Tandem mass spectrometry identifies many mousebrain O-GlcNAcylated proteins including EGFdomain-specific O-GlcNAc transferase targetsJoshua F. Alfaroa, Cheng-Xin Gongb, Matthew E. Monroea, Joshua T. Aldricha, Therese R. W. Claussa, Samuel O. Purvinea,Zihao Wangc, David G. Camp IIa, Jeffrey Shabanowitzd, Pamela Stanleye, Gerald W. Hartc, Donald F. Huntd,Feng Yanga,1, and Richard D. Smitha,1

aPacific Northwest National Laboratory, Richland, WA 99352; bDepartment of Neurochemistry, New York State Institute for Basic Research in DevelopmentalDisabilities, Staten Island, NY 10314; cDepartment of Biological Chemistry, Johns Hopkins University School of Medicine, Baltimore, MD 21205; dDepartmentof Chemistry, University of Virginia, Charlottesville, VA 22904; and eDepartment of Cell Biology, Albert Einstein College of Medicine, New York, NY 10461

Edited by Richard L. Huganir, The Johns Hopkins University School of Medicine, Baltimore, MD, and approved March 20, 2012 (received for review January11, 2012)

O-linked N-acetylglucosamine (O-GlcNAc) is a reversible posttransla-tional modification of Ser and Thr residues on cytosolic and nuclearproteins of higher eukaryotes catalyzed by O-GlcNAc transferase(OGT). O-GlcNAc has recently been found on Notch1 extracellular do-main catalyzed by EGF domain-specific OGT. AberrantO-GlcNAcmod-ification of brain proteins has been linked toAlzheimer’s disease (AD).However, understanding specific functions of O-GlcNAcylation in ADhas been impeded by the difficulty in characterization of O-GlcNAcsites on proteins. In this study, we modified a chemical/enzymaticphotochemical cleavage approach for enriching O-GlcNAcylated pep-tides in samples containing ∼100 μg of tryptic peptides from mousecerebrocortical brain tissue. A total of 274 O-GlcNAcylated proteinswere identified. Of these, 168were not previously known to bemod-ified by O-GlcNAc. Overall, 458 O-GlcNAc sites in 195 proteins wereidentified.Manyof themodified residues are either knownphosphor-ylation sites or located proximal to known phosphorylation sites.These findings support the proposed regulatory cross-talk betweenO-GlcNAcylation and phosphorylation. This study produced the mostcomprehensive O-GlcNAc proteome of mammalian brain tissue withboth protein identification and O-GlcNAc site assignment. Interest-ingly,weobservedO-β-GlcNAc on EGF-like repeats in the extracellulardomains of five membrane proteins, expanding the evidence for ex-tracellular O-GlcNAcylation by the EGF domain-specific OGT. We alsoreport a GlcNAc-β-1,3-Fuc-α-1-O-Thr modification on the EGF-like re-peat of the versican core protein, a proposed substrate of Fringe β-1,3-N-acetylglucosaminyltransferases.

chemical/enzymatic photochemical cleavage enrichment | glycosylation |mouse cerebral cortex

Asingle O-linked N-acetylglucosamine (O-GlcNAc) attached toSer and Thr residues of cytosolic and nuclear proteins is a re-

versible posttranslational modification (PTM) found in some bac-teria, some protozoans, filamentous fungi, viruses, and all highereukaryotes. The enzymes that catalyze the dynamic cycling ofO-GlcNAc modification, O-GlcNAc transferase (OGT) andO-GlcNAc hexosaminidase (OGA), are more highly expressed inthe pancreas and brain than in other tissues (1, 2). In addition,many proteins involved in neuronal communications, synaptictransmission, and synaptic plasticity are O-GlcNAcylated (3, 4),suggesting a role for this modification in brain function.O-GlcNAccycling is highly sensitive to nutrients and stress and is regulated bynearly every metabolic pathway. Aberrant O-GlcNAc modificationhas been linked to Alzheimer’s disease (AD) (5, 6) in which brainglucose metabolism is impaired (7). The reducedO-GlcNAcylationin AD contributes to hyperphosphorylation of tau protein andformation of the neurofibrillary tangles characteristic of AD andrelated neurodegenerative disorders (8, 9).Current understanding of the function of O-GlcNAcylation in

neurodegeneration has been impeded by difficulties in identifyingthis modification, even using MS (10, 11). First, the substoichio-

metric levels of O-GlcNAc modification at given sites on proteinsubstrates necessitate enrichment of O-GlcNAcylated proteins orpeptides before sequence analysis by MS (11). Additionally, iden-tifying specific Ser and Thr residues that are O-GlcNAcylated isdifficult, becauseO-GlcNAc is readily lost as an oxonium ion duringcollision-induced dissociation (CID), a widely used fragmentationmode for peptide sequencing byMS (12). Alternative higher-energycollisional dissociation (HCD) and electron transfer dissociation(ETD) MS methods have improved detection or have facilitatedsite-specific identification, but challenges remain (13).Several methods to enrich O-GlcNAcylated proteins or pep-

tides have led to the identification of a limited number ofO-GlcNAcylation sites by MS. For example, lectin weak-affinitychromatography (4, 10, 14) enabled identification of up to 142O-GlcNAcylation sites in 62 proteins frommouse embryonic stemcells (15). Immunoprecipitation, using O-GlcNAc-specific mono-clonal antibodies (13, 16), identified 83 O-GlcNAcylated sitesfrom a HEK293T cell extract (13). A recent metabolic labelingstudy that used alkyne-modified GlcNAc incorporated into OGTsubstrates and Cu(I)-catalyzed [3 + 2] azide–alkyne cycloaddition(CuAAC) to a chemically cleavable biotin–azide probe, enabledidentification of 374 putative O-GlcNAc modified proteins, butyielded no information on sites ofO-GlcNAcmodification (17). Inaddition to limited coverage ofO-GlcNAcylation sites, a drawbackto using these approaches is that they typically require milligramquantities of protein (4, 10, 13, 14, 16) or are limited to culturedcells (17), which makes them generally ill-suited for clinically de-rived tissue samples often available in small amounts.The chemical/enzymatic photochemical cleavage (CEPC)method

(18, 19) improves upon the highly selective chemical/enzymaticapproaches forO-GlcNAcylated proteins/peptides enrichment (20,21) and increases analytical sensitivity by introducing a photo-chemical cleavable-biotin probe that allows efficient release ofenriched peptides from the avidin affinity column. In this method(Fig. 1A), O-GlcNAcylated peptides are first enzymatically labeledwith azidogalactosamine (GalNAz). The free azido group in theGalNAz is then conjugated to the alkyne group in a photocleavablebiotin probe (PC-PEG-biotin-alkyne) through CuAAC. The bio-tinylated peptides are enriched using avidin affinity chromatogra-phy, and subsequently released through photochemical cleavage.O-GlcNAc-modified peptides enriched by this method aretagged with a basic aminomethyltriazolacetylgalactosamine (AMT-

Author contributions: J.F.A. and F.Y. designed research; J.F.A., T.R.W.C., and F.Y. per-formed research; J.F.A., M.E.M., J.T.A., S.O.P., Z.W., F.Y., and R.D.S. contributed newreagents/analytic tools; J.F.A., J.S., P.S., G.W.H., D.F.H., and F.Y. analyzed data; and J.F.A.,C.-X.G., D.G.C., P.S., G.W.H., D.F.H., F.Y., and R.D.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence may be addressed. E-mail: feng.yang@pnnl.gov or rds@pnnl.gov.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1200425109/-/DCSupplemental.

7280–7285 | PNAS | May 8, 2012 | vol. 109 | no. 19 www.pnas.org/cgi/doi/10.1073/pnas.1200425109

GalNAc) that facilitates ETD identification and site localization ofO-GlcNAc–modified peptides (10, 18, 19, 21). This approach en-abled identification of 141 O-GlcNAcylation sites in 64 proteinsfrom <15 μg of spindle and midbody proteins that were enrichedfrom HeLa cells (18, 19). However, to date, this method has notbeen used for complex tissues or global proteomic analyses. Inaddition, there remain challenges in reducing contaminants fromthe CuAAC reaction, which are detrimental to liquid chromatog-raphy (LC)-MS measurements.In the study reported herein, we modified the CEPC protocol

and used it to enrichO-GlcNAcylated peptides from tissue samplesobtained from the brain cortex of oneWTmouse and one 3xTg-ADmodel (22) mouse. CEPC enrichment of ∼100 μg mouse braincortex peptides from∼3mgWT tissue resulted in the identificationof 249 O-GlcNAcylated proteins and 358 O-GlcNAc site assign-ments from a single LC-MS/MS analysis, using an alternating CID/ETD approach. Enrichment of six 100-μg samples (three each forWT and 3xTg-AD mice) enabled identification of 274 O-GlcNA-cylated proteins and 458 O-GlcNAc sites, many of which werepreviously unknown. Important findings include data that supportextensive cross-talk between O-GlcNAcylated and phosphorylatedproteins involved in cerebrocortical processes, and unexpected sitesof extracellular O-GlcNAc modification. Specifically, we observedO-GlcNAc on a secreted cytokine AIMP1, on EGF-like repeats inthe extracellular domain of five other proteins, and on GlcNAc-β-1,3-Fuc-O-EGF of versican that reflects the action of a Fringeβ-1,3-GlcNAc transferase.

Results and DiscussionModified CEPC Approach for O-GlcNAc Peptide Enrichment. To max-imize sensitivity and selectivity in the CEPC O-GlcNAc peptideenrichment method (Fig. 1A), we modified the enzymatic reactions(SI Methods and SI Results and Discussion) to ensure high yields intransfer of GalNAz to O-GlcNAcylated peptides by the mutantβ-1,4-galactosyltransferase (GalT1Y289L), and effective removal ofN-glycans that may contain terminal GlcNAc by peptide:N-glycosi-dase F (PNGase F). We also incorporated additional aqueouswashes and an organic wash [70% (vol/vol) methanol in water]during avidin affinity enrichment (dashed boxes in Fig. 1A; workflowin Fig. 1B). The added washes increased the number of O-GlcNAc

protein identifications by ∼16% (Fig. 1B) and reduced the largeamount of hydrophobic Tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA) that remained from the CuAAC reaction (Fig. S1).

Evaluation of MS Fragmentation Methods for Identifying CEPC-Enriched Tagged O-GlcNAcylated Peptides. In a parallel effort weevaluated the capability of individualMS/MSmethods (HCD,CID,and ETD) for identifying tagged O-GlcNAc (AMT-GalNAz-GlcNAc modified) in 100 μg WT mouse cerebrocortical peptidesenriched using CEPC. Of the three fragmentation methods, onlyETD provided information regarding the location of the modifi-cation sites. The basic tag (AMT-GalNAz) increased peptidefragment efficiency of both ETD and HCD compared with that foruntagged, nativeO-GlcNAc peptides (13). Overall, ETD generatedthe greatest number of O-GlcNAcylated peptides, followed byHCD and CID (Fig. S2). In addition, the AMT-GalNAz-GlcNAcmodification produced three major diagnostic oxonium fragmentions for the intact sugars (204.09, 300.13, and 503.21m/z) comparedwith one (204.09 for HexNAc) from untagged, native O-GlcNAcpeptides. The AMT-GalNAz tag allowed CID to detect at leasttwo of the three major oxonium ions from intact sugars (Fig. S3)despite the one-third low molecular-weight cutoff rule (23). Com-pared with CID, HCD consistently detected the three major di-agnostic ions with higher mass accuracy and generally higherintensity (Fig. S4). The results of this evaluation indicate thatcombining tandemMS methods to obtain both site-specific (ETD)and diagnostic (HCD or CID) information can increase the con-fidence of O-GlcNAcylated peptide identification and site locali-zation, consistent with previous studies (10, 13). In addition,alternating CID-ETD analyses have been advocated as a superiormethod for identification of labile phosphopeptides (24). Our studyfurther demonstrates the need to combine the orthogonal methodsof fragmentation for assigning labile PTMs.

Known and Previously Unreported O-GlcNAcylated Proteins and Sites.In this study, we analyzed cerebrocortical tissue fromone 3xTg-ADmouse and one WT mouse (both female, 1 y old). The 3xTg-ADmouse, which overexpresses mutated human amyloid-β precursorprotein, tau, and preselinin-1, is a commonly used model of AD(22). Three samples from each mouse were analyzed (Fig. 1B).

Lysisand

trypsin diges�on

600 µg desalted pep�des

*O-GlcNAc Sites:

O-GlcNAc Proteins:

½ sampleCID-ETD

SCX, Bio�n-avidinEnrichment

OriginalWash Method

Labeling withBio�n-PEG-PC-Alkyne

½ sampleHCD-CID/ETD

½ sampleHCD-CID/ETD

1/3 1/3

WT WT WT3xTg-AD 3xTg-AD 3xTg-AD

249

358 152

133 126

160 135

104 144

182 139

122

GalNAz transfer by GalT1 Y289L, PNGase F

treatment, CIP

desal�ng1/3 Saved

SCX, Bio�n-avidinEnrichmentModified

Wash Method

UV cleavage UV cleavage

WT mouse20 mg cerebral cor�cal �ssue

3xTg-AD mouse20 mg cerebral cor�cal �ssue

BA

Fig. 1. Overview of the CEPC method (A) and experimental work flow (B) used in this study for the identification of O-GlcNAc proteins and modificationsites. (A) Steps modified in the enrichment strategy are depicted in dashed boxes. PNGase F and calf intestine phosphatase (CIP) are added to the reactionmixture to ensure selective and complete derivatization with GalNAz (18). During the CuAAC reaction, Cu(I) is stabilized with TBTA. (B) The original andmodified wash methods for the biotin-avidin enrichment step were compared using peptides from one WT mouse and one 3xTg-AD mouse (female, 1 y old).The number of O-GlcNAc sites and proteins identified from LC-MS/MS analysis of individual samples is depicted at the bottom of the figure.

Alfaro et al. PNAS | May 8, 2012 | vol. 109 | no. 19 | 7281

BIOCH

EMISTR

Y

We identified 1,575 unique O-GlcNAcylated peptides (TableS1) corresponding to 555 unique peptide sequences (Dataset S1)and 274 O-GlcNAcylated unique proteins (Dataset S2). Overall,458 unambiguous O-GlcNAc sites (≤5% false localization rate;Dataset S3) were assigned to 195 proteins, tripling the number ofO-GlcNAc sites reported in any single study (15, 19). Analysis ofthe sequence around these O-GlcNAc sites revealed severalstatistically significant (P ≤ 1E-6) motifs for O-GlcNAcylation(Fig. S5), among which P-X-gT-X-A and P-V-gS are enriched 14-and 23-fold, respectively, compared with a dynamic statisticalbackground of the entire mouse protein database. Some of themotif sequences agree with the previously reported OGT pre-ferred sequence P/V-P/V-V-gS/T-S/T (10, 19). Ontology analysisof the 274 O-GlcNAcylated proteins supports the involvement ofO-GlcNAcylation in numerous cellular functions in the brain,such as cytoskeleton organization, neurogenesis, synaptic trans-mission, learning, and memory (Dataset S4).We localized known and previously unreported O-GlcNAc

sites to specific residues in many of the 106 previously reportedO-GlcNAcylated proteins, which include numerous neuronal pro-teins implicated in AD, such as synapsin I and II, synaptopodin,

α-synuclein, several microtubule-associated proteins (Dataset S5),and proteins involved in neurogenesis.Because neurogenesis is impaired in AD (25), the O-GlcNA-

cylation status of these identified proteins (Dataset S4) may playa role in AD. Comparison of the O-GlcNAcylated peptidesobtained from 3xTg-AD vs. WT cerebrocortical tissue showedthat fewer (179 vs. 259) O-GlcNAcylated proteins were identifiedfrom the 3xTg-AD mouse (Fig. 1 and Fig. S6). These results areconsistent with previous observations of down-regulation of brainprotein O-GlcNAcylation in AD due to metabolic impairment(8). In addition, some of the O-GlcNAcylated proteins presentonly in the 3xTg-AD tissue may represent aberrantly O-GlcNA-cylated proteins (Dataset S2). Future quantitative proteomicstudies that analyze brain tissue from several WT and 3xTg-ADmice will help to suggest roles for O-GlcNAcylation in AD.We also identified 168 O-GlcNAcylated proteins previously

unknown to beO-GlcNAc-modified, including 114 proteins (Table1) that have confidently localized O-GlcNAc sites and 54 proteins(Table 2) that have defined tryptic peptide regions for theO-GlcNAc modification, but their modification residues areambiguous. Many of these O-GlcNAc-proteins are cytoskeleton

Table 1. Identified O-GlcNAcylated proteins previously unknown to be O-GlcNAc-modifiedwith confidently localized sites (≤5% false localization rate)

AccessionGene

symbol Site AccessionGene

symbol Site AccessionGene

symbol Site P51141 Dvl1 S383 Q91Z69 Srgap1 S982 Q80U40 Rimbp2 S681, T683P62484 Abi2 T297 O08553 Dpysl2 S507 Q80UY2 Kcmf1 S262 Q03173 Enah S362 O08599 Stxbp1 S511 Q80X80 C2cd2l T438, T447Q2PFD7 Psd3 S245 O54967 Tnk2 T832, T833 Q811L6 Mast4 S2165Q4JIM5 Abl2 T872 P0C090 Rc3h2 S592 Q8BG95 Ppp1r12b T542Q811P8 Arhgap32 S1027 P0C7T6 Atxn1l S40 Q8BHL3 Tbc1d10b T43, T44, T162Q8BJ42 Dlgap2 T633 P0CG14 Chtf8 S381 Q8BJM5 Slc30a6 T374

Q8BL65 Ablim2T363, S373, S381,

S412 P28652 Camk2b T325 Q8BU25 Pamr1 T267Q8BRT1 Clasp2 T476 P31230 Aimp1 T91 Q8BXL9 Iffo1 T175Q8CH77 Nav1 T543, T617, T619 P55937 Golga3 T207 Q8C0T5 Sipa1l1 T1402, S1403Q8VDQ8 Sirt2 T366 P59644 Inpp5j S117 Q8CDG3 Vcpip1 T1072 Q8VHG2 Amot T196 P59764 Dock4 T1806 Q8CFE4 Scyl2 S741 Q922J3 Clip1 T150 P97789 Xrn1 S1668 Q8CGI1 Fam193a T706Q9QXS1 Plec T2762 Q05793 Hspg2 T847 Q8K021 Scamp1 T59

Q9QYC0 Add1 T11, T16, T17, T540,

S557, T558, T559 Q2M3X8 Phactr1 S337 Q8K3X4 Irf2bpl S159 Q9WUM3 Coro1b S421 Q2VWQ2 Nell1 T542 Q8R1X6 Spg20 T478P06537 Nr3c1 S43 Q3UH68 Limch1 T506 Q8R3Y5 S354 P0C6A2 Mamld1 S253 Q3UNH4 Gprin1 T343 Q8VHW2 Cacng8 T381P16951 Atf2 T272 Q3V0G7 Garnl3 S905 Q91V09 Wdr13 S140 P42227 Stat3 T717 Q499E5 Stox2 T863 Q91X58 Zfand2b T167P58462 Foxp1 T446 Q4VAA2 Cdv3 T178 Q91XV3 Basp1 S169 P70365 Ncoa1 T401 Q571K4 Tab3 T385, S412 Q91Z67 Srgap2 S990 Q02780 Nfia T362 Q5FWH2 Unkl T459 Q922Y1 Ubxn1 T192Q3UCQ1 Foxk2 S540 Q5QNQ6 Osbp2 T140 Q99KN9 Clint1 S328 Q61026 Ncoa2 T964 Q5SRX1 Tom1l2 T187, T188 Q9DAI6 Fam135b T989Q62441 Tle4 T330 Q5SUE8 Ankrd40 S198, T199 Q9DAM7 T69, T72Q64336 Tbr1 S647 Q61001 Lama5 S2140 Q9DCT8 Crip2 T88

Q68ED7 Crtc1 T417 Q62419 Sh3gl1 T284 Q9EPN1 Nbea T1276, T1796,

T1797 Q6NXK2 Znf532 S455 Q64332 Syn2 S79 Q9ERQ3 Znf704 T468Q8BT14 Cnot4 S316, T573 Q68FF7 Slain1 T411 Q9EST3 Eif4enif1 S416

Q8BW22 Ss18l1 T48 Q68FH0 Pkp4S225, S226,

S1087 Q9JI46 Nudt3 T159Q91W39 Ncoa5 T521 Q69ZI1 Sh3rf1 T512 Q9QWY8 Asap1 T823P70392 Rasgrf2 S763 Q6A058 Armcx2 T328 Q9QWZ1 Rad1 T232P83510 Tnik S539, T577 Q6A0A2 Larp4b T51 Q9QY01 Ulk2 T613, T727Q80YA9 Cnksr2 S329 Q6NXJ0 Wwc2 S520, T528 Q9QZR5 Hipk2 S1009Q8CF89 Tab1 S393 Q6PFX7 T468 Q9R0Z9 Dlc1 S174

Q8CHG7 Rapgef2 S807, T1006, T1007 Q6RHR9 Magi1T1093, T1094 Q9WTS4 Odz1 S237, T685

Q91WJ0 Frs3 S439 Q80TN7 Nav3 S1210 Q9WUU8 Tnip1 T103

Proteins are shaded in different colors according to their functional category. Blue, cytoskeleton; red, regu-lation of transcription; brown, signaling; unshaded, other; underlined entries, kinases. Protein names are in-cluded in Table S2.

7282 | www.pnas.org/cgi/doi/10.1073/pnas.1200425109 Alfaro et al.

proteins, signaling proteins, or proteins involved in the regulationof transcription—all classes of proteins known to be O-GlcNA-cylated (26, 27). We observed three O-GlcNAcylation sites(T1276, T1796, and T1797) on neurobeachin (Nbea), a proteinimplicated in membrane protein traffic and autism, and requiredfor the formation and functioning of central synapses (28). Thisprotein is a known phosphoprotein and binds to protein kinase A;however, its O-GlcNAcylation status was previously unknown(29). Literature provides examples of cross-talk betweenO-GlcNAcylation and phosphorylation (19, 26, 30), as wellas examples of O-GlcNAc regulating ubiquitination throughE1 ubiquitin-activating enzyme (26, 31). We observed manyO-GlcNAcylated proteins previously shown to be involved in thecycling of either phosphorylation or ubiquitination, which providesindirect evidence of cross-talk between these modifications. Theseproteins were not previously reported to be O-GlcNAc-modifiedand include 13 kinases (underlined entries in Tables 1 and 2), oneputative phosphatase (Dnajc6), three proteins involved in phos-phatase regulation (Phactr1, Mprip, and Ppp1r12b), three E3ubiquitin-protein ligases (SH3RF1, HECTD1, and KCMF1), andone deubiquitinating protein (VCPIP1; Tables 1 and 2).All previously identified O-GlcNAcylated proteins are also

known to be phosphorylated (26), which supports cross-talk be-tween the two modifications. There are 268 (>98%) of the 274O-GlcNAc-modified proteins identified in this study also knownas phosphoproteins (32) (Dataset S2). We found that ∼24% ofthe identified O-GlcNAc sites have either reciprocal or proximal(±10 aa residues) phosphorylation sites (Dataset S3), whichagain suggests possible cross-talk between phosphorylation andO-GlcNAcylation in cerebrocortical processes. Some of thesepotential cross-talk cases that may regulate protein interactions(e.g., EMSY with BRCA2 and HCFC1 with SIN3A) were alsoobserved in HeLa cells (19). Our findings include two mappedO-GlcNAc sites (S539 and T577) on TNIK, a serine/threoninekinase previously unknown to be O-GlcNAcylated. TheseO-GlcNAc sites (S539 and T577) with proximal known phos-phorylation sites (S541, S545, and T552) (33) on TNIK are alllocated within its predicted (34) binding region to NEDD4-1, anE3 ubiquitin ligase. O-GlcNAcylation of human NEDD4-1 wasrecently reported (17), but with no site localization.We identified an O-GlcNAc site (T375) that is located within

the experimentally determined TNIK binding region of NEDD4-1(35). A known (S381) and a predicted (S385) (34) proximal phos-

phorylation site onNEDD4-1 also occur within this binding domain.NEDD4-1, TNIK, and Rap2A are known to form a complex thatregulates NEDD4-1–mediated Rap2A ubiquitination (35). To-gether, our findings suggest that cross-talk between phosphory-lation and O-GlcNAcylation may be involved in the NEDD4-1/TNIK/Rap2A (35) signaling pathway that regulates neurite growth.We also suggest that this cross-talk may extend to ubiquitination,given that TNIK is required in this complex to enable ubiquiti-nation of Rap2A (35) and that its interaction with NEDD4-1 maybe regulated by O-GlcNAcylation/phosphorylation.

O-GlcNAcylation on a Secreted Protein and on the ExtracellularDomains of Membrane Proteins. Fig. 2 shows subcellular localiza-tion of the 274 identified O-GlcNAcylated proteins. Previousstudies have demonstrated that O-GlcNAc modifies numerousnuclear and cytoplasmic proteins (26). Unexpectedly, we alsoidentified O-GlcNAc modification on the extracellular EGF do-main of five membrane proteins (Table 3), and on one secretedcytokine. Fig. S3 shows modification and site assignments derivedfrom MS/MS spectra for the tryptic peptide CACLAGYTGQRfrom the EGF domain of Pamr1. To our knowledge, there are onlytwo previous reports of extracellular O-GlcNAcylation, i.e., theextracellular domain of NOTCH and Dumpy inDrosophila (36, 37).The O-GlcNAc transferase that attaches O-β-GlcNAc to

NOTCH is a distinct enzyme that is genetically unrelated toOGT (36). It resides in the endoplasmic reticulum (ER) withinthe secretory pathway and is termed EGF domain-specificO-GlcNAc transferase (EOGT) (36, 37). EOGT is conserved

05

1015202530

% o

f O-G

lcN

Ac

Pro

tein

s

Fig. 2. Cellular component gene ontology annotation of identified O-GlcNAcylated proteins.

Table 2. Identified O-GlcNAcylated proteins previously unknown to be O-GlcNAc-modified withmodification site localized to a specific tryptic peptide region

Q8CHY6 Gatad2a 311–339 Q501J7 Phactr4 190–213 Q8VIG0 Zcchc14 615–645 Q9JL19 Ncoa6 1218–1246 Q5SWP3 Nacad 1190–1210 Q91YD3 Dcp1a 497–533 O35099 Map3k5 1216–1246 Q69ZH9 Arhgap23 559–591 Q9CR95 Necap1 195–217 P97379 G3bp2 225–252 Q69ZR2 Hectd1 1339–1370 Q9DBG5 Plin3 69–84 Q3UHD9 Agap2 73–111, 226–258 Q6NS60 Fbxo41 377–392 Q9EP53 Tsc1 1057–1079

Accession Gene

symbol O-GlcNAcylation

region* Accession Gene

symbol O-GlcNAcylation

region* AccessionGene

symbol O-GlcNAcylation

region* A2AHC3 Camsap1 370–396 Q62073 Map3k7 373–387 Q7TN29 Smap2 180–202 A2AKB4 Frmpd1 1206–1230 Q6PAJ1 Bcr 133–155 Q7TPM1 Prrc2b 1293–1311 P97434 Mprip 175–201 Q6PGG2 Gmip 578–591 Q80TL4 Kiaa1045 328–355 Q6DFV3 Arhgap21 312–331 Q9QZS8 Sh2d3c 420–439 Q80TZ3 Dnajc6 600–640

Q6PFD5 Dlgap3 552–575 B1AZP2 Dlgap4 258–290 Q80U78 Pum1 798–817 Q8BIE6 Frmd4a 926–953 O08919 Numbl 243–292 Q80VP1 Epn1 468–498 Q9JL04 Fmn2 243–264 P13595 Ncam1 894–938 Q80WC7 Agfg2 178–204, 445–479 Q9Z1K7 Apc2 2080–2096 P63250 Kcnj3 6–40 Q80YR4 Znf598 560–581

A2A884 Hivep3 973–991 Q03141 Mark3 536–552 Q8BXR9 Osbpl6 193–219

P42128 Foxk1 570–606 Q3UHC0 Tnrc6c 748–773 Q8BZB3 377–401

P45481 Crebbp 135–156 Q3UIL6 Plekha7 112–138 Q8C120 Sh3rf3 588–624 Q61818 Rai1 531–550 Q3UQN2 Fcho2 476–504 Q8CG79 Tp53bp2 323–343 Q80TZ9 Rere 1209–1246 Q4G0F8 Ubn1 988–1009 Q8R361 Rab11fip5 538–550

Proteins are shaded in different colors according to their functional category. Blue, cytoskeleton; red, regulation oftranscription; brown, signaling; unshaded, other; underlined entries, kinases; bold, peptide sequences containing theAsp-Xaa-Ser/Thr motif. Protein names are included in Table S3.*Tryptic peptide residue range.

Alfaro et al. PNAS | May 8, 2012 | vol. 109 | no. 19 | 7283

BIOCH

EMISTR

Y

from Drosophila to mammals (37, 38). There is no evidence thatintracellular forms of OGT or O-GlcNAcase occur in the ex-tracellular or luminal spaces. Four O-GlcNAcylated peptidesfrom these proteins were located within an EGF-like domain,sharing a similar motif sequence CXXGXS/TGXXC to thereported extracellular O-GlcNAcylation on NOTCH and Dumpyin Drosophila and Notch1 in mammals (36–38). We also identifiedanO-GlcNAc site on the EGF-like domain of NOTCH2 protein ofthe Notch signaling pathway, which we infer is on the T residue inthe YSCVCSPGFTGQR sequence, consistent with the CXXGXS/TGXXCmotif (Table 3). Our findings suggest that this EOGT (36)has additional substrates. In fact, we determined that 91 mouseproteins, 104 human proteins, and 18 Drosophila proteins containthe CXXGXS/TGXXC motif (Datasets S6, S7, and S8). Theseproteins are involved in the Notch signaling pathway, extracellularmatrix (ECM)-receptor interactions, and other signaling pathways.Among the proteins in both the human and mouse proteome thatcontain this motif, ∼30% are localized within the ECM. The im-portance of this modification within the ECM has been demon-strated in Drosophila, where loss of EOGT causes defects in theapical ECM (37). The proteins containing the CXXGXS/TGXXCmotif appear conserved across species, with 83 proteins conservedbetween human and mouse, and 15 of 18 Drosophila proteinshaving orthologs in both human and mouse.We mapped an O-GlcNAc site (T91) on AIMP1 that is known

to regulate angiogenesis, inflammation, and wound healing (39,40), and report O-GlcNAc modification on a secreted cytokine. Arecent study (41) showed AIMP1 plays a glucagon-like role inglucose homeostasis and its secretion is induced by TNFα or heatshock (40, 42). Motif analysis indicates two proximal phosphory-lation sites (S99 and S107) of T91 are potential targets of theintracellular kinase MAPK. In addition, because AIMP1 is presentin multiple subcellular locations besides the extracellular space(34) and lacks an EGF repeat, the AIMP1 form we detected frommouse cerebrocortical tissue is likely cytosolic, and the T91 is thesubstrate of OGT instead of EOGT. The O-GlcNAc site (T91),together with its reported five proximal phosphorylation sites (S99,S101, T105, S107, and S138), are all located within the HSP90B1interaction region of AIMP1. The potential cross-talk betweenthese sites may be relevant to AIMP1 and HSP90B1 interactionsthat regulate KDELR1-mediated retention of HSP90B1/gp96 inthe endoplasmic reticulum (34).Interestingly, we also identified a GlcNAc-β-1,3-Fuc-α-1-O-Thr

site (T3103) in the EGF-like domain 2 (EGF2) of versican coreprotein (Fig. S3) within the tryptic peptide sequence NGAT#CVDGFNTFR (# indicates the modification). The addition ofO-fucose to EGF repeats is catalyzed by Pofut1 (43), and elongationof the monosaccharide is initiated by Fringe, an O-fucose–specificβ-1,3-N-acetylglucosaminyltransferase (44, 45) to form the di-saccharide modification we detected. GlcNAc-β-1,3-Fuc may be anintermediate species before subsequent elongation by galactosyl-transferase and sialyltransferase to form a tetrasaccharide (46).The Thr modified by O-fucose is in a predicted consensus site(C2XXXXS/TC3) between C2 and C3 of EGF repeats for O-fuco-sylation (47, 48) and within EGF2 (C2RNGATC3) of versican.

Summary. To our knowledge, the present study has produced themost comprehensive O-GlcNAc proteome to date in terms ofboth protein identifications (274) and O-GlcNAc site assign-ments (458) for mouse brain tissue, and used much smallersamples (∼100 μg tissue peptide per enrichment) than in pre-vious studies (10, 15, 20). Our studies suggest roles for extensivecross-talk between O-GlcNAc and other posttranslational mod-ifications in not only the regulation of normal neuronal func-tions, but also in the etiology of neurodegeneration. We dem-onstrate the suitability of the CEPC method for global proteomicanalysis of O-GlcNAcylated peptides, and the potential for rapidlyexpanding the O-GlcNAc proteome in brain and other tissues.This approach will enable high-resolution spatial mapping ofO-GlcNAcylation patterns in brain samples from laser-capturemicrodissection to gain insights into roles for O-GlcNAcylationin neurodegenerative disease.

MethodsSample Preparation. Mouse cerebrocortical tissue was homogenized in asolution containing 6 M guanidine HCl, 10 mM DTT, 50 mM ammoniumbicarbonate [NH4HCO3 (pH 8.1)] with 1% (vol/vol) phosphatase inhibitormixture 2% (vol/vol) (Sigma), and 100 nM PUGNAc. Details of proteindigestion are described in SI Methods.

O-GlcNAc Enrichment. Enrichment was performed as previously described (18)except where specified otherwise in SI Methods.

LC-MS/MS Analysis. Samples were analyzed using a LTQ Orbitrap Velos MS(Thermo Scientific) coupled to an automated dual-columnmetal-free nanoLCplatform (49). Details of the separation and mass spectrometer parametersare described in SI Methods.

Data Analysis. We used two database search engines, SEQUEST and ProteinProspector, to obtain more comprehensive peptide identifications. Datawere searched against a decoy protein database. Fully tryptic peptide iden-tifications were filtered in a way that no reversed hits were left, with anestimated zero false discovery rate (Datasets S9 and S10) (50). All identi-fications were within ±5 ppm measured mass accuracy and required obser-vation of oxonium ion fragments (204.0872, 300.1308, and 503.2101 m/z forHCD; 300.1308 and 503.2101 m/z for CID) in their corresponding HCD or CIDscans with a mass tolerance of 0.0025 Da for HCD and 0.3 Da for CID, whichwere extracted using our updated MASIC software (51). Ascore (52) and SLIPscore (built into the Protein Prospector search engine) (53) were used to es-timate the confidence ofO-GlcNAc modification site assignment for SEQUESTand Protein Prospector search results, respectively. See SI Methods and SIResults and Discussion for more details. All of the peptides identified fromNEDD4-1, TNIK, AIMP1, and EGF-like repeats in the extracellular domain ofsix proteins were manually confirmed by authors J.F.A., F.Y., and J.S. The sixN-linked GlcNAc peptides (Dataset S11) were also manually confirmed.

Bioinformatics. Gene ontology annotation, cellular component, and bi-ological process was performed using the Software Tool for ResearchingAnnotations of Proteins, or STRAP (54). The pathway analysis was performedusing DAVID Bioinformatics Resources 6.7 as previously described (55).Briefly, all identified O-GlcNAc modified proteins were queried against themouse proteome as a background. The statistical enrichment was calculatedfor KEGG pathways identified from a protein list obtained during this study,

Table 3. O-GlcNAc modification sites on a secreted protein and on the extracellular domains ofmembrane proteins

Scan AccessionGenesymbol Modification site Motif sequence Localization

ETD P31230 Aimp1 T91 PLQTNCTASESVV Interaction region with HSP90B1ETD Q05793 Hspg2 T847 ACAPGYTGRRCES EGF-like 3 domainETD Q2VWQ2 Nell1 T542 VCPSGFTGSHCEK EGF-like 4 domainETD Q61001 Lama5 S2140 TCPPGLSGERCDT EGF-like 22 domainETD Q8BU25 Pamr1 T267 ACLAGYTGQRCEN EGF-like domainHCD* O35516 Notch2 T673? VCSPGFTGQRCNI EGF-like 17 domain

*The O-GlcNAc site identified by HCD is ambiguous due to the labile nature of the modification under thiscollision mode condition.

7284 | www.pnas.org/cgi/doi/10.1073/pnas.1200425109 Alfaro et al.

and pathways with P ≤ 0.05 were reported as significant. Prealigned se-quence was generated with six residues on either side of all of the un-ambiguous O-GlcNAc sites. The sequence was subject to motif analysis onlineby motif-x (http://motif-x.med.harvard.edu/) (56), and the motifs were builtthrough comparison with a dynamic statistical background of mouse proteindatabase. The occurrences threshold was set at 20, and the significance(P value) was 1E-6.

ACKNOWLEDGMENTS. We thank Dr. Joshua Adkins at Pacific NorthwestNational Laboratory (PNNL) for helpful suggestions regarding the manuscript,Robert Chalkley (University of California, San Francisco) for help with usingProtein Prospector, and Ronald J. Moore for discussions regarding MS analysis.

This work was funded by PNNL Laboratory Directed Research Developmentfunding (to F.Y.); three National Institutes of Health (NIH) grants (to R.D.S.),National Center for Research Resources Grant 5P41RR018522-10, NationalInstitute of General Medical Sciences Grants 8 P41 GM103493-10 andAG027429; NIH Grant GM 037537 (to D.F.H); NIH Grants N01-HV-00240, R01CA42486, and P01HL107153 (to G.W.H); AG027429 and TW008123 (toC.-X.G.); and NIH National Cancer Institute Grant R01 36434 (to P.S.). Sampleswere analyzed using capabilities developed under the support of the NIHNational Center for Research Resources Grant RR018522 and the US De-partment of Energy Biological and Environmental Research (DOE/BER). Workwas performed in the Environmental Molecular Science Laboratory, a DOE/BER national scientific user facility at PNNL in Richland, WA. PNNL is oper-ated for the DOE by Battelle under Contract DE-AC05-76RLO-1830.

1. Kreppel LK, Blomberg MA, Hart GW (1997) Dynamic glycosylation of nuclear andcytosolic proteins. Cloning and characterization of a unique O-GlcNAc transferasewith multiple tetratricopeptide repeats. J Biol Chem 272:9308–9315.

2. Okuyama R, Marshall S (2003) UDP-N-acetylglucosaminyl transferase (OGT) in braintissue: Temperature sensitivity and subcellular distribution of cytosolic and nuclearenzyme. J Neurochem 86:1271–1280.

3. Murrey HE, Hsieh-Wilson LC (2008) The chemical neurobiology of carbohydrates.Chem Rev 108:1708–1731.

4. Vosseller K, et al. (2006) O-linked N-acetylglucosamine proteomics of postsynapticdensity preparations using lectin weak affinity chromatography and mass spectrom-etry. Mol Cell Proteomics 5:923–934.

5. Dias WB, Hart GW (2007) O-GlcNAc modification in diabetes and Alzheimer’s disease.Mol Biosyst 3:766–772.

6. Gong CX, Liu F, Grundke-Iqbal I, Iqbal K (2006) Impaired brain glucose metabolismleads to Alzheimer neurofibrillary degeneration through a decrease in tau O-GlcNAcylation. J Alzheimers Dis 9:1–12.

7. Hoyer S (2004) Causes and consequences of disturbances of cerebral glucose metabolismin sporadic Alzheimer disease: Therapeutic implications. Adv Exp Med Biol 541:135–152.

8. Liu F, et al. (2009) Reduced O-GlcNAcylation links lower brain glucose metabolism andtau pathology in Alzheimer’s disease. Brain 132:1820–1832.

9. Yuzwa SA, et al. (2008) A potent mechanism-inspired O-GlcNAcase inhibitor thatblocks phosphorylation of tau in vivo. Nat Chem Biol 4:483–490.

10. Chalkley RJ, Thalhammer A, Schoepfer R, Burlingame AL (2009) Identification ofprotein O-GlcNAcylation sites using electron transfer dissociation mass spectrometryon native peptides. Proc Natl Acad Sci USA 106:8894–8899.

11. Wang Z, Hart G (2008) Glycomic approaches to study GlcNAcylation: Protein identifi-cation, site-mapping, and site-specific O-GlcNAc quantitation. Clin Proteomics 4:5–13.

12. Greis KD, et al. (1996) Selective detection and site-analysis of O-GlcNAc-modifiedglycopeptides by beta-elimination and tandem electrospray mass spectrometry. AnalBiochem 234:38–49.

13. Zhao P, et al. (2011) Combining high-energy C-trap dissociation and electron transfer dis-sociation for proteinO-GlcNAcmodification site assignment. J Proteome Res 10:4088–4104.

14. Vosseller K, Wells L, Hart GW (2001) Nucleocytoplasmic O-glycosylation: O-GlcNAcand functional proteomics. Biochimie 83:575–581.

15. Myers SA, Panning B, Burlingame AL (2011) Polycomb repressive complex 2 is nec-essary for the normal site-specific O-GlcNAc distribution in mouse embryonic stemcells. Proc Natl Acad Sci USA 108:9490–9495.

16. Teo CF, et al. (2010) Glycopeptide-specific monoclonal antibodies suggest new rolesfor O-GlcNAc. Nat Chem Biol 6:338–343.

17. Zaro BW, Yang YY, Hang HC, Pratt MR (2011) Chemical reporters for fluorescentdetection and identification of O-GlcNAc-modified proteins reveal glycosylation ofthe ubiquitin ligase NEDD4-1. Proc Natl Acad Sci USA 108:8146–8151.

18. Wang ZH, et al. (2010) Enrichment and site mapping of O-linked N-acetylglucosamineby a combination of chemical/enzymatic tagging, photochemical cleavage, andelectron transfer dissociation mass spectrometry. Mol Cell Proteomics 9:153–160.

19. Wang Z, et al. (2010) Extensive crosstalk between O-GlcNAcylation and phosphory-lation regulates cytokinesis. Sci Signal 3:ra2.

20. Rexach JE, Clark PM, Hsieh-Wilson LC (2008) Chemical approaches to understandingO-GlcNAc glycosylation in the brain. Nat Chem Biol 4:97–106.

21. Khidekel N, et al. (2007) Probing the dynamics of O-GlcNAc glycosylation in the brainusing quantitative proteomics. Nat Chem Biol 3:339–348.

22. Oddo S, et al. (2003) Triple-transgenic model of Alzheimer’s disease with plaques andtangles: Intracellular Abeta and synaptic dysfunction. Neuron 39:409–421.

23. Cunningham C, Jr., Glish GL, Burinsky DJ (2006) High amplitude short time excitation:A method to form and detect low mass product ions in a quadrupole ion trap massspectrometer. J Am Soc Mass Spectrom 17:81–84.

24. Kim MS, Zhong J, Kandasamy K, Delanghe B, Pandey A (2011) Systematic evaluationof alternating CID and ETD fragmentation for phosphorylated peptides. Proteomics11:2568–2572.

25. Lazarov O, Mattson MP, Peterson DA, Pimplikar SW, van Praag H (2010) When neu-rogenesis encounters aging and disease. Trends Neurosci 33:569–579.

26. Butkinaree C, Park K, Hart GW (2010) O-linked beta-N-acetylglucosamine (O-GlcNAc):Extensive crosstalk with phosphorylation to regulate signaling and transcription inresponse to nutrients and stress. Biochim Biophys Acta 1800:96–106.

27. Hart GW, Slawson C, Ramirez-Correa G, Lagerlof O (2011) Cross talk between O-GlcNAcylation and phosphorylation: Roles in signaling, transcription, and chronicdisease. Annu Rev Biochem 80:825–858.

28. Medrihan L, et al. (2009) Neurobeachin, a protein implicated in membrane proteintraffic and autism, is required for the formation and functioning of central synapses.J Physiol 587:5095–5106.

29. Wang XL, et al. (2000) Neurobeachin: A protein kinase A-anchoring, beige/Chediak-hi-gashiproteinhomolog implicated inneuronalmembrane traffic. JNeurosci20:8551–8565.

30. Hunter T (2007) The age of crosstalk: Phosphorylation, ubiquitination, and beyond.Mol Cell 28:730–738.

31. Guinez C, et al. (2008) Protein ubiquitination is modulated by O-GlcNAc glycosylation.FASEB J 22:2901–2911.

32. Hornbeck PV, Chabra I, Kornhauser JM, Skrzypek E, Zhang B (2004) PhosphoSite: Abioinformatics resource dedicated to physiological protein phosphorylation. Proteo-mics 4:1551–1561.

33. Huttlin EL, et al. (2010) A tissue-specific atlas of mouse protein phosphorylation andexpression. Cell 143:1174–1189.

34. Apweiler R, et al.; UniProt Consortium (2010) The Universal Protein Resource (UniProt)in 2010. Nucleic Acids Res 38(Database issue):D142–D148.

35. Kawabe H, et al. (2010) Regulation of Rap2A by the ubiquitin ligase Nedd4-1 controlsneurite development. Neuron 65:358–372.

36. Matsuura A, et al. (2008) O-linked N-acetylglucosamine is present on the extracellulardomain of notch receptors. J Biol Chem 283:35486–35495.

37. Sakaidani Y, et al. (2011) O-linked-N-acetylglucosamine on extracellular protein do-mains mediates epithelial cell-matrix interactions. Nat Commun 2:583.

38. Sakaidani Y, et al. (2012) O-linked-N-acetylglucosamine modification of mammalianNotch receptors by an atypical O-GlcNAc transferase Eogt1. Biochem Biophys ResCommun 419(1):14–19.

39. Ko YG, et al. (2001) A cofactor of tRNA synthetase, p43, is secreted to up-regulateproinflammatory genes. J Biol Chem 276:23028–23033.

40. Park SG, et al. (2005) The novel cytokine p43 stimulates dermal fibroblast pro-liferation and wound repair. Am J Pathol 166:387–398.

41. Park SG, et al. (2006) Hormonal activity of AIMP1/p43 for glucose homeostasis. ProcNatl Acad Sci USA 103:14913–14918.

42. Barnett G, et al. (2000) Prostate adenocarcinoma cells release the novel proin-flammatory polypeptide EMAP-II in response to stress. Cancer Res 60:2850–2857.

43. Wang Y, et al. (2001) Modification of epidermal growth factor-like repeats with O-fucose. Molecular cloning and expression of a novel GDP-fucose protein O-fucosyl-transferase. J Biol Chem 276:40338–40345.

44. Brucker K, Perez L, Clausen H, Cohen S (2000) Glycosyltransferase activity of fringemodulates Notch-Delta interactions. Nature 406:411–415, and erratum (2000) 407:654.

45. Moloney DJ, et al. (2000) Fringe is a glycosyltransferase that modifies Notch. Nature406:369–375.

46. Moloney DJ, et al. (2000) Mammalian Notch1 is modified with two unusual forms ofO-linked glycosylation found on epidermal growth factor-like modules. J Biol Chem275:9604–9611.

47. Rampal R, Luther KB, Haltiwanger RS (2007) Notch signaling in normal and diseasestates: Possible therapies related to glycosylation. Curr Mol Med 7:427–445.

48. Rana NA, Haltiwanger RS (2011) Fringe benefits: Functional and structural impacts ofO-glycosylation on the extracellular domain of Notch receptors. Curr Opin Struct Biol21:583–589.

49. Zhao R, et al. (2009) Automated metal-free multiple-column nanoLC for improvedphosphopeptide analysis sensitivity and throughput. J Chromatogr B Analyt TechnolBiomed Life Sci 877:663–670.

50. Elias JE, HaasW, Faherty BK, Gygi SP (2005) Comparative evaluation ofmass spectrometryplatforms used in large-scale proteomics investigations. Nat Methods 2:667–675.

51. Monroe ME, Shaw JL, Daly DS, Adkins JN, Smith RD (2008) MASIC: A software pro-gram for fast quantitation and flexible visualization of chromatographic profilesfrom detected LC-MS(/MS) features. Comput Biol Chem 32:215–217.

52. Beausoleil SA, Villén J, Gerber SA, Rush J, Gygi SP (2006) A probability-based approachfor high-throughput protein phosphorylation analysis and site localization. Nat Bio-technol 24:1285–1292.

53. Chalkley RJ, Baker PR, Trinidad JC (2011) Modification site localization scoring in-tegrated into a search engine. Mol Cell Proteomics 10(7):1–9.

54. Bhatia VN, Perlman DH, Costello CE, McCombME (2009) Software tool for researchingannotations of proteins: Open-source protein annotation software with data visual-ization. Anal Chem 81:9819–9823.

55. HuangW, ShermanBT, Lempicki RA (2009)Bioinformatics enrichment tools: Paths towardthe comprehensive functional analysis of large gene lists. Nucleic Acids Res 37:1–13.

56. SchwartzD,Gygi SP (2005)An iterative statistical approach to the identificationofproteinphosphorylation motifs from large-scale data sets. Nat Biotechnol 23:1391–1398.

Alfaro et al. PNAS | May 8, 2012 | vol. 109 | no. 19 | 7285

BIOCH

EMISTR

Y