Targeting aldolase A in hepatocellular carcinoma leads to imbalanced glycolysis and energy stress due to uncontrolled FBP accumulation

Snaebjornsson, Marteinn T.; Poeller, Philipp; Komkova, Daria; Röhrig, Florian; Schlicker, Lisa; Winkelkotte, Alina M.; Chaves-Filho, Adriano B.; Al-Shami, Kamal M.; Caballero, Carolina Dehesa; Koltsaki, Ioanna; Vogel, Felix C. E.; Frias-Soler, Roberto Carlos; Rudalska, Ramona; Schwarz, Jessica D.; Wolf, Elmar; Dauch, Daniel; Steuer, Ralf; Schulze, Almut

doi:10.1038/s42255-024-01201-w

Download PDF

Article
Open access
Published: 20 January 2025

Targeting aldolase A in hepatocellular carcinoma leads to imbalanced glycolysis and energy stress due to uncontrolled FBP accumulation

Nature Metabolism volume 7, pages 348–366 (2025)Cite this article

16k Accesses
12 Citations
65 Altmetric
Metrics details

Subjects

Abstract

Increased glycolytic flux is a hallmark of cancer; however, an increasing body of evidence indicates that glycolytic ATP production may be dispensable in cancer, as metabolic plasticity allows cancer cells to readily adapt to disruption of glycolysis by increasing ATP production via oxidative phosphorylation. Using functional genomic screening, we show here that liver cancer cells show a unique sensitivity toward aldolase A (ALDOA) depletion. Targeting glycolysis by disrupting the catalytic activity of ALDOA led to severe energy stress and cell cycle arrest in murine and human hepatocellular carcinoma cell lines. With a combination of metabolic flux analysis, metabolomics, stable-isotope tracing and mathematical modelling, we demonstrate that inhibiting ALDOA induced a state of imbalanced glycolysis in which the investment phase outpaced the payoff phase. Targeting ALDOA effectively converted glycolysis from an energy producing into an energy-consuming process. Moreover, we found that depletion of ALDOA extended survival and reduced cancer cell proliferation in an animal model of hepatocellular carcinoma. Thus, our findings indicate that induction of imbalanced glycolysis by targeting ALDOA presents a unique opportunity to overcome the inherent metabolic plasticity of cancer cells.

Bioinformatics analysis of the role of aldolase A in tumor prognosis and immunity

Article Open access 08 July 2022

MKL-1-induced PINK1-AS overexpression contributes to the malignant progression of hepatocellular carcinoma via ALDOA-mediated glycolysis

Article Open access 09 December 2022

The fructose-bisphosphate, Aldolase A (ALDOA), facilitates DNA-PKcs and ATM kinase activity to regulate DNA double-strand break repair

Article Open access 13 September 2023

Main

Increased glucose uptake and aerobic glycolysis are among the central features of metabolic reprogramming of cancer cells¹; however, in contrast to the historical view of the ‘Warburg effect’, it is now well recognized that most cancer cells also rely on oxidative metabolism in the mitochondria, both to produce energy as well as to generate various biosynthetic precursors derived from the tricarboxylic acid (TCA) cycle². The high dependency of cancer cells on glucose uptake is therefore likely caused by the need for biomass production via pathways that branch off from glycolysis, such as the pentose phosphate pathway (PPP), the hexosamine pathway and serine synthesis pathway, rather than energy generation per se³. Indeed, it has been shown that various cancer cells are able to proliferate in the complete absence of glucose when supplemented with uridine, which can be catabolized to ribose-5-phosphate (R5P) and enter the PPP^4,5,6. The conversion of fructose 6-phosphate (F6P) to fructose 1,6-bisphosphate (FBP) by the enzyme phosphofructokinase 1 (PFK1) is one of the rate-limiting steps in glycolysis and therefore tightly regulated⁷. PFK1 is generally highly active in cancer cells due to allosteric activation⁸, but modulation of its activity, for example by the TP53-induced glycolysis and apoptosis regulator (TIGAR)⁹ or the 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 4 (PFKFB4)¹⁰ can also promote cancer cell viability by supporting flux into the PPP. Furthermore, synthesis of FBP by PFK1 represents the last step of the investment phase and the beginning of the payoff phase of glycolysis in which the breakdown of FBP to pyruvate results in the net yield of two ATP molecules per molecule of glucose. Notably, the steady-state level of FBP has been identified as an intracellular indicator for glycolytic flux in a wide variety of systems, such as prokaryotes¹¹, yeasts^12,13, mouse embryos¹⁴ and cancer cells⁷.

Glycolysis is an autocatalytic pathway and carries an inherent risk of the investment phase outpacing the payoff phase, a state termed ‘imbalanced glycolysis’, which leads to a rapid depletion of intracellular ATP^15,16 and subsequent cell cycle arrest or cell death. In yeast this has been observed upon the loss of the enzyme trehalose 6-phosphate (T6P) synthetase (TPS1), which catalyses the first ATP-dependent step in the conversion of glucose-6-phosphate to trehalose, with the product of this reaction T6P functioning as an allosteric inhibitor of hexokinase (HK)¹⁷. Mammalian cells lack the trehalose pathway and it is the glycolytic intermediate glucose 6-phosphate (G6P) which regulates HK activity by inhibiting the enzyme. Indeed, overexpression of glucokinase in insulinoma cells¹⁸ and HK2 in iBMK cells was shown to induce rapid ATP depletion⁷ reminiscent of imbalanced glycolysis. Glucokinase is insensitive to G6P inhibition and HK2, which is upregulated in various cancers¹⁹, has been shown to alleviate this inhibition by interacting with voltage-dependent anion channels at the mitochondrial outer membrane^20,21,22,23. It is thus clear that the allosteric regulation of hexokinase via the trehalose pathway in yeast and via G6P in mammals has evolved as a natural safeguard mechanism against the formation of imbalanced glycolysis. Moreover, as high glucose uptake and elevated glycolytic flux are unifying features of cancers due to the constitutive activation of upstream signalling pathways, inducing imbalanced glycolysis could represent an attractive therapeutic strategy.

Here we report that liver cancer cells show a unique sensitivity toward depletion of fructose-bisphosphate aldolase A (ALDOA) despite other enzymes of the glycolytic pathway being largely dispensable. We find that depletion of ALDOA only has a minor impact on basal glycolytic flux but induces a pronounced build-up of FBP and severe energy exhaustion, indicative of a state of imbalanced glycolysis. Blocking FBP production by ablating the enzyme G6P isomerase (GPI) prevents the induction of imbalanced glycolysis and rescues proliferation in ALDOA-depleted cells. Using a mathematical model of glycolysis, we confirmed that ALDOA inhibition results in imbalanced glycolysis and depletion of inorganic phosphate (Pi). Based on the prediction that lowering glycolytic flux upstream of ALDOA increases the threshold for entering imbalanced glycolysis, we established that reducing glucose availability prevents FBP accumulation and energy exhaustion in ALDOA-depleted cells. Furthermore, downregulation of ALDOA induces pyrimidine deprivation and cell cycle arrest in cancer cells and reduced tumour growth in a mouse model of hepatocellular carcinoma (HCC). These results provide insight into a long-standing question regarding the essentiality of glycolytic enzymes and identify ALDOA as a key metabolic vulnerability of cancer cells engaging in aerobic glycolysis.

Results

Common ALDOA dependency across oncogenotypes and conditions

The metabolic activity of cancer cells is determined by several factors, including the metabolic characteristics of the parental tissue, the differential activity of oncogenic drivers and the metabolic milieu of the tumour microenvironment^24,25. To identify metabolic dependencies of cancer cells derived from the same parental tissue but induced by different defined genetic lesions, we employed cell lines derived from murine HCC driven by Akt1/c-Myc/p53^−/− (Akt1^myr;Myc^OE;Trp53^–/–) or by Nras/c-Myc/p53^−/− (Nras^G12V;Myc^OE;Trp53^–/–) generated by transposon-mediated gene transfer in p53-deficient mice^26,27. The cultures were first adapted to physiological nutrient levels by culturing them in mouse plasma-like medium (MPLM)²⁸ for ten population doublings. Next, cells were transduced with a focused short hairpin RNA (shRNA) library targeting approximately 450 genes related to metabolism (average of five hairpins per gene; Extended Data Fig. 1a and Supplementary Table 1). After selection, cells were exposed to different environmental conditions (FM-N, 10% fetal calf serum 21% O₂; LS-N, 1% fetal calf serum 21% O₂; and LS-H, 1% fetal calf serum 0.5% O₂) to simulate tumour-like metabolic stress²⁹, cultured for an additional ten population doublings and analysed by next-generation sequencing (Fig. 1a). Comparative evaluation of the ten genes with the highest essentiality score from each condition revealed that only Aldoa showed essentiality across both cell lines and under all conditions (Fig. 1b, Extended Data Fig. 1b and Supplementary Table 2). We confirmed that all Aldoa shRNA sequences used in the screen were efficient in reducing cell viability (Extended Data Fig. 1c). Furthermore, Aldoa-depleted cells showed reduced proliferation and a more spindle-like morphology (Extended Data Fig. 1d,e). In addition, CRISPR/Cas-mediated knockout of Aldoa caused a profound reduction in the proliferation of Myc^OE;Akt1^myr;Trp53^–/– cells (Extended Data Fig. 1f).

**Fig. 1: ALDOA represents a unique vulnerability in glycolysis independent of oncogenotype and environmental condition.**

The protein encoded by the Aldoa gene is a central enzyme in glycolysis. As many cancer cell lines adopt a glycolytic phenotype when cultured in vitro, we interrogated the essentiality scores of all other glycolytic enzymes represented in the library. Notably, none of the genes coding for glycolytic enzymes, apart from Aldoa, showed strong essentiality scores in both cell lines and all conditions used (Fig. 1c). Furthermore, interrogation of genetic dependencies across 1,800 cell lines reported in the Cancer Dependency Map (depmap.org) revealed common essentiality (CE) for several glycolytic enzymes, including ALDOA, GAPDH, PGK1, PGAM4 and PKM found by CRISPR-mediated deletion; however, only ALDOA was considered to be a CE gene also after gene silencing using RNAi, a targeting strategy unlikely to result in full elimination of the enzyme and only reducing its expression (Fig. 1d). In particular glucose-6-phosphate isomerase (Gpi/GPI), for which there is no paralogue in the mouse or human genome, was not identified as essential in our mouse screen nor scored as CE in the DepMap database (Fig. 1c,d). These results indicate that the essentiality of ALDOA in cancer cells cannot be explained by an overall requirement of aerobic glycolysis. Instead, ALDOA seems to represent a unique metabolic dependency of cancer cells across multiple entities and conditions.

ALDOA depletion exposes a metabolic bottleneck in glycolysis

The metabolic function of ALDOA is the reversible conversion of FBP into glyceraldehyde 3-phosphate (GAP) and dihydroxyacetone phosphate (DHAP) as part of the glycolytic pathway (Fig. 2a). Many glycolytic enzymes including ALDOA have noncatalytic functions, a phenotype known as ‘moonlighting’³⁰. Noncatalytic functions of ALDOA include the regulation of actin dynamics³¹, assembly of lysosomal vATPase³² and glucose sensing by AMP activated protein kinase (AMPK)³³. To investigate whether the essentiality of ALDOA in liver cancer cells involves catalytic or noncatalytic functions, we made use of a shRNA sequence targeting the 3′ nontranslated part of the Aldoa mRNA (shAldoa.1280) to combine inducible gene silencing with re-expression of either wild-type (WT) or mutant forms of the enzyme from the same lentiviral vector. Re-expression of WT ALDOA fully restored the viability of Akt1^myr;c-Myc^OE;Trp53^−/− cells (Fig. 2b,c). Of note, mutation of an arginine residue in the actin-binding domain of ALDOA (R43A), which had previously been shown to be essential for actin binding³⁴, had no effect on the ability of ALDOA to support cell viability (Extended Data Fig. 2a,b). In contrast, two different mutants in which the catalytic activity had been deleted (D34S and K146Q) were unable to restore cell viability (Fig. 2b,c and Extended Data Fig. 2a,b). Similar results were also obtained for Nras^G12V;Myc^OE;Trp53^–/– cells and cells derived from Akt1^myr ;Myc^OE driven, Cdkn2a^ARF-deficient tumours (Akt1^myr;Myc^OE;Cdkn2a^ARF–/–) (Extended Data Fig. 2c,d). We therefore concluded that ALDOA essentiality is determined by its catalytic activity rather than its moonlighting functions.

**Fig. 2: Depletion of ALDOA has minimal effect on basal glycolytic activity.**

We next explored the effect of ALDOA depletion on metabolic activity. Of note, Seahorse analysis revealed that ALDOA-depleted cells, or cells expressing the catalytically dead mutant D34S, only showed a minor reduction in basal glycolysis, whereas maximal glycolysis and the activation of the glycolytic reserve after blocking mitochondrial ATP production with oligomycin were severely impaired (Fig. 2d). Silencing of Aldoa resulted in a small increase in ATP-linked respiration, while basal and maximal respiration were largely unaffected (Fig. 2e). Analysis of the cell energy phenotype before and after oligomycin treatment (Fig. 2f) indicates that, while basal metabolism is not affected, cells depleted of ALDOA are unable to respond to increased energetic demand by increasing their glycolytic flux.

Consistent with the minor effect on basal glycolysis, we also observed that ALDOA depletion had no major effect on glucose or glutamine uptake or oxidative stress (Fig. 2g,h). In contrast, ALDOA-depleted or D43S-expressing cells showed a reduced NAD⁺:NADH ratio and an increased NADP⁺:NADPH ratio (Fig. 2i), potentially as a consequence of metabolic reprogramming in response to ALDOA inactivation.

Disruption of glycolysis at GPI does not block proliferation

While high glucose utilization is a common metabolic phenotype in cancer, increasing evidence supports the importance of oxidative metabolism for tumour growth³⁵. Given that no other glycolytic enzymes showed consistent essentiality in our screens and the DepMap data (Fig. 1c,d), we reasoned that it should be possible to disrupt the glycolytic pathway in cancer cells and still maintain viability. We decided to target the step that converts G6P to F6P, as this preserves the flux into the PPP and hexosamine pathways (Fig. 3a). Furthermore, GPI, the enzyme catalysing this step, has no paralogue in mammalian cells. Indeed, it has previously been shown that depleting GPI suppresses aerobic glycolysis in cancer cells without affecting cell proliferation in vitro and only a minimal effect on tumour growth³⁶.

**Fig. 3: Deletion of *Gpi* does not impair cell proliferation.**

Genetic deletion of Gpi was achieved using CRISPR/Cas9 technology either with two combined guide RNA (gRNA) sequences (sgGpi#1/#4) in a pooled cell population or by selecting a knockout clone generated by a single gRNA (sgGpi#4 cl.1) (Fig. 3b). In concordance with previous findings³⁶, deletion of Gpi had no effect on cell proliferation of murine liver cancer cells cultured in physiological medium conditions in normoxia (Fig. 3c); however, these cells were unable to adapt to hypoxic conditions (Extended Data Fig. 3a). To exclude effects caused by clonal adaptation, we also employed gene silencing using doxycycline-inducible shRNA expression. This confirmed that acute silencing of Gpi did not impair cell viability (Extended Data Fig. 3b,c). Similarly, GPI knockout did not impair proliferation in three human liver cancer cell lines (HLE, HLF and SNU-387) (Extended Data Fig. 3d,e). Seahorse analysis revelated that stable Gpi knockout or acute silencing caused a severe reduction in basal glycolysis and completely blocked the mobilization of the glycolytic reserve in both murine and human liver cancer cells (Fig. 3d and Extended Data Fig. 3f,g).

The impact of Gpi deletion on glycolytic flux was also investigated by [1,2]¹³C-glucose tracing into 3-phosphoglycerate (3PG) (Extended Data Fig. 3h). Stable isotope distribution at steady state (240 min) showed a higher proportion of M + 1 labelled 3PG, confirming rewiring of glycolytic flux via the oxidative and nonoxidative arms of the PPP (Extended Data Fig. 3i). However, while the glycolytic flux was evidently rewired, the overall rate of label incorporation into glycolytic metabolites was strongly reduced in Gpi knockout cells (28.6% versus 3.8% ¹³C enrichment per min), confirming inhibition of the pathway (Extended Data Fig. 3j). In parallel to the reduction in glycolytic flux, Gpi knockout cells showed a marked upregulation of basal, ATP-linked and maximal respiration (Fig. 3e), indicating that cells undergo a shift from a glycolytic to an oxidative phenotype. This was in contrast to ALDOA depletion, where no increase in OXPHOS was observed (Fig. 2e). In addition, Gpi knockout cells were no longer able to induce glycolysis in response to oligomycin treatment (Fig. 3f). We also observed a strong reduction in glucose uptake in Gpi-deleted cells. This was not accompanied by a compensatory upregulation of glutamine uptake (Fig. 3g), indicating that the cells generate ATP much more efficiently by complete oxidation of glucose. Alternatively, Gpi-deleted cells could compensate by upregulating fatty acid oxidation. Despite the increased respiration observed in Gpi knockout cells, there was no increase in oxidative stress (Fig. 3h). Furthermore, we observed an increase in the NAD⁺:NADH ratio which is in agreement with increased mitochondrial metabolism. In contrast, the NADP⁺:NADPH ratio is decreased in Gpi knockout cells, which could be a consequence of increased flux through the oxidative PPP (Fig. 3i). Together, these results suggest that disruption of glycolysis at the level of GPI has no major impact on cell viability and proliferation by allowing cancer cells to undergo a metabolic switch to ATP production via increased mitochondrial respiration.

Blocking ALDOA catalytic activity induces severe energy stress

Having observed the stark differences in cellular response to targeting glycolysis at the level of either GPI or ALDOA, we next determined changes in cellular metabolite levels using liquid chromatography (LC) linked to mass spectrometry (MS). For this, Akt1^myr;c-Myc^OE;Trp53^−/− cells were first depleted of ALDOA by inducing RNAi expression for 48 h and then treated with fresh medium for 6 h before analysis. LC–MS analysis revealed that the most strongly upregulated metabolite was FBP, the substrate of ALDOA, with levels being >30-fold higher compared with control cells (Fig. 4a). FBP was also the strongest upregulated metabolite in Nras^G12V;c-Myc^OE;Trp53^−/− and Akt1^myr;c-Myc^OE;Cdkn2a^ARF^–/–cells following ALDOA depletion (Extended Data Fig. 4a). Re-expression of WT ALDOA, but not the catalytic mutant D34S, prevented the strong accumulation of FBP (Fig. 4b). Closer inspection of metabolic intermediates of glycolysis showed that while GAP/DHAP and glycerol 3-phosphate (G3P) were downregulated, metabolites of lower glycolysis were only mildly affected (Fig. 4b,c). In addition, ALDOA depletion resulted in marked reduction of ATP, a significant decrease in the ATP:ADP ratio and increased levels of AMP (Fig. 4a,d), indicating profound exhaustion of cellular energy. Moreover, phosphocreatine, a readily available source of high-energy phosphoryl-groups used for the rapid regeneration of ATP from ADP, was strongly reduced (Fig. 4a,d). The dramatic loss of phosphocreatine after depletion of ALDOA was restored by re-expression of the WT enzyme but not the catalytically inactive mutant (Fig. 4d).

**Fig. 4: Depletion of ALDOA exposes a metabolic bottleneck in glycolysis leading to severe energy exhaustion.**

In contrast, cells depleted of GPI and treated in the same manner showed a significant downregulation of several glycolytic intermediates, most notably FBP (Fig. 4e). However, despite the complete disruption of glycolytic flux, Gpi knockout cells showed no significant change in ATP/ADP ratio or phosphocreatine levels, with only a small increase in AMP levels (Fig. 4f). Comparable results were also obtained after acute silencing of Gpi using inducible shRNA expression, with the notable exception that levels of FBP were not lowered, potentially due to a compensatory upregulation of G6P. Moreover, phosphocreatine levels were slightly reduced, indicating incomplete metabolic adaptation after acute GPI depletion (Extended Data Fig. 4b).

As disruption of the catalytic activity of ALDOA caused a severe reduction in cellular energy charge, we next investigated whether this also resulted in the activation of AMPK, a master regulator of metabolic homeostasis in response to low energy³⁷. Indeed, FBP and ALDOA have been shown to be involved in glucose sensing by AMPK independent of AMP/ADP³³. We observed that depletion of ALDOA induced serine 79 phosphorylation of acetyl-CoA carboxylase (ACC), indicative of AMPK activation. Furthermore, re-expression of the catalysis-defective D34S mutant, which retains the activity to bind FBP and prevent AMPK activation in glucose-starved HEK293 cells and MEFs³³, did not avert AMPK activation in murine liver cancer cells exposed to glucose-replete medium (Fig. 4g). This finding supports the conclusion that blocking the catalytic activity of ALDOA in cancer cells under conditions where glucose is readily available leads to a severe state of energy deprivation that overrides the glucose-sensing mechanism of AMPK.

Depletion of ALDOA induces imbalanced glycolysis

Glycolysis is controlled at the level of PFK1 through allosteric activation and inhibition by AMP and ATP, respectively, matching glycolytic flux with cellular energy demand (Fig. 5a). As strong FBP accumulation and severe energy stress in ALDOA-depleted cells was observed after 6 h of exposure to fresh medium containing all nutrients, including glucose, we reasoned that it could be the consequence of high glucose availability. We therefore determined FBP levels in control and ALDOA-depleted cells after 48 h of silencing in the same medium and over a time course of medium re-addition. In line with previous findings³³, control cells respond to renewed glucose availability by a moderate and transient increase in FBP (Fig. 5b). Remarkably, ALDOA-depleted cells showed the same initial low levels of FBP as control cells but responded to medium addition with strong and sustained FBP accumulation (Fig. 5b). Furthermore, ATP levels were initially similar in control and ALDOA-depleted cells, but were strongly reduced in ALDOA-depleted cells at 6 h after addition of fresh medium (Fig. 5c).

**Fig. 5: Deletion of ALDOA leads to imbalanced glycolysis.**

The initially counterintuitive finding that provision of fresh medium containing glucose to ALDOA-depleted cells could, in a matter of hours, lead to a strong reduction in ATP levels coinciding with a prominent build-up of FBP is reminiscent of imbalanced glycolysis, a phenotype caused by the loss of HK regulation by glycolysis and previously observed in yeast¹⁶. To further explore the relationship between glucose availability, glycolytic flux and ALDOA catalytic activity we made use of a coarse-grained mathematical model of glycolysis based on ordinary differential equations (Supplementary Information). The model describes flux through the glycolytic pathway, as well as levels of ATP, FBP and inorganic phosphate (Pi), and is capable of reproducing the observed results for a broad range of parameters. Specifically, at 100% ALDOA activity, the model gives rise to a viable stable steady state with low FBP levels (Fig. 5d, left graph), consistent with what was observed in WT cells (Fig. 5b); however, when ALDOA activity is reduced to 20%, the model gives rise to an imbalanced glycolytic state that is characterized by accumulation of FBP to high levels and severe depletion of ATP and Pi. As reported previously¹⁶, and also observed here (Fig. 5d, middle graph), the marked drop in Pi is caused by trapping high-energy phosphate groups in metabolites of upper glycolysis, most notably FBP. Consistent with the model, experimental ALDOA depletion in Akt1^myr;c-Myc^OE;p53^−/− cells resulted in a marked reduction in Pi (Fig. 5e). In contrast, knockout of Gpi caused an increase in Pi levels (Fig. 5f), inversely matching the low FBP levels observed in these cells (Fig. 4f). Similarly, acute silencing of Gpi, which did not lower FBP, did not affect Pi levels (Extended Data Fig. 5a).

Based on these findings, we considered potential mechanisms to rebalance glycolysis in ALDOA-depleted cells. The coarse-grained model indicates that reducing glycolytic flux by lowering glucose availability restores the balance between upper and lower glycolysis and prevents FBP accumulation (Fig. 5g). For example, 20% of ALDOA activity is sufficient to support balanced glycolysis when glucose concentrations are lowered to 1 mM (Fig. 5d, right graph). We therefore investigated whether reduced glucose provision could prevent the induction of imbalanced glycolysis. We therefore exposed control and ALDOA-depleted cells to fresh medium containing either at physiological (4.4 mM) or reduced (0.44 mM) concentrations of glucose for 6 h. This revealed that lowering the glucose availability prevented the accumulation of FBP and fully restored intracellular Pi levels in ALDOA-depleted cells (Fig. 5h); however, cells were unable to recover energy levels under glucose starvation, as the ATP:ADP ratio was reduced also in control cells (Fig. 5h) and AMPK was activated under all conditions (Fig. 5i). Moreover, glucose deprivation did not restore the induction of the glycolytic reserve in ALDOA-depleted cells (Extended Data Fig. 5b). We next explored the effect of different glucose concentrations on proliferation in control and ALDOA-depleted cells. While solvent-only-treated cells or cells expressing a nontargeting shRNA (shRen) responded to glucose starvation by reducing growth, proliferation of ALDOA-depleted cells was slightly enhanced under low-glucose concentrations (Fig. 5j and Extended Data Fig. 5c). These results indicate that, while glucose starvation prevents the induction of imbalanced glycolysis, it is not sufficient to support the high metabolic demand of cancer cells during proliferation.

Combined block of GPI and ALDOA prevents imbalanced glycolysis

To confirm that inhibition of cell viability following ALDOA depletion is mechanistically linked to the induction of imbalanced glycolysis, we hypothesized that genetic deletion of Gpi should prevent the trapping of high-energy phosphate groups in FBP (Fig. 6a). We therefore combined CRISPR-mediated deletion of Aldoa and Gpi in murine cancer cells using several combinations of gRNAs (Fig. 6b). As predicted by our model, deletion of Gpi fully restored cell viability in Aldoa knockout cells (Fig. 6c). This was accompanied by a complete block in FBP accumulation and a strong reduction in GAP/DHAP levels (Fig. 6d). We also established levels of ATP and ADP in single and double-knockout cells and found that combined deletion of Gpi and Aldoa did not fully restore the ATP:ADP ratio to that of control cells, most likely as Gpi deletion already reduced this ratio despite having no effect on cell proliferation. We also observed that Aldoa deletion reduced the amounts of both ATP and ADP, which was restored by co-deletion of Gpi (Extended Data Fig. 6a). Nevertheless, deletion of Gpi strongly increased phosphocreatine levels in Aldoa-deleted cells, indicative of restored cellular energy levels (Fig. 6d,e). Furthermore, Seahorse analysis showed that cells depleted of GPI and ALDOA showed impaired glycolytic activity but a strong upregulation of mitochondrial activity (Fig. 6f). This confirms that blocking glycolytic flux upstream of PFK prevents the engagement of the regulatory feedback loop that leads to the establishment of imbalanced glycolysis and severe energy stress. It also allows upregulation of mitochondrial metabolism to compensate for glycolytic impairment. Notably, knockout of Gpi also prevented the induction of ACC phosphorylation in response to Aldoa deletion, further confirming the restoration of cellular energy levels (Fig. 6b). We also extended this analysis to human liver cancer cell lines (HLE, HLF and SNU-387). Crucially, deletion of GPI rescued the effect of ALDOA deletion by significantly increasing cell number and, similar to the murine system, attenuating AMPK activation (Extended Data Figs. 6b,c).

**Fig. 6: Combined depletion of GPI and ALDOA rescues imbalanced glycolysis and restores proliferation.**

Together, these data demonstrate that blocking glycolysis at the step catalysed by ALDOA results in the trapping of high-energy phosphate in upper glycolysis. Under these conditions, cells enter a state of imbalanced glycolysis that results in profound energy stress. Imbalanced glycolysis is aggravated by allosteric activation of PFK, inducing a vicious cycle that further increases the synthesis of FBP, thereby depleting intracellular phosphate stores and blocking proliferation. Preventing the accumulation of large amounts of FBP by blocking GPI averts imbalanced glycolysis and restores viability.

Rewiring of metabolism in ALDOA-depleted cells

To obtain deeper insight into the metabolic consequences of imbalanced glycolysis we applied stable-isotope tracing. We initially conducted time-resolved experiments by treating control and ALDOA-depleted cells with 4.4 mM of fresh glucose for 30 or 180 min to induce imbalanced glycolysis. After this time, the medium was replaced with medium containing ¹³C₆-glucose and cells were incubated for a further 5, 10 or 15 min. This allowed us to determine flux for the reaction from glucose to G6P as well as changes in the abundance of the M + 6 isotopologues of G6P, F6P and FBP. In addition, we monitored the abundance of the M + 3 isotopologue of G3P, produced from glycolytic intermediates downstream of ALDOA, and the M + 5 isotopologue of R5P, a product of the oxidative PPP. It should be noted that it was not possible to determine flux for all reactions of upper glycolysis, as these are not in metabolic steady state under these conditions.

The results of this analysis showed that the flux from glucose into G6P was higher in ALDOA-depleted cells compared with controls, albeit with reduced abundance of the M + 6 isotopologue (Extended Data Fig. 7a and Fig. 7a). Furthermore, we observed a strong accumulation of the M + 6 isotopologue of FBP in ALDOA-depleted cells, particularly at the 30-min time point, confirming that ALDOA-depleted cells continue to synthesize this metabolite (Fig. 7a). In contrast, the production of metabolites downstream of ALDOA was severely impaired, as we observed a strong reduction in the M + 3 isotopologue of G3P. Interestingly, this analysis also revealed a substantial reduction in oxidative PPP, as the M + 5 isotopologue of R5P was strongly reduced. Together, these results confirm our model in which the continued production of FBP in ALDOA-depleted cells leads to the trapping of high-energy phosphate and severe energy stress.

**Fig. 7: Depletion of ALDOA retains flux in upper glycolysis but impairs oxidative PPP and pyrimidine synthesis.**

We also conducted experiments using labelling at isotopic steady state by culturing control and ALDOA-depleted cells with ¹³C₆-glucose or ¹³C₅-glutamine. After treating cells with doxycycline for 48 h to achieve ALDOA silencing, cells were induced to enter imbalanced glycolysis by addition of fresh medium containing 4.4 mM glucose for 1 h before addition of the labelled nutrient for 10 h. Isotopologue analysis showed that incorporation of carbons derived from glucose or glutamine into the TCA cycle was reduced in ALDOA-depleted cells and the total pool size of all measured TCA cycle intermediates was strongly diminished (Extended Data Fig. 7b,c). In contrast, levels and isotopologue distribution of glutamate remained mostly unchanged (Extended Data Fig. 7d). This indicates that the flux of metabolites into the TCA cycle is reduced but intermediates are still being drained into biosynthetic processes.

Nucleotide synthesis is among the cellular metabolic processes that require large amounts of energy and are regulated by the availability of phosphate³⁸. In particular, the first steps of pyrimidine synthesis respond to changes in ATP and Pi due to allosteric regulation of phosphoribosyl pyrophosphate synthetase and the multifunctional protein CAD. Given the strong antiproliferative effect of ALDOA depletion, we next investigated the synthesis of pyrimidine nucleotides, which require TCA cycle-derived carbons in the form of aspartate and nitrogen derived from the glutamine amide moiety. Of note, total metabolite levels of aspartate and glutamine were increased in ALDOA-depleted cells (Fig. 7b), indicative of a reduced consumption of these metabolites by the first and second step of pyrimidine biosynthesis, catalysed by the CAD enzyme. Subsequent metabolites of the pyrimidine biosynthesis pathway, including dihydroorotate and orotate, were drastically depleted. Moreover, re-expression of ALDOA WT, but not the catalysis-defective D34S mutant, restored levels of these metabolites (Extended Data Fig. 7e). CAD activity is regulated by the availability of phosphoribosyl pyrophosphate (PRPP), the synthesis of which requires large amounts of both ATP and R5P and is controlled by intracellular Pi³⁹. Consistently, we observed a strong accumulation of hypoxanthine, an intermediate of the purine salvage pathway (Fig. 7c), indicating that the recycling of purine bases by hypoxanthine phosphoribosyl transferase, which requires PRPP as substrate, is blocked in ALDOA-depleted cells. Together, these results support the conclusion that imbalanced glycolysis leads to an inhibition of nucleotide biosynthesis due to a lack of PRPP and ATP.

ALDOA depletion induces S-phase arrest and blocks tumour growth

Having observed a strong effect of imbalanced glycolysis on nucleotide biosynthesis, we next determined the effect of ALDOA depletion on cell cycle distribution (Fig. 8a,b). This revealed a marked reduction in the proportion of cells undergoing active DNA synthesis (from 70% to 20%) accompanied by an increase in the proportion of cells in G0/G1 and G2. Of note, cells also showed evidence of S-phase arrest, a phenotype associated with nucleotide depletion⁴⁰; however, addition of nucleosides did not prevent S-phase arrest and did not restore proliferation in ALDOA-depleted cells (Extended Data Fig. 8a–c).

**Fig. 8: Targeting of ALDOA blocks cell cycle progression and increases survival in an autochthonous mouse model of liver cancer.**

We next explored the consequences of ALDOA depletion in a well-established genetically engineered liver cancer mouse model²⁷. We used hydrodynamic tail vein injection of transposons encoding c-Myc and Akt1^myr together with shRNAs targeting Aldoa or nontargeting control (pT-CaMIA-shRNA). Both Aldoa shRNAs (shAldoa.1280 and shAldoa.558) strongly reduced tumour development and prolonged survival of tumour-bearing animals in comparison to a nontargeting control shRNA (shRen) (Fig. 8c), indicating that ALDOA is essential for liver tumour development. This is particularly remarkable, as this model was previously shown to display resistance toward the multikinase inhibitor sorafenib²⁷, a first-line therapy for HCC. Moreover, tumours arising after ALDOA depletion showed a marked reduction in proliferation, indicated by lower Ki67 positivity (Fig. 8d,e). Finally, analysis of public data indicates that ALDOA is highly expressed in The Cancer Genome Atlas-liver hepatocellular carcinoma (TCGA-LIHC) samples (Fig. 8f,g) and a strong predictor of poor survival (Fig. 8h). This is in agreement with earlier reports indicating an isoform switch from ALDOB to ALDOA in liver cancer^41,42. Together, these results suggest that inducing imbalanced glycolysis by targeting ALDOA could be a therapeutic strategy for cancers that engage in aerobic glycolysis.

Discussion

In this work we found that the glycolytic enzyme aldolase A is essential for proliferation in glycolytic cancer cells even though the glycolytic pathway itself is dispensable. Indeed, ALDOA was the only glycolytic enzyme that showed CE both in our RNAi screen in murine HCC cell lines with different oncogenotypes cultured under diverse environmental conditions and across the multiple cancer cell lines represented in the Cancer Dependency Map. Disrupting the catalytic activity of ALDOA was accompanied by strong accumulation of FBP and severe energy stress despite only causing a minor reduction in basal glycolysis. In contrast, blocking glycolysis at the level of Gpi was fully tolerated by murine and human cancer cells, rendering them also insensitive to ALDOA depletion. Moreover, we found that depletion of ALDOA caused reduced cancer cell proliferation and extended survival in a mouse model of HCC.

The strong dependency of cancer cells on the catalytic activity of ALDOA is caused by the autocatalytic nature of glycolysis and the crucial position of ALDOA in the sequence of reactions of the glycolytic pathway, controlling the transit of metabolites from the investment phase of upper glycolysis into the energy-producing reactions (payoff phase) of lower glycolysis. Inhibition of ALDOA catalytic activity in highly glycolytic cancer cells thus leads to the induction of an imbalanced glycolytic state that effectively transforms glycolysis from an ATP-producing into an ATP-consuming pathway. Mechanistically, when glucose enters glycolysis in a cell depleted of ALDOA, the glycolytic intermediate FBP cannot be further metabolized by lower glycolysis and rapidly accumulates. There are three major reasons why imbalanced glycolysis is an extremely problematic situation for the cell. (1) The conversion of F6P to FBP catalysed by PFK is essentially irreversible under physiological conditions. In addition, the only enzymes that utilize FBP as substrate are aldolase and FBPase, the latter being exclusively expressed in gluconeogenic tissues (such as the healthy liver) and frequently downregulated in cancer^2,43,44. (2) FBP, the end product of the investment phase of glycolysis, is an energy-rich molecule, as it contains both of the two phosphates invested into glycolysis by hexokinase and PFK1, respectively. Thus, strong accumulation of FBP, as seen in ALDOA-depleted cells, leads to profound trapping of high-energy phosphate groups and subsequent energy depletion. (3) Energy depletion caused by imbalanced glycolysis activates PFK, thus causing a vicious cycle in which reduced ATP levels lead to increased ATP consumption by upper glycolysis and a further build-up of FBP. As a consequence, cells can only escape this state of imbalanced glycolysis by limiting the influx into upper glycolysis, thus reducing the production of FBP, for example under conditions of reduced glucose availability or in response to Gpi deletion.

Previous studies describing imbalanced glycolysis as a result of loss of HK inhibition in yeast have reported that the resulting uncontrolled activation of upper glycolysis leads to a strong reduction in intracellular levels of Pi and ATP due to FBP synthesis acting as a trap for high-energy phosphate groups. Furthermore, the reduced availability of inorganic phosphate creates a blockage at the level of glyceraldehyde-3-phosphate dehydrogenase (GAPDH), which further compounds the problem^16,45. A similar relationship between FBP accumulation and reduced ATP and Pi levels was also observed in our coarse-grained kinetic model of glycolysis. Experimentally, we confirmed the glucose-dependent increase in FBP as well as a reduction in ATP and Pi as predicted by the model. Furthermore, ALDOA-depleted cells continued to consume glucose and convert it into FBP, as demonstrated by glucose-uptake experiments and stable-isotope tracing; however, unlike in both our model and the model describing imbalanced glycolysis caused by removal of HK inhibition¹⁶, we did not observe a complete depletion of ATP and Pi levels in ALDOA-depleted cells. The likely cause for this discrepancy is that the models do not account for compartmentalization of these metabolites between the cytosol and mitochondria (and possibly other compartments) or the contribution of respiration toward ATP production. Indeed, we found that although ALDOA-depleted cells had strongly reduced ATP levels, the overall activity of respiration was largely unaffected, even displaying a small but significant increase in ATP-linked respiration. This is in stark contrast to the effect of Gpi deletion, which caused a pronounced increase in both basal and ATP-linked respiration, in line with previous observations in melanoma and colon adenocarcinoma cells³⁶. A similar shift from glycolysis to oxidative phosphorylation has also been observed in cells in which glycolysis was inhibited by culture in acidic medium^46,47, which inhibits glycolysis at the level of PFK1 (refs. ^47,48). This phenotype is reminiscent of an inversion of the Crabtree effect, a principle describing the suppression of oxidative phosphorylation by high glycolytic activity⁴⁹. Inhibition of upper glycolysis at the level of Gpi could relieve the inhibition of oxidative phosphorylation by the Crabtree effect, thus allowing cells to compensate for the loss of glycolytic ATP production. Notably, targeting glycolysis at the level of ALDOA does not allow this compensatory activation of oxidative phosphorylation. This could potentially be mediated by high FBP levels, as studies in yeast have indicated that FBP inhibits components of the electron transport chain and has already been discussed as a mediator of the Crabtree effect^50,51; however, other metabolites, including inorganic phosphate⁵² or intermediates of lower glycolysis (GAP, 3-phosphoglycerate and phosphoenolpyruvate)⁵³, have also been proposed to induce electron transport chain inhibition.

In addition to its catalytic function in glycolysis, ALDOA has important noncanonical functions. Aldolases have been shown to bind to the actin cytoskeleton^34,54,55 and this interaction was found to be inhibited by the presence of FBP⁵⁶. However, we observed that expression of the R43A mutant of ALDOA, which renders the enzyme catalytically active but unable to bind to actin⁵⁴, was fully capable of supporting cell proliferation in cells depleted of endogenous ALDOA. This indicates that the ability of ALDOA to bind actin is nonessential for proliferation in our system. A second noncanonical function of ALDOA involves the regulation of AMPK and mTORC1 (refs. ^33,57,58). It was shown that aldolase bound to its substrate FBP inhibits AMPK activity. Of note, the D34S mutation in ALDOA mimics FBP-bound ALDOA, as it is catalytically inactive but can nevertheless mediate signalling to prevent AMPK activation. We found that expression of ALDOA D34S did not prevent AMPK activation upon in cells depleted of endogenous ALDOA, indicating that AMPK can still be activated by energy stress even though it receives the signal that glycolysis is active. In contrast, we found that Gpi deletion led to basal AMPK activation even though the energetic state of the cells was only mildly affected, suggesting that AMPK might also sense low glycolytic activity via the absence of FBP, as previously reported³³, as FBP levels were very low in GPI-deleted cells. These two modes of AMPK activation (via sensing of low glycolysis or via detection of energy stress) could lead to different metabolic outputs and GPI-deleted cells could provide an experimental system in which this question could be addressed.

Using stable-isotope tracing, we confirmed that flux in upper glycolysis is maintained in ALDOA-depleted cells and that the cells continue to synthesize FBP; however, we also found that ALDOA depletion leads to a reduction in the oxidative PPP, which could be caused by allosteric activation of PFK. In addition, we observed a strong reduction in the synthesis of pyrimidine nucleotides, most likely caused by the combination of reduced levels of ATP, loss of inorganic phosphate and depletion of R5P due to imbalanced glycolysis. This resulted in acute cell cycle inhibition, including S-phase arrest. Pyrimidine biosynthesis is governed by the CAD enzyme, which is allosterically activated by PRPP³⁹. PRPP itself is synthesized from ATP and R5P by PRPP synthetase whose activity is highly dependent on Pi, thus providing a link between intracellular phosphate availability and proliferation⁵⁹. Moreover, salvage of purine nucleotides was also blocked at the level of PRPP incorporation. We therefore conclude that the antiproliferative effect of imbalanced glycolysis in cancer cells could be due to inhibition of nucleotide biosynthesis.

Our finding that ALDOA is a common essential enzyme in HCC cell lines across different oncogenotypes and during exposure to diverse environmental conditions highlights the high susceptibility of cancer cells to imbalanced glycolysis. This could also be a mechanistic explanation for the unique essentiality of ALDOA observed across hundreds of human cancer cell lines in the Cancer Dependency Map, a phenotype not shared by other glycolytic enzymes. Aerobic glycolysis is a metabolic hallmark in many cancer entities, including HCC, and could be exploited for cancer therapy⁶⁰. Preclinical studies have revealed promising effects of targeting glycolysis by deletion of HK2 in lung and breast cancer¹⁹. However, it is clear that the ability of cancer cells to shift from glycolysis to oxidative phosphorylation for ATP production presents a major challenge for therapeutic strategies blocking glycolysis and may require combinatorial treatments⁶¹. Targeting ALDOA presents an opportunity to overcome the inherent metabolic plasticity of cancer cells, as it not only removes the ability of cells to utilize glycolysis as an energy-generating pathway but converts glycolysis into a trap for high-energy phosphate. Inducing imbalanced glycolysis by targeting ALDOA could thus be highly efficient in eliminating cancer cells.

Methods

Cell culture

Murine liver cancer cells were cultured in murine plasma-like medium (MPLM), an ‘in-house’-made medium based on the composition of mouse plasma as described in ref. ²⁸ with 10% dialysed fetal bovine serum, 100 U ml⁻¹ penicillin and 100 μg ml⁻¹ streptomycin. The formulation of MPLM is provided in Supplementary Table 4. To assess the impact of nucleoside addition on proliferation and cell cycle, the medium was supplemented with 1× Embryomax (ES-008-D Merck). Human liver cancer cell lines were kindly provided by J.P.F. Angeli and cultured in DMEM with 10% fetal bovine serum, 4 mM l-glutamine, 100 U ml⁻¹ penicillin and 100 μg ml⁻¹ streptomycin (all Sigma). All cell lines were cultured at 37 °C in a humidified incubator at 5% CO₂.

Library preparation and genetic screening

The shRNA library consists of 2,258 shRNAs targeting 457 metabolic genes. The mirE-based shRNAs were designed by J. Zuber (Research Institute of Molecular Pathology, Vienna Biocenter) ordered as 97-mer DNA oligonucleotides from IDT and cloned into pLT3GEPIR using XhoI or EcoRI restriction sites⁶². Sequence representation was confirmed by next-generation sequencing. shRNA sequences are provided in Supplementary Table 1. Akt1^myr;c-Myc^OE;Trp53^−/− and c-Myc^OE/Nras^G12V;c-Myc^OE;Trp53^−/− cells were transduced by lentiviral infection (multiplicity of infection of 0.3 and clonal redundancy of 1,000) and selected with puromycin for 3 days, at which stage a sample of genomic DNA was taken (T₀). Cells were then cultured in different conditions for ten population doublings while maintaining shRNA representation and genomic DNA was prepared (T₁). Experiments were performed in biologically independent triplicate. Normalized read counts generated from next-generation sequencing were analysed to identify essential genes using MAGeCK⁶³. Essentiality scores are provided in Supplementary Table 2.

Genetic modification using shRNA and CRISPR

For inducible gene silencing, shRNA sequences targeting murine Aldoa were cloned into pLT3GEPIR (Addgene, #111177). Viral particles were produced in HEK293T cells and stable populations were selected using puromycin. Sequences are provided in Supplementary Table 1. For CRISPR/Cas9-mediated deletion of murine and human GPI and human ALDOA, sgRNA sequences were designed using CHOPCHOP (https://chopchop.cbu.uib.no) or VBC score (https://www.vbc-score.org) and cloned into lentiCRISPRv2 (Addgene, #52961) or LentiGuide-TagRFP-2A-BSD (Addgene, #167930). Murine Aldoa sgRNA sequences originate from ref. ⁶⁴ (Addgene, #1000000052). Cells were infected and selected with the respective agent and either used as pools or single clones, as indicated. Silencing and knockout efficiency was confirmed using western blotting. Sequences are provided in Supplementary Table 3.

Analysis of cell viability and proliferation using crystal violet or live imaging

Cells were seeded in 24- or 96-well plates and treated as indicated. After incubation, cells were washed with PBS and fixed for 10 min in 3.7% paraformaldehyde. Cells were washed and stained with 0.1% crystal violet solution (Sigma) for 1 h. Plates were rinsed in water, dried and extracted using 10% acetic acid. Absorbance was measured at 550 nm. To monitor cell proliferation, cells were seeded on 96-well plates and confluency was monitored over time using live-cell imaging (Incucyte, Sartorius).

Western blot analysis

Cells were lysed in RIPA buffer (150 mM NaCl, 50 mM Tris, pH 8.0, 1% (v/v) NP-40, 0.5% (w/v) sodium deoxycholate and 0.1% (w/v) SDS) with protease and phosphatase inhibitors for 30 min and cleared by centrifugation. Proteins were quantified using BCA (Thermo Scientific). Proteins were separated on SDS–PAGE and blotted onto PVDF membrane (Immobilon), treated with blocking solution (5% BSA) and incubated with primary and secondary antibodies in 5% BSA. Signals were detected on a ChemiDoc (Bio-Rad). The antibodies used were anti-ALDOA (11217-1-AP, Proteintech), anti-GPI (15171-1-AP, Proteintech or CSB-PA00367A0Rb, Cusabio), anti-ACC (3662, Cell Signalling), anti-phospho-ACC (3661, Cell Signalling) (all diluted 1:1,000) and anti-vinculin (Sigma, V9131) (diluted 1:2,000). HR-coupled secondary antibodies were from GE Healthcare.

Analysis of DepMap and Cancer Genome Atlas data

CRISPR (DepMap Public 23Q2+Score, Chronos) and RNAi (Achilles+DRIVE+Marcotte, DEMETER2) essentiality scores for glycolytic enzymes were downloaded from depmap.org. Survival analyses in LIHC were performed using GEPIA2 (ref. ⁶⁵).

Seahorse analysis

Cells (2 × 10⁴ cells per well) were seeded onto Seahorse XFe96 microplates (Agilent) coated with Cell-Tak (Corning, 354240) using a centrifugation protocol (200g for 1 min, zero brake setting). A Glycolysis Stress Test and Mitochondrial Stress Test was performed according to the manufacturer’s protocols. For the Glycolysis Stress Assay, 10 mM glucose was injected, followed by 2 µM oligomycin and 50 mM 2-deoxyglucose. For the Mitochondrial Stress Assay, 2 µM oligomycin was injected, followed by 1 μM FCCP and 0.5 μM rotenone/antimycin A. Energy maps were generated by plotting oxygen consumption rates and extracellular acidification rates before and after addition of oligomycin.

Metabolomics

For extraction of polar metabolites, a solid phase extraction was used, Cells (~0.5–1 × 10⁶) were washed with cold ammonium acetate (154 mM), snap frozen in liquid nitrogen and scraped off in 0.25 ml ice-cold methanol:H₂O:acetonitrile (50:20:30 v/v) containing internal standards (MSK-CAA-1, DLM-7654, DLM-9476, DLM-9045, DLM-9071, DLM-6068, DLM-831 and DLM-3487, Cambridge Isotope Laboratories). Samples were subsequently run through a C18 8B-S001-DAK solid phase column (previously activated using acetonitrile and equilibrated using methanol:H₂O:acetonitrile (50:20:30 v/v)), the eluate was dried in a vacuum concentrator. Dried metabolite extracts were dissolved in 100 μl 5 mM NH₄OAc in CH₃CN/H₂O (75:25, v/v) and 3 µl of each sample was applied to an amide-HILIC column (2.6 μm, 2.1 × 100 mm, Thermo Fisher, 16726-012105). Metabolites were separated at 30 °C by LC using a DIONEX Ultimate 3000 UPLC system and the following solvents: solvent A consisting of 5 mM NH₄OAc in CH₃CN:H₂O (5:95, v/v) and solvent B consisting of 5 mM NH₄OAc in CH₃CN:H₂O (95:5, v/v). The LC gradient programme was: 98% solvent B for 2 min, followed by a linear decrease to 40% solvent B within 5 min, then maintaining 40% solvent B for 13 min, then returning to 98% solvent B in 1 min and then maintaining 98% solvent B for 5 min for column equilibration before each injection. The flow rate was maintained at 350 μl min⁻¹. The eluent was directed to the hESI source of the Q Exactive mass spectrometer (QE-MS; Thermo Fisher Scientific) from 1.85 min to 18.0 min after sample injection. The scan range was set to 69.0–1,000 m/z with a resolution of 70,000 and polarity switching (negative and positive ionization). Peaks corresponding to the calculated metabolites masses taken from an in-house metabolite library (MIM ± 5 ppm) were integrated using the El-MAVEN software v.12.1-beta⁶⁶. For the targeted quantification of FBP, extraction and run were performed as stated above with the following exceptions: 6 µM U-¹³C₆ FBP (Cambridge Isotope Laboratories, CLM-8962) was used as standard. The LC gradient programme was: 98% solvent B for 2 min, followed by a linear decrease to 30% solvent B within 3 min, then maintaining 30% solvent B for 15 min, then returning to 98% solvent B in 1 min and then maintaining 98% solvent B for 5 min for column equilibration before each injection. The scan range was set to 200–500 m/z with a resolution of 70,000 and only conducted in negative mode.

Stable-isotope tracing and glycolytic flux measurements

For glucose and glutamine tracing cells were cultured in MPLM medium containing solvent control (ethanol) or doxycycline (1 μg ml⁻¹) for 48 h at which point the cells received fresh MPLM medium for 1 h before being incubated in MPLM medium containing either U-¹³C-glucose (CLM-1396-1) or U-¹³C-glutamine (CLM-1822-H) at 4.4 mM or 1 mM respectively, for 10 h before metabolite extraction and metabolomics analysis. For 1,2-¹³C-glucose tracing, cells first received fresh medium for 1.5 h before incubation in MPLM containing 4.4 mM 1,2-¹³C-glucose (CLM-504-PK Cambridge Isotope Laboratories) for 5, 10, 30 or 240 min before metabolite extraction and metabolomics analysis. Isotopologue distribution of 3PG was determined at isotopic steady state (240 min). Peaks were integrated using the El-MAVEN software and the R package IsoCorrectoR⁶⁷ was used to correct for natural ¹³C abundance. Fractional labelling over time from 1,2-¹³C-glucose tracing was used to determine glycolytic flux. To this end, flux f and maximum enrichment E were estimated in R using nonlinear fitting of a negative exponential function using the function drc() in the package ‘drc’ and the function DRC.negExp() in the package ‘aomisc’, where e is the enrichment at time point t (e = E(1−e^-ft))^68,69

Mathematical modelling

To evaluate the relationship between glucose availability, glycolytic flux and ALDOA catalytic activity, we extended a previously published minimal kinetic model of glycolysis¹⁶. The model consists of five coarse-grained reactions (upper glycolysis, including glucose uptake, aldolase (ALD), lumped lower glycolysis, cellular ATP utilization (ATPase) and provision and utilization of Pi (PiT)) and describes the dynamics of the intracellular concentrations of FBP, triosephosphates (TP), ATP and free Pi using ordinary differential equations.

$$\frac{{\rm{d}}\left[{FBP}\right]}{{\rm{d}}t}=+{v}_{{UG}}-{v}_{{ALD}}$$

(1)

$$\frac{{\rm{d}}\left[{TP}\right]}{{\rm{d}}t}=+2{v}_{{ALD}}-{v}_{{LG}}$$

(2)

$$\frac{{\rm{d}}\left[{ATP}\right]}{{\rm{d}}t}=-2 {v}_{{UG}}+2 {v}_{{LG}}-{v}_{{ATPase}}$$

(3)

$$\frac{{\rm{d}}\left[{Pi}\right]}{{\rm{d}}t}=+{v}_{{ATPase}}+{v}_{{PiX}}-2 {v}_{{LG}}$$

(4)

All reactions are modelled by kinetic rate equations; parameterization follows van Heerden et al.¹⁶, with modifications (details of rate equations are described in Supplementary Information). In contrast with van Heerden et al.¹⁶, the model explicitly describes the dependence of flux on external glucose and expression of aldolase. The model was evaluated with respect to its stability for three different conditions (1) physiological AldoA (100%) and high glycolytic flux ([Glc] = 5 mM); (2) depleted AldoA (20%) and high glycolytic flux ([Glc] = 5 mM); and (3) depleted AldoA (20%) and low glycolytic flux ([Glc] = 1 mM). To test for the robustness of the result with respect to the choice of kinetic parameters, we performed random sampling of parameters (Monte-Carlo analysis), demonstrating that the described instability is an inherent feature of the pathway, independent of the precise choice of parameters. Numerical details are provided in the Supplementary Information.

Phosphate, ATP, ATP/ADP, NAD(P)⁺/NAD(P)H measurements

Pi was measured using phosphate assay kit (ab65622, Abcam) according to the manufacturer’s instructions, results were then normalized to protein content. Intracellular ATP was measured using a CellTiter-Glo luminescence assay (G7571, Promega) according to the manufacturer’s instructions; results were then normalized to cell density/or protein as assayed on an identical plate plated in parallel using crystal violet or BCA assay. The ATP:ADP ratio was measured using an ADP:ATP ratio assay kit (MAK135, Sigma) according to the manufacturer’s instructions. The NAD(P)⁺:NAD(P)H ratio was measured using NAD:NADH-Glo and NADP:NADPH-Glo assays (G9071 and G9081, Promega) according to the manufacturer’s instructions.

Oxidative stress measurements

Cells were cultured in medium containing 2 µM CellROX deep red reagent (C10491 Molecular Probes) and red fluorescence was monitored every 2 h using an Incucyte live-cell imaging instrument (Sartorius). As a positive control, cells were treated with 100 µM tert-butyl hydroperoxide (C10491 Molecular Probes).

Flow cytometry

Cells were treated with 10 µM bromodeoxyuridine (BrdU) for 30 min before being collected by trypsinization. The pellet was washed with cold PBS and fixed in 80% ethanol. Subsequently, 500,000 cells were washed in PBS and incubated for 30 min in 2 M HCl/0.5% Triton X-100 at room temperature. Cells were centrifuged at 400g, resuspended in 0.1 M Na2B4O7 (pH8.5) centrifuged again at 400g and then incubated for 30 min in 100 μl 1% BSA in PBS-T + 10 μl anti-BrdU V450 (BD, 560810) antibody. The cells were then washed in 1% BSA in PBS-T before being incubated in 38 mM citrate solution containing propidium iodide at a concentration of 54 µM and then analysed on a BD FACS-Canto II flow cytometer using BD FACSDIVA software. Cell cycle analysis was performed using FlowJo (v.10).

Animal experiments

All animal experiments were approved by the committees of the regional authority of the state of Bavaria (Regierungspräsidium Unterfranken, RUF 55.2.2-2532-2-751) and the state of Baden-Württemberg (Regierungspräsidium Karlsruhe, G-107/20). All mice were housed and maintained under pathogen-free conditions with a 12-h light–dark cycle, 45–65% relative humidity, temperature of 22 ± 2 °C and ad libitum access to water and food in accordance with institutional guidelines. Hydrodynamic tail vein injection was performed in 6–7-week-old male or female C57BL/6N WT mice. Pilot experiments confirmed no significant difference between survival of male and female mice in this model. Mice were randomly allocated to the different groups and injected with 25 µg of transposon plasmids (pT-CaMIA-shRen, pT-CaMIA-shAldoa.1280 or pT-CaMIA-shAldoa.558) together with 5 µg of plasmid coding for the Sleeping Beauty (SB) transposase diluted in 0.9% NaCl solution to a final volume of 10% of mouse body weight. DNA for hydrodynamic tail vein injection was generated using the QIAGEN EndoFreeMaxi kit. Animals that were not efficiently injected within 10 s were excluded from the experiments. Mice were observed until the predefined humane end point had been reached and differences in survival were determined by Kaplan–Meier analysis using a log-rank test. The SB transposase (pSB13) and the pT-CaMIA transposon plasmids have been described previously^27,70. The shRNAs were shuttled into pT-CaMIA using XhoI and MluI–AscI fragments.

Immunohistochemistry

Paraffin-embedded sections of murine liver samples were cut into 3-µm sections with a microtome (HM 355S). Slides were deparaffinized, hydrated and submitted to antigen retrieval and blocking. Slides were stained with anti-Ki67 (IHC-00375, 1:300 dilution, Bethyl Laboratories), developed with Alkaline Phosphatase Streptavidin (SA-5100, 1:200 dilution, Vector Laboratories) and counterstained with haematoxylin (Carl Roth, T865.3). Slides were scanned at ×20 resolution using a slide scanner (Zeiss Axioscan 7) and analysed using QuPath (v.0.3.2).

Statistical analysis

Statistical analyses of data obtained from metabolomics were performed with MetaboAnalyst v.6.0 (www.metaboanalyst.ca). Statistical analyses of the other data were carried out with GraphPad Prism v.9. Groups were compared by multivariate (principal-component analysis) and/or univariate analyses (t-test or analysis of variance (ANOVA), P < 0.05, false discovery rate (FDR)-applied) followed by post hoc tests, unless otherwise stated in the figure legend. No statistical methods were used to predetermine sample sizes for cell culture experiments. Power analysis was performed to determine size of animal cohorts based on pilot experiments (type 1 error of 5%; type 2 error of 20%; Software R, v.4.3.0). Data distribution was assumed to be normal but this was not formally tested. Randomization was performed for animal experiments. Image analysis of tissue samples was performed in a blinded manner using automated methods. No animals or data points were excluded. All data were presented as mean ± s.d. of independent biological replicates with the number of replicates stated in the figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Source data are provided as supplementary materials. All other data and materials will be made available upon request. Source data are provided with this paper.

References

Ward, P. S. & Thompson, C. B. Metabolic reprogramming: a cancer hallmark even warburg did not anticipate. Cancer Cell 21, 297–308 (2012).
Article PubMed PubMed Central CAS Google Scholar
DeBerardinis, R. J. & Chandel, N. S. We need to talk about the Warburg effect. Nat. Metab. 2, 127–129 (2020).
Article PubMed Google Scholar
Vander Heiden, M. G., Cantley, L. C. & Thompson, C. B. Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science 324, 1029–1033 (2009).
Article Google Scholar
Wice, B. M., Reitzer, L. J. & Kennell, D. The continuous growth of vertebrate cells in the absence of sugar. J. Biol. Chem. 256, 7812–7819 (1981).
Article PubMed CAS Google Scholar
Skinner, O. S. et al. Salvage of ribose from uridine or RNA supports glycolysis in nutrient-limited conditions. Nat. Metab. 5, 765–776 (2023).
Article PubMed PubMed Central CAS Google Scholar
Nwosu, Z. C. et al. Uridine-derived ribose fuels glucose-restricted pancreatic cancer. Nature 618, 151–158 (2023).
Article PubMed PubMed Central CAS Google Scholar
Tanner, L. B. et al. Four key steps control glycolytic flux in mammalian cells. Cell Syst. 7, 49–62 e48 (2018).
Article PubMed PubMed Central CAS Google Scholar
Ros, S. & Schulze, A. Balancing glycolytic flux: the role of 6-phosphofructo-2-kinase/fructose 2,6-bisphosphatases in cancer metabolism. Cancer Metab. 1, 8 (2013).
Article PubMed PubMed Central Google Scholar
Bensaad, K. et al. TIGAR, a p53-inducible regulator of glycolysis and apoptosis. Cell 126, 107–120 (2006).
Article PubMed CAS Google Scholar
Ros, S. et al. Functional metabolic screen identifies 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 4 as an important regulator of prostate cancer cell survival. Cancer Discov. 2, 328–343 (2012).
Article PubMed CAS Google Scholar
Kochanowski, K. et al. Functioning of a metabolic flux sensor in Escherichia coli. Proc. Natl Acad. Sci. USA 110, 1130–1135 (2013).
Article PubMed CAS Google Scholar
Huberts, D. H., Niebel, B. & Heinemann, M. A flux-sensing mechanism could regulate the switch between respiration and fermentation. FEMS Yeast Res. 12, 118–128 (2012).
Article PubMed CAS Google Scholar
Ortega, A. D. et al. A synthetic RNA-based biosensor for fructose-1,6-bisphosphate that reports glycolytic flux. Cell Chem. Biol. 28, 1554–1568 e1558 (2021).
Article PubMed CAS Google Scholar
Miyazawa, H. et al. Glycolytic flux-signaling controls mouse embryo mesoderm development. eLife https://doi.org/10.7554/eLife.83299 (2022).
Teusink, B., Walsh, M. C., van Dam, K. & Westerhoff, H. V. The danger of metabolic pathways with turbo design. Trends Biochem. Sci. 23, 162–169 (1998).
Article PubMed CAS Google Scholar
van Heerden, J. H. et al. Lost in transition: start-up of glycolysis yields subpopulations of nongrowing cells. Science 343, 1245114 (2014).
Article PubMed Google Scholar
Blazquez, M. A., Lagunas, R., Gancedo, C. & Gancedo, J. M. Trehalose-6-phosphate, a new regulator of yeast glycolysis that inhibits hexokinases. FEBS Lett. 329, 51–54 (1993).
Article PubMed CAS Google Scholar
Wang, H. & Iynedjian, P. B. Acute glucose intolerance in insulinoma cells with unbalanced overexpression of glucokinase. J. Biol. Chem. 272, 25731–25736 (1997).
Article PubMed CAS Google Scholar
Patra, K. C. et al. Hexokinase 2 is required for tumor initiation and maintenance and its systemic deletion is therapeutic in mouse models of cancer. Cancer Cell 24, 213–228 (2013).
Article PubMed PubMed Central CAS Google Scholar
Hashimoto, M. & Wilson, J. E. Membrane potential-dependent conformational changes in mitochondrially bound hexokinase of brain. Arch. Biochem. Biophys. 384, 163–173 (2000).
Article PubMed CAS Google Scholar
Pedersen, P. L., Mathupala, S., Rempel, A., Geschwind, J. F. & Ko, Y. H. Mitochondrial bound type II hexokinase: a key player in the growth and survival of many cancers and an ideal prospect for therapeutic intervention. Biochim. Biophys. Acta 1555, 14–20 (2002).
Article PubMed CAS Google Scholar
Majewski, N. et al. Hexokinase-mitochondria interaction mediated by Akt is required to inhibit apoptosis in the presence or absence of Bax and Bak. Mol. Cell 16, 819–830 (2004).
Article PubMed CAS Google Scholar
Pastorino, J. G. & Hoek, J. B. Regulation of hexokinase binding to VDAC. J. Bioenerg. Biomembr. 40, 171–182 (2008).
Article PubMed PubMed Central CAS Google Scholar
Yuneva, M. O. et al. The metabolic profile of tumors depends on both the responsible genetic lesion and tissue type. Cell Metab. 15, 157–170 (2012).
Article PubMed PubMed Central CAS Google Scholar
Faubert, B., Solmonson, A. & DeBerardinis, R. J. Metabolic reprogramming and cancer progression. Science https://doi.org/10.1126/science.aaw5473 (2020).
Dauch, D. et al. A MYC-aurora kinase A protein complex represents an actionable drug target in p53-altered liver cancer. Nat. Med. 22, 744–753 (2016).
Article PubMed CAS Google Scholar
Rudalska, R. et al. LXRα activation and Raf inhibition trigger lethal lipotoxicity in liver cancer. Nat. Cancer 2, 201–217 (2021).
Article PubMed CAS Google Scholar
Cantor, J. R. et al. Physiologic medium rewires cellular metabolism and reveals uric acid as an endogenous inhibitor of UMP synthase. Cell 169, 258–272 e217 (2017).
Article PubMed PubMed Central CAS Google Scholar
Young, R. M. et al. Dysregulated mTORC1 renders cells critically dependent on desaturated lipids for survival under tumor-like stress. Genes Dev. 27, 1115–1131 (2013).
Article PubMed PubMed Central CAS Google Scholar
Snaebjornsson, M. T. & Schulze, A. Non-canonical functions of enzymes facilitate cross-talk between cell metabolic and regulatory pathways. Exp. Mol. Med 50, 1–16 (2018).
Article PubMed CAS Google Scholar
Lew, C. R. & Tolan, D. R. Targeting of several glycolytic enzymes using RNA interference reveals aldolase affects cancer cell proliferation through a non-glycolytic mechanism. J. Biol. Chem. 287, 42554–42563 (2012).
Article PubMed Central Google Scholar
Lu, M., Sautin, Y. Y., Holliday, L. S. & Gluck, S. L. The glycolytic enzyme aldolase mediates assembly, expression, and activity of vacuolar H+-ATPase. J. Biol. Chem. 279, 8732–8739 (2004).
Article PubMed CAS Google Scholar
Zhang, C. S. et al. Fructose-1,6-bisphosphate and aldolase mediate glucose sensing by AMPK. Nature 548, 112–116 (2017).
Article PubMed PubMed Central CAS Google Scholar
Kusakabe, T., Motoki, K. & Hori, K. Mode of interactions of human aldolase isozymes with cytoskeletons. Arch. Biochem. Biophys. 344, 184–193 (1997).
Article PubMed CAS Google Scholar
Pachnis, P. et al. In vivo isotope tracing reveals a requirement for the electron transport chain in glucose and glutamine metabolism by tumors. Sci. Adv. 8, eabn9550 (2022).
Article PubMed PubMed Central CAS Google Scholar
de Padua, M. C. et al. Disrupting glucose-6-phosphate isomerase fully suppresses the “Warburg effect” and activates OXPHOS with minimal impact on tumor growth except in hypoxia. Oncotarget 8, 87623–87637 (2017).
Article PubMed PubMed Central Google Scholar
Garcia, D. & Shaw, R. J. AMPK: mechanisms of cellular energy sensing and restoration of metabolic balance. Mol. Cell 66, 789–800 (2017).
Article PubMed PubMed Central CAS Google Scholar
Lane, A. N. & Fan, T. W. Regulation of mammalian nucleotide metabolism and biosynthesis. Nucleic Acids Res. 43, 2466–2485 (2015).
Article PubMed PubMed Central CAS Google Scholar
Hove-Jensen, B. et al. Phosphoribosyl diphosphate (PRPP): biosynthesis, enzymology, utilization, and metabolic significance. Microbiol. Mol. Biol. Rev. https://doi.org/10.1128/MMBR.00040-16 (2017).
Agarwal, M. L. et al. A p53-dependent S-phase checkpoint helps to protect cells from DNA damage in response to starvation for pyrimidine nucleotides. Proc. Natl Acad. Sci. USA 95, 14775–14780 (1998).
Article PubMed PubMed Central CAS Google Scholar
Castaldo, G. et al. Quantitative analysis of aldolase A mRNA in liver discriminates between hepatocellular carcinoma and cirrhosis. Clin. Chem. 46, 901–906 (2000).
Article PubMed CAS Google Scholar
Wang, Y. et al. Identification of four isoforms of aldolase B down-regulated in hepatocellular carcinoma tissues by means of two-dimensional western blotting. Vivo 25, 881–886 (2011).
CAS Google Scholar
Huangyang, P. et al. Fructose-1,6-bisphosphatase 2 inhibits sarcoma progression by restraining mitochondrial biogenesis. Cell Metab. 31, 1032 (2020).
Article PubMed PubMed Central CAS Google Scholar
Li, F. et al. FBP1 loss disrupts liver metabolism and promotes tumorigenesis through a hepatic stellate cell senescence secretome. Nat. Cell Biol. 22, 728–739 (2020).
Article PubMed PubMed Central Google Scholar
Hohmann, S., Bell, W., Neves, M. J., Valckx, D. & Thevelein, J. M. Evidence for trehalose-6-phosphate-dependent and -independent mechanisms in the control of sugar influx into yeast glycolysis. Mol. Microbiol. 20, 981–991 (1996).
Article PubMed CAS Google Scholar
Michl, J. et al. CRISPR-Cas9 screen identifies oxidative phosphorylation as essential for cancer cell survival at low extracellular pH. Cell Rep. 38, 110493 (2022).
Article PubMed PubMed Central CAS Google Scholar
Wu, H. et al. T-cells produce acidic niches in lymph nodes to suppress their own effector functions. Nat. Commun. 11, 4113 (2020).
Article PubMed PubMed Central CAS Google Scholar
Erecinska, M., Deas, J. & Silver, I. A. The effect of pH on glycolysis and phosphofructokinase activity in cultured cells and synaptosomes. J. Neurochem. 65, 2765–2772 (1995).
Article PubMed CAS Google Scholar
Crabtree, H. G. Observations on the carbohydrate metabolism of tumours. Biochem. J. 23, 536–545 (1929).
Article PubMed PubMed Central CAS Google Scholar
Rosas Lemus, M. et al. The role of glycolysis-derived hexose phosphates in the induction of the Crabtree effect. J. Biol. Chem. 293, 12843–12854 (2018).
Article PubMed PubMed Central Google Scholar
Diaz-Ruiz, R. et al. Mitochondrial oxidative phosphorylation is regulated by fructose 1,6-bisphosphate. A possible role in Crabtree effect induction? J. Biol. Chem. 283, 26948–26955 (2008).
Article PubMed CAS Google Scholar
Koobs, D. H. Phosphate mediation of the Crabtree and Pasteur effects. Science 178, 127–133 (1972).
Article PubMed CAS Google Scholar
Darshi, M. et al. Crabtree effect in kidney proximal tubule cells via late-stage glycolytic intermediates. iScience 26, 106462 (2023).
Article PubMed PubMed Central CAS Google Scholar
Wang, J., Morris, A. J., Tolan, D. R. & Pagliaro, L. The molecular nature of the F-actin binding activity of aldolase revealed with site-directed mutants. J. Biol. Chem. 271, 6861–6865 (1996).
Article PubMed CAS Google Scholar
Hu, H. et al. Phosphoinositide 3-kinase regulates glycolysis through mobilization of aldolase from the actin cytoskeleton. Cell 164, 433–446 (2016).
Article PubMed PubMed Central CAS Google Scholar
Kao, A. W., Noda, Y., Johnson, J. H., Pessin, J. E. & Saltiel, A. R. Aldolase mediates the association of F-actin with the insulin-responsive glucose transporter GLUT4. J. Biol. Chem. 274, 17742–17747 (1999).
Article PubMed CAS Google Scholar
Li, M. et al. Transient receptor potential V channels are essential for glucose sensing by aldolase and AMPK. Cell Metab. 30, 508–524 e512 (2019).
Article PubMed PubMed Central CAS Google Scholar
Li, M. et al. Aldolase is a sensor for both low and high glucose, linking to AMPK and mTORC1. Cell Res. 31, 478–481 (2021).
Article PubMed CAS Google Scholar
Fridman, A. et al. Cell cycle regulation of purine synthesis by phosphoribosyl pyrophosphate and inorganic phosphate. Biochem. J. 454, 91–99 (2013).
Article PubMed CAS Google Scholar
Hay, N. Reprogramming glucose metabolism in cancer: can it be exploited for cancer therapy? Nat. Rev. Cancer 16, 635–649 (2016).
Article PubMed PubMed Central CAS Google Scholar
DeWaal, D. et al. Hexokinase-2 depletion inhibits glycolysis and induces oxidative phosphorylation in hepatocellular carcinoma and sensitizes to metformin. Nat. Commun. 9, 446 (2018).
Article PubMed PubMed Central Google Scholar
Fellmann, C. et al. An optimized microRNA backbone for effective single-copy RNAi. Cell Rep. 5, 1704–1713 (2013).
Article PubMed CAS Google Scholar
Li, W. et al. MAGeCK enables robust identification of essential genes from genome-scale CRISPR/Cas9 knockout screens. Genome Biol. 15, 554 (2014).
Article PubMed PubMed Central Google Scholar
Sanjana, N. E., Shalem, O. & Zhang, F. Improved vectors and genome-wide libraries for CRISPR screening. Nat. Methods 11, 783–784 (2014).
Article PubMed PubMed Central CAS Google Scholar
Tang, Z., Kang, B., Li, C., Chen, T. & Zhang, Z. GEPIA2: an enhanced web server for large-scale expression profiling and interactive analysis. Nucleic Acids Res. 47, W556–W560 (2019).
Article PubMed PubMed Central CAS Google Scholar
Melamud, E., Vastag, L. & Rabinowitz, J. D. Metabolomic analysis and visualization engine for LC-MS data. Anal. Chem. 82, 9818–9826 (2010).
Article PubMed PubMed Central CAS Google Scholar
Heinrich, P. et al. Correcting for natural isotope abundance and tracer impurity in MS-, MS/MS- and high-resolution-multiple-tracer-data from stable isotope labeling experiments with IsoCorrectoR. Sci. Rep. 8, 17910 (2018).
Article PubMed PubMed Central CAS Google Scholar
Ritz, C., Baty, F., Streibig, J. C. & Gerhard, D. Dose-response analysis using R. PLoS ONE 10, e0146021 (2015).
Article PubMed PubMed Central Google Scholar
Hothorn, T., Bretz, F. & Westfall, P. Simultaneous inference in general parametric models. Biom. J. 50, 346–363 (2008).
Article PubMed Google Scholar
Carlson, C. M., Frandsen, J. L., Kirchhof, N., McIvor, R. S. & Largaespada, D. A. Somatic integration of an oncogene-harboring Sleeping Beauty transposon models liver tumor development in the mouse. Proc. Natl Acad. Sci. USA 102, 17059–17064 (2005).
Article PubMed PubMed Central CAS Google Scholar

Download references

Acknowledgements

We thank S. Haid, I. Germer and B. Dankworth for technical assistance. We thank the Light Microscopy and FACS Core Units at the German Cancer Research Center (DKFZ) for support. We also thank L. Zender for providing murine liver cancer cell lines (University Hospital Tübingen) and J.P.F. Angeli (Rudolf Virchow Zentrum Würzburg) for providing human liver cancer cell lines. This work was funded by the German Research Foundation (FOR-2314, SPP-2306 and 453048493), the German Cancer Aid (70114554) and the Federal Ministry of Education and Research (SMART-NAFLD). K.A.-S. acknowledges funding by the German Academic Exchange Service. D.D. acknowledges funding by the Deutsche Forschungsgemeinschaft under Germany’s Excellence Strategy (EXC 2180 – 390900677; Image-guided and functionally instructed tumour therapies, iFIT).

Funding

Open access funding provided by Deutsches Krebsforschungszentrum (DKFZ).

Author information

Authors and Affiliations

Division of Tumor Metabolism and Microenvironment, German Cancer Research Center (DKFZ), Heidelberg, Germany
Marteinn T. Snaebjornsson, Philipp Poeller, Lisa Schlicker, Alina M. Winkelkotte, Adriano B. Chaves-Filho, Kamal M. Al-Shami, Carolina Dehesa Caballero, Ioanna Koltsaki, Felix C. E. Vogel, Roberto Carlos Frias-Soler & Almut Schulze
Faculty of Biosciences, Heidelberg University, Heidelberg, Germany
Philipp Poeller, Alina M. Winkelkotte, Kamal M. Al-Shami & Carolina Dehesa Caballero
Institute for Theoretical Biology, Humboldt University of Berlin, Berlin, Germany
Daria Komkova & Ralf Steuer
Department of Biochemistry and Molecular Biology, Theodor-Boveri-Institute, University of Würzburg, Würzburg, Germany
Florian Röhrig, Jessica D. Schwarz & Elmar Wolf
Department of Medical Oncology and Pneumology, University Hospital Tübingen, Tübingen, Germany
Ramona Rudalska & Daniel Dauch
IFIT Cluster of Excellence EXC 2180 ‘Image-Guided and Functionally Instructed Tumor Therapies’, University of Tübingen, Tübingen, Germany
Ramona Rudalska & Daniel Dauch
Biochemical Institute, University of Kiel, Kiel, Germany
Elmar Wolf
Tübingen Center for Academic Drug Discovery & Development (TüCAD2), Tübingen, Germany
Daniel Dauch
Peter Debye Institute for Soft Matter Physics, Leipzig University, Leipzig, Germany
Ralf Steuer

Authors

Marteinn T. Snaebjornsson
View author publications
Search author on:PubMed Google Scholar
Philipp Poeller
View author publications
Search author on:PubMed Google Scholar
Daria Komkova
View author publications
Search author on:PubMed Google Scholar
Florian Röhrig
View author publications
Search author on:PubMed Google Scholar
Lisa Schlicker
View author publications
Search author on:PubMed Google Scholar
Alina M. Winkelkotte
View author publications
Search author on:PubMed Google Scholar
Adriano B. Chaves-Filho
View author publications
Search author on:PubMed Google Scholar
Kamal M. Al-Shami
View author publications
Search author on:PubMed Google Scholar
Carolina Dehesa Caballero
View author publications
Search author on:PubMed Google Scholar
Ioanna Koltsaki
View author publications
Search author on:PubMed Google Scholar
Felix C. E. Vogel
View author publications
Search author on:PubMed Google Scholar
Roberto Carlos Frias-Soler
View author publications
Search author on:PubMed Google Scholar
Ramona Rudalska
View author publications
Search author on:PubMed Google Scholar
Jessica D. Schwarz
View author publications
Search author on:PubMed Google Scholar
Elmar Wolf
View author publications
Search author on:PubMed Google Scholar
Daniel Dauch
View author publications
Search author on:PubMed Google Scholar
Ralf Steuer
View author publications
Search author on:PubMed Google Scholar
Almut Schulze
View author publications
Search author on:PubMed Google Scholar

Contributions

M.T.S. and A.S. conceptualized the study and experimental design. M.T.S., P.P., F.R., L.S., A.M.W., A.B.C.-F., K.M.A.-S., C.D.C., I.K., F.C.E.V. and R.C.F.-S. conducted experiments and analysed data. D.K. and R.S. generated the mathematical model. J.D.S. and E.W. provided essential materials. D.D. and R.R. provided crucial methods and discussion. M.T.S., R.S. and A.S. wrote the paper with critical input from all authors.

Corresponding author

Correspondence to Almut Schulze.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Metabolism thanks Michael Lukeyand, Alpaslan Tasdogan the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Alfredo Giménez-Cassina, in collaboration with the Nature Metabolism team.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 In vitro screen identifies ALDOA as essential in HCC cells.

a) Distribution of individual shRNAs in the shRNA library used for the in vitro screen. b) A table showing the 10 top essential genes from the screen for each condition with essentiality scores. Full list is in supplementary table 2. c) Viability of Akt1^myr;Myc^OE;Trp53^–/– cells after 96hrs of AldoA depletion using the five shRNAs against AldoA from the screen. Data are presented as mean values ± SD. Significance was calculated using a two-tailed student’s t-test with FDR correction (n = 8 biologically independent replicates). d) Proliferation of control (shRen) or AldoA-depleted (shAldoA.1280) Akt1^myr;Myc^OE;Trp53^–/– cells in medium containing 1 µg/ml doxycycline (Dox). Data are presented as mean values ± SD (n = 3 biologically independent replicates). e) Representative images of control (shRen) or AldoA-depleted (shAldoA.1280) of Akt1^myr;Myc^OE;Trp53^–/– cells treated with 1 µg/ml doxycycline (Dox) for 48 h. f) Cell viability of Akt1^myr;Myc^OE;Trp53^–/– cells after knockout of Aldoa. Data are presented as mean values ± SD. Significance was calculated using one-way ANOVA with Dunnett’s post-hoc test (n = 3 biologically independent replicates).

Source data

Extended Data Fig. 2 The catalytic function of ALDOA is required for proliferation.

a) Viability of Akt1^myr;Myc^OE;Trp53^–/– cells after AldoA depletion (shAldoA) or control (shRen) or after silencing and re-expression of wild type AldoA (shAldoA_WT), a mutant unable to bind to actin (shAldoA_R43A) or a catalytically inactive mutant (shAldoA_K146Q). Inducible silencing and re-expression were achieved by treating cells with doxycycline (1 µg/ml) for 96 h. Data are presented as mean values ± SD. Significance was calculated using a two-tailed student’s t-test with FDR correction (n = 4 biologically independent replicates). b) Lysates from the cells shown in a) were analysed for expression of AldoA by Western blotting. Vinculin is shown as loading control. c) Viability of Nras^G12V;Myc^OE;Trp53^–/– cells after AldoA depletion (shAldoA) or control (shRen) or after silencing and re-expression of wild type AldoA (shAldoA_WT) or a catalytically inactive mutant (shAldoA_D34S). Inducible silencing and re-expression were achieved by treating cells with doxycycline (1 µg/ml) for 96 h. Data are presented as mean values ± SD. Significance was calculated using a two-tailed student’s t-test with FDR correction (n = 3 biologically independent replicates). d) Viability of Akt1^myr;Myc^OE;Cdkn2a^ARF^-/- cells after AldoA depletion (shAldoA) or control (shRen) or after silencing and re-expression of wild type AldoA (shAldoA_WT) or a catalytically inactive mutant (shAldoA_D34S). Inducible silencing and re-expression were achieved by treating cells with doxycycline (1 µg/ml) for 96 h. Data are presented as mean values ± SD. Significance was calculated using a two-tailed student’s t-test with FDR correction (n = 3 biologically independent replicates).

Source data

Extended Data Fig. 3 Deletion of GPI leads to metabolic reprogramming.

a) Proliferation of control (sgControl) and GPI knockout Akt1^myr;Myc^OE;Trp53^–/– cells (sgGPI #1/#4 and sgGPI #4 cl.1) cultured under hypoxia (0.5% O₂). Data are displayed as mean ± SD (0 h n = 4, all others n = 8 biologically independent replicates). b) Western blot probing for GPI and Vinculin after 48hrs doxycycline (1ug/mL) treatment of shRen, shGPI.213, shGPI.1604 and shGPI.1544 expressing Akt1^myr;Myc^OE;Trp53^-/- cells. c) Proliferation of shRen, shGPI.213, shGPI.1604 and shGPI.1544 expressing Akt1^myr;Myc^OE;Trp53^–/– cells treated with doxycycline (1 µg/mL) or EtOH (solvent). Confluency was detected using live cell imaging of parallel wells. Data are presented as mean values ± SD (shRen and shGPI.155 EtOH n = 11, all others n = 12 parallel wells). d) Deletion of Gpi in human liver cancer cells (HLF, HLE and SNU-387) using CRISPR/Cas9. e) Proliferation of human liver cancer cells after depletion of GPI (sgGPI #1/#3) or control (sgCtrl). Cell mass was detected by crystal violet staining of independently seeded wells. Data are presented as mean values ± SD (HLE: n = 9; SNU387 sgGPI #1/#3 92 h n = 2, all others n = 3; HLF n = 8). f) Glycolytic activity of control (sgControl) and GPI knockout cells (sgGPI #1/#3) Cells were analysed using the glycolytic stress test (Seahorse). Data are presented as mean values ± SD (n = 15 parallel wells). g) Glycolytic activity of shRen, shGPI.213, shGPI.1604 and shGPI.1544 expressing Akt1^myr;Myc^OE;Trp53^–/– cells. Cells were analysed using the glycolytic stress test (Seahorse). Data are presented as mean values ± SD. Significance was calculated using one-way ANOVA with Dunnett’s post-hoc test (shRen and shGPI.1604 n = 21, all others n = 23 parallel wells). h) Diagram indicating the flux of ¹³C-labelled carbon atoms from 1,2-¹³C-Glucose into glycolysis and the pentose phosphate pathway (PPP). i) Isotopologue distribution at steady state for 3-phophoglycerate (3PG) in control cells (sgControl) and Gpi deleted cells (sgGpi #1/#4 or sgGPI #4 cl. 1). Table shows the estimated isotopologue distribution for 3PG assuming 100% flux through glycolysis or 100% flux through PPP⁶⁹ in addition to the observed experimental results. Data are presented as mean values ± SD (n = 2 biologically independent replicates). j) Non-linear fit of relative enrichment of ¹³C labelled 3PG after 5, 10, 30 and 240 min of culture in 1,2-¹³C-Glucose containing medium, (E) is maximum enrichment at isotopic steady state and (f) is calculated flux. Data are displayed as mean ± SD (n = 2 biologically independent replicates).

Source data

Extended Data Fig. 4 Metabolic reprogramming after ALDOA and GPI depletion.

a) Volcano plots comparing intracellular metabolites of AldoA depleted cells (shAldoA.1280) and AldoA re-expression (shAldoA_WT) in Akt1^myr;Myc^OE;Trp53^–/–, Nras^G12V;Myc^OE;Trp53^–/– and Akt1^myr;Myc^OE;Cdkn2a^ARF-/- cells after 48 h of doxycycline treatment. Cells were harvested 6 h after addition of fresh MPLM media. Names of selected metabolites are shown. Significance was calculated using a two-tailed students t-test with FDR correction (n = 3 biologically independent replicates). b) Intracellular metabolite levels for G6P, FBP, phosphocreatine and AMP as well as the ATP/ADP ratio in shRen, shGPI.213, shGPI.1604 and shGPI.1544 expressing Akt1^myr;Myc^OE;Trp53^–/– cells treated with doxycycline (1 µg/mL) for 48hrs. Data are presented as mean values ± SD. Significance was calculated using one-way ANOVA with Dunnett’s post-hoc test (n = 4 biologically independent replicates).

Source data

Extended Data Fig. 5 The effects of ALDOA depletion are partially ameliorated by culture in low glucose.

a) Intracellular levels of inorganic phosphate (Pi) in shRen, shGPI.213, shGPI.1604 and shGPI.1544 expressing Akt1^myr;Myc^OE;Trp53^–/– cells treated with doxycycline (1 µg/mL) for 48hrs. Data are presented as mean values ± SD. Significance was calculated using one-way ANOVA with Dunnett's post-hoc test (n = 4 biologically independent replicates). b) Glycolytic activity of control (shRen) and AldoA knockdown cells (shAldoA) under the indicated glucose concentrations. Cells were analysis using the glycolytic stress test (Seahorse). Data are presented as mean values ± SD (n = 6 parallel wells). c) Proliferation of control (shRen) or AldoA (shAldoA.1280) cells in medium containing ethanol (solvent) and indicated glucose concentrations. Data are presented as mean values ± SD (n = 3 biologically independent replicates).

Source data

Extended Data Fig. 6 Deletion of GPI rescues the impact of ALDOA deletion in human HCC cells.

a) ATP and ADP levels in Akt1^myr;Myc^OE;Trp53^–/– cells after deletion of AldoA or after combined deletion of Aldoa and Gpi. Data are presented as mean values ± SD. Significance was calculated using one-way ANOVA with Dunnett's post-hoc test (n = 3 biologically independent replicates). b) Cell viability of human liver cancer cells (HLE, HLF and SNU-387) after deletion of GPI (sgGPI #1/#3) either alone or on combination with ALDOA- (sgALDOA #1, sgALDOA #2). Data are presented as mean values ± SD. Significance was calculated using a two-tailed student’s t-test with FDR correction (n = 3 biologically independent replicates). c) Immunoblot confirming the deletion of GPI (sgGPI #1/#3) either alone or on combination with ALDOA (sgALDOA #1, sgALDOA #2) in human liver cancer cells (HLE, HLF and SNU-387). Activation of AMPK was determined by analysing S79 phosphorylation on ACC. Vinculin is shown as loading control.

Source data

Extended Data Fig. 7 Depletion of ALDOA has minimal effect on glycolytic flux but impairs the TCA cycle.

a) Non-linear fit of absolute enrichment of ¹³C labelled G6P after 5, 10 or 15 min of culture in U-¹³C-Glucose containing medium, (f) is calculated flux. Data are displayed as mean ± SD (n = 3 biologically independent replicates). b) Akt1^myr;Myc^OE;Trp53^–/– cells expressing shRNA sequences targeting AldoA (shAldoA.558 and shAldoA.1280) or non-targeting controls (shRen) were treated with 1 µg/ml doxycycline (D) or solvent (ethanol, E) for 48 hrs before fresh medium containing 4.4 mM glucose including doxycycline or solvent was added for 1 h. Subsequently, cells were exposed to medium containing either 4.4 mM U-¹³C₆-glucose or 1 mM U-¹³C₅-glutamine for 10 h. Metabolites were extracted and analysed by LC-MS. Graphs show pool size and labelling of TCA cycle intermediates. Data are displayed as mean ± SD (shAA588_E n = 2, all others n = 3 biologically independent replicates). c) Diagram of the TCA cycle showing carbon labelling from glucose and glutamine. d) Pool size and labelling of glutamate in the same samples as in (b). Data are displayed as mean ± SD (shAA588_E n = 2, all others n = 3 biologically independent replicates). e) Intracellular metabolite levels for carbamoyl aspartate, dihydroorotate and orotate in AldoA depleted (shAldoA) or control (shRen) cells and in cells re-expressing wild type (shAldoA_WT) or catalytically inactive AldoA (shAlddoA_D34S). Data are presented as mean values ± SD. Significance was calculated using one-way ANOVA with Dunnett’s post-hoc test (n = 3 biologically independent replicates).

Source data

Extended Data Fig. 8 ALDOA depletion cannot be rescued by nucleoside addition.

a) Cell cycle distribution of control (shRen) or Aldo depleted (shAldoA.1280) Akt1^myr;Myc^OE;Trp53^–/– cells after 48 h of doxycycline treatment (1 µg/ml) followed by addition of fresh medium with or without nucleosides added for 6 h. Cells were incubated with 10 µM of BrdU for 0.5 h before harvest. b) Quantification of cell cycle distribution in control and AldoA depleted cells with or without addition of nucleosides. Data are presented as mean values ± SD. Significance was calculated using multiple two-tailed unpaired t-tests with Benjamini-Krieger-Yekutieli FDR correction (n = 3 biologically independent replicates, *p = 0.0075, **p = 0.0027, ***p = 0.00057, ****p = 0.0008, #p = 0.003, ##p = 0.0006). c) Proliferation of shRen and shAldoA Akt1^myr;Myc^OE;Trp53^–/– cells treated with doxycycline (1 µg/mL) and cultured with or without nucleosides. Data are displayed as mean ± SD (shRen n = 3, all others n = 4 parallel wells).

Source data

Supplementary information

Supplementary Information

Supplementary Modelling and Supplementary Figures.

Reporting Summary

Supplementary Table 1

Sequences of shRNA library used for dropout screen.

Supplementary Table 2

Results of dropout screen as MAGeCK essentiality scores.

Supplementary Table 3

Guide RNA sequences used in this study.

Supplementary Table 4

Composition of the mouse plasma-like medium used in this study.

Source data

Source Data Fig. 2

Statistical source data.

Source Data Fig. 3

Statistical source data.

Source Data Fig. 4

Statistical source data.

Source Data Fig. 5

Statistical source data.

Source Data Fig. 6

Statistical source data.

Source Data Fig. 7

statistical source data.

Source Data Fig. 8

Statistical source data.

Source Data Fig. 8

Gating strategy.

Source Data

Original immunoblots.

Source Data Extended Data

Original immunoblots.

Source Data Extended Data Fig. 1

Statistical source data.

Source Data Extended Data Fig. 2

Statistical source data.

Source Data Extended Data Fig. 3

Statistical source data.

Source Data Extended Data Fig. 4

Statistical source data.

Source Data Extended Data Fig. 5

Statistical source data.

Source Data Extended Data Fig. 6

Statistical source data.

Source Data Extended Data Fig. 7

Statistical source data.

Source Data Extended Data Fig. 8

Statistical source data.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Snaebjornsson, M.T., Poeller, P., Komkova, D. et al. Targeting aldolase A in hepatocellular carcinoma leads to imbalanced glycolysis and energy stress due to uncontrolled FBP accumulation. Nat Metab 7, 348–366 (2025). https://doi.org/10.1038/s42255-024-01201-w

Download citation

Received: 27 November 2023
Accepted: 06 December 2024
Published: 20 January 2025
Issue date: February 2025
DOI: https://doi.org/10.1038/s42255-024-01201-w

This article is cited by

Glucose metabolism and its direct action in cancer and immune regulation: opportunities and challenges for metabolic targeting
- Bo-Syong Pan
- Che-Chia Hsu
- Hui-Kuan Lin
Journal of Biomedical Science (2025)
A computational model elucidates the effects of oncogene-induced expression alterations on the energy metabolism of neuroblastoma
- Mareike Simon
- Uwe Benary
- Jana Wolf
Scientific Reports (2025)
Aldolase A: the broker of glycolysis
- Luiza Martins Nascentes Melo
- Feyza Cansiz
- Alpaslan Tasdogan
Nature Metabolism (2025)
An FBP1-regulated metabolic switch reverses liver senescence and drives hepatocellular carcinoma progression
- Yi Zhang
- Xinyue Zhang
- Zhaojian Gong
Science China Life Sciences (2025)

Subjects

Abstract

Similar content being viewed by others

Main

Results

Common ALDOA dependency across oncogenotypes and conditions

ALDOA depletion exposes a metabolic bottleneck in glycolysis

Disruption of glycolysis at GPI does not block proliferation

Blocking ALDOA catalytic activity induces severe energy stress

Depletion of ALDOA induces imbalanced glycolysis

Combined block of GPI and ALDOA prevents imbalanced glycolysis

Rewiring of metabolism in ALDOA-depleted cells

ALDOA depletion induces S-phase arrest and blocks tumour growth

Discussion

Methods

Cell culture

Library preparation and genetic screening

Genetic modification using shRNA and CRISPR

Analysis of cell viability and proliferation using crystal violet or live imaging

Western blot analysis

Analysis of DepMap and Cancer Genome Atlas data

Seahorse analysis

Metabolomics

Stable-isotope tracing and glycolytic flux measurements

Mathematical modelling

Phosphate, ATP, ATP/ADP, NAD(P)+/NAD(P)H measurements

Oxidative stress measurements

Flow cytometry

Animal experiments

Immunohistochemistry

Statistical analysis

Reporting summary

Data availability

References

Acknowledgements

Funding

Author information

Authors and Affiliations

Contributions

Corresponding author

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Extended data

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

This article is cited by

Search

Quick links

Phosphate, ATP, ATP/ADP, NAD(P)⁺/NAD(P)H measurements