RecQ DNA helicases germline variants in Lynch-like syndrome

Abstract

Purpose

Lynch-like syndrome (LLS) is associated with an incomplete understanding of its molecular basis. Some LLS tumors harbor somatic variants in mismatch repair genes, whereas germline variants in other DNA repair genes point to a hereditary predisposition in certain cases. Clinically, LLS patients tend to develop colorectal cancer (CRC) earlier, have a stronger family history of CRC, and respond better to immunotherapy compared with sporadic CRC cases. This study seeks to identify cancer-susceptibility genes in LLS patients, which could affect diagnosis, treatment, and surveillance.

Methods

Germline exome sequencing was performed on 102 participants, including 14 previously reported cases. In vitro models were created to study the effect of loss-of-function (LoF) variants in DNA repair genes. Gene/protein expression was evaluated using quantitative real-time PCR and Western blotting, and DNA repair capacity was assessed via H2AX phosphorylation after ionizing radiation exposure.

Results

Nine rare LoF variants were identified in DNA repair genes, including 3 in RecQ DNA helicases (WRN HGNC:12791, RECQL5 HGNC:9950). LLS patients had a higher mutational burden in DNA repair genes across various cancer phenotypes. In vitro, cells with these LoF variants showed reduced gene-protein expression and increased H2AX phosphorylation after irradiation, indicating impaired DNA repair.

Conclusion

A single LoF variant in RecQ DNA helicase genes can result in significant DNA repair deficiency, leading to genomic instability and contributing to CRC development in LLS. Therefore, we present evidence for incorporating RECQL5 into genetic testing panels for CRC risk assessment, which would enhance both diagnosis and treatment outcomes.

Introduction

Microsatellite instability (MSI) is a defining characteristic of a specific type of colorectal cancer (CRC) that occurs because of a deficiency in mismatch repair (MMR) mechanisms. In sporadic tumors, MSI arises from inactivation of the MMR gene MLH1 (HGNC:7127) through somatic methylation of its promoter region. In contrast, patients with Lynch syndrome (LS) (MIM 120435) inherit a pathogenic germline variant in an MMR gene and MSI develops in tumors after somatic inactivation of the remaining wild-type allele. Interestingly, there are individuals with suspected LS without an identifiable germline pathogenic variant in any MMR genes, leading to their classification as having Lynch-like syndrome (LLS). Tumors from LLS patients are MSI but lack a germline MMR variant or evidence of MLH1 somatic methylation. The risk of cancer in families with LLS probands is higher compared with those with sporadic CRCs and probands are typically diagnosed with CRC at a younger age.^1,2 In most cases of LLS, there is evidence of biallelic somatic inactivation of MMR genes. However, it remains unclear whether there are germline mechanisms that contribute to cancer development in these individuals.3, 4, 5, 6, 7 Our previous work identified germline pathogenic variants in non-MMR DNA repair genes, such as BARD1 (HGNC:952), WRN (HGNC:12791), MCPH1 (HGNC:6954), and REV3L (HGNC:9968) in LLS cases.⁸ Additionally, other studies have reported germline variants in MUTYH (HGNC:7527),⁹ EXO1 (HGNC:3511), POLD1 (HGNC:9175), RFC1 (HGNC:9969), RPA1 (HGNC:10289), MLH3 (HGNC:7128),¹⁰ as well as loss-of-function (LoF) variants in WRN in participants with an MMR-deficient tumor.¹¹ Furthermore, heterozygous germline variants in POLE (HGNC:9177) or POLD1 have been linked to the presence of hypermutated tumors, including MSI CRC.¹² These findings suggest that some LLS participants may harbor germline pathogenic variants in non-MMR DNA repair genes, thereby increasing their risk of developing cancer. Distinguishing between inherited and sporadic tumors is crucial for effective patient management, cascade family testing, and targeted cancer surveillance. Therefore, the study’s objective is to analyze a significant number of LLS cases to identify recurrent events in non-MMR DNA repair genes and investigate their molecular impact on tumor development. By doing so, we aim to enhance our understanding of the genetic factors that contribute to MSI in this specific cancer phenotype.

Materials and Methods

Study population

The study population consisted of 102 participants identified as having suspected LLS. These participants were tested through their original institutions, using the institution’s chosen molecular clinical testing methodologies. At the time of assessment, these participants did not carry a pathogenic variant in any MMR gene, or any pathogenic variant in a known cancer predisposition gene. Germline samples from peripheral blood or normal gastrointestinal mucosa were analyzed. Their respective institutions determined that all participants had tumors that were mismatch repair deficient, lacking MLH1 methylation (Supplemental Table1). Participants were recruited through various institutions and consortia. Additionally, 14 LLS participants previously characterized by our research group were included. Kinship analysis using PLINK confirmed no close relationships among participants.¹³

For comparative analysis, the 102 LLS cohort was evaluated alongside 3 additional cohorts. We included other types of cancer phenotypes, not LLS, to determine the frequency of germline genetic variants among a diverse population of participants with other cancer types: 39 participants with microsatellite stable hereditary nonpolyposis CRC,¹⁴ 156 participants with breast cancer and a family history of cancer without pathogenic variants in a known cancer predisposition gene, 75 participants with other cancer types without pathogenic variants in a known cancer predisposition gene. The latter 2 cohorts were evaluated at the Yale Cancer Genetics and Prevention Program. All participants were recruited under protocols approved by human research review committees, and data were shared in a deidentified format.

The study also involved analyzing The Cancer Genome Atlas (TCGA) colorectal cancer samples from the PanCancer Atlas via cBioPortal, including 535 tissue samples categorized as MSS (MANTIS ≤ 0.4) or MSI (MANTIS ≥ 0.6).¹⁵

Exome sequencing

The sequencing was performed using the Illumina HiSeq 2000 platform, achieving 40× sequencing depth with 95% coverage. Reads were aligned to the GRCh37.p13 (hg19) reference genome (Genome Reference Consortium, via UCSC Genome Browser) using BWA-MEM (v0.7.17). Duplicate marking, realignment, quality score recalibration, variant-calling, and annotation followed GATK best practices. Variant filtering criteria included a QS of >100 for SNVs and >1000 for InDels.

Germline mutational analysis

Focused on 168 DNA repair genes, this analysis only included LoF variants (excluding missense) with a population frequency ≤ 2 × 10⁻⁵ according to the ExAC database. Splicing variants were assessed using ASSP (http://wangcomputing.com/assp/) NetGene2, (https://bio.tools/netgene2) and VarSEAK (https://varseak.bio) tools. Frequency evaluation of identified variants was done using gnomAD v3.1.2 (noncancer), with visual verification through IGV_2.15.4.

RNA extraction and quantitative real-time PCR (RT-qPCR)

RNA extraction used the RNeasy Mini Kit (Qiagen). Reverse transcription of 1 (μg) RNA used iScript RT Supermix (Bio-Rad). RT-qPCR was then performed in triplicates using a Bio-Rad CFX96 with SsoFast EvaGreen Supermix (Bio-Rad) and specific PrimeTime qPCR primers from IDT (Hs.PT.58.26041956.gs and Hs.PT.58.24767201). Ct values were normalized to RNA 18S housekeeping expression after extrapolation to a standard curve.

Protein extraction and western blot

Cell lysis was done using RIPA buffer with Protease Inhibitor Cocktail (Roche), followed by sonication (Branson SFX150) and rotation at 4 (°C) for 30 minutes. Protein concentrations were determined using the BCA Protein Assay Kit (Biobasic). After normalization, samples underwent denaturing at 95 °C for 10 minutes and electrophoresis using Mini-PROTEAN TGX 4-20% Stain-Free gels (BIO-RAD), followed by transfer onto Immobilion PVDF membranes (Millipore). Western blots utilized anti-RECQL5 [NBP1-83438] (NOVUS Biologicals) and anti-WRN [NB100-471] (NOVUS Biologicals), with fluorescence-labeled secondary antibodies from LI-COR Biotechnology. Visualization was achieved with the Odyssey Infrared Imaging System (LI-COR Biosciences), and densitometry analysis was conducted using NIH ImageJ software (http://rsbweb.nih.gov/ij/). For quantification of Western blot analysis, protein levels are expressed in arbitrary units after correction by loading control.

In vitro cell systems

Lymphoblastoid cell lines (LCL) from a werner syndrome family

Generation

Two LCLs were established from a family affected by Werner syndrome. One was from an unaffected family member (WRN +/+), and the other from a heterozygote (WRN +/−) who harbored a stop-gain variant p.(Leu1246Ter).¹⁶ These cell lines were provided by Dr Ray Monnat.

Characteristics

The p.(Leu1246Ter) variant in the WRN +/− cell line is closely related (4 amino acids downstream) to the WRN variant p.(Gln1242Ter) identified in an LLS participant,⁸ suggesting similar protein effects.

Growth conditions

The WRN LCL cells were cultured in suspension in RPMI1640 medium, supplemented with 10% heat-inactivated fetal bovine serum. Cell density was maintained between 2 × 10⁻⁵ and 2 × 10⁻⁶ cells/mL before subculturing.

CRISPR-Cas engineered MCF10A cell lines

Generation

Three MCF10A cell lines were engineered using CRISPR-Cas technology by the commercial company genOway SA. After generation, the 3 cell lines were sequenced to confirm no significant off-target variants.

Characteristics

The 3 cell lines were specifically designed to study the effects of the RECQL5 c.1812+2T>C variant, identified in 2 unrelated LLS participants. These cell lines included a control (RECQL5 +/+), a heterozygous line (RECQL5 +/−), and a line with both alleles inactivated (RECQL5 −/−).

Growth conditions

The RECQL5-MCF10A cells were grown in DMEM/F12 medium, enriched with 10% horse serum, 20 (ng/mL) EGF, 10 (μg/mL) insulin, 50 (μg/mL) hydrocortisone, and 1% streptomycin-penicillin (Sigma-Aldrich). Cells were passaged upon reaching 70% confluence and incubated in a humidified atmosphere with 5% CO₂ at 37 (°C).

H2AX phosphorylation assay

Irradiation of cells

Cells, both untreated (control) and irradiated, were used. Twenty-four hours after subculture, in their exponential growth phase, cells were exposed to 5 (Gy) of ionizing radiation from an X-Rad 320 Biological Irradiator (320 (kV) and 12.5 (mA) at a dose of 2.6 (Gy/min)). Samples were collected at various time points after exposure, tailored to the DNA repair process and the specific cell type (WRN cells at 1, 4, 7, and 24 hours; RECQL5 cells at 1, 7, 24, and 48 hours).

Flow cytometry for WRN LCLs

About 10⁶ cells were harvested at each time point. Cells were stained (ViaKrome 405 Fixable Viability Dye [Beckman Coulter C36614]), fixed, and permeabilized (1% Paraformaldehyde [PFA] and 70% ethanol and stored at −20 (°C) until staining). Cells were rehydrated in PBS and then stained with 1:100 PE-conjugated anti-phosphoH2AX (Cell Signaling #5763S) and 0.5 (μM) Vybrant DyeCycle Ruby Stain (ThermoFisher V10309) for 1 hour at 37 (°C) in the dark. Stained samples were filtered and analyzed by flow cytometry with a 3-laser CytoFLEX (Beckman Coulter, Indianapolis, IN). Doublets and debris were excluded, and at least 50,000 cells were acquired for each condition. The gating strategy is shown in Supplemental Figure 1. The quantity of γH2AX molecules per cell was determined using QuantiBRITE PE beads (BD Biosciences), with a standard curve created to extrapolate and quantify the molecules.

Immunofluorescence for RECQL5-MCF10A cells

Cells were grown on coverslips (10⁵cells/well × 24 hour), washed (1× PBS), fixed (formalin × 5 min at room temperature), permeabilized (0.1% Triton X-100 in PBS × 5 min at room temperature), and washed thrice (1× PBS) and blocked (4% BSA in PBS × 10 min at room temperature). Overnight incubation (4 (°C) with the primary γH2AX antibody [BW301 05-636 Sigma-Millipore] at 1:500 in 4% BSA), washed thrice (0.1% Triton X-100 in PBS shaking). Incubation (40 min at 37 (°C) in a dark moist chamber with goat antimouse secondary antibody conjugated to the fluorophores Alexa-488 [1:1000]), washed thrice (1× PBS). Incubation (30 nM DAPI × 5 minutes), and mounted on slides with FluorSave reagent (Sigma-Millipore).

Immunofluorescence images were captured using a confocal laser microscope (40× objective, Leica). Image analysis was conducted using CellProfiler software (https://cellprofiler.org/) to determine the number of γH2AX foci per nucleus under each condition. The image analysis pipeline used was as follows: IdentifyPrimaryObjects, EnhanceOrSupressFeatures, Mask Image, IdentifyPrimaryObjects, MeasureObjectIntensity, and RelateObjects. Nuclei were identified discriminating the background by IdentifyPrimaryObjects with the Minimum Cross-Entropy method, and a diameter size of 20 to 50 pixels. Foci were identified using the Global-Manual method with a threshold of 0.035 and a diameter size of 1 to 15 pixels. The intensity of each spotlight was measured in the MeasureObjectIntensity module. Data were then exported and compared between different conditions. In each image, more than 100 cells were subjected to analysis.

The experiments were performed in triplicates, and suitable nonirradiated controls were incorporated for reference.

Statistical analysis

For comparison between categorical variables, we used the 2-sided Fisher’s Exact test. For comparison of continuous variables, we used the 2-sided Wilcox nonparametric test. All statistical comparisons were performed using R and Prism (Version 9.5.1 [528]).

Results

DNA repair germline variants in LLS

We analyzed 168 genes involved in DNA repair. Out of 102 suspected LLS participants, 7 participants (6.8%) were found to have variants in an MMR gene, with 4 of them harboring well-established pathogenic MMR gene variants. All participants had undergone prior clinical testing for MMR genes. The MMR variants identified in these 7 participants are new findings uncovered via reanalysis using our current exome sequencing pipeline. We excluded these 7 participants from further analysis (Supplemental Table 2). Among the remaining 95 LLS participants, 8 (8.4%) were found to have 9 LoF variants in non-MMR DNA repair genes. These variants included nonsense variants in FBXW7 (HGNC:16712) and RAD50 (HGNC:9816), an in-frame deletion in GEN1 (HGNC:26881), frameshift variants in RECQL5 (HGNC:9950) and RBBP8 (HGNC:9891), and splicing variants in CHEK2 (HGNC:16627), POLQ (HGNC:9186), and RECQL5 (Table 1). All these variants were rare in the general population, with a minor allele frequency of less than 7.9 × 10⁻⁵ in gnomAD (v3.1.2 noncancer) suggesting that these variants occur infrequently among the general population and may have a potential contribution to pathogenicity. To assess the tolerance level of these genes to loss-of-function variants, we determined the mutational load by counting the number of unflagged LoF variants in gnomAD (v3.1.2 noncancer) for each gene (Table 2). Interestingly, the mutational load in the LLS cohort was found to be 8 times higher compared with the general population (gnomAD v3.1.2 noncancer) (P value = 2.77 × 10⁻⁶; 9/181 vs 883/147,163), indicating an enrichment of variants in DNA repair genes in LLS, as we previously described.⁸

Table 1.

Germline loss-of-function DNA repair variants identified in Lynch-like samples

ID	Gene (HGNC)	HGVS_Genomic_GRCh37	HGVS_transcript	HGVS_protein	gnomAD AF	ACMG	ClinVar (n)	ClinVar ID
LLS-60	CHEK2 (16627)	NC_000022.10:g.29115382C>A	NM_007194.4:c.683+1G>T	NP_009125.1:p.?	Absent	PM2, PVS1	Pathogenic/Likely pathogenic (8)	187334
LLS-70	FBXW7 (16712)	NC_000004.11:g.153268144G>A	NM_001349798.2:c.664C>T	NP_001336727.1:p.(Arg222Ter)	Absent	PM2, PVS1 Likely pathogenic	__
LLS-27	GEN1 (26881)	NC_000002.11:g.17962698_17962700del	NM_001130009.3:c.2219_2221del	NP_001123481.3:p.(Gly740del)	6.8 × 10−6	PM2	Uncertain Significance (1)	1015331
LLS-60	POLQ (9186)	NC_000003.11:g.121230878T>A	NM_199420.4:c.1469-2A>T	NP_955452.3:p.?	2 × 10−5	PVS1, PM2 Likely pathogenic	__
LLS-57	RAD50 (9816)	NC_000005.9:g.131973850C>T	NM_005732.4:c.3553C>T	NP_005723.2:p.(Arg1185Ter)	3.4 × 10−5	PM2, PVS1, PP5	Pathogenic (4)	216097
LLS-99	RBBP8 (9891)	NC_000018.9:g.20596887del	NM_002894.3:c.2454del	NP_002885.1:p.(Tyr819IlefsTer82)	Absent	PVS1, PM2 Likely pathogenic	__
LLS-99	RECQL5 (9950)	NC_000017.10:g.73627744_73627759dup	NM_004259.7:c.1237_1238insTGCTCCAGGTGCCGCC	NP_004250.4:p.(His413LeufsTer60)	Absent	PVS1, PM2 Likely pathogenic	__
LLS-76 LLS-48	RECQL5 (9950)	NC_000017.10:g.73626262A>G	NM_004259.7:c.1812+2T>C	NP_004250.4:p.?	7.9 × 10−5	PVS1, PM2 Likely pathogenic	__

This table summarizes the germline LoF variants identified in 95 LLS. AF is gnomAD v3.1.2 noncancer.

ACMG, American College of Medical Genetics and Genomics; AF, allele frequency; LLS, Lynch-like syndrome; LoF, loss-of-function; (n), number of reports.

Table 2.

Allele frequency of unflagged LoF variants in DNA repair genes in LLS

Gene	LoF gnomAD	AF	P value
CHEK2	67	4.53 × 10−4
FBXW7	4	2.70 × 10−5
GEN1	72	4.87 × 10−4
POLQ	185	1.25 × 10−3
RAD50	286	1.94 × 10−3
RBBP8	42	2.84 × 10−4
RECQL5	227	1.54 × 10−3
Total	883	6.00 × 10−3	2.77 × 10−6
LLS	9	4.97 × 10−2

This table displays the AF of unflagged LoF variants in DNA repair genes identified in LLS. The AF values are derived from the gnomAD database (v3.1.2 noncancer) and indicate the prevalence of these variants in the general population. The LLS data set exhibits a significantly higher mutational load compared with the general population (P = 2.77 × 10-6; 9/181 vs 883/147,163), highlighting the enrichment of mutations in DNA repair genes in LLS.

AF, allele frequency; LLS, Lynch-like syndrome; LoF, loss-of-function.

Germline variants in RecQ helicase are enriched in the LLS phenotype

Our analysis revealed that RECQL5, a RecQ DNA helicase, was the only gene with recurring variants, comprising 3/9 (33.3%) variants identified in LLS. As described above, we previously reported 5 DNA repair germline variants in a series of 14 LLS participants. Interestingly, 1 of those LLS participants carried a nonsense variant in another RecQ helicase—WRN.⁸ After combining these datasets, we observed a significant 10-fold enrichment of LoF variants in RecQ helicase genes among participants with LLS in our cohort (4/109) compared with the expected frequency of such variants in the general population using gnomAD v3.1.2 noncancer data (P = 6.9 × 10⁻⁸). suggests a potential association between RecQ helicase variants and the LLS phenotype.

To further investigate this potential association, we compared the mutational frequency of DNA repair genes and particularly RecQ helicases of our combined LLS series with participants with other cancer phenotypes. Only LLS participants had germline variants in RecQ helicases, and they exhibited a higher mutational burden of LoF variants in DNA repair genes compared with the other cohorts analyzed (Table 3). Among the 5 RecQ helicases, we only identified pathogenic variants in RECQL5 and WRN. This result suggests that variants in RecQ helicases may contribute specifically to the development of the LLS phenotype.

Table 3.

Comparison of LoF variants in DNA Repair Genes and RECQ helicases among cancer phenotypes

Data Set	LoF Variants in DNA Repair Genes	AF	LoF Variants in RecQ Helicases	AF
LLS (n = 109)	13	0.063	4	0.018
Breast cancer (n = 156)a	5	0.016	0	0
Nonbreast tumors (n = 75)a	4	0.027	0	0
MSS HNPCC (n = 39)a	2	0.026	0	0

This table compares the LoF variants in DNA repair genes and RecQ helicases across different cancer phenotypes. It provides the number of LoF variants in DNA repair genes, AFs, and focuses on WRN and RECQL5 variants.

AF, allele frequency; HNPCC, hereditary nonpolyposis colorectal cancer; LLS, Lynch-like syndrome; LoF, loss-of-function; MSS, microsatellite stable.

^aP < .05 (Fisher’s exact test). Only variants in WRN and RECQL5 were identified among RecQ helicases.

Probands and family members with RECQL5 variants

The RECQL5 variant p.(His413LeufsTer60) was identified in a 66-year-old female proband with no family history of cancer (family A, Figure 1), who had a 40-year-old unaffected son carrying the same variant. In family B, the proband was a male participant diagnosed with CRC at age 64 years with the c.1812+2T>C RECQL5 variant. His father had CRC at 63, his sister had cervical cancer at 64, and his brother was diagnosed with CRC at 56. Among the proband’s sons, a 44-year-old son carried the RECQL5 variant, but the 48-year-old son did not. Neither had a history of cancer. In family C, which harbored the same RECQL5 variant as family B, the proband was a female diagnosed with CRC at 63 years of age. Her mother was diagnosed with CRC at 83, and her mother's brother had CRC at 80 years old. Additionally, 1 of her maternal cousins had breast cancer at 80. The proband’s brother carried the variant and was diagnosed with small bowel cancer at 80.

Figure 1.

Pedigrees of families A, B, and C illustrating the inheritance patterns of RECQL5 variants and associated cancer history. Family A: RECQL5 variant in a 66-year-old female proband with no family history of cancer. Family B: CRC-affected proband with strong cancer history. Family C: CRC-affected proband with familial cancer cases. Somatic RECQL5 variants detected in proband’s tumor samples. Cosegregation observed in family C.

Tumor testing from participants with heterozygous LoF variants in WRN and RECQL5

We had tumor tissue exome sequencing data available for the proband from family C and from the participant with a WRN variant. This sequencing effort was part of our previously published work.¹⁷

In the family C proband’s tumor, a missense RECQL5 variant, p.(Thr31Lys), was identified. Five in silico algorithms predicted this variant as likely damaging. The variant is not listed in COSMIC or cBioPortal; therefore, additional information about its pathogenicity is not available. There was no observed loss of heterozygosity in the 17q25 region, where RECQL5 is located. Regarding mismatch repair genes, this tumor showed loss of heterozygosity in MLH1 but no other somatic variant in the same gene.

Conversely, the participant with the WRN variant showed no somatic WRN variant, nor loss of heterozygosity at 8p12. This tumor showed loss of heterozygosity in MLH1 and a missense variant, p.(Thr117Arg), classified as likely pathogenic in ClinVar.

Heterozygous LoF variants in WRN and RECQL5 result in gene product downregulation

To investigate the potential impact of heterozygous LoF variants on the gene product, we performed protein extraction on cell lines carrying the WRN p.(Leu1246Ter) and RECQL5 c.1812+2T>C variants. Both WRN +/− and RECQL5 +/− cell lines showed significantly reduced WRN and RECQL5 protein levels when compared with the corresponding wild-type cell line (Figure 2A and C). We further confirmed the downregulation of RECQL5 expression through RT-qPCR and immunofluorescence (Figure 2A).

Figure 2.

Gene and protein expression analysis in RECQL5 and WRN cell lines and Family B. Heterozygous loss-of-function (LoF) variants in WRN and RECQL5 lead to downregulation of the gene product. A. MCF10A cells lines modified by CRISPR-Cas (RECQL5 +/+, +/− and −/−): their RECQL5 gene expression analyzed by qPCR, their RECQL5 protein expression measured by Western Blot and fluorescence intensity by immunofluorescence. B. Family B proband’s sons: RECQL5 gene expression levels measured by qPCR. In family B, the RECQL5 variant heterozygote displayed decreased protein expression, consistent with findings in RECQL5 +/− cell lines. C. WRN lymphoblastoid cell lines (WRN +/+ and +/−): their WRN protein expression by Western Blot. Protein analysis of WRN +/− and RECQL5 +/− cell lines confirmed approximately one-third reduction in WRN and RECQL5 protein levels, respectively. ∗P < 0.05.

We then extracted germline RNA from the 2 sons of the proband of family B. We observed a downregulation of RECQL5 RNA in the heterozygote of the variant (Figure 2B) in comparison with the individual who had wild-type RECQL5. These results are congruent with our observations in the RECQL5 +/− cell line (Figure 2A). Altogether, these results indicate that having a single deficient allele can significantly affect protein and gene expression.

Heterozygous LoF variants in RECQL5 and WRN are sufficient to result in a significant increase in DNA damage

To test our hypothesis that carrying a heterozygous variant in a RecQ DNA helicase significantly impairs DNA damage repair, we subjected cells to 5 (Gy) of ionizing radiation and measured their ability to resolve DNA damage over time.

Upon irradiation, both WRN +/+ and WRN +/− cells displayed a significant increase in γH2AX foci at 1 hour after irradiation (Figure 3). Although wild-type WRN cells demonstrated efficient DNA damage repair, evidenced by the progressively decreasing levels of γH2AX foci, WRN +/− cells maintained significantly elevated levels of γH2AX at 4 hours and 7 hours after irradiation. Notably, at 1 hour after irradiation, the wild-type cell line displayed a significantly higher number of foci than the heterozygous cell line. However, an examination of cell death revealed more than double the death in the WRN +/− cell line compared with the wild type at 1 hour after irradiation, potentially explaining the foci difference (Figure 3B). Overall, the substantial difference in cell death between WRN +/+ and WRN +/− indicates an increased susceptibility to DNA damage in the latter (Figure 3B). Altogether, these results suggest that WRN +/− cells have an impaired DNA damage repair capacity, potentially contributing to their heightened sensitivity to irradiation.

Figure 3.

WRN cells. Histone γH2AX expression in WRN+/+cells and WRN+/−. A. Schematic representation of the experimental procedures used to assess Histone γH2AX expression in WRN cells following irradiation. B. Flow cytometry plots depict γH2AX expression in relation to the cell cycle in WRN +/+ and WRN +/− cells. The DNA repair graph illustrates the mean number of γH2AX foci per cell at each time point after irradiation. The quantification of γH2AX foci per cell was determined using BD Biosciences Quantibrite Phycoerythrin (PE) Beads and the PE geometric mean for each cell population. The cell death graph represents the percentage of dead cells. ViaKrome Fixable Viability Dye (Beckman Coulter) was used for the discrimination of live and dead cells based on their fluorescence intensity. Error bars represent the standard error of the mean. N = 50,000. ∗P < 0.05; ∗∗∗P < 0.001.

In line with WRN cells, RECQL5 cells also displayed a prominent increase in γH2AX foci 1 hour after irradiation, which gradually decreased over time but remained elevated 48 hours after irradiation (Figure 4B). Notably, RECQL5 +/− cells displayed significantly higher foci numbers and percentages of foci-positive cells than RECQL5 +/+ cells 24 and 48 hours after irradiation, though these were lower than in RECQL5 −/− cells. These results suggest that having just 1 deficient allele in a RecQ helicase is enough to impair DNA damage repair. Interestingly, we also observed an elevated number of micronuclei in both RECQL5-deficient cells 48 hours after irradiation. Micronuclei formation has been associated with genomic instability and provides a plausible link to cancer predisposition, potentially contributing to genomic instability in both RECQL5+/− and −/− cells over time (Figure 4C).

Figure 4.

RECQL5 cells. Analysis of γH2AX foci and micronuclei in RECQL5+/+, RECQL5+/−and RECQL5−/−cell lines after irradiation. A. Workflow depicting the methods used to measure γH2AX foci and micronuclei in RECQL5 cells following irradiation. B. Immunofluorescence images showing γH2AX foci at 1 hour and 48 hours after irradiation in RECQL5 cell lines. Foci-positive cells, defined as cells with 10 or more foci, are indicated by purple dots. Graph illustrating the percentage of foci-positive cells over time in RECQL5 +/+, RECQL5 +/−, and RECQL5 −/− cell lines. C. Immunofluorescence images depicting the presence of micronuclei (indicated by yellow arrows) in RECQL5 cell lines 48 (hours) after irradiation. Barplot representing the percentage of cells displaying micronuclei in RECQL5 +/+, RECQL5 +/−, and RECQL5 −/− cell lines. ∗P < 0.05; ∗∗∗P < 0.001; ∗∗∗∗P < 0.0001.

Taken together, our findings provide strong molecular evidence that carrying a heterozygous pathogenic variant in a RecQ DNA helicase gene significantly impairs DNA damage repair.

Somatic RECQL5 variants are associated with the MSI phenotype

Lastly, we aimed to investigate the status of RECQL5 among MSI tumors overall, not specifically tumors from LLS participants. We performed a comparison of somatic variants in RECQL5 between MSI and MSS cases from the TCGA colorectal cohort. Our analysis revealed a significantly higher frequency of variants in RECQL5 in MSI tumors (P value = 5.15 × 10⁻⁸) (Supplemental Figure 2). It is worth noting that although MSI samples inherently exhibit a higher number of variants because of frameshift variants, the difference observed in variant frequency was primarily attributed to single nucleotide variants rather than insertions or deletions, which are not linked to the MSI phenotype (Supplemental Figure 2). Furthermore, we observed lower expression levels of RECQL5 in MSI tumors, suggesting a potential functional impact of these variants on the gene’s activity. Interestingly, the variants identified were distributed in all domains of RECQL5, indicating that they are not localized to specific microsatellite regions. This distribution suggests that these variants were present before the development of MSI and are not directly caused by subsequent microsatellite instability.

Discussion

The current understanding of the carcinogenic processes underlying the development of CRC in patients classified as LLS is evolving. Although many of these tumors exhibit somatic biallelic variants in the DNA mismatch repair genes, as documented in various studies,3, 4, 5^,7 our research and others’ findings have unveiled an association of pathogenic germline variants in non-MMR DNA repair genes,^8,10, 11, 12^,18,19 which challenges the notion of these tumors being solely sporadic in nature.

In the context of LLS patients, their cancer risk remains unclear, with studies presenting conflicting results regarding the age of CRC diagnosis and the risks of CRC and other cancers among both individuals and their first-degree relatives.^1,20, 21, 22, 23 Therefore, it is crucial to accurately identify whether a tumor arises in an inherited or sporadic context for appropriate patient management, including the implementation of family cascade testing and targeted cancer surveillance. This endeavor holds clinical significance, as the LLS phenotype could potentially be as prevalent as Lynch syndrome in the general population.²⁴

We propose that a subset of LLS tumors originates from a baseline of genomic instability triggered by pathogenic germline variants in non-MMR DNA repair genes. Over time, this initial instability can lead to the acquisition of somatic variants in CRC gatekeepers and caretakers genes, including the MMR genes. This process can cause microsatellite instability and cancer development. This initial instability appears to be a crucial factor because not all LLS tumors definitively acquire 2 somatic hits in an MMR gene, as observed in our current and previous work.¹⁷ Haploinsufficiency may contribute to increased cancer risk seen in LLS patients with these germline pathogenic variants. Haploinsufficiency occurs when a single copy of a gene is inactivated, and the remaining functional copy is insufficient to produce the necessary gene product for optimal function. Although it was traditionally believed that tumor suppressor genes require inactivation of both alleles, haploinsufficiency in cancer predisposition has gained attention. Evidence from mouse models suggests that a single functional copy of certain genes may be adequate to elevate cancer risk. For example, MLH1 +/− deficient mice exhibited increased deletions at mononucleotide repeats compared with MLH1 +/+ mice.²⁵ Similarly, heterozygosity of BRCA1 and BRCA2 (HGNC:1100, HGNC:1101) may suffice to heighten breast cancer risk in mice, even without undergoing loss of heterozygosity.²⁶

In this study, we present findings from a comprehensive analysis of a large multi-institution cohort of families including probands with an LLS classification. Remarkably, we identified a subgroup of participants with germline MMR gene variants that were previously not identified, reinforcing the utility of data reanalysis after a negative genetic test. Once again, we observed that a proportion of these cases harbor pathogenic germline variants in various DNA repair genes. Notably, among LLS participants, we identified RECQL5 as a recurrently altered gene.

Our functional investigations of RECQL5 revealed that a single heterozygous LoF variant led to a significant downregulation of gene expression. This downregulation was associated with a noticeable impairment in the DNA damage response, as evidenced by the accumulation of γH2AX foci and the formation of micronuclei after irradiation. Although tumor material was not available for all RECQL5 variant heterozygotes, the tumor available did not present with loss of heterozygosity or a truncating somatic variant. This aligns with prior evidence, described above, from mouse models showing that monoallelic loss in DNA repair genes can lead to genomic instability and increased cancer susceptibility. Altogether, these findings shed light on the role of RECQL5 in maintaining genomic stability and highlight its importance in the pathogenesis of LLS.

The RecQ DNA helicase family encompasses RECQL (HGNC:9948), BLM (HGNC:1058), WRN, RECQL4 (HGNC:9949), and RECQL5, with 4 of these helicases associated with genomic instability disorders. Biallelic pathogenic variants in BLM, WRN, RECQL4, and RECQL cause Bloom (MIM 604610), Werner (MIM 604611), Rothmund-Thomson (MIM 268400), and RECON syndrome (MIM 620370), respectively. A few case-control studies have shown an increased risk of pathogenic variants in BLM and colorectal cancer^27,28 and a few other studies between variants in the other RecQ helicases and breast cancer.29, 30, 31

Despite sharing the helicase domain, each RecQ helicase possesses unique functions that cannot be compensated for by their homologs, leading to diverse clinical manifestations. RECQL5 participates in repairing double-strand breaks through homologous recombination via interaction with RAD51 (HGNC:9817)³² and is also associated with transcription regulation through direct interaction with RNA polymerase II.³³ RECQL5 plays distinctive roles in safeguarding genomic integrity and preventing various diseases, including cancer.

Although biallelic variants in RECQL5 have not been linked to any known human recessive disease, compelling evidence connects defects in this gene to cancer predisposition. RECQL5-deficient mice develop normally but exhibit increased cancer incidence at later stages of life compared with wild-type mice.³⁴ Furthermore, RECQL5 germline variants have been associated with an OR > 2 among individuals with breast cancer.^29,31

In the context of CRC, a recent study has described germline RECQL5 missense variants among early-onset participants with multiple LS-associated cancers, including a participant with an MSI tumor.³⁵ Additionally, RECQL5 has demonstrated a tumor suppression role in the mouse gastrointestinal tract.³⁶ Based on TCGA data, RECQL5 is significantly downregulated in MSI CRC compared with MSS tumors. Similarly, a previous study characterizing the expression of RecQ helicases proteins in CRC found lower expression of RECQL5 and RECQL in MSI CRCs compared with matched normal colonic tissue.³⁷ These findings suggest a potential link between deficient RECQL5, CRC, and MSI; however, further investigation is necessary to determine whether this is a causal relationship or merely an association.

A limitation of our study is the inability to definitively prove cosegregation of germline variants in RECQL5 with cancer in multiple families. Although we observed a clear cosegregation pattern with cancer in family C, we cannot truly evaluate the cancer development in other families. This is due to the probands being younger than 50 years old and the potential for cancer to develop later. Another limitation is the use of different cell models in our study. One model is a lymphoblastic cell line, and the other is a CRISPR-based model. The differences in these cell models could potentially influence the observed results and their interpretation.

In conclusion, our findings offer valuable insights into plausible genetic predisposition among participants classified as having LLS. Our data suggest that heterozygous LoF variants in a RECQL5 and WRN genes could serve as a moderate penetrance genetic predisposition factor in CRC tumors with MSI. The evidence supporting haploinsufficiency as a mechanism of action in cancer predisposition collectively enhances our understanding of DNA repair genes as tumor suppressors. Further investigation is warranted to elucidate the precise mechanisms underlying the contribution of RECQL5 and other RecQ helicases to genomic instability and their potential implications in the development of LLS and associated cancers.

Collectively, our data provide the first evidence supporting RECQL5 as a molecular contributor to a CRC predisposition gene among individuals with MSI tumors and justify its inclusion on germline cancer risk- testing panels. Validation in independent patient cohorts will be critical to define the full cancer spectrum and penetrance associated with RECQL5 pathogenic variants. Importantly, identifying heterozygous carriers of germline RECQL5 variants with MSI tumors may reveal a subgroup with enhanced sensitivity to immunotherapy,³⁸ opening new avenues for precision treatment. Moving forward, detailed mechanistic studies of RECQL5 and other RecQ helicases across diverse MSI phenotypes will be essential to refine hereditary CRC risk models and guide both surveillance and therapeutic strategies.

Conflict of Interest

Sapna Syngal is a consultant for Myriad Genetics and DC Health Technologies and has rights to an inventor portion of the licensing revenue from PREMM5. All other authors declare no conflicts of interest.

Additional Information

Supplemental Tables S1 and S2 and Figures

F1 and F2