Tumor-associated neovascularization is necessary for the growth and metastatic spread of solid tumors.1 We now know there are a variety of angiogenesis inducers and inhibitors that jointly dictate the extent of tumor vascularization.1 Among these, thrombospondin-1 (TSP-1) was the first naturally occurring inhibitor of angiogenesis to be identified.2 TSP-1, by binding to endothelial cell surface receptor CD36, induces endothelial cell apoptosis through a p59fyn, caspase-3, and p38 MAPK signaling pathway.3 Because of its antiangiogenic properties, TSP-1 overexpression in tumor cells inhibits tumorigenesis in various mouse models, including breast and skin carcinomas, fibrosarcoma, melanoma and glioblastoma.4, 5 In addition, the growth of melanoma and mammary carcinoma cells is increased in TSP1-deficient mice, indicating that the TSP-1 derived from endogenous host cells can also make the tumor microenvironment less permissive for growth.6, 7
Assessing the relevance of such findings to the clinical progression of human tumors presents a challenge. In most human cancers, the expression of TSP-1 is decreased in tumor cells, whereas it is increased in the normal fibroblasts that are associated with the tumor stroma reaction.4, 8, 9, 10, 11 Yet in human bladder, gallbladder, thyroid, endometrial, cervical, esophageal, pancreatic and colon carcinomas, both pro- and antiangiogenic correlations to stromal TSP-1 have been reported.4, 9, 10, 11, 12 Similarly, increased stromal TSP-1 expression has been associated with either good or poor prognosis, pending of the cancer studied.4, 9, 10, 11, 12 In node-negative breast cancer, only one clinical study has examined the prognostic significance of the expression of TSP-1 and it was found to have no prognostic value.13
In an effort to obtain a more integrated understanding of the role of TSP-1 in human breast cancer, we have combined an experimental study in a model of breast tumorigenesis and a clinical study in node-negative and node-positive breast cancer patients to investigate the relationship between TSP-1 expression, angiogenesis and disease progression.
Material and methods
Human breast cancer material
Breast samples were obtained from 77 patients with invasive ductal carcinoma NOS. Formalin-fixed and paraffin-embedded tumor specimens were analyzed by immunohistochemistry as described later in this article.
TSP1-inducible cell line
The mouse fibroblastic NT26 TSP-1-inducible cell line has been described elsewhere.14 Murine TSP-1 production by NT26 cells is repressed in the presence of doxycycline when added in vitro into the cell culture medium (100 ng/ml) or in vivo in the water supply (100 μg/ml) for the animals.14 Conditioned media were prepared from NT26 cells treated or not treated with doxycycline in the absence of other antibiotics. The effects of these conditioned media were then tested in vitro on capillary-like tube formation and on migration and invasion of human MDA-MB-231/B02 breast carcinoma cells. Characteristics of the MDA-MB-231/B02 cell line are described elsewhere.15
Capillary-like tube formation assay
The surface of 24-well plates was coated with the basement membrane matrigel (400 μl/well; 10 mg/ml). Endothelial cells (4 × 104 cells/ml/well) were laid on matrigel-coated plates in the presence of conditioned media from NT26 cells. After a 16-hr incubation at 37°C in a 5% CO2 incubator, the wells were photographed. The total length of the tube network on photographs was measured using a semiautomatic analyzer (Videoplan, Kontron, Munich, Germany) and results were expressed as μm/cm2.
Breast cancer cell migration and invasion assays
Cell migration and invasion experiments were performed using Bio-Coat cell chambers (Becton Dickinson, San Jose, CA), which consist of a 24-well companion plate with the cell culture inserts containing an 8 μm diameter pore size filter. For the cell invasion assay, filters were coated with matrigel (30 μg/filter). The experimental procedure was as previously described.15, 16 Briefly, MDA-MB-231/B02 cells resuspended in RPMI/BSA medium (5 × 104/500 μl) were added to the insert (upper chamber), and serial dilutions of the chemoattractant (NT26 conditioned media) were placed in the lower chamber (750 μl/well). After incubation at 37°C for 8 hr (or 48 hr for cell invasion), the migrant cells on the undersurface of the membrane were fixed, stained and counted. All experiments were run in duplicate and cell migration/invasion was expressed in terms of cells per mm2.
Tumorigenicity assays were performed on 6-week-old female Balb/c nude mice (Iffa Credo, France). Studies involving animals, including housing and care, method of euthanasia and experimental protocols, were conducted in accordance with a code of practice established by the local animal ethical committee in Villejuif, France.
MDA-MB-231/B02 cells were mixed with NT26 fibroblasts previously cultured in the presence of doxycycline (10 ng/ml). A mix of 2 × 106 MDA-MB-231/B02 cells and 3 × 106 NT26 cells per site was injected subcutaneously in the hind quarters of animals in 50% (vol/vol) matrigel. Animals were randomized into 2 groups: 1 group received doxycycline (0.2 mg/ml) in the drinking supply, whereas the other group had no doxycycline. The size of the tumors was determined by external measurements of the tumors in 2 dimensions with a caliper. Volume (V) was estimated as V = L × l2, where L is the widest diameter and l the smallest. As previously reported,14 using this equation, there was a good correlation between the calculated tumor volume and the weight of the tumor at the time of the sacrifice of animals. After weight measurements, tumors were fixed and paraffin-embedded for further immunohistochemical analysis. Blood was collected for measurement of vascular endothelial growth factor (VEGF) in plasma using ELISA kit for human protein (R&D Systems) according to the manufacturer's instructions.
Paraffin-embedded serial tissue sections (5 μm thick) were deparaffinized, rehydrated and immunostained using an automated immunostaining apparatus (NexEs, Ventana, Strasbourg, France) with standardized duration and temperature of all the steps.14 For TSP-1 detection in human and murine tumors, immunostaining was performed with mouse monoclonal antibody MA-II17 and a rabbit polyclonal antibody (Ab-8; Neomarkers, Union City, CA), respectively. For microvessel detection, immunostaining was performed with a rabbit polyclonal antibody against von Willebrand factor (Dako, Trappes, France) or type IV collagen (batch 289 g; Novotec, Lyon, France). VEGF was immunodetected with a rabbit polyclonal antibody (Ab-1; Neomarkers). For the measurement of the proliferative activity, immunostaining was performed with mouse monoclonal antibody against PCNA (clone PC10; Upstate, Lake Placid, NY). Simultaneous detection of TSP-1 and microvessels in each human breast tumor section was performed by double labeling according to the manufacturer's instructions (Ventana). For scoring of the microvessel density, highly vascularized regions of the tumors were selected at a low magnification (100× magnification) and microvessels were counted in 3 nonoverlapping high-power fields (HPF; 400× magnification; area 0.283 mm2 per field).
For preclinical experiments and measurement of microvessel density in human breast tumors, data were analyzed using unpaired Student's t-test. For patients, the association between TSP-1 expression and clinicopathologic variables was assessed by the chi-square test. ANOVA followed by a Fisher's protected least significant difference (PLSD) test was used to examine the association between TSP-1 expression, clinicopathologic variables and microvessel density. Univariate analyses of time to death due to breast cancer and time to recurrence (relapse-free survival) were performed using the Kaplan-Meier method. Complete information on patient survival, recurrences, time and cause of death was available in 68 cases. The median follow-up time for the 68 evaluable patients was 67.5 months (range, 7–155 months). Differences between categories were tested by the log-rank and Breslow-Gehan-Wilcoxon tests. These analyses were performed using the StatView 5.0 software. p < 0.05 was considered statistically significant. All p-values were 2-sided.
Results and discussion
TSP-1 inhibits capillary-like tube formation but not breast carcinoma cell migration and invasion in vitro
TSP-1 in the stroma of invasive breast ductal carcinomas is produced by fibroblasts, but not by tumor cells themselves.8 We have used here a mouse fibroblastic NT26 cell line stably transfected with a tet-off regulatory expression vector for mouse TSP-114 to mimic in our tumorigenesis animal model the stroma reaction usually seen in human breast carcinomas. As shown by both Western blotting and immunofluorescence, the synthesis and secretion of TSP-1 by NT26 cells were repressed in the presence of doxycycline (Fig. 1a). The conditioned medium from NT26 cells drastically inhibited capillary-like tube formation only when TSP-1 expression was induced by doxycycline withdrawal (Fig. 1b). In contrast, serum-free conditioned media from untreated and doxycycline-treated NT26 cells stimulated the invasion (Fig. 1c, left) and migration (Fig. 1c, right) of MDA-MB-231/B02 breast cancer cells regardless of the expression of TSP-1. These results were in agreement with the fact that TSP-1 is an angiogenesis inhibitor2, 3 and that stromal fibroblasts play an important role in breast cancer progression.18
TSP-1 delays onset of tumorigenesis but has no effect on subsequent growth rate of breast tumor xenografts in animals
The subcutaneous coinoculation of MDA-MB-231/B02 cells and NT26 fibroblasts into doxycycline-fed nude mice where TSP-1 was repressed induced the formation of tumors reaching a volume of 2,068 ± 823 mm3 at day 31 (Fig. 2a). In contrast, when TSP-1 was expressed, there was a 64% decrease in tumor volume at day 31 (769 ± 647 mm3; n = 5; p = 0.024; Fig. 2a). Upon TSP-1 expression, there was an initial delay in tumor development of about 7 days. These results were in agreement with previous findings4, 5, 6, 7 showing that TSP1-expressing tumors were smaller than TSP1-negative tumors when compared at the same time point. Accordingly, there is also a decrease in the tumor vascularization of TSP1-expressing tumors.4, 5, 6, 7 However, as exemplified in Figure 2(a, inset), the logarithmic linearization of the growth curves indicated that, once tumors started to grow, the tumor growth rates were similar in the presence or absence of TSP-1, suggesting that TSP-1 no longer affected tumor vascularization. To address this question, volume-matched tumors from untreated and doxycycline-fed animals (harvested at day 36 and day 31, respectively) were analyzed by immunohistochemistry (Fig. 2b). In the presence of doxycycline to suppress TSP-1, TSP-1 was seen only as a thin pale line delineating the lumen of blood vessels. In contrast, in the absence of doxycycline, a striking increase in TSP-1 immunoreactivity in the stroma surrounding unstained clusters of malignant cells was observed (Fig. 2b). However, examination of tumors grown either with or without TSP-1, which were matched for volume, showed equivalent degrees of vascularization (+dox: 33 ± 8 vessels/mm2; −dox: 36 ± 10 vessels/mm2; Fig. 2b). Similarly, the proliferative index in volume-matched tumors was the same, as judged by the PCNA staining (Fig. 2b). The ability of TSP-1-expressing tumors to achieve normal vascularization was probably related to the increased VEGF immunoreactivity observed in these tumors compared to those harvested from animals in which stromal production of TSP-1 had been repressed (Fig. 2b). Supporting this assertion, the VEGF plasma level in TSP-1-producing tumor-bearing animals was significantly increased compared to that observed in animals where TSP-1 has been suppressed (72.9 ± 28 vs. 12.3 ± 13.3 ng/ml; n = 5; p = 0.0046).
In the light of these preclinical results, we suggest that TSP-1 produced by stromal fibroblasts inhibits angiogenesis during the early stage of breast tumor formation, resulting in an initial tumor growth delay. However, upon prolonged turnover in a TSP-1-rich environment, breast tumors can markedly increase their secretion of angiogenic factors such as VEGF that counterbalance the inhibitory effects of TSP-1. This possibility is supported by the similar behavior of fibrosarcoma and glioblastoma tumor cells where prolonged exposure to TSP-1 leads to the emergence of tumor cell variants that overexpress VEGF and/or become resistant to the antimitogenic effect of TSP-1-activated TGF-β.14 Another example is the outgrowth of A431 squamous carcinoma cell variants that become refractory to C225 antiangiogenic therapy, overexpress VEGF and are highly angiogenic.19 Similarly, endostatin induces tumor hypoxia and subsequent upregulation of VEGF in Mca-4 murine mammary carcinomas.20
Association of TSP-1 immunostaining with clinicopathologic classification and microvessel density in human breast tumors
To test the relevance of these laboratory findings to human primary tumors, TSP-1 was examined in 77 breast invasive ductal carcinomas not otherwise specified (NOS). TSP-1 was expressed in the tumor stroma (but not in tumor cells themselves; Fig. 3), reflecting the situation in the experimental tumors growing in nude mice that were analyzed above. Tumor sections were classified as S0 when the stromal TSP-1 immunoreactivity was negligible (Fig. 3a) or did not exceed 50% of the tumor section areas. Tumors with a moderate to strong TSP-1 immunoreactivity that exceeded 50% of the tumor section areas were classified as S1 (Fig. 3b). The relationship between TSP-1 score and clinical and biologic characteristics of breast cancer patients is shown in Table I (TSP-1 score column). As we previously reported for TSP-1 mRNA expression in breast cancer,21 there was no meaningful association between TSP-1 protein expression and any other tumor characteristics, with the exception of vascular density (Table I).
Table I. Clinical and Biological Characteristics, TSP-1 Expression and Microvessel Density in 77 Patients with Invasive Breast Ductal Carcinoma NOS
The immunostaining of blood vessels in tumor sections adjacent to those scored as S0 for TSP-1 expression (Fig. 3a) showed a high microvessel density (Fig. 3c). Conversely, a low microvessel density (Fig. 3d) was consistently observed in tumor sections adjacent to those scored as S1 (Fig. 3b). In this respect, tumors with a moderate to strong TSP-1 immunoreactivity had significantly (p = 0.003) lower vessel counts (26 ± 11 vessels per 400 × HPF; n = 53) than low TSP1-expressing tumors (37 ± 21 vessels per 400 × HPF; n = 22). Double immunostaining of a subset of 31 tumors using antibodies against TSP-1 and von Willebrand factor further demonstrated that, on the same tumor section, a higher microvessel density was consistently observed in TSP1-poor areas (Fig. 3e) when compared to that observed in TSP1-rich areas (Fig. 3f).
Using the ubiquitous VEGF as an example of the many angiogenic stimulators that can be produced by human breast cancer cells,22 we observed a high VEGF expression level in primary human breast tumors and metastatic axillary lymph nodes (not shown). In addition, TSP-1 no longer inhibited angiogenesis in metastatic axillary lymph nodes (Fig. 3g). Indeed, in axillary lymph nodes scored S1, tumor angiogenesis was somewhat (although not significantly) increased (43 ± 19 vessels per 400 × HPF; n = 13) compared to that observed in lymph nodes scored S0 (29 ± 7 vessels per 400 × HPF; n = 6). These results are in agreement with the fact that TSP-1 fails to inhibit tumor lymphangiogenesis,23 and they suggest that metastatic cells have adapted to circumvent TSP-1 inhibition.
Association between TSP-1 expression in primary tumors and clinical outcome
The observation that the overexpression of some angiogenic factors can compromise the antiangiogenic activity of TSP-1 over time is also strongly supported by the fact that TSP-1 expression in primary tumors was associated with a significant increase of disease recurrence in breast cancer patients. Analysis of the node-negative and node-positive patients (n = 68) indicated that the estimated 5-year probability of relapse-free survival was 0.61 (95% CI = 0.46–0.75) for patients with a S1 score and 0.85 (95% CI = 0.86–1.0) for patients with a S0 score (Breslow-Gehan-Wilcoxon p = 0.05). When stratifying according to the lymph node status, TSP-1 expression continued to be significantly associated with disease recurrence in node-positive patients (Fig. 3h). There was no significant association between TSP-1 expression and the overall survival (results not shown).
We have previously reported that in humans a strong stromal TSP-1 expression is present early, occurring in premalignant breast lesions and in in situ carcinomas.8 This information taken together, our results suggest that TSP-1 functions in human breast tumors as we have shown it to function in our experimental tumors described above and in previous work,6 as an early rate-limiting step resulting from its inhibition of angiogenesis. However, data presented here also suggest that TSP-1 produced by stromal cells in the tumor bed can foster the increased expression of angiogenic factors by breast cancer cells and that these tumor cells subsequently have a greater propensity to metastasize successfully to distant organs. These findings did not only extend previous experimental works obtained with fibrosarcoma and glioma cells.14 They show for the first time that a resistance may develop early in human primary breast tumors as a result of high in situ exposure to stromal TSP-1, leading to the development of highly angiogenic and metastatic tumors. Although such an observation does not apparently argue for antiangiogenic therapy, it must be pointed out that selection for drug resistance inevitably occurs as a consequence of tumor regression following effective therapy.24 In this respect, there may be no cells that are sufficiently resistant to tolerate a combination of very effective therapies. This contention is supported by the fact that impairing the synthesis of VEGF severely limits resistance of fibrosarcoma cells to TSP-1 antiangiogenic activity.25 In this line, we believe that breast cancer patients whose in situ TSP-1 expression in primary tumors is high could benefit from an early antiangiogenic treatment (such as anti-VEGF therapy) in combination with conventional treatments in order to minimize the induction of such tumor escape.
The authors thank Dr. Noel Bouck for critical reading of the manuscript and helpful suggestions and Dr. Jack Lawler for the gift of anti-TSP1 antibody MA-II. Supported by grants from INSERM, Association pour La Recherche sur le Cancer (ARC) and the European Commission (LSHC-CT-2004-503049; all to P.C.).