Recombination Regulation by Transcription-Induced Cohesin Dissociation in rDNA Repeats

See allHide authors and affiliations

Science  02 Sep 2005:
Vol. 309, Issue 5740, pp. 1581-1584
DOI: 10.1126/science.1116102


Organisms maintain ribosomal RNA gene repeats (rDNA) at stable copy numbers by recombination; the loss of repeats results in gene amplification. Here we report a mechanism of amplification regulation. We show that amplification is dependent on transcription from a noncoding bidirectional promoter (E-pro) within the rDNA spacer. E-pro transcription stimulates the dissociation of cohesin, a DNA binding protein complex that suppresses sister-chromatid-based changes in rDNA copy number. This transcription is regulated by the silencing gene, SIR2, and by copy number. Transcription-induced cohesin dissociation may be a general mechanism of recombination regulation.

In most organisms, recombination is necessary for DNA repair, chromosome segregation, and the rescue of stalled replication forks. If not properly regulated, however, recombination can lead to genomic instability (1) and can be toxic to cells (2). It is not clear how cells maintain only the positive effects of recombination.

In repeated-gene families, such as the ribosomal RNA (rRNA) gene repeats (rDNA), recombination helps maintain copy number (3) and the evolutionary stability of the repeats (4). The number of rDNA copies is tightly regulated; if repeats are deleted or inserted, copy number is quickly restored to that of the wild type (5, 6). One way that copy number is maintained is by gene amplification after deletional recombination. In the yeast Saccharomyces cerevisiae, this amplification is dependent on the replication-fork blocking protein, FOB1, and a ∼520-base pair (bp) cis-acting factor called EXP, which is found in the rDNA intergenic spacer (IGS) (Fig. 1A) (6, 7). In a recent Saccharomyces species phylogenetic foot-printing study, we found a highly conserved sequence that corresponds to a previously identified bidirectional RNA polymerase II (pol II) promoter (8) in EXP (9). This EXP promoter (named E-pro) does not appear to be associated with any coding function, and its position and conservation suggested it might play a role in rDNA amplification.

Fig. 1.

Bidirectional E-pro transcription is required for rDNA amplification. (A) rDNA occupies ∼60% of chr XII in S. cerevisiae. The 35S and 5S rRNA genes, IGS1 and 2, the origin of replication (rARS), the RFB, EXP, E-pro, and the CAR are indicated. Locations of Northern probes (NL and NR) and ChIP primers (E1 to E7) are shown. (B and C) CHEF gel showing chr XII sizes of various two-copy strains ∼45 generations after FOB1 or control plasmid (vector) transformation. WT, wild-type two-copy strain (TAK201). E-pro was replaced with the following: (i) an empty cassette (Δ, strain TAK222); (ii) a bidirectional Gal promoter in both orientations (GAL1/10 +/-, strains TAK223 and TAK224); and (iii) a unidirectional Gal promoter (GAL7), with transcription in the RFB (R, strain TAK225) and rARS (L, strain TAK226) directions (25). Two independent transformants were analyzed for each mutant. M is the Hansenula wingei marker. (B) is an ethidium bromide-stained gel and (C) is an autoradiogram of (B), probed with an rDNA probe showing chr XII position. (D) CHEF gel showing the effects of repressing the GAL1/10 promoter and of transcription termination on rDNA amplification. In the left panel, amplification was induced as in (B), and after ∼45 generations, half of the cultures were shifted to glucose media. Chr XII sizes were observed as in (C) at various generations after FOB1 transformation. The right panel shows GAL1/10 transcription inhibited in each direction by a pol II (URA3) terminator (fig. S1). Termination (Ter.) +/- (strains TAK227 and TAK228) indicate directions of transcription termination. C indicates the strain before FOB1 transformation.

To determine whether E-pro is involved in rDNA amplification, we replaced it with galactose-inducible pol II promoters (unidirectional GAL7 and bidirectional GAL1/10 promoters) (fig. S1) in an S. cerevisiae strain containing only two rDNA copies (two-copy strain), and we observed the effects on amplification. Reintroduction of a plasmid-borne FOB1 gene into the two-copy strain stimulated rDNA amplification, and the resulting rDNA copy-number increase can be visualized by an increase in the size of chromosome XII (chr XII) by using pulsed-field gel electrophoresis [contour-clamped homogeneous electric field (CHEF)] (7). The deletion of E-pro abolished amplification ability (Fig. 1, B and C), and when E-pro was replaced with the GAL7 promoter in either direction, amplification ability was not rescued. However, when E-pro was replaced with the bidirectional GAL1/10 promoter (GAL1/10 strain), the introduction of FOB1 resulted in amplification.

To confirm that amplification depends on E-pro transcription, we changed the carbon source from galactose to glucose to inhibit transcription in the GAL1/10 strain. The size of chr XII continued to increase in galactose-grown cells, but did not increase in glucose-grown cells over 150 generations (Fig. 1D). Furthermore, the chr XII bands of cells grown in glucose were sharp, indicating the inhibition of rDNA recombination (6). To investigate whether read-through transcription or another function of E-pro is required for amplification, we blocked each direction of GAL1/10 transcription by using a transcriptional terminator. Blockage in either direction resulted in the loss of amplification ability (Fig. 1D). Therefore, bidirectional E-pro transcription is essential for rDNA amplification.

How can transcription from E-pro trigger recombination and hence amplification? One way is through cohesin association. Cohesin is a multifunctional protein complex involved in chromatin structure (10), and its localization is inversely correlated with transcription, suggesting that transcription disrupts cohesin association (11, 12). Cohesin association is thought to hold chromatids in place, leading to equal (versus unequal) sister-chromatid recombination and thereby preventing changes in copy number after the formation of double-strand breaks (DSBs) (13). Thus, E-pro transcription may result in cohesin dissociation, allowing a change in copy number. Chromatin immunoprecipitation (ChIP) assays were performed with a GAL1/10 strain carrying hemagglutinin (HA) epitope-tagged Mcd1p (a cohesin complex component) in conjunction with seven rDNA primer combinations (Fig. 2). In the wild-type strain grown in both glucose and galactose, the cohesin associating region (CAR) gave the strongest signal of cohesin association, as found previously (13, 14), and the pattern in the galactose-grown GAL1/10 strain was similar. However, when grown in glucose, the GAL1/10 strain showed much stronger cohesin association throughout the IGS. Thus, the repression of E-pro transcription leads to increases in cohesin association on both sides of E-pro, not just in CAR. This increase is consistent with bidirectional transcription dissociating cohesin in both IGS1 and 2, and it suggests that unidirectional transcription leaves cohesin association on the opposite side, inhibiting unequal sister-chromatid recombination. We also tested the effect of a cohesin mutation, smc1-2 (15), in the GAL1/10 strain, and we confirmed that the amplification rate was increased (fig. S2).

Fig. 2.

Cohesin association within the IGS using ChIP assays. Wild-type (WT) (strain TAK1005) and GAL1/10 strains (TAK1006) with HA-tagged MCD1, as well as control strains without tag (labeled “NO TAG”), were grown in glucose and galactose (D and G, respectively). After formaldehyde treatment, rDNA fragments (sheared to ∼500 bp) were coimmunoprecipitated using antibodies to HA. Seven rDNA regions (E1 to E7, Fig. 1A) were analyzed by polymerase chain reaction (PCR) in two groups. Primer set E1 was used in both as a control. PCR reactions were terminated in the logarithmic phase of amplification. (A) Representative ethidium bromide-stained gels of the PCR products. IP, tagged immunoprecipitated samples; INPUT, nonimmunoprecipitated controls. (B) Quantification of cohesin association from three independent experiments. PCR products were quantified, values were corrected using NO-TAG results, and then normalized using respective INPUT values.

The silencing gene SIR2 suppresses rDNA copy-number change through effects on cohesin association, because SIR2 loss results in the loss of cohesin association in the IGS (13). SIR2 represses pol II-transcribed genes integrated in the rDNA (16, 17). We therefore speculated that SIR2 regulates recombination by repressing E-pro transcription. To test this, we used Northern blots to measure E-pro transcription levels by using wild-type and SIR2-disrupted strains with endogenous E-pro (Fig. 3, A and B). E-pro transcript levels were increased ∼16.5-fold in the ribosomal autonomously replicating sequence (rARS) direction and ∼9.5-fold in the replication fork barrier (RFB) direction when SIR2 was deleted. Transcripts could be detected in an RNA pol I mutant, suggesting that E-pro is transcribed by RNA pol II (fig. S3).

Fig. 3.

Regulation of E-pro transcription and rDNA stability by SIR2 and rDNA copy number. (A and B) Northern blot analysis of E-pro transcripts. The total RNA from wild-type (WT) (strain NOY408-1b), Δsir2 (labeled SIR2, strain TAK190), and amplifying low-copy (low) strains was hybridized with probes to the rARS-facing (Left) and RFB-facing (Right) transcripts, and an ACT1 control probe. (A) shows a representative Northern blot and (B) shows the quantification of E-pro transcript levels. Results were normalized using ACT1. (C) CHEF gel probed with an rDNA probe showing the effects of SIR2 deletion on chr XII stability. GAL1/10 (strains TAK2004 and TAK2005) and wild-type strains with and without SIR2 were grown in glucose (two independent colonies were analyzed for each). DNA loading was similar in each lane. Separation conditions differ from Fig. 1. Positions of H. wingei markers are indicated at the left. (D) rDNA copy number (6), the level of cohesin association using ChIP analyses with pooled E2 to E5 PCR products (Fig. 1A), and the cohesin association corrected by copy number were determined for wild-type (TAK1005) and low-copy (TAK1008) strains. Cohesin association is relative to wild-type. Absolute copy numbers are in parentheses.

If SIR2 is responsible for copy-number change by regulating E-pro transcription, then the deletion of SIR2 should not effect recombination in a glucose-grown GAL1/10 strain, because transcription is already repressed. To test this, we analyzed the rDNA stability of GAL1/10 and wild-type strains with and without SIR2 (Fig. 3C). As previously observed (13), the chr XII bands of a wild-type strain lacking SIR2 are smeared. In contrast, the chr XII bands in the glucose-grown GAL1/10 strain remain sharp, even after SIR2 deletion.

Finally, we tested the relationship between rDNA copy number and E-pro regulation. A strain undergoing amplification (low-copy strain) should show increased E-pro transcription and decreased cohesin association. To generate low-copy strains, we introduced a plasmid-borne FOB1 gene into two-copy strains. After 45 generations, there were ∼45 rDNA copies, indicating active amplification. Analysis with Northern blots showed that E-pro transcription in the low-copy strain was enhanced ∼4.5-fold over that in the wild-type in both directions, and the value per rDNA unit will be even higher (Fig. 3B). Furthermore, ChIP analysis with an HA-tagged MCD1 low-copy strain revealed that cohesin association was reduced to 35% of the wild-type level per unit of rDNA (Fig. 3D). Therefore, when the rDNA is amplifying, E-pro is activated and cohesin dissociates, indicating that E-pro is the major regulator of rDNA amplification.

These results suggest a model of amplification regulation where transcription of E-pro stimulates unequal recombination by disrupting cohesin association in the rDNA, thus allowing for a change in copy number (Fig. 4). Sir2p is a negative regulator of E-pro transcription, and in normal situations, its activity allows cohesin to associate throughout the IGS and thereby prevents unequal sister-chromatid recombination that leads to copy-number change. This model explains the stimulatory effects of SIR2 deletion and why the absolute level of rDNA recombination is the same, regardless of SIR2 status (13). The model is supported by evidence that Sir2p alters chromatin structure within EXP (18). DSBs in the IGS are expected to recruit cohesin (19), countering the effect of transcription. However, because not many DSBs form in the rDNA repeats (20), this effect is likely to be minor. Also, DSB formation at the RFB was similar in the GAL1/10 strain when grown on glucose/galactose (0.89/1.00, relative values). Therefore, cohesin dissociation appears to be the major role of E-pro transcription activity.

Fig. 4.

Transcription-induced cohesin dissociation model of rDNA amplification. (A) In normal situations, such as wild-type rDNA copy number, SIR2 represses E-pro activity, allowing cohesin to associate throughout the IGS. DSBs, formed by replication forks pausing at the RFB site, are repaired by equal sister-chromatid recombination, with no change in the rDNA copy number. (B) When SIR2 repression is removed, such as with sir2 mutation or low copy number, E-pro becomes active and transcription displaces cohesin. Unequal sister chromatids can then be used as templates for DSB repair, resulting in changes in the rDNA copy number. Lines represent single chromatids (double-stranded DNA). The IGS in which the replication fork is paused is expanded in the bracket.

Transcription-induced cohesin dissociation provides a potential mechanism for the well-established link between transcription and recombination (21), the molecular mechanism(s) of which have remained controversial. For instance, in immune cells, antibody gene recombination requires the transcription of flanking genes (22, 23), and this transcription-dependent recombination may be mediated through cohesin dissociation. Given the large amount of noncoding transcription recently found in higher eukaryotes (24), some of these transcripts may be involved in the regulation of cohesin association, allowing cells to regulate recombination.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Table S1


References and Notes

View Abstract

Navigate This Article