Versican overexpression in human breast cancer lesions: Known and new isoforms for stromal tumor targeting


  • Philippe Kischel,

    Corresponding author
    1. Department of Biology, Metastasis Research Laboratory, GIGA Cancer, University of Liege, Liège, Belgium
    • Metastasis Research Laboratory, Pathology Building, Bat. B23 (+4), University Hospital of Liege, 4000 Liege, Belgium
    Search for more papers by this author
    • Philippe Kischel and David Waltregny contributed equally to this work.

    • Fax: +3243662975

  • David Waltregny,

    1. Department of Biology, Metastasis Research Laboratory, GIGA Cancer, University of Liege, Liège, Belgium
    Search for more papers by this author
    • Philippe Kischel and David Waltregny contributed equally to this work.

  • Bruno Dumont,

    1. Department of Biology, Metastasis Research Laboratory, GIGA Cancer, University of Liege, Liège, Belgium
    Search for more papers by this author
  • Andrei Turtoi,

    1. Department of Biology, Metastasis Research Laboratory, GIGA Cancer, University of Liege, Liège, Belgium
    2. Laboratory of Mass Spectrometry, Department of Chemistry, University of Liege, Liège, Belgium
    Search for more papers by this author
  • Yannick Greffe,

    1. Department of Biology, Metastasis Research Laboratory, GIGA Cancer, University of Liege, Liège, Belgium
    Search for more papers by this author
  • Stephanie Kirsch,

    1. Laboratory of Mass Spectrometry, Department of Chemistry, University of Liege, Liège, Belgium
    Search for more papers by this author
  • Edwin De Pauw,

    1. Laboratory of Mass Spectrometry, Department of Chemistry, University of Liege, Liège, Belgium
    Search for more papers by this author
  • Vincent Castronovo

    1. Department of Biology, Metastasis Research Laboratory, GIGA Cancer, University of Liege, Liège, Belgium
    Search for more papers by this author


Proteoglycans play a key role in cancer development and progression by participating in the constitution of a specific fertile tumor microenvironment. As they are largely overexpressed in the malignant stroma, proteoglycans provide a reservoir of potential new targets for anticancer therapies, because they can serve to convey toxic payloads in the close proximity of cancer cells and subsequently destroy them. In this context, versican, a proteoglycan largely overexpressed in several solid cancers, bears the potential to be such an ideal target. As 4 main versican isoforms have been characterized, we sought to determine which isoform could represent the best target in human breast cancer. We used a series of 10 primary breast cancer lesions that were characterized as overexpressing the versican protein, when compared with matched normal breast tissues, using shotgun mass spectrometry and immunohistochemistry experiments. Quantitative polymerase chain reaction and western-blotting experiments were used to evaluate versican isoform expression in breast cancer/normal tissue pairs for which ARN quality was excellent. All known isoforms were significantly overexpressed in the malignant lesions, both at the mRNA and at the protein levels. In the course of this study, we also identified and cloned a new alternatively spliced versican isoform, referred to as V4, which was also found to be upregulated in human breast cancer. This study provides for the first time a comprehensive mRNA and protein analysis of versican isoforms expression in human breast tissues, and offers insights into which therapeutic strategy would be best suited to target versican in human breast cancer lesions.

The concept that stroma may play an active role in cancer progression is now unanimously recognized. The stroma of most human malignancies is fundamentally different from the stroma of the corresponding normal tissue, and is suspected to promote cancer progression.1, 2 Among hallmarks of stroma remodeling are fibroblast proliferation along with myofibroblastic differentiation, accumulation of connective tissue (desmoplasia), and overexpression of many proteoglycans. These latter proteins contain covalently linked glycosaminoglycans (GAG), and have been implicated in several biological processes involved in tumor progression, including cell adhesion and proliferation.3 Notably, chondroitin sulfate proteoglycans (CSPGs), components of both cell membranes and extracellular matrix, were demonstrated to be upregulated in various neoplastic tissues.4–6 Stromal therapy has emerged as a new strategy2, 7, 8 and is based on ligands binding with high affinity and specificity to target molecules that are found to be overexpressed at sites of disease. A molecule that is upregulated in cancer, accessible to anticancer treatments, and absent in most normal tissues would represent an ideal target for suitable high-affinity ligands, such as systemically delivered monoclonal antibodies.8 In this context, proteoglycans could represent highly promising targets. Using a new chemical proteomic approach, we recently identified the CSPG2 (aka versican) protein as a specific and accessible biomarker of human breast cancers, constituting therefore a potential new target for antibody-based anticancer therapies.9 Versican is a large “versatile proteoglycan,”10 a member of the lectican/hyalectin family composed of large aggregating proteoglycans (see Ref. 11 for review). In normal tissues, versican is highly expressed in the early stages of development,12 and remains expressed in the adult in a few tissue locations (e.g., central nervous system and placenta9). CSPG2 encodes, within 15 exons, the large (3396 amino acids) versican V0 protein.13 The structure of the versican protein consists of a GAG attachment domain of variable size, surrounded by 2 external globular domains, G1 and G3. The aminoterminal G1 domain is composed of Ig repeats and a hyaluronan binding domain, whereas the carboxyterminal selectin-like G3 domain is composed of EGF, lectin and complement regulatory protein elements.11 The alternate splicing of 2 exons, namely GAGα (encoded by exon 7) and GAGβ (encoded by exon 8), generates 3 variants, V1,10 V214 and V3.15 While the full length V0 isoform contains both GAGα and GAGβ exons, V1 and V2 isoforms lack GAGα and GAGβ, respectively. V3 lacks both GAG [chondroitin sulfate (CS)] attachment regions,15 resulting thereby in a side chain polymorphism in the different versican variants. Although the expression of versican isoforms in normal human tissues have been investigated at the mRNA level by reverse transcription-polymerase chain reaction (RT-PCR),16 information about protein expression levels of the different isoforms in different tissues and diseases has remained scarce, especially for the lower molecular weight V3 isoform. V0 and V1 are usually documented as the main isoforms found in tissues reported to express versican, whereas V2 is the main and restricted isoform in the mature brain.14, 17, 18

As we and others have reported a very strong versican overexpression in human breast cancer tissues following immunohistochemical analyses,9, 19 a detailed study of the versican expression was mandatory to determine which versican isoform was overexpressed, and help choosing which one could be best suited for targeted therapies. To this end, we have evaluated the mRNA and protein expression levels of both high- and low-molecular weight versican entities using quantitative RT-PCR and 1D-PAGE, respectively, in breast tumors and matched adjacent normal counterparts. Along with these information, we herein provide strong evidence for the overexpression of a new alternatively spliced versican isoform, referred to as “V4,” in human breast cancer. We found that this new isoform is secreted and upregulated in human breast fibroblasts in response to TGFβ1. We demonstrate an upregulation of all versican isoforms in breast tissues, and show the need to target versican domains that are not isoform specific.

Experimental Procedures

Tissue harvesting

Matched cancerous and nontumoral human breast tissue samples were obtained from 15 mastectomy specimens. Samples were either flash-frozen in liquid nitrogen typically <30 min after the excision in the operating theatre, or directly immersed in formalin and then processed for further histological and histochemical investigations. The Ethics Committee of the University Hospital of Liege reviewed and approved the specific protocol used in this study, and written informed consent was obtained from all patients. Six of the 10 breast cancer lesions were shown to overexpress versican using shotgun mass spectrometry and immunohistochemistry.9

PCR and sequencing

Original primers used for PCR were previously described.20 Primers used for cloning are reported in Supporting Information Table 1a (the position of these primers is depicted in Supporting Information Figure 1), and primers and probes used for quantitative PCR are reported in Supporting Information Table 1b. PCR runs were performed with the following standard cycles: 30″ 94°C, 1′ 60°C, 2′ 72°C, 35× (68°C for extension when using the Accuprime Pfx proofreading enzyme). Whenever required, amplicons generated with proofreading Taq were gel purified using the QiaQuick gel extraction kit (Qiagen, Valencia, CA), and cloned using the Topo-TA Cloning kit (Invitrogen, Merelbeke, Belgium). Sequencing of the purified cDNAs was performed on a 3100 Applied Biosystems sequencing unit.

Quantitative real time reverse transcription-polymerase chain reaction (qRT-PCR)

Total RNA was extracted using the TriZol reagent (Invitrogen), according to the manufacturer's instructions. RNA purity was assessed both with A260/A280 ratios and quality of 18S and 28S in agarose gels. Samples with insufficient RNA quality were discarded, and 10 primary breast cancer lesions and matched normal tissues were kept for subsequent analyses. For cDNA synthesis, 2 μg of total RNA were reverse transcribed in a 20 μl reaction mixture, as previously described.21 The RT reaction was performed at 42°C for 50′ before a 15′ inactivation step at 70°C. Negative controls included omission of the enzyme.

Quantitative real-time RT-PCR was performed in triplicate using the ABI Prism 7300 Sequence Detection System (Applied Biosystems, IJssel, The Netherlands), according to the manufacturer's instructions. The sequences of primers and probes for the different versican isoforms were designed using the Primer Express software (Applied Biosystems). All probes included modifications in 5′ (FAM) and modifications in 3′ (BHQ-1). TaqMan primers and probes for the human glyceraldehyde-3-phosphate dehydrogenase (GAPDH), β-actin, 18S RNA and cyclophilin A were purchased from Applied Biosystems. All sets of primers and probes were selected to work under similar cycling conditions. cDNA samples (from 100 ng or 50 ng total RNA from tissues or cells, respectively) were mixed with 100 nM of each primer, 100 nM probe and TaqMan PCR Universal Master Mix (Applied Biosystems), in a total reaction volume of 25 μl. Real-time PCR was performed for each target cDNA (V0–V4), and transcript expression levels were normalized to those of GAPDH for qRT-PCR on cultured cells. All samples were run as triplicates. Acquired data were analyzed using the Sequence Detector software (Applied Biosystems).

Western blotting

Proteoglycans extraction from breast tissues was adapted from Deepa et al.22 Proteins were digested at 37°C overnight with 20 mU chondroitinase ABC (Sigma, St. Louis, MO) per 100 μg proteins. For cell extracts, SDS was removed from SDS lysates by precipitation (TCA, followed by acetone washes), and the pellet was resuspended in chondroitinase ABC buffer. Digestion with chondroitinase ABC (0.2 U/100 μg proteins) was allowed at 37°C for 3 hr, and SDS was added again for western blotting. The digests were subjected to SDS-PAGE in 4–12% (NuPAGE, Invitrogen) gels under nonreducing conditions. Separated proteins were electrotransferred to nitrocellulose membranes at 160 mA constant current overnight. Membranes were blocked in TBS containing 5% nonfat dried milk. Bound antibodies were visualized using ECL chemiluminescent substrate (GE Healthcare, Diegem, Belgium) and exposure to X-ray films (Fuji, Dusseldorf, Germany).


The 12c5 antiversican antibody, developed by R. Asher, was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa (Department of Biological Sciences, Iowa City, IA). This antibody was raised against the aminoterminal domain of versican (glial hyaluronate-binding region of versican, aka “GHAP”). The monoclonal antibody 2B1 (Seikagaku), which recognizes an epitope at the C-terminal G3 domain,23 was purchased from (AmsBio, Oxfordshire, UK). A polyclonal goat antiversican, recognizing the first ∼210 amino acids of the GAGβ (exon 8, amino acids 1344–1554 of the V0 accession # P13611) was purchased from (R&D Systems, Oxon, UK).

Mass spectrometry: Multiple reaction monitoring analyses

We used a method based on tandem mass spectrometry, which represents an alternative and additional value to the classical Western blot. This method commonly known as multiple reaction monitoring (MRM) represents the tool of choice for quantitative measurements of small organic molecules (antibiotics, steroids, pesticides, etc.) in complex matrices. In the past, several comprehensive studies24–27 have already proven the applicability of this method to peptides, offering the possibility to detect and quantify proteins present in physiological concentrations in complex samples (pmol/ml). Here, a similar approach was applied to detect the specific isoform of versican using a characteristic tryptic peptide sequence. The screening was applied to both tumor and normal breast tissue samples as well as to 7 different cell lines. Relative quantification of the new versican isoform was conducted with respect to a reference synthetic peptide that was spiked in known concentration to each of the samples examined. Reference synthetic peptide with isotopically labeled arginine (13C and 15N), corresponding to the specific tryptic peptide fragment characterizing the novel isoform of versican, was obtained from Thermo Electron GmbH (Ulm, Germany). The reference peptide bearing the sequence EEEGTTGPDR* displayed a shift of + 10 m/z with respect to natural peptide contained in the sample.

MRM ion selection was conducted on the basis of daughter ion scan of the directly injected 1.4 μg/ml solution of the synthetic peptide in the MS (Waters Micromass Quattro Ultima Platinum). Collision energies were optimized for each of the daughter ions. Four specific daughter ions (185.2, 241.2, 555.4 and 713.5 m/z) were selected to build a fingerprint that will be used to identify the target peptide. For this purpose, from y ion series 185.2, 555.4 and 713.5, a total of 10 m/z was subtracted (difference synthetic to natural peptide) whereas from b ion 241.2 was left unchanged as it was not carrying the “heavy” arginine. The selection of these specific fragments was done based on their relative high abundance, good ionization and fragmentation properties. Only samples showing all 4 transitions (mother to daughter ions) were counted as positive and the quantification was conducted on their respective total ion chromatograms with respect to those of the synthetic peptide (Supporting Information Figure 4 and Table 3). The synthetic peptide was added to all the samples analyzed (total concentration per sample = 9.6 pg/μL).

Total protein digest of the samples was performed using trypsine. The HPLC system used was Waters Alliance 2690. Approximately 4.5 μg of the digested sample was injected on 2.1 mm × 150 mm C-18 column [(Polaris C-18 A media with 3 μm beads, 200 Å pore size (Varian, Palo Alto, CA)], separated at a constant flow of 400 μl/min. Linear gradient was used: t = 0 min, 100% water (+0.1% v/v acetic acid) and t = 20 min, 60% water and 40% acetonitrile (+0.1% v/v acetic acid). The eluent was split 50% before introduction into the ESI ionization chamber. The MS instrument (Waters Micromass Quattro Ultima Platinum) was set in the MRM mode, capillary voltage was 2.6 kV, desolvatation temperature set at 115 and 250°C, cone desolvatation gas at 50 and 630 l/hr and the collision cell pressure at 2 μbar. The data were accrued and evaluated using the MassLynx (v4.0) software.

Cloning of the versican V4 isoform

Using RT-PCR, total RNA from human breast tumor and the Accuprime Pfx DNA polymerase (Invitrogen), we attempted to amplify directly the versican V4 isoform with the forward primer “1” and with the reverse primer “2” (see Supporting Information Table 1 and Figure 1b). As the forward and reverse primers are not isoform specific, and because V4 is nearly 3.2 kb, V3 was preferentially amplified over V4. Therefore, we generated V4 by PCR as 3 smaller fragments, namely fragments A, B and C, using reverse transcription products of RNA from breast cancer tissue. The primers pairs used were (“1” and “6”) for fragment A amplification; (“5” and “4”) for fragment B, (“7” and “2”) for fragment C. Fragments A and B were combined and amplified using primers “1” and “4” to generate fragment AB. Fragments B and C were combined and amplified using primers “5” and “2” to generate fragment BC. Finally, fragments AB and BC were combined and amplified using primers “1” and “2” (the use of primer “2bis” allowed the removal of the original TGA stop codon and ensured the presence of the carboxyterminal V5 epitope and hexahistidine sequence) to generate the full length V4. The versican V4 coding sequence was inserted in frame into the pcDNA3.1D/V5-His-Topo (Invitrogen). The V4 construct was checked by DNA sequencing. The same cloning procedure applied for V3, generated directly with primers “1” and “2.”

Cell lines, reagents and treatments

NIH-3T3 cells were cultured in DMEM medium (Invitrogen) supplemented with 10% fetal calf serum (AEScientific, Marcq, Belgium) and 2 mM glutamine (Invitrogen) at 37°C in a humidified 5% CO2 incubator. Mammary fibroblasts were isolated from normal human breast tissues using a previously described technique.28 Briefly, fresh tissue was obtained from a patient undergoing reduction mammoplasty, with informed consent. Confluent monolayer cultures were obtained after 10 days, and the cell population was characterized phenotypically and immunocytochemically as >95% fibroblasts. TGFβ1 (Roche Diagnostics, Mannheim Germany) was used at a final concentration of 2.5 ng/ml in PBS.


Murine NIH-3T3 or human U87-MG cells grown in 6-wells plates were transfected at 90% confluency with 4 μg of pcDNA3.1D/V5-His plasmid, containing either V4 or V3, and lipofectamine 2000 transfection reagent (Invitrogen), according to the manufacturer's protocol. Mock transfection with transfecting reagent alone and transfection with empty pcDNA3.1D/V5-His plasmid (Invitrogen) served as controls. Cells were cultured for 24 h [for RNA extraction, using RNeasy kit (Quiagen)] or 96 h (for protein extraction, lysed in 1% SDS buffer). Detection of proteins was performed either by the 12c5 antibody against the aminoterminal portion of versican (GHAP), or against the V5-tag for appropriate constructs.

Lactate dehydrogenase release assay

To assess any potential cytotoxicity associated with plasmid transfection and thus, potential release of intracellular V4 or V3 recombinant proteins into the conditioned medium, lactate dehydrogenase (LDH) activity in culture supernatants (triplicate per each condition) was determined using the fluorescent-base CytoTox-One homogenous membrane integrity assay (Promega, Leiden, The Netherlands), according to the manufacturer's instructions.


mRNA analysis of versican isoforms

Quantitative SybrGreen-based real-time RT-PCR experiments were initially performed using previously described primers20 to determine which versican isoforms are upregulated in breast cancer. The 4 primers pairs were first tested by classic RT-PCR, to ensure their respective specificity. Each primer pair gave the expected amplicon (∼100 bp), though an unexpected band at ∼1200 bp showed up with primers designed to amplify V3 (Fig. 1b). These forward and reverse primers, located at the 3′ end of exon 6 and at the 5′ portion of exon 9, respectively, did not overlap exonic junctions (Fig. 1a). Theoretically, as these primers were not isoform specific, they were potentially able to generate (i) a 71 bp amplicon without GAG exons corresponding to V3, as well as (ii) amplicons comprising only GAGα (3032 bp, V2), comprising only GAGβ (5333 bp, V1) or comprising both GAG exons (8294 bp, V0), provided cycling conditions allowed the polymerase to generate such longer amplicons. To the best of our knowledge, the only report of an additional band in the literature using PCR primers designed to scan across exons 6 and 9 was made by Lemire et al.29 As the ∼1200 bp band did not correspond to an amplicon related to any known versican isoform, we performed error-free PCR amplification, cloning and sequencing. It appeared that the primers were able to amplify not only the end of exon 6 and the beginning of exon 9, as predicted, but also the first 1194 bp of exon 8 (Fig. 2). These results suggested the existence of a new versican isoform, hereafter referred to as “V4.” At this stage, the most probable hypothesis was that this alternate splicing, due to a cryptic splice acceptor site inside exon 8, produced only the insertion of a segment of exon 8 between intact G1 and G3 segments, as represented in Figure 2. Assuming the alternate splicing actually adds only 1194 bp between the G1 and G3 domains, we expected the size of V4 to be 3162 bp. PCR with external primers encompassing the ATG start codon (primer “1”) and the TGA stop codon (primer “2,” Supporting Information Figure 1) generated, in breast tumor tissues, a band at ∼2000 bp (consistent with the size of V3), and only a faint, inconsistent band above 3000 bp (data not shown). To ensure that V4 was not truncated, we sought to amplify specific fragments of the V4 isoform. For this purpose, the reverse and forward primers “8” and “9” (Supporting Information Figure 1b) were designed to overlap the fragment of exon 8 present in V4 and the exon 9 of the G3 domain, the only V4 specific overlapping region. These primers “8” and “9” are both over 30 nucleotides long, but display only a 10 bp overlap at their respective 3′ ends (i.e., 10 nt overlap with exon 8 for primer “8” and 10 nt overlap with exon “9” for primer 9), ensuring therefore the highest specificity for V4 detection. The use of the generic forward primer “1” and the V4 specific reverse primer “8” generated a ∼2300 bp fragment consistent with V4 amplification (2257 bp) in breast tumors (Supporting Information Figure 1c, Lane 2), as well as in breast fibroblasts (Supporting Information Figure 1c, Lane 3). These 2 PCR products suggested that V4 has a classical G1 domain. The other V4 specific forward primer “9” and the nonspecific reverse primer “2” gave the expected ∼1000 bp amplicon PCR product in breast tumors (Supporting Information Figure 1c, Lane 8), as well as in breast fibroblasts (Supporting Information Figure 1c, Lane 9). This result is consistent with V4 having a classical G3 domain (expected size of 936 bp). These 2 latter primers were not able to amplify V3 inserted in a pcDNA3.1 plasmid, confirming specific V4 amplification with the PCR conditions used (data not shown). Together, these results confirmed the structure of the V4 transcript depicted in Figure 2.

Figure 1.

Schematic overview of the 4 known mRNA splice variants. (a) The known versican splice variants (V0, V1, V2 and V3) result from the alternate splicing of exons 7 and 8. The various splicing variants contain both (V0), 1 (V1 and V2) or neither of these 2 exons (V3). The sizes of the mRNA segments are indicated in the color-coded boxes (values from ATG start codon to the TGA stop codon). Arrows indicate the position of the forward and reverse primers used to generate a V3 amplicon with a size compatible with qRT-PCR. (b) Representative RT-PCR detection of the V3 isoform transcript in normal breast tissues (Nx) and matched breast tumor lesions (Tx). Two different bands are detected in the tumor samples: the amplicon at the expected size (73 bp), designated as “V3,” and an additional band at ∼1250 bp, designated as “?.” This additional band is very faint in N2 and N4. [Color figure can be viewed in the online issue, which is available at]

Figure 2.

Genomic structure of the versican gene and scheme of the new versican V4 isoform. The 15 exons of the versican gene are depicted as filled rectangles linked by introns depicted by the black lines. Red rectangles show 5′UTR (before ATG start codon) and 3′UTR (after TGA stop codon). Orange and blue rectangles represent GAGα (exon 7) and GAGβ (exon 8), respectively. The green rectangle inserted into GAGβ represents the portion of exon 8 found in the V4 isoform. Red boxed electropherogram displays 18 nucleotides surrounding the junction between exons 6 and 8 (identical to the junction found in V1). Blue boxed electropherogram displays 18 nucleotides surrounding the junction between the truncated exons 8 and 9. The cryptic splice donor site inside exon 8 corresponds to the following underlined sequence: GAACTACAGGTACGGCTT). [Color figure can be viewed in the online issue, which is available at]

To quantify each of these 5 isoforms (V0, V1, V2, V3 and the new V4 isoform), a series of new primers and probes was designed (Supporting Information Table 1b). Specific exon overlapping probes were easily found for V0, V2, V3 and V4. Based on the structure of the isoforms, no specific probe could be designed for V1. We therefore designed 2 probes: a first probe overlapping exons 6 and 8 and able to amplify also V4, and a second probe overlapping exons 8 and 9 and able to amplify also V0. The quantitation of the V1 isoform was therefore made by subtracting either V4 or V0 RNA quantities from the results obtained with these 2 probes. From now on and for the sake of clarity, the only V1 probe we will refer to is the probe overlapping exons 6 and 8, also able to amplify V4. We nevertheless always checked for data consistency between the 2 sets of primers and probes.

Quantitative real-time RT-PCR was performed for all versican isoforms. Results were generated from 10 normal-tumor paired samples for which we had good quality RNA (A260/A280 ratios > 1.8, and sharp 18S and 28S bands, as assessed by agarose gel electrophoresis). Figure 3a shows the absolute cycle threshold (Ct) values for the 10 breast cancer samples analyzed in this study and their corresponding normal breast tissues (see Supporting Information Table 2 for the clinico-pathological characteristics of the breast carcinomas analyzed in the study). For each isoform, Ct values were always lower in tumor samples (the lower the Ct value, the higher the gene expression level). Considering all isoforms, versican transcripts were detected 8.01 ± 1.3 cycles (mean ± standard error to the mean) earlier in tumors than in their normal counterparts. We sought to normalize the expression levels of the different versican isoforms using various “so-called” house-keeping genes, including 18S, β-actin, GAPDH and cyclophilin A. However, despite the addition of strictly similar amounts of total RNA in the RT experiments, we have noted a high variability in the expression of house-keeping genes between matched normal and tumor samples for the same patient, as well as between samples from different patients. This variability has been previously demonstrated in breast cancer for GAPDH,30 as well as for 9 other “reference” genes and 18S RNA.31 For this reason, we decided to show only absolute Ct levels for all versican isoforms expression in breast tumor samples, without other normalization than starting cDNA quantities, as suggested elsewhere.31 Figure 3a unambiguously shows that “absolute” versican transcript levels for all isoforms, including the new V4 isoform, were higher in tumor samples, demonstrating the overexpression of all versican isoforms in breast lesions.

Figure 3.

Quantitation of the different versican isoforms by quantitative real-time RT-PCR. (a) Ct values obtained for the 5 different versican isoforms, determined in normal breast and in adjacent tumor breast tissues. Total RNA from malignant and matched adjacent normal breast tissues from 5 different patients was assayed in qRT-PCR with the primers and probes described in Supporting Information Table 1b. Replicates were averaged, and values from normal and malignant breast tissues are represented by open triangles (▵) and black filled squares (▪), respectively. Averaged values from the 10 patients are represented by empty and black filled bars for normal and tumor samples, respectively. (b) Expression of the normalized versican V4 transcript levels in cultured cells treated with TGFβ: breast fibroblasts (isolated from normal mammary tissues), human smooth muscle cells (HSMCs) and endothelial cells (HUVECs). Absolute Ct values were 26.24 ± 0.05, 25.23 ± 0.05 for fibroblasts (without and with TGFβ treatment, respectively). Absolute Ct values were 27.40 ± 0.05, 26.26 ± 0.06 for HSMCs (without and with TGFβ treatment, respectively). Absolute Ct values were 29.38 ± 0.06, 29.77 ± 0.11 for HUVECs (without and with TGFβ treatment, respectively). Cells were treated as described in Material and Methods section. Each TGFβ treated condition was normalized to its own control (the corresponding cell line without TGFβ).

qRT-PCR analyses were also conducted on different cell lines, to assess which cell type could be responsible for V4 synthesis. We examined the basal V4 mRNA expression in human breast cancer cells, breast fibroblasts, endothelial cells (HUVECs) and smooth muscle cells (HSMCs). qRT-PCR experiments were conducted on each cell line using GAPDH as reference gene. In terms of expression, V4 was barely detectable in MDA-MB231 breast cancer cells, whereas primary breast fibroblasts (isolated from normal mammary tissues) were found to be the most potent V4-expressing cells (mean Ct ± SE: 26.25 ± 0.08), with expression levels twice and 8 times higher than those in HSMCs (Ct ± SE: 27.4 ± 0.087) and HUVECs (Ct ± SE: 29.37 ± 0.097), respectively (Fig. 3b). As it is known that the versican gene is TGFβ responsive, cells were treated with 2.5 ng/ml TGFβ (Roche). After a 24 h treatment, a 4-fold and a 3-fold increase of V4 mRNA expression was observed in primary breast fibroblasts and in HSMCs, respectively, whereas HUVECs did not respond to the treatment (Fig. 3b). Similar results were obtained with V1 specific primers and probes (data not shown).

Analyses of versican isoforms expression at the protein level

We next sought to determine the expression of the various versican isoforms at the protein level in normal and adjacent tumoral breast tissues. We therefore compared samples digested or not with chondroitinase ABC. The 12c5 antibody was used to reveal the versican isoforms and the degradation products bearing the aminoterminal hyaluronate-binding region of versican (see also Table 1 for a detailed list of potential bands recognized by this antibody) expressed in normal and tumoral breast tissues. Figure 4a (Lanes 1–3) shows a western blot of a representative breast tumor and a matched normal tissue sample, in which major bands are numbered from 1 to 9. Chondroitinase efficiency was confirmed by the appearance of discrete high-molecular weight proteins in normal and tumor samples. Adult rat brain is known to express large amounts of V2 isoform.22 Band #3 would therefore correspond to V2 (Fig. 4a, Lane 4), and bands #2 and #1 would logically correspond to V1 and V0, as per their respective increasing molecular weight. V0, V1 and V2 can be detected in normal breast (Fig. 4a, Lane 5), and are always largely overexpressed in the matched breast cancer tissues. Band #4 is just above 150 kDa. Several studies have reported on the presence of such a band around 160 kDa, which most certainly corresponds to the aminoterminal degradation product of V0.32 However, this band is also recognized by the 2B1 antibody (Lane 2a).

Table 1. Designation, predicted molecular weight and actual size in 1D-PAGE of isoforms or G1-containing proteolytic fragments recognized by the 12C5 antibody
inline image
Figure 4.

Versican expression in human tissues and transfected cells in culture. Western blots revealed with the 2B1 antiversican antibody (a), the 12c5 antiversican antibody (a to d) and the anti-V5 tag antibody (d). (a) 4–12% PAGE of a representative normal and matched breast tumor sample. Lanes 1, 2 and 3 were loaded with 15 μg of normal, matched tumor digested with chondroitinase ABC, and associated tumor without chondroitinase ABC digestion, respectively (see “Results” section for the description of each of the 9 bands). (1a, 2a, 3a) and (1b, 2b, 3b) represent the same samples probed with 2 different antibodies: the 2B1 (a series) and 12c5 (b series). Lane 4: rat brain. Lane 5 corresponds to normal breast digested with chondroitinase ABC. Bands 3 and 8 correspond to V2 and GHAP, respectively. (b) 4–12% PAGE showing the migration level of recombinant V4 (Lane 2), tagged V4 (Lane 3), and V3 (Lane 4, initially Lane 6 from the same blot) produced in murine NIH-3T3 cells versus the migration level of versican entities in a representative breast tumor sample digested with chondroitinase ABC (Lane 1). (c) 4–12% PAGE showing the migration level of recombinant V4 (Lane 2), tagged V4 (Lane 3) produced in human U87MG cells versus the migration level of versican entities in a representative breast tumor sample digested with chondroitinase ABC (Lane 1). Lane 4 represents U87MG cells transfected with a tagged V4 and probed with the anti-V5 antibody. NB: Normal breast, TB: tumor breast, ChABC: with chondroitinase ABC treatment.

The predicted translation product of the full V4 sequence is a 118,151 Da protein. Cleavage of a signal sequence is predicted after alanine 20 and leucine 2110 (checked also with SignalP 3.0 webtool33), and would result in a core protein of 115,737 Da. The presence in V4 of a quarter of exon 8, bearing SG or GS sequences consensus for CS chain addition, could potentially modify the predicted molecular weight. If detectable in western blotting, the V4 isoform should anyway correspond to either band #5 or #4. After cloning and transfection, exogenous V3 and V4 were expressed in murine NIH-3T3 fibroblasts. The transcript expression of these isoforms was checked by qRT-PCR (data not shown), and whole cell lysates were analyzed by western blotting (Fig. 4b). Recombinant V4 and V3 migrated closely to bands #5 and #7, respectively (Lanes 2 and 4, Fig. 4b). The V4 tag-expressing clone (Lane 3) migrated a little bit slower than the “naked” V4, and was recognized by an anti-V5 antibody (data not shown). V4 could therefore correspond to band #5. Importantly, this band was recognized by the polyclonal antiversican supposed to recognize only V0, V1 and V4 (and any of their degradation products bearing the beginning of exon 8, Supporting Information Figure 2). As shown in Figure 4b, a smear appeared above the sharp V4 band (as well as above the tagged-V4 band), and most probably represents the processed form of V4 (with GAG added), because whole cell lysates were not digested with chondroitinase ABC. This was confirmed by the actual chondroitinase ABC digestion of these smears (Supporting Information Figure 5). Expectedly, no smear appeared in the whole cell lysate from 3T3 cells overexpressing recombinant V3, which does not possess any GAG attachment sites. We next expressed V4 in human cells (U87-MG), to ascertain that the processing of human V4 in NIH-3T3 cells of murine origin was not aberrant. We found that the core protein was indeed migrating faster, and localized as expected at the level of band #5 (Fig. 4c). More importantly, the processed forms were much more abundant in U87-MG, yet the pattern was quite similar between the murine and human cell lines (see also Supporting Information Figure 5).

Band #8 most likely represents the glial hyaluronate binding protein (GHAP). Indeed, it has been shown that human versican V2 can be cleaved by proteases at Glu405-Gln406 to generate GHAP.34 As only V2 is expressed in adult rat brain, the sole degradation product appearing in this Lane 4 (Fig. 4a) should come from V2 and thus represents GHAP.

Band #9 was observed in both normal and tumoral breast tissues. Moreover, under reducing conditions, 12c5 did not recognize any versican isoform but still recognized this band #9, strongly suggesting that this band can be considered as aspecific staining. The size (in the 55 kDa range) and the aspect of this latter band suggest that it represents human IgGs from breast tissues.

Among these 9 most frequently observed protein entities showing reactivity with the 12c5 antibody in breast cancer samples, only one of them remained difficult to identify, namely band #6, which migrated around 100 kDa.

V4 detection using MRM

We used MRM to detect a tryptic peptide that is isoform specific (sequence EEEGTTGPDR). The screening was applied to both tumor and normal breast tissue samples as well as to different cell lines, including the U87-MG glioma cell line, skin and breast fibroblasts, and MDA-MB-231 and MCF-7 breast cancer cell lines. Relative quantification of the new versican isoform was conducted with respect to a reference “heavy” synthetic peptide that was spiked in known concentration to each of the samples examined. Results are shown in Supporting Information Figures 4a and 4b, and Table 3. The specific peptide was detected in almost all samples (yet at various concentrations) except in normal breast tissues. These results demonstrate the actual presence of V4 in tumor breast tissues.

Is V4 actually secreted?

To ascertain the secretion of the new V4 isoform, we analyzed the conditioned medium of human U87-MG cells transfected with the V4 expression plasmids (with or without V5/His tags). Controls included lipofectamine only and empty plasmid. V4 was found to be secreted in the conditioned medium without concentration as well as after concentration onto a 100 kDa molecular weight cut-off centrifugal device. As depicted in Supporting Information Figure 3a, only the processed forms of V4 were detected in the conditioned medium, as opposed to total cell extracts in which both naked and processed V4 were found. The V4 found in the conditioned medium was most likely secreted, but not released by dead cells, as assessed by the LDH release assay (Supporting Information Figure 3b).


Up to now, numerous studies have shown that ECM proteoglycans are upregulated and accumulated in the stroma associated to cancerous lesions.35, 36 The CSPG versican is overexpressed in several cancer types, including prostate,37 brain,17 ovary38 and breast35, 39 cancers as well as melanoma.40 We have recently demonstrated that versican is overexpressed in breast cancer when compared both with adjacent normal breast tissue and with other normal tissues and organs within the human body.9 In this latter study, the 12c5 antibody was used. As this antibody recognize the G1 domain and thus all isoforms, it was not possible to determine which isoform(s) was (were) overexpressed (antibodies that are specific for each isoform are not currently available). In an attempt to determine which versican isoform(s) is (are) actually the best suited for anticancer targeted therapies in human breast cancers tissues, we have discovered a new versican variant. Sequence analysis of this new variant revealed that the first 1194 bp of exon 8 (GAGβ) are sandwiched between the end of exon 6 and the beginning of exon 9. This new variant, generated by alternate splicing involving the classical GT-AG rule for splicing junctions,41 has been termed “V4.” Alternative splicing seems to be a process that rather frequently involves genes coding for extracellular matrix proteins.42 We have herein shown that alternate splicing can generate, at least in human breast tissues, the transcript of the new V4 isoform.

Following mRNA analyses of the different versican isoforms, we demonstrate, in cancer lesions, the upregulation of all the previously known versican isoforms, as well as the upregulation of the new V4 isoform that is otherwise barely detectable in breast normal tissues. We sought to determine from which cells the new V4 could be produced, because versican expression was reported in fibroblasts,43 stimulated endothelial cells,16 smooth muscle cells29 and only marginally by cancer cells.44 We found that the most potent versican V4-expressing cells were fibroblasts. These latter cells were treated with either TGFβ1 or conditioned medium from MDA-MB 231 breast cancer cells, and V4 transcript abundance was found to be increased in both treatment conditions.

As mRNA expression does not necessarily translate into protein expression,45, 46 we have further sought to unveil versican isoform expression levels by western blotting. Identification of versican entities under 200 kDa remains difficult, even with the use of different available antibodies, because (i) very few data are available in the literature, (ii) degradation products from V0 and V1,32 as well as from V234 potentially hamper the identification of lower molecular weight entities in western blotting experiments (see Table 1) and (iii) potential, not yet identified splice variants may be expressed in a tissue and/or stage-specific manner. Recombinant expression of “low” molecular weight versican isoforms V3 and V4 helped us to identify the different isoforms.

As shown by the western blots, it is obvious that the global transcript overexpression of versican isoforms in the malignant breast lesions translates into protein overexpression, because multiple bands showed up in the lesions when compared with adjacent normal breast tissues. The migration pattern of the 3 high-molecular weight isoforms, namely V0, V1 and V2, has already been described, and rat brain lysate has been used as positive control for V2.22 Consistent with data from the literature, these 3 isoforms migrated far above their predicted molecular weight, even after chondroitinase ABC treatment14, 22 and reasons for this have already been proposed.47 The migration level of the 2 lower molecular weight isoforms (V3 and V4) were deduced by comparing the migration levels of recombinant proteins. The slightly higher migration levels of both recombinant proteins can be explained by the origin of cells (murine or human, normal fibroblasts or glioma cells) or the altered glycosylation patterns in cancer development.48 Of note, the migration of V3 can slightly differ simply after 2 different treatments, even on the same cell line.16

Our results demonstrate the overexpression of all versican isoforms, including the new one, V4, at the protein level. Based on these results, the best strategy for targeting versican would be to focus on regions that are not isoform specific. However, to become a potent target, a versican isoform has to be at least secreted in the extracellular matrix. Thus, to validate V4 in targeting approaches, we wondered whether V4 can be secreted or not. It has been demonstrated that GAG attachment appears to be important for the secretion process.49, 50 V4 features a portion of GAGβ that contains several serine-glycine consensus sequences for GAG attachment,51 yet the actual GAG attachment is not obvious because versican's GAG profile is dependent on tissue type location and versican variants.47 We found that the V4 band is chondroitinase ABC sensitive, even if some variations occurred between samples, ranging from moderate to black and white. The assumption that V4 is actually a proteoglycan is also supported by the smears only above recombinant V4 (and not V3) in NIH-3T3 cells. The human cells were able to process the V4 protein very efficiently when compared with the murine cells. We also demonstrated that the V4 protein was actually secreted by both human and murine cells by a direct approach. This secretion process warrants further investigations on the role of this new isoform in the tumor development and progression. Finally, overexpression of the V4 isoform in the extracellular matrix by breast fibroblasts in response to TGFβ (released by breast cancer cells52) might contribute to the tumoral progression. The specific role of V4 during breast cancer progression deserves further investigations.

To summarize, we herein report on a comprehensive analysis of the 4 already known versican isoforms, as well as of a new splice variant termed “V4,” bearing GAG attachment sites. Tumor ECM components are generally more stable than tumor-associated cell surface antigens.53 ECM molecules are also abundant and theoretically easily accessible: this is the case for versican, not only in breast cancer,9 but also in kidney cancer54 and colon cancer.55 In this context, overexpressed versican very likely represents a suitable target for antibody-based anticancer therapies. ECM antigens like versican can indeed serve to bring toxic payloads to tumor cells and subsequently destroy them. This strategy has proven successful, for example, with the extra domain B of fibronectin56 and tenascin C.57 In view of our results, it is tempting to propose a targeting of versican domains that would not be isoform specific. Further work is needed to validate this promising target by in vivo animal models.


P.K. is a Research Fellow and D.W. is a Senior Research Associate of the Belgian National Fund for Scientific Research.