Essential Role of the Glycosyltransferase Sxc/Ogt in Polycomb Repression

See allHide authors and affiliations

Science  03 Jul 2009:
Vol. 325, Issue 5936, pp. 93-96
DOI: 10.1126/science.1169727


Polycomb group proteins are conserved transcriptional repressors that control animal and plant development. Here, we found that the Drosophila Polycomb group gene super sex combs (sxc) encodes Ogt, the highly conserved glycosyltransferase that catalyzes the addition of N-acetylglucosamine (GlcNAc) to proteins in animals and plants. Genome-wide profiling in Drosophila revealed that GlcNAc-modified proteins are highly enriched at Polycomb response elements. Among different Polycomb group proteins, Polyhomeotic is glycosylated by Sxc/Ogt in vivo. sxc/Ogt–null mutants lacked O-linked GlcNAcylation and failed to maintain Polycomb transcriptional repression even though Polycomb group protein complexes were bound at their target sites. Polycomb repression appears to be a critical function of Sxc/Ogt in Drosophila and may be mediated by the glycosylation of Polyhomeotic.

The Drosophila gene super sex combs (sxc) was originally identified because mutations in this gene caused lethality at the pupal stage and homeotic transformations of multiple body segments into segments normally present in more posterior body regions (1). This phenotype suggested that multiple homeotic (HOX) genes were expressed outside of their normal expression domains, and the HOX gene Ultrabithorax (Ubx) was indeed misexpressed in sxc mutant larvae (2). Consequently, sxc was classified as a Polycomb group (PcG) gene. To explore the role of sxc in PcG repression, we stained larvae that were transheterozygous for two different sxc-null mutations with antibodies against the protein products of the PcG target genes Ubx, Abdominal-B (Abd-B), Sex combs reduced (Scr), engrailed (en), Distal-less (Dll), and teashirt (tsh). Each of these genes was expressed outside of its normal expression domain (fig. S1). sxc is thus needed for repression of multiple PcG target genes in Drosophila larvae.

We fine-mapped the sxc gene to a 160-kb genomic interval (3). Sequencing candidate open reading frames in genomic DNA from sxc mutants (3) revealed that sxc1, sxc4, sxc5, and sxc7 each had individual base changes in the open reading frame of CG10392 that changed an amino-acid codon into a premature termination codon or a codon for a different amino acid (Fig. 1A). In sxc6, a point mutation destroys the splice acceptor site in the fourth intron of CG10392 (Fig. 1A). CG10392 encodes O-linked N-acetylglucosamine (O-GlcNAc) transferase (Ogt), the enzyme that catalyzes addition of GlcNAc to serine or threonine residues of a broad range of nuclear and cytosolic proteins in animals and plants (4, 5). The molecular lesions in five different sxc alleles therefore identified sxc as the gene that encodes Ogt, and in the following we call this gene sxc/Ogt. We next analyzed the Ogt protein in extracts from sxc mutant larvae with an antibody to Ogt. No Ogt protein was detected in sxc3, sxc6, or sxc7 mutants, suggesting that these are protein-null mutations (Fig. 1B). OgtN948I (the mutant Ogt protein encoded by sxc4) and OgtΔ1031-1059 (the C-terminally truncated Ogt protein encoded by sxc5) were both expressed as stable polypeptides in mutant larvae (Fig. 1B). The phenotype of sxc4 and sxc5 mutants is as severe as that of the protein null mutants sxc3, sxc6, or sxc7 (13). The expressed OgtΔ1031-1059 and OgtN948I proteins, containing lesions predicted to affect the fold of the glycosyltransferase domain (6, 7), therefore seem to be completely nonfunctional. An intact glycosyltransferase domain in Ogt is thus critical for Polycomb repression.

Fig. 1

Molecular characterization of the Drosophila PcG gene sxc/Ogt. (A) Structure of the Sxc/Ogt protein and lesions in sxc mutant alleles. The 13 tetratrico peptide repeats and the bipartite glycosyltransferase domain composed of CDI and CDII, also called N- and C-terminal GT-B subdomains (6, 7), are shown as white, gray, and black boxes, respectively. Base substitutions in sxc mutant alleles are indicated. OgtΔ1031-1059, encoded by sxc5, lacks the C-terminal helix (gray box) that forms part of the CDI domain. Asn948, mutated in sxc4, is conserved in fly, worm, mouse, and human Ogt. (B) Detection of Sxc/Ogt protein in wild-type (wt) and sxc mutant larvae. Immunoblot of imaginal disc extracts from third instar larvae of the indicated genotype, probed with antibodies to human Ogt (top) and α-tubulin (bottom) is shown. The 80-kD protein (asterisk) is a truncated protein encoded by sxc1. The viable allele sxc2 expresses a full-length protein (3). Lack of functional Ogt in sxc7/sxc1 mutants is further supported by the loss of O-GlcNAcylation on nucleoporins (fig. S6).

More than 100 different proteins have been reported to bear the O-GlcNAc modification in mammalian cells, including RNA polymerase II and several transcription factors and coregulators (4, 5). Labeling of Drosophila polytene chromosomes with wheat germ agglutinin (WGA), a lectin that binds with high affinity to GlcNAc residues (8, 9), suggested that GlcNAcylated proteins are also present on Drosophila chromosomes (10). We determined the genome-wide distribution of GlcNAc-modified proteins in the chromatin of imaginal disc cells from Drosophila larvae by performing chromatin immunoprecipitation (ChIP) assays with an antibody that recognizes the O-GlcNAc modification (11) and with a WGA-affinity resin (3). We hybridized the precipitated material to high-density whole-genome tiling arrays and analyzed them using TileMap (12) using a stringent cutoff. Only genomic regions significantly enriched by both antibody to O-GlcNAc and WGA-affinity resin were considered, resulting in a high-confidence set of 1138 genomic sites bound by GlcNAc-modified proteins (Fig. 2A) (3). We next compared the GlcNAc profile with the genome-wide binding profile of the PcG protein Ph that we had generated in parallel (3) and with that of the PcG protein complex PhoRC (13). 490 of the 1138 GlcNAc sites overlapped with Polycomb response elements (PREs) bound by Ph and/or PhoRC (Fig. 2A, left). Moreover, among the 1% top-ranked GlcNAc sites nearly all (111 of 114) overlapped with Ph- and/or PhoRC-bound regions (Fig. 2A, right). In PcG target genes misexpressed in sxc mutants (fig. S1), GlcNAc-modified proteins were specifically localized at PREs (Fig. 2B and fig. S2). To validate these findings, we performed ChIP assays with the antibody to O-GlcNAc in imaginal discs from wild-type or sxc mutant larvae and analyzed immunoprecipitates for enrichment of selected DNA sequences of the PcG target genes Ubx, Abd-B, Scr, en, Dll, tsh, and pannier (pnr) (Fig. 3A). GlcNAc ChIP signals were present at PREs in wild-type but not in sxc mutant larvae, demonstrating that the antibody to O-GlcNAc indeed specifically immunoprecipitated GlcNAc-modified proteins bound at these PREs (Fig. 3A). We also compared the GlcNAc profile at the Ubx gene in wing (Ubx repressed) and haltere/third leg–disc cells (Ubx active). GlcNAcylated proteins were bound at Ubx PREs both in cells in which the gene is repressed and in cells in which it is active (Fig. 3A), much like the PcG protein complexes PhoRC, Polycomb repressive complex 1 (PRC1), or PRC2 (14).

Fig. 2

O-GlcNAc–modified proteins are localized at PREs in Drosophila. (A) Venn diagrams showing the overlap of 1138 10% top-ranked O-GlcNAc sites (left) or 114 1% top-ranked O-GlcNAc sites (right) with 1689 Ph and 338 PhoRC-bound regions in larval imaginal discs (tables S1 and S2) (3). (B) ChIP profiles of O-GlcNAc, Ph, and PhoRC subunits Pho and dSfmbt at the Antennapedia complex in imaginal disc cells. Hybridization intensities for oligonucleotide probes are plotted as color bars above the genomic map (release 5, kilobase coordinates) of Drosophila melanogaster; significantly enriched regions are marked below plots (3). The HOX genes labial (lab), proboscipedia (pb), Deformed (Dfd), Scr, and Antennapedia (Antp) on the plus (above) and minus (below) strand are represented with exons (black boxes) and introns (thin black lines).

Fig. 3

O-GlcNAc modification and PcG protein binding in wild-type and sxc mutant chromatin. (A) ChIP analysis monitoring O-GlcNAc modification in wing (magenta) and haltere/third leg (pink) imaginal discs from wt and sxc7/sxc1 mutant (sxc) larvae. Graphs show results from independent immunoprecipitation reactions (n = 3 immunoprecipitation reactions) with antibody to O-GlcNAc. ChIP signals, quantified by means of quantitative polymerase chain reaction, are presented as mean percentage of input chromatin precipitated at each region; error bars indicate ±SD (3). Locations of PREs (purple boxes) and other regions relative to transcription start sites are indicated in kilobases; euchromatic (eu.) and heterochromatic (het.) control regions were mapped outside of these genes. In wt larvae, GlcNAc ChIP signals at the –30 kb Ubx PRE were comparable in wing and haltere/third leg chromatin, but at the +32kb PRE the signal in haltere/third-leg chromatin was two- to threefold lower than in wing chromatin, paralleling PRC1 and PRC2 binding (14). (B) ChIP signals in WT (blue bars) and sxc wing discs (orange bars) representing Pho, E(z), and Ph binding.

We next performed ChIP assays with antibodies against the PhoRC subunit Pho, the PRC1 subunit Ph, and the PRC2 subunit E(z) to test whether binding of any of these protein complexes might be affected in sxc mutants. Comparison of the ChIP profiles in wild-type and sxc mutant wing discs showed that binding of Pho and E(z) at PREs was largely unaffected in sxc mutants (Fig. 3B). Ph was also bound at high levels at all PREs in sxc mutant discs, but Ph ChIP signals were 1.5- to twofold reduced at most PREs as compared with those in wild-type discs (Fig. 3B), even though nuclear localization of Ph appeared comparable in wild-type and sxc mutant cells (fig. S3). However, not only Pho and E(z) but also Ph were all bound at wild-type levels at the Scr and tsh PREs in sxc mutant discs (Fig. 3B), and yet Polycomb repression at these two genes was lost in the mutant discs (fig. S1). These results suggest that loss of target gene repression is not solely due to reduction of PcG protein complex binding, although it is possible that binding of other PcG proteins is more severely affected in sxc mutants. Lastly, we performed ChIP analyses with antibodies against trimethylated lysine 27 of histone H3 (H3-K27me3), the PcG-specific chromatin modification generated by PRC2 (15). H3-K27me3 levels in target-gene chromatin were comparable in wild-type and sxc mutant larvae, and so were bulk H3-K27me3 levels (fig. S4). Sxc/Ogt is therefore apparently not critical for recruitment of PRC2 or for its ability to trimethylate H3-K27 in target gene chromatin.

The similarity between the GlcNAc and PcG protein ChIP profiles prompted us to test whether any of the PcG proteins themselves may carry the O-GlcNAc modification. We used WGA-agarose to affinity-purify GlcNAc-modified proteins from wild-type or sxc mutant larval extracts and probed the purified material with antibodies against the PhoRC subunits dSfmbt and Pho; the PRC1 subunits Ph, Pc, Ring, and Scm; and the PRC2 subunits Su(z)12 and Nurf55. Ph, but none of the other proteins tested, was strongly enriched after WGA-affinity purification from wild-type but not from sxc mutant larvae (Fig. 4). This enrichment of Ph was also observed under denaturing conditions (Fig. 4). This suggests that Ph itself is GlcNAcylated by Sxc/Ogt. PRC1 components Ring, Pc, or Scm did not copurify with GlcNAc-modified Ph in these WGA pull-down assays from larval extracts (Fig. 4), although Ring and Pc were readily detected together with Ph in such purifications from embryonic nuclear extracts (fig. S4). This finding could be explained by differential association of Ph with PRC1 components and/or different accessibility of GlcNAc moieties on Ph during embryonic and larval development. WGA-purifications failed to provide evidence for GlcNAcylation of the large subunit of RNA polymerase II (Fig. 4) Rpb1, which was previously proposed to be a mechanism by which Ogt might mediate transcriptional repression in mammalian cells (16).

Fig. 4

Sxc/Ogt glycosylates Ph in Drosophila. Extracts of wild-type (WT) or sxc1/sxc7 (sxc) mutant larvae were subjected to affinity-purification with WGA-agarose under native or denaturing (denat.) conditions (3) and probed with antibodies to PRC1 subunits Ph, Pc, Ring, and Scm; PhoRC subunits dSfmbt and Pho; PRC2 subunits Su(z)12 and Nurf55; the large RNA polymerase II subunit Rpb1; and Ogt. “I” indicates 0.5% of input extract and “E” indicates 10% of affinity-purified material. Ph, but none of the other proteins, is strongly enriched in WGA affinity–purified material (lanes 2 and 4); no enrichment of any protein is seen in material purified from sxc mutant larvae (lane 6). Weak enrichment of dSfmbt under native (lane 4) but not denaturing conditions (lane 2) probably reflects association with GlcNAc-modified Ph. Levels of Ph are increased in sxc mutant larvae as compared with those in WT larvae (compare lane 5 with lanes 1 and 3), possibly reflecting a failure to downregulate Ph expression by the PcG system through PREs in the Ph gene (23).

The work reported here shows that GlcNAcylation by Sxc/Ogt plays an essential role in PcG repression. Ogt is a single-copy gene in mice, flies, and worms, and the only known glycosyltransferase that adds GlcNAc moieties to nuclear and cytosolic proteins (4, 5). Ogt is essential in mice, in which it is required for the viability of embryonic stem cells (17, 18), but dispensable for the normal development of Caenorhabditis elegans (19). In contrast, Drosophila mutants lacking Sxc/Ogt and O-GlcNAcylation display a specific phenotype: loss of Polycomb repression. sxc mutants show no other obvious developmental defects, suggesting that PcG repression is the main process that critically depends on O-GlcNAcylation in Drosophila. We provide evidence that Ph is GlcNAcylated. Although it remains to be determined whether Ph is indeed the relevant Sxc/Ogt substrate in PcG repression, it is tempting to speculate that the function of Sxc/Ogt in gene silencing may be to GlcNAcylate Ph. The sxc null phenotype is not as severe as that of other PcG mutants, notably that of ph [(1, 20, 21) and this study]. Thus, if GlcNAcylation of Ph contributes to its function, Ph still retains partial repressor activity in the absence of this modification. One possibility would be that GlcNAcylation of Ph is needed for efficient anchoring of Ph to PREs and/or for the capacity of PRE-tethered Ph to maintain a repressed chromatin state at target genes.

All Drosophila PcG proteins are conserved in vertebrates, and the PcG system represses a large set of orthologous developmental regulator genes both during Drosophila development and in mammalian embryonic stem cells (13, 22). Thus, GlcNAcylation of Polyhomeotic homologs, or perhaps other PcG proteins, may also be an evolutionary ancient and essential function of Ogt in vertebrates.

Supporting Online Material

Materials and Methods

Figs. S1 to S6

Tables S1 and S2


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank C. Fritsch and B. Papp for initial efforts to map sxc; M. Nekrasov, A. Nowak, and N. Ly-Hartig for raising E(z), Nurf55, and Scm antibodies, respectively; R. Matos and A. Gaytan for reagents; J. deGraaf and V. Benes (of the EMBL Genomics Core Facility) for technical support; and J. Gagneur, C. Girardot, B. Honda, P. Ingham, and K. Osborne for discussions. Comments from E. Conti helped to improve the manuscript. The authors are supported by EMBL and by grants from the Deutsche Forschungsgemeinschaft. The microarray data have been deposited in the ArrayExpress database under the accession number E-TABM-697.
View Abstract

Navigate This Article