Fernández-Vegaetal.BMCCancer2013,13:24 http://www.biomedcentral.com/1471-2407/13/24 RESEARCH ARTICLE Open Access Specific genes involved in synthesis and editing of heparan sulfate proteoglycans show altered expression patterns in breast cancer Iván Fernández-Vega1, Olivia García2, Ainara Crespo3, Sonia Castañón3, Primitiva Menéndez1, Aurora Astudillo1 and Luis M Quirós4* Abstract Background: The expression ofa specific set of genes controls the different structures of heparan sulfate proteoglycans (HSPGs),which are involved in thegrowth, invasion and metastatic properties ofcancerous cells. The purpose of this study is to increase knowledge ofHSPG alterations inbreast cancer. Methods: Twenty-three infiltrating ductal adenocarcinomas (IDCs), both metastatic and non-metastatic were studied. A transcriptomic approach to thestructureof heparan sulfate (HS) chains was used,employing qPCR to analyze both theexpression of theenzymes involved intheir biosynthesis and editing, as well as theproteoglycan core proteins. Since some of these proteoglycans can also carry chondroitin sulfate chains, we extended the study to include thegenes involved in the biosynthesis of these glycosaminoglycans. Histochemical techniques were also used to analyze tissular expression of particular genes showing significant expression differences, of potential interest. Results: No significant change in transcription was detected inapproximately70% of analyzed genes. However,13 demonstrated changes in both tumor types (40% showing more intensederegulation inthe metastatic),while 5 genes showed changes only innon-metastatictumors.Changes were related to 3core proteins:overexpression of syndecan-1 and underexpressionof glypican-3 and perlecan. HS synthesis was affected by lower levels ofsome 3-O-sulfotransferase transcripts,the expression ofNDST4 and, onlyin non metastatic tumors, higherlevels of extracellular sulfatases.Furthermore, theexpression of chondroitin sulfate also was considerably affected, involving both the synthesis of the saccharidic chains and sulfations atall locations. However, the pro-metastatic enzyme heparanase did not exhibit significant changes in mRNAexpression, although inmetastatic tumors it appeared related to increased levels ofthemost stable form of mRNA. Finally,the expression of heparanase 2,which displays anti-metastatic features, experienced a strong deregulationin all patients analyzed. Conclusions: IDCs show alterations inthe expression of HSPG genes; principally theexpression and localization of proteoglycans and thesulfation patternsofglycosaminoglycan chains, depending onthe metastatic nature of the tumor. In addition, theanti-proliferative molecule heparanase 2experiences strong deregulation, thus highlighting it as a potentially interesting diagnostic factor. Keywords: Heparan sulfate, Breast cancer, Proteoglycan, Glycosaminoglycan, Invasive ductal carcinoma *Correspondence:[email protected] 4UniversityInstituteofOncologyofAsturias,andDepartmentofFunctional Biology,UniversityofOviedo,Oviedo33006,Spain Fulllistofauthorinformationisavailableattheendofthearticle ©2013Fernández-Vegaetal.;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsofthe CreativeCommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse, distribution,andreproductioninanymedium,providedtheoriginalworkisproperlycited. Fernández-Vegaetal.BMCCancer2013,13:24 Page2of16 http://www.biomedcentral.com/1471-2407/13/24 Background ligands such as cytokines, chemokines, growth factors, Breast cancer is the most common form of cancer enzymesandenzymeinhibitors,andECMproteins[5,6]. reportedinwomeninallmajorregionsoftheworld,com- Cells exerciseexquisite control over HSPG composition prising22.9%ofallnon-melanomaskincancers,andbeing andsequence,thoughthisvariesbetweencelltypes,devel- responsible for 13.7% of cancerdeaths in women [1]. The opment stages, and also as a result of cell transformation most frequent type of breast cancer is the invasive (or in- in pathological processes. It is therefore of interest to filtrating) ductal carcinoma (IDC), which accounts for analyzeindetailthecompletesetofchangesintheexpres- about 8 out of 10 invasive breast cancers. It starts in a sion of PGs and HS biosynthetic enzymes in cancer path- duct,breaksthroughthewalloftheduct,andinvadesthe ologiesas well as the effectofthese specific signatures on tissue of the breast, from where it is then able to promotinginvasionandmetastasis. metastasizetootherpartsofthebody. In several cancer cells, genes involved in the biosyn- Invasive carcinoma cells characteristically induce thesis of HSPGs are either up- or down- regulated. As changes in the adjacent stroma, and experimental evi- such, the upregulation of two cell-surface PGs, glypican- denceshowsthatthestromainfactactivelycontributesto 1 (GPC1) and syndecan-1 (SDC1), and of the extracellu- carcinoma progression [2]. Breast stroma accounts for lar sulfatase Sulf-2, have been described in malignant morethan 80% of resting breast volume and is composed breast cancer tissues [9,10] whilst the downregulation of of collagen, fibroblasts, endothelial cells, adipocytes and a some genes including SULF1 and HS3ST2 has also been molecular network of proteoglycans (PGs) [3]. The intra- reported [11-13]. However, to date no studies have ana- lobular stroma, unlike the non-specialized interlobular lyzed the entire set of genes involved in the synthesis of stroma, forms a specialized functional unit with an abun- thesemolecules inthispathology. dance of PGs that facilitate hormonally-induced changes Inthispaper,investigationofHSPG’sbiologicalfunction in breast volume and this constitutes the backdrop to the in IDCs was undertaken by analyzing the expression pat- early stage of cancer invasion when the malignant trans- terns of the genes involved in HSPG biosynthesis and formationoftissuetakesplace. comparing them with healthy tissues from the same PGsareadiversegroupofglycoconjugatescomposedof patients. The tumors studied were subdivided into two different core proteins post-translationally modified with groups according to presence or absence of metastases in linear, anionic polysaccharides called glycosaminoglycans lymph nodes since this element is a key predictor of pro- (GAGs)whichconsistofrepeatingdisaccharides.Heparan gression. Thestudy includedgenes coding for HSPG core sulfate proteoglycans (HSPGs) comprise a specific small proteinandforenzymesresponsibleofHSchainsynthesis group of proteins covalently linked to HS GAG and modification. Taking into account that some of these chains. HS is a complex biopolymer initially created PGs can also carry chondroitin sulfate (CS) chains, we as a chain of alternating D-glucuronic acid (GlcA) extended the study to the genes involved in the biosyn- and N-acetyl-D-glucosamine (GlcNAc). At various posi- thesis of this GAG. The aim of the work was to increase tions,themoleculeismodifiedbyaseriesofinterdepend- ourknowledgeofstructuralalterationsofHSPGsinbreast ent enzymatic reactions that include N-deacetylation of cancer, which could be of future benefit in the develop- GlcNAc, usually followed by N-sulfation to produce mentofnewchemicalbiologyapproachestotheretarding GlcNSO , thus creating sulfated S-domains. Within these of tumor progression through the modulation of deregu- 3 regions,GlcAcan be epimerized toiduronate(IdoA), and latedbiosyntheticpathways. O-sulfate groups can be added at C6 of GlcN and C2 of IdoA residues. Minor sulfations at C3 of GlcN and C2 of GlcA may also occur. Chain modification results in clus- Methods tersofflexiblehighlysulfatedIdoA-richregions,separated Materials by more rigid lowly or non-sulfated regions [4,5]. HSPGs The following materials were purchased from the manu- are ubiquitously present intissues,mainly associatedwith facturers indicated: RNeasy Kit and RNase-Free DNase the cell surface and the extracellular matrix (ECM) [4,5] from Qiagen (Hilden, Germany); High-Capacity cDNA and a variety of both normal and pathological functions Reverse Transcription Kit and PowerSYBR Green PCR have been ascribed to them, including cell adhesion and Master Mix from Applied Biosystems (Foster City, CA); 0 migration, organization of the ECM, regulation of prolif- GenElute PCR clean-up kit and 3-3 diaminobenzidine 0 eration, differentiation and morphogenesis, cytoskeleton from Sigma-Aldrich (St. Louis, MO); Biotin 3 End DNA organization, tissue repair, inflammation, vascularization Labeling Kit from Thermo Scientific (Waltham, MA); In andcancermetastasis[4-8],thefunctionultimatelydepend- Situ Hybridization Detection System For Biotinylated ™ ingonthefinestructureofthe chains.Specificsetsof vari- Probes, EnVision G|2 Doublestain System and Envi- ably modified disaccharides, usually within the sulfated sion FLEX target retrieval solution of high pH from domains, define binding sites for a multitude of specific Dako (Glostrup, Denmark);. All other chemicals were Fernández-Vegaetal.BMCCancer2013,13:24 Page3of16 http://www.biomedcentral.com/1471-2407/13/24 obtained from commercial sources and were of analyt- the manufacturer’s instructions. Finally, the aliquots con- icalgrade. taining the cDNA were diluted 1:20 with water and used The following antibodies were used in this study: Goat forqRT-PCRassaysorstoredat−20°Cuntiluse. Anti-heparanase 1 (L-19), rabbit anti-sulf1 (H-81), rabbit anti-perlecan (H-300) and rabbit anti-chondroitin 6- qRT-PCRreactions sulfotransferase-2(Z-24),allofwhichpolyclonalantibodies In all cases, specific oligonucleotides were designed werepurchasedfromSantaCruzBiotechnology,Inc(Santa on different exons or exon junctions, using the pro- Cruz, CA). Rabbit anti-heparanase-2 polyclonal antibody gram Primer3.(http://biotools.umassmed.edu/bioapps/ fromGeneTex(Atlanta,GA),andmousemonoclonalanti- primer3_www.cgi). The size of the amplicon was situ- syndecan1 from DakoCytomation (Carpinteria, CA). Anti- ated in all cases between 70 and 150 base pairs, en- mouse (sc-2020), anti-rabbit (sc-2004) and anti-goat suring wherever possible that the Tm was above 77°C. (sc-2005) secondary antibodies were also from Santa Cruz The theoretical Tm for each amplicon was deter- Biotechnology(SantaCruz,CA). mined using the program Biomath (http://www.pro- mega.com/biomath/calc11.).Primer sequences are Tissuesamples presented in Additional file 1. We analyzed a cohort of 46 snap frozen breast samples, At least four repetitions of all the qRT-PCR reactions obtainedfromtheTumorBankattheInstituteofOncology were carried out in a final volume of 10 μl, according to of Asturias (Asturias, Spain). Twenty three of the samples the manufacturer’s specifications, using 1 μl of the cDNA were from IDCs while the remaining twenty three were dilutionastemplate,with2μlofprimerpairmix(200nM from the corresponding surrounding healthy tissue from final concentration) and 5 μl of SYBR Green mix all thesamepatientsandwereusedascontrol.Diagnoseswere assembled in 96 well microtiter plates. The plates were evaluatedusinghematoxylin-eosin-stainedslidesofallsam- sealed with optical film and centrifuged at 2500 rpm for ples according to the World Health Organization (WHO) 5 min before being placed in a Real-Time ABI Prism criteria and the snap frozen tissues were stored at −80°C DetectionSystemdevice(AppliedBiosystems;FosterCity, priortoisolationoftheRNA.ApplyingtheTNMclassifica- CA) using the following cycling conditions: 95°C for tion, all tumors were at the T2 stage and were classified 10 min, 40 cycles of 95°C for 15 s followed by 60°C for intotwogroupsdependingonthepresence(atleastN1)or 60 s. Following the thermal cycling and data collection absence (N0) of lymph node metastases, which resulted in steps, amplimer products were analyzed using a melt 10 samples being included in the first group and 13 in the curve program (95°C for 1 min, 55°C for 1 min, then in- second. The study was approved by the Ethics Committee creasingby0.5°Cpercyclefor80cyclesof10seach).For on Clinical Investigation of the Hospital Universitario eachamplificationthepresenceofasinglepeakwithaTm CentraldeAsturiasandallpatientsgavetheirconsent. corresponding to that previously calculated was verified. In those cases in which the amplifications were not ad- equate, new primer pairs were designed. Actin was TotalRNAisolationandcDNAsynthesis included on each plate as a control gene to compare run ToobtaintheRNA,fragmentsoftissueofbetween20and variationandtonormalizeindividualgeneexpression. 30 mg in weight were used. Samples were homogenized usingapolytronPT2100(KinematicaInc;Bohemia,NY), and RNAwas isolatedusingtheRNeasy kit, followingthe Dataanalysis manufacturer’s specifications. To ensure removal of re- To calculate the efficiencies of amplification for each sidual contaminating DNA, samples were subjected to gene we used the program LinRegPCR (http://www. treatment with RNase-free DNase during the purification gene-quantification.de/download.html), using the best process itself. The concentration of RNA obtained was correlationcoefficient(consideringaminimumof3points determined spectrophotometrically by measuring absorb- within the window of linearity) and establishing the aver- ance at 260 nm of a 1:50 dilution using a BioPhotometer age of all positive amplifications. At least 4 replicates of (Eppendorf; Hamburg, Germany). The samples were eachreactionswerecarriedout,withthenumberofrepli- divided into aliquots of 10 μl and used for reverse tran- cates being increased in those reactions that showed am- scriptionreactionsorstoredat−20°Cuntilfurtheruse. biguity or dispersion of results. The values of differential cDNAsynthesiswascarriedoutusingtheHighCapacity expression of the genes of interest were expressed as has cDNATranscriptionKitfollowingthemanufacturer’sspe- been described previously [14]. A non parametric cifications.Thereactionswereperformedusingathermo- Wilcoxon test was used for the statistical analysis of the cycler iCycler IQ (BioRad; Hercules, CA), using 2 μg of experiments using a level of significance of p<0.05. All RNA as starting material. The reaction products were analyseswereperformed using the Statisticsfor Windows cleaned using the PCR Clean-Up GenElute kit following program(StatsoftInc;Tulsa,OK). Fernández-Vegaetal.BMCCancer2013,13:24 Page4of16 http://www.biomedcentral.com/1471-2407/13/24 Riboprobepreparation usedasachromogen.Selectedslideswerelightlycounter- Specific sense and antisense riboprobes for NDST4, stainedwithhaematoxylin. SULF2 and HS3ST4 were designed. The riboprobe sequences were: NDST4 sense 50-GCTGCTCCTG Results 0 0 CTCTGCTGTTGCTAGTGCTGCTGTGC3,antisense 5 Analysisofdifferentialgeneexpression GCACAGCAGCACTAGCAACAGCAGAGCAGGAGCA We investigated the differential expression of the genes GxC 30; HS3OST4 sense 50- TGTGGGGAGGGAGGA involved in defined steps of the biosynthesis of HSPGs AGTCAGGGGTTGTGGGATGA30,antisense50TCATCC in IDCs dividing the sample into two depending on the CACAACCCCTGACTTCCTCCCTCCCCACA 30; SULF2 presenceorabsenceofmetastasesinthelymph nodes. 0 sense5 CTCGCGCTCGCCTCCAGCCACACACATTTG 13 samples were obtained from patients lacking metas- CCATT30,antisense50AATGGCAAATGTGTGTGGCTG tases; their mean age was 60±12 years; histological grade 0 GAGGCGAGCGCGAG 3. In all cases, the length of the inallcaseswasmoderate-low;theaveragetumorsizewas probeswasadjustedtobetween34and36nucleotides,the 2,08±0,9 × 1,35±0,6 cm; 90% wereluminal A; 65% were contentofG+Ctobetween48%and62%andTmwasal- locatedintheupperouterquadrantand7.5%showedvas- ways above 73°C. The probes were labeled with Biotin kit cularinvasion. 3 0End DNA Labeling Kit according to the manufacturer’s In addition, 10 samples were obtained from patients specifications. who showed lymph node metastases in 100% of cases, their mean age was 58±14 years; histological grade was Chromogenicinsituhybridization(CISH) moderate-high; the average tumor size was 3,7±0,4 × To perform the hybridizations, tissue sections in paraffin 3,11±1,9 cm; The percentage of luminal A was 60%; were treated with xylene to render them diaphanous, the only 20% were located in the upper, outer quadrant and paraffin later being removed by passing it through de- 80%showed vascularinvasion. creasingalcoholconcentrationsuntilitwaterwasreached. We used qRT-PCR to perform a quantitative analysis ThesampleswerethenincubatedatpH9inbufferDAKO of mRNA expression. In many of the genes in which we K8005 for30 minutes at90°C tofacilitatethe exposureof were able to detect differences between normal tissues cellular ribonucleic acid. Subsequently, the preparations and tumors we complemented the studies determining were washed with sterile tris-buffered saline (TBS), and the expression by histological techniques by including in incubated with labeled probes at a dilution of 1:2.5 in situhybridizations andInmunohistochemistry. sterilewater in a DAKO hybridization ovenfor 5 minutes at 95°C, followed by 15.5 hours at 62°C. Then, the pre- Differentialexpressionofgenesencodingcoreproteins parations werewashed withTBS for 10 minutes, followed carryingHSchains by a second wash for a further 5 minutes. The entire Only 13 genes encode HSPG core proteins. Two gene procedurewascarriedoutusingtheInSituHybridization families, syndecans and glypicans, account for most cell Detection System for Biotinylated Probes according to surface HSPGs. Respectively these families comprise 4 the manufacturer’s specifications. Sections were fixed, (SDC1-4) and 6 (GPC1-6) different proteins. The three mounted and examined with a Leica DMR microscope remaining molecules are arranged in the extracellular (Wetzlar,Germany). Visualization was carried out using a matrix and include perlecan (PRCAN), agrin (AGRN) DFC295Leicacamera. and collagen type XVIII (COL18A1) [15]. Within the syndecans group, no significant differences in the tran- Inmunohistochemistry script levels of species 2, 3 and 4 could be detected Tissue sections were dewaxed as described in the previous (Figure 1A and 1B); however, Syndecan-1 displayed a section.Rehydratedsectionswererinsedinphosphatebuf- more than two fold overexpression in both metastatic feredsaline(PBS)containing1%tween-20.Fordetectionof and non metastatic (p=0.013 and 0.028 respectively) sulfatase-1, heparanase 2, chondroitin 6-sulfotransferase-2 tumors (Figure 1C); overexpression occurred in 60% of and perlecan, sections were heated in high pH Envision the cases ofnon-metastatic IDCs analyzed and in75% of FLEXtarget retrieval solution at 65°C for 20 minand then the metastatic. Changes in Syndecan-1 were also evalu- incubatedfor20minatroomtemperatureinthesameso- ated immunohistochemically using monoclonal anti- lution. For detection of heparanase the final step was SDC1. Healthy tissue analysis showed intensive staining omitted. onthebaso-lateral surface ofepithelialcells ofductsand Endogenous peroxidase activity (3% H O ) and non- acini in duct-lobular units, with local staining in myoe- 2 2 specific binding (33% fetal calf serum) were blocked and pithelial cells (Figure 2A). Furthermore, IDCs displayed thesectionswereincubatedovernightat4°Cwithprimary positive immunoreactivity on the basal side of ducts, as antibodies using a 1:100 dilution. Secondary antibodies well as intense staining of the stroma, regardless of the were used at a 1:100 dilution. 3-3’ diaminobenzidine was nature ofthetumor (Figure 2B and2C). Fernández-Vegaetal.BMCCancer2013,13:24 Page5of16 http://www.biomedcentral.com/1471-2407/13/24 evidenced a strong (approximately 10 fold and 12 fold re- spectively)sub-expression(Figure1C). WhenevaluatingtheextracellularmatrixPGs,nosignifi- cantdifferencesweredetectedforagrinandcollagenXVIII. However, perlecan experienced significant down regulation ofexpressionin70%ofnon-metastatic(p=0.002)and90% ofmetastaticcases(p=0.01).Thevaluesofreducedrelative abundance of mRNA ranged from over 3 times more in non-metastatic IDCs to up to 6 times more in metastatic tumors(Figure1C).Inthesamevein,immunofluorescence staining confirmed that expression of perlecan protein was reduced in tumor tissue compared to in the surrounding healthytissue(Figure2Dand2E). ExpressionofenzymesinvolvedinthebiosynthesisofHS chains HS and CS/DS chains are synthesized by cooperation of multiple biosynthetic enzymes in the Golgi. This study included the genes which code glycosyltransferases (GTs) involvedinHSchainpolymerization,includingEXTL2,re- sponsible for transferring the first GlcNAc residue, and EXT1 and EXT2, which encode copolymerases for chain extension. None of the genes showed changes in their transcriptlevelsinIDCs(Figure3). The result of the activity of all GTs involved in the syn- thesisofHSisanunmodifiedchainmadeofGlcA-GlcNAc repeatingunits.Asthechainpolymerizes,itundergoessev- eralmodifications;theinitialonesinvolveremovalofacetyl groups from GlcNAc residues, followed by sulfation of the amino group which is catalyzed by four different isoforms of N-deacetylase/N-sulfotransferases, NDST1, NDST2, NDST3andNDST4[4,16].Transcriptsofonlytwoofthese isoforms, NDST1 and 2, were able to be quantified in all healthy tissues, while NDST3 and 4 were undetectable in most patients (Figure 3A and 3B). When the analysis was conducted in tumor tissues, no significant differences of Figure1DifferentialtranscriptionofgenesencodingHSPGs. transcript levels of isoforms 1 and 2 were detected (A,B)RelativetranscriptabundanceofmRNAsforHSPGs.Relative (Figure3Aand3B).Inaddition,NDST3didnotshowclear abundanceforhealthytissues(graybars)andtumors(blackbars)are alteration patterns, although its low expression did not plottedonalogscaleforeachgeneassayedandthespreads allow for definitive conclusions to be drawn. Interestingly, representthestandarddeviations.(A)Non-metastaticIDCs.(B) NDST4 transcription increased, and its expression was MetastaticIDCs.(C)Relativeexpressionratioofgenesthatshow statisticallysignificantdifferencesinexpressioninnon-metastatic(●) detectedin50%ofpatients,bothmetastaticandnonmeta- ormetastatic(■)IDCs.ValuesontheYaxisarerepresentedona static (Figure 3). CISH studies confirmed this result, since logarithmicscale. tumors expressing NDST4 showed a positive hybridization oftumoralcellsthatwasabsentinnormalcells(Figure4A). AnalysisoftheexpressionlevelsofallthedifferentGlypi- Further modifications of the HS chain include the epi- cans showed substantial differences between them, much merization of GlcA into IdoA, catalyzed by the action of wider than for the Syndecans, reaching up to nearly 3 the enzyme C5-GlcA epimerase (GLCE), the addition of orders of magnitude, with Glypican-1 being the most sulfate groups at C2 of uronic acid, catalyzed by the abundant species. The qRT-PCR results were unable to enzyme HS 2-O-sulfotransferase (HS2ST1), and the detect significant diferences in the levels of transcripts ex- addition at C6 of glucosamine residues, catalyzed by HS cept for Glypican-3 (GPC3), in which 85% of non- 6-O-sulfotransferase isoforms 1–3 (HS6ST1, HS6ST2 and metastatic (p=0.003) and all metastatic tumors (p=0.005) HS6ST3)[4,16].Noneoftheseenzymesshowedstatistically Fernández-Vegaetal.BMCCancer2013,13:24 Page6of16 http://www.biomedcentral.com/1471-2407/13/24 Figure2InmunohistochemistryofHSPGs.(A-C)Localizationofsyndecan-1usinginmunhistochemistryinnormalandtumoralbreasttissue. (A)Normaltissuesshowingsydecanexpressionexclusivelyintheterminalductallobularunit,magnification200X.(B,C)Non-metastaticand metastatictumors,respectively,showinganincreasedexpressionofsyndecan-1inthedesmoplasticstroma,magnification100X.(D)Controlof syndecanantibodies.IDCinareaofchronicinflammatoryreaction,wheresomeplasmaticcellsarepresent.Theyshowstrongpositivityto syndecan(arrow).(E,F)Confocalmicrocopicvisualizationofperlecan(green)inIDC.ThecytoplasmwasstainedwithTriC-conjugatedphalloidin (red)andnucleicounterstainedwithDAPI(blue).(D)Normalbreasttissue.(E)Metastatictumortissue,magnification600X. significant alterations in transcriptional level (Figure 3A isoform 4 down regulation (p=0.02) increased morethan and3B). 30fold(Figure3C).WecarriedoutCISHstudieswithbio- The last step in the modification of HS chains during tinylated probes for HS3ST4 which confirmed a decrease biosynthesis in the Golgi involves the addition of sulfate in staining in the tumoral cells relative to normal tissue group at C3 of glucosamine. This reaction is catalyzed by (Figure4B). HS 3-O-sulfotransferase isoforms 1–6 (HS3ST1, HS3ST2, ThefinalmodificationoftheHSpatterningiscarriedout HS3ST3A1, HS3ST3B1, HS3ST4, HS3ST5 and HS3ST6) atthecellsurfacebytwocellsurfacesulfatases,SULF1and [17].InnonmetastaticIDCs,isoforms4,5and6exhibited SULF2, which remove GlcN-6S groups from specific anapproximately8,5and10folddownregulationrespect- regions [17]. The transcript levels of both enzymes were ively (p=0.009, 0.01 and 0.01). Meantime, in metastatic overexpressed in non metastatic IDCs (p=0.007 and 0.01 tumors, while isoform 5 was not altered, there was a 10 respectively, Figure 3). Interestingly, no significant fold reduction in expression of isoform 6 (p=0.008), and differences could be detected in metastatic tumors. Fernández-Vegaetal.BMCCancer2013,13:24 Page7of16 http://www.biomedcentral.com/1471-2407/13/24 Figure3DifferentialtranscriptionofgenesencodingenzymesinvolvedinthebiosynthesisofHSrepeatingunit.(A,B)Relativetranscript abundanceofmRNAsforenzymesinvolvedinthesynthesisofHSchains.Relativeabundanceforhealthytissues(graybars)andtumors(black bars)areplottedonalogarithmicscaleforeachgeneassayedandspreadsrepresentthestandarddeviations.(A)Non-metastaticIDCs.(B) MetastaticIDCs.(C)Relativeexpressionratioofgenesthatshowstatisticallysignificantdifferencesinexpressioninnon-metastatic(●)or metastatic(■)IDCs.ValuesontheYaxisarerepresentedonalogarithmicscale.NDST4wasnotdetectedinhealthytissues. Inmunostaining and CISH techniques applied for SULF1 which was downregulated approximately 4 and 7 fold in and SULF2 respectively corroborated this data, showing non-metastatic (p=0.01) and metastatic (p=0.02) IDCs stronger staining in tumor cells relative to healthy ones respectively(Figure5C). (Figure4Cand4D,E). CS chains are modified to a lesser extent than those of HS. Possible reactions include epimerization of GlcA in ExpressionofenzymesinvolvedinthebiosynthesisofCS CS chains, which results in dermatan sulfate (DS) chains, chains catalyzedbyDSE;additionofsulfategroupsatC2ofIdoA We analyzed the transcription levels of genes involved in residue of DS, catalyzed by chondroitin uronosyl sulfo- the polymerization of CS chains, including CSGAL- transferase(UST);sulfationatC4ofGalNAc,catalyzedby NACT2, responsible for transferring the first N-acetyl- different isoenzymes with specificity for CS or DS chains galactosamine (GalNAc) residue, and the chondroitin (CHS11, CHS12, CHS13, CHS14), and addition of sulfate synthases CHSY1, CHPF and CHSY3 [17]. With one ex- at C6 of GalNAc, also catalyzed by different isoenzymes ception,noneofthesegenesshowedchangesintheirtran- (CHS3, CHS7); sulfation at C6 may also occur in pre- script levels in either non-metastatic or metastatic IDCs sulfated residues catalyzed by a N-acetylgalactosamine (Figure 5A and 5B). The exception was CSGALNACT2, 4-sulfate6-O-sulfotransferase(CHS15)[18]. Fernández-Vegaetal.BMCCancer2013,13:24 Page8of16 http://www.biomedcentral.com/1471-2407/13/24 Figure4LocalizationofenzymesinvolvedinthemodificationofHSchains.(A)CISHforNDST4inIDC,negativefortwoductsofnormal appearance(arrows)andpositivefortherestofthetissue,mainlyconsistingoftumorcells,magnification200X.(B)CISHforHS3OST4inIDC.The terminalductallobularunitshowinganintensereactivity(arrow)whilethestainingpatternsarelessintenseintumorcells,magnification100X. (C)CISHforSULF2inIDC.Atthebottomrightunstainednormalbreastductsareseen(arrow).Someadiposetissueisalsopresent.Hybridization ispositivefortherestofthetissue,constitutedbytumorcells,magnification100X.(D,E)InmunohistochemistryforSULF1inIDC.(D)Healthy tissue,terminalductallobularunit,magnification200X(E)Intenseimmunostainingintumorcellsthatgrowintoirregularductswithlights, magnification400X. No changes affecting the epimerization of GlcA could in immunostaining as tumor cell proliferate in relation to be detected, but several reduced relative abundances of healthytissue(Figure6A-C). mRNAinvolvinggenesresponsibleforthesulfationatdif- ferent locations were. C2 of IdoA residues seemed to be Expressionofheparanases undersulfated since UST transcription decreased 4 fold Heparanase (HPSE) is an endo-β-D-glucuronidase that bothinnonmetastatic(p=0.01)andmetastatic(p=0.01) degradesHS,generatingbiologicallyactivefragments.This IDCs(Figure5).SulfationatC4ofGalNAcappearedtobe enzyme, along with the previously mentioned SULF1 and selectivelyaffectedsinceoneofthechondroitin4-O-sulfo- SULF2, constitutes what are known as the editing transferases, CHS12, decreased 3 fold in non metastatic enzymes,responsibleformodificationofHSfinestructure tumors (p=0.02) and around 10 fold (p=0.009) in meta- in physiological and pathological processes. We analyzed static IDCs (Figure 5). However, what did appear to be its transcription in relation to the presence or absence of affected in a generalized form was the generation of metastasis in IDCs. The structure of the HPSE gene is GalNAc(6S) residues given that the transcription of the represented in Figure 7; it spans over 50 kb and consists two genes coding the enzymes that catalyze this reaction, of14exons[19].PerformingqRT-PCRreactions,wefailed CHST3 and CHST7, decreased in both non metastatic to detect significant changes in HPSE transcription either (p=0.003 and 0.003 respectively) and metastatic in metastatic or non metastatic IDCs (Figure 8), thereby (p=0.009 and 0.01) IDCs (Figure 5). Inmunohistochem- apparently disagreeing with previous reports indicating icalstudieswereperformedforCHST3usingspecificanti- overexpression of this enzyme, especially in metastatic bodies; these analyses allowed us to visualize the decrease tumors [20]. Our initial experiments were carried out Fernández-Vegaetal.BMCCancer2013,13:24 Page9of16 http://www.biomedcentral.com/1471-2407/13/24 indicated that the isoform HPSE1a was expressed in both healthyandtumortissues,whileHPSE1bremainedatvery loworundetectablelevels(Figure8Aand8B). Recent research has identified the existence of a 185-bp 0 sequence within the 3 untranslated region that mediates HPSEdown-regulation[21].Wedesignedprobestodetect 0 the presence of this alteration in the 3 end (Figure 7B). The results indicated that both forms could be detected, althoughtheshorter specieswasthemostabundant in all cases(morethantwofold,Figure8Aand8B).Comparison of expression levels of each of the transcripts did not evi- dence any significant differences between tumors and healthy tissue. However, when the relative expression levels of both forms (short form / long form expression ratio) were subjected to statistical analysis, there was in- deed no significant difference in non-metastatic tumors, but there was in fact a change in metastatic relative to healthy tissues (p=0.02),suggesting a higher relative pro- portion of the short isoform in this group of IDCs (Figure 8C). Alterations in the expression of HPSE were alsodeterminedintissuearraysbyimmunohistochemistry and demonstrated the existence of varying levels of pro- teinoverexpressionindifferentpatients(Figure9A-D). Heparanase 2 (HPSE2) is a homologue of HPSE that lacks HS-degrading activity, although it is still able to interact with HS with high affinity [20]. Using a pair of probes designed to anneal to common regions in all iso- forms predicted for this gene, we were able to detect a significant transcription alteration which appeared down-regulated approximately 30 fold (p=0.01 and 0.007) in all types of IDCs (Figure 8A and 8B). It has been previously described that wild-type heparanase 2 (HPSE2c) exhibits very high affinity for HS, and is able to compete with HPSE. To check the expression levels Figure5Differentialtranscriptionofgenesencodingenzymes of this isoform, we performed qRT-PCR reactions using involvedinthebiosynthesisofCSrepeatingunit.(A,B)Relative probes against exons 3 (absent in isoform HPSE2b) and transcriptabundanceofmRNAsforenzymesinvolvedinthe 4 (absent in isoforms HPSE2b and HPSE2a). The results modificationofCSchains.Relativeabundanceforhealthytissues were again of note, indicating as they did a 30 fold de- (graybars)andtumors(blackbars)areplottedonalogarathmic scaleforeachgeneassayedandspreadsrepresentthestandard crease in the expression of (p=0.01) in IDCs (Figure 8A deviations.(A)Non-metastaticIDCs.(B)MetastaticIDCs.(C)Relative and 8B). Immunohistochemical studies confirmed this expressionratioofgenesthatshowstatisticallysignificant result, with tissues from all patients analyzed showing a differencesinexpressioninnon-metastatic(●)andmetastatic(■) decrease in immunoreactivity for antibodies specific for IDCs.ValuesontheYaxisarerepresentedonalogarithmicscale. HPSE2(Figure 9E and9F). usingprimersplacedinexons5and6respectively;wealso Discussion performednewreactionsusingprimerslocatedinexons9 The expression of HSPGs is markedly altered during ma- and10,withsimilarresults(Figure8). lignant transformation and tumor progression, affecting Possible alterations in the noncoding regions of mRNA both the PG core proteins and the GAG chains [13]. The that could influence the expression levels were also ana- HS fine structure can be determined by cell-type specific lyzed. Two mRNA species containing the same ORF, expression of only certain isoforms of some biosynthetic HPSE1a, and HPSE1b, generated by alternative splicing, enzymes (Figure 10), notwithstanding the existence in have been described [19]. Both species differ in the struc- somespecificcasesofregulationattranslationleveloren- 0 tureattheir5 end,whichhelpstodifferentiatethemusing zymaticcatalysis[21-24].Inthispaperweinvestigatedthe appropriate primers (Figure 7A). qRT-PCR reactions expression patterns of the genes involved in HSPG Fernández-Vegaetal.BMCCancer2013,13:24 Page10of16 http://www.biomedcentral.com/1471-2407/13/24 Figure6Inmunolocalizationofchondroitin6-O- sulfotransferase1.(A)Typicalterminalductallobularunitofthe breastdisplayinganintensecytoplasmicimmunoreactivityfor CHST3.Perilobulillarinflammatorycellsarenotmarked.(B)Breast ductwithIDCshowingcytoplasmicimmunoreactivityonlyincellsof thebasallayer.(C)Groupofepithelialcellsthatproliferate haphazardly,infiltratingtheadjacentadiposetissue.ItisaIDCin whichthereislittleimmunoreactivityforCHST3,magnification400X. biosynthesis inIDCscomparedtothose of healthy tissues from the same patients. The tumors were subdivided into two groups according to presence or absence of metasta- sesinthelymphnodes,giventhisistheprincipalindicator ofcancerprogression. In human cells, there are 13 genes encoding “full- time” HSPGs, although a few more may appear as “part- time” ones [4,5]. Only SDC1 from the syndecan group appeared overexpressed more than two fold in both metastatic and non metastatic tumors. Upregulation of SDC1 has been previously described in breast cancer, al- though with values exceeding 10 times those of the nor- mal tissue, as determined by immunohistochemical staining quantification [25]. Our analysis of tumor and healthy tissue sections using monoclonal anti-SDC1 dis- played an intense Immunoreactivity of the tumoral stroma regardless of the nature of the tumor, although this was more intense in metastatic IDCs, and lesser stainingofthebasementmembraneofducts. This result represents a change in the location of the expression of SDC1 compared to healthy tissue, where immunoreactivity was displayed mostly on the baso- lateral surface of epithelial cells of ducts and acini in duct-lobular units. The shift of SDC1 from epithelial to stromal cells during progression of breast tumors has been previously reported, and suggested to stimulate carcinoma growth and angiogenesis [26,27]. Moreover, the existence of differences between transcript levels quantified by qRT-PCR and immunostaining found in the present study could be an indicator of additional post-transcriptional regulation of SDC1 expression, which has been described for certain cell types [24]. Upregulation of SDC1 has also been described in other tumors such as pancreatic, lung and brain cancer, and it has been postulated that this aberrant expression may play a key role in promoting growth factor signaling in cancer cells [13]. Interestingly, SDC1 is downregulated in various malignances, such as colorectal cancer, indi- cating that this HSPG may serve as a prognostic marker inacancer-type-specificmanner[13]. No significant differences in the levels of transcripts of isoforms 2, 3 and 4 were detected in this study, although overexpression of SDC4 has been previously described for an estrogen receptor-negative highly proliferative breast carcinoma subtype [28]; however, most samples