Part IV · Molecular biology and therapeutic vulnerabilities · Chapter 19

Metabolism, proteostasis, and stress responses

A pathway with alternates cannot be blocked the way a kinase can.

1 · Glycolytic and oxidative phenotypes across subtypes

Aerobic glycolysis is a statement about flux. It says that glucose is consumed and lactate is produced while oxygen is available. It does not say that mitochondria are idle, and in most breast tumours they are not. A great deal of confusion in this area comes from treating a flux observation as a claim about which organelle is working.

The subtype question is separately confounded by measurement. Fluorodeoxyglucose uptake reports transport and hexokinase activity in a voxel containing tumour, stroma and immune cells. A transcriptomic metabolic signature reports mRNA abundance averaged over the same mixture. Metabolomics reports pool sizes rather than rates. Isotope tracing reports rates, and it is almost never done on human tumours. These four methods answer four different questions, and they are routinely cited as though they answered one.

The cleanest available human data separate metabolic phenotype from transcriptional subtype. In a multi-omic series of 465 triple-negative tumours, three metabolic-pathway-based subtypes were resolved, validated by metabolomic profiling of 72 samples¹. One was lipogenic, with upregulated lipid metabolism. One was glycolytic, with upregulated carbohydrate and nucleotide metabolism. One was mixed. The lipogenic class was the more sensitive to inhibitors of fatty acid synthesis and the glycolytic class to inhibitors of glycolysis. The three classes were distributed unevenly across the expression-defined triple-negative subtypes². A metabolic partition and a transcriptional partition of the same tumours are related without being the same cut.

2 · Mitochondrial biogenesis and oxidative phosphorylation dependence

Some breast cancer cells depend on oxidative phosphorylation, and the dependent population is not the bulk of the tumour. It is the invasive fraction and the fraction that survives treatment.

Migratory and invasive cells favour mitochondrial respiration and raise ATP production³. They do so through the transcriptional coactivator PGC-1α, which drives mitochondrial biogenesis and oxygen consumption. In human invasive breast cancers, PGC-1α expression in the invasive cells correlated with the formation of distant metastases. Silencing PGC-1α suspended invasive potential and attenuated metastasis in those models without affecting proliferation, primary tumour growth or the epithelial to mesenchymal programme. The separation is the informative part. A dependency that governs dissemination but not bulk growth will be invisible to any endpoint measured as tumour shrinkage.

The same theme recurs after chemotherapy. MYC and MCL1 are frequently co-amplified in drug-resistant triple-negative disease after neoadjuvant treatment, and together they raise mitochondrial oxidative phosphorylation and reactive oxygen species⁴. Those two outputs maintained a chemotherapy-resistant stem-like compartment. A mutant MCL1 unable to reach mitochondria reduced that compartment without affecting the anti-apoptotic function of the protein. Function was separable from location.

Two cautions belong with this literature. Oxygen consumption is almost always measured in culture, in media whose substrate composition does not resemble interstitial fluid. And no inhibitor of oxidative phosphorylation has an established role in breast cancer, so the dependency is currently a description rather than a lever. The reasons that matters are developed in Metabolic plasticity as a resistance mechanism.

3 · Lipid metabolism and the adipose interface

The breast is a fat organ, and this makes breast cancer unusual among common solid tumours. The invasive front is in contact with mature adipocytes from the beginning. That contact is a metabolic relationship and not only a mechanical one.

Tumour cells cocultivated with mature adipocytes become more invasive, and the adipocytes change in return⁵. They lose lipid, lose adipocyte markers, and overexpress proteases and proinflammatory cytokines including interleukin 6. These modified cells are identifiable in human breast tumours, and the altered phenotype is most marked around larger tumours and those with nodal involvement.

The transfer itself has been traced. Free fatty acids released by adipocyte lipolysis are taken up by tumour cells and stored as triglyceride in lipid droplets⁶. They are then liberated over time through an adipose triglyceride lipase dependent pathway and oxidised. That oxidation was uncoupled from ATP production, which activated AMP-activated protein kinase and sustained the remodelled state. Blocking the coupled lipolysis and oxidation pathway abolished the invasive gain conferred by coculture.

Two lipid dependencies are therefore in play, and they are not interchangeable. De novo synthesis is one. Uptake of preformed fatty acid from the neighbouring tissue is the other. An inhibitor of fatty acid synthase addresses only the first, and a tumour sitting in fat has the second available. That distinction predicts where such an agent will fail, and it is consistent with the lipogenic subtype being the one that responds to synthesis inhibition¹. Host adiposity and insulin signalling are taken up separately in Obesity, insulin signaling, and metabolic dysfunction.

4 · Amino acid dependencies and one-carbon metabolism

Amino acid dependencies are attractive targets because they can be attacked by restricting supply rather than by inhibiting an enzyme. They are also the clearest example in this chapter of a dependency that is real, selective and still not actionable.

Phosphoglycerate dehydrogenase catalyses the first step of serine synthesis. It lies in a region of recurrent copy number gain in breast cancer, and its protein was elevated in 70% of the oestrogen receptor negative breast cancers examined⁷. Suppressing it reduced proliferation in cell lines with high expression and not in those without, which is what a genuine dependency looks like. The mechanism was not depletion of intracellular serine. It was loss of α-ketoglutarate, because in high-expressing cells the serine pathway carried roughly half of the total anaplerotic flux of glutamine into the tricarboxylic acid cycle.

Asparagine behaves differently again. Asparagine synthetase expression in a primary tumour correlated with later metastatic relapse, and limiting asparagine reduced metastasis without affecting growth of the primary tumour⁸. Three separate manipulations gave the same direction: knockdown of the synthetase, treatment with asparaginase, and dietary restriction. Raising availability moved metastasis the other way.

Serine also feeds the folate cycle, which supplies both nucleotide synthesis and the methyl donor pool. That is the point at which this chapter connects to Epigenetics, transcriptional control, and lineage identity. A cell's capacity to maintain DNA and histone methylation is a metabolic capacity, so a metabolic state can constrain an epigenetic one.

5 · Redox homeostasis, NRF2, and nucleotide synthesis

Reactive oxygen species are both a weapon and a signal. Proliferation generates them, several treatments generate more, and a cell that cannot buffer them dies. A cell that buffers them well survives conditions that would otherwise be lethal. Antioxidant capacity is therefore a survival trait under treatment rather than a general protective good.

NRF2 is the transcription factor that sets that capacity, and its behaviour in breast cancer models is unusually well characterised. Downregulating HER2 in breast cancer cells produced oxidative stress and a compensatory rise in NRF2⁹. NRF2 was activated during dormancy and in recurrent tumours. Constitutive activation accelerated recurrence and suppression impaired it. The reprogramming it drove did two things at once: it re-established redox homeostasis, and it upregulated de novo nucleotide synthesis. Those are the two requirements for a quiescent cell to start dividing again.

That result is why this section is flagged forward. NRF2 returns in Dormancy, residual disease, and late recurrence not as a redox mechanism but as a dormancy reactivation mechanism, and specifically in Metabolic control of dormancy and reactivation. The same transcription factor is doing the same thing in both chapters. Only the clinical question changes.

The persister literature points the same way. Cells that keep cycling under continuous drug exposure upregulate antioxidant gene programmes and shift toward fatty acid oxidation¹⁰. Impeding oxidative stress handling or the metabolic shift changed the fraction of persisters able to cycle. The same programmes were detectable in minimal residual disease in human tumours under several targeted therapies.

6 · Translational control, eIF4A, and selective protein synthesis

Translation is a control point because messenger RNAs are not equally easy to translate. Those with long or structured 5' untranslated regions need helicase activity to be read. Blocking that helicase therefore lowers a subset of proteins rather than all of them. Selectivity is the whole appeal.

The subset matters in this disease. Translation of the oestrogen receptor is independent of the cap-binding factor eIF4E and yet sensitive to inhibitors of the helicase eIF4A¹¹. Inhibiting eIF4A lowered receptor protein and cyclin D1 together. Growth was suppressed in ligand-independent models, including those driven by ESR1 mutants and by ESR1 fusions. Adding fulvestrant lowered receptor expression further and blocked tumour growth in vitro and in vivo.

The strategic point is worth stating plainly. Every endocrine agent in current use engages the receptor itself, which is why mutations in the ligand-binding domain matter so much in Estrogen receptor signaling and endocrine resistance and Temporal heterogeneity and clonal evolution. Lowering the amount of receptor synthesised is a different point of attack. It does not require the ligand-binding pocket to be intact, and it does not care which allele is present.

In an early-phase trial, the eIF4A inhibitor zotatifin was combined with fulvestrant, or with fulvestrant and abemaciclib, in patients with acquired resistance to those agents¹¹. Clinical responses were seen, including some durable regressions, with little toxicity. That is an uncontrolled early signal and nothing more. It is reported here because the mechanism is specific and the class is unfamiliar, not because the efficacy question is answered.

7 · Proteostasis, unfolded protein response, and autophagy

Proteostasis is a balance between load and capacity. An amplified oncogene raises synthetic load. Hypoxia, nutrient limitation and several drugs lower folding capacity. The unfolded protein response is the sensor that reconciles the two, and in breast cancer it is not a passive stress reaction.

XBP1 is the arm of that response with the clearest oncogenic role here. It is activated in triple-negative disease, and depleting it inhibited tumour growth and relapse and reduced the CD44-high CD24-low population¹². The mechanism was transcriptional rather than proteostatic in the narrow sense. XBP1 assembled a complex with HIF1α and recruited RNA polymerase II to hypoxia-inducible factor targets. An XBP1 expression signature correlated with hypoxia signatures and with poor prognosis in independent cohorts.

Autophagy is the disposal arm, and it matters most in cells that are not dividing. It was required for the survival of disseminated dormant breast cancer cells in mouse and human models, with ATG7 essential in vivo¹³. Blocking autophagic flux caused damaged mitochondria and reactive oxygen species to accumulate, and the cells then underwent apoptosis. This is the mechanistic argument behind interest in intercepting residual disease, developed in Dormancy, residual disease, and late recurrence.

Autophagy is not a single-signed process, and that limits what can be done with it. It restrains transformation early and supports survival in established and dormant disease. Inhibit autophagy is therefore not a coherent instruction without a stated disease stage. The available clinical tool compounds the problem. Hydroxychloroquine blocks lysosomal acidification, which affects autophagy without being specific to it. A negative trial using it does not cleanly test the hypothesis it was chosen to test.

8 · Metabolic plasticity as a resistance mechanism

A kinase inhibitor blocks a node. A metabolic inhibitor blocks a node that has alternates. This is the reason metabolic targeting has repeatedly failed to behave like targeted therapy, and it is the same redundancy argument made structurally in Pathway redundancy, feedback, and why single-node inhibition fails.

The evidence that plasticity is the mechanism of survival rather than a byproduct of it is now reasonably consistent across systems. Chemotherapy-resistant stem-like cells in triple-negative disease depend on mitochondrial oxidative phosphorylation⁴. Dormant cells that resume growth depend on NRF2-driven redox and nucleotide reprogramming⁹. Persisters that keep cycling depend on antioxidant programmes and fatty acid oxidation¹⁰. In each case the surviving population moved to a different metabolic state rather than continuing in the one that was blocked.

The inversion is the useful part. A cell that has committed to a state has acquired the liabilities of that state. Drug-tolerant persister cells arising across a wide range of cancers and treatments acquire a dependency on the lipid hydroperoxidase GPX4¹⁴. Removing GPX4 function killed those cells by ferroptosis and prevented tumour relapse in mice. Plasticity generates resistance and creates a new vulnerability in the same movement.

Against that biology sits the clinical record, which is thin and should be stated without softening. Metformin was tested against placebo for five years in 3649 patients with high-risk non-metastatic breast cancer, with the primary analysis in the 2533 whose tumours were hormone receptor positive¹⁵. Invasive disease-free survival events occurred at 2.78 per 100 patient-years with metformin and 2.74 with placebo, a hazard ratio of 1.01 with a 95% confidence interval of 0.84 to 1.21. Grade 3 non-haematological adverse events were more frequent with metformin, at 21.5% against 17.5%.

Where this chapter is heading is worth naming. The therapeutic question is taken up in Metabolic and mitochondrial targeting and the resistance framing in Mechanisms of therapeutic resistance. The honest summary is that metabolic dependencies in breast cancer are real, are subtype-associated and state-associated, and have so far produced one large negative randomised trial and no established agent.

References

1. Gong Y, Ji P, Yang YS, et al. Metabolic-pathway-based subtyping of triple-negative breast cancer reveals potential therapeutic targets. Cell Metab 2020 33:51-64. PMID 33181091
2. Lehmann BD, Bauer JA, Chen X, et al. Identification of human triple-negative breast cancer subtypes and preclinical models for selection of targeted therapies. J Clin Invest 2011 121:2750-2767. PMID 21633166
3. LeBleu VS, O'Connell JT, Gonzalez Herrera KN, et al. PGC-1alpha mediates mitochondrial biogenesis and oxidative phosphorylation in cancer cells to promote metastasis. Nat Cell Biol 2014 16:992-1003. PMID 25241037
4. Lee KM, Giltnane JM, Balko JM, et al. MYC and MCL1 cooperatively promote chemotherapy-resistant breast cancer stem cells via regulation of mitochondrial oxidative phosphorylation. Cell Metab 2017 26:633-647. PMID 28978427
5. Dirat B, Bochet L, Dabek M, et al. Cancer-associated adipocytes exhibit an activated phenotype and contribute to breast cancer invasion. Cancer Res 2011 71:2455-2465. PMID 21459803
6. Wang YY, Attane C, Milhas D, et al. Mammary adipocytes stimulate breast cancer invasion through metabolic remodeling of tumor cells. JCI Insight 2017 2:e87489. PMID 28239646
7. Possemato R, Marks KM, Shaul YD, et al. Functional genomics reveal that the serine synthesis pathway is essential in breast cancer. Nature 2011 476:346-350. PMID 21760589
8. Knott SRV, Wagenblast E, Khan S, et al. Asparagine bioavailability governs metastasis in a model of breast cancer. Nature 2018 554:378-381. PMID 29414946
9. Fox DB, Garcia NMG, McKinney BJ, et al. NRF2 activation promotes the recurrence of dormant tumour cells through regulation of redox and nucleotide metabolism. Nat Metab 2020 2:318-334. PMID 32691018
10. Oren Y, Tsabar M, Cuoco MS, et al. Cycling cancer persister cells arise from lineages with distinct programs. Nature 2021 596:576-582. PMID 34381210
11. Boyer JA, Rosen EY, Sharma M, et al. eIF4A controls translation of estrogen receptor alpha and is a therapeutic target in advanced breast cancer. Proc Natl Acad Sci U S A 2025 122:e2424286122. PMID 40690678
12. Chen X, Iliopoulos D, Zhang Q, et al. XBP1 promotes triple-negative breast cancer by controlling the HIF1alpha pathway. Nature 2014 508:103-107. PMID 24670641
13. Vera-Ramirez L, Vodnala SK, Nini R, Hunter KW, Green JE. Autophagy promotes the survival of dormant breast cancer cells and metastatic tumour recurrence. Nat Commun 2018 9:1944. PMID 29789598
14. Hangauer MJ, Viswanathan VS, Ryan MJ, et al. Drug-tolerant persister cancer cells are vulnerable to GPX4 inhibition. Nature 2017 551:247-250. PMID 29088702
15. Goodwin PJ, Chen BE, Gelmon KA, et al. Effect of metformin vs placebo on invasive disease-free survival in patients with breast cancer: the MA.32 randomized clinical trial. JAMA 2022 327:1963-1973. PMID 35608580