Membrane association is a determinant for substrate ...Membrane association is a determinant for substrate recognition by PMT4 protein O-mannosyltransferases Johannes Hutzler*, Maria

Membrane association is a determinant for substraterecognition by PMT4 protein O-mannosyltransferasesJohannes Hutzler*, Maria Schmid*†, Thomas Bernard‡, Bernard Henrissat‡, and Sabine Strahl*§

*Department of Cell Chemistry, Institute of Plant Sciences, University of Heidelberg, 69120 Heidelberg, Germany; and ‡Architecture et Fonction desMacromolécules Biologiques, Centre National de la Recherche Scientifique, Universités Aix-Marseille I and II, 13288 Marseille Cedex 9, France

Edited by Christian R. H. Raetz, Duke University Medical Center, Durham, NC, and approved March 18, 2007 (received for review January 15, 2007)

Protein O-mannosylation represents an evolutionarily conserved,essential posttranslational modification with immense impact on avariety of cellular processes. In humans, O-mannosylation defectsresult in Walker–Warburg syndrome, a severe recessive congenitalmuscular dystrophy associated with defects in neuronal migrationthat produce complex brain and eye abnormalities. In mouseand yeasts, loss of O-mannosylation causes lethality. Protein O-mannosyltransferases (PMTs) initiate the assembly of O-mannosylglycans. The evolutionarily conserved PMT family is classified intoPMT1, PMT2, and PMT4 subfamilies, which mannosylate distincttarget proteins. In contrast to other types of glycosylation, signalsequences for O-mannosylation have not been identified to date.In the present study, we identified signals that determine PMT4-dependent O-mannosylation. Using specific model proteins, wedemonstrate that in yeast Pmt4p mediates O-mannosylation ofSer/Thr-rich membrane-attached proteins. The nature of themembrane-anchoring sequence is nonrelevant, as long as it isflanked by a Ser/Thr-rich domain facing the endoplasmic reticulumlumen. Our work shows that, in contrast to several other types ofglycosylation, PMT4 O-mannosylation signals are not just linearprotein’s primary structure sequences but rather are highly com-plex. Based on these findings, we performed in silico analyses ofthe Saccharomyces cerevisiae proteome and identified previouslyundescribed Pmt4p substrates. This tool for proteome-wide iden-tification of O-mannosylated proteins is of general interest be-cause several of these proteins are major players of a wide varietyof cellular processes.

glycosylation � mannosyl glycans � POMT � yeast � mannosylation

G lycosylation is an essential and abundant protein modification(1). More than half of all proteins in biological systems areglycosylated, and the glycan chains influence a large number ofbiological processes (2).

Pro- and eukaryotes modify proteins with a variety of carbo-hydrate residues. Regarding the two major types of proteinglycosylation, N- and O-glycosylation, carbohydrate moieties areattached either to the amide group of asparagine (Asn) residuesof the sequon AsnXSer/Thr or to hydroxy amino acids, mostlyserine (Ser) and threonine (Thr) residues, respectively. In thecase of O-glycosylation, a variety of monosaccharides such asN-acetylgalactosamine, N-acetylglucosamine, fucose, glucose,xylose, or mannose are found in O-glycosidic bonds in differentorganisms (2, 3).

Protein O-mannosylation represents an evolutionarily conservedmodification among eukaryotes and mycobacteria (1, 4). In yeastsand filamentous fungi, O-mannosylation serves a variety of differ-ent functions. It is required for stability, sorting, and localization ofproteins, thereby affecting protein function and being indispensablefor cell wall integrity, cell polarity, and morphogenesis (1). InDrosophila melanogaster, reduced O-mannosylation results in al-tered muscle structures and alignment of adult cuticle (5). In mouse,lack of O-mannosylation results in embryonic lethality (6) and inhumans in congenital muscular dystrophies with neuronal migra-tion defects, such as Walker–Warburg syndrome and muscle–eye–brain disease (7).

O-mannosyl glycans are short linear oligosaccharides linkedvia an �-glycosidically-bound mannose to Ser and Thr residues(7). Biosynthesis is initiated at the endoplasmic reticulum (ER)by the transfer of mannose from dolichyl phosphate-activatedmannose to Ser or Thr residues of proteins that are entering thesecretory pathway (1). Further chain elongation takes place inthe Golgi apparatus using nucleotide activated sugars as donors.The initial mannosyltransfer reaction is catalyzed by an essentialfamily of dolichyl phosphate-D-mannose:protein O-mannosyl-transferases (PMTs) that is evolutionarily conserved from yeastto humans (5, 8–11). PMTs have been identified and extensivelycharacterized in yeast. In Saccharomyces cerevisiae, seven PMTfamily members (Pmt1p to -7p) are present (8, 9) that areintegral ER membrane proteins with seven transmembrane-spanning domains (12). Phylogenetic analyses indicate that thePMT family is subdivided into the PMT1, PMT2, and PMT4subfamilies, whose members include transferases closely relatedto S. cerevisiae Pmt1p, Pmt2p, and Pmt4p, respectively (1, 13).

In yeast members of the PMT family show a high degree ofconservation. Despite the high sequence homology, several featuressuggest that PMT1/PMT2 and PMT4 members form distinct func-tional subclusters. First, all PMT family members share three highlyconserved sequence motifs that, nonetheless, show significant vari-ations between PMT1/PMT2 and PMT4 subfamily members (13).Second, members of the PMT1 subfamily physically interact withmembers of the PMT2 subfamily, whereas the unique representa-tive of the PMT4 subfamily forms homomeric complexes (14).Third, Pmt1p/Pmt2p and Pmt4p complexes mannosylate differentacceptor proteins in vivo (15, 16). In S. cerevisiae, a subset ofO-mannosylated proteins such as Kre9p, Cts1p, Bar1p, Pir2p, andAga2p is exclusively mannosylated by Pmt1p/Pmt2p complexes(15). In contrast, Kex2p, Gas1p, Axl2p, and Fus1p are O-mannosylated by Pmt4p (15, 17). A third group of proteins includ-ing the WSC-family members, Mid2p and Ccw5p/Pir4p, is glyco-sylated by both complexes; however, Pmt1p/Pmt2p and Pmt4pmannosylate different domains of the protein (16, 18). In contrastto other types of glycosylation, signals causing O-mannosylation ofSer and Thr residues by PMT family members and determinants ofthe different substrate specificities among the PMT complexes areunknown.

Author contributions: S.S. designed research; J.H. and M.S. performed research; T.B.contributed new reagents/analytic tools; B.H. analyzed data; and J.H. and S.S. wrote thepaper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviations: TM, transmembrane domain; PMT, protein O-mannosyltransferase; ER,endoplasmic reticulum; GPI, glycosylphosphatidylinositol; OT, oligosaccharyl transferase.

†Present address: Department of Molecular Cell Biology, Max Planck Institute of Biochem-istry, Martinsried, Germany.

§To whom correspondence should be addressed at: Heidelberger Institut für Pflanzenwis-senschaften, Abteilung Zellchemie, Ruprecht-Karls-Universität Heidelberg, Im Neuenhei-mer Feld 360, 69120 Heidelberg, Germany. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/cgi/content/full/0700374104/DC1.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0700374104 PNAS � May 8, 2007 � vol. 104 � no. 19 � 7827–7832

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

July

8, 2

021

http://www.pnas.org/cgi/content/full/0700374104/DC1http://www.pnas.org/cgi/content/full/0700374104/DC1

Here, we present signals for Pmt4p-dependent O-mannosylation.Using specific model proteins, we demonstrate that S. cerevisiaePmt4p mediates O-mannosylation of Ser/Thr-rich membrane at-tached proteins, whereas Pmt1p/Pmt2p complexes act on both,soluble and membrane bound secretory proteins. The nature of themembrane-anchoring sequence is nonrelevant, as long as it isflanked by a Ser/Thr-rich domain facing the ER lumen. Based onthese results, an in silico analysis was performed, which identifiedpreviously uncharacterized Pmt4p substrates in S. cerevisiae. Ourwork shows that, in contrast to many other types of glycosylation,Pmt4p O-mannosylation signals are not just linear sequences ofproteins but instead are highly complex.

ResultsIn quest of determinants that bring about O-mannosylation and/ordefine specificity of PMTs toward different protein substrates, weperformed in silico analyses of known PMT substrates. No obvioussignals could be detected at the level of the primary and secondarystructure of the proteins (data not shown). However, we realizedthat all Pmt4p substrates characterized to date are membrane-associated. Thus, we hypothesized that membrane association is aprerequisite of Pmt4p-catalyzed O-mannosylation.

Disruption of Membrane Attachment of FUSwTMZZ Changes Its Spec-ificity for PMT Complexes. To follow up our hypothesis, we designedmodel substrates derived from the Pmt4p substrate Fus1p (Fig. 1A–G). Fus1p is a type I integral membrane protein involved in cellfusion during mating (19, 20). It was shown that Pmt4p mannosy-lation of the Ser/Thr-rich extracellular domain is crucial for cellsurface delivery (17).

We constructed a tandem protein A-tagged (ZZ) version of theFus1 protein lacking the cytosolic C-terminal domain (Fus1p withtransmembrane domain, FUSwTMZZ; Fig. 1B). To determinewhether this construct was O-mannosylated we expressed it in WTand various pmt� mutants (pmt1�, pmt2�, pmt1pmt2�, pmt3�,pmt4�, and pmt6�). In Western blot analyses of FUSwTMZZexpressed in WT cells (Fig. 2A), we detected four specific bandssimilar to the pattern described for the full-length protein which aredue to the processing of Fus1p (17). The same pattern was observedwhen FUSwTMZZ was expressed in pmt1�, pmt2�, pmt1pmt2�,

pmt3�, and pmt6� mutants, indicating that in these mutants theprotein was processed normally (Fig. 2A). In contrast, in pmt4�cells FUSwTMZZ migrated at the level of the predicted molecularmass of the unglycosylated protein (27.6 kDa). This band was notlabeled by the lectin Con A, which binds to O-mannosyl glycans[supporting information (SI) Fig. 8]. To further confirm that the27.6-kDa protein is nonglycosylated FUSwTMZZ, we expressed theprotein in the thermosensitive sec53 mutant. SEC53 encodes aphospho-mannomutase and when sec53 mutant cells are incubatedat 37°C (restrictive temperature), both N- and O-glycosylation areblocked (21, 22). In sec53 cells, FUSwTMZZ accumulates at 37°C asa single band with the same mobility as the band detected in pmt4�cells (Fig. 2B). Our data show that the vast majority of FUSwTMZZis specifically mannosylated by Pmt4p. In pmt4� cells, a very minorfraction is still processed to the mature form detected in WT cells(Fig. 2A). This residual activity is probably due to a compensatoryaction of one or more of the remaining Pmt proteins, whichmannosylate a fraction of FUSwTMZZ that accumulates in theabsence of Pmt4p.

To further analyze FUSwTM, a green fluorescent protein fusion(FUSwTMGFP) was expressed in WT and pmt4� cells. Whereas inWT cells, FUSwTMGFP localized to the plasma membrane (Fig. 3),transport of FUSwTMGFP to the cell surface was reduced in pmt4�cells, and the protein accumulated intracellularly (data not shown).

In summary, our data show that the deletion of the cytosolicdomain of Fus1p has no influence on O-mannosylation by Pmt4p.FUSwTMZZ shows very similar features as native Fus1p andtherefore represents an ideal basic model protein to analyze signalsfor Pmt4p-mediated O-mannosylation.

To test the role of the transmembrane domain (TM) for Pmt4p-dependent mannosylation, we deleted the TM from our modelsubstrate, resulting in FUSw/oTMZZ (Fus1p without transmem-brane domain; Fig. 1C). The protein was expressed in WT and inthe pmt� mutants and analyzed by Western blotting (Fig. 2C). Cellfractionation showed that in WT cells FUSw/oTMZZ is present in

Fig. 1. Schematic representation of the fusion proteins used in this study.The boxed sequence represents the entire Fus1p extracellular domain; signalpeptide is underlined. Constructs are fused to a C-terminal protein A-tag (ZZ).TM, transmembrane domain; AA, amino acids. Stippled, Fus1p; hatching,Axl2p; black, Kre9p; cross-hatching, Gas1p.

Fig. 2. Membrane attachment is a prerequisite for Pmt4p-mediated O-mannosylation. Western blot analyses of FUSwTMZZ (A and B) and intracellularFUSw/oTMZZ (C and D) expressed in WT and different mutant strains asindicated. Crude membranes isolated from �4 � 106 cells were analyzedunless indicated differently. (A) FUSwTMZZ is hypoglycosylated in the pmt4�mutant. (Left) Ten micrograms of membrane protein was analyzed. (B) Theform of FUSwTMZZ that is produced in the pmt4� mutant migrates at the levelof the unglycosylated protein produced in sec53 mutant cells at 37°C. (C andD) Intracellular FUSw/oTMZZ is not O-mannosylated (C) and migrates at thelevel of the unglycosylated protein produced in sec53 mutant cells at 37°C (D).

7828 � www.pnas.org�cgi�doi�10.1073�pnas.0700374104 Hutzler et al.

Dow

nloa

ded

by g

uest

on

July

8, 2

021

http://www.pnas.org/cgi/content/full/0700374104/DC1

the intracellular soluble and the membrane fraction (Fig. 2C anddata not shown). The intracellular protein shows an apparentmolecular mass of 23.3 kDa. Accumulation of this protein in mutantsec53 at 37°C (Fig. 2D) and lacking ConA staining (SI Fig. 8)confirmed that it is the unglycosylated FUSw/oTMZZ. Expressionof FUSw/oTMGFP in WT cells (Fig. 3) showed that the proteinlocalizes intracellularly as previously described for native Fus1p inpmt4� cells (17). These data demonstrate that deletion of the TMabolishes Pmt4p-mediated O-mannosylation.

Interestingly, we found that WT cells secrete a minor fraction ofFUSw/oTMZZ into the culture medium. Extracellular FUSw/oTMZZ showed an apparent molecular mass of �55 kDa (Fig. 4A).A partially glycosylated precursor of the secreted protein accumu-lated in temperature sensitive sec18 cells (23, 24), when ER exit isblocked at restrictive temperature (Fig. 4B). In the mutants pmt3�,pmt4�, and pmt6�, FUSw/oTMZZ with the same molecular mass asobserved in WT cells is secreted (Fig. 4A). In contrast, a shift to alower molecular mass is observed in the mutants pmt1�, pmt2�, andpmt1pmt2� (Fig. 4A). These data suggest that FUSw/oTMZZ isO-mannosylated by PMT1/PMT2-family members. Because only aminor fraction is glycosylated, FUSw/oTMZZ does not containpotent signals for PMT1/PMT2-family mediated O-mannosylation.

In summary, disruption of membrane integration of our modelsubstrate leads to a complete loss of Pmt4p-mediated O-mannosylation and permits modification by Pmt1p/Pmt2p com-plexes. We conclude that membrane localization of target proteinsplays an important role for recognition by Pmt4p.

The Distance of O-Mannosylation Sites from the Membrane Does NotInfluence Pmt4p-Dependent Glycosylation. To address the questionwhether Pmt4p complexes preferentially mannosylate Ser/Thr-richprotein domains that are in immediate proximity of the ERmembrane, we introduced spacer sequences between the Ser/Thr-rich domain and the TM of FUSwTMZZ. We generated twoconstructs, FUSwTM4AAZZ (Fig. 1D) and FUSwTM10AAZZ(Fig. 1E), which carry 4- and 10-aa spacers of alternating glycineand alanine residues, respectively. As expected, in WT and pmt4�cells, the proteins localized to the membrane fraction (data notshown). For both constructs, a pattern similar to that of Fus1p andFUSwTMZZ was observed (compare Figs. 5A and 2A) showingaccumulation of hypoglycosylated protein in the pmt4� mutant,which indicates that FUSwTMZZ, FUSwTM4AAZZ, andFUSwTM10AAZZ are processed in a similar, Pmt4p-dependentway. We conclude that the distance of the target sequence to the ERmembrane has no influence on O-mannosylation by Pmt4p, at leastin a range of up to 10 aa.

Membrane Localization but Not Intrinsic Features of the TM MediateO-Mannosylation by Pmt4p. Our experiments showed that the TM ofFUSwTMZZ is crucial for O-mannosylation by Pmt4p. Thus, weasked whether intrinsic signals for Pmt4p-dependent O-

Fig. 3. Cellular localization of FUSwTMGFP and intracellular FUSw/oTMGFP inWT cells. (Upper) FUSwTMGFP localizes mainly to the plasma membrane.(Lower) Deletion of the TM results in intracellular accumulation of FUSw/oTMGFP.

Fig. 4. Disruption of membrane attachment changes specificity for PMTcomplexes. (A) Western blot analyses of culture medium from WT and pmt�mutants expressing FUSw/oTMZZ. The extracellular protein is hypoglycosy-lated in pmt1�, pmt2�, and pmt1pmt2�. (B) Extracellular FUSw/oTMZZ accu-mulates in the ER of sec18 mutant cells at 37°C.

Fig. 5. The nature of the membrane-anchoring sequence does not affectPmt4p-mediated mannosylation. Shown are Western blot analyses of modelproteins expressed in WT and pmt� mutants. (A and C) Crude membranes (10�g of protein) were analyzed. (B and D) Crude membranes isolated from �4 �106 cells were analyzed. (A) Introduction of 4 or 10 spacing amino acids did notalter the Pmt4p-dependent processing of FUSwTMZZ. (B and C) Exchange ofthe Fus1p TM for TMs derived from Axl2p or Can1p did not disrupt O-mannosylation by Pmt4p. Both, FUS-AXL2TMZZ (B) and FUS-CAN1TMZZ (C) arehypoglycosylated in pmt4�. (D) Attachment of the TM from Can1p to thesoluble Pmt1p/Pmt2p substrate Kre9p triggers O-mannosylation by Pmt4p.

Hutzler et al. PNAS � May 8, 2007 � vol. 104 � no. 19 � 7829

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

July

8, 2

021

http://www.pnas.org/cgi/content/full/0700374104/DC1

mannosylation exist in the TMs of Pmt4p substrates. For thisreason, we exchanged the TM of FUSwTMZZ for TMs of otherPmt4p substrates (Axl2p) and non-substrates (Can1p).

The construct FUS-AXL2TMZZ (Fig. 1F) resulted from thefusion of the extracellular N-terminal domain of Fus1p with the TMand the cytosolic part of the type I membrane protein Axl2p. Axl2pis a Pmt4p substrate and O-mannosylation affects stability andlocalization of the protein thus resulting in abnormal buddingpattern (25). Cell fractionation experiments showed that FUS-AXL2TMZZ is exclusively membrane-associated in WT and pmt4�mutants (data not shown). Western blot analyses revealed under-glycosylation of the protein in the pmt4� mutant (Fig. 5B), dem-onstrating that FUS-AXL2TMZZ is mannosylated by Pmt4p. Thus,the exchange of the Fus1p TM for the TM of another Pmt4p targetdoes not affect the interaction with Pmt4p.

The next step in the analysis was to substitute the TM ofFUSwTMZZ with a TM derived from a protein that is not O-mannosylated. We selected TM XII of the plasma membranearginine permease Can1p which is oriented in the same way as theFus1p TM (26, 27). The Can1p TM was fused to the extracellularN-terminal domain of Fus1p resulting in FUS-CAN1TMZZ (Fig.1G). In WT and pmt4� cells, FUS-CAN1TMZZ localized specifi-cally to the membrane fraction (data not shown) and showed asimilar band pattern as observed for FUSwTMZZ (compare Figs.5C and 2A). In the pmt4� mutant, a major fraction of the proteinwas hypoglycosylated (SI Fig. 8) and showed the molecular mass ofthe unglycosylated protein (26.2 kDa), suggesting a similar process-ing of FUSwTMZZ, FUS-CAN1TMZZ, and native Fus1p. Theseresults show that FUS-CAN1TMZZ is a substrate for Pmt4p,although the TM is derived from a non-substrate. Our experimentsexclude the possibility of primary sequence based signals forPmt4p-mediated O-mannosylation encoded in TMs of specificPmt4p substrates.

To further confirm this result, we designed a gain-of-functionapproach by converting a non-substrate protein into a Pmt4psubstrate. Therefore, we fused TM XII of Can1p (see above) toa part of the soluble secreted protein Kre9p. Kre9p is involvedin �1,6-glucan assembly and is O-mannosylated exclusively byPmt1/Pmt2p complexes (15). The resulting construct KRE9-CAN1TMZZ (Fig. 1H) covered the Kre9p-derived N-terminalsignal sequence, the Ser/Thr-rich target sequence for Pmt1p/Pmt2p-mediated O-mannosylation and TM XII from Can1p. InWT, pmt4�, and pmt1pmt2� mutant strains, the fusion proteinKRE9-CAN1TMZZ localized to the membrane fraction (datanot shown). In WT cells, we detected two bands with markedlydecreased mobility compared with the predicted molecular massof 46.3 kDa of the nonglycosylated protein, indicating thatKRE9-CAN1TMZZ is O-mannosylated (Fig. 5D). In mutantpmt4� or pmt1pmt2�, respectively, KRE9-CAN1TMZZ is hypo-glycosylated and in addition to the dominant bands around �50kDa, further hypoglycosylated and/or degraded low molecularmass forms appear (Fig. 5D). These results indicate that Pmt1p/Pmt2p complexes as well as Pmt4p complexes participate in theprocessing of KRE9-CAN1TMZZ.

In summary, the addition of a single TM (derived from anon-substrate of O-mannosylation) to a soluble Ser/Thr-rich pro-tein domain (derived from a Pmt1p/Pmt2p substrate) triggered therecognition and modification by Pmt4p.

Deletion of the Glycosylphosphatidylinositol (GPI) Anchor of Gas1pDiminishes Pmt4p-Dependent O-Mannosylation. Next we addressedthe question whether the type of membrane association isrelevant for the modification by Pmt4p and examined theGPI-anchored plasma membrane protein Gas1p (Fig. 1I), a�1,3-glucanosyltransferase required for cell wall assembly (28).The protein is O-mannosylated by Pmt4p (15) and additionallyN-glycosylated. We performed Western blot analyses of theendogenous Gas1p from WT, pmt4�, and pmt1pmt2� mutants to

confirm these findings (Fig. 6 Upper). Gas1p is hypoglycosylatedonly in the pmt4� mutant, and Pmtl/Pmt2p complexes obviouslydo not recognize the protein as substrate for mannosyl transfer.To ensure that the change in molecular mass of Gas1p in thepmt4� mutant is due to a reduced amount of O-linked glycans,N-linked carbohydrates were removed by treatment with endo-N-acetylglucosaminidase H (EndoH). Although EndoH treat-ment reduced the apparent molecular mass by �30 kDa, themass difference between Gaslp from WT and pmt4� strainsremained (Fig. 6 Upper).

To test the impact of the GPI anchor on the O-mannosylationstate of Gas1p, we constructed a tagged truncated version of theprotein lacking the C-terminal hydrophobic GPI anchor signal thusgiving rise to the soluble secretory protein GAS1�GPIZZ (Fig. 1J).In WT, a major band with an apparent molecular mass of �120 kDacould be detected (Fig. 6 Lower). After EndoH treatment, theapparent molecular mass was decreased by �30 kDa indicatingsimilar N-glycosylation as compared with the native Gas1p protein.In pmt4� and pmt1pmt2� mutant cells, the same results wereobtained, indicating that deletion of the GPI anchor attachment sitealso eliminated Pmt4p O-mannosylation. Interestingly, hypoglyco-sylated GAS1�GPIZZ was not secreted into the medium but ratheraccumulated intracellularly (Fig. 6 Lower and data not shown).

From these results, we conclude that disruption of membraneattachment of Gas1p strongly affects its ability to interact withPmt4p complexes and that, in accordance with previously publishedresults (17), O-mannosylation, especially that mediated by Pmt4p,plays an important role in the secretion of proteins.

In Silico Identification of Pmt4p Substrates. Our data demonstratethat Pmt4p specifically acts on secretory proteins with an ER-luminally oriented Ser/Thr-rich region flanked by a membraneanchor. Based on these results, we performed an in silico search forputative Pmt4p substrates in S. cerevisiae. We screened the pro-teome of S. cerevisiae for proteins with at least one TM (found witha sliding window of 18 aa) and containing a region of �20 aa wherethe percentage in Ser/Thr is 40% or higher. Of 5,888 sequencesscanned, 51 confirmed to these criteria (SI Table 1). Twenty of theseproteins are putative GPI-anchored proteins and 31 are integralmembrane proteins. Out of the last group, we cloned taggedversions of Opy2p [an integral membrane protein that functions inthe signaling branch of the high-osmolarity glycerol pathway (29)],Prm5p [a pheromone-regulated protein that is predicted to haveone TM and is induced during cell integrity signaling (30, 31)],

Fig. 6. The type of membrane association is nonrelevant for Pmt4p-mediated modification. Shown are Western blot analyses of endogenousGas1p and GAS1�GPIZZ from WT and pmt� mutants. N-glycans were removedby EndoH as indicated. (Upper) Gas1p is hypoglycosylated only in pmt4�. (Lower)GAS1�GPIZZ is no longer mannosylated and accumulates intracellularly.

7830 � www.pnas.org�cgi�doi�10.1073�pnas.0700374104 Hutzler et al.

Dow

nloa

ded

by g

uest

on

July

8, 2

021

http://www.pnas.org/cgi/content/full/0700374104/DC1http://www.pnas.org/cgi/content/full/0700374104/DC1

Rax2p [which is involved in the maintenance of bud site selectionduring bipolar budding (32)], and YNL176c and analyzed them inWT and pmt4� mutant strains. For the tagged proteins Opy2p,Prm5p, Rax2p, and YNL176c, we could find a shift to lowermolecular weights or an accumulation of incompletely processedforms of the proteins in the pmt4� mutant, indicating Pmt4p-dependent processing (Fig. 7). The in silico identification of previ-ously uncharacterized Pmt4p substrates confirms that membraneassociation is a prerequisite of Pmt4p-mediated protein O-mannosylation.

DiscussionUsing specific model proteins, we identified signals that determinePmt4p-dependent O-mannosylation in yeast. We demonstrate that,in contrast to many other types of glycosylation, Pmt4p O-mannosylation signals are not linear sequences of proteins. Pmt4pmediates O-mannosylation of proteins, which are membrane at-tached and bear a Ser/Thr-rich domain facing the ER lumen.

The only Pmt4p substrate described so far, which at the firstglance did not comply with our predictions, was S. cerevisiaeCcw5/Pir4p (16). Ccw5p is a cell wall mannoprotein that is attachedto �-1,3-glucan and does not contain an obvious TM or GPIattachment site. However, cell fractionation studies revealed thatCcw5p is membrane-associated during secretion to the cell surface(J.H. and S.S., unpublished data), thus meeting the prerequisites ofa Pmt4p substrate. Further, our conclusions are in agreement withthe previously published observation that chimeric membraneproteins derived from Fus1p, Mid2p and invertase (Suc2p) wereO-mannosylated in a Pmt4p-dependent manner when they carrieda Ser/Thr-rich target sequence (17). Moreover, in mammals, theonly well characterized substrate of the Pmt4p homologue POMT1(�-dystroglycan) is membrane-associated (33). Dystroglycan istranslated as a type I ER membrane propeptide that is then furtherprocessed into two subunits (� and �) during secretion (33). TheO-mannosylated Ser/Thr-rich domain of dystroglycan propeptide isseparated from the TM by �266 aa, suggesting that in mammalsfurther determinants might have evolved.

Our data prove that Pmt4p mannosylates exclusively membrane-bound proteins whereas Pmt1p/Pmt2p complexes act on bothsoluble and membrane proteins. What could be the molecular baseof that specificity? There are several options. (i) The first optioncould be differences in the kinetics of O-mannosyl transfer cata-lyzed by Pmt4p versus Pmt1p/Pmt2p complexes and thus residencetime of protein substrates at the ER membrane. Similar to N-glycosylation, O-mannosyl transfer occurs while proteins are trans-located into the ER lumen (1). Membrane anchoring increases theresidence time of nascent proteins at the ER membrane and thus

might aid and abet Pmt4p-mediated O-mannosylation. (ii) Thesecond option could be association of PMTs with different trans-locon complexes in the ER. It was recently reported that twoisoforms of the oligosaccharyl transferase (OT) complex associatespecifically with two different translocon complexes. Ost3p con-taining OT complexes associate specifically with the Sec61 trans-locon, whereas Ost6p containing OT complexes with the Ssh1translocon (34). Because O-mannosylation and N-glycosylation arecompetitive processes (16), it is conceivable that distinct PMTcomplexes associate specifically with translocon–OT supercom-plexes providing the molecular base of substrate specificity. (iii) Athird option could be positioning of Pmt4p complexes in specificmicrodomains of the ER membrane. In yeast, the existence ofdistinct membrane microdomains has been reported that are al-ready formed in the ER (35, 36). Interestingly, the Pmt4p substratesFus1p, Gas1p, Wsc1p, and Mid2p are associated with such deter-gent resistant membrane fractions (36, 37). Thus, Pmt4p complexesand the corresponding substrates might colocalize in specific ERmembrane microdomains. Because we cannot eliminate any ofthese possibilities at present, one of the main future tasks will be toelucidate the molecular mode of operation of the O-mannosylationmachinery in the ER.

Based on our results, we screened the S. cerevisiae proteindatabase for novel Pmt4p substrates and identified 51 candidates(20 putative GPI-anchored and 31 integral membrane proteins).Among them are the so far known Pmt4p substrates Fus1p (17),Axl2p (25), Gas1p, Kex2p (15), Mid2p, Wsc1p, and Wsc2p (18).Analyses of the glycosylation status of selected putative substrates(Opy2p, Prm5p, Rax2p, and YNL176c) indeed confirmed theirO-mannosylation by Pmt4p. Pmt4p-mediated O-mannosylation iscrucial for the stability, sorting and/or performance of proteinswhich are functionally of high relevance for cell growth anddevelopment as demonstrated for Axl2p, Mid2p and Fus1p (19, 25,38). Thus, our findings provide a tool to identify proteins that arepotential major players in various cellular processes in yeast andother fungi. The identification of such proteins is highly significantbecause fungal pathogens represent the major eukaryotic diseasecausing agents in humans as well as in agricultural important cropplants. Thus, it is important to identify new targets for the devel-opment of novel antifungal drugs. In addition, we are now able toscreen genomes of higher eukaryotes for putative PMT substrateswhich might help to elucidate the relevance of O-mannosyl glycansfor early stages of development and for vital physiological functionsof proteins in mammals and humans (1).

Materials and MethodsYeast Strains and Plasmids. Used for this study were the S.cerevisiae WT SEY6210 (MAT�, his3-�200, leu2-3, -112, lys2-801, trp1-�901, ura3-52, suc2-�9) (39); sec mutants SEY5188(MAT�, sec18-1, suc2-09, leu2-3,112 ura3-52) (24) andSFNY28-6C (MATa, sec53, ura3-52) (17); and pmt deletionstrains pmt1� (isogenic to SEY6210, pmt1::HIS3) (40), pmt2�(isogenic to SEY6210, pmt2::LEU2) (40), pmt3� (isogenic toSEY6210, pmt3::HIS3) (41), pmt6� (isogenic to SEY6210,pmt6::LEU2) (41), and pmt1pmt2� (isogenic to SEY6210,pmt1::HIS3, pmt2::LEU2) (40). Strain STY100 (isogenic toSEY6210, pmt4::KanMX) was derived by disruption of the PMT4gene by using a KanMX disruption cassette released frompSB119 by NotI digestion.

Yeast strains were grown under standard conditions and trans-formed following the method of Gietz et al. (42). Standard proce-dures were used for all DNA manipulations. S. cerevisiae genesindicated below were amplified by PCR on genomic DNA.

For plasmid pSB119, megaprimers were amplified by PCR on S.cerevisiae genomic DNA. The megaprimers were used for PCR onpFA6a-GFP-kanMX6 (43) to amplify the KanMX cassette. ThePCR product was cloned into pGEM-Teasy (Promega, Mannheim,Germany). For plasmid pMS3, Schizosaccharomyces pombe adh1

Fig. 7. In silico screening identified previously uncharacterized Pmt4p sub-strates. Shown are Western blot analyses of the protein A-tagged proteinsOpy2p, Prm5p, Rax2p, and YNL176c from WT and mutant pmt4�. Incom-pletely processed forms of the proteins accumulate in mutant pmt4�.

Hutzler et al. PNAS � May 8, 2007 � vol. 104 � no. 19 � 7831

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

July

8, 2

021

promoter sequence derived from pREP3-adh (44) was cloned as aHindIII/XhoI fragment into pYEplac195ZZ (gift from J. Stolz).For plasmids pMS7.1 (FUSw/oTMZZ) and pMS7.2 (FUSwTMZZ),fragments of the FUS1-coding sequence (bp 1–290 or 1–342) wereamplified by PCR and cloned into pMS3 by using XhoI/BamHI. Forplasmids pMS9.1 (FUSw/oTMGFP) and pMS9.2 (FUSwTMGFP), S.pombe adh1 promoter and FUS1-coding sequences (bp 1–290 or1–342) derived from pMS7.1 and pMS7.2, respectively, were clonedas BamHI/HindIII fragments into YEplac195-GFP (gift from J.Stolz). For plasmid pJH1(FUS-AXL2TMZZ), a fragment of theAXL2-coding sequence (bp 1405–2469) was amplified by PCR andcloned into pMS7.1 by using BamHI/NheI. For plasmids pJH6.1and pJH6.2 (FUSwTM4AAZZ and FUSwTM10AAZZ), bp 220–342of the FUS1-coding sequence were amplified by PCR and clonedinto pMS7.1 by using BamHI/NheI, resulting in plasmid pJH5.Oligonucleotide 573 (5�-GATCCGGCGCCGGCGCCGGCGC-CGGCGCCGGCGCCG-3�) was annealed and ligated intoBamHI-digested pJH5. For plasmid pJH8 (FUS-CAN1TMZZ),part of the CAN1 coding sequence (bp 1561–1653) was amplified byPCR and cloned into pMS7.1 by using BamHI/NheI. For plasmidpJH12 (KRE9-CAN1TMZZ), part of the KRE9-coding sequence(bp 1–768) was amplified by PCR and cloned into pJH8 by usingXhoI/BamHI. For plasmids pJH15 (OPY2ZZ), pJH16 (PRM5ZZ),pJH17 (RAX2ZZ), and pJH20 (YNL176cZZ), coding sequenceslacking the termination codon were amplified by PCR and productswere cloned into pMS3 digested with XhoI/BamHI. For plasmidpJH23 (GAS1�GPIZZ), part of the GAS1-coding sequence (bp1–1584) was amplified by PCR and cloned into pMS3 by usingXhoI/BamHI.

Preparation of Cell Extracts. Yeast cells (5 � 108) from an expo-nentially growing culture were harvested and washed with 20 ml 10mM NaN3. Cell fractionation was performed as described (14). Toenrich secreted proteins, culture media were concentrated by usingVivaspin 500 centrifugal filter units (Sartorius, Goettingen, Ger-many) with a molecular weight cut-off of 10,000.

Protein Analyses in Temperature-Sensitive sec Mutants. Yeast cul-tures were grown to logarithmic phase at 25°C. At time 0, 108 cellswere harvested, washed with NaN3 (10 mg/ml), and frozen in liquid

nitrogen. Cultures were then shifted to restrictive temperature(37°C). Samples were taken at different times, and cell extracts wereprepared as described above.

Deglycosylation with Endo-�-N-Acetylglucosaminidase H. Ten tothirty micrograms of protein were treated with endo-�-N-acetylglucosaminidase H (EndoH; Calbiochem, Darmstadt, Ger-many) according to the manufacturer’s instructions. Mock incuba-tions were carried out without EndoH.

Western Blot Analysis. Proteins were fractionated by SDS/PAGEand transferred to nitrocellulose. Polyclonal anti-Gas1p antibodieswere used at a dilution of 1:2,500. Peroxidase-coupled anti-mouse-IgG antibody from rabbit (Sigma, Munich, Germany) and perox-idase-coupled anti-rabbit-IgG antibody from goat (Sigma) wereused at a dilution of 1:10,000. Protein–antibody complexes werevisualized by using the SuperSignal West Pico ChemiluminescentSystem (Pierce, Bonn, Germany).

Light Microscopy. Cells were immobilized by 0.8% agarose beforemicroscopic observation. Specimens were viewed by using anLSM510-Meta confocal microscope (Carl Zeiss, Jena, Germany)with �100 PlanApochromat objective (numerical aperture 1.4).Fluorescence signal of GFP (excitation 488 nm, Ar laser) wasdetected by using a bandpass emission filter 505–530 nm.

In Silico Identification of Pmt4p Substrates. We wrote a computerprogram that, in a first step, identified the proteins in the data setthat have a transmembrane segment, using the Kyte and Doolittlealgorithm (45). The selected proteins were then subjected to asearch for a region of at least 20 aa rich in serine or threonineresidues (at least 40%) adjacent to the transmembrane segment.

We thank A. Metschies and B. Jesenofski for excellent technicalassistance; L. Popolo (University of Milan, Milan, Italy), K. Simons (MaxPlank Institute of Molecular Cell Biology and Genetics, Dresden,Germany), and J. Stolz (University of Regensburg, Regensburg, Ger-many) for generously providing strains, plasmids, or antibodies; M.Lommel for many helpful discussions; and J. Stolz and M. Büttner forcritical reading of the manuscript. This work was supported by EuropeanUnion Grant FUNGWALL (EU-Project LSHB-CT-2004-511952).

1. Lehle L, Strahl S, Tanner W (2006) Angew Chem Int Ed Engl 45:6802–6818.2. Van den Steen P, Rudd PM, Dwek RA, Opdenakker G (1998) Crit Rev Biochem

Mol Biol 33:151–208.3. Spiro RG (2002) Glycobiology 12:43R–56R.4. VanderVen BC, Harder JD, Crick DC, Belisle JT (2005) Science 309:941–943.5. Martin-Blanco E, Garcia-Bellido A (1996) Proc Natl Acad Sci USA 93:6048–6052.6. Willer T, Prados B, Falcon-Perez JM, Renner-Muller I, Przemeck GK, Lommel

M, Coloma A, Valero MC, de Angelis MH, Tanner W, et al. (2004) Proc Natl AcadSci USA 101:14126–14131.

7. Willer T, Valero MC, Tanner W, Cruces J, Strahl S (2003) Curr Opin Struct Biol13:621–630.

8. Strahl-Bolsinger S, Immervoll T, Deutzmann R, Tanner W (1993) Proc Natl AcadSci USA 90:8164–8168.

9. Gentzsch M, Tanner W (1996) EMBO J 15:5752–5759.10. Jurado LA, Coloma A, Cruces J (1999) Genomics 58:171–180.11. Willer T, Amselgruber W, Deutzmann R, Strahl S (2002) Glycobiology 12:771–783.12. Strahl-Bolsinger S, Scheinost A (1999) J Biol Chem 274:9068–9075.13. Girrbach V, Zeller T, Priesmeier M, Strahl-Bolsinger S (2000) J Biol Chem

275:19288–19296.14. Girrbach V, Strahl S (2003) J Biol Chem 278:12554–12562.15. Gentzsch M, Tanner W (1997) Glycobiology 7:481–486.16. Ecker M, Mrsa V, Hagen I, Deutzmann R, Strahl S, Tanner W (2003) EMBO Rep

4:628–632.17. Proszynski TJ, Simons K, Bagnat M (2004) Mol Biol Cell 15:1533–1543.18. Lommel M, Bagnat M, Strahl S (2004) Mol Cell Biol 24:46–57.19. Trueheart J, Boeke JD, Fink GR (1987) Mol Cell Biol 7:2316–2328.20. Bagnat M, Simons K (2002) Proc Natl Acad Sci USA 99:14183–14188.21. Kepes F, Schekman R (1988) J Biol Chem 263:9155–9161.22. Ruohola H, Ferro-Novick S (1987) Proc Natl Acad Sci USA 84:8468–8472.23. Kaiser CA, Schekman R (1990) Cell 61:723–733.

24. Graham TR, Emr SD (1991) J Cell Biol 114:207–218.25. Sanders SL, Gentzsch M, Tanner W, Herskowitz I (1999) J Cell Biol 145:1177–

1188.26. Ahmad M, Bussey H (1986) Curr Genet 10:587–592.27. Regenberg B, Kielland-Brandt MC (2001) Yeast 18:1429–1440.28. Popolo L, Vai M (1999) Biochim Biophys Acta 1426:385–400.29. Wu C, Jansen G, Zhang J, Thomas DY, Whiteway M (2006) Genes Dev

20:734–746.30. Heiman MG, Walter P (2000) J Cell Biol 151:719–730.31. Jung US, Levin DE (1999) Mol Microbiol 34:1049–1057.32. Chen T, Hiroko T, Chaudhuri A, Inose F, Lord M, Tanaka S, Chant J, Fujita A

(2000) Science 290:1975–1978.33. Barresi R, Campbell KP (2006) J Cell Sci 119:199–207.34. Chavan M, Lennarz W (2006) Trends Biochem Sci 31:17–20.35. Malinska K, Malinsky J, Opekarova M, Tanner W (2003) Mol Biol Cell 14:4427–

4436.36. Bagnat M, Keranen S, Shevchenko A, Simons K (2000) Proc Natl Acad Sci USA

97:3254–3259.37. Malinska K, Malinsky J, Opekarova M, Tanner W (2004) J Cell Sci 117:6031–6041.38. Rajavel M, Philip B, Buehrer BM, Errede B, Levin DE (1999) Mol Cell Biol

19:3969–3976.39. Robinson JS, Klionsky DJ, Banta LM, Emr SD (1988) Mol Cell Biol 8:4936–4948.40. Lussier M, Gentzsch M, Sdicu AM, Bussey H, Tanner W (1995) J Biol Chem

270:2770–2775.41. Mrsa V, Seidl T, Gentzsch M, Tanner W (1997) Yeast 13:1145–1154.42. Gietz D, St Jean A, Woods RA, Schiestl RH (1992) Nucleic Acids Res 20:1425.43. Wach A, Brachat A, Alberti-Segui C, Rebischung C, Philippsen P (1997) Yeast

13:1065–1075.44. Willer T, Brandl M, Sipiczki M, Strahl S (2005) Mol Microbiol 57:156–170.45. Kyte J, Doolittle RF (1982) J Mol Biol 157:105–132.

7832 � www.pnas.org�cgi�doi�10.1073�pnas.0700374104 Hutzler et al.

Dow

nloa

ded

by g

uest

on

July

8, 2

021