During carcinogenesis, stromal fibroblasts undergo certain changes in concert with their neoplastic neighbors, an interaction that progressively leads to a cancer-associated state. However, despite the increasing appreciation of the importance of stromal/tumor interactions in the progression of cancer, little is known about the factors responsible for regulating the crosstalk between stromal fibroblasts and neoplastic cells. Here we show that the stage of the disease in primary human breast lesions affects p21 expression in the fibroblasts. In stromal fibroblasts of benign fibroadenomas, p21 exhibits a periductal pattern of staining, which is abolished in malignant adenocarcinomas in which p21 immunopositivity exhibits a mosaic pattern that eventually is abolished in more aggressive types of the disease. In order to address the role of fibroblasts’ p21 in tumorigenesis, we have reconstituted MCF7 human breast cancers in mice, with fibroblasts differing in the p21 status. These experiments showed that p21 deficiency in stromal fibroblasts accelerates tumor growth through cell non-autonomous mechanism(s). In addition, even a transient, siRNA-mediated p21 suppression in fibroblasts sufficiently stimulates MCF7 and MDA-MB-231 growth in vivo. We propose that p21 regulation is intimately linked with the ability of stromal cells to affect tumor growth.
Fibroblasts represent the major cellular component of cancer-associated stroma. Although, however, their role in accelerating cancer growth and in certain instances even to cause malignant conversion has been demonstrated, the molecular factors regulating this process remain largely unknown (1–4). Recently, by performing a series of tumor reconstituition experiments, involving the co-inoculation of breast cancer cells with fibroblasts differing in their p53 status, we showed that p53 mutations in stromal fibroblasts exert a positive effect on cancer growth (5). This finding was subsequently confirmed by others in a transgenic mouse model of prostate cancer involving the assessment of primary malignant lesions (6). Noteworthy, p53 mutations represent a detectable genetic lesion in the stroma of primary breast and probably other cancers, suggesting that at least in certain cases the transition of stromal fibroblasts into a cancer-associated state may be associated with the acquisition of p53 deficiency (7–9).
Although, however, the accumulation of certain genetic lesions by the stromal fibroblasts provides an attractive molecular mechanism that may account for the transition of the fibroblasts into a cancer-associated state, it can be challenged by the fact that its operation should involve the clonal selection of the mutant fibroblasts. The latter, although formally possible, appears unlikely during the early stages of the disease at which a rapid and more global response should be evoked in the stroma. Therefore, it is conceivable that other mechanisms, operating at the level of the (de-)regulation of gene expression, might be involved in this process. In the present study, we tested the hypothesis that differential expression of p21waf1/Cip1 (p21) in stromal fibroblasts is involved in tumorigenesis by non-autonomous mechanisms. Our hypothesis was based on two lines of evidence: (i) In part, the effects of p53 in cell cycle are mediated by p21 and thus, an overlap between the consequences of these two cell cycle regulators is anticipated; and (ii) p21 is only rarely—if ever—mutated in primary human tumors; however, it is subjected to complex, yet unclear, regulation during carcinogenesis (10–12).
MATERIALS AND METHODS
Cells, mice and xenograft development
Human mammary epithelial adenocarcinoma cells MCF7 and MDA-MB-231 were originally obtained from American Type Culture Collection (VA, USA) and maintained in DMEM containing 10% FBS and antibiotics/antimycotics. SCID and p21-deficient mice (13) were originally obtained by Jackson laboratories (ME, USA) and subsequently maintained in our laboratory. Care of animals was in accord with institutional guidelines. For xenograft development, 1.5 × 106 MCF7 cells and 0.5 × 106 MEFs per mouse were re-suspended in 0.2 ml of mixture at 1:1 ratio of serum-free DMEM and ice-cold Matrigel (BD Bioscience) and then injected s.c. into SCID female virgin mice, ∼6 weeks old. Subsequently, animals were observed at least twice per week for tumor development. Then, either the time period until the onset of palpable tumors was scored or upon tumor onset, tumor diameter was assessed by using a microcaliper. Animals were sacrificed when tumor exceeded 10% of body weight or when mice showed signs of distress.
SiRNA and western blot analysis
Suppression of p21 in the fibroblasts was achieved by transfecting wild-type MEFs with siRNA specific for mouse p21 (clone ID 160142, Ambion) according to the manufacturer's instruction. Cells were lysed 48 h after transfection by using RIPA reagent and total protein was subjected to western blot analysis. Antibodies for p21 and actin were obtained from Sigma.
Primary human breast tumors, histology and immunohistochemistry
For histological analyses, xenografts were fixed in 10% formalin, paraffin embedded and stained with hematoxylin/eosin for light microscopy. Immunohistochemistry for Ki-67 was performed with a rabbit polyclonal antibody obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA) by using the Kwik-DAB kit (ThermoShandon, Pittsburgh, PA, USA), according to the manufacturer's instructions. Before observation, a weak counter stain with hematoxylin was performed. Primary human breast tumors, including six specimens from fibroblastic disease, 20 fibroadenomas (FAs), five ductal carcinomas in situ (DCIS) and 32 invasive ductal adenocarcinomas (DCs), were selected from the archives of Aretaieion University Hospital on the basis of being rich in stromal fibroblasts. Then, 0.5 µ sections were subjected to immunohistochemical analysis for p21, p53 and vimentin using antibodies obtained from Sigma followed by the application of the Kwik-DAB kit. Images shown were obtained by Pro-image Analysis Software (Media Cybernetics, Inc., MD, USA).
Statistical analysis of the animal experiment results was performed by using Student's t-test, whereas the results of human tumor analysis have been analyzed by the χ2 test.
Initially, we have attempted to assess the growth of human cancer cells in mice suffering from severed combined immunodeficiency (SCID) that differed in the status of p21. Unfortunately, double mutant SCID/p21-null mice exhibited increased mortality prior to the age of 8 weeks old, preventing us from performing this experiment (data not shown). Therefore, we have concentrated on evaluating the role of p21 directly in fibroblasts. To address the role of p21 of stromal fibroblasts in tumor growth, we carried out tumor reconstitution experiments in which we evaluated the behavior of cancer cells that prior to their inoculation in mice were mixed with embryonic fibroblasts that differed in the status of p21. For tumor inoculation, we have used SCID mice. We chose this particular strain of immuno-incompetent mice because these animals can be used for tumor transplantation studies, have a normal reproductive cycle and can support the growth of MCF7 cells in addition to other hormone-dependent cells (5,13). Furthermore, we have selected MCF7 cells because alone they are hardly tumorigenic in SCID mice, whereas their tumorigenicity is dramatically increased when the MCF7 inoculates include fibroblasts, reflecting perhaps a sensitivity to stromal factors (5 and our unpublished observations). The experimental strategy we adopted is shown in Figure 1A. Reconstitution of mammary adenocarcinoma in SCID mice by fibroblasts lacking p21 and MCF7 human breast cancer cells resulted in tumors that were growing faster than those derived by inoculating MCF7 cells and wild-type fibroblasts (Fig. 1B, see legend and methods for more details).
Consistently with their increased growth rate, immunohistochemical analysis of the tumors using Ki67, a marker for proliferating cells (14), indicated that MCF7 xenografts reconstituted with p21-null fibroblasts contained more actively dividing cells than those reconstituted with wild-type p21 stromal fibroblasts (Fig. 1B, insets). Similar results were obtained when p21 was knocked-down by siRNA at 15 nm in wt fibroblasts and tumors were reconstituted with MDA-MB-231 breast cancer cells in addition to the MCF7 cells (Fig. 2).
In order to confirm the biological relevance of this paracrine modulation of tumor growth by fibroblasts' p21 in our experimental system, we assessed the expression of stromal p21 in a set of 63 primary human breast tumors, including six specimens from fibroblastic disease, 20 FAs, five DCIS and 32 invasive DCs, which were selected on the basis of being rich in stromal cells. Our results show that although in the fibroblastic disease specimens p21 immunopositivity was minimal in the stromal fibroblasts (one of six), in FAs it increased to 60% (12 of 20, P < 0.001)) and subsequently decreased progressively in grade II (55%, P < 0.001) and grade III (17%) adenocarcinomas (P < 0.005) (Figs 3 and 4 and Table 1). Besides, however, these differences in the overall staining intensity and the percentage of positive cells between the benign and the invasive breast lesions tested, an apparent difference that became readily detectable was related to the localization of the p21 immunopositive cells. In the FAs, immunopositivity was localized predominantly in the stromal fibroblasts adjacent to the epithelial cells, whereas in the invasive cancers it was random and independent of the relative position of the fibroblasts to the epithelial cells (P < 0.001). In FAs, only in one out of 12 positive specimens positivity was diffuse and mosaic when compared with the 10 of 11 grade II and all two grade III adenocarcinomas analyzed (Table 1). Furthermore, positivity for p21 in stromal fibroblasts correlated with p21 expression in the epithelium (P < 0.001) (Fig. 5).
|FD||FA||DCIS||DC II||DC III|
|FD||FA||DCIS||DC II||DC III|
The pattern of staining, being diffuse (mosaic) or localized (periductal), is shown.
A subset of the specimens, 16 FAs, 14 grade II tumors and seven grade III tumors, was stained for p53. Consistent with previous findings, p53 immunopositiviy was detectable in the stroma of both benign and malignant breast tumors (Fig. 3). Immunopositivity was increased in the epithelium in more advanced breast cancers when compared with the FAs. p53 immunopositivity in the stroma was not associated with the type of the disease (Fig. 6). However, p53 and p21 immunopositivity were slightly associated in the epithelium (P < 0.01) but not in the stroma of the specimens (Fig. 5). Finally, for both p21 (P < 0.001) and to a lesser extend p53 (P < 0.05), immunopositivity correlated in the stromal and the epithelial component of the same specimens (Fig. 5).
Selected specimens were also stained for vimentin, which represents a widely accepted fibroblast marker, in order to identify fibroblasts in the stromal tissue of the tumors (Fig. 3).
Certain genetic lesions, exemplified by the inactivation of p53 tumor suppressor, account for a subset of molecular alterations seen in stromal fibroblasts during carcinogenesis (7,9). These mutations have been shown capable of affecting both the latency of tumorigenesis and the efficacy of anticancer therapy (15). However, considering that the activation of stromal fibroblasts occurs early in the development of the disease, the clonal selection of fibroblasts carrying such genetic lesions poses some considerations related to the extent of the contribution of such mechanism. The latter is due to the fact that such mechanism must be operational early in carcinogenesis, even before malignant conversion and clonal growth of the cancer cells occur. That p21 operates—though not exclusively—downstream of p53 in combination with the observation that it is not targeted by mutations but is subjected to complex regulation during carcinogenesis, raises the possibility to play an important role in the control of the response of stromal fibroblasts in neoplastic growth. In the present study, we confirmed this hypothesis, providing evidence that p21 in fibroblasts modulates the profile of tumorigenesis by a non-autonomous mechanism in a xenograft model of the disease. The fact that even a transient, siRNA-mediated suppression of p21 sufficiently stimulates tumor growth suggests that the role of stromal fibroblasts’ p21 is particularly important in the very early stages of xenograft growth. Furthermore, we have provided evidence that the onset of a benign neoplastic lesion, such as the FA but for reasons unclear to us not from the fibroblastic disease, triggers the induction of p21 expression adjacent to the epithelial cells. This expression pattern of p21 is abolished, however, when lesions become malignant, as indicated by the progressive elevation of diffuse staining in grade II adenocarcinomas and its final abolishment in grade III tumors. Taking together this observation on the p21 expression in stromal fibroblasts in primary breast lesions, with the negative role of fibroblasts’ p21 in tumorigenesis, our results indicate that during very early stages of the disease when the tumor is still benign, a state exemplified by the FAs, stromal p21 is induced, rendering fibroblasts protective against neoplastic growth. This is also consistent with the observation that fibroblasts distal from the neoplastic lesion in FAs as well as those located closely to the malignant epithelial cells in invasive carcinomas both express low levels of p21, despite that they possess distinct—and in fact opposite—activities with respect to their role in carcinogenesis. How p21 in stromal fibroblasts affects tumor growth by paracrine mechanisms remains obscure; however, it is unlikely to contribute directly to the CAF-like phenotype as smooth muscle actin expression, a landmark of CAFs, remains unaffected following p21 knock-down (data not shown).
Collectively, our results attribute a direct role in p21 of stromal fibroblasts in tumorigenesis. Whether these non-autonomous effects of p21 are reversible implying certain therapeutic implications or the tumor enters a state that becomes independent of stromal p21 expression remains to be seen.
This work was supported by the grant EPAN (05NONEU-13) from GSRT, the KESY Oncology program 05 from the Greek Ministry of Health and Kapodistrias 06 from ELKE, University of Athens.
We are grateful to the personnel of the Laboratory of Experimental Surgery (Director, Prof. D. Perrea) for excellent animal housing care.
Conflict of Interest statement. None declared.