articles
Centromeric histone H2B monoubiquitination promotes noncoding transcription and chromatin integrity
npg
© 2014 Nature America, Inc. All rights reserved.
Laia Sadeghi1,2, Lee Siggens1,2, J Peter Svensson1 & Karl Ekwall1 Functional centromeres are essential for proper cell division. Centromeres are established largely by epigenetic processes resulting in incorporation of the histone H3 variant CENP-A. Here, we demonstrate the direct involvement of H2B monoubiquitination, mediated by RNF20 in humans or Brl1 in Schizosaccharomyces pombe, in centromeric chromatin maintenance. Monoubiquinated H2B (H2Bub1) is needed for this maintenance, promoting noncoding transcription, centromere integrity and accurate chromosomal segregation. A transient pulse of centromeric H2Bub1 leads to RNA polymerase II–mediated transcription of the centromere’s central domain, coupled to decreased H3 stability. H2Bub1-deficient cells have centromere cores that, despite their intact centromeric heterochromatin barriers, exhibit characteristics of heterochromatin, such as silencing histone modifications, reduced nucleosome turnover and reduced levels of transcription. In the H2Bub1-deficient cells, centromere functionality is hampered, thus resulting in unequal chromosome segregation. Therefore, centromeric H2Bub1 is essential for maintaining active centromeric chromatin. H2Bub1 levels in chromatin generally correlate with RNA polymerase II (Pol II) and transcription1,2, but H2Bub1 has also been implicated in diverse cellular functions, such as DNA replication3, differentiation4–6 and DNA-damage responses7–10. The modulation and expansion of heterochromatin regions has been reported at several loci in the absence of H2Bub1 (refs. 11,12). Early observations implicated H2B in centromere maintenance13 and, although mutations affecting the inner region of H2B can simultaneously decrease H2Bub1 and cause centromeric defects14, a direct role has not been demonstrated for histone ubiquitination in the maintenance of active centromeres. In budding yeast (Saccharomyces cerevisiae), methylation of kinetochore protein Dam1 has been found downstream of H2Bub1 (ref. 15). Fission yeast (S. pombe) and higher eukaryotes have regional centromeres; in S. pombe, these occupy 40–100 kb of the chromosome. The kinetochore attaches to the central domain of the centromere, thus allowing accurate segregation of sister chromatids. Rather than being specified by the DNA sequence, centromeres are defined epigenetically by the presence of the centromere-specific histone H3 variant CENP-A. The deposition of CENP-A is required for kinetochore assembly16–18, in a process that requires transcription19–24 (reviewed in ref. 25). Centromeres in most eukaryotes have two distinct domains: the centromeric CENP-A chromatin and the flanking pericentric heterochromatin marked by methylation of H3 Lys9 (H3K9). Both domains are needed for centromere functions (reviewed in ref. 26). In S. pombe and mice, the centromeric and pericentromeric domains are clearly defined and are associated with specialized DNA sequences. In other species, including humans, centromeric
DNA is mainly made up of repetitive α-satellite sequences (ALR), and, although the compartments are distinct, the DNA sequence does not distinguish between the two types of centromeric chromatin27,28. For the formation of pericentric heterchromatin in S. pombe, evidence of a role for noncoding RNA has become well established after the key discovery of RNA interference (RNAi) in heterochromatin formation29. From a clinical perspective, dysregulation of the H2Bub1 machinery has been linked to leukemia, breast cancer and other neoplasms30,31. The E3 ligase RNF20, which targets H2B, functions as a tumor suppressor in vivo30,32, and reduction in levels of RNF20 in mammalian cells causes compromised homologous recombination resulting in chromosomal instability and replication stress7. Depletion of RNF20 leads to altered gene expression, such as downregulation of p53 (ref. 30). Normally, in adult somatic cells, centromeres are not transcribed or are transcribed at low levels. However, the centromeres of tumor cells often have increased levels of transcription33,34. To understand the role of H2Bub1 in centromere establishment and propagation, we perturbed the H2Bub1 machinery in both S. pombe and human cells and studied the effects throughout the cell cycle. We show that noncoding centromere transcription is dependent on H2Bub1. Lack of H2Bub1, either in a mutant that cannot monoubi quitinate H2B or in cells lacking the E3 ligase (RNF20 in human cells or Brl1 in S. pombe), leads to chromosome mis-segregation. A pulse of H2Bub1 coupled to transcription of the central core of the centromere occurs before cells enter the M phase of the cell cycle. In the absence of centromeric H2Bub1, heterochromatin forms de novo over the central core, despite the maintained chromatin borders.
1Department of Biosciences and Nutrition, Karolinska Institutet, Huddinge, Sweden. 2These authors contributed equally to this work. Correspondence should be addressed to J.P.S. ([email protected]) or K.E. ([email protected]).
Received 18 December 2013; accepted 22 January 2014; published online 16 February 2014; doi:10.1038/nsmb.2776
236
VOLUME 21 NUMBER 3 MARCH 2014 nature structural & molecular biology
articles
Heterochromatin
0 –0.5 –1.0
0 0. 5 1. 0
.5
.0
–1
–2
.0 –0 .5
–1.5
5
H3 (enrichment)
10
15 10 5
g
3
CENP-ACnp1 (enrichment)
0
0
O tr2
O tr2
O tr2
0
15
C nt 2
0
1
20 CENP-ACnp1 (fold enrichment)
5
2
brl1∆
O tr2
10
WT 20
C nt 2
H3K9me2 (fold enrichment)
15
3
H3 (fold enrichment)
f
H3K9me2 (enrichment)
0.5
–1
H3K9me2 in htb1-K119R
d 1.0
htb1-K119R WT
–1
C nt 2
0
500
H3K9me2 in WT
CENP-ACnp1 (fold enrichment)
2
O tr2
4
90 75 60 45 30 15 0
250
0
htb1-K119R
C nt 2
H3 (fold enrichment)
6
0
–250
C nt 2
WT
C nt 2
H3K9me2 (fold enrichment)
2 1,599 bp 1 WT 0 2 1 0
htb1-K119R
2 1 0 2 1 0
Ratio
2 1 0
htb1-K119R
2 1 0
Ratio
2 1 0
WT
Heterochromatin Core
1,648 bp
WT
O
2
nt
C
RESULTS H2Bub1 prevents heterochromatin formation at centromeres To examine the role of H2Bub1 in epigenetic stability, we mapped the genomic distribution of heterochromatin marked by dimethylated H3 Lys9 (H3K9me2) in H2Bub1-proficient (wild-type, WT) and H2Bub1-deficient (htb1-K119R) S. pombe. By chromatin immunoprecipitation and subsequent genome-wide analysis of the recovered DNA (ChIP-chip), we identified regions in the WT strain enriched or depleted for H3K9me2 and found 197 and 513 regions, respectively35. Given previous studies11, we expected an expansion of the H3K9me2modified regions in the htb1-K119R mutant. However, in our data set, most boundaries of H3K9me2-enriched regions appeared to be intact (Fig. 1a). Also, the borders of the H3K9me2-depleted regions were not affected by the loss of H2Bub1 on a genome-wide scale (Fig. 1b). To correlate the heterochromatin distributions of the strains with different H2Bub1 status, we shattered the yeast genome in silico into fragments of 1,000 bp and compared the H3K9me2 signals for each fragment (Fig. 1c). When presented in this way, the genome-wide H3K9me2 profiles of the WT and mutant cells were similar (Pearson’s correlation coefficient = 0.83). The carefully studied heterochromatin of the pericentromeric regions yielded among the strongest signals for H3K9me2-enriched fragments in both strains. However, one cluster deviated between the strains. Genome annotation of this cluster revealed fragments from the central domain of the centromere. Examination of the three centromeres showed marked differences between the strains: H3K9me2-marked heterochromatin was increased in H2Bub1-deficient cells compared to WT (Fig. 1d and Supplementary Fig. 1). Normally, in S. pombe H3K9me2modified heterochromatin is distinctly found in the pericentric regions surrounding the central domain (Fig. 1d,e). In contrast, the central domain in the H2Bub1-deficient mutant was highly
tr2
Figure 1 Heterochromatin is formed in the centromere in H2Bub1-deficient cells despite 2 htb1-K119R the maintenance of a distinct central domain. (a,b) Alignment of regions enriched 1 2 0 (n = 197) (a) or depleted (n = 513) (b) of H3K9me2 in WT. The H3K9me2 ChIP values 1 relative to input are shown for WT (blue) and htb1-K119R (red). (c) Scatterplot of 2 Ratio 1 heterochromatin mark H3K9me2 in WT and H2Bub1-deficient cells (htb1-K119R). 0 0 Circled in blue are fragments of the pericentric regions; circled in orange are fragments of the centromere central cores. (d) Genome-browser view of centromere 2 probed with anti-H3K9me2, anti-H3 and anti-CENP-ACnp1 antibodies, relative to input. WT (top), htb1-K119R mutant (middle) and htb1-K119R/WT ratio (bottom) are shown. (e) ChIP levels of H3K9me2, H3 and CENP-ACnp1 at two centromeric loci (central core domain, cnt2; outer repeat, otr2) in WT and htb1-K119R, determined by ChIP-qPCR relative to beads. Error bars, s.e.m. (n = 3 independent experiments). (f) ChIP levels of H3K9me2, H3 and CENP-ACnp1 at two centromeric loci in WT and brl1∆, determined by ChIP-qPCR relative to beads. Bars indicate range (n = 2 independent experiments). (g) ChIP-qPCR of H2Bub1 in WT (light gray) and brl1∆ (dark gray) at two centromeric loci. Bars indicate range (n = 2 independent experiments). H2Bub1 (fold enrichment)
© 2014 Nature America, Inc. All rights reserved.
H3K9me2 (enrichment)
0 –500 –250 0 250 500 Position relative to region start (bp)
c
Position relative to region start (bp) –500
1
e
npg
b
htb1-K119R WT
O tr2
H3K9me2 (enrichment)
a
enriched with H3K9me2, at levels comparable to those found in the pericentric regions. We next examined whether the CENP-ACnp1 incorporation was defective in H2Bub1-deficient cells. By performing ChIP-chip with antibodies against CENP-ACnp1 and total H3 in cells lacking H2Bub1, we found that centromeric nucleosomes have slightly reduced levels of CENP-ACnp1 and higher levels of H3 compared to WT levels (Fig. 1d). This indicated that Lys9 dimethylation after H2Bub1 depletion occurred not only at the H3 normally present at centromeres but also at additional centromeric H3 nucleosomes. We confirmed the levels of the chromatin components in the different compartments of the centromere by using ChIP and quantitative PCR (ChIP-qPCR) (Fig. 1e and Supplementary Fig. 1c,d). Further, we determined that the E3 ligase responsible for the centromeric H2Bub1 was Brl1. Cells lacking Brl1 (brl1∆) showed levels of H3K9me2, H3 and CENPACnp1 similar to those in htb1-K119R (Fig. 1f). The centromere levels of H2Bub1 in brl1∆ cells were below our detection limit (Fig. 1g). Together, the data suggest that the exchange of CENP-ACnp1 and H3 is defective in H2Bub1-deficient cells. Heterochromatin borders maintained in the absence of H2Bub1 The observed heterochromatin formation of the central domain of the centromere could arise from a role of H2Bub1 in preventing de novo heterochromatin formation or promoting heterochromatin barrier function. The barriers between the centromeric central domain and pericentric heterochromatin in S. pombe are defined by RNA polymerase III (Pol III)-transcribed tRNA loci36. These loci are associated with the TFIIIC complex, which contains several subunits including Sfc6 (Fig. 2a)37. First, the pericentric borders appeared to remain functional in H2Bub1-deficient cells, because the CENP-ACnp1 was largely confined to the central domain (Fig. 1d). Second, chromatin
nature structural & molecular biology VOLUME 21 NUMBER 3 MARCH 2014
237
articles
a
Sfc6 (fold enrichment)
d
**
0.28 0.21 0.14 0.07
12 6
C
a1 Su
nt
2
G
M 2-
Ig
2
La
G
G
2 S-
G
S
G Ig
M
2
La
G
2-
2
G te
G
Heterochromatin
WT htb1-K119R IgG
18
0
0
S
*
WT htb1-K119R
g
Heterochromatin
6
– + – + – + – +
Primer
Actin
0
1
2
3
1
2
3
2 0
tr2
htb1-K119R
2
0.2
4
nt
WT
Ladder
+ – + – + – + –
0.4
O
2 3
0.6
H3-HA enrichment (2 h/0 h)
1
0.8
C
M 2-
2 G
G
2 La
RNA expression (% of actin)
*
te
G
S
Cnt2 Otr2
0.35
Core
WT htb1-K119R
S-
2
tr2
0.2
Core
5 4 3 2 1 0
4
O
0.4
f
6
PoI II (% of input)
c
0.6
Sura4+ Heterochromatin Heterochromatin
htb1-K119R × ∆ala
We monitored the consecutive stages of the cell cycle by flow cytometry and measured gene expression by using markers specific to cell-cycle stage (slp1, ste11 and cdc25). As hypothesized, the level of H2Bub1 transiently increased in the centromeric central domain, and this increase occurred at the G2-M transition (P < 0.05, t test) (Fig. 3b). An active form of RNA Pol II, phosphorylated at Ser5 of the C-terminal domain (Pol II pSer5), was present at the centromere central domain and showed an identical profile with a pronounced peak around the G2-M transition (Fig. 3c). This specific cell cycle–dependent pulse of Pol II was H2Bub1 dependent; we observed reduced Pol II occupancy in htb1-K119R cells relative to WT at the central centromere chromatin (Fig. 3d). From earlier studies, we know that active histone marks and Pol II–mediated transcription of the central domain are required for centromere function19–24. H2Bub1 is generally associated with an opening up of the chromatin structure and transcription by RNA Pol II (ref. 1). We observed a correlation between H2Bub1 and both Pol II and RNA (Pearson’s correlation coefficient 0.54 and 0.60, respectively; Supplementary Fig. 2a,b), in accordance with previous results. Consistently with previous observations40, the RNA levels in the htb1-K119R mutant were only slightly perturbed
0
e
Heterochromatin
0
*
0.8
c
WT htb1-K119R
8
te
WT
Cnt2 Otr2
1.0
b
Pol II pSer5 (% of input)
2 1 0
b
Core Heterochromatin Heterochromatin 3,747 bp 3,793 bp
H2Bub1 (% of input)
H2Bub1 (enrichment)
a
RNA expression (% of actin)
© 2014 Nature America, Inc. All rights reserved.
npg
H2B monoubiquitination leads to centromeric transcription To directly link H2Bub1 to centromeric chromatin maintenance, we mapped the genome-wide levels of H2Bub1. Though H2B is found at the centromere38,39, H2Bub1 levels in centromeric chromatin from asynchronously growing cells were relatively low (Fig. 3a). In brief, although centromeric nucleosomes generally did not contain H2Bub1, cells lacking H2Bub1 displayed an altered centromeric chromatin structure. To account for this apparent paradox, we hypothesized that H2B is transiently ubiquitinated in a narrowly defined stage of the cell cycle, consistently with the reported cell cycle–dependent centromere transcription19,23. Here, we synchronized cells in G2-M and, after release, measured centromeric H2Bub1 levels (Fig. 3b).
Heterochromatin
Sfc6
immunoprecipitation of myc-tagged Sfc6 identified equal localization to the tRNA locus in both strains, regardless of H2Bub1 status. This result showed that the heterochromatin barrier was intact in the H2Bub1-deficient mutant (Fig. 2b). Third, when we combined the htb1-K119R strain with a mutant lacking one of the barrier-defining tRNAs (SPATRNAALA.05, here denoted ala) at centromere 1 (∆ala; described in ref. 36), the double mutant was not viable (Fig. 2c), results suggesting that H2B monoubiquitination and the tRNA barrier work in parallel pathways to maintain active centromeres. Taken together, these data suggest that centromere silencing in H2Bub1deficient cells is caused by de novo heterochromatin formation.
ala Core Sfc6
Figure 2 Heterochromatin boundaries are maintained without H2Bub1. (a) Depiction of centromere surroundings with the pericentric heterochromatin (dark gray) being prevented from spreading into the central core by the TFIIIC-complex subunit Sfc6 at the tRNA loci (green). (b) ChIP-qPCR of the boundary-associated Sfc6 in WT and htb1-K119R cells at the alanine tRNA locus (ala). Error bars, s.e.m. (n = 3 independent experiments). (c) Tetrad dissection of a cross between htb1-K119R and a mutant (∆ala) lacking the alanine tRNA gene (SPATRNAALA.05) at centromere 1. Arrows indicate nonviable cells.
Figure 3 Cell-cycle regulation of H2Bub1 levels leads to chromatin transcription in M phase. (a) H2Bub1 levels at the centromere in WT asynchronous cells, as determined by ChIP-chip. (b) Quantification of H2Bub1 levels at the centromeric core (cnt2) and outer repeat (otr2) by ChIP-qPCR in synchronized cdc25-22 cells, *P < 0.05 (t test, two-tailed, between cnt2 and otr2). Error bars, s.e.m. (n = 5 cell cycles of four independent experiments). (c) Quantification of Pol II pSer5 levels at the centromeric core and outer repeat by ChIP-qPCR in synchronized cdc25-22 cells, **P < 0.01 (t test, two-tailed, between cnt2 and otr2). Error bars, s.e.m. (n = 3 cell cycles of two independent experiments). (d) Pol II levels at the centromere in asynchronous WT and htb1-K119R cells, as determined by ChIP-qPCR. Bars indicate range (n = 2 independent experiments). (e) Transcription of the ura4+ reporter integrated into centromere 2 relative to actin (act1), on the basis of cell-cycle stage in WT and htb1-K119R cells with a cdc25-22 background, *P < 0.05 (t test, two-tailed, between WT and htb1-K119R). Error bars, s.e.m. (n = 5 cell cycles of four independent experiments). (f) Endogenous centromere transcripts relative to actin (act1) on the pfs2-ts background in which mRNA cleavage is compromised, in asynchronous WT and htb1-K119R cultures. (g) Centromeric nucleosomal turnover in WT and htb1-K119R cells, as measured by ChIP-qPCR using antibody to hemagglutinin (HA) 2 h and 0 h after induction of H3-HA expression, *P < 0.05 (t test between WT and htb1-K119R). Error bars, s.e.m. (n = 3 independent experiments).
238
VOLUME 21 NUMBER 3 MARCH 2014 nature structural & molecular biology
articles Figure 4 H2Bub1 is required for functional centromeres. (a) TBZ sensitivity of WT and htb1-K119R cells or cells lacking either or both Ubp8 or Ubp16 deubiquitinases. (b) Immunofluorescent staining and quantification (c) of chromosomal mis-segregation in WT and htb1-K119R cells. White arrows indicate aberrant chromosomal segregation. Only mitotic cells with at least 5 µm of tubulin mark were scored. Error bars, s.e.m. (n = 3 independent experiments).
a
WT
htb1-K119R brl1∆ upb8∆
npg
© 2014 Nature America, Inc. All rights reserved.
upb16∆
H2Bub1 needed for chromosome segregation Next, we sought to confirm the requirement for H2B ubiquitination for centromere function leading to kinetochore formation and mitotic progression. Previous experiments have suggested a role of H2Bub1 in mitotic progression by using htb1 mutants with affected ubiquitination status, among other issues14. First, we tested whether H2B
upb8∆ upb16∆ 10 µg/ml TBZ
–TBZ
b
15 µg/ml TBZ
Normal segregation
c
10 µM
Abnormal segregation (%)
htb1-K119R 20 16 12 8 4 0 he tc re St
gg
in
d
g
Lagging chromosome
WT
La
(Supplementary Fig. 2c). To rule out the possibility that the observed effects were a consequence of altered transcription of kinetochore proteins in the htb1-K119R mutant, we took particular interest in the RNA levels of all genes encoding kinetochore proteins. The results revealed that all kinetochore transcripts remained similarly expressed (log2(htb1-K119R/WT)−1) (Supplementary Fig. 2c and Supplementary Table 1). On a global scale, the lack of H2Bub1 made the chromatin more compact compared to that of WT, as observed by reduced sensitivity to micrococcal nuclease (MNase) digestion (Supplementary Fig. 2d). This MNase resistance of H2Bub1-deficient cells has previously been observed in S. cerevisiae41,42. Possibly, the centromeric role of H2Bub1 contributes to central-domain decompaction and transcription. To determine how centromeric chromatin transcription relates to cell-cycle progression, we synchronized WT and H2Bub1-deficient strains with a reporter gene (ura4+) incorporated into the central core of centromere 2. In WT cells, and to a lesser extent in H2Bub1deficient cells, the reporter gene was expressed at low levels throughout the cell cycle, with a clear peak in M phase. This result implies a cell cycle– and H2Bub1-dependent chromatin alteration allowing transient transcription of the reporter gene (Fig. 3e). The endogenous centromeric core transcripts are rapidly cleaved and degraded (Supplementary Fig. 3a)20. Through inactivation of Pfs2, we compromised mRNA cleavage20,43 in asynchronous cells and observed H2Bub1-facilitated transcription of the endogenous centromeric central domain (Fig. 3f). Centromeric core transcription has been associated with the eviction of H3-containing nucleosomes20. Under this hypothesis, the H2Bub1-deficient cells, compared to WT cells, would be expected to have altered nucleosome turnover in the centromeric central domain. To this end, we measured the nucleosomal turnover by using epitopetagged H3 under an invertase-inducible promoter44. As we expected from other studies45, few newly produced H3-containing nucleosomes were incorporated at the pericentric regions after sucrose induction of the H3 promoter, results indicating nucleosomes with low turnover. However, in WT cells, the central domain was associated with enhanced accessibility for newly produced H3-containing nucleosomes (P < 0.05, t test; Fig. 3g). In the H2B ubiquitination–deficient mutant, the central domain and the pericentric region of the centromere had similar nucleosomal turnover rates (Fig. 3g) equivalent to the rates in WT heterochromatin. Also, at other genomic loci the changes in incorporation of new H3 correlated with changes in expression in the htb1-K119R mutant (Supplementary Fig. 3b). This result offers an explanation of the increased centromeric levels of H3 in the htb1-K119R mutant and further suggests that, at the centromere in WT cells, H2Bub1 leads to H3 instability.
Stretched chromosome
DAPI
α-tubulin
Merge
ubiquitination affects centromere function by growing mutant strains in the presence of the microtubule-destabilizing drug thiabendazole (TBZ) (Fig. 4a). The H2Bub1-deficient cells, both the htb1-K119R strain and cells lacking the E3 ligase Brl1, were sensitive to TBZ, thus revealing that H2B monoubiquitination acts synergistically with TBZ during chromosome segregation. In S. cerevisiae, two H2Bub1deubiquitinating proteins have been identified46,47. We measured the global levels of H2Bub1 in deletion mutants of the two homologs in S. pombe, Ubp8 and Ubp16, and determined that the main contributor to H2Bub1 deubiquitination was Ubp8 (Supplementary Fig. 4). Consistently with H2Bub1 removal by nucleosome eviction and not deubiquitination, cells lacking Ubp8 or Ubp16 or both were not affected by TBZ exposure (Fig. 4b,c). For a direct measurement of genome stability, we analyzed the chromosome segregation in M phase cytologically (Fig. 4b). We observed aberrant segregation in 26% of the cells (16% lagging chromosomes and 10% stretched chromosomes), compared to 2% in the WT cells (1% lagging and 1% stretched). These results confirm that H2Bub1 is needed for centromere function and genome stability. H2Bub1 found at centromeres in human cells To test whether the centromeric role of H2Bub1 is conserved in human cells, we examined the presence of H2Bub1 at ALR and in proximity to the CENP-A–defined chromatin in human centromeres. First, by chromatin immunoprecipitation, we detected a substantial enrichment of H2Bub1 at ALR in HeLa cells (Fig. 5a). Second, by in situ proximity ligation, a method using two different antibodies that when in proximity (
Centromeric histone H2B monoubiquitination promotes noncoding transcription and chromatin integrity
npg
© 2014 Nature America, Inc. All rights reserved.
Laia Sadeghi1,2, Lee Siggens1,2, J Peter Svensson1 & Karl Ekwall1 Functional centromeres are essential for proper cell division. Centromeres are established largely by epigenetic processes resulting in incorporation of the histone H3 variant CENP-A. Here, we demonstrate the direct involvement of H2B monoubiquitination, mediated by RNF20 in humans or Brl1 in Schizosaccharomyces pombe, in centromeric chromatin maintenance. Monoubiquinated H2B (H2Bub1) is needed for this maintenance, promoting noncoding transcription, centromere integrity and accurate chromosomal segregation. A transient pulse of centromeric H2Bub1 leads to RNA polymerase II–mediated transcription of the centromere’s central domain, coupled to decreased H3 stability. H2Bub1-deficient cells have centromere cores that, despite their intact centromeric heterochromatin barriers, exhibit characteristics of heterochromatin, such as silencing histone modifications, reduced nucleosome turnover and reduced levels of transcription. In the H2Bub1-deficient cells, centromere functionality is hampered, thus resulting in unequal chromosome segregation. Therefore, centromeric H2Bub1 is essential for maintaining active centromeric chromatin. H2Bub1 levels in chromatin generally correlate with RNA polymerase II (Pol II) and transcription1,2, but H2Bub1 has also been implicated in diverse cellular functions, such as DNA replication3, differentiation4–6 and DNA-damage responses7–10. The modulation and expansion of heterochromatin regions has been reported at several loci in the absence of H2Bub1 (refs. 11,12). Early observations implicated H2B in centromere maintenance13 and, although mutations affecting the inner region of H2B can simultaneously decrease H2Bub1 and cause centromeric defects14, a direct role has not been demonstrated for histone ubiquitination in the maintenance of active centromeres. In budding yeast (Saccharomyces cerevisiae), methylation of kinetochore protein Dam1 has been found downstream of H2Bub1 (ref. 15). Fission yeast (S. pombe) and higher eukaryotes have regional centromeres; in S. pombe, these occupy 40–100 kb of the chromosome. The kinetochore attaches to the central domain of the centromere, thus allowing accurate segregation of sister chromatids. Rather than being specified by the DNA sequence, centromeres are defined epigenetically by the presence of the centromere-specific histone H3 variant CENP-A. The deposition of CENP-A is required for kinetochore assembly16–18, in a process that requires transcription19–24 (reviewed in ref. 25). Centromeres in most eukaryotes have two distinct domains: the centromeric CENP-A chromatin and the flanking pericentric heterochromatin marked by methylation of H3 Lys9 (H3K9). Both domains are needed for centromere functions (reviewed in ref. 26). In S. pombe and mice, the centromeric and pericentromeric domains are clearly defined and are associated with specialized DNA sequences. In other species, including humans, centromeric
DNA is mainly made up of repetitive α-satellite sequences (ALR), and, although the compartments are distinct, the DNA sequence does not distinguish between the two types of centromeric chromatin27,28. For the formation of pericentric heterchromatin in S. pombe, evidence of a role for noncoding RNA has become well established after the key discovery of RNA interference (RNAi) in heterochromatin formation29. From a clinical perspective, dysregulation of the H2Bub1 machinery has been linked to leukemia, breast cancer and other neoplasms30,31. The E3 ligase RNF20, which targets H2B, functions as a tumor suppressor in vivo30,32, and reduction in levels of RNF20 in mammalian cells causes compromised homologous recombination resulting in chromosomal instability and replication stress7. Depletion of RNF20 leads to altered gene expression, such as downregulation of p53 (ref. 30). Normally, in adult somatic cells, centromeres are not transcribed or are transcribed at low levels. However, the centromeres of tumor cells often have increased levels of transcription33,34. To understand the role of H2Bub1 in centromere establishment and propagation, we perturbed the H2Bub1 machinery in both S. pombe and human cells and studied the effects throughout the cell cycle. We show that noncoding centromere transcription is dependent on H2Bub1. Lack of H2Bub1, either in a mutant that cannot monoubi quitinate H2B or in cells lacking the E3 ligase (RNF20 in human cells or Brl1 in S. pombe), leads to chromosome mis-segregation. A pulse of H2Bub1 coupled to transcription of the central core of the centromere occurs before cells enter the M phase of the cell cycle. In the absence of centromeric H2Bub1, heterochromatin forms de novo over the central core, despite the maintained chromatin borders.
1Department of Biosciences and Nutrition, Karolinska Institutet, Huddinge, Sweden. 2These authors contributed equally to this work. Correspondence should be addressed to J.P.S. ([email protected]) or K.E. ([email protected]).
Received 18 December 2013; accepted 22 January 2014; published online 16 February 2014; doi:10.1038/nsmb.2776
236
VOLUME 21 NUMBER 3 MARCH 2014 nature structural & molecular biology
articles
Heterochromatin
0 –0.5 –1.0
0 0. 5 1. 0
.5
.0
–1
–2
.0 –0 .5
–1.5
5
H3 (enrichment)
10
15 10 5
g
3
CENP-ACnp1 (enrichment)
0
0
O tr2
O tr2
O tr2
0
15
C nt 2
0
1
20 CENP-ACnp1 (fold enrichment)
5
2
brl1∆
O tr2
10
WT 20
C nt 2
H3K9me2 (fold enrichment)
15
3
H3 (fold enrichment)
f
H3K9me2 (enrichment)
0.5
–1
H3K9me2 in htb1-K119R
d 1.0
htb1-K119R WT
–1
C nt 2
0
500
H3K9me2 in WT
CENP-ACnp1 (fold enrichment)
2
O tr2
4
90 75 60 45 30 15 0
250
0
htb1-K119R
C nt 2
H3 (fold enrichment)
6
0
–250
C nt 2
WT
C nt 2
H3K9me2 (fold enrichment)
2 1,599 bp 1 WT 0 2 1 0
htb1-K119R
2 1 0 2 1 0
Ratio
2 1 0
htb1-K119R
2 1 0
Ratio
2 1 0
WT
Heterochromatin Core
1,648 bp
WT
O
2
nt
C
RESULTS H2Bub1 prevents heterochromatin formation at centromeres To examine the role of H2Bub1 in epigenetic stability, we mapped the genomic distribution of heterochromatin marked by dimethylated H3 Lys9 (H3K9me2) in H2Bub1-proficient (wild-type, WT) and H2Bub1-deficient (htb1-K119R) S. pombe. By chromatin immunoprecipitation and subsequent genome-wide analysis of the recovered DNA (ChIP-chip), we identified regions in the WT strain enriched or depleted for H3K9me2 and found 197 and 513 regions, respectively35. Given previous studies11, we expected an expansion of the H3K9me2modified regions in the htb1-K119R mutant. However, in our data set, most boundaries of H3K9me2-enriched regions appeared to be intact (Fig. 1a). Also, the borders of the H3K9me2-depleted regions were not affected by the loss of H2Bub1 on a genome-wide scale (Fig. 1b). To correlate the heterochromatin distributions of the strains with different H2Bub1 status, we shattered the yeast genome in silico into fragments of 1,000 bp and compared the H3K9me2 signals for each fragment (Fig. 1c). When presented in this way, the genome-wide H3K9me2 profiles of the WT and mutant cells were similar (Pearson’s correlation coefficient = 0.83). The carefully studied heterochromatin of the pericentromeric regions yielded among the strongest signals for H3K9me2-enriched fragments in both strains. However, one cluster deviated between the strains. Genome annotation of this cluster revealed fragments from the central domain of the centromere. Examination of the three centromeres showed marked differences between the strains: H3K9me2-marked heterochromatin was increased in H2Bub1-deficient cells compared to WT (Fig. 1d and Supplementary Fig. 1). Normally, in S. pombe H3K9me2modified heterochromatin is distinctly found in the pericentric regions surrounding the central domain (Fig. 1d,e). In contrast, the central domain in the H2Bub1-deficient mutant was highly
tr2
Figure 1 Heterochromatin is formed in the centromere in H2Bub1-deficient cells despite 2 htb1-K119R the maintenance of a distinct central domain. (a,b) Alignment of regions enriched 1 2 0 (n = 197) (a) or depleted (n = 513) (b) of H3K9me2 in WT. The H3K9me2 ChIP values 1 relative to input are shown for WT (blue) and htb1-K119R (red). (c) Scatterplot of 2 Ratio 1 heterochromatin mark H3K9me2 in WT and H2Bub1-deficient cells (htb1-K119R). 0 0 Circled in blue are fragments of the pericentric regions; circled in orange are fragments of the centromere central cores. (d) Genome-browser view of centromere 2 probed with anti-H3K9me2, anti-H3 and anti-CENP-ACnp1 antibodies, relative to input. WT (top), htb1-K119R mutant (middle) and htb1-K119R/WT ratio (bottom) are shown. (e) ChIP levels of H3K9me2, H3 and CENP-ACnp1 at two centromeric loci (central core domain, cnt2; outer repeat, otr2) in WT and htb1-K119R, determined by ChIP-qPCR relative to beads. Error bars, s.e.m. (n = 3 independent experiments). (f) ChIP levels of H3K9me2, H3 and CENP-ACnp1 at two centromeric loci in WT and brl1∆, determined by ChIP-qPCR relative to beads. Bars indicate range (n = 2 independent experiments). (g) ChIP-qPCR of H2Bub1 in WT (light gray) and brl1∆ (dark gray) at two centromeric loci. Bars indicate range (n = 2 independent experiments). H2Bub1 (fold enrichment)
© 2014 Nature America, Inc. All rights reserved.
H3K9me2 (enrichment)
0 –500 –250 0 250 500 Position relative to region start (bp)
c
Position relative to region start (bp) –500
1
e
npg
b
htb1-K119R WT
O tr2
H3K9me2 (enrichment)
a
enriched with H3K9me2, at levels comparable to those found in the pericentric regions. We next examined whether the CENP-ACnp1 incorporation was defective in H2Bub1-deficient cells. By performing ChIP-chip with antibodies against CENP-ACnp1 and total H3 in cells lacking H2Bub1, we found that centromeric nucleosomes have slightly reduced levels of CENP-ACnp1 and higher levels of H3 compared to WT levels (Fig. 1d). This indicated that Lys9 dimethylation after H2Bub1 depletion occurred not only at the H3 normally present at centromeres but also at additional centromeric H3 nucleosomes. We confirmed the levels of the chromatin components in the different compartments of the centromere by using ChIP and quantitative PCR (ChIP-qPCR) (Fig. 1e and Supplementary Fig. 1c,d). Further, we determined that the E3 ligase responsible for the centromeric H2Bub1 was Brl1. Cells lacking Brl1 (brl1∆) showed levels of H3K9me2, H3 and CENPACnp1 similar to those in htb1-K119R (Fig. 1f). The centromere levels of H2Bub1 in brl1∆ cells were below our detection limit (Fig. 1g). Together, the data suggest that the exchange of CENP-ACnp1 and H3 is defective in H2Bub1-deficient cells. Heterochromatin borders maintained in the absence of H2Bub1 The observed heterochromatin formation of the central domain of the centromere could arise from a role of H2Bub1 in preventing de novo heterochromatin formation or promoting heterochromatin barrier function. The barriers between the centromeric central domain and pericentric heterochromatin in S. pombe are defined by RNA polymerase III (Pol III)-transcribed tRNA loci36. These loci are associated with the TFIIIC complex, which contains several subunits including Sfc6 (Fig. 2a)37. First, the pericentric borders appeared to remain functional in H2Bub1-deficient cells, because the CENP-ACnp1 was largely confined to the central domain (Fig. 1d). Second, chromatin
nature structural & molecular biology VOLUME 21 NUMBER 3 MARCH 2014
237
articles
a
Sfc6 (fold enrichment)
d
**
0.28 0.21 0.14 0.07
12 6
C
a1 Su
nt
2
G
M 2-
Ig
2
La
G
G
2 S-
G
S
G Ig
M
2
La
G
2-
2
G te
G
Heterochromatin
WT htb1-K119R IgG
18
0
0
S
*
WT htb1-K119R
g
Heterochromatin
6
– + – + – + – +
Primer
Actin
0
1
2
3
1
2
3
2 0
tr2
htb1-K119R
2
0.2
4
nt
WT
Ladder
+ – + – + – + –
0.4
O
2 3
0.6
H3-HA enrichment (2 h/0 h)
1
0.8
C
M 2-
2 G
G
2 La
RNA expression (% of actin)
*
te
G
S
Cnt2 Otr2
0.35
Core
WT htb1-K119R
S-
2
tr2
0.2
Core
5 4 3 2 1 0
4
O
0.4
f
6
PoI II (% of input)
c
0.6
Sura4+ Heterochromatin Heterochromatin
htb1-K119R × ∆ala
We monitored the consecutive stages of the cell cycle by flow cytometry and measured gene expression by using markers specific to cell-cycle stage (slp1, ste11 and cdc25). As hypothesized, the level of H2Bub1 transiently increased in the centromeric central domain, and this increase occurred at the G2-M transition (P < 0.05, t test) (Fig. 3b). An active form of RNA Pol II, phosphorylated at Ser5 of the C-terminal domain (Pol II pSer5), was present at the centromere central domain and showed an identical profile with a pronounced peak around the G2-M transition (Fig. 3c). This specific cell cycle–dependent pulse of Pol II was H2Bub1 dependent; we observed reduced Pol II occupancy in htb1-K119R cells relative to WT at the central centromere chromatin (Fig. 3d). From earlier studies, we know that active histone marks and Pol II–mediated transcription of the central domain are required for centromere function19–24. H2Bub1 is generally associated with an opening up of the chromatin structure and transcription by RNA Pol II (ref. 1). We observed a correlation between H2Bub1 and both Pol II and RNA (Pearson’s correlation coefficient 0.54 and 0.60, respectively; Supplementary Fig. 2a,b), in accordance with previous results. Consistently with previous observations40, the RNA levels in the htb1-K119R mutant were only slightly perturbed
0
e
Heterochromatin
0
*
0.8
c
WT htb1-K119R
8
te
WT
Cnt2 Otr2
1.0
b
Pol II pSer5 (% of input)
2 1 0
b
Core Heterochromatin Heterochromatin 3,747 bp 3,793 bp
H2Bub1 (% of input)
H2Bub1 (enrichment)
a
RNA expression (% of actin)
© 2014 Nature America, Inc. All rights reserved.
npg
H2B monoubiquitination leads to centromeric transcription To directly link H2Bub1 to centromeric chromatin maintenance, we mapped the genome-wide levels of H2Bub1. Though H2B is found at the centromere38,39, H2Bub1 levels in centromeric chromatin from asynchronously growing cells were relatively low (Fig. 3a). In brief, although centromeric nucleosomes generally did not contain H2Bub1, cells lacking H2Bub1 displayed an altered centromeric chromatin structure. To account for this apparent paradox, we hypothesized that H2B is transiently ubiquitinated in a narrowly defined stage of the cell cycle, consistently with the reported cell cycle–dependent centromere transcription19,23. Here, we synchronized cells in G2-M and, after release, measured centromeric H2Bub1 levels (Fig. 3b).
Heterochromatin
Sfc6
immunoprecipitation of myc-tagged Sfc6 identified equal localization to the tRNA locus in both strains, regardless of H2Bub1 status. This result showed that the heterochromatin barrier was intact in the H2Bub1-deficient mutant (Fig. 2b). Third, when we combined the htb1-K119R strain with a mutant lacking one of the barrier-defining tRNAs (SPATRNAALA.05, here denoted ala) at centromere 1 (∆ala; described in ref. 36), the double mutant was not viable (Fig. 2c), results suggesting that H2B monoubiquitination and the tRNA barrier work in parallel pathways to maintain active centromeres. Taken together, these data suggest that centromere silencing in H2Bub1deficient cells is caused by de novo heterochromatin formation.
ala Core Sfc6
Figure 2 Heterochromatin boundaries are maintained without H2Bub1. (a) Depiction of centromere surroundings with the pericentric heterochromatin (dark gray) being prevented from spreading into the central core by the TFIIIC-complex subunit Sfc6 at the tRNA loci (green). (b) ChIP-qPCR of the boundary-associated Sfc6 in WT and htb1-K119R cells at the alanine tRNA locus (ala). Error bars, s.e.m. (n = 3 independent experiments). (c) Tetrad dissection of a cross between htb1-K119R and a mutant (∆ala) lacking the alanine tRNA gene (SPATRNAALA.05) at centromere 1. Arrows indicate nonviable cells.
Figure 3 Cell-cycle regulation of H2Bub1 levels leads to chromatin transcription in M phase. (a) H2Bub1 levels at the centromere in WT asynchronous cells, as determined by ChIP-chip. (b) Quantification of H2Bub1 levels at the centromeric core (cnt2) and outer repeat (otr2) by ChIP-qPCR in synchronized cdc25-22 cells, *P < 0.05 (t test, two-tailed, between cnt2 and otr2). Error bars, s.e.m. (n = 5 cell cycles of four independent experiments). (c) Quantification of Pol II pSer5 levels at the centromeric core and outer repeat by ChIP-qPCR in synchronized cdc25-22 cells, **P < 0.01 (t test, two-tailed, between cnt2 and otr2). Error bars, s.e.m. (n = 3 cell cycles of two independent experiments). (d) Pol II levels at the centromere in asynchronous WT and htb1-K119R cells, as determined by ChIP-qPCR. Bars indicate range (n = 2 independent experiments). (e) Transcription of the ura4+ reporter integrated into centromere 2 relative to actin (act1), on the basis of cell-cycle stage in WT and htb1-K119R cells with a cdc25-22 background, *P < 0.05 (t test, two-tailed, between WT and htb1-K119R). Error bars, s.e.m. (n = 5 cell cycles of four independent experiments). (f) Endogenous centromere transcripts relative to actin (act1) on the pfs2-ts background in which mRNA cleavage is compromised, in asynchronous WT and htb1-K119R cultures. (g) Centromeric nucleosomal turnover in WT and htb1-K119R cells, as measured by ChIP-qPCR using antibody to hemagglutinin (HA) 2 h and 0 h after induction of H3-HA expression, *P < 0.05 (t test between WT and htb1-K119R). Error bars, s.e.m. (n = 3 independent experiments).
238
VOLUME 21 NUMBER 3 MARCH 2014 nature structural & molecular biology
articles Figure 4 H2Bub1 is required for functional centromeres. (a) TBZ sensitivity of WT and htb1-K119R cells or cells lacking either or both Ubp8 or Ubp16 deubiquitinases. (b) Immunofluorescent staining and quantification (c) of chromosomal mis-segregation in WT and htb1-K119R cells. White arrows indicate aberrant chromosomal segregation. Only mitotic cells with at least 5 µm of tubulin mark were scored. Error bars, s.e.m. (n = 3 independent experiments).
a
WT
htb1-K119R brl1∆ upb8∆
npg
© 2014 Nature America, Inc. All rights reserved.
upb16∆
H2Bub1 needed for chromosome segregation Next, we sought to confirm the requirement for H2B ubiquitination for centromere function leading to kinetochore formation and mitotic progression. Previous experiments have suggested a role of H2Bub1 in mitotic progression by using htb1 mutants with affected ubiquitination status, among other issues14. First, we tested whether H2B
upb8∆ upb16∆ 10 µg/ml TBZ
–TBZ
b
15 µg/ml TBZ
Normal segregation
c
10 µM
Abnormal segregation (%)
htb1-K119R 20 16 12 8 4 0 he tc re St
gg
in
d
g
Lagging chromosome
WT
La
(Supplementary Fig. 2c). To rule out the possibility that the observed effects were a consequence of altered transcription of kinetochore proteins in the htb1-K119R mutant, we took particular interest in the RNA levels of all genes encoding kinetochore proteins. The results revealed that all kinetochore transcripts remained similarly expressed (log2(htb1-K119R/WT)−1) (Supplementary Fig. 2c and Supplementary Table 1). On a global scale, the lack of H2Bub1 made the chromatin more compact compared to that of WT, as observed by reduced sensitivity to micrococcal nuclease (MNase) digestion (Supplementary Fig. 2d). This MNase resistance of H2Bub1-deficient cells has previously been observed in S. cerevisiae41,42. Possibly, the centromeric role of H2Bub1 contributes to central-domain decompaction and transcription. To determine how centromeric chromatin transcription relates to cell-cycle progression, we synchronized WT and H2Bub1-deficient strains with a reporter gene (ura4+) incorporated into the central core of centromere 2. In WT cells, and to a lesser extent in H2Bub1deficient cells, the reporter gene was expressed at low levels throughout the cell cycle, with a clear peak in M phase. This result implies a cell cycle– and H2Bub1-dependent chromatin alteration allowing transient transcription of the reporter gene (Fig. 3e). The endogenous centromeric core transcripts are rapidly cleaved and degraded (Supplementary Fig. 3a)20. Through inactivation of Pfs2, we compromised mRNA cleavage20,43 in asynchronous cells and observed H2Bub1-facilitated transcription of the endogenous centromeric central domain (Fig. 3f). Centromeric core transcription has been associated with the eviction of H3-containing nucleosomes20. Under this hypothesis, the H2Bub1-deficient cells, compared to WT cells, would be expected to have altered nucleosome turnover in the centromeric central domain. To this end, we measured the nucleosomal turnover by using epitopetagged H3 under an invertase-inducible promoter44. As we expected from other studies45, few newly produced H3-containing nucleosomes were incorporated at the pericentric regions after sucrose induction of the H3 promoter, results indicating nucleosomes with low turnover. However, in WT cells, the central domain was associated with enhanced accessibility for newly produced H3-containing nucleosomes (P < 0.05, t test; Fig. 3g). In the H2B ubiquitination–deficient mutant, the central domain and the pericentric region of the centromere had similar nucleosomal turnover rates (Fig. 3g) equivalent to the rates in WT heterochromatin. Also, at other genomic loci the changes in incorporation of new H3 correlated with changes in expression in the htb1-K119R mutant (Supplementary Fig. 3b). This result offers an explanation of the increased centromeric levels of H3 in the htb1-K119R mutant and further suggests that, at the centromere in WT cells, H2Bub1 leads to H3 instability.
Stretched chromosome
DAPI
α-tubulin
Merge
ubiquitination affects centromere function by growing mutant strains in the presence of the microtubule-destabilizing drug thiabendazole (TBZ) (Fig. 4a). The H2Bub1-deficient cells, both the htb1-K119R strain and cells lacking the E3 ligase Brl1, were sensitive to TBZ, thus revealing that H2B monoubiquitination acts synergistically with TBZ during chromosome segregation. In S. cerevisiae, two H2Bub1deubiquitinating proteins have been identified46,47. We measured the global levels of H2Bub1 in deletion mutants of the two homologs in S. pombe, Ubp8 and Ubp16, and determined that the main contributor to H2Bub1 deubiquitination was Ubp8 (Supplementary Fig. 4). Consistently with H2Bub1 removal by nucleosome eviction and not deubiquitination, cells lacking Ubp8 or Ubp16 or both were not affected by TBZ exposure (Fig. 4b,c). For a direct measurement of genome stability, we analyzed the chromosome segregation in M phase cytologically (Fig. 4b). We observed aberrant segregation in 26% of the cells (16% lagging chromosomes and 10% stretched chromosomes), compared to 2% in the WT cells (1% lagging and 1% stretched). These results confirm that H2Bub1 is needed for centromere function and genome stability. H2Bub1 found at centromeres in human cells To test whether the centromeric role of H2Bub1 is conserved in human cells, we examined the presence of H2Bub1 at ALR and in proximity to the CENP-A–defined chromatin in human centromeres. First, by chromatin immunoprecipitation, we detected a substantial enrichment of H2Bub1 at ALR in HeLa cells (Fig. 5a). Second, by in situ proximity ligation, a method using two different antibodies that when in proximity (
