Development of TGF-beta antagonists for cancer therapy

基于有希望的临床前结果，多种tgf - β途径拮抗剂正处于治疗晚期癌症的早期临床试验中。然而，鉴于tgf - β的复杂生物学特性，用于癌症治疗的tgf - β拮抗剂的成功开发将取决于对这些药物如何起作用的清晰理解，以及如何选择将从这种治疗中受益的患者的相关问题。采用12只转移性乳腺癌小鼠同种异体移植模型，以转移负荷为主要终点，我们发现了对tgf - β拮抗剂的异质性反应。tgf - β通路阻断在某些模型中抑制转移，而在其他模型中对转移没有影响或刺激。我们正在继续应用靶向和基于发现的方法来解决治疗反应异质性的分子和生物学机制，并产生有用的预测性生物标志物。为此，我们从基因组学、转录组学和临床特征方面对转移模型进行了广泛的表征，并将这些信息公开。利用这些信息，我们已经证明了现有文献中提出的合理的候选生物标志物(例如。p53突变状态，低claudin肿瘤表型)似乎与对tgf - β拮抗剂的反应无关。出乎意料的是，原发肿瘤中tgf - β表达水平和tgf - β通路激活程度都不能预测抗tgf - β治疗的反应。然而，来自模型的肿瘤对tgf - β拮抗剂表现出理想的反应，其特征是tgf - β途径在未治疗状态下激活的转录组学证据。因此，在体内，tgf - β信号的输出可能比任何输入参数（如配体水平和信号转导组分的状态）更敏感地指示tgf - β通路的激活状态。我们继续探索对tgf - β拮抗剂的不良前转移反应的机制，我们已经表明它们是独立于免疫系统的，似乎涉及tgf - β对肿瘤细胞本身的直接转移抑制作用的破坏。通过模型组对tgf - β的体外反应分析，我们假设tgf - β的转移抑制作用专门针对癌症干细胞亚群，并且我们有初步证据表明这在体内也是正确的。我们也在继续研究提高tgf - β拮抗剂治疗效果的方法。在一种方法中，我们使用同种异构体选择性中和抗体来确定是否所有三种tgf - β同种异构体都应该作为最佳反应的目标。我们有大量相关数据表明tgf - β 1和tgf - β 3可能在乳腺癌中发挥相反的作用。在临床前小鼠模型中使用tgf - β同种型选择性抗体，我们发现保留tgf - β 3会导致代谢谱的改变和与人类乳腺癌数据集中良好预后相关的基因的上调，这表明这种方法可能具有治疗益处。

Based on promising preclinical results, a variety of TGF-beta pathway antagonists are in early phase clinical trials for the treatment of advanced cancer. However, given the complex biology of TGF-beta, the successful development of TGF-beta antagonists for cancer therapy will depend on a clear understanding of how these agents work, and the related question of how to select patients who will benefit from this type of treatment. Using a panel of 12 mouse syngeneic allograft models of metastatic breast cancer, with metastatic burden as the primary endpoint, we uncovered heterogeneous responses to TGF-beta antagonism. TGF-beta pathway blockade inhibited metastasis in some models, while having no effect on or stimulating metastasis in other models. We are continuing to apply targeted and discovery-based approaches to address molecular and biological mechanisms underlying the heterogeneity of therapeutic response and to generate useful predictive biomarkers. To do this, we have extensively characterized the panel of metastatic models with respect to genomic, transcriptomic and clinical features, and we have made this information publically available. Using this information we have shown that plausible candidate biomarkers suggested by the existing literature (eg. p53 mutation status, claudin-low tumor phenotype) do not appear to correlate with response to TGF-beta antagonism. Unexpectedly, neither TGF-beta expression levels nor extent of TGF-beta pathway activation in the primary tumor predict response to anti-TGF-beta therapy. However, tumors from models showing a desirable response to TGF-beta antagonism are characterized by transcriptomic evidence of TGF-beta pathway activation in the untreated state. Thus in vivo, the output of TGF-beta signaling may be a more sensitive indicator of the activation status of the TGF-beta pathway than any of the input parameters such as ligand levels and status of signal transduction components. We have continued to explore the mechanisms underlying the undesirable prometastatic responses to TGF-beta antagonists and we have shown that they are independent of the immune system and appear to involve disruption of direct metastasis-suppressing effects of TGF-beta on the tumor cells themselves. Analysis of in vitro responses to TGF-beta across the model panel led us to hypothesize that the metastasis-suppressing effect of TGF-beta is targeted specifically to the cancer stem cell subpopulation, and we have preliminary evidence that this is also true in vivo. We are also continuing to address ways of improving the therapeutic efficacy of TGF-beta antagonism. In one approach, we are using isoform-selective neutralizing antibodies to determine whether all three TGF-beta isoforms should be targeted for optimal response. We have generated substantial correlative data to suggest that TGF-beta1 and TGF-beta3 may play opposing roles in breast cancer. Using TGF-beta isoform-selective antibodies in preclinical mouse models, we find that sparing TGF-beta3 results in an altered metabolic profile and upregulation of genes that are associated with good prognosis in human breast cancer datasets, suggesting that this approach may have therapeutic benefit.