IDH1 Inhibition Reawakens the Immune Response against Cholangiocarcinoma

Cholangiocarcinoma is an aggressive malignancy of the bile ducts with dismal outcomes (1). Intrahepatic cholangiocarcinoma (ICC), located above the secondary biliary tree inside the liver, is characterized by a 10% to 20% prevalence of IDH1/2 mutations, which convert α-ketoglutarate to the oncogenic R-2-hydroxyglutarate (R-2HG) through NADPH-dependent reduction (2). Currently, patients with ICC have few therapeutic options, and early immune checkpoint blockade trials with single-agent anti–PD-1 show limited efficacy, with an overall response rate (ORR) ranging from 5.8% to 17% (3). Therefore, identification of novel therapeutic modalities is a pressing need.

Ivosidenib (AG120) is an oral and selective small-molecule mIDH1 inhibitor that can significantly decrease intracellular and circulating R-2HG. Excitingly, AG120 has recently been approved by the FDA for the treatment of advanced or metastatic mIDH1 ICC, as well as for adults with relapsed or refractory mIDH1 acute myeloid leukemia (4). However, the final results of the phase III ICC trial (n = 185) indicated a statistically significant but rather limited clinical benefit: AG120 produced only a 2% ORR and 51% stable disease (SD) versus 0% ORR and 28% SD for placebo (5), and a more promising difference of 10.3 versus 5.1 months for overall survival when adjusted for cohort crossover. Similarly low response rates were seen in a phase I study of advanced mIDH1 glioma, although the median progression-free survival compared favorably to temozolomide (6). These results suggest that combination therapies with AG120 have room to improve on clinical responses, particularly for solid cancers.

Reflecting single-agent mIDH1 inhibitor limitations, in a mouse model of mIDH1-driven glioma, removal of mIDH1 after tumor establishment did not induce tumor regression (7). In ICC, pathologic examination of pre- and post-AG120 biopsies from patients experiencing clinical benefit identified consistent morphologic tumor changes, a liver differentiation signature, and increased PD-L1 (8). Together, these findings suggest that there may be a limited degree to which solid tumors rely on mIDH1 for tumor maintenance, and that rational combination therapies may require regulation of the tumor immune microenvironment to invoke strong responses.

To begin to address these issues, Wu and colleagues developed Alb-Cre-Kras^G12D-IDH1^R132C/H genetically engineered mouse models of ICC (9). Compared with IDH1^R132H, mice harboring IDH1^R132C induce higher levels of R-2HG, a higher proportion of ICC, and lower tumor-free survival. These results are consistent with the much higher prevalence of IDH1^R132C mutations in patients with ICC compared with IDH1^R132H. After the establishment of ICC cell lines from the IDH1^R132C mice, AG120 significantly reduced R-2HG levels and inhibited tumor growth in syngeneic allografts but did not induce regressions, consistent with the clinical data. By contrast, no effect of AG120 on tumor growth was seen in severely immunodeficient NOD-scid IL2rg^−/− (NSG) mice harboring the same mIDH1 cell lines, confirming the involvement of an immune response in the antitumor effect of mIDH1 inhibition.

Accordingly, RNA sequencing and IHC analyses of the AG120-treated immunocompetent mouse tumors showed upregulation of IFNγ response genes and an increase in CD8⁺ T-cell infiltration; correspondingly, lower CD8⁺ tumor infiltration was seen in mIDH1 mouse and human ICCs compared with matched IDH1 wild-type ICCs. Critically, antibody-mediated CD8⁺ T-cell depletion completely abolished AG120-mediated growth inhibition in the mouse model. Moreover, R-2HG inhibited CD8⁺ T-cell activation in vitro, providing a mechanism. These results therefore show a central role of T-cell immunity evasion in mIDH1 ICC and its consequent central role in the AG120 response.

Given the known inhibition of TET2 by mIDH1-induced R-2HG, Wu and colleagues explored the role of TET2, which hydroxymethlates DNA, in mIDH1 ICC. Indeed, IDH1^R132C mouse ICCs showed the lowest level of global 5-hydroxy-methylcytosine compared with IDH1^R132H and IDH1^WT tumors, and this can be reversed by AG120 treatment. Using CRISPR, TET2 was then knocked out (KO) in the IDH1^R132C mouse ICC cell line. When TET2-KO tumors were treated with AG120, IFNγ response genes were no longer induced and tumor growth inhibition was abolished. Surprisingly, this occurred without affecting AG120-induced CD8⁺ T-cell infiltration. Accordingly, AG120 was also found to increase MHC class I molecules in TET2 wild-type but not TET2-KO cells. These data suggested that TET2 mediates the immune response to mIDH1 inhibition by epigenetically activating IFNγ response and MHC-I genes independently of CD8⁺ T-cell recruitment.

Because PD-L1 has been shown to increase after AG120 treatment, the synergistic effect between anti–PD-1 blockade and AG120 was explored; however, their combination was not synergistic. It was hypothesized that immunosuppressive regulatory T cells (Treg) may be compensatory, as they increased in number after treatment. To target Tregs, the combination of anti-CTLA4 antibodies with AG120 was tested and showed strong synergy in tumor growth inhibition, tumor cell apoptosis, and Treg suppression. Indeed, 9 of 10 mice with the combination treatment showed significant tumor regression including 3 with a complete response. Critically, the 3 complete responses showed full protection against reinjection of mIDH1 ICC cell lines compared with tumor-naïve mice, demonstrating a retained immune memory.

This study is the first to identify a robust combination therapy for mIDH1 ICC and is also the first to demonstrate the central role of the immune microenvironment in mIDH1-mediated tumor maintenance. This provides one potential explanation for the low objective responses in clinical trials with AG120 and also solves the conundrum of the previously reported lack of AG120 effect on immunodeficient xenograft ICC models and on cell lines in vitro. These findings place mIDH1 in stark contrast to more traditional oncogenes such as BRAF, whose inhibitors only partially instead of completely rely on an immune response for efficacy. The identification of CTLA4 as a robust combination therapy, if clinically successful, would therefore convert mIDH1 from weak to strong oncogene addiction by incorporating key microenvironmental vulnerabilities. However, it remains to be seen whether results from a single mouse model can be more broadly generalizable, and if so, whether it could also be applicable to other mIDH1 solid cancers such as subsets of glioma, melanoma, and chondrosarcoma.

No disclosures were reported.