Immune Desertic Landscapes in Hepatocellular Carcinoma Shaped by β-Catenin Activation

Cancerous tissues in almost every malignant disease show heterogeneity with regard to density of T lymphocyte infiltration, from desertic to abundantly crowded, referred to in immunotherapy lingo as cold or hot tumors, respectively. Beneficial prognostic implications of the abundance of CD8⁺ T lymphocytes have been found for many tumors, with the only exception being clear cell renal cell carcinoma (1). Furthermore, with the advent of efficacious immunotherapies based on PD-1/PD-L1 inhibitors, an increasingly common trend is that the pretreatment levels of T-cell infiltration or gene signatures denoting pretreatment cytotoxic T-cell infiltration and IFNγ production correlate well with the clinical benefit of the immunotherapy intervention. These findings are reminiscent of previous observations indicating that survival following cancer vaccines was also correlated with preexisting immune infiltrates (2). In the literature of cancer immunotherapy, we refer to these tumors with low or no T-cell infiltrates as immune deserts, even if they are usually populated by abundant myeloid leukocytes (3). An important question was, and still is, which are the genetic or epigenetic mechanisms in tumors that result in differential presence of T-cell infiltrates. The answer is that there are probably multiple mechanisms, but studies in melanoma have started to shed light on this matter. Gajewski and colleagues observed that mutations leading to activation of the β-catenin pathway correlated with poor infiltration: In elegant experiments of carcinogenesis in gene-modified mice, this group was able to show that β-catenin activation in transformed melanocytes gave rise to poor T-cell infiltration and escape from checkpoint inhibitor immunotherapy approaches (4). Along these lines, PTEN loss of function is also correlative with poor T-cell infiltrates and response to immunotherapy, as discovered by Hwu and colleagues (5). In lung cancer, interesting negative correlations were found by coexisting RAS and LKB1 mutations with T-cell infiltrates and primary resistance to immunotherapy (6).

The correlative findings of β-catenin activation and sparse T-cell infiltration have been clearly extended to a good number of tumor types that included hepatocellular carcinoma (HCC; ref. 7). In experimental melanoma, the mechanism of weak T-cell infiltration was mostly due to the lack of infiltration by conventional type 1 dendritic cells (cDC1) that are specialized in cross-presenting exogenous antigens, such as tumor antigens, to CD8⁺ T cells. The ultimate reason why cDC1s failed to enter was the repression of the chemokine CCL4 that otherwise would have attracted cDC1 cells in such a way that, in turn, these antigen-presenting cells would have attracted CD8 T cells (4).

Advanced HCC constitutes a terrible medical condition. As first-line treatment, the tyrosine kinase inhibitors sorafenib and lenvatinib are used with a minor fraction of patients achieving objective radiologic responses and based on an overall survival benefit of 3 to 5 months in this setting. PD-1 checkpoint blockade with nivolumab or pembrolizumab is approved as second-line treatment based on overall response rates around 15% and an impressive median overall survival of more than 16 months (8). Efforts are greatly needed to understand who are the patients who actually benefit from immunotherapy treatment. Positive predictive biomarkers based on PD-L1 expression, prefiltration by CD3⁺ cells, or IFNγ gene signatures show some degree of correlation with response and survival, but these percentages are certainly insufficient to properly select or exclude patients from treatment (Melero and colleagues, AACR 2019, abstract 2675). As a negative predictor, a small series of patients studied at Memorial Sloan Kettering Cancer Center showed that those patients with HCC with mutations in CTNNB1 did not respond at all to PD-1 blockade (9).

In this issue of Cancer Discovery, Lujambio and colleagues report that in mouse HCC, β-catenin activation gives rise to a T-cell exclusion phenotype (10). The model is based on hydrodynamic gene transfer to adult hepatocytes of three plasmids: (i) a transposon plasmid coding for MYC and a more or a less antigenic variant of luciferase, (ii) a plasmid coding for the transposase, and (iii) a CRISPR/Cas9 plasmid to target p53. In such a model, immunosurveillance attenuates tumorigenesis when the antigenic form of luciferase is transferred and successful escape tumor variants show β-catenin activation. Moreover, successful tumorigenesis in the hydrodynamic gene-transfer setting was restored when a gain-of-function mutation of the CTNNB1 gene was cotransferred. β-catenin activity resulted in fewer infiltrated tumors that did not respond to anti–PD-1–based immunotherapy. In this experimental setting, the mechanism was traced to the silencing of the chemokine CCL5, whose protein product also attracts cDC1 cells into the tumor microenvironment. Again, authors report on a small series of patients with HCC with activation of β-catenin in whom no objective responses to anti–PD-1 immunotherapy were recorded, in clear contrast to those with an intact repressed β-catenin pathway.

An important hypothesis is put forward in the sense that the third of patients with advanced HCC whose tumors show β-catenin gain-of-function mutations are very unlikely to benefit from checkpoint inhibitors, and this should be considered a potential negative predictor to be thoroughly studied as a biomarker candidate, useful for stratification in clinical trials and perhaps helpful to orient physicians choosing among second-line treatment options for patients with HCC.

The field of HCC immunotherapy is thrilled by recent results indicating that the combination of nivolumab with ipilimumab achieves objective responses in more than 30% of cases (Yau and colleagues, ASCO 2019, abstract 4012). It would be interesting to know whether cases responding to this immunotherapy combination also exclude those with β-catenin activation. Moreover, the randomized phase III clinical trial Checkmate 459 comparing sorafenib versus nivolumab as first-line treatment in advanced HCC is due to reveal its results soon. It would be important in it to ascertain whether the patients who would not draw benefit from nivolumab are at least in part those cases with activation of β-catenin.

The importance of β-catenin as a pathway repressing the chemokines that attract BATF3-dependent cDC1s into the tumor microenvironment and thereby avoiding CD8⁺ T-cell infiltration emerges as a general mechanistic feature in cancer (Fig. 1). This important piece of information warrants efforts to pharmacologically target the β-catenin pathway to turn cold tumors into hot tumors, with the ultimate objective of synergizing with immunotherapy strategies.

A recent press release has communicated that the Checkmate 459 trial failed to meet its primary endpoint by a short margin (HR = 0.85 [95% CI: 0.72–1.02]; p = 0.075). These results clearly stress the need for patient selection in such a way that alterations in the β-catenin pathway represent a clear biomarker opportunity, and available pretreatment biopsies from patients in the Checkmate 459 trial should be interrogated in this regard (https://news.bms.com/press-release/bmy/bristol-myers-squibb-announces-results-checkmate-459-study-evaluating-opdivo-nivol).

I. Melero has received commercial research grants from BMS, Roche, Alligator, Bioncotech, and AstraZeneca, has received honoraria from the speakers bureaus of MSD and Roche, and is a consultant/advisory board member for BMS, AstraZeneca, Roche, Numab, Molecular Partners, Bayer, MSD, Catalym, Merck Serono, and Genmab. No potential conflicts of interest were disclosed by the other authors.

I. Melero's scientific research is supported by MINECO SAF2017-83267-C2-1-R (AEI/FEDER, UE), Asociación Española Contra el Cancer (AECC) under grant GCB15152947MELE, Fondo de Investigación Sanitaria-Fondo Europeo de Desarrollo Regional (FEDER) under grants PI14/01686, PI13/00207, and PI16/00668, and H2020 PROCROP project under grant 635122. P. Berraondo is supported by Miguel Servet II (CPII15/00004) contract from Instituto de Salud Carlos III. M.C. Ochoa receives a contract from CIBERONC, and I. Olivera receives a scholarship for FPI program (MICINN).