PGAM1 rheostat in glycolysis and autophagy

Abstract

How tumour cells rewire metabolic processes in response to changing nutrient availability in the tumour microenvironment is complex. Here, Zhang et al. define a role for PGAM1 in promoting autophagy in response to nutrient deprivation that competes with its catalytic function in glycolysis, which may explain selection for PGAM1 overexpression in cancer.

Phosphoglycerate mutase 1 (PGAM1), and its yeast homologue Gpm1, are known for their canonical function in catalysing the conversion of 3-phosphoglycerate (3PG) to 2-phosphoglycerate (2PG) in the second to last step of aerobic glycolysis¹. In addition to supplying 2PG for subsequent steps in glycolysis that feeds the tricarboxylic acid (TCA) cycle, PGAM1 indirectly promotes the pentose phosphate pathway (PPP) owing to the inhibitory effect of 3PG on the PPP while enhancing serine biosynthesis owing to 2PG stimulation of phosphoglycerate dehydrogenase (PHGDH)², with its catalytic function also required to sustain nucleotide levels required for DNA damage repair³. These functions of PGAM1 in cellular metabolism have been assumed to be the key role of PGAM1 in promoting cancer². Together with its overexpression in many human cancers, including, breast, lung, prostate and liver, this has led to the development of PGAM1 inhibitors for therapeutic purposes^4,5. However, in this issue of Nature Cell Biology, a study reveals an unexpected and critical role for PGAM1 in promoting autophagy via interactions with ATG9 and ATG14 that stimulates autophagosome formation and LC3 processing in response to starvation⁶. This role for PGAM1 is independent of its catalytic activity and is critical for its ability to promote tumour cell growth. This function for PGAM1 is stimulated by starvation due to phosphorylation by ULK1 (Atg1 in yeast), promoting its interaction with ATG14 and causing PGAM1 to localize to ATG14-positive puncta. At the same time, glycolysis is suppressed under starvation conditions, indicating that PGAM1 may have a role in coordinating glycolysis with autophagy in response to nutrient availability in the tumour microenvironment (Fig. 1).

Fig. 1 |. Model summarizing the role of PGAM1 in coordinating glycolysis with autophagy in response to changes in nutrient availability.

PGAM1 is a key enzyme in glycolysis, catalysing conversion of 3PG to 2PG, where it is reported to function as a dimer. Work reported by Zhang et al.⁶ shows that PGAM1 is phosphorylated by ULK1 (Atg1) in response to nutrient deprivation (starvation), which induces its interaction with ATG14 and ATG9 at autophagosomes, where it has a crucial role in recruitment of PI3K complexes, is required for LC3 processing and promotes p62 degradation (all key autophagy functions). This role of PGAM1 in autophagy is accompanied by decreased activity in glycolysis despite the catalytic activity of PGAM1 not being required for its function in autophagy — its role in glycolysis and autophagy are shown to be separable. Thus, it is postulated that PGAM1 levels may be limiting under starvation conditions, and that this favours its role in autophagy. However, both PGAM1 functions (glycolysis and autophagy) are important in cancer and this may explain the selective pressure to upregulate PGAM1 (either via loss of p53 or via increased mTORC1–HIF1α signalling). The overexpression of PGAM1 in cancers may thus allow it to promote both glycolysis and autophagy simultaneously.

The authors started out seeking to explain how phosphatidylinositol 3-kinase (PI3K) complex I was recruited to the yeast pre-autophagosomal structure (PAS) by performing pulldown experiments for Atg14 in yeast grown under nitrogen starvation conditions that are known to induce autophagy. Although they pulled down some expected autophagy regulators, such as Vps34, they were surprised to pull down Gpm1 with Atg14, and went on to show that acute depletion of Gpm1 in yeast blocked starvation-induced autophagy⁶. Deletion of other glycolytic enzymes in yeast did not elicit the same effect, and mutation of Gpm1 to abolish its catalytic activity in glycolysis had no impact on its ability to promote starvation-induced autophagy, indicating that the functions of Gpm1 in glycolysis and in autophagy were separable. They went on to show that Gpm1 was required for the interaction of Atg1, Atg9 and Vac8 with Atg14 to cooperatively promote PAS recruitment of PI3K complex I, and that Gpm1 formed a bridge between Atg14 and Atg9, with amino acid residue K27 in Gpm1 required for its interaction with Atg9 and amino acid S56 required for its interaction with Atg14. Mutation of K27 to alanine in Gpm1 did not affect binding to Atg14 but did prevent Atg14 puncta formation and PAS recruitment of PI3K complex I in response to starvation. Similarly, mutation of S56 to alanine in Gpm1 inhibited its binding to Atg14 without affecting binding to Atg9, and again this inhibited Atg14 puncta formation and PAS recruitment of PI3K complex I in response to starvation. Critically, the authors went on to show that S56 is phosphorylated by Atg1 in response to starvation and that this promotes the interaction of Gpm1 with Atg14 and enhances PAS recruitment of PI3K complex I. Resonating with earlier work in the autophagy field⁷, these findings in yeast define a role for Gpm1 in promoting autophagy at the expense of glycolysis in response to nitrogen starvation.

The importance of the work is greater still given that the authors go on to show that this mechanism is conserved from yeast to humans⁶. Overexpression of wild-type PGAM1 induced autophagy in human cells (U2OS and HCT116) in a dose-dependent manner, whereas knockdown of PGAM1 expression caused a deficit in LC3 processing and a failure to degrade p62 in response to amino acid deprivation. As they observed in yeast, starvation promoted the interaction of PGAM1 with both ATG9 and ATG14, with PGAM1 localizing to ATG14 puncta while its glycolytic activity was suppressed. This indicated a role for PGAM1 in coordinating induction of autophagy with rates of glycolysis under starvation conditions in mammalian cells. The critical serine in yeast Gpm1 (S56) is conserved in human PGAM1 at S58 and, consistently, mutation of S58 to alanine blocked phosphorylation of PGAM1 by ULK1 and inhibited its interaction with ATG14 without affecting binding to ATG9. Conversely, the phosphomimetic mutant (S58D) promoted interaction with ATG14 and increased starvation-induced LC3 processing and p62 degradation. Importantly, PGAM1 catalytic mutants (H186R and Y92F) unable to promote glycolysis retained the ability to promote autophagy, whereas mutation of S58 or depletion of ULK1 did not affect PGAM1 functions in glycolysis, confirming that autophagy and glycolysis are separable activities of PGAM1. The S58A and H186R mutants both decreased tumour cell growth in colony assays in vitro and in mouse xenografts in vivo, indicating both functions of PGAM1 (in autophagy and glycolysis) are important for its ability to promote tumorigenesis. Meanwhile, the S58D mutant increased tumour growth, consistent with it retaining catalytic activity in glycolysis alongside greater ability to promote autophagy, again confirming that both roles of PGAM1 (glycolysis and autophagy) promote tumorigenesis.

The authors then overexpressed three different PGAM1 mutants (A63T, T66A and D139T) identified in human clinical cancer specimens and showed that whereas these mutants retained glycolytic activity, they had enhanced ability to bind ATG14, process LC3, degrade p62 and induce autophagy. Overexpression of these naturally occurring PGAM1 mutants promoted tumour growth, indicating again that the ability to induce autophagy is a major aspect of their enhanced tumour-promoting activities. Finally, the authors created double mutants combining the autophagy-promoting mutant S58D with the catalytically inactive mutant H186R, or alternatively the naturally occurring tumour D139T mutant with H186R, and showed that increasing the autophagy-promoting activity of PGAM1 compensated for the loss of function in glycolysis with respect to promoting tumour growth in vitro and in vivo, again indicating that the ability of PGAM1 to induce autophagy is its more critical function in promoting tumorigenesis. Based on these compelling and conceptually innovative findings, the authors propose that PGAM1 governs a metabolism–autophagy checkpoint, with its overexpression in tumours promoting glycolysis but also lowering the threshold for induction of autophagy⁶.

The data presented here indicate that PGAM1 catalytic activity and rates of glycolysis go down under starvation conditions, whereas its role in autophagy is conversely amplified via ULK1-induced phosphorylation promoting its interaction with ATG14. This raises the question of how these two different functions of PGAM1 are regulated. Mutation of S58 to aspartic acid (S58D) did not affect the glycolytic activity of PGAM1, indicating that its decreased catalytic activity in glycolysis under starvation conditions is not due to ULK1 phosphorylation. The interaction with ATG14 may spatially affect how PGAM1 functions to limit glycolysis as it gets sequestered at autophagosomes. Alternatively, PGAM1 interactions with ATG14 and ATG9 may sterically impede its interactions with its substrate (3PG) or with modulatory kinases, such as PKM2^8,9. Perhaps a more likely explanation for how cells switch from glycolysis to autophagy under starvation conditions is that levels of PGAM1 are limiting under starvation conditions. If there is insufficient PGAM1 protein available under such conditions to complete both its catalytic role in glycolysis and its role in autophagy, the affinity of its interactions with ATG14 and ATG9 may be greater, causing its autophagy functions to outcompete its role in catalysing 3PG conversion to 2PG. One additional consideration here is that PGAM1 is reported to function as a homodimer in glycolysis¹⁰, which was not addressed in the current study⁶ but raises the possibility that if PGAM1 functions as a simple monomer in autophagy, this may contribute to why its function in glycolysis is diminished under starvation conditions if levels of PGAM1 are limiting. Lack of sufficient PGAM1 levels to permit its multiple activities may explain selection for its overexpression in tumours that would allow PGAM1 to promote both glycolysis and autophagy simultaneously (Fig. 1). Elevated PGAM1 in tumours has been previously attributed to loss of p53, resulting in de-repression of PGAM1 expression¹¹. PGAM1 is also required for mTORC1-driven tumour growth, mediated via HIF1α that regulates PGAM1 transcription directly via E-boxes in the PGAM1 promoter¹². Hypoxia is a major factor in malignant progression of cancers and promotes both glycolysis and autophagy¹³, so it will be interesting to assess how hypoxia affects PGAM1 levels in vivo and how PGAM1 levels determine responses to hypoxia in terms of relative rates of glycolysis and autophagy.

Non-catalytic roles for PGAM1 have been previously reported, with its interaction with α-smooth muscle actin (ACTA2) shown to promote actin filament formation and tumour cell migration¹⁴. Like its function in autophagy, the function of PGAM1 in ACTA2 binding and cell migration was independent of its catalytic activity in glycolysis. It is unclear at this time whether PGAM1 binding ACTA2 to modulate the actin cytoskeleton plays into the role of PGAM1 in promoting autophagy, and conversely whether the ability of autophagy to promote cell migration¹⁵ depends in any way on PGAM1. The region of PGAM1 (amino acids 201–220) involved in ACTA2 binding is distinct from regions of the protein involved in either ATG9 and ATG14 binding or its catalytic activity, suggesting that PGAM1 could be tri-functional in promoting glycolysis, autophagy and cell motility, which would make it an even more compelling therapeutic target than appreciated up to now. As mentioned, several allosteric inhibitors of PGAM1 have been developed, with some targeted at inhibiting its glycolytic activity^2,5 while another targeted its interaction with ACTA2⁴. Whether either of these inhibitors affect the ability of PGAM1 to promote autophagy remains to be determined, which will be critical in assessing how efficacious these drugs are likely to be in the clinic. Overall, this work underlines how important a metabolic player PGAM1 is in cellular responses to nutrient availability in the tumour microenvironment that in turn modulates the aetiology and progression of cancers.