Cellular Mechanisms for Low-Dose Ionizing Radiation–Induced Perturbation of the Breast Tissue Microenvironment

The stromal microenvironment in which the tumor cells develop controls many steps of tumorigenesis (1). Normal stroma can suppress tumor growth by imposing growth constraint on carcinoma whereas tumor stroma can promote tumor progression. One of the most notable changes in tumor stroma is the activation of stromal fibroblasts, which respond to invading cancer cells by adopting distinct phenotypic and functional changes that render them the ability to stimulate the growth of cancer cells (2, 3). Interestingly, besides the stimulation provoked by cancer cells, fibroblasts can also be activated by growth factors or chemical carcinogens. Once activated, stromal fibroblasts can secrete various trophic factors and extracellular matrix remodeling enzymes leading to perturbation of local microenvironment (3, 4).

Senescence is an age-associated event. Because of the irreversibility, senescent cells persist over time and may accumulate in aging tissues (5). Ample evidence indicates that cells undergoing replicative senescence increase their production of degradative enzymes, cytokines, and growth factors, which closely resemble carcinoma-activated fibroblasts (6, 7). In light of this secretory phenotype, it has been proposed that senescent cells accumulated in the stroma may serve as a means of stromal activation (8). In addition to replicative senescence, recent evidence indicates that cumulative stress to tissues may serve as an alternative route for the gradual accumulation of senescent cells in the stroma, as fibroblasts respond to a variety of chronic stress insults, such as oxidative stress, UV, ethanol, inflammatory cytokines, and various genotoxic agents, by engaging in a senescence-like phenotype (9). Cellular senescence thus may represent a long-term sequel of subcytotoxic stress, which entails genotypic and phenotypic changes specific to different stress insults (5).

Using low-dose ionizing radiation as a model of environmental stress, we show that human mammary fibroblasts were induced to display a senescence-like phenotype in an accumulative and dose-dependent manner. By use of a three-dimensional coculture system to model at the cellular levels the critical interactions of different mammary cell types with their neighbors and with their environment, we provide a direct experimental proof that ionizing radiation–induced senescence-like fibroblasts can dysregulate three-dimensional growth characteristics of mammary epithelial cells at different transformation stages. Our results suggest that protracted low-dose ionizing radiation exposures can foster an oncogenic environment that disrupts mammary morphogenesis and compromises growth regulations operated at the tissue architectural level.

Cell culture and materials. Primary human mammary fibroblasts prepared from reduction mammoplasty tissue specimens, a gift from Dr. R.A. Weinberg (Whitehead Institute for Biomedical Research, Cambridge, MA), were maintained in DMEM supplemented with 15% fetal bovine serum (FBS). Human mammary fibroblasts at passages 4 to 6 were used for the experiments. MDA-MB-231 cells (American Type Culture Collection, Rockville, MD) were maintained in DMEM supplemented with 10% FBS. MCF-10A cells, a gift from Dr. J.S. Brugge (Harvard Medical School, Boston, MA), were grown and manipulated as described (10). Growth factor reduced Matrigel, high concentration rat tail collagen I, and mouse collagen IV were purchased from BD Biosciences (San Diego, CA). Function-inhibiting antibodies against α3 integrin (P1B5), α5 integrin (P1D6), α6 integrin (GoH3), β1 integrin (P4C10), and β4 integrin (ASC3) were all purchased from Chemicon (Temecula, CA). PD98059, cytochalasin D (Sigma, St. Louis, MO), LY294002, rapamycin, and matrix metalloproteinase (MMP) inhibitor II (Calbiochem, San Diego, CA) were dissolved in DMSO. N-Acetylcysteine (Sigma), nuclear factor κB (NFκB) SN50, and tissue inhibitor of metalloproteinase (TIMP)-1 (Calbiochem) were dissolved in PBS. FITC-conjugated anti–pan-cytokeratin monoclonal antibody (mAb), anti-actin mAb, and propidium iodide were purchased from Sigma; anti–Bcl-X_L mAb from Santa Cruz (Santa Cruz, CA); anti-Beclin mAb from BD Biosciences; anti–Thy-1 mAb and anti-p16 mAb from Calbiochem; anti–phospho-Akt (Ser473) mAb from Cell Signaling (Beverly, MA); Alexa Fluro-488–labeled phalloidin, Alexa Fluor-488–labeled goat anti-rat or anti-mouse immunoglobulin G (IgG), and DQ collagen I and IV from Molecular Probes (Eugene, OR).

Ionizing radiation treatment and induction of senescence-like phenotypes. Ionizing radiation was administered with a Philips MG160 high stability constant potential X-ray system (Philips Industrial X-ray, Hamburg, Germany). Detailed experimental design with respect to the dose and treatment schedule is described in Supplementary Fig. S1. Senescence-associated β-galactosidase activity was detected as described (11). The population doubling rates were calculated as log 2(D₇ / D₀), where D₇ is the density of cells after being cultured in DMEM plus 15% FBS for 7 days and D₀ is the density of cells at the time of seeding. For consistency, senescence-like fibroblasts induced by the same ionizing radiation regimen (5 cGy every 12 hours; total 10 Gy) were used for all the coculture experiments.

Retrovirus-mediated gene transfer and silencing. Amphotrophic retrovirus was produced by transfecting pBABE.hygro Bcl-X_L (J. Yuan, Harvard Medical School, Boston, MA), pLZRS.neo Akt (K179M), pLZRS.neo MyrAkt (V.M. Weaver, University of Pennsylvania, PA), or pLZRS.blasticidin N17Rac1 (P. Marinkovich, Stanford University, Stanford, CA) cDNA into Phoenix ampho cells (G. Nolan, Stanford University, Stanford, CA). PCG-gagpol and PCG-VSV-G (R. Mulligan, Harvard Medical School, Boston, MA) were cotransfected to improve titer. MCF-10A cells were infected with pBABE.hygro Bcl-X_L retrovirus and infected cell populations were selected using 300 μg/mL hygromycin (Calbiochem). Cells were then treated with culture supernatant from the amphotropic PA317-LXSN16E7 packaging cell line (D.A. Galloway, Fred Hutchinson Cancer Research Center, Seattle, WA) and selected using 300 μg/mL G418 (Invitrogen, Carlsbad, CA). A stable down-regulation of Beclin-1 or ATG-7 was achieved by a retrovirus-mediated RNA interference system (pSUPER.retro.puro, OligoEngine, Seattle, WA) using oligonucleotide sequences previously described (12). Infected cell populations were selected using 1 μg/mL puromycin. MDA-MB-231 cells were infected with retrovirus produced from pLZRS.neo Akt (K179M), pLZRS.neo MyrAkt, or pLZRS.blasticidin N17Rac1 and selected using 1 mg/mL G418 or 8 μg/mL blasticidin (Invitrogen).

Three-dimensional coculture assay. The three-dimensional heterotypic coculture assay was modified from an overlay three-dimensional culture method as described (10). MCF-10A cells, nonsenescent human mammary fibroblasts, and ionizing radiation–induced senescence-like human mammary fibroblasts were maintained in DMEM/F12 media containing 3.5% horse serum, 5 ng/mL epidermal growth factor, 0.5 μg/mL hydrocortisone, 100 ng/mL cholera toxin, 10 μg/mL insulin, and antibiotics for 72 hours, after which they were trypsinized and harvested in serum-free trypsin-neutralizing solution (Cambrex, Walkersville, MD) and then resuspended in DMEM/F12 supplemented media plus 2% growth factor reduced Matrigel. Nonsenescent fibroblasts (6,000/cm²) or senescence-like fibroblasts (10,000/cm²) and MCF-10A cells (600/cm²) were then seeded in eight-well (for morphologic and immunofluorescence analysis) or four-well (for cell quantification or immunoblotting analysis) chamber slides (Nalge Nunc, Naperville, IL) coated with 60 or 150 μL of growth factor reduced Matrigel and type I collagen mixed at a 3:1 ratio. Type I collagen (final concentration, 2.0 mg/mL) was included in the matrix to facilitate the acquisition of spindle-shaped morphology and migration of human mammary fibroblasts in three dimensions. The cultures were kept in 5% CO₂ at 37°C and were replenished with fresh medium every 3 days for 12 days followed by more frequent medium changes (every 2 days) thereafter. MCF-10A cells were seeded at a low density (∼400 μm between neighboring cells) to interrogate the growth characteristics of individual cells. Nonsenescent fibroblasts slowly proliferated in the above culture condition and therefore were seeded at a lower concentration such that the cell concentrations of nonsenescent fibroblasts and senescence-like fibroblasts were approximately the same at the end of coculture (i.e., day 20).

To recapitulate the stromal microenvironment of the cancer invasion front (13), an extracellular matrix gel mix comprising 50% growth factor reduced Matrigel and 50% type I collagen (final concentration, 4.0 mg/mL) was used for the coculture involving breast carcinoma cells. Malignant mammary epithelial cells and human mammary fibroblasts were maintained in low-serum growth medium containing DMEM/F12 medium plus 1% FBS and antibiotics as monolayers for 72 hours, after which they were trypsinized and harvested in trypsin-neutralizing solution and resuspended in low-serum growth medium containing 2% growth factor reduced Matrigel. Nonsenescent fibroblasts or senescence-like fibroblasts (6,000/cm²) and malignant mammary epithelial cells (1,000/cm²) were then seeded in extracellular matrix gel–coated chamber slides and the culture was maintained in 5% CO₂ at 37°C with medium change every 3 days.

Microarray and reverse transcription-PCR analysis. Nonsenescent fibroblasts or senescence-like fibroblasts were maintained in low-serum growth medium for 72 hours, after which total RNA was extracted and cDNA microarray analysis was then done as described (14). Reverse transcription-PCR (RT-PCR) analysis was conducted on selected metalloproteinases using the Qiagen OneStep RT-PCR kit (Qiagen, Valencia, CA). The primers were designed by Primer3 software (Whitehead Institute for Biomedical Research, Cambridge, MA). Primer sequences are available on request. PCR products were electrophoretically resolved on 2% agarose gels.

Quantitative extracellular matrix degradation assay and in situ imaging of proteolysis. The collagen degradative activity was analyzed by a quantitative three-dimensional fluorometric fluorescein (FITC) release assay as described (15), with minor modifications. Briefly, fibroblasts or mammary epithelial cells (10⁴ cells) were copolymerized with unlabeled collagen I (final concentration, 2 mg/mL) or phenol red–free growth factor reduced Matrigel and 2% DQ collagen I or DQ collagen IV, respectively. After incubation at 37°C for 72 hours, solid-phase matrices and cells were pelleted and FITC released into the supernatant was analyzed by spectrofluorometry. Background fluorescence was analyzed by pelleting cell-free collagen or Matrigel lattices containing 2% DQ collagen. For the in situ imaging of proteolysis and matrix degradation, the coculture was done in extracellular matrix gel containing 25 μg/mL of DQ-collagen IV and the fluorescent degradation products were imaged as described (16).

Functional inhibition assay. Integrin function–blocking antibodies were included in the extracellular matrix gel at the time of gelation as well as in the cell suspensions at a final concentration of 10 or 20 μg/mL (for GoH3). The cultures were replenished with fresh medium containing the respective antibodies every 3 days. MMP inhibitor II (1-10 μmol/L), TIMP-1 (10 μg/mL), PD98059 (10 μmol/L), LY294002 (4 μmol/L), rapamycin (20 nmol/L), or NFκB SN50 (18 μmol/L) was added to the medium on establishment of cocultures and was replenished in the culture medium every 3 days thereafter. Control cultures were treated with vehicle only.

Immunofluorescence analysis and confocal microscopy. Immunofluorescent staining and image acquisition were carried out as described (10). For confocal analysis, the three-dimensional structures were fixed, immunostained, and examined by use of a Leica TCSNT confocal laser scanning microscope (Leica Microsystems, Bannockburn, IL). For selected three-dimensional structures, serial optical sections were collected from the apical surface at 0.8 μm step intervals and rendered in three dimensions using VoxelView software (Vital Images, Minnetonka, MN) to examine the spatial distribution of mammary epithelial cells and human mammary fibroblasts within the three-dimensional organoid structures.

Isolation of mammary epithelial cells and autophagy assay. Single cells were recovered from the extracellular matrix gel following digestion with dispase according to the recommendations of the manufacturer (BD Biosciences). Human mammary fibroblasts were labeled by incubating with MACS Anti-Fibroblast MicroBeads (Miltenyi Biotech, Auburn, CA) and the magnetically labeled human mammary fibroblasts were depleted by the MidiMACS Separation Unit. The recovery and purity of mammary epithelial cells were ∼95% and 100%, respectively. Cell number was quantified by a Coulter Counter Apparatus (Beckman Coulter, Fullerton, CA). The relative doubling rate of mammary epithelial cells, which was used to quantify the growth-stimulating effect of senescence-like fibroblast as compared with that of nonsenescent fibroblast, was calculated as log 2(D_7-SF / D_0-SF) / log 2(D_7-NF / D_0-NF), where log 2(D_7-SF / D_0-SF) is the doubling rate (per week) of mammary epithelial cells cocultivated with senescence-like fibroblasts and log 2(D_7-NF / D_0-NF) is the doubling rate (per week) of mammary epithelial cells cocultivated with nonsenescent fibroblasts. For analysis of autophagy, human LC3 was amplified and ligated into pEGFP-C1 expression vector (BD Biosciences). The autophagy-associated LC3-II was detected as described (17).

Statistical analysis. Data are expressed as mean ± SE. Group comparison was judged by Mann-Whitney U tests. P < 0.05 was considered statistically significant.

Accumulation of senescence-like human mammary fibroblasts induced by protracted low-dose ionizing radiation exposures. Consistent with previous reports that acute exposure of human diploid fibroblasts to ionizing radiation induced permanent cell cycle arrest in a subpopulation of irradiated cells (18), we observed that protracted low-dose ionizing radiation exposures (5 cGy every 12 hours) caused permanent cell cycle arrest in primary human mammary fibroblasts in a dose-dependent manner, which was more effective than single-dose exposures at comparable doses (Fig. 1D). Aside from being growth arrested, the irradiated human mammary fibroblasts also exhibited multiple characteristics of replicative senescence including cytoplasmic senescence-associated β-galactosidase activities, autofluorescence, and increased p16 protein levels (Fig. 1A-C). Treatment of human mammary fibroblasts with or without a cell-permeable antioxidant N-acetylcysteine indicates that ionizing radiation–induced senescence is partially mediated by oxidative stress (Supplementary Fig. S1B). The senescence-like fibroblasts remained growth arrested but viable if fed routinely with serum-containing medium for more than 1 year (not shown).

Senescence-like fibroblasts disrupt mammary epithelial cell–derived ductal morphogenesis. Because normal breast epithelium undergoes repetitive glandular morphogenesis throughout the reproductive life of women, we asked whether ionizing radiation–induced senescence-like fibroblasts could interfere with the morphogenesis of mammary epithelial cells. To mimic the epithelial-stromal interactions in vivo, we cultivated senescence-like fibroblasts with MCF-10A cells in a three-dimensional heterotypic coculture system. MCF-10A cells were used because they are untransformed mammary epithelial cells that still retain the morphogenetic characteristics of normal epithelial cells. As described (19), MCF-10A cells cultured in three-dimensional extracellular matrix gel formed polarized acini with central cell-free lumens at days 15 to 20 (Fig. 2C,, insets). Under the same condition, nonsenescent fibroblasts exhibited a spindle-shaped morphology with elongated pseudopodia that gradually formed interconnections with neighboring fibroblasts (Fig. 2A). When cocultivated with nonsenescent fibroblasts at an optimal ratio and cell density, MCF-10A cells formed randomly oriented branching ductal-acinar structures along with scattered fibroblasts (Fig. 2B,, right), an architecture that closely resembles the mammary glandular epithelium (20). Confocal image analysis revealed that the MCF-10A cell–derived structures were composed of a single layer of polarized epithelial cells and a cell-free lumen (Fig. 2C and D) because of gradual loss of centrally located cells during the morphogenetic process (not shown). Observations on the coculture at early stages revealed that the pseudopodia networks of nonsenescent fibroblasts seemed to serve as platforms for the directional or branching outgrowth of MCF-10A cells (Fig. 2B , left). Consistently, pretreatment of nonsenescent fibroblasts with actin polymerization inhibitor cytochalasin D or low-dose ionizing radiation, which selectively disrupted the formation of pseudopodia, significantly impaired ductal morphogenesis of MCF-10A cells without affecting their ability to form acinus (Supplementary Fig. S2).

Compared with their normal counterparts, senescence-like fibroblasts, when cultured three-dimensionally, were increased in size and displayed small cytoplasmic protrusions without interconnections with neighboring fibroblasts (Fig. 2E). When cocultivated with senescence-like fibroblasts, MCF-10A cells failed to develop into ductal structures but instead formed enlarged and disorganized cysts (Fig. 2F). Despite their large size, these cystic structures were still composed of a single layer of polarized epithelial cells surrounding a hollow, cell-free cavity (Fig. 2G).

In vivo, aged individuals may accumulate varying amounts of senescence-like stromal cells due to different levels of prior radiation exposures or other stress insults (9). We therefore examined whether senescence-like fibroblasts, in the presence of their normal counterparts, could still interfere with ductal morphogenesis of adjacent mammary epithelial cells. To address this possibility, MCF-10A cells were cocultivated with nonsenescent fibroblasts and senescence-like fibroblasts mixed at different ratios. Quantification of the fraction of ductal component in the resultant three-dimensional structures revealed that ductal morphogenesis decreased with increasing proportions of senescence-like fibroblasts (Fig. 2H).

Senescence-like fibroblasts promote sustained growth of mammary epithelial cells with dysregulated cell cycle and cell death pathways. Studies have shown that senescent cells can exert growth-promoting and antiapoptotic effects on neighboring cells (21, 22). To test whether ionizing radiation–induced senescence-like fibroblasts disrupted morphogenesis in MCF-10A cells by promoting their growth, we analyzed growth curves. Unexpectedly, there was no significant difference in the growth rate of MCF-10A cells that were cocultivated with nonsenescent fibroblasts and those cocultivated with senescence-like fibroblasts (Fig. 3A). The MCF-10A cell numbers became constant after day 10 to day 15 when the ductal-acinar structures were fully developed, which was associated with a gradual loss of centrally located cells that lacked cell-extracellular matrix contact (not shown). We therefore reasoned that, if the cellular pathways governing this site-specific cell clearance in mammary epithelial cells were compromised, the trophic signals derived from senescence-like fibroblasts might induce their uncontrolled growth. Previously, coexpression of the human papilloma virus 16 (HPV16) E7 oncoprotein and the antiapoptotic protein Bcl-2 has been shown to lead to filling of the luminal space in mammary epithelial cell–derived organoids formed in three-dimensional homotypic culture (19). However, cocultivation of E7 and Bcl-2–expressing or E7 and Bcl-X_L–expressing MCF-10A cells (MCF-10^E7/Bcl-XL cells or MCF-10^E7/Bcl-2 cells) with senescence-like fibroblasts still resulted in the formation of enlarged cystic structures with single layered and polarized epithelial cells (not shown). Recent works have suggested that autophagy-dependent cell death contributes to the luminal formation of mammary epithelial cell–derived three-dimensional organoids (19, 23) and heterozygous deletion of the beclin-1 autophagy gene leads to hyperproliferative lesions in mammary gland with partially filled lumen (24). In an attempt to further block the luminal formation process, we down-regulated the expression of beclin-1 in MCF-10A^E7/Bcl-XL cells by retrovirus-mediated RNA interference (Fig. 3B). MCF-10A^{E7/Bcl-XL/Beclin-1−} cells exhibited deficient activation in autophagy (Fig. 3C) and formed apolar structures devoid of lumen when cultured alone three-dimensionally (Fig. 3D,, left). When cocultivated with nonsenescent fibroblasts, these cells formed disorganized cell cords and aggregates with filled lumens, whereas misshaped masslike structures were developed in the presence of senescence-like fibroblasts (Fig. 3D,, middle and right). Compared with cells that expressed vector controls, MCF-10A^{E7/Bcl-XL/Beclin-1} cells displayed faster and sustained proliferation in three-dimensional culture (not shown). Moreover, they grew faster when cocultivated with senescence-like fibroblasts than with nonsenescent fibroblasts (Fig. 3E). Similar results were obtained when another autophagy regulator, ATG-7, was silenced by RNA interference (MCF-10A^{E7/Bcl-XL/ATG-7−} cells; not shown).

Senescence-like fibroblasts induce invasive growth of breast carcinoma cells. Advanced carcinoma cells that undergo partial or complete epithelial-mesenchymal transition acquire the ability to invade surrounding extracellular matrix and exert altered growth pattern three-dimensionally (25). The strong influence of senescence-like fibroblasts on MCF-10A cells led us to ask whether senescence-like fibroblasts could also influence the growth and behaviors of mammary epithelial cells with higher malignant potentials. To address this possibility, we cocultivated nonsenescent fibroblasts or senescence-like fibroblasts with an invasive breast carcinoma cell line, MDA-MB-231, in a multicomponent extracellular matrix gel that partially mimics the extracellular matrix composition of the cancer invasion front (13). When seeded at a low density in a low-serum growth condition, MDA-MB-231 cells assumed an epithelioid configuration and were growth arrested (Fig. 4A,, top left). Higher serum concentration (10%) moderately stimulated the growth of MDA-MB-231 cells, which formed cell aggregates without apparent extracellular matrix invasion (Fig. 4A,, top right). In the presence of nonsenescent fibroblasts, MDA-MB-231 cells displayed moderate increase in growth and formed compact cellular aggregates (Fig. 4A,, bottom left). In contrast, MDA-MB-231 cells responded to senescence-like fibroblasts by assuming a polarized migratory appearance and enormously spreading around the fibroblasts (Fig. 4A,, bottom right). Confocal imaging analysis revealed that MDA-MB-231 cells formed heterocellular aggregates with either nonsenescent fibroblasts or senescence-like fibroblasts at an early stage (day 4) in coculture (Fig. 4B,, left). At day 7, cancer cells within the cell aggregates containing senescence-like fibroblasts started to invade surrounding extracellular matrix and became diffusively distributed thereafter (Fig. 4B,, bottom), whereas the cell aggregates containing nonsenescent fibroblasts moderately increased in size with no significant extracellular matrix invasion (Fig. 4B,, top). Quantification of the cell proliferation rates of MDA-MB-231 cells revealed that both nonsenescent fibroblasts and senescence-like fibroblasts stimulated the growth of MDA-MB-231 but the effect mediated by senescence-like fibroblasts was 1.5- to 2-fold greater than that from nonsenescent fibroblasts (Fig. 4C). Senescence-like fibroblasts also elicited a faster growth in other breast carcinoma cell lines including MCF-7, MDA-MB-435, and MDA-MB-468 to varying extents (not shown).

To further address the possibility that a subset of senescence-like fibroblasts present in the breast stroma could augment the growth and invasion of adjacent breast carcinoma cells, MDA-MB-231 cells were cocultivated with cell mixtures containing varying proportions of nonsenescent fibroblasts and senescence-like fibroblasts. The growth-promoting effects on MDA-MB-231 cells exerted by the fibroblast mixtures correlated with increasing proportions of senescence-like fibroblasts (Fig. 4D). Remarkably, as few as 5% senescence-like fibroblasts were able to augment the invasive growth of adjacent cancer cells to a discernible extent (Fig. 4E).

Senescence-like fibroblasts dysregulate three-dimensional growth of mammary epithelial cells by matrix metalloproteinase–dependent matrix degradation. To gain insights into the mechanisms underlying the growth-modulating ability of senescence-like fibroblasts, we analyzed and compared the transcriptional profiles of nonsenescent fibroblasts and senescence-like fibroblasts that were maintained in a low-serum growth condition. Of 7,680 known human genes, 232 genes satisfied our selection criteria and were significantly up- or down-regulated in ionizing radiation–induced senescence-like fibroblasts (not shown). Interestingly, six of the induced genes in senescence-like fibroblasts encode secreted-type MMPs including MMP-1, MMP-3, MMP-7, MMP-9, MMP-10, and MMP-12. RT-PCR analysis confirmed that these MMPs were significantly induced in two independently prepared batches of ionizing radiation–induced senescence-like fibroblasts (Fig. 5A). The expression of secreted MMPs would predict an increased matrix-remodeling activity of ionizing radiation–induced senescence-like fibroblasts. Indeed, a quantitative collagenolysis assay with either collagen I or collagen IV as substrate revealed a 2- to 3-fold higher activity in senescence-like fibroblasts than in nonsenescent fibroblasts (Fig. 5B). At the same cell concentration as fibroblasts, MDA-MB-231 cells displayed little proteolytic activity, which may explain why they were unable to assume mesenchymal-type configurations when cultured three-dimensionally at low densities (Fig. 4A,, top left). Moreover, in situ imaging of proteolysis showed a significantly increased activity in the heterocellular aggregates formed by MDA-MB-231 cells and senescence-like fibroblasts over the cell aggregates of nonsenescent fibroblasts and cancer cells (Fig. 5C).

Recent studies have shown that extracellular matrix can regulate the growth and behaviors of cells in three-dimensional contexts (26). Without pericellular matrix proteolysis, extracellular matrix will enmesh the cells in a three-dimensional physical cage and prevent the cytoskeletal rearrangements necessary for cell proliferation and migration. In light of the profound matrix-remodeling capability of senescence-like fibroblasts, we asked whether senescence-like fibroblasts modulate the three-dimensional growth characteristics of adjacent mammary epithelial cells by their secretion of MMPs into surrounding extracellular matrix. In line with this possibility, addition of the broad-spectrum MMP inhibitor II at micromolar ranges (5-10 μmol/L) that were sufficient to inhibit senescence-like fibroblast–mediated collagenolysis (Fig. 5B) markedly suppressed the senescence-like fibroblast–dependent invasive growth of MDA-MB-231 cells (Fig. 5D,, top) without affecting their long-term viability (not shown). Likewise, the growth of MDA-MB-231 cells in response to senescence-like fibroblasts also decreased by ∼70% when a physiologic inhibitor of secreted MMPs, TIMP-1, was included (Fig. 5D). Interestingly, in the presence of MMP inhibitor II, MCF-10A cells, when cocultivated with senescence-like fibroblasts, formed cystic structures of smaller size (not shown); similarly, MCF-10A^{E7/Bcl-XL/Beclin-1} cells formed smaller cellular aggregates in response to senescence-like fibroblasts (Fig. 5D , bottom), and their growth rate was undistinguishable from those cocultivated with nonsenescent fibroblasts (not shown). Together, these data indicate that the growth-promoting and proinvasive effects of senescence-like fibroblasts depend on the MMP-dependent matrix degradation.

Integrin/phosphatidylinositol 3-kinase/Akt/mammalian target of rapamycin as downstream effectors of the senescence-like fibroblast–mediated invasive growth. The three-dimensional coculture model provided us an opportunity to dissect the molecular pathways that mediate the senescence-like fibroblast–derived heterotypic influences on mammary epithelial cells. As pericellular matrix degradation is the prerequisite for the integrin-dependent mesenchymal-type migration of tumor cells in three-dimensional contexts (27), we examined the possibility that certain integrin receptors in mammary epithelial cells might respond to the remodeled extracellular matrix mediated by senescence-like fibroblasts. Malignant mammary epithelial cells frequently show atypical patterns of integrin expressions (28) and MDA-MB-231 cells are known to retain the expression of the α3, α5, α6, β1, and β4 integrins (29–31). We thus tested the role of these integrins by including function-blocking antibodies in three-dimensional coculture. Control nonimmune mouse IgG or anti–α5 integrin antibody did not affect the invasive growth of MDA-MB-231 cells induced by senescence-like fibroblasts [Fig. 6A  (top) and B]. Interestingly, in the presence of anti–α3 integrin antibody, MDA-MB-231 cells were unable to invade surrounding extracellular matrix whereas their growth was still induced by senescence-like fibroblasts as evidenced by the formation of enlarged spherical cell aggregates. In contrast, functional inhibition of α6 integrin or β4 integrin substantially abrogated the ability of both invasion as well as proliferation in MDA-MB-231 cells, whereas inhibition of β1 integrin was associated with a partial inhibition by ∼50%. All anti-integrin antibodies at the concentration used did not influence the formation of heterotypic cell aggregates when MDA-MB-231 cells were cocultivated with nonsenescent fibroblasts (not shown). Together, these results indicate that α6β4 integrin and, to lesser extents, α3β1 integrin and α6β1 integrin in MDA-MB-231 cells are responsible for relaying the growth-modulating signals from ionizing radiation–induced senescence-like fibroblasts.

To further delineate downstream pathways that relay the heterotypic signals from senescence-like fibroblasts, we inhibited the activation of different integrin-dependent signaling pathways including the extracellular signal-regulated kinase/mitogen-activated protein kinase (MAPK) pathway, the phosphatidylinositol 3-kinase (PI3K)/Akt pathway, and the Rac/NFκB pathway (32–34) in MDA-MB-231 cells. Functional inhibition of PI3K with LY294002, but not MAPK (by PD98059) or NFκB (by NFκB SN50), significantly compromised senescence-like fibroblast–induced invasive growth [Fig. 6A, (bottom) and B]. Blockade of Rac1 function through stable expression of a dominant-negative N17Rac1 had little effect on the invasive behaviors of MDA-MB-231 cells. In contrast, inactivation of Akt kinase activity, a major downstream effector of PI3K, by stable expression of a dominant-negative Akt (K179M) in MDA-MB-231 cells almost completely blocked senescence-like fibroblast–dependent invasive growth, indicative of a critical role of the PI3K/Akt pathway in mediating this heterotypic interaction. In support of the activation of the PI3K/Akt pathway, MDA-MB-231 cells, when cocultivated with senescence-like fibroblasts, show increased Akt phosphorylation (Fig. 6C). Furthermore, inhibition of mammalian target of rapamycin (mTOR), a downstream effector of PI3K/Akt (35), with rapamycin achieved a similar effect to those from LY294002 or functional inhibition of Akt. It is worth noting that the activation of PI3K/Akt is necessary but not sufficient for the three-dimensional invasive growth of MDA-MB-231 cells, as stable overexpression of a constitutively active form of Akt did not lead to significant extracellular matrix invasion when they were cultured in extracellular matrix gel in the absence of senescence-like fibroblasts (Fig. 6D). Together, our data show that MMP-mediated matrix degradation and subsequent activation of the integrin/PI3K/Akt/mTOR pathway are both required for the senescence-like fibroblast–induced invasive growth of breast carcinoma cells.

Ionizing radiation–induced remodeling of the stromal microenvironment, such as changes in the extracellular matrix composition and in situ activation of transforming growth factor-β, has been implicated in cancer progression (36, 37). Here we provide a novel mechanism by which low-dose ionizing radiation can perturb the tissue microenvironment through the accumulation of senescence-like fibroblasts in the stroma. As the stress-induced premature senescence phenotype is stable and irreversible, we envisage that the accumulation of senescent fibroblasts in the breast stroma can serve as a tissue memory of prior stress insults and foster an oncogenic environment that dysregulates cell-cell and cell-extracellular matrix interactions in affected tissues.

During mammary gland development, ductal morphogenesis mainly occurs during virginhood as a result of branching and elongation of the actively growing terminal end buds (20). Normal mammary epithelial cells, when cotransplanted with human mammary fibroblasts into mouse mammary fat pads, undergo ductal morphogenesis in a sex hormone–independent fashion (38, 39). Our three-dimensional coculture model that reflects the scale of mammary tissue architecture allowed us to recapture and dissect at the cellular level the process of mammary ductal morphogenesis in vitro. Interestingly, the pseudopodia networks of fibroblasts serve as a platform guiding mammary epithelial cells to grow into branching ductal structures, a phenomenon that might be mediated by the “contact guidance” between fibroblast and epithelial cells (40). This fibroblast-dependent ductal formation is disrupted by ionizing radiation–induced senescence-like fibroblasts due to their substantial cytoskeletal alterations. Nevertheless, in the presence of senescence-like fibroblasts, MCF-10A cells still formed structures that retained tissue polarity and lumen formation. Presumably, untransformed mammary epithelial cells may retain growth regulation at the tissue architectural level. Interestingly, expression of HPV16 E7 oncoprotein and concomitant inactivation of apoptosis and autophagy in mammary epithelial cells prevented the lumen formation and disrupted tissue polarity of the resultant three-dimensional organoid. Importantly, compromised three-dimensional growth control of mammary epithelial cells under this genetic background rendered them responsive to senescence-like fibroblast–derived growth stimulatory cues. Our results thus reveal an important interplay between genetic and epigenetic (tumor environment) alterations during the early stages of breast carcinogenesis.

MMP-dependent matrix remodeling is involved in inflammation, wound healing, and cancer (41). In human tumors, MMPs are largely expressed in recruited stromal cells and the tumor-promoting effects of stromal fibroblasts have been attributed to their production of MMPs (42). Certain MMPs, such as MMP-3 and MMP-7, play an important role in breast cancer initiation and growth (43, 44). MMP-3 has recently been shown to mediate the branching morphogenesis of mammary epithelial cells in response to senescent fibroblasts (45). MMPs have also been linked with invasion and metastasis of breast cancer in vivo (46). Our results suggest that mammary stroma fibroblasts acquire a MMP-secreting phenotype after prolonged exposure to environmental stresses like low-dose ionizing radiation. With the advent of the three-dimensional coculture system, we provide a direct experimental proof that MMP-dependent matrix degradation represents a major mechanism by which ionizing radiation–induced senescence-like fibroblasts promote growth of surrounding epithelial cells. This is in agreement with recent findings that extracellular matrix, in addition to serving as a substratum for adhesion cells, plays an important role in the control of cell growth and behaviors in three-dimensional contexts (26, 47). Consistently, MDA-MB-231 cells, which display low proteolytic activities per se, were entrapped by the extracellular matrix meshwork and assumed a growth-arrested state. The malignant potential of the highly invasive MDA-MB-231 cells did not manifest until being cocultivated with senescence-like fibroblasts, which relieves the extracellular matrix growth constraints by MMP-dependent proteolysis. Together, our findings lend a strong support to the role of stromal activation in radiation-induced breast cancer progression.

Cell-extracellular matrix interactions via integrins play important roles during tumor progression and metastasis (28). In breast cancer, the expression and distribution of various integrin subunits are frequently altered and a positive correlation has been observed between integrin functions and the malignant state (29–31). Consistent with the established role of α6β4 integrin in the migration and invasion of advanced breast carcinoma cells (48), we show that α6 integrin and β4 integrin are the primary extracellular matrix receptors responding to senescence-like fibroblasts. Of note, although β-1 integrin has been shown to play an important cooperative role during mammary tumor induction (49, 50), functional inhibition of β1 integrin only partially blocks the senescence-like fibroblast–mediated invasive growth of MDA-MB-231 cells. It is likely because, in carcinoma cells with altered β4-integrin expression, α6 integrin preferentially dimerizes with β4 integrin over β1 integrin (29), and the ligation of α6β4 integrin might be specifically induced by the remodeled extracellular matrix environment mediated by senescence-like fibroblasts. Consistent with the fact that α6β4 integrin activates PI3K preferentially over other integrins (48), the PI3K/Akt/mTOR pathway, but not other integrin-dependent signaling pathways, is selectively activated by senescence-like fibroblasts. It remains possible that the signaling pathway elicited by senescence-like fibroblasts may differ depending on the integrin used and the corresponding downstream effectors in different “target” mammary epithelial cells.

In summary, our results show an important role of low-dose ionizing radiation–induced senescent-like fibroblasts in fostering an oncogenic microenvironment, which disrupts glandular morphogenesis in untransformed mammary epithelial cells, exerts mitogenic effects on cells with dysregulated cell death pathways, and promotes invasive growth of advanced malignant mammary epithelial cells. All these effects can be attributed to the dysregulated cell-cell and cell-matrix interactions caused by senescence-like fibroblasts. Future studies to determine whether senescence-like fibroblasts exist in the stroma of the breast tissue are warranted.

Grant support: Harvard School of Public Health center grant from National Institute of Environmental Health Sciences, Roadmap Training Grant from NIH, and the Glenn/AFAR Scholarship from the American Federation for Aging Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Robert A. Weinberg for kindly providing us early passage human mammary fibroblasts; Dr. Joan S. Brugge for providing us MCF-10A cells and for the technical advice on three-dimensional culture; Dr. Valerie M. Weaver for providing us the pLZRS.neo Akt (K179M) and pLZRS.neo MyrAkt cDNAs; Dr. Peter Marinkovich for providing us the pLZRS.blasticidin N17Rac1 cDNA; Dr. Junying Yuan for providing us the pBABE.hygro Bcl-X_L cDNA. We acknowledge the use of the Bioimaging Facility and Flow Cytometry Facility at Harvard School of Public Health, and in particular the expert technical support of Jean Lai for confocal microscopic imaging and three-dimensional reconstitution.