CD44 enhances tumor formation and lung metastasis in experimental osteosarcoma and is an additional predictor for poor patient outcome

Authors


Abstract

Formation of metastases in the lungs is the major cause of death in patients suffering from osteosarcoma (OS). Metastases at presentation and poor response to preoperative chemotherapy are strong predictors for poor patient outcome. The elucidation of molecular markers that promote metastasis formation and/or chemoresistance is therefore of importance. CD44 is a plasma membrane glycoprotein that binds to the extracellular matrix component hyaluronan (HA) and has been shown to be involved in metastasis formation in a variety of other tumors. Here we investigated the role of CD44 expression on OS tumor formation and metastasis. High CD44 expression, evaluated with a tissue microarray including samples from 53 OS patients and stained with a pan-CD44 antibody (Hermes3), showed a tendency (p < 0.08) to shortened overall survival. However, nonresponders and patients with lung metastases and high CD44 expression had significantly poorer prognosis than patients with low CD44 expression. Overexpression of the standard CD44 isoform (CD44s) and its HA-binding defective mutant R41A in osteoblastic SaOS-2 cells resulted in HA-independent higher migration rates and increased chemoresistance, partially dependent on HA. In an orthotopic mouse model of OS, overexpression of CD44s in SaOS-2 cells resulted in an HA-dependent increased primary tumor formation and increased numbers of micrometastases and macrometastases in the lungs. In conclusion, although CD44 failed to be an independent predictor for patient outcome in this limited cohort of OS patients, increased CD44 expression was associated with even worse survival in patients with chemoresistance and with lung metastases. CD44-associated chemoresistance was also observed in vitro, and increased formation of lung metastases was found in vivo in SCID mice. © 2013 American Society for Bone and Mineral Research.

Introduction

Osteosarcoma (OS) is the most common primary tumor of bone in children and adolescents. The presence of malignant spindle cells that produce osteoid and/or immature bone is characteristic for this highly aggressive cancer type.1 The incidence of OS in the general population is three cases per million per year, but is higher in adolescence, in which it reaches eight to 11 cases/million/year at 15 to 19 years of age.2 OS has a great tendency to spread to the lungs, and less frequently to the bones. Formation of bone metastases occurs only after pulmonary metastases have already been established.3 At the time of diagnosis, up to 15% to 20% of patients already have detectable metastases. However, 80% of patients initially presenting with localized disease develop metastases after surgical resection.4 The combination of multi-agent chemotherapy with surgery introduced in the late 1970s remarkably improved the overall survival of patients with nonmetastatic disease, whose 5-year survival rate is now 70%, as opposed to only 20% few decades ago. In contrast, the patients with metastatic or recurrent disease did not benefit from these clinical advances and they unfortunately face a very poor prognosis, with a survival rate that still remains at 20%.5 The failure of treatment in these patients is often associated with gained resistance to chemotherapy.6 Nowadays, the most powerful and reproducible prognostic indicators for OS patients are metastatic lesions at presentation and histological response to preoperative chemotherapy.7 Thus, it is of substantial relevance to identify molecular markers associated with the increased metastatic potential or chemoresistance, which may serve as diagnostic or prognostic factors. Acquiring insight into the basic biology of OS progression will make the identification of such new therapeutic targets possible, with the final goal to develop treatment strategies that eradicate metastases, the major cause of death in OS.

CD44 has been linked with increased metastatic spread in various types of cancers.8 CD44 designates a family of broadly distributed type I transmembrane glycoproteins that serve as cell-cell and cell-matrix adhesion molecules and as principal receptors for hyaluronan (HA), a major component of the extracellular matrix in many tissues including bone.9 Existence of multiple isoforms, generated through alternative splicing, and extensive posttranslational modifications underlie the wide repertoire of CD44 biological functions in development, wound healing, inflammation, hematopoiesis, immune response, and tumor progression.10 Tissue-specific splicing results in the formation of the standard CD44 isoform (CD44s), which lacks all variant exons, in cells of mesenchymal origin; thus, the expression of this isoform may be relevant for sarcoma tumor progression.11, 12 CD44 has been shown to promote tumor and metastasis development both in an HA-dependent and HA-independent fashion.13, 14 Both CD44 and HA have been implicated in resistance to anticancer drugs.15–17

Only a few immunohistochemical studies using OS tissue specimens have addressed the contribution of CD44 to OS progression and metastasis, giving rise to conflicting data.18–20 Interestingly, to our knowledge, no gain- or loss-of-function studies investigating the role of CD44 in established OS cell lines or xenograft mouse models have so far been reported. On the other hand, a limited number of reports indicated the relevance of HA in OS tumor progression after making use of established cell lines. Treatment with HA oligosaccharides suppressed the formation of cell-associated matrix, leading to inhibited tumorigenicity of the human MG63 and murine LM8 OS cell lines in vitro.21 In in vivo studies, intratumoral injection of HA oligosaccharides into subcutaneous LM8-derived tumors reduced the accumulation of HA in tumor tissue and resulted in significant suppression of lung metastases.21 In vivo administration of 4-methylumbelliferon, an inhibitor of HA synthesis, inhibited the retention of HA in the periphery of the primary tumors and markedly reduced the number of metastatic lung lesions formed by LM8 OS cells.22

In this study we show that CD44 can be used as an additional prognostic factor for OS patients' outcome. With the aim of investigating the biological relevance of CD44s expression, and its interaction with HA, to OS progression and metastasis, we overexpressed the CD44s isoform and its HA-binding defective mutant CD44s R41A in the low metastatic human OS SaOS-2 cells, which display an osteoblastic phenotype most commonly observed in human patients. Using an intratibial xenograft OS mouse model we demonstrated that CD44s enhances primary tumor growth and formation of pulmonary metastases in an HA-dependent manner. In conclusion, the results presented here highlight CD44-HA interaction as a potential target for therapeutic intervention in OS.

Patients and Methods

Human OS tissue microarray analysis

OS tissue specimens were collected between June 1990 and December 2005 from 53 patients during primary tumor resection in accordance with the regulations of the local ethics committee. Clinical data for the patients are presented in Table 1. All patients received neoadjuvant chemotherapy and the subsequent response was determined histologically on resected tumor specimens according to Salzer-Kuntschik.23 Salzer-Kuntschik grades I, II, and III were considered to be a good response, whereas grades IV, V, and VI were classified as a poor response. The tumor samples were decalcified and the tissue microarray was arranged as described.24 Microarray sections of 4.5 µm were processed as reported25 and stained with a pan-CD44 antibody Hermes3 (generously provided by Dr. Sirpa Jalkanen, Turku, Finland) (1:1000) and counterstained with hematoxylin. Tissue microarray grading was performed based on the intensity and area percentage of the positive stain using Supplemental Table 1. The intensity of the stain was judged by eye (weak, moderate, and strong). The percentage of staining was calculated using a custom MATLAB (v2009b; Mathworks Inc., Natick, MA, USA) program. Positive (brown) and negative (blue) staining were separated using color deconvolution theory.26 The area percentage of the stain was defined as the positive-stained area (number of brown pixels) over total tissue area (number of blue and brown pixels) (Supplemental Fig. 1). Grade 1 was considered as negative and grades 2 and 3 were considered as positive for CD44 expression. Kaplan-Meier analysis was used to correlate CD44 expression with overall and metastasis-free survival of OS patients.

Table 1. Clinical Characteristics of Osteosarcoma Patients
Clinical characteristics (n = 53)n%
  1. SK = Salzer-Kuntschik.

Gender  
 Male3464
 Female1936
Age (years)  
 <101019
 10–192649
 20–291121
 30–3924
 40–4924
 50–5924
Tumor type  
 Osteoblastic3872
 Chondroblastic713
 Fibroblastic48
 Telangiectatic48
Anatomic site  
 Extremities4687
 Spine and pelvis611
 Face12
Chemotherapy response  
 Responders (SK-I to SK-III)3057
  Salzer-Kuntschik I611
  Salzer-Kuntschik II1325
  Salzer-Kuntschik III1121
 Nonresponders (SK-IV to SK-VI)2343
  Salzer-Kuntschik IV1121
  Salzer-Kuntschik V917
  Salzer-Kuntschik VI36
Metastasis  
 No metastasis3464
 Total metastasis1936
 Metastasis at diagnosis24
 Metastasis after diagnosis1732

Cell culture and transduction

Human OS SaOS-2 (HTB-85) cells obtained from American Type Culture Collection (ATCC, Rockville, MD, USA) were cultured in DMEM (4.5 g/L glucose)/HamF12 (1:1) medium (Invitrogen, Carlsbad, CA, USA) supplemented with 10% heat-inactivated fetal calf serum (FCS) at 37°C in a humidified atmosphere of 5% CO2/95% air. In order to enable visualization of tumor cells within mouse tissues, SaOS-2 cells were transduced with the LacZ gene (SaOS-2/LacZ cells) as described.27 Stable expression of CD44s and its HA-binding defective mutant CD44s R41A was achieved by retroviral gene transfer. Briefly, Murine Stem Cell Virus (pMSCV) vectors containing human CD44s and CD44s R41A coding sequences, generously provided by Prof. Ivan Stamenkovic (Lausanne, Switzerland), were used to fuse V5 and His6 epitopes to the COOH-terminal ends of CD44s and CD44s R41A, giving rise to CD44s-V5/His6 and CD44s R41A-V5/His6 encoding sequences, which were subsequently subcloned into the retroviral expression vector pQCIXH (Clontech, Paolo Alto, CA, USA) containing a hygromycin resistance gene. All expression constructs were verified by sequencing of both strands. Retroviral particles containing pQCIXH empty vector (EV), pQCIXH CD44s-V5/His6, and pQCIXH CD44s R41A-V5/His6 were produced in HEK 293T cells, and were subsequently used to infect SaOS-2/LacZ cells as described.27 Selection for hygromycin resistance in medium containing 400 µg/mL of hygromycin (Merck Chemicals LTD, Nottingham, UK) revealed SaOS-2 EV, SaOS-2 CD44s, and SaOS-2 CD44s R41A cell lines.

Western blot analysis

Cells were lysed by agitation on a carrousel at 4°C for 1 hour in lysis buffer containing 50 mM Tris/HCl (pH 7.5), 150 mM NaCl, 1% NP40, 0.5% deoxycholic acid, 0.1% sodium dodecyl sulfate (SDS), 1 mM dithiothreitol (DTT), 1 mM phenylmethylsulfonyl fluoride (PMSF), and 10 mg/mL aprotonin. Cellular debris was removed by centrifugation at 16,060g and 4°C for 20 minutes. Equal amounts of proteins of individual cell extracts were separated by 8% SDS-PAGE. The proteins were then transferred by semi-dry blotting to Hybond-ECL membranes (GE Healthcare, Glattbrugg, Switzerland). Endogenous and V5-tagged CD44 proteins and GAPDH were detected with respective mouse monoclonal Hermes3 antibody (concentration 1 µg/mL), V5 antibody (1:5000; Invitrogen), and rabbit polyclonal anti-GAPDH antibody (1:3000; Santa Cruz Biotechnologies, Santa Cruz, CA, USA) and corresponding horseradish peroxidase (HRP)-conjugated secondary antibodies purchased from Santa Cruz Biotechnologies. Peroxidase activity was visualized with the Immobilon chemiluminescence substrate (Millipore, Billerica, MA, USA) and detected with a VersaDoc Imaging System (Bio-Rad, Hercules, CA, USA).

Adhesion assay

Ninety-six–well (96-well) plates were coated with 333 µg/cm2 of high molecular weight HA (HMW-HA) (Sigma-Aldrich, St. Louis, MO, USA) at 4°C overnight. They were then washed with PBS and blocked with heat-denatured 1% bovine serum albumin (BSA) (HD-BSA). Noncoated wells or wells coated with HD-BSA alone were used as controls. Adhesion assays were carried out with cells grown in tissue culture medium to subconfluence in 25-cm2 tissue culture flasks. They were then detached with Accutase (Sigma-Aldrich), resuspended in medium, and seeded at 1 × 104 cells per well and allowed to adhere at 37°C for 30 minutes. In CD44 blocking experiments, adhesion was performed in the presence of 10 µg/mL Hermes1 antibody (kindly provided by Dr. Sirpa Jalkanen, Turku, Finland) that blocks HA binding of 10 µg/mL rat IgG2A antibody (R&D Systems, Minneapolis, MN, USA) as a control. Nonadherent cells were removed by washing with PBS and adherent cells were fixed with 10% formalin in PBS at room temperature (RT) for 15 minutes and then stained with 0.05% crystal violet in H2O at RT for 15 minutes. Images of randomly selected areas of 3.6 mm2 were taken with an AxioCam MRm camera connected to the Zeiss Observer.Z1 inverted microscope (Carl Zeiss MicroImaging GmbH, Göttingen, Germany) set at ×4 magnification. The number of adherent cells in the analyzed area was determined with ImageJ software (NIH; http://rsb.info.nih.gov/ij). The total number of adherent cells per well was calculated and the percentage of adherent cells was obtained by dividing the number of adherent cells by the total number of seeded cells and multiplying by 100. The experiments were performed in triplicates and repeated three times.

Transwell migration assay

Cell culture inserts (Becton Dickinson, San Jose, CA, USA) with 8-µm porous filters in 24-well plates were used for a transwell migration assay. Cells grown to subconfluence were detached with Accutase (Sigma-Aldrich) and 2 × 104 cells in 300 µL of serum-free cell culture medium supplemented with penicillin/streptomycin/amphotericin B (PSA; 1:100; Invitrogen) were added to the upper compartment of the inserts. The lower compartments were filled with 700 µL of medium containing 10% FCS complemented with PSA. After incubation at 37°C for 24 hours, nonmigrating cells on the upper side of the insert were removed with a cotton swab. Cells that had migrated to the lower side of the filters were fixed with 10% formalin, permeabilized with 50 µM digitonin (Merck Chemicals LTD), and stained with 300 nM Picogreen (Invitrogen) in PBS at RT for 15 minutes. Three images per insert (two inserts per cell line) showing an area of 0.58 mm2 were taken with an AxioCam MRm camera connected to the Zeiss Observer.Z1 inverted microscope adjusted to ×10 magnification and equipped with an appropriate filter block for blue excitation. The number of cells on the images was counted with the ImageJ software, and the percentage of migrated cells was calculated as described for the adhesion assay. The experiments were performed at least three times.

In vitro cell proliferation assay

Subconfluent cells in the logarithmic growth phase were trypsinized and 5 × 104 cells, resuspended in 2.5 mL of cell culture medium, were seeded in 12.5-cm2 cell-culture flasks. The cells were allowed to grow for between 1 and 7 days and counted in intervals of 48 hours in triplicates. Cells in individual flasks were detached by trypsinization and counted in a Neubauer chamber. The doubling time during logarithmic growth was calculated according to the equation N = N0ekt (N0 = number of seeded cells; N = number of cells at time t). The experiments with individual cell lines were carried out three times.

Cytotoxicity assay

A total of 3 × 103 cells per well were seeded in 96-well plates and allowed to adhere overnight. The cells were then incubated in duplicates with increasing concentrations of cisplatin (0.01–25 µg/mL) for 72 hours. Cisplatin was purchased from Sigma-Aldrich. After drug treatment, the cells were incubated with 10 µL/well of WST-1 reagent (Roche, Basel, Switzerland) for 3 hours and the cell viability was then assessed as reported.28 Prism 4 network software (GraphPad, La Jolla, CA, USA) was used to calculate the half-maximal growth inhibitory concentration (IC50) of the drug. The experiments were repeated three times.

Intratibial OS xenograft model in SCID mice

The animal experiments were performed as described25 according to the guidelines of the “Schweizer Bundesamt für Veterinärwesen” and as approved by the authorities of the Kanton Zürich. Briefly, on day 0 of the experiment, 10 µL with 5 × 105 of SaOS-2 cells engineered as indicated in PBS/0.05% EDTA were injected intratibially into SCID/CB17 immunocompromised mice obtained from Charles River Laboratories (Sulzfeld, Germany). After the injection, the health condition of the mice was closely monitored. The development of primary tumors was visualized biweekly during the first 1 to 9 weeks of the experiment and then weekly until the end of the experiment by X-ray with a MX-20 DC Digital Radiography System (Faxitron X-Ray Corporation, Lincolnshire, IL, USA). The tumor volume was further estimated by measuring the length and the width of the tumor leg with a caliper, and the volume was calculated according to the formula V = length × width2/2. The volume of the non-injected leg was used as a reference value. The mice were euthanized in week 12 after tumor cell injection and the lung was perfused in situ as described.25 Primary tumors and lungs were fixed at RT in 2% formaldehyde for 30 minutes and processed for X-gal staining as reported.27 Indigo-blue stained metastases on the surface of lung whole mounts were counted at ×4 magnification under the Nicon Eclipse E600 microscope (Nikon Corporation, Tokyo, Japan) equipped with an integrated size grid. Metastatic foci smaller than 0.1 mm in diameter were considered as micrometastases and foci bigger than 0.1 mm as macrometastases. Two independent animal experiments were performed and the data were pooled.

Statistical analysis

Overall and metastasis-free patient survival was calculated using Kaplan–Meier curves and statistical significance was assessed with the log-rank test. Differences between means were analyzed by the Student's t test and p < 0.05 was considered significant. The results are presented as mean ± SEM.

Results

CD44 expression in human OS tumor samples is an additional prognostic factor in nonresponders and in patients with lung metastases

Human OS tissue-microarray sections including tumor specimens from 53 patients were analyzed immunohistochemically for total CD44 expression (Fig. 1). Representative images of tissue microarray sections with non-detectable, moderate and intense CD44 immunostaining are presented in Fig. 1A. The adequacy of our patient cohort was evaluated by determining the correlation of chemotherapy response and the presence of metastases with the overall survival, because these were identified as key determinants of prognosis in OS.29 Indeed, nonresponders and metastases-positive patients had significantly shorter overall survival (p < 0.05, p < 0.0001, respectively) than responders and metastases-free patients (not shown). Patients responding poorly to chemotherapy developed lung metastases more rapidly, and had a mean metastasis-free survival of 14 ± 2 months compared to 40 ± 2 months in patients with good response (p < 0.05; Fig. 1B). A Kaplan-Meier analysis revealed a tendency of shorter overall mean survival (50 ± 8 months) of patients with positive CD44 staining in tumor resections than patients with undetectable CD44 staining (88 ± 8 months; p = 0.0758; Fig. 1C). Nonresponders who were CD44-positive had a tendency for shorter overall survival than CD44-negative patients (p = 0.0732; Fig. 1D). In addition, nonresponders with CD44-positive staining in their tumor samples had significantly shorter mean metastasis-free survival of only 8.3 ± 1.4 months compared to patients with undetectable CD44 in tumor sections with mean metastasis-free survival of 16.5 ± 3 months (p < 0.05; Fig. 1E). Strikingly, all patients that were positive for both CD44 expression and metastases died within 22 months, significantly earlier than CD44-negative patients (p < 0.0001; Fig. 1F). Despite the small number of patients examined, the findings implicate that CD44 is associated with worse outcome in subsets of OS patients who responded poorly to chemotherapy or had metastases.

Figure 1.

Kaplan-Meier analyses of osteosarcoma tissue microarray immunostained with pan-CD44 antibodies. (A) Representative images of tumor tissue microarray sections showing entire spots (upper panel) and higher magnification (lower panel) with nondetectable (left), moderate (middle), and intense (right) CD44 immunostaining. (B) Metastasis-free survival of responders (Salzer-Kuntschik and colleagues,23 grade I–III) or of nonresponders (grade IV–VI) to neoadjuvant chemotherapy. (C) Overall survival of patients with negative (CD44 neg) or positive (CD44 pos) immunostaining of tumor tissue. Overall (D) and metastasis-free (E) survival of nonresponders to neoadjuvant therapy with CD44 neg or CD44 pos tumors. (F) Overall survival of patients with metastases and CD44 neg or CD44 pos tumors.

Overexpression of CD44 in an osteoblastic OS cell line increases the adhesion to HA, promotes cell migration, and induces chemoresistance

Based on the tissue microarray results, we hypothesized that CD44 may have a significant impact on the metastatic activity and the chemoresistance of OS cells. We therefore overexpressed by retroviral gene transfer the C-terminally V5/His6-tagged standard isoform CD44s in the human low metastatic SaOS-2 OS cell line with low endogenous CD44 expression. The standard isoform, with all the variant exons excised, was chosen for overexpression because it was found expressed as the predominant isoform in other human OS cell lines (not shown). The V5/His6-tagged HA binding-defective mutant CD44s R41A was included in the study to assess the relevance of CD44/HA interactions in the regulation of the metastatic ability and chemoresistance of SaOS-2 tumor cells. Western blot analysis of whole-cell extracts with a pan-CD44 antibody (Hermes3) indicated overexpression of CD44s and CD44s R41A in respective SaOS-2 CD44s and SaOS-2 CD44s R41A cells compared to control SaOS-2 EV (Fig. 2A). The protein components detected by Hermes3 and V5 antibodies had the expected size of CD44s.

Figure 2.

Overexpression of CD44s in SaOS-2 cells increases the adhesion to HA, promotes cell migration and enhances cisplatin chemoresistance, but does not affect proliferation. (A) SaOS-2 cells were retrovirally transduced with empty vector (EV) or with constructs encoding CD44s or HA binding-defective CD44s R41A, and the expression of respective proteins was confirmed on Western blots of cell extracts analyzed with CD44 antibody Hermes3 (H3), V5 antibody, and antibody to GAPDH (protein loading control). (B) Adhesion (% of seeded cells) to immobilized hyaluronan within 30 minutes of empty-vector (EV)-transduced, CD44s-expressing, or CD44s R41A–expressing cells. (C) Adhesion (% of seeded cells) of CD44s expressing cells to hyaluronan in the absence (–H1) or in the presence (+H1) of the CD44-blocking Hermes1 antibody or of control IgG. Transwell migration (% of seeded cells) (D), proliferation (E) of EV-transduced, CD44s-expressing, or CD44s R41A–expressing cells and cisplatin cytotoxicity (F) in respective cell lines determined in a WST-1 assay. IC50 = half-maximal growth inhibitory concentrations of cisplatin. Values in BF are expressed as means ± SEM of at least three independent experiments; *p < 0.05.

A significant (p < 0.01) 4.2-fold higher percentage of HA-adhering SaOS-2 CD44s cells compared to SaOS-2 EV cells in an assay examining short-term (30 minutes) adhesion demonstrated the functional expression of CD44s at the cell surface (Fig. 2B). Consistent with the binding defect of CD44s R41A, the adhesion of SaOS-2 CD44s R41A cells to HA was indistinguishable from that of SaOS-2 EV cells. A significant reduction in the percentage of short-term adhering SaOS-2 CD44s cells by preincubation with Hermes1 CD44s blocking antibodies, which was not observed with control immunoglobulin G (IgG), further confirmed that increased adhesion of SaOS-2 CD44s cells to HA was indeed mediated by direct interaction of overexpressed CD44s at the cell surface with HA (Fig. 2C). Interestingly, pretreatment of SaOS-2 CD44s cells with Hermes1 antibodies inhibited their adhesion to HA to a percentage comparable to that of control SaOS-2 EV cells, indicating that CD44s blocking was almost complete. It is also important to note that pretreatment of SaOS-2 EV and of SaOS-2 CD44s R41A cells with Hermes1 did not affect short-term adhesion of the two cell lines to HA (not shown).

Tumor cell migration, another indicator of metastatic potential in vitro, was investigated with the CD44s expression-manipulated cells in a transwell migration assay. The migration rates of SaOS-2 CD44s and SaOS-2 CD44s R41A cells were fourfold (p < 0.05) and threefold (p < 0.05) higher than that of SaOS-2 EV cells (Fig. 2D), suggesting that the CD44s expression-related increase in migration activity was not dependent on CD44s/HA interactions. On the other hand, effects of CD44s overexpression on the proliferation of SaOS-2 cells were not observed. The calculated doubling times were 38.2 ± 1.6 hours for SaOS-2 EV cells, 42.1 ± 4.7 hours for SaOS-2 CD44s, and 46.7 ± 5.2 hours for SaOS-2 CD44s R41A cells (Fig. 2E). Thus, overexpression of CD44s in SaOS-2 cells enhanced the in vitro metastatic properties such as adhesion and migration in an HA-dependent and HA-independent manner, respectively, without affecting cell proliferation.

The tissue microarray analysis of OS resections also suggested that expression of CD44 in primary tumor tissue is related to and may even directly enhance the resistance to commonly used chemotherapeutics in OS patients. The results presented here of cytotoxicity experiments with CD44s and CD44s R41A overexpressing and control SaOS-2 EV cells supported this hypothesis. The half-maximal growth inhibitory concentration (IC50) of cisplatin was 2.4-fold higher (p < 0.01) in SaOS-2 CD44s cells than in control SaOS-2 EV cells (Fig. 2F). There was also a tendency (p > 0.05) for increased IC50 in SAOS-2 CD44s R41A cells when compared to that of SAOS-2 EV cells, implying that increased cisplatin resistance was only partially HA-dependent.

CD44s enhances intratibial primary tumor growth and pulmonary metastasis in a xenograft OS mouse model in an HA-dependent manner

The in vitro experiments showed that CD44s, when overexpressed in SaOS-2 cells, enhances adhesion and migration, cellular processes essential for the metastatic progression. We therefore compared the growth and metastasis of SaOS-2 EV, SaOS-2 CD44s, and SaOS-2 CD44s R41A cell line-derived intratibial primary tumors in SCID mice. Osteoblastic lesions were first observed by X-ray in mice that were injected with SaOS-2 CD44s cells 50 days after tumor cell injection (Fig. 3A). SaOS-2 EV and CD44s R41A cell-derived tumors developed more slowly and the first osteoblastic lesions became visible 2 weeks after those in mice injected with SaOS-2 CD44s cells. At the end of the experiment, on day 90 after tumor cell injection, SaOS-2 CD44s cell-derived tumors showed more extensive bone structures on X-ray images than tumors of SaOS-2 EV and SaOS-2 CD44s R41A cells. The animals became moribund in the 12th week after tumor cell injection and were subsequently euthanized. Finally, CD44s xenografts with a mean primary tumor volume of 226 ± 29 mm3 were significantly (p < 0.05) larger than SaOS-2 EV (121 ± 25 mm3) and SaOS-2 CD44s R41A (130 ± 31 mm3) xenografts (Fig. 3B). Moreover, the ex vivo analysis of whole mounts of lungs revealed 2.5 and 2.2 times higher numbers of X-gal–stained macrometastases in SaOS-2 CD44s than in SaOS-2 EV and SaOS-2 CD44s R41A tumor-bearing mice, respectively (p < 0.01; Fig. 4A and B). Similarly, the mean number of micrometastases was 2 times and 2.2 times higher on the lungs of mice with SaOS-2 CD44s cell-derived tumors than on the respective lungs dissected from SaOS-2 EV and SaOS-2 CD44s R41A tumor bearing animals, respectively (p < 0.05; Fig. 4C). The significant differences in tumor size and in the mean number of lung macro- and micrometastases in SaOS-2 CD44s compared to SaOS-2 CD44s R41A tumor-bearing mice clearly indicated that the observed CD44s-promoted enhanced malignancy of SaOS-2 CD44s cells was dependent on interaction with HA.

Figure 3.

Overexpression of CD44s in SaOS-2 cells enhances intratibial osteoblastic primary tumor growth in SCID mice in an HA-dependent manner. (A) Representative X-ray images of tumor-bearing legs taken on indicated days after intratibial injection of cells transduced with empty vector (EV) (11 mice) or cells expressing CD44s (8 mice) or CD44s R41A (9 mice). (B) Mean ± SEM primary tumor volume in respective mice 12 weeks after tumor cell injection. *p < 0.05.

Figure 4.

CD44s overexpression in SaOS-2 cells enhances pulmonary metastasis of intratibial primary tumors in an HA-dependent manner in SCID mice. (A) Representative images of blue X-gal–stained metastatic nodules in the lungs of mice bearing tumors derived from empty vector (EV)-transduced, CD44s-expressing, or CD44s R41A–expressing cells. Bars = 250 µm. Quantification of pulmonary macrometastases (B) and micrometastases (C) on whole-lung mounts prepared after sacrifice in week 12 after tumor cell injection. Values are the mean ± SEM; *p < 0.05.

Discussion

Osteosarcoma is the second leading cause of cancer-related death in pediatric age group and young adults.30 This high mortality is due to development of pulmonary metastases that are already detectable in 15% to 20% of the patients at the time of diagnosis. In order to devise successful disease management strategies and to improve the survival rate of OS patients, it is essential to gain a profound understanding of OS and metastasis progression as well as associated chemoresistance mechanisms. Identification of novel OS molecular markers that will serve as therapeutic targets is crucial for the design of effective treatment of this devastating disease.

In the present study, we investigated the prognostic value of CD44 expression for OS patients' outcome, as well as the biological relevance of CD44-HA interactions in an orthotopic xenograft OS mouse model. In our tissue microarray analysis, we evaluated the predictive value of total CD44 protein expression. CD44s is probably the main isoform expressed in OS samples, because it is the predominant isoform expressed in several established OS cells lines (not shown) that are of mesenchymal origin. This is in line with a study conducted by Brown and colleagues31 in which the authors demonstrate that a switch in expression from CD44 variant isoforms to CD44s is essential for epithelial-mesenchymal transition and that mesenchymal CD44s is upregulated in advanced human breast carcinomas. On the other hand, our suggestion contrasts with two immunohistochemical studies that found variant CD44 isoforms to be correlated with poor survival of OS patients after performing analysis on tumor samples derived from 50 and 39 patients, respectively.18, 19 We did not find a significant correlation between CD44 expression and overall survival, which is in good agreement with recent reports of Boldrini and colleagues,20 in which 34 OS specimens were immunohistochemically analyzed, and with the report of and Xu-Dong and colleagues,32 in which RT-PCR was applied to study the mRNA levels of CD44 in 32 OS patient samples. Interestingly, Xu-Dong and colleagues32 observed that OS patients with high expression of CD44 gene were more prone to have metastases. Our Kaplan-Meier analysis, however, revealed a significant correlation between CD44 expression in OS surgical tumor specimens of patients with a poor response to preoperative chemotherapy and shorter metastasis-free survival. Moreover, patients with metastasis and CD44-positive tumors had a significantly shorter overall survival when compared to patients with metastatic disease who had CD44-negative tumors. We analyzed the surgical resections, although the initial biopsy samples are generally considered to be the preferable type of specimen to study in order to determine the prognostic value of a given biomarker, because they contain tissue not previously exposed to therapy. Nevertheless, the limitation of using OS biopsy material relates to its small sample size. Because OS displays a high degree of heterogeneity, even within the same patient,4 very small tissue samples may not adequately reflect the biological properties of the entire tumor; therefore, larger resected tumor specimens collected at the time of definitive surgery were considered more appropriate for this analysis. However, the potential disadvantage of tissue from resections is that they had been treated with polychemotherapy, which may affect protein expression. On the other hand, and most importantly, the resected tumor samples are routinely used in current clinical practice to assess the response to chemotherapy, one of the most significant and most commonly used prognostic determinants in OS patients. Despite the limited number of patient samples analyzed due to the low incidence of OS, our data indicate that the expression of CD44 in OS primary tumor tissue is associated with worse prognosis for nonresponders and for patients with metastatic disease. The finding that CD44 expression correlates with poor outcome in both groups of patients is not unexpected when one keeps in mind that patients with a poor response to preoperative chemotherapy are also at higher risk for developing metastases, which was also observed in our cohort of patients. In the emerging era of personalized oncology care, the ultimate goal is to tailor therapy to individual patients according to distinct characteristics of their tumors in order to enhance clinical efficiency while minimizing the adverse toxic side effects of antineoplastic drugs.33 We suggest CD44 as a molecular marker that can be used to identify subgroups of patients with different prognosis and that can further be a basis for individualized treatment. Stratification of patients into different risk categories may guide the choice of postoperative chemotherapy. Our findings identify CD44-positive patients within nonresponders and patients with metastatic disease as high-risk patients who therefore may be considered as candidates for intensified chemotherapy or novel therapeutic strategies.

Based on our tissue microarray analysis, we speculated that CD44 may be an important player in the metastatic behavior and chemoresistance of OS cells. In fact, making use of a xenograft orthotopic mouse model of OS that more faithfully reproduces the clinical features of human disease than subcutaneous models, we demonstrated that CD44s has a tumor- and metastatic-promoting activity upon overexpression in a human osteoblastic SaOS-2 cell line. The malignancy-enhancing effect of CD44s was HA-dependent and was reflected in the significant increase of primary tumor volume and the numbers of pulmonary micro- and macrometastases when compared to that of SaOS-2 EV and SaOS-2 CD44s R41A cells. Apparently, CD44s affected both dissemination of tumor cells and their growth re-establishment at the secondary site. In vitro experiments showed that overexpression of CD44s promoted adhesion to HA and facilitated migration of SaOS-2 cells. Interestingly, Richter and colleagues34 showed that CD44 expressed on various tumor cells interacts with HA under physiologic flow conditions, analogous to that of lymphocytes. They reported deposition of HA in the periarterial space in the mouse lung, which we also detected (not shown). Hence, the HA-dependent increase in the metastatic potential in SaOS-2 CD44s cells most likely relies on the adhesion of CD44s overexpressing cells and HA within pulmonary microvasculature. The observed increase in migration rates upon overexpression of CD44s did not rely on the binding capacity to HA. This observation is not surprising given the fact that CD44 also functions as a co-receptor of several receptor tyrosine kinases, such as c-Met (Mesenchymal-Epithelial Transition Factor) and members of the ERBB or Epidermal Growth Factor receptor family, and can thereby modulate the signaling of associated growth factor receptors.10

We did not observe any effects of CD44 overexpression on proliferation in vitro, as opposed to the in vivo experiments in which CD44s overexpressing cells formed larger tumors in comparison to EV and CD44s R41A cells upon intratibial injection. These findings once again show that in vitro conditions do not fully replicate the physiological situation and highlight the major impact of tumor microenvironment on tumor development, which is already well established. Furthermore, additionally to promoting migration, CD44s induced in vitro chemoresistance of SaOS-2 cells to cisplatin, which is in good agreement with our tissue microarray results. The CD44s effect on response to this cytotoxic drug appeared to be only partially dependent on HA binding ability, which implicates involvement of additional mechanisms. In the past years, several lines of evidence have indicated that HA and CD44 promote chemoresistance in a wide spectrum of tumor cell types, including breast, lung, head, and neck carcinomas and lymphoma.35 It was shown that CD44/HA interaction regulates the expression of drug transporters, such as MDR1 (Multidrug Resistance Protein 1 or P-glycoprotein)17, Multidrug Resistance-associated Protein 2 (MRP2),36 and Breast Cancer Resistant Protein (BCRP).37 For example, in breast and ovarian cancer cells, MDR1 expression is regulated by the HA/CD44-activated Nanog-Stat-3 signaling pathway.38 Furthermore, HA/CD44 interaction induces ankyrin binding to MDR1, leading to the efflux of chemotherapeutics. In colon carcinoma cells, CD44-dependent chemoresistance is mediated through Lyn kinase and phosphoinositide 3-kinase/Akt.39 CD44 is associated with multidrug transporters also in malignant peripheral nerve sheath tumor cells, and the disruption of CD44/HA interactions abrogated drug resistance in these cells.40 Recently, Xu and colleagues16 reported evidence that CD44 attenuates the activation of the Hippo signaling pathway and consequently promotes resistance of glioma cells to cytotoxic agent–induced stress. At this point, it is important to notice that in numerous types of human carcinomas CD44 is a common marker for cancer stem cells (CSC) which are highly resistant to chemotherapy.15 The underlying mechanism of CD44 induced chemoresistance in OS still remains to be elucidated. Future research should aim at confirming CD44-mediated chemoresistance in animal models of OS.

In summary, our study identified CD44 as an additional negative predictor for OS patients' outcome that enables stratification of patients into subgroups, which may lead to more efficient personalized therapy. With the SaOS-2 model we developed a system that allows investigating the relevance of individual CD44 isoforms for the tumor biology of osteoblastic OS in a cellular background of insignificant endogenous CD44 expression. Findings presented here underscore the important role of CD44s/HA interaction in determining tumor malignancy in experimental OS. Taken together, our results point to CD44/HA interaction as a promising target for therapeutic intervention in OS.

Disclosures

All authors state that they have no conflicts of interest.

Acknowledgements

This study was supported in part by the Zurcher Krebsliga (Zurich, Switzerland), the Swiss National Science Foundation (SNF Grant Nr.31003A-120403), the University of Zurich, the Schweizerischer Verein Balgrist (Zurich, Switzerland), the Walter L. & Johanna Wolf Foundation (Zurich, Switzerland), and the Lydia Hochstrasser Stiftung (Zurich, Switzerland). We thank PD Dr. Beata Bode (University Hospital Zurich, Switzerland) for the support in generating and staining of the tissue microarray and Prof. Ivan Stamenkovic (University of Lausanne, Switzerland) for helpful discussion. We appreciate the excellent technical assistance provided by Josefine Bertz and Christopher Bühler.

Authors' roles: Study design: AG, RM, BF, and WB; Experiment conduct: AG; Animal work: AG, MJEA, CC, and PB; Contributed material: KH; Contributed analysis tools: YL; Data analysis: AG; Manuscript preparation: AG, RM, WB, and BF.

Ancillary