Basal Autophagy Is Required for the Efficient Catabolism of Sialyloligosaccharides*

Background: The role of autophagy in glycan catabolism remains to be clarified. Results: In Atg5−/− cells, defective in autophagosome formation, sialyloligosaccharides accumulate specifically in the cytosol. Conclusion: Basal autophagy is essential for lysosomal catabolism of sialyloligosaccharides. Significance: This result not only underscores the importance of autophagy in glycan catabolism but also suggests that basal autophagy is required for proper function of lysosomes. Macroautophagy is an essential, homeostatic process involving degradation of a cell's own components; it plays a role in catabolizing cellular components, such as protein or lipids, and damaged or excess organelles. Here, we show that in Atg5−/− cells, sialyloligosaccharides specifically accumulated in the cytosol. Accumulation of these glycans was observed under non-starved conditions, suggesting that non-induced, basal autophagy is essential for their catabolism. Interestingly, once accumulated in the cytosol, sialylglycans cannot be efficiently catabolized by resumption of the autophagic process, suggesting that functional autophagy is important for preventing sialyloligosaccharides from accumulating in the cytosol. Moreover, knockdown of sialin, a lysosomal transporter of sialic acids, resulted in a significant reduction of sialyloligosaccharides, implying that autophagy affects the substrate specificity of this transporter. This study thus provides a surprising link between basal autophagy and catabolism of N-linked glycans.

Glycosylation of proteins critically affects many cellular phenomena, including protein folding, subcellular transport of proteins, cell-cell interactions, and development (1,2). The explosive progress of glycobiology over recent decades has unveiled most of the biosynthetic pathways for N-glycans in mammals. In sharp contrast, how these N-glycans are degraded in cells is relatively unclear, except for the degradation processes in the lysosomes (3,4). The discovery of cytoplasmic peptide:N-glycanase in mammalian cells (5,6), however, led to characterization of an alternative degradation pathway for N-glycans, called the "non-lysosomal" catabolic pathway (7). In this novel pathway, endo-␤-N-acetylglucosaminidase (8,9) and ␣-mannosidase (Man2C1) (10,11) were shown to be involved in the degradation of high mannose-type free oligosaccharides (fOSs), 2 some of which are released by peptide:N-glycanase, in the cytosol of mammalian cells. On the other hand, the precise mechanism by which sialylated, complex-type glycans are catabolized in the cytosol of cells and tissues remains unclear (12,13).
Macroautophagy (hereafter referred to as autophagy) is an intracellular bulk degradation process that delivers cytoplasmic content to lysosomes for degradation (14,15). Proteins and other components, such as lipids or organelles, are also recycled by this pathway (16,17). Although fOSs accumulate in the cytosol, involvement of autophagy in their degradation has not been rigorously examined.
Since the 1990s, a series of autophagy-related (ATG) genes required for the autophagic process have been identified (18,19). Atg5 is a component of the Atg12-Atg5-Atg16L1 complex, which plays a pivotal role in the elongation and closure of the isolation membrane to form an autophagosome (15).
In this study, we analyzed the structures and amounts of accumulated fOSs in the cytosol of mouse embryonic fibroblasts (MEFs) that lack a component required for autophagosome formation (Atg5 Ϫ/Ϫ ) (20). Surprisingly, we found that sialyloligosaccharides were accumulated specifically in Atg5 Ϫ/Ϫ cells. Further analyses indicated that the basal autophagy process is required for keeping these sialylglycans from accumulating in the cytosol. Moreover, knockdown of sialin, a lysosomal transporter for sialic acid, in Atg5 Ϫ/Ϫ cells resulted in delayed accumulation of sialyloligosaccharides in the cytosol, suggesting that the function of sialin protein is somehow compromised. Our results therefore suggest that basal autophagy is important for lysosomal function, affecting the functional properties of lysosomal proteins.

EXPERIMENTAL PROCEDURES
Cell Cultures-Cells (MEF WT-A; MEF KO-B (Atg5 Ϫ/Ϫ (20)), m5-7 (Atg5 Tet-off cell line in KO-B background (21)); Atg9a Ϫ/Ϫ (22)) were cultured with Dulbecco's modified Eagle's medium (DMEM; Nacalai Tesque Co., Kyoto, Japan) supplemented with 10% fetal bovine serum (FBS) and antibiotics (100 units/ml penicillin G and 100 ng/ml streptomycin; Nacalai Tesque Co.) in humidified air containing 5% CO 2 at 37°C. Atg9a Ϫ/Ϫ cells were kindly provided by Profs. Shizuo Akira and Tatsuya Saitoh (Osaka University). Where indicated, 100 ng/ml doxycycline (Sigma) was added. In the case of removal of Dox from confluent (non-dividing) cells, m5-7 cells treated with Dox for either 4 or 10 days (to allow accumulation of sialyl-fOSs) were collected, and upon the removal of Dox from the medium, cells were first cultured at 40 -50% confluence (to allow resumption of the autophagic process). Cells were then collected again after 4 h of incubation by trypsinization and were seeded at full confluence and incubated for the next 2 days. Growing (dividing) cells were seeded at 20 -30% confluence upon the removal of Dox and were further incubated for 1 or 2 days. Cells were then extracted as described below, and sialyl-fOSs were quantitated. For starvation experiments, cells were washed with PBS, and medium was replaced with amino acid-free DMEM for 6 h in the same incubator.
Isolation of Free Oligosaccharides from Cultured Cells-Under standard conditions, 5 ϫ 10 7 cells were collected, washed once with PBS, and resuspended with 800 l of homogenization buffer (10 mM Hepes/NaOH buffer (pH 7.4) containing 5 mM dithiothreitol, 250 mM mannitol, 1 mM EDTA, 1ϫ complete protease inhibitor mixture (EDTA-free; Roche Applied Science), 1 mM Pefabloc (Roche Applied Science)), followed by homogenization in a motor-driven Potter-Elvehjem homogenizer. Debris was then cleared by centrifugation at 1,000 ϫ g for 3 min at 4°C, and the supernatant was centrifuged at 9,000 ϫ g for 10 min at 4°C to separate the supernatant and pellet (P9) fraction. Supernatant was further centrifuged at 100,000 ϫ g for 60 min at 4°C to obtain the pellet (microsome fraction, P100) and cytosol fractions (S100). To the S100 fraction was added 1.5 volumes of ethanol, and protein precipitate was removed by centrifugation at 17,000 ϫ g for 20 min at 4°C. To the P9 and P100 fractions were added 100 l of homogenization buffer, and protein was removed by adding 3 volumes of cold ethanol, followed by centrifugation at 17,000 ϫ g for 20 min at 4°C. Supernatants obtained were evaporated to remove excess ethanol, and samples were desalted using a PD-10 column (GE Healthcare) eluted with 5% ethanol according to the manufacturer's protocol. These desalted samples (P9, P100, and S100) were evaporated to dryness, and samples were subjected to labeling with 2-aminopyridine (PA-labeling), followed by removal of excess reagent using a MonoFas column (GL Sciences Inc., Tokyo, Japan). Detailed methods for PA-labeling, as well as removal of excess reagent were as described previously (23), and samples thus prepared were used for subsequent structural analysis.
Preparation of PA-labeled Standard Glycans-Authentic samples of PA-labeled Gn1-type (bearing only a single GlcNAc at the reducing terminus (24)) and high mannose-type glycans were prepared from various sources (25,26). 3 For standard Gn2-type glycans (bearing an N,NЈ-diacetylchitobiose structure at their reducing termini (24)), standard samples (high mannose-or complex-type glycans) were obtained from Takara Bio Inc. (Ohtsu, Japan) or Masuda Chemical Industry Co. (Takamatsu, Japan). When required, appropriate glycosidase digestion was carried out to generate new standard PA sugars. For PA-labeled, Gn1-type biantennary complex-type glycans, we first digested 25 mg of human IgG (Nihon Pharmaceutical Co., Tokyo, Japan) with 250 units of peptide:N-glycanase F (Roche Applied Science) in 1 ml of 50 mM NH 4 HCO 3 at 37°C overnight. Released glycans were collected using a Centricon TM device with an Ultracel YM-50 membrane (EMD Millipore Corp., Billerica, MA), and filtrate was desalted using a PD-10 column according to the manufacturer's protocol. Desalted samples were evaporated to dryness. The Gn2-type biantennary complex-type glycans thus obtained were dissolved in 200 l of distilled water, and samples were further digested with Endo F2/F3 (Sigma; 100 milliunits each) with the buffer provided at 37°C overnight. The reaction was stopped by adding 3 volumes of ethanol, followed by centrifugation at 17,000 ϫ g for 20 min at 4°C. Supernatants were desalted using a PD-10 column and were subjected to PA-labeling. PA-labeled Gn1-type glycans were then separated from Gn2-type glycans using an ODS column, as described previously (25). Structures of these glycans were confirmed by HPLC, MALDI-TOF MS, and glycosidase digestion analyses. When required, appropriate glycosidase digestion was carried out to generate new standard glycans.
For Gn1-type triantennary glycans, the remaining half of the free Gn2-type glycans obtained as described above was treated with 0.1 M NaOH at 50°C for 6.5 h in order to allow conversion of Gn2-type glycans to Gn1-type (28). Samples were desalted using PD-10 and labeled with PA, and Gn1-type glycans were isolated using an ODS column (25). Our experimental conditions gave a 10 -20% Gn2-to-Gn1 conversion efficiency.
HPLC Analysis-PA oligosaccharides were fractionated by various HPLC analyses. Anion exchange chromatography using a TSKgel DEAE-5PW column (7.5 ϫ 75 mm; Tosoh, Tokyo, Japan) was carried out as reported previously (cf. Fig.  1A) (29). For routine quantitation of sialylated glycans, the following conditions were used (cf. Fig. 5): solvent A, 10% acetonitrile, 0.01% triethylamine; solvent B, 10% acetonitrile, 3% acetic acid, 7.4% triethylamine; flow rate, 1.0 ml/min. Elution conditions were as follows: 0 -10 min, 100% A; 10 -20 min, 15% solvent B. Peaks were monitored by fluorescence with ex ϭ 310 nm and em ϭ 380 nm. The amounts of sialylated glycans were calculated based on the peak area observed from 12 to 16 min, which was subtracted from the data obtained using the sialidase-treated control. Each fraction was then quantitated using the peak area of standard PA-glucose hexamer in the PA-glucose oligomer (2 pmol/l; Takara Bio Inc.) as a reference.
Conditions for separation of sialylated oligosaccharides on the amide-80 column were as reported previously (27). Separation and identification of PA-sugars on the amino and ODS columns were performed as described previously (25). Detailed sugar structures were determined from the elution profiles of the ODS column (expressed as glucose units), and deduced structures were further confirmed by MALDI-TOF MS analysis (AXIMA-CFR or AXIMA-QIT; Shimadzu Corp., Kyoto, Japan) and by various glycosidase digestions (see below). With respect to the separation of sugar isomers using the TSKgel Sugar AXI column (4.6 ϫ 150 mm; Tosoh), PA-sugars were separated by isocratic elution with a 0.7 M borate-KOH (pH. 9.0)/acetonitrile ratio of 9:1 (v/v) at 65°C at a flow rate of 0.3 ml/min. High mannose-type glycans were quantitated using size fractionation HPLC (amino column).
Quantitative Gene Expression Analyses by Real-time PCR-Total RNA was isolated from cells using the Qiagen RNEasy kit (Qiagen); 4 g of RNA was reverse-transcribed with random hexamers using the SuperScript III RT kit (Invitrogen) according to the manufacturer's protocol. For control of Neu2 expression, cDNA from skeletal muscle was purchased from Clontech. Real-time PCR was performed using a TaqMan Universal PCR Master Mix system (Applied Biosystems). The final 20-l real-time PCR mixture contained 10 l of TaqMan Mix buffer, 1 l of probes, and 9 l of cDNA (0.2 g). cDNA was amplified by an initial step at 95°C for 10 min, followed by 40 cycles at 95°C for 5 s and 55°C for 20 s using an ABI PRISM 7900HT sequence detection system (Applied Biosystems). Samples were analyzed in duplicate, and the mean number of cycles to the threshold of fluorescence detection was calculated for each sample. GAPDH (glyceraldehyde-3-phosphate dehydrogenase) expression was quantified to normalize the amount of cDNA in each sample. Probes for sialin, Neu2, and GAPDH used for realtime PCR were obtained from Applied Biosystems.

Accumulation of Sialyloligosaccharides in the Cytosol of Atg5
Ϫ/Ϫ MEF Cells-To examine the effect of autophagy on the catabolism of fOSs, we utilized MEF cells derived from Atg5 Ϫ/Ϫ KO mice (20). It was previously reported that autophagy is unnecessary for degradation of high mannose-type fOSs, which are the oligosaccharides mainly observed in the cytosol of mammalian cells (31). This observation was consistent with our results, in which no drastic changes in the structures or amounts of these glycans was noted (Table 1). However, specific peaks for Atg5 Ϫ/Ϫ cells were seen on size fractionation HPLC (Fig. 1A). We pooled the peaks and performed MS analysis and, to our surprise, saw mass numbers corresponding to sialyl-fOSs (Fig. 1B). Accumulation of sialylglycans in mammalian cell cytosol has rarely been observed (12,13).
Our results suggest that sialyl-fOSs accumulate in Atg5 Ϫ/Ϫ cells but not in wild-type MEFs. To unequivocally show that this is due to the lack of functional Atg5 protein, we utilized m5-7 cells (21), Atg5 Ϫ/Ϫ isogenic cells with Atg5 gene reintroduction under a Tet-off promoter. Sialyl-fOSs, peaks of which disappeared after the sialidase treatment (third panel), were significantly reduced in Atg5-expressing m5-7 cells (Fig. 2A). In addition, when Dox was added to the m5-7 cells to impair   SEPTEMBER 13, 2013 • VOLUME 288 • NUMBER 37

JOURNAL OF BIOLOGICAL CHEMISTRY 26901
autophagosome formation (Fig. 2B, top), time-dependent accumulation of sialyl-fOSs was also seen (Fig. 2B, bottom). These results confirmed that sialyl-fOS accumulation is due to a lack of a functional Atg5 gene.
Biochemical fractionation using wild-type cells confirmed that 51% of the sialyl-fOSs were recovered in the P9 fraction (containing lysosomes; Fig. 3), whereas most of the increased sialyl-fOSs in the Atg5 Ϫ/Ϫ cells were recovered from the cytosol fraction (Table 2). These results suggest that most, if not all, of the sialyl-fOSs are accumulated in the cytosol of Atg5 Ϫ/Ϫ cells.
Basal Autophagy Is Required for Efficient Catabolism of Sialyloligosaccharides-The simplest explanation for this phenomenon (i.e. cytosolic accumulation of sialyl-fOSs under defective autophagy) is that they are specifically recruited to lysosomes via the autophagic process for degradation. To validate this hypothesis, m5-7 cells were first treated with Dox to allow accumulation of sialyl-fOSs. Autophagy was then resumed by removal of Dox. If autophagy specifically targeted sialyl-fOSs into lysosomes, one would expect them to be quickly catabolized upon resumption of autophagy. Reactivation of autophagy upon Dox removal was clearly seen, as evidenced by the occurrence of the LC3-II band (Fig. 4A). In contrast to our expectations, however, more than 70% of the sialyl-fOSs remained in the cytosol, even at 1 day after removal of Dox (Fig.  4B), whereas autophagy was reactivated within 4 h after Dox removal (21). Furthermore, it was found that reduction of sialyl-fOSs was observed for dividing cells but not for cells cultured under full confluence (i.e. non-dividing) conditions, even after 2 days of incubation (Fig. 4B). These results strongly indicate that clearance of sialyl-fOSs by basal autophagy, once accumulated in the cytosol, is a very slow process, and most, if not all, of the reduction is due to the dilution effect upon cell division ( Fig. 4B; compare third and fourth columns). We therefore hypothesized that the autophagic process is important for proper lysosomal function (i.e. efficient lysosomal catabolism of sialylglycans) rather than the active retrieval of sialyl-fOSs from the cytosol into lysosomes.
Although we showed that Atg5 Ϫ/Ϫ cells accumulate sialyl-fOSs in an Atg5 gene-dependent manner, it is possible that this accumulation is not related to autophagy, instead representing an as yet unclarified function of Atg5 protein. To validate this possibility, we examined the MEF cells derived from Atg9a Ϫ/Ϫ cells (22). Atg9a protein is involved in a distinct pathway with Atg5 and is important for the early stages of autophagosome formation (15). As shown in Fig. 5A, we also observed significant accumulation of sialyl-fOSs in Atg9a Ϫ/Ϫ cells, clearly indicating that their accumulation in Atg5 Ϫ/Ϫ cells is not due to their specific function but rather is due to a general defect in the autophagy process.
Starvation-induced Autophagy Is Involved in Non-selective Catabolism of Free Oligosaccharides-Specific accumulation of sialyl-fOSs was found to occur under non-starvation conditions. To examine the effects of starvation-induced autophagy on sialyl-fOS catabolism, m5-7 cells were first cultured with Dox to shut off Atg5 expression, and autophagy was then induced by removal of Dox and concurrent incubation with starvation medium (amino acid-free DMEM) for 6 h. Sialyl-fOSs were reduced to ϳ60% upon starvation induction; inter-   estingly, cytosolic high mannose-type glycans were also reduced to a similar extent (Fig. 5B). As controls, starvation treatment alone or Dox removal in rich medium did not result in significant differences in the total amount of sialyl-fOSs (data not shown), indicating that the observed effects are due to the induction of autophagy (i.e. removal of Dox) under starvation conditions. These results collectively suggest that induced autophagy is involved in the non-selective catabolism of cytoplasmic fOSs. This effect is clearly distinct from non-starvation conditions (i.e. specific accumulation of sialyl-fOSs).
Sialin May Be Involved in Efficient Catabolism of Sialyloligosaccharides under Normal Autophagy-After lysosomal degradation of glycoconjugates, monomeric sialic acids have been shown to be exported out of lysosomes by a sialic acid transporter, sialin (32). This transporter is able to transport sialic acids as well as acidic sugars, such as iduronic acid or glucuronic acid (3), which suggests that substrate specificity for this transporter is rather broad. Nevertheless, sialyl-fOSs do not appear in the cytosol unless autophagy is compromised ( Fig.  2A), thus suggesting that sialyl-fOSs are not substrates for this transporter under normal conditions. Because the accumulation of fOSs was found to be specific for sialylglycans, we wondered whether sialin is involved in their export from lysosomes into cytosol in Atg5 Ϫ/Ϫ cells. To validate this hypothesis, sialin expression was suppressed using the shRNA technique. We successfully established two sialin knockdown cell lines with distinct shRNA constructs from m5-7 cells (73% suppression for construct 1 and 77% suppression for construct 2). We then examined the effects of sialin knockdown on sialyl-fOS accu-  mulation. To this end, sialin knockdown cells were tested for sialyl-fOS accumulation 4 days after the addition of Dox to shut off autophagy. Sialyl-fOS accumulation was significantly reduced for both sialin knockdown cells (Fig. 6), indicating that sialin is somehow involved in the accumulation of sialyl-fOSs in Atg5 Ϫ/Ϫ cells. This also supports the notion that sialyl-fOSs originate in lysosomes.

DISCUSSION
This study clearly demonstrates a previously unknown relationship between autophagy and glycan catabolism. The effect appears to be specific to sialylglycans originating from N-gly-cans. Surprisingly, clearance of sialyl-fOSs upon resumption of autophagy was very inefficient, which suggests that basal autophagy is important to the proper functioning of lysosomes, thereby preventing sialyl-fOSs from accumulating in the cytosol, but the active involvement of autophagy in their catabolic pathway appears not to be significant.
There are few reports on the occurrence of sialyl-fOSs in the cytosol of mammalian cells (12,13). Most recent studies, however, have shown that similar sialyl-fOSs are accumulated in human pancreatic (33) or prostate cancers (34) but not in normal epithelial cells, thus suggesting that these glycans can serve as useful tumor markers. The mechanisms by which these glycans are accumulated in cancer cells are not understood. It is possible that the autophagic process is involved in the accumulation of sialyl-fOSs in those cancer cells.
Our subcellular fractionation studies indicated that the sialyl-fOSs accumulate in the cytosol. The cytosolic sialidase Neu2 is known to catabolize cytosolic sialyl-fOSs (13). We confirmed by quantitative PCR that Neu2 expression is, if it exists at all, very low in both WT cells and Atg5 Ϫ/Ϫ MEF cells (data not shown). Neu2 enzyme is known to have very low expression in most tissues (35,36), and it is possible that the availability of Neu2 protein may be critical for the efficient catabolism of the cytosolic sialyl-fOSs. In the absence of Neu2, however, our results also show that starvation-induced autophagy can promote catabolism of cytosolic fOSs in a structure-independent fashion, and therefore the active autophagy process can account for, at least to some extent, the catabolism of cytosolic fOSs, including sialylated ones.
Although further studies must be performed in order to unequivocally determine the source of sialyl-fOSs, our hypothesis is that they are most likely derived from lysosomes, for the following reasons: 1) in wild-type MEFs, over 50% of sialyl-fOSs are observed in the P9 fraction (containing lysosomes); 2) sialyl-fOS structures resemble those of lysosome-derived fOSs found in sialidosis patients (37,38); and 3) down-regulation of sialin resulted in a significant reduction in the accumulation of sialyl-fOSs. Because sialin does not appear to mediate transport of FIGURE 7. Proposed mechanisms for the role of autophagy in catabolism of sialyl-fOSs. Left, in normal basal autophagy, the substrate specificity of sialin, a sialic acid transporter, was strictly maintained, possibly due to its modification or binding proteins (indicated by a star), for which the normal basal autophagy process is essential. As a result, sialyl-fOSs did not accumulate in the cytosol and instead properly catabolized in lysosomes. Right, in the absence of functional autophagy, specific modifications or binding proteins were lost, and sialin lost strict substrate specificity; sialyl-fOSs were released into cytosol by the action of sialin. sialyl-fOSs under normal conditions, one could speculate that autophagy is necessary for this protein to retain strict substrate specificity; thus, only monomeric sugars can be exported out of lysosomes when autophagy is functional (Fig. 7). The detailed mechanisms, however, remain to be determined.
Irrespective of the mechanism, our data clearly indicate that autophagy is important for cells to prevent sialyl-fOSs from accumulating in the cytosol. It is also important to note that, in our study, the autophagy process was not induced by starvation, thus suggesting that basal autophagy under non-induced conditions is important for catabolism of sialyl-fOSs. Such basal autophagy has been described previously (39). Reportedly, defective autophagy may lead to compromised lysosomal enzyme function (40), which suggests that basal autophagy is essential for the proper function of lysosomes. Future studies will clarify how basal autophagy can contribute to the catabolism of lysosomal glycans in a structure-specific fashion.