Development of a detection system for determining the ratio of homologous recombination versus illegitimate gene integration, HR/NHR.
For comparing homologous recombination to nonhomologous gene integration (NHR), a system had to be created that evaluates the ratio and the efficiency of both processes. We transformed strain CW15 with the plasmid pSI61 (Fig.
1A) and generated the new
C. reinhardtii recipient strain, T61-9. It contains an inserted genomic DNA element comprising the
ble gene, the
gfp gene, and a 3′-truncated
aphVIII gene (Δ3′
-aphVIII, 175 nucleotides missing at the 3′ end), all in frame.
ble was used for selection of this strain on Zeocin (
19,
36), Δ3′
-aph for later selection of homologous recombinants on paromomycin, and
gfp for potential monitoring of
aphVIII expression as a fusion protein on protein blots or in living cells.
First, strain CW15 was transformed with a functional
aphVIII gene (plasmid pSI103; Fig.
1B) (
34). The expression was driven by a hybrid promoter, namely a combination of those of the
HSP70 and
RBCS2 genes (HA promoter) (
29) and further enhanced by the
RBCS2 3′ end. We obtained about 3,000 clones/1 μg of DNA, which is similar to the numbers obtained in earlier experiments (
34). Similar numbers were also achieved with the recipient strain, T61-9 (Table
1, no. 1). The clones were not analyzed any further because from earlier experiments we could anticipate that most transformants were based on nonhomologous DNA integration into the genome. Almost the same number of clones was generated with 1 μg of a 1.8-kb SacI-KpnI fragment of pSI103 comprising the
aphVIII coding region plus HA promoter and 3′ region of the
RBCS2 gene (Table
1, no. 2). The molar amount of DNA was roughly four times higher than that of cyclic DNA used, but with respect to μg of DNA, there was no difference between transformations with linear or circular double-stranded DNA.
Next,
Chlamydomonas cells were transformed with a plasmid that contained two diphtheria toxin A fragment genes,
dtA, plus HA promoter on both sites of the
aphVIII in order to suppress illegitimate gene integration (negative selection). The
dtA gene was resynthesized and adapted to the nuclear codon usage of strongly expressed
C. reinhardtii genes according to reference
9 (accession no. AY611535 ). Similar to in rice (
38), the total number of clones declined by a factor of about 8 in both strains CW15 and T61-9. Thus, the introduction of
dtA did not sufficiently reduce the transformation rate so as to make it likely that the HR/NHR ratio was significantly increased. Therefore, transformants were not analyzed within this context and a detailed analysis might be reported elsewhere (B. Zorin, unpublished data).
To determine the frequency of homologous recombination, we transformed
Chlamydomonas cells with an
aphVIII gene that was truncated at the 5′ end by 120 nucleotides (plasmid pSI201). This Δ5′
-aphVIII gene (Fig.
1C) could only generate paromomycin-resistant (Pm
r) clones after recombination with the Δ3′-
aphVIII of the recipient but not after integration elsewhere into the genome (I. Sizova, unpublished data). Only two T61-9 transformants were generated with 70 μg DNA (Table
1, no. 3). In yeast, integrative recombination is enhanced by introducing a double-stranded break in the region of homology shared between the transforming vector and the targeted gene (
24). In line with these early findings,
C. reinhardtii was transformed with a plasmid that has been linearized within the 5′Δ
-aphVIII gene with SalI (named pSI201:Sal). The free double-stranded ends obviously promoted recombination only slightly, namely by a factor of 2.5 in five independent experiments, which was at the border of significance (Table
1, no. 4). PCR such as that in Fig.
2A has shown that, in all seven transformants tested (no. 3 and 4), the
aphVIII gene has been successfully repaired. Primers from the
ble part and the 175-bp fragment that is missing in the recipient strain T61-9 were used (see Fig.
2A). Two clones generated with pSI201 and five clones generated with pSI201 linearized with SalI were analyzed by DNA blot hybridization. One of each is shown in Fig.
2B and
2C. In both transformants, we observed one BamHI-PstI fragment that is enlarged compared to the recipient (from 4 to 4.2 kb) and consistent with the repair of the
aphVIII gene. The larger shift of the PstI fragment (7.6 to 8 kb in Fig.
2B) again seen in all transformants is consistent with a single crossover leading to integration of plasmid DNA as outlined in Fig.
2D and E. However, the second PstI fragment including the Δ5′Δ3′
aphVIII region is different in all five analyzed transformants, indicating an unclear and variable integration of the 3′end. In one transformant generated with circular plasmid, the Δ5′Δ3′
aphVIII region is seen as 3.7-kb fragment (Fig.
2B), as expected from the PstI site position in the recipient. The interpretation was supported by probing the DNA blot with DIG-labeled pBluescriptIIKS
− (vector only), which identified the 8-kb band but not the 3.7-kb PstI band (data not shown). In the five transformants resulting from linearized plasmid, the Δ5′Δ3′
aphVIII region is located within short unexpected PstI fragments (1.3 kb in Fig.
2C and E) or is absent from all (data not shown). We suspect that the 0.5-kb BamHI fragment of Δ5′Δ3′-
aphVIII has run off the gel. These observations show that even if a gene is effectively targeted through HR, DNA rearrangements can occur. But, under our experimental conditions, clones with 5′ rearrangements did not survive. Large rearrangements have been found after gene targeting and repair of
NIT8 (
22).
In summary, we have compared transformation rates with dsDNA carrying a similar length of homology, but either able or unable to transform
Chlamydomonas via illegitimate recombination (experiments 1 versus 3 and 2 versus 4 of Table
1): we find that for linearized and circular DNA, the ratio HR/NHR is in the order of (4 × 10
4 to 10
5: [(3 × 10
3) × 70]/5 and [(3 × 10
3) × 70]/2).
Gene targeting using ssDNA.
By comparing natural homologous recombination processes with illegitimate gene integration processes as they occur in various systems like transposons, P-elements of Drosophila, retroviruses, retrotransposons, or T-DNA from agrobacteria, we have drawn the conclusion that all nonhomologous integration processes involve double-stranded DNA, whereas homologous recombination always occurs via single-stranded DNA or single-stranded reaction intermediates. We have hypothesized that the use of single-stranded DNA might prevent illegitimate integration processes.
We transformed
C. reinhardtii T61-9 cells with a fully functional single-stranded
aphVIII gene (sense strand, SacI-KpnI fragment of pSI103) without plasmid extensions but including HA promoter and the 3′ end from the
RBCS2 gene (Fig.
1B). Linear ssDNA was prepared by linear PCR from the ds-SacI-KpnI template and careful purification of the ss product. Ten micrograms of DNA generated only 70 transformants in T61-9 cells (Table
1, no. 5) instead of 3 × 10
4, which would appear after transformation with comparable amounts of the same double-stranded construct (Table
1, no. 2). From these numbers, it was not clear whether ssDNA is less efficient for transformation in general and to what extent HR and NHR had occurred. Moreover, all transformants could be based on nonhomologous gene integration caused by residual traces of dsDNA. All 70 transformants were tested by PCR. One transformant was a homologous recombinant. To test the frequency of homologous recombination with single-stranded DNA more directly,
Chlamydomonas cells were transformed with an ss-
aphVIII fragment that was deleted at the 5′ end (Table
1, no. 6). Four homologous recombinants were found. Two—not necessarily independent—were analyzed by DNA blotting (Fig.
3A). Both were based on double-crossover events but showed multiple integrations of the
aphVIII genes into the
aphVIII site of the recipient according to Fig.
3A and D.
It was conceivable that ssDNA degradation contributed to the reduction of the total number of transformants. Especially, degradation of the promoter would cause a reduced number of nonhomologous integrations in the case of construct no. 5 of Table
1. To test this possibility, transformation was repeated with cyclic single-stranded DNA (Table
1, no. 7). ss-
aphVIII DNA was produced in pBluescriptIIKS(−) phagemid. Ten micrograms of ss phage DNA resulted in four transformants. In one of them, the 3′ deletion of the
aphVIII had been repaired as verified by PCR and sequencing of the
aphVIII PCR product. In the other three transformants,
aphVIII had integrated somewhere else, possibly into homologous plasmid sequences of the T61-9 recipient outside the
aphVIII or into the endogenous
rbcS2 5′ or 3′ sequences. This has not been experimentally verified. Nevertheless, the result looked highly promising. But, unfortunately, we were unable to reproduce this finding. In several subsequent repeats, the number of clones was higher and no recombinant was found (see Table
1, no. 7). To get more reliable numbers for the frequency of HR after transformation with cyclic ssDNA, experiments were repeated with 5′-deleted single-stranded
aphVIII in plasmid (Table
1, no. 8) and with 3 times more DNA (30 μg ss-Δ5′-
aphVIII). Two clones, both homologous recombinants, were found. DNA blotting revealed that the two transformants generated with circular ss-Δ5′-
aphVIII gives rise to only one repaired
aphVIII gene in each transformant with no double-deleted copy (Fig.
3B and E). For the clone analyzed in Fig.
3B, the double-truncated Δ3′Δ5′
aphVIII copy (small PstI fragment), as created by a single “crossover” event using ds-circular DNA (Fig.
2B and C), did not appear. In addition, the size of the large PstI fragment increased only slightly, from 7.6 to 7.8 kb, in agreement with repair only. Both observations are consistent with a double-crossover event as outlined in Fig.
3E (further discussed below).
Single-stranded plasmid linearized within the
aphVIII gene comparable to the double-stranded plasmid in Table
1, no. 4, was not used for two reasons. First, linearization is difficult with ssDNA and only possible by using adapter primers. Second, it is thought that linearized double-stranded plasmid integrates into the host genome by the two 3′ ends of the opposite plasmid strands (
18). ssDNA with only one 3′ end cannot integrate via such mechanism (further discussed below).
By comparing transformation efficiencies, we draw the conclusion that the efficiency of homologous recombination is identical with ssDNA and dsDNA within the experimental errors. In contrast, the frequency of nonhomologous integration was dramatically reduced when ssDNA was applied.
It was not unlikely that the small absolute number of recombinants was correlated with the shortness of homology 5′ of the
aphVIII deletion that had to be repaired. In a final set of experiments, we transformed with a promoterless full-length
aphVIII connected to 720 nucleotides of
gfp(ss-M13-BZ301), resulting in a 1.4-kb sequence of homology 5′ contiguous to the recipient deletion. In former experiments, promoter deletion from double-stranded
aphVIII caused a 5- to 140-fold reduction of transformants compared to homologues that were linked to promoters of different strength (
34). Promoterless
aphVIII is able to jump in frame into any other gene, the transcription of which is driven by a moderate promoter. However, the production of the circular ss-
gfp-aphVIII plasmid proved to be difficult in our hands using the phagemid system. With increasing plasmid size, the yield of ssDNA of interest drastically dropped. Therefore,
gfp-aphVIII was directly cloned into M13mp18 (New England BioLabs) phage (plasmid M13-BZ301). After two independent transformations with 30 μg DNA, in every case 30 transformants appeared; 4 and 1 were homologous recombinants (Table
1, no. 10). Two were analyzed by DNA blotting (Fig.
3C). The blots are similar, and the transformants obviously did not originate from independent events. Both showed single integration and repair of the
aphVIII gene. As outlined in Fig.
3F, the 5′ crossover was homologous, whereas the 3′ crossover is less obvious. It must have occurred within the 8-kb Pst fragment of the recipient but outside the Pst-fragment of the transforming plasmid. Thus, the resulting Pst fragment was smaller than that of the recipient. We do not know to what respect the DNA regions outside the
RBCS2 site are homologous since the sequence of the recipient is not known. But, other than in the case of no. 8, a rearrangement of the 3′ end cannot be excluded. By comparing the number of clones that had appeared after transformation with the single-stranded M13-BZ301 vector (Table
1, no. 10) and double-stranded replicative form (Table
1, no. 9), the number of nonhomologous recombinants is reduced about 300 times with promoterless constructs. One should keep in mind that with promoterless constructs, only recombinations that occurred in frame into an active exon become visible as a clone. This reduced the number of nonhomologous recombinants and the NHR/HR ratio.