Hyper-Recombination and Genetic Instability in BLM-Deficient Epithelial Cells

The human Bloom syndrome gene (BLM) encodes a homologue of the Escherichia coli RecQ DNA helicase (1, 2). DNA helicase unwinds double-stranded DNA molecules, a process required for various aspects of DNA metabolism, including transcription, DNA repair, and replication (3). Mutations in the BLM homologues of mouse, Drosophila, yeast, E. coli, and Caenorhabditis elegans all cause a pronounced genomic instability phenotype manifest by gross chromosomal rearrangements, chromosomal nondisjunction, elevated levels of somatic recombination, and losses of heterozygosity (LOH). In Drosophila, mutations at the mus309 locus (DmBLM) result in 10-fold increased levels of nondisjunction as well as in whole chromosome loss (4). In Sgs1 mutant Saccharomyces cerevisiae, 20-fold increases in gross chromosomal rearrangements as well as elevated rates of homologous recombination have been observed (5, 6). Saccharomyces pombe mutants for Rqh1 are endowed with a hyper-recombinogenic and chromosome nondisjunction phenotype (7). Data from mouse models display all of the above mentioned forms of instability (8, 9). Gross chromosomal rearrangements have been reported in a few hematological neoplasms of Bloom syndrome patients who carry a defective BLM protein (10). Moreover, the incidence of exfoliated epithelial cells containing micronuclei is elevated in Bloom syndrome patients compared with individuals with a heterozygous BLM gene mutation status (11). BLM heterozygosity has been found to increase the risk for colorectal cancer in the Ashkenazi population (12). Colorectal adenomas from a Bloom syndrome patient have previously been analyzed cytogenetically (13). Interestingly, the great majority of cells displayed a diploid karyotype. This is in stark contrast to non-Bloom’s adenomas, which generally tend to show chromosomal gains/losses even at an early stage (14). This unexpected finding stimulated us to rigorously evaluate chromosomal instability in human epithelial cells in a well-defined experimental system.

HCT116 cells (American Type Culture Collection, Manassas, VA) were used to generate Blm−/− cell lines. Targeting was performed using the pFred two-vector targeting system described previously in Jallepalli et al. (15). Briefly, the left homology arm was amplified using primers L1 and L2, and cloned into SalI and EcoRI sites of the pFred-B construct. The right homology arm was amplified with R1 and R2, and cloned into BamHI and NotI sites of the pFred-A construct. HCT116 cells were cultured in McCoy’s 5A Medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. Using LipofectAmine (Invitrogen, Carlsbad, CA) HCT116 cells were transfected with the two plasmids described above. Clones were selected after 3 weeks of growth under 0.4 mg/ml geneticin (Invitrogen) selection. Recombinant clones were identified by PCR screening. Primer sequences for the left and right homology arms (lowercase sequences contain restriction sites for cloning): L1: 5′-aataattagtcgacTTTTAGCAAATTGGTGACATGA-3′; L2: 5′-attatgaattcGGTCCTCAAAGTTGTCCAGAAA-3′; R1: 5′-attaattggatccCAGATAAGTTTACAGCAGCAGC-3′; and R2: 5′-ataagaattatgcggccgcTTGAAATTGGGGTGGAAGGAC-3′.

Cells were grown in 50 μm bromodeoxyuridine for ∼35 h (16). After bromodeoxyuridine labeling, cells were incubated with 0.1 μg/ml Colcemid for 5 h. Cells were subjected to standard 0.075 M KCl hypotonic treatment and fixation with cold methanol/glacial acetic acid (3:1). Cells were dropped onto slides and stained with 0.5 μg/ml Hoechst 33258 in PBS for 5 min. Slides were mounted in MacIlvaine’s buffer, and SCEs were assessed through standard fluorescent microscopy.

The quantitative analysis of chromosomal instability using fluorescence in situ hybridization (FISH) analysis with chromosome-specific centromeric probes was performed as described previously by Lengauer et al. (17). Multiplex-FISH was performed as described previously by Speicher et al. (18). Metaphase spreads were prepared by treating cells with 0.1 μg/ml colcemid (KaryoMax; Invitrogen). Micronuclei were determined after staining cells with Hoechst 33258 as described above.

Site-directed homologous recombination was measured using two different constructs, which vary by an order of magnitude in their recombination frequency. The constructs tested were p53-ATG-neo (19) and pDnmt-hygro (20). Transfections and culturing conditions were performed as described above with the exception of selection for pDnmt1-hygro clones with 0.1 mg/ml hygromycin. Recombinants were identified through PCR screening.

LOH was determined through the typing of 16 diallelic markers identified previously by Weber et al. (21). Typing of single clones was performed using fluorescently labeled primers for each marker in a PCR reaction. PCR products were resolved on a SpectruMedix SCE-9610 (SpectruMedix, State College, PA).

To evaluate genomic instability in a major site of neoplasia in Bloom syndrome (12, 22), we genetically inactivated the BLM locus in the human colorectal cancer cell line HCT116 (Fig. 1 A). These epithelial cells have been shown previously to have a near-diploid karyotype and no measurable chromosomal instability (17). Knockout cells (BLM−/−) from three different clones had growth characteristics indistinguishable from parental cells (data not shown). Two of the BLM−/− clones were characterized in detail with respect to genomic instability.

It has been reported that mouse cells with a deficient BLM gene frequently show gross chromosomal rearrangements (9). To evaluate such changes in epithelial and fibroblast cells deficient in BLM, we performed multiplex-FISH karyotyping (Ref. 18; Fig. 2). Twenty-one of 24 studied metaphases of parental HCT116 cells displayed a clonal population with the characteristic karyotype of 45,X,-Y, der(10)dup(10)q24q26)t(10;16)(q26;q24), der(16)t(8;16)(q13;p13), and der(18)t(17;18)(q21;p11.3; Ref. 23). The heterozygote clone had a similarly stable karyotype (38 of 40 metaphases with characteristic karyotype) as did the two homozygote knockout clones (40 of 45 and 30 of 37 metaphases with the characteristic karyotype, respectively). These results were markedly different from those obtained after disruption of the hSecurin gene in the same cells, where only 10 of 40 cells retained the characteristic karyotype (P < 0.005; Refs. 15, 17; Fig. 2). We conclude that the prevalence of gross chromosomal rearrangements is not appreciably changed by disruption of the BLM gene in colorectal epithelial cells.

To evaluate the rate of chromosomal gains and losses in these cells, we analyzed interphase nuclei from parental, heterozygote, and homozygote cells after passage for 50 generations (17). Karyotypic losses and gains were assessed by counting the number of centromeres per nucleus for chromosomes 7, 12, 17, and X. The numbers of losses and gains were equally low for wild-type and BLM−/− homozygous knockout cells; the fraction of cells with centromere numbers different from the modal number was <5% for all of the chromosomes tested in HCT116 parental cells and for 7 of 8 chromosomes tested in two different BLM−/− knockouts clones (Table 1 A). As a positive control, we again evaluated HCT116 cells with an inactivated hSecurin gene. Almost half of the hSecurin-deficient cells (49%) exhibited centromere numbers different from the modal number (P < 10⁻⁶; Refs. 15, 17). In addition, we examined Bloom syndrome patient fibroblasts (GM08505C) together with matching control fibroblasts (GM00637H, both from the Coriell Cell Repository) for whole chromosome gains and losses. Interestingly, the percentage of cells with centromere numbers different from the mode in BLM fibroblasts was significantly higher (9%) than in control fibroblasts (2%; P < 10⁻⁶) but not as high as in cancer cells with chromosomal instability (17).

Cells were also assessed for the presence of micronuclei, implicated previously in genomic instability (11). Micronuclei levels were elevated in the BLM−/− cells compared with BLM+/+ controls. The BLM−/− clones demonstrated 5.8% and 6.5% micronuclei versus 1.2% in parental controls (P < 10⁻⁸). BLM patient fibroblasts demonstrated 11.5% micronuclei versus 4.0% control fibroblasts (P < 0.005). hSecurin−/− cells conversely displayed 13.5% and 16.5% micronuclei in two independently derived clones compared with 1.2% in parental HCT116 cells (15).

We then evaluated SCEs, the prototypical feature found in Bloom syndrome mesenchymal cells (Ref. 24; Fig. 1,B). All of the HCT116 BLM−/− clones displayed a significantly elevated number of SCEs. The knockout cells contained an average of 11 SCEs events per metaphase, compared with 5 and 4 SCEs/metaphase in BLM+/− and wild-type BLM+/+ cells, respectively (Table 1 B).

Increased levels of SCEs have been directly correlated with increased rates of targeted homologous recombination (9, 25). To test and compare rates of homologous recombination in the BLM−/− cells, we transfected them with targeting constructs for two different human genes (19, 20). The two genes chosen for these experiments had been shown previously to have targeting efficiencies that varied by 1 order of magnitude in parental HCT116 cells. Transfection of ∼2 × 10⁷ parental cells with a DNMT1 targeting construct yielded ∼1 targeted clone per 100 geneticin-resistant clones, whereas transfection with a p53 targeting construct gave rise to ∼1 positive clone per 1100 clones tested. These frequencies were in line with earlier studies (19, 20). However, homologous recombination in the BLM−/− background was elevated for both genes by >15 fold (P < 10⁻⁵ and P < 0.0002; Table 1 C).

Increased rates of homologous recombination have been associated with LOH events in mouse tumor models (9). Also, prior work has demonstrated an elevation in LOH in single lymphoblastoid clones derived from a Bloom’s patient (26). Therefore, we sought to measure the rate of background LOH in BLM-deficient cells. The analysis was performed using 16 heterozygous diallelic markers distributed along chromosomes 2 and 3 (Table 2). These markers were typed in 95 individual BLM−/− and 190 BLM+/+ clones. Eight (8 of 95) noncontiguous LOH events were observed in the BLM−/− clones and 1 (1 of 190) in the BLM+/+ clones (P < 0.0003).

It was surprising that the major manifestations of instability we found in BLM−/− colorectal cancer epithelial cells involved only those known to be associated with increased homologous recombination. The rates of whole chromosome gains and losses were essentially unaffected, and gross cytogenetic structural abnormalities could only rarely be identified. In contrast, there was a marked increase in LOH that did not involve whole chromosomes in SCE and in targeted homologous recombination. Hyper-recombinogenic states may prove to be a major mechanism for chromosomal instability in other hereditary and sporadic cancer types. However, our data show that such increased levels of recombination do not necessarily result in elevated levels of structural and numerical chromosome changes. Data supporting the idea that hyper-recombination is responsible for promoting tumorigenesis in BLM−/− cells has been obtained in BLM mouse models (9).

Apart from the implications for the pathogenesis of Bloom syndrome, the observations reported above have practical applications. The higher rate of recombination found in human BLM−/− HCT116 cells should prove useful for creating human somatic knockouts, an otherwise time-consuming, expensive, and labor-intensive task (27). Reversible knockout of the BLM gene, using cre-lox technologies, could be applied to any suitable cell type, such as human stem cells. Our results show that BLM-deficient human cells can be karyotypically stable, thereby facilitating subsequent evaluations and applications of the engineered clones.

We thank Leslie Meszler from the Cell Imaging Core Facility for help with imaging. We also thank Kurt Bachman and Harith Rajagopalan for helpful discussion and technical assistance.