O-Linked β-N-Acetylglucosamine (O-GlcNAc) Regulates Emerin Binding to Barrier to Autointegration Factor (BAF) in a Chromatin- and Lamin B-enriched “Niche”*

Background: Nuclear membrane protein emerin binding to nuclear intermediate filaments (lamins) and BAF contributes to forming a nuclear “lamina” structure. Results: Emerin is O-GlcNAc-modified at eight sites: two (Ser-53 and Ser-54) influence further O-GlcNAcylation, and one (Ser-173) regulates association with BAF in the chromatin/lamin B “niche.” Conclusion: O-GlcNAc transferase, a nutrient-responsive enzyme, regulates emerin. Significance: Emerin hyper-O-GlcNAcylation may contribute to cardiomyopathy and other conditions. Emerin, a membrane component of nuclear “lamina” networks with lamins and barrier to autointegration factor (BAF), is highly O-GlcNAc-modified (“O-GlcNAcylated”) in mammalian cells. Mass spectrometry analysis revealed eight sites of O-GlcNAcylation, including Ser-53, Ser-54, Ser-87, Ser-171, and Ser-173. Emerin O-GlcNAcylation was reduced ∼50% by S53A or S54A mutation in vitro and in vivo. O-GlcNAcylation was reduced ∼66% by the triple S52A/S53A/S54A mutant, and S173A reduced O-GlcNAcylation of the S52A/S53A/S54A mutant by ∼30%, in vivo. We separated two populations of emerin, A-type lamins and BAF; one population solubilized easily, and the other required sonication and included histones and B-type lamins. Emerin and BAF associated only in histone- and lamin-B-containing fractions. The S173D mutation specifically and selectively reduced GFP-emerin association with BAF by 58% and also increased GFP-emerin hyper-phosphorylation. We conclude that β-N-acetylglucosaminyltransferase, an essential enzyme, controls two regions in emerin. The first region, defined by residues Ser-53 and Ser-54, flanks the LEM domain. O-GlcNAc modification at Ser-173, in the second region, is proposed to promote emerin association with BAF in the chromatin/lamin B “niche.” These results reveal direct control of a conserved LEM domain nuclear lamina component by β-N-acetylglucosaminyltransferase, a nutrient sensor that regulates cell stress responses, mitosis, and epigenetics.

The nuclear envelope includes diverse membrane proteins, some of which mechanically link the cytoskeleton to the nucleoskeleton (1) or have roles in signaling or genome management (2). Among these are proteins encoded by seven human genes that share the ϳ40-residue LAP2 (lamina-associated polypep-tide 2), emerin, MAN1 domain (LEM domain) 2 (3,4). Emerin and most other LEM domain proteins localize predominantly at the nuclear envelope inner membrane. All characterized LEM domain proteins directly bind nuclear intermediate filaments (A-type lamins encoded by LMNA and/or B-type lamins encoded by LMNB1 and LMNB2) and barrier to autointegration factor (BAF) (encoded by BANF1) (3,4). Lamins, BAF, and LEM domain proteins are essential to reassemble nuclear structure after mitosis (5)(6)(7).
Emerin, other LEM domain proteins, and lamins organize and dynamically tether silent chromatin at the nuclear envelope (20 -27). For example, emerin and LAP2␤ each directly bind HDAC3 (histone deacetylase 3), a chromatin-silencing enzyme, and stimulate HDAC3 activity (28,29). At the IgH locus, the regulated tethering of lamina-associated domains of chromosomes is mediated by DNA sequence repeats bound to a transcriptional repressor (cKrox) in complex with HDAC3 and LAP2␤ and is established during mitosis (30). Notably, a pair of LEM domain proteins (emerin and LEM 2) plus lamin B for 24 h and then either harvested for immunoprecipitation or fixed for indirect immunofluorescence staining (see below).
Stable populations of mouse OGT (F/Y) embryonic fibroblasts expressing either GFP or mER-Cre-2A-GFP (provided by Natasha Zachara, Johns Hopkins University School of Medicine) (58) were cultured in Complete Medium. To induce OGT deletion, cells were plated (1 ϫ 10 6 cells/15-cm plate); 6 h later, we added ethanol with or without 4-hydroxytamoxifen (0.5 M final; Sigma) for 24 h and then washed in sterile PBS and cultured 16 h longer in Complete Medium prior to harvest (58).
Insoluble His-tagged pET29b His-emerin 1-220 polypeptides were purified using the Qiaexpressionist protocol for insoluble proteins (Qiagen). After qualitative SDS-PAGE and Coomassie staining, selected fractions were pooled, dialyzed first against 6 M urea (pH 7.4 resuspension buffer) and then against 4 M urea (pH 7.4 resuspension buffer), and stored at 4°C.
Mass Spectrometry-Recombinant purified His-tagged emerin from three separate preparations (50 -100 g each) was reduced (1 h) with DTT (Sigma) at a molar ratio of 20:1 (DTT/cysteine); carbamidomethylated (45 min) with iodoacetamide (Sigma) at a molar ratio of 3:1 (iodoacetamide/DTT); digested with trypsin (Promega), endoproteinase LysC (Roche Applied Science), and/or endoproteinase AspN (Roche Applied Science) (7 h) at a molar ratio of 20:1 (emerin/protease); and then quenched with glacial acetic acid to pH 3-4 for storage at Ϫ35°C. One preparation was digested on beads in 100 mM ammonium bicarbonate, pH 7.9, using LysC and then digested with either AspN or trypsin. A second preparation was digested with LysC on beads in 4 M urea, 0.1 M sodium phosphate, 10 mM Tris-HCl, pH 7.4, and then diluted with 100 mM ammonium bicarbonate to a urea concentration of 500 mM and digested with trypsin. A third preparation was digested in solution in 400 mM urea, 100 mM sodium phosphate, 10 mM Tris-HCl, pH 7.4, with LysC followed by AspN; a separate fraction was trypsin-digested in solution.
Peptides were analyzed by mass spectrometry as described (65) with the following changes. Precolumns were packed with 4 cm of 5-20-m irregular diameter (YMC) or 15-m regular diameter, 12-nm pore size, C 18 (Reprosil), and behind a 1-mm Kasil frit. Analytical columns were packed with 6 -9 cm of 5-m regular diameter, 12-nm pore size, C 18 (YMC or Reprosil) behind a 1-mm Lithisil frit. Typically, samples corresponding to ϳ1 pmol of emerin in Ͻ10 l of 0.1% (v/v) acetic acid were loaded on column for each mass spectrometric analysis. Samples were separated and eluted using reversed-phase HPLC (Agilent 1100) and analyzed online with a modified electron transfer dissociation (ETD)-enabled high resolution linear trap quadrupole-Fourier transform or linear trap quadrupole-Orbitrap mass spectrometer (Thermo Scientific). Mass analyses were completed with one high resolution MS1 scan (50,000 (Fourier transform) or 60,000 (Orbitrap) full width at half-maximum at 400 m/z), followed by 8 -10 collision-activated disso-ciation (CAD) or ETD MS2 scans acquired with the LTQ operating in either the data-dependent or targeting mode. ETD (30 -50-ms reaction time) was performed with azulene as reagent. ETD was employed for all O-GlcNAc site identifications, and ETD targeting mode was employed in site-mapping Ser-53, Ser-54, and Ser-87. Database analysis was performed using The Open Mass Spectrometry Search Algorithm (OMSSA), version 2.1.1, to search (parameters: Ϯ0.01 Da precursor mass tolerance, Ϯ0.35 Da fragment ion mass tolerance) against the human emerin sequence (NCBI accession number NP_000108). Either "Trypsin" or "No enzyme" were used as protease options, with up to three missed cleavages allowed when applicable. Specified variable modifications were carbamidomethylation of Cys, oxidation of Met, and O-GlcNAcylation of Ser and Thr residues. Signals corresponding to charge-reduced species were removed from the spectra by OMSSA before the database was searched. Although OMSSA searches were used as a guide, all O-GlcNAc site identifications were validated by manual interpretation of raw data.

RESULTS
We found that lysates from cells or isolated nuclei typically contained only about half of the endogenous emerin, the "easily released" fraction. The other half was recovered (solubilized) only when samples were sonicated prior to centrifugation. Using the protocol depicted in Fig. 1A (see "Experimental Procedures"), we isolated the cytoplasm (C), "easy" nuclear lysate (E) and sonicated nuclear lysate (S) from HeLa cells. Alternatively, HeLa cells were directly lysed in 0.3% Triton X-100 and 300 mM NaCl, sonicated, and centrifuged to obtain "total" whole cell lysates (WCL T ; Fig. 1B). Similarly, "total" nuclear lysates (N T ) were prepared by sonicating the low speed (nuclei/ organelle) pellet (Fig. 1C). All fractions were resuspended in equal volumes. Equal percentages of each fraction (2%) were resolved by SDS-PAGE and immunoblotted with antibodies to endogenous emerin, lamin A/C, lamin B, actin, BAF, or histone H3 (Fig. 1D). Emerin was detected at very low levels in isolated cytoplasm (Fig. 1D, lane 1) and was abundant in all fractions that included nuclear proteins (Fig. 1D, lanes 2-7). We quantified the emerin distribution in each nuclear fraction (e.g. N S ) as a percentage of each corresponding sum (e.g. N E ϩ N S ; Fig. 1E). Starting with isolated HeLa cell nuclei, on average, 56% of emerin was easily extracted, and 44% was sonication-dependent (Fig. 1E, n ϭ 4). Similarly, 65% of whole cell lysate emerin was easily extracted, and 35% was sonication-dependent ( Fig.  1E, n ϭ 4). We concluded that 35-44% of endogenous emerin is insoluble unless sonicated, potentially due to association with chromatin or insoluble nucleoskeletal structures, because histone H3 and lamin B were both found predominantly in sonicated fractions (Fig. 1D, lanes 2 and 3 and lanes 5 and 6). Two nonspecific bands recognized by the lamin B antibody provided internal loading controls (Fig. 1D). Lamins A and C (70 and 60 kDa, respectively) were present at low levels in the cytoplasm and were abundant in both nuclear fractions (Fig. 1D, lanes 2 and 3). More than half of BAF was cytoplasmic (Fig. 1D, lane 1), as expected (66 -68). In HeLa cells, most (ϳ74%) endogenous nuclear BAF was sonication-dependent ( Fig. 1D, lanes 2 and 3), suggesting avid association with chromatin or other N S -specific component(s) in vivo. Actin was abundant in the cytoplasm, as expected (Fig. 1D, lane 1), and was highly enriched in N E compared with N S (Fig. 1D, lanes 2 and 3). We concluded that the N E and N S fractions were enriched for distinct nucleoskeletal "niches"; both niches included emerin, A-type lamins, and BAF, whereas actin was primarily easy, and two other components, chromatin and lamin B, were primarily sonication-dependent.
GFP-Emerin and BAF Associate Only in the Sonication-dependent Fraction of HEK293T Cells-To compare these distributions in a different cell type, we used human HEK293T cells. Equal percentages (1%) of untransfected HEK293T cell cytoplasm (C), N E , N S , and N T fractions were resolved by SDS-PAGE and immunoblotted for endogenous emerin, A-and B-type lamins, BAF, and histone H3 (Fig. 1F). Consistent with HeLa cells, chromatin and lamin B were almost exclusively son-ication-dependent in HEK293T lysates; we therefore view chromatin and lamin B as defining components of the N S fraction. Emerin, found at roughly similar levels in N E and N S , also behaved consistently. By contrast, the distributions of A-type lamins and BAF in HEK293T nuclei shifted; a higher proportion of A-type lamins was sonication-dependent, whereas a higher proportion of nuclear BAF was easy (Fig. 1F, lanes 2 and 3; n ϭ 3).
We used HEK293T cells, which transfect and express exogenous proteins more efficiently than HeLa cells, to evaluate GFP-emerin binding to endogenous BAF in each fraction. Starting with HEK293T cells that expressed wild type GFPemerin for 24 h, we isolated nuclei and prepared equal volumes of either the separate N E and N S fractions or sonicated total

. Distribution of endogenous nuclear proteins in unsonicated versus sonicated lysates from HeLa or HEK293T cells ("N E /N S partitioning").
A-C, schematic diagrams of cell fractionation protocols used to isolate the easy versus sonication-dependent fractions of cells or pelleted nuclei (A). Alternatively, we directly lysed and sonicated whole cells (B) or pelleted nuclei (C) to generate total lysates. D and E, equal percentages (2%) of lysates from isolated nuclei or whole HeLa cell lysates (C, cytoplasm; E, easily extracted; S, sonicated; T, total) were resolved by SDS-PAGE and immunoblotted for endogenous cytoplasmic and nuclear proteins. *, B-type lamins; two nonspecific bands were also detected by the lamin B antibody in HeLa cells. The emerin distribution from D was quantified in E as the percentage of emerin in the easy versus sonication-dependent fractions (n ϭ 4; error bars, S.E.). F, same experiment as A for HEK293T cells (1% of fractions; n ϭ 3). G, GFP-emerin and endogenous BAF are present in both the N E and N S fractions of HEK293T cells but associate (co-immunoprecipitate) only in N S . Easy, sonicated, and total nuclear fractions (N E , N S , and N T , respectively) from HEK293T cells 24 h posttransfection with GFP-emerin (wild type or mutant; control lanes show input lysates (1%)) immunoprecipitated with bead-conjugated GFP antibodies or beads alone, resolved by SDS-PAGE, and blotted with antibodies to GFP (WB: GFP) or BAF (WB: BAF; bottom panel shows shorter exposure). Each blot shown is a representative of three independent experiments (n ϭ 3).
nuclear lysates (N T ). Control immunoblots of input lysates revealed GFP-emerin and endogenous BAF in both N E and N S (Fig. 1G, input lanes 1 and 2). In cells expressing GFP-emerin, endogenous BAF distributed more evenly between the N E and N S fractions (Fig. 1F (lanes 2 and 3) versus Fig. 1G (lanes 1 and 2, short exposure (short exp.)). We used agarose-conjugated llama GFP antibodies or agarose alone to immunoprecipitate an equal percentage of each fraction. Precipitates were resuspended in SDS-sample buffer, and 20% of each sample was resolved by SDS-PAGE and immunoblotted using antibodies specific for BAF or GFP (Fig. 1G, n ϭ 3). The agarose controls were negative (Fig. 1G, lanes 4 -6), confirming specific precipitation of GFP-emerin and BAF. Although both proteins were present in both fractions, endogenous BAF co-immunoprecipitated with GFP-emerin only from sonicated samples (Fig. 1G,  lane 7 versus lane 8). This unexpected result suggested that emerin-BAF association is biologically restricted (e.g. by posttranslational modification of one or both proteins) to the chromatin-containing niche. Around this time, we discovered a new posttranslational modification of emerin, O-GlcNAc, described next.
Endogenous Emerin is O-GlcNAcylated in Mammalian Cells-To explore whether endogenous emerin is O-GlcNAcylated, we treated HeLa cells for 4 h with or without 1 M TMG to specifically inhibit the enzyme (OGA) that removes O-GlcNAc (69). Equal protein concentrations of unsonicated whole HeLa cell lysates were then immunoprecipitated using either emerinspecific (␣-Emr) or nonspecific (IgG) rabbit antibodies ( Fig.  2A). Immunoprecipitates were resolved in parallel gels (SDS-PAGE), transferred to PVDF, and immunoblotted with an antibody specific for the O-GlcNAc modification (70), with or without competing 100 mM GlcNAc ( Fig. 2A). The O-GlcNAc antibody detected rabbit IgG light chain (ϳ25 kDa; Fig. 2A) as expected, and HeLa protein signals were higher in TMGtreated input lysates than in untreated lysates ( Fig. 2A, input), confirming sugar-specific detection (71). The emerin antibody specifically immunoprecipitated a ϳ34 kDa GlcNAcylated band from both the TMG-treated and untreated lysates (Fig. 2A, boxed region; n ϭ 3); specificity was confirmed by stripping each blot and reprobing with an emerin-specific monoclonal antibody (NCL-emerin; Fig. 2A, bottom panels). Most signals, including the 34 kDa band, were reduced or eliminated by competition with 100 mM GlcNAc ( Fig. 2A, right panel; n ϭ 3 To determine if the sonication-dependent population was also O-GlcNAc-modified, we immunoprecipitated endogenous emerin from equal volumes of HeLa cell N E and N S fractions ( Fig. 2B; 1% input shown), resolved by SDS-PAGE, immunoblotted for O-GlcNAc, and then stripped and reprobed for emerin (Fig. 2B). Specific O-GlcNAc signals at 34 kDa (the main emerin band) were detected at similar levels in both fractions ( Fig. 2B; n ϭ 2). Thus, both populations of emerin (N E and N S ) are O-GlcNAcylated in vivo.
To test emerin O-GlcNAcylation in a different species, the above experiment was repeated using easy whole cell lysates from the inducible mouse estrogen receptor-Cre-LoxP OGT F/Y line of mouse embryonic fibroblasts (mER-Cre-2A-GFP-MEFs) (58). When these cells are pretreated with the estrogen receptor-activating drug 4-hydroxytamoxifen (0.5 M), a cytoplasmic fusion protein (Cre-recombinase fused to mouse estrogen receptor; mER-Cre) translocates to the nucleus, where Cre excises the LoxP-flanked OGT gene. These cells constitutively express a GFP reporter, confirming that Ͼ90% of cells contain the mER-Cre-inducible protein (58). As controls, we used OGT F/Y MEFs that constitutively express GFP alone (58). Both lines were treated with 4-hydroxytamoxifen/ethanol or ethanol alone for 24 h and then washed to remove drug and cultured for 16 h in complete medium prior to immunoprecipitation and immunoblot analysis. The emerin antibody precipitated emerin specifically (no signal in the Ig control; Fig. 2C, Ig versus ␣-emerin) and at similar levels from both the GFP-expressing and OGT-deficient MEFs (Fig. 2C, WB: Emr). However, in cells with reduced levels of OGT enzyme, the level of O-GlcNAcylated emerin decreased 92% (Fig. 2C, left box; Fig. 2D, p Ͻ 0.004). Control blots with competing free sugar (Fig. 2C, right box) verified that the O-GlcNAc signal on emerin was specific. Immunoblotting of input lysates with antibodies to OGT protein, O-GlcNAc, or ␣-tubulin (loading control) verified reduction of both OGT itself and overall O-GlcNAcylation levels (Fig. 2E). These results independently validated emerin as endogenously O-GlcNAcylated in a different species (mouse).
Emerin Serines 53,54,87,171,and 173 and Three Other Residues are O-GlcNAcylated in Vitro-To identify modified residues, recombinant His-tagged emerin 1-220 (includes all Ser/Thr residues; Fig. 4) was O-GlcNAcylated in vitro and analyzed by mass spectrometry (see "Experimental Procedures"). His-tagged emerin was digested to generate either tryptic or LysC/AspN peptides for analysis by on-line HPLC MS on a high resolution tandem mass spectrometer. All 10 proteolytic peptides that contained Ser/Thr residues (87% coverage of emerin residues 1-220) were detected by accurate mass measurement and sequenced by CAD or ETD mass spectrometry. Mono-and di-O-GlcNAcylated forms of the tryptic peptide SSLDLSYYPTSSSTSFMSSSSSSSSWLTR (residues 175-203; Fig. 4) were detected by elution profile, accurate mass, and characteristic O-GlcNAc CAD signature ions (72). In summary, using mass spectrometry, we demonstrated that at least eight emerin residues were O-GlcNAcylated in vitro and successfully identified five of these sites: Ser-53, Ser-54, Ser-87, Ser-171, and Ser-173 (Fig. 4, A and B).

JOURNAL OF BIOLOGICAL CHEMISTRY 30199
These in vitro findings validated the mass spectrometry results and implied that Ser-53 and Ser-54 were either abundant or influenced further O-GlcNAcylation in vitro, because loss of either Ser-53 or Ser-54 (of at least eight O-GlcNAc sites) had a disproportionately large impact, reducing emerin O-GlcNAcylation by ϳ50% in vitro (Fig. 4, C and D). This result for Ser-54 was noteworthy because the S54F mutation is suffi-cient to cause EDMD, although the S54F protein is expressed and the majority localizes normally (76 -78) (see below).

Ser-53, Ser-54, and Ser-173 Are Relevant to Emerin O-GlcNAcylation in Cells-Before studying O-GlcNAc-site
mutations further in vivo, we did controls to determine if they affected GFP-emerin localization. HeLa cells that expressed GFP or GFP-emerin (wild type or mutant) for 24 h were visual-ized by direct GFP fluorescence imaging (Fig. 5A). GFP alone diffused throughout the nucleus and cytoplasm, as expected (79). When expressed at moderate levels, wild type GFPemerin and all five mutants (S53A, S54F, S52-54A, S173A, and S173D) localized mainly at the nuclear envelope and in discrete puncta outside the nucleus (Fig. 5A). We concluded that these mutations did not block the nuclear envelope localization of GFP-emerin.
To determine if the mutations affected emerin O-GlcNAcylation in vivo, unsonicated whole cell lysates from HeLa cells that expressed GFP or GFP-emerin (wild type or mutated) for 24 h were immunoprecipitated (equal protein concentration or cell-equivalent lysate) using a rabbit antibody to GFP, resolved by SDS-PAGE, and immunoblotted first with antibodies to O-GlcNAc and then with the GFP antibody (Fig.  5B). Results for most mutants were quantified by densitometry as the ratio of O-GlcNAc to GFP, normalized to wild type GFPemerin (Fig. 5C). Cells that expressed GFP alone gave only background O-GlcNAc signals around 60 kDa (Fig. 5B, boxed), whereas specific O-GlcNAc signals were detected for wild type, S173A-mutated, and S173D-mutated GFP-emerin (Fig. 5B).
The O-GlcNAc signal was significantly reduced by S53A (by 40%; p Ͻ 0.002; n ϭ 4) and S54F (by 45%; p Ͻ 0.005; n ϭ 4) and by the triple mutant S52-54A (by 66%; p Ͻ 0.007; n ϭ 5; Fig. 5, B and C). We concluded that Ser-53 and Ser-54 are individually important sites of emerin O-GlcNAcylation both in vitro (Fig.  4D) and in the easy niche of living cells (Fig. 5, B and C; N S fraction was not tested).
O-GlcNAcylation of the quadruple mutant (S52-54A plus S171A or S173A) was reduced to the same extent as the triple S52-54A mutant in these easy fractions (Fig. 5C). By contrast, the single S173A mutation caused the easy GFP-emerin population to be slightly more O-GlcNAcylated than wild type (18% more; p Ͻ 0.04; n ϭ 4; Fig. 5C). This showed that Ser-173 was relevant to emerin O-GlcNAcylation in the easy fraction; however, this particular outcome (increased O-GlcNAcylation presumably elsewhere in emerin) might have been due to loss of an alternative modification (e.g. phosphorylation) at Ser-173. Indeed, the phosphomimetic S173D mutation appeared to reduce emerin O-GlcNAcylation, but this reduction was not significant (Fig. 5C; n ϭ 3).
We wondered if fusion to GFP or O-GlcNAc site mutations disrupted the N E /N S distribution of emerin in HeLa cells. Therefore, as controls, we tested the distribution of endogenous emerin, GFP-emerin (wild type, S53A, S54F, S173A, or S173D) and endogenous nuclear markers by immunoblotting the isolated cytosol (C), N E , and N S fractions of HeLa cells (24 h posttransfection) for emerin (rabbit serum 2999), lamins A/C, lamin B, BAF, actin, or histone H3 ( Fig. 6A; 2% of each fraction loaded, n ϭ 3). Wild type and mutant GFP-emerins all partitioned similarly to each other (evenly split between N E and N S ; Fig. 6A) and did not disrupt partitioning of other nuclear proteins. Specifically, endogenous emerin and lamins A/C were present in both N E and N S , actin was abundant in cytosol and N E , BAF was abundant in cytosol and N S , and histone H3 and lamin B were predominantly in N S (Fig. 6A; compare Fig. 1A).

JOURNAL OF BIOLOGICAL CHEMISTRY 30201
resuspended in SDS-sample buffer, and 20% of each sample was resolved by SDS-PAGE and immunoblotted using antibodies specific for either lamin B or ␤-catenin ( Fig. 6B; n ϭ 2 each), or A-type lamins or BAF (Fig. 6B; n Ն 4 each). Among these various emerin mutations and known partners, we detected only one significant difference; the phosphomimetic GFP-emerin S173D polypeptide showed significantly reduced binding to endogenous BAF (58% less than wild type GFP-emerin; Fig. 6, B and C; n ϭ 7; p Ͻ 0.0002). BAF association was not significantly affected by any other mutation tested (Fig. 6, B and C; n ϭ 6 and  7). This suggested that GFP-emerin Ser-173 phosphorylation (mimicked by S173D) might specifically inhibit binding to Bar, 10 m. Images were contrast-adjusted using Adobe Photoshop. B and C, immunoblots of unsonicated whole cell protein lysates from HeLa cells (WCL E ) that transiently expressed GFP or GFP-emerin (wild type or missense-mutated) for 24 h, resolved by SDS-PAGE either before (1% input) or after immunoprecipitation with rabbit GFP antibodies. Immunoblots were probed first with the O-GlcNAc antibody (WB: O-GlcNAc; top) and then stripped and reprobed for GFP (WB: GFP; bottom). A vertical black line separates different parts of the same blot. C, results from B quantified by densitometry and graphed as the O-GlcNAc-to-GFP-emerin ratio for each mutant, normalized to wild type GFP-emerin. Error bars, S.E.; significance was determined by Student's t test. **, p Ͻ 0.007; *, p Ͻ 0.04 (n ϭ 3-5). D and E, similar to B, but HeLa cells were treated (or not) for 4 h with 1 M TMG (which inhibits OGA) to increase O-GlcNAc signals prior to harvest, and whole cell lysates were sonicated (WCL T ) prior to immunoprecipitation and immunoblotting first with antibodies to O-GlcNAc (D, top) and then GFP antibodies (D, bottom). Inset, contrast-enhanced (backlit scanned), close-up view of boxed region. E, results for each quadruple mutant (S52-54A plus S171A or S173A) graphed relative to the triple mutant S52-54A. Error bars, S.E.; n ϭ 3; *, p Ͻ 0.04 by Student's t test).
BAF. This result was initially puzzling, because the LEM domain (residues 4 -44) is sufficient to bind BAF (3), and in early studies, emerin binding to BAF was disrupted only by LEM domain mutations, not by the few tested mutations elsewhere in emerin (9).
However, several results provided a clue as to the mechanism. Emerin migrates slowly in SDS gels when hyperphosphorylated (62,81,82), and slowly migrating bands were obvious in the endogenous easy emerin population (e.g. Figs. 1F and 7A) and detectable in the endogenous N S population (e.g. see Fig.  2B). The S173D mutation appeared to increase the slowly migrating GFP-emerin signal (e.g. see Figs. 5B (GFP blot) and 6A (lane 17)). Furthermore, previous studies showed that emerin-BAF association in human cell lysates or mitotic Xenopus cell lysates was inexplicably affected by the Y161F or S175A mutations, respectively (62,80), near Ser-173. We therefore examined potential posttranslational consequences of emerin Ser-173 mutations.
In Vivo Consequences of GFP-Emerin S173A versus S173D-To explain how S173D reduced BAF association in vivo, we hypothesized that this phosphomimetic mutation "primed" emerin for further posttranslational modifications that, in turn, influenced BAF association. To test this model, we first quantified the obvious upper band in easy WCL. Equal protein amounts of unsonicated whole cell lysates (WCL E ) from HeLa cells that expressed GFP or GFP-emerin (wild type, S173A, or S173D) were resolved by SDS-PAGE and immunoblotted for GFP (Fig. 7A). The S173D mutation caused a consistent ϳ5-fold increase in the slowly migrating signal, compared with wild type and S173A-mutated GFP-emerin (Fig. 7, A and B; p Ͻ 0.009; n ϭ 3) and affected ϳ30% of the easy GFP-emerin population (Fig. 7B).
To test potential hyperphosphorylation of GFP-emerin in the N E and N S fractions, we tested sensitivity to -phosphatase (preferentially dephosphorylates Ser/Thr residues) with or without CIP, which preferentially dephosphorylates Tyr residues. N E and N S fractions from isolated HeLa nuclei were treated with or without -phosphatase or CIP. Samples were resolved by 10% BisTris SDS-PAGE to improve resolution of hyperphosphorylated bands and immunoblotted for emerin and lamins A/C (Fig. 7C). In the N E fraction, we detected two upper bands for GFP-emerin S173D (ϳ60 and ϳ58 kDa); -phosphatase eliminated both and produced a -resistant ϳ56 kDa band (Fig.  7C, lane 13 versus lane 15). CIP eliminated both upper bands of FIGURE 6. S173D mutation disrupts GFP-emerin association with BAF. A, the N E versus N S distributions of GFP or GFP-emerin (wild type, S53A, S54F, S173A, or S173D) were tested in HeLa cells 24 h posttransfection. Shown are equal percentages (2%) of each cytoplasm (C) and separated N E and N S fractions, resolved by SDS-PAGE and immunoblotted with antibodies to emerin, lamin B, lamins A/C, BAF, actin, or histone H3 (n ϭ 3). B and C, sonicated total nuclear lysates (N T ) from HEK293T cells 24 h posttransfection with GFP-emerin (wild type or mutant), precipitated with bead-conjugated GFP antibodies or beads alone, SDS-PAGE-resolved, and probed with antibodies to GFP, lamins A/C, or BAF. Controls show input N T lysate (1%) and beads-only precipitation from wild type GFP-emerin lysates. Results are quantified in C as the amount of BAF that co-immunoprecipitated with each GFP-emerin mutant, relative to wild type GFP-emerin (n ϭ 7; **, p Ͻ 0.0002 by Student's t test; bars, S.E.). OCTOBER 18, 2013 • VOLUME 288 • NUMBER 42

JOURNAL OF BIOLOGICAL CHEMISTRY 30203
GFP-emerin S173D and yielded trace amounts of a ϳ55 kDa "CIPresistant" band (Fig. 7C, lane 13 versus lane 14). Treatment with both enzymes gave the same results as CIP alone (Fig. 7C, lane 16 versus lane 14), suggesting that CIP effectively dephosphorylated both Tyr and Ser/Thr residues in emerin.
The S173D mutation also increased emerin hyperphosphorylation in the N S niche, but non-identically to its effects in the easy niche; we detected only one major S173D upper band in the N S fraction (ϳ58 kDa ; Fig. 7C, lane 13, bracket) (not two upper bands, as seen in N E ) that was similarly sensitive to and CIP (Fig. 7C, lanes 14 -16 versus lane 13; bracket). The slight downshift of the main GFP-emerin band in all CIP-treated samples (e.g. see Fig. 7C, lane 5 versus lane 6) was attributed to co-migration of 2 g of CIP (ϳ56 kDa; samples treated with both phosphatases had 1 g each). CIP treatment also caused endogenous A-type lamins in the N S niche, particularly lamin C, to migrate faster (Fig. 7C, lane 1 versus lane 2), consistent with many reported phosphorylation sites in A-type lamins (83). The minor (ϳ50 kDa) band (Fig. 7, A and C) is probably a GFP-emerin breakdown product.  's t test). C, potential Ser/Thr or Tyr phosphorylation of GFP-emerin upper bands tested by treatment with CIP (which preferentially dephosphorylates Tyr) with or without -phosphatase (which preferentially dephosphorylates Ser/Thr) or with neither as a control. Untransfected HeLa cells (UN) and HeLa cells 24 h posttransfection with GFP-emerin (wild type, S173A, and S173D) were fractionated; separate N E and N S fractions were treated with (or without) -phosphatase or CIP or both and then resolved using 10% BisTris SDS-PAGE (to improve band resolution) and probed with antibodies to emerin (serum 2999) or lamins A/C (N S fractions only). *, main GFP-emerin band. Brackets, slowly migrating GFP-emerin bands (n ϭ 2). D, schematic summary and model of our main results. Emerin mono-and di-O-GlcNAcylation at Ser-53 and Ser-54 is proposed to control emerin conformation near the LEM domain and favor O-GlcNAcylation at additional (unidentified) sites. Ser-173 O-GlcNAcylation is proposed to promote BAF association in the sonication-dependent niche. Potential (nonexclusive) mechanisms include (a) control of emerin conformation at S173; (b) control of an unidentified partner that influences emerin-BAF association; (c) inhibition of an alternative fate, namely Ser-173 phosphorylation and consequent hyperphosphorylation at unidentified sites.
Our estimate that the S173D mutation increased GFPemerin hyperphosphorylation at least 5-fold in the N E fraction (Fig. 7A) was conservative, because it did not include potentially phosphorylated molecules in the main emerin band (62). We conclude that emerin Ser-173 phosphorylation, mimicked by S173D, increases emerin phosphorylation in vivo in different ways, depending on whether emerin occupies the easy or sonication-dependent niche. Considered together with the S173A mutation, which increased O-GlcNAcylation in the easy WCL E niche (Fig. 5C) but decreased overall emerin O-GlcNAcylation in combination with the S52-54A mutation, we propose that emerin Ser-173 (and possibly neighboring residues) functions as a molecular "switch" that influences downstream O-GlcNAcylation and phosphorylation events and controls BAF association in vivo.

DISCUSSION
These results show emerin is regulated by O-GlcNAcylation in vivo. At least eight residues in human emerin are O-GlcNAcylated in vitro, including Ser-53, Ser-54, Ser-87, Ser-171, and Ser-173. Mutations at three sites (Ser-53, Ser-54, and Ser-173) had significant posttranslational or functional consequences, discussed below. We also report a fractionation protocol that separates emerin (along with BAF and A-type lamins) into two functionally and posttranslationally distinct populations. One population included the vast majority of lamin B and chromatin; only in this context did GFP-emerin associate with endogenous BAF. We propose that this sonication-dependent fraction corresponds, in living cells, to a niche in which chromatin is fundamentally organized by association with emerin, BAF, and B-type lamins. This niche also included variable levels of A-type lamins that might, we speculate, tether silent chromatin. By contrast, the paucity of emerin-BAF association in the easy fraction, where A-type lamins were also abundant, suggests that all three proteins have alternative functions in the easy niche. Potential alternative functions include roles in signaling, proliferation, epigenetic regulation, and association with other nucleoskeletal components (1,44,83). These two fractions significantly extend the nuclear matrix and nucleoskeleton as operational concepts by separating the nuclear membrane protein emerin and other nuclear lamina proteins based on their association with the genome; B-type lamins and a subset of emerin, BAF, and A-type lamins partition with chromatin, whereas easy emerin, BAF, and A-type lamins are either fully chromatin-independent or associate with trace amounts of chromatin (H3) present in this fraction. The unique functions and regulation of lamins, emerin, and BAF in each niche and their relationships to chromatin are important new questions for future work. Furthermore, our results collectively show that OGT, an essential enzyme, regulates at least two regions in the emerin molecule and controls emerin association with BAF (and hence, we propose, chromatin) at the nuclear envelope.
BAF Associates with Emerin Only in the Lamin B-and Chromatin-containing Niche-Unsonicated nuclear lysate supernatants (N E ) contained roughly half the emerin, BAF, and A-type lamins and most nuclear actin. Sonication was required to release the N S populations of emerin, BAF, A-type lamins, and ϳ82-96% of histone H3 and lamin B. Chromatin was no surprise as a defining component of the N S fraction, because sonication shears chromosomal DNA. We speculate that lamin B filaments associate more avidly with chromatin or other stable nucleoskeletal component(s) (1). Differential solubility of A-type lamins has ample precedent (84 -86), including FRAP evidence that "internal" GFP-lamins are more mobile than envelope-associated GFP-lamins (87,88) and evidence that lamins "loosen" locally during egress of nascent ribonucleoprotein particles (89) and herpes simplex virus-1 particles (90 -92).
We expected to find BAF predominantly in the easy fraction, because GFP-BAF is diffusionally mobile in HeLa cell nuclei (FRAP recovery half-time, 80 -260 ms) (93) and C. elegans nuclei (halftime, 2.24 s) (94). Instead, most (ϳ74%) endogenous nuclear BAF was sonication-dependent in HeLa cells, whereas ϳ22% was sonication-dependent in HEK293 cells, suggesting that N E /N S partitioning is influenced by cell type or malignancy status (67).
The sonication-dependent population of endogenous BAF is, we propose, closely associated with chromatin. BAF directly cross-bridges and "loops" dsDNA in vitro (95)(96)(97), condenses chromatin (66), mediates chromatin attachment to the nuclear lamina (98,99), associates with telomeres and "core" regions of chromatin during mitosis (5,6,100), and protects chromatin from DNA damage (101). Our discovery that GFP-emerin and BAF associate only in the N S fraction, when both were also abundant in N E , provides a vital new tool for isolating endogenous BAF and testing its proposed role in linking chromatin to other major "lamina" components (emerin and lamins) in mammalian cells.
Emerin O-GlcNAcylation at Ser-53 and Ser-54; Implications for Conformation, Mitosis, and EDMD-Mutating either Ser-53 or Ser-54 significantly (ϳ50%) reduced emerin O-GlcNAcylation, both in cells (where other enzymes and partners might influence outcome) and by purified OGT in vitro. We conclude that residues Ser-53 and Ser-54 are each individually important for OGT regulation of emerin. Interestingly, O-GlcNAc and phosphate can have opposite effects on peptide conformation; O-GlcNAc stabilized and tightened a canonical type-II ␤-turn, whereas phosphate opened and extended this turn (102). We therefore hypothesize that emerin Ser-53/Ser-54 O-GlcNAcylation either (a) influences OGT activity or substrate specificity (103,104) or (b) alters the conformation of emerin and thereby facilitates OGT access to additional sites. The hypothesis that O-GlcNAcylation alters emerin conformation will be tested in future work.
Emerin Residues 161-175, a Proposed "Nexus" for Complex Posttranslational Control of Emerin Binding to BAF-In striking contrast to Ser-53 and Ser-54 mutations, which disrupted emerin O-GlcNAcylation by purified OGT in vitro, mutations at Ser-173 had detectable consequences only in cells. Furthermore, these consequence(s) depended on whether the mutation was phosphomimetic (S173D versus S173A) and differed between the easy and sonication-dependent populations. The single S173A mutation enhanced GFP-emerin O-GlcNAcylation in the easy fraction. However, combined with the triple S52-54A mutation, S173A reduced emerin O-GlcNAcylation in the sonicationdependent fraction. The phosphomimetic S173D mutation increased hyperphosphorylation in both fractions but nonidentically (one versus two slowly migrating bands). BAF association was unaffected by S173A, suggesting that BAF requires neither a Ser nor O-GlcNAcylation at residue 173. However, BAF association was reduced significantly by the phosphomimetic S173D.
These consequences of emerin Ser-173 mutations and their effects on BAF can be explained by at least three nonexclusive mechanisms, depicted in Fig. 7D. First, phosphorylation versus O-GlcNAcylation at Ser-173 might have opposite effects on emerin conformation, as discussed above. Second, Ser-173 modifications might control binding to a partner(s) that influences emerin association with BAF, for example either sterically (e.g. GCL competes with BAF) (15) or posttranscriptionally (e.g. tyrosine protein phosphatase 1B) (110). This hypothetical partner(s) is unknown. Third, Ser-173 O-GlcNAcylation might promote BAF association by blocking an alternative fate in the sonication-dependent niche: Ser-173 phosphorylation and subsequent hyperphosphorylation. Our model that Ser-173 modifications, far from the LEM domain, somehow control binding to BAF is supported by previous evidence that the Y161F and S175A mutations also influence BAF association in whole cell lysates (62,80). These findings collectively define emerin residues 161-175 as a nexus for complex posttranslational regulation of emerin binding to BAF.
OGT Regulation of the Nuclear Lamina?-OGT responds to cell nutrient (including glucose) status, and its targets include histones, HDAC complexes, RNA polymerase II, and key transcription factors (57,64) and now also the conserved LEM domain protein, emerin. OGT regulation of emerin provides a direct molecular mechanism for cross-talk between cell nutrient status and emerin association with BAF, an enigmatic nuclear lamina component that influences histone postranslational modifications (111). Furthermore, OGT misregulation of emerin may be relevant to aging, heart disease, and diabetes, all of which are characterized by elevated protein O-GlcNAcylation (57). How O-GlcNAcylation influences the emerin molecule and its roles in mitosis, signaling, and gene silencing are major open questions for future work.