Repurposing non-invasive prenatal testing data: Population study of single nucleotide variants associated with colorectal cancer and Lynch syndrome

Abstract

In our previous work, genomic data generated through non-invasive prenatal testing (NIPT) based on low-coverage massively parallel whole-genome sequencing of total plasma DNA of pregnant women in Slovakia was described as a valuable source of population specific data. In the present study, these data were used to determine the population allele frequency of common risk variants located in genes associated with colorectal cancer (CRC) and Lynch syndrome (LS). Allele frequencies of identified variants were compared with six world populations to detect significant differences between populations. Finally, variants were interpreted, functional consequences were searched for and clinical significance of variants was investigated using publicly available databases. Although the present study did not identify any pathogenic variants associated with CRC or LS in the Slovak population using NIPT data, significant differences were observed in the allelic frequency of risk CRC variants previously reported in genome-wide association studies and common variants located in genes associated with LS. As Slovakia is one of the leading countries with the highest incidence of CRC among male patients in the world, there is a need for studies dedicated to investigating the cause of such a high incidence of CRC in Slovakia. The present study also assumed that extensive cross-country data aggregation of NIPT results would represent an unprecedented source of information concerning human genome variation in cancer research.

Introduction

Colorectal cancer (CRC) is the third most common cancer worldwide and the second most common cancer in Europe, causing an estimated 9.4% of all cancer deaths in Europe. CRC is also a serious societal problem in Slovakia, with an incidence rate of 15,7% and a mortality rate of 15,6% [(1,2) GLOBOCAN 2020 data]. Many risk factors and causes are associated with the likelihood of developing CRC, but the main reason is still not fully understood. The considerable geographical variability suggests that CRC is a complex polygenic disease caused by genetic and environmental factors and their interactions. Age, sex, lifestyle, and dietary habits (3,4), including meat and alcohol consumption (5,6), tobacco smoking (7–9), obesity and lack of physical activity (4,10,11) play a major role in the pathogenesis of CRC. Other well-known risk factors may also be inflammatory bowel diseases, acromegaly, renal transplantation with long-term immunosuppression, diabetes mellitus and insulin resistance, cholecystectomy, or androgen deprivation therapy (3,4).

Beside these, inherited susceptibility plays a significant role in the etiology of CRC because it can be responsible for about 35% of all cases of colorectal cancer. However, high-penetrance germline variants in known genes (APC, BRCA2, KRAS, NTS, SMAD4, POLE, BRAF, BMPR1A, POLD1, STK11, MUTYH and DNA mismatch repair genes), which are associated with severe hereditary syndromes, such as familial adenomatous polyposis and Lynch syndrome (also called hereditary non-polyposis colorectal cancer), account for only 5–7% of total CRC cases (4,12). Therefore, the remaining unknown heritability is probably explained by the interaction of common, low-penetrance variants identified through genome-wide association studies (GWAS). GWAS in ethnic/racial minority populations offers the opportunity to uncover genetic susceptibility factors and discover new genomic regions and loci that contribute risk for CRC development. Since 2007, more than 100 common risk variants have been successfully identified in GWAS, which have helped to elucidate the etiology of CRC (13–18).

Lynch syndrome (LS) is an autosomal dominant hereditary cancer syndrome that accounts for approximately 3% of all colorectal cancer cases (19). From a clinical point of view, 10–82% (20) of LS cases are associated with a lifetime risk of developing CRC, unless the risk is significantly lower in other types of cancer (21,22). LS is caused by pathogenic germline mutations in a class of genes called DNA mismatch repair (MMR) genes, mainly MLH1, located in 3p22.2 chromosome, and MSH2, located in 2p21 chromosome (23), which represent 70–85% of cases of LS (24). Mutations found in MSH6 (2p16.3), PMS2 (7p22.1) (25) and MLH3 genes have lower incidence (26). Molecular investigations have also shown that MSH3 (27) and germline 3′deletions of the EPCAM gene, which lead to epigenetic silencing of MSH2 (28), are also implicated in the pathogenesis of LS. As a consequence of MMR pathway inactivation and loss of expression of MMR proteins, DNA replication errors accumulate typically resulting in microsatellite instability (MSI), which is generally detected in LS patients' tumor tissues (29). The diagnosis of LS involves three main steps, identification of patients and their familial history that meet the Amsterdam or Bethesda guidelines, presence of MSI in tumors and immunohistochemical analysis (IHC) of MMR protein expression. A definitive diagnosis of LS must be confirmed by detecting the germline mutations in MMR genes (30).

Non-invasive prenatal testing (NIPT) based on low-coverage massively parallel whole-genome sequencing of plasma DNA from pregnant women generates a large amount of data that provides the resources to investigate human genetic variations in the population. In our previous studies, we described the re-use of the data from NIPT for genome-scale population specific frequency determination of small DNA variants (31) and CNVs (32). Since pregnant women represent a relatively standard sample of the local female population, we assumed this NIPT data could also be used in the population study of CRC, the most common cancer in Slovakia. Some research concerning on genomic analysis of plasma from NIPT has also demonstrated NIPT data's efficiency and utility for viral genetic studies (33), genetic profiling of Vietnamese population (34) or detection of CNV aberrations (32,35).

The main aim of our study was a detailed analysis of common variants (MAF >0,05) that showed evidence of association with CRC in GWAS datasets and characterization of population variability from data generated by NIPT. We assumed that the genetic factors, mainly the increased specific population frequency of CRC and LS variants could be responsible for the high incidence of CRC in Slovakia. To test this hypothesis, allele frequencies of risk CRC variants identified in the Slovak population were compared with allele frequencies of risk CRC variants in 6 worldwide populations. As LS is among the most common hereditary CRC syndromes, the aim of our study was also to analyze population allele frequencies and describe clinical impacts of relevant variants located in known LS predisposing genes. To our knowledge, this was the first population study of CRC using NIPT data conducted exclusively in the Slovak population.

Materials and methods

Data source

The laboratory procedure used, to generate the NIPT data, were as follows: DNA from plasma of peripheral maternal blood was isolated for NIPT analysis from 1,501 pregnant women after obtaining a written informed consent consistent with the Helsinki declaration from the subjects. The population cohort consisted from women in reproductive age between 17–48 years with a median of 35 years. Genomic information from a sample consisted of maternal and fetal DNA fragments. Each included individual agreed to use their genomic data in an anonymized form for general biomedical research. The NIPT study (study ID 35900_2015) was approved by the Ethical Committee of the Bratislava Self-Governing Region (Sabinovska ul.16, 820 05 Bratislava) on 30th April of 2015 under the decision ID 03899_2015. Blood samples were collected to EDTA tubes and plasma was separated in dual centrifugation procedure. DNA was isolated from 700 µl of plasma using DNA Blood Mini kit (Qiagen) according to standard protocol. Sequencing libraries were prepared from each sample using TruSeq Nano kit HT (Illumina) following standard protocol with omission of DNA fragmentation step. Each sample was normalized to 4 nm library and the final concentration of libraries was 2,8 pM. Individual barcode labelled libraries were pooled and sequenced using low-coverage whole-genome sequencing on an Illumina NextSeq500 platform (Illumina) by performing paired end sequencing of 2×35 bases (36).

Data analysis

The datasets generated and/or analyzed during the current study are available in the DSpace repository, https://dspace.uniba.sk/xmlui/handle/123456789/27 (31).

Analyses of common variants previously reported to be risk variants for CRC

We combined genotype data from all previously reported GWAS studies available online (https://www.gwascentral.org/) for the years 2007–2020, specifically 66 GWAS studies of CRC risk variants that included individuals with European, Asian and African American ancestry. Using data from these GWAS datasets, we identified 116 risk variants associated with CRC, which were then merged with our data of identified variants from NIPT. Risk variants that were not found in NIPT data were excluded from the analysis. All identified variants in the Slovak population used for further analyses were common (MAFs >0.05). Subsequently, allele frequencies of CRC risk variants for each population (East Asian, South Asian, African, American, Finnish European and non-Finnish European) were extracted from the gnomAD database available online (v3.0, downloaded from http://gnomad.broadinstitute.org/downloads) and compared with our frequencies determined for the Slovak population from NIPT data. Allele frequency in each population and allele frequency differences were plotted using boxplots. Outliers of boxplots that represent variants with highly different frequencies between Slovak and non-Finnish populations were annotated via published literature and studies [in dbSNP (https://www.ncbi.nlm.nih.gov/snp/) and GWAS (https://www.gwascentral.org/)]. To assess the relations between allele frequency of CRC risk variants in each population, we also used Principal Component Analysis (PCA) using matplotlib.pyplot library, which reduces the dimension of the data to a graphically interpretable 2D or 3D dimension. Consequently, we obtained information on which populations have similar or different allele frequencies of the identified CRC risk variants.

Analyses of variants located in genes associated with LS

After analyzing variants associated with CRC, we focused on the study of variants associated with LS. First, we filtered out a group of variants located in 7 genes known to be associated with LS (MLH1, PMS2, MSH6, MLH3, MSH2, TGFBR2, EPCAM). The genomic locations of genes were determined by the GeneCards database (https://www.genecards.org/). From the dataset of identified variants in LS associated genes, we excluded variants in low complexity genomic regions (soft masked in the reference FASTA file), eliminating the variants that could represent sequencing artifacts or repetitive regions. All variants that were used for further analysis were annotated using Ensembl Variant Effect Predictor (VEP, version 101_GRCh38). In our dataset, based on ClinVar database annotation of the most common types of pathogenic and likely pathogenic variants associated with LS (https://www.ncbi.nlm.nih.gov/clinvar), we selected variants including frameshift, missense, nonsense, splice site, non-coding and UTR variants. After this filtering, allele frequencies for both groups of variants (all variants identified in LS genes and selected types of variants) for each population (East Asian, South Asian, African, American, Finnish European and non-Finnish European) were extracted from the gnomAD database (v3.0, downloaded from http://gnomad.broadinstitute.org/downloads) and compared with our frequencies determined for the Slovak population from NIPT data. Allele frequency in each population and allele frequency differences were plotted using boxplots and PCA analysis using matplotlib.pyplot library. Outliers of boxplots representing variants with allele frequency differences more than 10% were annotated via published literature and studies [in dbSNP (https://www.ncbi.nlm.nih.gov/snp/) and GWAS (https://www.gwascentral.org/].

Results

Analyses of common variants previously reported to be risk variants for CRC

In the analysis of the 66 GWAS studies that included all identified risk variants associated with colorectal carcinogenesis from 2007–2020, we investigated 116 variants. There were 25 independent CRC risk variants in Asian population, 62 risk variants found in European population, 27 risk variants in both European and Asian population and 2 risk variants located in African American population that were previously reported in GWAS (Table I).

Table I.

Identification of 116 risk variants associated with colorectal cancer from 66 genome-wide association studies between 2007 and 2020.

First author, year	rs_ID	Chr	POS	POP	Gene	PMID	(Refs.)
Wang et al, 2017	rs7252505	19q13	33,084,158	AFR A	GPATCH1	28295283	(46)
Wang et al, 2017	rs56848936	19q13.3	46,321,507	AFR A	SYMPK	28295283	(46)
Lu et al, 2019	rs7542665	1p31.3	62,673,037	ASN	L1TD1	30529582	(13)
Law et al, 2019	rs12143541	1p32.3	55,247,852	ASN	TTC22	31089142	(43)
Lu et al, 2019	rs201395236	1q44	245,181,421	ASN	EFCAB2	30529582	(13)
Lu et al, 2019	rs7606562	2p16.3	48,686,695	ASN	PPP1R21	30529582	(13)
Lu et al, 2019	rs113569514	3q22.2	133,748,789	ASN	SLCO2A1	30529582	(13)
Lu et al, 2019	rs12659017	5q23.2	125,988,175	ASN	ALDH7A1, PHAX	30529582	(13)
Law et al, 2019	rs639933	5q31.1	134,467,751	ASN	C5orf66, LOC105379188	31089142	(43)
Jia et al, 2013	rs647161	5q31.1	134,499,092	ASN	PITX1	23263487	(58)
Law et al, 2019	rs6933790	6p21.1	41,672,769	ASN	TFEB	31089142	(43)
Zeng et al, 2016	rs4711689	6p21.1	41,692,812	ASN	TFEB	26965516	(14)
Schmit et al, 2018	rs6906359	6p21.31	35,528,378	ASN	FKBP5	29917119	(18)
Lu et al, 2019	rs3830041	6p21.32	32,191,339	ASN	NOTCH4	30529582	(13)
Lu et al, 2019	rs6584283	10q24.2	101,290,301	ASN	NKX2-3	30529582	(13)
Lu et al, 2019	rs77969132	12p11.21	31,594,813	ASN	DENND5B	30529582	(13)
Zeng et al, 2016	rs11064437	12p13.31	6,982,162	ASN	SPSB2	26965516	(14)
Lu et al, 2019	rs2730985	12q12	43,130,624	ASN	PRICKLE1	30529582	(13)
Lu et al, 2019	rs1886450	13q22.1	73,986,628	ASN	KLF5, KLF12	30529582	(13)
Lu et al, 2019	rs4341754	16q23.2	80,039,621	ASN	WWOX, MAF	30529582	(13)
Lu et al, 2019	rs1078643	17p12	10,707,241	ASN	PIRT	30529582	(13)
Law et al, 2019	rs73975588	17p13.3	816,741	ASN	NXN	31089142	(43)
Law et al, 2019	rs9797885	19q13.2	41,873,001	ASN	TMEM91	31089142	(43)
Law et al, 2019	rs6055286	20p12.3	7,718,045	ASN	None	31089142	(43)
Jia et al, 2008	rs2423279	20p12.3	7,812,350	ASN	HAO1	23263487	(58)
Law et al, 2019	rs2179593	20q13.12	42,660,286	ASN	TOX2	31089142	(43)
Lu et al, 2019	rs13831	20q13.32	57,475,191	ASN	GNAS	30529582	(13)
Law et al, 2019	rs61776719	1p34.3	38,461,319	EUR	None	31089142	(43)
Peters et al, 2013	rs10911251	1q25.3	183,112,059	EUR	LAMC1	23266556	(59)
Whiffin et al, 2014	rs10911251	1q25.3	183,112,059	EUR	LAMC1	24737748	(60)
Houlston et al, 2010	rs6687758	1q41	222,164,948	EUR	None	20972440	(61)
Houlston et al, 2010	rs6691170	1q41	222,045,446	EUR	DUSP10	20972440	(61)
Law et al, 2019	rs11692435	2q11.2	98,275,354	EUR	ACTR1B	31089142	(43)
Law et al, 2019	rs11893063	2q33.1	199,601,925	EUR	LOC105373831	31089142	(43)
Law et al, 2019	rs7593422	2q33.1	200,131,695	EUR	None	31089142	(43)
Orlando et al, 2016	rs992157	2q35	219,154,781	EUR	TMBIM1	27005424	(62)
Law et al, 2019	rs9831861	3p21.1	53,088,285	EUR	None	31089142	(43)
Law et al, 2019	rs12635946	3q13.2	112,916,918	EUR	None	31089142	(43)
Houlston et al, 2010	rs10936599	3q26.2	169,774,313	EUR	MYNN	20972440	(61)
Schmit et al, 2018	rs1370821	4q22.2	94,943,383	EUR	None	29917119	(18)
Law et al, 2019	rs17035289	4q24	106,048,291	EUR	None	31089142	(43)
Law et al, 2019	rs75686861	4q31.21	145,621,328	EUR	HHIP	31089142	(43)
Schmit et al, 2014	rs35509282	4q32.2	163,333,405	EUR	FSTL5	25023989	(63)
Schmit et al, 2018	rs58791712	5p13.1	40,281,797	EUR	PTGER4	29917119	(18)
Schmit et al, 2018	rs2735940	5p15.33	1,296,486	EUR	TERT	29917119	(18)
Peters et al, 2012	rs2853668	5p15.33	1299910	EUR	TERT	21761138	(64)
Schmit et al, 2018	rs62404968	6p12.1	55,714,314	EUR	BMP5	29917119	(18)
Dunlop et al, 2012	rs1321311	6p21.2	36,622,900	EUR	CDKN1A	22634755	(48)
Law et al, 2019	rs9271770	6p21.32	32,594,248	EUR	LOC107987449	31089142	(43)
Law et al, 2019	rs3131043	6p21.33	30,758,466	EUR	HCG20	31089142	(43)
Law et al, 2019	rs2070699	6p24.1	12,292,772	EUR	EDN1	31089142	(43)
Law et al, 2019	rs6928864	6q21	105,966,894	EUR	None	31089142	(43)
Law et al, 2019	rs10951878	7p12.3	46,926,695	EUR	None	31089142	(43)
Law et al, 2019	rs3801081	7p12.3	47,511,161	EUR	TNS3	31089142	(43)
Tomlinson et al, 2008	rs16892766	8q23.3	117,630,683	EUR	EIF3H	18372905	(64)
Law et al, 2019	rs1412834	9p21.3	22,110,131	EUR	CDKN2B-AS1	31089142	(43)
Schmit et al, 2018	rs10994860	10q11.23	52,645,424	EUR	A1CF	29917119	(18)
Al Tassan et al, 2015	rs10904849	10p13	16,955,267	EUR	CUBN	25990418	(15)
Law et al, 2019	rs4450168	11p15.4	10,286,755	EUR	SBF2	31089142	(43)
Dunlop et al, 2012	rs3824999	11q13.4	74,345,550	EUR	POLD3	22634755	(48)
Tenesa et al, 2008	rs3802842	11q23.1	111,171,709	EUR	COLCA2	18372901	(66)
Houlston et al, 2010	rs7136702	12q13.12	50,880,216	EUR	LARP4	20972440	(61)
Law et al, 2019	rs7398375	12q13.3	57,540,848	EUR	LRP1	31089142	(43)
Schmit et al, 2018	rs72013726	12q24.21	115,890,835	EUR	MED13L	29917119	(18)
Schmit et al, 2018	rs10161980	13q13.2	34,093,518	EUR	STARD13	29917119	(18)
Law et al, 2019	rs12427600	13q13.3	37,460,648	EUR	SMAD9	31089142	(43)
Law et al, 2019	rs45597035	13q22.1	73,649,152	EUR	KLF5	31089142	(43)
Law et al, 2019	rs1330889	13q22.3	78,609,615	EUR	LINC00446	31089142	(43)
Law et al, 2019	rs7993934	13q34	111,074,915	EUR	COL4A2	31089142	(43)
Tomlinson et al, 2011	rs1957636	14q22.2	54,560,018	EUR	BMP4	21655089	(67)
Houlston et al, 2008	rs4444235	14q22.2	54,410,919	EUR	BMP4	19011631	(68)
Tomlinson et al, 2011	rs4444235	14q22.2	54,410,919	EUR	BMP4	21655089	(67)
Tomlinson et al, 2008	rs11632715	15q13.3	33,004,247	EUR	None	18372905	(65)
Tomlinson et al, 2008	rs16969681	15q13.3	32,993,111	EUR	SCG5	18372905	(65)
Tomlinson et al, 2011	rs16969681	15q13.3	32,993,111	EUR	SCG5	21655089	(67)
Tomlinson et al, 2008	rs4779584	15q13.3	32,994,756	EUR	CRAC1	18372905	(65)
Law et al, 2019	rs4776316	15q22.31	67,007,813	EUR	SMAD6	31089142	(43)
Law et al, 2019	rs10152518	15q23	68,177,162	EUR	None	31089142	(43)
Law et al, 2019	rs7495132	15q26.1	91,172,901	EUR	CRTC3	31089142	(43)
Houlston et al, 2008	rs9929218	16q22.1	68,820,946	EUR	CDH1	19011631	(68)
Law et al, 2019	rs61336918	16q23.2	80,007,266	EUR	None	31089142	(43)
Schmit et al, 2018	rs2696839	16q24.1	86,340,448	EUR	FOXF1	29917119	(18)
Broderick et al, 2007	rs4939827	18q21	46,453,463	EUR	SMAD7	17934461	(69)
Tenesa et al, 2008	rs4939827	18q21	46,453,463	EUR	SMAD7	18372901	(66)
Law et al, 2019	rs285245	19p13.11	16,420,817	EUR	None	31089142	(43)
Houlston et al, 2008	rs10411210	19q13.11	33,532,300	EUR	RHPN2	19011631	(68)
Law et al, 2019	rs12979278	19q13.33	49,218,602	EUR	MAMSTR	31089142	(43)
Tomlinson et al, 2011	rs4813802	20p12.3	6,699,595	EUR	BMP2	21655089	(67)
Peters et al, 2012	rs4813802	20p12.3	6,699,595	EUR	BMP2	21761138	(64)
Houlston et al, 2008	rs961253	20p12.3	6,404,281	EUR	BMP2	19011631	(68)
Schmit et al, 2018	rs2295444	20q11.22	33,173,883	EUR	PIGU	29917119	(18)
Schmit et al, 2018	rs1810502	20q13.13	49,057,488	EUR	PTPN1	29917119	(18)
Law et al, 2019	rs3787089	20q13.33	62,316,630	EUR	RTEL1	31089142	(42)
Houlston et al, 2010	rs4925386	20q13.33	60,921,044	EUR	LAMA5	20972440	(61)
Schumacher et al, 2015	rs8124813	3p14.1	43,476,841	EUR, ASN	LRIG1	26151821	(70)
Schumacher et al, 2015	rs35360328	3p22.1	40,924,962	EUR, ASN	CTNNB1	26151821	(70)
Lu et al, 2019	rs1476570	6p22.1	29,809,860	EUR, ASN	HLA-G	30529582	(13)
Tomlinson et al, 2008	rs2450115	8q23.3	117,624,093	EUR, ASN	EIF3H	18372905	(67)
Tomlinson et al, 2008	rs6469656	8q23.3	117,647,788	EUR, ASN	EIF3H	18372905	(67)
Haiman et al, 2007	rs6983267	8q24.21	128,413,305	EUR, ASN	POU5F1B	17618282	(71)
Tomlinson et al, 2007	rs6983267	8q24.21	128,413,305	EUR, ASN	POU5F1B	17618284	(72)
Hutter et al, 2010	rs6983267	8q24.21	128,413,305	EUR, ASN	POU5F1B	21129217	(73)
Cui et al, 2011	rs6983267	8q24.21	128,413,305	EUR, ASN	POU5F1B	21242260	(4)
Law et al, 2019	rs12255141	10q25.2	114,294,892	EUR, ASN	VTI1A	31089142	(43)
Zhang et al, 2014	rs704017	10q22.3	80,819,132	EUR, ASN	ZMIZ1	24836286	(16)
Whiffin et al, 2014	rs1035209	10q24.2	101,345,366	EUR, ASN	SLC25A28	24737748	(60)
Zeng et al, 2016	rs4919687	10q24.32	104,595,248	EUR, ASN	CYP17A1	26965516	(14)
Zhang et al, 2014	rs11196172	10q25.2	114,726,843	EUR, ASN	TCF7L2	24836286	(16)
Wang et al, 2014	rs12241008	10q25.2	114,280,702	EUR, ASN	VTI1A	25105248	(75)
Zhang et al, 2014	rs1535	11q12.2	61,597,972	EUR, ASN	FADS2	24836286	(16)
Zhang et al, 2014	rs174550	11q12.2	61,571,478	EUR, ASN	FADS1	24836286	(16)
Zhang et al, 2014	rs4246215	11q12.2	61,564,299	EUR, ASN	FEN1	24836286	(16)
Zhang et al, 2014	rs174537	11q12.2	61,552,680	EUR, ASN	MYRF	24836286	(16)
Law et al, 2019	rs10849438	12p13.31	6,412,036	EUR, ASN	None	31089142	(43)
Zhang et al, 2014	rs10849432	12p13.31	6,385,727	EUR, ASN	PLEKHG6	24836286	(16)
Peters et al, 2013	rs3217810	12p13.32	4,388,271	EUR, ASN	CCND2	23266556	(59)
Whiffin et al, 2014	rs3217810	12p13.32	4,388,271	EUR, ASN	CCND2	24737748	(60)
Jia et al, 2013	rs10774214	12p13.32	4,368,352	EUR, ASN	CCND2	23263487	(58)
Zhang et al, 2014	rs12603526	17p13.3	800,593	EUR, ASN	NXN	24836286	(16)
Zhang et al, 2014	rs7229639	18q21.1	46,450,976	EUR, ASN	SMAD7	24836286	(16)
Zhang et al, 2014	rs1800469	19q13.2	41,860,296	EUR, ASN	TMEM91	24836286	(16)
Zhang et al, 2014	rs2241714	19q13.2	41,869,392	EUR, ASN	B9D2	24836286	(16)
Schumacher et al, 2015	rs606682520	20q13.13	897,353	EUR, ASN	PREX1	26498495	(70)
Schumacher et al, 2015	rs6066825	20q13.13	47,340,117	EUR, ASN	PREX1	26151821	(70)
Dunlop et al, 2012	rs5934683	Xp22.2	9,751,474	EUR, ASN	SHROOM2	22634755	(48)

Chr, chromosome; POS, position; POP, population; AFR A, African American; ASN, Asian; EUR, European.

After merging all identified variants from GWAS (116 risk variants) with our NIPT data, we identified 106 common risk CRC variants (Table SI), while 10 risk variants that were not called in the Slovak population were excluded from further analysis. The allele frequencies of 106 variants identified in our population sample (Slovak population) and the allele frequencies of variants for 6 world populations (East Asian, South Asian, African, American, Finnish European and non-Finnish European) obtained by gnomAD database (Table SII) are shown in graphical comparison by Boxplots (Fig. 1) and principal component analysis (PCA) (Fig. 2). As shown in Fig. 1, the MAF ranged from 0.0–0.963109 in 6 world populations that is comparable with the Slovak population (0.0521–0.931). The median allele frequency for the Slovak population reached the value of 0.4072, which is closest to the value of the median of the American population (MED=0.3991) and Finnish population (MED=0.4166). PCA placed our sample set most closely to the two gnomAD population sample sets, i.e., to the Finnish and non-Finnish European population.

Figure 1.

Boxplots show allele frequency of 106 risk colorectal cancer variants identified from genome-wide association studies for the Slovak and other six world populations. AFR, African population; AMR, American population; EAS, East Asian population; FIN, European (Finnish) population; NFE, European (non-Finnish) population; SAS, South Asian population; SVK, Slovak population.

Figure 2.

Principal Component Analysis plot illustrates the allele frequency of 106 risk colorectal cancer variants identified from genome-wide association studies for the Slovak and other six world populations. PC1, Principal Component 1; PC2, Principal Component 2; AFR, African population; AMR, American population; EAS, East Asian population; FIN, European (Finnish) population; NFE, European (non-Finnish) population; SAS, South Asian population; SVK, Slovak population.

Next, we compared known allele frequencies of 106 CRC risk variants in our sample set from the Slovak population to allele frequencies of CRC variants in six world populations. The final findings of allele frequency differences are shown in Fig. 3. The median allele frequency for comparing the Slovak population and non-Finnish European population reached the value of 0.002285. Together, we identified 14 outliers in Fig. 3 (3 of 14 variants reached a similar value and overlapped, so they are not clearly visible in Fig. 3). Since the same variant rs4246215 was identified in 4 different population comparisons (Slovak-American, Slovak-Finnish European, Slovak-non-Finnish European, and also identified in Slovak-East Asian population comparison) and the rs3131043 variant in 2 population comparisons (Slovak-Finnish European and Slovak-non-Finnish European population comparison), we identified a total of 10 variants, whose difference in population allele frequency was more than 10%. Table II includes annotation information about these variants by dbSNP NCBI, ClinVar database, and population comparison in which they were identified.

Figure 3.

Boxplots show allele frequency differences of Slovak and other six world populations for 106 risk colorectal cancer variants identified from genome-wide association studies. AFR, African population; AMR, American population; EAS, East Asian population; FIN, European (Finnish) population; NFE, European (non-Finnish) population; SAS, South Asian population.

Table II.

Outliers identified in boxplots that show allele frequency differences of Slovak and the other six world populations for 106 risk colorectal cancer variants identified from genome-wide association studies.

rs_ID	Population comparison	Chr	POS	Variant type	Gene	Consequence	Clinical significance
rs5934683	Slovak-East Asian	chrX	9783434	SNV	GPR143	Intron Variant	Not Reported in ClinVar
rs7252505	Slovak-African	chr19	33084158	SNV	GPATCH1	Intron Variant	Not Reported in ClinVar
rs4779584	Slovak-East Asian	chr15	32702555	SNV	None	None	Not Reported in ClinVar
rs174550	Slovak-American	chr11	61804006	SNV	FADS1	Intron Variant	Not Reported in ClinVar
rs4246215	Slovak-American	chr11	61796827	SNV	FEN1	3′ UTR Variant	Not Reported in ClinVar
	Slovak-European (Finnish)	chr11	61796827	SNV	FEN1	3′ UTR Variant	Not Reported in ClinVar
	Slovak-European (non-Finnish)	chr11	61796827	SNV	FEN1	3′ UTR Variant	Not Reported in ClinVar
	Slovak-East Asian	chr11	61796827	SNV	FEN1	3′ UTR Variant	Not Reported in ClinVar
rs10904849	Slovak-European (non-Finnish)	chr10	16955267	SNV	CUBN	Intron Variant	Not Reported in ClinVar
rs6928864	Slovak-African	chr6	105519019	SNV	None	None	Not Reported in ClinVar
rs3131043	Slovak-European (non-Finnish)	chr6	30790689	SNV	HCG20	Intron Variant	Not Reported in ClinVar
	Slovak-European (Finnish)	chr6	30790689	SNV	HCG20	Intron Variant	Not Reported in ClinVar
rs12659017	Slovak-East Asian	chr5	126652483	SNV	None	None	Not Reported in ClinVar
rs397775554	Slovak-European (non-Finnish)	chr5	40281696-40281704	Indel	None	None	Not Reported in ClinVar

Chr, chromosome; POS, position; SNV, single nucleotide variant.

Analyses of variants located in genes associated with LS

In the analysis of LS, we identified 1212 variants in our sample set from NIPT that were located in genes known to be associated with LS, i.e., MLH1, PMS2, MSH6, TGFBR2, MLH3, MSH2 and EPCAM. After excluding variants from low complexity regions, we obtained 648 variants that were finally annotated by VEP and used for further analysis. The allele frequencies of 648 variants identified in our population sample (Slovak population) and the allele frequencies of variants for 6 world populations (East Asian, South Asian, African, American, Finnish European and non-Finnish European) obtained by gnomAD database (Table SIII) are shown in graphical comparison by Boxplots (Fig. 4) and principal component analysis (PCA) (Fig. 5). As shown in Fig. 4, the MAF ranged from 0.0–1.0 in 6 world populations. In the Slovak population, all variants were with MAF>0.05 (0.0502–1.0). The median allele frequency for the Slovak population reached the value of 0.2204, which is closest to the value of the median of the South Asian population (MED=0.221274). PCA placed our sample set most closely to the non-Finnish European population (Fig. 5).

Figure 4.

Boxplots show allele frequency of 648 variants located in seven Lynch syndrome genes for the Slovak and other six world populations. AFR, African population; AMR, American population; EAS, East Asian population; FIN, European (Finnish) population; NFE, European (non-Finnish) population; SAS, South Asian population; SVK, Slovak population.

Figure 5.

Principal Component Analysis plot illustrates the allele frequency of 648 variants located in genes associated with Lynch syndrome for the Slovak and other six world populations. PC1, Principal Component 1; PC2, Principal Component 2; AFR, African population; AMR, American population; EAS, East Asian population; FIN, European (Finnish) population; NFE, European (non-Finnish) population; SAS, South Asian population; SVK, Slovak population.

In the next step, to identify variants having significantly different frequencies, we compared known allele frequencies of 648 variants located in genes associated with LS identified in our sample set from the Slovak population to allele frequencies of these variants in six gnomAD world populations. The final findings of allele frequency differences are shown in Fig. 6. The median allele frequency for the comparison of the Slovak population and non-Finnish European population reached the value of −0.01093. By comparing the allele frequency of variants of the Slovak and non-Finnish populations, we identified a total of 64 outliers. Most outliers were found in the MSH2 gene, others in MSH6, TGFBR2, PMS2, MLH1 and EPCAM. We did not identify any outlying variant in the MLH3 gene. The variation type of all outliers was ‘intronic variant’ and the clinical significance of all outliers was not reported in ClinVar. All this annotation information, also including chromosome position, variants ID, reference and alternative allele, genes, is available in Table SIV.

Figure 6.

Boxplots show differences of Slovak and the other six world populations in allele frequency for 648 variants located in genes associated with Lynch syndrome. AFR, African population; AMR, American population; EAS, East Asian population; FIN, European (Finnish) population; NFE, European (non-Finnish) population; SAS, South Asian population.

Our analysis also included allele frequency comparison of selected variants from 648 variants identified in our sample set of NIPT data. We focused on frameshift, missense, nonsense, splice site, non-coding and UTR variants, annotated in the ClinVar database as the most common types of pathogenic and likely pathogenic variants associated with LS. However, from these selected types of variants, we found only UTR and non-coding variants in our dataset of 648 variants. Other types of variants (downstream, upstream, and intron) were excluded from further analysis. Finally, we selected 18 variants, 10 UTR and 8 non-coding variants (all selected variants with annotation information by VEP and ClinVar are available in Table III). We compared known allele frequencies of these 18 selected variants identified in our population sample (Slovak population) to the six gnomAD world populations. The final findings of allele frequency differences are shown in Fig. 7. The median allele frequency for the comparison of the Slovak population and non-Finnish European population reached the value of 0.014215, which is closest to the value of the median allele frequency comparison of the South Asian and Slovak population (MED=0.014186). By comparing the allele frequency of variants of the Slovak and six gnomAD world population, we identified a total of 4 outliers-rs10951973, rs10951972 (identified in Slovak-American population comparison), rs6791557 (in Slovak-American and Slovak-non-Finnish European population comparison) and rs9852378 (in Slovak-South Asian population comparison). All outliers were non-coding variants, rs10951973 and rs10951972 located in the PMS2 and rs6791557 located in the TGFBR2 were not reported in ClinVar. The rs9852378 SNP, detected in the MLH1, was reported as benign by ClinVar.

Table III.

Identification of 18 selected variants (UTR and non-coding) from all 648 variants in genes associated with Lynch syndrome from non-invasive prenatal testing data in the Slovak population.

rs_ID	Chr	POS	Variant type	Gene	Consequence	Clinical significance
rs11901645	chr2	47510079	SNV	MSH2	3′ UTR variant	Not reported in ClinVar
rs11891189	chr2	47510259	SNV	MSH2	3′ UTR variant	Not reported in ClinVar
rs876936	chr2	47513059	SNV	MSH2	3′ UTR variant	Not reported in ClinVar
rs72872839	chr2	47565070	SNV	MSH2	3′ UTR variant	Not reported in ClinVar
rs17036769	chr2	47633503	SNV	MSH2	3′ UTR variant	Not reported in ClinVar
rs2969774	chr2	47661684	SNV	MSH2	3′ UTR variant	Not reported in ClinVar
rs2952372	chr2	47661919	SNV	MSH2	3′ UTR variant	Not reported in ClinVar
rs2969773	chr2	47662141	SNV	MSH2	3′ UTR variant	Not reported in ClinVar
rs2705765	chr2	47662389	SNV	MSH2	3′ UTR variant	Not reported in ClinVar
rs12328344	chr2	47662542	SNV	MSH2	3′ UTR variant	Not reported in ClinVar
rs10427209	chr2	47709297	SNV	MSH6	Non-coding transcript exon variant	Not reported in ClinVar
rs10427344	chr2	47709476	SNV	MSH6	Non-coding transcript exon variant	Not reported in ClinVar
rs3136240	chr2	47784947	SNV	MSH6	Non-coding transcript exon variant	Not reported in ClinVar
rs6791557	chr3	30614676	SNV	TGFBR2	Non-coding transcript exon variant	Not reported in ClinVar
rs1817338	chr3	30631239	SNV	TGFBR2	Non-coding transcript exon variant	Not reported in ClinVar
rs9852378	chr3	36997280	SNV	MLH1	Non-coding transcript exon variant	Benign in ClinVar
rs10951972	chr7	6002187	SNV	PMS2	Non-coding transcript exon variant	Not reported in ClinVar
rs10951973	chr7	6002205	SNV	PMS2	Non-coding transcript exon variant	Not reported in ClinVar

Chr, chromosome; POS, position; SNV, single nucleotide variant.

Figure 7.

Boxplots show differences of Slovak and the other six gnomAD world populations in allele frequency for 18 selected variants (UTR and non-coding variants) from 648 identified variants located in Lynch syndrome risk genes. AFR, African population; AMR, American population; EAS, East Asian population; FIN, European (Finnish) population; NFE, European (non-Finnish) population; SAS, South Asian population.

Finally, we analyzed allele frequencies of pathogenic and likely pathogenic variants associated with Lynch syndrome annotated in the ClinVar database. From 229 SNPs with pathogenic and likely pathogenic clinical significance, only 15 have non-zero AF records in the gnomAD database. As shown in Fig. 8, all found AF are significantly below 5% (Table IV and Fig. 8).

Figure 8.

Boxplots show allele frequency of 15 single nucleotide polymorphisms of Lynch syndrome genes with pathogenic and likely pathogenic clinical significance for six world populations. It was found that allele frequency was <5%. AFR, African population; AMR, American population; EAS, East Asian population; FIN, European (Finnish) population; NFE, European (non-Finnish) population; SAS, South Asian population.

Table IV.

Identification of 15 pathogenic and likely pathogenic variants associated with Lynch syndrome with non-zero allele frequency in gnomAD database.

					Allele frequency in population
rs_ID	Chr	POS	REF Allele	ALT Allele	African	American	East Asian	European (Finnish)	European (non-Finnish)	South Asian
rs63750615	2	47403333	G	T	0	0	0	0	0	3.28×10−5
rs1194793421	2	47414417	AG	A	0	0	0	0	2.75×10−5	0
rs63750636	2	47476492	C	T	0	0	0	0	1.55×10−5	0
rs63749873	2	47795903	C	G	0	0	0	0	3.10×10−5	0
rs587783056	2	47799684	GTT	G	4.76×10−5	0	0	0	0	0
rs63751017	2	47800714	C	T	0	0	0	0	3.10×10−5	0
rs876660943	2	47806359	G	T	0	0	0	0	1.55×10−5	0
rs63751221	3	37001045	C	T	2.38×10−5	0	0	0	0	0
rs587779338	7	5977589	G	A	4.86×10−5	0	0	0	1.58×10−5	0
rs267608161	7	5982885	C	T	4.80×10−5	0	0	9.61×10−5	1.55×10−5	0
rs63751422	7	5986838	G	A	2.38×10−5	0	0	0	0	0
rs63750250	7	5986933	A	AT	0	0	0	0	9.30×10−5	0
rs200640585	7	5992018	G	A	0	0	0	0	1.55×10−5	0
rs267608154	7	5995572	ACTGT	A	0	0	0	0	1.55×10−5	0
rs63750871	7	6002590	G	A	0	7.33×10−5	0	0	0	0

Chr, chromosome; POS, position; REF, reference; ALT, alternative.

Discussion

Population genetic studies currently have a huge impact on the study of genomics (37). The detection of risk variants in a population and identifying their genetic relationships have advanced our understanding of the human genome's variability and led to the elucidation of many factors that influence cancer risk. In recent years, NGS technologies have played a key role in colorectal cancer research and have become a useful tool for cancer diagnostics and screening (38–41). Due to the high incidence of colorectal cancer in the Slovak population, it is crucial to determine the possible causes of the high incidence of this disease in Slovakia.

Non-invasive prenatal testing of common fetal chromosomal aberrations, using low-coverage massively parallel whole-genome sequencing of maternal plasma cfDNA of pregnant women, has become the fastest low-cost genomic DNA test that is rapidly implemented in clinical practice. Currently, more than 3 million NIPT tests are carried out worldwide each year, and the large amount of data generated during NIPT provides the resources to investigate human genetic variations in the population (31). In our study, we analyzed low-coverage massively parallel whole-genome sequencing data of total plasma DNA from pregnant women generated for NIPT screening to characterize the variants in genes associated with CRC and LS in the Slovak population. To our knowledge, the present study is the first population analysis of CRC and LS variants worldwide and also in the Slovak population using NIPT data. We illustrate the utility of these genomic data for clinical genetics and population studies.

Over the past two decades, GWAS offer the opportunity to uncover genetic susceptibility factors for CRC and provide insights into the biological basis of CRC etiology. These studies have demonstrated that only a fraction of CRC heritability is explained by known risk-conferring genetic variation, whereas the remaining genetic risk of CRC may be accounted for by a combination of high-prevalence and low-penetrance of common genetic variants. To date, a large number of common genetic variants have been identified by the GWAS approach, which has intimately connected to the onset of CRC (13,18,42–46).

By pooling GWAS data of risk variants associated with colorectal carcinogenesis from 2007–2020 and data variants in our population sample from NIPT, we have identified 106 common risk CRC variants. When we compared allele frequencies of these variants to allele frequencies in six gnomAD world population, finally 13 common risk variants were found that showed statistically significant differences in population allele frequencies-rs5934683, rs7252505, rs4779584, rs1535, rs174550, rs4246215, rs11196172, rs10904849, rs6928864, rs3131043, rs1476570, rs12659017, rs397775554.

The SNP rs5934683 is located on chromosome Xp22.2 between two genes, GPR143 (G protein-coupled receptor 143), which is expressed by melanocytes and retinal pigment epithelium and SHROOM2 (shroom family member 2), a human homolog of the Xenopus laevis APX gene that has important functions in cell morphogenesis including endothelial and epithelial tissue development (44). Missense mutations in this gene have been detected in large-scale screens for recurring mutations in cancer cell lines. Both GPR143 and SHROOM2 play a role in melanosome biogenesis and retinal pigmentation. It is known that abnormal retinal pigmentation, similar to the congenital hypertrophy of retinal pigment epithelium lesions, are typical of the familial adenomatous polyposis syndrome (FAP), one of the inherited syndromes of CRC (47). The relationship between Xp22.2 and CRC risk represents the first evidence for the role of X-chromosome variation in predisposition to non-sex-specific cancer (48).

The SNP rs7252505, located in the 19q13 risk locus, is in an intron of the gene GPATCH1 (G-patch domain containing 1). Although GPATCH1 is expressed in the colon, little is known about its function other than the fact that it contains a G-patch domain, a domain typically associated with RNA processing. One study found that rs7252505 was associated with CRC in African Americans (46,49).

Intergenic variant rs4779584 in chromosomal region 15q13.3 lies between SCG5 and GREM1, and the association between this SNP to CRC has been identified in several GWAS studies (13,50).

The rs4246215 polymorphism is located in the FEN1 in the long arm of chromosome 11 (11q12.2). The association between this SNP and the potential risk of different types of cancers, including esophageal, lung, gastrointestinal, gallbladder, breast cancer in Chinese and Iran populations, glioma and childhood leukemia, has been previously studied. The rs4246215 variant was also associated with colorectal cancer in East Asians and the Chinese population (51,52).

To identify variants that may predispose to LS and may cause the high incidence of CRC in Slovakia, we used NIPT data, including variants with at least 5% AF and coverage at least 100 reads per variant. To verify the reliability of the found variants using NIPT, we selected 15 variants with AF below 5% and validated them using Sanger sequencing. For this reason, it is not possible to find rare variants with AF under 5%. Initially, we selected gene variants known to be associated with LS and we focused on their population AF in gnomAD database and as well as on pathogenicity as reported in public database ClinVar. No publications are available for all variants showing statistically significant differences in population allele frequencies and selected 18 variants. The rs9852378 SNP was reported as benign by ClinVar, and other variants were not reported in ClinVar.

Our study has several key shortcomings. None of the variants identified in this study are pathogenic or likely pathogenic due to their extremely low frequency in the general population (Fig. 8). From the total number of pathogenic or likely pathogenic variants annotated in the ClinVar database, we could determine the population frequency of only 15 variants even when using the gnomAD database (Table IV). Second, the sample size was relatively small and it is strongly biased towards females. We assume that even larger sample sets will also not offer opportunities to detect such low frequencies of LS variants in the population using NIPT data. Third, a substantial portion of identified variants was removed from analyses due to technical limitations, mainly because of their location in low complexity regions. Although these could be technical artifacts (53), they could also be real variants having biological effects that are yet generally hardly determinable, but likely existing (54). Moreover, colorectal cancer is a disease caused by a combination of multiple genes and environmental factors. To assess the relationship between the variants identified in population and CRC development, it is very important in future research to study the interaction between genes and also the environment on the colorectal cancer risk. Although suitable for the determination of general population frequencies of independent variants, NIPT data are unsuitable for calculations (such as polygenic risk score determinations) based on exact combinations of these variants in individuals, which may de facto determine the risk of individuals to develop certain diseases.

The underlying mechanism for a high incidence of CRC in the Slovak population is still unclear at the moment; however, it is possible that genetic factors, like the most common inherited syndrome-LS, play a crucial role in colorectal etiology. We have performed a literature search in PubMed focused on population studies of CRC and LS in Slovakia from 2010–2020 using next-generation sequencing. In the Slovak population, only a few population studies of risk variants have been conducted to elucidate the etiology of CRC (55–57). In general, little is known about risk variants associated with CRC or LS in the Slovak population.

Identifying mutations associated with CRC in populations with high mortality rate, such as the Slovak population, is important to reduce the incidence of this multifactor disorder. The findings from these studies suggest a lack of understanding of the mechanism of many risk variants of CRC. Due to study limitations, we could not identify any pathogenic variants associated with LS in the Slovak population using NIPT data. On the other hand, NIPT data is not a major obstacle to better results, as pathogenic variants have extremely low frequencies in the general population. Even in most cases, the frequencies are not known. However, we identified several promising common risk variants associated with CRC previously reported in GWAS studies that represent variants with highly different frequencies between Slovak and non-Finnish populations in boxplots. Since NIPT expands rapidly to millions of individuals each year, the reuse of these data reduces the cost of large-scale population studies and likely provides an acceptable background for information about genomic variation. Finally, future population studies on larger sample sets with various types of mutations are needed to reveal new mechanisms of pathogenicity and links to new biological pathways, which may be useful in designing preventive strategies and treatment of CRC.

Funding Statement

The present study was supported by the PANGAIA project H2020-MSCA-RISE-2019 (grant no. 872539) funded under H2020-EU.1.3.3. Programme, the OP Integrated Infrastructure for the project ‘Long term strategic research and development focused on the occurrence of Lynch syndrome in the Slovak population and possibilities of prevention of tumors associated with this syndrome’ (grant no. 313011V578) co-financed by the European Regional Development Fund (ERDF), the Operational Program Integrated Infrastructure (grant no. 313011F988) co-financed by the ERDF, and the Scientific Grant Agency of the Ministry of Education, Science, Research and Sport of the Slovak Republic and the Slovak Academy of Sciences (grant no. 1/0305/19).

Funding

The present study was supported by the PANGAIA project H2020-MSCA-RISE-2019 (grant no. 872539) funded under H2020-EU.1.3.3. Programme, the OP Integrated Infrastructure for the project ‘Long term strategic research and development focused on the occurrence of Lynch syndrome in the Slovak population and possibilities of prevention of tumors associated with this syndrome’ (grant no. 313011V578) co-financed by the European Regional Development Fund (ERDF), the Operational Program Integrated Infrastructure (grant no. 313011F988) co-financed by the ERDF, and the Scientific Grant Agency of the Ministry of Education, Science, Research and Sport of the Slovak Republic and the Slovak Academy of Sciences (grant no. 1/0305/19).

Availability of data and materials

The datasets generated and/or analyzed during the current study are available in the DSpace repository, https://dspace.uniba.sk/xmlui/handle/123456789/27.

Authors' contributions

NF and JG performed data analysis. NF was responsible for the literature search and manuscript writing. JG, JB and JR were responsible for designing the study and supervising the work. TS conceived the idea of the project and TS, JB and JR performed proofreading of the manuscript. JG and JB confirm the authenticity of all the raw data. All authors have read and approved the final manuscript.

Ethics approval and consent to participate

The non-invasive prenatal testing study (study ID. 35900_2015) was approved by the Ethical Committee of the Bratislava Self-Governing Region (Sabinovska ul.16, 820 05 Bratislava) on 30th April 2015 (approval no. 03899_2015).

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.