Hansenula polymorpha Pmt4p Plays Critical Roles in O-Mannosylation of Surface Membrane Proteins and Participates in Heteromeric Complex Formation

O-mannosylation, the addition of mannose to serine and threonine residues of secretory proteins, is a highly conserved post-translational modification found in organisms ranging from bacteria to humans. Here, we report the functional and molecular characterization of the HpPMT4 gene encoding a protein O-mannosyltransferase in the thermotolerant methylotrophic yeast Hansenula polymorpha, an emerging host for the production of therapeutic recombinant proteins. Compared to the deletion of HpPMT1, deletion of another major PMT gene, HpPMT4, resulted in more increased sensitivity to the antibiotic hygromycin B, caffeine, and osmotic stresses, but did not affect the thermotolerance of H. polymorpha. Notably, the deletion of HpPMT4 generated severe defects in glycosylation of the surface sensor proteins HpWsc1p and HpMid2p, with marginal effects on secreted glycoproteins such as chitinase and HpYps1p lacking a GPI anchor. However, despite the severely impaired mannosylation of surface sensor proteins in the Hppmt4∆ mutant, the phosphorylation of HpMpk1p and HpHog1p still showed a high increase upon treatment with cell wall disturbing agents or high concentrations of salts. The conditional Hppmt1pmt4∆ double mutant strains displayed severely impaired growth, enlarged cell size, and aberrant cell separation, implying that the loss of HpPMT4 function might be lethal to cells in the absence of HpPmt1p. Moreover, the HpPmt4 protein was found to form not only a homomeric complex but also a heteromeric complex with either HpPmt1p or HpPmt2p. Altogether, our results support the function of HpPmt4p as a key player in O-mannosylation of cell surface proteins and its participation in the formation of heterodimers with other PMT members, besides homodimer formation, in H. polymorpha.

Hansenula polymorpha is a thermotolerant methylotrophic yeast that can grow rapidly using methanol as its sole source of carbon and energy, and at high temperatures up to 48°C [31]. This yeast has been developed as a favorable host for the production of recombinant pharmaceuticals [32], since it has several important advantages as a versatile cell factory. It contains several strong and inducible promoters derived from genes of the methanol metabolic pathway, specifically methanol oxidase (MOX), dihydroxyacetone synthase (DHAS) and formate dehydrogenase (FMDH) [33,34]. Furthermore, H. polymorpha does not hyperglycosylate secreted proteins, which often causes a problem in heterologous protein production in S. cerevisiae [35,36]. Systematic studies on the H. polymorpha genes involved in N-linked glycosylation facilitated the development of glycoengineered H. polymorpha strains for the production of human-type N-glycans [37][38][39][40]. Our recent study reported that H. polymorpha has five PMT genes (HpPMT1, HpPMT2, HpPMT4, HpPMT5, and HpPMT6) encoding protein O-mannosyltransferases [41]. Lack of HpPmt1p results in increased temperature sensitivity and decreased O-glycosylation of the chitinase protein [41,42]. The Hppmt1pmt5Δ and Hppmt1pmt6Δ double mutants showed increased sensitivity to cell wall stressors and aberrant O-mannosylation of the surface glycoproteins HpWsc1p and HpMid2p [41]. Here, we report the functional and molecular characterization of the H. polymorpha PMT4 gene, which is the only member of the PMT4 subfamily. We present a line of evidence supporting our conclusion that HpPmt4p plays a critical role essentially in the O-mannosylation of surface membrane proteins. Moreover, we show that Pmt4p forms not only a homodimeric complex but also forms heterodimeric complexes either with Pmt1p or Pmt2p in H. polymorpha.

Materials and Methods
Yeast strains, plasmids, primers, and culture conditions The H. polymorpha strains and plasmids used in this study are listed in Table 1 and S1 Table, respectively. The primers used for the construction of plasmids and strains are listed in S2 Table. Yeast cells were grown at 37°C in YPD medium (1% yeast extract, 2% Bacto peptone, and 2% glucose). The transformation of H. polymorpha was performed according to the modified lithium acetate-dimethyl sulfoxide method [43]. A selective synthetic complete (SC) medium (0.67% yeast nitrogen base without amino acids, 2% glucose, 0.77 g/L drop-out amino

Construction of pmt mutants
The 1BQ-u strain was constructed in two steps from the 1B strain, a derivative of H. polymorpha CBS4732 (Ogataea polymorpha) [44]. First, the MOX ORF was replaced with the N302Q mutant of the ORF encoding human urokinase-type plasminogen activator [45]; then, in the 1BQ strain obtained, the URA3 gene was disrupted with the ADE2 selectable marker. The HpPMT1 gene in the 1BQ-u strain was disrupted using the pSS18-derived HpPMT1 disruption cassette, which was constructed in a previous study [42], generating the Hppmt1Δ mutant strain (1BU-u-pmt1). The HpPMT4 disruption cassette was constructed by introducing inverse recombination arms into a HpLEU2-equipped integrative vector according to a previously reported approach [46]. Briefly, H. polymorpha CBS4732 genomic DNA was digested with HindIII, self-ligated, and used as a template for PCR with the primers HpPMT4D_F and HpPMT4D_R. The PCR product containing inverse recombination arms was digested with EcoRI and XhoI for cloning between the EcoRI and SalI sites of the pCHLX plasmid [47]. The resulting plasmid pCSS27 was digested with HindIII to obtain the HpPMT4 disruption cassette. The plasmid pDUM3P600-HpPMT4D, used to construct the H. polymorpha pmt1pmt4Δ double mutant strain bearing an HpPMT4 allele under the control of the HpMET3 promoter (1BQ-u-pmt1/P MET3 -PMT4), was obtained as follows: The 600-bp HpMET3 promoter and the 750-bp partial 5 0 ORF region of HpPMT4 were PCR amplified from the H. polymorpha CBS4732 wild-type genomic DNA using the SalI-HpMET3_F/HpMET3_R-EcoRI and EcoR-I-HpPMT4D_F/HpPMT4D-ClaI_R primer pairs, respectively. The SalI/EcoRI-digested HpMET3 promoter was subcloned into SalI/EcoRI sites of pDUMOX-msdS (HA-HDEL) [37], generating pDUM3P600-msdS. Subsequently the EcoRI/ClaI-digested HpPMT4 fragment was cloned into EcoRI/ClaI-digested pDUM3P600-msdS, resulting in pDUM3P600-HpPMT4D, which possessed the incomplete HpPMT1 ORF under control of the HpMET3 promoter. The KpnI-digested plasmid pDUM3P600-PMT4D was integrated into the HpPMT4 locus of the 1BU-u-pmt1 by a single crossover recombination. The correct integration was verified by PCR.

Construction of expression vectors and strains for epitope-tagged Pmt4 proteins
To construct an expression vector for FLAG-tagged HpPmt4p, the DNA fragment containing the 523-bp promoter and full-length ORF without a stop codon of HpPMT4 was PCR amplified from H. polymorpha CBS4732 wild-type genomic DNA using the primer pair NheI-HpPMT4_F/ HpPMT4_R-AscI. The PCR product was digested with NheI/AscI and ligated into the NheI/AscIdigested pHIGAHT-4FLAG [41], resulting in the plasmid pHINHT-HpPMT4F. The 2.9 kb NheI/ClaI fragment containing the HpPMT4 promoter and ORF tagged with four units of FLAG was obtained from pHINHT-HpPMT4F and ligated with the ClaI/SalI-digested pDUMOX-MsdS (HA-HDEL) [37], generating an H. polymorpha expression vector for a FLAG-tagged HpPmt4p, pDUN-HpPMT4F. To generate an Hppmt4Δ strain expressing a FLAG-tagged HpPmt4p (Hppmt4Δ/PMT4F), the circular plasmid pDUN-HpPMT4F was introduced into the Hppmt4Δ strain.
To construct an expression vector for HA-tagged HpPmt1p, pHINZ-HpPMT1H, the PCR fragment containing the full-length ORF without the stop codon of the HpPMT1 gene was obtained from H. polymorpha CBS4732 genomic DNA using the primer pair BglII-HpPMT1_F/HpPMT1_R-XhoI. The PCR product was cloned as a BglII/XhoI-digested fragment into pHIGAZ-HpPMT1H [41], generating pHINZ-HpPMT1H, which was cut at the PstI site within the HpPMT1 ORF and integrated by homologous recombination into the genomic HpPMT1 locus of the 1BU-u-pmt1 (Hppmt1Δ) strain, resulting in a Hppmt1Δ/PMT1H strain. For the construction of an H. polymorpha wild-type strain expressing a FLAG-tagged HpPmt2p strain (1BQ-LA/PMT2F), the plasmid pHIHT-HpPMT2F [41] was cut at the AvaI site within the HpPMT2 ORF and integrated by homologous recombination into the genomic HpPMT2 locus of the 1BQ-LA strain. For construction of a H. polymorpha 1BU-u-pmt4 (Hppmt4Δ) mutant strain expressing a FLAG-tagged HpPmt2p (Hppmt4Δ/PMT2F), pHI-GAZ-HpPMT2F [41] was cut at the EcoRI site within the HpPMT2 ORF for linearization and integrated by homologous recombination into the genomic HpPMT2 locus of the Hppmt4Δ strain. To generate strains co-expressing either HA-tagged HpPmt1p and FLAG-tagged HpPmt4p (Hppmt1Δ/PMT1H/PMT4F) or FLAG-tagged HpPmt2p and HA-tagged HpPmt4p (Hppmt4Δ/PMT2F/PMT4H) in the CBS4732 strain background, the circular plasmid pDUN-HpPMT4F or the EcoRI-digested pHINZ-HpPMT4H were integrated in the Hppmt1Δ/ PMT1H or the Hppmt4Δ/PMT2F strain, respectively.

Construction of expression vectors and strains for epitope-tagged glycoproteins
For the construction of pHINZ-HpWSC1H, an expression vector of the HA-tagged HpWsc1p (HpWsc1p HA ), the DNA fragment containing the 500-bp promoter and the full-length ORF without the stop codon of HpWSC1 was amplified by PCR from CBS4732 genomic DNA using the primer pair BglII-HpWSC1_pm/HpWSC1_R-AscI. The PCR product was digested with BglII/AscI and ligated with BglII/AscI-digested pHIGAZ-6HA, generating pHINZ-HpWSC1H. To construct plasmid pHINZ-HpMID2F, an expression vector of FLAG epitope-tagged HpMid2p (HpMid2p FLAG ), the PCR fragment containing the 500-bp promoter and the fulllength ORF without the stop codon of the HpMID2 gene was obtained using the primer pair BglII-HpMID2_pm/HpMID2_R-AscI. The PCR product was cloned as a BglII/AscI-digested fragment into pHIGAZ-4FLAG digested with BglII and AscI enzymes, generating pHINZ-Hp-MID2F. The EcoRI-digested pHINZ-HpWSC1H or ScaI-digested pHINZ-HpMID2F was integrated by homologous recombination into the genomic HpWSC1 or HpMID2 locus of the wild-type, Hppmt1Δ, and Hppmt4Δ strains, respectively. For complementation of the Hppmt4Δ mutants expressing either HpWsc1p HA or HpMid2p FLAG by reintroduction of HpPMT4, the Hppmt4Δ/MID2F and Hppmt4Δ/WSC1H strains were transformed with the DraI-digested pHINZU-HpPMT4H or the circular plasmid pDUN-HpPMT4F, respectively. To generate pDLUMOX-HpYPS1H, the lacZ-HpURA3-lacZ cassette was excised from pTHpURA3-LZ digested with EcoRI, treated with the Klenow fragment, and inserted into pMOX-YPS1ct-His [49] at the unique PsiI site. The resulting circular plasmid pDLUMOX-HpYPS1H was transformed into wild-type, Hppmt1Δ, and Hppmt4Δ strains.

MAP kinase phosphorylation analysis
To detect the expression of phospho-Mpk1, the H. polymorpha strains were grown to an OD 600 of 1.0 in YPD medium at 37°C, and then one-half of the culture was harvested, while the second half was treated with 0.05% SDS, 2 μg/mL caspofungin (CAS), 2.5 μg/mL tunicamycin (TM), 0.2 mg/mL calcofluor white (CFW), 20 mM caffeine, 10 mg/mL Congo red (CR), or 0.5 M sodium chloride for 2 hr. The cells were harvested and lysed by vortexing with glass beads (425-600 μm in diameter, Sigma) in phosphatase inhibitor lysis buffer [41]. Supernatants were separated by centrifugation at 2,500 × g for 5 min at 4°C and analyzed with 10% SDS-PAGE. For detection of phosphorylated Mpk1 and phospho-Hog1 proteins, the anti-phospho-p44/ p42 MAPK and anti-phospho-p38 MAPK antibodies (Cell Signaling Technology) were used, respectively, and anti-ScHog1 antibody (Santa Cruz Biotechnology) was used to detect Hog1 protein. The anti-beta actin antibody (Abcam) was used as a loading control.

Preparation of membrane fractions and co-immunoprecipitation
Membrane fractions were isolated as previously described by Kim et al. [41]. HpPmt4p HA protein was immunoprecipitated from sodium deoxycholate (SDC) extract containing 300 μg of proteins by using the ProFound HA Tag IP/Co-IP Kit (Thermo Scientific). Similarly, HpPmt4p FLAG protein was immunoprecipitated from SDC extract containing an equal amount of protein in the final volume of 1 mL of lysis buffer using the FLAG-Tagged Protein IP Kit (Sigma).

Microscopy analysis
Yeast cells were grown to an OD 600 of~2 in SD medium, exposed to 2 mM cysteine for 24 hr, and harvested. Cells were adjusted to a final concentration of 1 × 10 7 cells/mL, fixed with 3.7% formaldehyde in a rotator for 10 min at RT, and washed with 1x PBS buffer three times. For cell wall material staining, cells were stained with fluorescent brightener 28 (CFW; Sigma Aldrich) at a final concentration of 5 mg/mL in a rotator for 10 min and washed with 1x PBS buffer three times. Light and fluorescence microscopy pictures were taken using a Zeiss LSM700 confocal microscope equipped with Axio Observer. Images were processed with ZEN2011 software (Zeiss).

Construction and growth phenotype analysis of an Hppmt4 null mutant strain
To obtain information on the function of PMT4 in H. polymorpha, we constructed an Hppmt4 null mutant strain in the CBS4732 derivative and analyzed its growth phenotypes. The sensitivity of the Hppmt4Δ mutant strain to various cell stresses, as well as antibiotic and antifungal drugs, was analyzed and compared to that of the Hppmt1Δ mutant strain containing a deletion of HpPMT1 (Fig 1A). The growth of the Hppmt1Δ and Hppmt4Δ mutant strains was significantly inferior to that of the wild-type strain in the presence of cell wall destabilizers such as CAS, CR, or SDS. However, the H. polymorpha pmt mutants showed marginal sensitivity to antifungals, such as amphotericin B (AMB) and CFW, and to an ER stress inducer TM, an inhibitor of Asn-linked glycosylation causing ER stress [50]. The retarded growth of both pmt mutants was more apparent in the presence of another ER stress inducer DTT, an inhibitor of disulfide bond formation [51], than in the presence of TM. Notably, the Hppmt4Δ mutant strain was more susceptible to cell wall stressors, such as caffeine and the antibiotic hygromycin B (HgB), and to the Pmt1p inhibitor R3A-1c, than the Hppmt1Δ strain. At elevated temperatures, the growth of the Hppmt4Δ mutant was marginally decreased, whereas the Hppmt1Δ strain displayed a profound temperature-sensitive growth defect. Such increased sensitivities of the Hppmt4Δ mutant strain to cell wall stressors and at high temperatures were also consistently observed in another independent Hppmt4Δ clone (S1A Fig). Altogether, these phenotypes suggest an important role for HpPMT4 in maintaining cell wall integrity. It is noteworthy that the Hppmt mutant strains displayed increased susceptibility to various salt stresses ( Fig 1B). When incubated at high salt concentrations, such as 1 M KCl or 0.5 M NaCl, the Hppmt1Δ and Hppmt4Δ mutant strains grew significantly slower than the wild-type strain. Interestingly, the Hppmt4Δ mutant appeared to be more susceptible to high concentrations of salts than the Hppmt1Δ strain. It is notable that the growth of the Hppmt1Δ and Hppmt4Δ mutants of the H. polymorpha CBS4732 strain was inhibited even by sorbitol and mannitol, which are known as osmotic stabilizers. For that reason, the increased sensitivity to cell wall disturbing reagents of the Hppmt1Δ and Hppmt4Δ mutants could not be complemented by the addition of 1 M sorbitol (Fig 1C). The supplementation with other osmotic protectors, such as 1 M glycerol or 1 M sucrose, also did not recover the growth defects of Hppmt1Δ and Hppmt4Δ mutants at high temperatures and in the presence of cell wall stressors (S1B and S1C Fig). Taken together, these results suggest that the loss of HpPmt4p results in significant alteration of cell wall integrity, generating increased sensitivity to cell wall and osmotic stresses.

O-mannosylation analysis of glycoproteins in Hppmt4Δ mutants
To investigate the influence of Hppmt4 mutation on O-glycosylation activity in vivo, the electrophoretic mobility of the endogenous chitinase, an extracellular glycoprotein that is exclusively O-glycosylated without N-glycosylation [37,42], was analyzed. While the chitinase of the Hppmt1Δ mutant migrated slightly faster than that of the wild-type strain, no apparent change was observed in the electrophoretic mobility of chitinase of the single Hppmt4Δ mutant (Fig  2A), suggesting that Pmt4p is not involved in O-mannosylation of chitinase in H. polymorpha. We further investigated effects of pmt4 and pmt1 mutations on glycosylation patterns of secreted His-tagged Yps1p (HpYps1p HIS ), which lacks a GPI anchor. Whereas HpYps1p is an aspartic protease largely localized on cell surface via GPI, HpYps1p HIS having a deletion of a GIP signal (deletion from ω+4 position to C-terminus) is efficiently secreted as a soluble form [49]. The hyperglycosylated forms of HpYps1p HIS were detected by immunoblot analysis, using an anti-His antibody (Fig 2B). After cleavage of the signal peptides, the molecular weight (MW) of secreted HpYps1p HIS was predicted to be approximately 55 kDa, while its N-glycosylated form in the ER is~75 kDa. The secreted HpYps1p HIS displayed a single broad band of higher MW (~90-110 kDa) due to the addition of O-mannosylation in the wild-type strain. An underglycosylated form of HpYps1p HIS of~70 kDa was apparently observed as a minor fraction in the Hppmt1Δ mutant, whereas the underglycosylated form of HpYps1p HIS was barely detected in the Hppmt4Δ mutant as a faint smear. After PNGase F treatment, which removes N-linked glycans from glycoproteins, the differential migration patterns of secreted HpYp-s1p HIS in the wild-type and Hppmt mutant strains were still seen, indicating that the difference in electrophoretic mobility is caused by differential O-glycosylation. Thus, this suggested that the secreted form of C-terminally truncated Yps1p is preferentially modified by Pmt1p, although Pmt4p may also be partly involved in the modification of Yps1p in H. polymorpha. We then examined whether HpPmt4p is required in O-mannosyl modification of cell surface glycoproteins, such as Wsc1p and Mid2p. For this analysis, either HpWsc1p with six copies of a C-terminal HA tag (HpWsc1p HA ) or HpMid2p with four copies of the C-terminal FLAG tag (HpMid2p FLAG ) was expressed in the H. polymorpha wild-type and pmt mutant strains. Although the deduced MW of HpWsc1p HA is 41.8 kDa, HpWsc1p HA migrated on SDS-PAGE with an apparent MW of approximately 100 kDa in the wild-type strain due to intensive Olinked mannosylation (Fig 2C). In the Hppmt1Δ mutant, we did not detect any significant difference in the migration pattern of HpWsc1p HA compared to that for the wild-type cells. In contrast, we clearly observed a remarkable reduction in the MW of HpWsc1p HA in the Hppmt4Δ mutant. We detected an abundant form of HpWsc1p HA at 72 kDa, which was not observed for the wild-type and Hppmt1Δ mutant strains (Fig 2C). Moreover, we observed a band of HpWsc1p HA smaller than 35 kDa, which is assumed to be the aberrant proteolytic product of unmodified HpWsc1p [27]. A similar pattern of MW change due to a lack of Omannosylation was obtained for HpMid2p FLAG in the Hppmt4Δ mutant strain. In the wildtype strain, smeared bands of HpMid2p FLAG of approximately 120-200 kDa, which is much larger than the predicted size of 46.8 kDa, were detected by anti-FLAG antibody. In the Hppmt1Δ mutant, the levels of the high MW forms of HpMid2p FLAG were slightly lower than those in the wild-type cells (Fig 2D, left panel). However, in the absence of HpPmt4p, the total amount of these high MW forms was dramatically decreased while small fragments of HpMid2p FLAG were generated (Fig 2D, right panel). The re-introduction of functional HpPmt4p in the Hppmt4Δ mutant strains expressing HpWsc1p HA or HpMid2p FLAG were shown to completely restore the altered O-mannosylation patterns and thereby the instability of HpWsc1 and HpMid2 proteins to those observed in the wild-type strains (S2 Fig). From these data, we conclude that HpPmt4p plays a major role exclusively in O-mannosylation of membrane associated proteins, which is critical for maintaining the stability of these membrane proteins.

Analysis of MAP kinase activation in the Hppmt4Δ mutant
In yeast, cell wall damage leads to the induction of the cell wall integrity (CWI) pathway, mediated by the mitogen-activated protein kinase (MAPK) Mpk1p, which results in numerous cellular responses [52]. To investigate whether the CWI pathway is activated due to cell wall damage caused by defective O-mannosylation in the Hppmt4Δ mutant, we analyzed the phosphorylation of HpMpk1p in the wild-type, Hppmt1Δ, and Hppmt4Δ mutant strains by western blot analysis using the anti-phospho-p44/p42 antibody. Intriguingly, the basal level of phosphorylated HpMpk1p, a key kinase in the cell wall integrity pathway, was markedly reduced in the Hppmt4Δ mutant strain, while the phosphorylation of HpMpk1p was slightly induced in the Hppmt1Δ mutant compared to that in the wild-type strain in the absence of other external stresses (Fig 3A and 3B). However, despite the decrease in the basal level of phosphorylated HpMpk1p in the Hppmt4Δ strain, the phosphorylation level was increased upon treatment with various cell wall disturbing agents, just as in the wild-type strain. Compared to moderate activation of HpMpk1p found in response to CFW, CR, CAS and SDS, the phosphorylation of HpMpk1p increased more significantly in the presence of caffeine (Fig 3A and S3 Table). Particularly, treatment with TM led to the most dramatic activation of HpMpk1p phosphorylation among tested cell wall stressors (Fig 3B and S3 Table). These results suggest that the CWI signaling pathway can still be activated by cell wall damage, even when the O-mannosylation of surface sensors such as Wsc1p and Mid2p, normally catalyzed by the Pmt4 protein, is severely reduced.
A previous study in S. cerevisiae showed that the high osmolarity glycerol (HOG) pathway was activated in response to zymolyase, which lyses cell walls of viable yeast cells [53], suggesting cooperative cross-talk between the CWI and HOG signaling pathways [54,55]. We thus explored whether reduced O-mannosylation in H. polymorpha would also lead to stimulation of the HOG pathway, which is mostly dedicated to the adaptation of yeast cells to osmotic stress. Under normal conditions, the levels of phosphorylated HpHog1p were found to be slightly higher in the Hppmt1Δ and Hppmt4Δ mutant strains than in the wild-type (Fig 3C), implying that the defect in cell wall integrity in these pmt mutants activates the HOG pathway in H. polymorpha even under normal growth conditions. The wild-type, Hppmt1Δ, and Hppmt4Δ mutant strains all showed further activation of the HOG pathway upon NaCl treatment. Altogether, these results indicate that, although the loss of HpPmt4p causes severe defects in O-mannosylation of cell surface sensor proteins, the mutant strain still retains its ability to activate the CWI and HOG MAPK signaling pathways in response to cell wall stress conditions.

Analysis of a conditional Hppmt1pmt4Δ double mutant
Our previous study showed that the Hppmt1pmt5Δ, Hppmt1pmt6Δ, Hppmt5pmt6Δ double mutants, and even a Hppmt1pmt5pmt6Δ triple mutant, were viable [41]. To investigate the effect of HpPMT4 deletion in the absence of HpPmt1p, we tried to construct an Hppmt1pmt4Δ double mutant strain. Despite several trials of transformation of a H. polymorpha pmt4 mutant strain with a PMT1 disruption cassette, we could not obtain the Hppmt1pmt4Δ double mutant strain. Along with the observation that Hppmt4Δ mutant showed an extremely severe sensitivity to the Pmt1p inhibitor R3A-1c (Fig 1A), this implied that the simultaneous loss of PMT1 and PMT4 functions is most likely lethal to cells in H. polymorpha. To verify this hypothesis, the conditional PMT4 allele under control of the HpMET3 promoter was introduced in the Hppmt1Δ mutant strain. This was accompanied with inactivation of the chromosomal wildtype PMT4 allele. Expression of the PMT4 conditional allele could be down-regulated by the controllable HpMET3 promoter, whose expression is repressed by the supplementation of cysteine but is induced under sulfur-limited conditions [56]. The growth of the Hppmt1Δ [P MET3 -PMT4] mutant strain was severely retarded on SD medium containing more than 2 mM cysteine, compared to that of the Hppmt1Δ single mutant (Fig 4A and 4B). These results supported the theory that the concerted action of HpPmt1p and HpPmt4p is indispensable for growth of H. polymorpha.
We further investigated cell morphology of the Hppmt mutants through microscopic observation. As shown in Fig 4C, the Hppmt1Δ and Hppmt4Δ mutants showed an aberrant swollen or enlarged shape and formed chains of cells, whereas the wild-type cells exhibited normal growth as single yeast cells with simple buds. Moreover, the conditional Hppmt1pmt4Δ double mutant displayed not only a large and dysmorphic phenotype but also prominent cell aggregation with an apparently aberrant (or random) budding pattern, reflecting inability of the cells to separate normally. To investigate the presence of chitin, the cells were stained with CFW. The birth scar, in which chitin has been degraded [57], was clearly observed in single cells of the wild-type strain. While the Hppmt1Δ and Hppmt4Δ single mutants showed a chitin deposition pattern similar to that of the wild-type cells, the conditional Hppmt1pmt4Δ double mutant cells displayed unusual accumulation of chitin between connected cells (Fig 4C, arrow), which might result in a failure of daughter cells to dissociate properly from mother cells.

Homo-and heterodimeric complex formation of HpPmt4 proteins
In S. cerevisiae, Pmt4p forms homomeric complexes in vivo [3]. To investigate whether this complex formation also occurs in H. polymorpha, we performed co-immunoprecipitation experiments using an HA epitope-tagged version of HpPmt4p (HpPmt4p HA ) and FLAG epitope-tagged HpPmt4p (HpPmt4p FLAG ). In order to confirm whether the epitope-tagged HpPmt4 proteins are functional in vivo, we tested functional complementation of the chemical reagent-sensitivity phenotype of the Hppmt4Δ mutant by expression of epitope-tagged HpPmt4 proteins. The Hppmt4Δ mutant cell was transformed with a plasmid containing HAor FLAG-tagged HpPmt4p. As shown in Fig 5A, the expression of HA-or FLAG-tagged HpPmt4p in the Hppmt4Δ mutant complemented the sensitivity to several cell wall  Hppmt1Δ (Δ1), and Hppmt4Δ (Δ4) mutant strains, which were grown to early exponential phase and then treated for 2 hr with 20 mM caffeine, 0.2 mg/mL CFW, 10 mg/mL CR, 2 μg/mL CAS, 0.05% SDS, or 2.5 μg/mL TM. The phosphorylated HpMpk1p was detected using the antiphospho-p44/42 MAPK antibody, and the protein loading was monitored using an anti-β-actin antibody. The relative values of HpMpk1p phosphorylation to the basal level value of the wild-type were indicated, which were obtained by measuring the signal intensity of western blots with Quantity One 4.6.6 software (Bio-Rad). Error bars represent standard deviation of duplicate measurements. (C) Analysis of HpHog1 MAP kinase activation. The total soluble protein samples were obtained from the wild-type (WT), Hppmt1Δ (Δ1), and Hppmt4Δ (Δ4) mutant strains, which were grown to early exponential phase in YPD containing 0.5 M NaCl for 2 hr. Immunoblotting was conducted using anti-phospho-p38 antibody to detect phospho-Hog1p and with anti-Hog1 antibody as a loading control. The strains used were wild-type 1BQ-LA (WT) and the Hppmt1Δ (Δ1) and Hppmt4Δ (Δ4) mutants.  destabilizers and ER stressors. These complementation experiments demonstrate that C-terminal epitope-tagging does not inhibit the function of Pmt4 protein in vivo in H. polymorpha. Next, to detect Pmt4p-Pmt4p complexes, HpPmt4p HA and HpPmt4p FLAG proteins were coexpressed in the Hppmt4Δ mutant strain. Proteins solubilized from the crude membrane fractions by treatment with Triton X-100 and sodium deoxycholate were analyzed by co-immunoprecipitation experiments using anti-HA agarose or anti-FLAG M2 affinity gel. The immunoprecipitate obtained was resolved with 8% SDS-PAGE and detected by western blotting using antibodies against HA or FLAG. The results from the experiment with the Hppmt4Δ mutant strain co-expressing HpPmt4p HA and HpPmt4p FLAG demonstrated that HpPmt4p forms homomeric complexes (Fig 5B).
Unlike S. cerevisiae PMT members, PMT2 and PMT4 subfamily members in mammals and flies form a heteromeric complex [58,59]. Furthermore, a very recent study showed that A. nidulans PmtC (subfamily PMT4) forms not only homodimers but also heteromeric complexes with either PmtA (subfamily 2) or PmtB (subfamily 1) [21]. To investigate whether HpPmt4p can interact with either HpPmt1p or HpPmt2p in H. polymorpha, we constructed H. polymorpha strains co-expressing either an HA epitope-tagged version of HpPmt1p (HpPmt1p HA ) and HpPmt4p FLAG , or FLAG epitope-tagged HpPmt2p (HpPmt2p FLAG ) and HpPmt4p HA . Coimmunoprecipitations were performed with anti-FLAG M2 affinity gel by using the solubilized membrane proteins derived from both strains, and then the corresponding immunoprecipitates were probed with anti-FLAG and anti-HA antibodies, respectively. When probed with anti-FLAG antibody, two protein bands of approximately 86 kDa and 89 kDa were detected in the strains bearing HpPmt4p FLAG (Fig 5C, upper panel) and HpPmt2p FLAG (Fig 5D, upper  panel), respectively. Immunoblots probed with anti-HA antibody displayed the bands of HpPmt1p HA (~91 kDa) and HpPmt4p HA (~86 kDa) in the samples from the HpPmt1p HA and HpPmt4p FLAG co-expressing strain (Fig 5C, lower panel) and HpPmt2p FLAG and HpPmt4p HA co-expressing strain (Fig 5D, lower panel), respectively. Furthermore, the HpPmt1p-HpPmt4p and HpPmt2p-HpPmt4p heterodimer complexes were also observed in the H. polymorpha DL-1 strains (S3 Fig), although their band intensities were much lower than the band intensity detected in the strain harboring HpPmt1p HA and HpPmt2p FLAG . These results strongly suggest that H. polymorpha Pmt4 proteins also participate in the formation of heteromeric complexes with either HpPmt1p or HpPmt2p.

Discussion
Protein O-mannosylation, initiated by PMT family members in the ER, is known to play important roles in various biological processes, including cell morphology, protein stability [27,60], protein secretion [2,25], cell wall integrity [5], the budding process [24], sorting [61] and localization of proteins, and virulence of pathogens [9,62]. Unlike the PMT1/PMT2 subfamilies, which are highly redundant, the PMT4 subfamily has only one representative per species. In the present study, we report the functional and molecular features of HpPmt4p, a unique representative of the PMT4 subfamily in a thermotolernat methylotrophic yeast H. polymorpha with high industrial potential. We showed that the function of HpPmt4p, as a key player in O-mannosylation of cell surface proteins, is critical for resistance to cell wall and osmotic stresses. Moreover, we presented a novel finding that HpPmt4p forms not only a homomeric complex but also a heteromeric complex with either HpPmt1p or HpPmt2p in H. polymorpha, which is distinctive from S. cerevisiae that is known to form exclusively heterocomplex between the PMT1 and PMT2 subfamilies.
The physiological effect of PMT4 deletion has been investigated in several organisms from yeast to humans. In S. cerevisiae, disruption of PMT4 exhibited no obvious growth defect phenotype, and only multiple PMT gene disruptions including a pmt4 mutation partially showed conspicuous phenotypes [5]. For example, the S. cerevisiae pmtlpmt3pmt4Δ triple mutant was only slightly affected in spore germination, whereas the S. cerevisiae pmt2pmt4Δ double mutant can grow only when osmotically stabilized. The S. cerevisiae pmt1pmt4Δ and pmt1pmt3pmt4Δ mutants are temperature sensitive, whereas S. cerevisiae pmtlpmt2pmt4Δ and pmt2pmt3pmt4Δ mutations seemed to be lethal. In contrast to S. cerevisiae PMT4, just a single deletion of PMT4 homologs of other yeast and filamentous fungal species generated apparent defects in cell morphogenesis. In S. pombe, the oma4 single mutant strain, which is a deletion of the ScPMT4 homolog, forms an abnormal cell wall and septum, thereby exhibiting severely altered cell morphology and cell-cell separation [63]. Hyphal formation of a C. albicans pmt4 mutant was defective on defined Lee's medium but increased under embedded or hypoxic conditions [9]. In the opportunistic human pathogen C. neoformans, CnPMT4 is also crucial for cell morphogenesis and virulence [10,62]. In another opportunistic human pathogenic fungus, A. fumigatus, deletion of AfPMT4 not only affects cell growth, hyphal morphology, and conidiation but also causes hypersensitivity to echinocandins. In the plant pathogen U. maydis, UmPMT4 is required for fungal pathogenesis through its essential role in appressorium formation and plant penetration [15]. Similarly, in another plant pathogenic fungus, B. cinerea, PMT4 plays a specific role in virulence [8]. Furthermore, mutations of the D. melanogaster POMT1 (PMT4 subfamily) causes defects in muscle attachment and changes in muscle contraction during larval development [64]. In mice, targeted deletion of the POMT1 gene results in embryonic lethality, which is due to structural and functional defects in the first basal membrane of the embryo [65]. Mutations in the human POMT1 cause not only Walker-Warburg syndrome (WWS) [66] but also muscle-eye-brain disease (MEB) [67]. In the present study, we also showed that just a single deletion of HpPMT4 resulted in severe defects in cell wall integrity and stress resistance of H. polymorpha, indicating certain unique roles of HpPmt4p in Omannosylation of glycoproteins, which cannot be fully complemented by the function of other PMT members.
Previous studies in S. cerevisiae and C. albicans showed the transcriptional alteration of PMT genes in response to defective O-mannosylation [52,68]. To explore the transcriptional regulation of the H. polymorpha PMT genes, we determined the mRNA levels of all PMT subfamily genes in the wild-type and pmt4Δ mutant strains. In contrast to the increased levels of the PMT1 subfamily (PMT1 and PMT5) transcripts in the pmt4 mutant of C. albicans [68], any compensatory up-regulation of other PMT genes was not detected in the H. polymorpha pmt4Δ mutant (S4A Fig). In addition, the transcript levels of all H. polymorpha PMT subfamilies were not significantly changed even in the presence of TM or a Pmt1p inhibitor. On the other hand, all five PMT genes of P. pastoris were reported to be extremely down-regulated during growth on methanol as compared to that on glycerol due to lower energy availability [11]. Thus, we In vivo complementation analysis of epitope-tagged HpPmt4 proteins. A series of 10-fold dilutions of yeast strain expressing the epitope-tagged HpPmt4 proteins were spotted on YPD plates containing 10 mM caffeine, 0.1 mg/mL CFW, 5 μg/mL HgB, 0.01% SDS, 0.5 μg/ mL TM, or 10 mM DTT and incubated for 2 days at 37°C. (B) Co-immunoprecipitation (co-IP) experiments to analyze interactions between HpPmt4 proteins. HpPmt4p HA and HpPmt4p FLAG were expressed individually or co-expressed in pairs in the H. polymorpha wild-type strain. SDC extracts were prepared and IP was performed using anti-HA agarose or anti-FLAG M2 affinity gel. Precipitates were resolved by 8% SDS-PAGE and analyzed by western blotting. Blots were sequentially probed with anti-HA and anti-FLAG antibodies. (C-D) Co-IP experiments to analyze interactions between HpPmt4p and either HpPmt1p (C) or HpPmt2p (D). HpPmt1p HA and HpPmt4p FLAG or HpPmt2p FLAG and HpPmt4p HA were expressed individually or coexpressed in pairs in the H. polymorpha wild-type strain. SDC extracts were prepared and co-IP was performed using anti-FLAG M2 affinity gel. Precipitates were analyzed by western blotting with anti-Flag or anti-HA antibody. The asterisk indicates a nonspecific protein band. tested the expression levels of H. polymorpha PMT genes on different culture media such as sulfur-free B media (synthetic media with 2% glucose without any sulfur source) [69], SD medium (synthetic minimal medium), and YPM medium containing 2% methanol as the sole carbon source (S4B Fig). The qRT-PCR analysis showed that although there was no significant alteration in the expression levels of the HpPMT genes in B and SD media, the PMT1, PMT2, and PMT6 genes were dramatically repressed by more than 2-fold on methanol-containing media. It is notable that, among the HpPMT genes coding for three major isoforms Pmt1p, Pmt2p, and Pmt4p, the transcript level of HpPMT4 was found to be maintained at a relatively constant level under various culture conditions tested in this study.
In yeast, damage to the cell wall is sensed by plasma membrane sensors in the WSC family (Wsc1p to Wsc4p) and Mid2p, which are highly O-mannosylated surface proteins [27]. We demonstrate that HpPmt4p mediates O-mannosylation primarily of cell surface proteins, such as HpWsc1 and HpMid2 proteins, which are involved in cell wall integrity signaling. Previously, we reported that the basal level of Mpk1 MAPK phosphorylation highly increased in the Hppmt1pmt5Δ and Hppmt1pmt6Δ mutant strains, in which hypomannosylation was also observed for HpWsc1 HA and HpMid2 FLAG proteins [41]. However, in the present study, we observed that the basal level of the phosphorylated Mpk1p in the Hppmt4Δ mutant strain was dramatically lower than that in the wild-type strain under normal conditions. This suggests that HpWsc1p and HpMid2p might be mannosylated by both PMT1/2 family members and Pmt4p, but in distinct regions, as shown in S. cerevisiae [27], and that the O-mannosylation of these sensor proteins by Pmt4p is required for maintaining basal activity of cell wall sensors under normal cell conditions. Until now, it has generally been considered that a member of the PMT1 subfamily interacts with a member of the PMT2 subfamily, whereas the PMT4 subfamily member self-dimerizes for maximal O-mannosyltransferase activity in yeasts and other fungi. However, in contrast to the findings for fungal species, the PMT2 subfamily and PMT4 subfamily are reported to form a heterodimeric complex in mammals. The co-immunoprecipitation experiments in S. cerevisiae showed that Pmt1p-Pmt3p and Pmt2p-Pmt5p are present in minor amounts, although Pmt1p-Pmt2p and Pmt5p-Pmt3p were the predominant complexes [3]. Similarly, in H. polymorpha, Pmt2p can also interact with Pmt5p, a minor member of the PMT1 family, in the absence of the preferred partner Pmt1p [41]. Both of these yeasts have a redundant member of the PMT1 and PMT2 subfamilies, which could substitute if the dominant members are lost. Notably, a very interesting report on complex formation in all pairwise combinations of three A. nidulans Pmts, one from each subfamily, was recently presented [21], describing the possibility of complex formation between PMT1 or PMT2 subfamily and PMT4 subfamily members in other fungal species. Our data from the co-immunoprecipitation analysis of the H. polymorpha wild-type strains co-expressing HpPmt1p HA and HpPmt4p FLAG or HpPmt2p FLAG and HpPmt4p HA strongly support that Pmt1p-Pmt4p and Pmt2p-Pmt4p complexes are also present in H. polymorpha. To our knowledge, this is the first case where the formation of Pmt1p-Pmt4p and Pmt2p-Pmt4p complexes is shown to occur in a yeast species that possesses a redundant member of the PMT1 and PMT2 subfamilies. Compared to the Pmt proteins of the budding yeasts S. cerevisiae and H. polymorpha, which contain two to three redundant members, many filamentous fungi, including A. nidulans, possess only one representative from each PMT subfamily. Animals have only the PMT2 and PMT4 subfamilies, except that all Pmts are missing in the nematode Caenorhabditis elegans [3]. Based on these findings, it can be speculated that PMT gene families have most likely evolved to lose redundant genes, while the complex formation necessary for enzymatic activity has diversified with simplification of the redundant PMT genes (Fig 6).
In this study, we showed that the function of H. polymorpha Pmt4 protein is critical, particularly for O-mannosylation of membrane proteins, which is crucial for protein stability. Likewise, in S. cerevisiae, the integral transmembrane proteins Axl2/Bud10 require Pmt4pmediated O-mannosylation for their proper folding and the avoidance of degradation in the Golgi apparatus [24]. In addition, un-O-mannosylated Fus1p accumulates in late Golgi structures, suggesting that O-mannosyl modification by Pmt4p functions as a sorting determinant for cell surface delivery of Fus1p [42]. These results point to different roles of Pmt4p from those of PMT1/2 subfamily members in O-mannosylation of glycoproteins. It seems that PMT1/2 subfamily members are mainly responsible for O-mannosylation of secretory proteins, while Pmt4p is instead focused on surface membrane proteins. Along with N-glycosylation engineering, O-glycoengineering has been attempted in order to produce recombinant secretory proteins with mammalian-type glycosylation patterns in several yeast species, including S. cerevisiae [70,71], P. pastoris [72], and Ogataea minuta [73]. Thus, understanding the specific functions of each PMT member in protein O-mannosylation in yeast is important to elucidate their differential effects on different target proteins, which should be considered when developing yeast expression systems either for production of secretory glycoproteins or for surface display of membrane glycoproteins.   HpPMT2, HpPMT4, HpPMT5, and HpPMT6) were determined by quantitative real-time PCR using SYBR Premix Ex Taq II (TAKARA) and 10 pmol of each forward and reverse oligonucleotide primer (S2 Table). Transcript levels relative to the ACT1 transcript levels were calculated by the modified formula 2 mean Ct ACT1 /2 mean Ct target . Error bars represent standard deviation of triplicated data. (A) The relative levels of five HpPMT transcripts under several stress conditions. H. polymorpha cells were grown to an OD 600 of 1.0 in YPD medium at 37°C, and then treated with 2.5 μg/mL TM or 60 μM Pmt1p inhibitor (R3A-1c) for 2 hr. (B) The relative levels of PMT mRNA in various media. The H. polymorpha wild-type cells cultured overnight were inoculated at an initial OD 600 of 0.4, grown to mid-logarithmic phase (OD 600 of 1.0) in YPD, and then transferred into YPM medium containing 2% methanol or B or SD minimal medium and were further cultivated for 2 hr. (DOCX) S1