Research Article 2851

IntroductionThe post-translational modification of cellular proteins by addinga single b-N-acetylglucosamine (GlcNAc) moiety is a highlyevolutionarily conserved process. The turnover rate of thismodification is much faster than for the protein itself, affectingmany downstream cellular activities, and thus serving as apreviously unidentified signaling mechanism. Two enzymes, O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA) regulate theO-GlcNAc modification, by adding and removing O-GlcNAc,respectively. UDP-GlcNAc, the substrate of OGT, is derived fromthe hexosamine biosynthetic pathway, a pathway modulated bycellular nutrient levels. Although the target sequence of thismodification does not conform to a clear consensus, O-GlcNAcylation occurs at serine or threonine residues and on manyphosphoproteins (Wang et al., 2008; Love and Hanover, 2005;Slawson and Hart, 2003). More than 600 proteins have beenidentified as O-GlcNAc-modifiable targets. The functional roles ofthis modification are not yet fully understood, but growing evidenceindicates that O-GlcNAc impacts multiple cellular activities. O-GlcNAcylated proteins might modulate a variety of downstreamsignaling pathways by influencing gene transcription, proteinstability, transport and proteasome activity (Love et al., 2010;Marshall, 2006; Zachara and Hart, 2004). The O-GlcNAcmodification is known to be responsive to metabolic status, andperturbations in O-GlcNAc cycling have been linked to majorchronic metabolic diseases such as Alzheimer’s, diabetes and cancer(Love et al., 2010; Lazarus et al., 2006; Love and Hanover, 2005).Additionally, a growing body of evidence suggests that elevatedlevels of O-GlcNAc, associated with cellular injury, are protectivein several models of stress (Zachara et al., 2010).

The present study focuses on splice variants of OGA, whichremoves O-GlcNAc on target proteins. OGA was first identified asmeningioma-expressed antigen 5 (MGEA5) and is found in alltissues examined, with the highest expression in brain, placenta andpancreas (Heckel et al., 1998; Comtesse et al., 2001). Human MGEA5is present on chromosome 10 at cytological position 10q24.1, whichhas been identified as a diabetes susceptibility locus in the Mexican-American population (Duggirala et al., 1999; Lehman et al., 2005).The product of this gene, OGA (E.C. 3.2.1.169, protein O-GlcNAcase), has been referred to as hexosaminidase C, because it isdistinct from the lysosomal b hexosaminidases A and B. Comparedwith the acidic hexosaminidases, OGA has a neutral pH optimumand shows selectivity for the removal of GlcNAc rather than GalNAc(Gao et al., 2001). Two splice variants derived from the MGEA5gene were originally identified in human tissues (Comtesse et al.,2001). The full-length OGA long form (OGA-L) has 916 aminoacids transcribed from 16 exons (NM_012215) and a short isoform(OGA-S) contains 677 amino acids transcribed from first 10 exons(AF307332). Both isoforms are identical at the N-terminalhyaluronidase domain, but differ in that OGA-L has a histone acetyltransferase (HAT)-like domain at the C-terminus (Schultz and Pils,2002; Toleman et al., 2004; Whisenhunt et al., 2006). The OGA-Sisoform continues transcription through exon 10, skipping the splicejunction and terminating with a unique 14 amino acid tail. Thecrystal structure of the human O-GlcNAcase has not been resolved,but two groups have reported the three-dimensional structure of twobacterial O-GlcNAcase homologs and have elucidated their catalyticmechanism (Dennis et al., 2006; Rao et al., 2006).

The functional activities of the hyaluronidase and HAT domainsin OGA-L were described by Wells et al., Toleman et al. and

SummaryProtein-O-linked N-Acetyl-b-D-glucosaminidase (O-GlcNAcase, OGA; also known as hexosaminidase C) participates in a nutrient-sensing, hexosamine signaling pathway by removing O-linked N-acetylglucosamine (O-GlcNAc) from key target proteins. Perturbationsin O-GlcNAc signaling have been linked to Alzheimer’s disease, diabetes and cancer. Mammalian O-GlcNAcase exists as two majorspliced isoforms differing only by the presence (OGA-L) or absence (OGA-S) of a histone-acetyltransferase domain. Here wedemonstrate that OGA-S accumulates on the surface of nascent lipid droplets with perilipin-2; both of these proteins are stabilized byproteasome inhibition. We show that selective downregulation of OGA-S results in global proteasome inhibition and the strikingaccumulation of ubiquitinylated proteins. OGA-S knockdown increased levels of perilipin-2 and perilipin-3 suggesting that O-GlcNAc-dependent regulation of proteasomes might occur on the surface of lipid droplets. By locally activating proteasomes during maturationof the nascent lipid droplet, OGA-S could participate in an O-GlcNAc-dependent feedback loop regulating lipid droplet surfaceremodeling. Our findings therefore suggest a mechanistic link between hexosamine signaling and lipid droplet assembly andmobilization.

Key words: O-GlcNAcase isoforms, O-GlcNAc cycling, Perilipin-2, Lipid droplets, Proteasome

Accepted 24 March 2011Journal of Cell Science 124, 2851-2860© 2011. Published by The Company of Biologists Ltddoi:10.1242/jcs.083287

A lipid-droplet-targeted O-GlcNAcase isoform is a keyregulator of the proteasomeChithra N. Keembiyehetty*, Anna Krzeslak*,‡, Dona C. Love and John A. Hanover§

Laboratory of Cell Biochemistry and Biology, National Institute of Diabetes and Kidney Diseases, National Institute of Health, Bethesda, MD 20892,USA*These authors contributed equally to this work‡Present address: Deparment of Cytobiochemistry, Universiti of Lodz, 90-237 Lodz, Poland§Author for correspondence (jah@helix.nih.gov)

Jour

nal o

f Cel

l Sci

ence

Butkinaree et al. (Wells et al., 2002; Toleman et al., 2004;Butkinaree et al., 2008). The OGA-L isoform has been shown tobe a part a gene co-repressor complex and is subject to O-GlcNAcylation by OGT (Lazarus et al., 2006; Whisenhunt et al.,2006). Until recently, the OGA-S isoform was not extensivelystudied. Previous reports using subcellular fractionation andantigenicity described OGA-L as a cytoplasmic–nuclear proteinand OGA-S as a nuclear variant (Comtesse et al., 2001). The O-GlcNAcase activity of OGA-S was demonstrated in our laboratoryusing a highly sensitive and specific fluorogenic substrate (Kim etal., 2006) as was the selective inhibition of the enzyme by a-GlcNAc thiosulfonate (Kim et al., 2007). The O-GlcNAcase activityof the short isoform was confirmed independently (Macauley andVocadlo, 2009) but its functional significance and its cellularlocalization remain to be established.

The present study explores the localization and functionalactivities of the two major OGA isoforms, OGA-L and OGA-S.Our results indicate that they both play key roles in modulatingproteasome activity, although the isoforms have distinct intracellularlocalizations. We also demonstrate that OGA-S is targeted to thesurface of nascent lipid droplets, thereby revealing a possible linkbetween hexosamine signaling and the proteasome-dependentremodeling of the surface of lipid droplets.

ResultsTargeting and activity of OGA isoformsThe mammalian Mgea5 gene encodes two major isoforms, whichwe have designated OGA-S and OGA-L (Kim et al., 2006). OGA-L utilizes all 16 exons, contains a N-terminal hyaluronidase domain(Comtesse et al., 2001) and a C-terminal putative histone acetyltransferase domain (Fig. 1A). The shorter isoform comprises 10exons with an N-terminal hyaluronidase domain and a unique 14amino acid C-terminus. Because antibodies raised against OGA donot identify both isoforms (supplementary material Fig. S1),plasmid vectors were constructed with green fluorescent protein(GFP) fusions of both isoforms (OGA-L–GFP and OGA-S–GFP)to determine their intracellular targeting. The isoforms wereexpressed as full-length GFP fusions in HeLa cells (Fig. 1B,arrows), however, the number of cells expressing OGA-S–GFPwas substantially lower than that expressing OGA-L–GFP,suggesting differences in RNA or protein stability between the twoisoforms. Western blot analysis revealed a significant, yetdifferential decrease in O-GlcNAc-modified proteins upon overexpression of either OGA-L–GFP or OGA-S–GFP, indicating thatboth enzymes are catalytically active (Fig. 1C). The levels ofenzyme activity correspond to the levels of expression of bothenzymes; quantitative western blot analysis demonstrated thatOGA-L–GFP lowered global O-GlcNAc levels by approximately72% and OGA-S–GFP lowed the O-GlcNAc levels by 35%. Next,an increase in OGA enzymatic activity was measured directlyusing a specific fluorogenic substrate (Fig. 1D). Overexpression ofOGA-L–GFP increased total OGA activity by 446% comparedwith control, whereas OGA-S–GFP increased activity by 136%.We examined the intracellular localization of the two isoforms inHeLa cells. OGA-L–GFP localized diffusely throughout the nucleusand cytoplasm (Fig. 1E, top panel, green channel) whereas OGA-S–GFP was found associated with structures suggestive of lipiddroplets (Fig. 1E, bottom panel, green channel). We monitored thein situ activities of the overexpressed enzymes by staining HeLacells with the O-GlcNAc-specific antibody, RL2 (Fig. 1E, redchannel). OGA-L was present in the nucleus and cytoplasm (Fig.

1E, top row) and efficiently depleted cytosolic O-GlcNAc. We alsonoted that the nuclear O-GlcNAc compartment (denoted by anasterisk) was somewhat refractory to overexpression of the enzyme.By contrast, OGA-S was associated with circular structuresreminiscent of lipid droplets (Fig. 1E, bottom row). We noted thatalthough the steady-state distribution of OGA-S was distinct fromthat of OGA-L, overexpression of this isoform dramatically reduced

2852 Journal of Cell Science 124 (16)

Fig. 1. O-GlcNAcase is expressed as two differentially targeted and activeisoforms. (A)Genomic organization and schematic exon diagram of the O-GlcNAcase isoforms. The OGA long isoform (OGA-L) is encoded by 16exons and contains a N-terminal hyaluronidase catalytic domain and a C-terminal HAT domain. The OGA short isoform (OGA-S) is encoded by 10exons and contains a N-terminal hyaluronidase catalytic domain. The HATdomain is replaced by a unique 14 amino acid C-terminal tail. (B)Western blotimmunostained with a GFP-specific antibody showing the expression of OGA-L–GFP (~150 kDa) and OGA-S–GFP (~120 kDa). Each isoform migrated tothe expected molecular mass. (C)Overexpression of OGA-L–GFP or OGA-S–GFP causes a decrease in O-GlcNAc immunoreactivity compared withuntransfected cells (control). Using densitometry, a 4-fold and 1.34-folddecrease in O-GlcNAcylation was detected for OGA-L–GFP and OGA-S–GFP, respectively. (D)Total OGA enzymatic activity was measured fromOGA-L–GFP and OGA-S–GFP cellular lysates using the fluorogenicsubstrate, FD-GlcNAc. Values are means ± standard deviation; s.d. of threeseparate experiments. (E)Representative confocal images of OGA-L–GFP andOGA-S–GFP overexpression in HeLa cells showing nucleocytoplasmiclocation of OGA-L (upper panel) and exclusive cytoplasmic staining of OGA-S isoform (lower panel). Overexpression of both isoforms reduced O-GlcNAclevels as shown by diminished O-GlcNAc antibody staining. Cells expressingthe fusion proteins are outlined for comparison. Asterisks indicate the nucleusof each cell.

Jour

nal o

f Cel

l Sci

ence

O-GlcNAc levels in the cell, including the nuclear and cytoplasmiccompartments (Fig. 1E, bottom row). These data suggested thatOGA-S was uniquely targeted within the cell and, like OGA-L,was catalytically active.

To confirm the targeting of OGA-S to the lipid droplet, HeLacells were treated with 0.4 mM oleic acid to stimulate lipid dropletformation. These cells were infected with adenoviral vectorsexpressing either OGA-L–GFP or OGA-S–GFP. OGA-L–GFP wasfound throughout the cell in both nucleus and cytoplasm, and wasnot enriched on neutral lipid droplets stained with Nile Red (Fig.2A; top row; red channel). By contrast, the majority of OGA-S–GFP concentrated at the periphery of lipid droplets (Fig. 2A,middle row; green channel). OGA-S–GFP localization to lipiddroplets in HeLa cells was further verified by staining of perilipin-2 (also called ADRP, ADFP and adipophilin), a lipid-droplet-associated protein. Here, OGA-S–GFP colocalized with perilipin-2at the lipid droplet surface (Fig. 2A, bottom row; arrows and inset).To confirm that the targeting was specific and not due to the C-terminal fusion of GFP, we generated a construct fusing ahemagglutinin (HA) tag at the N-terminus of OGA-L and OGA-S(HA–OGA-S and HA–OGA-L). The localization of these HA-tagged proteins duplicated the localization of the GFP-taggedproteins, confirming that OGA-L and OGA-S are differentiallytargeted within the cell and that OGA-S preferentially accumulatesat lipid-rich structures (supplementary material Fig. S2).

Lipid accumulation enhances OGA activityThe specific targeting and apparent stabilization of OGA-S bylipid accumulation might be expected to alter O-GlcNAcase activity.Using the highly sensitive fluorogenic O-GlcNAcase substrate,di(N-acetyl-b-D-glucosaminide; FD-GlcNAc), O-GlcNAcaseactivity was measured in cell lysates from HeLa cells that were

overexpressing either OGA-L–GFP or OGA-S–GFP, with andwithout oleic acid. Total activity of OGA-L–GFP was greater thantwofold higher than that of OGA-S–GFP, presumably because ofmarkedly lower expression efficiency of OGA-S–GFP. Upontreatment of cells with oleic acid, the highest relative increase inenzyme activity was seen in the OGA-S–GFP-expressing cells(35%) compared with the OGA-L–GFP-expressing cells (6%; Fig.2B). Thus lipid accumulation appears to enhance OGA-S–GFPactivity preferentially. To test whether oleic acid supplementationalso enhances endogenous OGA activity, we performed a similarexperiment using untransfected cells. Cells were either serumdepleted or treated with 0.4 mM oleic acid for 24 hours to modulatelipid stores (Fig. 3A). Compared with serum depletion, oleic acidsupplementation increased OGA activity more than twofold (Fig.3B). Based on the localization of OGA-S, a substantial portion ofactivity should be associated with the lipid droplet fraction. Thelipid droplet fraction from the oleic-acid-treated cells containedgreater than 50% of the total OGA activity (Fig. 3B). Themodulation of OGA activity by increasing lipid droplet storesmight be the result of enhanced stability of OGA-S at the lipiddroplet surface, in a manner similar to that reported for perilipin-2 (Xu et al., 2005).

The unique targeting of OGA-S was tested in a physiologicallyrelevant system using the 3T3 L1 adipocyte model. 3T3 L1 pre-adipocytes were differentiated to mature adipocytes using a standardprotocol (Student et al., 1980) and whole-cell lysates from eachstep in differentiation were probed for changes in O-GlcNAcylation.Different subsets of proteins were detected by O-GlcNAc-specificantibodies over the course of adipogenic differentiation, indicatinga variation in O-GlcNAc cycling on protein targets during thisprocess (Fig. 4A; and see also supplementary material Fig. S3).During the differentiation process, we also detected a modulation

2853O-GlcNAcase isoform targeted to lipid droplets

Fig. 2. OGA-S accumulates at the surface of lipid droplets and its activity is elevated by increased lipid stores. (A)Transfected HeLa cells expressing eitherOGA-L–GFP or OGA-S–GFP were treated with 0.4 mM oleic acid to increase lipid stores. OGA-L–GFP is located diffusely in the nucleus and cytoplasm (greenchannel) with no specific connection with lipid droplets (red channel). By contrast, OGA-S–GFP targets to the surface of lipid droplets. Neutral lipid was stainedusing Nile Red (red channel). Co-staining the cells for perilipin-2 demonstrates that OGA-S–GFP colocalizes with this resident lipid droplet protein. The degree ofcolocalization is shown in the merged panel. A higher magnification of the area denoted by the arrows is shown in the insets. (B)Oleic acid increases the amount ofOGA activity. OGA activity was measured using the fluorogenic substrate, FD-GlcNAc, in cells overexpressing OGA-L–GFP or OGA-S–GFP and treated with andwithout 0.4 mM oleic acid. OGA-S–GFP activity increased approximately 30% in the presence of 0.4 mM oleic acid, whereas OGA-L–GFP activity increased by6% with oleic acid treatment. Data shown are the means of duplicate samples and are representative of two separate experiments.

Jour

nal o

f Cel

l Sci

ence

in OGA-S targeting. The PAT proteins (perilipin, ADRP, TIP47;the last two are alternative names for other perilipins) are knownas bona fide resident proteins at the lipid droplet surface and areessential for maintaining lipid stores (Londos et al., 1999;Brasaemle, 2007). Pre-adipocytes contain small lipid droplets thatare enriched with perilipin-2. HA-OGA-S colocalized with theperilipin-2 in these nascent lipid droplets (Fig. 4B, merged panel).As the adipocytes differentiate, the lipid droplets increase in sizeand perilipin-2 is replaced with perilipin-1 (Bickel et al., 2009).Localization of OGA-S–GFP was examined in mature adipocytes.In differentiated 3T3 L1 adipocytes, the lipid droplets becamemarkedly larger and completely coated with perilipin-1 (Fig. 4C,red channel). OGA-S–GFP localization, however, remainedrestricted to the smaller diameter lipid droplets (Fig. 4C, greenchannel; arrows) and was markedly reduced at the larger dropletscontaining increased amounts of perilipin-1 (Fig. 4C, red channel).There was very little colocalization between OGA-S–GFP andperilipin-1 (Fig. 4C, merged panel). These data suggested thatOGA-S associates with nascent, but not mature, perilipin-1-coatedlipid droplets indicating that it might play a role in the earlierstages of the adipocyte differentiation.

Isoform-specific knockdown of OGA by siRNAGiven the unique targeting characteristics of the OGA isoformsand the potential interaction with lipid droplets, we examined thefunctional specificity of the endogenous OGA isoforms. UniquesiRNAs were designed to selectively downregulate each of theisoforms in HeLa and HEK-293 cells (see supplementary materialTable S1 for oligonucleotide sequences; oligos). The selectivity ofthe isoform-specific siRNAs was confirmed by several techniques.First, cells transfected with either OGA-L–GFP or OGA-S–GFPwere treated with the isoform-specific siRNA oligos, and controlswere treated with scrambled siRNA. Incubating transfected cellswith oligos specific for either OGA-L–GFP or OGA-S–GFPresulted in dramatic reduction in GFP fluorescence for eachexpression vector when compared with scrambled siRNAtreatments (Fig. 5A, compare right and left panels). Second,untransfected HeLa cells were treated for 48 hours with oligosselective for each OGA isoform. Semi-quantitative PCR was usedto confirm mRNA knockdown of endogenous transcripts in (Fig.5B). Transcripts for OGA-L were diminished by 50% and for

OGA-S by >60%. In addition, the fluorogenic substrate FD-GlcNAcwas used to measure total O-GlcNAcase activity in lysates fromcells treated with the specific oligos. Downregulation of OGA-Lresulted in a lowering of overall OGA activity to 33% of controlcells, whereas 77% of total activity remained after downregulationof OGA-S (Fig. 5C). Modulation of O-GlcNAcase activity bydownregulation of each isoform is reflected also in modified totalO-GlcNAc protein levels. RNAi of OGA-L resulted in a 3-foldincrease in total protein O-GlcNAcylation, whereas RNAi of OGA-S only modestly altered O-GlcNAc levels (1.4-fold; Fig. 5D,graph). Treatment with both siRNAs raised O-GlcNAc levels in anadditive manner compared with that of the siRNA specific for asingle isoform alone. Taken together, these data suggest that OGA-L is responsible for roughly two-thirds of the cellular O-GlcNAcaseactivity with OGA-S contributing approximately one-third.

O-GlcNAcase activity regulates proteasome functionAlthough OGA-S appears to contribute modestly to global O-GlcNAc cycling, we sought to determine if OGA-S might havespecific targets. The proteasome presents an ideal target for severalreasons: the activity of the 26S subunit of the proteasome isinhibited by O-GlcNAc modification (Zhang et al., 2007) andOGA-S is targeted specifically to the lipid droplet surface whereproteasomal activity is required for lipid droplet remodeling (Ohsakiet al., 2006).

To identify the relative contribution of the OGA isoforms inmodulating proteasomal functions, we analyzed both live cells andcell lysates for proteasome activity. HEK-293 ZsGreen ProteasomeSensor cells were used as our live cell model. These cells stablyexpress a GFP fusion protein that is targeted for 26S proteasomaldegradation allowing quantitative assay of proteasomal perturbationby following the accumulation of the GFP fusion protein. Undercontrol conditions little fluorescence was detected because theGFP fusion protein was rapidly and continuously degraded (Fig.6A). However, targeting either the long form or the short form ofOGA for downregulation by RNAi resulted in a marked increasein steady-state GFP fluorescence, indicating that the proteasomalactivity was compromised upon downregulation of OGA. As apositive control, cells were treated with the proteasome inhibitorALLN. The number of cells with GFP accumulation was directlyquantified by flow cytometry (Fig. 6B). The cells exhibited a shift

2854 Journal of Cell Science 124 (16)

Fig. 3. Lipid accumulation enhances endogenous OGA activity. (A)Schematic diagram of lipid augmentation. HeLa cells, either serum starved for 24 hours ortreated with 0.4 mM oleic acid, were assayed using FD-GlcNAc to measure the OGA activity in the total lysate from each condition shown plus the lipid fractionfrom the oleic-acid-supplemented cells. (B)Bar graph representing the measured OGA activity from each sample in arbitrary fluorescence units. Data shown arethe means of duplicate samples and are representative of three separate experiments.

Jour

nal o

f Cel

l Sci

ence

of the curve to the right, indicating a tenfold increase in GFP-positive cells when treated with OGA-S (red line) or OGA-L (blueline) siRNA compared with control siRNA (green line). We notedthat downregulation of OGA-S was equally effective at perturbingproteasomal activity, as was OGA-L. Maximum proteasomeinhibition by ALLN further right-shifted the curve.

Previous reports suggested that the O-GlcNAc modification hasa greater influence on the chymotrypsin-like activity of theproteasome (Zhang et al., 2003; Zhang et al., 2007). Bothchymotrypsin and trypsin-like activity of the proteasome weremeasured in HeLa and HEK-293 cell lysates. Isoform-specificknockdown of OGA was performed before testing for proteasome

activity in the lysates. Although chymotrypsin-like activity of theproteasome, as measured by the degradation of the syntheticfluorogenic peptide (Suc–LLVY–AMC), was reduced by depletingeither OGA isoform, statistical significance was reached only withdepletion of OGA-S (Fig. 6C, upper panel). Trypsin-like activitywas not significantly altered by downregulation of either isoform(Fig. 6C, lower panel). Interestingly, the knockdown of OGA-Swas equally effective (75% of control vs 83% of control) as OGA-L at perturbing proteasomal activity, which is in agreement withresults from proteasome sensor cells described above. Similarresults were obtained using HeLa cells (data not shown).Proteasome functional assays in two different cell models suggestthat both isoforms target proteasome activity.

OGA downregulation results in increasedpolyubiquitinylationProteasomal inhibition is normally associated with accumulationof polyubiquitinylation on proteins. As a further confirmation ofproteasomal inhibition by OGA, siRNA-treated cell lysates wereimmunoblotted with anti-ubiquitin to quantify global polyubiquitinlevels. HeLa cells treated with control siRNA showed little

2855O-GlcNAcase isoform targeted to lipid droplets

Fig. 4. 3T3 L1 adipocyte differentiation alters O-GlcNAcylated proteinprofiles and OGA-S localization. (A)O-GlcNAcylated protein profile in 3T3L1 adipocytes changed during differentiation. Total cell lysates were collectedfrom 3T3 L1 cells at the confluence stage (D2) and at 2-day intervals duringthe differentiation process. Equal amounts of protein (30mg/lane) from eachsample were resolved using a 4–12% Bis-Tris gel and immunoblotted with anO-GlcNAc specific antibody (CTD110.6). To test the specificity of the O-GlcNAc antibody, immunoreactivity was competed with excess N-acetylglucosamine (supplementary material Fig. S1). Lysates were probed forperilipin-2 as a positive control for differentiation. (B)In 3T3 L1 pre-adipocytes, HA–OGA-S (green channel) colocalized with nascent, perilipin-2-positive lipid droplets. (C)In differentiated adipocytes, OGA-S–GFP (green)associated with the smaller lipid droplets and did not colocalize with the largerlipid droplets completely coated with perilipin-1 (red). Insets show a highermagnification of the area denoted by the arrows.

Fig. 5. Selective downregulation of the O-GlcNAcase isoforms usingsiRNAs. (A)RNAi inhibition of OGA-L–GFP and OGA-S–GFP expression.OGA-L–GFP or OGA-S–GFP was expressed in HeLa cells and subsequentlyco-transfected with scrambled dsRNA (control) or isoform-specific siRNAs.Cells were fixed and analyzed by fluorescence microscopy. siRNA treatmentreduced the expression of each isoform as shown by a reduction in GFPexpression. GFP expression is shown in green and DAPI-stained nuclei areshown in blue. (B)RNAi reduced mRNA transcription. Semi-quantitative RT-PCR for endogenous OGA-L and OGA-S was performed using total RNAisolated from HeLa cells treated for 48 hours with scrambled dsRNA (con) andsiRNAs specific for long (sirL), short (sirS) or both (sirS/L) O-GlcNAcaseisoforms. GAPDH transcript levels are shown as a control. (C)OGAenzymatic activity is reduced with siRNA treatment (con, scrambled siRNA;sirL, siRNA for OGA-L; sirS, siRNA for OGA-S). OGA activity was assayedusing the fluorogenic substrate (values are the percentage activity normalizedto control ± s.e.m.). (D)Decreasing OGA levels by siRNA treatment increasedthe total amount of O-GlcNAc-modification in the HeLa cell lysate as shownby western blot analysis using an O-GlcNAc-specific antibody. Below is a bargraph showing the cumulative band intensity (percentage of the control ±s.e.m.) from two different experiments.

Jour

nal o

f Cel

l Sci

ence

accumulation of polyubiquitinylated proteins (Fig. 7A, con).However, by selectively lowering OGA-L with siRNA treatment,a threefold increase in cumulative levels of polyubiquitinylationwas observed (Fig. 7B, sirL). Ubiquitinylated protein accumulationwas fivefold higher with OGA-S (Fig. 7B, sirS). The characteristicimmunoreactive ubiquitin smear is shown with a longer exposure(Fig. 7A). Downregulation of either OGA isoform inhibitsproteasomal activity as measured by a synthetic sensor (Fig. 6A,ZsProSonsor1) or accumulation of polyubiquitin (Fig. 7). Thesedata suggest that the lipid-droplet-targeted OGA-S could influenceproteasome-mediated protein remodeling of the lipid droplet surface(see below).

Complex interplay of the proteasome and OGA on lipid-droplet-associated proteinsExpression of OGA-S fusion proteins was always markedly lowerthan that of OGA-L in our experiments, regardless of the tag used(HA or GFP) or the location of the tag (i.e. N- or C-terminal endof the enzyme). Typically, only 20–30% of the cells showed OGA-S expression, whereas tagged versions of OGA-L were robustlyexpressed within the monolayer. This difference in expressionlevels was confirmed by western blot analysis (Fig. 1B) andmeasured total O-GlcNAcase activity (Fig. 1D). Because rapiddegradation of OGA-S could cause the low expression level, weexamined how proteasome activity could alter the stability of the

2856 Journal of Cell Science 124 (16)

Fig. 6. Knockdown of the O-GlcNAcase isoforms decreases proteasome activity in cells and cell lysates. (A)ZsGreen Proteasome Sensor cells were treatedwith scrambled dsRNA (Con), or dsRNAs specific for OGA-L (sirL) or OGA-S (sirS) for 48 hours. ZsGreen Proteasome Sensor cells express a GFP reporterprotein (ZsProSensor-1) targeted for degradation; accumulation of GFP fluorescence indicates impaired proteasome activity. Downregulation of either OGAisoform increased the number of GFP-positive cells (top row, green channel), indicating reduced proteasome activity. DAPI-stained nuclei are shown (bottom row,blue). As a positive control, the proteasome inhibitor ALLN was used at 5mg/ml for 12 hours. (B)The accumulation of ZsProSensor-1 protein was analyzedquantitatively using FACS analysis. Histogram plots show differences in fluorescence intensity between cells treated with scrambled dsRNA (green line), ALLN(black line), siRNA for long (blue line) and short (red line) isoforms. (C)OGA downregulation modulates proteasomal activities. The proteasome peptidaseactivities in HEK-293 cell lysates were measured by fluorescence from the cleavage of Suc–LLVY–AMC (chymotrypsin-like activity) and Boc–LSTR–AMC(trypsin-like activity). The proteasome activity was measured 48 hours after transfection with the indicated siRNAs. Treatment with ALLN (5mg/ml) for 12 hoursis shown for comparison. Values are the mean of triplicate samples (± s.e.m.). *Statistical difference compared with control or sirL (P<0.05, paired Student’s t-test).

Fig. 7. OGA downregulation results in an accumulationof polyubiquitinyled proteins. (A)Western blot analysiswas performed using anti-ubiquitin antibodies as describedin the Materials and Methods. Total cellular lysates fromcells treated with scrambled siRNA (lanes 1and 2, con),siRNA for OGA-L (lanes 3 and 4, sirL) and siRNA forOGA-S (lanes 5 and 6, sirS) are shown. A longer exposureof lanes 5 and 6 is shown on the right to demonstrate thecharacteristic ubiquitin immunoreactive smear. The amountof actin is shown as a loading control. (B)Bar graph oftotal band intensity relative to actin. Values are meanpercentages ± s.e.m.

Jour

nal o

f Cel

l Sci

ence

OGA isoforms. Proteasome inhibition by ALLN resulted in a dose-dependent accumulation of both isoforms, but had the largestimpact on OGA-S accumulation (Fig. 8A). Without ALLN, OGA-S–GFP was nearly undetectable, compared with OGA-L–GFP (Fig.8A and Fig. 1B). Measurement of the integrated intensity of thewestern blots showed that the expression of OGA-S–GFP increasedapproximately 10- and 14-fold with 25 and 50 mM ALLN treatment,respectively, compared with 3- and 5-fold for OGA-L–GFP,suggesting that OGA-S levels are tightly regulated by theproteasome. Similar results were obtained with the HA-taggedenzymes (data not shown). As a control for ALLN action, thestabilization of perilipin-2 is shown (Fig. 8A, bottom row).

If OGA-S is selectively acting on the proteasome at lipiddroplets, then downregulation of OGA-S, but not OGA-L shouldperturb proteasome function at the lipid droplet surface and resultin an increase in lipid-droplet-associated proteins. To test thishypothesis, we selectively downregulated each isoform andmeasured the accumulation of perilipin-2 and perilipin-3 (alsoknown as TIP47; a related lipid-droplet-associated protein). OGA-

S-specific RNAi treatment stabilized perilipin-2 levels (Fig. 8B,upper panels). Such stabilization was not observed when OGA-Lwas downregulated. Quantitatively similar results were obtainedwith perilipin-3 (Fig. 8B, lower panels). Our results suggest that ahighly regulated feedback control of OGA-S might modulate lipid-droplet protein stability (Fig. 8C). In this model, proteasome activitytightly regulates the levels of OGA-S and its associated O-GlcNAcase activity, present at the surface of the lipid droplet. O-GlcNAc cycling is also known to regulate proteasome activity byinhibiting the Rtp2 domain of the 26S proteasome (Zhang et al.,2007). Therefore, OGA-S activity, acting through proteasomeactivation at the surface of the lipid droplet, is ideally positionedto modulate the stability of other lipid-droplet-associated proteins.This regulation might serve to control the size and surfacecomposition of the maturing lipid droplet.

DiscussionO-GlcNAc cycling on a multitude of diverse substrates is carriedout by a single set of opposing enzymes (OGT and OGA). A keyquestion concerning O-GlcNAc cycling is how a single set ofenzymes can regulate a multitude of substrates in a coordinatedmanner. We speculate that isoform specificity or enzymecompartmentalization could play a role in regulating multiplesubstrates independently. For OGT, differentially spliced isoformswere described and shown to localize at different cellularcompartments (Love et al., 2003). For the O-GlcNAcase two majorisoforms have been reported, but the detailed properties of theshorter isoform were not studied (Comtesse et al., 2001).Knowledge of isoform-specific effects of OGA are important indeveloping inhibitors as potential drug targets in metabolic diseases.Therefore, we have examined the localization and distinct activitiesof the O-GlcNAcase isoforms.

Results demonstrate that both isoforms are active and aretranscribed in numerous cell types, although the protein stabilitymight affect the relative abundance of each isoform. We furtherdemonstrate that the isoforms are differentially localized. Thelonger isoform resides in the nucleus and cytoplasm, whereas theshorter isoform is found concentrated at the surface of lipid droplets.The increase of lipid stores correlates positively with OGA activity,and isolated lipid droplets contain a major fraction of total OGAactivity. Finally, we show that the two isoforms differentiallymodulate proteasome activity at the lipid droplet surface. Becauseproteasome activity is essential for the remodeling and assemblyof lipid droplets, we propose a role for the shorter OGA isoform inregulating lipid droplet assembly and the mobilization of neutrallipids.

OGA isoforms exhibit unique catalytic propertiesThe relative catalytic efficiency of the two major O-GlcNAcaseisoforms has recently come under scrutiny (Macauley andVocadlo, 2009; Kim et al., 2007). Upon overexpression, OGA-Sand OGA-L are each capable of dramatically lowering theamounts of modified O-GlcNAc proteins at the single cell level,as shown by reduced staining with anti O-GlcNAc antibody (Fig.1B). The number of cells with detectable OGA-S protein,however, was greatly reduced compared with cells transfectedwith OGA-L. Consistent with this result, OGA activity in siRNA-treated total cell lysates indicated that OGA-S makes only asmall contribution to the total pool of O-GlcNAcase activity inHeLa cells. In addition to lower expression levels, the Km forOGA-S is approximately 20-fold lower than that for OGA-L

2857O-GlcNAcase isoform targeted to lipid droplets

Fig. 8. OGA-S–GFP expression is stabilized by proteasome inactivationand modulates Perilipin-2 and perilipin-3 levels in HeLa cells. (A)Westernblot analysis showing the stabilization of OGA-S–GFP, OGA-L–GFP andperilipin-2 with proteasome inhibition. Whole cell lysates from ALLN-treated(25 and 50mg/ml) cells expressing either OGA-S–GFP or OGA-L–GFP wereprobed with an anti-GFP to identify the fusion protein. The amount of OGA-S–GFP increased by 10- (25mg/ml ALLN) and 14-fold (25mg/ml ALLN)compared with the control, whereas OGA-L–GFP increased by 3- (25mg/mlALLN) and 5-fold (25mg/ml ALLN) over the control. An increase in perilipin-2 is shown as a positive control for proteasome inhibition. The amount of actinis shown as a loading control. (B)Depletion of OGA-S, but not OGA-Lstabilizes perilipin-2 and -3. Western blot analysis showing siRNA-mediatedOGA-S downregulation (sir-S) selectively stabilized perilipin-2 and perilipin-3expression in HeLa cells but OGA-L siRNA (sir-L) had no effect. Treatmentwith scrambled siRNA (con) is shown as a negative control. (C)Modelshowing the possible regulatory role of OGA-S on lipid droplet surface PATproteins. As depicted, OGT inhibits the proteasome by O-GlcNAcmodification of the Rpt2 domain whereas OGA-S activates the proteasome byremoving the modification. Activated proteasomes in turn facilitate thedegradation of lipid droplet surface proteins (PAT) and also create a negativefeedback loop by degrading OGA-S to limit its availability in a mannerdependent upon lipid stores.

Jour

nal o

f Cel

l Sci

ence

(Kim et al., 2006). The data reported in this manuscript are inagreement with the several-fold lower activity of OGA-Scompared with OGA-L. Macauley and Vocadlo also reportedlower efficiency of O-GlcNAcase activity for OGA-S (OGA-NV) compared with OGA-L (Macauley and Vocadlo, 2009). Theknown difference of the C-terminal tail of the two isoforms (longisoform having a ‘HAT-like’ domain and the short form a unique14 amino acid tail) might offer a basis for the difference incatalytic behavior at the N-terminus of the protein.

OGA isoforms might have unique intracellular targetsThe distinctive intracellular localization of the OGA isoformssuggests that they have specific functional targets. OGA-L–GFPexpression is, in both nucleus and cytoplasm, consistent with itsactivity in many nuclear and cytoplasmic proteins. OGA-S–GFP,however, preferentially accumulates at the surface of lipid droplets.Because OGA-S is colocalized with perilipin-2 in HeLa cells and3T3 L1 adipocytes, the short isoform of O-GlcNAcase could playa distinct role in modulating the function of lipid-droplet-associatedproteins. Perilipin-2 is one of the PAT proteins on the lipid dropletsurface actively involved in lipid droplet formation (Bickel et al.,2009; Brasaemle, 2007; Londos et al., 1999). Oleic acid is knownto induce synthesis of perilipin-2, initiating nascent lipid droplets,whereas it is gradually replaced by perilipin-1 during lipid dropletmaturation in adipocytes (Bickel et al., 2009). This transition fromperilipin-2 to perilipin-1 is regulated by a proteasome-sensitivemechanism (Xu et al., 2005; Xu et al., 2006; Masuda et al., 2006).Interestingly, OGA-S demonstrates a robust effect on proteasomalactivity and we have pursued the hypothesis that OGA-S couldregulate perilipin-2 stability by activating proteasomes located atthe lipid droplet surface.

OGA isoforms differentially regulate the proteasomeThe 26S proteasome is a well-established target of O-GlcNAccycling, as documented in several studies. O-GlcNAc modificationof the proteasome in neuronal cells has been reported to inhibitproteasomal activity (Zhang et al., 2007; Zhang et al., 2003; Liu etal., 2004). O-GlcNAc modification of the Rpt2 ATPase located atthe 19S proteasome regulatory cap inhibited the access ofubiquitinylated proteins to the proteasomal core for degradation(Zhang et al., 2007). Here, we provide evidence that O-GlcNAcasemodulates proteasome activity in living cells, using isoform-specificsiRNA treatment. Although OGA-S appears to be a minorcontributor to total O-GlcNAcase activity, downregulation of thisisoform had a similar impact on proteasome activity as diddownregulation of OGA-L. Additionally, increased accumulationof ubiquitinylated proteins upon OGA-S siRNA treatment suggeststhat the proteasome is a preferred OGA-S target.

The chymotrypsin-like, trypsin-like and caspase-like activitiesare among the several different catalytic functions of proteasomes.Our in vitro assays demonstrated that OGA isoform downregulationhas a more dramatic effect on the chymotrypsin-like activity thantrypsin-like activity. This corroborates the observations of Zhanget al. showing that cleavage of the LLVY (chymotrypsin-likeactivity) peptide was altered by OGT, whereas there was no effectobserved on a tryptic peptide LSTR (Zhang et al., 2003). Thepresent findings suggest that O-GlcNAcylation might not inhibitglobal proteasome function, but could alter activity towards certainspecific targets. Similarly, specific inhibition of chymotrypsin-likeactivity of the proteasome has been reported with long chain fattyacids (Hamel, 2009).

OGA-S functions at the surface of maturing lipid dropletsGiven the localization of OGA-S on the surface of lipid dropletsand its dramatic effect on proteasome function, OGA-S appears toplay a key role in lipid droplet metabolism. The proteasome isknown to play a key role in the remodeling of the lipid dropletsurface. Perilipin-2 and perilipin-1 are ultimately degraded by theubiquitin–proteasome pathway (Xu et al., 2005; Xu et al., 2006).Ubiquitin ligases, necessary for proteasomal function, are targetedto the lipid droplet surface and influence lipid droplet size andnumber (Eastman et al., 2009). Additionally, protein degradationby the proteasome-ubiquitin pathway and autophagy converge atthe lipid droplet surface to regulate lipid metabolism (Ohsaki et al.,2006; Singh et al., 2009). Because these two catabolic pathwaysare interdependent and act at the surface of the lipid droplet, OGA-S activity and O-GlcNAc cycling probably influences bothpathways to regulate lipid homeostasis.

Our data suggest a feedback mechanism between the proteasomaldegradation pathway and O-GlcNAcase activity at the surface ofthe lipid droplet. Proteasome activity is modulated by the action ofOGA-S, whereas relative abundance of the O-GlcNAc isoforms ismaintained through proteasomal degradation. Thus OGA-S andthe proteasome regulate each other, suggesting a feedbackmechanism often observed with nutritionally responsive, metabolicregulation. The localization of OGA-S to nascent lipid droplets co-occupied by perilipin-2 and the ability of OGA-S to regulateproteasome activity suggests a complex interplay between thepresence of lipid stores and the relative abundance of the OGA-Sisoform. As such, OGA-S, perilipin-2, oleic acid and the proteasomemight all participate in the intricate mechanism required to maintaincellular lipid stores. This notion is consistent with previous findingson lipid metabolism. C. elegans mutants with either ogt-1 or oga-1 null alleles were shown to have decreased lipid stores andincreased dauer formation associated with defective O-GlcNAccycling (Hanover et al., 2005; Forsythe et al., 2006). Luo et al.have suggested that increased O-GlcNAc flux activates AMPKand stimulates fatty acid oxidation (Luo et al., 2007).

Summary and conclusionThe differently targeted isoforms of OGA analyzed in this studyare likely to perform distinct intracellular functions. The nuclearand cytoplasmic OGA-L contains a HAT domain, suggesting itparticipates in chromatin-related regulation of gene expression. Bycontrast, the lipid-droplet-associated OGA-S is likely to play a rolein regulating the process of lipid storage and mobilization. Bymodulating the local activity of proteasomes associated with lipiddroplets, OGA-S might actively participate in the perilipin-2regulation and remodeling of the protein on the lipid droplet surface(Fig. 8C). The results of this study strongly suggest that OGA-Smight be a key enzyme in maintaining the delicate balance betweencarbohydrate and lipid metabolism.

Materials and MethodsChemicals and antibodiesThe proteasome inhibitor Ac-leucine-leucine-norleucinal (ALLN; calpain inhibitorI) and proteasome substrates Suc–LLVY–AMC and Boc–LSTR–AMC werepurchased from Sigma-Aldrich. Primary antibodies RL2 (cat no. MAI-072) andperilipin A (perilipin-1; cat no. PAI-1051) were from Affinity Bioreagents; CTD110.6from Covance (cat no. MMS-248R); and ubiquitin from EBiosciences. Antibodiesto mouse perilipin-2, perilipin-1 and perilipin-3 were gifts from the late DrConstantine Londos (NIDDK, NIH). Pan actin, GFP and HA antibodies were fromCell Signaling Technology. Mouse and rabbit IgG and IgM secondary antibodiesconjugated with IRDye 800, or IRDye700 were from Rockland Immunochemicals(Rockland, PA).

2858 Journal of Cell Science 124 (16)

Jour

nal o

f Cel

l Sci

ence

Plasmid and siRNAA cassette containing either human OGA long (NM_012215) or short isoform(AF307332) was incorporated into the pEGFP-N1 vector (Clontech) between theKpnI and PinAI sites. OGA-L–GFP and OGA-S–GFP clones have GFP fused at theC-terminus. An adenovirus vector was constructed using the same plasmid construct(adenoviral-type 5 (dE1/E3, Vector Biolabs). Two other OGA constructs were madewith HA-tag fusions using the pCMV-HA vector (Clontech). HA-OGA-L and HA-OGA-S were generated with the HA tag at the N-terminus of the OGA isoforms byinserting the OGA sequence between the BglII and KpnI sites, in frame with the HAtag.

Specific siRNA was designed, using web-based software, to silence OGA-L andOGA-S isoforms. Selected 20-mers, were purchased from Qiagen as annealeddouble-strand DNA. Initial experiments were conducted to select siRNA thatefficiently targeted OGA isoforms, and to determine the duration of the treatments.Selected sequences of siRNA specific for human OGA-L and OGA-S are shown insupplementary material Table S1.

Cell cultureHeLa cells (ATCC) and HEK-293 ZsGreen Proteasome Sensor cells (BD Bioscience,cat. no 631535) were grown in Dulbecco’s modified Eagle’s medium (DMEM) with10% fetal bovine serum and 1% of streptavidin and penicillin. 3T3 L1 pre-adipocyteswere obtained from ATCC and differentiated into adipocytes according to a standardprotocol (Student et al., 1980) Briefly, confluent cells were treated with 3-isobutyl-1-methylxanthine (IBMX), dexamethasone and insulin combined medium for 2days. Insulin was added to the medium for an additional 2 days, and then themedium was changed to regular growth medium. Mature adipocytes were used fortransfections. Lipid droplets were isolated as described previously (Brasaemle andWolins, 2006).

Plasmid and siRNA treatmentsHeLa and HEK-293 cells were seeded in six-well plates or in two-well chamberslides for 24 hours before transfection. Cells were transfected with 1–2 mg ofplasmid DNA using LipofectamineTM 2000 (Invitrogen) according to themanufacturer’s instructions.

A mixture of equal molar amounts of siRNA was used for each siRNA target(OGA-L or OGA-S) at a final concentration of 2 mg/ml with HeLa cells and 1 mg/mlwith HEK-293 cells. Cells were transfected with siRNAs using RNAiFect reagent(Qiagen) according to the manufacturer’s instructions. Medium was changed 24hours after transfection and experiments were initiated at 48 hours. These timeframes and DNA concentrations were selected from preliminary experiments.

To observe the siRNA-mediated downregulation, HeLa cells were grown on glasscoverslips and first transfected with OGA-L–GFP or OGA-S–GFP, followed by therespective siRNA transfection for 48 hours. Cells were fixed with 4% formaldehydein BSA and examined by fluorescence microscopy to detect the downregulation ofGFP-fused protein expression.

Cellular localization of OGA isoforms by confocal microscopyDifferentiated 3T3 L1 mouse adipocytes were cultured in two-well chamber slidesor on glass coverslips and infected with adenoviral vectors encoding OGA-L–GFPor OGA-S–GFP at an MOI of 50. HeLa cell culture medium was augmented with0.4 mM oleic acid (Sigma) to form the lipid droplets. The cells were fixed andstained with perilipin-2, perilipin-1 or LipidTOX (a neutral lipid-specific dyeconjugated with Alexa Fluor 568; Invitrogen) and DAPI for nuclear staining. Cellswere imaged using a Zeiss Axiovert 200M confocal microscope and processed withVolocity software (Perkin Elmer).

OGA mRNA expressionTotal RNA was isolated from HeLa cells using a RNeasy Mini Kit (Qiagen). cDNAwas synthesized from 1 mg of total RNA with Omniscript reverse transcriptase(Qiagen) in 20 ml reaction volume according to the enzyme supplier’s instructions.Semi-quantitative PCR was performed in a 50 ml reaction mixture containing 5 mlcDNA, and Taq DNA polymerase (TaKaRa). The following primer sequence wasused for amplifying OGA-L sense 5�-AATTGAAGAATGGCGGTCAC-3� andantisense: 5�-CTCTAAAGGCCCAGGGTTCT-3�, and for OGA-S sense: 5�-CCCTGGAGGATTTGCAGTTA-3� and antisense: 5�-CCTGGTGCAC -CTACCTAACC-3�; GAPDH served as a control. The reaction mixture was denaturedat 94°C for 5 minutes and amplified by 35 cycles under the following conditions:denaturing for 30 seconds at 94°C, annealing for 30 seconds at 60°C andpolymerization for 30 seconds at 72°C, followed by a single extension for 5 minutesat 72°C.

Western blot analysisProteins were collected using m-PER buffer (Pierce–Thermo Fisher Scientific)containing a protease inhibitor cocktail and phosphatase inhibitors (Roche AppliedSciences), according to the manufacture’s protocol. Cell extracts were sonicatedthree times for 10 seconds each in an ice-cold water bath, centrifuged at 11,000 gfor 10 minutes and supernatant was collected. Protein was quantified by the BCA(Pierce–Thermo Fisher Scientific) method. An equal amount of each protein (30–50mg) was resolved using a 10% NUPAGE gel, and western blot analysis was performed

with antibodies specific for O-GlcNAcylated protein (RL2 or CTD 110.6). To studythe specificity of the antibody, half of the blot was incubated with 20 mM O-GlcNAcsolution together with the CTD110.6 antibody, followed by secondary antibody.Rabbit polyclonal antibody raised against OGA (MGEA5) was obtained fromProteinTech Group (Chicago, IL). Intensity of the antibody-bound proteins wasdetected using IR dye secondary antibodies, and scanned with an Odyssey infraredimager (LI-COR Biosciences). The band intensity was calculated using Odysseysoftware and was normalized to actin expression levels.

O-GlcNAcase activityO-GlcNAcase activity of HeLa cell lysates, transfected with different plasmids(OGA-L–GFP, OGA-S–GFP) and siRNA treatments (control, OGA-L- and OGA-S-specific siRNAs) were assayed using fluorescently labeled FD-GlcNAc substrate (1mM) as described by Kim et al. (Kim et al., 2006). Cell lysates were prepared in 20mM Tris-HCl at pH 7.5 with protease inhibitors. The assay was conducted in 0.5 Mcitrate phosphate pH 6.5 at 37°C for 30 minutes in the presences of 500 mM GalNActo saturate the GalNAcase activity. Fluorescence was read at 435 nm excitation and535 nm emission using a Victor2 multi-label counter (Perkin Elmer).

Live cell proteasome activity using ZsGreen Proteasome Sensor cellsTo assay proteasome function in living HEK-293 cells, ZsGreen Proteasome Sensorcells (BD Biosciences) were treated with OGA-specific RNAi. The fluorescentprotein Zs-ProSensor-1 is a fusion protein consisting of ZsGreen (a green fluorescentprotein from Zoanthus sp., a reef coral) and amino acids 410–461 of the mouseornithine decarboxylase (MODC). This chimeric protein is constitutively degradedby the proteasome in stably transfected HEK-293 cells and accumulates GFP underconditions that alter the activity of the proteasome. HEK-293 ZsGreen ProteasomeSensor cells were treated with siRNAs for the O-GlcNAcase isoforms, and controlsiRNA. ALLN is a well-characterized inhibitor of proteasome-dependent proteolysisand was used as the reference control. The green fluorescence of the chimericprotein was monitored by microscopy using an Axiovert 200M microscope;micrographs were taken with the same exposure times for all the samples, with GFPfilters.

Fluorescence-activated cell sorting (FACS) analysis was performed to quantifyGFP expression, which correlated with the quantitative measure of proteasomeinactivation. After transfection for 48 hours with siRNAs, HEK-293 ZsGreenProteasome Sensor cells were trypsinized and centrifuged for 5 minutes at 200 g at4°C. The cell pellet was washed with PBS and cells were fixed with 4% formaldehydeand resuspended in PBS. Samples were analyzed with a FACS Calibur flow cytometer(Becton-Dickinson) and data were analyzed using CellQuest 3.1 software (Becton-Dickinson). For each sample, 10,000 events were collected.

Proteasome peptidase activity assayProteasome activity was assayed in lysates of HEK-293 and HeLa cells usingpeptide substrates linked to the fluorogenic reporter aminomethylcoumarin (Suc–LLVY–AMC and Boc–LSTR–AMC; Sigma). OGA siRNA-transfected cells culturedin 12-well plates were collected, washed in PBS, lysed in ice-cold buffer A (50 mMTris-HCl pH 7.5, 100 mM NaCl, 5 mM EDTA, 1% Triton X-100) for 30 minutesand subjected to sonication three times for 15 seconds each. Supernatant was isolatedafter centrifugation and the protein quantified using the BCA method before analysisof proteasome activity. Aliquots of lysates (50 ml) were suspended in 50 ml of bufferB (50 mM Tris-HCl pH 7.5, 100 mM NaCl, 5 mM EDTA, 4 mM DTT) the reactionwas started by adding 100 ml of 100 mM substrate and continued for 90 minutes at37°C. The reaction was stopped with 800 ml of 1% SDS. Fluorescence intensity wasmeasured (excitation 360 nm, emission 450 nm) using a F2000 fluorescencespectrophotometer (Hitachi). Enzymatic activity was normalized for proteinconcentration and expressed as percent activity of the lysate treated with scrambledsiRNA. Each measurement was carried out using at least three independent dsRNAand plasmid transfections.

Proteasome inhibitor assayHeLa cells were cultured in six-well plates and transfected with long and short formsof GFP-tagged or HA-tagged plasmids. After 24 hours, ALLN was added at 25 or50 mM. This experiment was performed with and without 0.4 mM oleic acidtreatment for 24 hours. Cells lysates were extracted as described above, and westernimmunoblot analysis performed with antibodies to GFP or HA and to perilipin-2.

Statistical analysisExperiments were carried out in duplicate or triplicate and the data collected werenormalized to the control. Paired Student’s t-tests were used to test for statisticalsignificance.

We are grateful to the Londos laboratory for providing us withantibodies to perilipins 1, 2 and 3, and assistance with the lipid dropletisolation. This research was supported by the Intramural ResearchProgram of the National Institutes of Health, National Institute forDiabetes and Digestive and Kidney Diseases. Deposited in PMC forrelease after 12 months.

2859O-GlcNAcase isoform targeted to lipid droplets

Jour

nal o

f Cel

l Sci

ence

Supplementary material available online athttp://jcs.biologists.org/cgi/content/full/124/16/2851/DC1

ReferencesBickel, P. E., Tansey, J. T. and Welte, M. A. (2009). PAT proteins, an ancient family of

lipid droplet proteins that regulate cellular lipid stores. Biochim. Biophys. Acta 1791,419-440.

Brasaemle, D. L. (2007). Thematic review series: adipocyte biology. The perilipin familyof structural lipid droplet proteins: stabilization of lipid droplets and control of lipolysis.J. Lipid Res. 48, 2547-2559.

Brasaemle, D. L. and Wolins, N. E. (2006). Isolation of lipid droplets from cells bydensity gradient centrifugation. Curr. Protoc. Cell Biol. Chapter 3, Unit 3.15.

Butkinaree, C., Cheung, W. D., Park, S., Park, K., Barber, M. and Hart, G. W. (2008).Characterization of beta-N-acetylglucosaminidase cleavage by caspase-3 duringapoptosis. J. Biol. Chem. 283, 23557-23566.

Comtesse, N., Maldener, E. and Meese, E. (2001). Identification of a nuclear variant ofMGEA5, a cytoplasmic hyaluronidase and a beta-N-acetylglucosaminidase. Biochem.Biophys. Res. Commun. 283, 634-640.

Dennis, R. J., Taylor, E. J., Macauley, M. S., Stubbs, K. A., Turkenburg, J. P., Hart,S. J., Black, G. N., Vocadlo, D. J. and Davies, G. J. (2006). Structure and mechanismof a bacterial beta-glucosaminidase having O-GlcNAcase activity. Nat. Struct. Mol.Biol. 13, 365-371.

Duggirala, R., Blangero, J., Almasy, L., Dyer, T. D., Williams, K. L., Leach, R. J.,O’Connell, P. and Stern, M. P. (1999). Linkage of type 2 diabetes mellitus and of ageat onset to a genetic location on chromosome 10q in Mexican Americans. Am. J. Hum.Genet. 64, 1127-1140.

Eastman, S. W., Yassaee, M. and Bieniasz, P. D. (2009). A role for ubiquitin ligases andSpartin/SPG20 in lipid droplet turnover. J. Cell Biol. 184, 881-894.

Forsythe, M. E., Love, D. C., Lazarus, B. D., Kim, E. J., Prinz, W. A., Ashwell, G.,Krause, M. W. and Hanover, J. A. (2006). Caenorhabditis elegans ortholog of adiabetes susceptibility locus: oga-1 (O-GlcNAcase) knockout impacts O-GlcNAccycling, metabolism, and dauer. Proc. Natl. Acad. Sci. USA 103, 11952-11957.

Gao, Y., Wells, L., Comer, F. I., Parker, G. J. and Hart, G. W. (2001). Dynamic O-glycosylation of nuclear and cytosolic proteins: cloning and characterization of aneutral, cytosolic beta-N-acetylglucosaminidase from human brain. J. Biol. Chem. 276,9838-9845.

Hamel, F. G. (2009). Preliminary report: inhibition of cellular proteasome activity by freefatty acids. Metabolism 58, 1047-1049.

Hanover, J. A., Forsythe, M. E., Hennessey, P. T., Brodigan, T. M., Love, D. C.,Ashwell, G. and Krause, M. (2005). A Caenorhabditis elegans model of insulinresistance: altered macronutrient storage and dauer formation in an OGT-1 knockout.Proc. Natl. Acad. Sci. USA 102, 11266-11271.

Heckel, D., Comtesse, N., Brass, N., Blin, N., Zang, K. D. and Meese, E. (1998). Novelimmunogenic antigen homologous to hyaluronidase in meningioma. Hum. Mol. Genet.7, 1859-1872.

Kim, E. J., Kang, D. O., Love, D. C. and Hanover, J. A. (2006). Enzymaticcharacterization of O-GlcNAcase isoforms using a fluorogenic GlcNAc substrate.Carbohydr. Res. 341, 971-982.

Kim, E. J., Amorelli, B., Abdo, M., Thomas, C. J., Love, D. C., Knapp, S. andHanover, J. A. (2007). Distinctive inhibition of O-GlcNAcase isoforms by an alpha-GlcNAc thiolsulfonate. J. Am. Chem. Soc. 129, 14854-14855.

Lazarus, B. D., Love, D. C. and Hanover, J. A. (2006). Recombinant O-GlcNActransferase isoforms: identification of O-GlcNAcase, yes tyrosine kinase, and tau asisoform-specific substrates. Glycobiology 16, 415-421.

Lehman, D. M., Fu, D. J., Freeman, A. B., Hunt, K. J., Leach, R. J., Johnson-Pais, T.,Hamlington, J., Dyer, T. D., Arya, R., Abboud, H. et al. (2005). A single nucleotidepolymorphism in MGEA5 encoding O-GlcNAc-selective N-acetyl-beta-Dglucosaminidase is associated with type 2 diabetes in Mexican Americans. Diabetes 54,1214-1221.

Liu, K., Paterson, A. J., Zhang, F., McAndrew, J., Fukuchi, K., Wyss, J. M., Peng, L.,Hu, Y. and Kudlow, J. E. (2004). Accumulation of protein O-GlcNAc modificationinhibits proteasomes in the brain and coincides with neuronal apoptosis in brain areaswith high O-GlcNAc metabolism. J. Neurochem. 89, 1044-1055.

Londos, C., Brasaemle, D. L., Schultz, C. J., Segrest, J. P. and Kimmel, A. R. (1999).Perilipins, ADRP, and other proteins that associate with intracellular neutral lipiddroplets in animal cells. Semin. Cell Dev. Biol. 10, 51-58.

Love, D. C. and Hanover, J. A. (2005). The hexosamine signaling pathway: decipheringthe “O-GlcNAc code”. Sci. STKE 2005, re13.

Love, D. C., Kochan, J., Cathey, R. L., Shin, S. H. and Hanover, J. A. (2003).Mitochondrial and nucleocytoplasmic targeting of O-linked GlcNAc transferase. J. CellSci. 116, 647-654.

Love, D. C., Ghosh, S., Mondoux, M. A., Fukushige, T., Wang, P., Wilson, M. A., Iser,W. B., Wolkow, C. A., Krause, M. W. and Hanover, J. A. (2010). Dynamic O-GlcNAc cycling at promoters of Caenorhabditis elegans genes regulating longevity,stress, and immunity. Proc. Natl. Acad. Sci. USA 107, 7413-7418.

Luo, B., Parker, G. J., Cooksey, R. C., Soesanto, Y., Evans, M., Jones, D. and McClain,D. A. (2007). Chronic hexosamine flux stimulates fatty acid oxidation by activatingAMP-activated protein kinase in adipocytes. J. Biol. Chem. 282, 7172-7180.

Macauley, M. S. and Vocadlo, D. J. (2009). Enzymatic characterization and inhibition ofthe nuclear variant of human O-GlcNAcase. Carbohydr. Res. 344, 1079-1084.

Marshall, S. (2006). Role of insulin, adipocyte hormones, and nutrient-sensing pathwaysin regulating fuel metabolism and energy homeostasis: a nutritional perspective ofdiabetes, obesity, and cancer. Sci. STKE 2006, re7.

Masuda, Y., Itabe, H., Odaki, M., Hama, K., Fujimoto, Y., Mori, M., Sasabe, N., Aoki,J., Arai, H. and Takano, T. (2006). ADRP/adipophilin is degraded through theproteasome-dependent pathway during regression of lipid-storing cells. J. Lipid Res. 47,87-98.

Ohsaki, Y., Cheng, J., Fujita, A., Tokumoto, T. and Fujimoto, T. (2006). Cytoplasmiclipid droplets are sites of convergence of proteasomal and autophagic degradation ofapolipoprotein B. Mol. Biol. Cell 17, 2674-2683.

Rao, F. V., Dorfmueller, H. C., Villa, F., Allwood, M., Eggleston, I. M. and van Aalten,D. M. (2006). Structural insights into the mechanism and inhibition of eukaryotic O-GlcNAc hydrolysis. EMBO J. 25, 1569-1578.

Schultz, J. and Pils, B. (2002). Prediction of structure and functional residues for O-GlcNAcase, a divergent homologue of acetyltransferases. FEBS Lett. 529, 179-182.

Singh, R., Xiang, Y., Wang, Y., Baikati, K., Cuervo, A. M., Luu, Y. K., Tang, Y., Pessin,J. E., Schwartz, G. J. and Czaja, M. J. (2009). Autophagy regulates adipose mass anddifferentiation in mice. J. Clin. Invest. 119, 3329-3339.

Slawson, C. and Hart, G. W. (2003). Dynamic interplay between O-GlcNAc and O-phosphate: the sweet side of protein regulation. Curr. Opin. Struct. Biol. 13, 631-636.

Student, A. K., Hsu, R. Y. and Lane, M. D. (1980). Induction of fatty acid synthetasesynthesis in differentiating 3T3-L1 preadipocytes. J. Biol. Chem. 255, 4745-4750.

Toleman, C., Paterson, A. J., Whisenhunt, T. R. and Kudlow, J. E. (2004).Characterization of the histone acetyltransferase (HAT) domain of a bifunctional proteinwith activable O-GlcNAcase and HAT activities. J. Biol. Chem. 279, 53665-53673.

Wang, Z., Gucek, M. and Hart, G. W. (2008). Cross-talk between GlcNAcylation andphosphorylation: site-specific phosphorylation dynamics in response to globally elevatedO-GlcNAc. Proc. Natl. Acad. Sci. USA 105, 13793-13798.

Wells, L., Gao, Y., Mahoney, J. A., Vosseller, K., Chen, C., Rosen, A. and Hart, G. W.(2002). Dynamic O-glycosylation of nuclear and cytosolic proteins: furthercharacterization of the nucleocytoplasmic beta-N-acetylglucosaminidase, O-GlcNAcase.J. Biol. Chem. 277, 1755-1761.

Whisenhunt, T. R., Yang, X., Bowe, D. B., Paterson, A. J., Van Tine, B. A. andKudlow, J. E. (2006). Disrupting the enzyme complex regulating O-GlcNAcylationblocks signaling and development. Glycobiology 16, 551-563.

Xu, G., Sztalryd, C., Lu, X., Tansey, J. T., Gan, J., Dorward, H., Kimmel, A. R. andLondos, C. (2005). Post-translational regulation of adipose differentiation-related proteinby the ubiquitin/proteasome pathway. J. Biol. Chem. 280, 42841-42847.

Xu, G., Sztalryd, C. and Londos, C. (2006). Degradation of perilipin is mediated throughubiquitination-proteasome pathway. Biochim. Biophys. Acta 1761, 83-90.

Zachara, N. E. and Hart, G. W. (2004). O-GlcNAc a sensor of cellular state: the role ofnucleocytoplasmic glycosylation in modulating cellular function in response to nutritionand stress. Biochim. Biophys. Acta 1673, 13-28.

Zachara, N. E., Molina, H., Wong, K. Y., Pandey, A. and Hart, G. W. (2010). Thedynamic stress-induced ‘‘O-GlcNAc-ome’’ highlights functions for O-GlcNAc inregulating DNA damage/repair and other cellular pathways Amino Acids 40, 793-808.

Zhang, F., Su, K., Yang, X., Bowe, D. B., Paterson, A. J. and Kudlow, J. E. (2003). O-GlcNAc modification is an endogenous inhibitor of the proteasome. Cell 115, 715-725.

Zhang, F., Hu, Y., Huang, P., Toleman, C. A., Paterson, A. J. and Kudlow, J. E. (2007).Proteasome function is regulated by cyclic AMP-dependent protein kinase throughphosphorylation of Rpt6. J. Biol. Chem. 282, 22460-22471.

2860 Journal of Cell Science 124 (16)

Jour

nal o

f Cel

l Sci

ence