Recql5 and Blm RecQ DNA Helicases Have Nonredundant Roles in Suppressing Crossovers

Construction of Recql5 targeting vectors.

The recombineering technique (35, 64) was used to construct the Recql5 targeting vectors. First, a 2.4-kb genomic sequence was amplified by PCR and used as a probe to screen the RPCI-22 129S6/SvEv bacterial artificial chromosome (BAC) library. The primers used were Q5e3f (5′-ATGGCGCGCCGAGATGGCAGCTTCAGCCTCCTTCCA-3′) and Q5e5r (5′-ATGGCGCGCCCGGACATTAGCTTTGTCTACTCCCAT-3′). Clone RPCI-22-51D24 was found to contain the entire Recql5 gene by PCR and Southern blot analysis (data not shown) and was chosen as the starting material for constructing Recql5 targeting vectors. Subsequently, a targeting strategy and the first targeting vector were designed (Fig. 1a). In this strategy, a 7.3-kb genomic fragment was used to construct a targeting vector, with which a conditional Recql5 knockout allele was created. In this allele, exon 4 of Recql5 was replaced by a loxP-exon4-loxP-PGKNeo-loxP cassette. To construct this vector, an RPCI-22-51D24 BAC clone was introduced into DY380 cells by electroporation. In the meantime, a retrieval vector, pQ5RevTK, was constructed. In this vector, two small genomic fragments, 400 bp each and corresponding to the 5′ and 3′ ends of the 7.3-kb genomic fragment, respectively, were amplified by PCR and cloned into pMc1TK. The primers used were as follows: Q5RevP1, 5′-GCGTGATCTAGAAACCCCAAAAAGATGATGAGG-3′; Q5RevP2, 5′-GCGGAATTCGCTAGCATTCCTGAGGTAAGACA-3′; Q5RevP3, 5′-GCGGAATTCGGATCCAGTATTTGTCCTGACCTC-3′; and Q5RevP4, 5′-CGTCGAGCGGCCGCACTCCATAGTGGGCCAG-3′. These two genomic fragments provided the homologies for the recombination between the retrieval vector and the BAC clone, whereas the Mc1TK cassette was used as a negative selection marker in the gene targeting experiments in mouse ES cells. The retrieval vector was linearized by BamHI and electroporated into DY380 cells containing the RPCI-22-51D24 BAC clone. Induced recombination inside the DY380 cells resulted in the retrieval of a 7.3-kb genomic fragment from the BAC clone onto the retrieval vector, generating an intermediate vector, pQ5TK. Then, a modification cassette was created. First, a loxP-exon4-loxP-PGKNeo-loxP cassette was constructed. In the meantime, two small genomic fragments, 400 bp each and corresponding to the 5′ and 3′ ends of exon 4, respectively, were amplified by PCR and cloned into pBluescript (Stratagene). The primers used were Q5BMP1, 5′-CGGAATTCCGGCGCGCCCTTGGACAGAAGGCTGAGAACGGG-3′; Q5BMP2, 5′-GCTCTAGAGATAACTTCGTATAATGTATGCTATACGAAGTTATCAGCCCACTCACTAGAGCACAGCA-3′; Q5BMP3, 5′-GCTCTAGAGCTGATTGTACCACCCTTCCCCTAG-3′; and Q5BMP4, 5′-GTGGCGCGCCAAGGGTACCTGTTCTGTTGTGATCCA-3′. Insertion of the loxP-exon4-loxP-PGKNeo-loxP cassette between the two genomic fragments then gave rise to the plasmid pQ5mod containing the 5′arm-loxP-exon4-loxP-PGKNeo-loxP-3′arm modification cassette. This cassette was then excised from pQ5mod and introduced into DY380 cells together with pQ5TK. Induced recombination between the homologous sequences in pQ5TK and the modification cassette led to the replacement of exon 4 with the loxP-exon4-loxP-PGKNeo-loxP cassette and gave rise to the first targeting vector, pRecql5tv1 (Fig. 1a). This vector was partially sequenced to confirm that no unintended mutations were created in exon 4 during these manipulations. Two other targeting vectors, pRecql5tv2 (Fig. 2a) and pRecql5tv3 (information available upon request) were generated by replacing the PGKNeo cassette of pRecql5tv1 with a PGKHprt minigene cassette and a PGKPuro cassette, respectively.

Gene targeting in mouse ES cells.

Gene targeting experiments in the AB2.2 ES cells were conducted as described previously (37). For all the clones used in experiments, correct targeting events were verified by Southern hybridization. All ES clones were subsequently reseeded at low densities to obtain single-cell-derived clonal lines.

Mouse work.

Two ES clones heterozygous for the M2 Recql5 conditional knockout allele (+/M2) were used to generate chimeras and to obtain germ line transmission of the conditional knockout allele as previously described (38). Zp3-Cre transgenic mice (8) were used to excise exon 4 in the conditional allele of Recql5 in mouse oocytes to generate heterozygous Recql5 knockout mice (+/M3). Recql5-targeted alleles were maintained in an F2 mixed genetic background between 129/SvEv and C57BL/6J inbred strains.

Cell culture and analyses.

ES cells were cultured in Iscove's modified Dulbecco's medium (Invitrogen) containing 15% fetal bovine serum as previously described (38). For cell growth analysis, 10⁴ cells were plated in each well of six-well feeder plates. The number of cells in each well was counted 48 h after plating and every 24 h afterwards. Standard colonogenic survival assays were used to assess sensitivities to gamma radiation. Briefly, ES cells were trypsinized into single-cell suspensions and then exposed to different doses (0 to 10 Gy) of radiation from a Cs-137 irradiator. After irradiation, cells were plated onto six-well feeder plates at 10⁴ cells/well to allow formation of single-cell-derived colonies. Eight days after plating, colonies that were at least 1 mm in diameter were counted. Experiments were carried out in triplicate and repeated twice. In each experiment, appropriate controls for evaluating plating efficiencies of individual cell lines were also included. Statistical analysis was performed using Prism software (GraphPad Software).

RT-PCR.

Total RNA was isolated from ES cells grown in gelatinized plates using TRIzol (Invitrogen) according to the instructions provided by the manufacturer. All reverse transcription-PCRs (RT-PCRs) were carried out using Superscript II RT/Platinum Taq mixture (Invitrogen). Primer pair Q5e3f1 (5′-ATGGCGCGCCCGGACATTAGCTTTGTCTACTCCCAT-3′)-Q5e5r1 (5′-ATGGCGCGCCCGGACATTAGCTTTGTCTACTCCCAT-3′) was used in studying the expression of wild-type and targeted Reclq5 alleles. These primers are specific to exons 3 and 5, respectively. To quantify the expression of Recql5 in BAC-rescued Recql5 knockout ES cells by quantitative RT-PCR, primer pair Q5e3f2 (5′-GGCAATCTGAGGGACTTCTGCCT-3′)-Q5e4r1 (5′-CTGCGTGGTAAGCCTTGGCGTTC-3′) was used. Q5e4r1 is specific to exon 4, which is deleted in the Recql5 knockout allele. Thus, RT-PCR using this primer pair results in the amplification of transcripts from only the wild-type Recql5 and not the knockout alleles. The primers used for amplifying the internal control Gapdh are Gapdh.f1 (5′-GTGCTGAGTATGTCGTGGAGT-3′) and Gapdh.r1 (5′-CACACACCCATCACAAACATG-3′). Semiquantitative, low-cycle-number RT-PCR amplification with ³²P-labeled primers (40) was used to quantify the expression of Recql5 transcripts. Specifically, for each RT-PCR, first-strand synthesis of cDNA was carried out using 1 μg of the total RNA. After that, ³²P-labeled 5′ primers were added, and the PCRs were allowed to proceed for various numbers of cycles which were expected to be within the linear range of target amplification according to the results of trial experiments. The RT-PCR products were then fractionated on a nondenatured polyacrylamide gel. After electrophoresis, the gel was dried and analyzed with a PhosphorImager (Molecular Dynamics). Data from the RT-PCRs within the linear range were used to calculate the relative levels of Recql5 expression.

SCE analysis.

The SCE analysis was carried out using a classical 5-bromo-2-deoxyuridine (BrdU)-labeling protocol (17) with a minor modification: the concentrations of BrdU were reduced to 1.5 μg/ml for ES cells and 3 μg/ml for MEF cells, respectively. Also, to minimize the variations between experiments, all cell lines involved in the large-scale SCE analyses were processed simultaneously in a single experiment. Sister chromatids were visualized by staining with acridine orange (Sigma) solution (0.1 mg/ml). Specifically, to count the number of SCE, images of metaphase spreads were captured, and SCE of individual cell lines were scored blind. Statistical analysis was performed with Prism software (GraphPad Software).

Inactivation of Recql5 in mouse ES cells by gene targeting.

Human RECQL5 has been shown to express three different transcripts encoding three RecQ DNA helicase isoforms (α, β, and γ) (29, 51). The three isoforms share the conserved helicase domain encoded by exons 2 to 6 (45), but only RECQL5β, the longest isoform, contains a nuclear localization signal. The mouse Recql5 gene is very similar to its human homologue; however, its expression has not been well characterized. Thus, to ensure the functional inactivation of the mouse Recql5 gene, we decided to delete its 106-bp exon 4 to create a frameshift mutation at codon 260. Furthermore, in anticipation of the possibility that the deletion of Recql5 might lead to lethality, we created a conditional Recql5 knockout allele. Specifically, we replaced exon 4 of Recql5 with a loxP-exon4-loxP-PGKNeo-loxP cassette to obtain a primary Recql5-targeted allele, the M1 allele (Fig. 1a). Following that, a conditional knockout allele (the M2 allele) and a true knockout allele (the M3 allele) were derived from this M1 allele by Cre-mediated deletions (Fig. 1a). ES cell clones carrying the desirable Recql5 alleles were identified either by Southern (Fig. 1b and c) or by PCR (Fig. 1d) analysis. Two targeted ES cell clones with the M2 conditional knockout allele were used to generate conditional knockout mice (see below and Fig. 7). In the meantime, two clones with the M3 allele were used to generate homozygous Recql5 knockout ES cells by deletion of the remaining wild-type Recql5 allele with a different targeting vector, pRecql5tv2 (Fig. 2a). This vector was designed to replace exon 4 of Recql5 with a PGKHprt minigene cassette, leading to another Recql5 knockout allele, the M4 allele (Fig. 2a). Correctly targeted ES cell clones were identified by Southern analysis (Fig. 2b and c).

RT-PCR analysis showed that the homozygous Recql5 knockout (M3/M4) ES cells no longer express the wild-type Recql5 mRNA, but they do express an aberrant transcript (Fig. 2d). Sequencing analysis showed that this aberrant transcript originated from the M3 allele and is expected to encode a truncated polypeptide of 260 amino acids consisting of the first 259 amino acids of the 981-amino-acid Recql5β helicase (45) plus an additional alanine residue. Thus, this M3 allele is expected to produce either an N-terminal truncated polypeptide without any discernible domain or no protein product at all because of a nonsense-mediated decay. Therefore, this allele likely represents a null allele without any dominant negative effect. These data also indicate that the M4 allele does not express any detectable transcripts. Thus, these Recql5 knockout (M3/M4) ES cells represent Recql5 null ES cells. They are referred to as Recql5^−/− ES cells hereafter.

Recql5^−/− ES cells have an elevated frequency of SCE and an increased incidence of multiradial structures.

Recql5^−/− ES cells have growth characteristics similar to those of their parental wild-type ES cells (Fig. 3a and data not shown). The major objective of this study was to investigate whether Recql5 plays a role in suppressing crossovers in mitotic cells. Deletion of a gene can affect the frequency of crossovers by affecting either the recombination repair machinery or the regulation of recombination. Thus, we first assessed the repair capacity of the recombination repair machinery in Recql5^−/− ES cells by comparing their sensitivity to gamma irradiation with that of its parental cell line. The result showed that Recql5^−/− ES cells and its parental line have similar sensitivities to gamma irradiation (Fig. 3b), indicating that deletion of Recql5 does not significantly impair the recombination repair capacity in mouse ES cells.

To investigate whether Recql5 is involved in the suppression of crossovers in mouse ES cells, we first compared the frequencies of SCE among wild-type, Recql5^+/−, and Recql5^−/− ES cells. Initially, 20 metaphase spreads derived from wild-type, two Recql5^+/−, and two Recql5^−/− ES cell lines were examined. This experiment showed that wild-type and Recql5^+/− ES cells have similar SCE frequencies. However, the two Recql5^−/− ES cell lines have similarly higher frequencies of SCE that are about three times higher than those of wild-type and Recql5^+/− ES cells (data not shown). Subsequently, we conducted a large-scale experiment to compare the frequencies of SCE in wild-type and Recql5^−/− ES cells. As a positive control, we also included Blm^−/− ES cells (19, 37) in this experiment. Specifically, the blm (M3-M4) ES cells (19) were used. These ES cells express 12% of the wild-type Blm protein from the M3 hypomorphic allele (41). These were also the cells used to generate Blm Recql5 double-knockout ES cells (see below). The results showed that Recql5^−/− ES cells indeed have a frequency of spontaneous SCE that is significantly higher than that of the wild-type control (P < 0.001) but is similar to that in Blm^−/− ES cells (Fig. 4a to d and Table 1). These data clearly indicate that the deletion of the Recql5 gene resulted in a significant increase in the rate of crossovers between sister chromatids.

In cells from both Bloom syndrome patients and Blm knockout mice, elevated frequencies of SCE are accompanied by an increased incidence of multiradial structures (16, 41), which are indicative of interchromosomal exchange. Consistent with these previous observations, we observed six such structures from 200 metaphase spreads derived from Recql5^−/− ES cells (Fig. 4e and f) but none from >500 spreads from wild-type ES cells. Thus, it appears that the increase in the frequency of SCE in Recql5^−/− ES cells is also accompanied by an increased frequency of interchromosomal exchange. However, the frequencies of other types of chromosome aberrations, such as chromsome-chromatid breaks, are not significantly different between Recql5 knockout and wild-type ES cells (data not shown).

The SCE phenotype of Recql5^−/− ES cells is due to the inactivation of Recql5.

In chicken DT40 cells, deletion of the RECQL5 homologue does not affect the rate of spontaneous SCE (57). This raised the question as to whether the SCE phenotype seen in Recql5^−/− ES cells was indeed due to the loss of Recql5 or some other unintended effects that were brought about when this Recql5^−/− ES cell line was generated. To address this issue, we attempted to rescue the SCE phenotype of Recql5^−/− ES cells by using a genetic complementation strategy. We argued that if the SCE phenotype of our Recql5^−/− ES cells was indeed caused by the loss of Recql5 function, reintroduction of a single copy of a functional Recql5 gene into Recql5^−/− ES cells would be sufficient to rescue the SCE phenotype, since Recql5^+/− ES cells have a normal frequency of SCE.

To ensure the proper expression of the Recql5 gene, we introduced a Recql5-containing BAC vector that contains the coding region plus both 5′ and 3′ regulatory sequences of the Recql5 gene into Recql5^−/− ES cells. We obtained a Recql5-containing BAC clone, RPCI-22-51D24, by screening a mouse BAC library. This BAC clone contains the entire Recql5 coding sequence plus more than 10 kb of both 5′ and 3′ regulatory sequences (data not shown). A PGKNeo selection marker was introduced into this BAC clone by the recombineering technique (64). Then, this modified BAC clone was introduced into Recql5^−/− ES cells to obtain stable transfected lines (designated Recql5^−/− Res cells) (data not shown). RT-PCR analysis using total RNA isolated from two of these Recql5^−/− Res clones showed that they both expressed wild-type Recql5 transcript as expected (data not shown and see below).

Initial SCE analysis of 20 chromosome spreads from two of these clones showed that they have frequencies that are much lower than their parental Recql5^−/− ES cells, but similar to that of the wild-type control. Thus, a large-scale analysis was carried out to compare the frequencies of SCE among one of these two rescued (Recql5^−/− Res) ES cell line, the parental Recql5^−/− cells, and the wild-type ES cells. The results of this study confirmed that the SCE of the Recql5^−/− Res cell line is indeed rescued to the level of wild-type ES cells (Fig. 5b). However, it remained possible that the reduction of SCE frequency in the Recql5^−/− Res cell line was caused by a nonspecific effect, due to overexpression of Recql5. To address this issue, the expression of Recql5 transcripts in this Recql5^−/− Res cell line and in wild-type ES cells were quantified and compared. The results of semiquantitative RT-PCR experiments showed that Recql5 was not overexpressed in Recql5^−/− Res cells; instead, it was expressed at only about 5% of that in the wild-type ES cells (Fig. 5c). Thus, these data provide the definitive proof that the SCE phenotype of Recql5^−/− ES cells is caused by the loss of Recql5 function.

Recql5 and Blm have nonredundant roles in the suppression of crossovers in mouse ES cells.

The observation that deletions of Blm or Recql5 led to similar SCE phenotypes in mouse ES cells raised the question as to whether these two RecQ helicases are both required in the same genetic pathway or act independently through different pathways to suppress crossovers in mouse ES cells. To address this issue, we examined the effect of the deletion of Recql5 on the frequency of SCE in Blm knockout ES cells.

To generate Blm^−/− Recql5^−/− ES cells, we inactivated the two copies of the Recql5 gene in Blm^−/− ES cells, the blm (M3/M4) ES cells (19), by two consecutive rounds of gene targeting experiments. First, we generate a new Recql5-targeted allele in these Blm^−/− ES cells, the M4 allele, by the same strategy described previously for a similar experiment with wild-type ES cells (Fig. 2a) but with inactivation of the other Recql5 allele by another strategy which replaces exon 4 of Recql5 with a PGKPuro cassette (information available upon request). These experiments resulted in the generation of several Blm^−/− Recql5^−/− ES cell lines. Blm^−/− Recql5^−/− ES cells showed growth characteristics similar to those of wild-type ES cells (data not shown). RT-PCR using a pair of primers specific for exon 3 and exon 5, respectively, failed to detect any Recql5 transcript from these cells (information available upon request), indicating that neither mutant allele produces any detectable amount of Recql5 transcripts.

We then examined the effect of the deletion of Recql5 on the frequency of SCE in Blm^−/− ES cells. Specifically, we compared the frequencies of SCE among the Blm^−/−, Recql5^−/−, and Blm^−/− Recql5^−/− ES cells. To ensure the accuracy of the experiments, the SCE frequencies of all the cell lines described in this paper were derived from cells that were prepared simultaneously in one experiment. This study showed that Blm^−/− Recql5^−/− ES cells have a very high frequency of SCE. In some cells, the presence of many SCE events on some chromosomes made it difficult to accurately measure the number of SCE (Fig. 6a). Nevertheless, even based on an anticipated underestimation, the overall frequency of SCE of the double-knockout ES cells was significantly higher than that of either Blm^−/− or Recql5^−/− ES cells (P < 0.001 in both cases) (Fig. 6b). In particular, the percentage of cells with more than 40 SCEs (e.g., an average of 1 SCE per chromosome) is much higher in the Blm^−/− Recql5^−/− ES cells (44%) than in Blm^−/− ES cells (9.3%), Recql5^−/− ES cells (10.7%), and wild-type ES cells (<0.1%) (Fig. 6c). Therefore, these data clearly showed that in the mouse ES cells, Blm and Recql5 have nonredundant roles in the suppression of crossovers.

Primary embryonic fibroblast (MEF) cells derived from Recql5 knockout mice exhibited an elevated SCE frequency.

Our data from experiments using knockout ES cells clearly showed that Recql5 plays an important role in the suppression of crossovers in mouse ES cells. However, mouse ES cells are unique in many aspects of their cell cycle checkpoint controls (1), as well as DNA repair and recombination (10, 11). Thus, a mutant phenotype observed in knockout ES cells might not indicate a similar phenotype in differentiated cells of the corresponding knockout mice. To examine whether Recql5 also has a role in suppressing crossovers in differentiated mouse cells, we created Recql5 knockout mice and studied the frequencies of SCE in differentiated cells derived from Recql5 knockout mice.

We first used ES cells that carry the Recql5 conditional knockout allele and the M2 allele (Fig. 1a) to generate a Recql5 conditional knockout mouse model. Two M2 allele-containing ES cell lines, Recql5^tm2Luo-1, and Recql5^tm2Luo-2, were used to generate mice carrying the M2 conditional knockout allele. Germ line transmission of the M2 conditional knockout allele was obtained from both cell lines (Fig. 7a and b and data not shown). We then introduced this Recql5 conditional allele into female Zp3-Cre transgenic mice (8) to facilitate the deletion of exon 4 of Recql5. Removal of the exon 4 of the Recql5 gene by Cre-mediated deletion in the oocytes of Zp3-Cre transgenic mice produced heterozygous Recql5 knockout mutant mice carrying the M3 allele as designed (Fig. 7a and c). Heterozygous Recql5 knockout mutant mice are viable, fertile, and indistinguishable from their wild-type littermates (data not shown). Intercrosses among these heterozygous mutants gave rise to progeny, including homozygous mutants, at the expected ratio (Fig. 7d and data not shown). To date, 50 homozygous Recql5 knockout mice have been monitored for 3 to 7 months. These mice do not appear to have any gross developmental abnormalities. They also enjoy normal postnatal growth and development and have normal fertility (data not shown). Detailed phenotypic characterization of these mice is in progress. In the meantime, these knockout mice provide a source for differentiated Recql5 knockout cells to study the effect of Recql5 deficiency on the rates of crossovers in differentiated mouse cells.

Primary MEFs provide a readily available source of differentiated cells from knockout mouse models that are also compatible with SCE analysis. Thus, to examine whether Recql5 plays a role in suppressing crossovers in differentiated mouse cells, we compared the frequencies of MEFs derived from three wild-type and four Recql5 knockout embryos. The results of this study showed that Recql5 knockout MEFs have a significantly elevated frequency of SCE compared with that of wild type MEFs (P < 0.0001) (Fig. 8). Therefore, Recql5 is also required for the suppression of crossovers in differentiated MEF cells.

Role of Recql5 and its homologues in the suppression of crossovers in mitotic cells.

The observation that human RECQL5 physically interacts with both TOPOIII alpha and TOPOIIIβ prompted the speculation that RECQL5 and its homologues might play important roles in the suppression of crossovers. This was challenged by the finding that deletion of the chicken RECQL5 homologue in DT40 cells did not lead to an increase in the rate of spontaneous SCE (57). In contrast, we have shown here that the deletion of Recql5 alone in mouse ES cells is sufficient to cause a significant increase in the frequency of spontaneous SCE, comparable to the increase caused by the deletion of Blm. Furthermore, this elevated SCE frequency is accompanied by an increased incidence of multiradial structures, which are indicative of interchromosomal exchanges. The SCE phenotype observed in Recql5 knockout ES cells does not appear to be caused by either a dominant negative effect or gain-of-function effect due to the targeted mutation, since heterozygous knockout ES cells do not have this phenotype. Most importantly, the results from our genetic complementation experiment clearly demonstrated that the elevated SCE frequency is caused by the loss of Recql5. At present, we do not have a clear explanation for this apparent discrepancy between our findings and that based on DT40 cells (57). This could reflect either cell-type-specific or species-specific differences between chicken DT40 cells and mouse ES cells. However, our observations of similar SCE phenotypes in two different Recql5 knockout cells, ES cells and MEFs, argue against a cell-type-specific difference. Further experiments are needed to resolve this issue. Importantly, our findings confirm the speculation that multiple pathways exist for suppressing crossovers in mitotic cells, which should be expected given the great complexity of the mammalian genome and the potential deadly consequences, i.e., carcinogenesis, due to crossovers. Mouse Recql5β and human RECQL5β share 72% identities in amino acid composition and likely have similar functions. Thus, we believe that human RECQL5 is also required for the suppression of spontaneous crossovers in mitotic cells. Currently, it remains unclear whether these helicases play any roles in regulating meiotic crossovers.

Human RECQL5 produces three mRNAs as the result of alternative RNA processing and hence was predicted to encode three RecQ helicase isoforms (29, 49, 51). The expression of mouse Recql5 has not been very well studied, but it is predicted to produce three isoforms similar to those of humans, since human RECQL5 and mouse Recql5 genes are highly conserved. In all the Recql5 knockout alleles reported here, the targeted mutations were expected to delete the conserved helicase domain of all three predicted Recql5 helicase isoforms. Thus, the SCE phenotype that we have observed could be due to the lack of any of these isoforms. However, only RECQL5β has been found to interact with TOPOIIIα and IIIβ (51), to have the 3′-to-5′ helicase activity, and to have single-strand annealing activity (14). Although it is not yet clear whether the helicase activity or the single-strand annealing activity is related to the regulation of recombination, it has been shown that the interaction between BLM and TOPOIIIα plays an important role in the suppression of crossovers (26). Furthermore, RECQL5β is also the only isoform that contains a putative nuclear localization signal. Thus, we believe that the deletion of Recql5β is primarily responsible for the SCE phenotype observed in our Recql5 knockout ES cells. But future experiments are needed to verify or refute this prediction.

The nonredundant roles of Blm and Recql5 in the suppression of crossovers.

In Saccharomyces cerevisiae, SGS1 and SRS2 are known to be involved in nonredundant pathways to control crossovers, indicating the importance of such regulatory mechanisms in the homeostasis of eukaryotes (26). This also suggests that crossovers can be initiated from different events that are subjected to the regulations of different pathways. We showed here that Blm and Recql5 are apparently involved in independent, nonredundant pathways to suppress crossovers in mouse ES cells, confirming the speculation that mammals also have more than one pathway for suppressing crossovers in mitotic cells. Interestingly, to date, no mammalian homologues of SRS2 have been found.

Our present data do not explain how these two helicases have different specificities towards specific types of recombination events or pathways. It is conceivable that the specificities lie in the unique structural motifs found within them. Members of the RecQ helicase family all have a conserved helicase domain but also have unique motifs at their N or C termini, or both. Thus, the specificity of an individual RecQ helicase is likely dictated, at least in part, by its N- and/or C-terminal motifs. In this regard, it is noteworthy that BLM and RECQL5β are structurally distinct. In particular, BLM has distinct structural motifs at both its N and C termini, whereas RECQL5β lacks any discernible conserved motifs at its N terminus (51). BLM interacts with TOPOIIIα and several other components of the replication and recombination machineries via both its unique N- and C-terminal motifs (3, 58, 61). To date, the only known interacting partners for RECQL5 are TOPOIIIα and IIIβ (51). But given the lack of any identifiable N-terminal motifs in all predicted isoforms encoded by RECQL5 (51), it is conceivable that their interacting partners differ from those of BLM. Functionally speaking, both BLM and RECQL5β are able to unwind double-stranded DNA in a 3′-to-5′ direction in vitro, but only RECQL5β can stimulate the annealing of complementary single-stranded DNA molecules (14). Thus, it is clear that there exist both structural and functional differences between human BLM and RECQL5. Homologues of the RecQ family are highly conserved between mice and humans. In particular, human RECQL5β and mouse Recql5β share 72% identities. Thus, it is likely that mouse Blm and Recql5β also have structural and functional properties similar to those of their human counterparts.

An elevated frequency of crossovers could either be due to an overall increase in the frequency of recombination or from a skewing in the resolution of recombination repair intermediates that favors crossovers. Based on our present data, we cannot determine the mechanism by which Recql5 deficiency increases SCE frequency in mouse ES cells. It should be noted that this is also an unresolved issue with respect to BLM- and Blm-deficient cells. Future experiments are needed to address this issue.

RecQ DNA helicase-mediated suppression of crossovers and carcinogenesis.

Mutations in the single RecQ DNA helicase-encoding gene in budding yeast cause a multitude of phenotypes, including hyperrecombination and chromosome missegregation (59, 60). In humans, mutations in three of five RecQ-encoding genes give rise to three distinct cancer-prone genetic diseases: Werner (63), Rothmund-Thomson (30), and Bloom (9) syndromes. These observations imply that RecQ helicases have multiple distinct roles and that in mammals, individual RecQ helicase has acquired specialized roles in maintaining the stability of the genome. Cells from Werner syndrome patients have been shown to exhibit extensive deletions (12). The nature of genomic instability associated with Rothmund-Thomson syndrome has not been well characterized. However, we very recently showed that Recql4-deficient cells derived from Recql4 knockout mice exhibited a chromosomal instability phenotype as the result of defective sister chromatid cohesion (39). We showed here that two members of the mammalian RecQ gene family, Blm and Recql5, are involved in the suppression of crossovers. This new finding underscores both the essential role of recombination repair in mitotic cells and the importance of regulating this repair process to suppress crossovers, which can lead to deleterious consequences.

SCE, the outcome of crossovers between identical sister chromatids, results in high-fidelity recombination repair of DNA lesions. Thus, SCE is not mutagenic or oncogenic. However, other types of crossovers in mitotic cells can lead to potential oncogenic events, such as translocations, deletion, inversion, and somatic LOH. In particular, recent data have shown that crossovers between homologous chromosomes can lead to crossover-mediated LOH (2, 36). This crossover-mediated LOH represents a major mechanism of LOH in noncancerous human cells (20). Importantly, such a mechanism also has been shown to play a critical role in the loss of function of tumor suppressor genes at the early stage of colorectal cancer (22, 37, 52). Consistent with this, both Bloom syndrome patients and Blm knockout mice exhibit striking cancer susceptibility phenotypes (16, 37, 41). Thus, our finding that Blm and Recql5 are involved in nonredundant pathways to suppress crossovers in mouse ES cells suggests that Recql5 also plays important roles in the suppression of crossovers and perhaps carcinogenesis. Currently, human RECQL5 mutations have not been linked to any cancer-prone disorders. However, given the viable nature of Recql5 knockout mice and the anticipated functional conservation between human RECQL5 and mouse Recql5, it is reasonable to speculate that RECQL5 deficiency may be compatible with viability in humans. Thus, it remains possible that RECQL5 is the disease-causing gene for a cancer-prone syndrome. The mouse model reported here will provide a valuable tool for investigating the role of Recql5 as well as the relative contributions of the Blm-dependent and Recql5-dependent pathways to tumor suppression in mice. The results of such studies may also aid in the future identification of the human disease(s) that is caused by RECQL5 mutations, should there be any.