Biochemical Journal

Research article

Silencing of hHS6ST2 inhibits progression of pancreatic cancer through inhibition of Notch signalling

Kai Song , Qin Li , Yong-Bo Peng , Jie Li , Kan Ding , Li-Juan Chen , Cheng-Hao Shao , Li-Jun Zhang , Ping Li


Many of the ligands involved in developmental processes require HS (heparan sulfate) to modulate signal transduction. hHS6ST2 (human heparan sulfate D-glucosaminyl 6-O-sulfotransferase-2) is a Golgi-resident enzyme that usually acts on GlcA/IdoA(2S)-GlcNAc/NS disaccharide-6-sulfate modifications within the HS sequence. Emerging evidence indicates the importance of 6-O-sulfation in a number of developmental processes. However, any correlation with cancer-related events remains largely unexplored. In the present study, we found that hHS6ST2, but not other variants, was activated in human PC (pancreatic cancer). shRNA (short hairpin RNA)-mediated silencing of endogenous hHS6ST2 expression in the PC cell line PANC-1 inhibited cell invasion and migration. hHS6ST2 knockdown also resulted in markedly reduced tumorigenesis in immunocompromised mice. To specifically explore the molecular alterations resulting from depletion of hHS6ST2-generated 6-O-sulfation, we employed two-dimensional gel electrophoresis technology followed by nano-HPLC–ESI (electrospray ionization)–tandem MS to separate and identify total proteins from PC cells. Our data suggest that hHS6ST2 potentiates Notch signalling in PC cells. We also identified a role for hHS6ST2 in the growth and tumorigenicity of these cells which, at least in part, acts through Notch-mediated EMT (epithelial–mesenchymal transition) and angiogenesis. The results of the present study suggest that hHS6ST2 could be an attractive target for PC therapy.

  • angiogenesis
  • epithelial–mesenchymal transition (EMT)
  • heparan sulfate
  • Notch
  • pancreatic cancer
  • proteomics
  • 6-O-sulfotransferase


HS (heparan sulfate) is a macromolecular polysaccharide that is ubiquitously present in mammalian cells and is found both at the cell surface and in the extracellular matrix [1]. The major function of HS chains is to interact with a large variety of protein ligands, such as growth factors, morphogens and their receptors, thus influencing countless biological and pathological processes [2]. Notably, these functional HS sequences are not directly encoded by genes, but are synthesized by many different Golgi-resident enzymes. The HS chain is initially polymerized to form an N-acetylglucosamine–D-glucuronic acid (GlcNAc-GlcA) copolymer through the synthesis of a tetrasaccharide linker sequence to a target serine residue. Structural diversity is then generated through the action of a variety of HS-modifying enzymes. These include NDSTs (N-deacetylase/N-sulfotransferases), which replace the N-acetyl group of the GlcNAc residues with a sulfate group; C5-epimerase, which converts selected GlcA residues into L-iduronic acid (IdoA); and a number of O-sulfotransferases, which catalyse the transfer of sulfate groups on to different positions (e.g. 3-O- and 6-O-sulfation of glucosamines and 2-O-sulfation of uronic acid). The end-product is a highly heterogeneous complex, since the cumulative action of these biosynthetic enzymes is incomplete. The specificity of the interactions between HS and its various ligands largely depends upon the number and distribution of the sulfate groups that are present within the structurally diverse motifs.

A number of studies have demonstrated a critical role for 6-O-sulfated HS in various developmental processes. By subjecting a sulfated octasaccharide library to an affinity chromatographic assay, Ashikari-Hada et al. [3] showed that 6-O-sulfation of HS is important for the binding activity of FGF (fibroblast growth factor)-10, -4 and -7 [3]. In Drosophila, expression of dHS6ST [Drosophila HS6ST (heparan sulfate D-glucosaminyl 6-O-sulfotransferase)] is required for tracheal development, where it regulates FGF-dependent signalling pathways. RNA-mediated knockdown of dHS6ST leads to embryonic lethality and disruption of the primary branching of the tracheal system [4]. Mutations of the HS6ST gene in Caenorhabditis elegans cause defects in axonal and cellular guidance of the neurons [5]. Two separate phenotypes have been reported in zebrafish resulting from morpholino knockdown of HS6ST genes. Morphants show disturbed somite specification and impaired muscle differentiation (which is associated with higher expression of the Wnt target genes) [6] and abnormalities in the branching of the caudal vein [7]. 6-O-sulfation of HS is also important for retinal axon guidance in Xenopus [8]. HS6ST-null mice show development abnormalities and late embryonic lethality, probably due to an HS-dependent defect in VEGF (vascular endothelial growth factor) signalling [9].

Recently, there has been much interest in the role played by embryonic signalling pathways (such as Notch, Hedgehog and Wnt) in cancer. However, despite the significant developmental roles played by 6-O-sulfation and the various hHS6ST (human HS6ST) isoforms, their biological function in cancer remains to be elucidated. Notch is an ancient cell signalling system that regulates cell fate specification, preserves cell ‘stemness’ and controls cellular differentiation in embryonic and postnatal tissues. Alteration of these functions in adults is associated with various types of cancer, in which Notch may act as an oncogene. Notch ligands (Dll1, Dll3, Dll4 and Jagged1 and 2 in mammals) and receptors (Notch-1–4) are membrane-bound and contain a variable number of EGF (epidermal growth factor)-like repeats within their extracellular domains (NECDs). Upon ligand receptor binding between two neighbouring cells, Notch is transactivated via proteolytic cleavage of the receptor, a process that liberates the NICD (Notch intracellular domain) from the plasma membrane. This then travels to the nucleus and associates with a CSL (CBF-1/Suppressor of Hairless/Lag-1) DNA-binding protein, thus activating cellular transcription [10]. The target genes include basic helix–loop–helix transcription factors for the HES (Hairy and Enhancer of Split) family and the Snail family (Snail and Snail2).

PC (pancreatic cancer) is an extremely aggressive malignant disease for which there is no effective treatment. Many changes within Notch have been observed during the abnormal epithelial differentiation programmes that occur in PC [11]. Recent studies have shown that Notch controls Snail expression via two distinct, but synergistic, mechanisms. Notch directly up-regulated Snail expression (through recruitment of NICD to the Snail promoter) and acts indirectly via lysyl oxidase (which stabilizes Snail protein), providing a new mechanism whereby Notch contributes to tumour cell invasiveness, through regulation of EMT (epithelial–mesenchymal transition) [12].

Notch protein components also play a prominent role during tumour angiogenesis [13]. Dll4 is expressed at high levels in the tumour vasculature. The systemic administration of neutralizing anti-Dll4 antibodies and the systemic, or localized, administration of recombinant forms of Dll4, which have been modified to block Dll4/Notch signalling, inhibit the growth of several different solid tumours in mice. Inhibition of Notch signalling by GSIs (γ-secretase inhibitors) produces similar effects. Analysis of tumour growth provides an intriguing insight into the functional consequences of Dll4/Notch inactivation: excessive tip cells are formed, and the tumour vessels are poorly perfused and non-productive, causing tumour hypoxia and inhibiting tumour growth [14].

We therefore believe that novel strategies aimed at down-regulating Notch signalling may be a useful approach to inhibit the progression of PC. Several studies have identified many different modulators of Notch signalling (e.g. glycosylation [15]). In the present study, we used a proteomic approach to characterize minor differences in protein expression between cells with low or high levels of hHS6ST2 gene expression and found that hHS6ST2 gene silencing down-regulated Notch signalling. The results also showed that hHS6ST2 knockdown disrupted EMT and angiogenesis through various signalling pathways, some of which control tumour cell gene expression and phenotype, and may mediate the tumorigenic action of hHS6ST2-modified HS chains.



Rat monoclonal antibody against E-cadherin, rabbit polyclonal antibody against Snail, mouse monoclonal antibody against vimentin and rabbit polyclonal antibody against Jagged1 were purchased from Abcam. Monoclonal anti-(human/mouse Dll4) antibody, monoclonal anti-[mouse VEGFR2 (VEGF receptor 2)] antibody and recombinant human Notch-1–Fc and Notch-2–Fc chimaeras were obtained from R&D Systems. Rabbit monoclonal antibody against Notch-2 was obtained from Cell Signaling Laboratories. Mouse anti-β-actin antibody, FITC-labelled goat anti-(rabbit IgG) (H+L), HRP (horseradish peroxidase)-labelled goat anti-(rabbit IgG) (H+L) and HRP-labelled goat anti-(mouse IgG) (H+L) were purchased from Beyotime. PE (phycoerythrin)-conjugated anti-(mouse/human Notch-1), PE-conjugated control IgG1 and PE-conjugated streptavidin were from eBioscience. VEGF-A165 was obtained from PeproTech. Gelatin, rhodamine–phalloidin and Hoechst 33258 were purchased from Sigma–Aldrich, and Matrigel™ was obtained from BD Biosciences. All HS disaccharide standards, the heparinase enzymes I, II and III, and chondroitin ABC lyase were purchased from Sigma. The disaccharide ΔUA2S-GlcNCOEt6S (IP), used as an internal standard for quantification, was purchased from Iduron. Pronase was from Roche Applied Science and Benzonase was from Invitrogen. DEAE-Sephacel and PD-10 columns were from GE Healthcare. All other chemicals were purchased from either GE Healthcare or Sigma–Aldrich.

Isolation and characterization of HS

Cells were grown in 10-cm-diameter dishes and harvested after transfection. The medium was collected and the cell layer was lysed in 50 mM Tris/HCl, 1% Triton X-100 and 4 M urea. The extracts were centrifuged at 3000 g for 10 min, and the supernatants were recovered and combined with the medium. After digestion with pronase at 37 °C for 2 h, the supernatants were subjected to the DEAE-Sephacel column and desalting using PD-10 columns to obtain the crude GAG (glycosaminoglycan) fraction. The freeze-dried GAG samples were treated with chondroitinase ABC at 37 °C for 24 h, and then applied to the DEAE column and desalting. After freeze-drying, pure HS was obtained.

Pure HS samples were reconstituted in 20 mM ammonium acetate buffer (pH 7.5) (containing 2 mM calcium acetate), subsequently digested by heparinase enzymes I, II and III (10 m-units for each) and incubated at 37 °C for 24 h. The reactions were considered to be complete when the prolonged reaction time or additional enzymes failed to increase the absorbance of the product mixtures at 232 nm. The products were recovered by centrifugal filtration at 10000 g for 10 min, using a YM-3, 3000 Da molecular-mass cut-off membrane (Millipore), and the HS disaccharides were recovered in the flowthrough, freeze-dried and redissolved in 10 μl of HPLC-grade water for ion-pairing reversed-phase HPLC-MS analysis. The chromatographic equipment included a binary pump, an online degasser, an autoplate-sampler and a thermostatically controlled column compartment (Agilent 1200 HPLC system). A gradient was applied at a flow rate of 0.5 ml/min on an Agilent Zorbax Extend-C18 column (4.6 mm×50 mm, 1.8 μm) at 40 °C. A binary solvent system consisting of (A) 5% acetonitrile and (B) 65% acetonitrile was used for linear gradient elution over 40 min. Both solvents contained 20 mM tributylamine and 2.5 mM ammonium acetate, and were adjusted to pH 7.50 using acetic acid. The eluent was detected by an Agilent 6110 single-quadrupole mass spectrometer with an ESI (electrospray ionization) source in negative mode.

Cell culture

PANC-1 cells and a HUVEC (human umbilical vein endothelial cell) line (HUVEC-CS, CRL-2873™) were obtained from the A.T.C.C. and cultured in DMEM (Dulbecco's modified Eagle's medium) (Invitrogen) supplemented with 10% FBS (fetal bovine serum), 100 units/ml penicillin and 100 μg/ml streptomycin (all from Invitrogen). The cells were incubated at 37 °C in a humidified chamber containing 5% CO2.

Plasmid constructs and transfection

RNA interference was used to down-regulate the expression of hHS6ST2 in cells. The short hairpin oligonucleotides targeted to the sequence TCGTCCTCCAATACGTGTG within the ORF (open reading frame) region and CACTAACCTCCTGGCTGTA within the 3'-UTR (untranslated region) for hHS6ST2 (GenBank® accession number NM_147175) were synthesized and inserted into pSUPER+neo+gfp (Oligoengine). The plasmid was also named pSUPER-916 and the empty plasmid pSUPER was used as a control.

For gene knockdown experiments, cells were transfected with plasmid constructs using either Trans-EZ reagent (SunBio) or cationic liposomes (a gift from Professor Chen Li-Juan, Sichuan University). The optimized proportion of plasmid to Trans-EZ reagent was 1:7 (w/w, for HUVEC-CS) and to cationic liposomes was 1:6 (w/w, for PANC-1 cells). Transfection efficiency was monitored by the fluorescence to achieve over 50%, and cellular toxicity was minimized. Down-regulation of the gene was achieved 72 h after transfection.

F-actin (filamentous actin) and IF (immunofluorescence) staining

Cells were fixed with 4% paraformaldehyde for 10 min, and washed twice. For F-actin staining, cells were stained with rhodamine–phalloidin and counterstained with Hoechst 33258. For IF staining, cells were blocked and then incubated with rabbit polyclonal antibody against Snail (at a dilution of 1:200) at 4 °C overnight. After washing, cells were then incubated with FITC-labelled goat anti-(rabbit IgG) (H+L) at a dilution of 1:500 and counterstained with Hoechst 33258. IF was detected, and images were captured using a Nikon Eclipse Ti-U fluorescence microscope.

Wound healing and Matrigel invasion assay

An in vitro wound healing assay was used to assess cell migration. The transfected PANC-1 cells were trypsinized and seeded in six-well plates at a density of 6×105 per well, and grown in DMEM containing 10% (v/v) FBS overnight to reach 100% confluence. The cells were then gently scraped with a 200-μl pipette tube to produce a wound area. The migration and movement of cells throughout the wound area was observed and examined after 24 h.

The Matrigel invasion assay was determined by using a Boyden chamber (8 μm pore size, Millipore). A total of 1.5×104 PANC-1 cells were loaded to the upper chamber, and the cells invading through the Matrigel were stained by H&E (haematoxylin and eosin) to quantify numbers. Six random images were taken at 20× magnification on a light microscope for each transwell, and the average number of invading cells per image was calculated.

Tube-formation assay

To prepare the tube-formation experiment, a 96-microwell plate prechilled at 4 °C was carefully filled with 50 μl/well of liquid Matrigel (10 mg/ml). The Matrigel was polymerized after 1 h at 37 °C. The transfected HUVEC-CS (2×104 per well) were suspended and plated on to the surface of the polymerized Matrigel. The movement of cells was captured at 6 h after plating.

RT (reverse transcription)–PCR and qPCR (quantitative PCR)

Total RNA was extracted according to the manufacturer's instructions, using RNAiso plus reagent (TaKaRa). First-strand cDNA was synthesized from 1 μg of RNA using AMV (avian myeloblastosis virus) reverse transcriptase XL (TaKaRa). For quantitative PCR analysis, reactions were run in the Bio-Rad iCycler equipment in the presence of SYBR Green dye (Invitrogen). Results from qPCR were calculated as threshold cycles normalized to that of the internal control 18S gene using 2−ΔΔ Ct methods. The primers used are shown in Table 1.

View this table:
Table 1 Sequences of primers used in the present study

Western blotting

Total proteins were extracted for Western blotting analyses. The protein concentration was determined using the BCA (bicinchoninic acid) protein assay kit (Beyotime). The total cellular protein extracts were separated according to molecular mass on an SDS/10% polyacrylamide gel, and transferred on to a PVDF membrane. Membranes were blocked with 5% (w/v) fat-free dried milk powder in TBST (Tris-buffered saline with Tween 20: 50 mM Tris/HCl, pH 7.6, 150 mM NaCl and 0.05% Tween 20) at room temperature for 1 h, and incubated with primary antibody at 4 °C overnight. Protein bands were detected by incubation with HRP-conjugated antibodies, and visualized with enhanced chemiluminescence reagent (GE Healthcare).

Flow cytometry

Transfected HUVEC-CS were treated with trypsin-EDTA and suspended in PBS. The cells were first incubated with PBS containing 1% (w/v) glucose and 5% (v/v) calf serum to block non-specific binding. After washing, the cells were incubated on ice with a PE-labelled control IgG1, and biotinylated Notch-1–Fc. The cells were then incubated with PE-conjugated streptavidin. After washing again, the cells were subjected to flow cytometry on a FACScan (BD Biosciences), and the data were analysed with CellQuest software (BD Biosciences).

Transcriptional activation assay

PC cells were transfected with the indicated plasmids. At 72 h after transfection, cells were treated by introduction with mSnail-promoter-driven reporter system (a gift from Professor Angela Nieto, Universidad Miguel Hernández-CSIC, Alicante, Spain) and 5 μg/ml recombinant human Notch-1–Fc chimaera (R&D Systems). Snail activity was determined by luciferase activity. The values shown are means±S.D. for three independent determinations.

In vivo analysis of tumours

Female BALB/c nude mice (6–8 weeks old) were housed under standard conditions and were cared for humanely. All animals used in the experiments were treated according to the Institutional Animal Care and Use Committee guidelines. Approximately 7×106 PANC-1 tumour cells suspended in 100 μl of serum-free medium were implanted subcutaneously into the right flank of each mouse. The plasmid–cationic liposome complex (State Key Laboratory of Biotherapy and Cancer Center) was prepared according to the manufacturer's instructions. Mice were assigned randomly to one of the following two groups (n=5) when tumours were palpable: (i) mice treated with 5 μg of empty vector pSUPER/30 μg of cationic liposome complexes, (ii) mice treated with 5 μg of pSUPER-916/30 μg of cationic liposome complexes. Animals were treated by intravenous injection every 2 days for 16 days. Tumour volume and weight were observed, and tumour size was determined by measuring the largest and perpendicular diameters every 3 days. Tumour volume was calculated according to the formula V=length×width2×0.5236. After mice were killed, the tumours were excised and fixed in 10% neutral buffered formalin solution and embedded in paraffin for further histological analysis.

Histological analysis

Paraffin-embedded tumour tissue was sliced into 3–5 μm sections for H&E staining and IHC (immunohistochemistry) staining. The IHC staining was performed as follows: sections were deparaffinized in xylene, rehydrated in graded alcohols and washed in PBS. The endogenous peroxidase was inhibited by 3% H2O2 for 10 min, slides were immersed in 10 mM citrate buffer (pH 6.0) and boiled for 10 min to retrieve antigen. Non-specific binding was blocked using 5% (v/v) normal goat serum for 10 min. The slides were incubated with 1:100 dilution of antibodies against E-cadherin, vimentin and Snail at 4 °C overnight. The slides were sequentially incubated with biotinylated secondary antibodies and then streptavidin–peroxidase conjugate. Finally the immunoreactivity was visualized using peroxidase–DAB (3,3′-diaminobenzidine). The angiogenesis activity was detected using frozen sections (Reichert Histostat) fixed in acetone and incubated with monoclonal anti-CD31 antibodies (Bioworld Technology). After washing in PBS, sections were treated with secondary HRP-conjugated antibody. The sections were then counterstained with H&E. The vessel density was determined by counting the number of microvessels per field.

Statistical analysis

Data are presented as means±S.D. for the indicated number of independent experiments. Statistical differences between groups were calculated using Student's two-tailed t-test. Differences were considered statistically significant for P<0.05 (*) or P<0.01 (**).

Proteomic analysis

Protocols for proteomic analysis can be found in the Supplementary Online Data at


hHS6ST2 is up-regulated in human PC

Initial analysis was performed comparing expression differences between tumour and the paired non-cancerous tissue mRNA derived from the same pancreas of PC patients. Semi-quantitative PCR revealed differential expression of three hHS6ST transcripts in human PC samples from the archives of the Department of General Surgery, Changhai Hospital, Shanghai, China (patients' consent and approval were obtained to use these clinical materials for research purposes). The hHS6ST1 and hHS6ST3 transcripts were barely detectable. In contrast, gene expression of a short variant of hHS6ST2 was detected in cancer tissue samples and the mean levels of hHS6ST2 mRNA were greater than in paired non-cancerous pancreatic tissues (n=5) [16]. We were therefore able to narrow down to this particular variant hHS6ST2 that contributed to the highest up-regulation of HS 6-O-sulfation. High levels of the hHS6ST2 transcript were also detected in PANC-1 cells, a human PC cell line widely used for studying cancer biology (Figures 1A and 1B). Our data and a previous report that hHS6ST2 was a tumour marker for uterine cervical and corpus cancer [17] suggested that it was imperative to elucidate the effect of the re-activation of this embryonic gene in cancer. To explore the role of hHS6ST2 in pancreatic tumorigenesis, shRNA (short hairpin RNA) methodology was used to silence the expression of hHS6ST2. Three shRNA constructs, two targeting hHS6ST2 either at a conserved (pSUPER-916) or 3′ region (pSUPER-3856) and a control (pSUPER), were used in the experiments. pSUPER-916 produced a marked reduction in hHS6ST2 mRNA expression in PANC-1 cells compared with the controls, and there was no disruption of endogenous hHS6ST2 expression by the empty vector pSUPER (Figure 1B).

Figure 1 hHS6ST2 gene silencing and enzyme activity

(A) Expression of hHS6ST2 and 18S genes in primary tumour (T) and non-cancerous (N) tissues from different patients were determined by RT–PCR. (B) Analysis of hHS6ST2 expression by RT–PCR indicated that 72 h after transient transfection of shRNA (pSUPER-916), endogenous expression of hHS6ST2 in PANC-1 was effectively suppressed. (C) Total ion chromatograms for mock- (pSUPER) and pSUPER-916-transfected PANC-1 cell groups. Peak 1, ΔUA-GlcNAc; peak 2, ΔUA-GlcNAc6S; peak 3, ΔUA-GlcNS; peak 4, ΔUA-GlcNS6S; peak 5, ΔUA2S-GlcNS; peak 6, ΔUA2S-GlcNS6S. The disaccharide ΔUA2S-GlcNCOEt6S was used as an internal standard (IS) for MS quantification. The peaks marked with asterisks are impurities.

We next needed to determine whether the depletion of hHS6ST2 affected the structural modification of HS. To address this, GAGs were isolated from both hHS6ST2-knockdown PANC-1 cells and vector controls. The purified HS was digested into disaccharides and then separated by HPLC–MS. The peaks were identified by comparing their elution times and molecular masses with those of known HS disaccharide standards. The composition of HS disaccharides in cell samples is shown in Figure 1C and Table 2. The results shown in Table 2 are expressed as relative percentages of each disaccharide and normalized to 100%. The calculated values are the means±S.D. for three independent cell transfection experiments. The levels of one of the three 6-O-sulfated disaccharides, ΔUA-GlcNAc6S (peak 2), were significantly reduced in the pSUPER-916-treated group compared with the mock group. Two other 6-O-sulfated disaccharides, ΔUA-GlcNS6S and ΔUA2S-GlcNS6S (peaks 4 and 6), were also modestly reduced. The total 6-O-sulfation of the pSUPER-916-treated group showed a 35% reduction compared with the vector controls. Taken together, pSUPER-916 was demonstrated to work under the experimental conditions.

View this table:
Table 2 Disaccharide composition (relative percentage) of HS from mock- (pSUPER) and pSUPER-916-transfected PANC-1 cells (n=3)

Morphology changes of hHS6ST2-knockdown cells

To evaluate the role of hHS6ST2 gene in tumorigenesis, we first compared the morphological differences between the mock- and pSUPER-916-treated PANC-1 cells. hHS6ST2-knockdown cells maintained an epithelioid appearance in cell culture, and F-actin staining showed a decrease in the formation of lamellipodia-like structures, suggesting a reduction in cell motility and invasiveness (Figure 2A).

Figure 2 Silencing hHS6ST2 inhibits lamellipodia-like structures formation, cell migration and invasion, and tube formation

(A) PANC-1 cells were transfected with either empty plasmid pSUPER or pSUPER-916 for gene silencing of hHS6ST2. The cells were stained with phalloidin for detection of F-actin and counterstained with Hoechst 33258 to visualize the nuclei. The distribution of actin filaments in lamellipodia-like structures was observed. (B and C) Cell migration and invasion rates of pSUPER and pSUPER-916 groups for PANC-1 were compared via wound healing assay and Matrigel invasion assay. The migration of cells throughout the wound area was quantified by calculating the percentage of the closure of the wound area using ImageJ (NIH). The average number of invaded cells were calculated in six fields at 20× magnification on a light microscope using ImageJ and expressed as the means±S.D., *P<0.01. (D) In the tube-formation assay, HUVEC-CS were transfected with indicated plasmids, and carefully added on the top of polymerized Matrigel. Images from the tube-formation experiments were taken 6 h after plating.

Silencing of hHS6ST2 inhibits PANC-1 cell migration and invasion

Next, the migration of shRNA- and mock-treated PANC-1 cells was examined using a wound healing assay. hHS6ST2 shRNA treatment resulted in an 81.5% reduction in wound closure compared with control cultures (which were considered to be 100%) after 24 h, indicating that cells transfected with hHS6ST2 shRNA had reduced migration over time (Figure 2B and see Supplementary Figure S1A (at The invasiveness of hHS6ST2-knockdown cells was then analysed using a Matrigel invasion assay. When loaded into the upper chambers of the transwells, PANC-1 cells invaded across the matrix in a relatively homogeneous fashion. However, the number of invading hHS6ST2-knockdown cells was decreased by 85.4% (Figure 2C and see Supplementary Figure S1B). Taken together, these results indicate that hHS6ST2 is required for PC cell migration and invasion.

Suppression of hHS6ST2 altered in vitro angiogenesis

Angiogenesis also plays a critical role in the development of cancer. 6-O-sulfation of HS is essential for tubular formation in vitro [18]. To study the possible anti-angiogenic effects of hHS6ST2 gene silencing, a tube-formation assay was used to investigate the effect of hHS6ST2 in HUVEC-CS. As shown in Figure 2(D), HUVEC-CS cultured for 6 h in DMEM aligned to form enclosed lumens. In contrast, HUVEC-CS cultured in the presence of pSUPER-916 were unable to form a consistent network of interconnecting tubules, with many cells aggregating in clumps and generating excessive cell extensions or multinodal branch points. This strongly suggested that silencing of hHS6ST2 markedly altered in vitro angiogenesis by HUVEC-CS. HS6ST2, but no other variant, was required for the branching morphogenesis of the caudal vein in vertebrate angiogenesis. HS6ST2-mediated angiogenesis was partially attributed to VEGF and FGF signalling; the contribution of other signalling pathways and proteins is not well documented [7,18].

Proteomics studies indicate that suppression of hHS6ST2 is sufficient to reverse EMT

To investigate further the mechanisms involved in hHS6ST2 modulation, two-dimensional gel electrophoresis technology followed by nano-HPLC–ESI–MS/MS (tandem MS) was used to characterize the global proteomic changes and signalling pathways mediated by hHS6ST2-generated modification. Supplementary Figure S2(A) (at shows representative two-dimensional gel electrophoresis images for total proteins extracted from pSUPER-916- or pSUPER-transduced PANC-1 cells 72 h after transfection. Detailed alterations in protein expression were observed (indicated by the spots marked with arrows in Supplementary Figure S2B and summarized in Supplementary Table S1 at Some 21 up-regulated proteins and 17 down-regulated proteins were identified in cells in the pSUPER-916 group with relatively low levels of hHS6ST2 (Tables 3 and 4 respectively).

View this table:
Table 3 Up-regulated proteins identified in pSUPER-916-transfected PANC-1 cells using ESI–MS/MS
View this table:
Table 4 Down-regulated proteins identified in pSUPER-916-transfected PANC-1 cells using ESI–MS/MS

The proteins identified in the proteomics study could be direct targets for hHS6ST2 or downstream regulators. Bioinformatics software was employed to predict protein–protein interactions, and proteins showing significant alterations were linked together to form a single network (Figure 3). Our results showing that signalling can be modulated by hHS6ST2 are in agreement with those reported in other EMT studies. For example, a hallmark of cancer EMT is the gain of the mesenchymal marker vimentin, and keratin no longer plays a role in epithelial cell protection, but is involved in cancer cell invasion and metastasis [19]. Consistent with this idea, we found that vimentin and keratin 7, keratin 8 and keratin 18 were abundant in control cells with high levels of hHS6ST2. Aberrant expression of Jagged1 and Notch-1 is associated with poor outcome in breast cancer. The expression of Snail2, a transcriptional repressor identified as a novel Notch target, correlates with increased expression of Jagged1 and with Notch-1-mediated repression of E-cadherin (the important caretaker of the epithelial phenotype), which results in tumour growth and metastasis characterized by EMT [20]. The acquisition of an EMT-like phenotype by gemcitabine-resistant PC cells is reported to be linked to activation of the Notch-2 signalling pathway [21]. Amplification and overexpression of TPM3 (tropomyosin 3) in hepatocellular carcinoma was recently reported by Choi et al. [22], and overexpression of TPM3 activates Snail-mediated EMT [22]. Notably, the down-regulated Notch ligands and TPM3 in cells undergoing MET (mesenchymal–epithelial transition) was consistent with suppression of hHS6ST2 observed in the present study. hnRNP (heterogeneous nuclear ribonucleoprotein) A2/B1 is reported to be overexpressed in lung cancer and in other cancers such as breast, pancreas and liver. Silencing of hnRNP A2/B1 in the non-epithelial lung cancer cell lines A549 and H1703 correlated with an increase in expression of E-cadherin, and down-regulation of the E-cadherin inhibitors Twist1 and Snail [23]. Genetic and gene expression alterations of VDAC1 (voltage-dependent anion channel 1) [24] have been shown to be relevant to the EMT processes. These findings were compatible with our proteomics data showing that the expression of hnRNP A2/B1 and VDAC1 were down-regulated in hHS6ST2-deficient cells and could reverse EMT. Our proteomic profiling data were also consistent with previous studies suggesting that forced expression of annexin A1 in metastatic mouse and human mammary carcinoma cells reverses EMT and abolishes metastasis [25], and that siRNA (short interfering RNA)-mediated knockdown of cofilin1 potentiates TGF (transforming growth factor)-β-induced EMT [26]. Our data found that AnxA1 and cofilin1 were up-regulated in hHS6ST2-deficient cells and correlated with the maintenance of an epithelial phenotype. However, up-regulation of superoxide dismutase [27], heat-shock protein [28], NASP (nuclear autoantigenic sperm protein) [29] and α-enolase [30], upon loss of hHS6ST2, contradicted previous findings that they positively contributed to EMT. This sets the foundation for further exploration of their role as multifunctional regulators of epithelial tumorigenesis.

Figure 3 Schematic diagram of a minimum protein–protein interaction network involved in hHS6ST2-mediated Notch signalling

The total cellular protein extracted from PANC-1 cells 72 h after transfection with pSUPER-916 or pSUPER was loaded on the 3–10 NL strips to perform isoelectric focusing as the first dimension and standard SDS/PAGE as the second dimension. Representative two-dimensional gel electrophoresis images were analysed and the significantly and consistently altered protein spots that were changed more than ±2.0-fold were identified by HPLC–ESI–MS/MS. The black dots signify proteins verified in the experiments. Proteins in the network are interacting with each other via intermediate partners (identified from known protein–protein interaction information).

hHS6ST2 modulation positively controls Notch signalling in EMT

Our studies suggested that EMT could be potentiated by manipulating hHS6ST2. hHS6ST2 could be secreted into the culture medium to modify cell-surface HS or it could generate 6-O-sulfated HS existing in the intracellular membrane-bounded compartments [16]. In this case, EMT-related membranous receptor proteins that were influenced by hHS6ST2-modified HS might possibly be considered as direct targets. A salient finding of the proteomic profiling study was the involvement of membrane-bound Notch.

We now considered whether hHS6ST2 was implicated in Notch signalling regulation. qPCR analysis of PC cells showed that two downstream targets of Notch, Hes-1 and Hey-1, were down-regulated after hHS6ST2 knockdown (Figure 4A).

Figure 4 Notch signalling components within hHS6ST2-modified PANC-1 cells and the Jagged1–Notch-2–Snail axis

(A) Total mRNA extracted from cells transfected with pSUPER and pSUPER-916 was subjected to qPCR for Notch target genes (Hes1 and Hey1) expression. Results are means±S.D. for three independent experiments. (B) Notch signalling was determined by utilizing the mSnailluciferase reporter assay in the presence, or absence, of a recombinant human Notch-2–Fc chimaera, which could specifically abrogate Notch–ligand interaction. Results are means±S.D. for three independent runs. *P<0.05. A.U., arbitrary units. (C) PANC-1 cells were transfected with empty plasmid pSUPER or pSUPER-916, and then the cell lysates were subjected to SDS/PAGE followed by Western blotting with anti-Jagged1, anti-Notch-2 and anti-Snail antibodies; β-actin was used as a loading control. FL, full-length.

As some of the molecular mechanisms that regulate Snail expression have been identified, and Notch has been shown to stabilize and promote the activation of Snail in response to cleavage of the NECD and the release and nuclear import of the NICD, we next chose to study the effects of hHS6ST2 on Notch signalling by monitoring the transcriptional activation of Snail. Notch signalling activity was measured using the mSnailluciferase reporter system. As shown in Figure 4(B), the presence of hHS6ST2 increased the luciferase signal, whereas reduced expression of hHS6ST2 inhibited the transcription activity. We used soluble Notch-2–Fc to disrupt the interaction between Notch-2 ligand and its receptor. The co-operation of gene silencing of hHS6ST2 and pre-incubation with Notch-2–Fc resulted in an additive inhibition of mSnail transactivation.

The Snail transcription factor family plays a key role in the Jagged1/Notch signalling pathway and stimulates cell motility and invasiveness during EMT [20,21]. IF staining and immuno-blotting experiments showed that treatment with pSUPER-916 resulted in nuclear export of Snail (see Supplementary Figure S3 at, and reduced protein levels of both full-length and the extracellular domain of Notch-2, Snail and Jagged1 (Figure 4C).

These results suggest that hHS6ST2 modulation may function as a switch of Notch signalling in EMT processes.

hHS6ST2 modulation positively controls Notch signalling in angiogenesis

Notch-1 and VEGF have opposing roles in regulating EMT in ECs (endothelial cells). VEGF-A also inhibited TGF-β–induced EMT [31]. However, a mechanism explaining how biological systems maintain this delicate balance of Notch–VEGF cross-talk remains complicated.

Immunoblotting experiments indicated that HUVEC-CS cultured in vitro did not express the Notch ligand of Dll4 (results not shown). Dll4 is up-regulated by VEGF stimuli [13]. We therefore asked whether Dll4 expression could be altered by manipulating hHS6ST2. Western blotting showed that hHS6ST2 shRNA treatment abolished VEGF-induced up-regulation of Dll4 protein, compared with controls (Figure 5A). Since decreased levels of hHS6ST2 on some ECs could undermine Notch signalling, we thus speculated that hHS6ST2-modified HS mediated the interaction between Notch and Dll4. Flow cytometry was carried out in an attempt to analyse this interaction after forced expression of Dll4 in HUVEC-CS. Specifically, the results confirmed the ability of hHS6ST2 shRNA to facilitate the binding of soluble Notch-1–Fc, but not soluble Notch-2–Fc (results not shown), to Dll4 on HUVEC-CS (Figure 5B); at the same time, the trans-interaction of Notch-1 and Dll4 expressed on neighbouring cells was principally blocked. Consistent results that VEGFR2 mRNA and protein levels (which were laterally suppressed by endothelial Notch1) were elevated, as determined by Western blotting and qPCR assays, revealed again that ectopic Notch1 signal was inhibited in hHS6ST2-knockdown cells (Figures 5A and 5C) [13]. Of note, RT–PCR revealed that a dose of 20 ng/ml VEGF-A165 present in the culture medium could down-regulate gene expression of hHS6ST2 (Figure 5D). Taken together, we postulated a negative-feedback loop of Notch–VEGF cross-talk orchestrated by hHS6ST2 (Figure 5E).

Figure 5 Notch signalling components within hHS6ST2-altered HUVEC-CS cells and the VEGF–Notch-1 cross-talk orchestrated by hHS6ST2

(A) The expression of the Notch ligand Dll4 and VEGFR2 were determined by Western blotting. (B) The selective binding of Notch-1–Fc to HUVEC-CS expressing human Dll4 was confirmed by flow cytometry. Cells were stained with biotinylated Notch-1–Fc followed by PE-labelled streptavidin (outlined in bold) or biotinylated control IgG1 (non-bold). Inactivation of hHS6ST2 significantly affected the Dll4–Notch-1–Fc interaction and inhibited Dll4–Notch trans-interactions with neighbouring cells. (C) The mRNA level of VEGFR2 was detected by qPCR after hHS6ST2 knockdown in HUVEC-CS stimulated with VEGF-A165. Results are means±S.D. for three independent experiments. *P<0.05. (D) hHS6ST2 gene expression stimulated by VEGF-A165 using RT–PCR. (E) Schematic diagram of the proposed negative hHS6ST2-mediated feedback loop regulating the VEGF–Notch axis.

Silencing of hHS6ST2 inhibits tumour progression in vivo

To examine the effects of hHS6ST2 silencing on the tumorigenicity of PANC-1 cells in vivo and explore the therapeutic potential of hHS6ST2 gene silencing in PC, we compared the growth of tumours in immunocompromised mice (BALB/c) either with or without delivery of hHS6ST2 shRNA, along with cationic liposome. Tumour size was monitored every three days by external palpation. At the end of the experiment, animals were killed and the tumours weighed. As shown in Figure 6, tumours derived from the pSUPER-916 group grew at a markedly lower rate than those derived from the control group (pSUPER) (Figure 6A). The tumours harvested from the pSUPER-916 group also weighed less (Figure 6B).

Figure 6 Effect of hHS6ST2 gene silencing on tumour growth in PANC-1 tumour xenografts

(A) PANC-1 cells (7×106/animal) were implanted subcutaneously into the flank of BALB/c mice. Mice that received an implant of PANC-1 cells were randomly allocated into two groups (n=5) and administered intravenously with pSUPER or pSUPER-916 (5 μg)–cationic liposome (30 μg) complexes. The upper panel shows macroscopic appearance of PANC-1 tumours in the pSUPER and pSUPER-916 groups. Tumour growth curves present the mean±S.D. tumour volume. **P<0.01. (B) Excised tumours were weighed after imaging. Results are mean±S.D. masses of the tumours in the mock- and shRNA-treated groups. *P<0.05.

Next, we sought to investigate the mechanisms under which silencing of hHS6ST2 could inhibit the progression of PC. IHC examination of the PANC-1-derived tumours showed homogeneous staining for vimentin and Snail along with weak E-cadherin staining, indicating that the control PANC-1 xenografts had a more mesenchymal phenotype than the hHS6ST2-knockdown group. In contrast, silencing of hHS6ST2 reverses the EMT-like phenotype in vivo (Figure 7A). Also, IHC staining for the CD31 endothelial marker was performed on the systemic transient shRNA-transfected xenografts. As can be observed in Figure 7(B), the pSUPER-916 transfection resulted in apparent suppression of angiogenesis in tumours, compared with the vector control. The average numbers of microvessels observed at a magnification of 200× in the pSUPER and pSUPER-916 groups were 15.7±1.6 and 8.0±1.7 respectively. This finding strongly suggested that silencing of hHS6ST2 reduced angiogenic response in vivo.

Figure 7 IHC staining analysis of hHS6ST2 gene silencing on EMT and microvessel formation in PANC-1 tumour xenografts

(A) The expression of E-cadherin, vimentin and Snail were compared in the pSUPER and pSUPER-916 groups via IHC staining. Quantification was performed by calculating the percentage of the staining intensities using ImageJ (NIH). Results are means±S.D., *P<0.05. (B) CD31 staining on PC xenografts. Tumour angiogenesis was assessed using IHC staining with anti-CD31 antibody on frozen PC tumour xenograft sections. Microvessel counting was performed at 200× magnification using ImageJ. Results are means±S.D. (*P<0.05).


Identifying the proteins that are regulated by modified-HS-mediated signalling will increase our knowledge regarding the biology of the extracellular matrix and shed new light on the mechanisms underlying EMT and cancer progression. A number of studies have focused on the pivotal role of HS6ST in development. However, few studies have explored the role that it plays in cancer and/or have used proteomics to characterize the proteins and signalling pathways mediated by 6-O-sulfation of HS sugar sequences. Our results strongly suggested that Notch signalling through hHS6ST2 is the pathological link between EMT and progression of PC.

Previous studies suggested that Jagged1-mediated activation of Notch promotes carcinogenesis by facilitating cancer cell metastasis through the initiation of EMT [20]. Importantly, we have demonstrated that, in human PC, abrogation of Jagged1 and Notch expression correlates positively with the transient inhibition of hHS6ST2, thereby rescuing EMT and suppressing tumour invasion. We also found a positive correlation between inhibited Notch signalling and Snail down-regulation. The Snail/Snail2 gene has been reported previously to be a direct target for Notch [12,20,32]. Interestingly, decreased levels of Snail, but not Snail2, was observed after transient silencing of hHS6ST2 (results not shown). This was in agreement with previous studies showing that the transient expression of Snail is involved in inducing the invasion process, whereas Snail2 is involved in the maintenance of the migratory invasive phenotype. Snail2 is also associated with tumours in patients with metastatic or recurrent disease and thus probably reflects the diverse and partly redundant mechanisms involved in EMT modulation. Hence our findings indicate that Snail, but not Snail2, is a downstream target gene of hHS6ST2-mediated Notch signalling in pancreatic tumours. An early feature of PC is that the cells change their epithelial differentiation programme and exhibit invasion potential. Thus it is plausible that silencing of hHS6ST2 could prevent this ‘malfunction’ at an early stage.

Our finding that hHS6ST2 influences Notch signalling also reveals a hitherto unknown synergy in the modulation of EMT and angiogenesis. To explain how the co-operative control of EMT and angiogenesis is affected by the modulation of hHS6ST2-related signalling, several observations are worthy of further comment. A previous study showed that Notch signalling is required to convert the hypoxic stimulus into EMT [12]. The hypoxic response in epithelial cancers is mainly regulated by HIF-1 (hypoxia-inducible factor 1), a basic helix–loop–helix transcription factor. HIF-1 controls tumour cell proliferation and angiogenesis by transcriptional regulation of several genes encoding growth factors, including VEGF. Our finding that VEGF-A165 down-regulated endogenous hHS6ST2 offered an attractive explanation for the VEGF–Notch cross-talk in the ECs during the regulation of angiogenesis (Figures 5D and 5E). Our preliminary results suggested that the primary Notch ligand, Dll4, was up-regulated by overexpression of hHS6ST2. High expression of Dll4 in hHS6ST2-expressing ECs, which leads and guides new sprouts (tip), is thought to activate Notch and induce lateral inhibition, resulting in reduced VEGFR2 turnover and suppression of sprouting by adjacent ECs (stalk) [33]. Under normal circumstances, VEGF might repress hHS6ST2, thus forming a negative-feedback loop that inhibits Notch signalling. This may partially rescue the Dll4–Notch-1–VEGFR2 axis that then terminates tip/stalk cell selection and limits angiogenic activity (Figures 5B–5E). Our work provides another mechanism whereby inhibition of hHS6ST2 probably results in non-productive angiogenesis and reduces normal microvessel formation, through down-regulation of Dll4 and inhibition of Dll4/Notch signalling [14,34].

How does HS 6-O-sulfation control Notch signalling in PC? First, hHS6ST2-generated HS may stabilize the Notch protein. Secondly, hHS6ST2-modified HS epitopes may specifically induce Dll4–Notch-1 trans-interactions. To investigate the relationship between altered HS and endocytosis of Dll4, we confirmed that Dll4 could bind to Notch-1 on hHS6ST2-altered HUVEC-CS using recombinant human Dll4–Fc. Although hHS6ST2 has been shown to influence Notch directly and also has a secondary effect on the local expression of Dll4, we found that this was not the case regarding the endocytic activity of the Dll4 protein. We saw no apparent influence on binding after pre-incubation of the cells with recombinant Dll4–Fc, indicating that hHS6ST2 activity is required only in Notch signal transduction other than retaining Notch ligand (D114) on the cell surface (results not shown). Although our work demonstrated that hHS6ST2 modulation was implicated in Notch signalling, direct evidence of interaction between HS and Notch could not be obtained. However, several other heparin-binding proteins have been shown to regulate Notch signalling. Therefore hHS6ST2 may affect Notch activity through these molecules. Previous work shows that Notch signalling is very sensitive to alterations in membrane trafficking [35]. The β-secretase and γ-secretase complexes cleave the NECD, which is required for activation of the Notch receptor, and may potentially associate with HS. It is therefore possible that 6-O-sulfated HS interacts with presenilin1, a core component of the γ-secretase complex, to regulate the processing of Notch (K. Song, L. Zhang, J. Wang, P. Zhou and P. Li, unpublished work).

The present study identifies the co-operative roles played by hHS6ST2 in co-ordinating distinct signalling pathways to orchestrate angiogenesis and EMT. However, to understand fully the molecular mechanisms involved, analysing the composition of altered HS and HS–protein complexes, and developing new methods for visualizing direct responses to angiogenic signals and the establishment of morphogen gradient in vivo, presents difficult challenges. Also, because both 3-O- and 6-O-sulfation can alternatively be the final step of HS biosynthesis and a single modification may trigger additional modifications at nearby sugar codecs, thereby expanding the coding information, it is imperative to use a systematic approach to model the process.


Kai Song contributed to conception, design, acquisition of data of cellular and molecular biology experiments, and drafting the paper. Qin Li contributed to the analysis of HS and animal experiments. Yong-Bo Peng and Li-Jun Zhang contributed to acquisition and interpretation of proteomics data. Jie Li and Cheng-Hao Shao contributed to acquisition of the clinical sample data. Kan Ding, Li-Juan Chen and Ping Li completed the writing of the paper before submission.


This work was supported by Program for Changjiang Scholars and Innovative Research Team in University [grant number IRT0868] and the National Science and Technology Major Project ‘Creation of Major New Drugs’ from China [grant numbers 2009ZX09502-020, 2009ZX09301-001 and 2009ZX09103-071].


We greatly appreciate the gift of mSnail promoter plasmid from Professor Angela Nieto (Universidad Miguel Hernández-CSIC). We are very grateful to Dr Xiang Chen, Dr Jia Hu, Dr Xiao-Dong Wen, Dr Hai-Ling Liu, Mr Chao-wei Guo, Miss Zhen-zhen Jiang, Dr Xiao-fen Lv and Dr Mei-ting Ren for experimental assistance.

Abbreviations: DMEM, Dulbecco's modified Eagle's medium; EC, endothelial cell; EMT, epithelial–mesenchymal transition; ESI, electrospray ionization; F-actin, filamentous actin; FBS, fetal bovine serum; FGF, fibroblast growth factor; GAG, glycosaminoglycan; H&E, haematoxylin and eosin; HIF-1, hypoxia-inducible factor 1; hnRNP, heterogeneous nuclear ribonucleoprotein; HS, heparan sulfate; HS6ST, heparan sulfate D-glucosaminyl 6-O-sulfotransferase; dHS6ST, Drosophila HS6ST; hHS6ST, human HS6ST; HRP, horseradish peroxidase; HUVEC, human umbilical vein endothelial cell; IF, immunofluorescence; IHC, immunohistochemistry; MS/MS, tandem MS; NECD, Notch extracellular domain; NICD, Notch intracellular domain; PC, pancreatic cancer; PE, phycoerythrin; qPCR, quantitative PCR; RT, reverse transcription; shRNA, short hairpin RNA; TGF, transforming growth factor; TPM3, tropomyosin 3; VDAC1, voltage-dependent anion channel 1; VEGF, vascular endothelial growth factor; VEGFR2, VEGF receptor 2


View Abstract