Colorectal cancer genetic variants are also associated with serrated polyposis syndrome susceptibility.

BACKGROUND: Serrated polyposis syndrome (SPS) is a clinical entity characterised by large and/ormultiple serrated polyps throughout the colon and increased risk for colorectal cancer (CRC). The basis for SPS genetic predisposition is largely unknown. Common, low-penetrance genetic variants have been consistently associated with CRC susceptibility, however, their role in SPS genetic predisposition has not been yet explored.
OBJECTIVE: The aim of this study was to evaluate if common, low-penetrance genetic variants for CRC risk are also implicated in SPS genetic susceptibility.
METHODS: A case-control study was performed in 219 SPS patients and 548 asymptomatic controls analysing 65 CRC susceptibility variants. A risk prediction model for SPS predisposition was developed.
RESULTS: Statistically significant associations with SPS were found for seven genetic variants (rs4779584-GREM1, rs16892766-EIF3H, rs3217810-CCND2, rs992157-PNKD1/TMBIM1, rs704017-ZMIZ1, rs11196172-TCF7L2, rs6061231-LAMA5). The GREM1 risk allele was remarkably over-represented in SPS cases compared with controls (OR=1.573, 1.21-2.04, p value=0.0006). A fourfold increase in SPS risk was observed when comparing subjects within the highest decile of variants (≥65) with those in the first decile (≤50).
CONCLUSIONS: Genetic variants for CRC risk are also involved in SPS susceptibility, being the most relevant ones rs4779584-GREM1, rs16892766-EIF3H and rs3217810-CCND2. © Author(s) (or their employer(s)) 2020. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

Introduction

Colorectal cancer (CRC) is currently one of the most common neoplasms in developed countries, and it represents the second most fatal malignancy after lung cancer.1 This neoplasm is developed from different precursor lesions, including conventional adenomas and serrated polyps. This heterogeneity drives the progression to carcinoma through distinct pathways.2 While most CRC progress through the adenoma–carcinoma sequence, serrated polyps, previously known as hyperplastic polyps, are also considered precursor lesions with an alternative pathway to CRC.3 The term serrated polyp refers to a lesion with a serrated or ‘sawtooth’ appearance of the colonic crypts at microscopy. On the other hand, the risk of developing CRC is influenced by both environmental and genetic factors. Part of this germline CRC predisposition is already known including both rare, high-penetrance and common, low-penetrance genetic variants.4 Genome-wide association studies (GWAS) have been conducted since 2007 and have identified ~130 common, low-penetrance genetic variants associated with CRC risk.5

Interestingly, prior studies have evaluated the association between GWAS-identified SNPs linked to CRC susceptibility and polyp subtypes, including adenomas, serrated polyps and advanced/non-advanced polyps but not the association with serrated polyposis syndrome (SPS).6–9

SPS is a condition characterised by the development of large and/or multiple serrated polyps throughout the colon and increased risk for CRC. The prevalence of CRC in patients with SPS has been estimated to range between 15% and 35%.3 10 In order to help in identifying this clinical entity, the WHO established in 2010 the following criteria defining SPS as the presence of (I) at least five serrated polyps proximal to the sigmoid colon, of which two measure at least 10 mm in diameter and/or (II) any number of serrated polyps occurring proximal to the sigmoid colon in an individual who has a first-degree relative with SPS and/or (III) more than 20 serrated polyps spread throughout the colon.11 Despite recent developments in sequencing technologies, the genetic aetiology of SPS remains largely unknown. The only proposed gene for germline SPS predisposition is RNF43, which would explain a small proportion of SPS cases.12–15

Keeping in mind that serrated polyps are considered CRC precursor lesions and considering that SPS confers a relatively high risk for CRC, the current study aims to test if common, low-penetrance genetic variants for CRC risk are also involved in the predisposition to SPS.

Materials and methods

Study population

Our study comprised 219 Spanish patients ascertained in high-risk clinics for CRC that fulfilled the SPS clinical criteria.11 They included 130 men (59.36%) and 89 women (40.63%) aged between 20–78 (mean age 54.62±11.08 years old). Presence of conventional adenomas in SPS patients was frequent (around 70%). We also included 548 asymptomatic controls from the Barcelona CRC screening programme with a positive faecal immunochemical test but a negative colonoscopy showing no relevant findings related to CRC risk. The control group included 269 men (49.08%) and 279 women (50.92%), aged between 50–69 (mean age 59.49±5.57 years old). Cases and controls included in this study were negative for a family history of hereditary or familial CRC or adenomatous polyposis (ie, ≥2 first-degree relatives with CRC/adenomatous polyposis or one first-degree relative diagnosed before the age of 60). DNA was obtained from frozen peripheral blood by standard extraction procedures for all samples. RNF43 mutations were excluded in the cases cohort. Mutations in other high-penetrance CRC genes such as MUTYH could not be ruled out although their contribution may be considered scarce since cases did not present a relevant family history for hereditary or familial CRC or adenomatous polyposis. Written informed consent was obtained from all individuals.

SNP genotyping and quality control

Sixty-five SNPs previously associated with CRC susceptibility were selected and genotyped in all available DNA samples. SNPs were selected from previous GWAS published before mid-2017 (www.ebi.ac.uk/gwas/, GWAS Catalogue). Genotyping was performed with the Biomark 96.96 Genotyping dynamic array (Fluidigm, San Francisco, CA, USA) and TaqMan assays (Thermo Fisher Scientific, Waltham, MA, USA). Data quality was assessed using the Fluidigm SNP Genotyping Analysis and PLINK softwares. Samples and SNPs with genotyping success rate below 90% were removed from subsequent analyses (including rs7229639, rs3764482). One SNP was monomorphic and was also eliminated (rs10904849). In order to further test for genotyping quality, five samples were duplicated and genotype concordance was 100%. Deviation of the genotype frequencies in the controls from those expected under Hardy-Weinberg equilibrium (HWE) was assessed by the χ2 test (1df). Each SNP was in HWE (p value >0.01) in controls (data not shown), thereby excluding the possibility of genotyping artefacts and any hidden population stratification. After quality filtering, the final available dataset comprised 741 samples (215 cases and 526 controls) and 62 SNPs. The overall genotyping success rate in the remaining individuals was 99.07%.

Statistical analysis

Logistic regression analysis was used to evaluate the association between each SNP and SPS risk under an additive model. The total number of risk alleles was calculated for all samples and coded as 0, 1 or 2 for each SNP assuming an additive genetic effect. Two-sided t-test was applied to compare the mean number of risk alleles between cases and controls. Furthermore, an additive genetic risk score (GRS) was developed by using a general linear model to assess for genetic susceptibility. GRS was defined as the count of risk alleles across all available SNPs. An unweighted GRS was preferred. A weighted genetic risk score was also tested, using the originally published log-ORs as weights. Since the results were essentially the same, and the weights may be biassed due to the winner’s course effect, we opted to use the unweighted score. All statistical tests were two-sided, and p values <0.05 were considered statistically significant. All analyses were performed using PLINK v1.09 (V.3.4.1). Bonferroni correction was used for multiple testing adjustment (α adjusted=0.05/62=8.06×10−4).

Results

Association tests for individual SNPs

A total of 215 SPS patients and 526 controls were successfully genotyped for 62 SNPs previously linked with genetic susceptibility to CRC. First, the frequency of CRC risk alleles between SPS cases and controls was compared and those significantly enriched in the SPS cohort were detected. Results are shown in table 1. Seven CRC SNPs showed statistically significant associations with SPS (rs992157, rs16892766, rs704017, rs11196172, rs3217810, 4779584, rs6061231) and these genetic associations were in the same direction as previously reported for CRC susceptibility. Despite being not significant, most remaining SNPs (42/52), showed ORs in the same directions as those previously described in the literature for CRC susceptibility.

Table 1

Case-control association results obtained by logistic regression analyses

SNP*	Region	Nearby gene	Variant type	Risk allele	RAF cases/controls	OR	95% CI	P value
rs12080929	1p33	SLC5A9	Intronic variant	T	0.73/0.71	1.086	(0.85 to 1.39)	0.5093
rs72647484	1p36.12	CDC42/WNT	Intergenic variant	C	0.08/0.08	1.096	(0.72 to 1.66)	0.6656
rs10911251	1q25.3	LAMC1	Intronic variant	A	0.58/0.59	0.9767	(0.78 to 1.23)	0.8409
rs6691170	1q41	DUSP10	Intergenic variant	T	0.34/0.36	0.9123	(0.72 to 1.15)	0.4376
rs11903757	2q32.3	NABP1	Intergenic variant	C	0.13/0.15	0.8341	(0.61 to 1.15)	0.2653
rs992157	2q35	PNKD/TMBIM1	Intronic variant	A	0.58/0.5	1.36	(1.09 to 1.7)	0.007454
rs11676348	2q35	CXCR2	Intergenic variant	T	0.49/0.54	0.8345	(0.67 to 1.03)	0.09742
rs812481	3p14.1	LRIG1	Intronic variant	G	0.53/0.52	1.022	(0.82 to 1.28)	0.8472
rs35360328	3p22.1	CTNNB1	intergenic variant	A	0.14/0.14	0.9986	(0.72 to 1.38)	0.9931
rs10936599	3q26.2	TERC	Synonymous variant	C	0.78/0.75	1.174	(0.89 to 1.54)	0.2474
rs7136702	4q13.2	LARP4/DIP2B	Intergenic variant	T	0.32/0.34	0.9124	(0.72 to 1.16)	0.4501
rs3987	4q26	NDST3	Intergenic variant	G	0.45/0.46	0.9761	(0.78 to 1.22)	0.831
rs2736100	5p15.33	TERT	Intronic variant	A	0.5/0.52	0.9495	(0.76 to 1.19)	0.6472
rs647161	5q31.1	PITX1	Intronic variant	A	0.73/0.69	1.196	(0.94 to 1.52)	0.1458
rs1321311	6p1.2	CDKN1A	Intergenic variant	A	0.29/0.27	1.086	(0.84 to 1.4)	0.527
rs4711689	6p21.1	TFEB	Intronic variant	A	0.54/0.54	1.01	(0.81 to 1.27)	0.9282
rs11987193	8p12	DUSP4	Intergenic variant	C	0.7/0.74	0.8091	(0.63 to 1.05)	0.1055
rs16892766	8q23.3	EIF3H	Intergenic variant	C	0.08/0.05	1.686	(1.09 to 2.62)	0.01959
rs6469656	8q23.3	EIF3H	Intragenic variant	A	0.9/0.89	1.112	(0.78 to 1.58)	0.5505
rs6983267	8q24.21	MYC	Non-coding transcript variant/intronic variant	G	0.54/0.51	1.099	(0.88 to 1.37)	0.4035
rs719725	9p24	TPD52L3/UHRF2	Intronic variant	A	0.62/0.58	1.198	(0.95 to 1.51)	0.1215
rs10795668	10p14	ARN5SP299/GATA3	Intronic variant	A	0.3/0.33	0.8803	(0.69 to 1.12)	0.2995
rs704017	10q22.3	ZMIZ1-AS1	Intronic variant	G	0.57/0.5	1.307	(1.04 to 1.64)	0.02218
rs1035209	10q24.2	ABCC2/MRP2	Intergenic variant	T	0.16/0.16	0.9457	(0.7 to 1.28)	0.7188
rs11190164	10q24.2	SLC25A28	Intergenic variant	G	0.2/0.23	0.8207	(0.62 to 1.08)	0.1657
rs11196172	10q25.2	TCF7L2	Intronic variant	A	0.11/0.17	0.648	(0.46 to 0.91)	0.01186
rs12241008	10q25.2	VTI1A	Intronic variant	C	0.11/0.09	1.272	(0.87 to 1.86)	0.2129
rs4246215	11q12.2	FEN1	3'-UTR variant	T	0.3/0.33	0.8561	(0.67 to 1.09)	0.2136
rs174537	11q12.2	MYRF	Intronic variant	G	0.7/0.67	1.161	(0.91 to 1.48)	0.226
rs1535	11q12.2	FADS2	Intronic variant	A	0.69/0.66	1.168	(0.92 to 1.49)	0.206
rs174550	11q12.2	FADS1	Intronic variant	T	0.69/0.67	1.124	(0.88 to 1.43)	0.3422
rs3824999	11q13.4	POLD3	Intronic variant	G	0.53/0.49	1.146	(0.92 to 1.43)	0.2257
rs3802842	11q23.1	COLCA1/COLCA2	Intronic variant	C	0.26/0.26	0.9952	(0.78 to 1.28)	0.9696
rs2238126	12p13.2	ETV6	Intronic variant	G	0.14/0.16	0.8391	(0.61 to 1.15)	0.2764
rs10849432	12p13.31	CD9	Intronic variant	T	0.88/0.88	1.065	(0.75 to 1.52)	0.7295
rs11064437	12p13.31	SPSB2	Splice acceptor variant	C	1/0.99	2.453	(0.57 to 10.48)	0.2256
rs3217901	12p13.32	CCND2	Intronic variant	G	0.38/0.34	1.162	(0.92 to 1.47)	0.2046
rs10774214	12p13.32	CCND2	Intronic variant	T	0.37/0.35	1.125	(0.89 to 1.42)	0.3242
rs3217810	12p13.32	CCND2	Intronic variant	T	0.12/0.08	1.686	(1.16 to 2.45)	0.006237
rs11169552	12q13.12	DIP2B/ATF1	Intergenic variant	C	0.76/0.76	0.9557	(0.74 to 1.24)	0.7319
rs3184504	12q24.12	SH2B3	Mis-sense variant	C	0.58/0.56	1.066	(0.85 to 1.33)	0.5737
rs59336	12q24.21	TBX3	Intronic variant	T	0.54/0.52	1.092	(0.87 to 1.37)	0.4528
rs73208120	12q24.22	NOS1	Intronic variant	G	0.06/0.05	1.053	(0.64 to 1.74)	0.8393
rs4444235	14q22.2	BMP4	Intergenic variant	C	0.56/0.55	1.036	(0.83 to 1.3)	0.7544
rs1957636	14q22.2	BMP4	Intronic variant	T	0.41/0.4	1.044	(0.83 to 1.31)	0.7078
rs17094983	14q23.1	RTN1	Intergenic variant	G	0.86/0.86	0.9959	(0.72 to 1.38)	0.9803
rs4779584	15q13.3	GREM1	Intergenic variant	T	0.25/0.17	1.573	(1.21 to 2.04)	0.0006432
rs9929218	16q22.1	CDH1	Intronic variant	G	0.72/0.69	1.169	(0.91 to 1.49)	0.212
rs16941835	16q24.1	FOXL1	Intergenic variant	C	0.17/0.18	0.9193	(0.68 to 1.24)	0.5842
rs12603526	17q13.3	NXN	Intronic variant	C	0.01/0.01	1.368	(0.45 to 4.13)	0.5786
rs4939827	18q21.1	SMAD7	Intronic variant	T	0.56/0.55	1.063	(0.85 to 1.33)	0.5909
rs12970291	18q22.3	TSHZ1	Intergenic variant	A	0.04/0.03	1.127	(0.6 to 2.12)	0.7105
rs10411210	19q13.11	RHPN2	Intronic variant	C	0.87/0.87	0.9811	(0.71 to 1.36)	0.9099
rs2241714	19q13.2	B9D2	Mis-sense variant/2 kb upstream variant	C	0.67/0.65	1.096	(0.86 to 1.39)	0.4517
rs1800469	19q13.2	TGFB1	2 kb upstream variant/0.5 kb downstream variant	G	0.67/0.65	1.102	(0.87 to 1.4)	0.4215
rs2423279	20p12.3	HAQ1	Intergenic variant	C	0.28/0.31	0.8595	(0.67 to 1.1)	0.2333
rs961253	20p12.3	BMP2	Intergenic variant	A	0.34/0.32	1.085	(0.86 to 1.37)	0.4941
rs4813802	20p12.3	BMP2	Regulatory region variant?	G	0.36/0.32	1.216	(0.96 to 1.54)	0.1046
rs6066825	20q13.13	PREX1	Intronic variant	A	0.6/0.58	1.113	(0.89 to 1.39)	0.3445
rs6061231	20q13.33	LAMA5	Intergenic variant	C	0.78/0.72	1.412	(1.08 to 1.85)	0.01189
rs4925386	20q13.33	LAMA5	Intronic variant	C	0.73/0.68	1.261	(0.98 to 1.62)	0.06688
rs5934683	Xp22.2	SHROOM2	Intronic variant	C	0.59/0.61	0.9276	(0.71 to 1.22)	0.5873

Association results for 216 SPS cases and 526 controls. Results are based on the reported risk allele from previous CRC GWAS. Statistically significant associations are denoted in bold (p value <0.05).

*SNPs with genotyping success rate below 90% were removed from subsequent analyses (including rs7229639, rs3764482). We also removed a monomorphic SNP (rs10904849).

CRC, colorectal cancer; GWAS, genome-wide association studies; RAF, risk allele frequency; SPS, serrated polyposis syndrome; UTR, untranslated region.

Genetic risk score model

The presence of higher number of CRC risk alleles in the SPS cohort when compared with controls was also evaluated. The distribution of risk by allele number for the 62 genotyped SNPs is displayed in figure 1, both for cases and controls. The distribution of risk alleles followed a normal distribution in both SPS cases and controls with a shift towards a higher number of risk alleles in affected individuals consistent with a cumulative impact of CRC risk alleles on SPS predisposition. The mean number of risk alleles in control individuals was 56.20 compared with 57.46 in SPS cases and there was a statistically significant difference in the mean number of risk alleles between SPS cases and controls (difference: −1.26; two-sided t-test p value=0.0016).

Figure 1

Genetic risk score. Distribution of risk by allele number for the 62 SNPs genotyped. The presence of multiple CRC risk alleles is displayed for SPS cases (bold bars) and controls (stripped bars). CRC, colorectal cancer; SPS, serrated polyposis syndrome

Next, a GRS was calculated for SPS cases and controls when carrying an increasing number of CRC risk alleles. The median number of risk alleles in controls, 56, was considered as reference. SPS cases and controls were grouped considering subjects carrying ≤45 risk alleles and ≥65 alleles, because of the small number of subjects at these extremes. The risk score was higher in the SPS group compared with the control cohort (OR=1.05, 95% CI 1.02 to 1.09, p value=0.0019). We observed that there was a twofold increase in SPS risk for subjects in the highest quintile of risk alleles (≥62) compared with those in the first quintile (≤53) (OR=2.16, 95% CI 1.29 to 3.63, p=0.0034). Additionally, a fourfold increase in SPS risk was detected for subjects in the highest decile of risk alleles (≥65), compared with those in the first decile (≤50) (OR=4.24, 95% CI 1.76 to 10.21, p value=0.0013). As shown in figure 1, the increase in risk per allele was linear, indicating the independent additive contribution of each allele to SPS risk.

Discussion

The association of 65 common, low-penetrance CRC susceptibility variants in SPS development was assessed in a cohort of 768 samples (219 SPS cases and 548 controls) in order to check if common, low-penetrance genetic variants for CRC risk were also involved in SPS genetic susceptibility. Our results showed statistically significant association of seven CRC genetic variants with SPS (rs992157, rs16892766, rs704017, rs11196172, rs3217810, rs4779584 and rs6061231).

Among the detected SPS genetic associations, the most significant corresponded to rs4779584. The T risk allele frequency was remarkably higher in SPS cases (25%) than in controls (17%) (OR=1.573, 1.21 to 2.04, p value=0.0006). This association remained statistically significant even after multiple testing corrections. This SNP maps to chromosomal region 15q13.3, is intergenic and lies between SCG5 and GREM1. It was previously suggested that this variant captures two independent association signals represented by rs16969681 and rs11632715 upstream of GREM1.16 The rs16969681 variant lies upstream of GREM1 and is close to a regulatory element that acts as an allele-specific GREM1 enhancer. The rs16969681 risk allele differentially binds with a higher affinity the intestine-specific transcription factor CDX2 and the Wnt effector TCF7L2 producing a stronger GREM1 expression and promoting tumourigenesis.17 Noteworthy, a 40 kb duplication upstream of this gene was previously linked with increased GREM1 expression in individuals with hereditary mixed polyposis.18 This clinical entity shows some overlap with SPS since patients develop polyps of multiple and mixed morphologies including serrated lesions, Peutz-Jeghers polyps, juvenile polyps, conventional adenomas and CRC (OMIM # 601228). Moreover, this genetic variant had been linked with serrated polyps in previous studies.7 9

Additionally, our results show that rs16892766 at 8q23.3 is associated with SPS susceptibility. Previous fine mapping in this region and functional analysis identified rs16888589 as the potential effector of this association through increasing the expression of EIF3H, promoting carcinogenesis.19 Besides, we also detected an association with SPS for rs3217810 at 12p13.32 located in the intron of CCND2, a cyclin involved in cell cycle G1/S transition.20 Both SNPs had been previously linked with serrated polyps.9

Further novel associations with SPS were also detected in our cohort for rs992157 (2q35, PNKD/TMBIM1), rs704017 (10q22.3, ZMIZ1), rs11196172 (10q25.2, TCF7L2) and rs6061231 (20q13.33, LAMA5). Interestingly, rs11196172 was previously associated with a higher TCF7L2 expression.21 As previously commented, high levels of activated TCF7L2 increase GREM1 expression. Therefore, the rs11196172 risk allele may also have an impact on GREM1 expression. It is also important to highlight that the 20q13.33 region harbouring LAMA5 may be suggested as relevant for SPS susceptibility since two variants from that locus showed either a significant (rs6061231) or borderline significant association (rs4925386).

Regarding the results of our polygenic risk model, the distribution of CRC risk alleles follows a normal distribution in both SPS cases and controls, with a shift towards a higher allele number in SPS cases. These results show an enrichment of CRC susceptibility variants in the SPS cohort, therefore suggesting those variants may play a role in mediating CRC risk through serrated lesions and SPS predisposition. A similar study from our group identified CRC risk variants related to advanced adenoma, implying that part of CRC risk is mediated through susceptibility to this other polyp type.8

It should be noted that this study corresponds to the first analysis of common, low-penetrance CRC risk variants in a SPS cohort. However, we are aware that our results are preliminary and this study has several limitations including limited sample size, lack of environmental data or no replication in an independent SPS cohort. Therefore, further studies are needed to confirm our findings. On the other hand, specific GWAS for SPS susceptibility or serrated polyps are needed to further characterise more precisely the germline predisposition architecture of this clinical entity.

In summary, our study suggests that some CRC risk genetic variants could also be involved in SPS susceptibility. The rs4779584 variant, near GREM1, seems to be the most important effector of the common, low-penetrance SPS susceptibility identified so far. Likewise, rs16892766, and rs3217810 could be relevant since they have been identified linked to serrated lesions in previous studies. There are potential implications of our study in the near future. CRC SNPs predisposing to SPS and new SPS SNPs could be used to identify a subgroup of individuals with increased disease risk and modulate population-based CRC-prevention measures.