Induction and repair of zinc-finger nuclease-targeted double-strand breaks in Caenorhabditis elegans somatic cells

ZFN Cleavage in C. elegans.

As a preliminary test of the efficacy of engineered nuclease technology in C. elegans somatic cells, we used a previously constructed nuclease, QQR, which has been shown to be active in Xenopus oocytes, human cells and Arabidopsis thaliana (16, 29, 30). The zinc fingers of this protein bind the DNA sequence 5′-GGG GAA GAA-3′. We created a synthetic target for QQR, consisting of two inverted binding sites separated by the 6-bp MluI restriction endonuclease recognition site (Fig. 1). An expression construct for the QQR nuclease was prepared with the coding sequence under the control of the well-characterized C. elegans 16-48 heat-shock promoter (36).

The QQR target and expression plasmids were coinjected into the germ line to generate an extrachromosomal, multicopy transgene array (37). Such arrays typically consist of tens to hundreds of copies of the injected plasmids. L2–L3 larvae from a stable line were heat-shocked to induce expression of the nuclease. Larvae were used rather than adult animals to minimize the contribution of the germ line, because transgenes on arrays are effectively silenced in worm germ tissue (38). Because 10–100 germ cells are present even in these larval stages [Kimble, J. and Crittenden, S. L. (2005) in WormBook, ed. The C. elegans Research Community, WormBook, doi/10.1895/wormbook.1.13.1, www.wormbook.org.], the frequencies of nuclease-induced events are modestly underestimated. After recovery for several hours, DNA was isolated and the QQR target was amplified by PCR in a 1-kb fragment. The PCR products were treated with MluI to identify targets that were altered by inaccurate repair after cleavage. As shown in Fig. 2A, a substantial portion of the QQR targets had become resistant to MluI digestion in the heat-shocked nematodes, but not in controls.

To get a quantitative measurement of the frequency of MluI site loss, PCR products were cloned without prior digestion and assayed individually. QQR-induced mutations occurred in 26% (43 of 167 clones from four individual worms) of the targets examined. This is a minimum estimate, because some repair events would restore the MluI site, and some deletions may have been large enough to escape detection. By comparison, animals not subjected to heat shock exhibited no mutations at the QQR target, indicating that the observed mutations did not arise independently, and that uninduced QQR expression was minimal.

The MluI-resistant cloned PCR products were sequenced to determine the nature of the QQR-induced mutations. As shown in Table 1, the majority were simple insertions at the cleavage site. 58% (25 of 43) of the products had a duplication of the 4-bp sequence at the center of the MluI recognition site, which very likely arose by fill-in and blunt-end ligation of the 5′ overhangs created by QQR cleavage (14). Five products corresponded to partial fill-ins, and two showed fill-ins with an additional nearby alteration; thus, the new junction was likely created by blunt-end ligation in 74% of the cases. Among the remaining products, eight had rather large deletions (>100 bp), two had single base pair deletions, and one had a simple substitution in the MluI site. All together, five of the cloned products contained untemplated substitutions very close to the new junction, whereas none had substitutions elsewhere in the sequenced region, which extended for several hundred base pairs on either side of the cleavage site. Although the 25 identical 4-bp fill-ins were derived from only four nematodes, it is likely that they arose independently, because there are thousands of targets in each worm that should all be amplified with comparable efficiency, and most other alterations were recovered only once. The observed mutations all fall within the range expected for inaccurate repair after QQR cleavage.

Cleavage of a Genomic Target.

Having demonstrated that one engineered nuclease works in worms, we set out to target a nematode genomic locus with newly designed ZFNs. To facilitate analysis, we searched for sites in the form 5′-(NNC)₃N₆(GNN)₃-3′ in which the 6-bp spacer contained a convenient restriction site. We limited the search to GNN triplets, because the corresponding fingers have been most extensively tested. Among many such sites in the C. elegans genome, we chose a target at bases 3008453–3008476 of the X chromosome, which contains a HindIII site in the spacer (Wormbase, www.wormbase.org, release WS160, July 12, 2006). The sequence is shown below.

5′-ACC CAC TCC AAGCTT GGG GTG GCT-3′

3′-TGG GTG AGG TTCGAA CCC CAC CGA-5′

Because this target is located more than 1 kb away from the nearest predicted gene, we named it Nowhere (Nw). We constructed a new pair of ZFNs designed to bind the component 9-mers of this site, NwA for GGG GTG GCT, NwB for GGA GTG GGT, shown in bold above, and cloned them behind the 16-48 heat-shock promoter. Transgenic larvae were heat-shocked to express the nucleases recognizing each half site. The target region was amplified by PCR, and nuclease-induced mutations were identified by HindIII restriction enzyme digest of the 752-bp products. A substantial portion (≈20%) of the DNA was resistant to HindIII cleavage in the products from several individual nematodes (Fig. 2B); thus, cleavage at a chromosomal site by engineered nucleases is efficient.

In contrast to nuclease-induced double-strand breaks in the extrachromosomal arrays, repair of many of the breaks in the genomic target resulted in deletions. The frequency of induced mutations at Nw was quite similar to that for the QQR array: 18% (27 of 154 unselected clones) were HindIII-resistant (P = 0.10 for comparison of the QQR and Nw results). DNA sequence analysis of these and additional clones (Table 2) revealed relatively few simple fill-ins (13%, 7 of 54 sequences), whereas a much higher proportion was represented by small and large deletions. A small number of single-base pair substitutions was observed. A possible reason for the differences between the repair products at the genomic and extrachromosomal loci is that homologous repair is more frequent with the abundance of templates on the array, and this biases the outcome of end joining (see Discussion).

DNA Ligase IV Is Required for Blunt-End Joining.

The mutations generated by ZFN expression apparently resulted from inaccurate DNA repair via nonhomologous end joining. To test this hypothesis, we examined repair in mutants lacking DNA ligase IV, which is required for this process (39). An array carrying the QQR target and expression construct was established in a lig-4(ok716) strain. Nuclease induction and target analysis were performed as described above, except that the PCR products were considerably shorter. Fig. 3 shows a comparison of MluI digests of 207-bp PCR products from wild-type and lig-4 nematodes obtained after QQR induction. As observed before, a substantial proportion of the PCR products were MluI-resistant in the wild type. By contrast, the lig-4 worms yielded a very low level of such products. When unselected cloned products were characterized, 11% (25 of 239 clones) from the wild type were MluI-resistant, whereas only 0.4% (1 of 232 clones) were resistant in the lig-4 case. This difference is highly significant (P = 5 × 10⁻⁶). No MluI-resistant products were recovered from the wild type without heat shock (0 of 112 clones). As expected, ligase IV was required for efficient end joining; the remaining ends were presumably either lost or repaired by homologous recombination.

Sequence analysis of MluI-resistant products showed that the spectrum of alterations was dramatically affected by the lig-4 mutation (Table 2). As before, repair products from wild-type worms were dominated by small insertions at the cleavage site: full (89 of 127 sequences) and partial (16 of 127) fill-ins represented 83% of the products. The remainder were small deletions (9%) and localized substitutions (9%). (Larger deletions were presumably absent because of the limited size of the PCR product.) No insertions were seen in the lig-4 background (Table 2), indicating that DNA ligase IV is absolutely required for blunt-end joining. Deletions comprised 20% of the total (15 of 76); however, the most common alterations were single-base pair substitutions in the MluI site (80%, 61 of 76). Several of the deletions and substitutions (8 cases total) showed additional replacements close to the junction. These are not simply artifacts of inaccurate Taq polymerase amplification, because they were very rarely seen in conjunction with insertions in the wild-type PCR products or at sites more than 30 bp from the new junction. Rather, these substitutions likely arose during repair via error-prone DNA synthesis (see Discussion).

Repair of Double-Strand Breaks in the Germ Line.

Because expression of transgenes on extrachromosomal arrays is largely suppressed in the nematode germ line (38), we could only analyze repair of nuclease-induced double-strand breaks in somatic cells in the above experiments. To determine whether similar repair mechanisms heal breaks in the germ line, we investigated the joining of ends of exogenous DNA injected into the syncytial gonad of both wild-type and lig-4 nematodes. Repair products were recovered by PCR. Plasmid DNA was cut with two different restriction enzymes, leaving incompatible four-base 5′ overhangs. This configuration eliminates repair by homologous recombination with uncut templates and by simple religation, both of which were possible with the breaks generated by the induction of the ZFNs. Thus, in addition to examining the germ-line function of the nonhomologous end joining machinery, it does so in the absence of competition by gene conversion.

Products recovered from wild-type animals consisted largely of partial fill-ins of the 5′ overhang (Table 3). In only two of nine products were the ends resected past the single-stranded 5′ overhang, and then by only one or two bases. Only four of nine products showed possible microhomologies at the junctions, and these were only one or two bases long. These repair products are consistent with previously demonstrated lig-4-dependent ligation of blunt ends and resembled those recovered from somatic cells after nuclease cleavage.

By contrast, almost all products recovered from lig-4 mutants exhibited resection into double-stranded DNA. Three of eight had large deletions, and seven of eight had some loss of sequence from double-stranded regions. Furthermore, seven of eight showed junctional microhomologies of two or four bases. Again these products resemble those produced in somatic cells in lig-4 mutants.

Very similar results were obtained with an injected DNA having blunt ends (Table 7, which is published as supporting information on the PNAS web site). In the wild type, blunt joins were recovered, along with very small deletions, some of which had short microhomologies. One particular 9-bp deletion with a 4-bp microhomology was recovered multiple times. In lig-4 worms, this deletion was the most common product (7 of 16 clones). Another favored junction (5 of 16) had a 266-bp deletion based on a microhomology in which 8 of 9 bp matched. In contrast to the wild type, all junctions formed in the lig-4 mutants were deletions bounded by microhomologies of at least 3 bp. Thus, lig-4-independent repair processes in both the soma and germ line rely on a microhomology-based mechanism of end joining.

Nematode Strains.

The wild-type N2 was the parent strain for all others. The DNA ligase IV mutant strain was RB873 lig-4(ok716) III, which harbors a deletion of more than half the gene, including most of the ligase consensus sequences, that almost certainly abrogates ligase function.

Plasmid Constructions.

Plasmid pJM1-ZFN carries a ZFN coding sequence under heat-shock promoter control, in which the zinc fingers are easily replaced for attack of alternative targets. The pJL44.2 vector, containing the 16-48 heat-shock promoter driving MosI transposase expression (36), was modified by PCR-directed mutagenesis of the SpeI and NdeI sites. The transposase coding sequence was excised by MluI + NheI digestion and replaced with a DNA fragment containing a four-adenosine translational initiation consensus sequence, an ATG codon, codons for a nuclear localization signal (DPKKKRKV), and the coding sequence for the ben-1A ZFN (J.M. and D.C., unpublished results). The latter has a very short linker between the binding and cleavage domains (L = 0) (16), and the zinc-finger region is bounded by NdeI and SpeI sites.

The coding sequence for the QQR zinc fingers was excised from the Zif-QQR-F_N (L0) plasmid (16) by NdeI + SpeI digestion and placed in the pJM1-ZFN backbone to create pJM1-QQR. Coding sequences for the NwA and NwB zinc fingers were synthesized from long oligonucleotides (48) in a Zif268 framework (2). The resulting PCR products were cloned in the pJM1-ZFN backbone to create pJM1-NwA and pJM1-NwB. The amino acid sequences of the specificity-determining residues (6) of the three zinc fingers in NwA are: finger 1, QSSDLTR (recognizes GCT); finger 2, RSDALTR (GTG); finger 3, RSDHLSR (GGG). In NwB they are: finger 1, TSGHLVR (GGT); finger 2, RSDALTR (GTG); finger 3, QSGHLQR (GGA).

The QQR target was created by hybridizing the complementary oligonucleotides (5′-CTAGCTTCTTCCCCACGCGTGGGGAAGAA-3′ and 5′-AGCTTTCTTCCCCACGCGTGGGGAAGAAG-3′). The annealed fragment was ligated into HindIII + NheI-digested pL4440 (gift from Susan Mango, University of Utah) to create pJM2-QQRt.

Transgenic Arrays.

N2 nematodes were transformed with pJM1-QQR (5 ng/μl), pJM2-QQRt (15 ng/μl), pPD118.33 (Pmyo-2::GFP; 1 ng/μl) and 1-kb ladder (80 ng/μl; Invitrogen, Carlsbad, CA) to produce strain EG3526 oxEx654 that carries both the QQR target and the QQR expression construct. The RB873 lig-4(ok716) III strain was transformed with the same mixture, creating strain EG3521 lig-4(ok716) III; oxEx653. Strain EG3839 oxEx706, with the NwA and NwB ZFNs, was constructed by injecting N2 worms with pJM1-NwA (5 ng/μl), pJM1-NwB (5 ng/μl), pPD118.33 (1 ng/μl) and 1-kb ladder (100 ng/μl). Plasmid DNAs were linearized by ScaI digestion. Injections were performed on young adult N2 and RB873 nematodes by using standard techniques (37).

ZFN Induction.

Plates containing L2–L3 nematodes from transgenic strains were wrapped in Parafilm and heat-shocked in a water bath at 35°C for 1 h on two consecutive days, then allowed to recover at room temperature for several hours.

DNA Isolation and Analysis.

Heat-shocked worms were frozen at −80°C in single-worm lysis buffer [50 mM KCl/10 mM Tris·HCl (pH 8.3)/2.5 mM MgCl₂/0.45% Nonidet P-40/0.45% Tween-20/0.01% gelatin/200 μg/ml Proteinase K], lysed at 65°C for 1 h, followed by 95°C for 15 min. The QQR target DNA was amplified by PCR with Taq polymerase (New England Biolabs, Beverly, MA) in two different formats. To produce a 207-bp target fragment, the primers 5′-CCTGGCTTATCGAAATTAATACGA-3′ and 5′-CTATAGGGCGAATTGGGTACC-3′ were used. An alternative pair of primers, 5′-AACCGGCCCGGGTGAGATACCTACAGCGTGAGC-3′ and 5′-AACCGGCCCGGGGTCGAGGTGCCGTAAAGCAC-3′, gave a 1,006-bp product. The Nw target was amplified by using primers 5′-AACCGGGAATTCCAATCTATTTTTCGTGTAACGTG-3′ and 5′-AATTCCGGATCCACAAAATTGGCTTTCTTGTAACC-3′ to give a 752-bp product. After amplification, DNA was purified on MinElute columns (Qiagen, Valencia, CA), digested with MluI (QQR target) or HindIII (Nw target) and examined by electrophoresis in 1.5% or 2% agarose gels. DNA resistant to digestion was excised from the gels, purified by using Qiagen MinElute columns, and reamplified by PCR using Taq polymerase. When determining the mutation frequency, this initial restriction digest and analysis by gel electrophoresis were not performed. For sequence analysis, the short QQR product was digested with NotI and XhoI, which produces an 88-bp fragment from the unmodified target, and ligated into a similarly digested pBluescript backbone. The 1-kb QQR product was ligated directly into the pGEM-T vector backbone according to the manufacturer's instructions (Promega, Madison, WI). The Nw target was digested at the EcoRI and BamHI sites introduced on the PCR primers, and ligated into digested pBluescript. Ligation mixtures were used to transform Escherichia coli SS320 electrocompetent cells, which were allowed to grow overnight at 37°C on plates containing 100 μg/ml ampicillin. Individual transformants were screened by colony PCR (49) and digestion with MluI (QQR) or HindIII (Nw). Transformants yielding resistant PCR products were cultured, DNA was purified with Qiagen columns, and automated DNA sequences obtained at the University of Utah Core Sequencing Center. Sequence analysis was performed by using DNA Strider.

Germ-Line Injection.

For incompatible 5′ overhangs, pLitmus28 plasmid (New England Biolabs) was digested with BglII and HindIII, purified by using a Qiaquick spin column, dried in a SpeedVac concentrator and resuspended at a concentration of 100 ng/μl, and injected by using standard techniques (37). After 1–4 h, worms were frozen at −80°C in single-worm lysis buffer as described above. PCR was performed with primers 5′-GCCCCCGATTTAGAGCTTG-3′ and 5′-ACGTTGTTGCCATTGCTACAGG-3′, followed by gel purification of the resulting bands and DNA sequencing by using the primer 5′-TGTAAAACGACGGCCAGT-3′ (M13 forward). For blunt DNA fragments, plasmid pWD97 was used. This plasmid contains coding sequences for mRFP (50), with a PvuII-SnaBI fragment from pLitmus28 inserted into a PvuII site. This construct provides the multiple cloning site from pLitmus, flanked by mRFP sequences. The plasmid was digested with EcoRV and PvuII and purified, concentrated, and injected as described above. The worms were frozen, lysed and their DNA amplified as before, except the primers 5′-TGCTAGTTATTGCTCAGCGGT-3′ and 5′-TGCTAGTTATTGCTCAG-CGGT-3′ were used. Bands were purified from an agarose gel and TA-cloned by using the pGEM-T kit from Promega. Plasmids were purified from single colonies and sequenced by using a standard SP6 primer 5′-ATTTAGGTGACACTATAG-3′.