Activating
KRAS and
BRAF mutations are found in approximately 40% and 10% of human colorectal cancers (CRCs), respectively (
1). BRAF is a direct target of KRAS, and both activate the mitogen-activated protein kinase (MAPK) pathway. Clinical studies indicate that activating mutations in
KRAS and
BRAF predict resistance to epidermal growth factor receptor (EGFR)–targeting agents (
2–
4). Thus, novel therapies for
KRAS or
BRAF mutant CRCs are urgently needed.
Glucose uptake, as measured by [
18F]fluorodeoxyglucose positron emission tomography (FDG-PET), correlates with
KRAS or
BRAF mutations and GLUT1 overexpression in CRCs (
5,
6), consistent with our previous finding that
KRAS or
BRAF mutant CRC cells rewire glucose metabolism, in part by up-regulating
GLUT1 expression (
7). These data suggest a strategy for targeting
KRAS or
BRAF mutant cancers by exploiting the selective expression of
GLUT1 and the metabolic liability that comes with increased reliance on glycolysis.
Dietary vitamin C is transported across cellular membranes by sodium vitamin C cotransporters (SVCTs) and facilitative glucose transporters (GLUTs) (
8,
9). Whereas SVCTs transport vitamin C directly into the cell, GLUTs—mainly GLUT1 and GLUT3—transport the oxidized form of vitamin C, dehydroascorbate (DHA). After import, DHA is reduced to vitamin C at the expense of glutathione (GSH), thioredoxin, and nicotinamide adenine dinucleotide phosphate (NADPH) (
10). Given that GLUT1 levels in
KRAS and
BRAF mutant cells are elevated, we hypothesized that the increase in DHA uptake could disrupt redox homeostasis and compromise cellular viability. To test our hypothesis, we used a panel of isogenic CRC cell lines harboring wild-type or mutant alleles of
KRAS (HCT116 and DLD1) or
BRAF (VACO432 and RKO) (
7).
In cell culture media, vitamin C is oxidized to DHA (half-life ~70 min) unless reducing agents are added (fig. S1) (
11). Using
14C-radiolabeled vitamin C, we tested which form of vitamin C (reduced or oxidized DHA) is preferentially imported. Both HCT116 and VACO432 cells took up [
14C]vitamin C efficiently (
Fig. 1A). However, adding GSH to the media to prevent oxidation of vitamin C to DHA abrogated [
14C]vitamin C uptake (
Fig. 1A). Furthermore, [
14C]vitamin C uptake was significantly decreased in both HCT116 and VACO432 cells treated with a GLUT1-specific inhibitor, STF31, and in GLUT1 knockout cells (
Fig. 1, A and B). Glucose competed with DHA for uptake in CRC cells (fig. S2). These results indicate that CRC cells preferentially import DHA, rather than vitamin C, and that this uptake is mediated by GLUT1 (fig. S3, A and B).
Given the increased expression of GLUT1 in mutant cells, we investigated whether
KRAS or
BRAF mutations influenced vitamin C uptake. We found that the mutant lines took up significantly more [
14C]vitamin C than did their wild-type counterparts (
Fig. 1, B and C). Overexpression of GLUT1 in wild-type cells was sufficient to increase [
14C]vitamin C uptake to levels commensurate with those of the mutants (
Fig. 1B and fig. S3C). Moreover,
KRAS and
BRAF mutant cells imported DHA faster than they did [
14C]vitamin C (fig. S4), consistent with the observation that vitamin C must first be oxidized to DHA to enter cells through GLUT1. Together, these results indicate that GLUT1 is the primary means of vitamin C uptake in CRC cells and that elevated GLUT1 expression in
KRAS or
BRAF mutant cells drives increased DHA uptake.
We next asked whether the increased uptake of DHA in
KRAS and
BRAF mutant cells could affect their survival and growth. When plated at a low density and grown in low-glucose medium (2 mM), all cell lines grew at similar rates and formed colonies (fig. S5). However, 24 to 48 hours of vitamin C treatment inhibited
KRAS and
BRAF mutant cell growth and colony formation, with reduced effects on their wild-type counterparts (
Fig. 2A and fig. S5). Because of the competitive nature of DHA import, mutant lines were most sensitive to vitamin C under low-glucose conditions (2 mM). Nonetheless, selective cytotoxicity against the mutant lines was achieved even under higher-glucose conditions (5 to 20 mM) when treating with less than 1 mM vitamin C (fig. S6), indicating that vitamin C can selectively kill mutant cells under physiological glucose concentration (5 to 10 mM).
Plasma vitamin C concentrations greater than 10 mM are easily achieved in humans and in our murine pharmacokinetic study (fig. S7) without significant toxicity (
12,
13). Vitamin C was cytotoxic rather than cytostatic, as evidenced by increased staining for the apoptotic marker annexin V in the mutants (fig. S8A). Adding GSH to the culture medium was sufficient to rescue the death of each mutant line (
Fig. 2A).
PIK3CA is one of three frequently mutated oncogenes in CRCs in addition to
KRAS and
BRAF. Unlike
KRAS or
BRAF, the
PIK3CA genotype did not predict vitamin C sensitivity (fig. S8B). Notably, although overexpression of
GLUT1 in wild-type cells increased vitamin C uptake (
Fig. 1B), it did not sensitize wild-type cells to vitamin C (fig. S8C); these results indicate that high GLUT1 expression alone, without oncogene-induced metabolic reprograming, is not sufficient to make cells susceptible to vitamin C–dependent toxicity.
We next explored whether vitamin C altered the growth of
KRAS and
BRAF mutant CRC cells in mice. Mice bearing established xenografts derived from parental HCT116 and VACO432 cell lines were treated twice a day via intraperitoneal (ip) injection of high-dose vitamin C (4 g/kg) or phosphate-buffered saline (PBS; vehicle control) for 3 to 4 weeks, at which point control mice had to be killed because of tumor size. Vitamin C treatment significantly reduced tumor growth relative to vehicle control treatment (
Fig. 2B).
KRAS and
BRAF wild-type isogenic HCT116 and VACO432 cell lines cannot form xenograft tumors in mice. To directly test the impact of
Kras mutation on the sensitivity of tumors to vitamin C treatment, we generated a transgenic model of intestinal cancer, driven by either
Apc mutation, or combined
Apc and
Kras (G12D) mutations. Compound mutant mice were generated by crossing available
Apcflox mice (
14),
LSL-KrasG12D mice (
15), and
Lgr5-EGFP-creERT2 (
16) animals, enabling intestinal restricted alteration of
Apc and
Kras. Tumors were induced with a single ip injection of low-dose tamoxifen (20 mg/kg) and treated daily thereafter with high-dose vitamin C (ip, 4 g/kg) for 5 to 7 weeks. Whereas
Apcflox/flox mice showed no difference in polyp burden after vitamin C treatment,
Apcflox/flox/KrasG12D mice had significantly fewer and smaller small intestine polyps (76 versus 165 in control group), confirming that vitamin C selectively affected
Kras mutant tumors (
Fig. 2C and fig. S9). Consistent with experiments in CRC lines, tumors from
Apcflox/flox/KrasG12D mice showed higher GLUT1 expression and greater vitamin C uptake than did tumors from
Apcflox/flox mice (
Fig. 2, D and E, and fig. S10).
To investigate the mechanism by which vitamin C is selectively toxic to
KRAS and
BRAF mutant cells, we used liquid chromatography–tandem mass spectrometry (LC-MS/MS)–based metabolomics to profile metabolic changes after vitamin C treatment (
17). In untreated
KRAS and
BRAF mutant lines, the relative intracellular metabolite levels of glycolysis and the nonoxidative arm of the pentose phosphate pathway (PPP) were increased relative to their isogenic wild-type counterparts (fig. S11). Addition of a MEK1/2 (MAPK kinase) inhibitor to the parental
KRAS or
BRAF mutant cells also decreased glycolytic and PPP metabolite levels, indicating that the increased metabolite levels were driven by oncogene-induced MAPK activity (fig. S12) (
18). Notably, within 1 hour of vitamin C treatment, the metabolic profile of the mutant cells changed markedly. Glycolytic intermediates upstream of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) accumulated while those downstream were depleted, which suggests that GAPDH was inhibited (
Fig. 3A and fig. S13). Also, oxidative PPP metabolites increased (
Fig. 3A and fig. S13), indicating that the blockage may shift glycolytic flux into the oxidative PPP. Indeed, vitamin C treatment stimulated oxidative PPP-dependent
14CO
2 production from [1-
14C]glucose in both
KRAS and
BRAF mutant cells, and to a lesser degree in wild-type cells (fig. S14A). Decreased NADPH/NADP
+ ratios are known to activate glucose-6-phosphate dehydrogenase allosterically to enhance oxidative PPP flux. The increased flux is an attempt to restore cytosolic NADPH back to homeostasis to mitigate oxidative stress (
19). We reasoned that DHA uptake may deplete cellular GSH and NADPH as they are consumed in reducing DHA to vitamin C. If the capacity of this pathway to restore GSH levels is exceeded, cellular reactive oxygen species (ROS) increase because GSH is the major cellular antioxidant (
20). Indeed, the ratio of reduced to oxidized glutathione decreased as intracellular vitamin C increased (
Fig. 3B and fig. S14B). Cysteine, the major limiting precursor for GSH biosynthesis, was also depleted after vitamin C treatment (fig. S13). As expected, vitamin C treatment induced a substantial increase in endogenous ROS in
KRAS and
BRAF mutant cells (
Fig. 3C).
Given that cancer cells with
KRAS or
BRAF mutations are heavily dependent on glycolysis for survival and growth, and that pyruvate (the end product of glycolysis) is a major carbon source for the mitochondrial TCA cycle (
7,
21), we hypothesized that inhibition of GAPDH, a glycolytic enzyme,might deplete adenosine triphosphate (ATP) and thereby induce an energetic crisis ultimately leading to cell death. Vitamin C treatment caused a rapid decrease in the glycolytic rate in
KRAS and
BRAF mutant cells, but not in wild-type cells, as determined by the extracellular acidification rate, a proxy for lactate production (
Fig. 3D and fig. S15). Accordingly, vitamin C induced a significant drop in ATP levels, with a concomitant increase in adenosine monophosphate (AMP) levels (
Fig. 3E and fig. S16A). Within 1 hour, AMP-activated protein kinase (AMPK), a marker for energy stress, was activated; activation was strongest in the mutant lines (
Fig. 3F). The cell-permeable reducing agent and glutathione precursor
N-acetylcysteine (NAC) rescued both AMPK activation and cell death in the mutant lines (
Fig. 3, F and G). Consistent with the in vitro results, supplementing drinking water with NAC over the course of vitamin C treatment abolished the ability of vitamin C to reduce xenograft growth (
Fig. 3H). Similarly, pyruvate and oxaloacetate, both of which can enter the TCA cycle and thus provide ATP, rescued energy stress and cell death, as did Trolox (a water-soluble analog of the antioxidant vitamin E) (
Fig. 3G and fig. S16, B and C). Rotenone, a complex I inhibitor, attenuated the ability of pyruvate to rescue vitamin C–induced cytotoxicity (fig. S17), indicating that the lack of mitochondrial substrates caused by glycolytic inhibition also contributes to ATP depletion in mutant cells (
21).
We next sought to determine the mechanism by which vitamin C inhibits GAPDH. GAPDH is known to have an active-site cysteine (Cys
152) that is targeted by ROS (
22). The active-site cysteine can undergo reversible S-glutathionylation in which the oxidized cysteine forms a mixed disulfide with GSH (Cys-GSH), or undergo further irreversible oxidations that include sulfonic acid (Cys-SO
3H) (
23,
24). Both cases result in loss of GAPDH activity. We measured GAPDH S-glutathionylation after vitamin C treatment by immunoprecipitating endogenous GAPDH and blotting with an antibody that recognizes S-glutathionylation under nonreducing conditions. In both
KRAS and
BRAF mutant lines, GAPDH S-glutathionylation levels were higher in vitamin C–treated cells than in vehicle-treated cells by a factor of 2 to 3 (
Fig. 4A). However, GAPDH sulfonylation was not detected with a GAPDH-SO
3H antibody (
Fig. 4B). GAPDH activity was assayed in lysates of vitamin C treated cells to confirm inhibition by S-glutathionylation (fig. S18). Treatment with vitamin C for 1 hour decreased GAPDH activity by 50% in both
KRAS and
BRAF mutant cells. Combining NAC with vitamin C fully rescued GAPDH activity (fig. S18).
We reasoned that the 50% reduction in GAPDH activity after vitamin C treatment could be explained by S-glutathionylation (
Fig. 4A). However, given that the GAPDH substrates were added to the lysates to perform the activity assay, and in light of the striking accumulation (by as much as a factor of 19) of the GAPDH substrate glyceraldehyde-3-phosphate (G3P) (
Fig. 3A and fig. S13), we suspected that additional mechanisms may contribute to GAPDH inhibition. This led us to examine the levels of the NAD
+ substrate required for GAPDH-dependent oxidation of G3P. In contrast to G3P levels, intracellular NAD
+ levels were significantly diminished after vitamin C treatment (fig. S19). PARP activation due to ROS-induced DNA damage consumes NAD
+ to form adenosine diphosphate (ADP)–ribose polymers on acceptor proteins. We observed PARP activation and phosphorylation of H2AX, a marker of DNA damage, shortly after vitamin C treatment (
Fig. 4C); this finding suggests that PARP activation may diminish NAD
+ levels, thereby further inhibiting GAPDH activity by depleting substrate availability (
25). To investigate whether PARP activation or NAD
+ depletion contributes to vitamin C–induced cytotoxicity in
KRAS and
BRAF mutant cells, we treated cells with a PARP inhibitor, olaparib, or a cell-permeable NAD
+ precursor, nicotinamide mononucleotide (NMN), before vitamin C treatment. Cell viability after vitamin C treatment was partially rescued by inhibiting PARP or supplementing with NMN (
Fig. 4D). Taken together, these results indicate that in
KRAS and
BRAF mutant cells vitamin C–induced endogenous ROS inhibits GAPDH by both posttranslational modifications and NAD
+ depletion, ultimately leading to an energetic crisis and cell death (
Fig. 4E).
High-dose vitamin C cancer therapy has a controversial history. Although some early clinical studies indicated that vitamin C had antitumor activity (
26,
27), others have shown little effect (
28,
29). Recent studies suggest that the contradictory clinical data may be explained, at least in part, by differences in administration route; the millimolar vitamin C plasma concentrations cytotoxic to cancer cells are only achievable via intravenous administration, not via oral administration (
30,
31). Given these findings, a growing number of phase I and phase II clinical trials are reevaluating intravenous infusion of vitamin C to treat various cancers (
12,
13,
32,
33). However, despite the previous studies demonstrating that high-dose vitamin C is cytotoxic to cancer cells in vitro (
34–
36) and that it delays tumor growth in xenograft models (
37,
38), the mechanism by which vitamin C kills cancer cells while sparing normal cells has been unclear. Our findings address this fundamental question by suggesting that the oxidized form of vitamin C, DHA, is the pharmaceutically active agent, and that the selective toxicity of vitamin C to tumor cells stems from high GLUT1 expression combined with
KRAS or
BRAF oncogene-induced glycolytic addiction. Although it is unclear whether the results we have observed in our cell culture and mouse studies will translate to human tumors, our findings on the mechanism of action of vitamin C may warrant future investigation in clinical trials.
Acknowledgments
We thank B. Vogelstein and K. W. Kinzler for helpful suggestions; the Cantley lab members, B. Hopkins, F. Karreth, C. Lyssiotis, and G. DeNicola for comments on the manuscript; and M. Yuan, S. Breitkopf, J. Wong, and O. Mashadova for technical assistance. We apologize for publications not cited because of space limitations. L.C.C. owns equity in, receives compensation from, and serves on the board of directors and scientific advisory board of Agios Pharmaceuticals. Agios Pharmaceuticals is identifying metabolic pathways of cancer cells and developing drugs to inhibit such enzymes in order to disrupt tumor cell growth and survival. Supported by the Damon Runyon Cancer Research Foundation (J.Y. and I.I.C.C.), KL2 Career Development Awards (J.Y.), the U.S. Department of Defense (J.P.), National Cancer Institute grants P01 CA120964-07 and P01 CA117969-09 (L.C.C.), and NIH grant P01 CA120964 (J.M.A.). The authors declare no competing financial interests.