We thank Dr. Moriguchi for his interest in our work showing the antiangiogenic effect of a transforming growth factor beta (TGF-β) receptor I kinase.1 It is well known that TGF-β has a dual role in cancer. In hepatocellular carcinoma (HCC), the issue is even more complicated because of the presence of the underlying cirrhosis, because TGF-β also has a role in the fibrogenetic process. In addition, heterogeneity is a hallmark of HCC, and therefore it is difficult to identify specific pathways involved in tumor progression. An “early” and “late” signature have been described for the TGF-β pathway in HCC human samples, and these signatures are correlated to the clinical outcome.2 Consistent with this, another study shows that TGF-β is responsible for Wnt activation, leading to a more aggressive phenotype, so this subset of patients is marked by a molecular signature.3 This more aggressive phenotype can be successfully blocked by targeting TGF-βRI, as we have recently demonstrated.4 In view of the data currently available, it is hard to believe that a “magic bullet” can target all types of HCC, and this is why the molecular approach for clustering patients with HCC would be an important step toward defining a tailored therapy. It could be useful in those patients with loss of the TGF-β pathway, for example.5 However, further studies aimed at investigating TGF-β–related genes that can be targeted with drugs should include a protein kinase profile and functional biological approaches. This is important to discover the real significance of gene modulation in terms of HCC cell behavior and therefore how drugs will act. In addition, we should consider that during the natural history of HCC, and likely correlated with tumoral progression, the genomic profile may change. This has been suggested by Coulouarn et al., who propose that the signature is, in fact, temporal.2 Furthermore, an assessment of a drug's effectiveness based only on gene expression could, in our opinion, be lacking in important information regarding pharmacokinetics, and even more importantly, toxicity for specific tissues.
In our case, we have investigated drug effectiveness using an integrated preclinical model that includes functional, biochemical, and molecular approaches in in vitro, in vivo, and ex vivo human samples.4 Naturally, we do not believe that our approach is the only reliable one or that our xenograft model is more suitable than others. The intrinsic concept of “model” automatically shifts any experimental system further from reality. Thus, in our opinion the debate regarding the usefulness of different experimental models is purely academic, whereas a correct interpretation of the limits inherent to each model is more relevant. Avoiding an “integralistic” vision of the model is a must, because this will certainly generate robust data but unlikely to be fully applicable to humans. In conclusion, a holistic, open-minded consideration of the literature is essential to guarantee progress in the field.