Senescent Stromal-Derived Osteopontin Promotes Preneoplastic Cell Growth

Age is the greatest risk factor for the development of neoplasia (1, 2). The molecular and physiologic factors that underlie age-related increases in cancer incidence are diverse and include cell autonomous genetic and epigenetic alterations and coincident alterations in the tissue microenvironment. Genetic analysis of human cancers has revealed a myriad of mutations that disrupt diverse cellular signaling pathways. Cell culture systems and murine models have shown that disruption of pathways that regulate tissue homeostasis, cell growth, cell survival/death, and invasiveness cooperates to produce a transformed cell (3–5). Although these cell autonomous alterations are critical to tumorigenesis, preneoplastic and neoplastic lesions arise within the context of a complex stromal compartment that consists of numerous cell types. Among these are epithelial cells, fibroblasts, immune cells, and cells that contribute to angiogenesis that together work to enable transformation.

The importance of the host microenvironment in the tumorigenic process is supported by numerous studies. For example, it has been shown that chickens infected with the Rous sarcoma virus develop tumors at sites of injury (6) and that activation of mast cells enhances transformation in a skin model of carcinogenesis (7). These studies suggest that the inflammatory response plays a key role in tumorigenesis. In further support of this hypothesis, other studies have shown that the chronic inflammation caused by Helicobacter pylori infection or Crohn's disease predisposes affected individuals to cancer (8, 9). In addition to inflammatory cells, fibroblasts can also promote tumorigenesis. Under normal conditions, fibroblasts are responsible for matrix deposition and reorganization and facilitate recruitment of cells that participate in tissue homeostasis. However, fibroblasts present within neoplastic lesions display altered expression profiles compared with fibroblasts isolated from normal tissues. Indeed, the gene expression profiles found in cancer-associated fibroblasts (CAF) are reminiscent of those found within a wound and include expression of numerous growth factors, chemokines, and angiogenic factors (reviewed in refs. 10, 11). A role for CAFs in promoting transformation was shown in one study in which CAFs isolated from prostate carcinomas were found to stimulate the in vitro and in vivo growth of immortalized prostate epithelial cells, whereas prostate fibroblasts distal to tumors had no such capacity (12).

In addition to the increased mutational load and presence of preneoplastic cells found to arise in aging epithelia (13, 14), the stromal compartment also undergoes age-related changes. Indeed, senescent cells accumulate in tissues with age (15–17). Analysis of senescent fibroblasts reveals a significantly altered gene expression profile relative to their younger counterparts (18, 19). Interestingly, this altered gene expression profile mirrors that observed during wound repair (reviewed in refs. 10, 11). Prominent among genes whose expression is dramatically altered during wound repair are those that remodel the extracellular matrix (ECM) as well as growth factors and inflammatory cytokines. Moreover, senescent fibroblasts function analogously to CAFs in that they promote the in vitro and in vivo growth of preneoplastic cells (19, 20). This latter observation suggests that the age-related accumulation of senescent stromal cells and the alterations that they produce within the tissue collaborate with increasing numbers of preneoplastic cells to influence cancer incidence in older individuals. However, how senescent stromal cells function to promote tumorigenesis is still poorly understood because the panoply of factors secreted by these cells has yet to be fully identified.

Our goal was to identify senescent stromal factors that affect preneoplastic cell growth to begin to delineate the molecular mechanisms by which senescent stroma functions in tumorigenesis. To this end, we performed expression profiling of replicative senescent (RS) and stress-induced premature senescent (SIPS) fibroblasts and found that numerous factors were coordinately modulated. This study identifies one of these factors, osteopontin (OPN), as a critical mediator of stromal-epithelial interactions both in vitro and in vivo. We show that OPN is sufficient to stimulate the growth of preneoplastic cells. Finally, we show that senescent stroma expressing OPN can be found within premalignant lesions in vivo, supporting the physiologic relevance of these results and the potential role of OPN in promoting tumorigenesis.

Cell lines. Human foreskin BJ fibroblasts and 293T cells were grown as previously described (21). HaCaT cells are spontaneously immortalized human keratinocytes, are aneuploid and nontumorigenic, and retain a near-normal pattern of differentiation (22). They were grown in DMEM plus 10% heat-inactivated FCS. N.p.c.T human keratinocytes were a kind gift from James Rheinwald (Harvard Medical School, Boston, MA) and were grown as previously described (23). N.p.c.T. are human epidermal keratinocytes that were derived from newborn foreskin and transduced with human telomerase, mutant cyclin-dependent kinase 4 (R24C), and a p53 mutant (p53DD). These cells are immortal but retain the ability to undergo an epidermis-type suprabasal differentiation program as evidenced by the ability to form a granular cell layer and an enucleated stratum corneum (23).

Virus production and retroviral infections. Virus preparation was as described (23).

Microarray analysis. BJ fibroblasts were mock or bleomycin sulfate treated (100 μg/mL; Sigma) for 24 h. Replicatively senescent fibroblasts were obtained by continuous passage. After 72-h serum starvation, RNA was collected using Trizol (Invitrogen). Biotinylated cRNA was hybridized to Affymetrix Human Genome U133 Plus 2.0 GeneChips by the Washington University Microarray Facility. There were four, six, and six samples for SIPS, RS, and young, respectively. Microarray quality control, analysis, and clustering (UPGMA by Centroid) were performed using dChip (May 2008 release; ref. 24). All GeneChip comparisons had a cutoff basis of a lower bound of 90% confidence of fold change ≥1.5 and expression intensity difference ≥75. Gene Ontological categorization of probe sets was performed using the Gene Ontology (GO) database, current to publication date. GO terms used were the following: Immune & Inflammation (IDs: 6955, 9615, 42742, 6952, and 6954), Mitogens & Regulation of Proliferation (IDs: 7067, 85, 45840, 86, 82, 45930, 7049, 45786, 187, 186, 1558, and 8283), and Extracellular Matrix & Secreted Factors (IDs: 7596, 7160, 7155, 5604, 8243, 42730, 7596, 5201, 5581, 4252, 8237, 5125, and 8008).

Coculture conditions. Fibroblasts (1.2 × 10⁴) were plated in 96-well plates (Fisher Scientific) and treated as described above. HaCaT (1.0 × 10³) or N.p.c.T (2.0 × 10³) click beetle red (CBR) luciferase cells were plated on fibroblasts and incubated for 96 h. Luciferase (Chroma-Glo Luciferase System, Promega) readings were performed in a Bio-Tek microplate reader (Bio-Tek Instruments).

OPN short hairpin RNA and real-time PCR. Three short hairpin RNA (shRNA) plasmids targeting the human OPN gene sequence were provided in the pLKO.1 plasmid (sequences found on Mission shRNA Web site; Sigma). Standard protocol was followed for cDNA and reverse transcription-PCR (RT-PCR) using the following primers: OPN, TTGCAGCCTTCTCAGCCAA (forward) and CAAAAGCAAATCACTGCAATTCTC (reverse); glyceraldehyde-3-phosphate dehydrogenase, GCATGGCCTTCCGTGTCC (forward) and AATGCCAGCCCCAGCGTCAAA (reverse).

OPN immunoprecipitation. BJ fibroblasts were plated in serum-free medium for 72 h. NETN lysis buffer was added to medium [20 mmol/L Tris-HCl (pH 7.9), 1 mmol/L EDTA, 100 mmol/L NaCl, 0.5% NP40] and incubated overnight at 4°C with OPN antibody (AF1433; R&D Systems).

Recombinant human OPN. HaCaT cells were plated in a 96-well plate (1.0 × 10³) and incubated in DMEM/F12 supplemented with 0.1% FCS for 24 h. Recombinant human OPN (rhOPN; 100 ng/mL; R&D Systems) was added for 72 h. N.p.c.T cells were plated overnight (1.0 × 10³) and 100 ng/mL rhOPN was added in 5% keratinocyte serum-free medium/DMEM/F12 (Invitrogen) for 72 h.

Xenograft model and chemical carcinogenesis protocol. Female nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice (National Cancer Institute-Fredrick) were housed according to the guidelines of DCM, Washington University School of Medicine. Live in vivo imaging was as described (25). Two-step chemical carcinogenesis protocol was as previously described (26) on 129S6/SvEv female mice (Taconic). To obtain papillomas, outbred transgenic mice (RBP-j^flox/+ and K14CreERT; N1^flox/+; RBP-j^flox/+) were treated topically with 7,12-dimethylbenz(a)anthracene (DMBA; 25 μg) followed by twice weekly 12-O-tetradecanoylphorbol-13-acetate (TPA) treatments (4 μg) for 15 wk starting 1 wk after DMBA treatment.

Senescent-associated β-galactosidase. Staining was performed on cells and 10-μm frozen sections as previously described (16).

Immunohistochemistry. Dissected tissue was fixed for 1 h in 4% paraformaldehyde and embedded in paraffin or OCT freezing medium (Sakura Finetek U.S.A.) and stained with the following antibodies: OPN (1:400, AF808; R&D Systems), p16 (1:50, antigen retrieval in pH 9 Tris buffer; Santa Cruz Biotechnology), CD45 (1:60, 550539; BD Pharmingen), and F4/80 (MCA497G; Serotec). Detection was visualized with Dako and Vector kits.

Statistical analysis. Data represent the mean ± SE; statistical significance was assessed by a two-tailed Student's t test.

Senescent BJ fibroblasts stimulate the growth of preneoplastic cells in vitro and in vivo. The tumor microenvironment is an important contributor to neoplastic progression and an increasing body of evidence suggests that it also plays a pivotal role in the early stages of the transformation process. One factor that affects the stromal compartment is age. Indeed, the presence of senescent cells increases with age (15–17) and these cells possess the capacity to promote preneoplastic cell growth in several models (19, 20, 27, 28). To address the molecular mechanism by which senescent fibroblasts mediate preneoplastic cell growth, we first established a culture system that allowed us to quantitate cell growth in vitro. Because the presence of senescent cells within the skin of aging individuals increases with age (15–17) and our microarray analysis (see below) was carried on human skin fibroblasts, we choose to use preneoplastic immortalized keratinocytes (HaCaT and N.p.c.T.) for our studies (22, 23). To facilitate quantification of HaCaT and N.p.c.T. cell growth, we transduced them with CBR luciferase and grew them in the presence of young or senescent BJ fibroblasts.

Senescence can be induced by continued passage leading to RS, which is telomere based or by any number of other methods collectively referred to as SIPS (19, 29, 30), all of which have been shown to stimulate the growth of preneoplastic cells (19, 27, 28, 31). Therefore, we chose to use bleomycin treatment, which reliably induces senescence of BJ skin fibroblasts (Fig. 1A). To confirm that cells undergoing RS and SIPS equivalently stimulated preneoplastic growth, we compared their ability to stimulate the growth of HaCaT cells. As shown in Fig. 1B, RS and SIPS fibroblasts were equally proficient at inducing the growth of HaCaT cells compared with HaCaT cells grown in the presence of young BJ fibroblasts. In addition, live animal luciferase imaging revealed that BJ fibroblasts undergoing RS or SIPS efficiently stimulated the growth of HaCaT xenografts in vivo (Fig. 1C; data not shown). Histologic analysis of xenografts confirmed that observed by our live animal imaging. Injection of HaCaT cells and senescent fibroblasts resulted in large benign lesions. In contrast, injection of HaCaT cells alone or in the presence of young fibroblasts resulted in small, nearly undetectable lesions (Fig. 1D).

Analysis of the senescent transcriptome. Given the observed stimulation of HaCaT cell growth after coculture with both RS and SIPS fibroblasts, we next carried out microarray analysis to identify putative senescent-associated candidate genes. Given that fibroblasts undergoing RS or SIPS similarly induced the growth of preneoplastic cells (Fig. 1B; ref. 20), we hypothesized that a common core of genes was responsible for their growth-promoting activities. Therefore, we conducted microarray analysis under conditions that captured the cell culture conditions used in our coculture experiments (i.e., serum-free and 3% O₂). Using these conditions, RNA was isolated from BJ skin fibroblasts that had undergone RS or SIPS following bleomycin treatment and compared the expression profile to young BJ fibroblasts.

Analysis of the resultant microarray data revealed a significant alteration in gene expression in cells undergoing senescence versus their younger counterparts. As expected, gene expression differences were observed between cells undergoing RS versus SIPS. However, a significant overlap in the gene expression patterns of RS and SIPS fibroblasts was apparent (Fig. 2A). Indeed, when comparing genes differentially expressed in each senescent population compared with young fibroblasts, 354 were coordinately regulated in RS and SIPS fibroblasts (Fig. 2B). GO analysis of the data revealed that the coordinately regulated gene groups included proliferative and mitogenic genes and genes involved in ECM functions and inflammation (Fig. 2C). Many of these genes code for secreted growth factors, chemokines, wound repair proteins, and matrix remodeling proteins, which have been implicated in tumor stroma-epithelial interactions (10, 11). We also found that senescent fibroblasts increased expression of AREG and matrix metalloproteinase (MMP) 3, which affect preneoplastic cell growth and morphology, respectively (19, 29). These observations show that senescent fibroblasts possess a significantly altered gene expression pattern compared with young fibroblasts and support the hypothesis that activation of senescence is likely to alter the microenvironment, leading to enhanced preneoplastic cell growth and tumor formation in a manner analogous to CAFs.

To validate our microarray results, we chose a subset of target genes based on their reported secretory and mitogenic properties, reasoning that they would affect preneoplastic cell growth. Pregnancy-associated plasma protein A (PAPPA) is a metalloproteinase that cleaves insulin-like growth factor binding protein 4, an inhibitor of insulin growth factor (IGF; ref. 32). The direct result of the activity of PAPPA is increased bioavailability of IGF, which is a potent growth-regulatory protein. Amphiregulin (AREG) is a member of the epidermal growth factor family that stimulates the growth and proliferation of epithelial cells. It was previously documented that AREG is expressed in senescent fibroblasts and functions in a paracrine fashion to stimulate the growth of prostate epithelial cells (19). Plasminogen activator inhibitor-1, an inhibitor of tissue plasminogen activator and urokinase-type plasminogen activator, was previously shown to mediate induction of senescence (33). Finally, we examined the expression of OPN, a multifunctional secreted glycoprotein that has been implicated in diverse stages of carcinogenesis (reviewed in ref. 34). Quantitative RT-PCR analysis verified the up-regulation pattern observed in the microarray for all of the selected genes (Fig. 2D).

Identification of stromal-derived OPN in preneoplastic lesions coincident with senescent stroma. Our microarray data indicated that OPN was significantly up-regulated in senescent fibroblasts, suggesting that it affects preneoplastic cell growth. OPN was originally identified as a transformation-associated matricellular protein produced by cancer cells (35). OPN is a secreted protein that is extensively modified by phosphorylation and glycosylation (36) and is targeted by several metalloproteinases (26). Although OPN contributes to mineralization and ECM integrity, it also functions as a potent signaling cytokine (36). Intriguingly, OPN-null mice display a significant delay in the appearance and incidence of papillomas in a classic two-stage skin chemical carcinogenesis model (37), suggesting that OPN plays a key role in skin tumorigenesis. To date, studies have focused on the function of epithelial-derived OPN in tumorigenesis and its value as a prognostic marker (38). However, given the high levels of OPN expressed in senescent fibroblasts (Fig. 2) and the delayed cancer phenotype observed in OPN-null mice (37), we postulated that OPN expressed within the stromal compartment contributes to the transformation process.

Given the possibility that stromal-derived OPN contributes to transformation, we next sought to determine whether senescent stroma was present within a preneoplastic lesion and whether it was coincident with OPN-expressing cells. To address this possibility, we took advantage of the two-stage skin carcinogenesis model. We choose to focus on this model because previous studies showed that TPA treatment of murine skin leads to OPN expression (39). In addition, OPN-null mice display a delayed tumor phenotype when treated with this protocol (37). To determine whether DMBA and/or TPA were sufficient to induce senescence, we first examined skin sections obtained from mice that were treated with a single dose of DMBA followed by three treatments of TPA. We used senescence-associated β-galactosidase (SA-βGal) staining to assess the presence of senescent cells. We observed the appearance of SA-βGal–positive cells within the stromal compartment of mice treated with DMBA and TPA as well as TPA alone (Fig. 3A). Remarkably, the appearance of senescent stroma precedes the presence of overt hyperplasia as evidenced by the positive staining in TPA only–treated mice (AT) within 7 days of treatment (Fig. 3A). In addition, the same stromal area displayed p16-positive cells, another senescent marker (Supplementary Fig. S1). Using this model, previous work showed the presence of SA-βGal within the epithelial compartment of papillomas that arose several months after the initial DMBA/TPA exposure (40). Although we see very limited SA-βGal staining within the epithelial compartment of the skin, we postulate that the differences observed between our studies are due to the time at which the tissue was analyzed. Indeed, in our hands, SA-βGal staining of papillomas arising several months after initial DMBA treatment did reveal limited staining within the epithelial compartment (Supplementary Fig. S2).

We next analyzed tissue from papillomas obtained from mice treated with a single dose of DMBA followed by repeated twice-weekly treatments with TPA, which produces papillomas at predictable latency (41). As shown in Fig. 3C (top), OPN staining within the stromal compartment of a papilloma coincided with p16 staining, whereas limited staining was observed within the epithelial compartment. In addition, we also observed OPN-positive cells in the stromal compartment of DMBA/TPA-treated skin adjacent to papillomas (Fig. 3C , bottom). The presence of p16 expression is consistent with the presence of senescent cells (16, 42). SA-βGal staining was also evident within the stromal compartment of the papillomas, further supporting the possibility that the OPN expression was within senescent stromal cells (Supplementary Fig. S2).

Stromal-derived OPN stimulates the growth of preneoplastic cells. Senescent fibroblasts possess growth-promoting activities that stimulate preneoplastic cell growth (19, 31) and our data suggested that OPN functions as an important mediator of this activity. Therefore, to directly assess whether stromal-derived OPN affected preneoplastic cell growth, BJ fibroblasts were stably transduced with virally encoded RNA interference (RNAi) hairpin constructs targeting OPN or a control hairpin construct. We found that knockdown of OPN in senescent fibroblasts by three independent hairpins (shOPN-7, shOPN-8, and shOPN-9) resulted in dramatically reduced mRNA levels (Fig. 4A) and undetectable protein levels (Fig. 4B) when compared with cells transduced with a control hairpin (shSCR). Fibroblasts transduced with hairpins targeting OPN entered senescence upon bleomycin treatment at the same rate as observed in control cells, indicating that OPN expression is not necessary for induction of senescence (Fig. 4C).

Having established that our RNAi constructs efficiently reduced OPN to undetectable levels without affecting the induction of senescence, we next investigated how loss of OPN affected the growth-promoting properties of senescent fibroblasts. Senescent fibroblasts mock transduced or transduced with a control hairpin (shSCR) increased the growth of HaCaT and N.p.c.T cells compared with cells grown on young fibroblasts (Fig. 4D). In contrast, RNAi-directed knockdown of OPN in senescent fibroblasts resulted in a significant reduction in HaCaT and N.p.c.T cell growth (Fig. 4D). Indeed, we found that loss of OPN resulted in HaCaT and N.p.c.T cell growth similar to that observed when these cells were plated on young fibroblasts, indicating that OPN is necessary for stimulation of preneoplastic cell growth by senescent fibroblasts.

Stromal-derived OPN stimulates the growth of preneoplastic cells in xenografts. Loss of OPN significantly reduced the growth-promoting properties of senescent fibroblasts in vitro. Therefore, we next addressed the effect of loss of stromal-derived OPN on the growth of HaCaT cells in xenografts. Using live in vivo luciferase imaging, we found that injection of HaCaT cells alone or in combination with young fibroblasts resulted in little growth of xenografts. In contrast, coinjection of senescent fibroblasts and HaCaT cells resulted in robust growth in vivo (Fig. 5A). Similarly, senescent fibroblasts expressing a control hairpin (shSCR) significantly stimulated HaCaT cell growth. In contrast, OPN depletion abolished HaCaT cell growth, indicating that OPN derived from senescent fibroblasts is necessary in vivo (Fig. 5A). Immunohistologic analysis of CD45 and F4/80 in xenografts revealed a decrease in immune infiltration in lesions resulting from injection of senescent cells expressing a hairpin targeting OPN versus a control hairpin (Fig. 5B). Together, these experiments indicate that stromal-derived OPN affects the microenvironment and significantly stimulates preneoplastic cell growth.

rhOPN is sufficient to stimulate the growth of preneoplastic cells. The above experiments show that stromal-derived OPN significantly affects preneoplastic cell growth. We next sought to determine whether OPN is sufficient to promote preneoplastic cell growth. To address this question, we treated preneoplastic cells (HaCaT/N.p.c.T) with rhOPN and measured its effect on cell growth. We found that addition of rhOPN to both HaCaT and N.p.c.T cells conferred a modest but significant growth advantage compared with mock-treated cells, indicating that OPN acts as a paracrine factor that is sufficient for preneoplastic cell growth (Fig. 6). The increased growth observed on rhOPN treatment was less robust then when HaCaT or N.p.c.T cells are plated directly on senescent fibroblasts expressing OPN (Fig. 4). Because senescent fibroblasts express numerous secretory proteins, it is possible that the reduced response observed on rhOPN treatment reveals a requirement for additional factors. Alternatively, the reduced response could be attributed to a lack of adequate protein modifications on the rhOPN protein. As stated earlier, OPN is a highly modified protein (43) and careful inspection of a longer exposure of our Western blot (Supplementary Fig. S3) revealed the presence of additional OPN species in supernatants obtained from senescent fibroblasts that were not present in the rhOPN preparations.

Alterations in the tissue microenvironment in combination with cell autonomous mutations collaborate to drive the transformation process. Here, we identify OPN as a critical stromal-derived mediator of preneoplastic cell growth. Using microarray analysis, we found that OPN is up-regulated in senescent fibroblasts while being undetectable in their younger counterparts. We found that senescent fibroblast-derived OPN was necessary for preneoplastic cell growth in vitro and in vivo. In addition, recombinant OPN was sufficient to stimulate the growth of preneoplastic cells. Finally, examination of papillomas that arose following treatment of mice with a DMBA/TPA protocol revealed the presence of senescent stromal cells and OPN expression within the stromal compartment. Taken together, our data argue that senescent fibroblast-derived OPN contributes to tumorigenesis and reveal a previously uncharacterized stromal function for OPN that may represent a novel therapeutic target and/or biomarker for the early stages of cancer development.

Investigations into the cell types within the microenvironment that affect tumorigenesis have revealed that macrophages, B and T cells, mast cells, adipocytes, endothelial cells, and fibroblasts can affect primary growth and metastasis. CAFs exert their influence on neoplastic progression through several mechanisms, such as secretion of paracrine factors, which directly stimulate cancer cells and affect proliferation and survival. Although senescent fibroblasts differ from CAFs with regard to their proliferative capacity, they possess the ability to stimulate preneoplastic and neoplastic cell growth in a manner analogous to CAFs (19, 20). This latter observation raises the possibility that the accumulation of senescent cells within aged individuals alters the tissue microenvironment that in turn collaborates with cell autonomous mutations to drive the transformation process. Intriguingly, we show that senescent stroma appears before the emergence of a papilloma, arguing that stromal changes arise early in the transformation process. In support of this hypothesis, our microarray data indicate that in addition to OPN, senescent fibroblasts express numerous factors capable of remodeling the microenvironment and modulating cell growth in a paracrine fashion. We propose that these early changes, which include the appearance of senescent stroma and concomitant up-regulation of OPN, provide a rich microenvironment for tumorigenesis.

OPN is a pleiotropic protein to which numerous functions have been ascribed, including mineralization and calcification, osteoclast activation, endothelial cell survival, wound healing, migration, lymphocyte proliferation, and immunosurveillance (34). Interestingly, in human cancer, OPN expression levels correlate with tumor grade and elevated OPN serum levels are a poor prognostic indicator in a wide variety of tumors (34). Murine studies revealed that OPN-null mice display a significant delay in tumorigenesis when treated with the classic two-step skin carcinogenesis model of DMBA/TPA treatment (37), suggesting that OPN plays a key role in cellular transformation.

To date, investigation into the role of OPN in tumorigenesis has been limited to the effect of epithelial-derived OPN on transformation and progression. However, a functional role for stromal-derived OPN in tumorigenesis has not been investigated, although other cell types, including macrophages and neutrophils, express OPN under both normal and pathologic conditions (44). Notably, a recent microarray analysis of the stromal compartment of human breast cancers identified OPN as one of 26 genes that when overexpressed is a strong predictor of clinical outcome (45). Our results showing increased OPN in the stromal compartment of DMBA/TPA-treated mice suggest an even more general role for OPN in epithelial tumorigenesis. Together, these observations indicate that non–epithelial-derived OPN is present at various stages of tumorigenesis in human cancer as well as mouse models, and support a physiologic role for this source of OPN in transformation.

OPN is clearly important for the growth-promoting effects of senescent fibroblasts as OPN depletion from senescent fibroblasts drastically reduces growth of preneoplastic cells. However, our data indicate that rhOPN does not fully mimic the enhanced preneoplastic cell growth observed when cells are grown in the presence of senescent fibroblasts expressing OPN. The inability of rhOPN to recapitulate the magnitude of preneoplastic cell growth observed in our coculture conditions raises the possibility that other factors cooperate with fibroblast-derived OPN. Alternatively, posttranslational modifications of stromal-derived OPN may be important. OPN is a heavily modified protein that possesses MMP cleavage sites (26). Both protein modifications and cleavage affect OPN function (46, 47). Interestingly, analysis of our microarray data indicated that several MMPs are expressed at high levels in senescent fibroblasts. Western blot analysis of conditioned medium obtained from senescent fibroblasts revealed the presence of OPN species that were not present in our rhOPN preparation, raising the possibility that one or more of these species mediate the growth-promoting activities of senescent fibroblasts. In addition, despite the high levels of OPN mRNA present within senescent fibroblasts, we did not detect the protein in the cell pellets but instead were only able to detect OPN in the conditioned medium of senescent fibroblasts, indicating that a significant amount of the protein is present as a soluble factor within the extracellular space. Such data suggest that the presence of OPN in the stroma of premalignant lesions, such as papillomas, may be underappreciated.

Our data and that of others (19, 20) support a role for senescent stroma in preneoplastic lesions but do not preclude its involvement in the later stages of transformation and cancer development. Interrogation of our senescent microarray data revealed a prominent inflammatory profile, which could facilitate recruitment of other components of the microenvironment in addition to affecting the ECM, bone marrow cell recruitment, and angiogenesis. Intriguingly, CAFs also express many of these factors, some of which are able to promote vascularization by recruiting endothelial cells (48). Furthermore, a recent report shows that CAFs secrete ECM tracks on which tumor cells can migrate, thus facilitating metastasis (49). Although OPN is important for metastasis and is a validated prognostic marker in human neoplasias, it was recently shown to function as a systemic instigator by mobilizing and recruiting bone marrow–derived cells to the tumor stroma (50). Interestingly, immunohistochemical analysis of xenografts arising from injections with stromal cells in the absence of OPN revealed a decrease in immune infiltration compared with those arising in the presence of OPN-expressing stromal cells. This observation raises the possibility that senescent stromal-derived OPN affects recruitment of cells of the innate immune system potentially through bone marrow mobilization. These results underscore the importance of paracrine signaling and the interplay between different components of the tumor microenvironment at multiple stages of transformation.

No potential conflicts of interest were disclosed.

Grant support: Philip Morris USA, Inc.; Philip Morris International; and Molecular Imaging Center grant P50-CA94056. E. Pazolli was supported by the Lucille P. Markey and the Department of Defense Breast Cancer Research Program Predoctoral Traineeship Award W81XWH-06-1-0691. L. Salavaggione was supported by grant P30-CA91842. S. Brehm was supported by the Howard Hughes Medical Institute through the Undergraduate Biological Sciences Education Program and a Hoopes Award. S. Demehri was supported by RO1-GM055479 awarded to Dr. Kopan.

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 members of the Stewart Laboratory for valuable discussions, Charlotte Kuperwasser for advice and inspiration, Cynthia Hernandez-DeLuca for technical assistance, Abhishek Saharia for graphical assistance, Loren Michel and Julien Duxin for critical reading, and Lisa Coussens and Cecilia Giachelli for the CD45 and OPN staining protocol.