The induction of recombination by transcription activation has been documented in prokaryotes and eukaryotes. Unwinding of the DNA duplex, disruption of chromatin structure or changes in local supercoiling associated with transcription can be indirectly responsible for the stimulation of recombination. Here we provide genetic and molecular evidence for a specific mechanism of stimulation of recombination by transcription. We show that the induction of deletions between repeats in hpr1Δ cells of Saccharomyces cerevisiae is linked to transcription elongation. Molecular analysis of different direct repeat constructs reveals that deletions induced by hpr1Δ are specific for repeat constructs in which transcription initiating at an external promoter traverses particular regions of the DNA flanked by the repeats. Transcription becomes HPR1 dependent when elongating through such regions. Both the induction of deletions and the HPR1 dependence of transcription were abolished when a strong terminator was used to prevent transcription from proceeding through the DNA region flanked by the repeats. In contrast to previously reported cases of transcription‐induced recombination, there was no correlation between high levels of transcripts and high levels of recombination. Our study provides evidence that direct repeat recombination can be induced by transcriptional elongation.
An interesting aspect of the genetic control of recombination is its putative connection to transcription. Recombination has been shown to be stimulated by transcription in both prokaryotes and eukaryotes. The first evidence for this phenomenon was obtained by the isolation of the recombination hotspot HOT1 (Keil and Roeder, 1984), a sequence from the rDNA region of Saccharomyces cerevisiae which is involved in the regulation of transcription by RNA polymerase I (Voelkel‐Meiman et al., 1987). The ability of HOT1 to enhance recombination is related to its capacity to promote transcription (Stewart and Roeder, 1989). Stimulation of recombination by the activation of RNA polymerase II‐mediated transcription of GAL10 has been also shown (Thomas and Rothstein, 1989). Other examples of transcription‐induced recombination are recombination at ADH1 in Schizosaccharomyces pombe (Grimm et al., 1991), mating‐type switching (Klar et al., 1981) and Ty recombination in S.cerevisiae (Nevo‐Caspi and Kupiec, 1994), intra‐ and interchromosomal recombination (Nickoloff and Reynolds, 1990; Nickoloff, 1992) and gene targeting (Thyagarajan et al., 1995) in mammalian cells, and site‐specific recombination in the immunoglobulin genes (Blackwell et al., 1986; Leung and Maizels, 1992; Lauster et al., 1993; Oltz et al., 1993). Stimulation of recombination by transcription has also been reported in virus and bacteria (Bernardi and Bernardi, 1988; Dul and Drexler, 1988a,b; Bourgeaux‐Ramoisy et al., 1995; Vilette et al., 1995).
Certainly, transcription‐driven recombination may be an indirect consequence of the effect of transcription on DNA. Thus, the unwinding of the DNA duplex, the changes in the local supercoiling of DNA or the disruption of chromatin structure associated with transcription could provide a better accessibility of recombination proteins to the transcribed DNA, could lead to DNA structures hypersensitive to endogenous nucleases or could facilitate a pairing reaction. There are additional reports that indirectly suggest that these types of alterations in the DNA structure can induce recombination. Thus, it has been shown in S.cerevisiae that the natural sites for the formation of double‐strand breaks initiating meiotic recombination map in promoter regions (Nicolas et al., 1989), and that mutations in the DNA topoisomerase structural genes, TOP1, TOP2 (Christman et al., 1988) and TOP3 (Wallis et al., 1989), and in the transcriptional regulators involved in chromatin structure, SIR2 (Gottlieb and Esposito, 1989), SPT4 and SPT6 (Malagón and Aguilera, 1996), confer hyper‐recombination between different types of repeats. In addition, it has been shown in vitro that: (i) transcription‐driven site‐specific recombination is linked to the negative DNA supercoiling transiently built by an advancing RNA polymerase (Dröge, 1993), and (ii) transcription activates RecA‐promoted homologous pairing of nucleosomal DNA (Kotani and Kmiec, 1994).
However, a direct role for the transcription machinery itself in recruiting recombination proteins to transcribed genes cannot be dismissed. Thus, it has been shown that the binding of yeast transcription factors to a promoter increases meiotic recombination (White et al., 1991, 1993), that the meiotic recombination hotspot located in the Eβ gene of the mouse major histocompatibility complex is adjacent to the binding motifs of known transcription factors (Shenkar et al., 1991) and that the human recombination signal sequence‐binding protein RBP2N functions as a transcriptional repressor (Dou et al., 1994). In this sense, the recent identification of the RecA‐like protein RAD51 as a component of the human RNA polymerase II holoenzyme is particularly intriguing (Maldonado et al., 1996).
The genetic and molecular analysis of mutants affected in both recombination and transcription should contribute greatly to the understanding of the connection between both processes. One such mutant is hpr1Δ of S.cerevisiae. Null hpr1Δ strains show a strong induction (50‐ to 1000‐fold) of recombination, leading to deletions between direct repeats (Aguilera and Klein, 1990). This makes the HPR1 gene particularly interesting for its special relevance in the genomic stability of DNA repeats. In addition, hpr1Δ mutants show high frequencies of chromosome loss (Santos‐Rosa and Aguilera, 1994) and are thermosensitive (ts) for growth (Fan and Klein, 1994). Recently, it has also been reported that HPR1 functions as a transcription factor required for the activation of different promoters (Zhu et al., 1995). These data suggest that HPR1 participates in both transcription and recombination between repeats. This is supported further by the observations that mutations in different transcription factors suppress either the hyper‐recombination (Piruat and Aguilera, 1996; Santos‐Rosa et al., 1996) or the ts phenotypes (Fan et al., 1996; Uemura et al., 1996) of hpr1Δ and that hpr1Δ‐dependent deletions between GAL10 repeats are stimulated under conditions of activation of transcription (Fan et al., 1996).
In order to understand, on the one hand, how deletions between repeats occur in connection with transcription and, on the other hand, the mechanism by which deletions are induced in hpr1Δ mutants, we have performed a molecular analysis of the effect of hpr1Δ on a number of different direct repeat constructs. After showing that the hyper‐rec phenotype is not general but specific to particular repeats, we demonstrate that deletions occurring in hpr1Δ cells depend on the progression of transcription through particular DNA regions. This is the first time that it has been shown that transcription elongation can induce recombination between repeats.
The hyper‐recombination phenotype of hpr1Δ is only observed in particular DNA repeats
We have determined the effect of the hpr1Δ mutation on recombination of three previously characterized DNA direct repeat constructs (Prado and Aguilera, 1995) which are based on the same 0.6 kb repeat sequence and separated by intervening regions of 5.57 kb (LY), 31 bp (L) and 2.51 kb (LU construct). All these recombination substrates are contained on CEN‐based plasmids that replicate autonomously. Figure 1 shows that the frequency of recombination of hpr1Δ cells was 56.5 times above the wild‐type value for the LY construct. However, hpr1Δ does not have a significant effect on the levels of recombination (deletions) of the L and LU direct repeat constructs. These results indicate that the hyper‐recombination phenotype produced by hpr1Δ is not general, but particular to DNA repeat constructs, consistent with our previous hypothesis that the function of HPR1 should not be equally relevant for all chromosomal regions or sequences (Santos‐Rosa and Aguilera, 1994). Indeed, a very weak and basal hyper‐rec phenotype (<5‐fold) is observed in all repeat constructs studied in hpr1Δ cells. We will refer to those repeat constructs showing an increase of recombination of <5‐fold as ‘non‐hyper‐rec’, and to those showing an increase of >20‐fold as ‘hyper‐rec’.
The hyper‐recombination phenotype of hpr1Δ is not dependent on either the DNA sequence per se or the length of the intervening region
We have suggested that the hyper‐recombination phenotype of hpr1Δ cells should be caused by an increase in the frequency of initiation events (Aguilera and Klein, 1990; Santos‐Rosa and Aguilera, 1994) and that spontaneous deletions in our direct repeat constructs initiate primarily at the intervening region (Prado and Aguilera, 1995). Since the LY (hyper‐rec in hpr1Δ strains) and LU (non‐hyper‐rec) direct repeat constructs differ only in a 3.03 kb ClaI–SmaI fragment from pBR322, present in the intervening region of LY and absent from LU (see Figure 1), we considered the possibility that such a 3.03 kb fragment contained a cis‐element required for the hpr1Δ‐induced recombination events. In order to identify such a putative cis‐element, we made DNA repeat constructs carrying deletions of the intervening region of LY. The analysis of recombination of such DNA repeats is shown in Figure 1.
We first confirmed that when the 0.6 kb direct repeats were flanking almost exclusively the LY‐specific 3.03 kb fragment (LYΔNS construct), recombination was stimulated in hpr1Δ cells to the same levels as construct LY (47‐ versus 57‐fold). Further analysis revealed that the sequences conferring the hpr1Δ‐dependent hyper‐rec phenotype could be located in the right side of the 2.84 kb internal fragment (compare LR and LP constructs).
Unexpectedly, the analysis of 12 DNA repeat constructs carrying serial deletions of the 3.03 kb region covering the pBR322 sequences immediately to the left of URA3 (Δ‐repeat constructs) revealed that such a region could be deleted from LY without significantly affecting the hpr1Δ‐dependent hyper‐rec phenotype (data not shown). The two Δ‐repeat constructs shown in Figure 1 (Δ20 and Δ21) have frequencies of recombination 20–35 times above the wild‐type value, an increase significantly higher than that of LU (3.8‐fold). Since the Δ‐repeat constructs retained the 2.2 kb right side fragment of the intervening region of LY, it seemed possible that such fragments could contain a second cis‐element able to induce recombination in hpr1Δ cells (see Δ21 in Figure 1). The 0.8 kb internal region of the yeast URA3 sequence, contained in the aforementioned 2.2 kb fragment, was indeed essential for conferring the hyper‐rec phenotype (see LYΔP construct). However, these results raised the paradox of why the LU repeat did not show a strong hyper‐rec phenotype in hpr1Δ cells, even though it contained the 2.2 kb fragment.
It could be possible that LU contained an additional cis‐element with a dominant‐negative effect on the 2.2 kb fragment. One candidate for such a cis‐element was the extra 54 bp containing the putative URA3 terminator of transcription located on the left side of the intervening region of LU, which is absent from the Δ‐repeat constructs. The addition of such a 54 bp URA3 sequence to Δ21 (see LNB construct, Figure 1) abolished the strong hpr1Δ‐dependent hyper‐rec phenotype. However, LNA, identical to LNB but with the intervening DNA region in the opposite orientation, shows a strong hpr1Δ‐dependent hyper‐rec phenotype. These results indicate that the DNA sequence alone is not sufficient to explain the hyper‐rec phenotype conferred by hpr1Δ, since the same intervening sequence confers hyper‐recombination in one orientation (LNA) but not in the opposite orientation (LNB).
Another factor that is not sufficient to explain the hpr1Δ‐induced hyper‐rec phenotype is the length of the intervening region. Not only do constructs with the same size of intervening region (Δ21, LNA and LNB) show different recombination phenotypes, but the length of the intervening region of hpr1Δ‐dependent hyper‐rec repeats ranges from 2.2 to 5.57 kb, and that of the non‐hyper‐rec constructs ranges from 0.6 to 2.54 kb. Consistent with this conclusion, a DNA repeat construct (LA) carrying the 0.6 kb leu2 repeats and a 3.6 kb ADE2 region as intervening DNA does not show hpr1Δ‐dependent hyper‐rec phenotype (frequencies of 5×104 and 8×104 in hpr1Δ and wild‐type cells, respectively).
Therefore, the different recombination behaviour of the DNA direct repeats used in this study in hpr1Δ cells has to be caused by other factors besides the presence of specific DNA sequences or the length of the intervening region.
The strong hpr1Δ hyper‐rec phenotype is linked to transcription progression through particular DNA regions
Since it has been suggested that HPR1 is a global positive regulator of transcription (Zhu et al., 1995), and mutations in the SRB2 general transcription factor completely suppress the hyper‐recombination phenotype of hpr1Δ (Piruat and Aguilera, 1996), we decided to determine the transcription pattern of the different repeat constructs analysed. We have mapped and quantified by Northern analysis the transcripts produced by nine different DNA repeat constructs and a recombined LEU2+ plasmid, used as control. (It is important to note that all recombined LEU2+ plasmids are identical, regardless of the DNA repeat from which they are derived.) By using different DNA probes (YIp5, pBR322, URA3, ACT1, the LEU2 internal repeat and a LEU2 5′ end‐specific fragment), we were able to map and quantify all the different transcripts produced by each repeat construct. All transcripts covered either the LEU2 or the URA3 gene. With the exception of the URA3 promoter‐driven transcripts, all other transcripts detected in each DNA repeat construct were driven from the LEU2 promoter located outside of the repeats, upstream of the leu2Δ3′ truncated copy (see Figure 1). This conclusion was suggested by the observation that all such transcripts hybridized with both LEU2 probes, regardless of whether or not they also hybridized with pBR322 (data not shown), and was confirmed by RNase A protection analysis (data not shown). All episomal LEU2‐derived RNAs initiate at the LEU2 transcription initiation region (Andreadis et al., 1984) and are transcribed in the same direction as the LEU2 gene. In consequence, the different length of the transcripts is caused by different termination sites.
The analysis of the repeats that do not show a strong hpr1Δ‐dependent hyper‐rec phenotype (Figure 2A) reveals that the major transcripts derived from each construct terminated either downstream of the first repeat, directly after entering the intervening region, or at the LEU2 terminator located downstream of the second leu2 repeat. The latter case was found mainly for DNA repeats containing short intervening regions (L and LP). However, in the repeat constructs with a strong hpr1Δ‐dependent hyper‐rec phenotype (Figure 3A), all major transcripts covered the intervening DNA, traversing DNA regions that were not present in the non hyper‐rec constructs. Among these regions, we find the sense strands of either the bacterial Amp (LR, LY) or Tet genes (LNA), or the nonsense strand of the yeast URA3 gene (Δ20).
There was no difference between wild‐type and hpr1Δ strains in the amount of transcripts produced by the non‐hyper‐rec constructs (Figure 2B). However, in the strongly hyper‐rec constructs, the levels of all transcripts driven from the external LEU2 promoter in hpr1Δ were 2–3 times lower than in wild‐type (Figure 3B). Therefore, our results indicate that transcription becomes HPR1 dependent when it progresses through particular regions located between the repeats, and this property is unique to DNA repeat constructs that show a strong hyper‐rec phenotype in hpr1Δ strains.
A comparative analysis of transcription of all repeat constructs (Figures 2 and 3) reveals that the size of all transcripts driven from the LEU2 external promoter are identical in both wild‐type and hpr1Δ cells, indicating that the sites of transcription initiation and termination do not depend on HPR1. Also, the overall levels of transcripts terminating at known yeast transcription termination sequences were much higher (>10‐fold in wild‐type cells) than those of the transcripts terminating at bacterial or yeast sequences which do not naturally function as transcription terminators in yeast (compare LU and LNB from Figure 2 with LY, Δ20 and LNA from Figure 3). This is independent of the hpr1Δ mutation. In addition, it is interesting to note that the LEU2+ transcript of the LEU2+ recombined plasmid is significantly less abundant than that of LU or LNB, even though they all terminate at natural yeast terminators. We do not have a clear explanation for this. A particular topological configuration of the LEU2+ recombined plasmid or some other structural features as well as the primary sequence of the LEU2+ mRNA might be among the factors that could be affecting either transcription efficiency or transcript stability of the episomal LEU2+ mRNA.
The relevant differences between the hyper‐rec and non‐hyper‐rec repeat constructs observed in this study are illustrated by the results of the non‐hyper‐rec LNB (Figure 2) and the hyper‐rec LNA (Figure 3) repeat constructs, which differ only in the orientation of the intervening region. Whereas in LNB the strong URA3 terminator of the intervening sequence is located downstream of the first leu2 repeated sequence, in LNA the URA3 terminator is at the other end of the intervening region, upstream of the second leu2 repeated sequence. As a consequence, the transcript of LNB, initiated at the external LEU2 promoter, only covers the first leu2 repeated sequence (1.05 kb transcript). However, all transcripts of LNA initiating at the same LEU2 promoter proceed through the first leu2 repeated sequence, traverse the intervening region and terminate at a region at the 5′ end of URA3 (2.25 kb transcript). (Only a minor proportion of the transcripts do not terminate at such a region.) In addition, the LNB transcripts were equally abundant in hpr1Δ and wild‐type strains, whereas the LNA transcripts were ∼3 times less abundant in hpr1Δ than in wild‐type strains.
Our results also suggest that the lower levels of transcripts observed for the hyper‐rec constructs in hpr1Δ strains, as well as the hyper‐rec phenotype, are not a consequence of differences in transcription termination between hpr1Δ and wild‐type cells. Figures 2 and 3 show that transcription terminates at the same 5′ region of URA3 in both the non‐hyper‐rec LYΔP and the hyper‐rec LNA construct. (This region contains transcription termination sequences of the upstream open reading frame.) However, although the hpr1Δ mutation does not affect either recombination or the level of transcripts in LYΔP, it conferred a hyper‐rec phenotype as well as a significant reduction (2‐ to 3‐fold) in the overall levels of transcripts in LNA.
A transcription terminator impeding transcription from elongating through the intervening region abolishes hyper‐recombination
We decided next to confirm directly the hypothesis that the progression of transcription through the intervening region is required for a strong stimulation of recombination and a significant reduction in the level of transcripts in hpr1Δ versus wild‐type cells (Figure 3). To do so, we put a strong transcription terminator (CYC1ter) downstream of the leu2Δ3′ repeat in the hyper‐rec construct LNA. Figure 4 shows that the presence of CYCter (LNAT construct) efficiently terminates transcription initiated at the LEU2 promoter, impeding, therefore, the progression of transcription through the intervening region. Consistent with our predictions, such a strong terminator abolished both the hyper‐recombination phenotype and the expression pattern of LNA (the new 1.15 kb transcript observed in LNAT was now equally abundant in wild‐type and hpr1Δ strains). Also, the overall expression levels of the 1.15 kb transcripts of LNAT were eight (wild‐type strains) and 25 times (hpr1Δ) above the levels of the 2.25 kb transcripts of LNA, consistent with our observation that termination of transcription at strong and natural yeast terminators led to high levels of transcripts in all other repeat constructs.
To confirm that transcription termination was not responsible for the hyper‐rec and the transcription phenotypes of hpr1Δ cells and that the CYCter sequence itself was not responsible for the lack of hpr1Δ‐dependent hyper‐rec phenotype of LNAT, we decided to place the CYCter sequence into the intervening region of LNA, 0.6 kb further downstream of the 3′ end of the leu2Δ3′ repeat (LNATD construct). As expected, in this new construct, the LEU2‐driven transcript is 0.6 kb longer than in LNAT and also terminates at CYCter (Figure 4). As can be seen in Figure 4, the LNATD construct shows 11 times more recombinants in hpr1Δ than in wild‐type cells and led to a 1.75 kb transcript which is 1.7 times less abundant in hpr1Δ than in wild‐type cells. Therefore, the new repeat construct becomes sensitive to hpr1Δ, even though to a lesser extent than LNA, suggesting that the extra 0.6 kb sequence that is transcribed covers part of the region responsible for the phenotypes observed in LNA. In addition, it is clearly observed that transcript levels in both wild‐type and hpr1Δ cells are significantly less abundant in LNATD as compared with LNAT. These results indicate that the elongation of transcription through the 0.6 kb fragment of the intervening region causes both the general decrease in transcript levels and the hpr1Δ‐dependent hyper‐recombination and transcription patterns of LNATD. Therefore, these experiments exclude any role for transcription termination in the transcription and recombination phenotypes of hpr1Δ.
Recombination induced by hpr1Δ does not initiate at regulated promoters
If HPR1 were a general activator of transcription (Zhu et al., 1995), we would expect it to perform its function at promoter regions. If recombination events in hpr1Δ were a direct consequence of transcriptional activation defects, they should be induced by regulated promoters. To assess this possibility, we constructed 0.6 kb LEU2 direct repeats carrying as intervening regions the GAL1‐ or the PHO5‐regulated promoters. The new DNA repeat constructs (LGAL1‐1, LGAL1‐5 and LPHO5) should show strong hyper‐rec phenotypes under activated conditions of transcription and wild‐type levels of recombination under repression conditions. Figure 5 shows that this is not the case. In the three DNA repeat constructs tested, there is a 2.0‐ to 3.6‐fold stimulation of recombination under repressed conditions that increases up to 7.3‐ (repeats with GAL1p as intervening region) or 9.1‐fold (PHO5p) under induced conditions in hpr1Δ strains, implying that the increase in recombination caused by hpr1Δ in connection with transcription activation of promoters is at most 3–fold. Therefore, the major recombination events observed in hpr1Δ cells do not initiate at regulated promoter sequences nor are they related to any role HPR1 might have in transcription activation.
The central conclusion of this study is that recombination leading to deletions between DNA repeats can be induced by transcription elongation. We provide evidence that the recombination events occurring in yeast hpr1Δ cells are associated with the elongation of transcription through particular DNA regions flanked by repeated sequences.
Recombination induced by hpr1Δ is associated with transcriptional elongation
We have shown that the recombination phenotype of hpr1Δ is not general but is specific to particular DNA repeat constructs. This is consistent with our previous hypothesis that the hyper‐rec phenotype of hpr1Δ cells should not be general for all chromosomal regions (Santos‐Rosa and Aguilera, 1994). Whether or not a direct repeat construct has a strong hyper‐rec phenotype (50‐fold stimulation of recombination) or not (<5‐fold) does not depend on either the DNA sequence per se or the length of the intervening region (Figure 1), but on transcription elongation. All tested repeat constructs are transcribed from the same upstream LEU2 external promoter. However, a characteristic unique to the strongly hyper‐rec constructs is that the transcripts driven from such a LEU2 promoter elongate through specific regions contained in the intervening DNA flanked by the repeats (Figure 3). The constructs in which transcription do not proceed through such DNA regions, either because they are not contained in the intervening DNA or because a terminator impedes transcription from entering the intervening region, are not strongly hyper‐rec (Figure 2). Indeed, a strongly hyper‐rec construct could be converted into non‐hyper‐rec by inserting a strong transcription terminator between the upstream repeat unit and the intervening region (Figure 4).
Further evidence that hyper‐recombination is linked to transcription elongation is provided by the observation that the levels of transcripts driven from the LEU2 promoter are significantly lower (3‐fold) in hpr1Δ than in wild‐type cells in the strongly hyper‐rec constructs, but identical in non‐hyper‐rec constructs. This rule has no exception for any of the repeat constructs studied. Thus, even the few transcripts that elongate through the intervening region in the non‐hyper‐rec constructs are equally abundant in wild‐type and hpr1Δ cells (Figure 2). Therefore, it is the passing of transcription through specific regions in the intervening DNA which is linked to both the lower levels of transcripts and the higher incidence of recombination in hpr1Δ cells.
The reduction in the levels of transcription of the hyper‐rec constructs in hpr1Δ cells could be consistent with the recent observation implying that HPR1 participates in transcriptional activation of promoters (Fan and Klein, 1994; Zhu et al., 1995). However, expression from the LEU2 promoter is not dependent on HPR1 (see endogenous LEU2 in Figures 2,3,4), nor is it obvious how different transcribed DNA sequences can affect the strength of its promoter directly. Our results suggest that the lower levels of transcription observed in hpr1Δ cells in the hyper‐rec constructs are caused by defective elongation. Such defective elongation can either slow down the rate of transcription (Krassimir et al., 1994) or lead to abortive transcripts, presumably undetectable by Northern analysis, given the instability of mRNA not terminating at the appropriate termination signals (Brawerman, 1987; Ross, 1995). Indeed, we have observed that, regardless of the strain background used, the transcripts driven from the same LEU2 promoter accumulate at higher levels (>10‐fold) when terminating at natural yeast terminators than when terminating at weak non‐natural terminators. The exception is construct LNATD which shows lower levels of transcripts terminating at the CYCter sequence (Figure 4). This suggests that the elongation through particular DNA regions can also lead to a reduction in transcript levels.
Transcription termination is not responsible for the different recombination and transcription phenotypes reported here. Transcription terminates at the same region in two direct repeats (LYΔP and LNA) with opposite patterns of recombination. In addition, the CYCter terminator of transcription abolishes the recombination and transcription phenotypes of hpr1Δ when located immediately downstream of the transcribed repeat (LNAT), but not when located 0.6 kb further downstream (LNATD) in the intervening region (Figure 4). We have also shown that the suggested promoter–activator function of HPR1 (Zhu et al., 1995) is not related to the hyper‐recombination phenotype of hpr1Δ cells either, since two different regulated promoters are not able to strongly induce deletions between repeats when used as intervening regions.
Therefore, our result suggests that a defective elongation of transcription is a cause of the hyper‐recombination phenotype and its associated reduction in transcript levels in hpr1Δ cells. This defective elongation is not equally relevant in all DNA regions. Only when the elongation through specific DNA regions is clearly altered (either retarded, stalled or arrested) does it produce a detectable reduction in the levels of transcripts (3‐fold) and a strong hyper‐rec phenotype (>50‐fold). One strand of the Amp, Tet or URA3 regions are candidate regions where the RNA polymerase could pause in the absence of HPR1. In this sense, the LNATD construct is particularly revealing, because it shows that transcription through the 0.6 kb region including the Tet gene reduces the transcript levels in wild‐type and, to a major extent, in hpr1 cells. It is likely that the proper progression of transcription throughout this region is HPR1 dependent.
How does transcription elongation induce deletions between repeats?
The characteristics of the induction of recombination by transcription in hpr1Δ cells are different from those of previously reported cases in which high recombination rates are associated with high transcription levels (Voelkel‐Meiman et al., 1987; Stewart and Roeder, 1989; Thomas and Rothstein, 1989; Nickoloff, 1992; Vilette et al., 1992). In hpr1Δ‐induced recombination, high transcription levels (LU, LNB, LNAT) are associated with low recombination levels (Figure 2). Therefore, the strong hpr1Δ hyper‐rec phenotype cannot be explained easily by an indirect effect of transcription on DNA structure, such as the unwinding of the DNA duplex, the accumulation of local negative supercoiling behind the advancing RNA polymerase or the disruption of the chromatine structure.
We believe that HPR1 is required to allow the RNA polymerase II to traverse specific DNA structures that would otherwise pause or block transcription in the absence of HPR1. A stalled transcription complex might confer nuclease‐hypersensitive sites, generate recombinogenic lesions (DNA breaks) or interfere with the replication machinery, causing deletions if flanked by direct repeats (Piruat and Aguilera, 1996). In this sense, illegitimate deletions in Escherichia coli have been suggested to be stimulated by collisions between converging replication and transcription machineries (Vilette et al., 1995). Whether or not the transcription machinery could facilitate the recruitment of recombination proteins, as in nucleotide excision repair (Basthia et al., 1996; Hoeijmakers et al., 1996) is certainly unknown. Regardless of the mechanism used, our results suggest that transcription elongation may be a source of mitotic recombination in S.cerevisiae.
Materials and methods
Strains and plasmids
Media and growth conditions
Standard media such as rich medium YEPD, synthetic complete medium (SC) with bases and amino acids omitted as specified, and sporulation medium were prepared as described previously (Sherman et al., 1986). Yeast strains were transformed using the lithium acetate method (Ito et al., 1983) modified according to Schiestl and Gietz (1989).
Genetic analysis and determination of recombination frequencies
Genetic analysis was performed as previously published (Sherman et al., 1986).
Recombination frequencies were calculated as the median value of 3–4 fluctuation tests, each one performed with six independent colonies for each transformant studied. Yeast strains were grown on SC‐trp. After 3 days, independent colonies were plated on SC‐leu to determine the median frequency of Leu+ recombinants. The viable cell number was determined on SC‐trp.
Nested deletions were obtained with E.coli ExoIII (Pharmacia) according to published procedures (Henikoff, 1984). Plasmid pRS314‐LY was digested with SmaI and NsiI prior to ExoIII treatment. Clones carrying deletions differing in length by 200–400 bp were selected for further studies.
[32P]dCTP‐labelled DNA probes were prepared as previously described (Feinberg and Vogelstein, 1984).
RNA–DNA hybridization was performed in 50% formamide 5× SSPE, 1× Denhardt's solution, 1% SDS at 42°C for 16 h. DNA probes used were labelled with [32P]dCTP.
RNase A protection assays were performed as described (Forrester et al., 1987) with the modifications of Almoguera et al. (1995). RNA probes were prepared by in vitro transcription from the T3 and T7 promoters of the pBSRS plasmid cut with either HindIII or EcoRI, respectively. RNA–RNA hybridization was carried out in 80% formamide, 0.4 M NaCl, 1 mM pH 8 EDTA, 40 mM pH 6.7 PIPES at 48°C for 16 h. The RNA–RNA hybrids were treated with RNase A at 30°C for 30–60 min. The resulting RNA fragments were analysed in 4% PAGE gels. The 768 bp RNAs derived from the SspI–EcoRI LEU2 region transcribed from either a T3 or T7 promoter (see text) were used as RNA probes.
RNA was quantified from Northern experiments using an InstantImager (Packard, USA). All data from different samples were normalized in relation to the values obtained for the ACT1 and LEU2 mRNAs obtained from the endogenous chromosomal genes.
We thank S.Chávez for helpful discussions, S.Chávez and E.Santero for critical reading of the manuscript, M.Funk and F.Estruch for generous gifts of plasmids, C.Almoguera for technical advice on RNase A protection assays, and W.Reven for style correction. This work was supported by DGICYT grant PB93‐1176‐C02‐01 from the Ministry of Science and Education of Spain and a grant from the Regional Government of Andalucía (Spain) to A.A. F.P. and J.I.P were recipients of predoctoral training grants from the Ministry of Science and Education of Spain.
- Copyright © 1997 European Molecular Biology Organization