BGPT: Paper Review: Metformin inhibits ovarian cancer via decreasing H3K27 trimethylation

Explore by Goal

Quick Explanation Copied

Paper focus: Metformin (in vitro, epithelial ovarian cancer cell lines) is reported to activate AMPK and reduce H3K27me3 and PRC2 component levels (EZH2/SUZ12/EED), with an EZH2 overexpression “rescue” used to argue that H3K27me3 reduction is functionally required for metformin’s antitumor phenotypes.

Long Explanation

Metformin inhibits ovarian cancer via decreasing H3K27 trimethylation — Visual Critical Review

Target marks/pathway: AMPK activation → PRC2 (EZH2/SUZ12/EED) down → H3K27me3 reduction → altered proliferation/migration/apoptosis

How to read Figure: Values plotted are the percent decreases in proliferation reported in the Results text for each line under 25 mM vs 5.5 mM glucose following 10 mM metformin for 24 h, measured by EdU.

What is “binary” here? This heatmap encodes only what is stated in the text:

Proliferation: metformin reduced EdU proliferation under both glucose conditions in all three lines.
Apoptosis (10 mM metformin): increased apoptosis in SKOV3 and ES2 under normoglycemia, but not under hyperglycemia; A2780 is not described as increasing apoptosis under hyperglycemia in the provided text.
Migration: migration was inhibited in both conditions, with more intensive inhibition under normoglycemia.

Skeptical note: The paper text excerpt does not provide enough numeric apoptosis/migration values here to compute effect sizes; this is therefore a faithful-to-the-text summary rather than a full quantitative meta-comparison.

Important: This plot is intentionally not numeric for AMPK because the excerpt does not provide quantitative p-AMPK/AMPK values; the paper only states dose-dependency and a “slight” increase under hyperglycemia.

1) Experimental design (what they actually did)

Cell models: SKOV3, ES2, A2780 (human epithelial ovarian cancer cell lines) cultured in DMEM with either 25 mM glucose (“hyperglycemic”) or 5.5 mM glucose (“normoglycemic”).

Phenotypes measured: proliferation (EdU), apoptosis (Annexin V/PI flow cytometry), migration (scratch wound healing; Transwell migration).

Mechanism readouts: AMPK pathway activity (AMPKα and phospho-AMPKα Thr172 by western blot), PRC2/H3K27me3 (H3K27me3; EZH2/SUZ12/EED mRNA by RT-qPCR and protein by western blot).

Key perturbations:

EZH2 overexpression (lentivirus) to increase H3K27me3 and test “requirement” for metformin’s effects.
Compound C pretreatment (20 µM) to inhibit AMPK and test whether AMPK activation is needed for PRC2/H3K27me3 suppression.
2-deoxy-D-glucose (2-DG) as another AMPK activator to see whether it mimics PRC2/H3K27me3 changes.
Glucose modulation (5.5 vs 25 mM) to probe how metabolic context changes sensitivity.

2) Results mapped to the proposed causal chain

2.1 Metformin alters core tumor phenotypes

Proliferation: EdU shows metformin (10 mM, 24 h) reduces proliferation in SKOV3, A2780, ES2 under both 25 mM and 5.5 mM glucose.
Apoptosis: flow cytometry reports apoptosis increases under normoglycemia (not hyperglycemia) in SKOV3 and ES2; the study frames apoptosis promotion as condition-dependent.
Migration: both scratch wound healing and Transwell migration assays show migration inhibition by metformin, stronger under normoglycemia; the paper also uses 2.5 mM to argue migration inhibition can occur without increased apoptosis.

2.2 Mechanistic readouts align with the proposed epigenetic mechanism

AMPK activation: metformin increases phospho-AMPKα (Thr172) and the p-AMPKα/AMPKα ratio, with a stronger effect in normoglycemia.
PRC2 and H3K27me3 reduction: metformin reduces H3K27me3 in a dose-dependent manner, especially under low glucose, and decreases EZH2/SUZ12/EED protein and mRNA levels with more modest suppression in hyperglycemia for some components.

2.3 Perturbation logic: “requirement” and “dependence” tests

EZH2 overexpression ‘rescue’: increasing EZH2/H3K27me3 is reported to enhance proliferation and migration and to abrogate metformin’s inhibitory effects on proliferation/apoptosis/migration in SKOV3 and ES2.
Compound C block of AMPK dependence: Compound C pretreatment is reported to prevent metformin-induced AMPK phosphorylation and block reduction in EZH2/SUZ12/EED and H3K27me3.

3) Critical appraisal (skeptical, evidence-weighted)

3.1 Strengths (what supports the claim)

Mechanistic triangulation: The study uses (i) upstream pathway activation (AMPK phosphorylation), (ii) epigenetic marker readouts (H3K27me3/PRC2 components), and (iii) two perturbations with opposite directionality—EZH2 overexpression and AMPK inhibition—to connect marker changes to phenotypes.
Metabolic context control: The paper explicitly tests glucose concentration and reports condition-dependent AMPK/epigenetic effects, which is mechanistically plausible given AMPK as an energy sensor.

3.2 Major limitations / blind spots (what could break the causal story)

In vitro-only evidence: The core experiments are cell line assays without xenograft/in vivo tumor microenvironment testing in the provided text. That limits generalizability to tumors’ metabolic heterogeneity and immune/ECM context.
Epigenetic mechanism is marker-level, not locus-level: The paper measures global H3K27me3 and protein/mRNA abundance of PRC2 components, but it does not (in the provided text) perform genome-wide occupancy/target gene identification (e.g., ChIP-seq/ATAC integration). Thus, “which genes” drive proliferation/migration changes remains unresolved.
Potential tool-specific caveats: Compound C and overexpression systems can have off-target effects or pathway rewirings. The paper reports Compound C blocks metformin’s AMPK phosphorylation and H3K27me3 effects, but it does not (in the provided text) include additional AMPK genetic perturbations or multiple inhibitors for triangulation.
Glucose interpretation ambiguity: The study interprets low glucose as increasing metformin sensitivity, consistent with its AMPK readout differences. However, cell culture glucose changes multiple metabolic pathways beyond AMPK (glycolysis flux, mitochondrial metabolism, ROS), and the excerpt does not establish which metabolic variables causally link to PRC2/H3K27me3.
Statistics transparency (as provided): The Methods state mean ± SD and one-way ANOVA with Tukey post hoc, p<0.05, with assays repeated at least three times. The provided text excerpt does not give sample sizes per assay beyond “triplicate” in some assays, nor does it show exact n for all comparisons.

3.3 What evidence would most likely disprove the paper’s central mechanism?

AMPK-independent metformin effect on H3K27me3: If an AMPK-independent route exists that still reduces H3K27me3 and phenotypes even when AMPK is blocked, the “AMPK required” claim would be weakened. (The paper’s Compound C result argues against this, but tool specificity could matter.)
EZH2/H3K27me3 not causal for phenotypes: If overexpressed EZH2 fails to restore H3K27me3 sufficiently at the relevant time points or fails to rescue migration/apoptosis in repeat experiments, the “H3K27me3 essential” interpretation would weaken.

4) Relationship map (pathway graph)

This directed graph is a structural rendering of the paper’s causal model as described in the abstract/results excerpt: metformin → AMPK activation → PRC2/H3K27me3 reduction → altered proliferation/migration/apoptosis; EZH2 overexpression and Compound C are depicted as perturbations that the paper reports change/abrogate these links.

Next exploratory steps (run BGPT)

EVOLVE (optional)

Author review buttons

Feedback:

Updated: April 23, 2026