Free Oligosaccharides in the Cytosol of Caenorhabditis elegans Are Generated through Endoplasmic Reticulum-Golgi Trafficking*

Free oligosaccharides (FOSs) in the cytosol of eukaryotic cells are mainly generated during endoplasmic reticulum (ER)-associated degradation (ERAD) of misfolded glycoproteins. We analyzed FOS of the nematode Caenorhabditis elegans to elucidate its detailed degradation pathway. The major FOSs were high mannose-type ones bearing 3-9 Man residues. About 94% of the total FOSs had one GlcNAc at their reducing end (FOS-GN1), and the remaining 6% had two GlcNAc (FOS-GN2). A cytosolic endo-β-N-acetylglucosaminidase mutant (tm1208) accumulated FOS-GN2, indicating involvement of the enzyme in conversion of FOS-GN2 into FOS-GN1. The most abundant FOS in the wild type was Man5GlcNAc1, the M5A′ isomer (Manα1-3(Manα1-6)Manα1-6(Manα1-3)Manβ1-4GlcNAc), which is different from the corresponding M5B′ (Manα1-2Manα1-2Manα1-3(Manα1-6)Manβ1-4GlcNAc) in mammals. Analyses of FOS in worms treated with Golgi α-mannosidase I inhibitors revealed decreases in Man5GlcNAc1 and increases in Man7GlcNAc1. These results suggested that Golgi α-mannosidase I-like enzyme is involved in the production of Man5-6-GlcNAc1, which is unlike in mammals, in which cytosolic α-mannosidase is involved. Thus, we assumed that major FOSs in C. elegans were generated through Golgi trafficking. Analysis of FOSs from a Golgi α-mannosidase II mutant (tm1078) supported this idea, because GlcNAc1Man5GlcNAc1, which is formed by the Golgi-resident GlcNAc-transferase I, was found as a FOS in the mutant. We concluded that significant amounts of misfolded glycoproteins in C. elegans are trafficked to the Golgi and are directly or indirectly retro-translocated into the cytosol to be degraded.

It is known that N-linked oligosaccharides play important roles in the quality control of glycoproteins. An oligosaccharide composed of 14 sugars (Glc 3 Man 9 GlcNAc 2 ) is transferred en bloc from a dolichol-linked donor to the Asn residue of nascent polypeptide chains by oligosaccharyltransferase (OST), 2 and the oligosaccharides serve as tags indicating the folding state of glycoproteins in the endoplasmic reticulum (ER). Then, properly folded and assembled glycoproteins are transported to their destinations such as the extracellular space, plasma membrane, or inner compartments via the Golgi apparatus, with some accompanying modifications on the oligosaccharides.
The presence of free oligosaccharides (FOSs) has been reported in the cytosol of several types of animal cells (1)(2)(3)(4)(5)(6). These FOSs were thought to be generated from the ER by the following pathways. First, ER-associated degradation (ERAD) of glycoproteins is involved in generation of FOS. Misfolded or unassembled glycoproteins are recognized by ER degradationenhancing ␣-mannosidase-like protein (EDEM), retro-translocated into the cytosol through the Sec61 translocon (7) or the Hrd1 complex (8), and subjected to ubiquitin-proteasome degradation. The N-glycans on the misfolded glycoproteins are released by peptide:N-glycanase (PNGase) in the cytosol before the degradation by the proteasome (9,10). Second, FOS may be generated as the hydrolysis products of a part of the dolichollinked oligosaccharides in the luminal face of the ER by the action of OST. Free triglucosylated oligosaccharide is then hydrolyzed by ER ␣-glucosidases and ER ␣-mannosidase I; thereafter, the resulting Man 8 -9 GlcNAc 2 are transported into the cytosol in an ATP-dependent manner (11). Third, monophospho-oligosaccharides may be released from dolichollinked oligosaccharide intermediates on the cytosolic face of the ER membrane, possibly by a cytosolic pyrophosphatase. The N,NЈ-diacetylchitobiose moiety of these FOSs (FOS-GN2) in the cytosol is believed to be hydrolyzed by a cytosolic endo-␤-N-acetylglucosaminidase (ENGase) to produce FOS with a single GlcNAc residue (FOS-GN1) (12)(13)(14). FOS-GN1 species are further trimmed by cytosolic ␣-mannosidase to form M5BЈ (Man␣1-2Man␣1-2Man␣1-3(Man␣1-6)Man␤1-4GlcNAc) (15)(16)(17). Subsequently, they are transported to the lysosome to * This work was supported by the 21st Century COE (Centers of Excellence) be degraded by lysosomal ␣and ␤-mannosidases (3,18). Although it has been known that significant amounts of FOSs are produced in the ER and cytosol, their biological functions or physiological significance remain to be uncovered.
Caenorhabditis elegans is a genetically and developmentally well characterized multicellular eukaryote. In addition, C. elegans shares many fundamental biological pathways with higher eukaryotes, which makes it attractive for investigating functions of various biological compounds, such as glycoconjugates. The structures of N-linked oligosaccharides on glycoproteins from mixed stages of C. elegans have been examined by several groups (19 -22). According to those reports, the most abundant N-glycans on glycoproteins are those of the high mannose-type (Man 3-9 GlcNAc 2 ); there are also minor amounts of methylated, fucosylated, phosphorylcholine-substituted hybrid-or complex-type oligosaccharides. Although fucosylated and phosphorylcholine-substituted oligosaccharides are unique to this lower eukaryote, fundamental sections of the glycoprotein synthetic pathway are well conserved.
Previously, we cloned a cDNA encoding ENGase from C. elegans (eng-1, F01F1.10) based on the sequence homology with Endo-M from the fungus strain Mucor hiemalis (23,24). Although ENGase is predicted to be involved in FOS metabolism in the cytosol, there has been no report that shows direct evidence of this. To clarify this speculation, we employed C. elegans as a model animal and analyzed FOS in the cytosol. In this report, we present structural analyses of neutral FOS in the wild-type and two mutant worms and show evidence indicating that a significant amount of FOSs are generated through ER-Golgi trafficking.

EXPERIMENTAL PROCEDURES
Extraction and Labeling of FOS from C. elegans-C. elegans N2 Bristol strain (wild type) and the mutant strains tm1208 and tm1078 obtained from the National Bioresource Project for the Experimental Animal Nematode C. elegans (Tokyo Women's Medical University, Japan) were maintained and cultured on nematode-growing medium agar plates or in liquid medium with Escherichia coli OP50 as food. Worms were harvested and washed with M9 buffer (22 mM KH 2 PO 4 , 42 mM Na 2 HPO 4 , 86 mM NaCl, and 1 mM MgSO 4 ). Contaminating food and worm debris were removed by the sucrose floatation method. To obtain cell extract of the worms, a suspension of ϳ2.5-ml of worms was resuspended in 8 ml of ice-cold FOS extraction buffer (20 mM Tris-HCl (pH 8.3), 10 mM EDTA, 1 mM deoxymannojirimycin (DMJ), 0.5 mM swainsonine) and disrupted by sonication using an Insonater 201M (Kubota, Japan) on ice. The extract of worms was centrifuged (17,300 ϫ g) for 20 min at 4°C, and the supernatant was further centrifuged (106,000 ϫ g) for 60 min at 4°C. This supernatant was used as a cytosolic fraction. For purification of FOSs, the cytosolic fraction was quickly heated at 100°C for 5 min to inactivate enzymes followed by addition of ethanol up to a concentration of 60% to precipitate proteins. After removal of the precipitates by centrifugation at 17,300 ϫ g for 20 min, the supernatant was concentrated and applied onto a Sephadex G-15 column (2.8 ϫ 40 cm, Amersham Biosciences) pre-equilibrated with 50 mM NH 4 OH. The fractions eluted after the void volume and until the corresponding elution position of Man were collected and concentrated. The solution was desalted using a Dowex 50 ϫ 2 column (H ϩ form, Sigma), followed by a Dowex 1 ϫ 2 column (OH Ϫ form, Sigma). The desalted FOS was lyophilized and PAlabeled according to a previous method (25). PA-FOSs were further applied onto a ConA-Sepharose 4B column (1.5 ϫ 6 cm, Amersham Biosciences), and 0.3 M methyl-␣-mannopyranoside-eluted fractions were concentrated and desalted by gel filtration.
HPLC Analyses-To separate the FOS-GN1 and FOS-GN2 species, reversed-phase HPLC was performed using a Cosmosil 5C-18 AR-II column (4.6 ϫ 250 mm, Nacalai Tesque, Japan). The column equilibrated with eluant A (water containing 0.1% trifluoroacetic acid) was washed with eluant A for 25 min at a flow rate of 1.2 ml/min, and then elution was performed by linearly increasing the acetonitrile concentration of eluant B (acetonitrile containing 0.1% trifluoroacetic acid) from 0 to 7% in 25 min, at the same flow rate. Almost all FOS-GN1 species were eluted in the flow-through fractions within 33 min, and FOS-GN2 species were eluted after 35 min in these conditions. Size-fractionation HPLC was carried out using a Shodex NH2P-50 column (4.6 ϫ 250 mm, Showa Denko, Japan). Two eluants were used: eluant C, 80% acetonitrile; eluant D, 20% acetonitrile. The column was equilibrated with C:D ϭ 9:1 (v/v) at a flow rate of 0.7 ml/min. After injecting a sample, elution was carried out with the same buffer for 5 min, and then linear gradient elution was performed to attain C:D ϭ 2:3 for more 30 min. PA-FOSs were detected by fluorescence using an excitation of 310 nm and an emission of 380 nm.
ESI-MS-The mass spectrometry was carried out using a PerkinElmer Life Sciences Sciex API-III, triple-quadrupole mass spectrometer with an atmospheric pressure ionization ion source. The analyses were operated in the positive ion mode. Samples were dissolved in 50% acetonitrile containing 0.05% formic acid and introduced into the electrospray needle by mechanical infusion through a microsyringe at a flow rate of 5 l/min. 1 H NMR Spectroscopy-Prior to NMR analysis, PA-FOS was exchanged in D 2 O (99.95 atom %D) with lyophilization, and finally dissolved in 200 l of D 2 O (99.95 atom %D). 1 H NMR spectrum was recorded at 600 MHz on an Avance (Bruker Biospin) instrument at a probe temperature of 333 K referenced with HOD ␦ ϭ 4.32.
␣-Mannosidase Inhibitor Treatment-The wild-type worms were harvested from nematode-growing medium agar plates with M9 buffer, washed several times, and resuspended in 2 ml of M9 buffer. Golgi ␣-mannosidase I inhibitors, DMJ (Merck, Germany) or KIF (Merck), were added to the worm suspension to a final concentration of 1 mM or 0.1 mM, respectively, and incubated for 4 h at 20°C with gentle agitation. After incuba-tion, worms were frozen in liquid nitrogen and subjected to extraction of FOS.
Genomic Southern Blot Analysis-The genomic DNA from the wild-type worms (20 g) was digested with HpaI (Toyobo, Tokyo), and then separated by electrophoresis on a 0.6% agarose gel. Separated products were blotted onto a Hybond-N ϩ membrane (Amersham Biosciences), and hybridized with an ECL-labeled probe in gold hybridization buffer (Amersham Biosciences). The DNA probe was made by PCR using a primer pair of TKp57 (gacagaatcgaaagtgaatagtctgg) and TKp232 (cgccatctatctcaactacaac) and cosmid F01F1 as a template. The eng-1 DNA was detected using ECL detection reagents (Amersham Biosciences).
Western Blot Analysis of C. elegans Extracts-Crude whole extracts of worms were separated by SDS-PAGE and blotted onto a polyvinylidene difluoride membrane. After blocking with 2% (w/v) skim milk, the blot was incubated with rabbit anti-ENG-1 polyclonal antibody (1:3000) raised against the recombinant ENG-1 expressed in E. coli, and horseradish peroxidase-conjugated goat anti-rabbit IgG (1:5000, Santa Cruz Biotechnology). ENG-1 was visualized using the SuperSignal West Pico Chemiluminescent Substrate (Pierce).

Structural Analyses of Neutral FOS in the Wild-type C.
elegans-We attempted to obtain high mannose-type FOS in the cytosolic fraction of the wild-type C. elegans, because cytosolic FOS might be produced in an early step of N-glycan synthesis and might have primitive structures without complex Golgi modifications. First, the worms were disrupted to extract FOS, and they were pyridylamine (PA)-labeled and purified by ConA-Sepharose column chromatography. The ConA-binding fractions were applied to a reversed-phase HPLC. Because it was previously reported that both PA compounds of FOS-GN1 and FOS-GN2 could be separated by reversed-phase HPLC (27), in our case, we supposed that the FOS-GN1 species must be eluted at the flow-through fraction (5-33 min, Fraction A) and FOS-GN2 species must be eluted later (after 35 min, Fraction B) (Fig. 1A). Each fraction was collected, concentrated, and further analyzed by size-fractionation HPLC. Upon HPLC of Fraction A, several peaks (1a-1h) were detected at the elution position of 3-11 Glc units (Fig. 1B, upper chart). A similar elution pattern was observed on HPLC of Fraction B, although the retention time of each peak (2a-2h) was a little longer compared with those of the corresponding peaks (1a-1h) on HPLC of Fraction A, because of the presence of one more GlcNAc (Fig. 1B, lower chart). By comparison with the retention times of standard PA-oligosaccharides, peaks 1a-1h and 2a-2h were presumed to be M3ЈϳM9Ј, G1M9Ј (with one GlcNAc) and M3ϳM9, G1M9 (with two GlcNAcs), respectively (see Tables 1  and 2 for abbreviations). The molecular masses of 1a-1g were analyzed by ESI-MS to confirm the sugar composition (Table  3). To analyze their structures in detail, Fractions A and B were treated with several kinds of exoglycosidases and analyzed by size-fractionation HPLC. By treatment of Fraction A with jack bean ␣-mannosidase, which hydrolyzes ␣1,2-, ␣1,3-, and ␣1,6linked Man residues, only one peak was found at the corresponding elution position of M1Ј, although there were very minor species refractory to jack bean ␣-mannosidase (Fig. 1C, middle chart). This fact suggested that almost all FOSs of the worm were constituted of ␣-linked Man residues. Next, we carried out a treatment of Fraction A with ␣1,2-specific mannosidase, which resulted in an increase of peak 1c and a decrease of peaks 1d, 1e, 1g, and 1h (Fig. 1C, bottom chart). Peak 1c was completely resistant to 1,2-␣-mannosidase (Fig. 1D), suggesting that peak 1c represents the M5AЈ isomer (Man␣1-3(Man␣1-6)Man␣1-6(Man␣1-3)Man␤1-4GlcNAc), because M5AЈ is the only M5 isomer that does not contain any ␣1,2-linked Man resi- Traces with arrowheads indicate the FOS structures of peaks. D, size-fractionation HPLC of peak 1c. Non-treatment (upper) and 1,2-␣-mannosidase treatment (lower). E, size-fractionation HPLC of peak 1h. Non-treatment (upper) and ␣-glucosidase treatment (lower). The peak indicated with the asterisk was derived from an enzyme solution.
dues. We further confirmed the structure of peak 1c by 1 H NMR. Four anomeric shifts from ␣-Man (indicated as 4, 4Ј, A, and B) and one from ␤-Man (indicated as 3) were found at ␦ 5.057, 4.789, 5.085, 4.832, and 4.770, respectively (Table 4). These anomeric and other chemical shifts of peak 1c were largely similar to those of the published data of non-pyridylaminated M5AЈ (28) or M5A-PA (29), supporting the notion that the isomeric structure of peak 1c was M5AЈ. The peak on the corresponding elution position of peak 1f also seemed to be resistant to 1,2-␣-mannosidase treatment. Meanwhile, ␣-glucosidase treatment of peak 1h altered its elution position to subsequently correspond with that of M9Ј (Fig. 1E), confirming this peak as representing glucosylated M9Ј (G1M9Ј). Peak 1g was also sensitive to ␣-glucosidase (data not shown) and appeared to represent G1M8Ј. Thus, the remaining peak at the elution position of 1f after 1,2-␣-mannosidase treatment of Fraction A was assumed to represent G1M7Ј converted from G1M8Ј and G1M9Ј. Fraction B was also analyzed in the same way. The amount of each FOS calculated from the peak areas of the HPLC chart (Fig. 1B) are summarized (Tables 1 and 2). As a result, M5AЈ was found to be the most abundant FOS in C. elegans (42.3 pmol/mg of protein), and 94% of the total FOS was comprised of GN1 species (FOS-GN1, 108 pmol/mg of protein, FOS-GN2, 6.9 pmol/mg of protein).
ENG-1 Is Involved in Conversion of FOS-GN2 into FOS-GN1-A large part of FOSs in the wild type was the FOS-GN1 species, which is supposed to be produced by ENGase in the cytosol. Our previous study on the substrate specificity of ENGase from C. elegans (ENG-1 or Endo-CE) demonstrated that the enzyme preferably hydrolyzed high mannose-type PA-oligosaccharides than complex-type ones, which is consistent with the idea that ENG-1 is involved in the FOS metabolism. Therefore, we attempted to confirm that ENG-1 indeed acts on FOS-GN2 species. We obtained a mutant of eng-1(tm1208), which harbors a genomic deletion of 511 bp in length over the first intron

Free Oligosaccharides Generated in C. elegans
and exon 3 of eng-1 gene as revealed by Southern blot analysis (Fig. 2, A and B). This genomic deletion results in the abolition of protein expression in the mutant (Fig. 2C). ENGase activity toward a substrate Man 5 GlcNAc 2 -PA was not detected at all in the lysate of the mutant (data not shown). Although no apparent morphological and behavioral phenotypes were observed, life span was slightly shorter than that of the wild type (data not shown). The FOS-GN1 and FOS-GN2 in this mutant were analyzed by the same method used for analyses of FOS in the wild type (supplemental Table S1). In the wild type, the amount of FOS-GN1 species (sum of M5Ј-M9Ј) was major, whereas in the eng-1 mutant (tm1208), GN2 species (M5-M9) was accumulated, although the total amount of FOS was smaller than that of the wild type (Fig. 2D). ENG-1 has no apparent signal sequence, and the protein expressed in CHO-K1 cells was distributed throughout the cell (Fig. 2E), suggesting that ENG-1 substantially contributes to the conversion of FOS-GN2 into FOS-GN1 in the cytosol.
Effect of Golgi ␣-Mannosidase I Inhibitors on the Production of FOS-We found that the M5AЈ isoform is the major FOS in the wild-type C. elegans. On the other hand, in mammals the M5BЈ isoform (Man␣1-2Man␣1-2Man␣1-3(Man␣1-6)Man␤1-4GlcNAc) is the major FOS structure (3)(4)(5)30), whereas the M5AЈ isoform is very minor (30). The mammalian  M5BЈ isoform is considered to be a result of the actions of the cytosolic ␣-mannosidase, which is classified as a class 2 ␣-mannosidase. However, no candidate cytosolic ␣-mannosidase homologue was found in the C. elegans genome as revealed by phylogenetic analysis (Fig. 3). Furthermore, we could not detect any neutral ␣-mannosidase activity in the cytosolic fraction of C. elegans using p-nitrophenyl-␣-mannopyranoside as a substrate (data not shown), although its activity in mammals is detectable using the same substrate.
Concerning the oligosaccharide structure of M5AЈ, it is identical to the part of the intermediate of N-glycan processing, which is formed in the cis-Golgi compartment by Golgi ␣-mannosidase I. Therefore, we speculated that the M5AЈ isoform of FOS in C. elegans is produced in the Golgi compartment as a consequence of the action of Golgi ␣-mannosidase I, a class 1 ␣-mannosidase. To confirm this assumption, we employed specific inhibitors for Golgi ␣-mannosidase I to inhibit the enzyme activity, because there are at least two candidates of Golgi ␣-mannosidase I, C52E4.5 and D2030.1, in the C. elegans genome (Fig. 3). Wild-type worms were incubated with 1 mM DMJ or 0.1 mM KIF for 4 h in M9 buffer, and FOS-GN1 species were extracted and analyzed. In the case of no treatment, M5Ј was the most abundant of all the FOSs, and relatively low quantities of M6Ј and M7Ј were recorded (Fig. 4, upper chart). In contrast, in the case of treatment with inhibitors, an increase in the amounts of M6Ј and M7Ј and a concomitant decrease in the amount of M5Ј was observed. The amounts of M8Ј and M9Ј were not substantially altered (Fig. 4, middle and bottom  charts). These results suggest that a class 1 ␣-mannosidase (possibly Golgi ␣-mannosidase I) is involved in the processing of the oligosaccharide of M7Ј into those of M6Ј or M5Ј. On the other hand, the production of M7Ј from M8Ј and M9Ј may not be affected by DMJ and KIF, indicating that these oligosaccha-ride degradation processes are mediated by another ␣-mannosidase insensitive to these inhibitors. Therefore, these facts suggest that misfolded glycoproteins to be degraded and/or OSTreleased FOSs are transported to the Golgi apparatus, and there subjected to the action of Golgi ␣-mannosidase I. No cytosolic ␣-mannosidase is involved in the degradation of FOS in C. elegans, which is quite different from mammals.
Analysis of FOS in Golgi ␣-Mannosidase II-deficient Worms-Next, to further confirm that FOSs are generated through the ER-Golgi trafficking, we analyzed FOS from Golgi ␣-mannosidase II-deficient worms. After the processing by Golgi ␣-mannosidase I, M5A isoform N-glycans on glycoproteins should be modified by addition of a single GlcNAc onto an ␣1,3branched Man residue by ␤-N-acetylglucosaminyltransferase I (GnT-I), to produce GNM5A (Man␣1-3(Man␣1-6)Man␣1-6(GlcNAc␤1-2Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc), followed by removal of two Man residues by Golgi ␣-mannosidase II to form GNM3 (Man␣1-6(GlcNAc␤1-2Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc) (31). This trimming by Golgi ␣-mannosidase II is the essential step for divergence to the complex-type N-glycan. In C. elegans, the F58H1.1 gene product is most similar to mammalian Golgi ␣-mannosidase II (Fig. 3), and an enzyme-deficient mutant strain (tm1078) is available, which has a 304-bp deletion over exons 5 and 6 of the F58H1.1 gene. Indeed, tm1078 showed a drastically altered pattern of N-glycans on the glycoproteins compared with the wild type (32) 3 ; namely, M3, fucosylated and complex-type oligosaccharides were not found in tm1078 and accumulation of high mannosetype oligosaccharides and GNM5A was observed. This result is consistent with a lack of Golgi ␣-mannosidase II activity. Subsequently, we analyzed FOS-GN1 in this mutant and found that the amount of M6Ј was markedly increased compared with that in the wild type (Fig. 5A, upper and second charts). After the treatment of the FOS-GN1 mixture from tm1078 with 1,2-␣mannosidase, the amount of M6Ј position decreased and that of M5Ј increased. Because a comparable amount of FOS remained at the elution position of M6Ј even with elongation of the reaction time or an increase of the amount of enzyme, we supposed that the peak at M6Ј position contained another FOS than M6Ј (Fig. 5A, third chart). Interestingly, when the FOS mixture was treated with the combination of 1,2-␣-mannosidase and ␤-Nacetylhexosaminidase, FOS at the M6Ј elution position completely disappeared, suggesting the presence of an ␣1,2-mannosidase-resistant FOS, which contains ␤-GlcNAc at its non-reducing end (Fig. 5A, bottom chart). Then, the remaining 1,2-␣-mannosidase-resistant fraction at the M6Ј elution position was collected and analyzed by ESI-MS, confirming that this FOS was GNM5AЈ (Fig. 5B). This result shows that Glc-NAc-attached (at the non-reducing end) glycans exist as FOS in Golgi ␣-mannosidase II-deficient mutant worms and also suggests that glycoproteins with N-glycans once processed in the Golgi apparatus might be dislocated into the cytosol and degraded. This is the first report suggesting the occurrence of quality control in the Golgi apparatus from the viewpoint of cytosolic FOS. Chemical shifts (ppm) of structural reporter group of PA-oligosaccharide of peak 1c observed on 600 MHz 1 H NMR spectroscopy referenced to HOD ␦ ‫؍‬ 4.32 a The numbers and letters correspond to the schematic structure depicted above. b The values at a probe temperature of 333 K reported by Priem et al. (28). c The values at 298 K with use of acetone as an internal standard reported by Kimura et al. (29).

DISCUSSION
In this study, we have shown the structures of FOS in C. elegans. The major FOS in the wild type were high mannosetype bearing one GlcNAc residue at their reducing ends (FOS-GN1 species), which composed 94% of the total FOS species. By contrast, the eng-1 mutant accumulated a large amount of FOS-GN2 species. These results directly showed that FOSs are generated primarily in the form of FOS-GN2 species by the action of PNGase or OST, followed by removal of one GlcNAc residue of the chitobiose moiety by a cytosolic ENGase, ENG-1, to form FOS-GN1 species. Besides, ENG-1 was suggested to convert newly released FOS-GN2 into FOS-GN1 very rapidly, because FOS-GN2 species are minor in the wild type. The substrate specificity of recombinant ENG-1 was consistent with this result in terms of acting preferentially on high mannose-type oligosaccharides (23). This is the first evidence for involvement of cytosolic ENGase in the processing of FOS in vivo. In our observation, the total amount of FOS in the eng-1 mutant was reduced as compared with that of the wild type. It was reported that mammalian PNGase was strongly inhibited by FOS-GN2 (33); therefore, the same phenomenon might occur and the rapid processing of FOS-GN2 by ENG-1 may be important for the action of PNGase in C. elegans.
Similarly in other multicellular organisms, M5Ј is the most abundant FOS in C. elegans. However, the isomeric structure was found to be M5AЈ, not M5BЈ, which is quite different from other organisms: the major mammalian M5Ј is M5BЈ (3,5). Although M5AЈ could be found in mouse liver, the levels are very low (30). This fact indicates that C. elegans has a different pathway of FOS catabolism from that reported thus far in other multicellular organisms. However, there are reports that M5AЈ is found as a FOS in appreciable amounts in some plant cells (27,34), implying that FOS catabolism may be similar between plants and lower eukaryotes.
In mammalian cells, M5BЈ might be formed from M8Ј or M9Ј by a cytosolic ␣-mannosidase (10). This neutral/cytosolic ␣-mannosidase was first isolated from rat liver as an ER ␣-mannosidase (35). Recently, mouse and human ␣-mannosidases (Man2C1) homologous to rat liver ER ␣-mannosidase was cloned and characterized as a cytosolic enzyme responsible for FOS degradation (36,37). These cytosolic ␣-mannosidases belong to the same cluster in the phylogenetic tree of ␣-mannosidases, however, we could not find any gene in the C. elegans genome classifiable into the cluster of cytosolic ␣-mannosidases, and we  could not detect ␣-mannosidase activities in the neutral pH range when the cytosolic fraction of C. elegans was assayed. Furthermore, treatment of wild-type worms with swainsonine, an inhibitor for class 2 ␣-mannosidases, including cytosolic ␣-mannosidase, caused a prominent accumulation of M5AЈ as well as a small amount of GNM5AЈ, but not high mannose-type species larger than M5AЈ (data not shown). These results are probably due to the inhibition of lysosomal ␣-mannosidase and Golgi ␣-mannosidase II. Taken together, C. elegans may have no mammalian-type cytosolic ␣-mannosidase, although we cannot completely rule out the existence of "cytosolic" ␣-mannosidase on the basis of our data. Recently, enzymatic characterization of recombinant class 2 ␣-mannosidases in C. elegans showed that one of the enzymes, AMAN-3 (F48C1.1), had a pH optimum of 6.5-6.7 and was Co 2ϩ -dependent, which is similar to mammalian cytosolic ␣-mannosidase (32). It may be necessary to investigate the expression and intracellular localization of this enzyme.
M5AЈ is suggested to be derived from M5A produced by Golgi ␣-mannosidase I, which is involved in the processing of N-glycan on glycoproteins in the cis-Golgi compartment. Golgi ␣-mannosidase I is specific to ␣1,2-linked Man residue and is inhibited by DMJ or KIF. Indeed, treatment of the wild-type worms with these inhibitors resulted in an altered ratio of FOS-GN1; namely, an increase of M7Ј and decrease of M5Ј. This fact suggests that M5AЈ is generated from M7 N-glycan by Golgi ␣-mannosidase I trimming followed by cytosolic PNGase and ENGase (ENG-1), and also that misfolded glycoproteins might be transported to the Golgi and subjected to the degradation pathway in the cytosol. Phylogenetic analysis revealed that the C. elegans genome contains seven genes classified into the glycoside hydrolase family 47 (class 1 ␣-mannosidases). Among them, four genes show similarity to the Golgi or ER ␣-mannosidase I (two each), and three genes are homologous with the EDEM proteins. It will be interesting to determine whether all Golgi ␣-mannosidase I enzymes are involved in the trimming of glycoproteins to be discarded, or whether a particular enzyme specifically acts on misfolded glycoproteins.
Although the present data suggest that FOSs are generated through Golgi processing, there is the possibility that some quantity of Golgi ␣-mannosidase I may move and remain in the ER, or that some other class 1 ␣-mannosidase may exist in the ER and produce M5A that is degraded into an M5AЈ isoform. It should be considered that EDEM homologs may process misfolded glycoproteins, because recent reports showed that  mammalian EDEM3 and EDEM1 had 1,2-␣-mannosidase activity in vivo (38,39).
To further confirm the Golgi processing of FOS, we used a Golgi ␣-mannosidase II-deficient tm1078, and successfully found that GNM5AЈ in FOS-GN1 species of the mutant. This isoform is derived from GNM5A N-glycan, which is the substrate for Golgi ␣-mannosidase II. Its ␤1,2-linked GlcNAc residue at the Man␣1-3 branch of the trimannosyl core is transferred by GnT-I, which is essential for synthesis of hybrid-and complex-type N-glycans and is localized in the medial-Golgi in mammals. In C. elegans, three GnT-I genes, gly-12, gly-13, and gly-14, were characterized, and immunolocalization experiments suggested that all three GnT-I proteins are localized in the Golgi complex (40,41). Our finding of GNM5AЈ in the tm1078 mutant supports the idea that even glycoproteins transported to the Golgi may be degraded in the cytosol. However, we could not detect GNM5AЈ in the wild-type worms, suggesting that the GnT-I step is critical for glycoprotein quality control in the physiological condition.
The importance of ER-Golgi trafficking in quality control was shown in yeast and mammals (42)(43)(44)(45). Most of the studies on ERAD used exogenous model substrates containing one or more mutations and were sometimes performed under nonphysiological conditions. Several reports also showed that N-glycans of folding-defective glycoproteins in mammalian cells were processed to Golgi forms of M5A or M6, prior to ERAD (46). However, very low levels of possible Golgi-derived FOS (M5AЈ, M6Ј, and M7Ј) could be detected in mammalian cells (ϳ1% of total FOS from mouse liver) (30). In this report, we observed FOS from native glycoproteins in C. elegans and found that a major portion of the FOS (ϳ60%) is considered to be Golgi-derived. These data show a unique metabolism of FOS during ERAD of glycoproteins in the nematode C. elegans.