An Interstitial Deletion at 3p21.3 Results in the Genetic Fusion of MLH1 and ITGA9 in a Lynch Syndrome Family

About 3% to 5% of all colorectal cancer cases in humans can be classified as hereditary nonpolyposis colorectal cancer (Lynch syndrome). The Lynch syndrome, an autosomal-dominant disorder, is caused by heterozygous defects in mismatch repair genes, mainly MLH1 (∼50%) and MSH2 (∼40%), also MSH6 or PMS2, and responsible for the early onset of colorectal and endometrial cancer (1). In addition, patients with Lynch syndrome are at increased risk to develop urinary tract, stomach, and small intestine cancer (2). Muir-Torre syndrome, a phenotypic subvariant of Lynch syndrome, is characterized by the association between one or more visceral malignancies and sebaceous tumors or keratoacanthomas. The Muir-Torre phenotype is not linked to particular mismatch repair gene alterations; hence, the Muir-Torre syndrome families present a wide phenotypic and genotypic variability (3, 4).

Mismatch repair deficiencies are either caused by point mutations, deletions (5, 6), or epigenetic processes, for example, silencing of specific mismatch repair promoters due to DNA methylation (7). Absence of key proteins or the presence of mutated repair proteins impairs the formation of functional repair initiating complexes. MLH1 binds either to PMS1 or PMS2, and both heterodimers (MutLβ and MutLα) bind either to the heterodimers MSH2/MSH6 (MutSα; recognition of mispairs) or MSH2/MSH3 [MutSβ; recognition of DNA insertions/deletions up to 12 nucleotides (nt) in length]. In mammalian cells, the association with proliferating cell nuclear antigen (which interacts with MSH3 and 6), RPA, exonuclease I, and DNA repair polymerases δ and ε is a crucial step for subsequent repair processes (8). Moreover, the proteins MLH1, MSH2, and MSH6 are part of the “surveillance and repair complex” involving ATM, BLM helicase, BRCA1, BRCA2, RAD50, MRE11, NBS1, Ku70, KU86, DNAPK_cs, and RFC (9). Therefore, many MLH1 and MSH2 mutations compromise the assembly of different protein complexes and result in the inability to cope with genotoxic stress situations. Mismatch repair deficiency causes the accumulation of numerous mutations in repetitive DNA sequences, leading to microsatellite instability, which is typically found in tumors of patients with Lynch syndrome (10).

The ITGA9 (integrin α 9) gene codes for a member of the integrin receptor family. Integrins form heterodimeric integral membrane glycoproteins and are composed of an α- and a β-chain that mediate cell-cell and cell-matrix adhesion (11). Integrin α 9 protein binds to integrin β 1 and thus forms a receptor for VCAM1, cytotactin, and osteopontin (12), also for vascular endothelial growth factors (13). Very high expression of ITGA9 has been shown in small cell lung cancer, and recently, ITGA9 has been described as a candidate cancer gene in breast and colorectal cancer (14, 15).

Here, we report on an in-frame gene fusion of MLH1 and ITGA9 in a Lynch syndrome family. This novel allele is unable to confer mismatch repair activity.

Family pedigree. The family pedigree and all data related to familiar cancer histories were collected by patient 25. After informed consent, peripheral blood samples of family members 9, 15, 18, 20, 21, 23, 25, and 28 were collected and used to isolate genomic DNA for further analysis.

Long-distance inverse PCR and genomic PCR experiments. Long-distance inverse PCR experiments were done as recently described (16). Genomic DNA was isolated from peripheral blood mononuclear cells. About 2 μg genomic DNA was digested with SpeI and NheI. An aliquot of 500 ng was religated in 150 μL and subjected to long-distance inverse PCR analysis. Different oligonucleotides were used to identify restriction polymorphic DNA fragments within the MLH1 exon 11/intron 11 region: MLH1•R4 (5′-ACCTTAGATAGTGGGAGGGGGAGAAAAAGC-3′) was used in combination with either MLH1•F5 (5′-AAGATTCATAAGTGGGTTCAGCTGGTGACA-3′), MLH1•F4 (5′-CCACACACTAAACAAGCAGTCAGTTCCAGA-3′), MLH1•F3 (5′-TGTTTGAGGATCTGAAAGTTGAGTGCAGTG-3′), or MLH1•F2A (5′-CAGTTAGGAAAGGGAGGGAAGAGACATGG-3′). PCR amplimers were cut out from the gel and subjected to sequence analysis.

The presence of the MLH1•ITGA9 allele was confirmed by using the following oligonucleotides: MLH1I11•3 (5′-GAGGAAGGGGAAGTGAGAAATTGTTTAATGG-3′) and ITGA9I16•5 (5′-CCAGTCAGTTATAGGATTAGGAGGGTAAAC-3′). The amount of genomic DNA was confirmed using the MLL-specific oligonucleotides MLL•F7 (5-TTGTGAGCCCTTCCACAAGTTTTGTTTAGAGG-3) and MLL•CR2 (5′-GTCCCAGGCACTCAGGGTGATAGCTGTTTCGG-3′).

The presence of all 19 MLH1 exons was analyzed using extracted DNA from paraffin embedded breast tumor and normal tissue of patient 25. PCRs for all 19 MLH1 exons were carried out using exon-specific primer combinations as previously published (17).

The presence of MLH1 exon 11, LRRFIP2 (leucine-rich repeat in flightless interaction protein 2) exon 28, GOLGA4 (Golgi autoantigen, golgin subfamily a, 4) exon 14, C3orf35/APRG1 (chromosome 3 open reading frame 35/AP20 region protein 1) exon 6, and ITGA9 exon 16 was verified with the following oligonucleotides: MLH1ex11•3 (5′-GGGCTTTTTCTCCCCCTCCC-3′), MLH1ex11•5 (5′-AAAATGTGGGCTCTCACG-3′), LRRFIP2•F1 (5′-TACGAACAGCACTGGACAAGAT-3′), LRRFIP2•R1 (5′-TAAAAGGGGTTTGTGTCAATGG-3′), GOLGA4•F1 (5′-CACCAGCAGCAAGTTGACAGTA-3′), GOLGA4•R1 (5′-TAGCAGATGCCTGCTGGAGTTC-3′), APRG1•F1 (5′-GACTGACTCAATCCCAGCTGCT-3′), APRG1•R1 (5′-TGCCAGAGCAATGCTGACCCAG-3′), ITGA9•F1 (5′- ATCGTGTTTGAAGCAGCCTACA-3′), and ITGA9•R1 (5′- TTTGGGCCATTATTTCTGCTCT-3′).

Cloning and expression of MLH1, ITGA9, and MLH1•ITGA9. The pcDNA3.1+ expression vector (Invitrogen) containing the entire open reading frame of MLH1 was a gift of Dr. Hong Zhang (Huntsman Cancer Institute, University of Utah, Salt Lake City, UT). cDNAs of ITGA9 and MLH1•ITGA9 were amplified from peripheral mononuclear cells of a healthy volunteer or family member 25, respectively. The obtained cDNA amplimers were cloned into the pcDNA3.1+ expression plasmid. Orientation and intact open reading frames were verified by sequence analyses.

Cell culture, transfection, and Western blot experiments. HEK 293T cells, obtained from Dr. Kurt Ballmer (Paul Scherer Institute, Villingen, Switzerland), were grown in DMEM with 10% FCS. Transient transfections were carried out using Gene Juice Transfection Reagent [Merck (Novagene) Schwalbach]. Total cell lysates (50 μg) were used for Western blot experiments along with 200 μg total protein isolated from peripheral mononuclear cells. For detection the following antisera were used: anti-MLH1 (clone LS-C566) raised against amino acids 387 to 403 of human MLH1 (Lifespan Biosciences), anti-ITGA9 (clone 3E4) raised against amino acids 785 to 887 from human ITGA9 (Abnova), and anti–β Actin (Sigma).

Mismatch repair assay. Because MutLα is not expressed in HEK 293T cells (18), these cells were transiently transfected according to standard procedures. Cells were harvested after 48 h; whole cell extracts, as well as nuclear extracts, were prepared; and mismatch repair reactions were done in vitro as described (19). Briefly, a substrate plasmid bearing a G-T mismatch within an AseI restriction site and a 3′-single-strand nick in 83-bp distance to the mismatch was used for in vitro DNA repair assays. Subsequent digestion with AseI was used to assess mismatch repair efficiency. Restriction digests were separated on 2% agarose gels and stained with ethidium bromide, and bands were quantified using Quantity One Software v4.6.1 (Bio-Rad).

Focus formation experiments. Focus formation experiments have been carried out using commercially available MEF/tTA cells (BD Clontech) as published recently (20, 21). Stably transfected cell lines expressing either empty vector, mutant RAS* (G6V mutation), or the MLH1•ITGA9 fusion protein were established. Briefly, MEF/tTA cells were grown in media containing 100 μg/ml G418. Transfection of pTRE2puro, pTRE2puro::RAS*, and pTRE2puro::MLH1•ITGA9 was done using the FUGENE 6 Transfection Reagent (Roche). Selection was started 48 h after transfection with 3 μg/mL puromycin, and cells were kept under selective conditions for a median time of 2 wks before single clones were isolated. Cell clones were tested by reverse transcription-PCR for their ability to induce transgene expression upon doxycycline withdrawal using the following oligonucleotides: RAS*•3 (5′-ATGACAGAATACAAGCTTGTGG-3′), RAS*•5 (5′-GCCAGGTCACACTTGTTGCCCA-3′), MLH1/ITGA9•3 (5′-GAATGGTTACATATCCAATG-3′), and MLH1/ITGA9•5 (5′-CCCAAAGCTAGATACAGGGTTA-3′). Transcriptional suppression of all transgenes was achieved by adding 10 μg/mL doxycycline to the media.

Mock-, MEF/tTA::RAS*-, and MLH1•ITGA9-transfected cell clones were seeded in small Petri dishes (1 × 10⁴ cells) and grown in the presence or absence of doxycycline for at least 21 d. All cells were washed once with PBS and then fixed with a 2% formaldehyde solution for 2 mins. The dishes were rinsed with water and stained with a 1 mg/mL methylene blue solution for 15 mins. Photographs were taken from sections of selected Petri dishes (magnification, ×100). All experiments have been repeated in triplicates.

Identification of the MLH1•ITGA9 fusion gene. A large family originating from French Guyana suffers from a variety of different tumors. This family was diagnosed for the Lynch syndrome after the detection of an MLH1 mutation in family member 9 by multiplex ligation-dependent probe amplification. The mutation was characterized as a deletion within the MLH1 gene, missing exons 12 to 19. The same deletion was verified by multiplex ligation-dependent probe amplification also in family member 25, the daughter of family member 9. The latter family member presented with a breast tumor at the age of 53, after she had already developed a sebaceous carcinoma at the age of 38. In addition, family member 25 was diagnosed to be positive for microsatellite instability. The available pedigree suggested that this MLH1 mutation represents a germline mutation that was transmitted over 5 generations (Fig. 1A). Twenty-two of 54 potential allele carriers were affected from different tumors involving endometrium, colon, liver, kidney, stomach, small intestine, pancreas, breast, brain, and sebaceous gland, or developed leukemia (Fig. 1B).

To characterize the genomic deletion in more detail, the MLH1 gene of family member 20 was characterized at the molecular level using a long-distance inverse PCR strategy. Because the genomic deletion was expected in vicinity of MLH1 exon 11, several oligonucleotides were designed to identify rearranged restriction fragments deriving from this genomic region. As shown in Fig. 1C, three restriction polymorphic PCR amplimers (lanes a-c) were obtained in four different PCRs. The smallest PCR amplimer (lane a) was subjected to DNA sequencing and revealed a genomic fusion between MLH1 intron 11 (at nt 5099 of the 5173-nt-long intron) and intron 16 (at nt 4334 of the 30074-nt-long intron) of ITGA9 (Fig. 1D). The fusion sequence contained a filler-DNA fragment of 13 nts that neither derived from MLH1 nor from ITGA9. Filler-DNA fragments at chromosomal fusion sites are a typical hallmark for nonhomologous end-joining DNA repair processes.

Verification of the MLH1•ITGA9 fusion allele in other family members. The observed genetic recombination resulted in the genetic fusion of MLH1 exons 1 to 11 and ITGA9 exons 17 to 28. The gene fusion was indicative of an interstitial deletion at 3p21.3 and encompasses an area of about 400 kb. The deleted area codes for three additional genes, LRRFIP2, GOLGA4, and C3orf35/APRG1, respectively (Fig. 2A). To verify these initial data, several members of the affected family were tested after informed consent to confirm the presence of the MLH1•ITGA9 fusion allele. Genomic DNA from family members 9, 15, 18, 20, 21, 23, 25, and 28 was isolated from peripheral mononuclear cells and tested in genomic PCR experiments using two oligonucleotides specifically detecting the genomic fusion between MLH1 intron 11 and ITGA9 intron 16. Seven of 9 tested family members were positively diagnosed for the presence of the MLH1•ITGA9 fusion allele (Fig. 2B, left lanes). A control PCR confirmed sufficient amounts of isolated genomic DNA (Fig. 2B, right lanes). Family members 9, 20, and 25 had already developed tumors (colon, endometrium, sebaceous gland, breast; marked with black symbols in Fig. 1A), whereas family member 28 exhibits a benign keratoacanthoma since birth. Except family members 21 and 24 (marked with open circles in Fig. 1A), all other investigated family members bear the identical MLH1•ITGA9 gene fusion (marked with asterisks in Fig. 1A). This confirmed that the novel MLH1•ITGA9 allele is germline transmitted in this Lynch syndrome family.

Deletion of the second MLH1 allele during tumor development. For family member 25, a breast tumor biopsy sample (90% tumor cells) was analyzed and compared with normal tissue of the same patient. Genomic PCR experiments revealed a complete loss of MLH1 exons 12 to 19 in the tumor biopsy sample (Fig. 3A). Thus, we concluded that the second MLH1 allele was deleted in the tumor cells. To analyze this secondary event in more detail, another biopsy sample of the same tumor, containing about 80% tumor cells, was analyzed for the homozygous deletion of LRRFIP2, GOLGA4, and C3orf35/APRG1, respectively. As shown in Fig. 3B, MLH1 exon 11 (present in two copies in normal and tumor cells) served as positive control. The other investigated genes were hemizygous in somatic cells (left), whereas tumor cells displayed the loss of all investigated genes (right). The observed faint PCR amplimers presumably derived from contaminating normal cells (containing still a hemizygous set of all investigated genes).

Expression of the MLH1•ITGA9 fusion protein in peripheral mononuclear cells of affected family member 25. The MLH1•ITGA9 fusion allele is driven by the MLH1 promoter, able to produce an MLH1•ITGA9 fusion mRNA, which can then be translated into the MLH1•ITGA9 fusion protein (Fig. 4A). The molecular weight of the fusion protein (85 kDa) is nearly identical to that one of the wildtype MLH1 protein (84 kDa) but differs from the ITGA9 protein (114 kDa). To verify the expression of the MLH1•ITGA9 fusion protein in somatic patient cells, MLH1, ITGA9, and the novel MLH1•ITGA9 fusion cDNA were cloned into an eukaryotic expression vector and transiently transfected into HEK 293T cells. All transfected cells (50 μg total protein) were analyzed together with peripheral mononuclear cells of family member 20 and a healthy volunteer (200 μg total protein). As shown in Fig. 4B, Western blot experiments confirmed the expression of recombinant ITGA9 (lane 2, 114 kDa), recombinant MLH1 (lane 3, 84 kDa), and the recombinant MLH1•ITGA9 fusion protein (lane 4, 85 kDa). Peripheral mononuclear cells of patient 25 also revealed a weak expression of the MLH1•ITGA9 fusion protein (lane 5, 85 kDa), also the expression of MLH1 (lane 5, 84 kDa), whereas a healthy volunteer showed only the expression of endogenous MLH1 protein but no expression of the ITGA9 protein in analyzed peripheral mononuclear cells. Therefore, we concluded that the MLH1•ITGA9 fusion gene is indeed expressed in somatic cells of affected individuals. Minor differences in the migration of transfected and endogenous MLH1•ITGA9 fusion protein may be due to (a) different amounts of protein used for the different lanes and (b) the expression of MLH1 and MHL1•ITGA9 that was simultaneously monitored in lane 5 when probed with an anti-MLH1 antibody.

MLH1•ITGA9 fusion protein is associated with mismatch repair deficiency. To analyze the MLH1•ITGA9 fusion protein for its mismatch repair capability, HEK 293T cells were transiently transfected with expression vectors coding for PMS2 in combination with either MLH1 or MLH1•ITGA9. Extracts of these transfected cells were prepared and used for an in vitro single-bp mismatch repair assay. As shown in Fig. 5A, cells expressing PMS2 and MLH1 are able to repair the single mispair of the substrate plasmid (faint band below linearized vector), whereas cells expressing PMS2 and the MLH1•ITGA9 fusion protein were not able to repair the substrate accordingly. Therefore, we concluded that the MLH1•ITGA9 fusion protein does not confer any mismatch repair capability. This is most likely due to its inability to bind to PMS1/2 (Fig. 4A).

Focus formation experiments. Stable cell lines transfected with empty vector, a mutant RAS* and MLH1•ITGA9 were grown either in the presence (no transgene expression) or absence of 10 μg/ml doxycycline (transgene expression). As shown in Fig. 5B, both transgenes were induced upon doxycycline withdrawal (left, lane f). A total of 10⁴ cells were seeded in small Petri dishes and grown for at least 21 days. Medium was exchanged every 3 days. Mock- and untransfected cells did not show any growth-transformed phenotype, whereas loss of contact inhibition was observed in a doxycycline-dependent manner for RAS*- and MLH1•ITGA9-expressing cells (right). The number of foci observed for MLH1•ITGA9 overexpressing cells was roughly half of the amount of foci observed for RAS* expressing cells.

We have investigated a Lynch syndrome family that suffers from a broad spectrum of tumors. The molecular analysis revealed an interstitial deletion of about 400 kb that fuses portions of the MLH1 (exons 1-11) with the ITGA9 gene (exons 17-28), partly disrupting the AP20 region at chromosome 3p21.3. The AP20 region is a genetic hotspot of homozygous deletions in renal, lung, and breast carcinomas (22), and has been implicated in major epithelial malignancies (23). Lynch syndrome patients diagnosed for a 3′-deletion of the MLH1 gene may represent individuals that carry interstitial deletions involving the AP20 region. Therefore, Lynch syndrome patients that exhibit an unusual tumor spectrum may be candidates for a molecular investigation of interstitial deletions at the AP20 region with the presented long-distance inverse PCR–based strategy. The established primer set (MLH1I11•3 and ITGA9I16•5) is highly specific for the described MLH1•ITGA9 fusion. They will be useful in the future to screen Lynch syndrome patients for the presence of that particular allele to validate the prognostic value of our finding.

Because of this germline deletion, loss of heterozygosity (LOH) was created for five genes (MLH1, LRRFIP2, GOLGA4, C3orf35/APRG1, and ITGA9). LOH of MLH1 will result in reduced mismatch repair activity. LRRFIP2 encodes a protein that has been identified as an activator of the canonical WNT signaling pathway and, in association with DVL3, exerts its activity upstream of CTNNB1/β-CATENIN (24). The GOLGA4 gene encodes a protein that has been postulated to play a role in Rab6-regulated membrane-tethering events in the Golgi apparatus (25). The C3orf35/APRG1 gene has been shown to exhibit a tumor suppressive function in breast cancer and is significantly down-regulated or lost in breast cancer and cervical carcinoma (26, 27). ITGA9 has been recently described as a candidate cancer gene that is overexpressed in human lung, breast, and colorectal cancer (14, 15). Noteworthy, ITGA9 is normally not expressed in peripheral mononuclear cells (28).

Unfortunately, we had no access to detailed clinical records of affected family members who belong to the first three generations. Nevertheless, some family members of the last two generations kindly provided their peripheral blood for further analyses. These analyses revealed that seven of nine tested family members were carriers of this particular fusion allele (family members 9, 15, 18, 20, 22, 25, 28). Four of them were already affected by tumors (endometrium, colon, small intestine, sebaceous adenoma, sebaceous carcinoma, breast cancer, and a benign keratoacanthoma), whereas family members 15, 18, and 22, although carriers, have no tumor history thus far. Tested family members 21 and 23 were negative and thus are not at risk to develop a Lynch syndrome because they do not carry the mutated MLH1 allele.

Although the family history fulfills the Amsterdam criteria II (29), the spectrum of tumors caused by the novel MLH1•ITGA9 gene fusion is not fully compatible with the spectrum of tumors normally observed in Lynch syndrome patients (1). The occurrence of sebaceous gland tumors is a typical feature of Muir-Torre syndrome patients (30). Breast (n = 4), liver (n = 2), and kidney carcinomas (n = 1) are generally not associated with the Lynch syndrome. Presumably, an impaired mismatch repair system in combination with either the deletion of AP20-region genes or a yet to show gain of function due to the presence of the MLH1•ITGA9 fusion protein may account for the unusual tumor spectrum in the investigated Lynch syndrome family.

Loss of the remaining MLH1 wildtype allele during tumor development was not unexpected. An experimental comparison of the genomic DNA isolated from paraffin embedded normal and tumor cells revealed also the loss of LRRFIP2, GOLGA4, and C3orf35/APRG1 at the second 3p21.3 allele. There are two possible explanations for this finding: it may (a) argue for an interallelic recombination event resulting in uniparental disomy (31, 32) or (b) be explained by a complete loss of the second chromosome 3. This should be investigated in more detail in the future, for example, by single-nucleotide polymorphism array analysis, because neither genomic PCR of genes surrounding the genomic fusion site nor microsatellite PCR will be informative. Genes surrounding the fusion site are still hemizygously on the derivative chromosomes, and microsatellite PCR experiments will not be of value in tumor cells that were positively tested to exhibit microsatellite instability (10).

Other important issues are potential functions of the MLH1•ITGA9 fusion protein. Because the MLH1•ITGA9 fusion protein does not exhibit a leader peptide, neither a membrane insertion nor the typical ITGA9-specific N-glycosylation can be expected. The remaining transmembrane domain and cytosolic tail present in the 3′-terminal portion of the MLH1•ITGA9 fusion protein may attract other cellular proteins, and thus, the presence of this particular fusion protein may provide complex effects on the biology of affected cells. Tumor cells without wildtype MLH1 alleles but at least one copy of the MLH1•ITGA9 fusion allele are unable to provide mismatch repair capabilities due to the missing PMS1/2 binding sites. Therefore, cells expressing this particular fusion protein have reduced ability to cope with genotoxic stress situations. Such a genetic situation will result in microsatellite instability and global genetic instability.

Because this is the first time that an MLH1 fusion protein has been identified in a Lynch syndrome family, we tested the hypothesis whether the fusion protein may provide a biological function. Focus formation experiments show that the doxycycline-dependent overexpression of the MLH1•ITGA9 fusion protein resulted in a loss–of–contact inhibition phenotype in murine embryonic fibroblasts. These preliminary data should be handled with care because the ectopic overexpression of many proteins displays similar phenotypes and is not necessarily representing a true biological phenotype. As a matter of fact, we used 200 μg of total cellular protein for the detection of the MLH1•ITGA9 fusion protein in peripheral mononuclear cells of one investigated family member. Thus, the fusion protein is only weakly expressed in somatic cells of affected patients. This may account for hyperplastic growth, which is presumably always a first step for any type of tumor developing along to a multistep carcinogenic pathway. Therefore, studies are currently being done, aiming to dissect the biological properties of this novel MLH1•ITGA9 fusion protein, which may provide clues to understand the clinical observations in more detail.

No potential conflicts of interest were disclosed.

We thank Prof. Michael Karas for critically reading the manuscript, Dr. Jochen Raedle and Dr. Brigitte Tandonnet for the medical help, and all family members who provided peripheral blood after informed consent.