Implanted adipose progenitor cells as physicochemical regulators of breast cancer

Tumor-Secreted Soluble Factors Regulate the Adipogenic and Proangiogenic Capability of ASCs.

To evaluate the effects of tumor cell-secreted soluble factors on ASC behavior, human ASCs were cultured in tumor-conditioned media (TCM) from commonly used highly (MDA-MB231) or less (MCF-7) aggressive human breast cancer cell lines. Additionally, mouse 3T3-L1 preadipocytes-a well-characterized cell model for studies of adipogenesis (15)-were used to verify the broad implications from primary ASC studies. Analysis of cell numbers and BrdU incorporation suggested that tumor-secreted factors promoted ASC and 3T3-L1 proliferation (Fig. 1A and Fig. S1A). In contrast, these factors inhibited adipose differentiation of both cell types as indicated by attenuated activity of the lipogenic enzyme glycerol-3-phosphate dehydrogenase (GPDH) (Fig. 1B), reduced activation of the key adipogenic transcription factor peroxisome proliferator-activated receptor gamma (PPARγ) (Fig. 1C), and inhibition of lipid droplet accumulation (Fig. S1B). An experiment in which ASCs and 3T3-L1s were first exposed to TCM and then evaluated for VEGF secretion further suggested that exposure to tumor-derived soluble factors enhances proangiogenic factor release by these cells (Fig. 1D). This increased proangiogenic potential is relevant to tumor angiogenesis as tumor-preconditioned ASCs and 3T3-L1s significantly enhanced VEGF-dependent migration of human umbilical vein endothelial cells (HUVECs) as compared to control cells (Fig. 1E). Notably, the magnitude of these changes directly correlated with tumor aggressiveness because MDA-MB231-TCM induced greater phenotypic changes than media collected from MCF-7 cultures (Fig. 1 and Fig. S1). Comparison of effects mediated by isogenically matched normal MCF10A, premalignant MCF10AT, and fully malignant MCF10ACA1a cells confirmed these results. Specifically, premalignant cells had no significant effect on ASC adipogenic differentiation and VEGF secretion but promoted ASC proliferation, whereas MCF10ACA1a cells mediated outcomes similar to MDA-MB231 (Fig. S1 C–E). The described observations were broadly relevant as TCM collected from two murine mammary tumor cell lines (16) also enhanced the proliferative and proangiogenic phenotype of adipose progenitor cells and inhibited adipose differentiation (Fig. S2 A–C).

To assess whether tumor cells actively recruit ASCs, we performed a transwell migration assay with TCM that revealed that tumor cells attract adipose progenitor cells towards their respective location by enhancing their directed migration (Fig. S2D). Collectively, these data suggest that tumor cell-secreted soluble factors enhance the proangiogenic cell population in the tumor stroma by guiding ASC behavior. As TCM similarly modulated ASC and 3T3-L1 behavior, and because data generated with primary cells are typically more robust and generalizable, the studies described in the following experiments were primarily performed with ASCs.

Tumor-Secreted Soluble Factors Enhance ASC Differentiation into ECM-Stiffening Myofibroblasts.

Myofibroblasts represent an abundant and proangiogenic cellular component of the tumor stroma whose differentiation is enhanced in the presence of reduced signaling by the adipogenic transcription factor PPARγ (17, 18). Accordingly, ASCs exposed to TCM exhibited increased expression of α-SMA relative to control conditions (Fig. 2A), whereby MDA-MB231 exerted a more pronounced effect than MCF-7 (Fig. S3A). Next, we evaluated whether tumor cell-derived TGF-β (19), which is pivotal to myofibroblast differentiation and secreted at higher levels by more malignant tumor cells (20), may be involved in inducing phenotypic changes of ASCs. Control media was supplemented with TGF-β at levels found in MDA-MB231-TCM, and this treatment induced ASC differentiation into myofibroblasts as suggested by immunofluorescence and Western Blot analysis (Fig. 2A and Fig. S3B). Furthermore, addition of a TGF-β epitope-blocking antibody to TCM abrogated myofibroblast differentiation. Whereas these data verify that tumor-secreted TGF-β functions as an upstream mediator of ASC differentiation into myofibroblasts (Fig. 2A), other molecules may also play a role. Comparison of control media as well as TCM from MCF-7 and MDA-MB231 via a cytokine antibody array revealed that interleukin-8 (IL-8), VEGF, tumor necrosis factor receptor 1 (TNFR1), and matrix metalloproteinase 3 (MMP3) were enhanced in the TCM (Fig. S3C). As IL-8 has been related to myofibroblast differentiation in prostate cancer (21) and was enhanced in MDA-MB231 vs. MCF-7 TCM (Fig. S3D), we assessed its effect on ASC myofibroblast differentiation. Results from these studies elucidated that IL-8 regulates ASC differentiation into myofibroblasts in a dose-dependent manner (Fig. S3E). These findings are of particular interest as myofibroblasts themselves can upregulate IL-8 (22), which may further promote ASC differentiation into myofibroblasts.

Myofibroblasts mediate tissue stiffening by altering ECM composition and enhancing contraction. In particular, increased collagen I deposition has been associated with increased tumor stiffness and malignancy (23). Our results support that ASC exposure to tumor-derived paracrine signals may contribute to this because ASCs produced significantly more procollagen I (6.0 ± 1.1-fold) and collagen I (4.2 ± 1.1-fold) when treated with TCM rather than control media (Fig. S4). Furthermore, ASCs seeded into microfabricated, free-floating 3D collagen disks differentiated into myofibroblasts, developed stress fibers, and contracted significantly more when cultured in the presence of TCM as compared to control media (Fig. 2 B and C). These variations in gel contraction were mediated by differences in TCM-mediated changes in myosin-mediated cell contractility rather than proliferation. Specifically, pharmacological inhibitors cytochalasin D (i.e., an agent preventing actin polymerization) or Y-27632 (i.e., an inhibitor of pROCK and thus RhoA signaling) added at levels not affecting cell viability restored the diameter of previously contracted gels (Fig. 2D), whereas disks cultured in TCM in the presence of mitomycin C contracted markedly more than similarly treated control disks (Fig. 2B). To fully elucidate whether the integrated effects of varied ECM deposition and contraction indeed contribute to enhanced tissue rigidity, the mechanical properties of ASC-seeded disks were analyzed following culture in control or TCM. The shear modulus of these constructs was determined by applying a shear force to one side of a sandwiched gel while measuring the displacement of a photobleached line within the gel (24). Our results indicate that TCM-treated collagen cultures were 1.5-fold stiffer than the control constructs (Fig. 2E) confirming the effect of soluble factors on ASC-mediated changes in tissue stiffness. These changes may be explained by varied matrix contraction and irreversible changes in ECM deposition mediated, for example, by increased collagen crosslinking due to enhanced lysyloxidase (LOX) expression by myofibroblasts (25) and/or deposition of additional fibrillar ECM components such as fibronectin (26).

ASCs Respond to ECM Mechanical Properties in a Tumor-Dependent Manner.

Normal and tumorigenic cells respond to increased matrix stiffness by adjusting integrin-dependent adhesion, traction forces, and subsequent downstream signaling (10, 27) with effects on cell proliferation and differentiation (28). To test the ability of ASCs to react to tumor-derived and/or self-imposed changes in matrix stiffness we evaluated their adhesion characteristics and corresponding activation of mechanoregulated signaling pathways using hydrogels of pathologically-relevant stiffness (8). Specifically, use of collagen-coated polyacrylamide gels allowed control over mechanical properties of the culture substrates without affecting adhesion ligand density that can independently affect cell behavior (29). ASCs spread significantly more on substrates mimicking breast tumor rigidity (approximately 10 kPa) (8) relative to substrates approximating premalignant (approximately 5 kPa) and adipose tissue (approximately 1 kPa) (8) stiffness (Fig. 3A). Traction force microscopy (TFM) confirmed that these morphological changes were accompanied by greater traction forces of ASCs on stiffer relative to more compliant matrices (Fig. 3A). These observations directly correlated with the recruitment of phosphorylated focal adhesion kinase (pFAK) to focal adhesions (Fig. 3B), a result of activated Rho-ROCK signaling due to force-induced changes in adhesion dynamics (30). Accordingly, ASC number was greater on stiffer matrices, and administration of cytochalasin D and Y-27632 inhibited this effect (Fig. 3C) confirming that Rho-ROCK mediated changes in cytoskeletal tension enable ASCs to respond to ECM stiffness.

To assess broadly stiffness-dependent cellular responses to a more relevant, biocompatible material that can be used for adipose tissue engineering and breast reconstruction applications (31, 32), we cultured adipose progenitors on Arg-Gly-Asp (RGD)-modified alginate disks similarly mimicking normal, premalignant, and cancerous breast tissue stiffness. Interestingly, increased stiffness recapitulated the effects of TCM and enhanced cell numbers and VEGF-secretion of these cells while inhibiting their adipogenic differentiation (Fig. S5). These results are consistent with previous studies that indicate elevated adipocyte conversion on more compliant matrices (28, 33) and with less spread, rounder cell morphology (34).

To investigate whether tumor-secreted soluble factors enhance ASC response to ECM rigidity, ASCs were cultured on polyacrylamide substrates of a broad range of stiffnesses (0.2–30 kPa) in the presence or absence of TCM from MCF-7 or MDA-MB231. Cell growth was commensurate with increasing stiffness, whereas the level of tumor malignancy affected ASC stiffness response. More specifically, TCM from the more malignant MDA-MB231 cells increased cell growth on gels of lower stiffness relative to TCM of the less malignant MCF-7 (Fig. 3D). Furthermore, the combined effects of ECM stiffness and tumor-derived soluble factors promoted cellular alignment on stiffer (but not soft) matrices, and this effect was more pronounced with TCM from MDA-MB231 than MCF-7 (Fig. 3E). Collectively, these results suggest that tumor-secreted factors dramatically enhance adipose-derived stem cell responses to stiffness leading to (i) an increased contractile cell population and (ii) directed force generation (35) that may ultimately exacerbate tumor rigidity and hence malignancy.

ASCs Modulate Tumor Progression In Vivo.

To determine the relevance of our in vitro findings to tumor growth in vivo, ASCs and MDA-MB231 cells were injected individually or in combination into the cleared mammary fat pad of immunocompromised mice. Coinjection of MDA-MB231 with ASCs yielded larger tumors than delivery of tumor cells alone in accordance with previous results (36, 37) (Fig. 4A). In contrast, injected ASCs largely formed adipose tissue (Fig. 4C) that was similar in size to explants from sham-injected media suggesting that the difference in tumor size was caused by varied tumor malignancy rather than the additional volume assumed by the coinjected ASCs (Fig. 4A). Furthermore, tumors resulting from the mixture of ASCs and MDA-MB231 cells appeared more locally invasive upon explantation relative to MDA-MB231-borne tumors that were easily demarcated for dissection. Pathological evaluation of H&E-stained cross-sections confirmed this assessment and revealed an enhanced degree of local invasion and desmoplasia for the coimplanted tumors vs. the control tumors (Fig. S6 A and B). Additionally, cytological features of malignancy (multinucleated cells and rosettes) were augmented in the coimplanted group as compared to the MDA-MB231 group (Fig. S6C). To correlate variations in tumor growth with differences in myofibroblast differentiation, immunohistochemical analysis was performed that confirmed that the density of α-SMA myofibroblastic cells (Fig. 4B) was increased in tumors originating from coinjection of both cell types rather than tumor cells alone. Circulating ASCs can be recruited to and associate with blood vessels in the form of α-SMA pericytes (38). In our coimplantation studies, however, immunohistochemical analysis of desmin, a pericyte marker (39), verified that the majority of α-SMA positive cells were myofibroblasts rather than pericytes as desmin staining showed no significant difference (p = 0.33) between coimplanted and control tumors. Nevertheless, blood vessel density was greater in the coimplantation group (Fig. 4C), and ELISA of tumor lysates suggested that ASC/myofibroblast secretion of proangiogenic VEGF and IL-8 (12, 22) may have contributed to these differences (Fig. S7).

To evaluate if varied myofibroblast content correlated with increased stiffness, we determined the mechanical properties of the different tumors via Dynamic Mechanical Thermal Analysis (DMTA) compression tests on tumor sections in the elastic regime of deformation. The mean Young’s modulus (E) was more than 50% higher in tumors generated by coimplantation of ASCs than in those generated in the absence of ASCs (3.1 ± 1.2 kPa vs. 2.0 ± 0.8 kPa, P < 0.04) (Fig. 4D and Fig. S8). Data from the full elastic and plastic regime of deformation (Fig. S8B) additionally indicated that tumors from the coimplanted group exhibit higher stiffness and a smaller range of elastic (reversible) deformation than tumors grown without ASCs. Consistent with these results, tumors resulting from coinjection contained more procollagen I and collagen (Fig. S9 A and B), were characterized by enhanced collagen fibril maturity and linearity (Fig. S9 C and D), and likely also comprised increased fibronectin (26). Such changes are indicative of enhanced desmoplasia and aggressiveness (23) and may further contribute to ASC-mediated changes in tumor growth; decellularized matrices of TCM-preconditioned adipose progenitors increased tumor cell growth relative to matrices from control cells in a manner dependent on generation of cytoskeletal tension (Fig. S9E). Nevertheless, additional conditions including paracrine signaling between ASC-derived cells and tumor cells may also contribute to the observed changes in tumor growth in our studies (36).

Cell Culture.

3T3-L1, MDA-MB231, MCF-7 (all from ATCC), and MCN1 and MCN2 mammary tumor cells isolated from the mammary epithelium of MMTV-Cre, p53^L/L, MMTV-Cre, and p53^L/L Rb^L/L, respectively (16), were routinely cultured in MEM (α-modification (αMEM) (Sigma) containing 10% FBS (Tissue Culture Biologicals) and 1% antibiotic (penicillin/streptomycin, Gibco). MCF-10A (ATCC), MCF10AT1, and MCF10ACA1a (both from Barbara Ann Karmanos Cancer Institute) (47) were maintained in DMEM/F12 supplemented with horse serum, epidermal growth factor (EGF), hydrocortisone, cholera toxin, insulin, and penicillin/streptomycin (all from Invitrogen) (48). ASCs and HUVECs (both from Lonza) were cultured in their corresponding growth media (ADSC-GM and EGM-2, respectively; Lonza). Lonza isolates ASCs from lipoaspirates based on expression of stem cell-associated surface markers (49), cryopreserves cells at passage one, and does not pool cells from multiple sources. Here, ASCs from two separate female donors were utilized at passages less than seven. ASC function in response to TCM and matrix mechanical properties was assessed (see SI Materials and Methods). Additionally, the effect of 3T3-L1-derived matrices on MDA-MB231 cell growth was measured (see SI Materials and Methods).

In Vivo Studies.

Media, ASCs and/or MDA-MB231 cells (1·10⁶ in 20 μL of DMEM/ Ham’s F-12, 10% FBS, 1% antibiotic) were injected into the cleared mammary fat pad of at least six three- week-old female SCID/NCr mice (Charles River Labs, 01S11) (two tumors per mouse) per condition in accordance with Cornell University animal care guidelines. Explants were harvested six weeks after implantation, imaged, weighed, and divided for subsequent formalin-fixation/paraffin-embedding and lysis in T-PER buffer (Pierce). H&E-stained sections were broadly evaluated for pathological features of malignancy. Cross-sections were characterized further via immunohistochemical staining for CD31, α-SMA, procollagen I, and desmin, whereas collagen was stained via Masson’s Trichrome. For staining procedures and image analysis methods, see SI Materials and Methods. VEGF and IL-8 content in lysates was measured via VEGF and IL-8 DuoSets (R&D Systems) and normalized to protein content as determined via bicinchoninic acid protein assay (Thermo Scientific). Tumor stiffness was determined using a DMTA as further outlined in SI Materials and Methods.

Statistical Analysis.

Statistical analysis was performed using GraphPad Prism 5. Student’s t tests were used to compare two data sets, whereas ANOVA was applied to determine the statistical significance of the differences between data sets of three or more using the Tukey method to make post-hoc pairwise comparisons. A p-value less than 0.05 was considered statistically significant. Data are represented as average ± standard deviations with the exception of the traction force measurements that are shown with standard errors.