O-Linked β-N-Acetylglucosaminyltransferase Substrate Specificity Is Regulated by Myosin Phosphatase Targeting and Other Interacting Proteins*

O-GlcNAc-transferase (OGT) substrate specificity is regulated by transiently interacting proteins. To further examine the regulation of OGT, we have identified 27 putative OGT-interacting proteins through a yeast two-hybrid screen. Two of these proteins, Trak1 (OIP106) and O-GlcNAcase, have been shown previously to interact with and regulate OGT. We demonstrate here that MYPT1 and CARM1 also interact with and target OGT. MYPT1 and CARM1 are substrates of OGT in vitro and in vivo. MYPT1 and CARM1 also function to alter OGT substrate specificity in vitro. Furthermore depletion of MYPT1 in Neuro-2a neuroblastoma cells alters GlcNAcylation of several proteins under basal conditions, suggesting that MYPT1 regulates OGT substrate specificity in vivo.

In metazoans, nuclear and cytoplasmic proteins are posttranslationally modified on serines and threonines with O-GlcNAc 3 in response to different extracellular signals, altering their biochemical and functional properties (for a review, see Ref. 1). The addition of the monosaccharide moiety is catalyzed by O-GlcNAc-transferase (OGT) using UDP-GlcNAc as the nucleotide sugar donor (2), whereas the hydrolysis of the O-GlcNAc residue is catalyzed by O-GlcNAcase (3).
In many respects, protein GlcNAcylation is analogous to Ser/ Thr phosphorylation. However, a primary difference is that, unlike phosphorylation, which is catalyzed by almost 400 Ser/ Thr kinases encoded by the human genome (4), GlcNAcylation is catalyzed by the products of a single human gene (5,6).
Although kinase substrate selectivity is partially genetically encoded, the mechanism of OGT substrate specificity is not clear. More than 600 GlcNAcylated proteins have now been identified (1), yet the question of how OGT is able to distinguish its substrates from other proteins is not well understood.
Using synthetic peptide substrates, OGT does seem to show sequence specificity in vitro. In addition, OGT is responsive to a wide range of UDP-GlcNAc concentrations, and the affinity of OGT for peptide substrates is altered with different UDP-GlcNAc concentrations (7). As a result, it has been suggested that the substrate specificity of OGT in vivo may be regulated by UDP-GlcNAc concentrations, which have been shown to be responsive to the nutritional state of the cell (8 -10). Although this hypothesis does partially explain hyperglycemia-mediated changes in protein GlcNAcylation, it cannot account for changes in GlcNAcylation of specific proteins that are independent of alterations in UDP-GlcNAc concentration.
The crystal structure of an OGT homolog has been solved recently along with a partial structure of OGT, providing insight into its regulation (11,12). The tetratricopeptide repeat domain of OGT forms a large superhelical structure very similar to that of importin ␣ (11,12), suggesting that OGT may interact with a diverse group of proteins through its N terminus (13,14).
In support of this hypothesis, it has been shown that Trak1 (previously known as OIP106) interacts with the tetratricopeptide repeat domain of OGT, recruiting OGT to RNA polymerase II (15). Other studies have since revealed OGT to be regulated by its interacting proteins (16 -18). Recently we have demonstrated that glucose deprivation in Neuro-2a cells increases the OGT-mediated GlcNAcylation of neurofilament H via targeting mediated by activated p38 (19). However, in this case, instead of interacting with the tetratricopeptide repeat domain of OGT, p38 interacts with the C terminus of OGT. Interestingly this same region of OGT has also been shown to interact with phosphatidylinositol 3,4,5-trisphosphate, leading to recruitment of OGT to the plasma membrane during insulin signaling (20).
Thus, it is becoming apparent that recruitment of OGT through its interaction domains may be a key mode of regulating OGT activity and substrate specificity. To further explore this hypothesis, we sought to identify additional OGT-interacting proteins using a conventional yeast two-hybrid screen. Here not only do we identify candidate OGT regulators, but also we demonstrate that MYPT1 and CARM1 can both function to alter OGT substrate specificity.
Plasmids, siRNAs, and Transfections-pJL59-OGT, the yeast expression plasmid encoding full-length rat OGT fused to GAL4 DNA binding domain, was created by standard PCR methods and transformed into the Saccharomyces cerevisiae AH109 (MATa) strain using the polyethylene glycol/lithium acetate procedure.
pGEX-MYPT1, the prokaryotic expression vector encoding full-length chicken MYPT1, was provided as a kind gift from Dr. Anne A. Wooldridge and Dr. Timothy A. J. Haystead (Duke University Medical Center, Durham, NC) (21). The mammalian expression vector pEF-HA-MYPT1 was created by subcloning from pGEX-MYPT1.
pSG5.HA-CARM1 was provided as a kind gift from Dr. Michael R. Stallcup (University of Southern California, Los Angeles, CA) (22). pGEX-CARM1 was created by subcloning from pSG5.HA-CARM1. pEF-HA-CARM1 was created by standard PCR cloning methods.
The prokaryotic expression vector encoding full-length human OGT was a kind gift of Dr. Suzanne Walker (Harvard Medical School, Boston, MA) (23).
Predesigned control non-targeting siRNAs and siRNAs specific for MYPT1 were obtained from Dharmacon. Neuro-2a cells were transfected with plasmid and siRNAs using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. siRNAs were transfected at a final concentration of 50 nM for a total of 48 h.
Recombinant Protein Expression and Purification-ncOGT was expressed and purified as described previously (23). GST was expressed using the pGEX-5x1 plasmid (GE Healthcare) and purified over glutathione-Sepharose (GE Healthcare) according to the manufacturer's instructions.
GST-MYPT1 was expressed and purified as described previously with minor modifications (21). Briefly bacteria were lysed in 25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 75 g/ml hen egg lysozyme with protease inhibitors. The lysate was purified over glutathione-Sepharose, and the eluant was fractionated over a Mono S high pressure liquid chromatography column developed with a linear gradient of NaCl up to 600 mM. The fractions containing GST-MYPT1 were pooled and desalted into 25 mM Tris-HCl, pH 7.5, 150 mM NaCl.
GST-CARM1 was expressed and purified over glutathione-Sepharose according to the manufacturer's instructions with minor modifications. Briefly the bacterial cell pellet was lysed with 10 mM Tris-HCl, pH 8.0, 200 mM NaCl, 1% Triton X-100 with protease inhibitors. The eluant was dialyzed against 10 mM Tris-HCl, pH 8.0, 200 mM NaCl.
Protein Analysis and Antibodies-Neuro-2a cells were washed with ice-cold phosphate-buffered saline, collected using cell scrapers, frozen on dry ice, and stored at Ϫ80°C until lysis. Neuro-2a cell pellets were lysed, immunoprecipitated, and immunoblotted as described previously (19).
Yeast Two-hybrid Screen-The human fetal brain MATCH-MAKER cDNA library fused to the GAL4 activation domain in pACT2 was obtained pretransformed into the S. cerevisiae Y189 (MAT␣) strain (BD Biosciences). The yeast two-hybrid screen was performed according to the manufacturer's instructions.
The AH109 strain containing pJL59-OGT was mated with the Y189 strain pretransformed with the pACT2 cDNA library. Mating efficiency was determined to be within the acceptable limits according to the manufacturer's protocol. Next diploid yeast were assayed for ␤-galactosidase reporter gene expression by filter assay. Yeast colonies positive for ␤-galactosidase expression were selected for growth on SD/ϪHis/ϪLeu/ϪTrp minimal medium (for low stringency selection). Yeast colonies were replica plated onto SD/ϪAde/ϪHis/ϪLeu/ϪTrp/5-bromo-4-chloro-3-indolyl-␣-D-galactopyranoside (X-␣-gal) minimal medium to select for medium (white) and high (blue) stringency clones. Positive clones were retested according to the manufacturer's instructions. Plasmid DNA purified from the clones surviving the high stringency selection that were confirmed by retesting were sequenced. A list of these clones is found in Table 1.
OGT and O-GlcNAcase Assays-OGT assays were performed as described previously with minor modifications (19). 3 g of ncOGT was assayed toward 1 g of GST, GST-MYPT1, or GST-CARM1 in buffer containing 50 mM Tris-HCl, pH 7.5, 1 g of purified bovine serum albumin, 1 unit of calf intestinal alkaline phosphatase, 0.5 Ci of UDP-[ 3 H]GlcNAc. Reactions were performed at room temperature for 1 h, stopped with Laemmli buffer, and separated by SDS-PAGE. The gels were then fixed, stained for protein using Coomassie Brilliant Blue G-250, treated with En 3 Hance autofluorography solution (PerkinElmer Life Sciences), dried, and exposed to film.
For OGT assays toward lysate, rat brain lysate was enriched over Sepharose covalently coupled to MYPT1 or CARM1. Proteins were eluted with 8 M urea, concentrated, and buffer-exchanged into 20 mM Tris-HCl, pH 7.5. 20 g of the enriched lysate was assayed in the above conditions with or without 1 g GST, GST-MYPT1, or GST-CARM1.
OGT activity measurements from whole cell lysate were performed as described previously with minor modifications (19). Cell lysate was concentrated and buffer-exchanged into 50 mM Tris-HCl, pH 7.5, using 30-kDa molecular mass cutoff centrifugal filter devices (Millipore). 140 g of concentrated lysate was assayed for [ 3 H]GlcNAc incorporation from UDP-[ 3 H]GlcNAc into a substrate peptide derived from casein kinase II (1 mM) for 30 min at room temperature. The reactions were purified over C 18 MacroSpin columns (The Nest Group) and counted by scintillation counting. O-GlcNAcase activity was assayed from whole cell lysates as described previously (19).
Statistical Measurements-All experiments were performed on at least three separate occasions (n Ն 3). Error bars represent S.E.
For densitometric lane profiles, the plot profile function in NIH Image v1.63 was used. The log(-fold change in O-GlcNAc over control) was calculated as follows. First, lane profiles were normalized against actin, and then the normalized lane profile for the experimental sample was divided by its paired control sample, resulting in the -fold change over control. The logarithm of the -fold change over the lane profile was taken. Values greater than 0 indicate increases over control, whereas values less than 0 indicate decreases compared with control. The plot in Fig. 5 represents the average of three independent experiments. The normal experimental deviation across control samples was measured to be Ϯ0.05 log(-fold change) units or ϳ12% and indicated on the plot by a gray bar between Ϫ0.05 and ϩ0.05.

RESULTS
Putative OGT-interacting Proteins Were Identified Using a Yeast Two-hybrid Screen-To identify candidate OGT regulatory proteins, we performed a conventional yeast two-hybrid screen of a human fetal brain cDNA library using full-length OGT as bait. Using high stringency reporter selection, we were able to identify 27 putative binding partners for OGT (Table 1). In our screen, two previously known interacting partners for OGT were successfully identified, namely O-GlcNAcase (25) and Trak1 (previously known as OIP106) (15). It is interesting to note that the proteins identified belong to a wide range of functional classes from ion transporting integral membrane proteins to signaling and transcriptional regulators.
To confirm the results of the yeast two-hybrid screen, we performed co-immunoprecipitations from several of the yeast clones isolated (Fig. 1A). As expected, all of the clones tested were confirmed to interact with OGT, whereas the TD1 negative control protein did not interact with OGT (Fig. 1B). Using this assay, we were also able to detect SORBS2 as a false positive result of the screen because the size of the protein failed to match the predicted size based on the cDNA sequence of the clone.
MYPT1 and CARM1 Both Interact with OGT-We then chose two putative OGT interactors from the yeast two-hybrid screen, MYPT1 and CARM1, for further analysis. MYPT1 is a known targeting regulatory subunit of PP1␤ (also known as PP1␦) (for a review, see Ref. 26), whereas CARM1 is a known transcriptional coactivating arginine methyltransferase (for a review, see Ref. 27).
We transfected Neuro-2a cells with plasmids expressing HA, HA-tagged MYPT1, or HA-tagged CARM1 and assayed their ability to interact with endogenously expressed OGT. Indeed both MYPT1 and CARM1 were found to co-immunoprecipitate with OGT (Fig. 2, A and B).
Next we confirmed that MYPT1 interacts with OGT in vivo. Using OGT-specific and MYPT1-specific antibodies, we were able to co-immunoprecipitate endogenous MYPT1 with OGT from rat brain lysate (Fig. 2C).

MYPT1 and CARM1
Are GlcNAcylated-Because both MYPT1 and CARM1 interact with OGT, we decided to determine whether they were substrates of OGT. Toward this end, we transfected Neuro-2a cells with plasmids expressing HA, HA-MYPT1, or HA-CARM1 and assayed their GlcNAcylation using an O-GlcNAc-specific antibody. Both MYPT1 and CARM1 are GlcNAcylated in vivo (Fig. 3, A and B).
To confirm that MYPT1 and CARM1 are OGT substrates, we performed in vitro OGT assays using purified recombinant GST-tagged proteins. Indeed both GST-tagged MYPT1 and GST-tagged CARM1, but not GST, are substrates for OGT in vitro (Fig. 3, C and D). MYPT1 appears to be a better substrate than CARM1 because there was greater [ 3 H]GlcNAc incorpo-  (14 -19). Thus, we sought to determine whether MYPT1 or CARM1 would be able to affect the substrate selectivity of OGT. We first assayed OGT activity toward whole rat brain lysate in the presence of either GST or GST-MYPT1; however, differences were difficult to detect and quantify because of the very large number of OGT substrates (data not shown). Therefore, we undertook an approach to enrich the rat brain lysate for MYPT1-binding proteins by passing the lysate over a MYPT1 affinity column. The proteins bound to this column were eluted and subsequently tested as substrates for OGT in the presence of GST or GST-MYPT1.   Indeed there are several proteins that seem to be better OGT substrates in the presence of GST-MYPT1 (Fig. 4A, arrowheads), suggesting that MYPT1 can alter OGT substrate specificity in vitro. Interestingly GST-MYPT1 itself was a better substrate for OGT in the presence of MYPT1-binding proteins, implying that there are proteins that also increase the affinity of OGT for MYPT1. We also tested to see whether MYPT1 could increase OGT activity toward purified myosin in vitro. MYPT1 had no effect on the GlcNAcylation of myosin in vitro (data not shown).
Next we tested whether CARM1 would function similarly in an in vitro OGT assay. However, it appears that there are only a few CARM1-binding proteins that are better OGT substrates in the presence of GST-CARM1 (Fig. 4B, arrowheads), confirming that CARM1 can alter OGT substrate specificity in vitro.
Interestingly it also appears that CARM1 increases GlcNAcylation of OGT itself.

MYPT1 Affects OGT Substrate Specificity in Vivo-
To determine whether MYPT1 alters OGT substrate selectivity in vivo, we decided to assay the effects of RNA interference-mediated knockdown of MYPT1 on protein GlcNAcylation. Although depletion of MYPT1 in Neuro-2a cells had little effect on OGT, O-GlcNAcase, PP1␤, or phosphorylated threonine levels, protein GlcNAcylation was altered significantly (Fig. 5A). In fact, it appears that depletion of MYPT1 decreased the in vivo GlcNAcylation of many proteins (Fig. 5A, black bars) without affecting the specific catalytic activity of OGT or O-GlcNAcase (Fig. 5, B and C), suggesting that MYPT1 is responsible for maintaining GlcNAcylation of many proteins under basal conditions. In contrast, knockdown of CARM1 had little to no detectable effect on protein GlcNAcylation under basal conditions (data not shown). However, association of OGT with CARM1 does seem to affect CARM1 activity toward histones. 4

DISCUSSION
The ability to selectively GlcNAcylate proteins in response to different conditions is essential for the function of OGT as a signaling molecule. One mechanism for achieving context-dependent substrate specificity is by using a number of bridging or adaptor proteins to recruit OGT directly to its targets (Fig. 6).
Here we describe the identification of putative OGT-interact-4 K. Sakabe and G. W. Hart, manuscript in preparation.  ing proteins and the characterization of MYPT1 and CARM1 as OGT recruitment factors. Although MYPT1 appears to be a key regulator of OGT under basal conditions, CARM1 is likely to regulate OGT under other conditions. Current efforts are focused on determining the cellular state(s) under which CARM1 regulates OGT. CARM1 is an arginine methyltransferase whose substrates include many components of the transcription initiation complex (27). When nuclear hormone receptors are activated, coactivator proteins recruit CARM1 to the site of activation to methylate proteins leading to full transcriptional activation (27). It is possible that under these same conditions CARM1 may recruit OGT to GlcNAcylate components of the transcriptional machinery to affect activation. Although OGT has been shown to regulate the activity of many transcription factors, there appears to be a number of different mechanisms used, making it difficult to generalize a single role for GlcNAcylation in transcription. Nonetheless understanding the context of the recruitment of OGT by CARM1 may help to elucidate the role of OGT and protein GlcNAcylation in the regulation of transcription by certain nuclear hormone receptors at individual promoters.
We are also currently working to identify the GlcNAcylated proteins that are recruited to OGT by MYPT1 and CARM1. The identification of these substrates promises to clarify the functional roles of MYPT1 and CARM1 in regulating OGT. Although myosin subunits are GlcNAcylated (28,29), MYPT1 did not affect the activity of OGT toward myosin. Also although MYPT1 and OGT are known to interact with PP1␤ (30, 31), MYPT1 does not seem to be required for the interaction of OGT with PP1␤ (data not shown). Although the function and sites of phosphorylation on OGT are not known (32), it is possible that MYPT1 may also serve to dephosphorylate OGT under certain conditions. It has been shown that MYPT1 is essential for cellular survival. MYPT1-null mouse embryos die prior to 7.5 days postcoitum, and cells completely lacking MYPT1 have yet to be isolated (33). In fact, it has recently been shown that knockdown of MYPT1 results in slight defects in mitotic progression in HeLa and SW962 cells consistent with a role in regulating mitosis and cytokinesis by regulating Polo-like kinase 1 (34 -36). We have recently shown that disruption of O-GlcNAc cycling by overexpression of O-GlcNAcase or OGT alters mitotic progression and cytokinesis also at least partially via Polo-like kinase 1 (24). Whether or not the mitotic defect seen during knockdown of MYPT1 is also related to altered protein GlcNAcylation is currently under investigation. Although we have shown here that MYPT1 regulates OGT during basal conditions, it is possible that MYPT1 may also regulate OGT during mitosis, recruiting OGT to a different subset of proteins.
The idea that a single protein may be capable of recruiting OGT to different sets of substrates depending on the cellular conditions is an intriguing one consistent with known mechanisms of regulating protein phosphatases. For example, MYPT1 recruits the dephosphorylation of myosin during cell migration (37), yet recruits the dephosphorylation of Polo-like kinase 1 during cell division (34). Identifying cell division machinery, which is targeted to OGT by MYPT1, may be a promising direction for future research.
Furthermore the disruption of the interaction of OGT with its individual binding partners may aid in studying GlcNAcylation of different proteins and signaling pathways, eventually leading to promising protein GlcNAcylation-targeted inhibitors and therapeutics. Here we present a number of candidate OGT-interacting proteins. It will be very interesting to see whether more of these proteins serve functional roles as OGT recruitment factors, helping to define substrate specificity.