The TERT gene harbors multiple variants associated with pancreatic cancer susceptibility

Abstract

A small number of common susceptibility loci have been identified for pancreatic cancer, one of which is marked by rs401681 in the TERT – CLPTM1L gene region on chr5p15.33. Since this region is characterized by low linkage disequilibrium (LD), we sought to identify additional SNPs could be related to pancreatic cancer risk, independently of rs401681. We performed an in-depth analysis of genetic variability of the telomerase reverse transcriptase (TERT) and the telomerase RNA component (TERC) genes, in 5,550 subjects with pancreatic cancer and 7,585 controls from the PANcreatic Disease ReseArch (PANDoRA) and the PanScan consortia. We identified a significant association between a variant in TERT and pancreatic cancer risk (rs2853677, OR=0.85; 95% CI=0.80–0.90, P=8.3×10⁻⁸). Additional analysis adjusting rs2853677 for rs401681 indicated that the two SNPs are independently associated with pancreatic cancer risk, as suggested by the low LD between them (r²=0.07, D´=0.28). Three additional SNPs in TERT reached statistical significance after correction for multiple testing: rs2736100 (P=3.0×10⁻⁵), rs4583925 (P=4.0×10⁻⁵) and rs2735948 (P=5.0×10⁻⁵). In conclusion, we confirmed that the TERT locus is associated with pancreatic cancer risk, possibly through several independent variants.

Introduction

The majority of pancreatic cancer patients die within a year of diagnosis¹. The poor prognosis is caused by various factors, including the lack of appropriate markers for early detection, the aggressiveness of the disease and the dearth of effective treatment possibilities available to the patients diagnosed at a late stage². Therefore, the best hope to reduce mortality among patients is early diagnosis, and a possible strategy to increase chances of early diagnosis is to identify people at high risk in the population and subject them to enhanced surveillance.

Only a few epidemiologic risk factors have been established for pancreatic cancer, including cigarette smoking, heavy alcohol intake, diabetes mellitus (although diabetes or glucose intolerance diagnosed up to three years before diagnosis of cancer may be a result of the malignancy rather than a risk factor)³, obesity, chronic pancreatitis and family history of pancreatic cancer^{4, 5}. Aside from ABO blood group^{6, 7}, even less is known about the genetic contribution to the disease, since only a rather small number of susceptibility loci have been identified through genome-wide association studies^8–11 and confirmed by follow-up studies¹².

The TERT-CLPTM1L gene region on chromosome 5p15.33 is one of these few identified loci for pancreatic cancer risk^{10, 13}. The TERT gene encodes the telomerase reverse transcriptase, which, together with the telomerase RNA component (encoded by the TERC gene), constitute the telomerase complex¹⁴. A correctly functioning telomerase is required for accurate de novo synthesis of telomeric ends. Even moderate changes in TERT and TERC activity can profoundly affect telomere homeostasis¹⁵. Telomeres are highly specialized structures that have key roles in various cellular processes such as chromosomal stability and cell growth^{16, 17} and in proper segregation of chromosomes to daughter cells¹⁸. Overwhelming evidence suggests that telomere dysfunction, mediated by telomerase activation, is a driving force in cancer development¹⁵.

Both TERT and TERC contain pleiotropic risk loci since SNPs in both genes are associated with risks of developing a number of types of human tumors. For example, TERT rs2736100 is associated with glioma, testicular cancer and lung cancer^19–22 while TERT rs401681 is associated with lung, bladder and pancreatic cancer^{10, 23, 24}, TERT rs10069690 with estrogen receptor-negative breast cancer²⁵, TERT rs2242652 with breast, prostate and ovarian cancer^26–28, and TERC rs10936599 with multiple myeloma and colorectal cancer^{29, 30}. The TERT locus is characterized by low linkage disequilibrium (LD), raising the possibility that additional SNPs could be, independently from rs401681 and rs2736098, related to pancreatic cancer risk, given the multiple polymorphic variants that are associated with other cancer types. To elucidate further the role of genetic variability in these two regions in pancreatic cancer risk, we examined 22 SNPs in TERT and 7 in TERC in 5,550 pancreatic ductal adenocarcinoma (PDAC) case subjects and 7,585 controls.

Material and methods

Study populations

We used a two-step strategy with a discovery phase consisting of biological samples from 1,885 PDAC case subjects and 4,048 controls collected in the context of the PANcreatic Disease ReseArch (PANDoRA) consortium, and a validation phase consisting of samples from 3,537 case subjects and 3,665 control subjects collected from studies participating in the PanScan consortium.

The PANDoRA consortium has been described in detail elsewhere³¹. Briefly, individuals with newly diagnosed PDAC were retrospectively identified in seven European countries (Italy, Poland, Germany, Czech Republic, England, Greece, Lithuania) between 1996 and 2012. Controls of Italian, Czech and Polish origin were recruited in the same hospitals, or at least the same geographical regions from where the case subjects were enrolled. British controls were selected from healthy volunteers recruited from the general population in the European Prospective Investigation on Cancer (EPIC), an ongoing prospective cohort study in ten European countries (http://epic.iarc.fr/). The German controls were enrolled in ESTHER, a prospective cohort with 9,953 participants recruited during a general health check-up between July 2000 and December 2002 in Saarland (a state in Southwestern Germany). All subjects signed a written consent from. Relevant characteristics of the populations are shown in table 1.

Table 1.

Description of the PANDoRA consortium population.

	Cases	Controls	Total
Geographic origin
Italy	789	1630	2419
Germanya	536	956	1492
Czech Republic	249	745	994
Greece	70	88	158
Lithuania	57	192	249
Poland	99	320	419
United Kingdomb	101	175	276
Total	1901	4106	6007
Genderc
Males	1093 (58%)	2228 (53%)	3321 (56%)
Females	787 (42%)	1808 (47%)	2595 (44%)
Median age (25%–75% percentiles)d
PANDoRA	64 (19–98)	58 (17–98)

^aCases from PANDoRA, controls from the ESTHER cohort.

^bCases from PANDoRA, controls from the EPIC cohort.

^cNumbers do not add up to the total of subjects because of missing information.

^dAge at diagnosis for cases, age at recruitment for controls.

For the validation phase, we used data from the PanScan consortium. The PanScan study has been fully described elsewhere^{8, 10}. Briefly, case and control data and DNA samples were collected from 12 cohort studies and 8 case-control studies. Cases were defined as those individuals having primary adenocarcinoma of the exocrine pancreas. Controls were frequency matched to cases and were free of pancreatic cancer at the time of enrolment. Matching criteria varied according to the studies within PanScan. Additional information on the matching criteria are given in the original publications^{8, 10}. All subjects signed a written consent form.

SNP selection

Common genetic variability in the TERT gene region was investigated following a hybrid functional and tagging approach to identify candidate SNPs. Within the region of TERT/CLPTM1L (chr5:1277490–1377121, NCBI36/hg18) all SNPs with a minor allele frequency (MAF) > 5% in Caucasians (International HapMap Project, version 28; http://www.hapmap.org) were considered. Tagging SNPs were selected with the use of the Haploview Tagger Program (http://www.broad.mit.edu/mpg/haploview/; http://www.broad.mit.edu/mpg/tagger/)³², using pairwise tagging with a minimum r² of 0.8. We selected additional SNPs significantly associated at a genome-wide level with cancer risk or with telomere length^{26, 28}. For the TERC gene we selected SNPs that have been previously associated with telomere length or cancer risk that also reside in chr3:170974797–170984874 (NCBI36/hg18)^{30, 33, 34}. The final selection consisted of 29 SNPs, 22 in the TERT region and 7 in TERC.

Genotyping

De novo genotyping for the discovery phase was carried out on 1,885 PDAC case subjects and 4,048 controls within PANDoRA at the German Cancer Research Center (DKFZ) in Heidelberg, Germany, on genomic DNA extracted from peripheral blood, using TaqMan (ABI, Applied Biosystems, Foster City, CA, USA) and KASPar (KBioscence, Hoddesdon, UK) technologies. The order of DNA samples from case and control subjects was randomized on plates in order to ensure that similar numbers of cases and controls were analyzed in each batch. For quality control, duplicates of 10% of the samples were interspersed throughout the plates. PCR plates were read on a ViiA7 real time instrument (Applied Biosystems). The ViiA7 RUO Software, version 1.2.2 (Applied Biosystems) was used to determine genotypes. The genotyping concordance between duplicate samples exceeded 99% and the average SNP call fraction was 97.5% (93.6%–99.8%), after all samples with a call fraction lower than 75% were discarded from the analysis. Genotype data used in the second phase were generated as part of PanScan at the National Cancer Institute (NCI) Cancer Genomics Research Laboratory (CGR), using Illumina HumanHap550 and HumanHap550-Duo SNP arrays (PanScan-I) and Illumina Human 610-Quad arrays (PanScan-II). Only SNPs with call rates >94% and samples with call rates >94% were included in the analysis. Participants with <80% European ancestry were excluded from the analysis. The final numbers of cases and controls included in stage 2 were 3,537 case subjects and 3,665 control subjects. An average discordance rate of 0.031% was observed for the 244 duplicate pairs used as quality control. Additional information on the genotyping performed in the PanScan studies is given in the original publications^{8, 10}.

Statistical analysis

Hardy Weinberg Equilibrium (HWE) was assessed in control subjects for each polymorphism. In the first phase, we included genotype data from 1,885 pancreatic cancer case subjects and 4,048 controls. Unconditional logistic regression methods were used to assess the main effects for the 29 selected genetic polymorphisms on PDAC risk, using allelic, co-dominant and dominant inheritance models. For each SNP, the more common allele in controls was assigned as the reference category. All analyses were adjusted for age (continuous), gender and geographic region of origin. In the validation phase, we examined SNPs that showed nominally statistically significant associations (P<0.05) with PDAC risk. For the validation phase, we used the summary results that were calculated in the PanScan-I and II projects, in meta-analysis with our phase-1 data. Of the 29 initial SNPs, 10 had been genotyped in PanScan, while 19 were imputed. Imputation was performed using the 1000 genomes reference dataset (1000G, Version 3, December 2012) (http://www.1000genomes.org/) and IMPUTE2³⁵. All 19 SNPs had quality scores (IMPUTE2 information score) >0.5. The significance threshold of the final analysis was adjusted, taking into account an estimate of the effective number of tests carried out as follows: since residual LD was possible, for each locus we calculated the effective number of independent SNPs, M_eff, using the SNP Spectral Decomposition approach (simpleM method)³⁶. The study-wise M_eff obtained was 18. Additionally, we corrected for the different inheritance models tested (allelic, co-dominant and dominant). Thus, the threshold for statistical significance was 9.26×10⁻⁴ (0.05/(18*3)).

Bioinformatic analysis

We used several bioinformatic tools to assess possible functional relevance for the three SNPs showing the most significant associations with risk of pancreatic cancer. RegulomeDB (http://regulome.stanford.edu/)³⁷ and HaploReg v2B³⁸ were used to identify the regulatory potential of the region nearby each SNP. Genevar (http://www.sanger.ac.uk/resources/software/genevar/)³⁹ was used to identify potential associations between the SNP and expression levels of nearby genes (eQTL).

Results

All analyzed SNPs were in HWE in controls (P>0.05) with the exception of rs16847897 that was then excluded from the following analysis.

SNP main effects

In the discovery phase, in which we genotyped DNA samples in the PANDoRA consortium, we noted 12 TERT and 5 TERC SNPs that were nominally associated with pancreatic cancer risk (P<0.05) considering any genotype comparison. The most significant finding was the association of the minor (G) allele of TERT rs2853677 with decreased pancreatic cancer risk (OR_homozygous=0.70; 95% CI 0.58–0.84; P=1.1×10⁻⁴; P_trend=8.1×10⁻⁵). We also confirmed the previously described association between rs401681 and pancreatic cancer risk (OR_homozygous=1.32; 95% CI=1.12–1.55; P=1.1×10⁻³; P_trend=1.1×10⁻³). The complete results for analysis of the TERT SNPs are shown in table 2. For the TERC gene, the most significant association was for the minor allele (T) of rs10936599 and decreased PDAC risk (OR_heterozygous=0.78; 95% CI=0.69–0.89; P=10⁻⁴, P_trend=8.9×10⁻³). The complete results for analysis of TERC SNPs are shown in table 3. Supplementary table 4 shows stratified analysis divided by country of origin.

Table 3.

Associations between pancreatic cancer risk and SNPs in the TERC gene regions (phase 1).

SNP	Alleles (M/m)a	Cases/Controlsb			M vs mc	Pallele	Mm vs MM	Phet	mm vs MM	Phom	Mm+mm vs MM	Pdom	Ptrend
		MM	Mm	mm
rs10936599	C/T	1126/2152	578/1395	101/235	0.84 (0.76–0.93)	0.001	0.78 (0.69–0.89)	0.0001	0.82 (0.64–1.06)	0.139	0.79 (0.70–0.89)	0.0001	0.0089
rs10936603	G/T	1045/2036	573/1364	119/280	0.87 (0.79–0.96)	0.005	0.81 (0.71–0.92)	0.001	0.86 (0.68–1.09)	0.205	0.82 (0.73–0.92)	0.001	0.1148
rs11709840	A/C	1009/1918	656/1498	150/346	0.87 (0.80–0.96)	0.004	0.83 (0.73–0.93)	0.003	0.83 (0.67–1.03)	0.085	0.83 (0.73–0.93)	0.001	0.1714
rs12696304	G/C	1001/1714	629/1284	124/262	0.89 (0.81–0.99)	0.025	0.85 (0.75–0.97)	0.015	0.87 (0.68–1.1)	0.244	0.85 (0.76–0.97)	0.012	0.0867
rs16854453	G/A	971/2092	553/1387	105/284	0.87 (0.79–0.96)	0.007	0.84 (0.74–0.96)	0.009	0.82 (0.64–1.05)	0.112	0.84 (0.74–0.95)	0.005	0.1399
rs1920116	G/A	1014/1935	702/1486	152/344	0.91 (0.83–1.00)	0.049	0.89 (0.79–1.01)	0.064	0.87 (0.70–1.08)	0.194	0.89 (0.79–1.0)	0.042	0.5162

^aM = major allele (i.e. more common in controls); m = minor allele (less common in controls).

^bNumbers may not add up to 100% due to genotyping failure, DNA depletion or covariate missing values.

^cM vs m = quantitative additive (allelic) model; Mm vs MM = heterozygous carriers vs common homozygous; mm vs MM = rare homozygous vs common homozygous; Mm+mm vs MM= heterozygous carriers + rare homozygous vs common homozygous (dominant model). Odds ratio (95% confidence interval). All analyses were adjusted for age at diagnosis/age at recruitment, gender and country of origin.

As a second step, we performed a meta-analysis between our discovery phase and previously generated PanScan data. We considered associations supported by P<9.26×10⁻⁴ as statistically significant. We identified one SNP in the TERT gene, rs2853677, that was significantly associated with PDAC risk (OR_allele=0.85; 95% CI=0.80–0.90; P=8.3×10⁻⁸). A second SNP in TERT, rs2736100, was associated with PDAC risk (OR_allele=0.90; 95% CI=0.85–0.94; P=3×10⁻⁵). In addition, we observed another statistically significant associations with pancreatic cancer risk in TERT: rs2735948 (OR_homozygous=1.27; 95% CI: 1.13–1.43; P=5×10⁻⁵). We also replicated the association between rs401681 and pancreatic cancer. A tendency for some SNPs to be associated with pancreatic cancer risk only in cohorts or only in case-control studies has already been observed in the context of PanScan^{8, 10}. Therefore, we performed an additional meta-analysis for rs4583925 excluding the cohorts. This analysis showed that the association with pancreatic cancer risk was stronger in the meta-analysis using only the case-control studies for rs4583925 (OR_{meta-case controls}=1.36; 95% CI: 1.17–1.57;P=4.0×10−5) and for rs13190087 (OR_{meta-case controls}=1.41; 95% CI: 1.17–1.71;P=0.0003). Table 4 shows results for all SNPs that reached study-wise significance (P<9.26×10⁻⁴). The results for the meta-analyses of PANDoRA and PanScan for all SNPs that were significant in phase one are shown in supplementary table 1.

Table 4.

Polymorphisms significantly associated with pancreatic cancer risk after adjustment for multiple testing.

Gene	SNP	Study	OR	95% CIa	P valueb
TERT	rs401681d	Pandora	1.32	(1.12–1.55)	0.001
		Panscan I + II	1.40	(1.23–1.60)	8.×10−7
		Meta-analysis	1.37	(1.24–1.42)	1.9×10−9
TERT	rs2853677c	Pandora	0.83	(0.76–0.91)	4.3×10−5
		Panscan I + II	0.86	(0.79–0.93)	1.2×10−4
		Meta-analysis	0.85	(0.80–0.90)	8.3×10−8
TERT	rs2736100c,f	Pandora	0.90	(0.83–0.98)	0.013
		Panscan I + II	0.90	(0.84–0.96)	0.0014
		Meta-analysis	0.90	(0.85–0.94)	3×10−5
TERT	rs4583925d,e	Pandora	1.38	(1.15–1.66)	0.001
		Panscan I + II	1.11	(0.96–1.30)	0.16
		Panscan I + II (case/control studies)	1.32	(1.04–1.70)	0.02
		Meta-analysis	1.21	(1.08–1.36)	0.001
		Meta-analysis (case/control studies)	1.36	(1.17–1.57)	4×10−5
TERT	rs2735948d	Pandora	1.30	(1.10–1.54)	0.002
		Panscan II	1.25	(1.06–1.47)	0.01
		Meta-analysis	1.27	(1.13–1.43)	5×10−5
TERT	rs13190087d,e	Pandora	1.57	(1.21–2.04)	0.001
		Panscan I+II	1.04	(0.85–1.27)	0.68
		Panscan I+II (case/control studies)	1.26	(0.96–1.66)	0.099
		Meta-analysis	1.27	(0.85–1.90)	0.251
		Meta-analysis (case/control studies)	1.41	(1.17–1.71)	0.0003

^a95% CI = 95% confidence intervals.

^bSince no heterogeneity was observed for the selected polymorphisms between the studies, we used a fixed-effects meta-analysis; for every SNP in the meta-analysis we considered the most significant association observed in phase one (i.e. homozygotes (co-dominant model) for the rare allele for rs401681 and rs2735948, carriers of the rare allele (allelic model) for rs2853677 and rs2736100, heterozygotes (co-dominant model) for rs4583925).

^cSNP genotyped in PanScan.

^dSNP imputed in PanScan.

^eResults reported in PanScan (Amundadottir 2009; Petersen 2010) prompted us to analyze separately the cohorts and case-control studies for all SNPs that after phase 1 were associated with risk at p<0.05. The complete results are reported in the Supplementary table 1.

^fThe reference allele in PANDoRA and in PanScan are inverted, therefore we changed it in PANDoRA in order to perform a correct meta-analysis

The rs2853677 and rs2736100 polymorphisms were moderately linked to each other (r²= 0.53) and in very low LD with the previously identified rs401681 PDAC risk locus (r²=0.07 and r²=0.01, respectively). rs4583925 and rs2735948 are not correlated with each other (r²=0.003 and D’=0.277) and rs2735948 showed moderate LD with rs401681 (r²=0.371 and D’=0.663). The last SNP rs13190087 has a moderate LD with all the other SNPs and its association with pancreatic cancer risk is probably only a reflection of this (supplementary table 2 shows the LD between the SNPs as calculated by the SNAP software⁴⁰).

Possible functional effects

We used several bioinformatic tools to predict possible functional relevance of the SNPs showing the most significant associations. Using Genevar, we observed that the A allele of rs2853677 was associated with increased gene expression of two genes in cis: the solute carrier family 6 member 18 (SLC6A18) and the zinc finger DHHC domain-containing protein 11 (ZDHHC11). These associations (P=0.014), however, were not below the threshold suggested by Genevar for significance (P<10⁻³). RegulomeDB showed a score of 5, indicating the possible presence of a transcription factor binding motif or a DNase sensitivity peak. For rs4583925 HaploReg suggested the presence of DNase sensitivity peak in pancreatic islets and in pancreatic adenocarcinoma tissues. In addition, this SNPs showed an association, statistically significant, with ZDHHC11 gene expression (P=10⁻⁴). Bioinformatics approaches did not reveal possible functional explanations for rs2735948 (supplementary table 3 shows the results from HaploReg).

Discussion

Multiple independent polymorphic variants in the 5p15.33 region, that includes the TERT and the CLPTM1L genes, are associated with the development of cancer in various organs^{10, 13, 19, 22, 23, 25, 27, 28, 41, 42}. This region is characterized by a low degree of linkage disequilibrium which allows for the possibility that several independent variants might be simultaneously associated with individual cancer sites, as has been shown for lung, prostate and bladder cancer²⁰. Thus, we sought to analyze this region in detail in a large-scale study, to determine whether multiple variants associate with risk of pancreatic cancer. Indeed, we report reduced risk associated with the G allele of rs2853677 (P=8.3×10⁻⁸). This SNP has previously been associated with glioma in Chinese subjects⁴³ and with lung cancer in Japanese subjects⁴⁴ although the allele associated with the increase in risk of the disease is the other one (A), a phenomenon that has been observed for other SNPs of this region. We found this SNP to be independent of rs401681 and rs2736098, the previously identified pancreatic cancer susceptibility loci, as clearly shown by the low LD between them (r²=0.07 between rs401681 and rs2853677; r²=0.23 between rs2736098 and rs2853677).

In TERT, rs2853677 is located in the first intron, a region that may play a role in the regulation of the gene expression, since it lies in a DNase I hypersensitive region. Bioinformatic analysis of rs2853677 using functional data from the Encyclopedia of DNA Elements (ENCODE) Project⁴⁵ obtained through HaploReg, Regulome DB and Genevar, suggested that the A allele may be associated with increased expression of two genes: the solute carrier family 6 member 18 (SLC6A18), a neutral amino acid transporter, and the zinc finger DHHC domain-containing protein 11 (ZDHHC11), the function of which is not clear yet. This suggestive association should be validated in an independent sample set.

On the other hand, rs2853677 is associated with leukocyte telomere length (LTL) and in particular the A (risk allele) is associated with longer LTL⁴⁶. It is interesting to note that in two recent prospective studies, longer LTL have been shown to be associated with increased risk of pancreatic cancer^{47, 48}. This is consistent with our finding that the G allele, which is associated with decreased pancreatic cancer risk in our study, is also associated with shorter telomeres in the study by Melin and colleagues. Thus it is possible that the link between rs2853677 and pancreatic cancer occurs via the variation of telomere length and in particular that the A allele leads to constitutively longer telomeres, that may in turn be responsible for the increase in pancreatic cancer risk. On the other hand, in another study, based on a retrospective case-control study, shorter telomeres were associated with increased risk of pancreatic cancer⁴⁹. The functional relevance of the association between rs2853677 and pancreatic cancer is therefore currently unclear and additional research is required.

Another SNP, rs2736100, that has been associated with risk of multiple cancer types⁵⁰, is in moderate LD (r²=0.538, D´=0.798) with rs2853677. In our study, rs2736100 shows an association with pancreatic cancer risk (P=3.0×10⁻⁵). These two SNPs are very close to each other (678bp) and the fact that both are strongly associated with the disease but that their clear functional effects cannot be demonstrated opens the possibility that there might be a yet unknown variant that is in LD with both SNPs and underlies the increased risk of the disease.

Another finding of potential significance is the association between the minor allele of rs4583925 and increased pancreatic cancer risk. This SNP is completely independent of both rs401681 and rs2853677, and bioinformatic analysis suggests that this SNP might also be involved in the regulation of ZDHHC11. The fact that two pancreatic susceptibility SNPs that are completely independent of each other (rs2853677 and rs4583925) could both influence the expression of the same gene suggests the possible involvement of ZDHHC11 in pancreatic cancer, although functional studies are needed to validate and better characterize this suggestive association. Moreover, for rs4583925, HaploReg shows that the SNP may lie in a pancreas-specific DNAse sensitivity region. This finding, if confirmed by functional studies, could be of importance in identifying a novel regulatory region for the TERT gene.

The major strength of this study is its size, since with a total of 5,550 subjects with PDAC and 7,585 control subjects; this is the largest genetic analysis of pancreatic cancer risk published to-date. Additionally, our selection of SNPs provides an extensive coverage of genetic diversity in the regions of interest, since we have represented, through tagging, more than 90% of common genetic variability in the TERT and TERC loci. Possible limitations of the study may be the fact that the vast majority of the subjects included were of Caucasian origin and therefore we cannot extend the findings to other populations and that patients and controls in PANDoRA were recruited in various centers across Europe and therefore there might be some population stratification. Additionally we used only bio-informatic tools to assess the possible functional effect of the SNPs

In conclusion, our results suggest that the TERT locus is significantly associated with pancreatic cancer risk, likely through more than one variant. We a possible new association between rs2853677 and risk of pancreatic cancer.

However we were not able to find mechanistic link between the association and the disease apart from a possible role in determination of telomere length and therefore our results have to be taken with caution. The next logic step to confirm the findings would be to perform functional studies in order to characterize the described associations.

Table 2.

Associations between pancreatic cancer risk and SNPs in the TERT gene regions (phase 1).

SNP	Alleles (M/m)a	Cases/Controlsb			M vs mc	Pallele	Mm vs MM	Phet	mm vs MM	Phom	Mm+mm vs MM	Pdom	Ptrend
		MM	Mm	mm
rs10069690	C/T	1025/2117	643/1429	127/257	0.95 (0.87–1.05)	0.299	0.91 (0.80–1.03)	0.123	0.99 (0.78–1.25)	0.908	0.92 (0.82–1.03)	0.16	0.8009
rs10078761	A/T	719/1478	804/1744	251/569	0.93 (0.85–1.01)	0.081	0.92 (0.81–1.05)	0.199	0.86 (0.72–1.03)	0.111	0.91 (0.80–1.02)	0.107	0.2013
rs13190087	T/G	1764/3445	118/154	3/5	1.52 (1.18–1.94)	0.001	1.57 (1.21–2.04)	0.001	1.30 (0.28–6.01)	0.74	1.56 (1.2–2.02)	0.001	0.028
rs2075786	C/T	839/1767	799/1689	239/409	1.10 (1.01–1.19)	0.036	1.03 (0.91–1.16)	0.676	1.28 (1.06–1.55)	0.011	1.07 (0.96–1.21)	0.223	0.0769
rs2242652	C/T	1254/2551	496/1170	80/150	0.89 (0.80–0.99)	0.034	0.82 (0.72–0.94)	0.003	1.01 (0.75–1.36)	0.967	0.84 (0.74–0.95)	0.007	0.2155
rs2735948	C/T	564/1199	861/1735	403/635	1.13 (1.04–1.23)	0.005	1.03 (0.90–1.18)	0.696	1.30 (1.10–1.54)	0.002	1.10 (0.97–1.25)	0.137	0.0205
rs2736098	G/A	980/1839	584/1307	126/251	0.96 (0.87–1.06)	0.381	0.89 (0.78–1.01)	0.08	1.04 (0.81–1.32)	0.778	0.91 (0.81–1.03)	0.147	0.1929
rs2736100	G/T	418/932	861/1763	445/817	1.11 (1.02–1.21)	0.013	1.09 (0.94–1.26)	0.259	1.24 (1.05–1.47)	0.013	1.14 (0.99–1.30)	0.071	0.008
rs2736109	C/T	664/1277	740/1700	242/600	0.91 (0.83–0.99)	0.03	0.87 (0.76–1.00)	0.047	0.84 (0.70–1.01)	0.067	0.86 (0.76–0.98)	0.024	0.0332
rs2736122	C/T	1099/2311	671/1483	114/259	0.95 (0.86–1.04)	0.245	0.96 (0.85–1.08)	0.472	0.88 (0.69–1.11)	0.276	0.94 (0.84–1.06)	0.327	0.3092
rs2853676	G/A	1012/2147	712/1606	136/296	0.95 (0.87–1.05)	0.32	0.92 (0.82–1.04)	0.202	0.96 (0.77–1.21)	0.737	0.93 (0.83–1.04)	0.217	0.2447
rs2853677	A/G	581/1021	804/1729	273/659	0.83 (0.76–0.91)	4.3×10−5	0.81 (0.70–0.93)	0.002	0.70 (0.58–0.84)	0.0001	0.78 (0.68–0.88)	0.0001	8.1×10−5
rs2853690	C/T	1272/2411	435/908	53/95	0.92 (0.82–1.04)	0.182	0.90 (0.78–1.03)	0.121	0.96 (0.67–1.37)	0.806	0.90 (0.79–1.03)	0.13	0.1892
rs2853691	T/C	866/1772	695/1646	141/342	0.89 (0.81–0.98)	0.014	0.87 (0.77–0.98)	0.026	0.82 (0.66–1.02)	0.078	0.86 (0.76–0.97)	0.013	0.0169
rs401681	C/T	497/1260	923/1977	437/811	1.15 (1.06–1.25)	0.001	1.17 (1.02–1.34)	0.023	1.32 (1.12–1.55)	0.001	1.21 (1.07–1.38)	0.003	0.0015
rs4246742	A/T	1265/2788	491/924	47/90	1.18 (1.06–1.32)	0.004	1.24 (1.09–1.42)	0.001	1.13 (0.77–1.65)	0.524	1.23 (1.08–1.4)	0.002	0.0091
rs4583925	G/A	1537/3101	243/386	6/15	1.31 (1.10–1.55)	0.002	1.38 (1.15–1.66)	0.001	0.81 (0.30–2.17)	0.679	1.36 (1.13–1.63)	0.001	0.0051
rs4635969	C/T	1075/2482	571/1302	85/168	1.01 (0.92–1.12)	0.811	0.99 (0.87–1.12)	0.823	1.10 (0.83–1.45)	0.501	1.00 (0.89–1.13)	0.991	0.7027
rs4975605	C/A	505/1080	783/1750	326/810	0.93 (0.85–1.01)	0.075	0.99 (0.86–1.14)	0.878	0.85 (0.71–1.01)	0.06	0.94 (0.83–1.08)	0.385	0.2505
rs655475	G/A	1367/2653	452/887	38/71	0.99 (0.87–1.11)	0.809	0.97 (0.84–1.11)	0.641	1.07 (0.70–1.63)	0.767	0.97 (0.85–1.11)	0.708	0.6902
rs7705526	C/A	820/1492	819/1630	195/406	0.91 (0.83–1.00)	0.045	0.91 (0.80–1.03)	0.147	0.83 (0.68–1.02)	0.076	0.90 (0.79–1.01)	0.07	0.0589
rs7726159	C/A	817/1472	784/1641	234/465	0.91 (0.83–0.99)	0.028	0.85 (0.75–0.96)	0.011	0.87 (0.72–1.05)	0.148	0.85 (0.76–0.96)	0.009	0.071

^aM = major allele (i.e. more common in controls); m = minor allele (less common in controls).

^bNumbers may not add up to 100% due to genotyping failure, DNA depletion or covariate missing values.

^cM vs m = quantitative additive (allelic) model; Mm vs MM = heterozygous carriers vs common homozygous; mm vs MM = rare homozygous vs common homozygous; Mm+mm vs MM= heterozygous carriers + rare homozygous vs common homozygous (dominant model). Odds ratio (95% confidence interval). All analyses were adjusted for age at diagnosis/age at recruitment, gender and country of origin.

Novelty & Impact Statements.

We found that four common variants in the TERT/CLPTM1L region show a significant impact on pancreatic cancer risk. The most statistically significant association found was at rs2853677, a SNP that has been associated with telomere length. The association between rs2853677 and pancreatic cancer risk is independent from the well known rs401681 SNPs that lies in the same region, which implies that pancreatic cancer risk is modulated by the TERT gene variation through multiple variants.