Transcriptional Induction of Periostin by a Sulfatase2 TGFb ... · Published OnlineFirst November 21, 2016; DOI: 10.1158/0008-5472.CAN-15-2556 insert that generates siRNAs but contains

Microenvironment and Immunology

Transcriptional Induction of Periostin by aSulfatase 2–TGFb1–SMAD Signaling Axis MediatesTumor Angiogenesis in HepatocellularCarcinomaGang Chen1,2, Ikuo Nakamura2, Renumathy Dhanasekaran2,3, Eriko Iguchi4,Ezequiel J. Tolosa4, Paola A. Romecin4, Renzo E. Vera4, Luciana L. Almada4,Alexander G. Miamen2, Roongruedee Chaiteerakij2,5, Mengtao Zhou1,Michael K. Asiedu2, Catherine D. Moser2, Shaoshan Han2, Chunling Hu2, Bubu A. Banini2,Abdul M. Oseini2, Yichun Chen2, Yong Fang2, Dongye Yang2, Hassan M. Shaleh2,Shaoqing Wang2,3, Dehai Wu2, Tao Song2, Ju-Seog Lee6, Snorri S. Thorgeirsson7,Eric Chevet8, Vijay H. Shah2, Martin E. Fernandez-Zapico2,4, and Lewis R. Roberts2

Abstract

Existing antiangiogenic approaches to treat metastatic hepa-tocellular carcinoma (HCC) are weakly effectual, promptingfurther study of tumor angiogenesis in this disease setting.Here, we report a novel role for sulfatase 2 (SULF2) indriving HCC angiogenesis. Sulf2-deficient mice (Sulf2 KO)exhibited resistance to diethylnitrosamine-induced HCC anddid not develop metastases like wild-type mice (Sulf2 WT).The smaller and less numerous tumors formed in Sulf2 KOmice exhibited a markedly lower microvascular density. Inhuman HCC cells, SULF2 overexpression increased endothe-lial proliferation, adhesion, chemotaxis, and tube formationin a paracrine fashion. Mechanistic analyses identified the

extracellular matrix protein periostin (POSTN), a ligand ofavb3/5 integrins, as an effector protein in SULF2-inducedangiogenesis. POSTN silencing in HCC cells attenuatedSULF2-induced angiogenesis and tumor growth in vivo. TheTGFb1/SMAD pathway was identified as a critical signalingaxis between SULF2 and upregulation of POSTN transcrip-tion. In clinical HCC specimens, elevated levels of SULF2correlated with increased microvascular density, POSTNlevels, and relatively poorer patient survival. Together, our find-ings define an important axis controlling angiogenesis in HCCand a mechanistic foundation for rational drug development.Cancer Res; 77(3); 632–45. �2016 AACR.

IntroductionHepatocellular carcinoma (HCC) is a highly vascular tumor

and angiogenesis plays an important role during malignantprogression of HCC. Angiogenesis is a process regulated bymultiple growth factors, including VEGF, platelet-derivedgrowth factor (PDGF), and basic fibroblast growth factor (FGF;ref. 1). HCCs require new vessel formation both to sustaintumor growth and to facilitate local invasion (2). Tumor cellssecrete proangiogenic growth factors that induce a self-perpet-uating cycle of progressive tumor growth. Drugs targetingangiogenesis have proven to be effective antitumor agents inmultiple cancers including colorectal cancer and renal cellcarcinoma (3, 4). In the case of HCC, traditional chemother-apeutics are not effective treatment options, and indeed sor-afenib, an antiangiogenic agent, is currently the only drugapproved for HCCmanagement (5). Sorafenib is a multi-kinaseinhibitor that inhibits VEGF receptors, PDGF receptors, c-KIT, and BRAF. Unfortunately, sorafenib shows only limitedeffectiveness against advanced HCC, and most patientshave only modest survival benefit with significant side effects.There is therefore an urgent need to develop more effective

1Division of Hepatobiliary Surgery, The First Affiliated Hospital of WenzhouMedical University, Wenzhou, Zhejiang, China. 2Division of Gastroenterologyand Hepatology, Mayo Clinic College of Medicine and Science, Rochester,Minnesota. 3Division of Gastroenterology and Hepatology, Stanford University,Palo Alto, California. 4Schulze Center for Novel Therapeutics, Mayo Clinic,Rochester, Minnesota. 5Department of Medicine, Faculty of Medicine, Chula-longkorn University and King Chulalongkorn Memorial Hospital, Thai Red CrossSociety, Bangkok, Thailand. 6Department of Systems Biology, MD AndersonCancer Center, Houston, Texas. 7Laboratory of Experimental Carcinogenesis,National Cancer Institute, Bethesda, Maryland. 8INSERM U1242, Chemistry,Oncogenesis Stress Signaling, Universit�e Rennes 1, and Centre de Lutte Contrele Cancer Eug�ene Marquis, Rennes, France.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

G. Chen, I. Nakamura, R. Dhanasekaran, E. Iguchi, E.J. Tolosa contributed equallyto this article.

Corresponding Author: Lewis R. Roberts, Division of Gastroenterology andHepatology, Mayo Clinic College of Medicine and Science, 200 First Street SW,Rochester, MN 55905. Phone: 507-538-4877; Fax: 507-284-0762; E-mail:[email protected]

doi: 10.1158/0008-5472.CAN-15-2556

�2016 American Association for Cancer Research.

CancerResearch

Cancer Res; 77(3) February 1, 2017632

on December 18, 2020. © 2017 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst November 21, 2016; DOI: 10.1158/0008-5472.CAN-15-2556

http://cancerres.aacrjournals.org/

antiangiogenic strategies for HCC derived from a basic under-standing of the pathogenesis of this malignancy.

Heparan sulfate proteoglycans (HSPG) play important rolesin cancer progression. HSPGs on the cell surface or in the extra-cellular matrix act as storage sites for key proangiogenic factorsincluding FGF, TGFb, and VEGF. Growth factors also utilizeHSPGs as coreceptors for their tyrosine kinase receptors (6–8).Sulfation of particular saccharide moieties of HSPGs is requiredfor growth factor signaling. Sulfatase 2 (SULF2) catalyzes theremoval of 6-O-sulfate groups from heparan sulfate disaccharideunits of HSPGs, decreasing the affinity of HSPGs for heparansulfate binding ligands and thus releasing the ligands fromsequestration sites and making them available for signaling(9). In our previous study, SULF2 was overexpressed in 60% ofhumanHCCs and was associated with worse prognosis andmorerapid recurrence. Furthermore, SULF2 expression promoted pro-liferation of HCC cells in vitro and correlated with larger tumorvolume and poor survival of nude mice bearing HCC xenograftsin vivo (10, 11). Mechanistically, SULF2 enhanced FGF-mediatedextracellular signal–regulated kinase (ERK) and WNT/b-cateninsignaling (10, 12). In this study, we define a novel pathwayregulated by SULF2, which stimulates angiogenesis in HCC.Using a Sulf2 knockout mouse model, we define a new role forSULF2 in tumor progression of HCC involving the regulation oftumor angiogenesis. Furthermore, we show that the mechanismby which SULF2 modulates angiogenesis is through the up-regulation of POSTN in a TGFb/SMAD-dependent manner,an important modulator of angiogenesis in different tumors(13, 14). Together, the findings of this study propose theSULF2/POSTN axis as a novel target for the development ofantiangiogenic therapies for HCC.

Materials and MethodsA comprehensive list of all the reagents, kits, antibodies, and

primers used in this study is found in Supplementary Tables S1and S2.

AnimalsSulf2 knockout (Sulf2-KO) mice were generated by gene trap

insertional mutagenesis [strain name: B6;129P2-Sulf2Gt(PST111)Byg/Mmucd]. The homozygous mice were shown tohave mild skeletal abnormalities, including premature fusion ofthe sacral vertebrae and occasional fusion in the sternebrae (15).Homozygous (Hm) KO mice were 13% smaller than their WTlittermates at 4 weeks but appeared otherwise normal. Crosses ofhomozygous mice had lower fecundity than heterozygouscrosses. There was no obvious liver abnormality on histology ofthe Hm KO mouse livers and liver transaminase levels werenormal. Sulf2-KO mice were crossed with B6;129PF2J mice (cat-alog No. 100903) from Jackson Laboratory tomaintain the strainbackground. The care anduse of the animals for these studieswerereviewed and approved by the Mayo Clinic Institutional AnimalCare and Use Committee.

Diethylnitrosamine-induced liver tumor in miceDiethylnitrosamine (DEN) was used as carcinogen to induce

primary liver tumors. DEN is a genotoxic drug and its use has beenwell-established in mouse models for hepatocarcinogenesis (16).At 14 days of age, mice received a single intraperitoneal injectionof DEN (15 mg/kg body weight). At 21 days of age, mice wereseparated by sex and genotyped. Littermates with negative geno-

types were used as WT controls. Only male mice were included inthe analysis, as female mice rarely developed tumors with DEN.All mice were sacrificed at 8 months, and their liver and lungsexamined for tumors. The liver weight, number of visibletumors, and size of visible tumors were recorded.

Analysis of human HCC microarray gene expression dataHCC tumor and adjacent benign tissues were obtained at

surgical resection from 139 individuals at centers in Asia, Europe,and the United States. The details of themicroarray gene expressionprofiling were previously reported by Lee and colleagues (17).Kaplan–Meier analysis was used to compare overall survivalbetween patients with high versus low POSTN expression. Parteksoftware was used to identify SULF2-associated angiogenic factors.

The Cancer Genome Atlas gene expression analysismRNA expression data from HCC specimens (200 tumors; 50

surrounding normal liver tissues) was obtained from The CancerGenome Atlas (TCGA). The file containing level 3 normalizedRSEM (RNA-Seq by Expectation Maximization) data from theFirehose run of the Broad Genome Data Analysis Center wasdownloaded on Dec 17, 2014 (http://gdac.broadinstitute.org).Fold change for POSTN transcript was calculated as the ratio ofgene expression values of tumor to normal, with the mediansample value of normal tissues used as baseline. A ratio of log2(tumor/normal) greater than or equal to 1.0 (a fold change oftumor vs. normal greater than or equal to 2) was consideredincreased, and a ratio of log2 (tumor/normal) less than or equal to�1.0 was considered decreased. Spearman analysis was used toanalyze correlation between POSTN and SULF2.

Cell lines and rhPOSTN and rhTGFb1 treatmentsHep3B and PLC/PRF/5 cell lines (SULF2-negative) were

obtained from ATCC and cultured in complete Minimum Essen-tial Media (MEM) with 10% FBS. Huh-7 cells (SULF2-positive)were obtained from the Japan Health Science Research ResourcesBank (HSRRB, Osaka, Japan) and grown inDMEMwith 10%FBS.Human hepatic sinusoid endothelial cells (HHSEC) and humanumbilical vein endothelial cells (HUVEC) were obtained fromScienCell Research Laboratories. BothHHSECs andHUVECswerecultured in EGM-2 culture medium. Cell authentication wasperformed by short tandem repeat (STR) analysis (GENEWIZ)on the basis of the Approved American National Standards(ANSI).

For recombinant human POSTN (rhPOSTN) treatments,20 ng/mL of rhPOSTN was used unless otherwise noted. For re-combinant TGFb1 (rhTFGb1) treatments, 2 to 10 ng/mL rhTGFb1was added to the cells directly in the medium. For the chromatinimmunoprecipitation (ChIP) assay, Huh-7 cells were incubatedwith 5 ng/mL rhTGFb1 in serum-free medium for 1 hour.

Cell transfectionA recombinant plasmid expressing full-length human SULF2

cDNA cloned into the pcDNA3.1 expression vector from Invitro-gen was used as a SULF2-expressing plasmid (17). Hep3B cellswere grown to 60% to 80% confluency and transfected withSULF2-expressing plasmid or vector plasmid using FuGENE6Transfection Reagent from Roche. shRNAs cloned into lentivirusvector pLKO.1-puro were chosen from the human library (MIS-SION TRC-Hs 1.0) and purchased from Sigma-Aldrich. Nontargetcontrol shRNAs (pLKO.1 NTC vector, Sigma) contain a hairpin

Role of Sulfatase 2 in HCC Tumorigenesis

www.aacrjournals.org Cancer Res; 77(3) February 1, 2017 633




insert that generates siRNAs but contains 5-base pair mismatchesto any known human gene. The target sequences used for SULF2shRNA constructs were HW11: CAAGGGTTACAAGCAGTGTAAand HW13: CCACAACACCTACACCAACAA. Lentivirus particleswere produced by transient transfection of these 2 differentshRNAs targeting HSulf-2 (pLKO.1-HSulf-2; shSULF2) andpLKO.1 NTC (scrRNA) along with packaging vectors (pVSV-Gand pGag/pol) in 293T cells as previously described (18). Huh-7cells were grown to 80% confluency and infected with lentivirusparticles. siRNAs targeting POSTN (siPOSTN) from Santa Cruzwere transfected following the manufacturer's protocol. Nontar-geting siRNAs (control siRNA) from Santa Cruz were utilized as anegative control.

Production of conditioned mediaHep3B cells transfected with SULF2-expressing plasmid or

empty vector plasmid and Huh-7 cells infected with lentiviruscontaining shSULF2 or scrRNA for 48 hours were grown to 70%to 80% confluency and washed 3 times with PBS and thenincubated in fresh media containing 0.5% FBS for 48 hours.Conditioned media (CM) were harvested, centrifuged at 1,000rpm for 10 minutes to remove cell debris, filtered through a0.22-mm filter, and stored at 4�C. Hep3B cells were alsocotransfected with SULF2 and siPOSTN or control siRNA andtheir CM were harvested.

Coculture experimentsSULF2-downregulated Huh-7 and SULF2-overexpressing

Hep3B cells and their respective controls were seeded in 6-wellcoculture dishes. HHSECs and HUVECs were cultured on 0.4-mmpore size cell culture inserts from BD Biosciences for 24 hours at 1� 105 cells per well and then maintained in medium containing0.5%FBS. Inserts were placed in the insert companion plate for 48hours, which allow diffusion of media components but preventcell migration. Independent experiments were carried out at least3 times.

RNA extraction and quantitative real-time PCR analysisTotal RNA was extracted using RNAeasy Mini Kit (Qiagen).

cDNA synthesis was performed using High Capacity cDNAReverse Transcription Kit from Applied Biosystems to transcribe2 mg of total RNA. mRNA levels were quantified by real-timereverse transcription PCR (RT-qPCR) in an ABI 7300 system. EachmRNA level was normalized by comparison to 18S ribosomalRNA in the same samples. All reactions were performed intriplicate. Primer sequences are listed in Supplementary Table S2.

Western blot analysisWhole-cell lysates and tissues from mice were extracted in

lysis buffer and quantified using BCA Protein Assay Kit (Pierce).Equal amounts of protein (20 mg/lane) were separated on 4%to 15% SDS-PAGE and then transferred to polyvinylidenefluoride membranes (Millipore). After blocking for 1 hour with5% BSA and washing with TBST, the membranes were probedwith polyclonal or monoclonal antibodies against SULF2,POSTN, phospho-FAK, FAK, phospho-AKT, AKT, and b-actinby incubation at 4�C overnight. Subsequently, goat anti-mouseor goat anti-rabbit horseradish peroxidase–conjugated second-ary antibodies were used to detect antigen antibody complexes.Immune complexes were visualized using the HyGLO HRPDetection Kit from Denville Scientific and exposed to X-ray

films. b-Actin was used to control for equal loading. Eachexperiment was repeated at least 3 times.

Immunohistochemical stainingFor IHC staining, tumor tissues were fixed in formalin, embed-

ded in paraffin, cut into 4-mm sections, and stained with antibodyagainst CD31 and VEGFR2. Negative controls were set up byreplacement of the primary antibody with 1% BSA/TBS. Forquantification of tumor microvascular density (MVD), CD31-positive cells were identified by a brown precipitate in the cyto-plasm of endothelial cells; vessels in each section were counted in5 microscope fields (19).

Immunofluorescent staining and confocal microscopyFrozen sections of tumor tissues from either Sulf2-KO or WT

mice were permeabilized with 0.1% Triton X-100 in PBS for30 minutes. Sections were then blocked for 1 hour at roomtemperature with 5% goat serum and incubated with anti–heparan sulfate antibody (RB4EA12) overnight at 4�C. Sectionswere then rinsed in PBS and incubated with Alexa Fluor 594–conjugated anti-rabbit IgG from Invitrogen for 1 hour at roomtemperature. The content of sulfated disaccharides was validat-ed using high-performance liquid chromatography (HPLC) aspreviously described. Disaccharide peaks were identified byreference to the consistent elution positions of authentic disac-charide standards.

HHSECs andHUVECs were seeded on 8-well culture slides andincubated for 48 hours with CM from Huh-7 cells infected withscrRNA or shSULF2, Hep3B cells transfected with empty vector orSULF2, and Hep3B cells transfected with siPOSTN or controlsiRNA. Cells were rinsed with PBS at room temperature and fixedfor 20 minutes with 2.5% formaldehyde in piperazine-N,N0-bis(2-ethanesulfonic acid) (PIPES). They were then rinsed with PBSand blocked with 5% normal goat serum for 1 hour at 37�C. Afterincubation with antibody against POSTN, p-FAK, p-AKT, VEGFB,MMP2, MMP9, PDGFA, or PDGFB for 1 hour at 37�C, they wererinsed in PBS and incubated with Alexa Fluor 488–conjugatedanti-mouse from Molecular Probes for 1 hour at room temper-ature, and subsequently mounted with Prolong Gold anti-fadereagent with DAPI from Invitrogen. Slides were examined byconfocal microscopy (Zeiss LSM-510).

ELISATGFb1 concentrations were measured using a human TGFb1

immunoassay kit from Abcam. POSTN concentrations were mea-sured using a human POSTN immunoassay kit from MyBio-Source. Samples were prepared following the manufacturerprotocol.

Cell proliferation, chemotaxis, and adhesion assaysHHSECs (1�103) andHUVECs (2�103) cellswere seeded in a

96-well culture plate for 3 hours and washed 3 times with PBS.After confirmation of cellular adhesion to the plates, the mediumwas replaced with each CM. Different concentrations of recom-binant humanPOSTN(rhPOSTN) and siPOSTNwere also used totreat HHSECs and HUVECs. Twenty-four, 48, and 72 hours aftertreatment, endothelial cell viability was determined using MTTassays from ATCC. The absorbance was determined at 570 nm.Experiments were carried out in triplicate.

The chemotaxis of endothelial cells to different CM was mea-suredby Transwell assay.HHSECs (5�104) orHUVECs (5�104)

Chen et al.

Cancer Res; 77(3) February 1, 2017 Cancer Research634




in 500 mL serum-free medium were added into 24-well chamberswith 8-mm pore size from BD Biosciences. Then, 750 mL of eachCM supplemented with 0.5% FBS was used as a chemoattractantin the lower compartment. After 24-hour culture at 37�C, themigrated cells, adhering to the lower surface of the membrane,fixed with 100% methanol, were stained with crystal violet, andcounted under a light microscope. Migrated cells were counted in5 randomly selected microscopic fields for each chamber, andchemotaxis was expressed as the number of migrated cells perfield. Each experiment was performed in triplicate.

For measurement of cell adhesion, HHSECs and HUVECswere cultured in CM for 12 hours and resuspended in respectiveCM at a density of 2 � 105 cells/mL, followed by addition of100 mL cell suspension to each well of 96-well plates. Followingincubation for 1 hour at 37�C, unattached cells were removedby rinsing twice with PBS. Cells were then stained with crystalviolet for 10 minutes. Plates were read at 570 nm. Differentconcentrations of rhPOSTN and siRNA were also used to treatHHSECs and HUVECs. Each experiment was performed intriplicate.

Xenografts model and adenovirus infection protocolHep3B cells (1.75 � 106) were seeded in MEM medium

supplemented with 10% FBS and cultured overnight. For theinfection, adenoviruses were obtained from Vector BioLabs andused at concentrations suggested by the manufacturer. Fourexperimental groups were designed: (i) Ad-CMV-Null:Ad-GFP-U6-scrmb-shRNA (Empty vectorþ scrRNA); (ii) Ad-h-SULF2:Ad-GFP-U6-scrmb-shRNA (SULF2 þ scrRNA); (iii) Ad-CMV-Null:Ad-GFP-U6-h-POSTN-shRNA (Empty vector þ shPOSTN); and(iv) Ad-h-SULF2: Ad-GFP-U6-h-POSTN-shRNA (SULF2 þshPOSTN). In all the groups, the ratio between the SULF2-over-expressing adenovirus and the adenovirus used to knockdownPOSTN was of 0.5:1.5, respectively. The efficiency of SULF2overexpression and POSTN knockdown was tested by RT-qPCRand Western immunoblotting (data not shown). Seventy-twohours after infection, cells were harvested by trypsinization,washed with PBS, and reconstituted with Matrigel (1:10; BDBiosciences) in MEM to a final volume of 0.1 mL/mouse. Twen-ty-four 4- to 5-week-old male NOD/SCID mice were injectedsubcutaneously in the rightflank (at sacroiliac joint level)with5�106 adenovirus-infected Hep3B cells/mouse. Tumor growth wasmonitored 3 times a week and measured using a digital Verniercaliper to calculate the tumor volumes [V¼ (a� b2)/2].Mice werecared for and handled in accordance with institutional and NIHguidelines and after 40 days were sacrificed and tissue collected.

Cell tube formation assayEvaluation of tube formation by endothelial cells was per-

formed using the In Vitro Angiogenesis Assay Kit from Millipore.Briefly, ECMatrix gel solution was thawed, mixed with diluentbuffer, and placed in a precooled 96-well plate. The plate was thenplaced at 37�C for 1 hour to enable the matrix solution to gel.HHSECs and HUVECs were cultured in CM for 12 hours andresuspended in respective CM at a density of 2 � 105 cells/mL.One hundred microliters of the cell suspension was loaded ontothe surface of the gelledmatrix and incubated at 37�C for 6 hours.Images were acquired using Zeiss Axioplan 2 (bright-field,�100)with attached Zeiss Axiocam and Axio Vision 4.6.3 software, andtubule lengthwasmeasured using ImageJ image analysis softwarefrom NIH. Different concentrations of rhPOSTN and siPOSTN

were also used to treat HHSECs and HUVECs. Each experimentwas performed in triplicate.

ImmunoprecipitationHuh7 and Hep3B cells grown in 10-cm dishes were washed

twice with ice-cold PBS and lysed on ice for 10minutes in 1mL ofradioimmunoprecipitation assay (RIPA) lysis bufferwithproteaseinhibitor. Cellular debris was removed by centrifugation at10,000 � g for 10 minutes. The supernatant was transferred toa new tube, and 1 mg of mouse IgG and 20 mL volume ofresuspended Protein A/G Plus-Agarose Sepharose added (SantaCruz). After incubation for 30 minutes on ice, the beads werepelleted by centrifugation at 25,000 rpm for 5minutes at 4�C, andthe supernatant was transferred to a new tube. The proteinconcentration was measured and diluted to approximately100 mg/mL of total cell protein with PBS. The lysate was incubatedwith 10 mL ofmouse anti-TGFBR3 antibody on ice for 1 hour, andTGFBR3 protein was immunoprecipitated by Protein A/G PLUS-Agarose Sepharose (20 mL) overnight at 4�C. Immune complexeswere pelleted by centrifugation for 5 minutes at 1,000� g at 4�C,washed 4 times with RIPA buffer, and eluted from the beads byboiling in 40mLof 1� samplebuffer for 3minutes. About 20mLofthe sample was analyzed by SDS-PAGE. Western immunoblotanalysis was performed using anti-TGFb1 and anti-TGFBR3antibodies.

ChIP assayChIP was conducted as previously described (20). Briefly, 1

hour after TGFb1 treatment, Huh-7 cells (10 � 106) were cross-linked with 1% formaldehyde, followed by cell lysis. DNA wassheared using a Bioruptor 300 (Diagenode) to fragment DNA toabout 600 bp. Aliquots of the sheared chromatin were thenimmunoprecipitated using magnetic beads and SMAD3 anti-body or normal rabbit IgG. Following immunoprecipitation,cross-links were removed, and immunoprecipitated DNA waspurified using spin columns and subsequently amplified byquantitative PCR. PCR primers were designed to amplify regionsof the POSTN promoter containing potential SMAD-bindingsites. The sequences of the primers are listed in SupplementaryTable S2. Samples for quantitative SYBR PCR were performed intriplicate using the C1000 Thermal Cycler (Bio-Rad). Results arerepresented as percentage of input.

Statistical analysesAll data represent at least 3 independent experiments using cells

from separate cultures and are expressed as the mean� SEM. Thec2 testwas used to compare categorical variables, and the Student ttest was used to compare continuous variables. Differencesbetween the Kaplan–Meier curves of patients with HCC withupregulated and downregulated SULF2 expression in their initialHCC tissue as compared with the adjacent benign tissue wereanalyzed using the log-rank and Wilcoxon test. P < 0.05 wasconsidered to be statistically significant.

ResultsSULF2 loss impairs DEN-induced HCC

Todefine the role of Sulf2 in a biologically relevantHCCmodel,we generated a Sulf2-KO mouse. Heterozygous (Hz) KO micewere crossed with each other, and progeny were given a singleintraperitoneal injection of DEN on day 14 (Supplementary






Fig. S1A). On day 21, the mice were separated by sex, genotyped,and RT-qPCRwas used to confirm that the expression of Sulf2wasprogressively and significantly lower in Hz KO and Hm KOmicethan in Sulf2 WT mice (P < 0.001, Supplementary Fig. S1B).Western immunoblotting was also used to confirm knockdownof Sulf2 protein expression in both the normal liver and tumorcells (Supplementary Fig. S1C). We used an anti–heparan sulfateantibody (RB4EA12) detecting 6-O-sulfated HSPG to confirm thefunctional consequences of Sulf2 inactivation in the liver. Asexpected, Sulf2 Hm KO mice had higher levels of 6-O-sulfatedHSPGs than Sulf2 Hz KO or Sulf2 WT mice (Supplementary Fig.S1D). Analysis of liver heparan sulfate glycosaminoglycans(HSGAG) by purification, digestion to their constituent disac-charides, and HPLC confirmed that liver tissue from Sulf2Hz KOmice had higher levels of 6-O-sulfated HSPGs than Sulf2WTmice(Supplementary Fig. S1E).

Assessment of the phenotype of Sulf2 inactivation on DEN-induced liver tumorigenesis revealed that compared to Sulf2 WTmice (n¼ 13), Sulf2Hz KOmice (n¼ 11) had substantially fewer

and smaller tumors, and Sulf2 Hm KO mice (n ¼ 10) had thefewest and smallest tumors (Fig. 1A, Supplementary Fig. S2A).Hematoxylin and eosin (H&E) staining showed the typical fea-tures of HCC in all 3 groups with enlarged round hyperchromaticnuclei, high nuclear–cytoplasmic ratios, and moderate micro- ormacrovesicular fat globules in the cytoplasm (Fig. 1B). Signifi-cantly smaller proportions of Hm and Hz KO mice had �5tumors/mouse and tumor volume �35 mm3 compared with WTmice (P < 0.01 in Hm KO and P < 0.05 in Hz KO, Fig. 1C). Sulf2Hm KO mice also had significantly lower body weight and liverweight thanWTmice (Supplementary Fig. S2B). Tumor incidencewas not different between the 3 groups (Supplementary Fig. S2B).Also, we found that 23% (3 of 13) of the Sulf2WT and 17% (2 of11) of the Hz KO mice developed lung metastases, in contrast,none of the HmKOmice had lung metastases (P < 0.05; Fig. 1D).

SULF2 promotes angiogenesis in HCCThe above findings provide strong evidence for a role of

Sulf2 in HCC tumor progression. Because angiogenesis is

Figure 1.

Knockout of SULF2 inhibits HCC progression. A, Representative images of livers from Sulf2 WT, Sulf2 Hz KO, and Sulf2 Hm KO mice carrying tumors afterDEN treatment. At 14 days of age, mice received a single intraperitoneal injection of DEN. All mice were sacrificed at 8 months of age and their liverand lungs were examined for tumors. B, H&E staining of representative tumors from Sulf2WT, Sulf2 Hz KO, and Sulf2 Hm KO mice showed typical HCC. C, Bargraphs showed that significantly smaller proportions of Sulf2 Hm KO and Sulf2 Hz KO mice presented �5 tumors/mouse (top) and tumor volume�35 mm3 (bottom) compared with Sulf2WTmice. D, Bar graph representing the incidence of lung metastasis (left) and H&E staining (right) showed that Sulf2WT and Hz KO mice developed lung metastasis but Sulf2 Hm KO did not. � , P < 0.05; �� , P < 0.01.

Chen et al.





considered to be crucial for tumor growth and dissemination,we then evaluated the influence of Sulf2 on angiogenesis. Tothis end, liver tumors generated in Sulf2 Hm KO mice wereexamined by IHC for the endothelial cell marker CD31 fol-lowed by calculation of microvascular density (MVD). HCCsfrom Sulf2 Hm KO mice had lower MVD than tumors fromSulf2WTmice (P < 0.001, Fig. 2A). Similar to CD31, expressionof the angiogenic marker VEGFR2 showed downregulation in

tumors from Sulf2 Hm KO mice compared with tumors fromSulf2 WT animals (Fig. 2B).

Tumor angiogenesis involves the proliferation and migra-tion of endothelial cells. Hence, we examined whether thesecellular functions were affected by SULF2 in a paracrine mannerwith in vitro experiments. Hep3B cells, which are constitutive-ly low SULF2-expressing cells, were transfected with controlvector or SULF2, and Huh-7 (high SULF2-expressing cells) were

Figure 2.

SULF2 expression promotes angiogenesis in HCCs. A, DEN-induced HCCs from Sulf2 Hm KO mice had decreased MVD compared to tumors from Sulf2WT mice, as shown by CD31 IHC staining (left, �200) and quantification of tumor MVD (right). For quantification of MVD, CD31-positive cells wereidentified by a brown precipitate in the cytoplasm of endothelial cells; vessels in each section were counted in 5 microscope fields. B, VEGFR2 IHC stainingof HCC tumors from DEN-induced Sulf2 WT and Sulf2 Hm KO mice showed increased staining in Sulf2-expressing tumors (�400). Tumor tissues werefixed in formalin, embedded in paraffin, sectioned, and stained with antibody against VEGFR2. Tube formation (C) and chemotaxis (D) of HHSECs andHUVECs, measured after 12-hour treatment with CM from high SULF2-expressing Hep3B or low SULF2-expressing Huh-7, showed that the expression statusof SULF2 in the CM can affect tumor angiogenesis. � , P < 0.05; �� , P < 0.01; �� , P < 0.001.






transfected with scrambled shRNA (scrRNA) or SULF2 shRNA(shSULF2). CM from the above cells was used to treat HHSECsand HUVECs. Tube formation and chemotaxis of HHSECsand HUVECs were increased after treatment with CM fromhigh SULF2-expressing HCC cells (Hep3B SULF2), comparedwith CM from low SULF2-expressing cells [Huh-7 shSULF2;P < 0.05 for tube length except HUVEC treated with Huh-7-derived CM (P < 0.01), P < 0.001 for chemotaxis except HHSECtreated with Hep3B-derived CM (P < 0.01), Fig. 2C and D].Similarly, viability and adhesion of HHSECs and HUVECsincreased after treatment with CM from high SULF2-expressingHCC cells compared with control cells (P < 0.01 except HUVECviability under Hep3B-derived CM (P < 0.05) and HHSECadhesion under Hep3B-derived CM (P < 0.05; Supplement-ary Fig. S3A and S3B). Thus, we were able to demonstrate thatSULF2 regulates angiogenesis both in vivo and in vitro byincreasing angiogenic potency of endothelial cells.

SULF2-induced angiogenesis is dependent on POSTNTo determine the mediators of SULF2-induced HCC angio-

genesis, we conducted an in silico study to identify potentialcandidate effectors of this phenomenon (Supplementary Fig.S4A). We identified 178 candidate genes selected from 738publications (Supplementary Tables S3 and S4). We then per-formed a correlation analysis of these 178 candidates withSULF2 expression in a microarray gene expression analysis of139 resected human HCCs and identified the top 10 angiogenicfactors most significantly correlated with SULF2 expression(Supplementary Fig. S4B). We also added 15 well-establishedangiogenic factors (Supplementary Fig. S4C), which were not inthe top 10 to create a comprehensive list of 25 proangiogeniccandidates (17, 21). We then evaluated the expression of these25 candidates in Hep3B cells overexpressing SULF2 or not andHuh-7 cells transfected with scrRNA or shSULF2. We found thatoverexpression of SULF2 in Hep3B cells increased expressionof 15 of the 25 candidates examined. Among these, POSTN, aproangiogenic extracellular matrix protein secreted by tumor orstromal cells (22–24), was the most highly correlated gene,being induced approximately 6-fold by the overexpression ofSULF2 (P < 0.001, Fig. 3A). In contrast, down-regulation ofSULF2 in Huh-7 cells significantly suppressed POSTN expres-sion (P < 0.001, Fig. 3B). In vitro, we confirmed using RT-qPCR(P < 0.001, Fig. 3C), Western blotting (Fig. 3D), immunoflu-orescence (Fig. 3E), and ELISA (P < 0.01, Fig. 3F), that SULF2overexpression in hepatocytes increased POSTN expressionand secretion. Analysis of the DEN-induced HCCs showed a4.1-fold higher expression of POSTN in WT mice than in Sulf2Hm KO mice (P < 0.01, Supplementary Fig. S5A and S5B).

To further study the effect of SULF2 on POSTN expression inthe tumor microenvironment, we used an in vitro coculturemodel of HCC cells with endothelial cells. Coculture withSULF2-expressing Hep3B cells increased POSTN expression andsubsequently increased levels of phospho-AKT and phospho-FAK, known downstream effectors of POSTN (12), in bothHHSECs and HUVECs when compared with coculture withSULF2-negative Hep3B cells. Conversely, suppression of SULF2expression in Huh7 cells resulted in decreased POSTN expres-sion and decreased levels of phospho-AKT and phospho-FAKin cocultured HHSECs and HUVECs (Fig. 4A and C, respec-tively). Immunofluorescence showed similar results (Fig. 4Band D). To confirm that the activation of PI3K pathway was

indeed POSTN-mediated, we cotransfected SULF2-expressingHep3B cells with either scrRNA or siPOSTN and showed thatthe previously observed increase in phospho-AKT and phos-pho-FAK was aborted when POSTN was suppressed (Fig. 4A–D,last 2 columns).

To define whether SULF2-induced POSTN expression wasrequired for the activation of endothelial cells, SULF2 was over-expressed inHep3B cells, and the cells were then subsequently co-transfected with either siRNA targeting POSTN or control scram-bled siRNA. CM was collected 24 hours later. Treatment ofHHSECs and HUVECs with CM from SULF2-expressing Hep3Bcells cotransfected with siPOSTN decreased endothelial cellviability (P < 0.05), adhesion (P < 0.01), chemotaxis (P < 0.01inHHSEC and P < 0.05 inHUVEC), and tube formation (P < 0.05in HHSEC and P < 0.001 in HUVEC) when compared with CMfrom SULF2-expressing Hep3B cells cotransfected with controlsiRNA (Fig. 5A). Further analysis using a similar experimentaldesign in a Hep3B xenograft model showed that inactivation ofPOSTN impairs SULF2-induced tumor growth (Fig. 5B) andvesseldensity (data not shown).

To further confirm the effect of POSTN on endothelialcells, we directly treated HHSECs and HUVECs with recom-binant human POSTN (rhPOSTN) and showed that this ledto increased angiogenic potency of endothelial cells as assess-ed by cell viability (P < 0.05 in HHSEC and P < 0.01 inHUVEC), adhesion (P < 0.001), and tube formation assay(P < 0.01; Fig. 5C).

Given the previous evidence that SULF2 promotes angiogen-esis by increasing the expression of several angiogenic factors inHCC (Fig. 3A and B), we hypothesized that the SULF2-mediatedincreased expression of angiogenic factors was partly mediatedby activation of the POSTN pathway. Therefore, we examinedthe mRNA expression profile of Hep3B cells after transfectionwith SULF2 plasmid followed by cotransfection with eithersiPOSTN or control siRNA. As hypothesized, targeting thePOSTN pathway in SULF2-expressing Hep3B cells by transfec-tion with siPOSTN led to a decrease in the mRNA expression ofthe angiogenic factors VEGFB, MMP2, MMP9, PDGFA, andPDGFB (P < 0.001 for PDGFB, P < 0.01 for MMP9 and P <0.05 for the others, Fig. 5D). In summary, these data indicatethat POSTN expression is a key mediator of SULF2-dependentangiogenesis.

SULF2 increases POSTN expression via activation of theTGFb pathway

Next, we investigated the mechanism by which SULF2induces POSTN expression. Because SULF2 acts via the mod-ulation of growth factor–dependent pathways and TGFb hasbeen reported to transcriptionally induce POSTN (25, 26), weevaluated whether SULF2 induced POSTN via activation of theTGFb pathway. We treated HCC cells with low or high SULF2expression with TGFb1 and found that expression of SULF2 inHep3B cells led to activation of TGFb1 downstream signaling,as evidenced by increase in phospho-SMAD2 and phospho-SMAD3 expression, and enhanced TGFb1-induced POSTNexpression (Fig. 6A). Similar findings were confirmed by exam-ining POSTN mRNA expression in HCC cells and POSTNprotein expression in the CM (Fig. 6B). We found by immu-nocytochemistry and Western immunoblotting higher levels ofTGFb1 and POSTN in SULF2-expressing cells compared withthe controls (Fig. 6C and D). We then explored the mechanism

Chen et al.





by which SULF2 activates the TGFb pathway. Of the 3 receptors,TGFBR1, TGFBR2, and TGFBR3/betaglycan, we identifiedTGFBR3 as a potential target for desulfation by SULF2, as itis an HSPG. TGFBR3 functions as a storage coreceptor for theTGFb1 ligand at the cell surface, sequestering the ligand and

thus inhibiting TGFb1 signaling (27). To confirm this effect inHCC cells, Huh-7 cells with low or high SULF2 expression weretreated with TGFb1 and their cell lysates were then immuno-precipitated using antibodies against TGFbR3. This revealedthat the interaction between TGFb1 and TGFbR3 was reduced in

Figure 3.

POSTN is a key downstream signaling molecule mediating SULF2 induced HCC angiogenesis. Relative mRNA expression of 25 proangiogenic factors in Hep3Bcells transfected with control or SULF2-overexpressing vectors (A) or Huh-7 transfected with scrRNA or shSULF2 (B). Together, these results showed thatPOSTN was the most highly correlated gene to SULF2 expression status. Relative expression of POSTN at mRNA level (C), protein level (D), and byimmunofluorescent staining (E) in Hep3B cells overexpressing SULF2, Huh-7 transfected with siSULF2, and their corresponding controls showed that theexpression of POSTN was dependent on the levels of SULF2. F, ELISA performed on the CM of the four different groups of cells described above showed thatSULF2 regulates the secretion of POSTN protein into the medium. � , P < 0.05; �� , P < 0.01; ��, P < 0.001.






SULF2-expressing cells (Fig. 6E), likely thus increasing TGFb1availability. This hypothesis is further confirmed by ELISAdemonstrating that CM from SULF2-expressing cells had higherconcentrations of TGFb1 in a time-dependent manner whencompared with parental control vector–transfected cells(Fig. 6F). After that, we confirmed that SULF2-mediated stim-ulation of POSTN is TGFb-dependent by demonstrating thatthe increase in POSTN expression and secretion in SULF2-expressing cells can be blocked by treating the cells with theTGFbR inhibitor SB431542 (Fig. 6G). Finally, we confirmed byChIP assay that POSTN is downstream of TGFb1 pathwayeffector such as Smad3 (Fig. 6H). To summarize, we demon-strate that SULF2 promotes the release of TGFb1 from TGFBR3,thereby leading to activation of the TGFb pathway, which inturn transcriptionally enhances expression and secretion ofPOSTN.

Correlated expression of SULF2 and POSTN in human HCCpredicts poor prognosis

Finally, to determine the translational significance of ourfindings, we examined the expression of SULF2 and POSTN inhuman HCCs. We first used microarray data from 139 resected

human HCC tumors and showed that SULF2 expression inhuman HCCs was positively correlated with more than 70(40%) of the 178 angiogenesis-related genes we identifiedearlier, including VCAM1, PDGFRa, and PDGFRb (Supplemen-tary Table S3). Furthermore, in tumors highly expressing SULF2,the mRNA levels of the neovascularization markers CD31 andCD34 were higher than in tumors with low SULF2 expression(P < 0.001 for CD31 and P < 0.05 for CD34, Fig. 7A). Thus,SULF2 expression appears to be associated with angiogenesis inhuman HCCs. Also, a subset of human HCCs overexpressedPOSTN (Fig. 7B, right), and POSTN expression was positivelycorrelated with SULF2 (P < 0.002, Fig. 7B, left). Survival analysisshowed that patients with HCC tumors expressing high levels ofPOSTN had a significantly worse prognosis than those withlow POSTN expression (P < 0.01, Fig. 7C). To validate thesefindings, we used an independent dataset of 200 HCC tumorsand 50 peritumoral benign liver tissues from TCGA project. Thedetails of analysis are described in Materials and Methods. Wewere able to confirm that POSTN was indeed significantly over-expressed in human HCC tumor tissue when compared withsurrounding normal liver and that POSTN expression signifi-cantly positively correlated with SULF2 expression (P < 0.0001,

Figure 4.

SULF2 modulates paracrine activation of the POSTN pathway in endothelial cells. Immunoblotting analysis (A) and immunofluorescence (B) of POSTN andits signaling molecules p-FAK and p-AKT expressed in HHSECs cocultured for 48 hours with Hep3B SULF2 or Huh-7 shSULF2 cells and their respectivecontrols, as well as cocultured with high SULF2-expressing Hep3B cells, in which POSTN is targeted by siRNA. Immunoblotting analysis (C) andimmunofluorescence (D) of POSTN and its signaling molecules p-FAK and p-AKT expressed in HUVECs cocultured for 48 hours with Hep3B SULF2 or Huh-7siSULF2 cells and their respective controls, as well as cocultured with high SULF2-expressing Hep3B cells in which POSTN is targeted by siRNA.

Chen et al.





Figure 5.

SULF2-mediated promotion of angiogenesis is POSTN-dependent. A, HHSECs and HUVECs cells were incubated for 12 hours with CM from SULF2-overexpressing Hep3B cells cotransfected with siPOSTN and then collected and tested for cell viability, adhesion, chemotaxis, and tube formation. Resultsshowed a decrease in the angiogenic potency of the cells when POSTN expression was donwregulated. B, Representation of the tumor volume ofsubcutaneous mouse xenografts. Combinations of Hep3B cells infected with adenovirus containing either shSULF2 or control vector (empty vector)along with shPOSTN or scramble control (scrRNA) were injected subcutaneously in the right flank of NOD/SCID mice (6 mice per group). Tumor volumewas measured three times a week up to the end point of the experiment. C, Viability, adhesion, and tube formation ability of HHSECs and HUVECs wereaugmented after treatment with increasing concentrations of recombinant human (rh) POSTN. D, Relative mRNA expression of SULF2-associatedangiogenic factors in Hep3B SULF2 cells cotransfected with either control siRNA or siPOSTN. Results of immunofluorescent staining confirmed theresults of RT-PCR, where the downregulation of POSTN expression led to a decrease in expression of the angiogenic factors VEGFB, MMP2, MMP9, PDGFA,and PDGFB. � , P < 0.05; �� , P < 0.01; �� , P < 0.001.






Figure 6.

SULF2 increases POSTN expression via activation of the TGFb pathway. A, Immunoblotting of Hep3B cell lysates with low or high SULF2 expression showedinduction of POSTN after recombinant (rh) TGFb1 treatment along with the activation of the TGFb1 downstream pathway. B, POSTN expression analyzedby RT-qPCR and ELISA in HCC cells with low or high SULF2 expression after TGFb1 treatment confirmed the results obtained at the protein level. C,Immunofluorescence showed higher levels of TGFb1 and POSTN in high SULF2-expressing HCC cells compared with the controls, for both times tested. D,Immunoblotting of Hep3B and PLC/PRF/5 cell lysates after treatment with rhTGFb1 at 0, 24, 48, and 72 hours. Results showed induction of TGFb1 and POSTNexpression in a time-dependent manner. (Continued on the following page.)

Chen et al.





Fig. 7D). Also, high POSTN expression was confirmed to beassociated with poorer overall survival (P ¼ 0.002, Fig. 7E).Thus, we show that in human HCC, SULF2 and POSTN appearto be coexpressed and confer a poor prognosis.

Discussion

In this study, we demonstrated for the first time that SULF2has a proangiogenic role in HCC, one of the major drivers of the

Figure 7.

SULF2 and POSTN are coexpressed in human HCC and associated with poor patient survival.A, Relative mRNA expression of CD31 and CD34 in HCC specimens. Forthis analysis, sampleswere divided in high SULF2 and lowSULF2-expressing tumors. Results, presented as log2 fold change between HCC tumor and adjacent benigntissues, showed that the mRNA levels of CD31 and CD34 were more elevated in high SULF2–expressing than in low SULF2–expressing HCC. B, POSTN isoverexpressed in a subset of human HCCs and its expression is positively correlated with SULF2 expression. C, Patients with HCC tumors expressing highlevels of POSTN had a significantly worse prognosis than those with low POSTN expression. D, Analysis of TCGA data shows that POSTN was significantlyoverexpressed in tumor tissues when compared with peritumoral benign tissues (left) and POSTN expression is positively correlated with SULF2 expression (right).E, High POSTN expression was confirmed to be associated with poorer overall survival in the TCGA dataset. � , P < 0.05; �� , P < 0.001.

(Continued.) E, Immunoprecipitation revealed loss of interaction between TGFb1 and TGFBR3 in Huh-7 cell lysates with high SULF2 expression. F, ELISAshowed higher concentrations of TGFb1 in the CM from high SULF2-expressing cells compared with controls and that the increment is time-dependent.G, POSTN expression analyzed by RT-qPCR and ELISA in HCC cells with high SULF2 expression after treatment with rhTGFb1, in the presence orabsence of a TGFbR inhibitor treatment (SB431542). Results showed that the SULF2-mediated stimulation of POSTN is TFGb1-dependent. H, Bioinformaticsanalysis of the POSTN promoter identified two regions with candidate SMAD (S)-binding sites upstream of the first exon and primers for q-PCR weredesigned to amplify these sites. ChIP assay performed in Huh-7 cells after 1-hour treatment with 5 ng/mL rhTGFb1 in the absence of FBS confirmed bindingof Smad3 transcription factor to the POSTN gene promoter in the Site#2. � , P < 0.05; �� , P < 0.01.






observed tumor progression. UponDEN treatment, Sulf2HmKOmice had substantially fewer and smaller tumors with absence oflung metastases and lower MVD. We demonstrate that SULF2expression in the tumor cells enhances the angiogenic potency ofendothelial cells in a paracrine fashion, thus promoting angio-genesis. Furthermore, our in vivo, in vitro, and human HCC dataidentify POSTN as a novel key downstream mediator of SULF2-dependent angiogenesis and show that RNAi targeting of POSTNin HCC cells attenuates SULF2-induced angiogenesis. Thus, weestablish the SULF2/POSTNaxis as a novel target for developmentof new antiangiogenic therapies for HCC.

Angiogenesis is stimulated by the enlarging tumor mass,which secretes proangiogenic factors which, in turn, induce theactivation and proliferation of endothelial cells in a paracrinefashion, resulting in the sprouting of new vessels from pre-existing ones. SULF2, a heparan sulfate 6-O-sulfatase, releasesgrowth factors from extracellular storage sites and facilitatestumor progression (8, 10, 12, 28), but its role in modulatingangiogenesis in HCC has not been studied before. In the presentstudy, we examined the role of SULF2 in mediating angiogenesisduring HCC progression using in vitro and in vivo assays. AsSULF2 functions by desulfating HSPG chains and thereby reg-ulating the interactions of HSPGs with heparan sulfate–bindinggrowth factors, cytokines, and receptors, we hypothesized thatSULF2 mediates the paracrine effects of HCC cells on angio-genesis by regulating the local concentration of angiogenicgrowth factors in the tumor microenvironment through HSPGdesulfation and release of HSPG-bound angiogenic factors onthe cell membrane and in the extracellular space. Althoughresults from our previous studies and from other researchgroups have implicated SULF2 in the regulation of severalgrowth factor signaling pathways, the key downstream pathwaymediating SULF2 signaling is still unknown and is likely to bedifferent in different cellular contexts and for different func-tional effects (10, 12, 29).

Through an extensive literature search and in vitro experiments,we identified for the first time that POSTN was the most highlyregulated downstream angiogenic target of SULF2 in HCC. Upre-gulation of SULF2 expression in HCC cells was accompanied byboth increased expression and secretion of POSTN, confirmingSULF2 can modulate POSTN levels in the extracellular matrix.POSTN is an extracellular matrix protein that is overexpressed in anumber of human cancers and has been shown to promoteangiogenesis by enhancing the proangiogenic potency of endo-thelial cells (30). In this study, we found downregulation ofPOSTN and its downstream signaling substrates phospho-AKTand phospho-FAK in DEN-induced HCCs in Sulf2 KO mice.Thus, the in vivo evidence suggests that POSTN expression andPOSTN pathway activation in HCC was SULF2-dependent. Fur-thermore, using in vitro coculture models, we demonstratethat increased expression of SULF2 stimulates POSTN expressioninbothHUVECs andHHSECs. Keeping in linewith this, we foundthat SULF2-dependent secretion of POSTN by HCC cells ledto activation of POSTN downstream signaling through phos-pho-FAK and phospho-AKT in endothelial cells. This SULF2-dependent activation of the POSTN pathway in endothelialcells resulted in enhancement of endothelial cell function asmeasured by endothelial cell viability, adhesion, chemotaxis, andtube formation. These effects were confirmed in HUVECs andwere also shown for the first time in HHSECs, liver-specificendothelial cells, which exist in the unique liver microenviron-

ment (31). Although POSTNwas found to be themost significantdownstreammediator of SULF2-dependent angiogenesis, we didfind that several other proangiogenic factors like VEGFB, MMP-2,MMP-9, PDGFA, and PDGFB were also overexpressed in SULF2-expressing cells (Fig. 3A and B). To further clarify the role ofPOSTN in the induction of the other angiogenic factors, weperformed in vitro experiments by blocking POSTN expressionin SULF2-expressing cells. This resulted in inhibition of theexpression of the SULF2-induced angiogenic factors mentionedabove, thus suggesting that SULF2 requires POSTN expression tofuel the ongoing production of angiogenic factors, thus promot-ing angiogenesis of HCC. Activation of this axis appears to bereflected in patient clinical outcomes, as patients with highPOSTN expression in their HCC tumor tissues have substantiallyworse prognosis.

After the approval of sorafenib for treatment of advancedHCC,there was great hope that other molecular-targeted therapies forHCC would soon follow. Unfortunately, thus far, multiple sub-sequent phase III clinical trials have failed to identify a safe andmore effective drug for HCC (32). Hence, it is critical to identifynovel antiangiogenic drug targets for treatment of HCC. SULF2is overexpressed in a majority of HCCs and associated withpoor prognosis (33). In view of its ability to regulate multipleangiogenic factors in HCC, SULF2 is likely to be upstream ofmany of the currently targeted angiogenic pathways. Interesting-ly, the heparin sulfate mimetic PI-88, an inhibitor of SULF2function, displays strong antiangiogenic potency (34). Further-more, this study uses 2 different datasets to demonstrate for thefirst time that POSTN is overexpressed in a subset of humanHCCsand that patients with high POSTN expression have poor prog-nosis.Hence, both SULF2andPOSTNmay serve as biomarkers formore aggressive disease. Consequently, inhibition of the SULF2-POSTN axis may be a useful therapeutic strategy in combinationwith other treatments for HCC.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

DisclaimerIts contents are solely the responsibility of the authors and do not necessarily

represent the official views of the NIH.

Authors' ContributionsConception and design: G. Chen, R. Dhanasekaran, E.J. Tolosa, A.G. Miamen,M.K. Asiedu, M.E. Fernandez-Zapico, L.R. RobertsDevelopment of methodology: G. Chen, I. Nakamura, R. Dhanasekaran, P.A.Romecin, A.G.Miamen,M. Zhou,M.K. Asiedu, S. Han, B.A. Banini, A.M.Oseini,H.M. Shaleh, M.E. Fernandez-ZapicoAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): G. Chen, I. Nakamura, R. Dhanasekaran, E. Iguchi,R.E. Vera, L.L. Almada, A.G. Miamen, R. Chaiteerakij, M.K. Asiedu, C.D. Moser,C. Hu, Y. Chen, H.M. Shaleh, S.S. Thorgeirsson, E. Chevet, V.H. ShahAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): G. Chen, I. Nakamura, R. Dhanasekaran, E. Iguchi,A.G. Miamen, R. Chaiteerakij, M.K. Asiedu, B.A. Banini, T. Song, J.-S. Lee, M.E.Fernandez-Zapico, L.R. RobertsWriting, review, and/or revision of the manuscript: G. Chen, R. Dhana-sekaran, E. Iguchi, E.J. Tolosa, L.L. Almada, A.G. Miamen, R. Chaiteerakij,M. Zhou, M.K. Asiedu, B.A. Banini, A.M. Oseini, D. Yang, D. Wu, E. Chevet,V.H. Shah, M.E. Fernandez-Zapico, L.R. RobertsAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): I. Nakamura, A.G. Miamen, C.D. Moser, S. Han,Y. Chen, Y. Fang, D. Yang, S.Wang, T. Song,M.E. Fernandez-Zapico, L.R. RobertsStudy supervision: M.E. Fernandez-Zapico, L.R. Roberts

Chen et al.





Grant SupportThis work was supported by NIH grants CA128633 to L.R. Roberts and

CA165076 to L.R. Roberts and M.E. Fernandez-Zapico; the Mayo Clinic Centerfor Cell Signaling in Gastroenterology (NIDDK P30DK084567) to L.R. Robertsand M.E. Fernandez-Zapico; the Mayo Clinic Cancer Center (CA15083) toL.R. Roberts; the Mayo Clinic Center for Translational Science Activities(NIH/NCRR CTSA Grant Number UL1 TR000135) to L.R. Roberts; and theNational Natural Science Foundation of China81201953 to G. Chen.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received September 16, 2015; revised October 7, 2016; accepted October 27,2016; published OnlineFirst November 21, 2016.

References1. Zhu AX, Duda DG, Sahani DV, Jain RK. HCC and angiogenesis: possible

targets and future directions. Nat Rev Clin Oncol 2011;8:292–301.2. Bishayee A, Darvesh AS. Angiogenesis in hepatocellular carcinoma: a

potential target for chemoprevention and therapy. Curr Cancer DrugTargets 2012;12:1095–118.

3. McCormack PL, Keam SJ. Bevacizumab: a review of its use in metastaticcolorectal cancer. Drugs 2008;68:487–506.

4. Powles T,Chowdhury S, Jones R,MantleM,NathanP, BexA, et al. Sunitiniband other targeted therapies for renal cell carcinoma. Br J Cancer 2011;104:741–5.

5. Llovet JM, Ricci S,Mazzaferro V,Hilgard P, Gane E, Blanc JF, et al. Sorafenibin advanced hepatocellular carcinoma. N Engl J Med 2008;359:378–90.

6. Campbell JS, Hughes SD, Gilbertson DG, Palmer TE, Holdren MS,Haran AC, et al. Platelet-derived growth factor C induces liver fibrosis,steatosis, and hepatocellular carcinoma. Proc Natl Acad Sci U S A 2005;102:3389–94.

7. Ogasawara S, Yano H, Iemura A, Hisaka T, Kojiro M. Expressions of basicfibroblast growth factor and its receptors and their relationship to prolif-eration of human hepatocellular carcinoma cell lines. Hepatology 1996;24:198–205.

8. Neaud V, Faouzi S, Guirouilh J, Le Bail B, Balabaud C, Bioulac-Sage P, et al.Human hepatic myofibroblasts increase invasiveness of hepatocellularcarcinoma cells: evidence for a role of hepatocyte growth factor. Hepatol-ogy 1997;26:1458–66.

9. Rosen SD, Lemjabbar-Alaoui H. Sulf-2: an extracellular modulator of cellsignaling and a cancer target candidate. Expert Opin Ther Targets 2010;14:935–49.

10. Lai JP, Oseini AM, Moser CD, Yu C, Elsawa SF, Hu C, et al. The oncogeniceffect of sulfatase 2 in human hepatocellular carcinoma ismediated in partby glypican 3-dependent Wnt activation. Hepatology 2010;52:1680–9.

11. Nawroth R, van Zante A, Cervantes S, McManus M, Hebrok M, Rosen SD.Extracellular sulfatases, elements of the Wnt signaling pathway, positivelyregulate growth and tumorigenicity of human pancreatic cancer cells. PLoSOne 2007;2:e392.

12. Lai JP, SandhuDS, YuC,Han T,Moser CD, JacksonKK, et al. Sulfatase 2 up-regulates glypican 3, promotes fibroblast growth factor signaling, anddecreases survival in hepatocellular carcinoma. Hepatology 2008;47:1211–22.

13. Shao R, Bao S, Bai X, Blanchette C, Anderson RM, Dang T, et al. Acquiredexpression of periostin by human breast cancers promotes tumor angio-genesis through up-regulation of vascular endothelial growth factor recep-tor 2 expression. Mol Cell Biol 2004;24:3992–4003.

14. ZhuM, FejzoMS, Anderson L, Dering J, Ginther C, Ramos L, et al. Periostinpromotes ovarian cancer angiogenesis and metastasis. Gynecol Oncol2010;119:337–44.

15. Ratzka A, Kalus I, Moser M, Dierks T, Mundlos S, Vortkamp A. Redundantfunction of the heparan sulfate 6-O-endosulfatases Sulf1 and Sulf2 duringskeletal development. Dev Dyn 2008;237:339–53.

16. Dhanasekaran R, Nakamura I, Hu C, Chen G, Oseini AM, Seven ES, et al.Activation of the transforming growth factor-b/SMAD transcriptionalpathway underlies a novel tumor-promoting role of sulfatase 1 in hepa-tocellular carcinoma. Hepatology 2015;61:1269–83.

17. Lee JS, Heo J, Libbrecht L, Chu IS, Kaposi-Novak P, Calvisi DF, et al. A novelprognostic subtype of human hepatocellular carcinoma derived fromhepatic progenitor cells. Nat Med 2006;12:410–6.

18. Khurana A, McKean H, Kim H, Kim SH, Mcguire J, Roberts LR, et al.Silencing of HSulf-2 expression in MCF10DCIS.com cells attenuate ductalcarcinoma in situ progression to invasive ductal carcinomain vivo. BreastCancer Res 2012;14:R43.

19. de Jong JS, van Diest PJ, Baak JP. Hot spot microvessel density and themitotic activity index are strong additional prognostic indicators in inva-sive breast cancer. Histopathology 2000;36:306–12.

20. Lo Re AE, Fernandez-Barrena MG, Almada LL, Mills L, Elsawa SF, Lund G,et al. A novel AKT1-GLI3-VMP1 pathwaymediates Kras induced autophagyin cancer cells. J Biol Chem 2012;287:25325–34

21. Hernandez-Gea V, Toffanin S, Friedman SL, Llovet JM. Role of the micro-environment in the pathogenesis and treatment of hepatocellular carci-noma. Gastroenterology 2013;144:512–27.

22. Bao S, Ouyang G, Bai X, Huang Z, Ma C, Liu M, et al. Periostin potentlypromotesmetastatic growth of colon cancer by augmenting cell survival viathe Akt/PKB pathway. Cancer Cell 2004;5:329–39.

23. Erkan M, Kleeff J, Gorbachevski A, Reiser C, Mitkus T, Esposito I, et al.Periostin creates a tumor-supportive microenvironment in the pancreasby sustaining fibrogenic stellate cell activity. Gastroenterology 2007;132:1447–64.

24. Tilman G, Mattiussi M, Brasseur F, van Baren N, Decottignies A. Humanperiostin gene expression in normal tissues, tumors and melanoma:evidences for periostin production by both stromal and melanoma cells.Mol Cancer 2007;6:80.

25. Kim JE, Jeong HW, Nam JO, Lee BH, Choi JY, Park RW, et al. Identificationof motifs in the fasciclin domains of the transforming growth factor-beta-induced matrix protein betaig-h3 that interact with the alphavbeta5integrin. J Biol Chem 2002;277:46159–65.

26. Horiuchi K, Amizuka N, Takeshita S, Takamatsu H, Katsuura M, Ozawa H,et al. Identification and characterization of a novel protein, periostin, withrestricted expression to periosteum and periodontal ligament andincreased expression by transforming growth factor beta. J Bone MinerRes 1999;14:1239–49.

27. Eickelberg O, Centrella M, Reiss M, Kashgarian M, Wells RG. Betaglycaninhibits TGF-b signaling by preventing type I-type II receptor complexformation: Glycosaminoglycan modifications alter betaglycan function.J Biol Chem 2002;277:823–9.

28. Nakamura I, Fernandez-Barrena MG, Ortiz-Ruiz MC, Almada LL, Hu C,Elsawa SF, et al. Activation of the transcription factor GLI1 by WNTsignaling underlies the role of SULFATASE 2 as a regulator of tissueregeneration. J Biol Chem 2013;288:21389–98.

29. UchimuraK,Morimoto-TomitaM, BistrupA, Li J, LyonM,Gallagher J, et al.HSulf-2, an extracellular endoglucosamine-6-sulfatase, selectively mobi-lizes heparin-bound growth factors and chemokines: effects on VEGF,FGF-1, and SDF-1. BMC Biochem 2006;7:2.

30. Liu AY, Zheng H, Ouyang G. Periostin, a multifunctional matricellularprotein in inflammatory and tumor microenvironments. Matrix Biol2014;37:150–6.

31. Van den Eynden GG, Majeed AW, Illemann M, Vermeulen PB, Bird NC,Hoyer-Hansen G, et al. The multifaceted role of the microenvironment inliver metastasis: biology and clinical implications. Cancer Res 2013;73:2031–43.

32. Llovet JM, Hernandez-Gea V. Hepatocellular carcinoma: reasons for phaseIII failure and novel perspectives on trial design. Clin Cancer Res 2014;20:2072–9.

33. Yang JD, Sun Z, Hu C, Lai J, Dove R, Nakamura I, et al. Sulfatase 1 andsulfatase 2 in hepatocellular carcinoma: associated signaling pathways,tumor phenotypes, and survival. Genes Chromosomes Cancer 2011;50:122–35.

34. Joyce JA, Freeman C, Meyer-Morse N, Parish CR, Hanahan D. A func-tional heparan sulfate mimetic implicates both heparanase and heparansulfate in tumor angiogenesis and invasion in a mouse model ofmultistage cancer. Oncogene 2005;24:4037–51.






2017;77:632-645. Published OnlineFirst November 21, 2016.Cancer Res Gang Chen, Ikuo Nakamura, Renumathy Dhanasekaran, et al. Hepatocellular CarcinomaSMAD Signaling Axis Mediates Tumor Angiogenesis in

−1βTGF−Transcriptional Induction of Periostin by a Sulfatase 2

Updated version

10.1158/0008-5472.CAN-15-2556doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2016/11/19/0008-5472.CAN-15-2556.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/77/3/632.full#ref-list-1

This article cites 34 articles, 8 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/77/3/632.full#related-urls

This article has been cited by 2 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/77/3/632To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-15-2556

http://cancerres.aacrjournals.org/content/suppl/2016/11/19/0008-5472.CAN-15-2556.DC1

http://cancerres.aacrjournals.org/content/77/3/632.full#ref-list-1

http://cancerres.aacrjournals.org/content/77/3/632.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/77/3/632


Documents

Transcriptional Induction of Periostin by a Sulfatase2 TGFb ... · Published OnlineFirst November 21, 2016; DOI: 10.1158/0008-5472.CAN-15-2556 insert that generates siRNAs but contains