Starve a cold, and perhaps a cancer

Abstract

Stem cells tightly link their metabolism to cell fate decisions; however, how cancers co-opt and bypass metabolic pathways for growth advantage remains unclear. New evidence in this issue highlights how cancer stem cells avoid epigenetically driven differentiation by shutting down endogenous serine synthesis and becoming serine auxotrophs.

For normal or cancer stem cells (SCs) to maintain their stemness, they must balance both cell intrinsic and extrinsic inputs, maintaining proliferation while inhibiting unwanted differentiation or immune recognition. Epigenetic regulators are effectors of these inputs, fine-tuning transcriptional responses that facilitate this tight-rope act, with alterations often leading to cell death and tissue exhaustion, or oncogenic transformation and growth¹.

Emerging data tightly link metabolic pathway inputs with the genome’s epigenetic state¹. Extensive work has assessed distinct cancer types for their differential metabolic vulnerabilities, with a goal to ‘starve a cancer’¹. Variables such as the tumour’s cell or tissue of origin, microenvironment, and mutational status can dramatically influence the reliance on distinct metabolic pathways. Moreover, oncogenes such as Ras, PI3K, MEK/ERK, AKT, mTOR, MYC, and p53 co-opt metabolic activities in their pro-tumourigenic capacities¹.

Extensive reviews of cancer metabolomics have linked extracellular nutrient availability, intracellular metabolic machinery, and the ability to alter metabolic preferences to cancer cell fate decisions and clinical outcomes. Growing evidence expands on this well-established balancing act involving the non-essential amino acids serine, glycine, and glutamine and their potential to alter repressive epigenetic histone and DNA methylation modifications^1,2. Serine, in particular, has been implicated with indirectly affecting tumour cell fate³.

All three amino acids are needed as building blocks for protein synthesis and cellular replication, and their scarcity results in slowed proliferation. The trio are also precursors for one-carbon metabolism that supplies the key methyl donor S-adenosylmethionine (SAM), enabling the Polycomb complex to promote lysine 27 methylation on histone H3 (H3K27me3) and DNA methyltransferases to methylate DNA^1,3. Interestingly, the same trio also supplies precursors in the opposing reaction mediated by Jumonji domain-containing histone demethylases (JHDM) and the ten–eleven-translocation (TET) enzymes. JHDM and TET promote demethylation contingent on the byproduct of endogenous serine biosynthesis, α-ketoglutarate (αKG)¹ (Fig. 1). SCs can use these metabolite inputs in conjunction with the aforementioned epigenetic modifiers to regulate their genome and subsequent cell fates. Quite interestingly, cancers and their cell fates are also influenced by this metabolite–epigenetic balance. However, cancer-associated mutations affecting these inputs can provide a convenient growth advantage in select tumour microenvironments.

Fig. 1 ∣. Differential serine metabolic pathways are dependent on epidermal cellular states.

a, Schematic summary of endogenous (2, 3, and αKG, red) and exogenous (1 and 3, red) serine metabolic pathways. b, Summary diagram of the key epidermal cellular states (WT, wild type) EpdSCs, oncogenic EpdSCs, and SCC) in close proximity to blood vessels, a source of extracellular serine. c, Graph depicting the differential needs of serine between the different epidermal cellular states and the potential metabolic paths from a utilised to respond to limited basal concentrations from the environment.

In this issue of Nature Cell Biology, Baksh et al. used murine epidermal stem cells (EpdSCs), oncogenic EpdSCs transitioning to squamous cell carcinoma (SCC), and SCC tumour models to illustrate that the source of serine, whether from the extracellular microenvironment or endogenously synthesised, represents a unique solution to the problem of balancing metabolic proliferation and differentiation⁴. EpdSCs prevented differentiation through H3K27me3 maintenance and DNA methylation, and maintained low levels of αKG and of JHDM and TET demethylation. Ectopic Sox2 expression in EpdSCs altered EpdSC cell fate towards an initial neoplastic SCC-like state (oncogenic EpdSCs). Because of higher serine needs from the elevated proliferative rate and to prevent differentiation, oncogenic EpdSCs circumvented elevated levels of αKG and αKG-dependent dioxygenases by shutting down endogenous serine synthesis and becoming serine auxotrophs (Fig. 1)⁴.

Through methods involving glucose tracing, and a series of inhibitors and knockdown studies in cells and transgenic mice mouse models that captured both the initial neoplastic and full SCC state, the authors beautifully demonstrated that serine auxotrophy occurred through affecting the redox status of the cytoplasm, with a high NAD⁺/NADH ratio inhibiting the rate-limiting enzyme phosphoglycerate dehydrogenase (PHGDH). Although the oncogenic EpdSCs grow slower due to a lack of other byproducts and a reduction in protein synthesis, these neoplastic cells successfully escaped unwanted differentiation and maintained H3K27me3 and the associated stem- like properties. In this trade-off, starving the SCC tumours through the dietary withdrawal of serine both further reduced growth and promoted differentiation, presumably because of the tumour’s need to engage serine synthesis and produce αKG⁴. Understanding how these needs and strategies directly regulate cell fate consequences sheds light on the critical interplay between metabolite availability, cellular metabolism, and epigenetic regulation (Table 1).

Table 1 ∣.

Summary table of distinct cell types (epidermal cell states, ES cells, and different cancers) and the cell fate consequences of their methylation status

Cell type/differentiation state	WT EpdSCs	Oncogenic EpdSCs	SCC	ES	Lymphoma/colon	Breast/melanoma
High αKG: Histone demethylation DNA demethylation	Differentiation	Differentiation	Less aggressive (more differentiated)	Pluripotency	Less aggressive?	More aggressive?
High 1C Histone methylation DNA methylation	Basal state	Basal state	More aggressive (less differentiated)	Differentiation	More aggressive?	Less aggressive?

1C, one carbon.

The work provides an opportunity to generalise and draw parallels with other tumours and acknowledge that although the ‘starving’ approach has merit, it remains dependent on tumour type and context. Metabolic and epigenetic regulation within the same cell type (EpdSC vs. oncogenic EpdSC) and a cancer-associated subtype (oncogenic EpdSC vs. SCC) vary in response to serine levels (Fig. 1)⁴. The context-specific roles of serine may, in part, be explained by the need for endogenous serine byproducts and the availability of extracellular serine. Microenvironmental serine comes in part from the vasculature, which plays an intimate role in providing nutrients to tumours. Vascular-derived serine in neoplastic stages could provide sufficient levels to promote oncogenic EpdSC growth and proliferation, with increased metabolic demands in larger tumours, providing a selective environment for vascularised SCC tumours to prevent differentiation (Fig. 1). Interestingly, when starved of serine, oncogenic EpdSCs choose not to engage in endogenous serine synthesis. Although their proliferation is serine dependent, we speculate that a ‘starving’ approach would likely not target these neoplastic lesions, as their proliferation rate is low enough and their nutrient supply sufficient to maintain an undifferentiated state.

The tumour context also applies to the oncogenic stimulus creating the tumour. The authors initially compared EpdSCs with those expressing the Sox2 oncogene and showed that serine starvation affects EpdSCs whether driven by activated oncogene H-ras or through a loss of the tumour suppressor p53 (ref. ⁴). However in other K-ras-driven models such as pancreatic tumours, starvation fails despite K-ras induction of endogenous serine synthesis enzymes because of compensatory changes preventing overt differentiation⁵. Other oncogenes also induce elevated serine synthesis enzymes, but in the absence of compensation, metabolically convert to serine auxotrophy to prevent differentiation.

The central role for serine in linking proliferation and differentiation highlights additional context-specific roles that vary in cancer type and stage^5-9. In SCCs, H3K27me3 demethylation itself could be an alternative method to regulate cell fate. In support of this strategy, human tumours that have high levels of H3K27me3 (associated with a less differentiated state) present as higher grade and more aggressive SCC tumours. Given the distinct responses to serine starvation between oncogenic EpdSCs and SCC models, alternative strategies that lead to removal of H3K27me3 and subsequent differentiation such as cell-permeable αKG, forced expression of histone demethylases such as JHDM, or EZH2 inhibitors could prove useful in both preventing neoplasia formation and treating more advanced SCCs. Extensive work has uncovered EZH2 as a regulator of stem cell fate, as well as a targetable oncogene or tumour suppressor, further suggesting differential roles of H3K27me3 in a context-specific manner^10,11.

While some SCs rely on maintaining H3K27me3 status, other normal and oncogenic SCs require the opposite. Human embryonic stem cells maintain their high proliferation and pluripotency by removing divalent marks of H3K27me3 and DNA methylation and maintaining high endogenous serine synthesis and αKG-mediated dioxygenase activity^1,12.

Work in certain melanoma and breast cancer types growing in serine-limited environments shows that increased serine synthesis provides a growth advantage⁹. The underlying basis remains poorly understood, but could stem from the fact that melanoma are derived from the neural crest rather than the surface ectoderm for SCC or the possibility that melanomas have developed an unidentified compensatory mechanism. Differential sensitivity and responses to limited environmental serine in distinct cancer types, stages, and contexts warrant further investigation. Additionally, the work presented by Baksh et al. indicated that distinct stages of tumour progression (oncogenic EpdSCs vs. SCCs) have distinct mechanisms to respond to lack of serine,suggesting that other cancers may have more complex stage-specific metabolic needs and constraints.

An intriguing aspect uncovered by Baksh et al. relates to the mechanistic switch to auxotrophy used by tumours. They showed that the trigger is low cytoplasmic NAD⁺/NADH ratios deriving from mitochondrial pyruvate metabolism. As NAD⁺ is required for the PHGDH enzymatic reaction, a lack of lactate dehydrogenase A (LDHA)-mediated NAD⁺ blocked both PHGDH and downstream αKG levels. While operative in oncogenic EpdSCs, LDHA levels become elevated during tumour progression in many cancers, suggesting that there are other mechanisms to switch off endogenous serine synthesis, or that increased LDHA elevation comes with concomitant compensatory mechanisms to regulate histone and DNA demethylation¹³. Additional complexity comes from the frequent overexpression and amplification in prostate, ovarian, and pancreatic cancers of the major NAD biosynthetic enzymes such as nicotinate phosphoribosyltransferase, nicotinamide phosphoribosyltransferase, and quinolinate phosphoribosyltransferase¹⁴. High levels of these enzymes would buffer against low NAD⁺/NADH ratios and the ability to switch to serine auxotrophy. A key question for future investigation is how the sequential selection of altered metabolism relates to the tumour’s epigenetic needs and the microenvironment in which it grows.

The compensatory mechanisms available to tumours to balance metabolic and epigenetic needs certainly require a precision medicine approach to identify key vulnerabilities. We may have to accept that there is no one targetable unifying principle, but rather a plethora of context-specific functions for metabolites and metabolic pathways. Further complicating our understanding is the growing appreciation of tumour heterogeneity and the finding that, like gene expression, the metabolic landscape at single-cell resolution is quite different than bulk analysis¹⁵. This difference is highlighted in breast cancer models, which indicate that metastatic cells have distinct metabolic programs compared to bulk tumour¹⁵. Like many platitudes, the idea of generally starving a cancer has some basis in truth but may not be generally applicable.