Versican Upregulation in SÉZary Cells Alters

ORIGINAL ARTICLE Versican upregulation in Sézary cells alters growth, motility and resistance to chemotherapy

K Fujii1,3, MB Karpova1,4, K Asagoe1,5, O Georgiev2, R Dummer1 and M Urosevic-Maiwald1

Sézary syndrome (SéS) represents a leukemic variant of cutaneous T-cell lymphoma, whose etiology is still unknown. To identify dyregulated genes in SéS, we performed transcriptional profiling of Sézary cells (SCs) obtained from peripheral blood of patients with SéS. We identified versican as the highest upregulated gene in SCs. VCAN is an extracellular matrix proteoglycan, which is known to interfere with different cellular processes in cancer. Versican isoform V1 was the most commonly upregulated isoform in SCs. Using a lentiviral plasmid, we overexpressed versican V1 isoform in lymphoid cell lines, which altered their growth behavior by promoting formation of smaller cell clusters and by increasing their migratory capacity towards stromal cell-derived factor 1, thus promoting skin homing. Versican V1 overexpression exerted an inhibitory effect on cell proliferation, partially by promoting activation-induced cell death. Furthermore, V1 overexpression in lymphoid cell lines increased their sensitivity to doxorubicin and gemcitabine. In conclusion, we confirm versican as one of the dysregulated genes in SéS and describe its effects on the biology of SCs. Although versican overexpression confers lymphoid cells with increased migratory capacity, it also makes them more sensitive to activation-induced cell death and some chemotherapeutics, which could be exploited further for therapeutic purposes.

Leukemia (2015) 29, 2024–2032; doi:10.1038/leu.2015.103

INTRODUCTION In this study, we identified versican as one of the highest Sézary syndrome (SéS), a leukemic variant of cutaneous T-cell upregulated genes in SCs using high-throughput gene expression lymphoma (CTCL), is characterized by erythroderma, generalized profiling. Overexpression of versican V1 exhibited influence on lymphadenopathy and the presence of clonal T lymphocytes in growth and migration behavior, cell proliferation and suscept- the skin, lymph nodes and peripheral blood.1 Tumor cells of SéS, ibility to chemotherapy of lymphoid cells, providing first evidence called Sézary cells (SCs), harbor spectrum of genetic abnormalities on versican functionality in a leukemic T-cell lymphoma. as well as alterations in apoptotic pathways, lymphocyte differentiation and cytokine signaling.2 Various studies have MATERIALS AND METHODS attempted to clarify the exact pathogenic events in SéS; however, Biologic samples and cell lines the exact mechanisms underlying accumulation of malignant SCs Peripheral blood mononuclear cells (PBMCs) and skin biopsy samples from in the skin and other target organs are still poorly understood. five patients with SéS were a surplus material obtained during routine One molecule described in context of SéS is versican (VCAN, diagnostic procedures, for which the patients have given their written also known as PG-M and CSPG2), a member of the large consent and obtained approval of the institutional ethical committee. aggregating chondroitin sulfate (CS) proteoglycan family repre- PBMCs were obtained by Ficoll-density gradient centrifugation (Sigma- senting one of the major components of extracellular matrix Aldrich, St Louis, MO, USA). All patients fulfilled diagnostic criteria for 13 (ECM).2–4 Versican is expressed in a variety of soft tissues, SéS. Using magnetic labeling system on Macs cell sorter (Miltenyi Biotec, 5 α Bergisch Gladbach, Germany), SCs were separated by using CD4 − and including skin. Alternative splicing of the CS domain into CS fi β and CSβ generates four versican isoforms, V0, V1, V2 and V3. second by using patient clone-speci c T-cell receptor (TCR) V -chain antibody.14,15 In this way, highly purified CD4+Vβ+ SC population Overexpression of V1 in vitro results in enhanced cell proliferation was generated. The residual, non-malignant CD4+Vβ − cell population and protection from apoptosis.6 V1-overexpression can simulta- 7 was employed as an internal control in addition to CD4+ cells from healthy neously cause apoptotic resistance and sensitivity. On the other donors. hand, V2 and V3 exhibit opposite biological activities by inhibiting Three CTCL cell lines, MyLa (derived from a plaque biopsy of a patient cell proliferation (V2, V3) as well as by increasing the apoptotic with mycosis fungoides), SeAx and Hut 78 (both derived from peripheral rate (V3). Unbalanced expression of versican isoforms is described blood of a patient with SéS) were used. All cell lines were kindly 16 in pathological conditions, such as increased expression of V0 and provided by Dr Keld Kaltoft, Aahus, Denmark. Cell lines were grown in V1 in cancer.8 Moreover, existing literature suggests that versican supplemented RPMI 1640 medium (Invitrogen, Carlsbad, CA, USA). has the ability to regulate migratory behavior of inflammatory cells, including lymphocytes.8–10 Versican is therefore a recognized Gene expression profiling and data analysis cell fate and motility modulator that may facilitate tumor cell Total RNA, double-stranded cDNA and labeled complementary RNA were invasion and metastasis.11,12 generated as previously described.17 After purification, complementary

1Department of Dermatology, University Hospital Zurich, Zurich, Switzerland and 2Institute of Molecular Life Sciences, University of Zurich, Zurich, Switzerland. Correspondence: Professor M Urosevic-Maiwald, Department of Dermatology, University Hospital Zurich, Gloriastrasse 31, Zurich 8091, Switzerland. E-mail: [email protected] 3Current address: Department of Dermatology, Kagoshima University Graduate School of Medical and Dental Sciences, Kagoshima, Japan. 4Current address: Roche Pharmaceutical Research and Early Development, Roche Innovation Center Penzberg, Penzberg, Germany. 5Current address: Division of Dermatology, National Hospital Organization Okayama Medical Center, Okayama, Japan. Received 14 January 2015; revised 13 March 2015; accepted 7 April 2015; accepted article preview online 27 April 2015; advance online publication, 15 May 2015 Versican upregulation in Sézary cells K Fujii et al 2025 RNA was hybridized to HG-U95Av2 GeneChip arrays (Affymetrix, Santa HotStart system in the LightCycler thermocycler (Roche Diagnostics). Clara, CA, USA). Chips were hybridized, washed and stained according to Versican V0–V3 primers were designed by us (Supplementary Table S1) protocols provided by Affymetrix. and synthetized by Microsynth (Balgach, Switzerland). Primers for versican The scaling, data normalization and polishing were performed according V0/1 and housekeeping genes were purchased from Search LC (Heidel- to the protocol previously described by us.17 Data analysis was carried berg, Germany). Platelet-derived growth factor (PDGF) 10 ng/ml (Gibco Life out with the GeneSpring Expression Analysis software (Silicon Genetics, Technologies, Carlsbad, CA, USA) was used for stimulation of VCAN Redwood City, CA, USA). Briefly, expression profiles of CD4+Vβ+ cells and expression during 48 h. CD4+Vβ − cells were compared using a non-parametric, Mann–Whitney test with Benjamini–Hochberg multiple-testing correction (Po0.005), fold- change cutoff ⩾ 1 and a detection call filter ⩾ 1 (at least one of the samples Western blotting in each group had to be called 'present'). Mining of functional gene Cells lysed with either TRIzol or RIPA lysis buffer were used for protein associations was performed using Cytoscape (http://www.cytoscape.org)18 analysis. Proteins were separated by sodium dodecyl sulfate- and GeneMANIA plug-in.19 Clustering analysis and display was performed polyacrylamide gel electrophoresis using the NuPAGE SDS-PAGE Gel with Gene Cluster and Tree View program, respectively (http://rana.lbl.gov). System (Invitrogen) on 10% Tris-glycine gels (Invitrogen) under reducing conditions according to the manufacturer’s instructions and transferred onto nitrocellulose membranes (Invitrogen). Versican V0/V1 antibody from Generation of versican V1 vector Dianova (ABR-27039, Hamburg, Germany) was diluted at 1:2000. In order to overexpress versican V1 in CTCL cell lines, we generated a lentiviral vector containing versican V1 (Supplementary Figure S1). Human fibroblasts (WH87) were stimulated by transforming growth factor (TGF)-β Assessment of cluster formation in V1-transfected CTCL cell lines to induce versican V1 expression. Following RNA extraction, mRNA Naive, control- and V1-transfected CTCL cell lines were washed and diluted fragments of V1 isoform were amplified, extracted and ligated, producing to 2 × 105/ml. After 24 h incubation, clusters of cells were evaluated by a coding sequence of 8224 base pairs, which was then cloned into phase-contrast microscopy at × 100 magnification and scanned. Diameters lentiviral pHIV-1 vector with blasticidin-resistance gene.20 To differentiate (pixel) of five clusters were measured using Adobe Photoshop CS5 (Adobe between endogenous and transduced versican V1, the vector was labeled Systems, San José, CA, USA). with VSV-G-tag. The same vector with green fluorescent protein instead of V1 was used as mock control (mRNA and protein expression data are Migration assay shown in Supplementary Figure S2). Cell migration was assayed with a modified Boyden chamber assay system. Transwells (BD, Franklin Lakes, NJ, USA) with polycarbonate membranes Real-time quantitative PCR and cytokine treatment containing 8-μm pores were used. CTCL cell lines were washed and diluted RNA was isolated using the TRIzol Reagent (Invitrogen) and reverse at a concentration of 2 × 105/ml in the migration medium. After incubation transcribed using the First Strand cDNA Synthesis Kit (Roche Diagnostics, at 37 °C, ca. 2 × 105 of suspended cells were added to the top chamber. Rotkreuz, Switzerland). PCR amplifications were carried out with the For each bottom chamber, migration medium ± 100 ng/ml of stromal cell-

Co-expression Co-localisation TCL1A CD24 Genetic interactions HD1_CD4 HD2_CD4 P2_CD4+Vbeta- P4_CD4+Vbeta- P1_CD4+Vbeta- P3_CD4+Vbeta- P5_CD4+Vbeta- P4_skin P1_skin P5_skin P2_skin P4_CD4+Vbeta+ P1_CD4+Vbeta+ P2_CD4+Vbeta+ P5_CD4+Vbeta+ P3_CD4+Vbeta+ HSPA2 E2F5 Pathway Physical interactions PRKAR2B LCN2 Predicted CCR1 ELANE TCL1A Shared protein domains CD19 DNASE1L3 NFE2 HSPA2 PPBP TSPAN3 CTSG AQP9 SERPINB2 VSIG4 CEBPB STEAP1 CD163 CLIC4 AKAP8L CAPG CXCL2 VCAN SERPINA1 VSIG4 TGFBI TNFAIP2 TBC1D9 PRKCD SERPINB2 SERPINA1 PLAU PLCG2 TGFBI EGR2 CD163 PIGA ARHGEF2 DHX29 PTGS1 WDR77 TNFAIP2 SEPHS1 EPB41L3 C9orf97 STX16 RPH3A FLNA CLIC4 STEAP1 VCAN RAB13 VCAN CD44 AQP9 Figure 1. Versican-associated gene networks in SéS. (a) Network of VCAN neighbors obtained by GeneMANIA plug-in in Cytoscape. Genes are identified by their gene symbols. Upregulated genes are shown in shades of red; downregulated genes in shades of green color. The legend of the connection type is displayed in the upper right corner. (b) Two-way clustering analysis using 145 differentially expressed probes from sorted CD4+Vβ+ (red) and CD4+Vβ − (blue) cells. Skin samples (yellow) and CD4 T-cells from healthy donors (gray) were also included. The similarity of gene expression profiles between examined samples is summarized in a dendrogram above the heatmap, in which the pattern and the length of the branches reflect the relatedness of the samples. Red color in the heatmap stands for high relative level of expression; green color stands for low relative level of expression. Genes are represented by their gene symbols.

© 2015 Macmillan Publishers Limited Leukemia (2015) 2024 – 2032 Versican upregulation in Sézary cells K Fujii et al 2026 derived factor 1α (SDF-1α, Gibco Life Technologies) ± 1 μg/ml AMD3100 (plerixafor, Sigma-Aldrich) was added. Hyaluronidase was used at 5 U/ml 80 CD4+Vb+ (Sigma-Aldrich). After 4 h of incubation at 37 °C, cells in the upper and 60 CD4+Vb- bottom chamber were counted with FACScanto (BD). Migration index was 40 calculated according to the formula: migration index (%) = cell number in 20 bottom, experimental chamber/cell number in the upper, control chamber. 5 All experiments were carried out in at least triplicates.

Cell proliferation assay, cell cycle and apoptosis assessment 1 0.05 Cells were seeded in 96-well plates at 2 × 104/well and treated with

phorbol 12-myristate 13-acetate (PMA, Sigma-Aldrich) plus ionomycin versican/GAPDH ratio mRNA 0 (Sigma-Aldrich), doxorubicin (LC Laboratories, Woburn, MA, USA, metho- Pt1 Pt4 Pt2 Pt3 Pt5 SeAx MyLa

trexate (Orion Pharma, Espoo, Finland) or gemcitabine (Actavis, Dublin, NJ, Hut78 USA) for 24 h. The cell survival fractions were determined by the MTT (3- CD4 HD1 CD4 HD2 [4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazoliumbromide) dye reduction assay. Plates were incubated in the presence of MTT dye to enable the Patients CD4 bulks & cell lines formation of formazan crystals. After dissolution of the crystals, the absorbance of samples was measured using a microplate reader at 570 nm Pt1 Pt2 Pt3 Pt4 Pt5 (Biorad, Hercules, CA, USA). Results are reported as a percentage of the value for the untreated control. V0/V1 For cell cycle analysis, cells were washed and diluted at 6 a concentration of 1 × 10 /ml with or without PMA plus ionomycin. After Loading 24 h stimulation, cells were fixated with cold 70% methanol and stained control with propidium iodide (Life Technologies). DNA content was assessed by FACScanto (BD) and analyzed using FlowJo (Tree Star Inc., San Carlos, CA, USA). For spontaneous and activation-induced apoptosis analysis, cell CD4+Vb- CD4+Vb- CD4+Vb- CD4+Vb- CD4+Vb- CD4+Vb+ suspensions were stained with annexin V (BD) according to the CD4+Vb+ CD4+Vb+ CD4+Vb+ CD4+Vb+ manufacturer's instructions and assessed by FACScanto. All samples were tested in at least triplicate. 80 V1 Immunohistochemistry V2 60 V3 Paraffin-embedded skin biopsies and acetone-fixated cytospins were used. Following antibodies were used: (1) anti-versican antibody (Boster, 40 Pleasanton, CA, USA), diluted at 1:200 and stained according to peroxidase protocol; (2) anti-p53(DO7) antibody (Cell Marque, Rocklin, CA, USA), prediluted and stained according to alkaline-phosphatase performed as 20 previously described.17 15

versican/GAPDH ratio mRNA 0 Statistical analysis CD4+Vb- CD4+Vb+ Statistical analysis of the data other than from microarray was performed * using the GraphPad Prism software (GraphPad Software, La Jolla, CA, USA). Employed statistical tests are described together with used experiments Patients (see below). P values o0.05 were considered statistically significant. Figure 2. Versican expression in SCs. (a) VCAN mRNA expression per patient determined by quantitative PCR in CD4+Vβ+ and RESULTS CD4+Vβ − cells from peripheral blood. Bars represent glyceralde- Gene expression profiling of SCs reveals versican upregulation hyde 3-phosphate dehydrogenase (GAPDH)-normalized expression of VCAN. (b) Versican and β-actin protein expression detected by We obtained CD4+Vβ+ (corresponding to SCs) and CD4+Vβ − cell western blotting in CD4+Vβ+ and CD4+Vβ − cells. (c) GAPDH- populations from peripheral blood of five patients with SéS and normalized expression of versican isoforms V1–V3 in CD4+Vβ+ and subjected them to gene expression profiling. Using our filtering CD4+Vβ − cells. Bars represent mean ± s.e.m. of three independent criteria, we identified 145 probes (139 genes) that were experiments. Significant P-values o0.05 are highlighted with one differentially expressed in CD4+Vβ+ cells. Of these, 56 probes asterisk (*). (54 genes) were upregulated and 89 probes (85 genes) were downregulated. In the group of upregulated genes, we observed a prominent upregulation of versican (VCAN, former CSPG2) In the next step, we included four skin samples from five (Supplementary Table S2). These differentially expressed genes analyzed Sézary patients and CD4 T-cells from healthy donors in were subjected to pathway analysis using GeneMANIA plug-in in the clustering analysis using above described differentially Cytoscape. In our data set, genes involved in the control of cell β cycle and intracellular organization had the highest rating (q-value expressed probes. CD4+V + cells and skin samples from the o0.0098) (Supplementary Figure S3). VCAN was included in the same Sézary patient clustered together and opposite to β − glycosaminoglycan-binding group. We then looked into genes CD4+V and CD4 T-cells obtained from healthy donors β that share any type of connection to VCAN by assessing its (Figure 1b). This implies the relatedness of the CD4+V + cells neighbors in GeneMANIA network (Figure 1a). Some of these found in the blood and in the skin of the Sézary patients. Skin 10 genes have already been associated with neoplastic lymphoid samples, normally harboring high levels of versican, showed cells, such as SERPINA1 (α1-AT or αPI, a major plasma serine lower VCAN expression than CD4+Vβ+ cells (darker red color proteinase inhibitor),21 TGFBI (Betaig-h3, a tumor-suppressor intensity in Figure 1b), suggesting true upregulation of VCAN in gene),22,23 TNFAIP2 (a tumor necrosis factor (TNF)-α inducible SCs. Likewise, CD4+Vβ − cells appear to be more related to CD4 protein),24 TCL1A (an oncogene)25 or induction of Th2-type T-cells from healthy donors as they showed decreased expression immunity (SERPINB2).26 of genes pertinent to SCs (green color in Figure 1b).

Leukemia (2015) 2024 – 2032 © 2015 Macmillan Publishers Limited Versican upregulation in Sézary cells K Fujii et al 2027 The induction of VCAN mRNA in CD4+Vβ+ cells could be further VSV-G-V1 (Figure 4). In both MyLa and SeAx cell lines, cluster confirmed by quantitative PCR. The level of VCAN gene expression forming was significantly reduced in comparison to mock- varied considerably between analyzed patients (Figure 2a). transfected cell line (t-test Po0.0001; Figures 4a and b). Because We additionally observed substantially lower VCAN mRNA apoptosis can be the cause for the decrease in cluster size, we expression in purified CD4+ T-cells of healthy donors and very assessed levels of spontaneous apoptosis by annexin V-surface low VCAN expression in CTCL cell lines (SeAx cell lines being expression, which did not significantly differ in V1- and mock- completely negative). Versican protein expression detected by transfected cells (t-test P40.05). western blotting appeared to be more prominent in CD4+Vβ+ cell population (Figure 2b). Knowing that alternative splicing can Versican V1 overexpression modulates migratory properties of generate different versican isoforms, we looked into the expres- lymphoid cells sion of versican isoforms in SCs from our patient set. Versican We then looked into the migratory properties of mock- and isoforms V1, V2 and V3 showed significantly higher expression in V1-transfected cell lines, as versican is implicated to modulate CD4+Vβ+ cells (analysis of variance (ANOVA) P = 0.0006), with migration of inflammatory cells.9,27 We first assessed spontaneous versican V1 being the highest expressed isoform in CD4+Vβ+ cells and SDF-1-directed migration of parent cell lines MyLa and SeAx, (Figure 2c). before and after stimulation with PDGF. PDGF28 as well as TGF-β,29 Immunohistochemistry revealed versican expression in the lithium-chloride30 and TNF-α,31 were described to be inductors of lymphocytic infiltrate in the biopsies of analyzed Sézary patients, VCAN expression in different cell types. In CTCL cell lines, only PDGF in addition to known epidermis and connective tissue positivity could produce relevant VCAN mRNA induction (data not shown), (Figures 3a and b). In the control, skin samples with atopic eczema which was the rationale to use it in the following experiments. As and psoriasis inflammatory cells lacked versican positivity with the migration of SeAx cell line did not significantly change, only overall less versican expression in epidermis and dermal tissue results obtained with MyLa are discussed. (Figures 3c and d). Treatment with PDGF increased spontaneous (t-test P = 0.036) as well as migration of MyLa toward SDF-1 (t-test P = 0.009), Versican V1 overexpression alters growth behavior of lymphoid cells a chemokine responsible for skin homing of lymphocytes As there are no known lymphoid cell lines with relevant versican (Figure 5a). Addition of AMD3100 (plerixafor), a C-X-C chemokine expression, we created a lentiviral vector pHIV-1-Blasti-VSV-G-V1 receptor 4 (CXCR4) inhibitor which abrogates SDF-1 effects, led to and used it to overexpress versican V1 in our CTCL cell lines. decrease in migratory activity. Nevertheless, MyLa cells treated Whereas Hut 78 and lymphocytes from healthy donors were with PDGF showed higher percentage of cells migrating to SDF-1, notoriously resistant to transfection, MyLa and SeAx cell lines despite inhibition with AMD3100 (t-test P = 0.002), implying that could be successfully transfected with pHIV-1-Blasti-VSV-G-V1 and PDGF-mediated upregulation of versican may be one factor were used in subsequent experiments. Upon transfection with facilitating migration of MyLa cells to SDF-1. Analogous findings pHIV-1-Blasti-VSV-G-V1, we observed changes in behavior of MyLa could be obtained with MyLa cells overexpressing V1 (Figure 5b). and SeAx cell lines. When growing in suspension, CTCL cell lines MyLa-V1 cells showed higher spontaneous migratory activity than tend to form cell aggregates or clusters. This capacity to build mock-transfected cells (t-test P = 0.0006). Likewise, migration clusters tends to be reduced upon transfection with pHIV-1-Blasti- of MyLa-V1 to SDF-1 was also enhanced (P = 0.0006), even

Figure 3. Versican expression in the skin of Sézary patients. (a and b) Immunohistochemistry of versican expression in a skin biopsies from Sézary patients (peroxidase protocol, original magnification × 10). Note intensive expression of versican in the dermis and dermal-infiltrating cells. Higher magnification zooms (×20) to demonstrate versican-expressing lymphocytic infiltrating cells are provided in the lower left corner of panels (a and b). Control skin samples from psoriasis (c) and atopic eczema (d) reveal considerably lower versican expression or rather its absence on the infiltrating cells.

1000 1000 MyLa mock MyLa-V1 SeAx mock SeAx-V1 800 800 *** *** 600 600

400 400

200 200 cluster size (pixel) cluster cluster size (pixel) cluster

0 0 Myla-mock Myla-V1 SeAx-mock SeAx-V1 Figure 4. Changes in cell cluster size of lymphoid cells upon upregulation of versican V1. CTCL cell lines MyLa (a) and SeAx (b) were transfected with pHIV-1-Blasti-VSV-G-V1 or mock vector. Scatter plots represent cell cluster size measured in pixels, while a horizontal line represents the mean value of multiple independent experiments. Visualization of morphological changes by phase-contrast microscopy (original magniﬁcation × 10) of bulk cell cultures of MyLa (a) and SeAx (b) before and after transfection pHIV-1-Blasti-VSV-G-V1 or mock. Signiﬁcant P-values o0.001 are highlighted with three asterisks (***).

* 30 30 *** ** ** * ** ***

20 20

10 10 migration index (%) migration index migration index (%) migration index

0 0 MyLa MyLa-V1 MyLa-mock MyLa, SDF1 MyLa, MyLa+PDGF MyLa-V1, Hase MyLa-V1, MyLa-V1, SDF1 MyLa-V1, MyLa-mock, Hase MyLa-mock, MyLa-mock, SDF1 MyLa-mock, MyLa+PDGF, SDF1 MyLa+PDGF, MyLa, SDF1, AMD3100 SDF1, MyLa, MyLa-V1, SDF1, AMD3100 SDF1, MyLa-V1, MyLa-mock, SDF1, AMD3100 SDF1, MyLa-mock, MyLa+PDGF, SDF1, AMD3100 SDF1, MyLa+PDGF, Figure 5. Changes in migratory behavior of lymphoid cells upon upregulation of versican V1. Scatter plots represent migration index with a horizontal line representing mean of at least three independent experiments. (a) Cell chemotaxis assays with/without SDF1 using parent MyLa cell line, before and after PDGF stimulation to induce versican expression. AMD3100 (plerixafor) is a CXCR4 inhibitor that blocks SDF1–CXCR4 axis. (b) Cell chemotaxis assays with/without SDF1 using MyLa cell line transfected with pHIV-1-Blasti-VSV-G-V1 or mock. Hase stands for hyaluronidase and is employed to abrogate versican-mediated effects. Signiﬁcant P-values are highlighted in each graph, *Po0.05; **Po0.01; ***Po0.001.

upon CXCR4 inhibition with AMD3100 (t-test P = 0.012). Overexpression of versican V1 inhibits cell proliferation by As V1-transfected MyLa cells have indigenously higher sponta- promoting activation-induced cell death and cell cycle arrest neous migration rate than the parent cells line (t-test Po0.0001), As versican exerts diverse influences on cellular processes, we next we assumed that this may be due to versican overexpression. sought to investigate the effect of versican V1 overexpression on Therefore, we treated our V1-transfected cells with hyaluronidase proliferation of CTCL cell lines. For this purpose, mock- and V1- to enzymatically degrade versican (Figure 5b). Hyaluronidase transfected cell lines were stimulated with PMA and ionomycin, treatment significantly diminished spontaneous migration in V1- and proliferation was measured by MTT assay. V1 transfection transfected (t-test P = 0.0004) in contrast to mock-transfected exhibited an inhibitory effect on MyLa and SeAx proliferation MyLa cells (t-test P = 0.159). It appears that V1 overexpression after (ANOVA Po0.0001 for both cell lines), while mock-transfected cell transduction, similar to V1 overexpression upon PDGF stimulation, lines proliferated readily to stimulation with PMA and ionomycin increases spontaneous as well as chemokine-directed migration of (Figure 6a). This is the first report on how versican V1 isoform lymphoma cells, which is likely to be one advantage for their affects proliferation of lymphoid cell lines originating from further spread. mycosis fungoides (MyLa) and Sézary patients (SeAx). Because

150 **** *** **** **** **** 80 * Myla-mock Myla-mock *** Myla-V1 Myla-V1 60 SeAx-mock *** 100 SeAx-mock SeAx-V1 * SeAx-V1

% 40 *** 50 *** 20 apoptotic cells (%)

0 0 control control PMA/Iono PMA 100ng + Iono 1uM PMA 50ng + Iono 0.5uM PMA 25ng + Iono 0.25uM

***G1, ****S 150 250 G1 MyLa *G1, ****S, *G2/M SeAx S 200 100 G2/M 150 % 100 50

p53+ cells/HPF 50

0 0 Mock V1 SeAx-V1, - SeAx-V1, SeAx-mock, - SeAx-mock, SeAx-V1, PMA/Iono SeAx-V1, SeAx-mock, PMA/Iono SeAx-mock, Figure 6. Effects of versican V1 overexpression on cell proliferation, cell death and cell cycle. (a) Cell proliferation analysis by MTT assay in MyLa and SeAx cell lines transfected with pHIV-1-Blasti-VSV-G-V1 or the mock control. Bars represent mean ± s.e.m. of at least three independent experiments. (b) Assessment of AICD in MyLa and SeAx cell lines transfected with pHIV-1-Blasti-VSV-G-V1 or mock. Bars represent mean ± s.e.m. of at least three independent experiments. (c) Analysis of cell cycle phases in SeAx cell line transfected with pHIV-1-Blasti-VSV-G- V1 or mock. Stacked bars represent mean percentage of cells in a given phase. (d) Quantitative immunohistochemistry of p53 protein expression in V1- or mock-transfected MyLa and SeAx cell lines. Scatter plots represent the number of p53-positive cells, with a horizontal line representing mean. Significant P-values are highlighted in each graph, *Po0.05; ***Po0.001; ****Po0.0001. stimulation with PMA and ionomycin can lead to activation- decrease of cells detected in S phase could be observed (30.65% induced cell death (AICD) and thus reduce the number of vs 18.65%, ANOVA Po0.0001). Changes in G1 and S phases were proliferating cells, we measured the number of apoptotic cells particularly evident in V1-transfected SeAx cells in contrast to following stimulation with these two substances. AICD could be mock (ANOVA Po0.001 and Po0.0001, respectively). These readily induced in both cell lines and was significantly higher in results imply an inhibitory effect of versican V1 overexpression on V1-transfected cells as compared with mock (t-test Po0.05, proliferation of lymphoid cells resulting in G1 arrest. One of the Figure 6b). However, the percentage of apoptotic cells was explanations for G1 arrest might be a constantly higher expression definitively higher in the SeAx cell line (t-test for mock P = 0.001, of p53 protein in SeAx than in MyLa cells (t-test P = 0.039), which V1 P = 0.003, Figure 6b), which would explain higher reduction of was not influenced by the transfection with either of the vectors cell numbers in SeAx following TCR triggering with PMA and (Figure 6d). ionomycin. The following analysis of the cell cycle by flow cytometry revealed that V1-transfected cell lines display remarkable altera- Overexpression of versican V1 changes chemosensitivity of tions of cell cycle phases. Stimulation with PMA and ionomycin led lymphoid cells to an increase of cells in G2/M phase, irrespective of the vector After having shown that V1 overexpression interferes with cell used. In MyLa, there was no difference between stimulated mock- cycle, we next analyzed the impact of versican overexpression on and V1-transfected cells with respect to a particular cell cycle sensitivity of CTCL cell lines to chemotherapeutics by employing phase (ANOVA P40.05, data not shown). Cell cycle machinery in MTT assay (Figure 7). We chose doxorubicin, gemcitabine and the SeAx cell line, on the other hand, reacted differently to PMA methotrexate for our experiments, as they are commonly used to and ionomycin stimulation (Figure 6c). V1-transfected SeAx cells treat SéS. Even though both MyLa and SeAx cell lines proliferated revealed an increased proportion of cells in G1 phase following less after treatment with doxorubicin, there was a further this stimulation, suggesting G1 arrest (control 48.57% vs stimula- significant decrease of cell proliferation in V1-transfected cells as tion 54.13%, ANOVA Po0.05). Correspondingly, a prominent compared with mock (ANOVA, Po0.0001). V1-transfected SeAx

150 Myla-mock SeAx-mock 150 Myla-mock SeAx-mock Myla-V1 SeAx-V1 Myla-V1 SeAx-V1

** **** *** **** 100 100 **** **** **** **** **** ** % %

50 50

0 0 control control Doxorubicin 2 ug Doxorubicin 4 ug Doxorubicin Gemcitabine 1 uM Gemcitabine 5 uM Doxorubicin 10 ug Doxorubicin 20 ug Doxorubicin Gemcitabine 0.1 uM Gemcitabine 0.5 uM Figure 7. Increased sensitivity to chemotherapeutics upon V1 overexpression in lymphoid cells. Cell proliferation analysis by MTT assay in MyLa and SeAx cell lines transfected with pHIV-1-Blasti-VSV-G-V1 or mock before and after treatment with doxorubicin (a) and gemcitabine (b). Bars represent mean ± s.e.m. of at least three independent experiments. Signiﬁcant P-values are highlighted in each graph, **Po0.01; ****Po0.0001.

cells, in particular, appeared to be highly sensitive to doxorubicin Our study reveals for the first time the effects of versican over- in a dose-dependent manner (SeAx-V1 vs MyLa-V1, ANOVA expression on lymphoid cells. Regulation of versican expression by Po0.0001, Figure 7a). Higher, more toxic concentrations of cytokines is variable depending on the cell type analyzed.8,28,29,35 doxorubicin (10 and 20 μg) did not seem to enhance this effect In our system using lymphoid cell lines, versican could be further in SeAx cells. Conversely, treatment with gemcitabine only positively regulated by PDGF. Expression of PDGF was rarely influenced proliferation in MyLa cells transfected with V1. Here a investigated in CTCL, with one report suggesting its enhanced constant reduction of cell proliferation could be observed expression in CTCL.36 Versican V1 overexpression in CTCL cell lines compared with mock control (ANOVA Po0.01, Figure 7b). In altered their growth and migratory behavior. Cell clusters that are SeAx cells, gemcitabine did indeed reduced the proliferation but usually observed in the bulk CTCL cell cultures were decreased independent of V1 overexpression. Finally, treatment with upon V1 overexpression. It may be that smaller cell clusters are methotrexate negatively impacted proliferation in both MyLa more likely to interact with other cell types involved in leukocyte and SeAx irrespective of V1 overexpression (data not shown). Our trafficking to regions of inflammation by using cell receptors that data reveal that versican V1 induction in lymphoid cells can, in bind to versican, similar to ECM components, different selectins addition to cell cycle changes, increase their sensitivity to and CD44 (downregulated in our array data set).9,10,37 We show doxorubicin and gemcitabine, which is a finding of clinical that versican V1 overexpression increases spontaneous and SDF-1- relevance for this patient group. directed migration of lymphoid cells, thus facilitating their homing to skin. Specific CS domains of versican can bind chemokines such as SDF-1 to attract other mononuclear cells,9,27 which in turn DISCUSSION would increase homing of these cells into skin. Versican and its Versican is a versatile ECM component involved in different binding partners in the ECM form filamentous cable-like structures cellular processes (adhesion, migration, differentiation, prolifera- that serve as scaffolds for the infiltrating cells and are so able to tion, apoptosis) both in health and disease, including cancer.8 modulate their phenotype and immune functions.10,37,38 In solid Gene expression profiling studies have identified versican (VCAN) cancer models, versican V1 was able to reduce the attachment as one of the dysregulated genes in CTCLs (reviewed in Dulmage and promote cancer cell migration and invasion.6,8,11,39 In the skin, and Geskin2) by showing its overexpression in PBMCs or CD4+ versican is normally expressed in the epidermis and dermal cells of patients with SéS.3,4 Our study reveals upregulation of connective tissue. We demonstrate a consistent versican expres- versican in sorted CD4+Vβ-restricted malignant cells of Sézary sion on the SCs in the blood and skin, supporting its relevance in patients. In CTCL, tumor cell population lies within CD4+ most of SéS. Recently, versican was shown to activate resident fibroblasts the time, more specifically within clonal TCR-Vβ-restricted and endothelial cells in the tumor stroma via TLR2/6 to elicit malignant cells.2 High versican (CSPG2) expression has also been production of various cytokines, including interleukin-8 (upregu- observed in adult T-cell leukemia, where PBMCs or CD4+ cells lated in our array data set), a proinflammatory C-X-C chemokine were used for gene expression analysis.32,33 Proteoglycans, with chemotactic, tumorigenic and proangiogenic properties.40 including versican, can be synthetized by leukocytes during No doubt versican does not act alone in directing SCs to the skin, inflammatory response.12 T lymphocytes, on the other hand, show and it will be important to identify other molecules with which no or very scarce expression of versican34 as our control samples versican interact. revealed. This, in conjunction with above mentioned reports, Accumulation of versican is associated with proliferative cell suggests that surface upregulation of versican (irrespective of the phenotype seen in cancer,11,12 even though some versican leukemic cell sorting method used) may be advantageous to isoforms may have opposing activities on cell proliferation.6,41,42 malignant T cells, yet this hypothesis was lacking comprehensive Versican V1 was shown to stimulate tumor cell proliferation in functional evidence until now. various in vitro studies.29,42–44 V1 overexpression in our CTCL cell To date, versican expression has been studied in the context of lines demonstrated an inhibitory effect on cell proliferation, ECM, as extracellular protein and seldom as surface protein. partially by promoting AICD. We observed no change in

Leukemia (2015) 2024 – 2032 © 2015 Macmillan Publishers Limited Versican upregulation in Sézary cells K Fujii et al 2031 steady-state apoptosis rate in V1-transfected cell lines, despite 8 Du WW, Yang W, Yee AJ. Roles of versican in cancer biology--tumorigenesis, previous reports that V1 overexpression may account for selective progression and metastasis. Histol Histopathol 2013; 28: 701–713. apoptosis resistance and sensitivity.7 Moreover, V1 overexpression 9 Evanko SP, Potter-Perigo S, Bollyky PL, Nepom GT, Wight TN. Hyaluronan and disturbed cell cycle by arresting CTCL cells in G1 and S phases, versican in the control of human T-lymphocyte adhesion and migration. Matrix 31 – without changes in their p53 status. TCR increased further the Biol 2012; :90 100. fl percentage of cells in G1, S and G2/M stages. The accumulation of 10 Wight TN, Kang I, Merrilees MJ. Versican and the control of in ammation. Matrix Biol 2014; 35: 152–161. cells in G2/M phases can be in part attributable to lentiviral 45 fi 11 Ricciardelli C, Sakko AJ, Ween MP, Russell DL, Horsfall DJ. The biological role and plasmid used for transfection, although similar ndings with regulation of versican levels in cancer. Cancer Metastasis Rev 2009; 28:233–245. increased cell numbers in S, G2 and M stages were reported for 12 Wight TN, Kinsella MG, Evanko SP, Potter-Perigo S, Merrilees MJ. Versican and the 46,47 versican-transfected breast cancer cells. Reduction of cell regulation of cell phenotype in disease. Biochim Biophys Acta 2014; 1840: proliferation and increased susceptibility to AICD would be 2441–2451. beneficial in CTCL, as CTCL tumor cells are highly resistant to 13 Kim YH, Willemze R, Pimpinelli N, Whittaker S, Olsen EA, Ranki A et al. TNM AICD induced by TCR stimulation.48 classification system for primary cutaneous lymphomas other than mycosis fun- Furthermore, different chemotherapeutics that target cell cycle goides and Sézary syndrome: a proposal of the International Society for Cuta- might have functional benefit in cells with high versican neous Lymphomas (ISCL) and the Cutaneous Lymphoma Task Force of the expression. Following DNA intercalation and topoisomerase II- European Organization of Reseace and Treatment of Cancer (EORTC). Blood 2007; 110:479–484. mediated DNA damage, doxorubicin similar to other anthracy- 14 Dummer R, Nestle FO, Niederer E, Ludwig E, Laine E, Grundmann H et al. Geno- clines induces growth arrest in G1/G2 phases, followed by 49 typic, phenotypic and functional analysis of CD4+CD7+ and CD4+CD7- T lym- programmed cell death. Gemcitabine, a cytidine analog, is an phocyte subsets in Sezary syndrome. Arch Dermatol Res 1999; 291: 307–311. 50 S-phase specific agent. Cytotoxic effects of doxorubicin and 15 Schwab C, Willers J, Niederer E, Ludwig E, Kundig T, Grob P et al. The use of gemcitabine seem to be potentiated in our V1-overexpressing anti-T-cell receptor-Vbeta antibodies for the estimation of treatment success and CTCL cell lines, identifying an additional shortcoming of versican phenotypic characterization of clonal T-cell populations in cutaneous T-cell V1 overexpression that provides sensitivity to chemotherapy and lymphomas. Br J Haematol 2002; 118: 1019–1026. could be exploited further for therapeutic purposes. 16 Kaltoft K, Bisballe S, Dyrberg T, Boel E, Rasmussen P, Thestrup-Pedersen K. In conclusion, we confirm versican as one of the dysregulated Establishment of two continuous T-cell strains from a single plaque of a patient 28 – genes in SéS and describe its effects on the biology of SCs. with mycosis fungoides. In Vitro Cell Dev Biol 1992; :161 167. Although versican overexpression confers lymphoid cells with 17 Urosevic M, Fujii K, Calmels B, Laine E, Kobert N, Acres B et al. Type I IFN innate immune response to adenovirus-mediated IFN-gamma gene transfer increased migratory capacity, it also makes them more sensitive to contributes to the regression of cutaneous lymphomas. J Clin Invest 2007; 117: AICD and some chemotherapeutics. With increased knowledge of 2834–2846. the roles of versican and its isoforms in lymphoma biology, it will 18 Saito R, Smoot ME, Ono K, Ruscheinski J, Wang P-L, Lotia S et al. A travel guide to be possible to identify strategies for the therapeutic targeting of Cytoscape plugins. Nat Methods 2012; 9: 1069–1076. this proteoglycan. 19 Montojo J, Zuberi K, Rodriguez H, Kazi F, Wright G, Donaldson SL et al. Gene- MANIA Cytoscape plugin: fast gene function predictions on the desktop. Bioin- formatics 2010; 26: 2927–2928. CONFLICT OF INTEREST 20 Kumar M, Keller B, Makalou N, Sutton RE. Systematic determination of the 12 – The authors declare no conflict of interest. packaging limit of lentiviral vectors. Hum Gene Ther 2001; : 1893 1905. 21 Lamant Le Reyniès A, Duplantier M-M, Rickman DS, Sabourdy F, Giuriato S et al. dGene-expression profiling of systemic anaplastic large-cell lymphoma reveals ACKNOWLEDGEMENTS differences based on ALK status and two distinct morphologic ALK+ subtypes. Blood 2007; 109: 2156–2164. We thank Ines Kleiber-Schaaf for technical assistance in immunohistochemistry. This 22 Fang H, Liu J, Guo D, Liu P, Zhao Y. Epigenetic regulation of putative tumor project was supported by research funding from Oncosuisse (Grant number suppressor TGFBI in human leukemias. Chin Med J (Engl) 2014; 127: 1645–1650. OCS-01934-08-2006 to MU-M) and Kurt und Senta Herrmann-Stiftung (to MU-M). Microarray experiments and data acquisition were performed at Functional Genomics 23 Liu C, Li B, Cheng Y, Lin J, Hao J, Zhang S et al. MiR-21 plays an important role in Center Zurich (http://www.fgcz.ch). radiation induced carcinogenesis in BALB/c mice by directly targeting the tumor suppressor gene Big-h3. Int J Biol Sci 2011; 7:347–363. 24 Kondratiev S, Duraisamy S, Unitt CL, Green MR, Pinkus GS, Shipp MA et al. REFERENCES Aberrant expression of the dendritic cell marker TNFAIP2 by the malignant cells of Hodgkin lymphoma and primary mediastinal large B-cell lymphoma distinguishes 1 Talpur R, Singh L, Daulat S, Liu P, Seyfer S, Trynosky T et al. Long-term outcomes these tumor types from morphologically and phenotypically similar lymphomas. of 1,263 patients with mycosis fungoides and Sézary syndrome from 1982 Am J Surg Pathol 2011; 35: 1531–1539. to 2009. Clin Cancer Res 2012; 18: 5051–5060. 25 Teitell MA. The TCL1 family of oncoproteins: co-activators of transformation. 2 Dulmage BO, Geskin LJ. Lessons learned from gene expression profiling of Nat Rev Cancer 2005; 5: 640–648. cutaneous T-cell lymphoma. Br J Dermatol 2013; 169:1188–1197. 26 Zhao A, Yang Z, Sun R, Grinchuk V, Netzel-Arnett S, Anglin IE et al. SerpinB2 is 3 Kari L, Loboda A, Nebozhyn M, Rook AH, Vonderheid EC, Nichols C et al. Classi- critical to Th2 immunity against enteric nematode infection. J Immunol 2013; 190: fication and prediction of survival in patients with the leukemic phase of cuta- 5779–5787. neous T cell lymphoma. J Exp Med 2003; 197: 1477–1488. 4 van Doorn R, Dijkman R, Vermeer MH, Out-Luiting JJ, van der Raaij-Helmer EM, 27 Hirose J, Kawashima H, Yoshie O, Tashiro K, Miyasaka M. Versican interacts 276 Willemze R et al. Aberrant expression of the tyrosine kinase receptor EphA4 and with chemokines and modulates cellular responses. J Biol Chem 2001; : – the transcription factor twist in Sezary syndrome identified by gene expression 5228 5234. analysis. Cancer Res 2004; 64: 5578–5586. 28 Cardoso LEM, Little PJ, Ballinger ML, Chan CK, Braun KR, Potter-Perigo S et al. 5 Cattaruzza S, Schiappacassi M, Ljungberg-Rose A, Spessotto P, Perissinotto D, Platelet-derived growth factor differentially regulates the expression and Morgelin M et al. Distribution of PG-M/versican variants in human post-translational modification of versican by arterial smooth muscle cells tissues and de novo expression of isoform V3 upon endothelial cell through distinct protein kinase C and extracellular signal-regulated kinase path- activation, migration, and neoangiogenesis in vitro. J Biol Chem 2002; 277: ways. J Biol Chem 2010; 285: 6987–6995. 47626–47635. 29 Nikitovic D, Zafiropoulos A, Katonis P, Tsatsakis A, Theocharis AD, Karamanos NK 6 Touab M, Villena J, Barranco C, Arumi-Uria M, Bassols A. Versican is differentially et al. Transforming growth factor-beta as a key molecule triggering the expres- expressed in human melanoma and may play a role in tumor development. sion of versican isoforms v0 and v1, hyaluronan synthase-2 and synthesis of Am J Pathol 2002; 160:549–557. hyaluronan in malignant osteosarcoma cells. IUBMB Life 2006; 58:47–53. 7 LaPierre DP, Lee DY, Li S-Z, Xie Y-Z, Zhong L, Sheng W et al. The ability of versican 30 Rahmani M, Read JT, Carthy JM, McDonald PC, Wong BW, Esfandiarei M et al. to simultaneously cause apoptotic resistance and sensitivity. Cancer Res 2007; 67: Regulation of the versican promoter by the beta-catenin-T-cell factor complex in 4742–4750. vascular smooth muscle cells. J Biol Chem 2005; 280: 13019–13028.

© 2015 Macmillan Publishers Limited Leukemia (2015) 2024 – 2032 Versican upregulation in Sézary cells K Fujii et al 2032 31 Tufvesson E, Westergren-Thorsson G. Alteration of proteoglycan synthesis in 41 Lemire JM, Merrilees MJ, Braun KR, Wight TN. Overexpression of the V3 variant of human lung fibroblasts induced by interleukin-1beta and tumor necrosis versican alters arterial smooth muscle cell adhesion, migration, and proliferation factor-alpha. J Cell Biochem 2000; 77:298–309. in vitro. J Cell Physiol 2002; 190:38–45. 32 Alizadeh AA, Bohen SP, Lossos C, Martinez-Climent JA, Ramos JC, Cubedo-Gil E 42 Sheng W, Wang G, Wang Y, Liang J, Wen J, Zheng P-S et al. The roles of versican et al. Expression profiles of adult T-cell leukemia-lymphoma and associations with V1 and V2 isoforms in cell proliferation and apoptosis. Mol Biol Cell 2005; 16: clinical responses to zidovudine and interferon alpha. Leuk Lymphoma 2010; 51: 1330–1340. 1200–1216. 43 Hernández D, Miquel-Serra L, Docampo MJ, Marco-Ramell A, Bassols A. Role of 33 Pise-Masison CA, Radonovich M, Dohoney K, Morris JC, O'Mahony D, Lee MJ et al. versican V0/V1 and CD44 in the regulation of human melanoma cell behavior. Int fi Gene expression pro ling of ATL patients: compilation of disease-related genes J Mol Med 2011; 27:269–275. and evidence for TCF4 involvement in BIRC5 gene expression and cell viability. 44 Onken J, Moeckel S, Leukel P, Leidgens V, Baumann F, Bogdahn U et al. Versican 113 – Blood 2009; : 4016 4026. isoform V1 regulates proliferation and migration in high-grade gliomas. J Neu- 34 Rottiers P, Verfaillie T, Contreras R, Revets H, Desmedt M, Dooms H et al. rooncol 2014; 120:73–83. Differentiation of EL4 lymphoma cells by tumoral environment is associated with 45 Planelles V, Jowett JB, Li QX, Xie Y, Hahn B, Chen IS. Vpr-induced cell cycle arrest is inappropriate expression of the large chondroitin sulfate proteoglycan PG-M and conserved among primate lentiviruses. J Virol 1996; 70: 2516–2524. the tumor-associated antigen HTgp-175. Int J Cancer 1998; 78:503–510. 46 Du WW, Yang BB, Shatseva TA, Yang BL, Deng Z, Shan SW et al. Versican G3 35 Arslan F, Bosserhoff AK, Nickl-Jockschat T, Doerfelt A, Bogdahn U, Hau P. The role promotes mouse mammary tumor cell growth, migration, and metastasis by of versican isoforms V0/V1 in glioma migration mediated by transforming growth influencing EGF receptor signaling. PLoS One 2010; 5: e13828. factor-beta2. Br J Cancer 2007; 96: 1560–1568. 36 Su IJ, Kadin ME. Expression of growth factor/receptor genes in postthymic T cell 47 Du WW, Yang BB, Yang BL, Deng Z, Fang L, Shan SW et al. Versican G3 malignancies. Am J Pathol 1989; 135:439–445. domain modulates breast cancer cell apoptosis: a mechanism for breast 6 37 Petrey AC, de la Motte CA. Hyaluronan, a crucial regulator of inflammation. Front cancer cell response to chemotherapy and EGFR therapy. PLoS One 2011; : Immunol 2014; 5: 101. e26396. β 38 Gill S, Wight TN, Frevert CW. Proteoglycans: key regulators of pulmonary 48 Klemke C-D, Brenner D, Wei E-M, Schmidt M, Leverkus M, Gülow K et al. Lack of – inflammation and the innate immune response to lung infection. Anat Rec T-cell receptor induced signaling is crucial for CD95 ligand up-regulation and (Hoboken) 2010; 293:968–981. protects cutaneous T-cell lymphoma cells from activation-induced cell death. 39 Wasa J, Nishida Y, Shinomura T, Isogai Z, Futamura N, Urakawa H et al. Versican V1 Cancer Res 2009; 69: 4175–4183. isoform regulates cell-associated matrix formation and cell behavior differentially 49 Perego P, Corna E, De Cesare M, Gatti L, Polizzi D, Pratesi G et al. Role of apoptosis from aggrecan in Swarm rat chondrosarcoma cells. Int J Cancer 2012; 130: and apoptosis-related genes in cellular response and antitumor efficacy of 2271–2281. anthracyclines. Curr Med Chem 2001; 8:31–37. 40 Wang W, Xu G-L, Jia W-D, Ma J-L, Li J-S, Ge Y-S et al. Ligation of TLR2 by versican: a 50 Mini E, Nobili S, Caciagli B, Landini I, Mazzei T. Cellular pharmacology of gemci- link between inflammation and metastasis. Arch Med Res 2009; 40:321–323. tabine. Ann Oncol 2006; 17 (Suppl 5): v7–v12.

Supplementary Information accompanies this paper on the Leukemia website (http://www.nature.com/leu)