Restored Ketosis Drives Anticancer Immunity in Colorectal Cancer

A large panel of dietary interventions, including supplementation with specific vitamins (1) or other bioactive factors (2), as well as the implementation of diets low in calories (2) or specific amino acids (3), has been shown to mediate antineoplastic effects in numerous preclinical tumor models. However, the underlying mechanisms appear to vary considerably in different experimental settings and globally remain poorly understood. Depriving cancer cells of nutrients that are essential to maintain the transformed state (3), directly inhibiting cancer cell metabolism (1), activating autophagic responses that promote anticancer immunosurveillance as a consequence of improved danger signaling (2), altering the gut microbiome (4), and normalizing systemic glucose metabolism (5) are only few of the mechanisms that have been suggested to explain the ability of dietary alterations to inhibit tumor growth (at least in mice). In this issue of Cancer Research, Wei and colleagues report that ketogenesis is generally inhibited in colorectal cancer and that restoring it with genetic interventions or a ketogenic diet (KD), which is rich in lipids and poor in carbohydrates, activates an immunometabolic axis, resulting in the downregulation of Krüppel like factor 5 (KLF5) in cancer-associated fibroblasts (CAF), leading to: (i) suppressed secretion of C-X-C motif chemokine ligand 12 (CXCL12), (ii) a switch from an immunosuppressive to an immunostimulatory tumor microenvironment (TME), and (iii) accrued sensitivity to an immune checkpoint inhibitor (ICI) specific for programmed cell death 1 (PDCD1, best known as PD-1; ref. 6).

Following up on previous findings from the same team suggesting that ketogenesis is involved in intestinal differentiation (7), Wei and colleagues tested the impact of this metabolic pathway on colorectal cancer. While the ketone body beta-hydroxybutyric acid (βHB) failed to affect the proliferation of human and mouse colorectal cancer cells maintained in vitro, a KD effectively reduced the growth of mouse MC38 colorectal cancers established in immunocompetent syngeneic (but not immunodeficient) mice, correlating with an increase in CD8⁺ cytotoxic T lymphocytes and a decrease in immunosuppressive myeloid cells in the TME. Single-cell RNA sequencing (scRNA-seq) studies confirmed these findings and revealed that (i) other immune effector cells, notably natural killer (NK) cells and Th1 CD4⁺ T cells were enriched in colorectal cancer tumors from mice fed a KD, and (ii) immunosuppressive CD4⁺CD25⁺FOXP3⁺ regulatory T (Treg) cells as well as myeloid-derived suppressor cells (MDSC, an immature and heterogeneous population of cells that mediate potent immunosuppression) were depleted from the TME upon administration of the KD (6). These findings indicate that the therapeutic activity of restored ketogenesis is largely mediated by the reactivation of immunosurveillance upon reconfiguration of the immunologic colorectal cancer microenvironment.

Based on the prominent role of CAFs in the establishment of an immunosuppressive TME in multiple tumors (8), Wei and colleagues hypothesized that the ability of ketogenic interventions to reconfigure the colorectal cancer microenvironment would involve stromal cells. Indeed, primary human CAFs genetically engineered to overexpress catalytically active (but not catalytically inactive) 3-hydroxy-3-methylglutaryl-CoA synthase 2 (HMGCS2) produced increased quantities of βHB and expressed limited amounts of the immunosuppressive cytokine CXCL12. Similar results were obtained by treating primary human CAFs (but not normal human fibroblasts) with βHB. Moreover, CAFs isolated from MC38 tumors exposed to a KD exhibited reduced Cxcl12 mRNA levels. Of note, the overexpression of wild-type (but not catalytically inactive) HMGCS2, as well as βHB treatment, caused the downregulation of the transcription factor KLF5 in primary human CAFs, at least in part via a pathway involving the inhibition of class I histone deacetylases (HDAC). Consistent with a mechanistic implication for KLF5 in CXCL12 production by CAFs, human primary CAFs genetically engineered to overexpress KLF5 secreted increased amounts of CXCL12 as compared with their control counterparts infected with a GFP-expressing construct. Conversely, pharmacologic and genetic inhibition of KLF5 suppressed CXCL12 synthesis by human primary CAFs. Moreover, the ability of βHB to limit CXCL12 secretion was lost upon transgene–driven KLF5 overexpression. Two putative KLF5-binding sites were identified in the Cxcl12 promoter, one of which was largely occupied by KLF5 (in chromatin immunoprecipitation assays) and accounted for most Cxcl12 promoter activity upon KLF5 overexpression (in binding site deletion assays; ref. 6). Altogether, these data suggest that restored ketogenesis suppresses Cxcl12 transcription in CAFs through downregulation of KLF5.

Next, Wei and colleagues assessed the proliferative response of human primary CAFs driven into ketogenesis, revealing a marked cytostatic effect following both HMGCS2 overexpression and βHB administration. In line with this notion, a KD reduced the expression of actin alpha 2, smooth muscle (ACTA, a marker of CAFs) in the microenvironment of MC38 tumors as well as CT26 tumors (another mouse model of colorectal cancer) growing in immunocompetent syngeneic hosts. Moreover, while KLF5 overexpression increased the proliferation of human primary CAFs, genetic KLF5 inhibition limited it, consistent with the existence of an axis linking ketogenesis to suppressed CAF functions (6).

Importantly, Wei and colleagues detected decreased βHB and/or HMGCS2 levels, as well as increased amounts of CXCL12 in a majority of samples from human colorectal cancer as compared with adjacent normal tissues. Moreover, HMGCS2 expression was found to negatively correlate with KLF5 and CXCL12 levels not only in primary colorectal cancer samples, but also in colorectal cancer liver metastases (6). These findings align with previous data demonstrating a negative prognostic impact for high intratumoral levels of KLF5 in patients with colorectal cancer (9). Further corroborating these observations, restoring ketogenesis with a KD not only elevated βHB and reduced CXCL12 in both the plasma and tumors of mice bearing MC38 or CT26 tumors, but it also amplified the therapeutic effects of a PD-1 blocker. Finally, conditioned medium from human primary CAFs overexpressing wild-type (but not catalytically inactive) HMGCS2 limited the migratory and invasive capacity of human colorectal cancer cells in vitro. Moreover, the metastatic dissemination of intravenously injected CT26 cells to the liver and lungs was limited in mice fed a KD, correlating with increased βHB and decreased CXCL12 at metastatic sites (6). While the involvement of the host immune system remains to be formally established, these findings link suppressed ketogenesis to colorectal oncogenesis (in humans) and accrued metastatic dissemination (in mice).

In summary, Wei and colleagues delineated a novel immunometabolic axis linking inhibited ketogenesis to the establishment of an immunosuppressive colorectal cancer microenvironment by CAFs. Such a mechanism of immunoevasion could be reverted (at least in mice) by restoring ketogenesis with nutritional interventions (Fig. 1). These findings are particularly relevant as only a fraction of patients with colorectal cancer exhibit an inflamed TME at baseline (and hence respond to ICIs; ref. 10), implying that clinically actionable strategies to reverse immunosuppression in colorectal cancer are needed. Despite a variety of obstacles including issues linked to compliance, ketogenic diets have recently attracted considerable attention in this respect, and >20 clinical trials are currently evaluating the utility of this approach in patients with various tumors according to www.clinicaltrials.gov. While the results of these studies are urgently awaited, the global focus on nutritional interventions as tools to ameliorate quality of life and/or treatment efficacy in patients with cancer aligns with the recent recognition of other, previously overlooked factors that may ultimately influence disease outcome in this patient population, including stress, lifestyle, and over-the-counter medications. We surmise that an increased consideration of these factors during the analysis of retrospective patient series, coupled to properly controlled preclinical experiments and prospective clinical studies, will improve clinical cancer care.

L. Galluzzi reports grants from Lytix and Phosplatin; personal fees from Boehringer Ingelheim, AstraZeneca, OmniSEQ, The Longevity Labs, Inzen, Luke Heller TECPR2 Foundation, and personal fees from Sotio outside the submitted work. No disclosures were reported by the other author.

D.C. Montrose is supported by NIH/NCI (#K22CA226033), a grant from the American Pulse Association, and startup funds from the Stony Brook Cancer Center (Stony Brook, NY) and Bahl Center for Metabolomics and Imaging (Stony Brook, NY). The L. Galluzzi lab is supported by a Breakthrough Level 2 grant from the US DoD BRCP (#BC180476P1), by the 2019 Laura Ziskin Prize in Translational Research (#ZP-6177, PI: Formenti) from the Stand Up to Cancer (SU2C), by a Mantle Cell Lymphoma Research Initiative (MCL-RI, PI: Chen-Kiang) grant from the Leukemia and Lymphoma Society (LLS), by a startup grant from the Dept. of Radiation Oncology at Weill Cornell Medicine (New York, NY), by a Rapid Response Grant from the Functional Genomics Initiative (New York, NY), by industrial collaborations with Lytix Biopharma (Oslo, Norway) and Phosplatin (New York, NY), and by donations from Phosplatin (New York, NY), the Luke Heller TECPR2 Foundation (Boston, MA), Sotio a.s. (Prague, Czech Republic), Onxeo (Paris, France), Ricerchiamo (Brescia, Italy), and Noxopharm (Chatswood, Australia).