Familial Adenomatous Polyposis: Experience from a Study of 1164 Unrelated German Polyposis Patients

The autosomal-dominant precancerous condition familial adenomatous polyposis (FAP) is caused by germline mutations in the tumour suppressor gene APC. Consistent correlations between the site of mutations in the gene and clinical phenotype have been published for different patient groups. We report our experiences of APC mutation analysis and genotype-phenotype correlations in 1166 unrelated polyposis families and discuss our results in the light of literature data. We show that the mutation detection rates largely depend on the family history and clinical course of the disease. We present a list of 315 different point mutations and 37 large deletions detected in 634 of the 1166 index patients. Our results confirm previously published genotype-phenotype correlations with respect to the colorectal phenotype and extracolonic manifestations. However, 'exceptions to the rule' are also observed, and possible explanations for this are discussed. The discovery of autosomal-recessive MUTYH-associated polyposis (MAP) as a differential diagnosis to FAP implies that some results have to be reinterpreted and surveillance guidelines in the families have to be reevaluated.

I In nt tr ro od du uc ct ti io on n Familial adenomatous polyposis (FAP, OMIM +175100) is a clinical diagnosis that is typically based on the presence of more than 100 colorectal adenomas. If untreated patients develop colorectal cancer at a mean age of 40 years [1]. Other gastrointestinal features (duodenal adenomas, fundic gland cysts) and extragastrointestinal manifestations, including congenital hypertrophy of the retinal pigment epithelium (CHRPE), desmoids and osteomas are frequently described.
FAP is an autosomal-dominant disorder caused by germline mutations in the tumour suppressor gene APC on chromosomal region 5q22 [2,3]. The APC gene encodes a multifunctional protein of 2843 amino acids that plays a major role in controlling cell cycle progression, migration, differentiation and apoptosis (review in [4,5]). Germline mutations in the APC gene were also detected in some families with an attenuated polyposis phenotype (AAPC, AFAP) who present with less than 100 adenomas and a later age at onset compared to patients with typical FAP [6].
During the last few years correlations between site of mutation in the APC gene and clinical phenotype have been reported on patient groups of different sizes [7][8][9][10][11][12]. Most of the published correlations proved to be statistically consistent, but there were also 'deviations from the rule' when individual cases were considered.
Recently a polyposis syndrome characterised by multiple adenomatous polyps and an autosomal-Waltraut Friedl, Stefan Aretz -recessive mode of inheritance has been identified that is caused by germline mutations in the base excision repair gene MUTYH (MYH) [13]. Biallelic MUTYH mutations have been identified in up to 40% of patients in whom no mutation in the APC gene was identified [14][15][16]. This new adenomatous polyposis condition is designated as MUTYH-associated polyposis (MAP; OMIM #608456) and has to be considered as a differential diagnosis of the autosomal-dominant FAP.
In this review we report on our experience of mutation analysis in the APC gene and genotype-phenotype correlations in 1166 unrelated patients with a clinical diagnosis of FAP or multiple adenomatous polyposis consistent with AAPC and discuss our results in the light of literature data.
P Ph he en no ot ty yp pe e c cl la as ss si if fi ic ca at ti io on n Since 1991, blood samples from 1166 unrelated patients with a clinical diagnosis of typical or attenuated FAP (AAPC) have been referred to the Institute of Human Genetics, University of Bonn for mutation analysis in the APC gene. Clinical information on 2066 patients from the 1166 families was obtained during genetic counselling sessions, from a questionnaire, or through telephone interviews and/or medical records. The study was approved by the Ethics Commission of University Hospital, Bonn. C Co ol lo or re ec ct ta al l p ph he en no ot ty yp pe e The classification of different FAP phenotypes was based on the number of colorectal adenomas, age at diagnosis of FAP and occurrence of CRC (Table 1). The FAP phenotype was classified as typical when the patient presented with more than 100 colorectal adenomas before the age of 35; in cases of unavailable or unclear colonoscopic data, the classification was based on the occurrence of clinical bowel symptoms before the age of 35, or a diagnosis of CRC before the age of 45. The diagnostic criterion for the attenuated phenotype (AAPC) was the occurrence of a smaller number of adenomas (10-100) after the age of 25 or more than 100 adenomas diagnosed for the first time after the age of 45. When the polyp number was unknown, AAPC was assigned if the first symptoms or diagnosis of CRC occurred after the age of 45. The phenotype was considered severe in patients who developed more than a thousand adenomas or polyposis-related bowel symptoms before the age of 15, or when diagnosed with polyps before the age of 10. The phenotype was classified as atypical in patients who did not meet the criteria for either typical or attenuated FAP, when an unambiguous attribution was impossible. Patients with an atypical course usually presented with more than 100 polyps, diagnosed between 35 and 45 years of age; in addition, cases with an obvious discrepancy between the age at diagnosis (symptoms) and the number of colorectal adenomas were also considered atypical. If clinical information on the colorectal disease was not available or not sufficient to determine the extent of colorectal polyposis, the phenotype was considered 'unknown'.
F Fa am mi il ly y h hi is st to or ry y Due to the existence of the recently discovered MUTYH-associated polyposis [13] we considered it important to allow for the mode of inheritance in the families: index patients were classified as familial cases with autosomal-dominant inheritance when at least two patients (one of them a parent of the index patient) were affected in the family. An autosomal-recessive mode of inheritance was considered when the index patient had at least one affected sibling, but no affected parents or children. Cases where the index patient was the only affected person known in the family, or no information on other family members (parents, siblings, children) was available, were classified as 'single cases'. Index patients were classified as de novo mutations only if an APC mutation was identified and both parents were either healthy until an advanced age, or when the mutation was excluded in both parents, irrespective of whether or not the index patient had affected children [17].
M Mu ut ta at ti io on n a an na al ly ys si is s i in n t th he e A AP PC C g ge en ne e P Po oi in nt t m mu ut ta at ti io on ns s Genomic DNA extracted from peripheral blood samples was used for mutation analysis. We started the search for germline mutations in the APC gene in 1991 by SSCP and heteroduplex analysis. To date, we apply the protein truncation test (PTT) for detection of mutations in exon 15 in four overlapping fragments, essentially as described [18], using an in vitro transcription translation kit (Promega, Mannheim, Germany) in the presence of 35 S-Methionine (Amersham). Denaturing high-performance liquid chromatography (DHPLC) is used for screening of exons 1-14 and the first 400 bp of exon 15 (WAVE, Transgenomics). PCR fragments showing variant bands by either method were sequenced on an ABI prism 377 or ABI 3100 automated sequencer (Applied Biosystems, Darmstadt, Germany) using the cycle sequencing procedure and the BigDye terminator kit version 2.0 or 1.1, respectively. Some of the APC mutation negative patients from the first series were reexamined using the more up-to-date procedures described.
We identified a germline point mutation in the APC gene in 597/1166 patients (51.2%) (Table 1). Overall, we found 315 different point mutations (Supplementary  table 1). Most of the mutations are spread over the first half of the gene. Of note, no mutation was detected in exons 1 and 2 ( Fig. 1). Two hot spot mutations at codons 1309 and 1061 are evident.
Most mutations were truncating, including 364 frameshift mutations, 182 nonsense mutations and 38 mutations in the highly conserved splice site sequences. In addition, in 13 patients we identified 8 different missense mutations and three different silent substitutions; seven of these single base substitutions were localised at exonic boundaries. For four substitutions localised in exon 4 and exon 14 we could demonstrate by RNA analysis that they actually affect splicing and thus cannot be considered missense mutations or silent variants [19].
L La ar rg ge e g ge en no om mi ic c d de el le et ti io on ns s DNA samples from patients negative for point mutations in the APC gene were examined for the presence of large genomic deletions or duplications using MLPA (multiplex ligation-dependent probe amplification) (SALSA P043 APC exon deletion test kit; MRC Holland, Amsterdam) according to the manufacturer's recommendations, as described [20].
Five large genomic deletions encompassing the APC gene had been detected earlier by haplotype analysis; one of them was cytogenetically visible [21,22]. By use of MLPA we have examined so far 401 out of the 569 patients negative for point mutations in the APC gene for the presence of large genomic deletions or duplications [20]. Altogether we have detected a total of 37 large deletions. The deletions extend over single exons up to the entire APC gene including the promotor region (see Supplementary table 2 [19] 1309 mutation increases to 10% (54/537). The mutation occurs at high frequency de novo: in 22 of the 56 index patients (39%) the 5 bp deletion at codon 1309 was excluded in both parents [17]. In the two patients diagnosed at ages of 43 and 53, the mutation had occurred de novo as well. This mutation was also reported at high frequencies in patients from Japan, Singapore and South Africa [23][24][25] and at moderate proportions in most European populations [26][27][28][29], but -interestingly -it has not been found in patient groups from Australia or Spain [30,31]. The hot spot mutation c. 3183_3187delACAAA (at codon 1061) was detected in 43 index patients: 3.7% of all patients and 7.1% (36/504) in the group of patients with typical FAP. This mutation occurred de novo in 7 of the 43 index patients (16.3%).
M Mu ut ta at ti io on n d de et te ec ct ti io on n r ra at te e a ac cc co or rd di in ng g t to o p ph he en no ot ty yp pe e a an nd d f fa am mi il ly y h hi is st to or ry y APC mutation detection rates in FAP patients varying between 30-85% have been published [26][27][28][29]32]. For evaluation of the APC mutation detection rate and prevalence of mutations it is worthwhile to consider the colorectal phenotype and the family history of index patients examined for mutations. Overall, point mutations and large deletions in the APC gene were detected in 634 of the 1166 index patients (54%) ( Table 1). The mutation detection rate was 82% in patients with typical FAP, but only 24% in patients with AAPC. Remarkably, 30 of the large genomic deletions (81%) occurred in patients with typical FAP.   The distribution of patients according to family history and the corresponding mutation detection rates are presented in Table 1. As expected, the mutation detection rate increases in all patient groups when only familial cases with an autosomal-dominant mode of inheritance are considered ( Table 1). The mutation detection rate in cases with de novo mutations is -by our definition -100%. Most of the patients with a de novo mutation had developed typical FAP.
G Ge en no ot ty yp pe e--p ph he en no ot ty yp pe e c co or rr re el la at ti io on ns s For evaluation of the correlation between the site of mutation in the APC gene and different colonic and extracolonic manifestations we included clinical information obtained from 2066 patients from the 1166 families.

C CH HR RP PE E
Presence of CHRPE has been reported in patients with mutations 3' to exon 9 [9, 10] but not beyond codon 1444 [8,33]. However, isolated cases constituting exceptions to this 'rule' have also been published [34].
Information on the results of retinal examination was available for 413 patients from 285 families with known germline mutations ( Table 2). Since patients with mutations in exon 9 were published both with and without CHRPE, we defined the limit of the region characteristic for CHRPE in codon 410, the first codon outside the alternatively spliced sequence of exon 9.
Overall Taken together, the previously published correlation between site of mutation and expression of CHRPE was confirmed in our large group of patients, but overall 18% of patients with point mutations (63+9/393) do not conform to the rule. Part of this deviation might be explained by the fact that some patients were seen by ophthalmologists who are not aware of the different types of CHRPE, including even small retinal lesions [35]. However, intrafamilial discrepant observations were also found in some patients who were examined by an ophthalmologist experienced in diagnosis of CHRPE over several years. Of note, 12 out of the 63 CHRPE-negative patients (19%) with mutations within codons 423-1367, but only 21 of the 244 CHRPE-positive patients (9%) had a de novo mutation (p=0.023; Fisher's exact test, two sided), suggesting that CHRPE may not be expressed in some patients with de novo mutations because of somatic mosaicism. C Co ol lo or re ec ct ta al l p ph he en no ot ty yp pe e According to data from the literature, a severe clinical phenotype is associated with mutations at codon 1309 [7,12], and an attenuated FAP is observed in patients with mutations 5' to codon 168 [6], in the alternatively spliced sequence of exon 9 [11,[36][37][38][39] or 3' to codon 1580 [40][41][42] while typical FAP is observed in patients with mutations at the remaining sites.
Progression of colorectal disease in FAP patients is a continuous process. Therefore, evaluation of the severity of the disease is largely dependent on time of  diagnosis. The most accurate parameters for evaluation of the natural course of the disease would be the age at occurrence of the first colonic adenomas and the age-related number of polyps. However, these data are available only for a small number of mutation carriers who started with regular colonoscopies at an early age. Age at diagnosis of adenomas in asymptomatic mutation carriers who underwent endoscopic screening is not a suitable parameter as it depends on the time of the first endoscopy and whether surveillance is regularly performed. Therefore, we used the age at diagnosis of FAP in patients who developed gastrointestinal symptoms (rectal bleeding, diarrhoea, etc.) prior to the first endoscopies as a parameter for the severity of the colorectal phenotype [28]. Age at diagnosis of FAP was known in 760 patients who were diagnosed after the onset of bowel symptoms (Fig. 2). The mean age at diagnosis in these patients was 35 years (range 3-78 years). Germline mutations in the APC gene were identified in 485 of the 760 patients diagnosed because of bowel symptoms. We confirmed in these patients the correlation between APC mutation site and mean age at diagnosis of FAP (Fig. 2). Patients with a mutation at codon 1309 developed bowel symptoms more than 10 years earlier (20 years of age) compared with patients with other mutations: 50 years of age in patients with mutations at sites characteristic of AAPC, 32 years of age in patients with mutations at other sites and 41 years of age in patients with no identified APC mutation (p<0.001 for all comparisons; two-sided Student's ttest). The 32 patients with large genomic deletions had a mean age at diagnosis of 33 years, comparable to that of patients with mutations characteristic of typical FAP (p=0.38). Despite this significant correlation a considerable variation of age at diagnosis in patients with the same mutation or even within the same family was observed, as we had illustrated for part of this patient group in a previous report [28].
As another parameter for severity of the disease we used the clinical phenotype classification that takes into account the age-related polyp number in addition to the age at diagnosis (see phenotype classification).
When grouping the 2066 patients according to the clinical phenotype and the mutation intervals outlined above, most patients conformed to the published genotype-phenotype correlation ( Table 3): 84 of the 101 patients with mutations at sites characteristic of AAPC (83%) were classified as having an attenuated phenotype, and another 8 patients had an atypical phenotype. Accordingly, 80% of patients with mutations at sites characteristic of typical FAP (587/735) were classified as having a typical FAP phenotype. Only 34% (28/83) of patients with mutations at codon 1309 were classified as having a severe phenotype, whereas 63% (52/83) had typical FAP.
I In nf fl lu ue en nc ce e o of f m mu ut ta at ti io on n t ty yp pe e o on n c cl li in ni ic ca al l p ph he en no ot ty yp pe e The genotype-phenotype correlations presented here and in reports on other patient groups roughly consider only the site and not the type of mutations. When we looked in more detail at mutations between codons 170-1578 in the 68 patients with attenuated phenotype who did not conform with the expected genotype-phenotype correlation, we observed some distinctive features.
In this patient group there was an unexpectedly high proportion of splice site mutations: 20/68=29% in patients with attenuated phenotype vs. 25/587=4% in patients with typical phenotype (p<0.001; Fisher's exact test, two sided). Some of these splice site mutations may lead to in-frame deletions or to partial exon deletions. We present some examples: we showed by RNA analysis that the substitution c.1312+3A>G in intron 9 detected in two unrelated patients with attenuated phenotype leads to partial loss of exon 9, while the substitution at position +5 of this intron (c.1312+5G>A) detected in a family with typical FAP leads to complete loss of exon 9. Moreover, the substitution at the highly conserved position -2 in intron 14 (c.1959-2A>G) is predicted to result in a splicing  defect. RNA analysis demonstrated a partial use of a cryptic splice site 12 bp downstream, resulting in a slightly shortened in-frame transcript and likely to lead to a functional APC protein [19]. Somatic mosaicism in patients with de novo mutations may be another reason for an attenuated phenotype. In two patients the nonsense mutations p. Arg216X and p. Gln1127X, respectively, were detected in blood samples only as faint signals, suggesting in these cases the presence of a somatic mosaic in the blood and possibly in the colonic epithelium. In the two patients with the 5 bp deletion at codon 1309 and an attenuated phenotype the mutation had occurred de novo; the mutant alleles were detected in their blood samples at normal levels (comparable to familial cases) but may be present as somatic mosaics in the intestinal epithelium.
While these examples can explain only some of the cases of nonconformity with the established genotype-phenotype correlations, they demonstrate that mutational consequences may be different even if mutations occur in the immediate vicinity of one another. Since the majority of mutations are truncating and most cases of possible mosaicism affect only the first generation, exceptions to the otherwise consistent genotype-phenotype correlations are relatively rare.

M Mo od di if fi ie er r g ge en ne es s
The intrafamilial variability frequently observed in large multigeneration FAP families [41,[43][44][45] cannot be explained by the reasons outlined above. Variations in other genes acting as modifiers at the level of APC synthesis and function or on other regulatory proteins in Wnt signalling and cell adhesion are discussed as causative factors of phenotypic variability within families or in patients with the same germline mutation. In animal experiments a variant in the secretory type II phospholipase A2 gene (Pla2s) designed as MOM1 (modifier of Min) has been proven as the first modifier influencing the number of intestinal polyps in Min mice [46]. Discordant findings regarding the existence of a modifying locus for severity of colonic or extracolonic manifestations in FAP on human chromosome 1p35-36, the region synthenic with the Pla2s locus on mouse chromosome 4, have been published [47][48][49][50]. Moreover, polymorphisms in enzymes playing a role in different metabolic and cell cycle controlling pathways, e.g. N-acetyl transferase type 1 and 2, glutathione S-transferases M1 amd T1, and others, have been examined for possible modifying effects on gastrointestinal and extragastrointestinal symptoms [47][48][49][50]; however, no concordant results were obtained.

D De es sm mo oi id d d di is se ea as se e a an nd d o os st te eo om ma as s
A higher frequency of desmoid tumours and osteomas has been repeatedly observed in FAP patients with a germline mutation 3' to codon 1444 [8,51]. Moreover, several families with mutations in the 3' part of the APC gene who developed desmoids in the absence of polyposis or with a very mild polyposis phenotype have been described [52,53].
In our patient sample presence or absence of desmoids and osteomas was only occasionally reported in the medical records, and we have not specifically asked for this information during the last few years. Therefore, our data on these extracolonic manifestations are mainly similar to those previously published [28]. The actual data are presented in Table  4. Information on desmoid and osteoma status was obtained from 380 and 195 patients with identified mutations, respectively. From these data it becomes evident that in a retrospective study like ours, where patients are not explicitly examined for desmoids or osteomas, there must be a selection bias: the presence of clinically manifest desmoids or osteomas in patients is more likely to be documented as compared to their absence. Therefore, when calculating the presence of these extracolonic manifestations on the basis of patients with documented disease status, the values are too high. We believe that the proportions calculated on the basis of all patients with mutations in different APC regions (figures given in parentheses in Table 4) better reflect the real situation regarding the presence of these extracolonic features. Even with this rough estimation the higher proportion of desmoids and osteomas in patients with mutations 3' to codon 1444 remains evident. Of note, CHRPE status is not so much affected by this selection bias, because determination of the presence or absence of CHRPE requires a special ophthalmological examination.

C Co on nc cl lu us si io on n
The discovery of the molecular basis of FAP by identification of germline mutations in the APC gene in polyposis patients has stimulated the interest of clinicians and molecular biologists in this disease. During the last years several groups of polyposis patients have been examined for mutations in the APC gene. APC mutation detection rates in FAP patients varying between 30-85% have been published. This variation largely depends on patient selection and their clinical phenotype: the discovery that attenuated FAP is also caused by germline mutations in the APC gene has increased the proportion of patients with few adenomas irrespective of the family history among patients examined for mutations. This has led to an overall decrease in the mutation detection rate. Our data show mutation detection rates up to 85% when patients with typical FAP are considered. Impressive correlations between the site of mutations in the large APC gene and the clinical phenotype including both the colonic polyposis phenotype and manifestation of extracolonic features have been published. Examination for CHRPE had been proposed as a useful marker for pointing to the gene region that should be examined for APC mutations. In addition, CHRPE status was thought to be used for predictive testing in families with no identified APC mutation, but where the index patient presents with CHRPE. However, findings of intrafamilial discrepancies in CHRPE status limit its use as a predictive marker in persons at risk. Moreover, due to the absence of CHRPE in most patients with an attenuated polyposis phenotype, ophthalmogical examinations are of little value in the actual diagnostic procedure.
The association of attenuated FAP with APC mutations localised in the 5' or 3' part of the gene or in the alternatively spliced region of exon 9 has also repeatedly been published and has been, in principle, confirmed in our sample of 1379 patients with known APC mutations. When examining this correlation it should be kept in mind that classification of the clinical phenotype into typical and attenuated FAP may be difficult in some cases. This classification is based ideally on the age at onset of polyp growth and age-related polyp numbers, but in most cases these data are not available because patients are diagnosed at different stages of disease development. On the other hand, the type of mutation may also play a role and may explain some exceptions to the expected relationship. In addition, even intrafamilial variation of the clinical phenotype has been described, pointing to the presence of other genetic or environmental factors. Therefore, identification of the APC mutation in a patient is of limited value for the prediction of the clinical course of the disease or for therapeutic decisions.
The discovery of MUTYH-associated polyposis (MAP) due to biallelic mutations in the MUTYH gene has complicated our perception of adenomatous polyposis. Initially it was thought that all cases of adenomatous polyposis follow an autosomal-dominant mode of inheritance and are caused by mutations in the APC gene, irrespective of whether or not the mutation was identified. The existence of an autosomal-recessive polyposis disease that clinically resembles FAP implies that some statements regarding risk predictions in the families based on indirect genotype analysis or apparent de novo mutations have to be reevaluated. The frequency of MUTYH-associated polyposis among unselected APC-mutation-negative polyposis patients ranges between 15% and 40%.
In this context the search for an APC mutation, especially in patients without a family history of FAP, is of importance for defining the recurrence risk in their families. However, mutation analysis is still laborious and expensive, and even with more up-to-date and sophisticated methods a mutation in the APC (or MUTYH) gene(s) cannot be excluded. Moreover, as other polyposis syndromes (e.g. familial juvenile polyposis or Peutz Jeghers syndrome, caused by mutations in other genes) have to be considered as differential diagnoses, we would like to emphasise the importance of the clinical diagnosis including polyp number, site of location in the colorectum and histology, and family history prior to requesting molecular diagnostics. These data, especially polyp histology, are necessary in order to differentiate between different types of polyposis and to provide mutation analysis of the appropriate genes.
The identification of the underlying germline mutation(s) in polyposis patients allows predictive diagnostics in persons at risk in their families. Genetic counselling should be offered to the patients and their families in order to explain the complex picture of polyposis disease, including the possibilities and limitations of molecular diagnostics for surveillance and therapy.
A Ac ck kn no ow wl le ed dg ge em me en nt ts s We are grateful to the patients who participated in the study and to the colleagues who helped us with clinical information about the patients. We thank K. Siberg and M. Vogel for data collection and M. Sengteller, D. Stienen and S. Uhlhaas for skilful technical assistance. The study was supported by German Cancer Aid (Deutsche Krebshilfe e. V. Bonn, project no. 106244).