Kex2 protease converts the endoplasmic reticulum α1,2-mannosidase of Candida albicans into a soluble cytosolic form

Cytosolic α-mannosidases are glycosyl hydrolases that participate in the catabolism of cytosolic free N-oligosaccharides. Two soluble α-mannosidases (E-I and E-II) belonging to glycosyl hydrolases family 47 have been described in Candida albicans. We demonstrate that addition of pepstatin A during the preparation of cell homogenates enriched α-mannosidase E-I at the expense of E-II, indicating that the latter is generated by proteolysis during cell disruption. E-I corresponded to a polypeptide of 52 kDa that was associated with mannosidase activity and was recognized by an anti-α1,2-mannosidase antibody. The N-mannan core trimming properties of the purified enzyme E-I were consistent with its classification as a family 47 α1,2-mannosidase. Differential density-gradient centrifugation of homogenates revealed that α1,2-mannosidase E-I was localized to the cytosolic fraction and Golgi-derived vesicles, and that a 65 kDa membrane-bound α1,2-mannosidase was present in endoplasmic reticulum and Golgi-derived vesicles. Distribution of α-mannosidase activity in a kex2Δ null mutant or in wild-type protoplasts treated with monensin demonstrated that the membrane-bound α1,2-mannosidase is processed by Kex2 protease into E-I, recognizing an atypical cleavage site of the precursor. Analysis of cytosolic free N-oligosaccharides revealed that cytosolic α1,2-mannosidase E-I trims free Man8GlcNAc2 isomer B into Man7GlcNAc2 isomer B. This is believed to be the first report demonstrating the presence of soluble α1,2-mannosidase from the glycosyl hydrolases family 47 in a cytosolic compartment of the cell.


INTRODUCTION
N-Glycosylation is one of the most common post-translational modifications of proteins in eukaryotic cells. The Nglycosylation pathway begins in the endoplasmic reticulum (ER), where the dolichol-bound Glc 3 Man 9 GlcNAc 2 oligosaccharide is assembled, then transferred to nascent proteins and processed by a-glycosidases which remove the three glucose residues and one mannose residue to generate the Man 8 GlcNAc 2 oligosaccharide (Herscovics, 1999a, b). In mammalian cells, N-oligosaccharides are further modified to complex glycans by different Golgi glycosyl hydrolases and transferases, whereas in lower eukaryotes such as Saccharomyces cerevisiae the only further modifications are by Golgi mannosyl transferases that lead to the biosynthesis of high mannose oligosaccharides (Herscovics, 1999a, b).
In lower eukaryotes, cytosolic a-mannosidase activity has not been reported; however, in mammals these enzymes have been studied extensively (Dutta & Majumder, 1984;Tulsiani & Touster, 1987;Haeuw et al., 1991;De Gasperi et al., 1992). These neutral cytosolic a-mannosidases participate in the hydrolysis of free N-oligosaccharides, which are generated by the ERAD of misfolded Nglycoproteins and by cleavage of dolichol-linked oligosaccharides (Spiro, 2004;Suzuki et al., 2006). This activity is not sensitive to 1-deoxymannojirimycin (1-DMJ), but is inhibited by swainsonine (Tulsiani & Touster, 1987;Haeuw et al., 1991) and activated by Co 2+ (Dutta & Majumder, 1984;Haeuw et al., 1991;De Gasperi et al., 1992;Weng & Spiro, 1996). Two cytosolic a-mannosidases have been described in rat liver. One trims Man 9 GlcNAc into Man 8 GlcNAc and Man 7 GlcNAc as the ER a1,2mannosidase hydrolyses this oligosaccharide, suggesting that it could be a soluble form of the ER enzyme (Grard et al., 1994). The other is a neutral a-mannosidase involved in the catabolism of cytosolic N-oligosaccharides released from ER (Grard et al., 1996). Studies of rat liver ER amannosidase II, which has catalytic and immunological properties similar to the cytosolic a-mannosidase, suggest that the cytosolic enzyme may be derived from the ER membrane-bound form (Bischoff & Kornfeld, 1986;Weng & Spiro, 1996). cDNA sequences of rat and mouse ER/ cytosolic a-mannosidase indicate that they encode soluble proteins, lacking signal sequences, and have homology to the vacuolar a-mannosidase from S. cerevisiae and other family 38 mannosidases (Bischoff et al., 1990;Costanzi et al., 2006). It has been suggested that this enzyme is synthesized in the cytoplasmic compartment, and then transported to the ER by a mechanism that involves proteolytic processing (Weng & Spiro, 1996;Spiro, 2004). Proteolytic processing also occurs with the vacuolar amannosidase from S. cerevisiae during import from the cytoplasmic compartment to the vacuolar lumen (Yoshihisa & Anraku, 1990).
We have previously purified and characterized two soluble a-mannosidases, E-I and E-II, from the human-pathogenic fungus Candida albicans. These represent 20 % and 80 % of the total soluble activity, respectively (Vázquez-Reyna et al., 1999;Mora-Montes et al., 2004). Based on the hydrolysis of natural substrates, we proposed that both proteins are a1,2mannosidases belonging to family 47 of glycosyl hydrolases, and that these may be involved in N-glycan processing (Mora-Montes et al., 2004, 2006. We have demonstrated that E-I and a membrane-bound a-mannosidase activity can be converted into E-II by a proteolytic activity sensitive to pepstatin A (Mora- Montes et al., 2006).
Here, we studied the intracellular localization of a1,2mannosidases of C. albicans. A 65 kDa membrane-bound a1,2-mannosidase was localized in the ER-and Golgiderived vesicles, and a1,2-mannosidase E-I in the cytosolic and Golgi compartments. However, we found that enzyme E-II is an in vitro artefact generated by proteolysis during the preparation of cell extracts. Our data indicate that ER membrane-bound a1,2-mannosidase is processed into cytosolic enzyme E-I by Kex2, recognizing an atypical cleavage site. Analysis of the cytosolic fraction of C. albicans revealed that E-I is involved in the processing of free M 8 B to Man 7 GlcNAc 2 isomer B (M 7 B) N-oligosaccharide. This study appears to be the first report of an a1,2mannosidase of family 47 in the cytosolic compartment of a lower eukaryote.
Preparation of cell-free extracts and enzyme purification. Cellfree homogenates were prepared while maintaining all solutions at 4 uC. C. albicans yeast cells were collected by low-speed centrifugation, washed twice and resuspended in 50 mM MES/Tris buffer, pH 6.0 (buffer A), with or without protease inhibitors and were broken in a Braun MSK cell homogenizer using 0.45 mm diameter glass beads. The homogenate was centrifuged at 1000 g for 10 min and the supernatant was collected and further centrifuged at 105 000 g for 1 h (ultracentrifugation). The high-speed supernatant was collected, freeze-dried and kept at -20 uC until use. In some experiments, the pellet, consisting of a mixed-membrane fraction, was resuspended in buffer A and used to assay a-mannosidase activity.
For enzyme E-I purification, the high-speed supernatant obtained in the presence of 1 mM pepstatin A was purified by size-exclusion, anionic-interchange and concanavalin A-Sepharose affinity chromatography as described previously (Mora-Montes et al., 2006). To purify the membrane-bound a-mannosidase from C. albicans, C. glabrata and S. cerevisiae, the mixed-membrane fraction was incubated for 1 h at 48 uC, followed by incubation at 220 uC for 30 min. This procedure solubilized about 50 % of the non-enzymic protein present in the mixed-membrane fraction, without significantly affecting the total a-mannosidase activity remaining in the pellet. Then, the preparation was ultracentrifuged, and the pellet was recovered, resuspended in 1 ml of buffer A with 3 mM EDTA, incubated 1 h at room temperature and ultracentrifuged as indicated. The pellet was resuspended in 1 ml buffer A containing 0.5 M NaCl, incubated for 1 h at room temperature and ultracentrifuged. Finally, the pellet was resuspended in 1 ml buffer A added with Triton X-100 in a protein : detergent ratio of 1 : 1, and was incubated for 1 h at room temperature and ultracentrifuged. The pellet was then resuspended in 1 ml buffer B (50 mM Tris, 0.2 M KCl, 1 mM CaCl 2 , 0.01 % Triton X-100, pH 7.2) and kept at 220 uC until use.
a-Mannosidase assay. Enzyme activity was measured using either the fluorogenic substrate MUaMan or the natural oligosaccharides M 9 , M 8 B or Man 5 GlcNac 2 -Asn, essentially as described previously (Mora-Montes et al., 2004).
Protein determination. Protein was measured by absorbance at 280 nm and by the method of Bradford (1976) using BSA as a standard.
Electrophoresis. SDS-PAGE was carried out using 10 % gels following standard protocols (Laemmli, 1970), and proteins were stained with Coomassie blue (Merril, 1990). Zymogram analysis in SDS-PAGE, using the fluorogenic substrate MUaMan, was performed for the enzyme detection as described previously (Mora-Montes et al.,

2004).
Protoplast generation and differential centrifugation of homogenates. Protoplasts were obtained following the protocol described by Ramírez et al. (1989). Briefly, yeast cells were collected by lowspeed centrifugation, resuspended at an OD 600 of 2-3 in buffer C (50 mM Tris/HCl buffer, pH 7.5, 1 M sorbitol, 0.8 M KCl and 10 mM MgSO 4 ), lyticase (0.25 mg ml 21 ) and 15 mM b-mercaptoethanol. After incubation at 37 uC for 30 min, nearly 100 % of the cells had been converted into protoplasts. These were washed twice with buffer C and resuspended in 10 mM phosphate buffer, pH 6.0, containing 3.0 mM MgCl 2 (buffer D). Samples of protoplasts were lysed with a Potter-Elvehjem and pestle (8-10 strokes) and the lysates were centrifuged at 1000 g for 10 min. The resulting supernatants were collected and 4 ml aliquots were loaded onto the top of a 35 ml, continuous 10-65 % (w/w) sucrose density gradient prepared with buffer D and centrifuged at 232 000 g for 4 h at 4 uC using a VTi 50 rotor (Beckman Coulter). Gradients were fractionated from the top and 1 ml fractions were collected. Alternatively, the protoplasts were incubated with 10 mM monensin for 1 h at 37 uC and lysed as described above.
Determination of free N-oligosaccharides in the cytosol.
Homogenates of protoplasts of C. albicans, C. glabrata or S. cerevisiae, treated or untreated with monensin, were fractionated in a continuous 10-65 % (w/w) sucrose density gradient as indicated above. The soluble fraction (typically fractions 1-4) was saved, freezedried, resuspended in 150 ml deionized water and applied onto a column (0.36105 cm) of Bio-Gel P-6. Fractions of 120 ml were collected and those enriched with the oligosaccharides were pooled, and analysed by high-performance anion-exchange chromatography as described previously (Mora-Montes et al., 2004).
Antibodies. The anti-a1,2-mannosidase antibody was generated from recombinant protein as follows. The open reading frame of the C. albicans MNS1 gene (GenBank/EBI accession AY167027) was cloned into the bacterial expression vector pET100/D-TOPO (Invitrogen), overexpressed in Escherichia coli and the recombinant protein was purified. Antibodies were raised in a male New Zealand White rabbit after intramuscular injection of 150 mg protein emulsified with complete Freund's adjuvant (day 0). Booster injections were given (150 mg protein emulsified with incomplete Freund's adjuvant) on days 15, 30, 45 and 60 and the animal was bled on day 75.
Immunoblots. Purified protein or samples (10 mg) of each amannosidase peak, separated after density-gradient fractionation of protoplast homogenates, were subjected to SDS-PAGE (10 %) and electrophoretically transferred to Hybond-C extra nitrocellulose membranes following standard protocols. The membranes were stained with Ponceau S Red to assess the efficiency of transfer. The membranes were blocked with 1 % BSA in PBS for 2 h, washed three times with 0.05 % Tween 20 in PBS for 10 min, and the primary antibody was added and incubated for 2 h at room temperature. Membranes were then washed three times with 0.05 % Tween 20 in PBS for 10 min, incubated for 2 h with the secondary antibody, and washed twice with 0.05 % Tween 20 in PBS for 10 min and once with PBS for 10 min. Bands were revealed with 3,39-diaminobenzidine (0.5 mg ml 21 ) and 3 % hydrogen peroxide in PBS. Reactions were stopped with deionized water.
Preparation of Kex2 enzymes. S. cerevisiae and C. glabrata secreted, soluble Kex2 proteins were produced in a commercial Pichia pastoris expression system (Invitrogen). The strain expressing S. cerevisiae Kex2 was a kind gift of G. Boileau (Lesage et al., 2001). The C. glabrata enzyme was produced in a similar manner. The C. albicans secreted, soluble form of the enzyme was expressed from the ACT1 promoter in C. albicans strain CAI4. All enzyme preparations were purified by a combination of size-exclusion and anion-exchange chromatography and tested with specific substrates and inhibitors to ascertain their specificity (O. Bader & B. Hube, unpublished data).
Proteolytic cleavage of membrane-bound a-mannosidase.
Twenty micrograms of protein from peak 2 and peak 3, obtained H. M. Mora-Montes and others after treatment with 10 mM monensin (see above), was resuspended in 50 ml buffer D added with 0.5 % Triton X-100, with or without protease inhibitors, and incubated for 1 h at 37 uC with gentle shaking. The reactions were analysed by immunoblotting using the anti-a1,2-mannosidase antibody. For the processing using recombinant Kex2, 20 mg of the partially purified membrane-bound amannosidase and 3 mg of recombinant Kex2 from S. cerevisiae, C. albicans or C. glabrata were resuspended in buffer B in a final volume of 20 ml, and incubated for 1 h at 37 uC with gentle shaking. Then, the reactions were applied to 4-12 % NuPAGE Bistris gels (Invitrogen) and analytic zymograms were carried out (see above).
N-terminal sequencing. The purified a-mannosidase E-I and the 52 kDa a-mannosidase generated by Kex2 were subjected to doubledimension electrophoresis using isoelectrofocusing ZOOM pH 4-7 strips and 4-12 % NuPAGE Bistris ZOOM gels (Invitrogen). The gels were washed three times with buffer A added with 1 % Triton X-100 for 20 min at room temperature, and a-mannosidase activity was detected after incubation with buffer A added with 40 mM MUaMan for 1 h at 37 uC. The proteins were transferred to a PVDF membrane and N-terminal sequencing was carried out by BioSynthesis.

Effect of protease inhibitors on intracellular distribution of a-mannosidase activity
Yeast cells of C. albicans were broken in the presence of protease inhibitors and the resulting homogenates were subjected to ultracentrifugation. a-Mannosidase activity was measured in the high-speed supernatant and in the mixed-membrane fraction. The results, summarized in Table 1, indicate that of the inhibitors tested, only pepstatin A affected the subcellular distribution of the enzyme. In the absence of pepstatin A, a-mannosidase was recovered mostly in the supernatant (78-82 %) while only 18-22 % remained associated with the membrane fraction. However, in homogenates prepared in the presence of pepstatin A, total activity was distributed in approximately equal proportions in the soluble (44 %) and membrane (56 %) fractions.
The ratio of the soluble E-I and E-II enzymes could be determined after separation of these two soluble activities by ion-exchange chromatography. In the presence of pepstatin A, only a single peak of a-mannosidase representing E-I was eluted by 0.1 M NaCl from a DEAE Bio-Gel A column. After eluting with 0.125 M NaCl, a second barely detectable peak corresponding to a-mannosidase E-II was obtained (Fig. 1).
The enzyme E-I was purified from pepstatin A-treated extracts using a combination of size-exclusion, ionexchange and affinity chromatography as described previously (Mora-Montes et al., 2006). Following this protocol, the enzyme E-I was purified 230-fold with a recovery of 34 % of the starting material (Table 2). Analytical electrophoresis of the purified sample revealed the presence of two polypeptides, of 23 and 52 kDa (Fig. 2a). Only the latter was active on the fluorogenic substrate MUaMan as revealed by zymogram analysis (Fig. 2b), and was recognized by the anti-a1,2-mannosidase antibody (Fig. 2c). No immunorecognition was observed in samples containing pre-immune serum or lacking the primary antibody (Fig. 2d, e).
The processing of natural substrates such as M 9 oligosaccharide by the purified enzyme was similar to that described previously (Mora-Montes et al., 2006). This resulted in the production of M 8 B and mannose as the sole products of hydrolysis after 12 h of incubation, and M 7 B and Man 6 GlcNAc 2 after 24 h of incubation (data not shown). As was described for a-mannosidases E-I and E-II (Mora- Montes et al., 2004Montes et al., , 2006, the purified enzyme did not hydrolyse Man 5 GlcNAc 2 -Asn oligosaccharide (data not shown) and was more sensitive to 1-DMJ than to swainsonine, with IC 50 values of 0.22 mM and 0.54 mM, respectively.
Immunoblot assays using antibodies against a1,2-mannosidase and several organelle-marker proteins were carried out to determine the presence of a-mannosidase E-I in the peaks observed in Fig. 3(a). The results, illustrated in Fig. 4, indicate that the anti-a1,2-mannosidase antibody detected a polypeptide of 52 kDa in peaks 1 and 2, which corresponds to the molecular mass of a-mannosidase E-I (see above). The same antibody detected trace amounts of a 65 kDa protein in peak 2 which was enriched in peak 3. Hexokinase I, a cytosolic marker (Huh et al., 2003), was only detected in peak 1, while a Golgi marker was barely detected in peak 1 but was abundant in peak 2. CPY, a marker of acid vacuoles (Bryant & Stevens, 1998), was present in peaks 2 and 3, and calnexin, an ER marker (Parlati et al., 1995), was only detected in peak 3. Mock controls containing pre-immune serum or lacking the primary antibodies run in parallel gave no detectable signals. Therefore E-I was enriched in cytosolic and Golgi compartments. The anti-a1,2-mannosidase antibodies detected a 67 kDa polypeptide in both a-mannosidase peaks separated on sucrose density gradients of extracts from protoplasts of S. cerevisiae (Fig. 4). Results with organelle markers were similar to those obtained with amannosidase peaks from the C. albicans preparation. These results confirmed that S. cerevisiae had only one a1,2mannosidase, whereas C. albicans had two isoforms of a1,2-mannosidase.
An anti-a1,2-mannosidase antibody was raised against the MNS1 product (see Methods), which is a 65 kDa a1,2mannosidase that is predicted to be localized in the ER (Mora- Montes et al., 2007). Western analysis showed that Mns1 and a1,2-mannosidase E-I were both detected by this antibody. In order to demonstrate that both a1,2mannosidase isoforms are encoded by MNS1, immunoblot assays using anti-a1,2-mannosidase antibodies and cell homogenate from a mns1D null mutant were carried out. Neither protein was detected in cell homogenates from mns1D null mutant (Fig. 5, lane 2), whilst two protein bands with molecular masses of 65 and 52 kDa were  recognized by the antibodies in the cell homogenates from the wild-type strain (Fig. 5, lane 1) and the reintegrant control (Fig. 5, lane 3). Control assays using pre-immune serum did not give any detectable signal (Fig. 5, lanes 4 and  5). Additionally, the mns1D null mutant did not show any hydrolytic activity towards the natural substrate M 9 (data not shown). Therefore, MNS1 encodes both 65 kDa and 52 kDa, a1,2-mannosidase isoforms.

a1,2-mannosidase
The results in Fig. 4 indicate that a1,2-mannosidase E-I is associated with the cytosolic and Golgi markers, whereas a 65 kDa a1,2-mannosidase was enriched in the ER and to a lesser extent in Golgi fractions. To determine whether E-I was generated from the 65 kDa protein, the effect of monensin on a1,2-mannosidase localization was analysed. Monensin, a monovalent ion-selective ionophore, blocks the transport of vesicles from the medial to the trans cisternae of the Golgi complex (Rosa et al., 1992), leading to an accumulation of proteins in the ER. Protoplasts of C. albicans were incubated with 10 mM monensin (see Methods) before their disruption and fractionation in a continuous 10-65 % sucrose density gradient. Under these conditions, the a-mannosidase activity was not affected (data not shown) and was again distributed in three peaks with corresponding densities of 1.032±0.003 g cm 23 for peak 1 (fractions 1-3), 1.139±0.007 g cm 23 for peak 2 (fractions 22-23), and 1.184±0.002 g cm 23 for peak 3 (fractions 27-28). The a-mannosidase activity associated with peaks 1 and 2 was significantly decreased to 3 % and 6 %, respectively, and the rest of the recovered amannosidase activity (91 %) was associated with peak 3 (Fig. 6a). Anti-a1,2-mannosidase antibodies barely detected a-mannosidase E-I in peaks 1 and 2, whereas the 65 kDa a1,2-mannosidase was abundant in peak 3 (not shown). When protoplasts of S. cerevisiae were treated in the same way, the a-mannosidase activity was distributed Fig. 3. Distribution of a-mannosidase activity after densitygradient centrifugation of homogenates from C. albicans and S. cerevisiae. Protoplasts from C. albicans (a) or S. cerevisiae (b) were prepared and homogenized as described in Methods. A 4 ml sample of each homogenate was loaded onto the top of a 35 ml, continuous 10-65 % (w/w) sucrose density gradient and centrifuged at 232 000 g at 4 6C for 4 h in a VTi 50 rotor. The gradient was fractionated from the top and 1 ml fractions were collected. These were used to monitor enzyme activity [nmol MU min "1 (mg protein) "1 ] using MUaMan as substrate (open symbols) and the sucrose concentration (filled symbols). Fraction 1 corresponds to the top of the gradient.

Fig. 2.
Analytical zymogram and immunodetection of purified amannosidase. The purified enzyme (10 mg) was analysed by SDS-PAGE (10 %) either after heat denaturation (a) or without heating (b) and separated by electrophoresis. Lane (a) was stained with Coomassie blue and lane (b) was incubated with MUaMan, as described in the text, to reveal the a-mannosidase activity. After separation by SDS-PAGE, proteins were electrotransferred to nitrocellulose membranes and immunodetected with anti-a1,2mannosidase antibody (c), rabbit pre-immune serum (d) or antirabbit IgG-horseradish peroxidase antibody (e).
in two peaks associated with fractions 22-23 (peak 2) and 27-28 (peak 3) with corresponding densities of 1.140±0.003 g cm 23 and 1.186±0.005 g cm 23 , respectively. As for the monensin-treated protoplasts of C. albicans, there was a redistribution of the a-mannosidase activity associated with peaks 2 and 3, representing 4 % and 96 % of total recovered activity, respectively (Fig. 6b).
These results suggest that a-mannosidase E-I is generated from the 65 kDa a1,2-mannosidase by proteolysis in the Golgi compartment. Supporting this, we found that proteolytic processing of the 65 kDa a1,2-mannosidase to the 52 kDa form occurred if we used ER vesicles from monensin-treated cells as a source of the a1,2-mannosidase, and Golgi vesicles as a source of the processing protease (Fig. 7). The generation of 52 kDa a1,2-mannosidase E-I depended on addition of Triton X-100 to the incubations to solubilize the proteins from the vesicles (data not shown).
In order to establish the nature of the proteolytic activity, similar experiments were carried out in the presence of protease inhibitors (Fig. 7). Addition of PMSF at 0. The Golgi protease activity that processed the 65 kDa a1,2mannosidase into a1,2-mannosidase E-I was sensitive to high concentrations of PMSF and EGTA. This inhibition profile is similar to that observed for the Ca 2+ -dependent protease Kex2 (Fuller et al., 1988(Fuller et al., , 1989. Hence, we compared the a-mannosidase activity distribution in a C. albicans kex2D null mutant and in the wild-type strain Fig. 4. Immunodetection of a1,2-mannosidase and some organelle markers in fractions separated as indicated in Fig. 3. Protein samples (10 mg) of a-mannosidase peaks from C. albicans and S. cerevisiae were separated by SDS-PAGE, electrotransferred to nitrocellulose membranes and immunodetected with anti-a1,2mannosidase and organelle markers as described in Methods. Controls used rabbit pre-immune serum (Pre-immune serum), antirabbit IgG-horseradish peroxidase (Rabbit IgG-HRP) or antimouse IgG-horseradish peroxidase (Mouse IgG-HRP).  , 2006); however in the kex2D null mutant [total a-mannosidase activity 90.8±4.8 nmol MU min 21 (mg total protein) 21 ] more than 99 % of the a-mannosidase activity was associated with the mixed-membrane fraction. When wild-type or kex2D null mutant cells were broken in the presence of 1 mM EGTA, the a-mannosidase distribution was similar to that of untreated cells. These results suggest that Kex2 participates in the generation of a-mannosidase E-I. To determine whether Kex2 was able to process the 65 kDa a1,2-mannosidase into E-I, incubations of purified 65 kDa a1,2-mannosidase from mixed-mem-brane fraction and recombinant Kex2 from S. cerevisiae were carried out, and analysed in zymograms using the fluorogenic substrate MUaMan (see Methods). The results, illustrated in Fig. 8, indicate that Kex2 processed the membrane-bound 65 kDa a1,2-mannosidase into a-mannosidase E-I (lane 3), and this proteolytic cleavage was sensitive to EGTA (lane 4). The 65 kDa a1,2-mannosidase and E-I were run as controls (lanes 1 and 2). Similar results were obtained when we analysed the products with the anti-a1,2-mannosidase antibodies, or when we performed the reactions using ER membrane-bound a1,2-mannosidase from C. glabrata or recombinant Kex2 from C. albicans or C. glabrata (data not shown). However, recombinant Kex2 failed to hydrolyse the ER membranebound a1,2-mannosidase from S. cerevisiae (data not shown).
a-Mannosidase of 52 kDa generated by Kex2 processing (Fig. 8, lane 3) and the previously purified a-mannosidase E-I (Fig. 2) were subjected to N-terminal sequencing, and for both proteins the identical amino acid sequence Asp-Trp-Ile-Lys-Asn-Asp-Leu-Asp-Tyr-Thr-Phe-Asp-Tyr-Asn-Val-Asn-Thr-Phe-Glu was obtained. This sequence was localized at position 103-121 within the sequence for the MNS1 product. Therefore, the same Kex2-dependent proteolytic product was apparently generated in vitro and in vivo.

Effect of monensin on free N-oligosaccharides present in the cytosol
To obtain further insights into the role of a1,2-mannosidase E-I in the cytosolic compartment, the soluble fraction of protoplasts of C. albicans was recovered after fractionation in a continuous, 10-65 % sucrose density gradient (typically fractions 1-4) and was analysed by highperformance anion-exchange chromatography (see Methods). M 8 B and M 7 B were both present in the cytosol, representing 21 % and 79 % of the total free N-oligosaccharides, respectively (Table 3). When protoplasts of C. albicans were treated with monensin, almost all the cytosolic a-mannosidase E-I was absent and the levels of M 7 B decreased to 3 % of the total content of cytosolic free N-oligosaccharides. This was accompanied by an increase in the M 8 B levels (Table 3). Similar results were obtained using protoplasts of C. albicans CAI4, SC5314 (data not shown) and C. glabrata (Table 3). When similar assays were performed using protoplasts of S. cerevisiae, C. albicans kex2D null mutant or C. glabrata kex2D null mutant, only M 8 B was detected in the cytosolic compartment (Table 3). Therefore the generation of M 7 B was associated with the presence of a1,2-mannosidase E-I in the cytosol.

DISCUSSION
Core-N-glycan processing enzymes are of major importance to the growth and viability of fungi and play key roles in the biology of all eukaryotes. Our previous studies of C. albicans a-mannosidases showed that two activities could be identified and that about 85 % of soluble a1,2mannosidase activity corresponds to enzyme E-II (Mora- Montes et al., 2004). Here, cell disruption in the presence of pepstatin A generated membrane fractions with increased levels of a-mannosidase activity at the expense of a proportional decrease in the amount of soluble activity, which was enriched with a1,2-mannosidase E-I. These results suggest that pepstatin A inhibited the formation of a-mannosidase E-II. Hence, it is most likely that this enzyme corresponds to a soluble form of the membranebound a-mannosidase which is generated during sample preparation. Addition of pepstatin A did not affect the biochemical properties of purified a1,2-mannosidase E-I, which showed a molecular mass and N-mannan core processing similar to that previously described (Mora-Montes et al., 2006). Accordingly, E-I can be classified in the glycosyl hydrolase family 47. The relationship of the 23 kDa polypeptide with the purified enzyme remains undetermined.
Density-gradient centrifugation assays indicated that a1,2mannosidase activity of C. albicans protoplasts was associated with fractions with densities in the range reported for cytosolic, Golgi and ER compartments (Beaufay et al., 1974;Chrispeels et al., 1982;Chrispeels, 1983;Harris & Waters, 1996). This distribution contrasts with that exhibited by a similar homogenate from S. cerevisiae centrifuged under the same conditions, where amannosidase activity was separated into two peaks with densities in the range reported for Golgi and ER membranes.
Hexokinase I distribution confirmed that peak I was enriched with components from the cytosolic compart-  ment. In agreement with the peak density range, a Golgi marker gave a strong signal in peak 2. This peak was also enriched with the Golgi form of CPY (Bryant & Stevens, 1998), which has a molecular mass (70 kDa) significantly higher than that of 61.1 kDa predicted for this enzyme (Mukhtar et al., 1992). Peak 3 gave strong signals with ER but not Golgi and cytoplasmic markers, thus supporting the nature of these vesicles. The 68 kDa protein detected with the anti-CPY serum most likely corresponds to the ER form. Thus, the presence of hexokinase I and the absence of CPY and calnexin indicate that peak 1 did not contain vesicles or elements derived from the ER or Golgi. Because the marker for the Golgi compartment corresponds to a protein associated with the cytosolic face of the Golgi complex (Bloom & Brashear, 1989), it is possible that the barely detectable band in peak 1 represents trace quantities of the protein that is detached from the Golgi membranes.
These results indicate that a1,2-mannosidase E-I is localized in the cytosolic cell fraction.
The anti-a1,2-mannosidase serum detected two polypeptides, of 52 kDa and 65 kDa, in Golgi and ER-associated vesicles, respectively. The molecular mass of 65 kDa is consistent with that predicted for the C. albicans MNS1 gene product, and our results confirmed that this protein is Mns1, the ER membrane-associated a1,2-mannosidase (GenBank/EBI accession AY167027; Mora-Montes et al., 2007). Its presence in the ER compartment probably results, as has been described for S. cerevisiae (Massaad & Herscovics, 2001), from a Rer1-dependent mechanism of localization. The enzyme E-I was not immunodetected in a C. albicans mns1D null mutant; thus, this gene encodes both a1,2-mannosidases. The members of glycosyl hydrolase family 47 that have so far been identified in the C. albicans genome (Arnaud et al., 2005) are MNS1 and orf19.834. The latter is homologous to S. cerevisiae MNL1, which encodes an ER mannosidase-like protein required for degradation of glycoproteins (Jakob et al., 2001;Nakatsukasa et al., 2001). However, these products have low values of identity (17.5 %) and similarity (28.5 %). MNS1 and AMS1, the putative gene encoding the vacuolar a-mannosidase, showed identity and similarity of only 11.6 % and 19 %, respectively. These data reinforce the observation that both 65 kDa and 52 kDa polypeptides are encoded by MNS1.
Monensin blocked the transport of ER a1,2-mannosidase to Golgi and led to a depletion in the levels of a1,2mannosidase E-I in the cytosolic compartment. These results suggest that the soluble a1,2-mannosidase is generated in the Golgi complex by proteolytic processing of the 65 kDa membrane-associated protein, and then is transported to the cytosolic compartment. Indeed, our results indicate that the 65 kDa membrane-bound enzyme was processed to a1,2-mannosidase E-I by Golgi Kex2 serine protease, which activates precursors of secreted proteins by an endoproteolytic cleavage at the C-terminus of dibasic motifs (Newport et al., 2003). Moreover, inhibition of Kex2 by EGTA during the cell-breaking process did not change the a-mannosidase distribution in wild-type cells, suggesting that a1,2-mannosidase E-I is generated in vivo rather than in vitro. However, the membrane-bound a1,2-mannosidase from S. cerevisiae was not processed by Kex2. This is in agreement with previous reports indicating that ER membrane-bound a1,2-mannosidase is the unique isoform present in vivo in S. cerevisiae (Jelinek-Kelly & Herscovics, 1988;Herscovics, 1999a).
The N-terminal sequence of both the purified a1,2mannosidase E-I and the enzyme E-I proteolytically generated by Kex2 was located between amino acids 103 and 121 of the MNS1 gene product. This predicts the generation of a soluble polypeptide of 463 amino acids with a molecular mass of 52.8 kDa, containing all the amino acids required for enzyme activity. These characteristics are fulfilled by a1,2-mannosidase E-I. Kex2 carries out the endoproteolysis at the C-terminus of dibasic amino acid motifs (Lys-Arg and Arg-Arg) (Fuller et al., 1988). However, the N-terminal end of E-I is not preceded by a typical Kex2 recognition sequence. Instead, the sequences Ser 99 -Arg 100 and Ala 101 -Arg 102 are present. In vitro and in vivo studies have demonstrated that Kex2 can recognize both of these non-canonical sequences, but with lower affinity (Bevan et al., 1998;Brenner & Fuller, 1992). This may explain why only around 45 % of the a-mannosidase activity was present as E-I. Hence, the results indicate that Kex2 recognizes an atypical cleavage site on the ER membrane-bound a1,2-mannosidase. The mechanism utilized to transport the a1,2-mannosidase E-I to the cytosolic compartment still remains to be determined. Fig. 9 shows a model for the processing of membrane-bound a1,2mannosidase into enzyme E-I.
To gain further insights into the role of the cytosolic a1,2mannosidase, we determined the levels of free cytosolic Noligosaccharides in protoplasts treated and not treated with the ionophore monensin, which blocks glycosylation by affecting divalent cation transport into the Golgi. In agreement with previous reports (Chantret et al., 2003), protoplasts of S. cerevisiae showed only the presence of M 8 B in the cytosolic preparations. It is believed that this Noligosaccharide is generated by deglycosylation of glycoproteins during the ERAD (Chantret et al., 2003). For protoplasts of C. albicans, M 8 B and M 7 B were detected in the cytosolic compartment, representing 21 % and 79 % of the total oligosaccharides, respectively. Protoplasts depleted of enzyme E-I had reduced amounts of M 7 B oligosaccharide and increased levels of M 8 B. These results suggest that enzyme E-I is required for the trimming of M 8 B into M 7 B. Similar results were obtained using protoplasts of C. glabrata, indicating that ER membranebound a1,2-mannosidase from this yeast can be processed to a cytosolic isoform by Kex2. Similar results have been reported in human cells, where the downregulation of cytosolic a-mannosidase activity directly influences the amount and the structure of the N-oligosaccharides present in the cytosol (Suzuki et al., 2006).
The limited proteolysis of ER a1,2-mannosidase by Kex2 could represent a regulatory mechanism similar to that recently described for the human ER a1,2-mannosidase (Wu et al., 2007), with implications for the regulation of ERAD, N-glycosylation and the degradation of cytosolic free N-oligosaccharides. To our knowledge this is the first report of a cytoplasmic a1,2-mannosidase in lower eukaryotes belonging to family 47 of glycosyl hydrolases.