Drosophila O-GlcNAcase Deletion Globally Perturbs Chromatin O-GlcNAcylation*

Gene expression during Drosophila development is subject to regulation by the Polycomb (Pc), Trithorax (Trx), and Compass chromatin modifier complexes. O-GlcNAc transferase (OGT/SXC) is essential for Pc repression suggesting that the O-GlcNAcylation of proteins plays a key role in regulating development. OGT transfers O-GlcNAc onto serine and threonine residues in intrinsically disordered domains of key transcriptional regulators; O-GlcNAcase (OGA) removes the modification. To pinpoint genomic regions that are regulated by O-GlcNAc levels, we performed ChIP-chip and microarray analysis after OGT or OGA RNAi knockdown in S2 cells. After OGA RNAi, we observed a genome-wide increase in the intensity of most O-GlcNAc-occupied regions including genes linked to cell cycle, ubiquitin, and steroid response. In contrast, O-GlcNAc levels were strikingly insensitive to OGA RNAi at sites of polycomb repression such as the Hox and NK homeobox gene clusters. Microarray analysis suggested that altered O-GlcNAc cycling perturbed the expression of genes associated with morphogenesis and cell cycle regulation. We then produced a viable null allele of oga (ogadel.1) in Drosophila allowing visualization of altered O-GlcNAc cycling on polytene chromosomes. We found that trithorax (TRX), absent small or homeotic discs 1 (ASH1), and Compass member SET1 histone methyltransferases were O-GlcNAc-modified in ogadel.1 mutants. The ogadel.1 mutants displayed altered expression of a distinct set of cell cycle-related genes. Our results show that the loss of OGA in Drosophila globally impacts the epigenetic machinery allowing O-GlcNAc accumulation on RNA polymerase II and numerous chromatin factors including TRX, ASH1, and SET1.

Epigenetic regulation of gene expression during development is essential for proper cell fate determination. Epigenetic modifiers act by modifying chromatin and thereby altering chromatin structure. The Polycomb (Pc) 2 repressor and Trithorax (Trx) and Compass activator complexes play major roles in maintaining gene expression profiles required for proper body plan formation. Methylation of several lysine residues of histone 3 are among the best understood epigenetic modifications. The trimethylation of histone 3 lysine 27 (H3K27me3) by Polycomb repressive complex 2 (PRC2) member Enhancer of zeste (E(z)) is a repressive transcription mark (1)(2)(3). In contrast, histone 3 lysine 4 monomethylation (H3K4me) performed by TRX, histone 3 lysine 36 dimethylation (H3K36me2) by ASH1, and histone 3 lysine 4 trimethylation (H3K4me3) by Compass member SET1 are activating modifications in Drosophila (4,5). Pc group member super sex combs (ogt/sxc) encodes the Drosophila O-GlcNAc transferase (OGT), which regulates Pc-mediated repression by post translationally O-GlcNAcylating and stabilizing Polyhomeotic (Ph), a member of PRC1 in Drosophila (6 -8). Ph forms large protein aggregates and cannot function in ogt mutants, leading to homeotic defects (7,9). Moreover, it has been shown that knockdown of mammalian OGT decreases H3K27me3 levels by affecting the stability of E(z) homolog 2 (EZH2) in the MCF7 breast cancer cell line (1).
The hexosamine biosynthetic pathway generates UDP-GlcNAc using glucose, glutamine, acetyl-CoA, and UTP. Therefore changes in the intracellular levels of these nutritionderived products directly influence the cellular concentration of UDP-GlcNAc making it sensitive to nutrient levels. OGT then catalyzes the addition of O-GlcNAc onto hydroxyl groups of serine/threonine residues of proteins using the nutrient sensor UDP-GlcNAc as a substrate. The O-GlcNAc modification on nucleocytoplasmic proteins is then removed by the enzyme O-GlcNAcase (OGA) in a dynamic fashion, modulating intracellular events ranging from transcription to cell cycle regulation (10,11). O-GlcNAcylation, like phosphorylation, can impact protein function, localization, and/or expression levels (12,13). The singularity of the OGT and OGA enzymes, the rapidity with which O-GlcNAc cycles, and the diversity of protein substrates poises this post-translational modification to play a critical role in modulating the rapid cellular changes required for proper development (14 -17).
O-GlcNAc was first detected on Drosophila polytene chromosomes (18) and later found at the promoter regions of Caenorhabditis elegans genes that are involved in a wide variety of pathways ranging from metabolism to aging (19). In addition to its role in Pc repression, OGT is thought to have additional roles in epigenetic regulation in mammals. First, OGT can directly O-GlcNAcylate chromatin remodelers like Sin3A and SET1DA (14,17). Although controversial, OGT is argued to directly O-GlcNAc modify histone 2B, thereby altering chromatin structure (20,21). Beyond affecting gene expression through directly or indirectly changing chromatin, O-GlcNAc influences transcription by affecting the activity and/or stability of key players including RNA polymerase II (RNA Pol II) and many transcription factors (11,16,(22)(23)(24)(25)(26). Furthermore, O-GlcNAc is known to play a role in cell cycle progression with the transcriptional co-regulator host cell factor 1 (HCF1) having been identified as an OGT target (27). Indeed, HCF1 needs to be O-GlcNAcylated and cleaved by OGT to regulate cell division (27). O-GlcNAcylation plays a key role in mitosis as overexpression of OGT or inhibition of OGA impairs cell cycle progression (28). Last, the O-GlcNAc modification of histones in a cell cycle-dependent manner may prime this post-translational modification to influence cell cycle and gene expression (29,30).
To better understand how O-GlcNAc cycling influences gene expression and which genomic regions are more susceptible to changing O-GlcNAc levels, we altered O-GlcNAc levels by knocking down either OGT or OGA expression by RNAi and performed ChIP-chip for O-GlcNAc and other chromatin associated factors followed by gene expression analysis in Drosophila Schneider 2 (S2) cells. An indicator of active transcription, phosphorylated serine 2 on the carboxyl-terminal tail of RNA polymerase II (RNA Pol II Ser2P) was generally at low levels at sites of O-GlcNAc modified chromatin including the Pho-enriched Hox gene clusters suggesting that the O-GlcNAc modification is mainly associated with transcriptionally silent regions. Interestingly, O-GlcNAc occupies additional sites on chromatin other than Pho co-occupied sites. A number of these Pho independent O-GlcNAc occupied chromatin regions were shared with RNA Pol II Ser2P underscoring that O-GlcNAc plays a role in active transcription as well. Gene expression profiling of these cells revealed that O-GlcNAc levels most significantly affect pathways including cell cycle and metabolism. Intrigued by the presence of O-GlcNAc on transcriptionally silent and active chromatin regions, we elected to study the consequences of a permanent increase in O-GlcNAc levels in the whole animal by generating a null allele of oga in Drosophila (oga del.1 ). In oga del.1 mutant animals O-GlcNAc cycling on chromatin was globally perturbed when visualized on polytene chromosomes. We determined that Trithorax members TRX and ASH1, and Compass member SET1 histone methyltransferases are O-GlcNAc modified in oga del.1 mutants. Furthermore, expression of specific cell cycle-related genes, including host cell factor, were altered in oga mutant ovaries. Our findings directly demonstrate that O-GlcNAc cycling is an important part of the epigenetic machinery in Drosophila.

Experimental Procedures
Drosophila S2 Cell Culture and RNAi Treatment-S2 cells were cultured in Schneider's Drosophila medium supplemented with 10% FBS.
RNAi knockdown of OGT and OGA in S2 cells was performed as described previously (31). Briefly, genomic DNA was isolated from S2 cells with DNAeasy kit (Qiagen). OGT, OGA, or GFP specific primers were designed to include a T7 RNA polymerase binding site. The PCR product is then in vitro transcribed to generate dsRNA (T7 MEGAscript Kit, Ambion). dsRNA was purified with the RNAeasy kit (Qiagen) and trans-fected to S2 cells. Cells were harvested to isolate RNA for transcriptomics or chromatin isolation for ChIP-chip analysis 3 days after transfection.
Antibodies Used for ChIP-chip-Mouse anti-O-GlcNAc (Thermo Scientific, MA1-076) and rabbit anti-RNA Pol II Ser2P (Abcam, ab5095) antibodies were used. Pho antibody was described earlier (32). After chromatin purification, Pho, O-GlcNAc, and RNA Pol II Ser2P ChIP-chip was performed by a minor modification of the method described previously (19).
Whole Genome Transcriptome Analysis-Transcriptome analysis was performed using Affymetrix Genechip Drosophila Genome 1.0 Arrays. cDNA was prepared using Smartscribe prior to library synthesis according to the manufacturer's instructions. Statistical analysis was performed as previously described (19).
ChIP-on-Chip Analysis-ChIP-chip analysis was carried out in Drosophila S2 cells by a modification of the method described previously using anti-O-GlcNAc antibody (mouse HGAC-85) (19).
Probe Signal and Enrichment Analysis-Analysis was performed using Affymetrix GeneChip Drosophila Tiling 1.0R Arrays and analyzed using Affymetrix build 5 (for NCBI). The CEL files (Cell Intensity Files; containing processed image data of the array scans) were analyzed using Affymetrix Tiling Analysis Software (TAS version 1.1.02). A two-sample analysis was performed comparing each CEL file of the ChIP/IP samples against the CEL file from the input DNA array. This analysis generates BAR (binary analysis results) files that contain the signal values for all probes on the arrays. Signal values are "estimates of-fold enrichment" of ChIP/IP-DNA, which in essence are ratios (in linear scale) between the intensity of the probes on the ChIP/IP array divided by the intensity of the corresponding probe on the input DNA array. To make the values more significant, however, these ratios are computed by applying averaging and ranking steps to a set of probes within a 400 -800-bp sliding window. The TAS parameters used for binary analysis results file generation are given in the summary file (sheet: TAS parameters; "Analyze Intensities").
Interval Analysis-An interval is a discrete genomic region, defined by the chromosome number and a start and end coordinate. Intervals represent the locations of signal peaks. For each binary analysis results file, intervals are calculated using Affymetrix TAS and compiled into BED files (browser extensible data). Ratios of normalized averaged signal intensities between Chips were used to calculate fold-enrichment between OGA, WT control (GFP), and OGT knockdown experiments. The co-enrichment of O-GlcNAc, Pho, RNA Pol II Ser2P, and other chromatin factors were determined using Affymetrix Tiling Analysis Software version 1.1.
Two-sample Analysis-In this analysis pipeline, for each tiling probe, an enrichment is estimated, and this involves combining two statistical approaches: the Wilcoxon signed-ran test (a nonparametric paired difference test) and the Hodges-Lehmann estimator (a robust and nonparametric estimator of the location parameter of a population). All of the peaks we reported had significant co-enrichments as determined by the statistical tests mentioned above; these are the values used to populate the co-enrichment table. The ChIP-chip and gene expression microarray experiments were both done in triplicate. Data were submitted in the GEO database as GSE74846.
Functional Annotation Clustering-Functional annotation clustering is a tool in version 6.7 of DAVID available (david.abcc.ncifcrf.gov) for annotation, visualization, and integrated discovery that analyzes enrichment in related gene sets into clusters by using a variety of assembled gene sets in biological pathways. For ChIP-chip analysis, we identified the group of genes that had O-GlcNAc occupied regions that showed little or no increase (0.5-1), moderately increased (1-1.5), and highly increased (1.5-2.25) with respect to O-GlcNAc levels following OGA RNAi using high classification stringency. For each clustering analysis only the most highly enriched two groups was shown. The same functional clustering with medium stringency was used to analyze genes whose expression was altered by changes in O-GlcNAc cycling. Genes that showed 1.5-fold or more change in expression were used for OGT RNAi, and genes that displayed altered expression of 1.2-fold were used for OGA RNAi.
Fly Stocks-13618 OGA P element insertion, ogt/sxc mutants, Tubulin Gal4, Actin-Gal4, Nanos Gal4, transposase lines, and the two deficiency lines spanning the oga gene, B9485 and B9487, were from the Bloomington Stock Center. The UAS-OGA-RNAi fly line was obtained from VDRC (33). The reported UAS-OGA overexpression lines were originally generated by Kaasik et al. (34). oga del.1 mutant was generated by standard P-element excision protocol (35). oga del.1 mosaics were generated using the FRT/FLP recombination system (36). Flies were maintained at 25°C in a humidified incubator. Drosophila MM media was purchased from KD Medical (Columbia, MD).
Polytene Chromosome Staining and Imaging-Polytene chromosomes were prepared as described previously (37). For staining, the slides containing polytenes were incubated with 100, 50, and 25% ethanol followed by PBS/Triton X-100 (0.1% Triton X-100). After 3 washes with PBS/Triton X-100, the slides were blocked with Odyssey blocking reagent for 1 h at room temperature, and incubated with ASH1, TRX, or SET1 at 1/50 dilution along with O-GlcNAc specific antibody at 1/100 dilution overnight at 4°C in a humidified chamber. On the next day slides were washed 4 times with PBS/ Triton X-100 and incubated with Alexa Fluor-conjugated secondary antibodies in the dark for 2 h at room temperature. The slides were then mounted in Slowfade mounting medium (Invitrogen) and visualized using a Zeiss LSM 700 confocal microscope with Zen imaging software (Zeiss). Primary antibodies and dilutions used for staining were: rabbit anti-ASH1 (Novus number 50100002), anti-RNA Pol II Ser2P (Abcam, ab5095), anti-Polycomb (Santa Cruz, number sc-25762), and mouse HGAC-85 anti-O-GlcNAc (Thermo Scientific, number MA1-076). Rabbit anti-SET1 and anti-TRX antibodies were a kind gift from Dr. Shilatifard (38). All secondary antibodies were Alexa Fluor 488-or Alexa Fluor 568-conjugated (Invitrogen) and used at 1/250 dilution.
Immunoprecipitation-Flies were fed with fresh yeast paste for 3 days and ovaries were dissected in ice-cold PBS. Ovary extracts were prepared in T-PER tissue extraction buffer (Thermo) containing protease inhibitors (Roche Applied Science) in an Eppendorf tube with a hand held pellet pestle (Kon-tes). Samples were then homogenized further on a rocker for 1 h at 4°C. After centrifugation for 5 min at 14,000 rpm at 4°C, the supernatant was used for immunoprecipitation. 300 g of protein was used for immunoprecipitation in 500 l of PBS, Triton X-100. ASH1 (Novus, number 50100002), BRE1 (Novus, number 40280002), SET1, and TRX antibodies were described (38). All primary antibodies were used at 1/100 dilution. Samples with antibodies were kept on a rocker overnight at 4°C. 25 l of Protein A/G beads were added the next morning and samples were rocked another 2 h. Samples were then centrifuged at 1,000 rpm (ϳ300 ϫ g) for 5 min at 4°C and washed three times with PBS/Triton X-100 for 15 min each. Immunoprecipitated proteins were loaded onto SDS-PAGE gel and Western blot was performed for the presence of O-GlcNAc.
O-GlcNAcase Activity Assay-OGA assay was performed as described (39). Briefly, ovary protein lysates were prepared in T-PER buffer. 100 g of WT or oga del.1 lysate were added to a mixture of 200 M fluorescein di(N-acetyl-␤-D-glucosaminide) (FDGlcNAc) and 50 mM N-acetylgalactosamine (GalNAc), in 50 mM citrate/phosphate buffer, pH 6.5. To control for any background fluorescence of the FDGlcNAc substrate itself, T-PER was added to a mixture of FDGlcNAc and GalNAc in the citrate/phosphate buffer ("no lysate"). All reactions were incubated in the dark at 37°C, with shaking at 100 rpm for 30 min. The reactions were quenched by adding Na 2 CO 3 to a final concentration of 400 mM. Fluorescence was measured in 1-s intervals at the excitation wavelength of 485 nm and emission wavelength of 535 nm on a Wallac 1420 fluorometer (PerkinElmer Life Sciences). All assays were performed in triplicate. Student's t test was used for data analysis. The signal detected in the no lysate reactions was averaged, and this value was subtracted from each lysate measurement. Data are presented as the mean Ϯ S.E.
RNA Extraction and RT-PCR Analysis-5-Day-old well fed females were used for this experiment. 15 female ovaries were dissected on ice-cold PBS and RNA was extracted using TRIzol (Life Technologies, Inc.) according to the manufacturer's instructions. cDNA was prepared with Q-Script cDNA master mixture (Quanta). RT-PCR analysis was performed using a Applied Biosystems instrument. Student's t test on selected pairs was used to compare gene expression levels.
Fecundity Assay-6 Newly hatched male and females were placed in embryo collection chambers (Genesee Sci) with apple juice agar plates and fresh yeast paste. Flies were acclimated to the chamber in constant darkness for 2 days and changing the apple juice agar plate every 24 h. The number of eggs on the plates was counted daily for the following 3 days. Experiments were done in triplicate and repeated 3 times. Statistical analysis was done by Student's t test on selected pairs. Data are presented as the mean Ϯ S.E.

Interfering with O-GlcNAc Cycling Alters Chromatin-associated O-GlcNAc Levels and Gene Expression in S2 Cells-The
Pc group consists of a set of proteins required for regulating proper body plan development by repressing the expression of Hox genes through compacting chromatin and making DNA inaccessible to RNA polymerase (40,41). OGT plays a role in Pc repression in Drosophila (6) and is essential for the function of Pc by O-GlcNAcylating Ph (7). Moreover, OGT is a member of the Pc group in Drosophila (6,8) and ogt mutant flies die as pharate adults displaying homeotic transformations (9). The undiscovered genomic regions responsible for the homeotic transformations observed in ogt mutant flies encouraged us to increase our understanding of the genome-wide relationship between O-GlcNAc and Pc repression in Drosophila. For this purpose, we analyzed O-GlcNAc, Pho, and active transcription indicator RNA Pol II Ser2P distribution on chromatin using ChIP-chip tiling arrays in S2 cells that have decreased or increased O-GlcNAc levels (following OGT or OGA RNAi, respectively). We reasoned that altered O-GlcNAc levels would change the distribution of repressed chromatin (to be observed by Pho distribution) and/or actively transcribed (to be observed by RNA Pol II Ser2P distribution) chromatin regions. ChIPchip was used because we found that it better reflects quantitative differences in occupancy than ChIP-Seq. The data we have reported here are available in the GEO database as GSE74846. Analysis of these data suggested that O-GlcNAc and Pho co-occupied many chromatin regions including the Hox gene clusters, whereas RNA Pol II Ser2P was excluded from those same regions (Fig. 1a). After OGA RNAi, O-GlcNAc levels were increased more than 2-fold (Fig. 1b). The number and amplitude of O-GlcNAc peaks on chromatin were also significantly increased (Fig. 1c, red lines) compared with the control sample (Fig. 1c, blue line).
Importantly, all ϳ8000 observable O-GlcNAc peaks were dramatically decreased by OGT RNAi (Fig. 1c, green line). The O-GlcNAc peak intervals associated with most genes showed a substantial increase (1.4-fold average) upon silencing of OGA. Based on the involvement of OGT in Pc repression and the homeotic transformations seen in ogt mutants, we expected to see altered O-GlcNAc occupancy on areas surrounding Hox genes upon disruption of O-GlcNAc cycling by loss of OGA. To our surprise, DAVID clustering analysis showed that the homeotic genes were 23-fold enriched in the small fraction of genes showing little if any change in O-GlcNAc levels after loss of OGA (OGA/WT ratio of 0.5-1.0-fold) (Fig. 1d) (42). Interestingly, DAVID clustering (42) also revealed that cell cyclerelated genes were among those enriched for increased O-GlcNAc peak intensity (Dataset 1, GEO GSE74846) and were among those genes that modestly increased (OGA/WT ratio 1.0 -1.5-fold, enrichment score of 12.55) (Fig. 1, c and d). Fig. 1d also shows that the genes that most dramatically increase following OGA knockdown (1.5-2.25-fold enrichment) are those associated with protein degradation and steroid hormone activation. These data suggest that O-GlcNAc cycling is more dynamic on particular regions of chromatin. Therefore, chromatin response to the loss of OGA is not global, rather gene specific. Further bioinformatics analysis of ChIP-chip data were also performed by "two-sample analysis" as described under "Experimental Procedures." This analysis revealed interesting correlations between the O-GlcNAc and the other ChIP-chip datasets (Table 1, Dataset 1, GEO GSE74846). In agreement with the previously reported O-GlcNAc modification of Ph (6, 7), we found that 93% of Pho-occupied regions are also occupied by O-GlcNAc (Table 1, row 1, column C). Knockdown of OGA resulted in an increase, such that 96% of Pho-occupied regions were co-occupied by O-GlcNAc (Table 1, row 1, column D). As expected, knockdown of OGT decreased Pho and O-GlcNAc co-occupancy to 62% (Table 1,  To analyze the affect of O-GlcNAc and RNA Pol II Ser2P chromatin co-occupancy on transcription, we utilized robust a Affymetrix analysis whole-genome Drosophila microarray fol- The O-GlcNAc signal was normalized to actin loading control. The graph represents average Ϯ S.D. c, the O-GlcNAc peak intensities of each of ϳ8000 regions similar to those shown in a were plotted in increasing order of intensity. Peak intensities increased an average of 1.4-fold in OGA knockdown (red) cells, whereas intensities decreased to near background levels in OGT knockdown (green) cells compared with GFP control RNAi (blue). A 50-gene moving average for OGA and OGT knockdown is highlighted in black, with blue denoting a GFP (WT) control. d, the ratio of peak intensity observed for OGA to WT samples on the Affymetrix Genechip arrays was calculated for each of the roughly 8000 O-GlcNAc occupied chromatin sites. Those intervals that did not change significantly upon OGA knockdown (0.5-1.0) were examined bioinformatically using DAVID functional annotation clustering (42) and found to be 23-fold enriched for Hox genes and other transcriptional regulators (p Ͻ 0.0001). Intervals exhibiting an intermediate response to OGA knockdown were 10 -12-fold enriched in genes associated with cell cycle and ATP utilization (p Ͻ 0.0001). Those showing the greatest change in intensity (1.5-2.5) between OGA and wild type showed a more modest 2-3-fold enrichment for genes associated with ubiquitin and steroid hormone response (p Ͻ 0.001).  lowing OGT or OGA knockdown. We found that the expression of genes related to development, cell cycle, and metabolism were affected (Table 2) when O-GlcNAc levels were perturbed by RNAi. The complete list of deregulated genes can be found in Dataset 2 (GEO GSE74846). We were surprised to find that whereas knockdown of OGT yielded deregulation of 321 unique genes more than 1.5-fold, knockdown of OGA yielded down-regulation of only OGA expression when the same stringency was applied (Table 2). Because RNAi was demonstrated to be effective at silencing both OGA and OGT (see Fig. 1b) this finding suggests that active transcription was more sensitive to loss of OGT than OGA. Upon further examination with reduced stringency, genes that were 1.2-fold up-regulated in OGA knockdown cells with p Ͻ 0.05 included those predicted by DAVID analysis (42) to be important for cell morphogenesis, oogenesis, and female gamete generation, suggesting that development or oogenesis might be more sensitive to OGA in flies. This possibility prompted us to examine the impact of loss of the oga gene in the context of Drosophila embryogenesis and development in the intact animal. Generation of OGA Knock-out Flies-To examine the impact of blocked O-GlcNAc cycling in Drosophila development we generated oga del.1 flies by P-element excision using a P-element insertion line generated in the gene disruption project (43). The P-element was originally inserted ϳ100 bp upstream of the oga ATG start site. Excision removed 657 bp through the oga gene including the promoter, first exon, and a portion of the second exon (Fig. 2a). The deletion removed 171 amino acids at the N terminus of the protein sequence, corresponding to over half of the predicted O-GlcNAcase domain including nearly all of the catalytically important residues (Fig. 2b). In addition, transcript levels were monitored by RT-PCR and reduced to near zero (%) at the deleted locus (data not shown). The rest of the gene is transcribed at levels comparable with WT (data not shown), therefore it is possible that a truncated protein containing an intact C-terminal domain including the pseudo-histone acetyltransferase domain could still be made in oga del.1 mutants. In animals lacking O-GlcNAcase activity, we expected an increase in the levels of O-GlcNAc-modified proteins. Indeed, we observed a significant (ϳ6-fold) increase in protein O-GlcNAcylation for oga del.1 mutant fly extracts compared with heterozygote (oga del.1 /TM6) or wild type (WT) fly extracts (Fig. 2c) as reflected by O-GlcNAc band intensities. UAS-driven OGA RNAi led to a more modest increase in O-GlcNAc levels than the deletion strain (Fig. 2c,  right panels). In control rescue experiments, ectopic expression of UAS-OGA in oga del.1 mutant flies using the actin-Gal4 promoter restored O-GlcNAc levels to WT (w 1118 ) levels (Fig. 2d).
To determine the level of O-GlcNAcase enzyme activity in oga del.1 flies, we analyzed ovary protein extracts for in vitro OGA activity. oga del.1 mutant ovary protein extracts displayed very little O-GlcNAcase-specific enzyme activity when compared with WT ( Fig. 2e) with residual activity attributed to additional hexosaminidases and chitinases present in the extract (39). Thus, we conclude that oga del.1 animals lack significant O-GlcNAcase activity. Phenotypically, homozygous oga del.1 mutants were viable and fertile. The mutants showed a semi-penetrant oogenesis defect that was rescued by overexpressing OGA under actin-Gal4 or ovary-specific nanos-Gal4 promoter to restore normal cycling of O-GlcNAc (supplemental Fig. S1). However, fertility was also normal when oga del.1 mutants were crossed to two different oga deficiency lines. The deficiency lines are missing at least 30 genes near the "NK" cluster of homeotic genes in Drosophila including oga. There are numerous possible interpretations of our findings. First, the dosage of truncated OGA protein or mRNA may be important for the defective oogenesis phenotype. Second, the dosage of another gene that is missing in the deficiency line may be important for the oogenesis defect in oga del.1 mutants. We are currently testing these and other possibilities.

Polytene Chromosomes Reveal Genome-wide Distribution of Chromatin Modifiers in O-GlcNAc Cycling Mutants-Polytene chromosomes allow visualization and global analysis of chromatin-associated factors in
Drosophila. This allowed us to extend our ChIP-Chip findings on S2 cells to chromosomes derived from genetically defined Drosophila larvae. We first

TABLE 2 DAVID clustering analysis of gene expression changes upon OGT and OGA RNAi
The change in gene expression for OGT RNAi-or OGA RNAi-treated Drosophila S2 cells compared to control (GFP RNAi) was performed. The genes that increased or decreased more than 1.5-fold were analyzed by the DAVID Functional Annotation tool. The expression of nearly 300 genes was changed upon OGT RNAi treatment and the most enriched pathways are noted. To examine the changes in gene expression upon OGA RNAi treatment, we lowered the threshold for gene expression to 1.2-fold, and saw an increase in pathways depicted in the table. No gene other than OGA was downregulated. All clustering analysis yielded a p value Ͻ 0.05 for each annotation and enrichment score noted.   (Fig. 3). We then colocalized these O-GlcNAc bands with other relevant markers of transcriptional activity. As shown in Fig. 4, we found O-GlcNAc was present at a limited number of discrete sites on WT polytene chromosomes compared with elongating RNA Pol II Ser2P, which was much more abundant (Fig. 4, left panels,  w 1118 ). The number and intensity of O-GlcNAc positive bands increased in oga del.1 mutant polytenes and the overlap with RNA Pol II Ser2P also was demonstrably increased (Fig, 4, middle panels, oga del.1 ). In contrast to the findings with RNA Pol II Ser2P, the major O-GlcNAc sites co-localized with Pc (Fig. 4, middle panel,  w 1118 ) and Pho (supplemental Fig. S2) in WT flies confirmed the high degree of overlap observed in ChIP-chip results in S2 cells ( Fig. 1 and Table 1). In the oga del.1 strain many more O-GlcNAc bands appeared that were not coincident with Pc (Fig. 4, right panel, oga del.1 ). This suggests that O-GlcNAc cycling normally occur at those sites.
These findings, and our previously described ChIP-chip analysis showed that many more O-GlcNAc occupied regions were free from Pho and coincident with RNA Pol II Ser2P upon OGA knockdown. The presence of O-GlcNAc at sites of active transcriptional elongation prompted us to question whether Pho-independent O-GlcNAc-enriched chromatin regions were co-occupied by Trithorax or Compass group members. Drosophila TRX, ASH1, and SET1 are all capable of methylating histone 3 (H3) thereby opening chromatin to allow for active transcription. TRX monomethylates H3K4, whereas SET1 and ASH1 are responsible for trimethylation of H3K4 and dimethylation of H3K36, respectively (44). After observing the dramatically increased chromatin occupancy of O-GlcNAc on oga del.1 mutant polytenes in comparison to WT (Fig. 4), we co-stained polytene chromosomes of oga del.1 mutants with O-GlcNAc and antibodies against each of the Trithorax members: TRX, ASH1, or Compass member SET1. In WT, we observed that a limited number of bands co-stained with O-GlcNAc and each of these proteins (Fig. 5). Strikingly, we observed more bands corresponding to each of these proteins on polytene chromosomes of oga del.1 mutants. Despite the increases in O-GlcNAc at sites associated with these proteins in the oga del.1 mutants, not all the bands stained with TRX, ASH1, or SET1 were co-stained with O-GlcNAc. ASH1, SET1, and TRX Histone Methyltransferases Are O-GlcNAc Modified-Direct O-GlcNAc modification of TRX, ASH1, and SET1 could potentially alter their stability, activity, or chromatin binding efficiency, which could lead to inappropriate gene expression in oga del.1 flies. To examine whether these Trithorax group and Compass proteins were modified directly by O-GlcNAc, we immunoprecipitated each from ovary extracts and probed for O-GlcNAc. All three proteins showed detectable levels of O-GlcNAc modification in oga del.1 ovary extracts, although none of them were detectable in WT. TRX has been shown to exist in as many isoforms that are separable on SDS-PAGE (45). We saw O-GlcNAcylation of these same isoforms for TRX following immunoprecipitation (Fig. 6). Importantly, BRE1, which is thought to modulate SET1 activity (46), did not show an increase in O-GlcNAc modification under the same IP conditions. These findings suggest that the O-GlcNAc modification of TRX, ASH1, and SET1 is normally at substoichiometric levels and increases when O-GlcNAcase activity is eliminated (Fig. 6). Furthermore, our results underscore that the oga del.1 animals generated herein provide an excellent tool to: 1) identify proteins that are O-GlcNAc modified and 2) examine the consequences of the post-translational modification on protein stability and function in vivo.
We next postulated that the irreversible O-GlcNAcylation of these three histone methyltransferases could change their activity thereby interfering with the proper spatiotemporal expression of their target genes. To test this theory, we examined H3 modifications in oga del.1 ovaries by immunoblotting and noted no significant changes in the H3 methylations analyzed (Fig. 7a). We suggest that subtle chromatin methylation differences between WT and oga del.1 would be hard to detect in whole ovary extracts because the ovary contains many different cell types including stem, nurse, oocyte, somatic cells, all of which are constantly developing to produce the mature egg. To examine possible cell type-specific H3 modification differences that might be present, we generated germline chimeras (Fig. 7b) using a FRT/FLP mosaic system (36), in which we were able to detect individual cells as oga del.1 mutant (GFP negative) or WT/heterozygote (GFP positive). Quantitation of the levels of H3 methylation in the germline stem cells show a slight decrease, which did not reach statistical significance between the oga del.1 and adjacent wild type or heterozygous cells. We also observed no difference in H3 methylation in somatic fol- licular cells between these genotypes. These findings are reminiscent of ogt mutants, which also did not show any change in H3K27me3 levels despite the fact that OGT is essential for Ph function and proper development (6,7).
Gene Expression Is Deregulated in oga del.1 Ovaries-Detection of ASH1, TRX, and SET1 as new O-GlcNAc targets raised the possibility of gene expression changes that may occur as a result of irreversible O-GlcNAcylation during development and differentiation in a specific tissue or cell type. These cell type-specific changes may not have been detected in S2 cells following OGA RNAi. To address that possibility, we chose a small group of highly O-GlcNAc-enriched genes from the ChIP-chip dataset (GEO database GSE74846, Dataset 1) and analyzed the expression of oogenesis-related (HTS, Capicua, PKA-C1, sxl, bam, and bazooka) and cell cycle-related (bam, bazooka, HCF, embargoed) genes by RT-PCR in ovaries, where there constant development and differentiation occur. Interestingly, the expression of oogenesis-related genes did not change, but cell cycle-related HCF and embargoed genes were up-regulated in oga del.1 mutant ovaries (Fig. 8). Importantly, not all the cell cycle genes we analyzed had differential expression in oga del.1 mutants. These data suggest that the expression of only a subset of genes were affected in oga del.1 mutants.    set (ϳ500 genes) either slightly decrease or stay constant upon OGA depletion (ratio of 0.5-1.0). Hox genes and other transcriptional regulators are highly over-represented in these genes that change little in response to OGA knockdown. These are also the major sites of Polycomb repression in Drosophila. The finding that these chromatin regions are insensitive to OGA loss of function is consistent with a model in which these domains are normally unavailable to OGA because of chromatin structure or the lack of a histone signature to recruit OGA to those regions. Intriguingly, the genes showing most dramatic elevation of O-GlcNAc chromatin occupancy upon OGA depletion were those associated with protein degradation (ubiquitin), and rapid gene activation (steroid hormone response).

Dynamic O-GlcNAc Cycling Is Highly Regulated and Occurs at All But a Small Subset of the Thousands of Genomic Sites in
OGA Loss of Function Differs from OGT Loss of Function in Both Phenotype and Gene Expression Changes-Whereas we expected to see a significant number of gene expression changes by whole genome microarray based on the overlap between O-GlcNAc and RNA Pol II Ser2P, we observed only few changes upon OGA knockdown in S2 cells. The knockdown of OGA was demonstrated to be highly efficient (Fig. 1) although we cannot exclude the possibility that it did not reach the threshold required for major changes in gene expression. The knockdown did achieve elevation of total O-GlcNAc levels and increase in O-GlcNAc occupancy at sites across the genome. The minor changes in gene expression after OGA knockdown raises the possibility that a subset of genes may be deregulated in a specialized cell type, or in a tissue that is sensitive to changes in O-GlcNAc levels. Alternatively, redundant bypass mechanisms may act to limit the impact on transcription. It should also be noted that loss of OGA had a more modest effect on transcription than loss of OGT in our previous analysis in C. elegans (19).
Both Ogt (49) and Oga (50,51) are essential for normal development in mice, and ogt is required for proper body development in Drosophila (9). We hypothesized that an oga knock-out O-GlcNAc Cycling Occurs at Sites Associated with Polycomb and Trithorax Complexes-Our phenotypic analysis suggested that Pc repression is not disturbed in oga mutants. Polytene chromosomes isolated from WT, ogt(sxc), and oga mutants were used to further examine the association of O-GlcNAc with polycomb and other chromatin modifiers. Polytenes from WT flies exhibited a banding pattern largely coincident with Pc and Pho; no O-GlcNAc-specific bands were detectable in polytenes prepared from ogt(sxc). oga deletion flies showed a polytene banding pattern with many more bands with higher overall intensity. These new bands were in addition to those seen in wild type. These new bands appearing on polytenes derived from oga deletion flies were shown to significantly overlap with members of the Trithorax group (TRX, ASH1) and Compass group (SET1) of transcriptional activators. Interestingly, only a subset of the sites stained positive for TRX, ASH1, or SET1 were also stained positive for O-GlcNAc, suggesting that a subset of their target genes were sensitive to O-GlcNAc levels. We then examined whether epigenetic activators in the TRX or Compass complexes were O-GlcNAc modified. We selected ovary tissue for use in our experiments as Drosophila ovaries have a high number of germ cells that are constantly differentiating from a stem cell to a mature egg making them ideal to study epigenetic machinery. Indeed, we found that TRX, ASH1, and SET1 are O-GlcNAcylated in oga del.1 ovaries. However, there were no global changes in the levels of histone modifications these enzymes perform. We then wondered whether this would result in any gene expression changes and selected to analyze a small set of genes whose promoters were highly O-GlcNAc modified in ChIP-chip analysis of S2 cells. Of the eight genes we analyzed, only the expression of HCF and embargoed were changed in oga del.1 mutant ovaries supporting our hypothesis that the expression of a specific subset of genes in a specific cell or tissue type are likely to be affected by irreversible O-GlcNAc modification of TRX, ASH1, or SET1.
HCF, which is necessary for proper cell cycle progression, interacts with both Pc and Trx groups in Drosophila (52). Moreover, SET1 activity is regulated by HCF1 in mammals and HCF1 activity is regulated by O-GlcNAcylation (14,27,53). In light of our findings, we speculate that irreversible O-GlcNAcylation of SET1 and increased expression of HCF could alter cell cycle progression in a specific cell type in oga del.1 mutants.
Regulation of gene expression by Pc group is different in male and female germlines in Drosophila (2). Recent work identified that PRC2 appears to control oogenesis by regulating the expression of cell cycle genes, whereas PRC1 members control sperm development (2). Similar to sex-specific germline regulation identified with the Pc complexes, TRX, ASH1, and SET1 could also display gender-specific phenotypes. Some evidence suggests that such regulation exists. For example, knockdown of SET1 causes an oogenesis defect, but does not affect neuronal development (54). It will be interesting to see whether O-GlcNAc plays a role in gender-specific or tissue-specific gene expression. Support for this idea comes from a study showing that OGT plays a critical role in neonatal epigenetic programming (55).
The O-GlcNAcase Mutant Fly as a Model for O-GlcNAc Cycling-The mammalian and fly OGA molecules show high sequence similarity and are likely to perform similar functions. Mammalian OGA (MGEA5) is suggested to play an O-GlcNAc independent role in activation of gene expression in that the pseudo-histone acetyltransferase domain of MGEA5 may play a role in gene activation (56). The deletion in oga del.1 mutants spans the O-GlcNAcase catalytic domain but leaves the C-terminal pseudo-histone acetyltransferase domain intact. The fly system therefore should provide a valuable platform to examine the roles of the domains of OGA in performing its many functions.
Our study introduces viable oga del.1 mutant flies as a valuable tool to study in vivo effects of increased O-GlcNAc levels. These mutants have greatly enhanced levels of chromatin-associated O-GlcNAc. Previous work has shown that RNA Pol II and other chromatin factors are substrates for O-GlcNAc on chromatin (15,19,40). Here we have shown that increased O-GlcNAc levels correlate with impairment of epigenetic modifications. We show that Trx members ASH1 and TRX, and Compass member SET1 histone methyltransferases are O-GlcNAc modified in ovaries and speculate that their stability or functions may be altered when they are irreversibly O-GlcNAcylated, which could ultimately change the expression of a subset of their target genes. The large number of potential O-GlcNAc targets on chromatin, and their increased modification upon interfering with O-GlcNAc cycling suggests a more general role for O-GlcNAcylation in stabilizing and activating epigenetic effectors.