From glucose dependence to lipid use: histone lactylation in metformin-resistant colorectal CSCs

Why metformin keeps failing in the clinic—and what CSCs have to do with it

Metformin inhibits mitochondrial complex I and activates AMPK, and preclinical work suggested anticancer synergy with chemotherapy and radiotherapy [6]. But late-phase randomized trials, most notably the MA.32 Phase III study of over 3,600 high-risk operable breast cancer patients, found no disease-free survival benefit with adjuvant metformin [6]. The authors of that perspective argue the trials were designed before patient selection, mechanism, and dosing were adequately understood—leaving open the question of which tumors, and which metabolic contexts, might still respond [6]. Cancer stem cells are a plausible missing variable: they self-renew, resist therapy, and dynamically switch between glycolysis and OXPHOS, which is precisely the plasticity that would let a tumor escape a complex I inhibitor [2]. Earlier work showed that OXPHOS-dependent CD133+ pancreatic and cholangiocarcinoma CSCs are metformin-sensitive, but a glycolytic CD133+ subset in PDAC resists it—establishing that CSC metabolic phenotype, not lineage marker alone, determines response [1][2].

The glucose-to-lipid switch: how metformin's own lactate becomes the resistance signal

The anchor study first established the dietary context in vivo: metformin significantly inhibited HCT116 and SW620 xenograft growth under a low-carbohydrate diet, but this effect was markedly diminished under a high-carbohydrate diet and completely abrogated in the HCT116 model [1]. Peripheral blood glucose stayed within physiological range in both diet groups, yet high-carbohydrate-fed mice had higher tumor interstitial fluid glucose—and metformin further increased it rather than lowering it [1]. The same pattern held in a CT26 syngeneic immunocompetent model and in mice given 30% dextrose water, with CD133+ cells enriched after metformin treatment in both settings [1]. In vitro, 25 mM glucose (designated high glucose) reversed metformin-mediated CD133 repression across three CRC cell lines, while 5.5 mM glucose suppressed it—and this glucose dependence was selective for metformin, not standard chemotherapeutics [1]. Mechanistically, metformin increased LDHA-dependent lactate production under high glucose; lactate enhanced H3K18la, which activated c-JUN and induced CD36, driving free fatty acid uptake, lipid droplet formation, and β-oxidation that generated ATP and NADPH and reduced energy stress and ROS [1]. Inhibiting either LDHA or CD36 restored metformin sensitivity in vitro and in vivo [1].

Where this fits in the lactylation literature—and where it diverges

Histone lactylation as a metabolic-epigenetic link was established before this paper. In non-small cell lung cancer, lactate attenuated glycolysis while maintaining mitochondrial homeostasis, and ChIP assays showed increased histone lactylation at HK-1 and IDH3G promoters—demonstrating that lactate can directly reshape metabolic gene expression [7]. In pancreatic cancer, global lactylome profiling identified H3K18la as the most prevalent histone modification with unfavorable prognosis, and a lactate/H3K18la/ACAT2/MTCH2 feedback loop amplified lactate accumulation and drove cholesterol-linked immunosuppression [8]. H3K18la has also been implicated in cisplatin resistance in bladder cancer, where targeted inhibition restored sensitivity, and in CD8+ T cell exhaustion in ovarian cancer via a FOXK1-TOX-lactylation axis [9][3]. The anchor paper's contribution is not the discovery of lactylation but its placement in a therapy-resistance circuit: it identifies metformin itself as the lactate source, and CD36 as the lactylation target that executes the metabolic switch [1]. This is a different logic from the pancreatic cancer feedback loop, where lactylation sustains its own lactate supply through ACAT2/MTCH2 [8]; here, lactylation redirects the cell toward lipid import and oxidation instead.

Competing explanations: fatty acid oxidation as a general CSC dependency, and combination metabolic targeting

The CD36/FAO axis is not unique to colorectal CSCs. In acute myeloid leukemia, stem cells display increased dependence on mitochondrial OXPHOS and fatty acid oxidation, facilitated by transport proteins including CD36—suggesting that lipid reliance may be a broader CSC property rather than a metformin-specific adaptation [4]. This raises the possibility that the anchor paper's findings reflect a pre-existing CSC metabolic preference that high glucose and metformin merely unmask, rather than a resistance mechanism induced de novo. A separate validation study showed that simultaneous inhibition of LDH and complex I produced synergistic effects across a large panel of tumor cells and patient-derived colorectal cancer organoids, and explicitly noted that OXPHOS inhibition often leads to compensatory glycolysis upregulation—which LDH inhibition can counteract [5]. That work supports the anchor paper's logic that LDHA inhibition complements metformin, but it also implies the combination may act through general metabolic collapse rather than the specific H3K18la-CD36 axis [5]. The anchor paper's own data partially address this: CD36 knockdown reduced lipid droplets below PBS control levels even under high glucose, indicating lipid accumulation was CD36-dependent rather than merely acidosis-driven [1].

What the evidence does not yet show—and what would need to be tested

The conclusions rest on CD133+ colorectal cancer cell lines and xenograft/syngeneic mouse models; patient-derived xenografts and clinical samples were not used to validate the LDHA-lactate-H3K18la-CD36 axis [1]. The authors acknowledge that distinguishing selective survival of pre-existing CD133+ cells from active dedifferentiation of non-stem cells will require lineage tracing, and that contributions from other acyl marks or parallel metabolic pathways cannot be excluded [1]. The lactylation field more broadly has relied heavily on cell-line and mouse models—the ovarian cancer study used SKOV3 and ID8 cells with n=3 mice per group and noted that mass spectrometry or IgG negative controls would be needed to confirm TOX lactylation precisely [3]. The bladder cancer study linking H3K18la to cisplatin resistance used single-cell RNA sequencing but is available only as an abstract, limiting assessment of its generalizability [9]. For the anchor paper's translational claim to hold, the key open questions are whether H3K18la and CD36 are elevated in metformin-treated patient tumors, whether dietary carbohydrate intake modifies metformin response in humans as it does in mice, and whether LDHA or CD36 inhibitors can be combined with metformin at tolerable doses [1][6].

About These Sources

This research page is built on 9 peer-reviewed studies — published from 2021 to 2026, 7 from 2024 or later, collectively cited 788 times — selected as the most relevant from 13 studies that passed quality screening, drawn from 57 papers retrieved from a database of over 500 million.

Sources used in this answer