The Fate of Linear DNA in Saccharomyces cerevisiae and Candida glabrata: The Role of Homologous and Non-Homologous End Joining

In vivo assembly of plasmids has become an increasingly used process, as high throughput studies in molecular biology seek to examine gene function. In this study, we investigated the plasmid construction technique called gap repair cloning (GRC) in two closely related species of yeast – Saccharomyces cerevisiae and Candida glabrata. GRC utilizes homologous recombination (HR) activity to join a linear vector and a linear piece of DNA that contains base pair homology. We demonstrate that a minimum of 20 bp of homology on each side of the linear DNA is required for GRC to occur with at least 10% efficiency. Between the two species, we determine that S. cerevisiae is slightly more efficient at performing GRC. GRC is less efficient in rad52 deletion mutants, which are defective in HR in both species. In dnl4 deletion mutants, which perform less non-homologous end joining (NHEJ), the frequency of GRC increases in C. glabrata, whereas GRC frequency only minimally increases in S. cerevisiae, suggesting that NHEJ is more prevalent in C. glabrata. Our studies allow for a model of the fate of linear DNA when transformed into yeast cells. This model is not the same for both species. Most significantly, during GRC, C. glabrata performs NHEJ activity at a detectable rate (>5%), while S. cerevisiae does not. Our model suggests that S. cerevisiae is more efficient at HR because NHEJ is less prevalent than in C. glabrata. This work demonstrates the determinants for GRC and that while C. glabrata has a lower efficiency of GRC, this species still provides a viable option for GRC.


Introduction
Plasmid construction is an essential technique in molecular biology. Assembly of plasmids, containing specific DNA fragments, can be carried out either in vitro or in vivo. In vitro cloning, which requires restriction enzymes and DNA ligase, can be costly and inefficient for high throughput methods. In vivo cloning, such as gap repair cloning (GRC), utilizes homologous recombination (HR) activity and can be cheaper and more efficient [1,2]. GRC uses available homologous ends of a linearized vector and a DNA fragment (usually generated by PCR) to fuse the two, creating a circular plasmid (Figure 1). Yeast species are appealing for GRC as many species appear to perform HR efficiently [3,4,5,6,7]. By understanding the minimal requirements for GRC, costs can be minimized.
Several studies have examined GRC in the budding yeast, Saccharomyces cerevisiae, and demonstrated this species is capable of efficient GRC [1,8,9,10]. These analyses specify that 30 base pairs of homology is sufficient for DNA integration into a linearized vector, but it may not be the minimum requirement [9,11]. Recently, a similar study indicated that GRC is a viable cloning technique in Schizosaccharomyces pombe [4]. Additionally, our laboratory routinely uses GRC in C. glabrata, but the specific requirements for efficient GRC are unknown.
Our studies with C. glabrata suggest that HR appears to be less efficient than in S. cerevisiae, as deletion of genes using HR is less efficient in C. glabrata [5](data not shown). We expected that a detailed analysis of GRC in the two species would elucidate the role of HR, the mechanism by which this technique takes place, and non-homologous end joining (NHEJ). Additionally, C. glabrata is closely related to S. cerevisiae and pathogenic to mammals, allowing for comparisons over evolutionary time. We examined the role of two genes involved in HR and NHEJ -RAD52 and DNL4. RAD52 has previously been identified as a gene involved in DNA double-strand break repair and it facilitates HR in S. cerevisiae and S. pombe [12]. When RAD52 is deleted, HR should be decreased and we would expect that GRC will either not take place or will be dramatically reduced. DNL4 is required for NHEJ, which is the repairing of double stranded DNA breaks without homologous ends via its ligase activity [13,14]. Mutations in dnl4 + in S. pombe lead to increased frequency of HR, and we predicted that loss of DNL4 should lead to higher frequency of GRC in S. cerevisiae and C. glabrata, as NHEJ would be reduced.
The goal of this study was to define the determinants of GRC in both S. cerevisiae and C. glabrata, with regard to the amount of homology and ratios of DNA concentrations. Additionally, we aimed to examine the role of HR in GRC and how NHEJ influences the frequency of GRC. Finally, we incorporated our data into a model of the fate of linear DNA when transformed into the two species.

Yeast Strains and Growth Conditions
Wild-type S. cerevisiae and wild-type C. glabrata were used as the host strains for GRC transformations as both are ura3 2 his3 2 (See Table 1 for strains). Deletions of RAD52 and DNL4 in both S. cerevisiae and C. glabrata were generated using antibiotic resistance genes KANMX6 and NATMX6 (conferring resistance to G-418 and nourseothricin, respectively) and homologous recombination to delete the ORFs [5,15,16], which was confirmed by PCR. For transformations, yeast strains were grown in YEPD medium at 30uC until logarithmic growth phase (OD 600 0.2-0.5). To select for plasmids, strains were grown in synthetic dextrose (SD) medium with CSM lacking the appropriate amino acids (either histidine or uracil) (Sunrise Science, San Diego, CA, USA). Transformations were performed using a standard lithium acetate protocol [17,18,19].

DNA for Transformations
CgURA3 was amplified from C. glabrata wild-type DNA with varying amounts of homology on each side using a mixture of Taq and Pfu polymerase ( Figure 1). To amplify CgURA3 with 40 bp of homology, primers with the sequences: 59-GCTCCACCGCGGTGGCGGCCGCTCTAGAACTAGTG-GATCCtgacttttacactaatgagg and 39-GGTCGACGGTATCGA-TAAGCTTGATATCGAATTCCTGCAGctagatattacatgcataac were used, where the lowercase letters correspond to the CgURA gene. For 20 bp and 10 bp of homology, truncated versions of these primers were used. pRS313 (HIS3 + ) [20] was linearized with SmaI or with EcoRV and SmaI. PCR products and linear vectors were subjected to gel electrophoresis, purified using a Geneclean II Kit (M.P. Biomedicals, CA, USA), and quantified with a NanoDrop 2000. The CgURA3 PCR products and linearized vectors were co-transformed at different molar ratios (0, 0.02, 0.2, and 1.0) into S. cerevisiae and C. glabrata strains. Based on the size of fragments, we estimated that equal amounts of DNA in nanograms were a molar ratio of 0.2, as the plasmid was ,56 the size of the insert. For all transformations, we used 50 ng of linearized vector. All cells from S. cerevisiae transformations and 20% of cells from C. glabrata were plated onto SD medium lacking histidine, so that individual colonies per plate were ,500 cfu. After 3 days of growth, these HIS3 + colonies were then replica plated to SD medium lacking uracil.

Sequencing
To sequence plasmids created by GRC, cells were grown overnight in selective medium and yeast plasmids were purified utilizing an ammonium acetate procedure [19]. Yeast plasmid preparations were transformed into chemi-competent XL1-Blue Escherichia coli cells and plasmids were isolated by an alkaline lysis protocol. We were successful in isolating plasmid DNA from yeast .90% of the time. Plasmids were sequenced (GENEWIZ, NJ, USA) with T7 and T3 primers to determine the sequence on each side of the insertion.

Investigation of GRC Efficiency in S. cerevisiae and C. glabrata
To compare the ability of S. cerevisiae and C. glabrata to carry out GRC, we performed transformation reactions using the same preparations of linearized vector, pRS313 (HIS3 + ), and the same C. glabrata gene, CgURA3 with promoter, in both species. pRS313 contains an autonomously replicating sequence (ARS) that is functional in both species and CEN6 from S. cerevisiae [19,20]. It is unclear whether CEN6 is functional in C. glabrata, but it is unlikely as there are very specific sequence requirements for centromeric sequences in C. glabrata [21,22] and we have observed relatively quick plasmid loss under non-selective conditions in C. glabrata  (data not shown). Each CgURA3 product was amplified to include 40 bp of vector homology on each side, which is sufficient homology for GRC. Using different molar ratios of CgURA3 to digested pRS313 (0, 0.02, 0.2, and 1.0), we titrated the concentration of DNA required for effective GRC. We measured GRC efficiency as the percentage of HIS3 + colonies (from linearized pRS313) that became URA3 + ( Figure 2). Not surprisingly, the higher the molar ratio of insert DNA (CgURA3 with homology) to digested vector, the better the GRC efficiency. We determined that an insert to vector molar ratio of 1.0 produced an average 55% DNA insertions in S. cerevisiae and an average 49% DNA insertion in C. glabrata, and adding 5-fold molar excess of insert DNA allowed GRC to approach 90% in both species (data not shown). We concluded that S. cerevisiae and C. glabrata both efficiently engage in GRC, and that GRC can occur even when there is a relatively low amount of insert. It is not surprising that equal molar amounts are not required for efficient GRC, as it is likely that multiple DNA fragments enter the cell during transformation, and all that is required for our assessment of GRC is for one linear DNA vector molecule to be repaired by one insert with homologous ends. It is worth noting that an alternative reason for observing URA3 + HIS3 + colonies could be integration of either piece of DNA into the genome or URA3 + integration coupled with re-circularization of the pRS313. To confirm that we were observing GRC plasmids, we grew 92 Ura + His + colonies from S. cerevisiae and 115 colonies from C. glabrata on non-selective (YEPD) medium overnight to allow for plasmid loss and then replica-plated to medium containing 5-FOA, which selects against cells that are expressing the URA3 + gene ( Figure S1 and Figure S2). In C. glabrata 113/115 colonies resulted in FOA R colonies, and in S. cerevisiae 91/92 colonies, suggesting almost all Ura + colonies were a consequence of URA3 + incorporation into a plasmid as opposed to stable integration into the genome. Importantly, all of the FOA R colonies except for one C. glabrata colony resulted in a His 2 phenotype, indicating that the vast majority of URA3 and HIS3 genotypes are coupled on a plasmid and not a consequence of chromosomal integration events.

Determination of Minimal Base Pair Homology for Detectable GRC Results
To investigate the minimal amount of homology required for GRC, we performed transformation reactions in both S. cerevisiae and C. glabrata with varying amounts of homology on each side of the CgURA3 PCR product: 40 bp, 20 bp, and 10 bp ( Figure 3A and 3B). We determined that decreasing the homology on both sides from 40 bp to 20 bp dramatically decreased the GRC efficiency. In both S. cerevisiae and C. glabrata, decreasing the homology from 40 base pairs to 20 base pairs on a single side reduces the efficiency by 50%-60%. Further decreasing the homology from 40 bp to 10 bp on both sides allowed for only extremely rare GRC activity. From this we conclude that decreasing the homology on a single side of the DNA insert significantly affects the efficiency of GRC. Furthermore, the data are not consistent with the hypothesis that one side of high homology (40 bp) can compensate for low homology (10 bp) on the other side -i.e. the 40/10 PCR products perform poorly relative to the 20/20, even though this is more total homology with the 40/10 PCR products, suggesting that there is not a simple counting mechanism for homologous recombination. Our data suggest that the minimum requirement for homology is 20 bp on each side and that each side of homology has an individual role in GRC. These requirements are similar in both S. cerevisiae and C. glabrata.
While performing these experiments, we noticed a different trend between the two species. When transforming 40 bp of homology on each side, the number of colonies produced by S. cerevisiae increased consistently as the insert to vector ratio increased from 0.0 to 1.0 ( Figure 4A); however, when transforming C. glabrata with 40 bp of homology on each side, the number of colonies produced remained relatively constant, despite the increasing insert to vector ratio ( Figure 4B). These data are consistent with a difference in HR activity vs. NHEJ activity in the two species. If HR is more active in S. cerevisiae relative to C. glabrata, then we would expect only GRC clones to be observed when insert is added. If NHEJ is more active, as in C. glabrata, then there should be approximately the same amount of colonies observed regardless of insert to vector ratio. Our data suggest that NHEJ activity, the mechanism by which a vector re-circularizes, with or without insert, is more prevalent in C. glabrata than it is in S. cerevisiae.

Effects of Inactivating Homologous Recombination or Non-homologous Recombination Pathways on the Frequency of GRC
To examine the role of HR and NHEJ on the rate of GRC, we deleted RAD52 and DNL4 in both S. cerevisiae and C. glabrata and repeated the transformation experiments with 40/40 bp and 40/ 20 bp of homology. RAD52 is a key gene in genetic recombination and DNA repair, and is required for most forms of HR, but deleting RAD52 does not completely eliminate HR [7,12,23,24]. DNL4 also mediates DNA repair, but through non-homologous repair mechanisms [14,25].
Decreasing HR activity through deletion of RAD52 considerably reduces the frequency of GRC in both species compared with wild-type strains ( Figure 5A and 5C -note difference in scale). In S. cerevisiae, the data support that deleting RAD52 does decrease the frequency of GRC, but does not completely eliminate it. Likewise, in the Cgrad52D strain, the efficiency of GRC is dramatically decreased but not completely eliminated.
Inactivating the pathway responsible for NHEJ activity has different effects on S. cerevisiae and C. glabrata ( Figure 5B and 5C). Notably, in Scdnl4D cells, the percent of DNA insertion into the plasmid is relatively unaffected compared with transformations in the wild-type -i.e. loss of NHEJ does not dramatically increase GRC efficiency. However, in C. glabrata, GRC increases in a Cgdnl4D strain, suggesting that NHEJ is more active, and loss of NHEJ drives the equilibrium of re-circularization towards HR.

Determination of Proportion of Resealed Vectors through HR or NHEJ in S. cerevisiae
In some transformations, such as those in a rad52D strain or where PCR product with low bp homology was used, the likelihood of acquiring plasmids with the URA3 insertion was low. To determine whether these rare URA3 + transformants were a consequence of HR or NHEJ, we rescued the plasmids in E. coli and sequenced the plasmids that resulted from the transformations. Not surprisingly, plasmids rescued from a S. cerevisiae wildtype transformation with 40 bp of homology on each side exhibited re-circularization via HR (data not shown); however, even with 10 bp of homology on each side, S. cerevisiae wild-type appeared to only re-circularize via HR ( Figure 6). Likewise, in a Scrad52D, which has decreased HR ( Figure 5A), transformations with 40 bp of homology on each side yielded plasmids that recircularized via HR, indicating that loss of RAD52 does not completely eliminate GRC.
We sequenced re-circularized S.cerevisiae plasmids that did not contain URA3 (HIS3 + only), and found that all appeared to not have been digested at all with SmaI. Knowing that restriction enzyme digestion may not proceed to completion, we hypothesized that these few HIS3 + URA3 2 plasmids could be a consequence of incomplete digestion. To test this hypothesis, we doubly-digested pRS313 with two blunt cutting enzymes, EcoRV and SmaI, expecting that if re-circularization is caused by NHEJ and not a consequence of partial restriction enzyme digestion, we should observe the loss of 13 bp in the plasmid sequence (Figure 1).  We only observed completely undigested plasmid in S. cerevisiae plasmids, suggesting either we are unable to observe NHEJ in S. cerevisiae in this experiment, or that a singly digested plasmid ligated with no change in sequence. Supporting the lack of detectable NHEJ in this GRC assay, we were unable to recover URA3 + HIS3 + plasmids in S. cerevisiae strains where RAD52 was deleted and the homology was decreased to 10 bp on each side.

Determination of Proportion of Resealed Vectors through HR or NHEJ in C. glabrata
In C. glabrata we also rescued the plasmids from the rare URA3 + HIS3 + transformants using DNA inserts with 10 bp of homology on each side or from Cgrad52D transformations. The sequencing data from these plasmids indicated that DNA was inserted via HR in C. glabrata wild-type (with 10 bp of homology) and in Cgrad52D (with 40 bp of homology), strengthening the conclusion that HR is the preferred mechanism in C. glabrata as well ( Figure 6). We then transformed the Cgrad52D strain using linear vector and CgURA3 with 10 bp of homology, and noted a difference between S. cerevisiae and C. glabrata. Whereas we never identified a URA3 + HIS3 + colony in Scrad52D with 10 bp homology, we identified colonies that were both URA3 + and HIS3 + in C. glabrata. We purified these plasmids and confirmed that 60% of the time the linear PCR product was inserted into the plasmid via NHEJ, because there was a 10 bp duplication of sequence flanking both sides of the URA3 + gene.
Following this experiment, plasmids were rescued from a transformation using a doubly digested plasmid and no insert. In S. cerevisiae, 100% of the sequences from these double-digested plasmids indicated that the vector was likely never fully digested. In C. glabrata, 100% of the sequences demonstrated that the rescued vectors were digested and re-circularized, lacking the 13 bp, and thus were a consequence of NHEJ. These data reveal that C. glabrata is more capable of performing NHEJ relative to S. cerevisiae.
Our results may appear to conflict with other published studies -i.e. our observation of no NHEJ activity in S. cerevisiae; however, it is worth pointing out four major differences between this study and the others. First, we used at least 406 less plasmid DNA (50 ng vs. 2 mg) than another study using plasmids making it unlikely that we have saturated the HR machinery in the cell and skewing results towards HR [13]. We chose these low concentrations to see subtle differences between the two species in a titration of different variables, but these conditions minimize NHEJ activity (Figure 2), as previous experiments have demonstrated less than 5 transformants/mg of non-homologous DNA [26]. Second, most NHEJ studies examine events in the chromosome, not in plasmids [5,26,27]. Third, many studies indicate high levels of HR in yeast species and so it is not surprising given our conditions that we are not observing relatively rare NHEJ events [10,13,24]. Finally, previous studies have indicated that direct ligation may be a mode of repair, but we have also examined the role of partial plasmid digestion (see below). Because we digest with two blunt end restriction enzymes, and observe ligation events in C. glabrata, but not in S. cerevisiae, we can conclude that this ligation/NHEJ activity Figure 5. Average insertion percentage of CgURA3 into pRS313 (HIS3 + ) in rad52D (A), dnl4D (B), and wild-type (C) strains in S. cerevisiae and C. glabrata. Average insertion percentage from GRC experiments using CgURA3 PCR products that have either 40 bp of homology on each side or a combination of 40 and 20 bp of homology. PCR to vector molar ratio was constant at 1. Average insertion percentage was calculated as URA3 + /HIS3 + and vector had no added PCR product. Note difference in scale for A relative to B and C. doi:10.1371/journal.pone.0069628.g005 is higher in C. glabrata relative to S. cerevisiae. Therefore, in our conditions we do not observe NHEJ in S. cerevisiae, but there are likely rare NHEJ events that we are not observing. In fact, in other studies even with higher concentrations of plasmid in S. cerevisiae the vast majority of ''re-circularization events'' were actually incomplete digestion events [10]. Figure 6. Outcomes of sequencing plasmids from GRC experiments. Sequences from rescued plasmids showed one of four different results: insertion via HR, insertion via NHEJ, vector ligation via NHEJ, or undigested vector. Each column represents a different GRC experiment with varying species, PCR homology, vector, and genotype. The two vector types include pRS313 (digested with only SmaI) and a doubly digested pRS313 (SmaI and EcoRV). The plasmids from the doubly digested vector were rescued from colonies that were HIS3 + and URA3 2 . The values given represent the number of plasmid sequences displaying each outcome out of the total number of plasmids sequenced. The shaded regions highlight outcomes. doi:10.1371/journal.pone.0069628.g006 Figure 7. Model of the fate of linear DNA in S. cerevisiae and C. glabrata. In S. cerevisiae, we were unable to observe re-circularization via NHEJ; however, it is likely that there is still a low level of NHEJ activity. Consequently, there is little change in the frequency of HR in both S. cerevisiae wildtype and dnl4D. In contrast, in C. glabrata, we observe insertion and vector ligation via NHEJ. Deletion of DNL4 results in an increase in HR, indicating that NHEJ is more prevalent in C. glabrata. doi:10.1371/journal.pone.0069628.g007 Assessment of the Impact of Chromosomal Integration on GRC A possible complication to our work is that URA3 + could be integrated into the chromosomal genome via HR or NHEJ at the same time that vector re-circularizes, overestimating the frequency of GRC. To determine whether integration into the genome of the URA3 + gene could cause a Ura + phenotype independent of plasmid re-circularization, we performed a reciprocal experiment. We transformed both species as before, but selected for URA3 + first and measured the frequency of these colonies that were HIS3 + . In S. cerevisiae, we did not observe any Ura + colonies in the absence of linearized vector and we only observed Ura + His + colonies during co-transformation with both PCR product and linearized vector, suggesting little/no NHEJ.
In C. glabrata, we observed Ura + colonies with PCR product alone; however, the frequency of these colonies was ,5% of the maximal number of colonies observed when linearized vector was present. Additionally, during co-transformation, we observed ,50% of colonies that were Ura + but not His + . Initially, this might suggest that NHEJ and chromosomal integration is very high; however, the CgURA3 PCR product is derived from C. glabrata, and we hypothesized that HR between the PCR product and the endogenous Cgura3::NATMX6 locus might be occurring in these transformations. Because we deleted CgURA3 in the genome with the NATMX6 cassette, we expected loss of the NATMX6 cassette if CgURA3 is integrated into the genome through HR. Importantly, all Ura + His + colonies were resistant to nourseothricin and all Ura + His 2 colonies were sensitive, indicating that HR is far more prevalent than NHEJ in C. glabrata.

Discussion
The goal of this project was to characterize GRC in S. cerevisiae and C. glabrata. Our results suggest that GRC can be effectively carried out in both yeast species. This has previously been demonstrated in S. cerevisiae but not in C. glabrata. We were also able to uncover the determinants of GRC for both species.
Our data indicate that 20 bp of homology on each side of the PCR product is required for detectable GRC to occur. An insert:vector molar ratio of 1.0 was shown to be sufficient for obtaining insertion of the PCR product at ,50%, while increasing this ratio to 5.0 drives GRC to almost 100%. We also found that S. cerevisiae is slightly more efficient at performing GRC. This result was supported by the GRC data we obtained using the rad52D mutants. The S. cerevisiae rad52D mutant was only able to perform GRC at a very low frequency and required 40 bp of homology on each side, indicating that this species is very dependent on HR for re-circularization. The C. glabrata rad52D mutant produced re-circularized plasmids that contained URA3 at a low frequency as well. However, the sequencing data suggests that NHEJ, in addition to HR, is an option for inserting URA3 into the vector ( Figure 6). The presence of NHEJ activity in C. glabrata may be an effect of slower DNA degradation in this species, although we have not tested that hypothesis in this work. Interestingly, sequencing from a separate experiment (data not shown) indicated that C. glabrata displayed a strong preference for performing HR. In this experiment, a sequence contained a fragment of salmon sperm DNA that had been inserted into the vector by GRC, with only 11 bp of homology on one side of the salmon sperm DNA fragment. This fortuitous finding suggests DNA may persist longer in C. glabrata, allowing more time for NHEJ or extremely rare HR events to occur, even with unintended targets, such as small fragments of carrier salmon sperm DNA.
The S. cerevisiae dnl4D mutant shows a similar rate of GRC as the wild-type strain. Here again, our data suggest that S. cerevisiae may not be capable of performing NHEJ at a detectable rate in this GRC assay, indicating HR is the only means to re-circularize DNA through insertion. In contrast to S. cerevisiae, the Cgdnl4D strain shows an increased frequency of GRC compared to wild-type, suggesting that NHEJ activity is detectable in this GRC assay and that C. glabrata has more NHEJ activity in general.
Based on results from all of the experiments that were performed, we generated a model for the fate of linear DNA in S. cerevisiae and C. glabrata (Figure 7). In S. cerevisiae, it appears that the linear DNA (both pRS313 and PCR product) takes two out of three avenues in the majority of instances: HR or vector degradation (the eventual fate of transformed linear DNA in cells if vector is not re-circularized). Removing NHEJ activity in S. cerevisiae, through deletion of DNL4, has little effect on GRC, because there is little NHEJ to begin with. On the other hand, the fate of linear DNA in C. glabrata is influenced by HR, NHEJ, or degradation. When NHEJ activity is deleted in C. glabrata, there is an increase in HR. Although both species are capable of performing GRC at a high level, this model suggests that S. cerevisiae is the preferred species in which to perform this technique. Our data indicate that NHEJ is more prevalent in C. glabrata; however we cannot fully determine whether this is because NHEJ is inherently more active in this species, or whether linear DNA is degraded more slowly in C. glabrata and the NHEJ machinery is able to work for a longer period. Figure S1 S. cerevisiae Ura + His + cells are primarily a consequence of GRC. Using 40 bp homology on each side, colonies that were Ura + His + were picked to a YEPD plate and then replica-plated to 5-FOA plates and colonies were scored for growth. Then, the colonies were replica-plated to medium lacking uracil or histidine to assess whether the two markers were coupled on a plasmid. (EPS) Figure S2 C. glabrata Ura + His + cells are primarily a consequence of GRC. The same experiment was performed as described in Figure S1. One colony out of the 113 was identified as His + and Ura 2 and is potentially a consequence of plasmid integration into the genome. (EPS)