NOS indicates not otherwise specified.
Letters indicate branch of family G.
Douglas JA, Gruber SB, Meister KA, Bonner J, Watson P, Krush AJ, Lynch HT. History and Molecular Genetics of Lynch Syndrome in Family GA Century Later. JAMA. 2005;294(17):2195-2202. doi:10.1001/jama.294.17.2195
Author Affiliations: Departments of Human Genetics (Drs Douglas and Gruber), Epidemiology (Dr Gruber), and Internal Medicine (Dr Gruber, Ms Meister, and Mr Bonner), University of Michigan Medical School, Ann Arbor; Department of Preventive Medicine and Public Health, Creighton University School of Medicine, Omaha, Neb (Drs Lynch and Watson); and Department of Internal Medicine and Human Genetics, Johns Hopkins University School of Medicine, Baltimore, Md (Ms Krush).
Context In 1895, Aldred Scott Warthin, MD, PhD, initiated one of the most thoroughly documented and longest cancer family histories ever recorded. The unusually high incidence and segregation of cancers of the colon, rectum, stomach, and endometrium in Dr Warthin’s family G was later followed up by his colleagues, most recently by Henry Lynch, MD. Described today as a Lynch syndrome family, family G was last documented in 1971, prior to the modern era of molecular diagnostics.
Objective To update family G.
Design, Setting, and Participants Historical prospective cohort study of family G members from 1895 to 2000.
Main Outcome Measures The primary outcomes were the frequencies and types of cancers, ages at diagnosis, and presence of the T to G transversion at the splice acceptor site of exon 4 of the mutS homolog 2, colon cancer, nonpolyposis type 1 (E coli ) (MSH2) gene in family G members. A secondary analysis compared cancer-specific incidence rates in family G with published national and regional cancer incidence rates through the standardized incidence ratio (SIR).
Results Family G now has 929 known descendants of the original progenitor first reported in 1913. Cancers of the colon and rectum (SIR, 3.20; 95% confidence interval [CI], 2.39-4.19) and endometrium (SIR, 3.51; 95% CI, 1.92-5.89) continue to predominate in family G. Five of 40 tested members of family G carry the MSH2 T to G mutation; as a result, 15 of their living relatives are at increased risk of developing 1 or more colorectal or Lynch syndrome–associated cancers. In contrast, 97 living members of family G can now be excluded as mutation carriers.
Conclusion Within the last decade, molecular diagnostic testing has transformed the care of family G and other Lynch syndrome families in which a pathogenic mutation has been identified.
The observation of a positive family history of cancer is a surrogate for genetic susceptibility in many epidemiological studies and is often a strong indicator for personal cancer risk. The first documentation of such a hereditary cancer family was reported nearly 100 years ago1 by Aldred Scott Warthin, MD, PhD (Figure 1) and remains one of the longest family cancer histories ever recorded. This family, referred to as family G, was originally ascertained in the medical record review of family histories of cancer patients treated at the University of Michigan hospitals between 1895 and 1913 and was at the time characterized by a distinct susceptibility to cancer of the uterus and stomach. Family G, updated by Warthin2 in 1925, Weller and Hauser3 in 1936, and Lynch4 in 1971, was subsequently described as a Lynch syndrome family.
Lynch syndrome, previously called hereditary nonpolyposis colorectal cancer, is characterized by an autosomal dominant mode of inheritance of colorectal cancer and is the most common form of hereditary colon cancer, comprising approximately 5% of all colorectal cancers.5 Primary clinical characteristics include early onset of colorectal cancer (average age at diagnosis is 45 years) and an excess of extracolonic cancers (or Lynch syndrome–associated cancers), including carcinoma of the endometrium (the second most common site), stomach, ovary, small bowel, pancreas, brain, hepatobiliary tract, and upper uroepithelial tract and frequent multiple primary cancers.6,7 Various clinical criteria based on a family history of colorectal and extracolonic cancers have been used for research purposes to identify Lynch syndrome families with mixed success, including the Amsterdam criteria.8,9 With the relatively recent identification of Lynch syndrome susceptibility genes, it is now possible to directly assay susceptibility for individuals and examine family history as it relates to the presence of specific mutations.
Lynch syndrome is associated with germline mutations in mismatch repair genes on chromosomes 2, 3, and 7, namely mutL homolog 1, colon cancer, nonpolyposis type 2 (Escherichia coli) (MLH1),10,11 mutS homolog 2, colon cancer, nonpolyposis 1 (E coli) (MSH2 ),12,13 and mutS homolog 6 (E coli) (MSH6),14,15 and less commonly postmeiotic segregation increased 2 (Saccharomyces cerevisiae) (PMS2).16 In 2000, Yan and colleagues17 identified a T to G transversion at the splice acceptor site of exon 4 in the MSH2 gene in one member of family G. Subsequent testing of other members of family G for the MSH2 T to G mutation is now possible and provides an opportunity to improve risk assessment and clinical management of this large historically high-risk cancer family. The history of family G highlights the change in clinical management of a single family over a century due to recent advances in molecular genetic testing and diagnostics. Our purpose herein is to update, document, and summarize the history and genetics of Lynch syndrome in family G.
A total of 929 members of family G in 7 generations are included in this study. All are descendants of a man who was born in Plattenhardt, Germany, in 1796 and subsequently immigrated to the United States in 1831. This progenitor, who was diagnosed with cancer (site unknown), had 9 children and thus began family G. Historically, all 9 children and their descendants were recorded as separate family branches, labeled A through I (Figure 2). For the purpose of this study, only biological descendants of the original male progenitor are included. Members of family G were excluded from certain analyses if essential data, such as sex, date of birth, and date of death were missing. Institutional review boards at the University of Michigan (Ann Arbor) and Creighton University (Omaha, Neb) approved the genealogical and molecular studies, and all living family G members participating in the study gave written, informed consent.
Cancer diagnoses were verified through medical records and all available pathology reports dating from 1895 to 2000 for family G members who received treatment at the University of Michigan Hospitals and St Joseph’s Hospital in Ann Arbor, Mich. Family members living outside the Ann Arbor area were contacted through the other surviving family G members and the Cancer Genetics Registry at Creighton University. Overall, cancer diagnoses were confirmed by a pathology report in 33% of cases and in 60% of cases in more recent generations. Additional sources of information included autopsies, death certificates, medical or family reports, and physical examinations. All skin cancers and benign or in situ lesions were excluded.
A research-oriented reunion for family G members was held at the University of Michigan in March 2000 to offer genetic testing and update family information. Twenty-seven family members attended this reunion, and 17 of them participated by signing the written, informed consent document. A board-certified genetic counselor disclosed the genetic test results to family members and recommended that the results be confirmed in a Clinical Laboratory Improvement Amendment (CLIA)—certified laboratory and shared with their personal physicians and other family members.
Mutation detection was performed by allele-specific oligonucleotide hybridization. Genomic DNA was extracted from blood using a PUREGENE Kit (Gentra Systems, Minneapolis, Minn) according to the manufacturer’s instructions. DNA was quantified by spectrophotometry and diluted in LoTE buffer. An amplicon spanning the known familial mutation was generated by polymerase chain reaction (PCR) using the following primers and reaction conditions: MSH2-4F: TGT AAA ACG ACG GCC AGT TTT TTG CTT TTC TTA TTC CTT TTC and MSH2-4R: CAG GAA ACA GCT ATG ACC TGA CAG AAA TAT CCT TCT AA, at 95°C for 5 minutes, then 35 cycles of 52°C for 1 minute, 72°C for 1 minute, then 72°C for 10 minutes. Allele-specific oligonucleotide hybridization was performed as previously described18 with the following modifications: prehybridization and hybridization were both performed at 38°C for 1 hour, wild-type and mutant labeled probes were MSH2-Ex4 WT: TTT CAA AAT AGA TAA TT and MSH2-Ex4 MU: TTT CAA AAG AGA TAA TT, and membranes were washed at 37°C for 20 minutes. Membranes were exposed to film and were then scored independently by 2 different laboratory technicians. All positive samples and a limited number of negative samples were confirmed by sequencing with 100% concordance.
The primary outcomes of this study were the frequencies and types of cancers, ages at diagnosis, and the MSH2 mutational status of family G members. For each branch of family G, cumulative cancer incidence curves were calculated according to the method of Kaplan and Meier and compared using the log-rank test. All family members were included from birth until their first diagnosis of a colorectal or Lynch syndrome–associated cancer or until the data were censored (in other words, the subject died or was considered to be alive and free of a cancer diagnosis as of March 2000).
A secondary analysis was carried out to estimate cancer-specific risk ratios in family G. The selection of cancer sites for analysis was based on previous reports of suspected associations with Lynch syndrome. Other reported cancers were also evaluated. A standardized incidence ratio (SIR) was calculated by dividing the observed number of cancers by the expected number. The expected number was determined by assuming the cancer incidence in family G is the same as the general population. Age-, sex-, period-, and tumor site–specific incidence rates were applied to the appropriate person-years at risk using national and regional databases for the years 1935 to 2000, extrapolating back to 1900. Person-years at risk were accumulated in 5-year age intervals for each family G member beginning with the date of birth or January 1900, which ever occurred later, and ending with the date of cancer diagnosis, date of death, or March 2000,whichever occurred first. Date of death was used as a conservative estimate of date of diagnosis whenever the date of diagnosis could not be determined. Observed numbers were compared with their expected values assuming a Poisson distribution. The 95% confidence intervals (CIs) for the SIRs were determined using Byar’s approximation of an exact Poisson test.19
National and regional cancer-specific rates were obtained from 2 sources: the Surveillance, Epidemiology, and End Results (SEER) 9 database and the Connecticut Tumor Registry.20 SEER incidence data by tumor site, age, sex, and year of diagnosis were used for the period 1973 to 2000 for whites only. Similar incidence data from the Connecticut Tumor Registry were used for the period 1935 to 1972. Cancer incidence rates were not available prior to 1935. Instead, age-, sex-, and tumor site–specific rates for the year 1935 were used as estimates for the period 1900 to 1934. In the SEER 9 database, age-specific rates for 80 to 84 years were used as estimates for individuals older than 84 years.
PAMCOMP21 version 1.4 was used to calculate person-years and SIRs. All other statistical analyses were performed using SAS version 8.2 and S-PLUS version 6.2 (SAS Institute Inc, Cary, NC). P values less than .05 were considered statistically significant.
There are presently 929 known descendants of the original progenitor of family G spanning 7 generations (Table 1). Of the 929 descendants, 74 were diagnosed with 1 or more colorectal or Lynch syndrome–associated cancers (Table 2). Not unexpectedly, colorectal was the most frequent cancer, with an average age at diagnosis of 55 years, followed by cancer of the endometrium, with an average age at diagnosis of 53 years. Variation in age at diagnosis is high. Indeed, the average age at diagnosis of colorectal cancer differs by branch, sex, and generation. Members of certain branches, namely D and I, men, and the later generations have earlier ages at diagnosis (data not shown). For example, in branch I the mean (SD) age at colorectal cancer diagnosis is 46 (14) years vs 59 (17) years in branch B. When comparing men with women, the mean (SD) age at colorectal cancer diagnosis is 48 (15) years for men vs 63 (14) years for women. Although later generations have on average earlier ages at diagnosis, there is little evidence for anticipation. For example, among 18 parent-offspring pairs (n = 25) diagnosed with colorectal cancer, the mean (SD) pair-wise difference in age at diagnosis between parent and offspring is approximately −1 (18) year. Indeed, among 8 of the 18 pairs, the parent was diagnosed at an earlier age than his or her offspring.
Eight family G members were diagnosed with multiple colorectal or Lynch syndrome–associated cancers (Table 3). Remarkably, one branch G member was diagnosed with 5 primary cancers—endometrium, sigmoid (twice), cecum, and stomach—over a 26-year period. Cancers of the colon, rectum, endometrium, and stomach appear across 4 successive generations (II-V) in family G (Table 2). Cancers of the brain and ovary appear less frequently and only intermittently among the generations (Table 2). Other invasive cancers, excluding all skin cancers, are reported in Table 4. Collectively, 115 members of family G were diagnosed with 127 invasive cancers: 74 individuals with 85 colorectal or Lynch syndrome–associated cancers and 41 individuals with 42 cancers at other anatomic sites.
The observed and expected numbers of cancers in family G members are shown in Table 5 for colorectal cancers and Lynch syndrome–associated sites. The incidence of colorectal cancer and endometrial cancer exceeded population expectations by over 3-fold. Results for colorectal cancer were similar in men and women (data not shown). Although cancers of the stomach and brain were also observed in excess of their expectations, the results were not statistically significant. The SIRs for ovarian cancer were less than but not significantly different from unity. Notably, there was a deficit of both breast and lung cancers in family G (data not shown). Only 4 breast cancers were observed compared with an expected 18.20 for an SIR of 0.22 (95% CI, 0.06-0.56), and only 3 lung cancers were observed compared with an expected 13.3 for an SIR of 0.23 (95% CI, 0.05 -0.66). The SIR for prostate cancer was not significantly different from unity (data not shown).
A description by branch of all colorectal and Lynch syndrome–associated cancers of family G is provided in Table 6 . Remarkably, 21 cancers of the colon and rectum and 4 cancers of the endometrium were diagnosed in branch I. Together these cancers comprise nearly 35% of all such cancers diagnosed in family G even though branch I represents less than 12% of the family G members. Another 11 colorectal or Lynch syndrome–associated cancers, 7 of the colon and rectum, 2 of the endometrium, and 2 of the stomach, were diagnosed in branch G in generations II, III, and IV. Like branch I, members of branch G represent only 5% of family G but 13% of all colorectal and Lynch syndrome–associated cancers. In contrast, only 1 invasive cancer (of the bladder) was diagnosed in the 36 members of branch C, whose female progenitor in generation II lived to be 84 years and was never diagnosed with cancer of any kind (Figure 2).
Figure 3 shows the differences by branch in the cumulative proportion of family G members free of a colorectal or Lynch syndrome–associated cancer diagnosis according to age. The differences between branches are statistically significant (P <.001). For example, only 78% of branch I family members (SE, 6%) were free of a colorectal or Lynch syndrome–associated cancer diagnosis at 50 years of age compared with 100% of branch E members. At 60 years of age, the value drops to a mean (SE) 49.7% (8.6%) for branch I while the value for branch E is still 100%. These findings are consistent with branch differences in the SIRs. Specifically, when examined by branch, the SIRs for colorectal cancer were significantly different and greater than unity only for branches D, G, and I (data not shown); likewise, significance of the SIR for cancer of the endometrium was only achieved for branches F and I (data not shown). There were no significant differences by sex in the cumulative proportion of family members free of a colorectal or Lynch syndrome–associated cancer diagnosis after adjustment for branch (P = .51).
The presence of the MSH2 T to G mutation at the splice acceptor site of exon 4 was tested in 40 members of the fourth, fifth, sixth, and seventh generations of family G. A majority of the consented and tested individuals were in branch I (n = 14). No one was tested from branch C. These numbers reflect a combination of factors, including family members who could not be contacted, were at low or high risk based on test results in immediate family members, and/or chose not to be tested. Five of the 40 tested members of family G (all in branch I) are carriers of the MSH2 T to G mutation (Table 7; Figure 4). The putative lines of transmission of the MSH2 mutation can be traced back from all 5 carriers to the common ancestor in generation III, who was diagnosed with endometrial cancer at age 55 years and cancer of the cecum at age 57 years. Testing results have been disclosed to family members with appropriate genetic counseling but are not shown in order to maintain confidentiality.
Of the 35 tested noncarriers in family G, only 3 developed an invasive cancer (prostate, ascending colon, and 1 involving lymph nodes). The family member with colon cancer tested negative for the mutation by conversion analysis (Bert Vogelstein, MD, written communication, January 2005). In total, 97 family G members can now be excluded as mutation carriers, either because they personally do not carry the mutation or because their parent or grandparent does not carry the mutation (Table 7). In contrast, 15 members of family G are at increased risk of carrying the MSH2 mutation by 25% to 50% because a first- or second-degree relative is a known carrier.
The genetic basis of Lynch syndrome in family G is now known, making molecular diagnosis possible for at-risk relatives. We tested 40 members of family G for the T to G transversion in MSH2, a mutation originally identified by Yan and colleagues.17 Five members in 2 generations were heterozygous for the mutation, and 3 of them were previously diagnosed with 1 or more colorectal or Lynch syndrome–associated cancers. All are related to the original progenitor of family G by 3 consecutive generations of family members who were diagnosed with cancer.
Our analysis of 929 members of family G identified a more than 3-fold increased incidence of cancers of the colon, rectum, and endometrium compared with the general US population. These findings support previous observations that these specific cancers are most commonly and consistently involved in families with Lynch syndrome. In contrast, the incidence of both breast and lung cancer was significantly less than expected. This is of interest given previous and conflicting reports in the literature of the incidence of breast22- 24 and lung cancer23 in families with Lynch syndrome. Aside from cancers of the colon, rectum, and endometrium, cancers of the stomach and prostate were the next most commonly reported cancers among family G members. These reported cancers are common in the Lynch syndrome tumor spectrum whereas prostate cancer is common in the general population.
Risk of colorectal cancer was comparable for both sexes in family G but delayed in women by at least 10 years. Sex differences in age at diagnosis may not be specific to family G since similar differences have been observed in other Lynch syndrome families.25 In general, the age at diagnosis of colorectal and Lynch syndrome–associated cancers in family G varied greatly even among mutation carriers. Overall, average age at colorectal cancer diagnosis was 10 years delayed in family G members relative to Lynch syndrome patients ascertained on the basis of Amsterdam criteria (55 vs 45 years). In at least 1 study, however, age at presentation with Lynch syndrome was 56 years among molecularly screened MLH1/MSH2 mutation–positive patients,26 similar to our findings in family G. Incidentally, we observed little evidence for anticipation among parent-offspring pairs affected with colorectal cancer although we had limited data with which to evaluate this hypothesis. Further studies are clearly required to better evaluate age at presentation with Lynch syndrome in an unbiased manner.
Our findings should be interpreted in light of several potential limitations. Foremost is that presently we do not know the carrier status for the majority of the family G members. Results were directly determined or indirectly inferred for 102 of 665 living family members, resulting in 5 carriers and 97 noncarriers. Given our inability to distinguish carriers from noncarriers over a period of more than 100 years, our risk estimates for colorectal and Lynch syndrome–associated cancers are necessarily attenuated. In general, the inclusion of both sporadic and genetic cases of colorectal and Lynch syndrome–associated cancers compromises the accuracy of cancer risk assessments and estimates of age at cancer diagnosis for mutation-positive Lynch syndrome families when mutation testing is incomplete, as it is currently for family G. Smaller than expected risk ratios for colorectal and Lynch syndrome–associated cancers (compared with mutation-positive carriers) and for breast and lung cancers (compared with population expectations) may also be a consequence of underreporting. The relative extent to which each of these factors has biased our results, however, is unknown.
Genetic testing has enormous clinical value for Lynch syndrome families like family G in which a pathogenic mutation has already been identified. In general, genetic testing for a specific, known mutation in other family members is essentially 100% specific and 100% sensitive, obviating the need to rely on surrogate indicators of hereditary cancer risk. Genetic testing of high-risk members of Lynch syndrome families permits focused and increased surveillance and prophylactic interventions on those truly at excess risk by virtue of their carrier status. Several studies have now established the benefits of repeated colonoscopic screening of high-risk and mutation-positive members of Lynch syndrome families, including a 62% reduction in the incidence of colorectal cancer,27 a 65% reduction in overall mortality,28 and a 7-year increase in life expectancy.29 Although we have not identified any mutation-positive members of family G outside of branch I, several lines of evidence suggest that the MSH2 T to G mutation was, and possibly continues to be, segregating in other branches of family G, including a significant excess of colorectal cancer in branches D and G relative to the general population and multiple primary cancers in descendents of branches B, D, and G.
Our results suggest that the uptake of genetic testing for Lynch syndrome is low, even within family G, a historically high-risk family with 6 generations of documented family cancer history. Our results, however, are consistent with a recent study of the uptake of genetic testing in other high-risk Lynch syndrome families,30 emphasizing the need to further examine and address the barriers to genetic testing. The extent to which other family G members are aware of their noncarrier or high-carrier risk status is currently unknown and complicated by the importance of maintaining the confidentiality of tested family members. Future research is needed to determine the impact of genetic testing on untested family members. The benefits of assessing carrier status in a hereditary cancer syndrome family with an identified mutation can only be achieved if family members are aware of their carrier risk or changes in their risk as a result of mutation testing in a relative.31 In mutation-positive families like family G, there are compelling reasons to determine who has and who does not have the mutation.
Corresponding Author: Julie A. Douglas, PhD, Department of Human Genetics, University of Michigan Medical School, Buhl Bldg, Room 5912, Ann Arbor, MI 48109-0618 (firstname.lastname@example.org).
Author Contributions: Dr Douglas had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Douglas, Gruber.
Acquisition of data: Gruber, Meister, Watson, Krush, Lynch.
Analysis and interpretation of data: Douglas, Gruber, Bonner, Watson, Lynch.
Drafting of the manuscript: Douglas, Gruber, Bonner.
Critical revision of the manuscript for important intellectual content: Douglas, Gruber, Meister, Watson, Krush, Lynch.
Statistical analysis: Douglas, Gruber, Bonner, Watson, Lynch.
Obtained funding: Gruber.
Administrative, technical, or material support: Gruber, Meister, Bonner, Krush.
Study supervision: Gruber.
Financial Disclosures: None reported.
Funding/Support: This article was funded by revenue from the Nebraska cigarette taxes awarded to Creighton University by the Nebraska Department of Health and Human Services. Funding was also received from grants U01 CA86389 and R01 CA81488 from the National Institutes of Health, as well as from the Weinstein Foundation and Ravitz Foundation. Dr Lynch received salary support through the Charles F. and Mary C. Heider Endowed Chair in Cancer Research.
Role of the Sponsor: The National Cancer Institute, the Nebraska Department of Health and Human Services, the Weinstein Foundation, and the Ravitz Foundation had no direct involvement in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
Disclaimer: The article’s contents are solely the responsibility of the authors and do not necessarily represent the official views of the State of Nebraska or the Department of Health and Human Services.
Acknowledgment: We thank all of the members of family G who participated in this study, and we gratefully acknowledge the Cancer Genetics Registries at Creighton University and the University of Michigan for providing data essential to this analysis. We also thank Richard Monson, Professor of Epidemiology at the Harvard University School of Public Health, for providing electronic data from the Connecticut Tumor Registry.