Common genetic variants associated with pancreatic adenocarcinoma may also modify risk of pancreatic neuroendocrine neoplasms

Common genetic variants associated with pancreatic adenocarcinoma may also modify risk of... Abstract Pancreatic neuroendocrine neoplasms (pNEN) account for less than 5% of all pancreatic neoplasms and genetic association studies on susceptibility to the disease are limited. We sought to identify possible overlap of genetic susceptibility loci between pancreatic ductal adenocarcinoma (PDAC) and pNEN; therefore, PDAC susceptibility variants (n = 23) from Caucasian genome-wide association studies (GWAS) were genotyped in 369 pNEN cases and 3277 controls from the PANcreatic Disease ReseArch (PANDoRA) consortium to evaluate the odds associated with pNEN risk, disease onset and tumor characteristics. Main effect analyses showed four PDAC susceptibility variants—rs9854771, rs1561927, rs9543325 and rs10919791 to be associated with pNEN risk. Subsequently, only associations with rs9543325, rs10919791 and rs1561927 were noteworthy with false positive report probability (FPRP) tests. Stratified analyses considering age at onset (50-year threshold), showed rs2736098, rs16986825 and rs9854771 to be associated with risk of developing pNEN at a younger age. Stratified analyses also showed some single nucleotide polymorphisms to be associated with different degrees of tumor grade, metastatic potential and functionality. Our results identify known GWAS PDAC susceptibility loci, which may also be involved in sporadic pNEN etiology and suggest that some genetic mechanisms governing pathogenesis of these two entities may be similar, with few of these loci being more influential in younger cases or tumor subtypes. Introduction Pancreatic neuroendocrine neoplasms (pNEN) arise from islet cells and comprise less than 5% of all new pancreatic neoplasms (1). Heterogeneous in tumor behavior and clinical symptoms, pNENs account for 10% of neuroendocrine tumors and have a 5-year mortality rate of 60% (1,2). Compared to the more common exocrine pancreatic ductal adenocarcinoma (PDAC), pNENs are characterized by a rather ‘silent’ clinical course and generally present at a relatively earlier age (median age at diagnosis 53–60 years, and 72 years for PDAC) (1). Majority (40–90%) of pNENs are non-functional (i.e. do not secrete hormones that cause systemic effects) thus constituting a clinical and prognostic challenge for physicians. Surgical resection is currently the primary curative therapy option for both malignancies, however, most patients unfortunately exhibit unresectable tumors at diagnosis (3), stressing the need for improved risk stratification and timely diagnostic biomarkers. Familial clustering observed in population-based studies among PDAC and other cancer patients provide evidence of an inherited basis for sporadic pNEN (1). Although there is no clear link between environmental exposures and risk of developing pNEN, an overlap of risk factors in PDAC and pNENs is likely (4). Increased interest in these neoplasms has become evident in the last decade, but despite this, molecular understanding of pNENs is still insufficient to drive clinical interventions. Recent exome and whole genome sequencing studies describing the genetic basis of pNENs among Caucasians (5,6) have implicated somatic mutations in four core pathways: activation of mTOR signaling, DNA damage repair, chromatin modification and altered telomere length in tumorigenesis. Genome wide association (GWAS) or candidate gene or pathway studies investigating the etiology of pNENs have been reasonably limited by small sample size, when compared with PDAC for which 23 common risk variants have so far been identified through GWAS or studies focusing on genes situated in well-known pleiotropic regions (7–11). Based on the rationale that known common genetic susceptibility loci for sporadic PDAC may also modify the risk of developing pNEN, we genotyped 369 pNEN cases from a retrospectively enrolled population within the PANcreatic Disease ReseArch (PANDoRA) consortium and compared genotype frequencies of all known PDAC GWAS susceptibility variants to over 3200 control subjects from the same consortium. We tested their possible association with overall sporadic pNEN risk and within pNEN subgroups. Materials and methods Selection of subjects and polymorphisms Demographic characteristics of subjects within the PANDoRA consortium have been described previously (12) although the number of neuroendocrine cases within PANDoRA has increased since that publication. Written informed consent and biospecimens (blood, tissue or genomic DNA) were obtained for all subjects and ethical approval for the PANDoRA study protocol received from the Ethics commission of the Medical Faculty Heidelberg. Only sporadic (non-familial) pNEN cases not associated with genetic syndromes such as multiple endocrine neoplasia type 1 (MEN-1), von Hippel–Lindau syndrome, von Recklinghausen disease (neurofibromatosis NF-1) and tuberous sclerosis complex were used in this study. Twenty-three single nucleotide polymorphisms (SNPs) reaching genome-wide significance and suggestive levels (up to 5 × 10–7) were selected from GWAS publications in PDAC among Caucasians (7–10). pNEN tumors were classified according to the World Health Organization 2010 classification system (13). Genotyping and quality control SNPs were genotyped in 369 pNEN cases and 3277 controls using TaqMan allelic discrimination (Applied Biosystems) and KASPar (LGC Genomics) assays according to manufacturers’ instructions. Post-PCR allelic discrimination was done on the 7900HT real-time PCR system (Life Technologies) and data analyzed using the affiliated SDS software (v.2.3). Internal replicates (7% of all samples) and negative controls were included to assess the fidelity of genotype calls. Deviation from Hardy–Weinberg equilibrium was checked using the Pearson’s χ2 test. Statistical analyses We evaluated associations [odds ratios (OR) and 95% confidence intervals (CI)] between SNPs and pNEN risk using unconditional multivariate logistic regression in STATA v.11 (StataCorp LP). Tests were done using a codominant inheritance model, with adjustment for potential confounding variables—age, geographical origin and gender. We used a nominal significance threshold (P < 0.05) since the selected SNPs were based on prior evidence of their association with pancreatic cancer. Noteworthiness of SNP associations with P < 0.05 were tested using the Bayesian false positive report probability (FPRP) test which takes the observed P value, statistical power of the test and prior probabilities for the associations into account (14). All selected SNPs bore a high prior possibility of being associated with pNEN risk (π = 0.2) as recommended by Wacholder et al. (14), and only SNP associations with P < 0.05 and π < 0.2 were considered ‘noteworthy’. To evaluate the effect of SNPs on pNEN onset, we first performed a case-only analysis comparing genotype distributions among cases (>50 years) with their younger counterparts (≤50 years) while adjusting for gender and geographical origin. We also compared genetic associations from three additional models: cases ≤50 years versus controls of all ages (EOP1), cases >50 years versus controls of all ages (TOP) and cases up to 50 years versus controls up to 50 years (EOP2). Associations between SNPs and tumor grade and stage were estimated by case-only analyses. For tumor stage, we compared stage IV tumors versus all other tumors (stages I, IIa, IIb, IIIa, IIIb) based on European Neuroendocrine Tumor Society (ENETS) classification system (15) and for grade, we compared well differentiated tumors (grades 1 and 2) versus poorly differentiated tumors (grade 3) based on the World Health Organization (WHO)/ENETS guidelines (16). Tumor functional status (e.g. insulinoma, gastrinomas, etc) was available for 54% of all cases in this study, out of which 62% were non-functioning. We compared genotype frequencies of all SNPs in the latter with functioning tumors. Microarray data screening To gain additional insight to the transcriptomic profiles of these gene regions in human pancreatic islet cells, we explored expression profiling data from the Gene Expression Omnibus (GEO) database (Accession number GSE43795); this study characterized mRNA and microRNA expression profiles in solid-pseudopapillary neoplasms of pancreas (n = 14), ductal adenocarcinoma (n = 6), pancreatic neuroendocrine tumors (n = 6) and non-neoplastic pancreatic tissue samples (n = 5) (17). Results were downloaded in text format and GEO2R queries (http://www.ncbi.nlm.nih.gov/geo/geo2r/) applied to assess differentially expressed genes in pNEN compared to non-neoplastic pancreas tissue samples. GEO2R is an R programming language-based dataset analysis tool that compares two or more groups of samples under the same experimental conditions and analyzes almost any GEO series (18). P values were adjusted to correct for false-positives due to multiple testing using Benjamini and Hochberg false discovery rate method and genes that met the cut-off criteria of Padjusted < 0.05 and |logFC| > 1.0 were screened out as differentially expressed genes. Results Baseline characteristics of the study population are summarized in Table 1. The median age at diagnosis in this study was 57 years, congruent with previous literature reports. Genotypes for most SNPs, except rs505922 (P = 0.04) were distributed in accordance to Hardy–Weinberg equilibrium among controls. Duplicated samples showed genotyping concordance rate of 99.6%. Genotyping data for those subjects with genotyping completion rates below 75% (n = 35) were excluded from all statistical analyses. Table 1. Clinical characteristics of cases and controls used in this study Region of origin  Germany  Greece  Italy  Poland  United Kingdom  Total subjectsa  Median age at diagnosisb, years (1st–3rd quartile)  51 (44–58)  45 (35–58)  61 (49–72)  46 (31–66)  66 (59–74)  57 (44–67)   ≤50 years  383  151  491  237  15  1277   >50 years  394  75  1238  185  209  2101  Gender   Males  356  135  1098  185  130  1906   Females  421  95  814  257  94  1681  Functional status   Yes  4  20  47  3  3  77   No  n/a  2  118  3  n/a  123   Unknown  31  0  82  9  47  169  Grade, G (WHO)   G1 and G2  n/a  21  149  9  29  208   G3  n/a  1  6  2  1  10   Unknown  35  0  12  4  20  71  Stage (ENETS)   I/II/III  17  n/a  133  12  n/a  162   IV  5  n/a  32  2  n/a  39   Unknown  13  22  82  1  50  168  Total cases  35  22  247  15  50  369  Total controls  797  208  1670  427  175  3277  Region of origin  Germany  Greece  Italy  Poland  United Kingdom  Total subjectsa  Median age at diagnosisb, years (1st–3rd quartile)  51 (44–58)  45 (35–58)  61 (49–72)  46 (31–66)  66 (59–74)  57 (44–67)   ≤50 years  383  151  491  237  15  1277   >50 years  394  75  1238  185  209  2101  Gender   Males  356  135  1098  185  130  1906   Females  421  95  814  257  94  1681  Functional status   Yes  4  20  47  3  3  77   No  n/a  2  118  3  n/a  123   Unknown  31  0  82  9  47  169  Grade, G (WHO)   G1 and G2  n/a  21  149  9  29  208   G3  n/a  1  6  2  1  10   Unknown  35  0  12  4  20  71  Stage (ENETS)   I/II/III  17  n/a  133  12  n/a  162   IV  5  n/a  32  2  n/a  39   Unknown  13  22  82  1  50  168  Total cases  35  22  247  15  50  369  Total controls  797  208  1670  427  175  3277  ENETS, European Neuroendocrine Tumor Society; WHO, World Health Organization; y, years; n/a, not available. aNumbers may not add up to 100% due to genotyping failure or unavailable covariate data. bpNEN cases only. View Large Overall pNEN risk analyses Minor allele frequencies, genotype distributions, risk estimates [odds ratios (OR) and 95% confidence intervals (CI)] of all polymorphisms investigated in this study are summarized in Table 2. rs9543325 was associated with higher risk of pNEN (OR 1.63, 95% CI (1.10–2.47), Phom = 0.02). rs10919791, rs1561927 and rs9854771 were associated with lower risk of pNEN (OR 0.65, 95% CI (0.46–0.93), Phet = 0.02; OR 0.71, 95% CI (0.55–0.92), Phet = 0.01 and OR 0.76, 95% CI (0.58–0.99), Phet = 0.04 respectively). Subsequent testing of these findings using FPRP with a prior of 0.2 showed that only associations with rs10919791 (FPRP = 0.13), rs9543325 (FPRP = 0.15) and rs1561927 (FPRP = 0.07) were noteworthy. The association with rs9854771 failed the noteworthiness test (FPRP = 0.25). Table 2. Associations between known GWAS PDAC susceptibility loci and pNEN risk SNP (nearby gene(s))  Genotypes (cases/controlsa)  MAF CEU 1KG/current studyb  MM versus Mmc  MM versus mmc  OR (95% CI)d  P  OR (95% CI)d  P  rs351365 (WNT2B)  CC (198/1549) CT (108/938) TT (26/172)  0.27/0.24  0.91 (0.69–1.20)  0.50  1.34 (0.82–2.18)  0.24  rs3790844 (NR5A2)  TT (140/1231) TC (64/721) CC (19/123)  0.36/0.23  0.82 (0.59–1.13)  0.22  1.27 (0.74–2.16)  0.38  rs1486134 (ETAA1)  TT (162/1569) TG (141/1120) GG (31/217)  0.27/0.27  1.20 (0.94–1.55)  0.13  1.36 (0.88–2.09)  0.16  rs9854771 (TP63)  GG (149/1131) GA (140/1311) AA (51/350)  0.29/0.36  0.76 (0.58–0.99)  0.04  1.00 (0.69–1.46)  0.98  rs2736098 (TERT)  GG (158/1514) GA (130/1131) AA (19/211)  0.27/0.27  1.08 (0.84–1.39)  0.56  0.84 (0.50–1.41)  0.51  rs17688601 (SUGCT)  AA (31/226) AC (130/1082) CC (176/1469)  0.17/0.28  0.98 (0.77–1.26)  0.89  1.19 (0.78–1.82)  0.41  rs6971499 (LINC-PINT)  AA (241/2222) AG (67/755) GG (5/72)  0.12/0.15  0.85 (0.64–1.14)  0.27  0.89 (0.35–2.27)  0.81  rs167020 (SHH)  AA (20/80) AG (108/392) GG (127/507)  0.15/0.26  1.04 (0.68–1.59)  0.85  0.92 (0.44–1.94)  0.83  rs1561927 (MIR1208, PVT1)  CC (25/238) CT (104/1297) TT (181/1643)  0.30/0.28  0.71 (0.55–0.92)  0.01  0.91 (0.58–1.43)  0.69  rs10991043 (SMC2)  CC (48/371) CT (146/1295) TT (133/1190)  0.43/0.36  1.02 (0.79–1.31)  0.90  1.17 (0.81–1.68)  0.41  rs505922 (ABO)  TT (110/935) TC (120/1019) CC (28/333)  0.35/0.39  1.11 (0.84–1.48)  0.45  0.76 (0.48–1.19)  0.23  rs7310409 (HNF1A)  GG (120/1008) GA (159/1409) AA (59/490)  0.41/0.39  0.90 (0.69–1.16)  0.41  1.04 (0.74–1.46)  0.84  rs9581943 (PDX1)  GG (110/1016) GA (149/1524) AA (51/510)  0.33/0.42  0.94 (0.72–1.22)  0.64  0.91 (0.64–1.31)  0.62  rs9543325 (KLF5/KLF12/PIBF1/DIS3/BORA)  CC (48/316) CT (113/960) TT (62/771)  0.46/0.39  1.35 (0.96–1.89)  0.08  1.63 (1.10–2.47)  0.02  rs8028529 (None)  TT (155/1194) TC (86/654) CC (16/106)  0.20/0.22  1.08 (0.79–1.48)  0.64  1.25 (0.69–2.25)  0.46  rs7190458 (BCAR1/CTRB1/CTRB2)  CC (294/2832) CT (18/215) TT (1/5)  0.10/0.04  0.78 (0.46–1.30)  0.34  2.25 (0.25–20.3)  0.47  rs11655237 (LINC00673)  CC (256/2113) CT (75/583) TT (3/37)  0.23/0.12  1.08 (0.80–1.45)  0.63  0.70 (0.16–3.02)  0.64  rs1517037 (GRP)  CC (227/1871) CT (107/896) TT (10/129)  0.24/0.20  0.89 (0.69–1.15)  0.36  0.59 (0.30–1.15)  0.12  rs16986825 (ZNRF3)  CC (193/1996) CT (100/919) TT (15/108)  0.20/0.19  1.07 (0.82–1.39)  0.63  1.21 (0.67–2.17)  0.53  rs2853677 (TERT)  AA (89/783) AG (155/1319) GG (52/532)  0.39/0.45  1.01 (0.74–1.38)  0.94  0.82 (0.54–1.22)  0.32  rs401681 (TERT)  TT (51/377) TC (129/949) CC (76/649)  0.41/0.43  1.10 (0.81–1.52)  0.52  1.11 (0.75–1.63)  0.61  rs10919791 (NR5A2)  GG (148/1272) GA (50/682) AA (15/111)  0.35/0.22  0.65 (0.46–0.93)  0.02  1.06 (0.59–1.90)  0.84  rs3790843 (NR5A2)  GG (124/1064) GA (72/847) AA (25/178)  0.40/0.29  0.76 (0.55–1.04)  0.09  1.20 (0.74–1.93)  0.46  SNP (nearby gene(s))  Genotypes (cases/controlsa)  MAF CEU 1KG/current studyb  MM versus Mmc  MM versus mmc  OR (95% CI)d  P  OR (95% CI)d  P  rs351365 (WNT2B)  CC (198/1549) CT (108/938) TT (26/172)  0.27/0.24  0.91 (0.69–1.20)  0.50  1.34 (0.82–2.18)  0.24  rs3790844 (NR5A2)  TT (140/1231) TC (64/721) CC (19/123)  0.36/0.23  0.82 (0.59–1.13)  0.22  1.27 (0.74–2.16)  0.38  rs1486134 (ETAA1)  TT (162/1569) TG (141/1120) GG (31/217)  0.27/0.27  1.20 (0.94–1.55)  0.13  1.36 (0.88–2.09)  0.16  rs9854771 (TP63)  GG (149/1131) GA (140/1311) AA (51/350)  0.29/0.36  0.76 (0.58–0.99)  0.04  1.00 (0.69–1.46)  0.98  rs2736098 (TERT)  GG (158/1514) GA (130/1131) AA (19/211)  0.27/0.27  1.08 (0.84–1.39)  0.56  0.84 (0.50–1.41)  0.51  rs17688601 (SUGCT)  AA (31/226) AC (130/1082) CC (176/1469)  0.17/0.28  0.98 (0.77–1.26)  0.89  1.19 (0.78–1.82)  0.41  rs6971499 (LINC-PINT)  AA (241/2222) AG (67/755) GG (5/72)  0.12/0.15  0.85 (0.64–1.14)  0.27  0.89 (0.35–2.27)  0.81  rs167020 (SHH)  AA (20/80) AG (108/392) GG (127/507)  0.15/0.26  1.04 (0.68–1.59)  0.85  0.92 (0.44–1.94)  0.83  rs1561927 (MIR1208, PVT1)  CC (25/238) CT (104/1297) TT (181/1643)  0.30/0.28  0.71 (0.55–0.92)  0.01  0.91 (0.58–1.43)  0.69  rs10991043 (SMC2)  CC (48/371) CT (146/1295) TT (133/1190)  0.43/0.36  1.02 (0.79–1.31)  0.90  1.17 (0.81–1.68)  0.41  rs505922 (ABO)  TT (110/935) TC (120/1019) CC (28/333)  0.35/0.39  1.11 (0.84–1.48)  0.45  0.76 (0.48–1.19)  0.23  rs7310409 (HNF1A)  GG (120/1008) GA (159/1409) AA (59/490)  0.41/0.39  0.90 (0.69–1.16)  0.41  1.04 (0.74–1.46)  0.84  rs9581943 (PDX1)  GG (110/1016) GA (149/1524) AA (51/510)  0.33/0.42  0.94 (0.72–1.22)  0.64  0.91 (0.64–1.31)  0.62  rs9543325 (KLF5/KLF12/PIBF1/DIS3/BORA)  CC (48/316) CT (113/960) TT (62/771)  0.46/0.39  1.35 (0.96–1.89)  0.08  1.63 (1.10–2.47)  0.02  rs8028529 (None)  TT (155/1194) TC (86/654) CC (16/106)  0.20/0.22  1.08 (0.79–1.48)  0.64  1.25 (0.69–2.25)  0.46  rs7190458 (BCAR1/CTRB1/CTRB2)  CC (294/2832) CT (18/215) TT (1/5)  0.10/0.04  0.78 (0.46–1.30)  0.34  2.25 (0.25–20.3)  0.47  rs11655237 (LINC00673)  CC (256/2113) CT (75/583) TT (3/37)  0.23/0.12  1.08 (0.80–1.45)  0.63  0.70 (0.16–3.02)  0.64  rs1517037 (GRP)  CC (227/1871) CT (107/896) TT (10/129)  0.24/0.20  0.89 (0.69–1.15)  0.36  0.59 (0.30–1.15)  0.12  rs16986825 (ZNRF3)  CC (193/1996) CT (100/919) TT (15/108)  0.20/0.19  1.07 (0.82–1.39)  0.63  1.21 (0.67–2.17)  0.53  rs2853677 (TERT)  AA (89/783) AG (155/1319) GG (52/532)  0.39/0.45  1.01 (0.74–1.38)  0.94  0.82 (0.54–1.22)  0.32  rs401681 (TERT)  TT (51/377) TC (129/949) CC (76/649)  0.41/0.43  1.10 (0.81–1.52)  0.52  1.11 (0.75–1.63)  0.61  rs10919791 (NR5A2)  GG (148/1272) GA (50/682) AA (15/111)  0.35/0.22  0.65 (0.46–0.93)  0.02  1.06 (0.59–1.90)  0.84  rs3790843 (NR5A2)  GG (124/1064) GA (72/847) AA (25/178)  0.40/0.29  0.76 (0.55–1.04)  0.09  1.20 (0.74–1.93)  0.46  MAF CEU, minor allele frequency in Caucasians; 1kG, 1000 Genomes Project (http://browser.1000genomes.org). Only association results for SNPs with P < 0.05 and with outcomes below FPRP-level 0.2 are marked bold. CI, confidence interval; OR, odds ratio. aNumbers may not add up to 100% due to genotyping failure or unavailable covariate data; bControls only; cM, major allele, m, minor allele where MM represents the reference category. dAnalyses were performed with adjustment for age, gender and country of origin. View Large Stratified analyses based on age at pNEN onset With stratification according to age at disease onset (≤50 years versus >50 years) among successfully genotyped pNEN cases, we observed significant differences between genotype groups for rs16986825 (ZNRF) (OR 2.67, 95% CI (1.57–4.55), Phet = 0.0003 and OR 2.27, 95% CI (1.37–3.77), P = 0.002 in the recessive model); rs2736098 (TERT) (OR 3.71, 95% CI (1.36–10.10), Phom = 0.01; and rs9854771 (TP63) (OR 1.85, 95% CI (1.11–3.10), Phet = 0.02 and OR 1.69, 95% CI (1.04–2.75), P = 0.03 in the recessive model (Table 3). FPRP tests were similarly applied using a prior probability of 0.2 and associations between early pNEN onset and rs16986825 (FPRP < 0.01 in both models), rs2736098 (FPRP = 0.08) and rs9854771 (FPRP = 0.14), were noteworthy. For rs9854771, our observed association for the recessive model was not noteworthy (FPRP = 0.22). Since pNENs are diagnosed roughly 10 years earlier than PDAC, we also performed the same stratified analyses with threshold at 40 years, and only the TERT SNPs (rs2736098 and rs2853677) were significantly associated with higher risk of developing pNEN at 40 years or younger, albeit with extremely small group sizes (results not shown). Results from alternative analyses of disease onset using three additional models: cases ≤50 years versus controls of all ages (EOP1), cases >50 years versus controls of all ages (TOP) and cases up to 50 years versus controls up to 50 years (EOP2) are shown in Supplementary Table I, available at Carcinogenesis Online. Table 3. Associations between TERT_rs2736098, ZNF_rs16986825, rs9543325 and risk of pNEN onset Locus  Gene  SNP ID  Genotypes (cases ≤50 years/>50 years)  OR (95% CI)a  P  5p15.33  TERT  rs2736098  GG (52/104)  0.83 (0.50–1.40)b  0.49  GA (36/89)  3.71 (1.36–10.10)c  0.01  AA (12/7)  1.04 (0.64–1.69)d  0.89  22q12.1  ZNRF3  rs16986825  CC (53/136)  2.67 (1.57–4.55)b  0.0003  CT (46/52)  0.67 (0.17–2.63)c  0.57  TT (3/11)  2.27 (1.37–3.77)d  0.002  3q28  TP63  rs9854771  GG (38/108)  1.85 (1.11–3.10)b  0.02  GA (53/87)  1.29 (0.63–2.66)c  0.49  AA (15/35)  1.69 (1.04–2.75)d  0.03  Locus  Gene  SNP ID  Genotypes (cases ≤50 years/>50 years)  OR (95% CI)a  P  5p15.33  TERT  rs2736098  GG (52/104)  0.83 (0.50–1.40)b  0.49  GA (36/89)  3.71 (1.36–10.10)c  0.01  AA (12/7)  1.04 (0.64–1.69)d  0.89  22q12.1  ZNRF3  rs16986825  CC (53/136)  2.67 (1.57–4.55)b  0.0003  CT (46/52)  0.67 (0.17–2.63)c  0.57  TT (3/11)  2.27 (1.37–3.77)d  0.002  3q28  TP63  rs9854771  GG (38/108)  1.85 (1.11–3.10)b  0.02  GA (53/87)  1.29 (0.63–2.66)c  0.49  AA (15/35)  1.69 (1.04–2.75)d  0.03  Only association results for SNPs with P < 0.05 and with outcomes below FPRP-level 0.2 are marked bold. aCases only—≤50 years (n = 114) versus >50 years (n = 247); where cases >50 years represent the reference category. Analyses were performed with adjustment for gender and country of origin. bMM versus Mm; cMM versus mm; d(MM + Mm) versus mm; where M, major allele and m, minor allele; y, years. View Large Stratified analyses based on tumor characteristics (grade, stage and functional status) With stratification based on disease grade, rs10919791 and rs16986825 were associated with well differentiated pNEN tumors (OR 5.11, 95% CI (1.13–23.14), P = 0.03 and OR 7.44, 95% CI (1.09–50.5), P = 0.04, respectively). rs1561927, rs9854771 and rs351365 were also associated with tumors with nodal metastasis (i.e. advanced stage) (OR 2.39, 95% CI (1.18–4.85), P = 0.02; OR 0.46, 95% CI (0.24–0.88), P = 0.02; and OR 4.62, 95% CI (1.16–18.38), P = 0.03), respectively. However, only the association between rs9854771 and stage appeared reliable, considering that the size of genotype groups was much smaller for other SNPs. FPRP tests applied using a prior probability of 0.2 showed the association between tumor grade and rs10919791 (FPRP = 0.19) to be noteworthy. For rs16986825, the association was not noteworthy (FPRP > 0.20). Regarding tumor stage, associations with rs1561927, rs9854771 and rs351365 (recessive model only) (FPRP = 0.11, 0.16 and 0.19, respectively) were noteworthy. Two variants rs7310409 (HNF1A) and rs1517037 (GRP) also appeared to be associated with tumor functional status [OR 0.36, 95% CI (0.16–0.79), P = 0.01 and OR 0.37, 95% CI (0.16–0.85), P = 0.02, respectively] (Table 3). Both associations were noteworthy at the 0.2 threshold (FPRP < 0.08). Associations between all 23 SNPs and tumor characteristics are shown in Supplementary Table II, available at Carcinogenesis Online. Differentially expressed genes based on microarray data screening Reasoning that the additional gene expression profiles of all 26 genes harboring or located close to the SNPs investigated in this study would amplify our ability to identify relevant underlying biological processes, we screened a GEO superseries dataset including six pNEN, five healthy pancreas (controls) and six ductal adenocarcinoma tissue samples. Based on the inclusion criteria of Padjusted < 0.05 and |logFC| > 1.0, five of these genes—nuclear receptor subfamily 5 group A member 2 (NR5A2), v-myc avian myelocytomatosis viral oncogene homolog (MYC), Krüppel-like factor (KLF) 5 (KLF5) and chymotrypsinogen genes, CTRB1 and CTRB2 were under expressed in pNEN compared to normal pancreas tissue (Table 5). Supplementary Table III, available at Carcinogenesis Online, shows expression data for all 23 SNPs. Our analyses are based on only a few samples, therefore caution should be applied in the interpretation of these results. Table 5. Differentially expressed genes obtained from the GSE43795 dataset (six pNEN tumors versus five non-neoplastic pancreatic tissue samples and six pNEN tumors versus six ductal adenocarcinoma tissues) Gene  pNEN versus normal  PDAC versus normal  Log2 fold-change  P  Adjusted-P*  Log2 fold-change  P  Adjusted-P*  MYC  4.38  9.47 × 10−7  5.88 × 10−5  2.65  4.55 × 10−4  5.93 × 10−3  NR5A2  4.91  7.82 × 10−10  2.38 × 10−7  3.51  2.00 × 10−4  3.31 × 10−3  KLF5  1.69  1.59 × 10−2  6.96 × 10−2  −1.41  1.59 × 10−1  3.18 × 10−1  CTRB1  7.90  1.10 × 10−9  3.04 × 10−7  7.55  2.56 × 10−8  5.18 × 10−6  CTRB2  7.70  6.75 × 10−8  8.08 × 10−6  7.33  1.71 × 10−6  1.13 × 10−4  PDX1  0.72  3.58 × 10−1  5.49 × 10−1  2.43  1.18 × 10−3  1.16 × 10−2  Gene  pNEN versus normal  PDAC versus normal  Log2 fold-change  P  Adjusted-P*  Log2 fold-change  P  Adjusted-P*  MYC  4.38  9.47 × 10−7  5.88 × 10−5  2.65  4.55 × 10−4  5.93 × 10−3  NR5A2  4.91  7.82 × 10−10  2.38 × 10−7  3.51  2.00 × 10−4  3.31 × 10−3  KLF5  1.69  1.59 × 10−2  6.96 × 10−2  −1.41  1.59 × 10−1  3.18 × 10−1  CTRB1  7.90  1.10 × 10−9  3.04 × 10−7  7.55  2.56 × 10−8  5.18 × 10−6  CTRB2  7.70  6.75 × 10−8  8.08 × 10−6  7.33  1.71 × 10−6  1.13 × 10−4  PDX1  0.72  3.58 × 10−1  5.49 × 10−1  2.43  1.18 × 10−3  1.16 × 10−2  Based on predefined cutoff value of adjusted-P < 0.05 and |log2 fold-change| > 1, genes which are differentially expressed in both groups are in bold. *P value adjusted using Benjamini and Hochberg method (False discovery rate). View Large Discussion By investigating known GWAS PDAC susceptibility loci (MAF ≥5%), our study has identified the possible role of these loci in the risk of developing pNEN. Combined with gene expression analyses, we report for the first time, few loci which may also be involved in genetic mechanisms influencing pNEN tumorigenesis, age at onset and tumor characteristics. These findings confirm that some susceptibility loci are shared by both entities. Interestingly, most known environmental risk factors associated with increased pNEN risk are also associated with PDAC risk (4). Although rarity and lack of predictive biomarkers for early-stage diagnosis or risk stratification of healthy individuals in both entities are two common features of these neoplasms, understanding their molecular mechanisms remains essential and advantageous. It is common albeit controversial knowledge that pancreatic ductal cells are ‘potential facultative stem cells’ capable of being reprogrammed into cells that closely resemble islet cells in vivo (19,20). Whether the shared genetic susceptibility is due to the origin of both these tumor types from a common pluripotent precursor cell, or a phenomenon linked with the reported capacity of islet cells to transdifferentiate to ductal cells (21,22) is currently unknown. Among explanations proposed for the transdifferentiation of pancreas cells, a largely favored one in pathogenesis is based on the hypothesis that tumor stroma mimics the regulatory role of pancreatic embryonic mesenchyme in duct and endocrine development, to induce the switch between ductules and endocrine cells (23,24). A second hypothesis based on multiple lines of evidence also suggests that tumor cells may originate from a common pluripotent cell which matures into two phenotypically different cell lines (25), and this pluripotent cell expresses transcription factors such as PDX1 and SOX9 which are involved in pancreatic development and homeostasis (26). Our study, which comprised of 369 pNEN cases and 3277 healthy controls from the international PANDoRA consortium, represents the largest of such studies performed to date. Recent attempts to delineate the mutational landscape of pNENs through whole genome sequencing (6), have implicated activation of mTOR signaling as one of the core pathways commonly altered in these neoplasms. In overall risk analyses, four SNPs (rs9854771, rs9543325, rs1561927 and rs10919791) showed associations with sporadic pNEN at P < 0.05. Adjustment for multiple comparisons was not applied to the nominal significance threshold (P < 0.05) since selected SNPs were based on prior evidence of their association with pancreatic cancer. Using the FPRP model in overall analyses, only associations with rs10919791, rs9543325 and rs1561927 reached noteworthiness with a prior probability of 0.2. rs1561927 in MIR1208/PVT1 was associated with a 29% lower risk of developing pNEN among heterozygotes, which is similar to the original report in PDACs [allelic OR 0.87 95% CI (0.83–0.92) (9)]. This intronic SNP maps to a locus associated with suggested long-distance interactions with MYC and PVT1 promoters in multiple cancers (9). rs9543325, an intergenic SNP in chr13q22.1 which is frequently deleted in several cancers (8) was associated with a 37% increased risk of developing pNEN (for subjects with CC genotype), higher than risk estimates from the original study in PDACs [OR 1.26 95% CI (1.18–1.35)] (8). Genes closest to this locus are KLF5, KLF12, PIBF1, DIS3 and BORA, which range up to 586 kb in distance from rs9543325. Krüppel-like factor (KLF) 5 and 12 are members of a family of zinc-finger transcription factors reported to exhibit tumor-suppressor and oncogenic activity, respectively, in various human cancers (27). KLF5 is involved in transcriptional activation of PI3K/Akt signaling, BORA (encoding aurora kinase A activator) is known to regulate cell proliferation and overexpressed in tumors (28). PIBF1 (encoding progesterone immunomodulatory binding factor 1) is speculated to be involved in progesterone-dependent immunomodulation with higher levels in tumors (28). Aberrant expression of DIS3 has been implicated in many cancers and plays a crucial role in gene regulation and small RNA processing (29). rs10919791 lies in the first intron of NR5A2 (chr1q32.1) which plays crucial roles in pancreatic function and development (30). In the current study, rs10919791 was associated with a 35% lower risk of developing pNEN among heterozygotes, similar to the original estimate among PDACs (ORhet 0.76 (0.68–0.84)) (8). The bioinformatics approach with a GEO2R microarray dataset including pancreatic neuroendocrine tumors (n = 6) and non-neoplastic pancreatic tissue samples (n = 5), was used to augment results from our study and explore the plausible role of these genes in pNEN development. Interestingly, MYC, KLF5, NR5A2, CTRB1 and CTRB2 appeared down-regulated among pNENs compared to non-neoplastic pancreatic tissue in the microarray dataset, further implicating these genes in pNEN tumorigenesis. To substantiate these findings, we performed a similar comparison of expression profiles of these genes in PDAC (n = 6) versus healthy pancreas, which showed differential expression of all genes but KLF5 in the same manner. To the best of our knowledge, this is the only publicly available dataset that includes expression data of pNEN. The data we observed in this dataset corroborate our findings of the association analysis. Nevertheless the very small sample size of this dataset is a cause for concern, and these data should be taken with caution. Additional stratified analyses based on age at onset and tumor characteristics (Tables 3 and 4, respectively), despite relatively small genotype strata for some SNPs, also shed light on susceptibility loci that may be useful for improved risk stratification and possible choice of therapy based on tumor characteristics. The contribution of rs7310409 to pNEN development is conceivable given that pleiotropic HNF1A plays a crucial role, as a tumor suppressor, in the transcriptional regulation in endocrine and exocrine pancreas development and homeostasis (31). This locus has also been reported as a prime candidate mediating susceptibility to type 2 diabetes among Caucasians (32). Although the role of p53 homolog, p63 in pancreatic cancer is poorly understood, it has been implicated in tumorigenesis and metastasis via cell cycle arrest and apoptotic mechanisms (33). Our observed association between heterozygous carriers of this polymorphism and metastatic pNEN tumors is therefore plausible. Table 4. Associations between SNPs and pNEN tumor characteristics Locus  Gene  SNP  Genotype groups  OR (95% CI)a  P  Grade (G1/G2/well differentiated tumors versus G3/poorly differentiated tumors)   1q32.1  NR5A2  rs10919791  GG/GA/AA 115/38/15  5.66 (1.18–27.11)b  0.03   22q12.1  ZNRF3  rs16986825  CC/CT/TT 113/62/9  7.34 (1.02–52.67)c 7.44 (1.10–50.53)d  0.05 0.04  Stage (I/II/III versus IV)   8q24.21  MIR1208, PVT1  rs1561927  CC/TC/TT 18/55/95  2.39 (1.18–4.85)b  0.02   3q28  TP63  rs9854771  AA/GA/GG 22/67/81  0.44 (0.22–0.90)b  0.02   1p13.1  WNT2B  rs351365  AA/GA/GG 11/58/99  4.15 (1.01–16.95)c 4.62 (1.16–18.38)d  0.05 0.03  Functional status (functional tumors versus non-functional tumors)   12q24.21  HNF1A  rs7310409  AA/GA/GG 32/79/66  0.36 (0.16–0.79)b  0.01   18q21.2  GRP  rs1517037  CC/CT/TT 118/56/4  0.37 (0.16–0.85)b  0.02  Locus  Gene  SNP  Genotype groups  OR (95% CI)a  P  Grade (G1/G2/well differentiated tumors versus G3/poorly differentiated tumors)   1q32.1  NR5A2  rs10919791  GG/GA/AA 115/38/15  5.66 (1.18–27.11)b  0.03   22q12.1  ZNRF3  rs16986825  CC/CT/TT 113/62/9  7.34 (1.02–52.67)c 7.44 (1.10–50.53)d  0.05 0.04  Stage (I/II/III versus IV)   8q24.21  MIR1208, PVT1  rs1561927  CC/TC/TT 18/55/95  2.39 (1.18–4.85)b  0.02   3q28  TP63  rs9854771  AA/GA/GG 22/67/81  0.44 (0.22–0.90)b  0.02   1p13.1  WNT2B  rs351365  AA/GA/GG 11/58/99  4.15 (1.01–16.95)c 4.62 (1.16–18.38)d  0.05 0.03  Functional status (functional tumors versus non-functional tumors)   12q24.21  HNF1A  rs7310409  AA/GA/GG 32/79/66  0.36 (0.16–0.79)b  0.01   18q21.2  GRP  rs1517037  CC/CT/TT 118/56/4  0.37 (0.16–0.85)b  0.02  Only association results for SNPs with P < 0.05 and with outcomes below FPRP-level 0.2 are marked in bold. Analyses were performed with adjustment for age, gender and country of origin. aAll G1 and G2 (i.e. well-differentiated) tumors, non-functional tumors and Stages I, II and III tumors (i.e. no nodal metastasis) represent the reference category for Grade, Functional status and Stage, respectively. bMM versus Mm; cMM versus mm; d(MM + Mm) versus mm; where M, major allele and m, minor allele. View Large Nevertheless, the small size observed in each stratum due to the rarity of these neoplasms, raises a valid concern on the possible interpretation of our findings. We therefore applied the FPRP approach instead of a much lower P value adjusted for multiple comparisons, as the latter would have resulted in unnecessarily low power for our high prior probability hypotheses, particularly in these neoplasms where collection of larger number of cases is unrealistic. Notwithstanding, all noteworthy findings reported here must be interpreted with caution and replicated in possibly larger and independent studies. The major strengths of this study are its relatively large size, considering the low incidence of pNEN and data analysis approaches applied to minimize the effect of potentially confounding factors and more importantly evaluate the noteworthiness of our findings. The integration of gene expression profiling provides valuable insight to shared genetic correlations and expands the probable margins of our results. A limitation of our investigations however, was that due to lack of clinical information, we could not stratify subjects based on familial history of PDAC or other non-genetic factors such as history of diabetes mellitus. Taken together, these results illuminate genetic similarities between the two main groups of pancreatic neoplasms, suggesting that common variations at three known PDAC susceptibility loci –13q22.1 (TP63, rs9543325), 1q32.1 (NR5A2, rs10919791) and 8q24.21 (MIR1208/PVT1, rs1561927) potentially influence tumorigenesis of pancreatic beta cells in a similar manner as in exocrine pancreatic cells. Our results also suggest that genetic predisposing factors vary on the basis of age at pNEN onset as well as tumor characteristics. Striking similarities between embryogenesis and tumorigenesis which have become more apparent in past decades taken together with our genetic findings may provide groundwork for future evaluation of the potential usability of these loci as predictive biomarkers and in risk assessment. Functional studies are also warranted to better understand the biological implications of these loci in pancreatic hormone-producing (endocrine) and enzyme-producing (exocrine) cell function. Funding This study was supported by intramural funds from the German Cancer Research Centre (DKFZ, Heidelberg, Germany) awarded to F.C. G.C. was supported by AIRC grant IG 2015, Id 17177. Abbreviations CI confidence interval FPRP false positive report probability GEO Gene Expression Omnibus GWAS genome-wide association studies KLF Krüppel-like factor OR odds ratio PANDoRA PANcreatic Disease ReseArch PDAC pancreatic ductal adenocarcinoma pNEN pancreatic neuroendocrine neoplasm SNP single nucleotide polymorphism Acknowledgements The authors acknowledge Angelika Stein (DKFZ, Heidelberg, Germany) for her technical support. Conception and design: D.C, F.C and O.O. Acquisition of samples and clinical data: all authors. Analysis and interpretation of data (e.g., experimental and statistical analysis): O.O, D.C and F.C. Writing, review and/or revision of the manuscript: O.O, G.C, D.C, F.C. and all other authors contributed to its critical revision. Conflict of Interest Statement: None declared. References 1. Capurso, G.et al.  ( 2009) Risk factors for sporadic pancreatic endocrine tumors: a case-control study of prospectively evaluated patients. Am. J. Gastroenterol ., 104, 3034– 3041. Google Scholar CrossRef Search ADS PubMed  2. Yao, J.C.et al.  ( 2008) One hundred years after “carcinoid”: epidemiology of and prognostic factors for neuroendocrine tumors in 35,825 cases in the United States. J. Clin. Oncol ., 26, 3063– 3072. Google Scholar CrossRef Search ADS PubMed  3. Li, D.et al.  ( 2004) Pancreatic cancer. Lancet , 363, 1049– 1057. Google Scholar CrossRef Search ADS PubMed  4. Haugvik, S.P.et al.  ( 2015) Diabetes, smoking, alcohol use, and family history of cancer as risk factors for pancreatic neuroendocrine tumors: a systematic review and meta-analysis. Neuroendocrinology , 101, 133– 142. Google Scholar CrossRef Search ADS PubMed  5. Jiao, Y.et al.  ( 2011) DAXX/ATRX, MEN1, and mTOR pathway genes are frequently altered in pancreatic neuroendocrine tumors. Science , 331, 1199– 1203. Google Scholar CrossRef Search ADS PubMed  6. Scarpa, A.et al.  ; Australian Pancreatic Cancer Genome Initiative. ( 2017) Whole-genome landscape of pancreatic neuroendocrine tumours. Nature , 543, 65– 71. Google Scholar CrossRef Search ADS PubMed  7. Amundadottir, L.et al.  ( 2009) Genome-wide association study identifies variants in the ABO locus associated with susceptibility to pancreatic cancer. Nat. Genet ., 41, 986– 990. Google Scholar CrossRef Search ADS PubMed  8. Petersen, G.M.et al.  ( 2010) A genome-wide association study identifies pancreatic cancer susceptibility loci on chromosomes 13q22.1, 1q32.1 and 5p15.33. Nat. Genet ., 42, 224– 228. Google Scholar CrossRef Search ADS PubMed  9. Wolpin, B.M.et al.  ( 2014) Genome-wide association study identifies multiple susceptibility loci for pancreatic cancer. Nat. Genet ., 46, 994– 1000. Google Scholar CrossRef Search ADS PubMed  10. Childs, E.J.et al.  ( 2015) Common variation at 2p13.3, 3q29, 7p13 and 17q25.1 associated with susceptibility to pancreatic cancer. Nat. Genet ., 47, 911– 916. Google Scholar CrossRef Search ADS PubMed  11. Campa, D.et al.  ( 2015) TERT gene harbors multiple variants associated with pancreatic cancer susceptibility. Int. J. Cancer , 137, 2175– 2183. Google Scholar CrossRef Search ADS PubMed  12. Campa, D.et al.  ( 2013) Genetic susceptibility to pancreatic cancer and its functional characterisation: the PANcreatic Disease ReseArch (PANDoRA) consortium. Dig. Liver Dis ., 45, 95– 99. Google Scholar CrossRef Search ADS PubMed  13. Bosman, F.T.et al.   ( 2010) WHO Classification of Tumours of the Digestive System . 4th ed. World Health Organization, Geneva. 14. Wacholder, S.et al.  ( 2004) Assessing the probability that a positive report is false: an approach for molecular epidemiology studies. J. Natl. Cancer Inst ., 96, 434– 442. Google Scholar CrossRef Search ADS PubMed  15. Ramage, J.K.et al.  ; UK and Ireland Neuroendocrine Tumour Society. ( 2012) Guidelines for the management of gastroenteropancreatic neuroendocrine (including carcinoid) tumours (NETs). Gut , 61, 6– 32. Google Scholar CrossRef Search ADS PubMed  16. Klimstra, D.S.et al.  ( 2010) The pathologic classification of neuroendocrine tumors: a review of nomenclature, grading, and staging systems. Pancreas , 39, 707– 712. Google Scholar CrossRef Search ADS PubMed  17. Park, M.et al.  ( 2014) Characterization of gene expression and activated signaling pathways in solid-pseudopapillary neoplasm of pancreas. Mod. Pathol ., 27, 580– 593. Google Scholar CrossRef Search ADS PubMed  18. Barrett, T.et al.  ( 2013) NCBI GEO: archive for functional genomics data sets–update. Nucleic Acids Res ., 41( Database issue), D991– D995. Google Scholar PubMed  19. Zhou, Q.et al.  ( 2008) In vivo reprogramming of adult pancreatic exocrine cells to beta-cells. Nature , 455, 627– 632. Google Scholar CrossRef Search ADS PubMed  20. Hezel, A.F.et al.  ( 2006) Genetics and biology of pancreatic ductal adenocarcinoma. Genes Dev ., 20, 1218– 1249. Google Scholar CrossRef Search ADS PubMed  21. Pour, P.M.et al.  ( 2003) What is the origin of pancreatic adenocarcinoma? Mol. Cancer , 2, 13. Google Scholar CrossRef Search ADS PubMed  22. Maitra, A.et al.  ( 2012) Disputed paternity: the uncertain ancestry of pancreatic ductal neoplasia. Cancer Cell , 22, 701– 703. Google Scholar CrossRef Search ADS PubMed  23. Deshpande, V.et al.  ( 2003) Ductulo-insular pancreatic endocrine neoplasms: clinicopathologic analysis of a unique subtype of pancreatic endocrine neoplasms. Am. J. Surg. Pathol ., 27, 461– 468. Google Scholar CrossRef Search ADS PubMed  24. Landsman, L.et al.  ( 2011) Pancreatic mesenchyme regulates epithelial organogenesis throughout development. PLoS Biol ., 9, e1001143. Google Scholar CrossRef Search ADS PubMed  25. Regitnig, P.et al.  ( 2001) Insulinoma of the pancreas with insular-ductular differentiation in its liver metastasis–indication of a common stem-cell origin of the exocrine and endocrine components. Virchows Arch ., 438, 624– 628. Google Scholar CrossRef Search ADS PubMed  26. Oliver-Krasinski, J.M.et al.  ( 2009) The diabetes gene Pdx1 regulates the transcriptional network of pancreatic endocrine progenitor cells in mice. J. Clin. Invest ., 119, 1888– 1898. Google Scholar CrossRef Search ADS PubMed  27. McConnell, B.B.et al.  ( 2010) Mammalian Krüppel-like factors in health and diseases. Physiol. Rev ., 90, 1337– 1381. Google Scholar CrossRef Search ADS PubMed  28. Azmi A. ( 2014) Molecular Diagnostics and Treatment of Pancreatic Cancer: Systems and Network Biology Approaches . Elsevier, Amsterdam. 29. Robinson, S.R.et al.  ( 2015) The 3’ to 5’ exoribonuclease DIS3: from structure and mechanisms to biological functions and role in human disease. Biomolecules , 5, 1515– 1539. Google Scholar CrossRef Search ADS PubMed  30. Kelly, V.R.et al.  ( 2010) Dax1 up-regulates Oct4 expression in mouse embryonic stem cells via LRH-1 and SRA. Mol. Endocrinol ., 24, 2281– 2291. Google Scholar CrossRef Search ADS PubMed  31. Hoskins, J.W.et al.  ( 2014) Transcriptome analysis of pancreatic cancer reveals a tumor suppressor function for HNF1A. Carcinogenesis , 35, 2670– 2678. Google Scholar CrossRef Search ADS PubMed  32. Voight, B.F.et al.  ; MAGIC investigators; GIANT Consortium. ( 2010) Twelve type 2 diabetes susceptibility loci identified through large-scale association analysis. Nat. Genet ., 42, 579– 589. Google Scholar CrossRef Search ADS PubMed  33. Bergholz, J.et al.  ( 2012) Role of p63 in development, tumorigenesis and cancer progression. Cancer Microenviron ., 5, 311– 322. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2017. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Carcinogenesis Oxford University Press

Loading next page...
 
/lp/ou_press/common-genetic-variants-associated-with-pancreatic-adenocarcinoma-may-rfkA8keM30
Publisher
Oxford University Press
Copyright
© The Author(s) 2017. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com.
ISSN
0143-3334
eISSN
1460-2180
D.O.I.
10.1093/carcin/bgx150
Publisher site
See Article on Publisher Site

Abstract

Abstract Pancreatic neuroendocrine neoplasms (pNEN) account for less than 5% of all pancreatic neoplasms and genetic association studies on susceptibility to the disease are limited. We sought to identify possible overlap of genetic susceptibility loci between pancreatic ductal adenocarcinoma (PDAC) and pNEN; therefore, PDAC susceptibility variants (n = 23) from Caucasian genome-wide association studies (GWAS) were genotyped in 369 pNEN cases and 3277 controls from the PANcreatic Disease ReseArch (PANDoRA) consortium to evaluate the odds associated with pNEN risk, disease onset and tumor characteristics. Main effect analyses showed four PDAC susceptibility variants—rs9854771, rs1561927, rs9543325 and rs10919791 to be associated with pNEN risk. Subsequently, only associations with rs9543325, rs10919791 and rs1561927 were noteworthy with false positive report probability (FPRP) tests. Stratified analyses considering age at onset (50-year threshold), showed rs2736098, rs16986825 and rs9854771 to be associated with risk of developing pNEN at a younger age. Stratified analyses also showed some single nucleotide polymorphisms to be associated with different degrees of tumor grade, metastatic potential and functionality. Our results identify known GWAS PDAC susceptibility loci, which may also be involved in sporadic pNEN etiology and suggest that some genetic mechanisms governing pathogenesis of these two entities may be similar, with few of these loci being more influential in younger cases or tumor subtypes. Introduction Pancreatic neuroendocrine neoplasms (pNEN) arise from islet cells and comprise less than 5% of all new pancreatic neoplasms (1). Heterogeneous in tumor behavior and clinical symptoms, pNENs account for 10% of neuroendocrine tumors and have a 5-year mortality rate of 60% (1,2). Compared to the more common exocrine pancreatic ductal adenocarcinoma (PDAC), pNENs are characterized by a rather ‘silent’ clinical course and generally present at a relatively earlier age (median age at diagnosis 53–60 years, and 72 years for PDAC) (1). Majority (40–90%) of pNENs are non-functional (i.e. do not secrete hormones that cause systemic effects) thus constituting a clinical and prognostic challenge for physicians. Surgical resection is currently the primary curative therapy option for both malignancies, however, most patients unfortunately exhibit unresectable tumors at diagnosis (3), stressing the need for improved risk stratification and timely diagnostic biomarkers. Familial clustering observed in population-based studies among PDAC and other cancer patients provide evidence of an inherited basis for sporadic pNEN (1). Although there is no clear link between environmental exposures and risk of developing pNEN, an overlap of risk factors in PDAC and pNENs is likely (4). Increased interest in these neoplasms has become evident in the last decade, but despite this, molecular understanding of pNENs is still insufficient to drive clinical interventions. Recent exome and whole genome sequencing studies describing the genetic basis of pNENs among Caucasians (5,6) have implicated somatic mutations in four core pathways: activation of mTOR signaling, DNA damage repair, chromatin modification and altered telomere length in tumorigenesis. Genome wide association (GWAS) or candidate gene or pathway studies investigating the etiology of pNENs have been reasonably limited by small sample size, when compared with PDAC for which 23 common risk variants have so far been identified through GWAS or studies focusing on genes situated in well-known pleiotropic regions (7–11). Based on the rationale that known common genetic susceptibility loci for sporadic PDAC may also modify the risk of developing pNEN, we genotyped 369 pNEN cases from a retrospectively enrolled population within the PANcreatic Disease ReseArch (PANDoRA) consortium and compared genotype frequencies of all known PDAC GWAS susceptibility variants to over 3200 control subjects from the same consortium. We tested their possible association with overall sporadic pNEN risk and within pNEN subgroups. Materials and methods Selection of subjects and polymorphisms Demographic characteristics of subjects within the PANDoRA consortium have been described previously (12) although the number of neuroendocrine cases within PANDoRA has increased since that publication. Written informed consent and biospecimens (blood, tissue or genomic DNA) were obtained for all subjects and ethical approval for the PANDoRA study protocol received from the Ethics commission of the Medical Faculty Heidelberg. Only sporadic (non-familial) pNEN cases not associated with genetic syndromes such as multiple endocrine neoplasia type 1 (MEN-1), von Hippel–Lindau syndrome, von Recklinghausen disease (neurofibromatosis NF-1) and tuberous sclerosis complex were used in this study. Twenty-three single nucleotide polymorphisms (SNPs) reaching genome-wide significance and suggestive levels (up to 5 × 10–7) were selected from GWAS publications in PDAC among Caucasians (7–10). pNEN tumors were classified according to the World Health Organization 2010 classification system (13). Genotyping and quality control SNPs were genotyped in 369 pNEN cases and 3277 controls using TaqMan allelic discrimination (Applied Biosystems) and KASPar (LGC Genomics) assays according to manufacturers’ instructions. Post-PCR allelic discrimination was done on the 7900HT real-time PCR system (Life Technologies) and data analyzed using the affiliated SDS software (v.2.3). Internal replicates (7% of all samples) and negative controls were included to assess the fidelity of genotype calls. Deviation from Hardy–Weinberg equilibrium was checked using the Pearson’s χ2 test. Statistical analyses We evaluated associations [odds ratios (OR) and 95% confidence intervals (CI)] between SNPs and pNEN risk using unconditional multivariate logistic regression in STATA v.11 (StataCorp LP). Tests were done using a codominant inheritance model, with adjustment for potential confounding variables—age, geographical origin and gender. We used a nominal significance threshold (P < 0.05) since the selected SNPs were based on prior evidence of their association with pancreatic cancer. Noteworthiness of SNP associations with P < 0.05 were tested using the Bayesian false positive report probability (FPRP) test which takes the observed P value, statistical power of the test and prior probabilities for the associations into account (14). All selected SNPs bore a high prior possibility of being associated with pNEN risk (π = 0.2) as recommended by Wacholder et al. (14), and only SNP associations with P < 0.05 and π < 0.2 were considered ‘noteworthy’. To evaluate the effect of SNPs on pNEN onset, we first performed a case-only analysis comparing genotype distributions among cases (>50 years) with their younger counterparts (≤50 years) while adjusting for gender and geographical origin. We also compared genetic associations from three additional models: cases ≤50 years versus controls of all ages (EOP1), cases >50 years versus controls of all ages (TOP) and cases up to 50 years versus controls up to 50 years (EOP2). Associations between SNPs and tumor grade and stage were estimated by case-only analyses. For tumor stage, we compared stage IV tumors versus all other tumors (stages I, IIa, IIb, IIIa, IIIb) based on European Neuroendocrine Tumor Society (ENETS) classification system (15) and for grade, we compared well differentiated tumors (grades 1 and 2) versus poorly differentiated tumors (grade 3) based on the World Health Organization (WHO)/ENETS guidelines (16). Tumor functional status (e.g. insulinoma, gastrinomas, etc) was available for 54% of all cases in this study, out of which 62% were non-functioning. We compared genotype frequencies of all SNPs in the latter with functioning tumors. Microarray data screening To gain additional insight to the transcriptomic profiles of these gene regions in human pancreatic islet cells, we explored expression profiling data from the Gene Expression Omnibus (GEO) database (Accession number GSE43795); this study characterized mRNA and microRNA expression profiles in solid-pseudopapillary neoplasms of pancreas (n = 14), ductal adenocarcinoma (n = 6), pancreatic neuroendocrine tumors (n = 6) and non-neoplastic pancreatic tissue samples (n = 5) (17). Results were downloaded in text format and GEO2R queries (http://www.ncbi.nlm.nih.gov/geo/geo2r/) applied to assess differentially expressed genes in pNEN compared to non-neoplastic pancreas tissue samples. GEO2R is an R programming language-based dataset analysis tool that compares two or more groups of samples under the same experimental conditions and analyzes almost any GEO series (18). P values were adjusted to correct for false-positives due to multiple testing using Benjamini and Hochberg false discovery rate method and genes that met the cut-off criteria of Padjusted < 0.05 and |logFC| > 1.0 were screened out as differentially expressed genes. Results Baseline characteristics of the study population are summarized in Table 1. The median age at diagnosis in this study was 57 years, congruent with previous literature reports. Genotypes for most SNPs, except rs505922 (P = 0.04) were distributed in accordance to Hardy–Weinberg equilibrium among controls. Duplicated samples showed genotyping concordance rate of 99.6%. Genotyping data for those subjects with genotyping completion rates below 75% (n = 35) were excluded from all statistical analyses. Table 1. Clinical characteristics of cases and controls used in this study Region of origin  Germany  Greece  Italy  Poland  United Kingdom  Total subjectsa  Median age at diagnosisb, years (1st–3rd quartile)  51 (44–58)  45 (35–58)  61 (49–72)  46 (31–66)  66 (59–74)  57 (44–67)   ≤50 years  383  151  491  237  15  1277   >50 years  394  75  1238  185  209  2101  Gender   Males  356  135  1098  185  130  1906   Females  421  95  814  257  94  1681  Functional status   Yes  4  20  47  3  3  77   No  n/a  2  118  3  n/a  123   Unknown  31  0  82  9  47  169  Grade, G (WHO)   G1 and G2  n/a  21  149  9  29  208   G3  n/a  1  6  2  1  10   Unknown  35  0  12  4  20  71  Stage (ENETS)   I/II/III  17  n/a  133  12  n/a  162   IV  5  n/a  32  2  n/a  39   Unknown  13  22  82  1  50  168  Total cases  35  22  247  15  50  369  Total controls  797  208  1670  427  175  3277  Region of origin  Germany  Greece  Italy  Poland  United Kingdom  Total subjectsa  Median age at diagnosisb, years (1st–3rd quartile)  51 (44–58)  45 (35–58)  61 (49–72)  46 (31–66)  66 (59–74)  57 (44–67)   ≤50 years  383  151  491  237  15  1277   >50 years  394  75  1238  185  209  2101  Gender   Males  356  135  1098  185  130  1906   Females  421  95  814  257  94  1681  Functional status   Yes  4  20  47  3  3  77   No  n/a  2  118  3  n/a  123   Unknown  31  0  82  9  47  169  Grade, G (WHO)   G1 and G2  n/a  21  149  9  29  208   G3  n/a  1  6  2  1  10   Unknown  35  0  12  4  20  71  Stage (ENETS)   I/II/III  17  n/a  133  12  n/a  162   IV  5  n/a  32  2  n/a  39   Unknown  13  22  82  1  50  168  Total cases  35  22  247  15  50  369  Total controls  797  208  1670  427  175  3277  ENETS, European Neuroendocrine Tumor Society; WHO, World Health Organization; y, years; n/a, not available. aNumbers may not add up to 100% due to genotyping failure or unavailable covariate data. bpNEN cases only. View Large Overall pNEN risk analyses Minor allele frequencies, genotype distributions, risk estimates [odds ratios (OR) and 95% confidence intervals (CI)] of all polymorphisms investigated in this study are summarized in Table 2. rs9543325 was associated with higher risk of pNEN (OR 1.63, 95% CI (1.10–2.47), Phom = 0.02). rs10919791, rs1561927 and rs9854771 were associated with lower risk of pNEN (OR 0.65, 95% CI (0.46–0.93), Phet = 0.02; OR 0.71, 95% CI (0.55–0.92), Phet = 0.01 and OR 0.76, 95% CI (0.58–0.99), Phet = 0.04 respectively). Subsequent testing of these findings using FPRP with a prior of 0.2 showed that only associations with rs10919791 (FPRP = 0.13), rs9543325 (FPRP = 0.15) and rs1561927 (FPRP = 0.07) were noteworthy. The association with rs9854771 failed the noteworthiness test (FPRP = 0.25). Table 2. Associations between known GWAS PDAC susceptibility loci and pNEN risk SNP (nearby gene(s))  Genotypes (cases/controlsa)  MAF CEU 1KG/current studyb  MM versus Mmc  MM versus mmc  OR (95% CI)d  P  OR (95% CI)d  P  rs351365 (WNT2B)  CC (198/1549) CT (108/938) TT (26/172)  0.27/0.24  0.91 (0.69–1.20)  0.50  1.34 (0.82–2.18)  0.24  rs3790844 (NR5A2)  TT (140/1231) TC (64/721) CC (19/123)  0.36/0.23  0.82 (0.59–1.13)  0.22  1.27 (0.74–2.16)  0.38  rs1486134 (ETAA1)  TT (162/1569) TG (141/1120) GG (31/217)  0.27/0.27  1.20 (0.94–1.55)  0.13  1.36 (0.88–2.09)  0.16  rs9854771 (TP63)  GG (149/1131) GA (140/1311) AA (51/350)  0.29/0.36  0.76 (0.58–0.99)  0.04  1.00 (0.69–1.46)  0.98  rs2736098 (TERT)  GG (158/1514) GA (130/1131) AA (19/211)  0.27/0.27  1.08 (0.84–1.39)  0.56  0.84 (0.50–1.41)  0.51  rs17688601 (SUGCT)  AA (31/226) AC (130/1082) CC (176/1469)  0.17/0.28  0.98 (0.77–1.26)  0.89  1.19 (0.78–1.82)  0.41  rs6971499 (LINC-PINT)  AA (241/2222) AG (67/755) GG (5/72)  0.12/0.15  0.85 (0.64–1.14)  0.27  0.89 (0.35–2.27)  0.81  rs167020 (SHH)  AA (20/80) AG (108/392) GG (127/507)  0.15/0.26  1.04 (0.68–1.59)  0.85  0.92 (0.44–1.94)  0.83  rs1561927 (MIR1208, PVT1)  CC (25/238) CT (104/1297) TT (181/1643)  0.30/0.28  0.71 (0.55–0.92)  0.01  0.91 (0.58–1.43)  0.69  rs10991043 (SMC2)  CC (48/371) CT (146/1295) TT (133/1190)  0.43/0.36  1.02 (0.79–1.31)  0.90  1.17 (0.81–1.68)  0.41  rs505922 (ABO)  TT (110/935) TC (120/1019) CC (28/333)  0.35/0.39  1.11 (0.84–1.48)  0.45  0.76 (0.48–1.19)  0.23  rs7310409 (HNF1A)  GG (120/1008) GA (159/1409) AA (59/490)  0.41/0.39  0.90 (0.69–1.16)  0.41  1.04 (0.74–1.46)  0.84  rs9581943 (PDX1)  GG (110/1016) GA (149/1524) AA (51/510)  0.33/0.42  0.94 (0.72–1.22)  0.64  0.91 (0.64–1.31)  0.62  rs9543325 (KLF5/KLF12/PIBF1/DIS3/BORA)  CC (48/316) CT (113/960) TT (62/771)  0.46/0.39  1.35 (0.96–1.89)  0.08  1.63 (1.10–2.47)  0.02  rs8028529 (None)  TT (155/1194) TC (86/654) CC (16/106)  0.20/0.22  1.08 (0.79–1.48)  0.64  1.25 (0.69–2.25)  0.46  rs7190458 (BCAR1/CTRB1/CTRB2)  CC (294/2832) CT (18/215) TT (1/5)  0.10/0.04  0.78 (0.46–1.30)  0.34  2.25 (0.25–20.3)  0.47  rs11655237 (LINC00673)  CC (256/2113) CT (75/583) TT (3/37)  0.23/0.12  1.08 (0.80–1.45)  0.63  0.70 (0.16–3.02)  0.64  rs1517037 (GRP)  CC (227/1871) CT (107/896) TT (10/129)  0.24/0.20  0.89 (0.69–1.15)  0.36  0.59 (0.30–1.15)  0.12  rs16986825 (ZNRF3)  CC (193/1996) CT (100/919) TT (15/108)  0.20/0.19  1.07 (0.82–1.39)  0.63  1.21 (0.67–2.17)  0.53  rs2853677 (TERT)  AA (89/783) AG (155/1319) GG (52/532)  0.39/0.45  1.01 (0.74–1.38)  0.94  0.82 (0.54–1.22)  0.32  rs401681 (TERT)  TT (51/377) TC (129/949) CC (76/649)  0.41/0.43  1.10 (0.81–1.52)  0.52  1.11 (0.75–1.63)  0.61  rs10919791 (NR5A2)  GG (148/1272) GA (50/682) AA (15/111)  0.35/0.22  0.65 (0.46–0.93)  0.02  1.06 (0.59–1.90)  0.84  rs3790843 (NR5A2)  GG (124/1064) GA (72/847) AA (25/178)  0.40/0.29  0.76 (0.55–1.04)  0.09  1.20 (0.74–1.93)  0.46  SNP (nearby gene(s))  Genotypes (cases/controlsa)  MAF CEU 1KG/current studyb  MM versus Mmc  MM versus mmc  OR (95% CI)d  P  OR (95% CI)d  P  rs351365 (WNT2B)  CC (198/1549) CT (108/938) TT (26/172)  0.27/0.24  0.91 (0.69–1.20)  0.50  1.34 (0.82–2.18)  0.24  rs3790844 (NR5A2)  TT (140/1231) TC (64/721) CC (19/123)  0.36/0.23  0.82 (0.59–1.13)  0.22  1.27 (0.74–2.16)  0.38  rs1486134 (ETAA1)  TT (162/1569) TG (141/1120) GG (31/217)  0.27/0.27  1.20 (0.94–1.55)  0.13  1.36 (0.88–2.09)  0.16  rs9854771 (TP63)  GG (149/1131) GA (140/1311) AA (51/350)  0.29/0.36  0.76 (0.58–0.99)  0.04  1.00 (0.69–1.46)  0.98  rs2736098 (TERT)  GG (158/1514) GA (130/1131) AA (19/211)  0.27/0.27  1.08 (0.84–1.39)  0.56  0.84 (0.50–1.41)  0.51  rs17688601 (SUGCT)  AA (31/226) AC (130/1082) CC (176/1469)  0.17/0.28  0.98 (0.77–1.26)  0.89  1.19 (0.78–1.82)  0.41  rs6971499 (LINC-PINT)  AA (241/2222) AG (67/755) GG (5/72)  0.12/0.15  0.85 (0.64–1.14)  0.27  0.89 (0.35–2.27)  0.81  rs167020 (SHH)  AA (20/80) AG (108/392) GG (127/507)  0.15/0.26  1.04 (0.68–1.59)  0.85  0.92 (0.44–1.94)  0.83  rs1561927 (MIR1208, PVT1)  CC (25/238) CT (104/1297) TT (181/1643)  0.30/0.28  0.71 (0.55–0.92)  0.01  0.91 (0.58–1.43)  0.69  rs10991043 (SMC2)  CC (48/371) CT (146/1295) TT (133/1190)  0.43/0.36  1.02 (0.79–1.31)  0.90  1.17 (0.81–1.68)  0.41  rs505922 (ABO)  TT (110/935) TC (120/1019) CC (28/333)  0.35/0.39  1.11 (0.84–1.48)  0.45  0.76 (0.48–1.19)  0.23  rs7310409 (HNF1A)  GG (120/1008) GA (159/1409) AA (59/490)  0.41/0.39  0.90 (0.69–1.16)  0.41  1.04 (0.74–1.46)  0.84  rs9581943 (PDX1)  GG (110/1016) GA (149/1524) AA (51/510)  0.33/0.42  0.94 (0.72–1.22)  0.64  0.91 (0.64–1.31)  0.62  rs9543325 (KLF5/KLF12/PIBF1/DIS3/BORA)  CC (48/316) CT (113/960) TT (62/771)  0.46/0.39  1.35 (0.96–1.89)  0.08  1.63 (1.10–2.47)  0.02  rs8028529 (None)  TT (155/1194) TC (86/654) CC (16/106)  0.20/0.22  1.08 (0.79–1.48)  0.64  1.25 (0.69–2.25)  0.46  rs7190458 (BCAR1/CTRB1/CTRB2)  CC (294/2832) CT (18/215) TT (1/5)  0.10/0.04  0.78 (0.46–1.30)  0.34  2.25 (0.25–20.3)  0.47  rs11655237 (LINC00673)  CC (256/2113) CT (75/583) TT (3/37)  0.23/0.12  1.08 (0.80–1.45)  0.63  0.70 (0.16–3.02)  0.64  rs1517037 (GRP)  CC (227/1871) CT (107/896) TT (10/129)  0.24/0.20  0.89 (0.69–1.15)  0.36  0.59 (0.30–1.15)  0.12  rs16986825 (ZNRF3)  CC (193/1996) CT (100/919) TT (15/108)  0.20/0.19  1.07 (0.82–1.39)  0.63  1.21 (0.67–2.17)  0.53  rs2853677 (TERT)  AA (89/783) AG (155/1319) GG (52/532)  0.39/0.45  1.01 (0.74–1.38)  0.94  0.82 (0.54–1.22)  0.32  rs401681 (TERT)  TT (51/377) TC (129/949) CC (76/649)  0.41/0.43  1.10 (0.81–1.52)  0.52  1.11 (0.75–1.63)  0.61  rs10919791 (NR5A2)  GG (148/1272) GA (50/682) AA (15/111)  0.35/0.22  0.65 (0.46–0.93)  0.02  1.06 (0.59–1.90)  0.84  rs3790843 (NR5A2)  GG (124/1064) GA (72/847) AA (25/178)  0.40/0.29  0.76 (0.55–1.04)  0.09  1.20 (0.74–1.93)  0.46  MAF CEU, minor allele frequency in Caucasians; 1kG, 1000 Genomes Project (http://browser.1000genomes.org). Only association results for SNPs with P < 0.05 and with outcomes below FPRP-level 0.2 are marked bold. CI, confidence interval; OR, odds ratio. aNumbers may not add up to 100% due to genotyping failure or unavailable covariate data; bControls only; cM, major allele, m, minor allele where MM represents the reference category. dAnalyses were performed with adjustment for age, gender and country of origin. View Large Stratified analyses based on age at pNEN onset With stratification according to age at disease onset (≤50 years versus >50 years) among successfully genotyped pNEN cases, we observed significant differences between genotype groups for rs16986825 (ZNRF) (OR 2.67, 95% CI (1.57–4.55), Phet = 0.0003 and OR 2.27, 95% CI (1.37–3.77), P = 0.002 in the recessive model); rs2736098 (TERT) (OR 3.71, 95% CI (1.36–10.10), Phom = 0.01; and rs9854771 (TP63) (OR 1.85, 95% CI (1.11–3.10), Phet = 0.02 and OR 1.69, 95% CI (1.04–2.75), P = 0.03 in the recessive model (Table 3). FPRP tests were similarly applied using a prior probability of 0.2 and associations between early pNEN onset and rs16986825 (FPRP < 0.01 in both models), rs2736098 (FPRP = 0.08) and rs9854771 (FPRP = 0.14), were noteworthy. For rs9854771, our observed association for the recessive model was not noteworthy (FPRP = 0.22). Since pNENs are diagnosed roughly 10 years earlier than PDAC, we also performed the same stratified analyses with threshold at 40 years, and only the TERT SNPs (rs2736098 and rs2853677) were significantly associated with higher risk of developing pNEN at 40 years or younger, albeit with extremely small group sizes (results not shown). Results from alternative analyses of disease onset using three additional models: cases ≤50 years versus controls of all ages (EOP1), cases >50 years versus controls of all ages (TOP) and cases up to 50 years versus controls up to 50 years (EOP2) are shown in Supplementary Table I, available at Carcinogenesis Online. Table 3. Associations between TERT_rs2736098, ZNF_rs16986825, rs9543325 and risk of pNEN onset Locus  Gene  SNP ID  Genotypes (cases ≤50 years/>50 years)  OR (95% CI)a  P  5p15.33  TERT  rs2736098  GG (52/104)  0.83 (0.50–1.40)b  0.49  GA (36/89)  3.71 (1.36–10.10)c  0.01  AA (12/7)  1.04 (0.64–1.69)d  0.89  22q12.1  ZNRF3  rs16986825  CC (53/136)  2.67 (1.57–4.55)b  0.0003  CT (46/52)  0.67 (0.17–2.63)c  0.57  TT (3/11)  2.27 (1.37–3.77)d  0.002  3q28  TP63  rs9854771  GG (38/108)  1.85 (1.11–3.10)b  0.02  GA (53/87)  1.29 (0.63–2.66)c  0.49  AA (15/35)  1.69 (1.04–2.75)d  0.03  Locus  Gene  SNP ID  Genotypes (cases ≤50 years/>50 years)  OR (95% CI)a  P  5p15.33  TERT  rs2736098  GG (52/104)  0.83 (0.50–1.40)b  0.49  GA (36/89)  3.71 (1.36–10.10)c  0.01  AA (12/7)  1.04 (0.64–1.69)d  0.89  22q12.1  ZNRF3  rs16986825  CC (53/136)  2.67 (1.57–4.55)b  0.0003  CT (46/52)  0.67 (0.17–2.63)c  0.57  TT (3/11)  2.27 (1.37–3.77)d  0.002  3q28  TP63  rs9854771  GG (38/108)  1.85 (1.11–3.10)b  0.02  GA (53/87)  1.29 (0.63–2.66)c  0.49  AA (15/35)  1.69 (1.04–2.75)d  0.03  Only association results for SNPs with P < 0.05 and with outcomes below FPRP-level 0.2 are marked bold. aCases only—≤50 years (n = 114) versus >50 years (n = 247); where cases >50 years represent the reference category. Analyses were performed with adjustment for gender and country of origin. bMM versus Mm; cMM versus mm; d(MM + Mm) versus mm; where M, major allele and m, minor allele; y, years. View Large Stratified analyses based on tumor characteristics (grade, stage and functional status) With stratification based on disease grade, rs10919791 and rs16986825 were associated with well differentiated pNEN tumors (OR 5.11, 95% CI (1.13–23.14), P = 0.03 and OR 7.44, 95% CI (1.09–50.5), P = 0.04, respectively). rs1561927, rs9854771 and rs351365 were also associated with tumors with nodal metastasis (i.e. advanced stage) (OR 2.39, 95% CI (1.18–4.85), P = 0.02; OR 0.46, 95% CI (0.24–0.88), P = 0.02; and OR 4.62, 95% CI (1.16–18.38), P = 0.03), respectively. However, only the association between rs9854771 and stage appeared reliable, considering that the size of genotype groups was much smaller for other SNPs. FPRP tests applied using a prior probability of 0.2 showed the association between tumor grade and rs10919791 (FPRP = 0.19) to be noteworthy. For rs16986825, the association was not noteworthy (FPRP > 0.20). Regarding tumor stage, associations with rs1561927, rs9854771 and rs351365 (recessive model only) (FPRP = 0.11, 0.16 and 0.19, respectively) were noteworthy. Two variants rs7310409 (HNF1A) and rs1517037 (GRP) also appeared to be associated with tumor functional status [OR 0.36, 95% CI (0.16–0.79), P = 0.01 and OR 0.37, 95% CI (0.16–0.85), P = 0.02, respectively] (Table 3). Both associations were noteworthy at the 0.2 threshold (FPRP < 0.08). Associations between all 23 SNPs and tumor characteristics are shown in Supplementary Table II, available at Carcinogenesis Online. Differentially expressed genes based on microarray data screening Reasoning that the additional gene expression profiles of all 26 genes harboring or located close to the SNPs investigated in this study would amplify our ability to identify relevant underlying biological processes, we screened a GEO superseries dataset including six pNEN, five healthy pancreas (controls) and six ductal adenocarcinoma tissue samples. Based on the inclusion criteria of Padjusted < 0.05 and |logFC| > 1.0, five of these genes—nuclear receptor subfamily 5 group A member 2 (NR5A2), v-myc avian myelocytomatosis viral oncogene homolog (MYC), Krüppel-like factor (KLF) 5 (KLF5) and chymotrypsinogen genes, CTRB1 and CTRB2 were under expressed in pNEN compared to normal pancreas tissue (Table 5). Supplementary Table III, available at Carcinogenesis Online, shows expression data for all 23 SNPs. Our analyses are based on only a few samples, therefore caution should be applied in the interpretation of these results. Table 5. Differentially expressed genes obtained from the GSE43795 dataset (six pNEN tumors versus five non-neoplastic pancreatic tissue samples and six pNEN tumors versus six ductal adenocarcinoma tissues) Gene  pNEN versus normal  PDAC versus normal  Log2 fold-change  P  Adjusted-P*  Log2 fold-change  P  Adjusted-P*  MYC  4.38  9.47 × 10−7  5.88 × 10−5  2.65  4.55 × 10−4  5.93 × 10−3  NR5A2  4.91  7.82 × 10−10  2.38 × 10−7  3.51  2.00 × 10−4  3.31 × 10−3  KLF5  1.69  1.59 × 10−2  6.96 × 10−2  −1.41  1.59 × 10−1  3.18 × 10−1  CTRB1  7.90  1.10 × 10−9  3.04 × 10−7  7.55  2.56 × 10−8  5.18 × 10−6  CTRB2  7.70  6.75 × 10−8  8.08 × 10−6  7.33  1.71 × 10−6  1.13 × 10−4  PDX1  0.72  3.58 × 10−1  5.49 × 10−1  2.43  1.18 × 10−3  1.16 × 10−2  Gene  pNEN versus normal  PDAC versus normal  Log2 fold-change  P  Adjusted-P*  Log2 fold-change  P  Adjusted-P*  MYC  4.38  9.47 × 10−7  5.88 × 10−5  2.65  4.55 × 10−4  5.93 × 10−3  NR5A2  4.91  7.82 × 10−10  2.38 × 10−7  3.51  2.00 × 10−4  3.31 × 10−3  KLF5  1.69  1.59 × 10−2  6.96 × 10−2  −1.41  1.59 × 10−1  3.18 × 10−1  CTRB1  7.90  1.10 × 10−9  3.04 × 10−7  7.55  2.56 × 10−8  5.18 × 10−6  CTRB2  7.70  6.75 × 10−8  8.08 × 10−6  7.33  1.71 × 10−6  1.13 × 10−4  PDX1  0.72  3.58 × 10−1  5.49 × 10−1  2.43  1.18 × 10−3  1.16 × 10−2  Based on predefined cutoff value of adjusted-P < 0.05 and |log2 fold-change| > 1, genes which are differentially expressed in both groups are in bold. *P value adjusted using Benjamini and Hochberg method (False discovery rate). View Large Discussion By investigating known GWAS PDAC susceptibility loci (MAF ≥5%), our study has identified the possible role of these loci in the risk of developing pNEN. Combined with gene expression analyses, we report for the first time, few loci which may also be involved in genetic mechanisms influencing pNEN tumorigenesis, age at onset and tumor characteristics. These findings confirm that some susceptibility loci are shared by both entities. Interestingly, most known environmental risk factors associated with increased pNEN risk are also associated with PDAC risk (4). Although rarity and lack of predictive biomarkers for early-stage diagnosis or risk stratification of healthy individuals in both entities are two common features of these neoplasms, understanding their molecular mechanisms remains essential and advantageous. It is common albeit controversial knowledge that pancreatic ductal cells are ‘potential facultative stem cells’ capable of being reprogrammed into cells that closely resemble islet cells in vivo (19,20). Whether the shared genetic susceptibility is due to the origin of both these tumor types from a common pluripotent precursor cell, or a phenomenon linked with the reported capacity of islet cells to transdifferentiate to ductal cells (21,22) is currently unknown. Among explanations proposed for the transdifferentiation of pancreas cells, a largely favored one in pathogenesis is based on the hypothesis that tumor stroma mimics the regulatory role of pancreatic embryonic mesenchyme in duct and endocrine development, to induce the switch between ductules and endocrine cells (23,24). A second hypothesis based on multiple lines of evidence also suggests that tumor cells may originate from a common pluripotent cell which matures into two phenotypically different cell lines (25), and this pluripotent cell expresses transcription factors such as PDX1 and SOX9 which are involved in pancreatic development and homeostasis (26). Our study, which comprised of 369 pNEN cases and 3277 healthy controls from the international PANDoRA consortium, represents the largest of such studies performed to date. Recent attempts to delineate the mutational landscape of pNENs through whole genome sequencing (6), have implicated activation of mTOR signaling as one of the core pathways commonly altered in these neoplasms. In overall risk analyses, four SNPs (rs9854771, rs9543325, rs1561927 and rs10919791) showed associations with sporadic pNEN at P < 0.05. Adjustment for multiple comparisons was not applied to the nominal significance threshold (P < 0.05) since selected SNPs were based on prior evidence of their association with pancreatic cancer. Using the FPRP model in overall analyses, only associations with rs10919791, rs9543325 and rs1561927 reached noteworthiness with a prior probability of 0.2. rs1561927 in MIR1208/PVT1 was associated with a 29% lower risk of developing pNEN among heterozygotes, which is similar to the original report in PDACs [allelic OR 0.87 95% CI (0.83–0.92) (9)]. This intronic SNP maps to a locus associated with suggested long-distance interactions with MYC and PVT1 promoters in multiple cancers (9). rs9543325, an intergenic SNP in chr13q22.1 which is frequently deleted in several cancers (8) was associated with a 37% increased risk of developing pNEN (for subjects with CC genotype), higher than risk estimates from the original study in PDACs [OR 1.26 95% CI (1.18–1.35)] (8). Genes closest to this locus are KLF5, KLF12, PIBF1, DIS3 and BORA, which range up to 586 kb in distance from rs9543325. Krüppel-like factor (KLF) 5 and 12 are members of a family of zinc-finger transcription factors reported to exhibit tumor-suppressor and oncogenic activity, respectively, in various human cancers (27). KLF5 is involved in transcriptional activation of PI3K/Akt signaling, BORA (encoding aurora kinase A activator) is known to regulate cell proliferation and overexpressed in tumors (28). PIBF1 (encoding progesterone immunomodulatory binding factor 1) is speculated to be involved in progesterone-dependent immunomodulation with higher levels in tumors (28). Aberrant expression of DIS3 has been implicated in many cancers and plays a crucial role in gene regulation and small RNA processing (29). rs10919791 lies in the first intron of NR5A2 (chr1q32.1) which plays crucial roles in pancreatic function and development (30). In the current study, rs10919791 was associated with a 35% lower risk of developing pNEN among heterozygotes, similar to the original estimate among PDACs (ORhet 0.76 (0.68–0.84)) (8). The bioinformatics approach with a GEO2R microarray dataset including pancreatic neuroendocrine tumors (n = 6) and non-neoplastic pancreatic tissue samples (n = 5), was used to augment results from our study and explore the plausible role of these genes in pNEN development. Interestingly, MYC, KLF5, NR5A2, CTRB1 and CTRB2 appeared down-regulated among pNENs compared to non-neoplastic pancreatic tissue in the microarray dataset, further implicating these genes in pNEN tumorigenesis. To substantiate these findings, we performed a similar comparison of expression profiles of these genes in PDAC (n = 6) versus healthy pancreas, which showed differential expression of all genes but KLF5 in the same manner. To the best of our knowledge, this is the only publicly available dataset that includes expression data of pNEN. The data we observed in this dataset corroborate our findings of the association analysis. Nevertheless the very small sample size of this dataset is a cause for concern, and these data should be taken with caution. Additional stratified analyses based on age at onset and tumor characteristics (Tables 3 and 4, respectively), despite relatively small genotype strata for some SNPs, also shed light on susceptibility loci that may be useful for improved risk stratification and possible choice of therapy based on tumor characteristics. The contribution of rs7310409 to pNEN development is conceivable given that pleiotropic HNF1A plays a crucial role, as a tumor suppressor, in the transcriptional regulation in endocrine and exocrine pancreas development and homeostasis (31). This locus has also been reported as a prime candidate mediating susceptibility to type 2 diabetes among Caucasians (32). Although the role of p53 homolog, p63 in pancreatic cancer is poorly understood, it has been implicated in tumorigenesis and metastasis via cell cycle arrest and apoptotic mechanisms (33). Our observed association between heterozygous carriers of this polymorphism and metastatic pNEN tumors is therefore plausible. Table 4. Associations between SNPs and pNEN tumor characteristics Locus  Gene  SNP  Genotype groups  OR (95% CI)a  P  Grade (G1/G2/well differentiated tumors versus G3/poorly differentiated tumors)   1q32.1  NR5A2  rs10919791  GG/GA/AA 115/38/15  5.66 (1.18–27.11)b  0.03   22q12.1  ZNRF3  rs16986825  CC/CT/TT 113/62/9  7.34 (1.02–52.67)c 7.44 (1.10–50.53)d  0.05 0.04  Stage (I/II/III versus IV)   8q24.21  MIR1208, PVT1  rs1561927  CC/TC/TT 18/55/95  2.39 (1.18–4.85)b  0.02   3q28  TP63  rs9854771  AA/GA/GG 22/67/81  0.44 (0.22–0.90)b  0.02   1p13.1  WNT2B  rs351365  AA/GA/GG 11/58/99  4.15 (1.01–16.95)c 4.62 (1.16–18.38)d  0.05 0.03  Functional status (functional tumors versus non-functional tumors)   12q24.21  HNF1A  rs7310409  AA/GA/GG 32/79/66  0.36 (0.16–0.79)b  0.01   18q21.2  GRP  rs1517037  CC/CT/TT 118/56/4  0.37 (0.16–0.85)b  0.02  Locus  Gene  SNP  Genotype groups  OR (95% CI)a  P  Grade (G1/G2/well differentiated tumors versus G3/poorly differentiated tumors)   1q32.1  NR5A2  rs10919791  GG/GA/AA 115/38/15  5.66 (1.18–27.11)b  0.03   22q12.1  ZNRF3  rs16986825  CC/CT/TT 113/62/9  7.34 (1.02–52.67)c 7.44 (1.10–50.53)d  0.05 0.04  Stage (I/II/III versus IV)   8q24.21  MIR1208, PVT1  rs1561927  CC/TC/TT 18/55/95  2.39 (1.18–4.85)b  0.02   3q28  TP63  rs9854771  AA/GA/GG 22/67/81  0.44 (0.22–0.90)b  0.02   1p13.1  WNT2B  rs351365  AA/GA/GG 11/58/99  4.15 (1.01–16.95)c 4.62 (1.16–18.38)d  0.05 0.03  Functional status (functional tumors versus non-functional tumors)   12q24.21  HNF1A  rs7310409  AA/GA/GG 32/79/66  0.36 (0.16–0.79)b  0.01   18q21.2  GRP  rs1517037  CC/CT/TT 118/56/4  0.37 (0.16–0.85)b  0.02  Only association results for SNPs with P < 0.05 and with outcomes below FPRP-level 0.2 are marked in bold. Analyses were performed with adjustment for age, gender and country of origin. aAll G1 and G2 (i.e. well-differentiated) tumors, non-functional tumors and Stages I, II and III tumors (i.e. no nodal metastasis) represent the reference category for Grade, Functional status and Stage, respectively. bMM versus Mm; cMM versus mm; d(MM + Mm) versus mm; where M, major allele and m, minor allele. View Large Nevertheless, the small size observed in each stratum due to the rarity of these neoplasms, raises a valid concern on the possible interpretation of our findings. We therefore applied the FPRP approach instead of a much lower P value adjusted for multiple comparisons, as the latter would have resulted in unnecessarily low power for our high prior probability hypotheses, particularly in these neoplasms where collection of larger number of cases is unrealistic. Notwithstanding, all noteworthy findings reported here must be interpreted with caution and replicated in possibly larger and independent studies. The major strengths of this study are its relatively large size, considering the low incidence of pNEN and data analysis approaches applied to minimize the effect of potentially confounding factors and more importantly evaluate the noteworthiness of our findings. The integration of gene expression profiling provides valuable insight to shared genetic correlations and expands the probable margins of our results. A limitation of our investigations however, was that due to lack of clinical information, we could not stratify subjects based on familial history of PDAC or other non-genetic factors such as history of diabetes mellitus. Taken together, these results illuminate genetic similarities between the two main groups of pancreatic neoplasms, suggesting that common variations at three known PDAC susceptibility loci –13q22.1 (TP63, rs9543325), 1q32.1 (NR5A2, rs10919791) and 8q24.21 (MIR1208/PVT1, rs1561927) potentially influence tumorigenesis of pancreatic beta cells in a similar manner as in exocrine pancreatic cells. Our results also suggest that genetic predisposing factors vary on the basis of age at pNEN onset as well as tumor characteristics. Striking similarities between embryogenesis and tumorigenesis which have become more apparent in past decades taken together with our genetic findings may provide groundwork for future evaluation of the potential usability of these loci as predictive biomarkers and in risk assessment. Functional studies are also warranted to better understand the biological implications of these loci in pancreatic hormone-producing (endocrine) and enzyme-producing (exocrine) cell function. Funding This study was supported by intramural funds from the German Cancer Research Centre (DKFZ, Heidelberg, Germany) awarded to F.C. G.C. was supported by AIRC grant IG 2015, Id 17177. Abbreviations CI confidence interval FPRP false positive report probability GEO Gene Expression Omnibus GWAS genome-wide association studies KLF Krüppel-like factor OR odds ratio PANDoRA PANcreatic Disease ReseArch PDAC pancreatic ductal adenocarcinoma pNEN pancreatic neuroendocrine neoplasm SNP single nucleotide polymorphism Acknowledgements The authors acknowledge Angelika Stein (DKFZ, Heidelberg, Germany) for her technical support. Conception and design: D.C, F.C and O.O. Acquisition of samples and clinical data: all authors. Analysis and interpretation of data (e.g., experimental and statistical analysis): O.O, D.C and F.C. Writing, review and/or revision of the manuscript: O.O, G.C, D.C, F.C. and all other authors contributed to its critical revision. Conflict of Interest Statement: None declared. References 1. Capurso, G.et al.  ( 2009) Risk factors for sporadic pancreatic endocrine tumors: a case-control study of prospectively evaluated patients. Am. J. Gastroenterol ., 104, 3034– 3041. Google Scholar CrossRef Search ADS PubMed  2. Yao, J.C.et al.  ( 2008) One hundred years after “carcinoid”: epidemiology of and prognostic factors for neuroendocrine tumors in 35,825 cases in the United States. J. Clin. Oncol ., 26, 3063– 3072. Google Scholar CrossRef Search ADS PubMed  3. Li, D.et al.  ( 2004) Pancreatic cancer. Lancet , 363, 1049– 1057. Google Scholar CrossRef Search ADS PubMed  4. Haugvik, S.P.et al.  ( 2015) Diabetes, smoking, alcohol use, and family history of cancer as risk factors for pancreatic neuroendocrine tumors: a systematic review and meta-analysis. Neuroendocrinology , 101, 133– 142. Google Scholar CrossRef Search ADS PubMed  5. Jiao, Y.et al.  ( 2011) DAXX/ATRX, MEN1, and mTOR pathway genes are frequently altered in pancreatic neuroendocrine tumors. Science , 331, 1199– 1203. Google Scholar CrossRef Search ADS PubMed  6. Scarpa, A.et al.  ; Australian Pancreatic Cancer Genome Initiative. ( 2017) Whole-genome landscape of pancreatic neuroendocrine tumours. Nature , 543, 65– 71. Google Scholar CrossRef Search ADS PubMed  7. Amundadottir, L.et al.  ( 2009) Genome-wide association study identifies variants in the ABO locus associated with susceptibility to pancreatic cancer. Nat. Genet ., 41, 986– 990. Google Scholar CrossRef Search ADS PubMed  8. Petersen, G.M.et al.  ( 2010) A genome-wide association study identifies pancreatic cancer susceptibility loci on chromosomes 13q22.1, 1q32.1 and 5p15.33. Nat. Genet ., 42, 224– 228. Google Scholar CrossRef Search ADS PubMed  9. Wolpin, B.M.et al.  ( 2014) Genome-wide association study identifies multiple susceptibility loci for pancreatic cancer. Nat. Genet ., 46, 994– 1000. Google Scholar CrossRef Search ADS PubMed  10. Childs, E.J.et al.  ( 2015) Common variation at 2p13.3, 3q29, 7p13 and 17q25.1 associated with susceptibility to pancreatic cancer. Nat. Genet ., 47, 911– 916. Google Scholar CrossRef Search ADS PubMed  11. Campa, D.et al.  ( 2015) TERT gene harbors multiple variants associated with pancreatic cancer susceptibility. Int. J. Cancer , 137, 2175– 2183. Google Scholar CrossRef Search ADS PubMed  12. Campa, D.et al.  ( 2013) Genetic susceptibility to pancreatic cancer and its functional characterisation: the PANcreatic Disease ReseArch (PANDoRA) consortium. Dig. Liver Dis ., 45, 95– 99. Google Scholar CrossRef Search ADS PubMed  13. Bosman, F.T.et al.   ( 2010) WHO Classification of Tumours of the Digestive System . 4th ed. World Health Organization, Geneva. 14. Wacholder, S.et al.  ( 2004) Assessing the probability that a positive report is false: an approach for molecular epidemiology studies. J. Natl. Cancer Inst ., 96, 434– 442. Google Scholar CrossRef Search ADS PubMed  15. Ramage, J.K.et al.  ; UK and Ireland Neuroendocrine Tumour Society. ( 2012) Guidelines for the management of gastroenteropancreatic neuroendocrine (including carcinoid) tumours (NETs). Gut , 61, 6– 32. Google Scholar CrossRef Search ADS PubMed  16. Klimstra, D.S.et al.  ( 2010) The pathologic classification of neuroendocrine tumors: a review of nomenclature, grading, and staging systems. Pancreas , 39, 707– 712. Google Scholar CrossRef Search ADS PubMed  17. Park, M.et al.  ( 2014) Characterization of gene expression and activated signaling pathways in solid-pseudopapillary neoplasm of pancreas. Mod. Pathol ., 27, 580– 593. Google Scholar CrossRef Search ADS PubMed  18. Barrett, T.et al.  ( 2013) NCBI GEO: archive for functional genomics data sets–update. Nucleic Acids Res ., 41( Database issue), D991– D995. Google Scholar PubMed  19. Zhou, Q.et al.  ( 2008) In vivo reprogramming of adult pancreatic exocrine cells to beta-cells. Nature , 455, 627– 632. Google Scholar CrossRef Search ADS PubMed  20. Hezel, A.F.et al.  ( 2006) Genetics and biology of pancreatic ductal adenocarcinoma. Genes Dev ., 20, 1218– 1249. Google Scholar CrossRef Search ADS PubMed  21. Pour, P.M.et al.  ( 2003) What is the origin of pancreatic adenocarcinoma? Mol. Cancer , 2, 13. Google Scholar CrossRef Search ADS PubMed  22. Maitra, A.et al.  ( 2012) Disputed paternity: the uncertain ancestry of pancreatic ductal neoplasia. Cancer Cell , 22, 701– 703. Google Scholar CrossRef Search ADS PubMed  23. Deshpande, V.et al.  ( 2003) Ductulo-insular pancreatic endocrine neoplasms: clinicopathologic analysis of a unique subtype of pancreatic endocrine neoplasms. Am. J. Surg. Pathol ., 27, 461– 468. Google Scholar CrossRef Search ADS PubMed  24. Landsman, L.et al.  ( 2011) Pancreatic mesenchyme regulates epithelial organogenesis throughout development. PLoS Biol ., 9, e1001143. Google Scholar CrossRef Search ADS PubMed  25. Regitnig, P.et al.  ( 2001) Insulinoma of the pancreas with insular-ductular differentiation in its liver metastasis–indication of a common stem-cell origin of the exocrine and endocrine components. Virchows Arch ., 438, 624– 628. Google Scholar CrossRef Search ADS PubMed  26. Oliver-Krasinski, J.M.et al.  ( 2009) The diabetes gene Pdx1 regulates the transcriptional network of pancreatic endocrine progenitor cells in mice. J. Clin. Invest ., 119, 1888– 1898. Google Scholar CrossRef Search ADS PubMed  27. McConnell, B.B.et al.  ( 2010) Mammalian Krüppel-like factors in health and diseases. Physiol. Rev ., 90, 1337– 1381. Google Scholar CrossRef Search ADS PubMed  28. Azmi A. ( 2014) Molecular Diagnostics and Treatment of Pancreatic Cancer: Systems and Network Biology Approaches . Elsevier, Amsterdam. 29. Robinson, S.R.et al.  ( 2015) The 3’ to 5’ exoribonuclease DIS3: from structure and mechanisms to biological functions and role in human disease. Biomolecules , 5, 1515– 1539. Google Scholar CrossRef Search ADS PubMed  30. Kelly, V.R.et al.  ( 2010) Dax1 up-regulates Oct4 expression in mouse embryonic stem cells via LRH-1 and SRA. Mol. Endocrinol ., 24, 2281– 2291. Google Scholar CrossRef Search ADS PubMed  31. Hoskins, J.W.et al.  ( 2014) Transcriptome analysis of pancreatic cancer reveals a tumor suppressor function for HNF1A. Carcinogenesis , 35, 2670– 2678. Google Scholar CrossRef Search ADS PubMed  32. Voight, B.F.et al.  ; MAGIC investigators; GIANT Consortium. ( 2010) Twelve type 2 diabetes susceptibility loci identified through large-scale association analysis. Nat. Genet ., 42, 579– 589. Google Scholar CrossRef Search ADS PubMed  33. Bergholz, J.et al.  ( 2012) Role of p63 in development, tumorigenesis and cancer progression. Cancer Microenviron ., 5, 311– 322. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2017. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com.

Journal

CarcinogenesisOxford University Press

Published: Mar 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

Monthly Plan

  • Read unlimited articles
  • Personalized recommendations
  • No expiration
  • Print 20 pages per month
  • 20% off on PDF purchases
  • Organize your research
  • Get updates on your journals and topic searches

$49/month

Start Free Trial

14-day Free Trial

Best Deal — 39% off

Annual Plan

  • All the features of the Professional Plan, but for 39% off!
  • Billed annually
  • No expiration
  • For the normal price of 10 articles elsewhere, you get one full year of unlimited access to articles.

$588

$360/year

billed annually
Start Free Trial

14-day Free Trial