Dual functionality of O-GlcNAc transferase is required for Drosophila development

Post-translational modification of intracellular proteins with O-linked N-acetylglucosamine (O-GlcNAc) catalysed by O-GlcNAc transferase (OGT) has been linked to regulation of diverse cellular functions. OGT possesses a C-terminal glycosyltransferase catalytic domain and N-terminal tetratricopeptide repeats that are implicated in protein–protein interactions. Drosophila OGT (DmOGT) is encoded by super sex combs (sxc), mutants of which are pupal lethal. However, it is not clear if this phenotype is caused by reduction of O-GlcNAcylation. Here we use a genetic approach to demonstrate that post-pupal Drosophila development can proceed with negligible OGT catalysis, while early embryonic development is OGT activity-dependent. Structural and enzymatic comparison between human OGT (hOGT) and DmOGT informed the rational design of DmOGT point mutants with a range of reduced catalytic activities. Strikingly, a severely hypomorphic OGT mutant complements sxc pupal lethality. However, the hypomorphic OGT mutant-rescued progeny do not produce F2 adults, because a set of Hox genes is de-repressed in F2 embryos, resulting in homeotic phenotypes. Thus, OGT catalytic activity is required up to late pupal stages, while further development proceeds with severely reduced OGT activity.


Introduction
Post-translational modification of more than 1000 proteins with O-linked N-acetylglucosamine (O-GlcNAc) has been shown to affect a diverse array of cellular functions in metazoa, including protein stability, intracellular localization, protein-protein interaction, phosphorylation and ubiquitylation [1][2][3]. The enzyme that catalyses the addition of a single GlcNAc onto serine/threonine residues on intracellular proteins is O-GlcNAc transferase (OGT). OGT is highly conserved in metazoa, consisting of a C-terminal glycosyltransferase domain and N-terminal tetratricopeptide repeats (TPRs). Recent studies on the structure of the catalytic domain of human OGT (hOGT) have provided an insight into the mechanism of catalysis and protein substrate recognition [4][5][6]. TPRs are 34-amino acid helical protein-protein interaction motifs that are typically clustered in 3-16 repeats. TPR motif-containing proteins participate in a diverse array of functions, often as part of multi-protein complexes [7]. In hOGT, the TPRs adopt a right-hand super-helical conformation protruding away from the catalytic domain [5,8]. This TPR super-helix creates a large surface area that is thought to allow OGT to interact with a variety of protein substrates. Most OGT interactors require its TPRs for binding, for instance GRIF-1, OIP106 [9], Sin3A [10] and TET2 [11], while some others interact with the C-terminal domain as in the case of MAPK [12]. OGT isoforms possessing varying numbers of TPRs or recombinant OGT lacking the full complement of TPRs have distinct substrate preferences [13,14]. However, there is no evidence implicating the involvement of OGT in an interaction that does not ultimately invoke its catalytic activity. Nevertheless, it is plausible that proteins interacting only with the most N-terminal TPRs of OGT are non-substrate interactors.
Most animal genomes appear to contain a single ogt gene, except zebrafish, which has two [15]. In several animal models, it has been shown that ogt is essential for embryogenesis and early development. ogt null mice are embryonic lethal, while tissue-specific ogt knock out in T cells, fibroblasts and neurons results in severe phenotypic abnormalities and perinatal death [16,17]. ogt knockdown in zebrafish and Xenopus laevis embryos produces severe growth defects, shortened body axis and retarded nervous system development [18,19]. In Drosophila, ogt is known as super sex combs (sxc), mutants of which do not develop beyond the pharate adult stage. sxc belongs to the Polycomb group (PcG) genes that play key roles in developmental regulation, stem-cell maintenance and genomic imprinting [20][21][22]. In Caenorhabditis elegans, animals homozygous for a partial deletion of ogt-1 are viable and fertile [23]. In Arabidopsis, unlike animal genomes, there are two genes, SPINDLY (SPY) and SECRET AGENT (SEC) coding for OGT with both overlapping and distinct functions. Double mutants of these OGT genes are embryonic lethal [24].
Many of the above studies used genetic approaches designed to generate organisms that are OGT protein null. Transheterozygotic Drosophila larvae with an sxc 1 allele have been reported to possess low levels of expression of a truncated form (lacking the C-terminal 165 amino acids) of OGT [20]. Another sxc allele, sxc 6 , is a protein null mutant. Larvae carrying two other sxc alleles, sxc 4 and sxc 5 , also express mutant OGTs at levels comparable with the wild-type [20,22]. These alleles carry a point mutation (N948I; sxc 4 ) or a 19-amino acid C terminal deletion (D1031-1059; sxc 5 ). The phenotypes of the sxc 4 and sxc 5 alleles are as severe as the null mutants, suggesting a role for the OGT catalytic domain in Drosophila development. While the catalytic activity of DmOGT D1031 -1059 has not been assessed, larval lysate from sxc 4 /sxc 4 possesses catalytic activity comparable with that of sxc 2637 /sxc 2637 , an OGT null mutant. Given the lack of mutant OGTs with impaired catalytic activity, phenotypes associated with partial loss of O-GlcNAc levels have not been explored.
Here, we use Drosophila melanogaster as a model organism to investigate the dependence of developmental pathways on O-GlcNAcylation. The structural and catalytic similarities between human OGT (hOGT) and DmOGT (DmOGT) were investigated using protein crystallography and enzymology. Structure-guided mutagenesis led to the identification of OGT mutants with varying levels of catalytic activity leading to generation of transgenic flies expressing catalytically impaired DmOGTs. Strikingly, pupal lethality observed in sxc mutants was rescued by overexpressing either the OGT WT or the catalytically inactive OGT D955A mutant. However, the sxc F1 progeny rescued with OGT D955A do not produce any F2 adults. F2 embryos from OGT D955A rescue display derepression of a subset of Hox genes. These experiments reveal that OGT activity is required for development to pupal stages, while a severely hypomorphic form of OGT is sufficient to support later developmental processes dependent on zygotic sxc products.

DmOGT is structurally similar to hOGT
Sequence similarity between hOGT and DmOGT is high, with 90% and approximately 80% sequence identity in the TPR region and the catalytic domain, respectively. The most variable domain is the 'intervening domain', a 100-amino acid insertion between the two Rossmann-folds that constitute the catalytic domain, the sequence identity being only 39%. No function has been attributed to the intervening domain despite structural information being available [5]. Additionally, the patch of basic residues (contained within hOGT 958-1001) proposed to interact with phosphatidylinositol-(3,4,5)-trisphosphate (PIP 3 ) is not conserved in DmOGT.
To determine how differences in sequence influence the overall structure of DmOGT and to provide a template for structure-guided mutagenesis, an N-terminally truncated construct starting at amino acid 353 in TPR 10 (D1-352) carrying a mutation (K872M) of a key catalytic lysine was expressed in Escherichia coli and crystallized in complex with the inhibitor/ substrate analogue UDP-5S-GlcNAc. The structure was solved by molecular replacement and refined against 2.7 Å synchrotron diffraction data (table 1), yielding clear electron density for UDP-5S-GlcNAc ( figure 1a,b). It appears that DmOGT adopts the canonical OGT fold with the bilobal arrangement of two Rossmann-like domains characteristic of the GT-B superfamily of glycosyltransferases, as well as the additional TPR-like helices (535-566) in the N-terminal of the catalytic domain, which lead into the TPR domain (figure 1a). As a result, the TPRs are in close association with the glycosyltransferase domain and the catalytic site is aligned with the channel along the main axis of the TPR superhelix. Superposition of the catalytic domain of DmOGT with the reported hOGT structure (PDBID 3PE4 (5)) highlights the structural similarities  rsob.royalsocietypublishing.org Open Biol. 5: 150234 loops were disordered, but the overall domain architecture mirrors that of the hOGT intervening domain (figure 1c).

Rational design of catalytically impaired DmOGT mutants
A recent study investigating the substrate sequence requirements of hOGT identified a peptide derived from human retinoblastoma-like protein 2 (RBL2; KENSPCVTPVSTA) as the best peptide substrate among the 720 Ser/Thr containing peptides [25]. RBL2 is a retinoblastoma family member-a family of proteins that are involved in a variety of cellular processes including the cell cycle, cell differentiation and apoptosis [26]. In order to explore the mechanistic similarities between the hOGT and DmOGT, Michaelis-Menten kinetic parameters of the D1-352 DmOGT were determined with the RBL2 peptide as an in vitro acceptor substrate in a radiometric assay. The Michaelis constant (K m ) of UDP-GlcNAc for D1-352 DmOGT is 17.8 + 0.7 mM, similar to the K m of the donor for D1-312 hOGT (6.6 + 0.3 mM, figure 1d). A substrate-assisted glycosyltransfer mechanism has been proposed for hOGT, wherein a non-bridging oxygen of the a-phosphate of UDP-GlcNAc serves as the catalytic base [6]. This unique catalytic mechanism explains the specific inhibition of hOGT with the thiosugar derivative of the donor substrate, UDP-5S-GlcNAc [27]. Activity of the D1-352 DmOGT is inhibited by UDP-5S-GlcNAc with an inhibition constant (K i ) of 36.2 mM, comparable with the K i of D1-313 hOGT at 13.6 mM (figure 1e). This observation further supports the hypothesis that DmOGT adopts the same catalytic mechanism as hOGT, corroborating the structural data and Michaelis-Menten kinetics.
To identify OGT mutants with a range of catalytic activities, four DmOGT point mutants were designed based on the structure of DmOGT and previous enzymatic studies on hOGT [6,28,29]. The residue Asp925 in hOGT (DmOGT D955 ) interacts with the ribose of the donor substrate (figure 1b) and an alanine mutant (D925A) has been shown to abolish hOGT enzymatic activity, primarily by disrupting UDP-GlcNAc binding [29]. Two other point mutants of hOGT, H558F (DmOGT H596F ) and K842M (DmOGT K872M ), have been reported to be enzymatically inactive while retaining the ability to bind the donor substrate UDP-GlcNAc with similar affinity as the wild-type enzyme [6]. In addition, H498A (DmOGT H537A ), a mutant of a residue that was previously thought to be important for catalysis, was chosen [5,6]. The D1-352 forms of DmOGT WT and the DmOGT point mutants were expressed and purified. Enzymatic activities of these proteins were determined by a radiometric assay using the RBL2 peptide as acceptor substrate. D1-352 DmOGT K872M and D1-352 DmOGT D955A appear to have no catalytic activity even upon extending the assay to 12 h (figure 1f,g). However, D1-352 DmOGT H537A and D1-352 DmOGT H596F exhibited 5.6% and 3.0% activity, respectively, relative to D1-352 DmOGT WT , and thus retained some degree of catalytic activity (figure 1f,g).

OGT catalytic activity is essential for early embryonic development in Drosophila
It has previously been demonstrated that transgenic expression of wild-type DmOGT in sxc transheterozygotes rescues their lethality at the pharate adult stage [22]. Using this readout, the developmental requirements of catalytic versus non-catalytic functions of OGT were dissected (figure 2a). Rescue of the sxc 1 /sxc 6 pupal lethality was performed by ubiquitously driving the full-length DmOGT WT or one of the DmOGT point mutants (H537A, H596F, K872M or D955A) with a tubulin:: GAL4 driver. The details of all the genotypes obtained in these rescue experiments are outlined in table 2. The rescue measured with full-length DmOGT WT was in agreement with the previous report that also used a tubulin::GAL4 driver [22]. In the absence of this rescue, as is the case with control crosses lacking either the driver or the transgene, no adult sxc 1 /sxc 6 transheterozygotes were recovered. Of all the flies scored from the rescue cross, the fraction of sxc 1 /sxc 6 transheterozygote adults recovered on driving the DmOGT WT form of OGT ubiquitously was 18.6% (figure 2a and table 2). Given the crossing scheme, on complete rescue of the lethality phenotype, the rescued flies would constitute 20% of total progeny, which is in agreement with the level of DmOGT WT rescue observed. The fraction of rescued sxc 1 /sxc 6 (table 3). To test whether the lack of F2 adults from the DmOGT D955A rescued sxc 1 /sxc 6 transheterozygous parents was a result of their infertility, rescued males/virgin females were crossed to wild-type virgins/males, respectively. While crosses using DmOGT D955A rescued sxc 1 /sxc 6 transheterozygote males produced adult progeny, the crosses with DmOGT D955A rescued sxc 1 /sxc 6 transheterozygote females only produced a few larvae that eventually died (table 3)  transheterozygotes rescued using DmOGT WT , DmOGT H537A or DmOGT H596F were comparable with that of the F1 rescues (figure 2j). To determine the specificity of the signals obtained using the O-GlcNAc antibody (RL2), total WT fly lysates were incubated with CpOGA, a potent enzyme that is known to remove O-GlcNAc from proteins [30,31]. CpOGA treatment revealed that the prominent 100 kDa and lower molecular weight proteins are non-specific signals detected in the assay (electronic supplementary material, figure S1A). In addition, competition with 0.5 M GlcNAc during primary antibody incubation confirmed the specificity of RL2 reactivity (electronic supplementary material, figure S1B). These control experiments confirm that the O-GlcNAc-specific reactivity of RL2 is to proteins greater than 60 kDa. sxc has shown to be involved in Polycomb dependent derepression of Hox genes [32]. As observations can be confounded by the presence of maternal OGT products in the F1 embryos or larvae, phenotypes were assessed in F2 embryos. The F2 embryos are sxc 1 homozygotes, sxc 6 homozygotes or sxc 1 /sxc 6 transheterozygotes. Nevertheless, it is not possible to determine the precise sxc genotype on the second chromosome in these embryos. By immunostaining for the HA tag, the embryos expressing the respective OGT transgene can be identified and all experiments in rescued F2 embryos described further are in embryos expressing the respective DmOGT transgene. O-GlcNAc levels were assessed in F2 embryos using the RL2 antibody. While O-GlcNAc levels in DmOGT WT rescued embryos were comparable with those in w1118 embryos, faint immunostaining in cells overexpressing DmOGT H537A was observed (figure 3a-c). However, no O-GlcNAc staining could be observed in embryos rescued with either DmOGT H596F or DmOGT D955A (figure 3d,e). As sxc mutants display phenotypes similar to PcG genes that are involved in Hox gene repression, expression patterns of the Hox genes, Sex combs reduced (Scr), Ultrabithorax (Ubx) and Abdominal-B (Abd-B) were assessed by immunostaining. Expression patterns of Ubx (figure 3k-o) and Scr (figure 3p-t) remained unchanged in F2 embryos rescued with any of the DmOGT transgenes and were comparable with the wild-type pattern. Interestingly, Abd-B was de-repressed anterior to its normal expression domain in most of the DmOGT-D955A rescued F2 embryos (figure 3j). The Abd-B expression pattern was unaltered in all DmOGT WT and most of the DmOGT H537A and DmOGT H596F rescued embryos (figure 3f-i).

Discussion
The role of sxc as a PcG gene that functions in homeotic gene repression in Drosophila has been established [21]. The homeotic transformation and lethality phenotypes of sxc have been ascribed to the catalytic glycosyltransferase activity of OGT [20,22,32], but OGT is a large multi-domain protein known to participate in numerous protein-protein interactions [9][10][11][12]. We aimed to dissect the requirement of OGT catalytic activity during Drosophila development using a hypomorphic Table 2. Rescue of sxc lethality by OGT point mutants. Crosses were set up with flies of the indicated genotypes and transferred into fresh vials every 3 -4 days. Adults emerging from the crosses were scored for the presence of second and third chromosome balancers/marker, CyO and MKRS or TM6. Flies that did not possess any of the balancers/markers (þ;þ) were the rescued sxc 1 /sxc 6 transheterozygotes. Control crosses with flies lacking either the driver (tubulin::GAL4) or any of the OGT transgenes do not yield any non-CyO adults. n.a., not applicable.  Table 3. Maternal requirement of OGT catalytic activity. Crosses were set up using rescued F1 flies of the indicated genotypes and scored for the presence (þ) or absence (2) of F2 adults or larvae. Wild-type males or females were also crossed with OGT D955A rescued flies females or males, respectively to assess fertility of the F1 adults. transgenic approach. The structure of DmOGT was determined by X-ray crystallography, revealing that the overall fold and domain architecture of DmOGT closely mirrors that of hOGT. The degree of flexibility of individual domains of OGT has been the subject of some speculation [5]. Central to this speculation is the assumption that OGT modifies intact, fully folded and predominantly large protein substrates. And yet, in the available structures of hOGT, the active site is not very accessible, being partly occluded by the TPR domain. This has led to the hypothesis of a 'hinge-like movement' between TPRs 12 and 13 that would expose the active site [5]. While such a substrate-induced conformational change cannot be excluded, the current structural evidence shows OGTs of three species (bacterial [29], human and D. melanogaster) adopting very similar conformations with limited access to the active site. In the absence of a complex between OGT and an entire protein, the question of substrate access remains unresolved. The almost complete identity of sequence and structure in the active site of DmOGT and hOGT suggested a similar catalytic mechanism and that any differences in substrate specificity are most probably attributable to regions of lower conservation beyond the active site. The largely unexplored intervening domain has the least structural similarity between hOGT and DmOGT. We also noticed a higher degree of disorder in this part of the structure. If the intervening domain was involved in the recruitment of protein substrates, it would explain the flexibility we observed in the absence of a binding partner. A key observation from the sequence comparison of DmOGT to hOGT is the non-conservation of the C-terminal array of lysines, which in hOGT was proposed to bind PIP 3 and hence mediate translocation of hOGT to the plasma membrane upon insulin stimulation [33]. However, it has been recently reported that hOGT does not bind PIP 3 [5]. The absence of a similar positively charged patch on the surface of DmOGT further weakens the hypothesis that PIP 3 binding in general could regulate the subcellular localization of OGT.
Analysis of DmOGT and hOGT structures guided the design of several OGT point mutants that were found to possess varying degrees of catalytic activities with DmOGT K872M and DmOGT D955A activities being undetectable. These mutations, by virtue of possessing almost no catalytic activity, provided tools to examine the role of OGT catalysis.
The TPR regions of OGT are better conserved than the rest of the protein. In keeping with this motif as a protein -protein interaction domain, there are several examples of OGT substrates interacting with the OGT TPR domain. Nevertheless, there are also examples of OGT catalytic domain-dependent binding of substrates. However, instances of OGT performing a scaffolding function completely independent of its catalytic activity are not known. Ectopic expression of the TPR domain was carried out in Arabidopsis spindly (spy) alleles, where lesions in TPRs are associated with gibberellin signalling defects [34]. However, it was reported that the TPR domain alone could not rescue the spy phenotype. When expressed in a wild-type background, the TPR domain mimics spy phenotypes, implicating that it disrupts interaction, oligomerization or localization of endogenous full-length SPINDLY [34]. Given that TPRs form the most prominent structural feature other than the catalytic core, OGT could also perform a non-enzymatic, scaffolding/adapter function as has been demonstrated for many other TPR proteins [35][36][37][38]. Nevertheless, one or more of the proteins assembling into such complexes could also be OGT substrates, thus complicating the dissection of these two functions. The only non-O-GlcNAc transfer role described for OGT is of aiding proteolysis of HCF1 and is dependent on the presence of the TPR domain as well as O-GlcNAcylation of HCF1 [39,40]. Notably, the key catalytic residue important for O-GlcNAc transfer, K842, is also essential for OGT proteolytic activity against HCF1.
Rescue of sxc pupal lethality is in agreement with a previous study [22] on rescuing with DmOGT WT . While both DmOGT K872M and DmOGT D955A possess no detectable  rsob.royalsocietypublishing.org Open Biol. 5: 150234 enzyme activity in vitro, the sxc phenotype could only be rescued in animals overexpressing DmOGT D955A . Mutant forms of the equivalent residues in hOGT (K842M and D925A) have been shown to be inactive in vitro against peptide substrates [29,39]. Whereas hOGT K842M can bind UDP-GlcNAc with similar affinity as the wild-type, hOGT D925A was proposed to be deficient in donor substrate binding [29]. The difference in the rescue potential of DmOGT K872M and DmOGT D955A may be due to undetectably low activity of DmOGT D955A . It therefore appears that adult Drosophila can survive with extremely low OGT catalytic activity. No F2 adults could be derived from sxc flies rescued with DmOGT D955A outlining the developmental requirement for catalytic activity in the absence of maternal products. However, minimal levels of OGT activity were sufficient to overcome this requirement, as evidenced by F2 progeny produced by DmOGT H596F sxc rescues.
Wing phenotypes in rescued F1 adults correspond to the level of OGT activity of the rescue construct, with reduced activity corresponding to more severe phenotypes. The most prevalent defect observed was for ectopic vein arising from the posterior cross-vein in rescues with almost all the constructs. Interestingly, similar ectopic vein phentotype, arising from the posterior cross-vein, is observed in Drosophila HCF (dHCF) mutants [41]. With dHCF being an OGT substrate, it is possible that sub-optimal O-GlcNAcylation of dHCF leads to the ectopic vein phenotype. Given that more than half of flies rescued even with DmOGT WT exhibit the ectopic vein phenotype, it implies that this phenotype is sensitive to precise levels of O-GlcNAcylation of an OGT substrate, possibly dHCF. Interestingly, other phenotypes observed in dHCF mutants [41] are not replicated in the rescued F1 adults.
Homeotic transformations in sxc mutants and the loss of PcG repression in larval imaginal discs have been demonstrated [20,21]. In Drosophila, components of polycomb repressive complex 1 (PRC1), a PcG complex, were found to be O-GlcNAcylated while the assembly of PRC2 was found to be upstream of O-GlcNAcylation in mouse ES cells [20,42]. More recently, however, it has been demonstrated in MCF-7 cells that the protein stability of EZH2, the histone methyl transferase component of PRC2, was dependent on its O-GlcNAcylation [43]. Nevertheless, recruitment of PRC proteins to DNA is not dependent on O-GlcNAcylation in Drosophila [20]. O-GlcNAcylation of YY1, a PcG protein, is essential for its interaction with retinoblastoma protein and hence transcriptional control by YY1 [44]. Polyhomeotic (Ph), a PcG protein, co-localizes with O-GlcNAcylation in polytene chromosomes and its recruitment to DNA at polycomb response elements is reduced in sxc mutants [20,22]. It has been demonstrated that when a Ser/Thr rich stretch in Ph is deleted, Abd-B is de-repressed, a phenotype also observed in sxc mat,zyg embryos [32]. Interestingly, Ubx, another Hox gene which was also reported to be de-repressed in sxc mat,zyg embryos [32], retains its normal domain of expression in the F2 embryos rescued with any of the DmOGT transgenes in this study ( figure 2k -o). This suggests that different mechanisms are possibly adopted by OGT to repress different Hox genes. Abd-B repression probably requires relatively higher OGT catalytic function as compared to repression of the other Hox genes.
In summary, we sought to determine the requirement for protein O-GlcNAcylation in Drosophila development. Rescue experiments in sxc mutants suggest the requirement of OGT catalytic activity up to early larval development. These experiments also demonstrate an extremely reduced requirement of OGT catalysis in pupal/adult stages. Association of the severity of wing phenotypes with reduced OGT activity implies a role for O-GlcNAcylated substrates in wing development. Embryos lacking maternal OGT catalytic activity, and consequently possessing significantly reduced O-GlcN-Acylated substrates in the embryo, also display de-repression of only a subset of Hox genes. While these results underline the importance of OGT catalysis for Drosophila embryonic and larval development, it is intriguing that further development occurs almost independent of OGT catalytic activity. The possibility that non-catalytic OGT function(s) are essential in the adult remains unexplored.

Expression and purification of DmOGT constructs
DmOGT D1 -352 K872M was expressed in ArcticExpress E. coli and purified as described for hOGT D1 -312 [6]. N-acetyl-Dglucosamine, 30% PEG8000/ethylene glycerol and 0.1 M Tris-Bicine pH 8.5) gave needle-shaped crystals after 1-2 days at 218C. The crystals were flash-frozen without further cryoprotection. Data were collected at the European Synchrotron Radiation Facility, Grenoble, France, on beamline ID29. Crystals belonged to space group P32 and contained 3 molecules per asymmetric unit. The structure was solved by molecular replacement with MOLREP [45] using the A chain of PDB ID 3PE4 and refined using iterative model building and refinement with COOT [46] and REFMAC5 [47] (table 1).

Enzyme kinetics on hOGT and DmOGT
Steady-state kinetics was performed with the acceptor peptide KENSPCVTPVSTA using a radiometric assay. Assays were conducted in a final reaction volume of 20 ml containing 50 nM hOGT (D1-312) or 100 nM DmOGT (D1-352), 0.25 mM RBL2 peptide and UDP-GlcNAc (100 mM, 50 mM, 20 mM, 5 mM, 2 mM, 0.5 mM and 0.05 mM) with final 0.3 Ci mmol 21 UDP-[ 3 H]-GlcNAc as a radioactive tracer, in buffer containing 100 mM Tris-HCl, pH 7.5, 150 mM NaCl and 1 mM DTT. The reaction was stopped by addition of 200 ml of 0.75 M ice-cold phosphoric acid containing a final concentration of 0.5% trifluoroacetic acid (TFA) and 5% acetonitrile (ACN). A C18 column was activated by 200 ml 100% methanol and equilibrated twice with 200 ml of equilibration buffer containing 0.5% TFA and 5% ACN. Samples were passed through the C18 column by centrifugation at 500 r.p.m. for 30 s and columns were washed six times with 400 ml equilibration buffer. Peptide bound to the column was eluted by the addition of 100 ml 100% methanol rsob.royalsocietypublishing.org Open Biol. 5: 150234 followed by evaporation with SpeedVac. Radioactivity was detected by the addition of 4 ml scintillation fluid and signal was read by a scintillation counter. Triplicate data points were fitted to the Michaelis-Menten equation. Experiments to determine the half maximal inhibitory concentration (IC50) for UDP-5S-GlcNAc was performed as described above with UDP-GlcNAc concentrations fixed at the respective K m for hOGT (5 mM) and DmOGT (15 mM). Duplicate data points were fitted to a three-parameter equation for dose-dependent inhibition. The inhibition constants (K i ) of UDP-5S-GlcNAc for both proteins were approximated by the Cheng-Prusoff equation.

Western blotting
For western blots, five anaesthetized adult flies were frozen on dry ice. The frozen flies were homogenized in 50 ml of lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% Triton-X-100, 1 mM GlcNAcstatin C, 5 mM sodium fluoride, 2 mM sodium orthovanadate, 1 mM benzamidine, 0.2 mM PMSF, 5 mM leupeptin and 1 mM DTT), following which an equal volume of three times SDS Laemmli buffer was added. Lysates were then heated for 5 min at 958C, centrifuged at 16 000g for 10 min and supernatants were collected. Protein concentrations were estimated using the 660 nm protein assay (Thermo Scientific). Thirty micrograms of the crude lysate were subjected to SDS-PAGE and transferred onto nitrocellulose membrane before immunoblotting with RL2 (1 : 1000), rabbit anti-OGT (H-300, 1 : 1000, SantaCruz Biotech), mouse anti-HA (1 : 5000, 12CA5) and/or anti-actin (1 : 5000, Sigma) and the respective infrared dye conjugated secondary antibodies (Li-Cor or Life Technologies, 1 : 10 000). To determine the specificity of the O-GlcNAc signal with the RL2 antibody, w 1118 adult fly lysates were treated with CpOGA [30] for 30 min at 308C before processing for western blotting. Alternatively, the RL2 antibody incubations were carried out in the presence of 0.5 M GlcNAc.