Smac mimetics LCL161 and GDC-0152 inhibit osteosarcoma growth and metastasis in mice

Osteosarcoma is the most common primary bone malignancy. These genomically unstable cancers develop due to oncogenic transformation, usually involving inactivation of p53 [1], of osteoblast lineage cells or their mesenchymal progenitors [2, 3]. Osteosarcomas typically arise in the extremities of teenagers. Osteosarcoma is rarer in older populations, and approximately half of elderly osteosarcoma patients acquire these cancers secondarily to Paget’s disease or bone irradiation [4]. Osteosarcoma preferentially metastasises to the lungs, and around a fifth of patients have detectable metastases at diagnosis [5, 6].

Interventions for osteosarcoma patients typically involve chemotherapy (usually methotrexate, doxorubicin and cisplatin) before and after amputation or limb-sparing surgery [7]. The introduction of chemotherapeutics to osteosarcoma treatment regimens in the 1970s and 1980s improved 5-year osteosarcoma survival rate from ~ 20% in the 1960s to ~ 60% by the 1980s [8], however there has been no significant improvement since [9], and current treatments are only effective for 20–30% of patients with metastatic disease [6, 9]. Better therapies are needed for non-responsive tumors. Various targeted therapeutic agents such as inhibitors of VEGFR, IGF1-R, mTOR, and immune checkpoint molecules, are presently being evaluated clinically for osteosarcoma [10].

“Smac mimetics” (also known as “IAP antagonists”) are small molecules developed to mimic the activity of the cellular protein Smac [11, 12]. They induce cell death by inhibiting the activity of pro-survival IAP proteins such as XIAP, cIAP1 and cIAP2 [13]. XIAP exerts its pro-survival activity through inhibition of pro-apoptotic caspase-3, − 7 and − 9 [14], and some IAP antagonists can relieve this inhibition by binding to XIAP. On the other hand, cIAP1/2 polyubiquitinate RIPK1, ultimately promoting NF-κB-mediated induction of genes that induce cell proliferation, migration and invasion in cells exposed to TNFα [15]. Smac and its mimetics promote cIAP1/2 auto-ubiquitination and degradation, which leads to de-ubiquitination of RIPK1, resulting in formation of the “ripoptosome” complex [16]. The pro-apoptotic protein caspase-8 is activated in this complex to induce cell death through activation of executioner caspases, if their inhibition by XIAP is relieved [16]. RIPK1 can also activate RIPK3 and MLKL to induce necroptosis, a form of caspase-independent cell death [17] that can be activated by TNFα in cells lacking caspase-8 and IAP activity [18].

Monovalent Smac mimetics, such as GDC-0152 [19] and LCL161 [20, 21], resemble the amino terminus of Smac, and can interact at one site of an IAP protein, whereas bivalent compounds like Birinapant [22] target two such sites conferring higher potency and affinity. Smac mimetics also differ in their affinities towards particular IAP proteins. Birinapant preferentially binds to cIAP1 and cIAP2 [22], however LCL161 and GDC-0152 bind with similar affinities to XIAP, cIAP1 and cIAP2 [19, 20]. Smac mimetics can induce cell death in some cell types as sole agents, through stimulation of the non-canonical NF-κB pathway to produce TNFα, which then stimulates TNFR1-mediated cell death pathways [23‐25]. Other cells types, including osteosarcoma cells [26], fail to produce autocrine TNFα and therefore are only efficiently killed by Smac mimetics when exposed to exogenous TNFα.

Smac mimetics have been shown to be well-tolerated in patients, however high doses of LCL161 triggered cytokine release syndrome due to autocrine TNFα production [20], and occasional patients administered Birinapant experienced Bell’s Palsy [27, 28]. As single agents, Smac mimetics induced complete or partial remissions in a minority of patients and stabilized disease in others [29]. Over a third of acute myeloid leukemia patients administered DEBIO1143 with chemotherapy experienced complete remissions, although half subsequently relapsed [30]. Pre-clinical studies revealed that Smac mimetics could also augment the cytotoxicity of other targeted therapies [22, 31‐43]. The utilities of some of these co-treatments are presently being assessed in clinical trials. As mentioned above, exposure to Smac mimetics only provokes autocrine TNFα production to facilitate sole agent killing in cells from a subset of tumors. This does not necessarily preclude effective Smac mimetic-based treatment of tumors composed of such cells though, as Smac mimetics can boost systemic TNFα levels, conceivably providing sufficient TNFα at the tumor site to enable Smac mimetics to activate cell death pathways. Oncolytic viruses that stimulated intratumoral inflammatory cytokine production synergized strongly with Smac mimetics in mouse models of glioblastoma, rhabdomyosarcoma, mammary carcinoma and colon cancer [44‐47]. Cooperation by inflammatory cytokines and Smac mimetics has been documented to stimulate anti-tumor immunity via both innate and adaptive mechanisms [48, 49]. Indeed, Smac mimetics enhanced the efficacy of immune checkpoint inhibitors in mice [47], even in a context in which the tumor cells lacked cIAP1 and 2 [50].

There have been very limited investigations into the possible utility of Smac mimetics for treating osteosarcoma, with no clinical trials registered or conducted to date, however several lines of evidence suggest that these agents may be efficacious for this malignancy. The major molecular targets of these drugs, cIAP1 and 2, have been documented to be upregulated in osteosarcoma, and their silencing impaired osteosarcoma growth in mice [51]. A subset of Smac mimetics (SM-164, LCL161 and GDC-0152) potently cooperated with TNFα to kill cells from many murine osteosarcomas in vitro, and this toxicity was potentiated by co-treatment with doxorubicin [26]. Other studies have also reported the sensitivity of osteosarcoma cells to SM-164 [52], GDC-0152 [53] and DEBIO1143/AT-406 [54] in vitro. So far, only two articles have reported effects of Smac mimetics on osteosarcomas in vivo. DEBIO1143, a Smac mimetic that exhibited poor anti-osteosarcoma in vitro [26], did not significantly affect the growth of KHOS/NP cells implanted into nude mice as a sole agent [54]. Co-treatment with doxorubicin yielded a slight but statistically significant reduction in tumor growth a week after treatment began, although that effect’s duration was not reported [54]. The other in vivo study examined the anti-osteosarcoma efficacy of LCL161, which was one of the most active Smac mimetics in vitro [26]. Disappointingly, those authors observed that LCL161 treatment only slightly reduced the growth of human osteosarcoma xenografts in SCID mice [21]. However, SCID mice have lower levels of TNFα than wild type mice [55], and since osteosarcoma cells were only sensitive to Smac mimetics in vitro when co-treated with TNFα [26], the SCID xenograft model may underestimate the efficacy of LCL161. Levels of TNFα within osteosarcomas have not been previously reported, but published data suggest that they may be high. Serum TNFα levels were documented to be elevated in osteosarcoma patients, with concentrations reflecting disease progression and primary tumor size [56, 57]. Osteosarcomas harbor a large population of macrophages [58‐60] that could secrete TNFα, and implantation of transformed mesenchymal cells into mice produced osteosarcomas that were infiltrated by TNFα-expressing macrophages [61]. The observation that osteosarcoma cells were sensitive in vitro to TNFα combined with physiologically achievable concentrations of Smac mimetics, coupled with these suggestions that osteosarcomas may contain high levels of TNFα, prompted us to examine the anti-osteosarcoma activity of selected Smac mimetics in vivo, as sole agents or in combination with doxorubicin, using nude mice implanted subcutaneously or intramuscularly with murine or human osteosarcoma cells.

We previously profiled the in vitro sensitivity of cells from a number of spontaneous primary and metastatic murine osteosarcomas to a panel of Smac mimetics. SM-164, GDC-0152 and LCL161 potently sensitized cells from most tumors to killing by TNFα, although we observed some inter-tumor variability in the magnitude of this effect [26]. We generated luciferase- and mCherry-expressing derivatives of a subset of those murine osteosarcoma cell lines, to monitor tumor growth and drug responses in vivo. A reporter gene-expressing derivative of the murine osteosarcoma cell line 1029H, which displayed intermediate in vitro sensitivity [26], was reproducibly tumorigenic upon subcutaneous implantation into nude mice, so was selected for initial evaluation of the in vivo efficacy of Smac mimetics. Of the three Smac mimetics that cooperated most potently with TNFα to kill osteosarcoma cells in vitro, LCL161 and GDC-0152 have progressed furthest towards clinical use [19, 20, 66], so they were selected for pre-clinical in vivo anti-osteosarcoma testing. Bioluminescence readings during the first 5 weeks after implantation demonstrated that GDC-0152 strongly suppressed tumor growth (Fig. 1a). Bioluminescence readings were unreliable after this time, presumably reflecting poor uptake of luciferin into large tumors. Tumors were resected and weighed post-mortem to assess and compare the ultimate outcome of the treatments. Tumors regrew after GDC-0152 treatment ceased, as reflected by the weights of the tumors and the bioluminescence reading taken a week after the last drug administration. Caliper measurements, ¹⁸FDG-PET and MRI were also used to evaluate tumor responses to GDC-0152 treatment (Fig. 1b-e). Confirming the anti-osteosarcoma activity of GDC-0152 detected using bioluminescence and via tumor weights at endpoint (Fig. 1a), tumors in GDC-0152-treated mice were less metabolically active and significantly smaller than the untreated tumors (Fig. 1b-e). Mice given the highest dose of GDC-0152, 50 mg/kg, lost around 5% of their body weight the day after each drug delivery, but gradually recovered to attain similar weights to their untreated peers within a week of each treatment (Fig. 1a, right panel). This was a more pronounced adverse effect than that reported by Flygare et al., who only noted a reduction in body weight when tumor-bearing nude mice were given 100 mg/kg of GDC-0152 [19]. The likelihood that further dose escalation would have been intolerably toxic precluded us from testing whether a higher dose of GDC-0152 may have produced a more durable anti-tumor response.

LCL161 treatment also significantly impeded osteosarcoma growth (Fig. 1f). A published regimen (100 mg/kg each weekday) was very effective but, in contrast to a previous report that failed to detect any toxicity associated with this treatment [67], we observed substantial weight loss. After noting cumulative weight loss after the initial five daily administrations, we reduced the administration frequency to twice weekly, which prevented further net weight loss, but the animals failed to reach normal weights (Fig. 1f, right panel). The intermediate dosing regimen, 50 mg/kg twice per week, was slightly less effective but better tolerated, although this dosing prevented normal weight gain by these young animals.

Doxorubicin had less impact on osteosarcoma growth than the Smac mimetics in this model. Only the highest dose of 6 mg/kg/week significantly impaired tumor growth (Fig. 1g). This was counter-intuitive, given the clinical efficacy of doxorubicin for treating osteosarcoma patients [68], and the in vitro sensitivity of 1029H cells to this agent [26]. Doxorubicin has been documented to penetrate poorly into tumors [69] so it is possible the marginal efficacy of doxorubicin in this context reflects a low bioavailability of this poorly penetrant drug within subcutaneous tumors that may not be extensively vascularized [70].

On average, tumor growth was more substantially hampered by co-treatment with medium to high doses of Smac mimetics and doxorubicin than by the drugs as sole agents (Fig. 2a-d, left panels), although tumors regrew after treatment cessation. Although this trend of cooperation was observed in multiple experiments, statistical analyses failed to rule out the possibility that these differences were due to chance. The suggestion of an improvement in efficacy associated with the co-treatment was however accompanied by enhanced toxicity (right panels). One mouse that received twice-weekly treatment with 50 mg/kg LCL161 plus weekly administration of 6 mg/kg doxorubicin lost more than 15% of its weight within a day, necessitating euthanasia. In subsequent experiments involving co-treatment with these drugs, we therefore reduced the frequency of LCL161 administration from twice-weekly to weekly.

Although GDC-0152 and LCL161 could theoretically kill cells via relieving XIAP-mediated caspase inhibition, their major mechanism of lethality involves stimulation of cIAP1/2 degradation, facilitating RIPK1 de-ubiquitination, which redirects TNFα-mediated TNFR1 signaling towards apoptotic or necroptotic pathways [71]. Unlike some other cell types that can produce autocrine TNFα in response to Smac mimetic treatment [23‐25], this class of drugs only killed osteosarcoma cells upon provision of exogenous TNFα [26]. The in vivo efficacy of GDC-0152 and LCL161 we observed in this study therefore implied either that the in vivo tumor microenvironment somehow imbued osteosarcoma cells with the ability to produce autocrine TNFα, or that host cells in or around the tumors secreted TNFα that cooperated with the administered Smac mimetics to kill the osteosarcoma cells in vivo. Our data support the latter model. Flow cytometry revealed that only 37% of the cells comprising a subcutaneous tumor expressed detectable mCherry fluorescence. Around half of the cells within this tumor were infiltrating host cells, mostly macrophages (Fig. 3a). The phenotypes of 12 % of the cells could not be determined with the antibody panel we used; some were probably osteosarcoma cells whose mCherry fluorescence was too weak to detect and others were probably other types of infiltrating host cells. We performed intracellular cytokine staining of fixed tumor cells from three untreated mice and three animals that received a single dose of GDC-0152 six hours prior to culling. Unfortunately, the fixation abolished mCherry fluorescence, so 1029H-Luc cells could not be distinguished from other cells that lacked markers detected by our antibodies. Approximately 2–4% of the cells within tumors, mostly immune cells, produced TNFα, and this proportion was very slightly higher in samples from mice that received GDC-0152 treatment (Fig. 3b). Hardly any cells that lacked immune cell markers, which presumably were mostly 1029H-Luc osteosarcoma cells (Fig. 3a), contained TNFα (Fig. 3b). Ex vivo incubation of the tumor cells with lipopolysaccharide (LPS), but not GDC-0152, induced the majority of the immune cells to express TNFα (Fig. 3b).

High concentrations of TNFα, presumably derived from tumor-associated immune cells, were detected within lysates of tumors resected from mice six hours after administration of a single dose of saline, GDC-0152 or LCL161 (Fig. 4a). If most of the TNFα in these tumors was within interstitial fluid, and this constituted around 10% of the tumor volume (as was reported for subcutaneous fibrosarcomas [72]) our data suggest that the tumor cells in this implantation model may be exposed to around 6–10 pg/ml of TNFα in vivo, a concentration that achieved approximately half-maximal cooperation with Smac mimetics to kill osteosarcoma cells in vitro [26]. Analysis of blood harvested six hours after either a single drug treatment (Fig. 4a) or the last of four weekly treatments (Fig. 4b) confirmed published observations [20, 73] that these drugs dramatically increased levels of TNFα in the blood. This effect, which was particularly pronounced for GDC-0152, was ameliorated by co-treatment with doxorubicin (Fig. 4b), reflecting its established myelosuppressive activity in humans [74] and mice [75]. In vivo treatment with GDC-0152 or LCL161 reduced tumor levels of cIAP1/2, confirming the drugs accessed the tumors and exerted their expected biochemical effect on tumor cells (Fig. 4c). The presence of macrophages in spontaneously-arising osteosarcomas has been published [58‐60], but to our knowledge the amount of TNFα within naturally-arising osteosarcomas have not previously been measured. To investigate TNFα levels in spontaneous osteosarcomas within immunocompetent animals, we harvested tumors and blood from mice that developed osteosarcomas due to an osteoblast lineage-specific deletion of the tumor suppressor genes p53 and Rb [65]. Blood from tumor-free animals was also collected for comparison. The spontaneous osteosarcomas, like the subcutaneously implanted tumors, contained abundant TNFα (Fig. 4d). In keeping with our observation that the anti-osteosarcoma potential of Smac mimetics hinges on the presence of TNFα produced by myeloid cells within the tumors, disaggregated cells from freshly-resected implanted tumors (consisting of both osteosarcoma and infiltrating non-cancerous cells) were efficiently killed in vitro by Smac mimetics as sole agents, whereas the corresponding in vitro-cultured osteosarcoma cells were only sensitive to Smac mimetics when co-treated with exogenous TNFα (Fig. 4e).

We were interested in whether cells from human osteosarcomas would exhibit similar Smac mimetic sensitivity profiles to their murine counterparts, in vitro and in vivo. To explore this, we determined the in vitro sensitivity of a panel of human osteosarcoma cell lines to GDC-0152 or LCL161, alone or with TNFα, using the “CellTiter-Glo” assay. In this assay, a reagent containing high concentrations of luciferase plus its substrate luciferin is applied to treated or untreated cells. The intensity of light emitted correlates with the amount of ATP in the well, which enables luciferase to catalyse the luminescent reaction. Two minimally passed human osteosarcoma cell lines, OS9 and OS17 [62], survived incubation with Smac mimetics as sole agents but responded to co-treatments with TNFα (Fig. 5a), like cells from most of the murine tumors we previously tested [26]. The responses of established human osteosarcoma cell lines (SaOS2, SJSA1, U2OS, 143B and KRIB) varied substantially, however. SJSA1 and U2OS were resistant, even to co-treatment with Smac mimetics plus TNFα. KHOS cells were somewhat sensitive to Smac mimetics alone, and addition of TNFα only slightly augmented this sensitivity. SaOS2 cells were slightly less sensitive than OS9 and OS17 to co-treatment with Smac mimetics and TNFα. Parental and luciferase-expressing derivatives of 143B and KRIB were slightly more sensitive to the combination treatment than OS9 and OS17 (Fig. 5b, data not shown). To model the expected exposure of the human tumor cells to Smac mimetics and TNFα following implantation into nude mice, we compared the extents to which Smac mimetics sensitized the luciferase-tagged KRIB and 143B human osteosarcoma cells to murine versus human TNFα. The CellTiter-Glo assay was used for these experiments. The CellTiter-Glo reagent was designed to contain sufficient luciferase to ensure that reaction rates are proportional to ATP concentrations across a large range of cell densities, so we suspect that the additional presence of some transgene-encoded luciferase in these cells would be unlikely to affect the reaction rate and hence the light emitted. However we cannot conclusively exclude the possibility that lower luminescent readings following drug treatment may reflect a reduction in cellular luciferase, as well as ATP levels, as cells died. Although published data suggest human TNF receptors bind murine TNFα with only slightly lower affinity than human TNFα [76‐78], the human osteosarcoma cells were significantly more sensitive to Smac mimetics coupled with human than murine TNFα (Fig. 5b).

Our observation that doxorubicin only slightly impaired the growth of subcutaneously implanted murine osteosarcomas raised the possibility that vascularization of these tumors may be poor, despite obviously being sufficient to mediate intratumoral access of immune cells and Smac mimetics. We therefore decided to use a different implantation route for testing in vivo drug efficacy against human osteosarcomas. We first considered orthotopic routes. Intrafemoral and intratibial osteosarcoma implantation models have been developed, but the technical challenges associated with these procedures can lead to highly variable tumorigenicity rates, and intraosseous tumors are not well tolerated by mice [79‐81]. These factors would have necessitated using large numbers of animals to discern significant drug effects, and the need to provide analgesia could have introduced potentially confounding drug-drug interactions. Given the requirement of the inflammatory cytokine TNFα for the anti-osteosarcoma activity of Smac mimetics, we were particularly keen to avoid analgesics with anti-inflammatory activities. We therefore decided to establish an intramuscular implantation model for testing the impact of Smac mimetics on human osteosarcoma xenografts. Intramuscular implantations of osteosarcoma cells, either into the upper hind paw [82] or gastrocnemius muscle [83, 84], were reported to be highly tumorigenic. To minimise the tumors’ impact on leg function, we chose to inject luciferase-expressing KRIB human tumor cells into the cranial tibial muscle of mice. This yielded reproducible primary tumor growth that was well tolerated by the mice (obviating the need for analgesia), and metastases to the lungs of all untreated mice within 7 weeks of implantation.

As mentioned above, KRIB cells were only sensitive to Smac mimetics in vitro in the presence of exogenous TNFα, and murine TNFα cooperated with these drugs less potently than human TNFα in vitro, implying that this xenograft model may underestimate the ability of Smac mimetics to eliminate human osteosarcoma cells in patients. Nevertheless, LCL161 limited the growth of intramuscular KRIB tumors (Fig. 6a). Doxorubicin was also effective in this model, and co-treatment was very effective (Fig. 6a). This model enabled monitoring of metastases development, as measured by in vivo lung bioluminescence (Fig. 6b) and qPCR-based quantitation of the lung tumor burden at the experimental endpoint (Fig. 6c). Weekly or twice-weekly LCL161 administration, and weekly co-treatment with LCL161 plus doxorubicin significantly delayed metastases development (Fig. 6b). The numbers of osteosarcoma cells within the lungs of mice within each treatment group varied substantially (Fig. 6c, d) so, although LCL161 administration or co-treatment with doxorubicin slowed metastasis development (Fig. 6b), we did not discern statistically significant effects of treatment on ultimate lung tumor burden (Fig. 6c). Two mice treated with doxorubicin and two co-treated with LCL161 plus doxorubicin failed to develop pulmonary metastases and experienced durable primary tumor regressions: primary tumors were undetectable from week 3 for the two co-treated mice and from weeks 4 and 5 for those two doxorubicin-treated animals (Fig. 6d).

These experiments revealed that Smac mimetics GDC-0152 and LCL161 impeded the growth of implanted osteosarcomas in nude mice. The in vitro sensitivity of the murine and human osteosarcoma cells used to create these tumors depended on supplied TNFα. The in vivo efficacy we observed was probably due to high levels of endogenous TNFα within the implanted tumors (Fig. 7). This introduced a slight complication into our experiments designed to test Smac mimetic efficacy against human osteosarcoma cells grown in mice: murine TNFα cooperated with Smac mimetics less potently than human TNFα, thus our mouse experiments may have underestimated the potential for LCL161 to treat osteosarcomas in patients. Like the implanted tumors, spontaneously arising osteosarcomas that arose in genetically engineered immunocompetent mice also bore high concentrations of TNFα, excluding the possibility that this phenomenon was an artefactual consequence of tumors implanted into nude mice. Immunophenotyping revealed that implanted osteosarcomas, like patient tumors [58‐60], were heavily infiltrated by immune cells, which our data suggest were responsible for producing most of the intratumoral TNFα. Although we did not formally test the requirement for TNFα in order for Smac mimetics to exert anti-osteosarcoma effects in our model, this conclusion is consistent with our data showing that (a) Smac mimetic sensitivity of osteosarcoma cells in vitro depended on exogenous TNFα, (b) Smac mimetics retarded growth of tumors derived from these cells in vivo, and (c) implanted osteosarcomas contained TNFα that was produced by intratumoral immune cells. We would predict that Smac mimetic treatments would be ineffective in osteosarcoma-bearing animals treated with TNFα-blocking agents, or TNFα-deficient mice. Indeed, the presumed deficiency in TNFα-producing tumor-associated myeloid cells within SCID mice probably explains the relatively poor anti-osteosarcoma efficacy of LCL161 in treating SCID mice bearing patient-derived xenografts [21]. Mice bearing implanted or spontaneous osteosarcomas had around twice as much TNFα in their blood as tumor-free animals. Although this difference was not statistically significant, it mirrored published data from humans: the TNFα concentration in sera of osteosarcoma patients was approximately double that in control individuals’ blood [56]. This implies that osteosarcoma cells within patients’ tumors may be exposed to enough TNFα to render them sensitive to the lethal effects of Smac mimetics, but direct measurement of TNFα within patients’ tumors would be necessary to confirm this suspicion.

Doxorubicin, the linchpin of current osteosarcoma regimens, exhibited marginal single agent efficacy against subcutaneously-implanted 1029H tumors, but was more effective at reducing the growth of intramuscularly-implanted KRIB tumors. In both contexts, doxorubicin tended to cooperate with Smac mimetics to retard tumor growth, although this trend was not statistically significant. This co-treatment was particularly effective and sustained in the KRIB intramuscular model. Encouragingly, LCL161, alone or with doxorubicin, also significantly delayed the appearance of pulmonary metastases in mice bearing KRIB intramuscular tumors. Further work will be needed to ascertain whether this anti-metastatic effect was because the treated mice had smaller primary tumors (which would presumably seed fewer tumor cells to the lungs) and/or reflected drug-mediated destruction of osteosarcoma cells located within the lungs.

Subsequent studies will also be required to accurately model the potential benefit of co-treatment of osteosarcoma patients with Smac mimetics plus doxorubicin, versus single agent treatment, and to consider the balance between efficacy versus toxicities conferred by co-treatment, relative to Smac mimetics alone or coupled with other chemotherapy drugs. Doxorubicin dampened the inflammatory effect of Smac mimetic treatment, as reflected in less drastically elevated serum TNFα levels in co-treated mice, relative to animals that just received GDC-0152, consistent with the established myelosuppressive activity of doxorubicin [74, 75]. Despite that amelioration of the Smac mimetics’ dose-limiting toxicity, mice that received doxorubicin plus Smac mimetics lost more weight than animals that only received one drug. Additional research will be needed to determine the mechanism underlying this cooperative toxicity, including exploration of the possibility that Smac mimetics may exacerbate doxorubicin’s cardiotoxicity [85]. It will be important to determine whether cooperative toxicities would be avoided by sequential exposure, in which case subsequent Smac mimetic treatment may be considered for patients whose tumors persist or recur after administration of the maximal cumulative dose of doxorubicin recommended to avoid dose-limiting cardiotoxicity. The outcome of those experiments may help define clinical contexts in which the anti-osteosarcoma efficacy of Smac mimetics could be maximized while their toxicities are managed.

The clinical responsiveness of individual osteosarcomas to Smac mimetics will presumably be strongly dictated by the ability of the drugs to cooperate with TNFα to trigger apoptotic or necroptotic death of the individual’s cancer cells. Further work will be needed to gain a comprehensive understanding of intertumoral variability in the in vitro sensitivity of human osteosarcomas to Smac mimetics. Cells from two minimally-passaged human osteosarcomas were quite sensitive to Smac mimetic/TNFα co-treatment, however established human osteosarcoma cell lines varied substantially in their sensitivity to Smac mimetics as sole agents and together with TNFα. This heterogeneity may reflect biological variability between different tumors, and/or it may be a consequence of genomic instability-mediated phenotypic drift during extended in vitro culturing [86]. If the latter is a major factor, the sensitive phenotypes of the minimally passaged lines (OS9 and OS17) may reflect the typical responsiveness of osteosarcoma cells within patients’ tumors better than the established cell lines, some of which were more resistant.

It is important to note that our experiments were conducted in nude mice. Although these mice possess innate immune cells, which can produce the TNFα required for Smac mimetic-mediated osteosarcoma cell destruction, they have almost no T cells [87]. If Smac mimetics can stimulate immune-targeting of osteosarcoma cells through boosting lymphocyte survival and activation, as has been demonstrated in other cancers [48], this may augment the direct osteosarcoma cell killing we observed in nude mice, to yield a more pronounced anti-osteosarcoma effect in immunocompetent animals or humans.

The Smac mimetics LCL161 and GDC-0152 cooperated with TNFα produced by infiltrating immune cells to limit osteosarcoma growth and metastasis in nude mice. These data illustrate the potential for Smac mimetics to target malignancies like osteosarcoma in which the cancer cells fail to produce autocrine TNFα in response to these agents. Results from this study suggest that safe regimens involving Smac mimetics like LCL161 or GDC-0152 may improve treatment outcomes for osteosarcoma patients.