is a cancer-predisposing gene with proven tumour suppressor properties. However, very little evidence on function, especially on pathways, is currently available. In this study, we were able to verify downregulated ARLTS1
expression in the blood-derived RNA of PCa patient samples, demonstrating for the first time the effect of germline alteration on ARLTS1
expression levels. Tumour suppressor function of the ARLTS1
gene has been previously proven by Calin GA et al., who found that transduction of full-length ARLTS1
to A549 cells in Nu/Nu mice decreased tumour growth when compared to empty vector 
. The ability of ARLTS1
to suppress tumour formation in preclinical models has also been observed with ovarian 
and lung cancer cells 
. However, these results are based on somatic mutations in cancerous cell lines, whereas our result reveals a novel expression difference at the germline level, supporting the role of ARLTS1
as a tumour suppressor gene. In the future, these types of findings could be used to enable screening and detection of at-risk patients even before clinical diagnoses.
We have previously genotyped ARLTS1
variants in prostate, breast and colorectal cancer 
and produced a prostate cancer follow-up study 
. In the first study 
, we reported a statistically significant association with ARLTS1
variants T442C, G194T and prostate cancer risk. However, after adjusting for multiple testing, none of the results were significant. In the follow-up study with larger sample size, we reported a statistically significant association with T442C variant and prostate cancer risk 
. We reported also a decreased or lost ARLTS1
RNA or protein expression in clinical prostate tumors, prostate cancer cell lines and xenografts, supporting the role of ARLTS
1 as a tumor suppressor gene. Thus, there is no conflict between the previous publications and this study that is confirming the tumor suppressor role of ARLTS1
Here, 14 eSNPs located at the 13q14 region were shown to significantly influence ARLTS1 transcript levels in PCa patients. The eQTL analysis was performed with two methods, a linear model approach and a directional test. One advantage of our directional test method over the commonly used linear model approach is its robustness against outliers and the weaker model assumptions. In the linear model approach, the expression values of the different genotype groups are considered to follow the same normal distribution up to a location shift parameter. If the location shift parameter is not equal to zero, it generates a testing problem. The directional test only assumes that the different expression values follow certain distribution functions, so a testing problem only occurs if the distributions of the genotype groups are stochastically ordered.
Interestingly, 5 of the eSNPs are located within the protein-coding genes SETDB2
. Two eQTLs were positioned in the SETDB2
(SET domain, bifurcated 2) gene, also known as the CLLD8
(chronic lymphocytic leukemia deletion region gene 8 protein) gene, which functions mainly in epigenetic regulation 
. The PHF11
gene, another hit for the eSNPs, is a positive regulator of Th1-type cytokine gene expression through nuclear factor kappa B, (NF-kB) 
. The third eSNP gene, SPRYD7
, (SPRY domain containing 7/chromosome 13 open reading frame 1 [C13or1]
) also known as chronic lymphocytic deletion region gene 6 protein, (CLLD6
, is proposed to have a tumour suppressor function because it has been detected to be downregulated in B-CLL patients 
. The SPRY/B30.2 protein domain occurs in a variety of cellular proteins, mediates protein-protein interactions and negatively regulates cytokine activities 
. The fourth eSNP gene, MLNR/MTLR1
, (G-protein-coupled receptor 38, GPR38) is a member of the G-protein coupled receptor 1 family and is identified as the motilin receptor 
. The found eSNPs were not affecting the expression of the gene they resided in, suggesting the role of these eSNPs as specific regulators targeting ARLTS1
expression. However, no explicit conclusions about the actual causal interaction between these variants and ARLTS1
can be made without further functional validation. The eQTL result gathered in this study was from a relatively small sample set in which analyses encompassed 1 Mb up- and downstream of the ARLTS1
gene. Thus, further validation in larger sample sets and genomic areas are warranted. However, in our results, one eSNP remained significant even after adjustments for multiple testing. Those adjustments are intricate in the case of eQTL analysis and studies with small sample sizes often suffer from a low detection rate after multiple testing adjustment 
. One possible strategy to deal with this is to accept a relatively high FDR for the sake of later validation in a larger population. Here, no potential candidate eSNPs were excluded from further studies because the α with largest ratio between observed and rejected tests was considered as optimal test size.
Lymphoblastoid cell lines (LCLs) are widely used in eQTL studies because they are an easily accessible source of patient samples of a single cell type. Particularly with prostate cancer, the tumour specimens of multiple foci are hard to collect in large amounts making it difficult to reliably detect disease-associated traits. By using prostate cancer specimens, it is possible to find tissue-specific genetic effects, but the use of lymphoblastoid cell lines or whole blood of familial prostate cancer patients enables one to detect the possible heritable germline differences which may contribute to prostate cancer susceptibility. It has been argued that SNP-transcript approaches using LCLs of small sample sets are underpowered 
, but many studies have been able to find overlapping eQTLs in cell lines and primary tissues 
. Our finding of the eSNP in the SPRYD7
gene was endorsed in the co-expression data, in which the ARLTS1
expression levels were significantly correlated with SPRYD7
expression, and in the prostate tumour specimens (). Additionally, in the data from the ENCODE consortium, the significance and usage of LCLs are justified because the lymphoblastoid cell line (GM12878) is one of the three cell lines in the higher priority Tier 1 cohort 
There may be a joint effect between the 13q14 genes. For example, one larger transcript variant that consists of more than one gene has been proposed with SETDB2
. Very little is known about the interactions in this area, particularly the genomic collaborators of the ARLTS1
gene. A recent study identified cellular retinoic acid binding protein 2 (CRABP2) and phosphoglycerate mutase 1 (PGAM1) as novel ARLTS1-binding proteins using the in-frame cDNA library technique 
. These proteins were not observed in the expression or genotype level of our data, so further studies are needed to elucidate the actual ARLTS1
interacting genes and proteins.
Considering the interactions between ARLTS1
and inflammation pathway genes, it was our hypothesis that ARLTS1
would have interactions with inflammatory genes, as chronic inflammation has been proposed as a prostate cancer risk factor. With MDR analysis, we aimed to investigate the previously observed ARLTS1
connection with immune system processes 
. Another aim was to test the hypothesis of genes functioning in inflammation processes affecting prostate cancer risk. However, we failed to substantiate our hypothesis, and statistically significant association between ARLTS1
T442C SNP and the inflammatory/immune system SNPs was not found within the 700,000 SNPs analysed. For the analysis, we extracted distinct SNP groups from 4,764 markers found from our data of 700,000 variants in total. One possible reason for the negative results may be the relatively limited sample set of 135 cases and controls because we were not able to generalise the results to the test data. We were also concerned about overfitting the data with MDR.
Our previously found association between ARLTS1 T442C and prostate cancer risk may be caused by the interaction network of different SNPs in the 13q14 region. Variant T442C may be included in a very rare haploblock, or there may still be a missing variant because the region has been connected to other cancers in addition to PCa.
In conclusion, in this study we have shown that the ARLTS1 expression level is influenced by changes in germline expression levels and that the ARLTS1 gene is a quantitative trait locus for 14 eSNPs in the 13q14 region. Thus, our results indicate a more complicated network of changes that increase PCa risk than merely one genetic variant in ARLTS1.