MLH1 Founder Mutations with Moderate Penetrance in Spanish Lynch Syndrome Families

Lynch syndrome (MIM 120435) is an autosomal-dominant condition caused by germline mutations in mismatch repair (MMR) genes MLH1, MSH2, MSH6, and PMS2 (1). It is characterized by early-onset colorectal cancer (CRC) and an increased risk of other cancers (1). At age 70, cumulative CRC risk has been estimated to be 27% to 66% for men and 22% to 47% for women; risk for endometrial cancer (EC) varies between 14% and 40% (2–5).

Several founder mutations have been described in MMR genes (6). In MLH1, founder mutations were identified in several populations (6–8), where they can explain a substantial fraction of Lynch syndrome occurrences (9, 10), facilitating genetic diagnosis. In Spain, founder mutations in MMR genes have only been identified in the MSH2 gene (11, 12).

MMR founder mutations reported thus far are heterogeneous in function and frequency (6). Most of them encode truncated proteins and are readily categorized as pathogenic mutations. However, founder MMR variants with uncertain pathogenicity have also been identified (11, 13, 14). To investigate the functional effect of MMR variants, numerous assays at the RNA and protein level have been developed (15, 16). The classification of a variant as a disease-causing mutation requires integration of data from different sources, including its frequency in control populations, its cosegregation with the disease in families, its clinicopathologic features, and its properties in functional studies (17).

We have identified and characterized two frequently occurring MLH1 variants in the Spanish population. To address their clinical significance, we examined their penetrance and performed functional studies. The founder effect hypothesis was further explored by haplotype analysis.

A total of 57 families were included in this study. Twenty-nine families were carriers of c.306+5G>A or c.1865T>A MLH1 variants (17 and 12 families, respectively) and were assessed through Spanish genetic counseling and molecular diagnostic laboratories. Twenty-eight families were carriers of other pathogenic MLH1 germline mutations assessed at the Catalan Institute of Oncology (ICO). Detailed family histories from at least three generations and geographic origins were obtained. Clinical data collection included tumor location, age at diagnosis, microsatellite instability (MSI) testing, and immunohistochemistry of MMR proteins in tumors. Genealogic investigations based on interviews failed to identify relationships between individuals from different families. The study protocol was approved by the Human Research Ethics Committees of participating centers, and informed consent was obtained from all subjects evaluated.

Genomic DNA was extracted from whole blood using the FlexiGene DNA kit (Qiagen) or from formalin-fixed, paraffin-embedded tissues using the QIAamp DNA Mini kit (Qiagen). DNA samples from 325 cases and 309 controls in a hospital-based CRC case-control study conducted to assess gene-environment interactions in relation to CRC risk (18) were used.

Screening for the c.306+5G>A and c.1865T>A MLH1 mutations was performed by conformation-sensitive capillary electrophoresis (CSCE) in DNA samples from the same CRC case-control study (Supplementary Table S1; conditions available on request; ref. 18). DNAs with patterns differing from those of controls were amplified using nonfluorescent primer and sequenced with BigDye Terminator v.3.1 Cycle Sequencing kit (Applied Biosystems). The methods for analysis of loss of heterozygosity (LOH) of the variant c.1865T>A of MLH1 in tumors are included in Supplementary Methods.

The following computational methods were used to predict modifications in splice-sites or exonic splicing enhancer sites: Spliceport (19), NNSplice (20), Rescue_ESE (21), and ESEfinder (22). The Polyphen (23) algorithm was used to predict the pathogenicity of p.Leu622His variant using default settings. The Clustalw program was used to align the amino acid sequence of MLH1 in 13 phylogenetically diverse species.

Total RNA was extracted from peripheral blood lymphocytes using Trizol (Invitrogen), and cDNA was synthesized using random primers (Invitrogen) and SuperScript II reverse transcriptase (Invitrogen). Specific primers were used to amplify the appropriate MLH1 coding region (Supplementary Table S1). The methods for allele-specific expression (ASE) analysis of the variant c.1865T>A of MLH1 are descibed in Supplementary Methods.

The pcDNA3.1+ vector containing wild-type (WT) MLH1 cDNA (Genbank accession no. NM_000249.2) was kindly provided by Dr. R. Kolodner (Ludwig Institute for Cancer Research, UC San Diego School of Medicine, La Jolla, USA). The mutant p.Leu622His-MLH1 cDNA was constructed using the QuikChange Site-Directed Mutagenesis kit (Stratagene) with primer 5′-AAGAAGAAGGCTGAGATGC(A)TGCAGACTATTTCTCTTTG-3′ (sense strand; mutagenesis site is in parentheses). Sequencing was used to verify the presence of the mutation. The cDNA insert between XhoI and BamHI sites was subcloned into WT MLH1-pcDNA3.

HCT116 [deficient for endogenous MLH1 (24)] and 293 cells were obtained from the American Type Culture Collection, resuscitated from stocks frozen at low passage within 6 months of purchase, and cultured as described (25, 26). Cell lines were routinely tested by Mycoplasma presence and verified by morphology, growth curve analysis, and expression of proteins (e.g., MLH1 and PMS2). HCT116 cells were transfected with pGFP (green fluorescent protein) and WT MLH1-pcDNA3.1 or p.Leu622His-MLH1-pcDNA3.1 vector using Lipofectamine 2000 (Invitrogen) and Plus (Invitrogen). Transfection efficiency was measured by cytometry 24 hours after transfection.

Protein extraction from peripheral blood lymphocytes and HCT116 and 293 cells was performed as described elsewhere (27). MLH1 and PMS2 expression levels were examined by SDS-PAGE, followed by Western blotting analysis with anti-MLH1 and anti-PMS2 antibodies (clones G168-15 and A16-4; BD Biosciences). β-Tubulin antibody (Sigma) was used to assess equal loading in all lanes. Band intensities were quantified using Quantity One software (Bio-Rad), and expression of MLH1 and PMS2 was normalized to β-tubulin expression. The decrease of protein expression was calculated by dividing the normalized protein expression in p.Leu622His-transfected cells by the expression in WT MLH1–transfected cells. The stability of WT and variant MLH1 was assessed by treating cells with cycloheximide, a global inhibitor of de novo protein synthesis, as described (25).

We used information on the occurrence of CRC and EC in relatives of MMR mutation-positive index cases to estimate age- and gender-specific incidences of CRC and EC (females) in MMR mutation carriers by maximum likelihood using modified segregation analysis (2). The method was implemented in MENDEL (v3.3.5; refs. 28, 29).

Relatives were assumed to be followed from 20 years of age and to be censored at the age of CRC diagnosis, at the age of death, at the age of last follow-up, or at age 70 years, whichever occurred earliest. We did not ignore the cases beyond age 70 but assigned them the same risk at age 70 to avoid the larger variances observed when only sparse data are available. In estimating the risk of CRC and EC, female relatives were censored at age of CRC or EC diagnosis, whichever was diagnosed first. Information on MMR mutation status in relatives was included whenever available. For individuals with missing age information, the age was imputed based on relationship with the proband, age of the proband, and deceased status at last follow-up (dead or alive).

To correct for ascertainment, we maximized the conditional likelihood of observing the phenotypes (CRC and/or EC) and genotypes (c.306+5G>A or c.1865T>A MLH1 variants) of the entire pedigree given the phenotypic and genotypic information of the index case (proband) and phenotypic information of the pedigree to account for multiplex ascertainment bias. Age-specific cumulative risks (penetrance) and hazard ratio (HR) estimates of CRC and EC risks in families with MMR gene variants were calculated assuming a proportional hazards model, with λ(t) = λ₀(t)exp[g(t)], where λ₀(t) is the background incidence. They were compared with the risks in the general population, which were assumed to follow the population incidence from the Spanish Cancer Registries (30). The age-specific relative risks in carriers compared with the population rates are modeled through the function exp[g(t)]. We estimated the age-specific log HR parameters for the two age intervals <50 and ≥50, assuming that

To construct confidence intervals (CI) for the log(HR) estimates, we assumed that the maximum likelihood estimates of the parameters were asymptotically normally distributed with covariance matrix given by the inverse of the Fisher information matrix. Cumulative risk (e.g., penetrance) and its 95% CI were calculated from the cumulative incidence Λ(t) given by

Wald-type tests and CIs were then used based on the point estimate and the estimated SEs (28, 29).

The analysis accounted for both genotyped and ungenotyped relatives. Missing genotype information was handled by including the allele frequency as a parameter in the likelihood and then maximizing the marginal likelihood of the phenotype and genotype data of the entire pedigree, summing over all possible configurations of the unobserved genotype matrix, given the observed genotypes. Cumulative risks and HR for CRC and EC were estimated separately for males and females; risk of EC was estimated only for females.

Haplotype analysis was performed using three MLH1 single-nucleotide polymorphisms and seven microsatellite markers (Supplementary Table S1; conditions available on request). One hundred and twenty-two DNA samples from members of the studied families and 50 control individuals randomly selected from the CRC case-control study (18) were analyzed, including individuals that come from the areas of origin of the founders. To deduce the mutation-associated haplotype, intrafamilial segregation analysis was performed under the assumption that the number of crossovers between adjacent markers was minimal. The frequency of disease haplotypes in the control population was estimated using PHASE2 (default settings; ref. 31).

We used a modification of the method of Schroeder and colleagues (32) to estimate the time to the most recent common ancestor (TMRCA) separately for the c.306+5G>A and c.1865T>A alleles. This method uses a count of the number of recombination events that have occurred on copies of the ancestral mutant haplotype, together with an estimate of the recombination map length of the haplotype, to estimate the total time length of the genealogy of sampled copies of a mutant allele in units of generations. In our application of the approach (32), to estimate TMRCA from the genealogy length, we relied on a multiplier c(n), a ratio of TMRCA to tree length that was estimated from coalescent simulations with sample size n. The TMRCA estimate was then taken as an estimate of the age of the mutation, and CIs were obtained by considering the uncertainty in converting the length estimate for the genealogy into an estimate of TMRCA. Because analysis of family history revealed no relationships among sampled families in the most recent three to five generations, our age estimation was performed from a reference point four generations in the past, so that at the end of the estimation process, the estimate was increased by four generations. See Supplementary Materials and Methods for further details.

Mutational screening of MMR genes performed in suspected Lynch syndrome patients at ICO led to the identification of two frequently occurring germline MLH1 variants: c.306+5G>A and c.1865T>A. No other alterations in MLH1 coding sequence or at exon-intron boundaries were detected in carriers. Of the 83 families with germline MMR alterations identified at ICO, 45 were carriers of an MLH1 alteration (28 pathogenic mutations and 17 variants of unknown significance). The c.306+5G>A and c.1865T>A variants were identified in nine and four families, constituting 20% and 9% of all MLH1 carrier families, respectively. An extended collaborative study with other referral centers (see legend of Table 1) enabled the identification of 16 additional carrier families. In all, we identified 17 families with c.306+5G>A and 12 with c.1865T>A MLH1 variants (Table 1). Neither of these two variants was detected by CSCE in a panel of paired CRC cases and controls, reducing the probability of being polymorphisms or CRC risk alleles.

Most of the included families fulfilled the modified Amsterdam criteria (Table 1). The majority of tumors diagnosed in carriers belonged to the Lynch syndrome spectrum. The median age at diagnosis was 49.5 years (range, 20–80) for c.306+5G>A carriers and 45.5 years (range, 24–79) for c.1865T>A carriers. CRC and EC tested were microsatellite unstable (MSI+), and informative cases showed loss of expression of the MLH1 protein (Table 1).

Both c.306+5G>A and c.1865T>A MLH1 variants cosegregated with cancer in 29 and 21 affected members, respectively, and were absent in 37 and 18 unaffected members, respectively. The c.306+5G>A variant was also identified in 21 unaffected members (median age, 34.0; range, 22–63), and c.1865T>A in 21 unaffecteds (median age, 29.0; range, 18–41).

For the c.306+5G>A variant, splicing prediction programs predicted the creation of a new donor site 5 bp upstream. Reverse transcription-PCR (RT-PCR) analysis on RNA from lymphocytes of carriers confirmed the generation of this aberrant mRNA transcript, r.303_307delugagg, expected to generate a truncated protein. This transcript was associated with an increased amount of the r.208_306del alternative constitutional transcript, corresponding to the in-frame skipping of MLH1 exon 3 (Fig. 1). Although the variant c.306+5G>A is pathogenic at the RNA level, neither abnormal bands nor differences in MLH1 protein expression were observed in lymphocytes from c.306+5G>A carriers, as assessed by Western blotting (data not shown).

The MLH1 c.1865T>A variant is predicted to generate the missense change p.Leu622His, which was classified as pathogenic by the Polyphen algorithm and MAPP-MMR and SIFT predictions (33, 34). The affected amino acid residue is highly conserved and is located at the interaction domain for MutL (the homologue of MLH1 in Escherichia coli; ref. 35). Spliceport and NNSplice did not predict any effect on mRNA processing, whereas ESEfinder and Rescue_ESE predicted the creation of a new potential binding site for SRp proteins. In lymphocytes of c.1865T>A carriers, differences neither at the mRNA level [as assessed by RT-PCR (data not shown) and allele-specific expression analysis (Supplementary Fig. S1)] nor at the protein expression level (data not shown) were observed. The transient transfection of the p.Leu622His variant into the HCT116 cell line (24), which lacks endogenous MLH1, resulted in diminished MLH1 expression when compared with WT MLH1–transfected cells (Fig. 2A). MLH1 heterodimerizes with PMS2, stabilizing its expression (25). As expected, the diminished p.Leu622His expression was associated with lower PMS2 expression compared with WT MLH1–transfected cells (Fig. 2A). After cycloheximide treatment, MLH1 expression was lower in both p.Leu622His-transfected cells and WT MLH1–transfected cells. However, the effect was greater in p.Leu622His-transfected cells, indicating that p.Leu622His protein was less stable than WT MLH1 (Fig. 2B). Five of the six tumors analyzed from c.1865T>A carriers lost the MLH1 WT allele (Supplementary Fig. S1), suggesting a growth advantage associated with loss of the WT protein. Together, these lines of evidence support the notion that the variant c.1865T>A is a pathogenic mutation via a mechanism in which MLH1 protein expression is decreased.

A total of 1,936 individuals were included in the penetrance analysis, 57 of which were probands and 388 of which were first-degree relatives of probands (Table 2). Age-specific cumulative risks of CRC by decade compared with risks in Spanish Cancer Registries (30) are shown in Table 3A. For c.306+5G>A carriers, the risk of CRC by age 50 significantly exceeds the cumulative risk in the general population. Risk of CRC continues to increase, and by age 70, lifetime risk for CRC is estimated at 20.1% (95% CI, 1.4–35.9%) for men and 14.1% for women (95% CI, 1.2–25.2%). The overall HR for CRC is 7.7 (95% CI, 2.9–20.6) for males and 9.0 (95% CI, 3.6–22.6) for females (Table 3B). The age-specific HR estimates in the age intervals 20 to 49 and 50 to 69 clearly indicate much higher relative risk compared with the general population, declining in the age range from 50 to 69 years. Cumulative risk for EC also exceeds that of the general population at age 70 years, equaling 7.2% (95% CI, 0–16.9%) with HR of 6.3 (95% CI, 1.4–27.9; Table 3).

Although the cumulative risk of CRC for c.1865T>A carriers exceeds the risk in the general population by age 60 years, it is lower than for c.306+5G>A. Risk of CRC continues to increase and by age 70 is 6.8% (95% CI, 0–15.8%) for males and 7.3% (95% CI, 0–16.5%) for females. The overall HR for CRC is 2.4 (95% CI, 06–10.2) for males and 4.5 (95% CI, 1.1–18.0) for females. Risk for EC tends to exceed that of the general population at age 70 years, equaling 3.4% (95% CI, 0–10.0%) with HR of 2.9 (95% CI, 0.4–23.6; Table 3).

In carriers of other Spanish MLH1 pathogenic mutations, estimated lifetime risk for CRC is 26.3% (95% CI, 7.8–41.1%) for men and 10.7% for women (95% CI, 0–16.8%), with HR of 10.5 (95% CI, 5.0–21.8) for males and 5.4 (95% CI, 2.0–14.9) for females, and cumulative risk for EC is 6.5% (95% CI, 0–13.1%) with HR of 5.7 (95% CI, 1.9–17.0; Table 3). Therefore, both c.306+5G>A and c.1865T>A mutations show a penetrance within the range of variability estimated in families with other MLH1 mutations, with a nonsignificant trend to lower penetrance for the c.1865T>A mutation.

Ancestors of families carrying the c.306+5G>A mutation came from the Ebro river valley in northern Spain, whereas ancestors of families with the c.1865T>A mutation originated from the region of Jaén in southern Spain (Fig. 3). Common geographic origins of carrier families suggested that each of the two mutations could have occurred as a unique event in a single founder individual. This hypothesis was confirmed by haplotype analysis (Table 4).

The existence of a clear shared haplotype in carrier individuals permitted estimation of mutation age. In this estimation for the c.306+5T>A mutation, we considered disease haplotypes in the region extending from D3S2369 to D3S1298 (D3S1612 to D3S1298 for c.1865T>A; Table 4), which has a recombination map length of 0.9796 cM (2.2578 cM for c.1865T>A). In the control population, the estimated frequency of the identified minimum common haplotype, excluding the disease mutation, was 0.069 for c.306+5T>A and 0.048 for c.1865T>A (Supplementary Table S2).

To estimate the age of the mutation, we relied on an estimate of the number of recombination events that occurred on the ancestral disease haplotype for each disease mutation. At the 5′ end of the common haplotype of c.306+5G>A, a likely recombination event at marker D3S2369 was detected in family A10; at the 3′ end (D3S1298), the common haplotype was lost in 6 of the 17 families, likely in two separate recombination events (Table 4A). Therefore, for c.306+5T>A, we counted a minimum of three recombination events necessary to explain all disease haplotypes in the region. For c.1865T>A, we counted a minimum of one event, which occurred at the 3′ end (D3S1298) and which explained the loss of the common haplotype by 5 of the 12 families (Table 4B).

From the map lengths (r) and recombination counts (k), we estimated genealogy lengths (L) using the formula

We have identified and characterized two frequent MLH1 variants, c.306+5G>A and c.1865T>A (p.Leu622His), among Spanish Lynch syndrome families. They represent the first founder MLH1 mutations identified in the Spanish population.

The predicted effect of the previously reported MLH1 variant c.306+5G>A (36) on splicing was confirmed, consistent with observations on the nearby mutations c.306+2dupT and c.306+4A>G (37, 38). At the protein level, no differences in expression were observed in mutation carriers when compared with controls, presumably because of a lack of stability of mutated MLH1 proteins or due to the limitations of the Western blotting assay.

The c.1865T>A variant (p.Leu622His) was previously described as a germline MLH1 mutation in two Spanish families (39, 40). Prior functional assays for p.Leu622His were inconclusive to assess its pathogenicity: In yeast-based assays, it was deficient in MMR activity (33, 41), whereas HCT116 cells cotransfected with MLH1 and PMS2 retained partial MMR activity (69.2%). The effect on protein expression levels was inconclusive (33). In vitro expression studies in MLH1-transfected HCT116 cells indicated that this mutation may substantially reduce MLH1 expression by affecting its stability, which in turn leads to a reduction in the expression of its counterpart PMS2. A pathogenic role for the p.Leu622His mutation was further supported by the frequent loss of the MLH1 WT allele observed in tumors (42). An additional effect of the mutation on the capacity to bind PMS2 or on the intracellular trafficking of MLH1/PMS2 could not be ruled out.

Once the functional effect of the founder mutations was shown, we explored whether penetrance estimates differed from other MLH1 mutations. The estimated penetrance of Spanish MLH1 mutations was moderate and consistent with data reported for Dutch families (5). The penetrance of founder mutations was within the range of variability of penetrance observed for Spanish MLH1 mutations. The penetrance for c.1865T>A missense mutation may be lower, a fact that may be attributed to the observed decrease in MLH1 expression. The penetrance estimates are sensitive to the type of ascertainment correction, and elevate with use of an unconditional likelihood or likelihood conditioned only on the proband, a phenomenon also noted in (2). Of note, penetrance estimates for Spanish Lynch syndrome families are lower than those reported in North American populations (2, 3). The fact that we have used the same method as Stoffel and colleagues (2) and that similar cumulative risks are observed in the United States and Spain suggests that geographic differences in cancer risk might exist in Lynch syndrome carriers. Several factors including lifestyle, environmental factors, mutational spectrum, or, as recently reported, the existence of distinct weak alleles capable of producing polygenic effects (43) may account for this observation.

Founder mutations in MMR genes have an important effect in traditional “founder populations”: Four different mutations in MLH1 and MSH2 genes in Finland, Newfoundland, and in Ashkenazi Jews account for between 25% and 50% of all Lynch syndrome cases in these populations (9, 10, 13, 44). Before our report, two founder mutations in the MSH2 gene had been identified in Spanish Lynch syndrome families, although their frequencies were low in our population (11, 12). In the cohort of Lynch syndrome families from ICO, the identified c.306+5G>A and c.1865T>A mutations account for 28.8% of the 45 families carrying MLH1 alterations (13 founders and 32 nonfounders) and represent 17.6% (13 of 74) of all families with mutations in MMR genes. In addition, the c.306+5G>A founder accounts for up to 25% of MLH1 mutations identified in the Ebro basin area (data not shown).

Our detection of founder mutations in MLH1 supports the view that they can occur not only in groups commonly viewed as founder populations but also in geographically localized subsets of larger populations. These two founder mutations were not detected in a prior study of 1,222 incident cases of CRC in Spain (EPICOLON; ref. 45). This study identified 11 germline MMR mutations, 4 in MLH1. The lack of detection of founder mutations may be due to the low number of identified mutations or the relatively low coverage of the areas of origin.

Families carrying the c.306+5G>A mutation cluster within the Ebro river basin, in northern Spain, and we estimate that the mutation arose ∼1,879 years ago. Taking into account the fact that the river valley is geographically isolated by mountain ranges, and the Ebro river was navigable until the 19th century, we hypothesize that the mutation arose somewhere in the valley and was distributed along the river over the years (Fig. 3). The c.1865T>A mutation is younger (384 years) than c.306+5G>A, and its distribution is restricted to the mountainous province of Jaén. It is noteworthy that probands have been identified in Madrid and Barcelona, frequent destinations of internal migratory movements during the period 1960 to 1970. Thus, the origin of the MLH1 founder mutations can be linked to specific geographic areas, and their current distribution can be explained by migration patterns. Allele age estimation is generally imprecise (46) due to several factors, including uncertainty in the count of recombination events on ancestral haplotypes, the accuracy with which the recombination count predicts the true genealogy length, and the accuracy with which TMRCA reflects allele age. However, the relatively recent values obtained are sensible for rare mutations restricted to relatively isolated geographic areas.

The high prevalence and moderate penetrance of the c.306+5G>A and c.1865T>A variants have implications for molecular and clinical approaches to founder mutations in many settings. In specific populations such as the Spanish population, initial screening for founder mutations increases the efficiency of testing of molecular diagnostic labs. Even more importantly, careful clinical characterization of founder mutations can lead to mutation-specific counseling and improved clinical care.

No potential conflicts of interest were disclosed.

We thank the staff at the Genetic Counseling Units who have participated in this study and Michael DeGiorgio for helpful discussions about the mutation age estimates.

Grant Support: Fundació La Caixa grant BM 04-107-0; Fundació Gastroenterologia Dr. Francisco Vilardell grant F05-01; Ministerio de Educación y Ciencia grants AGL2004-07579-04, SAF 06-6084, and SAF09-7319; Spanish Networks RTICCC grants RD06/0020/1050, 1051, 0021, and 0028; Acción en Cáncer (Instituto de Salud Carlos III); Fundació Roses Contra el Càncer; Fondo de Investigación Sanitaria (CP 03-0070); and NIH grant CA81488.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.