Propagation of Polycomb-repressed chromatin requires sequence-specific recruitment to DNA

See allHide authors and affiliations

Science  07 Apr 2017:
Vol. 356, Issue 6333, pp. 85-88
DOI: 10.1126/science.aai8266

DNA sequence and inherited gene silencing

Cell fate decisions require a gene's transcriptional status, whether on or off, to be stably and heritably maintained over multiple cell generations. For silenced genes, heterochromatin domains are associated with specific histone posttranslational modifications, and these histone marks are maintained during DNA replication and chromosome duplication (see the Perspective by De and Kassis). Laprell et al. show that parental methylated histone H3 lysine 27 (H3K27) nucleosomes in Drosophila are inherited in daughter cells after replication and can repress transcription, but that they are not sufficient to propagate the mark. Trimethylation of newly incorporated nucleosomes requires recruitment of the methyltransferase Polycomb repressive complex 2 (PRC2) to neighboring cis-regulatory DNA elements. Coleman and Struhl demonstrate that H3K27 trimethylated nucleosomes play a causal role in transmitting epigenetic memory at a Drosophila HOX gene through anchoring of PRC2 at the Polycomb response element binding site. Wang and Moazed examine fission yeast and show that both sequence-dependent and chromodomain sequence-independent mechanisms are required for stable epigenetic inheritance of histone modifications and the epigenetic maintenance of silencing. These studies highlight the crucial role of DNA binding for heritable gene silencing during growth and development.

Science, this issue p. 85, p. eaai8236, p. 88; see also p. 28


Epigenetic inheritance models posit that during Polycomb repression, Polycomb repressive complex 2 (PRC2) propagates histone H3 lysine 27 trimethylation (H3K27me3) independently of DNA sequence. We show that insertion of Polycomb response element (PRE) DNA into the Drosophila genome creates extended domains of H3K27me3-modified nucleosomes in the flanking chromatin and causes repression of a linked reporter gene. After excision of PRE DNA, H3K27me3 nucleosomes become diluted with each round of DNA replication, and reporter gene repression is lost. After excision in replication-stalled cells, H3K27me3 levels stay high and repression persists. H3K27me3-marked nucleosomes therefore provide a memory of repression that is transmitted in a sequence-independent manner to daughter strand DNA during replication. In contrast, propagation of H3K27 trimethylation to newly incorporated nucleosomes requires sequence-specific targeting of PRC2 to PRE DNA.

The ability of certain histone-modifying enzymes to bind to the modification they generated has led to models where such enzymes might propagate modified chromatin domains by a positive-feedback loop, independently of the underlying DNA sequence. Two paradigms of chromatin states have been proposed to be maintained by such an epigenetic inheritance mechanism: constitutive heterochromatin with histone H3 lysine 9 di- and trimethylation (H3K9me2/3) generated by Suv39/Clr4 enzymes (13) and Polycomb-repressed chromatin marked with H3 lysine 27 trimethylation (H3K27me3) by Polycomb repressive complex 2 (PRC2) (4, 5). In both chromatin states, these histone modifications are essential for repressing gene transcription (6, 7). To date, there is compelling evidence that H3K9me2/3- and H3K27me3-modified nucleosomes are transmitted to a daughter strand DNA during replication (2, 3, 8). However, the steps required to propagate these modifications are much less understood. Fission yeast Clr4 has the capacity to propagate ectopically induced H3K9me2/3 domains over many cell divisions by an H3K9me2/3-based positive-feedback loop but only in cells mutated for H3K9me2/3 demethylase activity (2, 3). In the case of PRC2, allosteric activation of the enzyme induced by binding to H3K27me3 has been proposed to be the foundation for propagating H3K27me3 chromatin (4, 5). In mammalian cells, transient DNA-tethering of PRC2 generates short ectopic H3K27me3 domains that were at least partially maintained for several cell divisions after release of DNA-tethered PRC2 (4, 9). However, in Drosophila, where PRC2 and other Polycomb group (PcG) protein complexes are targeted to Polycomb response elements (PREs) (10, 11), repression imposed by insertion of PRE DNA next to a reporter gene was lost upon excision of PRE DNA (12, 13). Here, we investigated how insertion and excision of PRE DNA at ectopic sites in Drosophila affects binding of PcG proteins and H3K27me3 at the molecular level.

We first analyzed two previously described strains that each carried a single copy of the >PRE>dppWE–TZ reporter gene, integrated at different chromosomal locations (Fig. 1) [compare (13)]. >PRE>dppWE–TZ contains a 1.6-kilobase (kb) DNA fragment of the bxd PRE from the HOX gene Ultrabithorax (Ubx), flanked by short flippase (Flp) recognition target (FRT) recombination sites (>PRE>) to permit excision of PRE DNA by Flp-mediated recombination (Fig. 1) [compare (13)]. Adjacent to the >PRE> cassette, the construct contains a reporter gene for the wing imaginal disc enhancer from decapentaplegic (dpp) (E), linked to the hsp70 TATA box minimal promoter (T), and LacZ sequences encoding β-galactosidase (Z) (Fig. 1) [compare (13)]. In the presence of the >PRE> cassette, the transgene was silenced, and no β-galactosidase activity could be detected in wing imaginal discs of >PRE>dppWE–TZ transgenic animals (+PRE in Fig. 1) [compare (13)]. In contrast, >dppWE–TZ transgenic animals, generated by excision of the >PRE> cassette in the germ line, showed strong β-galactosidase expression in the characteristic pattern driven by the dpp enhancer (–PRE in Fig. 1) [compare (13)]. The observation that silencing of the intact >PRE>dppWE–TZ reporter gene is lost in mutants lacking PRC2 function (13) prompted us to determine the H3K27 methylation profile and binding of PcG proteins across the transgene. In both lines, the transgene had inserted into a genomic location normally devoid of H3K27me3 and PcG protein binding (Materials and Methods). We performed chromatin immunoprecipitation (ChIP) assays on batches of wing imaginal discs from >PRE>dppWE–TZ and the corresponding >dppWE–TZ transgenic animals and analyzed the immunoprecipitates by quantitative real-time PCR (qPCR). For qPCR, we used primer pairs that selectively amplified transgene sequences (amplicons 1 to 9, Fig. 1) and sequences in the genomic regions flanking the transgene insert (amplicons F′ and F″; H′ and H″ for lines 17a and 36c, respectively) (Fig. 1 and fig. S1A). As controls, we used primer pairs that amplify sequences at the endogenous bx PRE in Ubx that are known to be bound by PcG proteins (C2) or enriched for H3K27me3 (C1 and C3) (14) and at two regions elsewhere in the genome (C4 and C5) without PcG protein binding or H3K27me3 (Fig. 1).

Fig. 1 Long-term propagation of H3K27me3 chromatin requires PRE DNA sequences.

(Left) Imaginal wing discs from >PRE>dppWE–TZ (+PRE) and dppWE–TZ (–PRE) animals, and from >PRE>dppWE–TZ animals at 12, 32, or 56 hours after induction of PRE excision, stained with X-gal; transgene insertion 17b in all cases. (Right) ChIP analysis in wing imaginal discs from the same animals, monitoring Pc, H3K27me3, and H3. See text and table S3 for information on qPCR amplicons. Each graph shows the results from independent ChIP reactions (N ≥ 3) that were each performed on an independently prepared batch of chromatin; ChIP signals are presented as percentage of input chromatin precipitated for each region; error bars, SD. After PRE excision, region 3 can no longer be analyzed (asterisk). At C2, the low H3K27me3 and H3 ChIP signals reflect reduced nucleosome occupancy in the center of the bx PRE (14); region 3 is located 0.7 kb off the center of the transgene PRE, and therefore, the reduced nucleosome occupancy in this PRE (14) is not detected.

The PRC1 subunits Polycomb (Pc) and Polyhomeotic (Ph) and the PRC2 subunit E(z) were specifically enriched at the transgene PRE in animals carrying >PRE>dppWE–TZ (Fig. 1 and fig. S1A), and as expected, no binding was detected in >dppWE–TZ animals (Fig. 1 and fig. S1A). In both >PRE>dppWE–TZ transgenic lines, H3K27me3 was present at high levels across a domain that extended about 4 to 5 kb to either side of the >PRE> cassette, spanning almost the entire construct (Fig. 1 and fig. S1B). No enrichment of H3K27me3 was detectable at the >dppWE–TZ transgene (Fig. 1 and fig. S1B). At >PRE>dppWE–TZ, PRC2 thus trimethylates H3K27 across a chromatin interval that spans about 8 to 10 kb.

To estimate to what extent nucleosomes at the >PRE>dppWE–TZ transgene are trimethylated at H3K27, we determined the H3K27me2 profile. H3K27me2 levels across the >PRE>dppWE–TZ transgene were much lower than at C4 and C5 and comparable to the levels at Ubx (regions C1 to C3) that is repressed and predominantly trimethylated at H3K27 in wing imaginal discs (fig. S1A). Conversely, across >dppWE–TZ, H3K27me2 levels were much higher and comparable to those seen at C4 and C5 (fig. S1A). This suggests that the nucleosomes across the >PRE>dppWE–TZ transgene are predominantly trimethylated at H3K27.

Excision of the >PRE> cassette from >PRE>dppWE–TZ transgenic animals by heat shock–induced expression of Flp during larval development results in the appearance of β-galactosidase expression in the dpp pattern 12 hours after the heat shock (Fig. 1) [compare (13)]. We measured the efficiency of PRE excision and found that 8 hours after a single 1-hour heat shock, excision had occurred in about 95% of wing imaginal disc cells (fig. S1C). The delayed increase of β-galactosidase expression over time (Fig. 1) (13) suggests a gradual, rather than abrupt, loss of repression. We performed ChIP analyses on chromatin prepared from batches of entire wing imaginal discs dissected from >PRE>dppWE–TZ transgenic animals 12, 32, or 56 hours after Flp induction. This allowed monitoring the consequences of PRE excision in cells that had undergone at least one cell division (+12 hours) or at least two (+32 hours) or more than four (+56 hours) cell divisions (15). Twelve hours after Flp induction, H3K27me3 levels were reduced by at least half across the entire transgene and further reduced by at least half at the 32 hours time point (Fig. 1 and fig. S1B). Fifty-six hours after Flp induction, H3K27me3 levels across the transgene were nearly as low as in >dppWE–TZ animals derived from >dppWE–TZ germ cells (Fig. 1 and S1B). The histone H3 profile was unaltered at all time points (Fig. 1), suggesting that PRE excision does not cause global disruption of nucleosome occupancy across the transgene. The loss of H3K27me3 after PRE excision suggests that PRC2 is unable to propagate H3K27me3 across the >dppWE–TZ transgene in the absence of PRE DNA.

In parallel, we monitored Pc protein binding after PRE excision. Pc, unlike Ph or other PRC1 subunits, is not only bound at PREs but also associates with the chromatin-flanking PREs (Fig. 1 and fig. S1A) [compare (14, 16)], likely by interaction with H3K27me3-modified nucleosomes (17, 18). Twelve hours after PRE excision, Pc binding at the transgene was already reduced almost to background levels (Fig. 1).

We then analyzed the H3K27me3 profile at the >PRE>dppWE–UZ transgene that contains a 4.1-kb fragment of the Ubx promoter instead of the hsp70 minimal promoter (13). At >PRE>dppWE–UZ, the H3K27me3 domain spans about 12 kb (fig. S1D) and is thus about 4 kb longer than at >PRE>dppWE–TZ. Nevertheless, after PRE excision, H3K27me3 at dppWE–UZ was lost at a rate comparable to that seen at dppWE–TZ (fig. S1D). Ubx promoter DNA thus does not enable H3K27me3 propagation. We conclude that even at a domain that spans 12 kb and therefore comprises about 60 nucleosomes, PRC2 is unable to propagate H3K27me3 in the absence of PRE DNA.

We next analyzed the H3K27me3 profile and reporter gene repression after PRE excision in animals in which DNA replication had been blocked. We reared larvae in liquid medium containing aphidicolin, an inhibitor of DNA polymerases A and D, which resulted in a complete block of DNA replication in imaginal discs (fig. S2A). In larvae reared in aphidicolin-containing medium, Flp-induced PRE excision from >PRE>dppWE–TZ was as efficient as under normal growth conditions (fig. S2B), but 12 hours after excision, H3K27me3 levels at the transgene were undiminished compared with +PRE control larvae (Fig. 2A, left). In larvae reared in liquid medium without aphidicolin, PRE excision resulted in the expected reduction by half of H3K27me3 levels after 12 hours (Fig. 2A, right). Together, this suggests that the loss of H3K27me3 nucleosomes after PRE excision in proliferating cells reflects their dilution as they become transmitted to DNA daughter strands during replication. Unlike under normal growth conditions (Fig. 1), aphidicolin-treated larvae lacked detectable β-galactosidase expression 12 hours after PRE excision (Fig. 2B). When these animals were permitted to recover in medium lacking aphidocolin, they resumed DNA replication (fig. S2A) and began expressing β-galactosidase (Fig. 2B). If DNA replication is blocked and H3K27me3 levels stay high, repression is thus also sustained in the absence of PRE DNA, possibly by PRC1.

Fig. 2 H3K27me3 and repression persist after PRE excision but are eliminated by DNA replication.

(A) (Left) H3K27me3 ChIP analysis in wing imaginal discs from >PRE>dppWE–TZ (line 17b) transgenic animals reared in aphidicolin-containing medium. Larvae were preincubated in aphidicolin medium for 6 hours to establish a complete block of DNA replication (fig. S2A), followed by induction of PRE excision and incubation in aphidicolin medium for an additional 12 hours (PRE excision +12 hours). The same growth regime was applied to larvae without PRE excision (+PRE). (Right) H3K27me3 ChIP analysis in control larvae reared for 16 hours in liquid medium lacking aphidicolin. (B) Imaginal wing discs from >PRE>dppWE–TZ (line 17b) transgenic animals subjected to the indicated growth regimes and heat shock (HS) treatment and stained with X-gal. The lower level of β-galactosidase expression 24 hours after heat shock in animals initially exposed to aphidicolin (bottom middle) compared with control animals (bottom right) suggests delayed loss of repression. (C) H3K27me3 ChIP analysis at the >PRE>dppWE–TZ transgene (line 17b) in wing imaginal discs from Utx/ Df (2L)BSC143 animals. (A) and (C) Results from at least three independent ChIP reactions, each performed on an independently prepared batch of chromatin. ChIP signals are presented as in Fig. 1.

Finally, we induced PRE excision from >PRE>dppWE–TZ in larvae that were hemizygous for UtxΔ, a null mutation in the single H3K27me3 demethylase in Drosophila (19). Twelve hours after Flp induction, H3K27me3 levels at the transgene were reduced by about half, as they are in wild-type animals (Fig. 2, B and C). This suggest that demethylation of H3K27me3 by Utx does not contribute to the disappearance of H3K27me3 from transgene chromatin after PRE excision.

These results lead to the following conclusions. First, PRE cis-regulatory DNA provides the genetic basis not only for generating but also for propagating H3K27me3-modified chromatin. This argues against a simple epigenetic model where PRC2 binding to parental H3K27me3 nucleosomes after replication would suffice to propagate H3K27 trimethylation in daughter strand chromatin. PRC2 needs to be recruited to PRE DNA first (14, 16), before allosteric activation through interaction with H3K27me3 nucleosomes (5, 20) in flanking chromatin may then facilitate methylation of newly incorporated nucleosomes. Second, after PRE excision and replication, parental H3K27me3 nucleosomes remain associated with the same underlying DNA in daughter cells and provide epigenetic memory. In particular, in replication-stalled cells, high levels of H3K27me3-modified nucleosomes support repression even in the absence of PRE DNA. In proliferating cells, however, the dilution of these modified nucleosomes is accompanied with loss of repression after one cell division. H3K27me3 nucleosomes only appear to provide short-term epigenetic memory of the repressed state. DNA targeting of PRC2 after replication to replenish H3K27me3 is therefore critical to preserve repression.

Drosophila HOX and other large-size PcG target genes often contain multiple PREs and H3K27me3 domains that span dozens of kilobases. Deletion of single PREs from these genes typically results in only minor diminution of the H3K27me3 profile (21) and misexpression is less severe (2123) than misexpression of the native genes in PcG mutants. Furthermore, when the same >PRE> cassette that was used here was excised from a Ubx-LacZ reporter gene with more extended Ubx upstream regulatory sequences, repression was lost with a longer delay (12); this suggests that additional elements with PRE properties in those Ubx sequences sustained repression and H3K27me3 through more cell divisions. The evolution of PRE DNA sequences and of their frequency and arrangement within target genes may thus ultimately determine stability and heritability of H3K27me3 chromatin and PcG repression.

Supplementary Materials

Materials and Methods

Figs. S1 and S2

Tables S1 to S3

References (24, 25)

References and Notes

  1. Acknowledgments: We thank G. Struhl for discussions. This work was supported by the Deutsche Forschungsgemeinschaft (SFB1064) and the Max Planck Society.
View Abstract

Navigate This Article