Germline Alterations in Patients With IBD-associated Colorectal Cancer

Abstract

Background

Patients with inflammatory bowel diseases (IBD), both ulcerative colitis (UC) and Crohn’s disease (CD), are at risk of developing a colorectal cancer (CRC). No information is available on the contribution of patients’ genetic background to CRC occurrence. This study investigates germline alterations in patients with IBD-associated CRC.

Methods

We profiled a panel of 39 genes potentially involved in cancer predisposition and searched for germline variants in IBD patients with CRC or high-grade dysplasia.

Results

After clinical exclusion of genetic cancer syndromes, 25 IBD patients (4 CD and 21 UC) with CRC or high-grade dysplasia were studied. After excluding variants with low likelihood of pathogenicity (classes 1 or 2 according the International Agency for Research on Cancer [IARC]), the panel identified pathogenic variants, likely pathogenic, or variants with unknown significance in 18 patients (72%). Six patients (24%) carried pathogenic or likely variants (IARC class 5 or 4). Of the identified variants, 4 encompassed the APC region, 3 the MLH1 gene, and the remaining ones the MSH2, MSH3, monoallelic MUTYH, EPCAM, BRCA1, CHEK2, POLD1, POLE, CDKN2A, and PDGFRA genes. Four patients carried at least 2 variants in different genes. Duration of IBD was significantly shorter in carriers of 4 or 5 IARC variants (7 years; range 0–21; P = .002) and in those with variants with unknown significance (12 years; range 0–22; P = .005) compared with patients without or with only benign variations (23.5 years; range 15–34).

Conclusions

In silico analysis and sequence-based testing of germline DNA from IBD patients with CRC or high-grade dysplasia detected 24% of variants positioned in pathogenic classes. In patients with type 3, 4, and 5 variants, the onset of high-grade dysplasia or CRC was significantly earlier than in patients with benign or unidentified variants. The screening for these genes could identify IBD patients requiring a more intensive endoscopic surveillance for earlier detection of dysplastic changes.

Introduction

Patients with inflammatory bowel diseases (IBDs) and colonic localization are at a higher risk of developing colorectal cancer (CRC), usually referred as IBD-associated or colitis-associated CRC (CA-CRC).¹ The cumulative risk increases over time from the initial diagnosis of IBD, with an estimated value of 0.02% to 2% at 10 years, and 5% to 18% at 20 years.^2–6 Although the cumulative risk in the affected patients has declined in the last years, CRC remains the most fearsome complication of IBD, accounting for 10% to 15% of deaths.^1,7

Chronic inflammation is assumed to play a key role in the cancer pathogenesis by inducing a chronic, persistent DNA damage in the colonic mucosa.⁸ It has been hypothesized that enhanced oxidative stress damaging the genes involved in the cell cycle may be the initial step in the carcinogenetic process,⁹ leading to a specific pathway, “dysplasia-cancer.” ⁸ As matter of fact, the recently reported decline in IBD-associated CRC incidence^10,11 has been attributed to the better utilization of medical therapy and overall immunosuppressant and immunomodulant agents, which are considered to be more effective in the control of the inflammation.¹² However, even in patients with a long-standing IBD, CRC occurs only in a minority of cases. This finding could suggest the presence of a genetic component that, besides inflammation, may predispose to CRC development. The concept of genetic predisposition to CRC in IBD patients is also suggested by other 2 lines of evidence: a familial recall of CRC by some patients, and an animal model showing that after the administration of colitis-inducing agents, genetically engineered cancer-prone mice were more likely to develop CA-CRC than wild-type animals.¹³

With the intent to minimize such a risk, efforts have been directed to control inflammation and attain an early diagnosis of dysplasia or cancer by a surveillance program with scheduled endoscopic examination.^14,15 In the last 2 decades, the availability of new therapies has allowed the achievement of mucosal and histological healing, and the control of inflammation has become a major therapeutic goal.¹⁶ However, none of the available therapeutic agents is able to provide a permanent control of the inflammation. In addition, patients report reluctance to undergo repetitive invasive procedures, as current guidelines mandate¹⁴; and given the general context of cost-saving in health care, a universal surveillance for dysplasia and CRC detection may not be cost-effective due to the low incidence of CA-CRC.¹²

In the last few years, genomic-wide associations studies (GWAs), meta-analyses, and transancestry studies pinpointed more than 240 germline loci associated with IBD susceptibility,^17,18 with some such as NOD2 variants strongly associated mainly to the CD ileal location.¹⁹ At somatic level, patients with CA-CRC showed alterations in TP53, APC, IDH1, MYC, and RAS, in addition to genes involved in histone modification and chromatin remodeling.^20–23 However, some somatic variants in specimens of CA-CRC were acquired, and expression of the crowded tumor mutational burden²⁴ was analyzed.

From a clinical standpoint, it could be of utmost interest to delineate germline variants in IBD patients who eventually manifest CRC because this knowledge might help to select those patients at high risk for CA-CRC to be entered into a stringent surveillance program. To date, this information is lacking in the literature.

By using a next-generation sequencing (NGS) approach, we conducted an analysis of germline variants in patients with CA-CRC or high-grade dysplasia (CA-HGD) after a meticulously carried out exclusion of known genetic syndromes associated with CRC development.

Methods

Patients and Samples

The study was carried out at the IBD Unit of the Fondazione IRCCS Casa Sollievo della Sofferenza in San Giovanni Rotondo, Italy. A cohort of about 2000 patients with IBD is currently being followed, and the 25 patients that participated in the project were sorted out from those attending the unit. The present investigation and the experimental protocols were approved by the local Ethics Committee. All investigations were carried out in accordance with the guidelines of the Declaration of Helsinki. All participants signed the informed consent before study entry.

The following demographic and disease features were collected from all patients: age at diagnosis, type, localization and extent of the disease, extraintestinal manifestations, comorbidities, previous resective surgery, smoking habits (at least 1 cigarette/day at diagnosis), personal and family history of cancer in any site. All data were recorded and stored in an anonymized database.

Blood samples were collected in tubes containing sodium citrate and stored at −30°C until DNA extraction. Genomic DNA samples were purified from peripheral blood leukocytes using the QIAamp DNA Blood Kit following the manufacturer’s protocol (Qiagen, Hilden, Germany).

Genomic Analysis

Two authors (G.B. and O.P.) profiled a panel of 39 genes potentially involved in high or moderate penetrance hereditary predisposition to cancer (Supplementary Table 1). Target enrichment was performed using a customized panel of amplicons (Truseq Custom Amplicon; Illumina, San Diego, CA). Per the manufacturer’s protocol, a library was created and subsequently sequenced on the Illumina MiSeq (Illumina) using 300 cycle V3 kits (Genomix4life S.R.L., Baronissi, Salerno, Italy). After demultiplexing short reads per sample, raw short-reads were aligned to the hg19 human reference sequence through the BWA-mem version 0.7 aligner.²⁵ Qualimap version 2.2.1²⁶ was used to assess the quality of mapped reads. A custom Python script was used to measure the target regions’ depth of coverage. Sample-specific variants were identified by the Genome Analysis Toolkit (GATK) version 4.1,²⁷ and the variant functional effect was determined through the Annovar software package²⁸ and the NCBI Reference Sequence Database.²⁹ Variants were checked for their presence in NCBI dbSNP version 151,³⁰ ExAC version 0.3,³¹ and gnomAD version 3.1.³²

The sequencing polymerase chain reaction (PCR) products, obtained using ABI PRISM BigDye Terminator v1.1 Cycle Sequencing Kit (Applied Biosystems, Thermo Fisher Scientific, Foster City, CA), were subjected to direct sequencing using an ABI Prism 3500DX Genetic Analyzer (Applied Biosystems) and then evaluated by means of the Sequencing analysis software 5.4. Parametric and nonparametric statistical tests were used as appropriate to test the differences between patient groups to identify possible genotype-phenotype correlations. A P value of < .05 was considered statistically significant.

In Silico Prediction of Variant Pathogenicity

Complying to consensus guidelines,^33–35 we sorted all identified germline sequence variants by pathogenicity. Variants were classified according to the International Agency for Research on Cancer (IARC) classification. Briefly, class 5 corresponds to the qualitative classification of definitely pathogenic, with a probability of being pathogenic of >0.99%; class 4 is likely pathogenic, with a probability ranging between 0.95% and 0.99%. Classes 5 and 4 variants were collectively termed as pathogenic. For variants belonging to class 3, there is too little information to make any recommendation; these variants were commonly simply reported as variants with unknown significance (VUS) or “unclassified,” with a probability of being pathogenic ranging between 0.05% and 0.949%. Class 2 variants (0.1%–5% likelihood of pathogenicity) were classified as “likely not pathogenic,” whereas class 1 corresponds to the qualitative classification “not pathogenic” or “no clinical significance.” To predict the biological impact of nonsynonymous single-nucleotide polymorphisms (nsSNPs), we looked at data from the predictor tools, namely SIFT,³⁶ Polyphen2,³⁷ Mutation Taster,³⁸ Mutation Assessor,³⁹ and CADD.⁴⁰ The predicted effect of nonsynonymous and splice-site variants were retrieved from dbNSFP version 3.4.⁴¹ Additional variant annotations were obtained using information from Exome Aggregation Consortium (ExAC; http://exac.broadinstitute.org) or from curated databases for the interpretation of sequence variants, such as the Leiden Open Variation Database (LOVD; https://www.lovd.nl) and ClinVar (https://www.ncbi.nlm.nih.gov/clinvar).

Results

By searching the IBD registry at our institution, we identified 25 patients (17 male, 8 female; 5 with CD and 20 with UC) who received a diagnosis of IBD associated-CRC (20 adenocarcinoma, 4 high-grade dysplasia, 1 neuroendocrine tumor). Median age of IBD diagnosis was 44 years (range 17–68), whereas the median age of HGD/CRC diagnosis was 62 years (range 38–83). Mean duration of IBD at time of detection of high-grade dysplasia or colorectal cancer (HGD/CRC) diagnosis was 14 years (range 0–34). No patient had a previous diagnosis of primary sclerosing cholangitis. Fourteen patients (56%) reported at least 1 first-degree relative with cancer (colon, n = 6 [24 %]; breast, n = 4 [16 %]; pancreas, n = 2 [8 %]; brain, n = 2 [8%]; other cancer, n = 8 [33 %]). Patients with previous diagnosis of Lynch syndrome or fulfilling Amsterdam-II criteria were excluded. The clinical features of the patients are reported in Table 1.

Analysis of Germline Variants

After excluding the IARC variants of class 1 or 2, the multigene panel testing identified pathogenic variants, likely pathogenic variants, or VUS (IARC class 5, 4, and 3) in 18 IBD patients (72%). Six out of 25 patients (24%) carried pathogenic variants (IARC class 5 or 4). We globally found 4 variants in the APC region, 3 in the MLH1 gene, and the remaining ones in MSH2, MSH3, monoallelic MUTYH, EPCAM, BRCA1, CHEK2, POLD1, POLE, CDKN2A, and PDGFRA genes. Four patients simultaneously carried at least 2 variants: APC/BRCA1, MSH2/CHEK2/MUTYH, MSH3/BRCA1, and CDKN2A/POLE. The spectrum of these variants is shown in Table 2.

Variants in Familial Adenomatous Polyposis (FAP and MAP) Genes

We highlighted 4 variants in the APC gene, a negative regulator of Wnt signaling: 3 of them were nonsynonymous SNPs (nsSNP; c.T3920A c.C5224T, c.C5402T), and 1 was a deletion (c.5501_5506delTCAGAG). The I1307K variant has a low penetrance, and in the investigated databases, its results are as conflicting interpretation of pathogenicity, since it ranked from likely benign to pathogenic, predominating in Ashkenazi Jews.^42,43 The c.C5224T (p.R1742C) variant is not included in the LOVD database. Although the R1742C variation occurs at a highly conserved position across species, and the in silico analyses predicted that it is likely damaging protein structure and function, it is considered a VUS in ClinVar. The c.C5402T (p.A1801V) variant is also considered a VUS in ClinVar. The in-frame deletion of 6 nucleotides c.5501_5506delTCAGAG (p.V1834_R1835del) is located in exon 16 of the APC gene. This leads to deletion of 2 amino acid residues in the APC protein. It is defined as a VUS from ClinVar, owing to a nonperfect allele segregation in 1 study.⁴⁴

Two variants were identified in the MUTYH, a gene involved in the oxidative DNA damage repair, being part of the base excision repair pathway. The c.1187G>A (p.G396D) variant is one of the 2 variants frequently detected in European patients with MUTYH-associated polyposis (MAP).⁴⁵ It is labeled as pathogenic or likely pathogenic in both the ClinVar and in the LOVD databases. The c.1567C>T (p.R523C) variant has been classified as a VUS.

Variants in DNA Mismatch Repair (MMRs) Genes

We identified alterations in several genes involved in DNA-mismatch repair system. Three variants were identified in the MLH1 gene. Two of them (c.C2173G and c.G1166A) were nsSNPs, and 1 was a variation in the splicing site (c.885-3C>T). The 2 mentioned nsSNPs were positioned in class 3 or termed as VUS in the LOVD database. No evidence was available for c.885-3C>T in the LOVD and ClinVar databases, although it could affect the strength of canonical splice site.

In the MSH2 gene a single variant, the c.T435G (p.I145M), was found. Although this carrier did not recall a family history for CRC, this variant has been reported to cosegregate with CRC in 1 family, meeting the Amsterdam Criteria for the Lynch syndrome.^46,47 Both LOVD and ClinVar classify it as VUS.

In the MSH3 and EPCAM genes, 2 VUS variants were identified. Karageorgos and colleagues considered the c.T2732G (p.L911W) variant in the MSH3 gene as a putative disease-causing variant but with residual doubt as to the pathological significance⁴⁸; while the nsSNP in EPCAM (p.I277M) was identified as VUS in ClinVar.

Variants Proofreading Function of DNA Polymerases Genes

Two nsSNPs were identified in the POLD1 (c.G2275A) and POLE (c.C139T) genes. These genes are 2 components of the proofreading function of DNA polymerases, and their genetic variants might respond to checkpoint inhibitors beyond defective MMR.⁴⁹ In the ClinVar database, the role of both the c.G2275A (p.V759I), which encompasses the POLD1 gene and c.C139T (p.R47W) in POLE gene, had conflicting interpretation of pathogenicity.

Variants Involving High or Moderate Penetrance Genes for CRC

Eight nsSNPs were identified in the BRCA1, CHECK2, and PDGFRA genes. These genes are not usually considered relevant for the development of a CRC, even if their variants have been associated with an increased risk of cancers in different organs. Information for the 2 nsSNPs encompassing the BRCA1 gene, the c.C4754T (p.P1585L), and c.G1567T (p.K1636N) are not present in all the investigated databases. The variant c.3928dupA (p.T1310Nfs) may be a pathogenic one, altering the frameshift and being reported as associated to hereditary breast and ovarian cancer syndromes.⁵⁰

A further pathogenic frameshift variant, c.1100delC (p.T367Mfs), was found in the CHEK2 gene, a gene involved in Li-Fraumeni syndrome.⁵¹ In the CHEK2 gene, we identified 2 nsSNPs; the first one, the c. T470C (p.I157T) variant, has been associated with an increased risk for several cancers⁵² and classified as having a conflicting pathogenetic relevance in the ClinVar database (class 4); the second variant, c.G1312T (p.D438Y), was considered a VUS.

The CDKN2A gene is linked to malignant melanoma and pancreatic cancer⁵³ as a homozygote mutant; the variation c.C430T (p.R144C) is labeled as VUS in ClinVar. Two variants, c.A2008 (p.I670V) and c.G1285A (p.G429R), were detected in the PDGFRA gene, which is a gene connected to the gastrointestinal stromal tumor.⁵⁴ The c.A2008 (p.I670V) variant has never been described before; however, the c.G1285A (p.G429R) variant has already been reported with conflicting interpretation of its pathogenicity.

Rare Variants Encompassing NOD2 Gene

Three new rare variants, c.T743G (L248R), c.A866G (p.N289S), and c.C2050T (p.A684T), in the NOD2 gene were identified. Furthermore, the major NOD2 variants (R702W, G908R, and L1007fs) were also detected. Although in ClinVar, the new variants were described as benign; and with conflicting interpretation of their pathogenicity, they may be involved in the IBD susceptibility (Table 3).

Genotype Phenotype Analysis

To pinpoint possible phenotypic IBD characteristics among carriers of 3 (VUS), 4, and 5 IARC variants, a genotype-phenotype analysis was performed. No significant associations were found for most of the phenotypic characteristics. Only when considering the duration of IBD at the time of CRC/HGD diagnosis was a significant difference found. Carriers of 4 or 5 IARC variants (7 years; range 0–21; P = .002) and those with VUS (12 years; range 0–22; P = .005) had a shorted disease duration compared with patients without or only benign variations (23.5 years; range 15–34; Figure 1 and Table 1).

Discussion

The aim of our study was to determine the prevalence and spectrum of germline alterations occurring in patients with IBD-associated CRC. Hence, in one-fourth of patients, we found variants that currently belong to classes 4 or 5 of the IARC classification and who should therefore be considered as pathogenically or likely pathogenically disposed for CRC development. The remaining variants found in half of the patients, however, belonged to class 3. These findings together with the high prevalence of a first-degree relative with CRC (24%) or cancer at any site (56%) and the “anticipation” of HGD/CRC onset underline the relevance of a possible genetic background for CRC development in our series.

According to the Knudson’s “2-hit” mechanism,⁵⁵ we are tempted to hypothesize that in a background of germline alterations (“the first hit”), the oxidative stress secondary to the chronic inflammatory process in the colonic mucosa (“the second hit”) would promote the progression towards CRC. In keeping with this, the carcinogenetic process should not be related to environmental factors alone, as commonly hypothesized, but would also recognize a host genetic predisposition, similar to what has already been established for many hereditary cancer syndromes.³⁷

Our data showed a progressive anticipation of the cancer development from the IBD diagnosis according to different degrees of pathogenicity of the mutations, ranging from a mean of 7 years for the carriers of 4 or 5 IARC variants, 12 years for those with VUS, and 23.5 years for patients without mutations or variants of 1 or 2 IARC classes. Interestingly, not only the carriers of pathogenic or likely pathogenic mutations had a shorter time from IBD diagnosis to CRC/HGD, but also those with VUS. Because VUS are currently considered “variants not associated with enough data to meet either the pathogenic or the nonpathogenic rules,” ⁵⁶ the management of VUS carriers remains to be determined. Whether the IBD patients carrying VUS should be entered into a more strict surveillance regimen, as in the case of variants in IARC classes 4 and 5, is open to discussion. Prospective surveys with a long follow-up time of IBD patients carrying these variants should help to clarify this issue and further refine the relevance of VUS.

In addition, our analysis allowed the identification of 4 patients with more than a single pathogenic variant or VUS. In one patient carrying the pathogenic CHECK2 (p.I157T) variant and 2 VUS in the MSH2 and MUTYH genes, no family history of cancer was reported. In the other 2 patients, the co-occurrence of VUS (APC + BRCA1 and MSH3 + BRCA1) was found. Finally, a single patient tested positive for a homozygous variant in the CDKN2A and POLE genes. For the carriers of compound variants without cancer family history, we may hypothesize either an additive role of more variants in the CA-CRC occurrence or a different parental gene segregation.

The NOD2 gene variants were identified in 7 of the 25 patients with IBD-CRC (28%). The prevalence of NOD2 variants is coherent with predominance of UC and colonic localization of CD in our series. Of note, we were able to enlighten 3 rare variants of this gene not previously described in IBD patients. More likely, NOD2 variants are not associated to the risk of IBD-CRC but only to IBD genetic susceptibility.

We acknowledge some limitations of this study, such as the limited number of patients and the relatively small number of evaluated genes. In addition, the study design was to identify private, very low frequent pathogenic mutations within genes with high or moderate penetrance in cancer predisposition, but the absence of a control group or a replication cohort represents a pitfall. The number of cases investigated is coherent with a single-center series, and surely larger prospective cohort might better disclose the possible genetic component underlying in the development of IBD-associated CRC. Finally, the panel of the explored genes cannot rule out the possibility of additional causative variants in genes not investigated in this study.

If our findings were further confirmed, the detection of variants with a pathogenic significance for IBD-CRC development could allow for the identification of high-risk subjects who could be entered into an intensive surveillance regimen. The detection of these variants could also contribute to selecting patients who would be good candidates for surgery. Current recommendations of performing colonoscopy for all patients with a long-standing IBD should be revised taking into account the presence of germline pathogenic variants, limiting intensive surveillance in patients without family history or germline pathogenic variants.

In conclusion, our study provides new insights on the genetic characterization of IBD patients with CRC, shifting the focus from the disease duration to genetic predisposition as a risk factor. Multigene panel testing facilitated the identification of germline pathogenic or likely pathogenic variants. In the era of precision-medicine, identifying genetic markers associated to CRC occurrence could possibly stratify some IBD patients into high-risk subsets and modify the surveillance accordingly. The disclosure of genetic predisposition would prompt a more tailored approach, such as an intensive endoscopic surveillance with new imaging techniques such as chromoendoscopy or optical coherence tomography (OCT) or eventually suggest a prophylactic colectomy.

Author Contributions

O.P. and A.L. were grant holders (Italian Ministry of Health); G.B., O.P., G.B., F.B., A.A., and A.L. conceived and designed the study; G.B., F.B., and A.A. recruited the participants; S.T. collected, collated, and analyzed the data; T.M. supervised the data analysis; O.P. interpreted the data; R.F., A.G., T.L., G.C., M.N., and G.M. performed sample collection and experiments of sequencing; O.P. and G.B. wrote the article; A.A. performed extensive editing of the article; all authors reviewed and approved the final manuscript for submission.

Funding

The study was funded by the “Ricerca Corrente” funding program of the Italian Ministry of Health.

Conflicts of Interest

All the authors have no relevant disclosures.

References