NCOAT associates with corepressors
Antibodies were raised against NCOAT, and an OGT antibody wasobtained (Kreppel et al., 1997) to characterize in vivo interactions.Antibodies were affinity purified and checked for specificityby western blot analysis (supplementary data, Figure 1). Althoughraised against recombinant full-length NCOAT purified from Escherichiacoli, two antisera were generated, one of which [-NCOAT(P)]was state specific and did not recognize phosphorylated NCOAT.Multiple forms of NCOAT were detected by both antibodies. Theserepresent various isoforms and modified versions of NCOAT andOGT (Comtesse et al., 2001; Love et al., 2003; Lazarus et al.,2006).
Protein complexes were immunoprecipitated from chinese hamsterovary (CHO) cells using normal rabbit IgG, -OGT, -NCOAT, and-mSin3A antibodies. The immunoprecipitated complexes and whole-celllysate (WCL) were analyzed by western blot probed for OGT, NCOAT,and mSin3A (Figure 1A). -OGT complexes contained NCOAT and mSin3A,-NCOAT complexes contained OGT and mSin3A, and -mSin3A complexescontained NCOAT and OGT. These proteins were not detected inthe control IgG as expected. The data indicate the in vivo associationsof OGT, NCOAT, and mSin3A. Complexes were unaffected by thephosphorylation state of NCOAT (supplemental data, Figure 2).
In vitro associations between NCOAT and mSin3A were not seen(data not shown); therefore, we mapped the regions of OGT andNCOAT required for the observed in vivo association. By in vitropull-down assays using GST–OGT fusion proteins (supplementaldata, Figure 3) and full-length NCOAT, we found that OGT stronglyinteracted with NCOAT through the N-terminus and first six tetratricopeptiderepeat (TPR) elements of OGT (Figure 1B). This region of OGTwas both necessary and sufficient for the interaction. The necessityof the N-terminus is, so far, unique to the OGT interactionwith NCOAT, because all other interactions with OGT only requirethe TPR motifs.
To conversely map the interaction domain (ID) in NCOAT, we incubatedGST–OGT with various length deletion constructs of NCOAT(supplemental data, Figure 2) (Figures 1C and D). The regionin NCOAT encompassing amino acids 404–548 [NCOAT(ID)]was found to be necessary, and amino acids 336–548 [NCOAT(ID+)]were sufficient for the interaction. Interactions across specieswere tested, showing that NCOAT and NCOAT(ID+) from humans,rats, and mice interchangeably interacted with rat OGT (Figure1D). A summary of the association between NCOAT and OGT alongwith their catalytic domains is shown in Figure 1E.
To demonstrate the strength of association between NCOAT andOGT, we compared in vitro pull-downs with transcription factorsHDAC1, N-CoR, SMRT, and p53 to OGT and NCOAT. [35S]methionine-labeledproteins were incubated with equal amounts of GST–OGT,GST–NCOAT, GST–NCOAT(ID+), and GST alone (Figure1F). All three corepressors were able to interact with GST–OGTand GST–NCOAT(ID+) but not with GST alone. HDAC1 showeda stronger interaction with GST–NCOAT than with GST–OGT.SMRT had a stronger interaction with GST–NCOAT(ID+) thaneither full-length proteins. Although p53 can be modified withO-GlcNAc (Shaw et al., 1996), it did not interact directly withany of the proteins. The interaction of NCOAT with itself islikely representative of intramolecular interactions betweenthe termini (data not shown). Notably, the association betweenNCOAT and OGT was the strongest demonstrated by the ratio ofinput to pulled-out protein with equal exposure. Most of thelabeled OGT or NCOAT was pulled out by its opposite, emphasizinga strong and direct association between them.
To test the associations of naturally occurring NCOAT splicevariants (Toleman et al., 2004) with OGT and corepressors, BSC40cells were transfected with tagged NCOAT constructs for thesevariants (Figure 1G). Transfection allowed us to control whichvariant was interacting and the dependence on the OGT/NCOATinteraction for binding with corepressors. Cells transfectedwith both HA–OGT and GST–NCOAT demonstrated strongassociation regardless of the splice variants of NCOAT usedin the GST–NCOAT construct. These variants all containedNCOAT(ID); however, the SD variant does not contain amino acids336–403 found in NCOAT(ID+). When the complex was precipitatedvia endogenous mSin3A, the transfected HA–OGT and GST–NCOATwere detected. Significantly, OGT interacted with endogenousmSin3A (Yang et al., 2002), but the interaction was much moreefficient if NCOAT was present. NCOAT also interacted with endogenousmSin3A despite the lack of direct in vitro association. Whenmore OGT was present, the association was more efficient. Thispreference of binding the NCOAT/OGT complex was especially notablewith the GK variant, which is inactive as an O-GlcNAcase (Tolemanet al., 2004). The need for both NCOAT and OGT proteins, constitutingthe O-GlcNAczyme, for stable complex formation with mSin3A isevidence for the specificity of the interactions and aids incomplete and correct corepressor complex formation.
NCOAT and OGT locate to repressed promoters
It was previously shown that mSin3A associates with repressedpromoters along with a higher level of O-GlcNAc as comparedwith activated promoters (Yang et al., 2002). We performed chromatinimmunoprecipitation (ChIP) assays to identify the recruitmentof both OGT and NCOAT, as a complex, to promoters. Cell lysatesfrom MCF7 cells, and MCF7 cells transfected with HA–OGTand Flag–NCOAT were collected for western blot analysisor ChIP assays (Shang et al., 2000). Western blot analysis showedFlag–NCOAT expression as well as OGT (data not shown)and mSin3A associations (Figure 2A). ChIP assay for cathepsinD (CatD), PS2, and p21 gene promoters were immunoprecipitatedusing the indicated antibodies to endogenous or tagged proteinsshowing association with the target proteins (Figure 2B). Thepromoters of the two estradiol (E2)-responsive genes, CatD andPS2, showed an inverse E2-dependent association with these targets.However, the association with the E2-independent gene, p21,was not significantly effected by E2. These results confirmedearlier experiments showing that, in the absence of estrogen,repressed promoters are hyper-O-GlcNAcylated and that this modificationis removed from these promoters with the addition of E2 (Yanget al., 2002). Furthermore, the enzyme responsible for the additionof this modification (OGT) is resident on the repressed promotersalong with other corepressors. Primers against a downstreamsection of the PS2 gene showed no association with the corepressorsOGT, mSin3A, and HDAC1 and showed a sustained level of O-GlcNAcmodification independent of E2.
As expected, based on corepressor associations, rather thanassociating with the activated promoters, NCOAT was among theproteins on the repressed promoters. It responded to E2 in asimilar fashion to that of OGT, mSin3A, and HDAC1. The controlprimers against a downstream section of the PS2 gene showeda slight increase in associated Flag–NCOAT, but not endogenousNCOAT, with the addition of E2. These observations confirm thatNCOAT specifically associates with repressed promoters ratherthan with activated promoters.
NCOAT activities should reverse the actions of its partners,OGT and HDAC1, but is found in a region of high-protein O-GlcNAcylation,suggesting that NCOAT activity is off during repression. NCOAT,the only known protein with the needed O-GlcNAcase activityto yield the decrease seen on activated promoters, must be specificallyactivated at these promoters in response to E2 and likely acetylatesneighboring histones (Toleman et al., 2004), before its exitfrom, and the activation of, the promoter. The fact that theO-GlcNAcase activity of NCOAT must be suppressed on repressedpromoters implies that for O-GlcNAcylation, there is no futilecycle.
NCOAT can recruit corepressors to repress transcription
Although NCOAT has the requisite enzyme activity for reversingthe hyper-O-GlcNAcylation of proteins seen on repressed promoters,we have provided compelling evidence that NCOAT is itself associatedwith repressed promoters as a component of the O-GlcNAczyme.The question, therefore, remained as to whether NCOAT couldreverse repression by undoing this modification or whether itsdominant feature was to nucleate the corepressor complexes.To determine which property of NCOAT dominated, activation byremoval of O-GlcNAcylation or repression through associationwith corepressors, the protein was fused to the Gal4 DNA-bindingdomain (Gal4-DBD) and cotransfected along with a Gal4–luciferasereporter plasmid (Figure 3A). Putting NCOAT on this syntheticpromoter caused a dose-dependent repression, rather than activation,of the reporter. The amount of Gal4–NCOAT, determinedby western blot analysis, was proportional to the amount ofplasmid transfected, whereas mSin3A as a loading control remainedthe same (Figure 3A, middle panel). The Gal4–GFP fusionhad no significant effect on transcription, whereas the Gal4-DBDwas mildly stimulatory. Gal4–NCOAT was as potent as Gal4–OGTin repression (Figure 3C).
With previous evidence that the activity of the transcriptionalactivator, Sp1, can be down-regulated by O-GlcNAcylation (Rooset al., 1997; Yang et al., 2001, 2002), reporter constructswere tested which contained Sp1-binding sites upstream or downstreamof the Gal4-binding sites. These constructs would allow Sp1to be modified and thus tests a specifically activated promoterrather than just a basal promoter. Again, repression with Gal4–NCOATwas observed regardless of the position of the Sp1-binding sitesrelative to the Gal4 sites (Figure 3B).
This repression by Gal4–NCOAT was additive with OGT (Figure3C). Even when catalytically inactive OGT (1–286) wasexpressed, some additivity remained because this fragment ofOGT could recruit both NCOAT (Figure 1E) and other corepressors,including wild-type OGT, to the promoter (Yang et al., 2002).When OGT was tethered to the DNA using the Gal4-DBD, free NCOATgave additional repression (Figure 3C, lanes 8 and 9) but onlyif NCOAT could interact with OGT (Figure 3C, lanes 10 and 11).A deletion of NCOAT that prevented interaction with OGT despitemaintaining potential enzymatic activities (Toleman et al.,2004) (Figure 3C, lane 12) was unable to repress or activatebasal transcription activity. Thus, the dominant feature ofNCOAT in this assay was transcriptional repression through itsassociation with OGT and repression complexes. The data suggestthat NCOAT is not a classical activator, but rather exists asa member of the repression complex, poised to derepress as needed.
A modified ChIP assay was preformed to confirm that the transcriptionrepression by the O-GlcNAczyme results from the recruitmentof corepressors to the DNA of the reporter, as can be done byOGT alone (Yang et al., 2002). MCF7 cells were transfected witha Gal4–luc reporter (Figure 3A) and plasmids encodingGal4–DBD fused to mSin3A, NCOAT, or OGT proteins. Detectionof the immunoprecipitated DNA was by Southern blot probed forthe presence of the luciferase reporter (Figure 3D). The -Gal4–DBDprecipitated significant amounts of cross-linked DNA in allsamples containing transfected DBD and DBD fusions. A significantdifference in the ability of -N-CoR to precipitate reporterDNA with endogenous N-CoR was only seen when the transfectedfusion proteins, Gal4–mSin3A, Gal4–NCOAT, and Gal4–OGT,were present. These results confirm previous reports of an associationbetween N-CoR and mSin3A (Heinzel et al., 1997; Nagy et al.,1997; Wong and Privalsky, 1998).
The ability of both Gal4–NCOAT and Gal4–OGT to nucleateendogenous corepressors like N-CoR to the DNA explains the observedrepression with both fusion proteins. This result implies thatNCOAT is not a passive transcription activator, but rather thatits enzymes must be specifically activated to reverse repression.
Regulating E2 response with O-GlcNAc
NCOAT(ID+) alone is responsible for the binding of NCOAT toOGT. To test whether the binding is saturable, [35S]-labeledNCOAT(wt) and NCOAT(ID+) were incubated with GST–OGT onbeads in differing ratios where the amount of NCOAT(wt) andGST–OGT remained constant. The inputs show how much proteinis used in the assay and are representative of 1 on the ratiosgiven in the panel. In the presence of an excess of NCOAT(ID+),wild-type binding to GST–OGT was competed away (Figure4A), indicating that NCOAT(ID+) is sufficient to compete NCOAT(wt)from OGT. NCOAT(ID+) has no enzymatic activities, and OGT/NCOATbinding is qualitatively saturable. Therefore, NCOAT(ID+) wouldact as a location-dominant negative by disrupting endogenousNCOAT association into corepressor complexes rather than a classicaldominant negative which would disrupt enzymatic activity.
The O-GlcNAczyme, formed by the interaction between OGT andNCOAT, binds corepressors better than the respective enzymesalone. Furthermore, NCOAT(ID+) binds these proteins even betterthan NCOAT(wt) (Figure 1G). The lack of enzymatic activitiescoupled with the greater interactions with corepressors wouldpredict that addition of NCOAT(ID+) to OGT would lead to a similaror greater repression than NCOAT(wt). This was tested usingthe Gal4–luc assay for transcription control (Figure 4B).The NCOAT(ID+) indeed repressed transcription better than NCOAT(wt).The repression with Gal4–OGT and NCOAT(ID+) is in starkcontrast to the activation seen when NCOAT(ID+) is expressedalong with the Gal4–DBD alone, again highlighting theinteraction between OGT and NCOAT(ID+). This result supportsthe proposed location-dominant negative potential of the NCOAT(ID+)on promoter regulation.
With only one gene for OGT and one gene for NCOAT, knockouts(O’Donnell et al., 2004) and drug-based inhibition arelethal (Liu et al., 2000; Konrad et al., 2002). Furthermore,because the OGT, bound to NCOAT, binds to its accompanying corepressorsbetter than OGT alone, we chose to use a dominant-negative approachto obviate problems inherent with knocking out the expressionof NCOAT. Because the binding of NCOAT to OGT is saturable,this approach was possible. We used the ability of the NCOAT(ID+)to inhibit formation of completely active O-GlcNAczyme complexeswith NCOAT activities to test its role in the regulation ofestrogen-responsive genes. MCF7 cells were transfected withindicated constructs at 85% transfection efficiency in estrogen-freemedia. Cells were treated with E2 for the indicated times, 24h after transfection. Northern blot analysis of the RNA withprobes to CatD, CAD (Khan et al., 2003), and glyseraldehyde-3-phosphatedehydrogenase (GAPDH) was performed (Figure 4C). Overexpressionof NCOAT(wt) and NCOATID did not repress the initiation of transcriptionbut did have an effect on sustained transcription in the presenceof E2. On the contrary, NCOAT(ID+) yielded an almost completeblock of the E2-induced transcription of the CatD and CAD genes.
A naturally occurring NCOAT splice variant was cloned from theGoto-Kakizaki rat [NCOAT(GK)] (Toleman et al., 2004). This particularNCOAT variant is able to interact with OGT (Figure 1G) and hasHAT activity in vitro but lacks O-GlcNAcase activity (Tolemanet al., 2004). Therefore, this variant could be used as a location-dominantnegative in the O-GlcNAczyme complex specific for O-GlcNAcaseactivity. We transfected MCF7 cells with Flag–NCOAT(wt),Flag–NCOAT(GK), or Flag–GFP. Western blot analysisshowed significantly higher levels of expression for the GKvariant over endogenous NCOAT to yield the theoretical location-dominantnegative (Figure 4D). The transfected cells were treated asbefore to perform ChIP assays. To distinguish between nontransfectedand transfected cells, we performed a two-step immunoprecipitation(IP). The cell lysate was incubated with -Flag beads to obtainpromoters with Flag associations. Following elution of the bead-boundcomplexes, a second IP was performed with the indicated antibodies(Figure 4E). Cells transfected with the Flag–NCOAT againshowed the E2-dependent promoter occupancy by Flag–NCOAT,and secondary IP indicated the Flag–NCOAT complexes containedOGT and O-GlcNAc which exited with the addition of E2 (Figure4E, top panel). On the contrary, cells transfected with theGK variant showed stable promoter occupancy in the presenceof E2 for Flag–NCOAT(GK), OGT, and O-GlcNAc (Figure 4E,middle panel). These data indicate that NCOAT(GK), lacking onlyO-GlcNAcase activity, does indeed behave as a location-dominantnegative. Interestingly, unlike NCOAT(wt), the Flag–NCOAT(GK)complexes were not found to be associated with the p21 promoter,and ongoing studies hope to elucidate this difference in promoterregulation. As well, Flag–GFP was not associated withany of the tested promoters (Figure 4E, bottom panel).
O-GlcNAcase activity in the O-GlcNAczyme complex is essential for proper estrogen signaling
GFP and GFP–NCOAT(GK) were expressed in MCF7 cells and
imaged by fluorescent microscopy (Van Tine et al
expression of GFP–NCOAT(GK) (Figure 5
A) shows a predominantly
cytosolic location, the presence of NCOAT(GK) on the DNA (Figure4
E) and the nuclear pattern seen with GFP–NCOAT(GK) imply
sufficient amounts are in the nucleus to regulate gene expression.
To test the role of the O-GlcNAcase activity in the O-GlcNAczyme,a conditional transgenic animal model was used. A TetRE–NCOAT(GK)transgenic mouse line was crossed with an MMTV–rtTA transgenicline to yield mammary-specific expression of the NCOAT(GK) splicevariant in the presence of doxycycline (Dox) (Roh et al., 2001,2005; Xie et al., 1999) (Figure 5B). Prior studies have confirmedthat this system is effective in dominant-negative expressionanalysis (Roh et al., 2001).
A bitransgenic MMTV–rtTA/TetRE–LacZ line was shown
to express ß-Galactosidase in the ductal epithelia
of mammary glands, following Dox induction (data not shown).
The MMTV–rtTA and TetRE–NCOAT(GK) mice were bred
to result in eight female pups, three of which were bitransgenic.
The NCOAT(GK) transgene was activated by administering Dox at
3 weeks of age and treatment continued until the animals were
sacrificed at 8 weeks of age, several weeks beyond onset of
puberty. Northern blot analysis confirmed transgene mRNA expression
C). Whole-mount analysis demonstrated reduced mammary
ductal side branching in comparison with littermate controls
D [1 and 2]). These mice also demonstrated reduced
end buds (Figure 5
D [3 and 4]), and H & E-stained sections
of these glands showed less ducts (Figure 5
D [5 and 6]). Immunohistochemical
examination of mammary ducts demonstrated reduced expression
of CatD (red) in bitransgenic females receiving Dox in comparison
with controls (Figure 5
D [7–12]) consistent with the results
shown in Figure 4
C. Transgenic animals are very complex, making
individual relationships hard to decipher. Although activation
of one E2-dependent gene was significantly inhibited, this inhibition
could have resulted from the nonpromoter-specific effects NCOAT(GK)
could play on hormone signaling (Nadal et al
the absence of O
-GlcNAcase activity in the O
demonstrate an in vivo
consistent with the previous data, blocking hormone-dependent
signaling and E2-induced gene activation.