HSC: Hepatic Stellate Cell; HB: Hepatoblastoma; HCC: Hepatocellular Carcinoma; IGF2: Insulin-Like Growth Factor 2; DLK1: Protein Delta Homolog 1; TGF-β1: Transforming Growth Factor-β1; MALAT1: Metastasis-Associated Lung Adenocarcinoma Transcript 1; MIG6: ERBB Receptor Feedback Inhibitor 1; HEPES: 4-(2-Hydroxyethyl)-1-Piperazineëthanesulfonic Acid

Hepatoblastoma (HB) is the most common form of liver tumor in infants and children [1] with the age of 2 months to 3 years old [2]. Compared to Hepatocellular carcinoma (HCC), which majorly affects adults and is caused by the infection of hepatitis virus. Hepatoblastoma is a result of accumulating mutations during the growth of liver cells. Although several genetic conditions are associated with the increased risk of developing hepatoblastoma, such as Beckwith-Wiedemann syndrome [3] and Familial Adenomatous Polyposis [4], the pathological details of hepatoblastoma are yet to be determined. A transcriptomic and genomic analysis demonstrated that the gene expression patterns between HB and HCC are different [5]. Several genes are over expressed in HB compared to in HCC, such as IGF2, Fibronectin, DLK1, TGF-β1, MALAT1 and MIG6.

HepG2 cell line is derived from hepatoblastoma and has been widely used as liver cancer model. For example, Blazquez et al. [6] transfected HepG2 cells with CCAAT; enhancer-binding protein β (CEBP β), hepatocyte nuclear factor 4α (HNF4α), and constitutive androstane receptor (CAR) to create a stable model to study bile acid biosynthesis. Investigations at proteome level using HepG2 cells are becoming popular as the progress of proteomics. Pattanakitsakul et al. [7] analyzed the host responses in HepG2 cells during Dengue virus infection and 17 differentially expressed proteins were identified by 2-D PAGE. Tong et al. [8] identified Hepatitis B virus X protein as an inducer for hyper methylation in HepG2 cells.

Glycosylation has long been recognized as one of the most common post-translational modifications of proteins [9]. Glycosylation not only plays a key role in protein stability, solubility, folding, and activity, but also participates in cellular processes such as protein trafficking, sorting and molecular recognition [10-12]. Aberrant glycosylation is a fundamental characteristic of cancer and other diseases [12-14], and many clinical biomarkers and therapeutic targets are glycoproteins [15,16]. N-linked glycoproteins are prevalent on the extracellular side of the plasma membrane, in secreted proteins, and in proteins retained in body fluids [17]. These membrane glycoproteins are easily accessible to therapeutic drugs, antibodies, and ligands; and the secreted glycoproteins in the body fluids are useful in patient diagnosis or clinical management. Therefore, glycoproteomic analysis of HepG2 and L02 cells will not only provide a deeper understanding of hepatoblastoma but also may lead to potential cancer biomarkers.

Many glycoprotein and glycopeptide enrichment methods have been developed for mass spectrometric analysis. These methods include lectin affinity [18,19], immunoaffinity [20,21], boric acid chemistry [22,23], size exclusion [24,25], hydrophilic interaction [26-28], and hydrazide chemistry [29,30]. Among these, hydrophilic interation and hydrazide chemistry methods are two important methods that can theoretically isolate all types of N-glycopeptides regardless of their attached glycans. Hydrophilic interaction solid-phase extraction method [31], which takes advantage of the higher hydrophilicity of glycopeptides vs. non-glycopeptide, is able to efficiently enrich glycopeptide from peptide mixtures. Hydrazide chemistry isolates glycopeptides from complex samples through the formation of hydrazone bonds between the hydrazide functional groups on solidphase support and aldehydes from oxidized glycans. Compared to the methods mentioned above to extract all N-glycopeptides, selective extraction of the sub-glycoproteome with glycans containing terminal sialic acid can be performed by a modified hydrazide chemistry method [32], which utilizes a reduced concentration of NaIO₄ and lowered reaction temperature at oxidation step. It has been reported that these methods were complementary and using a combination of these method could lead to a more comprehensive glycoproteome profiling than using individual method [33].

In this study, we used a combination of hydrazide chemistry and hydrophilic interaction methods to extract N-glycopeptides from HepG2 cells and L02 cells (Figure 1). The extracted N-glycopeptides were analyzed by LC-MS; MS and N-glycoproteins with the corresponding N-linked glycosylation sites were identified. A series of bioinformatics analysis were performed to annotate the identified glycoproteins. The dataset introduced in this study provides a basis for further functional studies such as analyzing glycan mediated recognition events.

Figure 1: Glycoproteomic approach combining hydrazide chemistry and hydrophilic affinity.

Materials

Gibco® RPMI 1640 medium was purchased from Invitrogen (Carlsbad, CA, USA). Non-essential amino acids were from Hyclone (Losan, UT, USA). Fetal bovine serum was from Zhejiang Tianhang Biological Technology Co. Ltd. (Zhejiang, China). Calf serum was from Minhai Biotechnology Co. Ltd. (Beijing, China). 4-(2-hydroxyethyl)- 1-piperazineëthanesulfonic acid (HEPES) was from Qingdao MDBio Biotech Co. Ltd. (Qingdao, China). Affi-Gel® Hz Hydrazide Gel was from Bio-Rad Laboratories (Hercules, CA, USA). Dithiothreitol (DTT) was from GE Healthcare (Waukesha, WI, USA). Sequencing grade modified Trypsin was from Promega (Madison, WI, USA). PNGase F was from New England Biolabs Inc. (Beverly, MA, USA). Other compounds such as iodoacetamide (IAA) and urea were from Sigma- Aldrich (St Louis, MO USA). Sep-Pak C18 1 cc Vac Cartridge, 50 mg Sorbent per Cartridge was from Waters (Milford, MA, USA).

Cell culture and sample preparation

Cell cultures of Human hepatoblastoma cell line HepG2 and human normal liver cell L02 were grown with RPMI 1640 medium supplemented with 10% fetal bovine serum and antibiotics. Cells were harvested at 90% confluence. To remove the possible contaminating proteins from culture medium, cells were washed with cold 0.1 M phosphate buffered saline (PBS), pH 7.4 for 3 times. Cells were then lysed with a buffer consisting of 8 M urea, 0.1 M sodium bicarbonate, and 0.1% sodium dodecyl sulfate (SDS) on ice for 30 minutes. Lysates were sonicated using an ultrasonic cell disruptor (Xinzhi, NingBo, China) on 100W for 5 minutes (60% duty cycle). The sample was clarified by centrifugation at 10,000 × g for 60 min. The total amount of proteins from supernatant was quantified by BCA method and kept at -80°C for future use.

Protein digestion

0.5 mg/ml total proteins were reduced by 5 mM DTT at 37°C for 60 min and alkylated by 20 mM iodoacetamide at room temperature in the dark for 60 min. After the urea concentration was diluted below 2 M with a 0.1M NH4HCO3 solution, 10 μg sequencing grade modified trypsin (Promega, USA) was introduced and the glycoproteins were incubated at 37°C overnight. The resulting peptides were purified by Sep-Pak^® Vac C18 cartridge according to manufacturer’s instruction. Peptides were dried with a Freeze Dryer (Christ Alpha 1-4 LSC, Germany) and stored at -20°C for future use.

Solid phase extraction of total N-glycopeptide (SPEG)

Total N-glycopeptides of plasma proteins were enriched according to Tian et al. [30]. Briefly, the cis-diols on glycans of tryptic glycopeptides were oxidized with 10 mM NaIO₄ in 0.1% trifluoroacetic acid (TFA) at room temperature in dark for 60 minutes. Then, excess amount of NaIO4 were removed by Sep-Pak^® Vac C18 cartridge and the oxidized glycopeptides were mixed with 25 μl hydrazide resin. After an overnight reaction, the oxidized glycopeptides were covalently attached to the hydrazide resin. N-glycopeptides were released by a glycan cleaving reaction using 1000 U of Peptide-N-Glycosidase F in 25mM NH₄HCO₃ at pH 7.8 overnight following extensively washing hydrazide resinglycopeptides complexes to remove non-glycopeptides. The formerly glycopeptides were then purified, dried and stored at -20°C for LC-MS; MS analysis

Enrichment of N-glycopeptide by hydrophilic affinity solid phase extraction method

Hydrophilic affinity solid phase extraction method took advantage of the difference in hydrophilicity between normal peptides and glycopeptides to enrich glycopeptides. 20 μl of Sepharose CL-4B resin was cleaned using Washing Buffer (n-Butanol: ethanol: water = 5: 1:1 (v;v;v)). 0.5 mg tryptic glycopeptides were dissolved in 500 μl Loading Buffer (n-Butanol: ethanol: water = 5: 1:1 (v;v;v) with 1mM MnCl2) and mixed with Sepharose CL-4B resin. The mixture was shocked at room temperature for 45 minutes. Glycopeptides were released by Elusion buffer (ethanol: water = 1:1 (v;v)) following washing the resin twice with Washing buffer to remove non-glycopeptides. Finally, the enriched glycopeptides were subjected to PNGase F digestion and kept at -20°C for LC-MS; MS analysis.

LC-MS; MS analysis

The previously isolated glycopeptides were separated on EasynLC II HPLC system (Thermofisher, USA) with a C18 EASY-Column (10cm, ID 75 μm, 3 μm, C18) and identified using LTQ Orbitrap XL mass spectrometer (Thermofisher, USA). The nano-LC separation was performed at a flow rate of 300 nl/min. The eluents used for the LC were (A) 99.9% H2O; 0.1% FA and (B) 99.9% ACN; 0.1% FA. A gradient was from 3% B to 9% B for 10 min, from 10% B to 41% B for 70 min and from 41% B to 81% B for 15 min, followed by holding at 81% B for 25 min. Collision induced dissociation (CID) was used to generate daughter ions from 10 parent ions of each MS spectrum, and dynamic exclusion was enabled. Spray voltage was set at 1.95 kV and the capillary temperature was 200°C.

Database searching and identification of glycosylation sites

MS;MS spectra were searched against human IPI database using Sequest algorithm of Protein Discoverer (version 1.2., Thermo Fisher, USA) with the following parameters. Precursor Mass Tolerance was set as 10 ppm and Fragment Mass Tolerance was 50 mmu. Trypsin was used as protease and two missed cleavage sites were allowed. Carbamido methylation (+57.021 Da (C)) was set as fixed modification. Oxidation (+15.995 Da (M)) and Deamidation (+0.984 Da (N, Q)) were set as dynamic modifications. A confidence level of 99% was used to generate the identification list. The N-glycosylation sites were confirmed by a deamidation event at asparagine residue in the typical glycosylation motif, which is N-X-T; S; C-X (X can be any amino acid residue except proline). To further analyze the frequency of amino acids at #2 and #4 positions of N-glycosylaitonsequon, a motif analysis was performed using WebLogo [34]. The frequency of amino acid showing at each position was analyzed manually

Label free relative quantification of glycopeptides

Expression level of identified glycoproteins between HepG2 and L02 cells were estimated using spectral counting methods as described by Tian et al. [35]. The quantification was performed at protein expression level. Total numbers of spectra for each identified glycoprotein was summed. Glycoproteins with spectral counts of at least 2 were used for quantification. Ratios of total numbers of spectra of same glycoprotein from HepG2 vs. L02 were reported as the relative expression level. For the glycoproteins that were only identified in HepG2 cells, the ratios were arbitrarily marked as 100.

Gene ontology analysis of identified glycoproteins

Gene Ontology (GO) analysis was performed according to the standard procedure of Blast2GO [36] to gain insight about identified glycoproteins in three aspects, namely cellular distribution, molecular function, and potential biological processes that glycoproteins participated in. The differences in GO terms between of the glycoproteins from HepG2 cell line and L01 healthy liver cells were identified.

Glycopeptides and glycoproteins identified from HepG2 and L02 cells

In this research, glycoproteins were released from cell lysate and digested into tryptic peptide mixtures. The enrichment of glycopeptides was performed using hydrazide chemistry and hydrophilic affinity methods. 394 non-redundant N-glycopeptides with 398 glycosylation sites from 237 N-glycoproteins were identified from HepG2 and L02 cells (Supplementary Table S1). Of these, 256 N-glycopeptides, from 164 N-glycoproteins, were identified from HepG2 cells. 331 N-glycopeptides, from 200 N-glycoproteins, were identified from L02 cells. 191 glycopeptides were confirmed in both cell types while 65 glycoproteins were shown in HepG2 cells only (Figure 2A).

Figure 2: N-glycopeptides and glycoproteins identified from HepG2 and L02 cells. A) Comparison of N-glycopeptides and glycoproteins isolated from HepG2 cell lines and L02 cells; B) comparison of N-glycopeptides and glycoproteins isolated using two different methods. “HA” and “HZ” represent the hydrophilic affinity solid phase extraction method and hydrazide chemistry solid phase extraction method, respectively. C) Motif analysis of identified glycosylation sites.

Hydrazide chemistry and hydrophilic affinity are complementary approaches to enrich glycopeptides. In the current research, 170 glycopeptides were identified by both methods, 75 glycopeptides were identified by hydrophilic affinity method only, and 149 glycopeptides were identified by hydrazide chemistry method only (Figure 2B).

To confirm the known N-glycosylation sites and to identify new N-glycosylation sites, each glycosylation site was mapped to the corresponding protein in Uniprot database [37]. 188 N-glycosylation sites (47.2%) were newly identified in the current research and 210 known glycosylation sites (52.8%) were confirmed.

Dual identity of glycoproteins; glycan binding proteins

Interestingly, 11 N-linked Glycoproteins identified from HepG2 cell line and; or L02 cell line have carbohydrate binding ability according to the annotation from Uniprot database [37]. These glycoproteins include Lysosomal alpha-mannosidase (O00754), Attractin (O75882), Thrombospondin-1 (P07996), Cation-independent mannose- 6-phosphate receptor (P11717), Cation-dependent mannose-6- phosphate receptor (P20645), CD44 antigen (P16070), Neural cell adhesion molecule L1 (CD171 antigen, P32004), Basigin (CD147 ,P35613), Follistatin-related protein 1 (Q12841), Nodal modulator 1 (Q15155), C-type mannose receptor 2 (CD280,Q9UBG0) (Table 1). The dual identity of those glycoprotein; glycobinding proteins may suggest the importance of a glycosylation based regulation network. For example, CD44 antigen, located on the surface of a cell, is the receptor for hyaluronic acid and regulates the cell-cell, cell-matrix interaction, which is crucial for cell migration and tumor growth. C-type mannose receptor 2 has calcium dependent glycol-binding ability and it participates in endocytosis of glycosylated ligand via clathrin [33].

Table 1: Identified N-linked glycoproteins with carbohydrate binding ability.

Certain enzymes of O-glycan biosynthesis are N-glycosylated

Glycosyl transferases are responsible for protein glycosylation. The aberrant glycosylation is due to the altered activity of glycosyl transferases. In this study, four enzymes, glucoside iv xylosyl transferase 1 (Q4G148), UDP glucuronosyl transferase 1 family A1 (P22309), glycotransferase 25 domain containing 1(Q8NBJ5) and procollagen-lysine, 2-oxoglutarate 5-dioxygenase 3 (O60568), in O-glycan biosynthesis were also identified by this research. Glucoside iv xylosyltransferase 1 acts on O-glucosylated Notch epidermal growth factor repeats, thus participating in Notch signaling[38]. The N-glycosylation sites of glucoside iv xylosyltransferase 1 were also identified recently [39]. The ASN347 of UDP glucuronosyl transferase 1 family A1 was confirmed by this research as an N-glycosylation site, which exclusively identified from normal liver cell (L02) but not in cancer cell line (HepG2). UDP glucuronosyl transferase 1 family A1 was known as a major component in toxic compounds metabolism by glucuronidation. Although mutations of UDP glucuronosyl transferase has been associated with Gilbert and Crigler-Najjar type II syndromes [40], the effect of N-glycosylation on enzyme activity has yet to be determined. O-linked glycosylation, such as O-linked N-acetyl glucosamine, has long been recognized as a widespread signaling mechanism in nucleus and cytoplasm [41,42]. It would be interesting to postulate the existence of a cross-talking mechanism between N-glycosylation and O-glycosylation through modulating the activity and; or substrate specificity of glycotransferases.

Motif analysis for N-glycosylation sites

N-glycosylation is protein sequence dependent. It has been reported that in eukaryotic species, the amino acid at second [43] and fourth [44] positions in Asn-Xxx-Thr; Ser; Cys-Xxx (where Xxx is any amino acid except proline) affect the frequency of glycosylation. The hydroxyl [45] or thiol at the third position is hypothesized to increase the nucleophilicity of the amide group of the asparagine, facilitating the glycosylation. From the motif analysis, neutral amino acids, such as AGILV, are more popular at the second and fourth position than the amino acids with phenol ring, such as WF (Figure 2C).

Relative quantification of identified glycopeptides between HepG2 and L02 cells using spectral counting

The glycopeptide extraction step in this study prevents an absolute quantification of certain protein in each sample due to the following reasons. First, non-glycoproteins are unable to be extracted by solid phase support. This alters the composition of peptide mixture, lowering the protein coverage. Second, formerly glycopeptides from different proteins may have different ionization property. Third, it is common to see miss cleavage during tryptic digestion. Same glycosylation site may exist in different peptides, which have different ionization property in ion source. To overcome these difficulties, label-free quantification has been achieved by non-glycopeptides from glycoproteins [46]. Labelfree relative quantification of glycopeptide between samples is plausible using isolated glycopeptides although an absolute quantification is very difficult. For example, spectral counts of identified glycosylation sites have been used to compare the expression level of sialylated glycoproteins in breast cancer [35].

Of 222 glycoproteins quantified in this study, 36 N-glycoproteins were uniquely identified from HepG2 while 72 N-glycoproteins were uniquely identified from L02 cells. 22 glycoproteins were up regulated (ratio > 2) in HepG2 cells while 32 glycopeptides were down regulated (ratio < 0.5) compared to those in L02 cells. A truncated list and the complete list of quantified glycoproteins were shown in Table 2 and Supplementary Table S2, respectively.

Table 2: Label-free quantification of identified glycoproteins (partial).

Gene ontology of identified glycoproteins

To understand the biological functions of identified glycoprotein, the identified glycoproteins were mapped into 31,798 gene ontology terms. 218 out of 235 N-glycoproteins (92.8%) were annotated. 62 N-glycoproteins have catalytic activity. 124 N-glycoproteins were located on to organelles and 59 were at extracellular region. More than 70% of the identified N-glycoproteins were binding partners. Other enriched molecular function terms include catalytic activity (35.7%), molecular transducer activity (19.7%), and transporter activity (11.0%). In terms of the biological processes, most of the identified N-glycoproteins participated in cellular process (61.0%), metabolic process (49.5%) and biological regulation (45.8%) (Figure 3A).

Figure 3: Glycoproteomic approach combining hydrazide chemistry and hydrophilic affinity. A) Gene Ontology analysis of total identified glycoproteins; B) GO enrichment analysis- molecular function; C) GO enrichment analysis- biological process.

To determine if there was any difference in terms of cellular distribution and biological functions between the identified N-glycoproteins from HepG2 and L02 cells, we performed a Gene Ontology Enrichment analysis, which demonstrated 12 significantly different GO terms (with p < 0.05 by Fisher’s exact test). There was no significant difference of the cellular distribution of N-glycoproteins between HepG2 and L02 cells. But an enhanced binding ability in N-glycoproteins from HepG2 cells, especially carbohydrate binding ability was observed. Also the receptor activity was elevated (Figure 3B). The regulation of biological quality and cellular process terms were also enriched in glycoproteins from HepG2 cells compared to the glycoproteins from L02 cells (Figure 3C).

Liver cells retain the ability of fast proliferation and differentiation during our entire life. That ability also comes with higher risk to accumulate genetic errors, which may lead to uncontrolled proliferation (cancer), in cells compared to other organs. Compared to hepatocellular carcinoma, which majorly caused by viruses, hepatoblastoma provides a perfect model to analyze the cancer pathology and protein interaction networks in cancer cells. Also, the early symptoms of hepatoblastoma in infants may due to unknown developmental errors which need to be addressed. It has been suspected that carcinogenesis happens even before birth [2].

Post-translational modifications are able to significantly modify the function of proteins. Among varies modifications, glycosylation is interesting due to the diversity of the structures of glycan and the drastic difference in molecular properties between peptides and carbohydrates. Carbohydrate turns a biomolecule into an information storage unit and an information processing unit when molecular recognition event happens. In terms of cancer pathology, aberrant glycosylation interrupts the normal information flow within cells or between cells by utilizing unusual glycosylation sites and/or generating abnormal glycan structures. These features may lead to the development of clinical biomarkers and therapeutic targets [15,16]. For example, Golgi membrane protein 1 (Golgi protein 73, GP73) has been proposed as a biomarker for hepatocellular carcinoma with 3 potential glycosylation sites. Glycosylation site N109 was confirmed from HepG2 and L02 cells in this study.

With the rapid progress of targeted glycoproteomics, various enrichment methods for glycoprotein; glycopeptides have been developed. Hydrazide chemistry is the most specific method due to the formation of covalent linkages between glycan and solid phase support during extraction of glycopeptides. Other merits include ease to handle and high reproducibility. It has been widely used in serum glycoproteome research, glycosylation analysis in influenza virus and disease-related glyco-biomarker discovery. Several variants of hydrazide chemistry method exist. For example, extraction of glycoproteins before tryptic digestion or extractions of glycopeptides after tryptic digestion are two commonly used methods. The latter method usually allows for more identification due to reduced steric hindrance effect around the N-glycosylation sites.

Label-free quantification methods are gaining popularity in glycoproteomics, although the best quantification methods nowadays remains the ones using stable isotope labeling, due to their convenience and enough accuracy as needed [47]. Spectral counting was first introduced by Liu et al. [48] and remains the most popular method. The accuracy of quantification is directly correlated with spectral counts. In this study, more than 51% of identified glycopeptides had at least 4 spectra matched to them, which provided a relatively good estimation for expression level. To further improve the peptide sampling during mass spectrometry, a label-free quantification of glycoproteins by non-glycopeptides using spectral count has been recently described by Chen et al. [46], with a good correlation between spectral count and glycoprotein content.

The glycoproteomic analysis of human hepatoblastoma cell lines by glycopeptide capture and mass spectrometry described above leads to almost 400 glycosylation sites identified from the two cell lines and more than 1; 3 of them were newly identified thus not included in UniProt database. Several glycoproteins highlighted by semi-quantitation assay may lead not only to new functional studies of individual proteins but also the dynamic network view of post-translational glycosylation. For example, Inactive tyrosine-protein kinase 7, whose glycopeptides were only identified from HepG2 cells, is a protein involved in Wnt signal pathway. It participates in cell adhesion, migration and proliferation and could be a proteolytic target for cancer cell invasion [49]. Nicastrin is an integral membrane protein of gamma-secretase complex in Notch signalling pathway. It is overexpressed in breast cancer and is able to promote invasion [50]. However, 5 glycopeptides, 35 spectra total, of Nicastrin were mapped only in L02 lysate in this study, suggesting a different role of Nicastrin in breast cancer and in hepatoblastoma.

Epidermal growth factor receptor (EGFR) is up-regulated in HepG2 cells than in L02 cells with a spectral count ratio above 4. This observation is in agree with Hoffmann et al. showing EGF activation enhances the proliferation of resistant cancer cell [51]. Beyond the several examples shown above, Golgi apparatus protein 1 (GLG1) is another interesting target which is up-regulated in ovarian cancer [52].

This work is supported by NFSC (Grant No. 81372365) and Foundation of Shaanxi Educational Committee (Grant No. 12JK0830).