Manipulating Mitotic Recombination in the Zebrafish Embryo Through RecQ Helicases

Abstract

RecQ DNA helicases resolve Rad-51-mediated recombination and suppress aberrant homologous recombination. RecQ gene loss is associated with cancer susceptibility and increased mitotic recombination. We have developed an in vivo assay based on a zebrafish pigment mutant for suppression of RecQ activity, and demonstrate that zebrafish RecQ genes have conserved function in suppressing mitotic recombination.

HELICASES function in DNA replication, repair, recombination, and RNA transcription (van Brabant et al. 2000); helicase-catalyzed strand separation is usually coupled to adenosine triphosphate (ATP) hydrolysis. RecQ-related helicases contain a characteristic RecQ C-terminal (RQC) domain, which likely interacts with other proteins (Hickson 2003; Khakhar et al. 2003), and the RNAseD C-terminal (HRDC) domain, thought to be involved in binding the nucleic acid substrate (Liu et al. 1999). Among other functions, RecQ helicases suppress homologous recombination at stalled replication forks (Khakhar et al. 2003). In response to lesions that impede replication, RecQ-deficient cells use homologous recombination to restart replication, leading to accumulation of recombination-dependent cruciform structures at damaged forks (Liberi et al. 2005).

Mutations in human RecQ-related genes, BLM, WRN, and RECQL4, give rise to Bloom syndrome, Werner syndrome, and Rothmund–Thompson syndrome, respectively (Ellis et al. 1995; Kitao et al. 1999; Yu et al. 1996). Patients share susceptibility to cancer, implicating RecQ genes in tumor suppression. The cellular hallmark of Bloom syndrome is increased mitotic recombination, evidenced by sister chromatid exchange (SCE) and recombination between homologs. Loss of BLM also correlates with an increased rate of targeted homologous recombination (Luo et al. 2000; Wang et al. 2000). Loss of another RecQ family member, Recql5, further increases the rate of SCE in Blm^−/− cells (Hu et al. 2005; Wang et al. 2003).

We identified ESTs encoding two zebrafish RecQ-related genes, most closely related to Blm and Recql5 genes from other species (data not shown). blm and recql5 were expressed maternally and ubiquitously through somitogenesis. Thereafter, they displayed complex, largely overlapping patterns of tissue-specific transcription, prominent in the heart, brain ventricles, otic vesicle, retina, and somites through 5 days post fertilization (dpf) (data not shown).

Mitotic recombination can lead to distal loss of heterozygosity and cells homozygous for a mutation initially present on one homolog. Mutants for the zebrafish pigment gene golden (gol) (Lamason et al. 2005) have reduced melanin accumulation, apparent in the retinal pigment epithelium (RPE) at 3 dpf. The induction of gol⁻ cells in gol^+/− heterozygotes has been used to estimate the size of the RPE precursor pool and the timing of their commitment (Streisinger et al. 1989) and as the basis of a genetic screen for mutants with genomic instability (Moore et al. 2004; Moore et al. 2006). We have used embryos heterozygous for gol^b1, a null allele maintained on the AB genetic background (Johnson and Zon 1999), in an assay for the suppression of Blm and Recql5 function through injection of putative dominant negative RNAs.

Several BLM mutations reportedly have dominant negative activity, on the basis of their ability to cause mislocalization of wild-type (WT) BLM (Wang et al. 2001) or to increase the rate of SCE in WT cells (Neff et al. 1999). However, one such mutation (BLM^C1055S) was first described as a homozygous mutation in a patient (Ellis et al. 1995), suggesting that it does not act dominantly in vivo. We generated constructs encoding mutated zebrafish proteins, blm^C1066S (blm^C>S) and blm^K706T (blm^K>T), corresponding to described human mutations, and mutant recql5^K46T (recql5^K>T) was similarly generated in the conserved ATP-binding site of the helicase domain (Figure 1). WT and mutant RNAs were injected into gol^+/− embryos at the one-cell stage, and the larvae assayed at 3 dpf for hypopigmented RPE cells (Figure 2). Injection of WT blm, blm^C>S, or WT recql5 RNAs gave rise to hypopigmented cells in the RPE at similar, low frequences (Table 1), indicating that Blm^C>S does not have a specific dominant negative effect. Rather, WT Recql5, WT Blm, and Blm^C>S may nonspecifically disrupt helicase function when ectopically expressed at high levels. In contrast, injection of blm^K>T or recql5^K>T RNAs induced RPE mosaicism at significantly higher frequencies (Table 1). Consistent with the mutant proteins acting as dominant negatives, suppression of recql5 function through injection of an antisense morpholino also led to RPE mosaicism (Figure 2F; Table 1).

Figure 1.—

Portions of Blm and Recql5 protein sequences for zebrafish (DR), mouse (MM), and human (HS), indicating positions of conserved residues altered in mutant RNAs for injections; identical residues are shaded. The conserved lysine residues in the ATP-binding sites, mutated in Blm^K706T (A) and Recql5^K46T (C), and the conserved cysteine residue in the RQC domain, mutated in Blm^C1066S (B), are indicated by asterisks. Two zebrafish ESTs (accession nos. CB352479 and BG883342) were homologous to the 5′- and 3′-ends of human BLM cDNA, respectively, with an internal overlap of 1.2 kb. The full-length zebrafish blm coding sequence was assembled by PCR. One EST (accession no. BG304090) containing the entire coding sequence of zebrafish recql5 was identified (data not shown), corresponding to the longest (β) splicing form containing a nuclear localization signal (Shimamoto et al. 2000).

Figure 2.—

Suppression of Blm or Recql5 activity induces somatic mosaicism. (A) Diagram of a mitotic recombination event producing gol^−/− cells. (B) Note the difference in eye pigmentation between a gol^+/− (top) and gol^−/− (bottom) larva at 3 dpf. (C–E) Examples of mosaicism resulting from injection of either blm^K>T or recql5^K>T RNA are shown. Some RPE cells are outlined to indicate the number of gol^−/− cells in the clones (C and D). (E) Dorsal view of one injected fish at 3 dpf, with hypopigmented RPE cells in one entire eye and several smaller clones in the other eye. (F) Suppression of Recql5 through antisense morpholino injection can also induce RPE mosaicism, seen in this dorsal view of a larva at 3 dpf. Point mutations were introduced into blm and recql5 by megaprimer PCR (Ke and Madison 1997). RNAs were transcribed in vitro and injected as previously described (Xie and Fisher 2005).

TABLE 1.

Suppression of Blm or Recql5 activity induces formation of gol− cells in gol+/− embryos

RNA injected (pg)	% RPE clones (no. of embryos)	% abnormal
blm (100)	0.0 (187)	11.0
blm (500)	0.1 (710)	16.0
blmC>S (100)	0.0 (198)	9.1
blmC>S (500)	0.5 (619)	14.0
recql5 (100)	0.0 (179)	6.1
recql5 (500)	0.6 (326)	15.0
blmK>T (60)	1.7* (178)	3.4
blmK>T (120)	1.9* (261)	7.3
blmK>T (200)	3.4* (146)	12.0
recql5K>T (60)	2.1* (327)	5.8
recql5K>T (120)	2.4* (252)	6.3
blmK>T+recql5K>T (15 each)	2.3* (219)	4.1
blmK>T+recql5K>T (30 each)	3.5* (171)	4.1
blmK>T+recql5K>T (60 each)	3.6* (611)	7.4
blmK>T+recql5K>T (100 each)	3.0* (267)	21.0
recql5-MO (1 ng)	1.6* (501)	N/A

The indicated RNAs were injected into gol^+/− embryos at the one-cell stage; at 3 dpf, embryos were scored for the presence of clones of hypopigmented cells in the RPE (% RPE clones). The number of embryos that developed with gross morphological abnormalities, such as shortened axis, reduced head size, or general developmental delay, were also scored (% abnormal). A standard one-sided test for equality of probabilities in binomial data indicated that all injections with blm^K>T and recql5^K>T RNAs or with the recql5 morpholino differed significantly from injection of WT RNAs. *P < 0.001.

Clones of gol^−/− cells in a gol^+/− embryo could arise by somatic mutation at the WT allele of gol, gene conversion, deletion of the region encompassing the WT gol locus, mitotic nondisjunction, or mitotic recombination between homologs proximal to gol. Prior studies on BLM point to mitotic recombination as the likely mechanism (Luo et al. 2000). To exclude some of these mechanisms, we genotyped gol^−/− RPE cells from mosaic embryos. We crossed gol^−/− fish on the AB genetic background to WIK fish (Johnson and Zon 1999) to generate gol^+/− embryos on a polymorphic background and injected them with blm^K>T and recql5^K>T RNAs. From larvae with large patches of hypopigmented RPE cells, genomic DNA was prepared separately from dissected RPE cells and the remainder of the embryo. In two independent cases, only AB-specific alleles were detected from the RPE tissue for markers close to gol, while a proximal marker gave rise to AB and WIK bands; the remainder of the embryo was heterozygous for all three markers (Figure 3). These results are inconsistent with gene conversion, somatic mutation, or mitotic nondisjunction. They are consistent with mitotic recombination events proximal to gol, although we also cannot rule out small deletions encompassing gol.

Figure 3.—

Molecular evidence for mitotic recombination in mosaic gol^+/− embryos. gol^−/− fish were crossed with WIK fish, and 30 pg each of blm^K>T and recql5^K>T RNAs was injected into the resulting embryos. Larvae were fixed, large patches of hypopigmented RPE cells dissected out, and DNA prepared from RPE cells and remainder of the larva. (A) For one mosaic larva, DNA samples from the parental AB strain (indicated as A above gel lanes), the bulk of the larva (A/W), and the dissected RPE cells (R) were typed for three SSLP markers on chromosome 18; their positions relative to gol are indicated in the map (left). For the proximal marker (z11944), both the RPE cells and the rest of the larva are heterozygous for the AB and WIK parental alleles. For distal markers (z13836 and z46013), RPE cells show predominantly the AB allele, while the rest of the embryo is heterozygous. This pattern is consistent with a recombination between z11944 and z13836, as indicated by the homozygous red chromosomal segment in the diagram. (B) DNA samples from a second mosaic larva were typed for a SNP in intron 5 of the gol gene, revealed by TaqI restriction digest. DNA from the entire embryo is heterozygous for the SNP, while the RPE cells display only the SNP associated with the gol^b1 allele. M, marker lane.

Mitotic recombination proximal to gol, in any cell that eventually gives rise to RPE, could give rise to a clone of gol⁻ cells that would be scored in our assay. At 3 dpf, the RPE consists of a single layer of ∼550 polygonal cells in each eye, which arise from ∼40 specified precursors (Streisinger et al. 1989). Mitotic recombination in one precursor cell would generate a clone of ∼14 gol^−/− cells (1100/40 × 0.5), consistent with the clone size in many mosaic embryos. However, in cases where gol^−/− cells composed one-fourth to one-half of total RPE cells (Figure 2E), recombination may have occurred as early as the one-cell stage. In addition, the smallest clone in our assay contained only two cells (Figure 2D), suggesting that recombination events continue to occur for some time after the RPE precursors are specified. Thus, it seems likely that suppression of Blm and Recql5 activities can result in mitotic recombination during any of the early cell cycles.

Genetic mosaics allow for the study of pleiotropic gene function and analysis of cell lineage and cell autonomy. These are created in other experimental organisms through site-specific Flp recombinase (Golic and Lindquist 1989; Xu and Rubin 1993) or Cre–loxP-mediated recombination (Gu et al. 1994; Zong et al. 2005). We show here that suppression of RecQ helicases can also generate somatic clones of homozygous mutant cells. Through Blm and Recql5 suppression in germ cells, a similar approach might also be used to generate zebrafish embryos entirely lacking a maternally required gene product from a heterozygous female, with a method less cumbersome than that currently employed (Ciruna et al. 2002). To this end, we fused mutant blm or recql5 coding sequences to the 3′ UTR of nanos-1 (nos1) (Koprunner et al. 2001). Chimeric RNAs were injected, together with gfp–nos1–3′ UTR RNA, to enable visualization of germ cells. At the lowest RNA levels, GFP⁺ germ cells appeared normal in the large majority of embryos (Table 2). Progressively fewer embryos displayed normal germ cells at higher RNA doses, but >80% of embryos were still grossly normal, showing that deleterious effects were largely confined to the germ cells. Injected larvae showing normal numbers of germ cells were raised to adults; these were fertile and their offspring developed normally (data not shown), indicating that germ cell function was not compromised.

TABLE 2.

Suppression of Blm and Recql5 activities allows normal germ cell development

	% abnormal (no. of embryos)	Normal
RNA injected (pg)		% GFP−	% GFP+
GFP (50)	4.8 (84)	2.4	93
blmK>T (80)	7.8 (51)	20	73a
recql5K>T (80)	11 (65)	7.7	82a
blmK>T+recql5K>T (40 each)	12 (43)	14	74a
blmK>T+recql5K>T (80 each)	10 (48)	29	60
blmK>T+recql5K>T (160 each)	17 (63)	67	16

The mutant blm or recql5 cDNA was inserted into pSP64GFP3′UTRnos1, replacing the gfp coding sequence and allowing transcription of chimeric RNAs containing the 3′ UTR sequence of nanos-1 gene (Koprunner et al. 2001). The indicated RNAs were injected as previously described; all injections included 50 pg of gfp-nos1 3′ UTR RNA to mark the germ cells with GFP. The embryos that developed with gross morphological abnormalities are indicated in the “% abnormal” column; within the grossly normal embryos, “GFP−” indicates those with no apparent germ cells, and “GFP+” indicates embryos with number and location of germ cells similar to control embryos.

^aEmbryos saved for stocks, to assess fertility.