Biallelic germline mutations of mismatch-repair genes

Abstract

BACKGROUND.

Heterozygous defects in mismatch-repair (MMR) genes cause hereditary nonpolyposis colorectal cancer (HNPCC). In this syndrome, tumors typically arise from age 25 years onward. Case reports have shown that homozygosity or compound heterozygosity for MMR gene mutations can cause multiple tumors in childhood, sometimes combined with neurofibromatosis type I (NF1)-like features. Therefore, the authors studied the role of homozygosity or compound heterozygosity (CZ) for MMR gene defects in children with multiple primary tumors.

METHODS.

A database that contained all pediatric oncology patients who were seen between 1982 and 2003 at the author's institution was queried to identify patients aged <16 years with more than 1 tumor for whom tissue of at least 1 tumor was available. On isolated DNA, microsatellite instability (MSI) and immunohistochemistry of MMR proteins were assessed.

RESULTS.

In total, 15 patients with more than 1 tumor were identified. Abnormal test results were obtained in 2 of them, including 1 patient who was diagnosed at age 4 years with a glioblastoma (MSI-stable; no human mutL homolog 1 [MLH1] or postmeiotic segregation increased, Saccharomyces cerevisiae 2 [PMS2] expression) and a Wilms tumor (high MSI; no MLH1 or PMS2 expression). Apart from >6 cafe-au-lait spots, he had no other signs of NF1. The patient had CZ identified for a pathogenic MLH1 mutation (593delAG frameshift) and an unclassified MLH1 variant (Met35Asn). There was strong evidence that this unclassified variant was a pathogenic mutation. The second patient was diagnosed with a non-Hodgkin lymphoma (no tissue available) and an anaplastic oligodendroglioma (low MSI; no MSH6 expression) at age 4 years and 6 years, respectively. His brother had died of a medulloblastoma at age 6 years (low MSI, no MSH6 expression). Both boys had cafe-au-lait spots. Further genetic testing was not possible.

CONCLUSIONS.

Carriage of biallelic MMR gene defects can be associated with multiple malignancies in childhood that may differ from the standard spectrum of HNPCC tumor types. In 15 pediatric patients with multiple malignancies, the authors identified 1 clear case and 1 possible case of biallelic MMR gene defect. Recognition of the inherited nature of the tumors in these patients is important for counseling these patients and their families. Cancer 2007. © 2007 American Cancer Society.

Patients with hereditary nonpolyposis colorectal cancer syndrome (HNPCC) or Lynch syndrome are at considerable risk for colorectal cancer and endometrial cancer.1 Various other organs and organ systems also can be affected in Lynch syndrome (eg, stomach, small bowel, ovary, hepatobiliary, pancreas, renal pyelum, and ureter).2, 3 Typically, affected individuals present with cancer in the fourth or fifth decade; however, occasionally, patients may be younger (early in the third decade) or older (with first presentations in the sixth or seventh decade).4 The molecular genetic basis of Lynch syndrome/HNPCC is a germline heterozygous defect in 1 of the mismatch repair (MMR) genes that inherits in an autosomal-dominant manner. Pathogenic mutations in either MMR gene human mutL homolog 1 (MLH1), MSH2, MSH6, or postmeiotic segregation increased, Saccharomyces cerevisiae 2 (PMS2) cause Lynch syndrome.5-9 It is estimated that Lynch syndrome is responsible for approximately 2% to 5% of the total burden of colorectal cancer in the Western world.10-13 In recent years, several case reports have been published of children with either compound heterozygosity (CZ) or homozygosity for an MMR gene defect.14-22 These patients had malignancies at a young age. Overviews are given by Menko et al.19 and Bandipalliam.23 Initially, hematologic malignancies, lymphomas, and brain tumors seemed to be more common in these patients; however, more recent reports also contain more Lynch syndrome-related tumors, such as gastrointestinal tumors.21, 22 Six patients with >1 malignancy have been described. Another striking feature of these patients is that nearly all of them display some features of neurofibromatosis type I (NF1). After identifying a patient with glioblastoma and Wilms tumor at age 4 years who was CZ for MLH1 variants, we searched for MMR deficiency in children with multiple tumors.

MATERIALS AND METHODS

To identify patients with multiple tumors, a prospectively collected database that contained all pediatric oncology and hematology patients in our institution was queried. From a total of 2230 patients in the database who were included between January 1, 1982 and January 1, 2004, 15 individuals, including our index patient, could be identified who had been diagnosed with multiple malignancies. The tumors and medical files of these patients were analyzed anonymously. Informed consent had been obtained previously from the family of our index patient.

Deoxyribonucleic Acid Isolation

The tumor and surrounding normal tissues were fixed routinely in formalin and embedded in paraffin. A hematoxylin and eosin (H & E)-stained section was used to select parts of the tissue that were composed of > 70% tumor cells, and, from the same paraffin block, normal tissue was selected. The selected tissue parts gently were punched out of the tissue block. After this procedure, a second H & E-stained section was made to verify the isolated tissue parts. The tissue fragments were digested, without deparaffinization, in 200 μL of 50 mmol/L Tris-HCl, pH 8.0, to which 20 μL of proteinase K (20 mg/mL) were added. After overnight incubation at 56°C, the lysates were boiled for 10 minutes and subsequently centrifuged. Because the DNA retrieved from routine processed tissues is highly degraded, we used small-amplicon (<250-base-pair [bp]) polymerase chain reaction to obtain molecular information.

Molecular Analyses

Microsatellite instability

To assess for microsatellite instability (MSI), we usedmarkers D2S123 (forward [F], 5V-AAACAGGATGCCTGCCTTTA-3V; reverse [R], 5V-GGACTTTCCACCTATGGGAC-3V; size, 211 bp), D5S346 (F, 5V-AGCAGATAAGACAGTATTACTAGTT-3V; R, 5V-ACTCACTCTAGTGATAAATCGGG-3V; size, 125 bp), D17S250 (F, 5′-GGAAGAATCAAATAGACAAT-3′; R, 5′-GCTGGCCATATATATATTTAAACC-3′; size, approximately 160 bp), Bat25 (F, 5V-TCGCCTCCAAGAATGTAAGT-3V; R, 5VTCTGCATTTTAACTATGGCTC-3V; size, 124 bp), Bat26 (F, 5V-TGACTACTTTTGACTTCAGCC-3V; R, 5V-TAACCATTCAACATTTTTAACCC-3V; size, 123 bp), and Bat40 (F, 5V-ACAACCCTGCTTTTGTTCCT-3V; R, 5V-GTAGAGCAAGACCACCTTG-3V; size, 107 bp). Polymerase chain reaction was performed with 1 μL of isolated DNA in a final reaction volume of 15 μL that contained 1.5 mmol/L MgCl₂, 0.02 mmol/L adenosine triphosphate (dATP), 0.2 mmol/L guanine triphosphate, 0.2 mmol/L thymidine triphosphate, 0.2 mmol/L cytidine triphosphate, 0.8 μCi α-³²P dATP (Amersham Biosciences, Buckinghamshire, UK), 20 pmol of each primer, and 0.2 U of Taq polymerase (Promega Benelux B.V., Leiden, the Netherlands). Polymerase chain reaction was performed for 35 cycles of denaturing at 95°C for 30 seconds, annealing at 55°C for 45 seconds, and extension at 72°C for 1 minute in a Biometra thermocycler (Biometra, Leusden, the Netherlands). A final extension was performed at 72°C for 10 minutes. Polymerase chain reaction products were diluted with loading buffer (95% formamide; 10 mmol/L ethylenediamine tetracetic acid [EDTA] [pH 8.0], 0.025% bromophenol blue, and 0.025% xylene cyanol), denatured at 95°C for 4 minutes, and snap-cooled on ice. Polymerase chain reaction products were separated on a denaturing 6% polyacrylamide gel. After electrophoresis, gels were dried on blotting paper on a vacuum gel dryer and exposed to X-ray film. Films were evaluated by visual inspection. Additional markers were used in MSI analysis, clustered around the PTEN and NF1 genes (Table 1).

Table 1. Results of Microsatellite Instability Analysis and Immunohistochemistry Testing in 15 Pediatric Patients with Multiple Tumors

Case	Tumor 1	MSI	IHC	Tumor 2	MSI	IHC
I	Glioblastoma	Stable	MLH1/PMS2 ↓	Nephroblastoma	High	MLH1/PMS2 ↓
II	Neurofibroma	X	X	Vestibular schwannoma	Stable	Normal
III	Pilocytic astrocytoma	Stable	Normal	Mandibular rhabdomyosarcoma	Stable	Normal
IV	Lymphoma	X	X	Oligodendroglioma	Low	MSH6 ↓
V	Nephroblastoma	Stable	X	Brain tumor	Stable	X
VI	Neurofibroma	Stable	Normal	Meningeoma	Stable	Normal
VII	Chiasma tumor	X	X	Glioblastoma	Stable	Normal
VIII	PNET	Stable	Normal	Kidney sarcoma	Stable	Normal
IX	Sinus tumor	X	X	Lymphoma	Stable	Normal
X	Nephroblastoma	Stable	Normal	Alveolar rhabdomyosarcoma	Stable	X
XI	Nephroblastoma	X	X	Rhabdomyosarcoma	Stable	Normal
XII	Lymphoma	Stable	Normal	Leukemia	X	X
XIII	Angiosarcoma	Stable	Normal	Lymphoma	Stable	Normal
XIV	Lymphoma	X	Normal	Malignant schwannoma	Stable	Normal
XV	Chloroma	Not informative	Normal	ALL	X	X

• MSI indicates microsatellite instability (in the international marker set); IHC, immunohistochemistry; MLH1, human mutL homolog 1; PMS2, postmeiotic segregation increased, Saccharomyces cerevisiae 2; ↓, decreased; X, material not available; MSH6, human mutS homolog 6; PNET, primitive neuroectodermal tumor; ALL, acute lymphocytic leukemia.

Mutation analysis

Mutation analysis for p53 was performed as described previously.24 In short, p53 gene exons 5 through 8 were amplified in 2 overlapping fragments each. PCR products were subjected to single-strand conformation polymorphism (SSCP) analysis. In addition, the PCR products were sequenced bidirectionally on an ABI sequencer. With the same SSCP and sequence procedures, the unclassified variant (UV) of the MLH1 gene (ATG→AAC; MET35ASN) in Case 1 was investigated. PCR primers flanking the UV were F-5′-AGACAGTGGTGAAACGCATC-3′ and R-5′-AGTCGTAGCCCTTAAGTGAG-3′. Mutations in BRAF were investigated after PCR amplification of exon 15 with primers F-5′-AAACTCTTCATAATGCTTGCTCTG-3′ and R-5′-GGCCAAAAATTTAATCAGTGGAA-3′. The PCR products were sequenced bidirectionally on an ABI sequencer.

Immunohistochemistry

Four-micrometer paraffin sections were dewaxed, and antigen retrieval was performed in 10 mmol/L Tris-EDTA buffer, (pH 9.0) in a microwave oven for 20 minutes at 100°C. Primary antibodies anti-MLH1 (Pharmingen BD, Alphen aan den Rijn, the Netherlands; clone G168-728; dilution, 1:100), anti-MSH2 (Pharmingen BD; clone G219-1129; dilution, 1:300), anti-MSH6 (Pharmingen BD; clone 44; dilution, 1:100), and anti-PMS2 (Pharmingen BD; clone A16-4; dilution, 1:200) were applied for 1 hour at 4°C. P53 immunohistochemistry was performed with the antibody DO-7 (DAKO B.V.; Heverlee, Belgium; dilution, 1:100).

After washing, immunoreactivity was visualized with the Envision kit (DAKO B.V.). Subsequently, the sections were counterstained with Mayer hematoxylin and evaluated under light microscopy.

RESULTS

The results of testing for MSI and immunohistochemistry are shown in Table 1. Despite the finding that some materials had been stored for >20 years, DNA extraction and amplification usually were successful as well as immunohistochemistry.

When patients were reviewed, it was possible to make a definite or presumptive diagnosis in nearly all cases based on tumor spectrum, phenotype, and family history. Two patients were identified who had either abnormal MSI results and/or abnormal immunohistochemistry findings. Both cases are described below. In the available tissue from the remaining 13 patients, no abnormal results were found with either technique.

Case I

A boy aged 4 years was diagnosed simultaneously with a Wilms tumor and a glioblastoma and was referred for treatment and further diagnostic evaluation. At physical examination, he had >6 cafe-au-lait spots that measured >0.5 cm, but he had no other features of NF1. The glioblastoma showed strong nuclear p53 expression, whereas the Wilms tumor was negative for p53 (Fig. 1). Mutation analysis revealed a tumor-specific p53 A138T missense mutation in the brain tumor that was not present in the Wilms tumor (Fig. 1). Loss of heterozygosity analyses of different loci revealed different patterns of genomic losses in both tumors (not shown), also indicating that the patient had 2 independent primary tumors.

Figure 1

Case I (CI). p53 Immunohistochemistry and sequence analysis of the Wilms tumor (WT) and brain tumor (BT). Note that no p53 expression and no mutation were identified in the WT, but strong p53 expression and a p53 A138T mutation were identified in the BT.

The glioblastoma was microsatellite stable, but the Wilms tumor appeared to have high MSI (Fig. 2). Both tumors were negative for MLH1 and PMS2 staining, and the normal cells lacked immunoreactivity for these proteins, whereas the tumor and blood vessels were positive for MSH2 and MSH6 (Fig. 3). In both tumors, microsatellite analysis was extended with CA dinucleotide repeat markers at PTEN, DMBT1, and NF1 loci, but the glioblastoma remained microsatellite stable (not shown). Somatic BRAF mutations associated with sporadic high-MSI tumors (V600E, E585K, and D593G) were analyzed in both tumors, and no mutation was detected.

Figure 2

Microsatellite instability (MSI) results from the Wilms (WT) and the brain tumor (BT) compared with normal (N) DNA from Case I. Note that MSI for the microsatellite markers BAT25, BAT26, and D2S123 was observed in the WT and not in the BT.

Figure 3

Histology (hematoxylin and eosin [H & E]) and immunohistochemistry for the mismatch-repair proteins human mutL homolog 1 (MLH1), postmeiotic segregation increased, Saccharomyces cerevisiae 2 (PMS2), human mutS homolog 2 (MSH2), and MSH6 in the Case I (CI) Wilms tumor (WT) and brain tumor (BT), in a sample from a paternal uncle's colorectal cancer (UCRC) with the MLH1 unclassified variant, and in a sample from a maternal aunt's endometrial cancer (AEC) with the pathogenic MLH1 mutation. Note the absence of MLH1 and PMS2 expression in both normal cells and tumor cells from CIWT and CIBT and in tumor cells from UCRC and AEC.

The mother of the boy was an asymptomatic MLH1 gene mutation carrier from a known Lynch syndrome family (593delAG). On further examination, it was revealed that his father was a member of a family with early-onset colorectal cancer (Fig. 4A, pedigree). The father's brother was diagnosed with colorectal cancer at age 42 years. This tumor displayed a high MSI phenotype. Immunohistochemistry showed no MLH1 or PMS2 staining in the colorectal cancer cells, whereas the normal cells were positive for both proteins (Fig. 3). No BRAF mutation was detected in the colorectal cancer. Both the father and his brother were carriers of an MLH1 mutation (Met35Asn), the clinical significance of which is not known (UV). Both the maternal mutation and the paternal UV were inherited by the patient. SSCP and sequence analyses for the UV in the index case glioblastoma and the Wilms tumor revealed no MLH1 gene allelic loss, whereas loss of the wild-type MLH1 allele was demonstrated in the uncle's colorectal cancer (Fig. 5). Unfortunately, no more DNA was available for NF1 mutation analysis in the index patient.

Figure 4

(A) Pedigree of Case I (index patient indicated by arrow). The mutations in MLH1 are depicted below the tested individuals, and the age of onset of cancer onset is indicated in years (yrs). Circles indicate female; square, male; slash forward, individual deceased; dark right half, affected by cancer; dark square, multiple cancers. CRC indicates colorectal cancer; BT, brain tumor; WT, Wilms tumor; CAF, cafe-au-lait spots. (B) Pedigree of Case IV. (index patient indicated by arrow). Ly indicates lymphoma.

Figure 5

Sequence analysis of the MLH1 UV Met35Asn in Case I normal DNA (CIN) and Wilms tumor (CIWT) and in uncle's normal DNA (UN) and colorectal cancer (UCRC). Note the presence of the UV in both patients' normal and tumor DNA with no loss in CIWT and loss of the wild type allele in the UCRC (*).

Case IV

The second case concerned a boy of Moroccan descent who was diagnosed with a non-Hodgkin lymphoma at age 4 years. At age 6 years, an oligodendroglioma was diagnosed. At physical examination, he had multiple cafe-au-lait spots. There were no further signs of NF1.

His brother died of a medulloblastoma at age 8 years. He also had multiple cafe-au-lait spots (Fig. 4B, pedigree). Mutational analysis of the NF1 gene was negative. No information with regard to consanguinity in this family was available.

Both the oligodendroglioma and the medulloblastoma displayed a low MSI phenotype (Fig. 6) and completely lacked MSH6 staining in both tumor cells and normal cells (Fig. 7). Five samples of normal brain tissue were stained as immunohistochemical controls, and all 5 were negative for MSH6, except for the blood vessel cells. To further evaluate the role of immunohistochemistry in brain tumors, 6 sporadic medulloblastomas and 6 sporadic oligodendrogliomas were investigated. All tumors stained positive for MSH2 and, to a lesser extent, for MSH6. Again, all normal brain tissues that could be evaluated in these patients showed no expression for MSH6. Unfortunately, no tissue was available from the lymphoma. Further genetic analysis was not possible, because no consent could be obtained.

Figure 6

Microsatellite instability (MSI) results from the Case IV (CIV) oligodendroglioma (O) and from his brother's medulloblastoma (BM) compared with matched normal DNA (N). Note MSI for the microsatellite markers D2S123 and BAT40.

Figure 7

Immunohistochemistry for MSH2 and MSH6 in the Case IV (CIV) oligodendroglioma (O), in his brother's medulloblastoma (BM), and in unrelated, normal control brain tissue (NB). Note the absence of MSH6 expression in tumor cells in both CIVO and BM and in normal brain cells. In blood vessels (arrows), few cells are positive for MSH6 (arrowhead) in normal brain tissue. No MSH6-positive cells in blood vessels were found in CIVO or BM.

DISCUSSION

In 15 pediatric patients who had more than 1 tumor, most of which were malignancies, we were able to identify 1 patient (Case I) with CZ and 1 patient (Case IV) with possible homozygosity or CZ for MMR gene defects. In Case I, the Wilms tumor had high MSI, and the glioblastoma was microsatellite stable. This patient was CZ for a pathogenic MLH1 gene mutation and an MLH1 gene UV. The absence of MLH1 and PMS2 expression points to a primary MLH1 gene defect and not a primary PMS2 defect.7 Despite the fact that the UV in the paternal MLH1 gene in Case I has not to our knowledge been described in HNPCC families, we obtained indications that this UV is a pathogenic mutation. These indications are that the methionine at position 35 is evolutionary conserved and is located in an evolutionary conserved and functionally important domain (ATPase/ATP binding activity). In addition, it has been demonstrated that a comparable mutation Met35Arg is important functionally in yeast.25-27 The UV is present in a family with early-onset colorectal cancer. The index case, age 4 years, is carrier of the UV and has 2 independent tumors, as demonstrated by histology and by p53 and loss of heterozygosity analyses. Both tumors and the normal tissue lack MLH1 and PMS2 protein expression without the loss of an MLH1 allele. A colorectal cancer from a family member with the UV was negative for MLH1 and PMS expression, whereas the normal cells were positive. This tumor demonstrates loss of the wild-type MLH1 allele. Obviously, no immunoreactive MLH1 protein is generated from the UV allele, as demonstrated by the absence of MLH1 expression in both the normal cells from the index case and in the tumor cells from an uncle with the UV. Finally, in both tumors from Case I and in the uncle's colorectal cancer, no oncogenic BRAF mutations were identified. BRAF mutations are found almost exclusively in sporadic cancers with high MSI and not in HNPCC-related tumors. However, an as yet undetected pathogenic MLH1 gene mutation, most likely in cis with the UV, cannot be ruled out completely.

The brain tumor did not display a high MSI phenotype. Possibly, MMR deficiency-associated brain tumors display a different pattern of microsatellite marker instability from colorectal cancers, as observed in patients with HNPCC-related endometrial cancer who are MSH6 mutation carriers.28 Further exploration of this hypothesis by testing additional markers clustered around the PTEN, DMBT1, and NF1 genes for MSI revealed no abnormalities. In addition, MSI was detected previously in pediatric brain tumors,29 and, in 5 patients, it was proven that this was caused by MLH1 or MSH2 germline mutations.30, 31 Another explanation may be that other selective advantages of MMR-deficient cells, such as resistance to apoptosis, played an important role in the development of our patient's brain tumor.32

In Case IV, both the oligodendroglioma and the medulloblastoma in the patient's brother had low MSI. Both tumors were positive for MSH2 expression, but no MSH6 expression was observed in either tumor cells or normal cells. We observed no MSH6 expression in brain cells from normal control samples. However, solitary MSH6-positive cells were observed in blood vessels. In sporadic medulloblastomas and oligodendrogliomas, MSH6 was expressed, possibly because of increased proliferation. In addition, the absence of MSH6 expression in the blood vessels from the oligodendroglioma and medulloblastoma in our patients points to a biallelic MSH6 gene germline defect. However, these findings have to be interpreted with caution because of the limited data on MSH6 staining in brain tissue and tumors. Also, in the patient who was described in the literature with a proven homozygous MSH6 mutation, the brain tumor in the affected patient did express MSH6, unlike our patient.19 We were unable to investigate the family of Case IV any further, because no consent was obtained. Based on the type of tumors and the occurrence of cafe-au-lait spots in the affected patients, the apparent autosomal-recessive phenotype, and the molecular results, we assume that a biallelic MMR deficiency causes this remarkable clustering of malignancies in this relatively small family.

In the other 13 patients, molecular analysis showed no evidence for biallelic MMR deficiencies; however, based on the tumor spectrum (Table 1), other genetic predispositions may have been responsible for the development of malignancies in these children, eg, NF1 (Cases VII and XIV), neurofibromatosis type II (NFII) (Cases VI and II), p53 mutation (Cases III, X and XI), or SNF5 (Cases V and VIII). Cases VII and XIV and Cases II and VI actually fulfilled the clinical diagnostic criteria for NF1 and NFII, respectively. In the other patients, no signs of NF1 were observed.

In conclusion, germline biallelic MMR gene mutations can cause central nervous system tumors, hematologic and lymphatic malignancies in addition to gastrointestinal malignancies and Wilms tumors. MSI and immunohistochemical analysis can produce abnormal results, but the results are not consistent enough to recommend either technique for identifying MMR-deficient children. Dermatologic evaluation for the presence of cafe-au-lait spots and family history may indicate an MMR defect. However, in patients who have small families or hypomorphic mutations (ie, mutations leading to an abnormal protein with residual function), with a reduced penetrance, pedigree analysis will not help to identify biallelic MMR gene defects.22

It is important for pediatricians and clinical geneticists to be aware of biallelic MMR deficiency and other genetic predispositions in children with multiple malignancies to obtain an accurate cancer risk assessment in these patients and their relatives aimed at risk-reducing interventions whenever possible. We advise performing mutation analysis of MLH1, MSH2, and MSH6 in all children who have multiple tumors, cafe-au-lait spots, and a negative family history or mutation analysis for NF1.