Research ArticleMetabolism

p53 is regulated by aerobic glycolysis in cancer cells by the CtBP family of NADH-dependent transcriptional regulators

See allHide authors and affiliations

Science Signaling  05 May 2020:
Vol. 13, Issue 630, eaau9529
DOI: 10.1126/scisignal.aau9529

Protective p53 response

To fuel their rapid proliferation, tumor cells increase their rates of glycolysis, which not only generates precursors for biosynthetic pathways but also results in an increase in the NADH:NAD+ ratio. Birts et al. found that, in highly glycolytic cancer cells in vitro, the increased NADH:NAD+ ratio resulted in an increase in the abundance of the NADH-bound forms of CtBP family transcriptional regulators, which prevented their interaction with the p53-binding protein HDM2, thus enabling the accumulation of p53. The resulting expression of p53 target genes was required to maintain metabolic homeostasis, highlighting a role for p53 in protecting cells from glycolytic stress.

Abstract

High rates of glycolysis in cancer cells are a well-established characteristic of many human tumors, providing rapidly proliferating cancer cells with metabolites that can be used as precursors for anabolic pathways. Maintenance of high glycolytic rates depends on the lactate dehydrogenase–catalyzed regeneration of NAD+ from GAPDH-generated NADH because an increased NADH:NAD+ ratio inhibits GAPDH. Here, using human breast cancer cell models, we identified a pathway in which changes in the extramitochondrial-free NADH:NAD+ ratio signaled through the CtBP family of NADH-sensitive transcriptional regulators to control the abundance and activity of p53. NADH-free forms of CtBPs cooperated with the p53-binding partner HDM2 to suppress p53 function, and loss of these forms in highly glycolytic cells resulted in p53 accumulation. We propose that this pathway represents a “glycolytic stress response” in which the initiation of a protective p53 response by an increased NADH:NAD+ ratio enables cells to avoid cellular damage caused by mismatches between metabolic supply and demand.

INTRODUCTION

Constitutive aerobic glycolysis (“the Warburg effect”) (1) is a hallmark of cancer cells that is commonly caused by mutations in oncogenes and tumor-suppressor genes (2). It has multiple consequences for tumor cells (2), including the ability to generate adenosine 5′-triphosphate (ATP), which decreases reliance on oxygen for ATP generation, thus reducing the generation of potentially damaging reactive oxygen species (ROS) by the mitochondrial electron transport chain. Through the provision of glucose-6 phosphate for the oxidative pentose phosphate pathway, glycolysis also facilitates the generation of reduced form of nicotinamide adenine dinucleotide phosphate (NADP+) (NADPH), which provides reducing equivalents for ROS-protective pathways (3). Glycolytic intermediates are also important precursors for anabolic pathways involved in DNA, lipid, and protein synthesis (4). The final steps of glycolysis, which generate both ATP and precursors for serine and nucleotide biosynthesis, are dependent on glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (5). GAPDH is, in turn, dependent on the coenzyme NAD+, which it reduces to NADH. To sustain GAPDH activity, a low extramitochondrial-free NADH:NAD+ ratio is maintained by oxidation of NADH to NAD+ by the mitochondrial electron transport chain and, in highly glycolytic cells, lactate dehydrogenase (LDH). Cellular export of the LDH-generated lactate is facilitated by the monocarboxylic acid transporters MCT1 and MCT4 (6). Despite these compensatory mechanisms, an increase in the NADH:NAD+ ratio occurs under physiological and pathophysiological conditions where the rate of glycolytic flux through GAPDH is not fully matched by the cells’ ability to regenerate NAD+; for example, under hypoxia, the NADH:NAD+ ratio increases more than threefold (7). Differences between nontransformed cells and tumor cells are comparable (8), and the ratio is consistently increased in cancer cells exhibiting the Warburg effect (810). This ratio is also increased by lactate (11), enhanced production of which is a defining feature of the Warburg effect (1, 2), and which accumulates in the tumor microenvironment to concentrations that have profound effects on cancer cell phenotype (12, 13). A clear demonstration that the flux of intermediates from the later stages of glycolysis into anabolic pathways can be rate limiting for cancer development is the genomic amplification and overexpression of the PHGDH gene, which encodes phosphoglycerate dehydrogenase, in breast cancer and melanoma, diverting glycolytic carbon into serine and glycine metabolism (14, 15). Furthermore, novel anticancer strategies that aim to target glycolytic cells by inhibiting either LDH or MCT1/2/4-mediated lactate export (2, 1619) further increase the extramitochondrial-free NADH:NAD+ ratio, with consequent negative effects on GAPDH-dependent glycolysis.

An important consequence of the inability of an anabolically active cell to maintain the NADH:NAD+ ratio at the low levels required for rapid glycolysis is an increased likelihood that the cell will be unable to appropriately match its supply of metabolites with its high metabolic demands, a scenario that we have termed “glycolytic stress.” Such a mismatch may result in increased ROS production because of the excessive channeling of pyruvate into the mitochondrial tricarboxylic acid cycle and electron transport chain (16), or nucleotide depletion because of reduced amounts of glycolytic intermediates, such as 3-phosphoglycerate, which are required for anabolic pathways (20, 21). These effects may cause DNA damage, potentially leading to either cell death, or the survival of cells with acquired genetic mutations. We therefore speculated about the existence of a stress-response pathway able to detect such glycolytic stress, which prevents such damage by rebalancing metabolic demand with metabolic capacity. The multistress-responsive transcription factor p53 (22, 23) is a strong candidate for an effector in such a pathway, because it induces a program of gene expression that matches the requirements of such an effector. First, this transcriptional program reduces metabolic demand by inhibiting cell proliferation and suppressing anabolic metabolism, thus minimizing the likelihood that glycolytic stress will result in cellular damage. Second, genes induced by p53 direct cellular metabolism away from GAPDH-dependent glycolysis into pathways that protect from ROS, notably the TIGAR-dependent suppression of phosphofructokinase-1 (PFK-1). This suppression of PFK-1 diverts metabolites into the pentose phosphate pathway and also promotes ATP synthesis by increasing fatty acid oxidation and mitochondrial oxidative phosphorylation (OXPHOS) (3, 2426). Consistent with a role in metabolic homeostasis, p53 regulates the cellular response to energy stress, serine restriction, and NADPH metabolism (2729). Furthermore, in a study of tumor material from patients with breast cancer, wild-type p53 gene status correlated with a lactic acidosis–induced gene signature (30), further supporting a role for p53 in directing a glycolytic stress-dependent transcriptional response in vivo. Whereas increased amounts of tumor lactate can be associated with poor clinical outcome, the presence of a lactic acidosis–induced gene signature is associated with favorable outcome (30), which is consistent with a potential role of this response in the prevention of metabolic imbalance–related genetic damage.

The extramitochondrial-free NADH:NAD+ ratio represents a potential signal for such glycolytic stress. The ratio in nonmalignant cells is ~1:700 (7, 31), and therefore after a typical ~3-fold change in the ratio induced by hypoxia or transformation (7, 8), it is only the concentration of NADH that changes sufficiently to act as such a signal (31). Two proteins are described as both NADH sensors and p53 regulators: NAD(P)H quinone dehydrogenase 1 (NQO1) and C-terminal binding proteins (CtBPs). The mechanism of p53 regulation by NQO1 is well studied (32), but NQO1 does not distinguish between NADH and NADPH [free extramitochondrial concentrations of NADPH exceed that of NADH by about fourfold (33)]. The CtBP proteins (CtBP1 and CtBP2) are near-ubiquitous transcriptional regulators, which are sensitive to NADH but not to NADPH (34, 35). Binding of NADH toggles CtBPs between monomeric (NADHfree) and dimeric (NADHbound) forms, as well as controlling their differential binding to cellular partner proteins (31, 3640). The KD of NADH for CtBP is ~100 nM (about equal to the extramitochondrial-free NADH concentration), whereas the dissociation constant (KD) of NAD+ is >100-fold greater (31). These biochemical studies predict that metabolic alterations that increase the NADH:NAD+ ratio will increase CtBP dimerization and induce a CtBP-dependent phenotypic response; cancer cell biology studies from multiple independent laboratories are entirely consistent with this hypothesis (7, 41, 42). We have previously shown that, when overexpressed, NADH-free (monomeric) forms of CtBPs can bind to the p53 regulator HDM2 [human form of murine double minute-2 (MDM2)] to suppress p53 function (43) and that small interfering RNA (siRNA)–mediated CtBP knockdown induces p53 protein accumulation (44). We have now undertaken further experiments to define the role of this pathway in the response to glycolytic stress.

RESULTS

p53 abundance and activity in proliferating cells are increased by high rates of glycolysis

To determine how high rates of aerobic glycolysis in cancer cells affect p53, it was necessary to generate matched pairs of cell lines that proliferated normally but differed in their glycolytic states. Starting with a highly glycolytic cancer cell line and completely removing glucose from the medium result in ATP depletion and the activation of an adenosine 5′-monophosphate (AMP)/AMP-activated protein kinase (AMPK)–dependent p53 response (28). Similarly, it is not possible to merely reduce the glucose concentration in the culture medium to an extent that would restrict glycolysis, because the cells then rapidly deplete the glucose concentration toward zero (45), unless complex flow cell systems are used to overcome this problem (46). We have thus used the approach of culturing cells in medium containing an alternative sugar, in this case fructose, which is imported into cells less rapidly than is glucose, but is still metabolized by glycolysis. Once cells have adapted their metabolism in response to this reduced intracellular sugar availability, this model enables cell cultures with stable low rates of glycolysis to be established (45, 47, 48). Three different breast cancer cell lines, MCF-7 cells and ZR-75-1 cells (both of which have wild-type p53) and MDA-MB-231 cells (which express a functionally compromised mutant p53), were used for this current study.

When MCF-7 cells, which are normally cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing standard 25 mM glucose (MCF-7GLU cells), were cultured in medium in which glucose was either lacking or replaced with 10 mM fructose, proliferation was markedly decreased (Fig. 1A). However, after adaptation to culture with fructose for 30 days (MCF-7FRU cells), proliferation rates were equal to, or greater than, those of MCF-7GLU cultures (Fig. 1A). Comparable results were obtained with ZR-75-1 cells (which also express wild-type p53) and MDA-MB-231 cells (which express a functionally compromised mutant p53) (fig. S1A).

Fig. 1 p53 protein abundance is increased in glycolytic cells.

(A) MCF-7GLU and MCF-7FRU cells had their culture medium replaced with one containing the indicated sugar and then were analyzed for proliferation using xCELLigence. Left: Cell index showing cell proliferation from 1 to 100 hours after plating. Data are means ± range of at least two experiments. Right: Proliferation rates (40 to 100 hours). Data are representative of two experiments. (B) Glucose- or fructose-adapted breast cancer cells (as indicated) were analyzed for rates of glycolysis-dependent extracellular acidification (ECAR) using a Seahorse XF96. Initial readings were taken in medium lacking sugar, and ports were injected with the sugar to which the cells had been adapted. Data are means ± SEM of four technical replicates. Data are representative of two experiments. 2-DG, 2-deoxyglucose. (C) Determination of the NADH:NAD+ ratios of glucose- and fructose-cultured MCF-7 cells by Peredox biosensor analysis. Data are means ± SEM of at least 80 cells from a representative of two experiments and were analyzed by unpaired t test. Calibration is shown in fig. S1C. (D) Left: MCF-7 cells adapted to the indicated sugar were lysed and subjected to Western blotting analysis with antibodies specific for the indicated targets. Numbers indicate the mean ± SEM of the normalized p53 band intensities from three experiments (R1 to R3) and were analyzed by paired t test. Middle: Immunofluorescence staining of p53 in MCF-7GLU and MCF-7FRU cells. A rainbow lookup table demonstrates p53 staining intensity (blue/low → red/high, ~15-fold maximum range between cells). Scale bars, 50 μm. Right: ZR-75-1 cells adapted to the indicated sugar were lysed and subjected to Western blotting analysis with antibodies specific for the indicated targets. Numbers indicate the mean ± SEM of the normalized p53 band intensities from three experiments and were analyzed by paired t test. p53-pSer15 was undetectable in these cells.

MCF-7GLU cells exhibited high rates of glucose-dependent extracellular acidification (ECAR) (Fig. 1B). In contrast, in MCF-7FRU cells, fructose resulted in barely detectable ECAR that could not even be increased when oligomycin was used to inhibit mitochondrial respiration (Fig. 1B). Furthermore, addition of glucose to temporarily glucose-starved MCF-7GLU cells decreased oxygen consumption (the Crabtree effect), whereas fructose did not similarly affect MCF-7FRU cells (fig. S1B). Thus, compared to MCF-7GLU cells, rates of fructose import into MCF-7FRU cells enabled very low levels of flux through the ATP- and NADH-generating stages of glycolysis and little conversion of pyruvate to lactate. In comparison, ZR-75-1GLU cells were less glycolytic than MCF-7GLU cells, and consequently, although ECAR was further decreased in ZR-75-1FRU cells, the effect was much less pronounced than that in MCF-7 cells (fig. S1B). As further confirmation of decreased glycolytic flux in the fructose-cultured cells, a fluorescent biosensor of the NADH:NAD+ ratio (Peredox, fig. S1C) (11) was expressed in MCF-7 cells and revealed a statistically significant decrease in NADH abundance after adaptation to culture in fructose (Fig. 1C).

Similarly to many nontransformed and cancer cells that retain expression of wild-type p53, MCF-7GLU cells had readily detectable p53 protein (Fig. 1D, left). We found that p53 was not homogeneously distributed but rather was present at high amounts in the nuclei of about 10 to 20% of the cells (Fig. 1D, middle). This is likely because of short pulses of p53 accumulation during cell cycle progression (49). In contrast, MCF-7FRU cells demonstrated a reduction in total p53 abundance to 39 ± 3% of that in MCF-7GLU cells (Fig. 1D, left) and a loss of the high p53-expressing cells (Fig. 1D, middle). Induction of p53 by the DNA-damaging agent doxorubicin was not suppressed in MCF-7FRU cells (fig. S1D). Culture of MCF-7GLU cells in medium lacking sugar resulted in p53 accumulation (fig. S1E), as was previously reported (28). Because MCF-7FRU cells proliferate at least as fast as MCF-7GLU cells, the pulses of p53 accumulation in proliferating cells may be dependent on the cycling cells having a glycolytic phenotype. In the less glycolytic ZR-75-1 cells, the change in carbohydrate source had a lesser effect on p53 abundance (Fig. 1D, right).

A detectable proportion of the p53 protein present in MCF-7GLU cells, but not in MCF-7FRU cells, was phosphorylated at Ser15 (Fig. 1D, left). Therefore, the p53 in MCF-7GLU cells may be active as a positive regulator of gene transcription. To determine whether this was the case, we examined the expression of a panel of known p53-responsive genes in the cells (Fig. 2A, left). When mRNA abundance in glycolytic MCF-7GLU cells was plotted relative to their abundance in nonglycolytic MCF-7FRU cells, a positive effect of glycolysis on the expression of multiple p53-responsive genes was revealed. Furthermore, in the mutant p53-expressing MDA-MB-231 cell line, the abundance of most of these mRNAs was insensitive to glycolysis. In the wild-type p53-expressing ZR-75-1 cells, a positive effect of glycolysis on the abundance of p53-responsive mRNAs was also observed (Fig. 2A, right), although the magnitude of the effect was lower than that in MCF-7 cells. This is consistent with the lower rates of glycolysis in ZR-75-1 cells and the smaller effect on p53 protein abundance; however, it does confirm that the effects of glycolysis on p53-dependent transcriptional programs occur in a cell line other than MCF-7 cells. To determine the p53 dependency of the expression of these genes in MCF-7GLU cells, p53 was either knocked down with siRNA (Fig. 2B, left; see fig. S2 for knockdown experiments) or further induced with the HDM2 inhibitor, nutlin-3 (Fig. 2B, right). This experiment confirmed that the expression of most of the transcripts (for example, DDB2, HDM-P2, and CDKN1A) was p53 dependent, but for some genes (for example, DRAM1), the regulation appeared to be more complex.

Fig. 2 Glycolysis promotes p53-dependent gene expression.

(A) Reverse transcription (RT)–qPCR analysis of the expression of a panel of known p53-responsive genes in glucose- and fructose-cultured cells. Data are means ± SEM of three experiments. (B) mRNAs were extracted from MCF-7 cells in which p53 had been either suppressed by p53-specific siRNA (72 hours after transfection; data are means ± SEM of three experiments) or induced by treatment with 5 μM nutlin-3 for 24 hours (data are means ± SEM of three experiments). o, outlier. (C and D) Whole-transcriptome, single-cell mRNA-seq was performed on MCF-7GLU and MCF-7FRU cells. Data are combined from three experiments. (C) Violin plots comparing MCF-7GLU cells with MCF-7FRU cells in terms of glycolysis gene expression score (left), percentage of all mRNA transcripts that are derived from mitochondrial DNA (middle), and p53-activity gene expression score (right). (D) Analysis of MCF-7GLU cells alone. Violin plots of p53 activity score (left) and glycolysis score (right), according to assigned cell cycle phase. The percentage of cells in each phase with a score greater than the threshold value is indicated.

Because the expression of p53 was restricted to a small subset of cells in the MCF-7GLU cultures, we then performed whole-transcriptome, single-cell mRNA sequencing to examine the effects of restricted glycolysis on the p53-dependent transcriptional response in single cells. Using data from 344 MCF-7GLU and 416 MCF-7FRU cells, we first generated a glycolysis gene expression score for each cell based on a 29-gene set (see the Supplementary Materials) and used this to identify cells with the highest glycolysis scores, which were thus likely to have the highest potential for the induction of glycolytic stress. We found that there was about a nine-fold increase in the number of cells with the highest scores (>0.35 cuffoff) in MCF-7GLU cells compared to that in MCF-7FRU cells (Fig. 2C, left) Conversely, in MCF-7FRU cells, the percentage of transcripts originating from mitochondrial DNA was relatively increased (Fig. 2C, middle), indicating a compensatory increase in mitochondrial biogenesis in those cells in which glycolysis was restricted. We then used a validated set of 116 p53-activated target genes (50) to generate a p53 activity score for each cell. Application of a cutoff threshold to identify cells with a high p53 activity score (>0.1) identified 6.7% of MCF-7GLU cells and 2.2% of MCF-7FRU cells (Fig. 2C, right). These data are thus consistent with the global transcriptional changes in p53 response genes that we identified (Fig. 2) being focused in a subset of cells that are more prevalent in MCF-7GLU cells than in MCF-7FRU cells. This finding is also consistent with the heterogeneity of protein abundance that we identified earlier (Fig. 1D).

We then proceeded to further examine the nature of the cells with a high p53 activity score in the MCF-7GLU cells only. Differential gene expression analysis between the cells with a high p53 activity score (>0.1) and those with a low score (≤0) identified the gene with the highest relative expression in the high p53 activity score cells to be CDKN1A and in the low-scoring group to be the proliferation marker MKI67 (fig. S3A). Consistent with this, application of a cell cycle phase score to the cells demonstrated that most of the cells with high p53 activity were in G1 phase (Fig. 2D, left). In contrast, the cells with high glycolysis scores were concentrated in the S and G2-M phases (Fig. 2D, right), and relatively few had a high p53 activity score (fig. S3B). Given that pulses of increased p53 protein abundance in proliferating cells are known to only last about 3 hours (49) and that these pulses by definition precede expression of the p53 target genes (51), a lack of overlap between the p53-activating signal and the peak expression of p53-target genes is to be anticipated. The data are consistent with a small fraction of highly glycolytic cells inducing a p53 response, potentially in S phase where cellular NADH:NAD+ ratios are greatest (52), followed by subsequent CDKN1A-induced protective G1 arrest, although the precise pattern of p53 induction is likely to be confounded by the effects of other transient stresses, such as DNA damage (49). Note that, with respect to the potential effects of p53 on metabolism, those cells with the greatest p53 activity were characterized by the expression of GDF15 and TP53INP1, both p53-responsive genes whose products mediate protection from ROS (53, 54). Furthermore, examination of the expression of genes encoding specific glycolytic enzymes in this dataset (fig. S3, C and D) showed that those involved in the latter stages of glycolysis were most consistently decreased in expression in the cells with high p53 activity, with TPI1 showing the most statistically significant differential. This also is consistent with these cells diverting glucose carbon into the pentose phosphate pathway to provide reducing equivalents for protection from ROS (55), in line with previously understood p53-induced protective responses (25).

Inhibiting LDH increases p53 abundance and activity

To investigate the nature of the pathway that signals from glycolysis to p53, we wished to suppress glycolysis at a different point in the pathway to induce a differential effect on glycolytic intermediates to those caused by culture in fructose. We therefore inhibited the LDH-dependent regeneration of NAD+ from NADH. Using the Peredox biosensor, we first confirmed that suppression of cellular LDH activity resulted in the expected inhibition of NAD+ regeneration from NADH in MCF-7GLU cells. LDH is a tetrameric enzyme consisting of LDHA- and LDHB-encoded subunits, with LDHA being the predominant form in highly glycolytic cells (17). LDHA-specific siRNA statistically significantly increased the NADH:NAD+ ratio (Fig. 3A, left). We found that p53-specific siRNA had a comparable effect (Fig. 3A, left), consistent with a role for p53 in the suppression of glycolysis and promotion of OXPHOS. The LDH inhibitor, sodium oxamate (50 mM), also increased the NADH:NAD+ ratio in MCF-7 cells (Fig. 3A, right). We then titrated sodium oxamate in MCF-7GLU cells to determine the optimal concentration for the inhibition of aerobic glycolysis (Fig. 3B). Glucose-dependent ECAR was suppressed by concentrations of oxamate between 10 and 50 mM. At 50 mM and greater concentrations, oxamate also modestly suppressed oxygen consumption, indicating that it had effects on targets other than LDH at these concentrations. A 2-hour incubation of MCF-7GLU cells with 25 mM oxamate did not reduce ATP abundance, and an apparent modest reduction at 50 mM was not statistically significant (Fig. 3C). The concentration range of oxamate at which ECAR was inhibited also overlapped with the dose required to inhibit MCF-7GLU cell proliferation (Fig. 3D) and colony-forming ability (Fig. 3D). Compared to MCF-7GLU cells, MDA-MB-231GLU cells required higher concentrations of oxamate to inhibit their colony-forming ability (Fig. 3E). LDHA-specific siRNA also decreased MCF-7GLU cell proliferation (Fig. 3D). After a time-course analysis of the effects of 50 mM oxamate on p53 protein abundance in MCF-7GLU cells (Fig. 3F, left), we performed an oxamate dose-response analysis with a 2-hour exposure that demonstrated that concentrations of 25 mM and greater induced a rapid accumulation of p53 protein (Fig. 3F, middle), with expression after exposure to 50 mM oxamate increasing to 218 ± 1% of untreated controls (Fig. 3F, right). The p53-inducible protein HDM2 was also increased by 50 mM oxamate (Fig. 3F, left). LDHA-specific siRNA also induced the accumulation of p53 protein (Fig. 3F, middle), as did 50 mM lactate (Fig. 3G). Both treatment with LDHA-specific siRNA (for 72 hours) (Fig. 3H) and 25 to 50 mM oxamate (for 6 hours) (Fig. 3I) induced the expression of multiple p53-responsive gene transcripts, including CDKN1A and PUMA. The relative magnitude of induction of the individual p53-responsive genes differed between these two approaches, which may be due to differences either in the mechanisms of action of the siRNA and the small-molecule inhibitor or the kinetics of LDH inhibition and transcriptional response studied. We also examined the effect of lactate on p53-dependent gene expression (fig. S4). We also confirmed that the oxamate-induced, p53-dependent transcriptional response did not involve AMP-AMPK signaling (fig. S5).

Fig. 3 Inhibition of LDH activates p53.

(A) NADH:NAD+ ratios in MCF-7GLU cells were measured using the Peredox biosensor. Left: Forty-eight hours after transfection with the indicated siRNAs. Data from control siRNA-transfected cells 6 min after the addition of 50 mM lactate to the medium (L:P 50:1) are included as a positive control. Right: No transfection with siRNA. The cells were imaged, as described earlier, and 6 min after the addition of medium containing 50 mM sodium oxamate or 50 mM lactate, and the fold change in green/red ratio was determined. Control, medium alone added. Data in both panels are means ± SEM of at least 35 cells from a representative of two or more experiments and were analyzed by unpaired t test. (B) The effect of increasing concentrations of sodium oxamate on ECAR and OCR in MCF-7 cells was determined using Seahorse XF96. Data are means ± SEM of four experiments. Data were analyzed by unpaired t test and compared to the 0 mM control. (C) ATP was measured in MCF-7GLU cells in 25 mM glucose (GLU) 2 hours after the initiation of treatment with oxamate or 2-deoxyglucose. Data are means ± SEM of three experiments and were analyzed by paired t test compared to the 0 mM control. (D) MCF-7GLU cells were transfected with the indicated siRNAs or were continuously exposed to the indicated concentrations of sodium oxamate, and the effect on proliferation was determined at 72 hours. Data are means ± SEM of three experiments and were analyzed by unpaired t test compared to either the 0 mM oxamate condition or to the control siRNA. (E) Cells were plated as single cells at a low density and exposed to the indicated concentrations of sodium oxamate for 10 days before colonies were counted. Data are means ± SEM of three experiments and were analyzed by unpaired t test and compared to the 0 mM oxamate condition. (F) MCF-7GLU cells. Left: Western blotting analysis of p53 and HDM2 protein abundance during exposure to 50 mM sodium oxamate for the indicated times. Data are representative of two experiments. Middle: Western blotting analysis of p53 protein abundance after 2 hours of exposure to the indicated concentrations of sodium oxamate or 48 hours after transfection with control of LDHA-specific siRNA. Right: Western blotting analysis of p53 abundance in three biological replicates (separated by vertical dotted lines) after 2 hours of exposure to 50 mM sodium oxamate. Numbers under the blot indicate the mean ± SEM of the relative intensity of the p53 bands in the oxamate-treated cells compared to that in the untreated cells, which was set at 1. Data were analyzed by paired t test. (G) MCF-7GLU cells (three biological replicates separated by vertical dotted lines) were cultured as described in (A) for 6 hours in the absence or presence of 50 mM lactate before being analyzed by Western blotting for p53 abundance. Numbers under the blot indicate the mean ± SEM of the relative intensity of the p53 bands in the lactate-treated cells compared to that in the untreated cells, which was set at 1. Data were analyzed by paired t test. (H) Reverse transcription qPCR (RT-qPCR) analysis of mRNAs extracted from MCF-7GLU cells in which LDHA was knocked down by LDHA-specific siRNA. Samples were analyzed 72 hours after transfection. Data are means ± SEM of three experiments. Knockdown of LDHA protein by the specific siRNA is shown in fig. S3. (I) RT-qPCR analysis of mRNAs extracted from MCF-7GLU cells in which LDH was inhibited by treatment with the indicated concentrations of sodium oxamate for 6 hours. Data are means ± SEM of three experiments.

Therefore, these two distinct methods of inhibiting glycolysis had opposing consequences for the activation of p53 in the cell. When glucose import was low such that there were low rates of flux through phospho-fructokinase into the latter stages of glycolysis, p53 abundance was decreased. However, when the block was at a later stage in the pathway such that the accumulation of metabolites in the later stages of glycolysis was induced, p53 abundance increased. As we have shown, NADH was oppositely affected by the two treatments, consistent with a potential role in the signal to p53. We therefore proceeded to examine whether an NADH-CtBP signaling pathway could be controlling p53 in response to changes in glycolytic metabolism.

A role of the CtBP family of NADH-regulated transcriptional regulators in the regulation of p53 abundance by glycolysis

We have previously established a method of cell synchronization, followed by transfection with CtBP-specific siRNA at the time of release, to determine the effects of CtBP suppression on specific stages of the cell cycle (56). We used this model to establish whether loss of CtBPs might influence p53 activity, independently of their effects on mitotic fidelity that we have previously described. As shown by both Western blotting (Fig. 4A) and single-cell immunofluorescence (Fig. 4B), when compared to control siRNA-transfected MCF-7GLU cells, cells in which the expression of CtBP1 and CtBP2 was knocked down by siRNA showed a marked increase in p53 protein abundance by 24 hours after release from G1 arrest. Comparison with the cell cycle analysis (Fig. 4A) demonstrated that CtBPs were required for the prevention of spontaneous accumulation of p53 in S phase. p53 also accumulated in response to CtBP knockdown in MCF-7FRU cells (Fig. 4C), which is consistent with a model in which CtBPs promote p53 turnover in nonglycolytic cells.

Fig. 4 CtBPs are required to prevent the accumulation of p53 protein in proliferating cells.

(A) MCF-7 cells were synchronized by serum starvation for 48 hours and then transfected with control of CtBP1/2-specific siRNAs at the point of restimulation with serum. Left: Cell cycle phase in the control siRNA-transfected cells was determined by propidium iodide flow cytometry (left axis), whereas mitotic events were determined by time-lapse video microscopy (right axis). Right: Western blotting of the indicated samples for CtBP1, CtBP2, and p53. Blots are representative of three experiments. Quantification shows the effect (fold increase) of CtBP1/2-specific siRNA on p53 abundance compared to that of control siRNA at the equivalent times. (B) MCF-7 cells were transfected and restimulated with serum as described in (A), and p53 was detected by immunofluorescence staining. Cells exposed to 5 μM nutlin-3 are included for comparison. Images are representative of two experiments. (C) MCF-7GLU and MCF-7FRU cells transfected with the indicated siRNAs were analyzed by Western blotting at the indicated times after culture under the indicated conditions. Blots are from a single experiment. Quantification of the effect of CtBP1/2-specific siRNA versus control siRNA on p53 abundance at the 30-hour time point is shown. (D) At the point of release from serum starvation, MCF-7 cells were microinjected with the indicated GST miniproteins. Twenty-four hours later, p53 in the injected cells was determined by immunofluorescence staining, and the percentages of p53-positive cells were determined. The numbers of microinjected cells analyzed are shown in parentheses. The broken x axis indicates that the results are from independently controlled experiments. (E) MCF-7 cells were treated with CP61-TAT, TAT only, or dimethyl sulfoxide (DMSO) for 48 hours before being subjected to Western blotting analysis for p53. Cells transfected with CtBP1/2-specific siRNA are included as a positive control. Blots are representative of three experiments.

To investigate which activities of CtBPs were responsible for their p53-suppressing activity, we used our approach of microinjection of dominant-negative fragments of CtBPs to dissect CtBP function (56). Microinjection of MCF-7GLU cells with an N-terminal fragment of CtBP2 (CtBPDN) designed to be targeted to the cell nucleus and disrupt the interaction of CtBPs with multiple protein partners including HDM2 resulted in the accumulation of p53 protein in most of the injected nuclei (Fig. 4D and fig. S6A). Comparable results were obtained by transfection of MCF-7 cells with a plasmid encoding an enhanced green fluorescent protein (EGFP)–CtBPDN fusion protein (fig. S6B). In contrast, a mutated form of this construct (CtBPDNΔNLS) lacking the N-terminal domain amino acids that constitute both a nuclear localization sequence and an HDM2-binding region failed to induce p53 when microinjected (Fig. 4D). In a previous analysis (56), we further dissected the contribution of the two functions of this region by microinjecting directly into the nucleus; however, this was not possible here, because nuclear microinjection alone induced p53 accumulation. A single point mutation (V72R) in the PxDLSx-binding domain with which it interacts with most of its protein partners also rendered CtBPDN unable to induce p53 accumulation, suggesting that complexes involving PxDLS motif–containing proteins are important in the CtBP-dependent regulation of p53 abundance (Fig. 4D). We then compared the effects of CtBPDN with those of CtBPDD, which contains the central dimerization domain of CtBP2 and is designed to inhibit CtBP dimerization. Whereas we previously showed that CtBPDD disrupts the fidelity of mitosis (47), it does not induce p53 protein accumulation. Consistent with this latter finding, our previously described cell-permeable cyclic peptide inhibitor of CtBP dimerization, CP61-TAT (47), also did not induce p53 protein accumulation (Fig. 4E). Together, these data suggest that monomeric forms of CtBP, in complex with other protein partners including, potentially, HDM2, are required for the suppression of p53 protein accumulation. Given that binding to NADH drives CtBPs into the dimeric form, we hypothesized that the accumulation of p53 in highly glycolytic cells or in cells in which NADH accumulates due to LDH inhibition was driven by the NADH-dependent loss of CtBP monomers. This conclusion is consistent with our earlier studies of the effects of overexpressed full-length CtBP2 proteins on p53-dependent transcription (43).

Evidence for an NADH-CtBP-p53 pathway that regulates cell proliferation and survival in response to glycolytic stress

We next determined whether the p53 protein that accumulated in response to the loss of CtBPs was active. siRNA-mediated CtBP depletion from MCF-7GLU cells resulted in the induction of most of the p53-response genes tested (Fig. 5A). Overall, the magnitude of the effect was greater than that obtained by changing the carbohydrate source or suppressing LDH but less than the response to nutlin-3. The relative magnitude of response for different genes also differed between the different interventions. For example, CDKN1A was one of the lesser-induced genes in response to CtBP1/2 knockdown, whereas it was consistently one of the more robustly induced genes in response to the other interventions. Note that several genes involved in the regulation of cellular redox and metabolic pathways (TP53INP1, SESN1, and TIGAR) were induced by CtBP loss. To determine the potential effect of CtBPs on cellular metabolism, we used the Seahorse instrument to study MCF-7GLU cells after knockdown of p53 and of CtBPs. Knockdown of CtBPs alone caused reductions in both ECAR and oxygen consumption rate (OCR) in glucose-cultured cells, whereas p53-specific siRNA caused only a small decrease in OCR (Fig. 5B, left and middle). A more substantial effect on OCR was caused by the combination of CtBP and p53 knockdown. This was not merely indicative of a reduced demand for ATP, because the oxidative capacity of the cells [as revealed by uncoupling mitochondrial respiration from ATP demand with carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP)] was markedly reduced relative to the glycolytic capacity (as revealed by inhibiting mitochondrial ATP synthesis with oligomycin, hence releasing glycolytic enzymes from ATP-mediated negative feedback) (Fig. 5B, middle and right). These two capacities are primarily determined by the expression of genes involved in the metabolic pathways, and p53-dependent stress responses control gene expression to suppresses glycolysis and promote oxidative phosphorylation (24, 26, 57). Therefore, the fact that this ratio was only disrupted when both CtBPs and p53 were knocked down simultaneously can be explained if there exists an autocompensatory relationship between CtBPs and p53, which regulates metabolic homeostasis through the control of gene expression. To further investigate the role of the NADH binding by CtBPs on their regulation of glycolysis, we stably overexpressed in MCF-7GLU cells either wild-type CtBP2 or an NADH-binding incompetent CtBP2 mutant (G189A) (Fig. 5C, left). Compared to the parental line, a panel of control-transfected clones and the wild-type CtBP2–expressing clone, CtBP2G189A cells exhibited an increase in basal glycolysis compared to that in the other lines (Fig. 5C, right). This finding lends further support to a model in which the loss of NADH-free forms of CtBPs in glycolytic cells acts as a negative feedback signal to maintain glycolytic homeostasis.

Fig. 5 Functional consequences of the regulation of p53 by CtBPs.

(A) mRNAs were extracted from MCF-7 cells 72 hours after they had been transfected with control or CtBP1/2-specific siRNAs. Data are means ± SEM of three experiments. Knockdown of CtBP proteins by the specific siRNA is shown in fig. S3. (B) MCF-7 cells were transfected with the indicated siRNAs. Forty-eight hours later, the effects on ECAR (left) and OCR (middle) were determined using the Seahorse XF96 instrument. Data are means ± SEM of five technical replicates and are representative of two experiments. Right: The ratio of glycolytic capacity to mitochondrial respiratory capacity. (C) MCF-7 cells were stably transfected with empty vector or with plasmids encoding CtBP2 or CtBP2G189A and individual clones were isolated. Left: Cells were analyzed by Western blotting with antibodies against the indicated proteins. Blots are representative of two or more experiments. Right: Glycolytic ECAR was determined by Seahorse XF96. (D) MCF-7 cells were transfected with the indicated siRNAs and then analyzed 48 hours later. Left: Western blotting analysis of the indicated proteins. Blots are representative of three experiments. Right: Equal numbers of cells were plated at low density for colony-forming assays. Colonies were counted 10 days later. Data are means ± SEM of three experiments and were analyzed by Fisher’s least significant difference test.

To determine the potential implications of this pathway for cancer cell survival in response to altered cellular metabolism, we examined the colony-forming capacity of MCF-7GLU cells after knockdown of CtBPs, p53, or LDHA alone or in combination (Fig. 5D). CtBP1/2-specific siRNA resulted in statistically significantly decreased colony numbers, and consistent with our previous work (44), cell survival was further decreased when p53 was also knocked down. Thus, the induction of p53 in response to CtBP loss is prosurvival. In contrast, whereas knockdown of LDHA resulted in decreased colony numbers, this effect was reversed when p53 was also knocked down. This implies that the effects of LDHA-specific siRNA on colony formation were either due to the suppression of glycolysis, which p53-specific siRNA overcame by increasing glycolysis, or due to the induction of a p53-dependent, antiproliferative response. In either case, it appears that, in the absence of p53, cells show a higher propensity to maintain proliferation under conditions where their rates of glycolysis would otherwise exceed the capacity of LDH to regenerate NAD+.

Mechanisms of regulation of p53 by CtBP proteins

CtBP2 has an HDM2-binding site in its N-terminal domain, preferentially binds to HDM2 in its NADH-free form, and inhibition of p53-dependent transcription by CtBPs can be HDM2 dependent (43, 56). Because we had established that CtBPs had a potentially direct effect on p53 protein abundance, we investigated their effect on p53 ubiquitylation, which is driven by HDM2. When MCF-7GLU cells in early S phase were analyzed by Western blotting to detect p53, overexposure of the Western blots showed a ladder of high–molecular weight bands (Fig. 6A), which are consistent with ubiquitylated p53 (58). Band intensity increased when the proteasome-mediated degradation of ubiquitylated p53 was blocked by MG132. In contrast, in cells transfected with CtBP1/2-specific siRNA, these high–molecular weight p53 bands were less readily detectable and did not substantially increase with MG132 treatment. Therefore, we conclude that, in normally proliferating cells, ongoing ubiquitylation of p53 was at least partially dependent on CtBPs. In addition, CtBPs bound to both p53 and HDM2 in these cells (Fig. 6B).

Fig. 6 Mechanisms of regulation of p53 by CtBPs.

(A) MCF-7 cells were serum starved for 48 hours transfected with the indicated siRNAs at the point of serum re-stimulation, and then treated with MG132 or DMSO 24 hours later. Cells were lysed and analyzed by Western blotting at the indicated times after serum restimulation. The asterisk indicates that the non-ubiquitylated p53 signal is overexposed to enable visualization of the ubiquitylated p53 bands. Blots are representative of two experiments. (B) MCF-7 cells were transfected with the indicated plasmids together with pEGFP-N1, treated with or without 10 μM nutlin-3, and protein complexes were analyzed by GFP-Trap immunoprecipitation. Blots are representative of two experiments. (C) MCF-7 cells were transfected with the indicated siRNAs at the point of release from 48 hours of serum starvation. Samples were prepared for Western blotting at the indicated times after release. Nutlin and doxorubicin were included as positive controls for p53 induction and modification. Quantification of the blots is shown in fig. S7. (D) MCF-7 cells were treated with 2 μM SB202190, 10 μM KU55933, or 5 μM wortmannin for 4 hours either 48 hours after transfection with the indicated siRNAs or in the presence of neocarzinostatin (200 ng/ml) or exposure to 20 J/m2 ultraviolet-C. Proteins were then analyzed by Western blotting. Quantification shows the effect of KU55933 on both p53 abundance and the p53-Ser15/p53 ratio under each treatment conditions. Blots are representative of three experiments. The dotted vertical line indicates noncontiguous blots. (E) MCF-7 cells were transfected with control or CtBP1/2-specific siRNA. Forty-eight hours later, the cells were treated with 10 μM KU55933 for a further 24 hours. MCF-7 cells pretreated with 10 μM KU55933 for 24 hours were treated with neocarzinostatin (200 ng/ml) or 10 μM nutlin-3 for a further 4 hours before being harvested and being subjected to RT-qPCR of the relative abundances of the indicated mRNAs. Data are means ± SD of two experiments. NCS, neocarzinostatin.

Blockade of HDM2-mediated p53 degradation can occur through multiple mechanisms depending on the stress involved. These include classic pathways such as those induced by ATP depletion and genotoxic stress being posttranslational modifications (PTMs) at the N and C termini of p53 that interfere with HDM2 binding and ubiquitylation, whereas other pathways, such as those induced by ribosomal stress, involve primary effects on protein-protein interactions. To investigate the mechanism whereby CtBP loss promoted p53 accumulation and activity, we examined the role of PTMs. We performed a time-course analysis of the accumulation of multiple p53 PTMs in response to CtBP knockdown in MCF-7GLU cells (Fig. 6C and fig. S7). We found that, compared to that in control siRNA-treated cells, p53 protein abundance began to increase within 24 hours of the release of CtBP-depleted cells from serum starvation. Phosphorylation at the major N-terminal PTM site, Ser15, also increased, although this was not readily apparent until the 48-hour time point. The pSer37 signal in these cells was weak relative to those in the nutlin-3–treated or doxorubicin-treated controls. The abundance of p53-pSer392 increased approximately in line with p53 accumulation, and there was a modest increase in the appearance of p53 acetylated at Ser382 by 48 hours (Fig. 6C). To determine whether phosphorylation at Ser15 was required for p53 accumulation, we added inhibitors of signaling protein kinases to the cells (Fig. 6D). The ataxia telangiectasia mutated (ATM) inhibitor KU55933, in part, suppressed the phosphorylation of p53 at Ser15 in cells treated with CtBP1/2-specific siRNA (Fig. 6D). However, whereas KU55933 blocked p53 protein accumulation in response to neocarzinostatin, p53 protein still accumulated in CtBP-depleted cells in which ATM-dependent phosphorylation of Ser15 was blocked. A lack of dependency on ATM signaling for CtBP knockdown–induced p53 protein activity was confirmed by analysis of the induction of p53 target genes (Fig. 6E). Regulation of p53 by CtBPs may therefore involve protein-protein interactions rather than classical PTM pathways, with the late induction of PTMs that we observed (Fig. 6C) being potentially due to the previously reported effects of CtBP depletion on mitotic fidelity (44).

DISCUSSION

The initial paradigm of p53 function as “guardian of the genome” (59) proposed that the function of p53 that was inhibited by p53-binding oncoproteins such as HDM2 was its ability to protect cells from the mutagenic consequences of genotoxic stress. p53 is now known to respond to, and regulate, many other pathways. Of these, p53-regulated gene expression networks that control cell metabolism have perhaps some of the most fundamental implications in health and disease (22, 24). p53 limits flux through the latter stages of glycolysis before GAPDH-dependent NAD+ reduction while promoting ATP generation by OXPHOS and protection from ROS by promoting flux through the pentose phosphate pathway. Furthermore, its induction of genes such as CDKN1A decreases metabolic demand by inducing cell cycle arrest. Consequently, p53 protects cells from metabolic stresses such as glucose starvation–mediated ATP depletion (28) and serine starvation (29). Our delineation of a protective p53 response in cells that are driven to undergo high rates of glycolytic metabolism that exceed their ability to maintain it through LDH-dependent NAD+ regeneration adds to our understanding of the role of p53 metabolic homeostasis (Fig. 7).

Fig. 7 Model for the role of the NADH-CtBP-p53 pathway in the regulation of glycolytic homeostasis.

Primary metabolic pathways are shown in blue (fructose model is in cyan), cellular phenotypes are in gray, and the CtBP-p53 pathway in shown in orange. Solid lines show metabolic reactions, whereas dashed lines show regulatory relationships. Regeneration of NAD+ from GAPDH-generated NADH by the mitochondrial electron transport chain is reduced by decreased O2 concentrations and increased by ATP generation through glycolysis. The ability of LDH to regenerate NAD+ is linked to rates of lactate export, which are inhibited by high tumor lactate concentrations. Oncogenic mutations drive both cell proliferation and increased rates of glycolysis for the synthesis of anabolic metabolites. The biochemical limitation of glycolysis due to an increased NADH:NAD+ ratio may lead to an imbalance between metabolic demand and supply, resulting in cellular damage. Sensing of this ratio by CtBPs, and the subsequent regulation of p53, matches metabolic demand with supply and reduces cellular damage. In the absence of functional p53, cells continue to proliferate in adverse microenvironments despite metabolic imbalances, and surviving cells acquire further mutations resulting in tumor progression.

Our mechanistic analysis of this pathway builds on our earlier work that showed that CtBPs bind to HDM2 in an NADH-dependent manner to suppress p53-dependent transcription from specific target promoters and that CtBP depletion by siRNA induces a p53 response that protects against the loss of mitotic fidelity that is also induced by CtBP loss (43, 44, 56). Here, our kinetic analysis demonstrates that the induction of p53 accumulation upon CtBP loss occurs before this effect on mitosis. It also precedes, and is independent of, the DNA damage-induced ATM-p53-pSer15 signaling cascade. Overall, our data support a model whereby CtBPs are required for effective HDM2-dependent ubiquitylation and degradation of p53 in proliferating cells. Their loss results in the accumulation of p53 in sufficient amounts to induce a robust transcriptional response. Because it is the NADH-free monomeric forms of CtBPs that interact with HDM2, the increase in NADH abundance under glycolytic stress results in a loss of CtBP-regulated HDM2 activity and thus increased p53 accumulation.

TP53 is one of the most frequently mutated genes in human cancer (60). Given that, in most cell types in the absence of stress, p53 is rendered essentially nonfunctional through rapid rates of proteasome-mediated degradation, the presence of stress signals that activate p53 must be a near-ubiquitous occurrence in human cancer; otherwise, p53-inactivating mutations would not be selected for (61). However, whereas there are exceptions, such as where DNA-damaging chemotherapy appears to select for tumor cells with treatment-induced p53 mutations (62), in most chemo-naïve tumors, the nature of the p53-activating stress is harder to define. Increased rates of aerobic glycolysis are also an extremely common event during many cancers, and they tend to increase further with cancer progression (2, 10). Furthermore, as tumors expand in the absence of a well-developed blood supply, the tumor microenvironment becomes acidified through the build-up of glycolytic lactate (13), which, combined with an increase in glycolysis, drives an increase in the intracellular NADH:NAD+ ratio by inhibiting LDH (63). Here, we showed that this inherent hallmark of most cancer cells acts as a p53-activating stress. This p53-dependent glycolytic stress response pathway could potentially serve as a selection pressure for the acquisition of p53 mutations in a high percentage of human tumors. p53 mutation–containing cells in these hostile microenvironments would continue to proliferate at rates that exceed the ability of glycolysis to provide metabolites for pathways such as nucleotide biosynthesis or ROS protection. Whereas cell death may initially increase in these cells due to increased cellular damage, the increased DNA mutation rates would further promote tumor progression in the cells that do survive (Fig. 7). Our findings are entirely consistent with the observations of Chen et al. (30) who identified a lactic acidosis–induced gene expression signature in normal mammary epithelial cells, and then, in a cohort of patients with breast cancer, showed that this signature is associated with both a wild-type p53 gene status and good prognosis.

MATERIALS AND METHODS

Cell culture, cell cycle analysis, and reagents

Cell culture and analysis of cell DNA content was performed, as previously described (44). Cells used were MCF-7 [American Type Culture Collection (ATCC) HTB-22], ZR-75-1 [ATCC CRL-1500], and MDA-MB-231 [ATCC HTB-26]}. For media with different sugars, 25 mM glucose or 10 mM fructose was added to glucose-free DMEM (Sigma-Aldrich no. D5030) supplemented with sodium bicarbonate (3.7 g/liter), 2 mM glutamine, 1 mM sodium pyruvate, and 10% fetal bovine serum. Cells were grown for ≥21 days with fresh medium added every 2 to 3 days. KU55933, SB202190, and wortmannin were obtained from Tocris. Sodium oxamate, MG132, and neocarzinostatin were obtained from Sigma-Aldrich. Nutlin-3 was obtained from Enzo Life Sciences.

Western blotting, immunofluorescence, and RNAi

Western blotting and immunofluorescence analyses and RNA interference (RNAi)–based experiments were performed, as described previously (44). The following antibodies were used: CtBP1 (E12; Santa Cruz Biotechnologies); CtBP2 (clone 16; BD Biosciences); LDHA (C4B5; Cell Signaling Technology); p53 (DO-1; AbD Serotec); pSer15-p53, pSer37-p53, pSer392-p53, and ac382-p53 (Cell Signaling Technology); HDM2 (2A10, Abcam); and GFP (Abcam). siRNA reagents targeting CTBP1/2 and TP53 were previously described (44). LDHA-specific siRNA (no. s229664) was obtained from Life Technologies. CP61-TAT and TAT were synthesized, as described previously (47).

Seahorse assays

ECARs and OCRs were calculated using a Seahorse XF96 (Agilent) with the glycolytic stress test or mitochondrial stress test according to the manufacturer’s guidelines. Cells were plated at a density of 18,000 per well 18 hours before the assay. Data were normalized to total protein per well. Glycolytic capacity = ([ECAR after oligomycin] − [ECAR after 2-deoxyglucose]); mitochondrial respiratory capacity = ([OCR after FCCP] – [OCR after rotenone + antimycin]).

Cell proliferation, ATP analysis, and colony-forming assays

Cell proliferation was assessed using the xCELLigence Real-Time Analysis system (ACEA Biosciences). Cells were plated at 5000 per well, and proliferation was followed for 100 hours. For ATP analysis, 5000 cells per well were plated in 96-well plates. Twenty-four hours later, metabolic inhibitors were added to the medium. After 2 hours of incubation, ATP was analyzed with an ATPlite Luminescence ATP Detection Assay System (PerkinElmer) according to the manufacturer’s instructions. Assay medium was transferred to white-welled plates and analyzed with a Varioskan Flash instrument (Thermo Fisher Scientific). After background subtraction, the means of technical replicates for each independent experiment were first calculated, and the data were presented as means ± SEM for three independent experiments. Colony-forming assays were performed by plating single cells at low density for 10 days before colonies were counted. Where the effects of siRNA-mediated knockdown on colony forming is shown, the mean of triplicate cell plates was first calculated, and the data were presented as means ± SEM from independent experiments.

Transfection, overexpression, and microinjection

The CtBP2 and CtBP2G189A constructs for overexpression analysis and the CtBP2-GFP construct for immunoprecipitation were generated, as described previously (43), and were used in the transfection of cells with FugeneHD (Promega). Immunoprecipitation was performed with GFP-Trap A (Chromotek). The pcDNA3.1-Peredox-mCherry-NLS plasmid was a gift from G. Yellen (Addgene plasmid no. 32384) and used to transfect cells at 2 μg of DNA per well with FugeneHD (Promega). Seventy-two hours after transfection, the cells were transferred to assay medium [serum-free and sodium bicarbonate–free DMEM (Sigma-Aldrich no. D5030), 2 mM Hepes (pH 7.4), 25 mM glucose, 1 mM sodium pyruvate, and 2 mM glutamine] before imaging. Fluorescence images were captured with an Olympus IX81 microscope. Fluorescence quantification from captured images was performed with ImageJ software. The changes in fluorescence in response to the NADH:NAD+ ratio were calculated, as described by Hung et al. (11). Glutathione S-transferase (GST) and GST-fusion proteins were expressed, purified, and microinjected, as described previously (43, 56).

RNA analysis

Total RNA was isolated from cells with the RNA ReliaPrep system (Promega) and treated with deoxyribonuclease 1. RNA was reverse transcribed with moloney murine leukemia virus (M-MLV) reverse transcriptase and oligo(dT) primer (Promega). Quantitative polymerase chain reaction (qPCR) analysis was performed with a 7500 Real-Time PCR system (Applied Biosystems) (for information on the primers used, see table S1). Relative RNA abundance was calculated by the ∆∆CT method using ACTB (Applied Biosystems no. 4326315E) as endogenous control. The mean of technical qPCR duplicates was first calculated for each sample, and data are presented as means ± SEM from independent experiments. One-sample t tests on the mean response of each gene in the panel were used to determine the statistical significance of the p53-induced gene expression signature (outliers were discarded using Grubb’s test).

Whole-transcriptome, single-cell mRNA sequencing (Drop-seq)

Using a custom microfluidic workstation (https://dropletkitchen.github.io/), experiments were performed according to the microfluidic device design and protocol of Macosko et al. (64) with the following adjustments: 13 PCR cycles and 500 pg of complementary DNA from a total of 1000 cells per culture condition, combined from three independent experiments, were used as a template for Nextera XT library preparation. Libraries were sequenced on the Illumina NextSeq 500 platform. Raw sequencing reads were aligned to hg19 and binned/collapsed onto the cell barcodes corresponding to individual mRNA capture beads using Drop-Seq tools (http://mccarrolllab.com/dropseq) (64). To exclude low-quality cells and likely cell doublets, cells with fewer than 2000 genes and greater than 15,000 unique molecular identifiers were removed. All genes that were not detected in at least three cells were discarded, and all mitochondrial DNA-encoded genes were excluded, leaving 15,946 genes across 760 cells (344 MCF-7GLU cells and 416 MCF-7FRU cells). The digital gene expression matrix was library-size normalized, scaled by the total number of transcripts, multiplied by 10,000, and natural log-transformed before further downstream analysis with Seurat (http://satijalab.org/seurat/) (64). R Scripts are provided in the Supplementary Materials.

Statistical analysis

GraphPad Prism software (GraphPad) was used for statistical analysis. When t tests were used, they were two sided. For all statistical analysis, *P < 0.05, **P < 0.01, and ***P < 0.001.

SUPPLEMENTARY MATERIALS

stke.sciencemag.org/cgi/content/full/13/630/eaau9529/DC1

Fig. S1. Effects of altered media sugar composition on proliferation, metabolism, and p53 abundance in breast cancer cells.

Fig. S2. Validation of siRNA efficacy in MCF-7 cells.

Fig. S3. Extended analysis of single-cell mRNA sequencing of MCF-7 cells.

Fig. S4. The effect of lactate on the induction of p53-responsive genes.

Fig. S5. The effect of AMPKα-specific siRNA on the induction of p53-responsive genes by oxamate.

Fig. S6. Effect of dominant-negative CtBP fragments on p53 abundance in MCF-7 cells.

Fig. S7. Quantification of the effect of CtBP1/2-specific siRNA on p53 PTMs.

Table S1. The qPCR primers and probes used in this study.

Glycolysis gene annotation list for Drop-seq analysis

R script for Drop-seq analysis

REFERENCES AND NOTES

Acknowledgments: Peredox-encoding plasmids were made available by G. Yellen (Harvard Medical School). Funding: Funding from Breast Cancer Now (2014NovPR341 and 2010NovPR12) and Cancer Research UK (C34999/A18087). Medical Research Council, UK, (MC_PC_15078) and a University of Southampton Institute For Life Sciences fellowship (J.W.) supported the establishment of single-cell mRNA sequencing. C.N.B. is now an Against Breast Cancer–funded lecturer at the School of Biological Sciences, University of Southampton. Author contributions: Conceptualization and methodology: J.P.B. and C.N.B. Investigation and analysis: C.N.B., A.B., M.D., C.R.D., S.N., S.K.N., R.P., M.J.J.R.-Z., J.W., and J.P.B. Writing (original draft): J.P.B. and C.N.B. Writing (review and editing): J.P.B., C.N.B., A.B., A.T., J.W., and M.J.J.R.-Z. Funding acquisition: J.P.B., C.N.B., and A.T. Competing interests: J.P.B. has served as a member of the Scientific Advisory Board of Breast Cancer Now. The other authors declare that they have no competing interests. Data and materials availability: The datasets generated analyzed during the current study are available in the Gene Expression Omnibus repository (www.ncbi.nlm.nih.gov/geo/), with the identifier GSE115869. All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article