Our study clearly demonstrates that treatment with TGF-β antagonists inhibits the ability of bone-as well as lung-tropic MDA-MB-231 cell lines to establish experimental metastases in vivo. This convincingly demonstrates that TGF-β signaling plays an important role in this process, largely independently of the organo-tropism of the tumor cells (Figure 4). Our results are consistent with several previous studies that have reported anti-metastatic activity of individual TGF-β antagonists in in vivo models of human mammary cancer. For example, Arteaga et al.  reported that intraperitoneal injections of the murine TGF-β neutralizing antibody, 2G7 (Genentech®), was able to suppress lung metastases of MDA-MB-231 breast cancer cells that had been inoculated intraperitoneally. More recently, using the same experimental metastasis assay we employed, Ehata et al.  reported that treatment with a TGF-β type I receptor kinase inhibitor, Ki26894, decreased bone metastases and prolonged survival of mice inoculated with highly bone-tropic human MDA-MB-231-D breast cancer cells. Similarly, Korpal et al.  recently reported that treatment with LY2106791 inhibited early skeletal metastases.
In our hands, both classes of TGF-β antagonist significantly reduced the burden of skeletal and pulmonary metastases (Figure 4). Prior to our study, little information was available to determine whether the anti-metastatic efficacy of TGF-β antagonists on human breast carcinoma was organ site-specific. Separate reports indicated that the anti-TGF-β antibody 1D11 appeared to inhibit skeletal-or pulmonary metastases of the murine 4T1 mammary carcinoma cells. Thus, treatment with 1D11 resulted in a significant reduction in the number of 4T1 lytic bone lesions . Using the same 4T1 cell line, Nam et al. showed that treatment with 1D11 significantly suppressed both the number and size of tumor metastases to the lungs [52–54]. Although one has to be cautious about direct comparisons across studies, the therapeutic effects of TGF-β neutralizing antibodies against 4T1-derived skeletal or pulmonary metastases appeared to be of a similar order of magnitude.
Although our results are consistent with previous reports of anti-metastatic activity of individual TGF-β antagonists in in vivo breast cancer models, none of the previous studies have conducted a comparison between two different pharmacological strategies to inhibit TGF-β signaling. Thus, our second most important finding is that both neutralization of active TGF-βs using the 1D11 antibody and inhibition of TGF-β receptor kinases using the dual receptor kinase inhibitor, LY2109761, resulted in quantitatively remarkably similar degrees of inhibition of experimental metastases to both bone and lungs. Besides inhibiting the TGF-β type I (and -II) receptor kinases, LY2109761 also inhibits the activin receptor kinases, Alk-4 and Alk-7. This is a property shared by all known other members of this class of compounds, raising the concern that their biological activity may be mediated by either TGF-βs or activins. On the other hand, 1D11 is specific for bioactive TGF-βs and does not neutralize any of the other TGF-β superfamily members, including activin or BMPs. Thus, the qualitatively and quantitatively similar anti-metastatic effects we observed using both compounds in both experimental metastasis assays strongly support a specific role for TGF-β in this process, and essentially exclude the possibility that the effects we observed were due to interference with either activin-or BMP signaling.
In vitro, treatment with exogenous TGF-β induced Smad2/3 phosphorylation in all six MDA-MB-231 subclones and both TGF-β antagonists were capable of blocking Smad2/3 signal activation (Figure 2). In addition, both compounds effectively cause Smad2/3 signal termination, albeit that LY2109761 induced dephosphorylation of Smad2 and -3 more rapidly than 1D11. Consistent with these in vitro findings, in vivo, phospho-Smad2 levels were reduced in lungs of animals treated with either compound compared to vehicle treated controls (Figure 5). Moreover, LY2109761 treatment partly inhibited mRNA expression of TGF-β target genes, consistent with blockade of endogenous TGF-β signaling in vivo. These results are consistent with our previous findings using the TGF-β type I receptor inhibitor, SD-208, in the syngeneic 4T1 mammary cancer model . In contrast, 1D11 treatment was not associated with a significant reduction in target gene transcript levels by in vivo, suggesting that this agent only neutralizes activated ligand and selectively spares endogenous TGF-β signaling.
We and others have recently reported that, besides Smad2 and -3, TGF-β also activates the BMP Smads, Smad1 and -5, in normal and malignant mammary and epidermal epithelial cells [9–11, 55, 56]. Moreover, the degree to which exogenous TGF-β induced Smad1/5 phosphorylation in the different subclones appears to reflect their metastatic ability in vivo (Figure 2). Thus, the activation state of BMP Smads should be explored as a predictive biomarker of response to TGF-β antagonists in a clinical setting.
A major unresolved question is whether and under which conditions the predominant role TGF-β plays is mediated by its tumor cell autonomous effects, or via its actions on the host microenvironment. We approached this question by comparing two types of bone-tropic MDA-MB-231 subclones. Following intracardiac inoculation with MDA-MB-231 cells, some animals developed skeletal metastases following a prolonged period of dormancy (Lu et al., In Preparation). Cell lines derived from these "post-dormancy" metastases retain clear bone-tropism when re-injected into secondary animals, but display a gene expression profile that is quite distinct from that found in the "primary" bone metastases (Lu et al. In Preparation) . However, when we treated mice that had been inoculated with post-dormancy bone tropic 2860 TR cells with the 1D11 TGF-β neutralizing antibody, the development of skeletal metastases was inhibited to a similar extent as in SCP2-TR inoculated mice (Figure 4). Thus, 1D11 appeared to be anti-metastatic independently of the intrinsic gene expression profile of individual bone tropic tumor cell clones derived from the same parental cell line. These results suggest that, at least in this MDA-MB-231 in vivo model, TGF-β's pro-metastatic activity may be mediated predominantly by its actions on host cells within the bone microenvironment, rather than by autocrine effects on the tumor cells themselves. Consistent with this idea, neither LY2109761 or 1D11 treatment inhibited tumor cell proliferation or induced tumor cell apoptosis, in vivo (Figure 6).
In response to activated TGF-β released from bone matrix, MDA-MB-231 cells secrete a number of signaling molecules, including PTHrP and RANK-L, that stimulate osteoclast activity . Osteoclast-mediated bone breakdown is thought to release TGF-β, thereby resulting in a "vicious cycle" that leads to progressive bone destruction . Thus, we predicted that treatment with TGF-β antagonists would decrease osteoclast activation in the context of MDA-MB-231 bone metastases. In fact, 1D11 treatment resulted in a significant reduction in the number of active osteoclasts at the tumor:bone interface (Figure 6). Similarly, Futakuchi et al.  recently reported that treatment with 1D11 inhibited osteoclast activation and osteolytic bone destruction by 4T1 mammary carcinoma cells in vivo. In this study, identical effects were obtained using a chemical TGF-β type I receptor kinase inhibitor . Consistent with these findings, Mohammad et al.  recently reported that treatment with the TGF-β type I receptor kinase inhibitor, SD-208, increased osteoblast differentiation and bone formation, while reducing osteoclast differentiation and bone resorption. In aggregate, these studies have clearly demonstrated that pharmacological blockade of TGF-β signaling shifts the balance from bone breakdown to bone (re)generation, thereby inhibiting tumor-associated osteolysis.
In the lung metastasis model, treatment with TGF-β pathway antagonists inhibited tumor angiogenesis, as reflected by a decrease in CD34-positive microvessel density. These findings are consistent with our own earlier studies of the effects of the TβR-I kinase inhibitor, SD-208, against 4T1 lung metastases . Similarly, Nam et al.  reported that treatment with 1D11 was associated with a statistically significant decrease in microvessel density in 4T1 murine mammary tumors. Consistent with these findings, treatment of 4T1 tumor bearing mice with the 2G7 anti-TGF-β neutralizing antibody significantly reduced circulating VEGF levels (Genentech, US Patent Application 2005/0276802 A1). Thus, at least in lung metastases, TGF-β pathway antagonists have been consistently found to exert modest anti-angiogenic effects against basal-like mammary cancer in vivo.
Even though both TGF-β antagonists clearly had a demonstrable anti-metastatic effect in the MDA-MB-231 human breast cancer models, neither of the two agents completely abolished skeletal or pulmonary metastases. In part, this may be due to the fact that we had to use immunodeficient mice as hosts for human tumor cells because TGF-β pathway antagonists have been shown to de-repress anti-tumor immunity in mouse models of mammary cancer [48, 49, 52–54]. For example, we ourselves demonstrated that treatment with the TGF-β type I receptor kinase inhibitor, SD-208, inhibited spontaneous pulmonary metastases of R3T mammary carcinoma cells much more strongly in syngeneic than in nude mice . Published studies have demonstrated that tumor-associated TGF-β not only suppresses NK cell activity and T-cell mediated anti-tumor responses, but also actively subverts the CD8+ arm of the immune system into directly promoting tumor growth by an IL-17-dependent mechanism [48, 49, 52–54]. As we utilized athymic nude mice as hosts, we cannot ascribe the observed anti-metastatic effects of TGF-β antagonists to stimulation of T-cell-dependent processes. Moreover, even though Arteaga et al. were able to detect an effect on NK cells, even in the MDA-MB-231 model , we were unable to detect an increase in NK cell infiltration into metastases of 1D11 or LY2109761 treated animals in the current study (data not shown). Thus, we predict that treatment with TGF-β antagonists will have significantly greater anti-metastatic impact when applied in the context of a syngeneic host, in which they will act by a cooperative mechanism that involves several different cellular compartments, including the CD8+ T cells, NK cells, the microvasculature, osteoclasts and the tumor cells themselves .
Finally, we should note that all of the pre-clinical studies of TGF-β pathway antagonists in mammary cancer reported to date, have employed cell lines derived from basal-like tumors. Thus, these studies preclude any conclusions regarding the possible anti-metastatic activity these compounds may or may not have in the context of estrogen-dependent or HER2-mediated breast cancers. In fact, a wealth of experimental and clinical evidence suggests that, as long as breast cancers remain dependent on estrogens, TGF-β protects against rather than promotes tumor progression . Thus, one has to be cautious in extrapolating the results from the current and other preclinical studies of TGF-β pathway antagonists to breast cancers other than those of the basal-like subtype.