Novel hMLH1 and hMSH2 Germline Mutations in African Americans With Colorectal Cancer FREE

Colorectal cancer remains a serious public health challenge in the United States with more than 130,000 new cases and 55,000 associated deaths estimated for 1999.¹ This challenge appears to be much greater among African Americans for whom multiple reports in the literature document increased colorectal cancer mortality compared with whites.^2- 5

Hereditary nonpolyposis colorectal cancer (HNPCC) is an autosomal dominant syndrome characterized by early-age onset colorectal cancer and extra-colonic malignancies that include endometrial, urinary tract, and upper gastrointestinal tract carcinomas.^6- 7 To facilitate collaborative studies and genetic linkage analysis, the International Collaborative Group on HNPCC, meeting in Amsterdam in 1991, established a set of phenotypic requirements for HNPCC families.⁸ The "Amsterdam criteria" stipulate that to be designated as HNPCC, an individual's family must include 3 members diagnosed as having histologically verified colorectal cancer in 2 successive generations. One of the 3 individuals must be a first-degree relative of the other 2, and 1 of the 3 must be diagnosed prior to 50 years of age.⁸ Five human homologues of the DNA mismatch repair (MMR) genes have been cloned and germline mutations in these genes, principally hMLH1 and hMSH2, have been shown to cosegregate with the colorectal cancer phenotype in multiple HNPCC pedigrees.^9- 16 The protein products of these genes have been shown in Escherichia coli and yeast to contribute to DNA replication fidelity by the identification and correction of mis-paired nucleotides during DNA replication.¹⁷ The human homologues of these genes are believed to function similarly.¹⁸

The overwhelming majority of reports on MMR gene–associated colorectal cancer are based on highly selected HNPCC pedigrees meeting Amsterdam criteria.^19- 29 However, MMR gene alterations associated with colorectal cancer are not restricted to these highly selected families.^30- 32 The characterization of DNA MMR genes offers the prospect of identifying the specific MMR gene mutation(s) segregating with colorectal cancer in at-risk families and the delineation of family members at increased risk for the disease.²⁵ The use of colonoscopy for colorectal cancer surveillance has been reported to reduce the rate of colorectal cancer in at-risk members of HNPCC kindreds.³³ The International Collaborative Group on HNPCC and the Cancer Genetics Studies Consortium have published clinical surveillance guidelines for HNPCC families that include periodic colonoscopy.^34- 35

Although an African American kindred exhibiting the HNPCC phenotype was described in 1987,³⁶ the world literature is without comment regarding either the extent of the HNPCC phenotype or the carrier frequency of mutations in any of the known MMR genes among African Americans.

In this article we present the results of a preliminary study of the frequency of germline hMLH1 and hMSH2 mutations in 11 unrelated African Americans with histologically confirmed colorectal cancer. Analysis was confined to hMLH1 and hMSH2 for this initial study as alterations in PMS1, PMS2, and GTBP (hMSH6) have been infrequently reported among populations studied to date.^23,25 Early-age onset is associated with heritable solid tumor malignancies at multiple organ sites, including the colon and rectum, which involve multiple genetic elements.^{11- 12,37- 38} The inclusion of individuals with early-age onset without recognized prior family history in our "family" cancer registry reflects our interest in further study of the molecular genetic basis for early-age onset colorectal cancer. Mismatch repair gene–associated early-age onset colorectal cancer without prior family history has been reported,^30- 31 but remains unstudied among African American probands. Of the hMLH1 and hMSH2 mutations detected in 3 probands, herein the results pertaining to the Amsterdam criteria family member have been published³⁹; however, none of the mutations detected have been reported by other investigators.⁴⁰

Forty African Americans with histologically confirmed colorectal cancer enrolled in the Roswell Park Family Cancer Registry⁴¹ were selected for study on the basis of family and clinical history alone. Eleven individuals, 8 women and 3 men, were available for and consented to study participation. Among the 29 nonparticipants, 9 had died since their diagnosis, 10 were no longer accessible by telephone or US mail, and 10 declined to participate when contacted. The age and sex distribution, family history, and stage at diagnosis characteristics of the nonparticipants were similar to those of participants. One study proband was a member of an Amsterdam criteria HNPCC pedigree.³⁹ The remaining 10 study participants were diagnosed prior to 50 years of age and had no prior family history of the disease. An additional 50 age (within 2-year age range) and sex-matched African American adults (composed of individuals who were hospital employees and members of the community) without personal or known family history of colorectal, endometrial, ovarian, urinary tract, or small bowel cancer were included as a polymorphism control population. Seventy-five of 102 individuals initially agreed to participate in the control group; however, 25 were excluded because of positive personal or family cancer history. The additional 27 individuals declined to participate at the time of initial invitation. Twenty control subjects received a modest financial incentive for participation. The remaining 30 participants were unpaid volunteers.

After obtaining institutional review board approval and subject informed consent, venous blood samples were obtained from each study subject. The Roswell Park Cancer Institute Division of Surgical Oncology employs 2 full-time genetic counselors who are available to all study subjects.

Genomic DNA isolation was performed using standard organic extraction methods. Polymerase chain reaction fragment generation, single-strand conformation polymorphism analysis, and direct nucleotide sequence analysis were performed as described previously.³⁹ All exons of each gene were analyzed for sequence variation in all study subjects. The Fisher exact test was used for comparisons.

Of the 11 probands studied, a total of 9 hMLH1 and hMSH2 alterations were detected (Table 1 and Table 2). These included 3 mutations detected in 3 unrelated study probands (Table 1). One carrier proband (166) was a member of an Amsterdam criteria HNPCC family and the remaining 2 mutation carriers were diagnosed prior to 50 years of age without prior family history. Each of the carriers were women. The remaining 6 alterations are presumed to be polymorphisms (Table 2). Proband 166 was diagnosed as having metachronous rectal and colon cancers at ages 47 and 49 years, respectively. This proband carried an A→T transversion resulting in a missense mutation characterized by a substitution of phenylalanine for isoleucine at codon 26 of the hMLH1 gene. Proband 404 was diagnosed as having metachronous uterine and rectal cancers at ages 35 and 38 years, respectively. This proband carried a GG→AT substitution across codons 177 and 178. This results in a change from glutamine to a Stop signal at codon 178. Proband 449 was diagnosed as having colon cancer at 44 years of age and carried the only hMSH2 mutation detected, a missense mutation characterized by a C→T transition at codon 389 resulting in the substitution of phenylalanine for leucine. The mutations seen in hMLH1 codon 26 and codons 177 and 178, and hMSH2 codon 389, were not detected among the 50 controls studied.

Probands 166, 410, and 460 carried an identical C→T transition at codon 718 of the hMLH1 gene resulting in the substitution of a polar-uncharged tyrosine for a positively charged histidine. Clinical characteristics of proband 166 were presented above. Probands 410 and 460 were diagnosed as having colon cancer at ages 42 and 43 years, respectively.

One of the 50 control subjects carried the hMLH1 codon 718 C→T transition. The control subject, a 40-year-old woman, reported that her father died of an uncertain malignant disease at 40 years of age and her father's mother died of a presumed abdominal malignancy at 45 years of age.

Ethnic variations in both the prevalence of the HNPCC phenotype and carrier frequencies of germline hMLH1 and hMSH2 mutations in selected HNPCC families have been extensively described^{9- 10,20- 21,23- 24,28- 29,42- 46}; however, the world literature is without comment on these points in regard to the African American population. This article presents the results of a preliminary study of germline hMLH1 and hMSH2 mutations in African American patients with colorectal cancer that includes a member of an Amsterdam criteria HNPCC family and 10 unrelated individuals diagnosed prior to age 50 years without known prior family history.

Three of the 11 unrelated probands tested carried 3 MMR gene mutations. Two of these alterations were detected in hMLH1, 1 in hMSH2. Two of the 3 mutations detected are missense in character; in previous reports, approximately 50% of hMLH1 and 25% of hMSH2 mutations associated with HNPCC were missense alterations.^{9- 16,19- 29,42- 46} Multiple authors have acknowledged the role low-penetrance mutations may play in colorectal cancer predisposition,^30- 31,47 and MMR gene–associated colorectal cancer in early-age onset individuals without prior family history has been documented.^30- 31 Significantly, in these reports,^30- 31 the MMR gene alterations observed are missense in character. Regarding the substitution of phenylalanine for isoleucine at hMLH1 codon 26 in proband 166, although both isoleucine and phenylalanine are nonpolar amino acids, the bulky ring structure of the phenylalanine is likely to contribute to altered protein structure and function. In addition, we have published evidence for microsatellite instability in the colorectal tumors of this patient, supporting the pathologic significance of the hMLH1 alteration detected.⁴⁸ Tissue from the rectal cancer of proband 404 also demonstrated microsatellite instability.⁴⁸ Tumor tissue from proband 449 was not available for study. None of the missense mutations reported in this study (Table 1) have been detected in prior polymorphism studies,³⁹ and they are not recorded in the Mutation/Polymorphism Data Base of the International Collaborative Group on HNPCC.⁴⁰ The results of our study support an association between each of the missense mutations reported and colorectal cancer predisposition.

Probands 404, 410, and 460 carried an identical exon 19 alteration at codon 718 (Table 2). This codon 718 alteration results in the substitution of a polar-uncharged tyrosine for a positively charged histidine. Although the number of hMLH1 exon 19 mutations reported is second only to the number reported for exon 16,³⁹ the codon 718 alterations detected in this study have not been reported.⁴⁰ We note with interest that the charge-altering substitution at codon 718 occurs at a highly conserved codon that is located within the PMS1 interactive domain of the hMLH1 gene.⁴⁹ Fisher exact test comparative analysis of the frequency of the codon 718 alteration in cases (3/11) vs controls (1/50) yielded a P value of .02. However, these data are not sufficient to interpret the 718 alteration as a disease-causing mutation. In a worst-case scenario of increased initial participation among our 40 cases with the finding of 3 carriers of the 718 alteration, the comparison of 3 of 40 cases vs 1 of 50 controls using the Fisher exact test results in a P value of .32, which is not significant. Because of participation rates, we cannot rule out the possibility of potential bias due to lack of generalizability of the subgroup studied relative to the initial sample. We strongly agree with White⁵⁰ that large, well-controlled studies will be required to determine the true pathologic significance of apparently variably penetrant alterations such as the adenomatous polyposis coli gene I1307K⁵¹ and possibly the hMLH1 codon 718 alteration we report in this study.

For affected individuals without prior history, explanations for the presence of disease-associated mutations include de novo mutations as well as variable penetrance, nonpaternity, and inadequate family history. Parents of the 3 carriers reported were not available for study, precluding evaluation of de novo mutation and variable penetrance. With regard to mutation detection methods used in this study, although single-strand conformation polymorphism analysis sensitivity is limited to approximately 80% to 85%, it enjoys wide acceptance in screening for MMR gene alterations^52- 54 and has been successfully used in larger population-based mutation detection studies.⁵⁵ Although efficient at detecting Stop codon alterations, protein truncation–based analysis will not detect missense alterations and has been reported to result in false-positive detection of deletion variants and alternative transcript results of questionable clinical significance.^26,56

African Americans diagnosed as having colorectal cancer carry an increased risk of dying of the disease compared with whites.^2- 5 We do not propose that MMR gene–associated colorectal cancer is necessarily the explanation for the mortality rates experienced by African American patients with colorectal cancer. More than half of the excess colorectal cancer mortality described for African Americans is attributable to late-stage diagnosis.⁴ It has been suggested that MMR gene–related colorectal cancer is associated with improved survival^57- 58; however, other reports differ on this point.^10,59 An initial study linking HNPCC to chromosome 3p included a family in which the disease demonstrated rapid progression and poor prognosis; all patients died within 2 years of diagnosis.¹⁰

While expectations are high that genetic predisposition testing combined with clinical screening of proven efficacy may reduce overall colorectal cancer morbidity and mortality rates,²⁵ the segment of the population apparently at highest risk of fatal colorectal cancer^4- 5 has not been included in studies conducted to determine the prevalence of the HNPCC phenotype or the carrier frequency of MMR gene mutations associated with colorectal cancer.^{9- 16,19- 29,42- 46} The results of this study indicate that MMR gene alterations may play a role for both African American Amsterdam criteria HNPCC families and patients with early-age onset colorectal cancer without prior family history. This latter group is of particular clinical relevance as these patients are unlikely to appreciate the potential increased risk of subsequent colorectal cancer for themselves and their parents, siblings, and children.⁶⁰ In addition, the characteristics of the mutations we have reported suggest their codon distribution is novel when compared with that of other ethnic and racial groups.^{9- 16,19- 29,42- 46} These results are reminiscent of recent findings of ethnic variability involving increased frequency among Ashkenazi Jews of specific mutations associated with the BRCA1 and BRCA2 genes⁶¹ and the adenomatous polyposis coli gene.⁵¹

In summary, these preliminary results suggest that further study of the contribution of germline mutations in the MMR genes to the colorectal cancer phenotype in African Americans would enhance our understanding of this disease and facilitate the directing of clinical surveillance to those at high risk. The observation that as much as 50% of the increased mortality sustained by this segment of the population is attributable to a higher frequency of late-stage diagnosis compared with the white population⁴ underscores the potential benefit that could be derived from increased clinical surveillance of members of this high-risk group. Current studies in progress in our laboratory include an enlarged survey of MMR gene–associated colorectal cancer among African Americans with the inclusion of MSH6, PMS1, and PMS2 in our proband analysis. We conclude that progress in the understanding of the molecular genetics of MMR gene–associated colorectal cancer will lead to an important contribution to cancer control in this understudied, high-risk patient population.