have identified a role for GPR15 in the development of inflammation in the colon.
There are substantial species-specific differences in the expression of GPR15 in T cells. In mice, Gpr15 is expressed in regulatory T cells, and in humans, GPR15 is expressed in Th2 cells.
We found a positive correlation (r = 0.59) between the expression levels of GPR15 and FOXP3 (Figure 3). This finding is in agreement with those from a recent study in which the regulatory role of GPR15 on the forkhead box P3–positive (FOXP3+) regulatory cells was described.
Interestingly, smoking seems to have the opposite effect on FOXP3 activation, depending on the subjects and pathological status. In patients with ulcerative colitis, smoking increases the prevalence of FOXP3+ cells, whereas in patients with Crohn disease, it increases the Th1 subsets.
In smokers with normal lung function, smoking prominently up-regulates the regulatory T cells in bronchoalveolar lavage fluid; this effect is absent in patients with chronic obstructive pulmonary disease.
Taken together, the co-regulation of GPR15 and FOXP3 in our study supports the functional relevance of the smoking-induced up-regulation of GPR15. Smoking seems to induce the appropriate transcriptional network necessary for GPR15 induction.
As mentioned in the previous paragraph, GPR15 is involved in the inflammation of the large intestine by mediating T-cell recruitment to the colon. GPR15 is also required for the trafficking of dendritic epidermal T cells to the epidermal tissues.
Therefore, GPR15 is involved in the homing of T cells into the epithelial barrier tissues. In addition to skin and colon, GPR15 has been described in the synovial tissue in patients with rheumatoid arthritis (RA).
The expression level of GPR15 in the peripheral blood and synovia was dependent on the presence of inflammation. These findings taken together suggest that GPR15 is an orphan receptor with a well-recognized role in the regulation of immune response, and smoking increases its expression.
Smoking is recognized as a significant health risk factor for various chronic inflammation. In the case of skin (psoriasis) and intestinal chronic inflammation (Crohn disease and ulcerative colitis), the impact of smoking is well known, albeit controversial.
Even large-scale meta-analyses have found that current smoking is a significant risk factor for Crohn disease but is significantly protective against ulcerative colitis.
The association of smoking with psoriasis is more clear. A recent population-based, cross-sectional study found a significantly increased prevalence of psoriasis in smokers and ex-smokers.
We did not find changes in the gene-expression level of AHR or AHRR. However, we found significant methylation differences in the case of AHRR, but these differences did not have an impact on RNA expression level (data not shown). Our approach was to start from the global gene-expression data to identify the differentially expressed genes and then to find molecular reasons for these differences. In the case of GPR15, this approach worked nicely—we saw differences in RNA levels and in methylation. These findings are a good example in which the results of RNA analysis coincide with epigenetic results, which provides additional support that the found differences have a biological impact. At the same time, the identification of GPR15 is functionally a relevant finding to explain, at least partially, the health effects of smoking.
Our study had some important limitations. Its cross-sectional design did not allow for determining whether the health outcomes were correlated with the expression of GPR15. We also did not have functional immune cells from the study groups to perform more detailed mechanistic analysis. Our results support a causative connection between smoking and GPR15 expression via hypomethylation, but we still were not able to show a full causative pathway from smoking to GPR15 changes and to the development of pathologies. These questions will be the focus of further study.
Conclusion
We found that smoking induces overexpression of GRP15 in the blood, and that this overexpression is caused by the hypomethylation of this locus. This change was evident in both men and women. GPR15 is the chemoattractant receptor that directs the homing of T cells to the colon and skin and is up-regulated in the synovial tissue in cases of rheumatoid arthritis. Therefore, GPR15 and its up-regulation are interesting candidates for the explanation of the health hazards of smoking in chronic inflammatory diseases. The induction of GPR15 expression might be, at least partially, the mechanism of the influence of smoking on the functions of the body.
Acknowledgments
We thank the Estonian Genome Center for providing data and samples.
G.K. analyzed data and prepared the manuscript. E.R. performed RNA sequencing. M.-L.U. and M.L. performed methylation analysis. P.P. performed methylation analysis and helped with data analysis. S.K. conceived the study, analyzed RNAseq data, and wrote the manuscript.
Determination of Methylated CpG Sites in the Promoter Region of Catechol-O-Methyltransferase (COMT) and their Involvement in the Etiology of Tobacco Smoking.
Epigenome-wide association study in the European Prospective Investigation into Cancer and Nutrition (EPIC-Turin) identifies novel genetic loci associated with smoking.
Supported by institutional research funding ( IUT20-46 ) from the Estonian Ministry of Education and Research , the Centre of Translational Genomics of the University of Tartu (SP1GVARENG), and the European Regional Development Fund , Center of Translational Medicine, University of Tartu.
00434-4&id=gr1.jpg){kind=link}
00434-4&id=gr2.jpg){kind=link}
00434-4&id=gr3.jpg){kind=link}
00434-4&id=gr4.jpg){kind=link}