Glycoproteomic analysis of the changes in protein N-glycosylation during neuronal differentiation in human-induced pluripotent stem cells and derived neuronal cells

N-glycosylation of glycoproteins, a major post-translational modification, plays a crucial role in various biological phenomena. In central nervous systems, N-glycosylation is thought to be associated with differentiation and regeneration; however, the state and role of N-glycosylation in neuronal differentiation remain unclear. Here, we conducted sequential LC/MS/MS analyses of tryptic digest, enriched glycopeptides, and deglycosylated peptides of proteins derived from human-induced pluripotent stem cells (iPSCs) and iPSC-derived neuronal cells, which were used as a model of neuronal differentiation. We demonstrate that the production profiles of many glycoproteins and their glycoforms were altered during neuronal differentiation. Particularly, the levels of glycoproteins modified with an N-glycan, consisting of five N-acetylhexosamines, three hexoses, and a fucose (HN5H3F), increased in dopaminergic neuron-rich cells (DAs). The N-glycan was deduced to be a fucosylated and bisected biantennary glycan based on product ion spectra. Interestingly, the HN5H3F-modified proteins were predicted to be functionally involved in neural cell adhesion, axon guidance, and the semaphorin-plexin signaling pathway, and protein modifications were site-selective and DA-selective regardless of protein production levels. Our integrated method for glycoproteome analysis and resultant profiles of glycoproteins and their glycoforms provide valuable information for further understanding the role of N-glycosylation in neuronal differentiation and neural regeneration.


Results
Experimental design for integrated glycoprotein analyses. According to the integrated glycoprotein analyses workflow (Fig. 1), iPSCs were differentiated into NSCs, NPCs, or dopaminergic neuron-rich cells (DA) (n = 5-7 in each stage) using a slightly modified method by Chambers et al. (2009) 38 . The differentiation stage of the cells was confirmed by immunostaining using stage-specific markers. Proteins were extracted from the cells in each stage and digested with trypsin. Each tryptic digest was divided into two tubes, one of which Immunofluorescence to estimate the differentiation stages of iPSC-derived cells. The differentiation stages of the cells were estimated by immunostaining using commonly used stage-specific markers. The iPSCs were confirmed by the presence of R-10G, rBC2LCN, SSEA4 on the cell surface (Fig. 2a). Nestin in the cytoplasm and PAX6/SOX1in the nucleus were stained after culturing for 7 and 15 days (Fig. 2b,c). The presence of SOX2 was observed in the nucleus of cells after 7 days, but it disappeared after 15 days (Fig. 2b,c). These observations suggest that the cells at 7 days and 15 days may be classified as NSC and NPC, respectively. The presence of tyrosine hydroxylase in the cytoplasm, FOXA2 in the nucleus, and neuron-marker Tuj1in the cytoplasm suggested differentiation into neurons containing dopaminergic cells (DA) (Fig. 2d).

Quantitative analysis of proteins in iPSC-derived cells. The proteins extracted from the iPSCs and
iPSC-differentiated cells were precipitated by adding chloroform/methanol and digested with trypsin. Some of the tryptic digests were subjected to label-free relative quantitative analysis with a data-dependent acquisition measurement (first run). A total of 12,780 unique proteins were identified (FDR < 1%) from the iPSCs, NSCs,   (Fig. 3a). Compared to the iPSCs, a total of 1,821 unique proteins increased (relative peak area > 2.0, p-value < 0.05), and 2,069 proteins decreased (relative peak area < 0.5, p-value < 0.05) during neural differentiation (Fig. 3b, Supplementary Table S1). The GO enrichment analysis was conducted to annotate the proteins based on their molecular function to characterize the proteins with increased levels in each stage. The proteins increased in the cells after 7 days and were predicted to be linked with brain development and axon guidance. The proteins that increased in the cells after 15 days were likely to be associated with synapse and dendrite morphogenesis. Based on the proteins that increased in the cells after 33 days, such as neurexins, synaptotagmins, and calciumdependent secretion activators, these cells were strongly predicted to be involved in neuronal function (Fig. 3c, Supplementary Table S1). The alteration of protein profiles suggests that the iPSC-derived cells after 7, 15, and 33 days have the characteristics of NSCs, NPCs, and neurons, respectively.  , and DAs (right) compared to the iPSCs (p-value < 0.05). Left: 1. glutathione derivative biosynthetic process, 2. brain development, 3. semaphorin-plexin signaling pathway involved in axon guidance, 4. glutathione metabolic process, 5. branchiomotor neuron axon guidance, 6. central nervous system development, 7. nitrobenzene metabolic process, 8. axon development, 9. regulation of cardiac muscle contraction by regulation of the release of sequestered calcium ion, 10. nervous system development. Middle: 1. cell adhesion, 2. mesenchyme migration, 3. protein localization to synapse, 4. dendrite morphogenesis, 5. muscle contraction, 6. fatty acid beta-oxidation, 7. integrin-mediated signaling pathway, 8. synapse assembly, 9. aging, 10. positive regulation of establishment of protein localization to plasma membrane. Right: 1. neurotransmitter secretion, 2. nervous system development, 3. synaptic vesicle exocytosis, 4. axon guidance, 5. chemical synaptic transmission, 6. glutamate secretion, 7. regulation of cardiac conduction, 8. learning, 9. ion transmembrane transport, 10. synapse assembly. www.nature.com/scientificreports/ and the peptide sequence of the 7,192 glycoforms, derived from 2,192 glycopeptides and 1,149 glycoproteins, were deduced from the MS/MS data using the commercially available data processing software Byonic. We manually excluded suspicious glycoproteins based on the presence of unfamiliar glycans, such as fucosyl highmannose type glycans, NeuGc-modified glycans, HexNAc 2 Hex 10-12 glycans, and glycans below the trimannosyl core, narrowing the analysis to 6,590 glycoforms. The relative quantification of each glycoform was achieved by comparing the peak area of the iPSCs to that of the other cells (Supplementary Table S1). The number of glycoproteins that increased in the NSCs, NPCs, and DAs compared to those in the iPSCs (relative peak area > 2.0, p-value < 0.05) are shown in Fig. 4a. The comprehensive analysis of the enriched glycopeptides with LC/MS/ MS succeeded in identifying 354 more glycoproteins that failed to be identified by the previous comprehensive analysis of the tryptic digests. The percentage of N-glycan forms that increased in the NSCs, NPCs, and DAs compared to those in the iPSCs (relative peak area > 2.0, p-value < 0.05) is shown in Fig. 4b. It was suggested that the increase in glycoproteins after neuronal differentiation was mostly modified with high-mannose type glycans. Compared to the iPSCs, the NSCs and NPCs had more HexNAc 2 Hex 5 (Man5) and HexNAc 2 Hex 8 (Man8), whereas DA had more Man5 and HexNAc 2 Hex 6 (Man6). Interestingly, the level of an N-glycan, composed of five HexNAcs, three Hexs, and a Fuc (HN 5 H 3 F), significantly increased only in the DAs. The HN 5 H 3 F-modified glycoforms were not detected in the iPSCs and were detected in low levels in the NPCs.

N-glycosylation analysis of membrane-associated proteins in iPSC-derived cells. N-Glyco
The structure of HN 5 H 3 F may be a triantennary glycan or a bisected biantennary glycan. We manually analyzed the product ion spectra of HN 5 H 3 F-modified proteins to determine their structures. In the representative product ion spectrum acquired from the plexin B2 glycopeptide (Fig. 4c), a product ion of m/z 1,790.8477 (charge: 1), which corresponds to a peptide bearing HN 5   www.nature.com/scientificreports/ presence of the diagnostic ion related to bisected glycopeptides was observed in the product ion mass spectra acquired from 23 glycopeptides (Supplementary Figure S1). The diagnostic product ions were not detected in the product ion spectra acquired from 87 glycopeptides, which is likely due to the lower intensity of the precursor ions.

Analysis of HN 5 H 3 F-modified glycoproteins.
We conducted a GO functional annotation to characterize the function of HN 5 H 3 F-modified glycoproteins detected in the DAs. Both the HN 5 H 3 F-modified glycoproteins and unmodified glycoproteins were predicted to be involved in cell adhesion and axon guidance. The HN 5 H 3 F-modified glycoproteins were presumed to be involved in the semaphorin-plexin signaling pathway and negative chemotaxis, but unmodified glycoproteins were not. In contrast, glycoproteins not modified with HN 5 H 3 F in the DAs were suggested to be associated with the ion transmembrane transport and extracellular matrix organization, whereas no association was found with HN 5 H 3 F-modified proteins ( Fig. 5a; Supplementary  Table S1). We performed the Kyoto Encyclopedia of Genes and Genomes (KEGG) 42 pathway enrichment analysis to further explore the signaling pathway in which HN 5 H 3 F-modified glycoproteins were involved. The HN 5 H 3 F-modified proteins were suggested to be significantly associated with cell adhesion (p-value = 8.4E−15) and axon guidance (p-value = 9.8E−10) ( Table 1). The representative glycoproteins associated with cell adhesion included contactin 1, neuronal cell adhesion molecules (NrCAM), L1 cell adhesion molecule (L1CAM), neural cell adhesion molecule 2 (NCAM2), and neogenin 1 (NEO1), which were classified into the immunoglobulin superfamily 43 .
Then, we explored whether the increase in the levels of HN 5 H 3 F-modified glycoforms was dependent on protein expressions or differentiation stages. The peak areas of peptides derived from the above glycoproteins in various differentiation stages were extracted from the LC/MS/MS data of the PNGase F-treated glycopeptides (third run), which is appropriate for quantifying glycoproteins because of the elimination of microheterogeneity. The levels of NrCAM, L1CAM, NCAM2, NEO1, plexins A1, A3, A4, B1, 2, C1, and D1 changed with neuronal differentiation (Fig. 5b). The levels of the cell adhesion proteins and plexins A1, A3, A4, and C1 increased in DAs, whereas those of plexins B1, B2, and D1 decreased in DAs. Using the LC/MS/MS data of N-glycosylated peptides (LC/MS/MS Data 2), we sorted the glycoforms derived from the glycoprotein to elucidate the glycosylation profile of each glycosylation site in each glycoprotein and extracted some glycoforms derived from NrCAM, L1CAM, NCAM2, NEO1, and plexins A1, A3, A4, B1, B2, C1, D1 in the iPSCs, NSCs, NPCs, and DAs ( Fig. 5c; Supplementary Figure S1). In the detected range, the HN5H3F modification was limited to the sites at N1006 in NrCAM; N288, N668, N723, and N977 in L1CAM; N217 and N445 in NCAM2; N461 and N628 in NEO1; and N74 and N1042 in plexin A1; N56, N1008, and N1113 in plexin A3; N1180 and N1248 in plexin B1; N124, N844, and N1084 in plexin B2; and N132 and N820 in plexin C1. Interestingly, HN 5 H 3 F-modified glycoforms of plexins B1 and B2 increased in DAs despite a reduction of protein levels. These findings suggest that the HN 5 H 3 F modification is neuron-specific and independent of protein levels.

Discussion
In this study, first, we conducted the quantitative analysis of proteins in the iPSCs, and iPSC-derived neuronal cells to elucidate the changes in protein production during neuronal differentiation. Label-free quantitative proteomic analysis showed that the level of the proteins associated with brain development and axon guidance was already enhanced after 7 days of differentiation. An increase in the proteins related to synapse and dendrite morphogenesis was found in the cells after 15 days of culture, where axon extension and dendritic growth had not been observed. These findings suggest that the proteins with increased levels may be used as neuronal differentiation markers to define NSCs and NPCs, which cannot be distinguished using current markers 44,45 , such as Nestin, PAX6, SOX1, and SOX2. Many proteins involved in neurotransmitter secretion, nervous system development, and synaptic vesicle exocytosis were identified in the DAs. The presence of gamma-aminobutyric acid (GABA) receptors and glutamine receptors as well as tyrosine hydroxylase suggests the differentiated cells are a mixture of various neurons. The lack of glial fibrillary acidic protein implies the absence of astrocyte in DAs 46 .
For cell-based therapies, the World Health Organization requires information on the identity, purity, and activation state of the critical cell type. The identity of the primary cell populations needs to be described at the phenotypic level (i.e., describe the cell surface expression profile with at least two cell surface markers and the functional level 47 ). Our approach demonstrating the expression profile at various differentiation stages could help to define the cell surface expression profile and identify appropriate cell surface markers. Consequently, it can be possible to identify the primary cells and evaluate their purity and maturation, predict their differentiation, or compare different manufacturing processes.
Many glycoproteins are involved in neuronal differentiation and function; thus, glycoproteomics analysis was performed to explore the changes in N-glycosylation and anticipate its role in neuronal differentiation by using two improved methods. One of the methods was N-glycopeptide enrichment by microcrystalline cellulose, which successfully eliminated the coexisting peptides that caused the ion-suppression of glycopeptides and enabled the acquisition of appropriate mass spectra for glycosylation analysis. The peptide sequences and carbohydrate composition of each glycoform were deduced from the LC/MS/MS data by using the software Byonic followed by manual validation. Some unreliable results, such as fucosylated high-mannose type, NeuGc-modified glycans, and glycans that were shorter than trimannosyl core or longer than Man10 were manually excluded. Furthermore, the glycopeptides bearing N-glycans of interest were manually analyzed to confirm the peptide sequence and carbohydrate composition by examining the product ion spectra.
The other improvement in methodology was the sequential LC/MS/MS analysis for determining more glycoprotein structures, which consisted of quantitative proteomic, glycoproteomic, and deglycosylated proteomic  analyses. In addition to abundant glycoproteins, such as integrins and laminins, the method also elucidated structures of lower levels of proteins related to neural functions such as semaphorin subtypes and plexin subtypes. We demonstrated that glycosylation changed at each glycosylation site and in each glycoprotein during differentiation. The N-glycan structures tended to be simplified with reduced fucosylation and sialylation, and the levels of glycoproteins modified with an N-glycan, consisting of five HexNAc, three Hex, and a Fuc (HN 5 H 3 F), increased in DAs. Shimizu et al. discovered a brain-specific glycan in the mouse brain and reported that it was a fucosylated and bisected biantennary glycan (BA2) by HPLC with fluorescence detection of released N-glycans 48 . The HN 5 H 3 F glycan is considered identical to BA2 based on the presence of a bisected glycan-related fragment in the product ion spectra of the glycoforms. An increase in bisected glycans during neural differentiation has been also reported in mouse iPSCs and embryonic stem cells 49 , suggesting that this phenomenon may be common to mouse and human neural differentiation. Previous research has used lectins 50,51 and released N-glycans 48,49,52 but could not conclusively identify the proteins modified with BA2 and other bisected glycans. In the biosynthesis of N-glycans, beta-1,4 N-acetylglucosaminyltransferase (GnT-III) catalyzes the formation of bisecting GlcNAc. It has been reported that GnT-III is associated with Alzheimer's disease [53][54][55] . Akasaka-Manya et al. demonstrated that the level of GnT-III mRNA had significantly increased in AD brains compared to controls 54 , and Kizuka et al. indicated that GnT-III-deficient AD model mice showed reduced amyloid-β (Aβ) accumulation in the brain by suppressing the function of a key Aβ-generating enzyme, β-site APP-cleaving enzyme-1 (BACE1), and greatly improved AD pathology 55 . They found the altered BACE1 subcellular localization in GnT-III-deficient cells, from early endosomes to lysosomes and speculated that bisecting GlcNAc serves as a trafficking tag for the movement of modified proteins to an endosomal compartment. It is of interest whether BA2 affects the localization of proteins during neuronal differentiation.
In this study, we identified BA2-modified proteins and explored the functions of the proteins using GO analysis and KEGG pathway analysis. The BA2 modified proteins were predicted to be involved in neural cell adhesion, axon guidance, and the semaphorin-plexin signaling pathway. Particularly, proteins in the immunoglobulin superfamily were included in cell adhesion molecules. Here, it is known that axon guidance proteins in the axon's growth cone are associated with neurite outgrowth in neuronal cells [56][57][58] . The plexin proteins are transmembrane receptors for semaphorin ligands. The semaphorin-plexin signaling pathway is thought to be essential in neuronal development such as axon guidance and cell migration 59,60 ; however, roles of individual plexin and semaphorin subtypes in neuronal differentiation are still unclear. The levels of the cell adhesion proteins and plexins A1, A3, A4, and C1 increased in DAs, and BA2 modifications occurred at only limited sites of the proteins in DAs. In contrast, the levels of plexins B1, B2, and D1 increased in NPC and decreased in DAs. Interestingly, BA2-modified glycoforms in plexins B1 and B2 increased in DAs. These results suggest that the BA2 modification is protein-, site-, and differentiation stage-selective regardless of protein production levels, implicating that BA2 may be associated with neuronal differentiation.
Our integrated method for glycoproteome analysis, and resultant glycoprotein and glycoform profiles provided valuable information for understanding the role of N-glycosylation in neuronal differentiation. As only one iPSC line was used in this study, further validation with different genotypes and multiple independent clones will be needed to confirm the phenomenon found in this study and apply it to quality control of cell therapy products. The neural differentiation of iPSCs was based on the method of Chambers et al. 38 . NSCs were generated by culturing subconfluent iPSC 201B7 in a differentiation medium composed of 48.5% DMEM/F12 (FUJIFILM Wako Pure Chemical Inc., Osaka, Japan), 48.5% Neurobasal medium (Thermo Fisher Scientific), 1% N 2 supplement (Thermo Fisher Scientific), 2% B-27 supplement (Thermo Fisher Scientific), 1% nonessential amino acids (NEAA; FUJIFILM Wako Pure Chemical), 2 mM L-Alanyl-L-Glutamine (FUJIFILM Wako Pure Chemical), 100 μM 2-Mercaptoethanol (β-ME; FUJIFILM Wako Pure Chemical), 100 nM LDN193189 (FUJIFILM Wako Pure Chemical), and 10 μM SB431542 (FUJIFILM Wako Pure Chemical) at 37 °C in 5% CO 2 for 7 days. After washing with Dulbecco's phosphate buffered saline without calcium and magnesium (DPBS (-); Nacalai tesque Inc., Kyoto, Japan) twice, the cells were incubated in EDTA (FUJIFILM Wako Pure Chemical)/DPBS (-) at 37 °C in 5% CO 2 for 15 min; the unattached cells were removed. Fresh EDTA/DPBS (-) was added to the wells, and the attached cells were collected by pipetting. The supernatant was removed by centrifuging at 300 g for 5 min. The NSC was maintained in the culture medium consisting of 97% DMEM/F12, 1% N2 supplement, 2% B-27 supplement, 2 μg/L fibroblast growth factor 2 (FGF2; FUJIFILM Wako Pure Chemical), 2 μg/L epidermal growth factor (EGF; Table 1. KEGG pathway analysis of the HN 5 H 3 F-modified proteins in the DAs. *The presence of diagnostic ion for bisecting GlcNAc was confirmed in the product ion spectrum.

Immunostaining. The cells in
Deglycosylation. The dried 20 μg glycopeptides were redissolved in 50 mM Tris-HCl pH 8.0, and 2 units of peptide-N-glycosidase F (PNGaseF; Roche Inc., Basel, Switzerland) were added and incubated at 37 ℃ for 16 h. Desalting was performed with a C-tip SDB (Nikkyo Technos Int., Tokyo, Japan). The deglycosylated peptides were dried with the Speed Vac concentrator (Sakuma). Lastly, the dried glycopeptides were resuspended in 3% acetonitrile/0.1% formic acid.  (Thermo Fisher Scientific) against the UniProt human database (July 2020). Peptide sequencing was performed on the fully trypsin-digested proteins with a maximum of two missed cleavages using a 5-ppm mass tolerance for precursor ions and 0.02 Da of fragment ion tolerance. The carbamidomethylation of Cys was used for static modification. For the deglycosylated peptide analysis, asparagine deamidation was added. Dynamic modifications of proteins included the oxidation of methionine, the acetylation, methionine-loss, and methionineloss + acetylation of N-Terminus. Label-free relative quantitation comparing iPSC proteins was performed using the Proteome Discoverer 2.4. The qualitative and quantitative analysis of enriched glycopeptides was conducted using the Byonic software version 2.10 (Protein Metrics Inc., Cupertino, CA, USA) and Proteome Discoverer 2.4. The databases used were the UniProt human database (July 2017) and the glycan database containing 309 mammalian N-glycans. Glycopeptide sequencing was performed with a configuration involving trypsin digestion with a maximum of two missed cleavages, with a 5-ppm mass tolerance for precursor ions, and 0.01 Da of product ion mass tolerance. Static modifications included the carbamidomethylation of Cys and modification of GlcNAc with Asn. The results were evaluated manually, and peptides lacking N-glycosylation consensus sequence were excluded. Finally, the GO enrichment analysis and KEGG PATHWAY analysis were performed with DAVID 61 ver.6.8.