Altered N-Glycan Expression Profile in Epithelial-to-Mesenchymal

Apr 14, 2014 - ABSTRACT: Epithelial-to-mesenchymal transition (EMT) is an essential biological process that occurs in embryonic development, metastati...
1 downloads 0 Views 6MB Size
Article pubs.acs.org/jpr

Altered N‑Glycan Expression Profile in Epithelial-to-Mesenchymal Transition of NMuMG Cells Revealed by an Integrated Strategy Using Mass Spectrometry and Glycogene and Lectin Microarray Analysis Zengqi Tan,† Wei Lu,† Xiang Li,‡ Ganglong Yang,† Jia Guo,† Hanjie Yu,§ Zheng Li,§ and Feng Guan*,† †

The Key Laboratory of Carbohydrate Chemistry & Biotechnology, Ministry of Education; School of Biotechnology and ‡Wuxi Medical School, Jiangnan University, 1800 Lihu Avenue, Wuxi, Jiangsu 214122, China § Laboratory for Functional Glycomics, College of Life Sciences, Northwest University, 229 Taibai Beilu, Xi’an 710069, China S Supporting Information *

ABSTRACT: Epithelial-to-mesenchymal transition (EMT) is an essential biological process that occurs in embryonic development, metastatic diseases, and cancer progression. Altered expression of glycans is known to be associated with cancer progression. No studies to date have presented global analysis of the precise variation of N-glycans in EMT. We describe here the profile of Nglycans and glycogene expression in the EMT process induced by transforming growth factor-β1 (TGFβ1) in a normal mouse mammary gland epithelial (NMuMG) cell model. An integrated strategy with a combination of mass spectrometry, glycogene microarray analysis, and lectin microarray analysis was applied, and results were confirmed by lectin histochemistry and quantitative real-time PCR. In TGFβ-induced EMT, levels of high-mannose-type N-glycans were enhanced, antennary N-glycans, and fucosylation were suppressed, and bisecting GlcNAc N-glycans were greatly suppressed. The expression of seven N-glycan-related genes was significantly changed. The products of glycogenes ALG9, MGAT3, and MGAT4B appeared to contribute to the observed alteration of N-glycans. The findings indicate that dysregulation of N-glycan synthesis plays a role in the EMT process. Systematic glycomic analysis based on the combination of techniques described here is expected to facilitate the discovery of the aberrant N-glycosylation in tumor progression and provide essential information in systems glycobiology. KEYWORDS: N-glycan, glycogene, mass spectrometry, gene microarray, lectin microarray, high throughput



strated in various cancer cell models.7−10 Altered expression of other glycoconjugates is involved in certain oncogenic processes; for example, ganglioside GD2 was identified as a novel biomarker for breast cancer stem cells (CSCs),11,12 and gangliotetraosylceramide (Gg4) and its gene β3GalT4 were shown to be down-regulated in epithelial-to-mesenchymal transition (EMT).13,14 EMT, a process whereby differentiated epithelial cells are transformed to spindle-shaped mesenchymal cells, is characterized by the loss of epithelial cell marker molecules (e.g., Ecadherin) and acquisition of mesenchymal markers (e.g., fibronectin, vimentin). EMT plays an essential role in embryonic development and in adults is involved in wound healing, fibrosis, metastatic diseases, and cancer progression.15 Cells produced by EMT in oncogenesis acquire the migratory phenotype of malignant cells and related characteristics such as the ability to evade apoptosis and anoikis.16 The alteration of glycans during the EMT process has been characterized using high-throughput technologies. The expression of glycans in hepatocyte growth factor (HGF)-induced

INTRODUCTION Glycans (assemblies of monosaccharides that form complex branching topologies) are an essential class of biomolecules in living organisms, along with proteins, nucleic acids, and lipids. Glycans are often attached to proteins and lipids to form glycoproteins, glycolipids, glycosaminoglycans, or other glycoconjugates. Such glycan-based molecules play key roles in a variety of biological processes, including cell adhesion, molecular trafficking and clearance, receptor activation, signal transduction, and endocytosis.1 Alterations of protein glycosylation have been implicated in development of many diseases or identified as biomarkers for certain tumors.2,3 In particular, N-glycans are well-known biomarkers in cancer progression, and the analysis of N-glycans has been facilitated by techniques based on specific release by PNGase F. Recent studies have documented increased fucosylated N-glycans and decreased sialylated N-glycans in sera of late-stage breast cancer patients,4 increased trisialylated triantennary glycan containing α-1,3linked fucose in total sera of advanced breast cancer patients,5 and increased sialyl Lewis x (SLex) in sera of ovarian cancer patients.6 Aberrant expression of glycosyltransferases (enzymes responsible for glycan synthesis), for example, GnT-III, GnT-V, FUT8, ST8SiA2, ST8SiA4, and GALNT3, has been demon© 2014 American Chemical Society

Received: December 4, 2013 Published: April 14, 2014 2783

dx.doi.org/10.1021/pr401185z | J. Proteome Res. 2014, 13, 2783−2795

Journal of Proteome Research

Article

PER Reagent containing a protease inhibitor (0.1% aprotinin), incubated for 30 min on ice, homogenized, and centrifuged for 15 min at 12 000 rpm. The supernatant was harvested and stored at −80 °C. Protein content was determined by BCA assay (Beyotime Institute of Biotechnology; Haimen, China).

EMT of hepatocellular carcinoma HUH7 cells was analyzed using lectin microarrays,17 and glycogenes in transforming growth factor-β (TGFβ)-induced EMT process of pancreatic cancer cells were determined using microarrays.18 Altered expression of other glycoconjugates and glycosyltransferases has been observed in EMT process.13,14,19−21 However, no study to date has presented global analysis of the expression of N-glycans and their genes in a TGFβ-induced EMT process in combination with high-throughput techniques such as mass spectrometry and lectin and glycogene microarray analysis. We integrated high-throughput techniques as previously described to profile the expression of glycans and their related genes during TGFβ-induced EMT in a normal mouse mammary gland epithelial (NMuMG) cell model. N-glycans linked to glycoproteins were released by PNGase F and analyzed by MALDI-TOF−MS, and mRNA levels of glycanrelated genes were analyzed using glycogene microarrays. Altered N-glycans on glycoproteins were further analyzed using lectin microarrays. This integrated strategy (summarized schematically in Figure 1) provides a global approach for identifying patterns of altered N-glycans and their genes during EMT and will provide essential information in systems glycobiology.

Western Blot Analysis

The total proteins (20 μg) from control and TGFβ-treated samples were separated by 7.5% SDS-PAGE. The gels were transferred onto polyvinylidine difluoride (PVDF) membranes (Bio-Rad; Hercules, CA) with Trans-Blot Turbo Transfer System (Bio-Rad). The membranes were soaked in 5% (w/v) skim milk in TBST (20 mM Tris-HCl, 150 mM NaCl, 0.05% Tween 20, pH 8.0) for 2 h at 37 °C. The membranes were probed with primary antibodies against GnT-III and MAN2A1 (Santa Cruz Biotechnology; Santa Cruz, CA) overnight at 4 °C and incubated with appropriate secondary antibody conjugated with HRP. The bands were visualized with enhanced chemiluminescence detection kit Westar Nova (Cyanagen; Bologna, Italy). RNA Isolation

NMuMG cells (1 × 105 per well in a six-well plate) were cultured and treated as previously described. Total RNA was isolated using an RNApure Tissue Kit (CWbiotech; Beijing, China) according to the manufacturer’s instructions. Glycogene Microarray Analysis

GlycoV4 oligonucleotide microarray analysis was performed at Gene Microarray Core E, Consortium for Functional Glycomics, Scripps Research Institute (La Jolla, CA). Openaccess data available at www.functionalglycomics.org/static/ consortium/resources.shtml were analyzed as previously described.13,22,23 In brief, the raw values were normalized using the robust multichip average (RMA) expression summary. Data were processed using the Biocondutor project and R program software. Fold changes were estimated by fitting a linear model for the genes, and linear modeling was performed with the Limma package in R software for differential expression analysis. Differentially expressed transcripts of control versus TGFβ-treated samples were compared using cut-offs of fold change >1.5, fold change 5 from total glycoproteins were annotated using GlycoWorkbench software

(Figure 2). The control and TGFβ-treated samples showed 27 and 24 distinct m/z N-glycans, respectively. There were 19 Nglycan structures found in both control and TGFβ-treated samples but with different intensities. Eight structures were 2788

dx.doi.org/10.1021/pr401185z | J. Proteome Res. 2014, 13, 2783−2795

Journal of Proteome Research

Article

(41.7%) than in control samples (48.1%). An enhanced degree of fucosylation has been observed in many types of cancer.36,37 Up-regulation of fucosyltransferase-8 (FUT8), which catalyzes α-1,6-fucosylation in mammals, has been reported in malignant liver, ovarian, thyroid, and colorectal cancers.38−40 To clarify the differences between the results of the present versus previous studies, we determined the relative degrees of fucosylation recognized by LCA and LTL lectins and expression of the FUT8-encoding gene Fut8. Sialic acids are usually attached to the end of N-glycans and play essential roles in many biological processes. The expression of sialic acids undergoes significant changes during cancer progression.41,42 Sialic acids are unstable during MALDI-MS analysis, and in the absence of modifications, information on sialic acids on N-glycans is lost. Acetohydrazide is an amidation reagent that modifies both α-2,3-linked and α2,6-linked sialic acids, in contrast with the incomplete modification of α-2,3-linked sialic acids that occurs during conventional permethylation, methylesterification, and amidation.43 To investigate the role of sialo-N-glycans during EMT, we used acetohydrazide to modify the sialic acids on N-glycans. Sialo-N-glycans were annotated using Glycoworkbench software (Figure 4). The MS/MS spectra of sialo-N-glycans are shown in Figure S3 in the Supporting Information. A total of 16 sialo-N-glycans were detected by MALDI-TOF−MS: 8 monosialo-N-glycans, 6 bisialo-N-glycans, and 2 trisialo-Nglycans. The numbers of distinctive sialo-N-glycan in control and TGFβ-treated samples were 10 and 11, respectively, and 5 sialo-N-glycans were found in both control and TGFβ-treated samples (Table S1 in the Supporting Information).

unique to control samples, and 5 structures were unique to TGFβ-treated samples (Table 1). MALDI-TOF−MS analysis does not yield defined glycan compositions; therefore, MALDITOF/TOF−MS/MS was performed to obtain detailed information regarding substitutions and branching patterns of the monosaccharide constituents. Two general types of cleavages were observed by MALDI-TOF/TOF−MS/MS: (i) glycosidic cleavages that generated B-, Y-, C-, and Z-type ions and (ii) cross ring cleavages that generated A- and X-type ions. Among these, B- and Y-type ions were easily cleaved and commonly observed to obtain detailed sequence and branching information because of their low-energy requirement. Composition and linkage information were obtained from A- and Xtype ions. MS/MS spectra of precursor ions with m/z 1743.613, 1850.666, 1905.634, and 2012.719 are shown in Figure 3. Other MS/MS spectra are shown in Figure S2 in the Supporting Information. Structures of the mannose branches were revealed by fragments ions B4Y3β (833.253) and B4Y5β (1157.437) in the m/z 1743.613. The presence of fucose was indicated by 2,5X6βZ1β (1582.089) in the m/z 1850.666 and by B5 (1645.581) in the m/z 2012.719. In the m/z 1850.666, Nacetylglucosamine (GlcNAc) of the pentasaccharide core was four-linked to mannose, as shown by 2,4A5Y4α (1137.445). In the m/z 2012.719, the N-glycans were core fucosylated, as indicated by the fragment ions B 4 (1442.406) and B 5 (1645.581). In conclusion, MS profiling in combination with tandem mass spectrometry revealed a clear alteration of Nglycans during the EMT process with high resolution and sensitivity. Relative variation of the major types of N-glycans in EMT is summarized in Table 2. The proportion of high-mannose-type

Variation of Glycogene Expression in TGFβ-Treated Samples

Table 2. Relative Variation of Different Types of N-Glycans in Control and TGFβ-Treated Samples glycan type hybrid high-mannose complex complex-type glycans biantennary triantennary tetra-antennary penta-antennary fucosylated

control

TGFβ-treated

11.1% 33.3% 66.7%

8.3% 50.0% 50.0%

40.7% 37.0% 22.2% 3.7% 48.1%

33.3% 29.2% 16.7% 0.0% 41.7%

The differences in glycogene expression between control and TGFβ-treated samples were investigated using a GlycoV4 chip that covers 1246 mouse glycogenes.13 Among the differentially expressed genes, 79 glycogenes involved in the metabolism of glycoconjugates (including glycosphingolipids, glycopeptides, glycoproteins, and glycosaminoglycans) were analyzed and visualized as a “heatmap” using Cluster and Tree View software (http://rana.lbl.gov/EisenSoftware.htm). Glycoconjugate-related genes from control and TGFβ-treated samples were clustered side-by-side in the dendrogram (Figure 5A). On the basis of DAVID software analysis (http://david.abcc.ncifcrf. gov/), the 79 glycogenes were classified into 5 groups based on function, as follows: 7 N-glycan-related genes (9% of the total), 6 O-glycan-related genes (8%), 15 glycosphingolipid-related genes (19%), 26 glycosaminoglycan-related genes (33%), and 31% other glycan-related genes (Figure 5B). Among the Nglycan-related genes, 6 (ALG9, MAN2A1, MGAT3, NEU1, NEU2, MANBA) were down-regulated and one (MGAT4B) was up-regulated in the TGFβ-treated samples (Table 3). The differentially expressed glycogenes involved in metabolism of other glycoconjugates are listed in Table S2 in the Supporting Information. To confirm the results of glycogene microarray analysis, we determined expression of the above seven genes and of Fut8 (representing an invariable result in microarray analysis) by qRT-PCR. In TGFβ-treated cells, the expression of five genes (ALG9, MGAT3, MANBA, NEU1, NEU2) was significantly decreased, and the expression of MGAT4B was greatly increased (Figure 5C), consistently with results from microarrays. The expression of MAN2A1 showed a significant

N-glycan structures was greater (45.8%) in TGFβ-treated samples than in control samples (33.3%). The accumulation of these structures indicates a blocked step in N-glycan synthesis, resulting in incomplete glycosylation in EMT. Frank et al. reported elevated levels of the high-mannose-type N-glycan GlcNAc2Man9 in both mouse and human sera in the presence of breast cancer.34 However, complex-type N-glycan structures were suppressed in TGFβ-treated samples (50.0 vs 66.7% in control samples) and displayed the following properties: (i) The bi-, tri-, tetra-, and penta-antennary N-glycan structures (particularly triantennary) were suppressed. Similar suppression of tri- and tetra-antennary N-glycan structures was reported in hepatocellular carcinoma.35 These findings indicate a general suppressed expression of tri- and tetra-antennary N-glycan structures in TGFβ-induced EMT processes. (ii) The proportion of core fucosylated complex-type N-glycan structures was slightly lower in TGFβ-treated samples 2789

dx.doi.org/10.1021/pr401185z | J. Proteome Res. 2014, 13, 2783−2795

Journal of Proteome Research

Article

Figure 4. MALDI-TOF−MS analysis of acetohydrazide derivatization of sialic acid of N-glycans. Total proteins were denaturated by urea, DTT and IAM. One hundred μL of 1 M acetohydrazide, 20 μL of 1 N HCl, and 20 μL of 2 M EDC were added to the ultrafiltration units, and the mixture was incubated at RT for 4 h. After modification of sialo-N-glycans, N-glycans were desalted, and analyzed by MALDI-TOF/TOF-MS/MS. Experiments were performed in biological triplicate, and representative sialo-N-glycan spectra are shown. The data were processed based on MALDI-TOF−MS analysis of neutral N-glycans.

N-Acetylglucosaminyltransferase-III (GnT-III) encoded by the MGAT3 gene catalyzes the addition of GlcNAc by β-1,4 linkage to the core mannose to form bisecting GlcNAc structures in N-glycans.46 The expression of GnT-III was decreased at both mRNA (Figure 5A,C) and protein (Figure 5D) levels. N-Acetylglucosaminyltransferase-IVb (GnT-IVb) encoded by the MGAT4B gene catalyzes the transfer of GlcNAc by branching β-1,4 linkage to the Manα1−3 arm of the N-glycan core to form triantennary structures. The decreased expression of MGAT3 was more striking than the increased expression of MGAT4B in EMT (Table 3), suggesting that MGAT3 plays the dominant role in N-glycosylation during EMT. Thus, the suppression of bisecting antennary N-glycan structures (products of MGAT3) is dominant over the enhancement of branching antennary N-glycan structures (products of MGAT4B). Similarly, the relative proportion of antennary N-glycan structures, as detected by mass spectrometry, was suppressed in EMT, and this suppression was correlated with the trend of bisecting antennary N-glycan structures. The suppression of antennary N-glycan structures

decrease in microarray analysis but not in qRT-PCR. However, the expression of MAN2A1 protein was markedly downregulated during EMT, as evaluated by Western blot (Figure 5D). There was no clear difference in Fut8 expression between control and TGFβ-treated cells (Figure 5C). In the N-glycan biosynthesis process, α-1,2-mannosyltransferase (encoded by ALG9) catalyzes the addition of the seventh and ninth mannose residues (DolPP-GlcNAc2Man6 and DolppGlcNAc2Man8, respectively) to lipid-linked oligosaccharides. A deficiency of this enzyme leads to accumulation of GlcNAc2Man6, GlcNAc2Man8, and Glc3Man9GlcNAc2 structures of N-glycans44 and causes glycosylation-IL disease (CDGIL), a congenital disorder.45 Similarly to results of previous studies, ALG9 expression was reduced in TGFβ-induced EMT, as determined by glycogene microarray analysis. ALG9 deficiency may increase the amounts of GlcNAc2Man6, GlcNAc2Man8, and other related high-mannose-type N-glycan structures in TGFβ-induced EMT, as suggested by our MALDI-TOF−MS spectra (Figure 2). 2790

dx.doi.org/10.1021/pr401185z | J. Proteome Res. 2014, 13, 2783−2795

Journal of Proteome Research

Article

Figure 5. Gene expression during EMT revealed by glycogene microarray analysis. (A) Differentially expressed genes involved in the metabolism of glycosphingolipids, glycoproteins, and glycosaminoglycans in control versus TGFβ-treated samples are shown as a heatmap. Red: genomic activation. Green: inhibition. Black: no clear link. Gray: missing data. (B) Differentially expressed genes determined by DAVID software were classified, and the relative proportions are shown as a pie diagram. N-glycan-related, 9%. O-glycan-related, 8%. Glycosphingolipid-related, 19%. Glycosaminoglycanrelated, 33%. Other glycan-related, 31%. (C) Gene expression for ALG9, MAN2A1, MGAT3, MGAT4B, MANBA, NEU1, NEU2, and Fut8 was analyzed by quantitative real-time PCR, as described in M&M. Experiments were performed in biological triplicate. Relative expression was analyzed using the 2−ΔΔCt method and represented as Log2 Relative expression in relation to control samples. Expression of genes above Log2(3/2) was significantly up-regulated. Expression of genes below Log2(2/3) was significantly down-regulated. (D) Western blot analysis of GnT-III and MAN2A1. Tubulin was used as the loading control.

Table 3. Differential Expression of 7 N-Glycan-Related Genes gene name

GenBank acc. #

fold change

description

ALG9 MAN2A1

NM_133 981 NM_008 549

0.66 0.66

MGAT3

NM_010 795

0.29

MGAT4B MANBA

NM_145 926 NM_027 288

1.54 0.55

NEU1

NM_010 893

0.61

NEU2

NM_015 750

0.36

The encoded enzyme catalyzes the addition of α-1,2-mannose to Man6GlcNAc2-PP-Dol or Man8GlcNAc2−PP-Dol. The encoded enzyme hydrolyzes the terminal 1,3- and 1,6-linked α-D-mannose residues in the mannosyloligosaccharide Man5(GlcNAc)3. The encoded enzyme, GnT-III, transfers GlcNAc from UDP-GlcNAc to a β-1,4-mannose to form a bisecting GlcNAc linkage in N-glycans. The encoded enzyme, GnT-IVb, adds GlcNAc to a branching β-1,4-mannose to form triantennary structures. The encoded enzyme, β-mannosidase, hydrolyzes the terminal, nonreducing β-D-mannose residues in β-Dmannosides. The encoded enzyme, α-neuraminidase1, hydrolyzes α-2,3-, α-2,6- and α-2,8- glycosidic linkages of terminal sialic acid residues in oligosaccharides, glycoproteins, glycolipids, colominic acid, and synthetic substrates. The encoded enzyme, α-neuraminidase 2, performs the same function as α-neuraminidase1.

detected by mass spectrometry corresponded primarily to the bisecting type. These findings indicate that expression of MGAT3 and bisecting N-glycan structures is suppressed in TGFβ-induced EMT, consistently with results of previous studies.19,20,30

MAL-II) and fucosylation (recognized by LTL and LCA) were consequently lower in the TGFβ-treated samples than in control samples. The expression of sialic acids was previously reported to be reduced in sera of breast cancer patients versus healthy subjects.4 We performed complete hierarchical clustering and visualization using Hierarchical Clustering Explorer 3.0 software (http://www.cs.umd.edu/hcil/hce/); the resulting heatmap shows control, and TGFβ-treated samples are clustered side-by-side in the dendrogram (Figure 6B). We further investigated and confirmed the glycan profiles by histochemistry using lectins SJA, MAL-I, DBA, LCA, and PHA-E (Figure 7). The TGFβ-treated samples showed greatly increased fluorescence signal intensities of SJA and MAL-I and decreased signal intensities of DBA and LCA, consistently with the results of lectin microarray analysis (Figure 6). The observed decline in fluorescence intensities of PHA-E, which recognizes bisecting GlcNAc N-glycan structure, corresponds to findings from mass spectrometry, glycogene microarrays, and

Comparison of Glycosylation in TGFβ-Treated versus Control Samples by Lectin Microarray Analysis

Lectin microarrays are commonly used for analysis of fine glycan structures of glycoproteins. Using a microarray containing 37 lectins (Table S3 in the Supporting Information), we observed significant changes (>1.5-fold or