Glucose Deprivation Contributes to the Development of KRAS Pathway Mutations in Tumor Cells

See allHide authors and affiliations

Science  18 Sep 2009:
Vol. 325, Issue 5947, pp. 1555-1559
DOI: 10.1126/science.1174229


Tumor progression is driven by genetic mutations, but little is known about the environmental conditions that select for these mutations. Studying the transcriptomes of paired colorectal cancer cell lines that differed only in the mutational status of their KRAS or BRAF genes, we found that GLUT1, encoding glucose transporter-1, was one of three genes consistently up-regulated in cells with KRAS or BRAF mutations. The mutant cells exhibited enhanced glucose uptake and glycolysis and survived in low-glucose conditions, phenotypes that all required GLUT1 expression. In contrast, when cells with wild-type KRAS alleles were subjected to a low-glucose environment, very few cells survived. Most surviving cells expressed high levels of GLUT1, and 4% of these survivors had acquired KRAS mutations not present in their parents. The glycolysis inhibitor 3-bromopyruvate preferentially suppressed the growth of cells with KRAS or BRAF mutations. Together, these data suggest that glucose deprivation can drive the acquisition of KRAS pathway mutations in human tumors.

Mutations in oncogenes and tumor suppressor genes endow cancer cells with the ability to outgrow their neighboring cells in situ (1). Though numerous studies have identified the downstream effects of such mutations and their biochemical mediators, relatively little is known about the microenvironmental conditions that provide the selective advantage that allows cells with such mutations to clonally expand. Mutations in KRAS commonly occur in colorectal, pancreatic, and some forms of lung cancer, whereas BRAF mutations occur commonly in melanomas, as well as in colorectal tumors without KRAS mutations (24). BRAF and KRAS mutations are mutually exclusive; that is, they do not occur in the same tumor, suggesting a common origin and effect. KRAS binds to and activates BRAF, thereby activating mitogen-activated protein kinase (MAPK) signaling pathways (5, 6). Despite advances in the molecular delineation of the RAS/RAF pathway, the specific environmental pressures that drive KRAS and BRAF mutations and how KRAS and BRAF mutations alleviate these pressures are unknown.

To explore this issue, we developed isogenic colorectal cancer (CRC) cell lines in which the endogenous wild-type (WT) or mutant alleles had been inactivated through targeted homologous recombination (figs. S1 and S2 and table S1) (7). We chose to use targeted homologous recombination instead of the more commonly used overexpression or small interfering RNA–dependent systems because only the former permits examination of cells expressing normal or mutant proteins at physiological, normally regulated levels (8). For the investigation of BRAF mutations, we used RKO and VACO432, CRC lines with valine-to-glutamate mutations at codon 600 (V600E) of BRAF. This is the most common BRAF mutation in human tumors, accounting for more than 90% of BRAF mutations (3). To investigate KRAS in an analogous manner, we used HCT116 and DLD1, CRC lines with glycine-to-aspartate mutations at codon 13 (G13D). This mutation is one of the most common in CRC, accounting for ~20% of KRAS mutations (2). These paired lines essentially differ in only one base pair: the base that is mutated or wild type in KRAS or BRAF. At least two independent clones of each of the derivatives of each of the four parental cell lines were developed (table S1). In all cases, independent clones with the same genotype behaved similarly in the assays described below.

On the basis of the mutual exclusivity of KRAS and BRAF mutations and knowledge of the KRAS pathway described above, we hypothesized that a common set of transcripts would be deregulated in response to mutations in either gene. We performed expression analysis on clones of various genotypes with microarrays, as well as with massively parallel sequencing of serial analysis of gene expression (SAGE) tags. Only three genes were found to be more than twofold up-regulated in all four lines containing mutant KRAS or BRAF alleles compared with their isogenic counterparts containing WT alleles: GLUT1 (also known as SLC2A1), DUSP5, and DUSP6 (fig. S3). DUSP5 and DUSP6 are known feedback regulators of the MAPK signaling pathway, up-regulated when the pathway is active (9), and thus were unlikely to be positive effectors of KRAS and BRAF tumorigenesis. On the other hand, GLUT1 was intriguing, as it encodes a glucose transporter known to be overexpressed in many types of cancer, and its high expression in tumors has been associated with poor prognosis (10, 11). We confirmed the results of the microarray and SAGE expression analyses through quantitative polymerase chain reaction. GLUT1 transcript expression was always higher, ranging from 3- to 22-fold, in the clones with mutant KRAS or BRAF alleles compared to the isogenic clones with WT alleles (Fig. 1A and fig. S3). Accordingly, we found that the expression of the GLUT1 protein was markedly higher in cells with mutant KRAS or BRAF alleles (Fig. 1B). Targeted disruptions of both alleles of GLUT1 in RKO and DLD1 cells (fig. S4 and table S1) were used as negative controls to ensure the specificity of the antibodies to GLUT1 (Fig. 1B). As expected, the GLUT1 protein was found in the membrane fraction of cells, regardless of KRAS or BRAF mutational status. Of the 12 human glucose transporter homologs present in the human genome (10), only GLUT1 was up-regulated in the mutant KRAS- or BRAF-containing lines compared to those with WT alleles.

Fig. 1

Expression of GLUT1 in matched pairs of isogenic clones. (A) Expression levels of GLUT1 transcripts were determined by real-time PCR and normalized to those of β-actin. Each panel includes the parental line (Parent), which harbors both mutant and WT alleles of KRAS or BRAF, two independent clones with only mutant alleles (MUT1 and MUT2), and two independent clones with only WT alleles (WT1 and WT2). The data represent the mean and SD (error bars) of triplicate experiments. The differences between MUT and WT clones were statistically significant in all cases (P < 0.05, Student’s t test). (B) Expression of GLUT1 membrane-associated protein levels, as determined by immunoblotting. Na+,K+-ATPase, a membrane-associated protein, was used as a loading control.

To test the specificity of GLUT1 up-regulation, we evaluated the expression of this protein in cell lines in which the mutant or WT alleles of PIK3CA had been disrupted by targeted homologous recombination (12). PIK3CA has been implicated in the RAS/RAF pathways, as well as in metabolic regulation, and is commonly mutated in cancers (12, 13). Unlike KRAS and BRAF, the PIK3CA genotype did not have a clear effect on GLUT1 protein expression (Fig. 1B). We also tested lines with targeted disruptions of both alleles of HIF1A (14) (table S1). Though HIF1A has been shown to regulate transcription of GLUT1 in hypoxic conditions (1517), GLUT1 expression was found to be largely independent of HIF1A status when cells were grown in normal oxygen concentrations (Fig. 1B).

We hypothesized that the up-regulation of GLUT1 would result in increased glucose uptake in the clones with mutant KRAS or BRAF alleles. To test this idea, we incubated cells with 2-deoxy-D-[3H] glucose (2-DG), a nonhydrolyzable glucose analog, and measured its uptake. We found that the up-regulation of GLUT1 was accompanied by a significant increase in glucose uptake in all cells with mutant KRAS or BRAF alleles compared to the isogenic cells with WT alleles (Fig. 2A). Disruption of GLUT1 substantially inhibited glucose uptake, demonstrating that GLUT1 was the major glucose transporter in these cancer cells (Fig. 2A).

Fig. 2

Glucose uptake and lactate production in cells with KRAS or BRAF mutations. (A) Glucose uptake, as determined with the use of [3H] 2-deoxyglucose, was normalized to the amount of total protein. Differences between MUT and WT clones were statistically significant (P < 0.01, Student’s t test). (B) Lactate production was normalized to cell number. The differences between MUT and WT clones were statistically significant in all cases (P < 0.03, Student’s t test). The data represent the mean and SD (error bars) of triplicate experiments.

We next determined whether the increased glucose transport was associated with increased lactate production. Lactate production was significantly increased in cells with mutant KRAS or BRAF alleles, indicating an increased rate of glycolysis and consistent with higher glucose uptake (Fig. 2B). Lactate production was very low in cells without GLUT1 genes, as would be expected if GLUT1 functions as the major glucose transporter in these cells (Fig. 2B). On the other hand, oxygen consumption in cells with mutant KRAS or BRAF alleles was similar to that in cells with WT alleles of these genes, suggesting that mitochondrial function and oxidative respiration were not affected by KRAS or BRAF mutation (fig. S5). Accordingly, there were no consistent differences in cellular adenosine triphosphate (ATP) concentrations or ATP/ADP (adenosine diphosphate) ratios in cells with mutant KRAS or BRAF alleles compared with their WT counterparts (figs. S6 and S7).

These results suggested that the increase in glucose uptake and glycolysis might provide a growth advantage to cells with KRAS or BRAF mutations in low-glucose environments. When grown in standard, commercially available media (25 mM glucose), all cell lines, including those without the GLUT1 gene, grew reasonably well (fig. S8) and formed colonies when plated at low density. However, when placed in media containing low-glucose concentrations (0.5 mM), only cell lines with KRAS or BRAF mutant alleles survived (Fig. 3A). This growth was dependent on GLUT1, as cells in which the GLUT1 gene was inactivated by targeted homologous recombination lost their ability to form colonies in low-glucose environments, even though they contained mutant KRAS or BRAF genes (Fig. 3A). In contrast, such growth was independent of HIF1A, as cells with mutant KRAS or BRAF alleles survived in low-glucose conditions when the HIF1A gene was inactivated by targeted homologous recombination (Fig. 3A).

Fig. 3

KRAS and BRAF mutations confer a selective growth advantage in hypoglycemic conditions. (A) Cells were subjected to a low-glucose environment (0.5 mM) for 2 (RKO and VACO432) or 4 (HCT116 and DLD1) days, then dissociated and plated in media containing standard concentrations of glucose (25 mM). Colony counts were normalized to those obtained in cells subjected to the same experimental procedure, with the exception that standard glucose levels were substituted for low-glucose levels. See (7) for details. The differences between MUT and WT clones were statistically significant in all cases (P < 0.004, Student’s t test). Error bars indicate SD. (B) MUT and WT clones were mixed at the indicated ratios and grown in media with 0.5 mM glucose for 2 (RKO) or 5 (DLD1) days. The media was replaced with one containing 25 mM glucose, and the cells were incubated for another 10 to 16 days. RNA was purified from the cells that survived, and the KRAS or BRAF genes were PCR-amplified and sequenced. G and A nucleotides at the underlined positions in the sequencing chromatograms represent WT and mutant alleles of KRAS, respectively, in DLD1 cells. T and A nucleotides represent WT and mutant alleles of BRAF, respectively, in RKO cells. (C) DLD1 cells in which the mutant KRAS allele had been deleted by targeted recombination [KRAS (–/+)] were plated in low-glucose conditions (0.5 mM). After 25 to 30 days, the few clones that survived were grown in standard glucose (25 mM) conditions and assessed for GLUT1 expression and the sequence of the KRAS gene. Clones that harbored mutant alleles of KRAS (G12D, G13D, or G13C) are indicated, as are clones in which KRAS remained as wild type. As controls, the same cells [KRAS (–/+)] were plated at limiting dilution in media containing 25 mM glucose and individual clones assessed for GLUT1 expression (Control clones). The parental cells used for these experiments (DLD1, WT) are also included, as were their isogenic counterparts in which the WT rather than the mutant allele was disrupted by homologous recombination (DLD1, MUT). All clones had been growing in media containing 25 mM glucose for at least 20 days when harvested for the assessment of GLUT1 expression by immunoblotting. Na+,K+-ATPase was used as a loading control. A diagram of the selection scheme is provided in fig. S10, and detailed methods are provided in (7).

We then determined whether clones with mutant KRAS or BRAF genes could selectively outgrow cells without these mutations. For this purpose, cells with mutant KRAS or BRAF alleles were mixed with an excess of cells containing WT KRAS or BRAF alleles, respectively, and were incubated in either low (0.5 mM) or standard (25 mM) concentrations of glucose. Cells with mutant KRAS or BRAF alleles preferentially survived in low-glucose conditions and overtook the population within two weeks after the medium was changed to one containing 25 mM glucose. In contrast, the cells with WT alleles remained predominant when they were not exposed to low-glucose conditions (Fig. 3B).

To mimic situations that might occur in vivo, we subjected cells with only WT alleles (obtained by disrupting the mutant KRAS or BRAF alleles) (figs. S1 and S2) to a low-glucose environment in vitro and isolated the few colonies that survived (fig. S10). We reasoned that in the ~35 generations that had elapsed between targeted disruption and this experiment, a small fraction of cells would have spontaneously acquired mutations in genes that could potentially permit them to survive in medium containing low-glucose concentrations. The fact that the two cell lines used for this experiment were both mismatch-repair deficient should have facilitated the development of such de novo mutations (18). We found that the fraction of DLD1 KRAS (–/+) or RKO BRAF (–/–/+) cells that could form colonies in low-glucose conditions was ~0.05%. Once formed, the colonies were grown in a medium containing standard concentrations of glucose (25 mM). We found that more than 75% of the clones derived from either cell line after selection in low-glucose, stably expressed high levels of GLUT1 protein, even when subsequently grown in standard medium (25 mM glucose) (Fig. 3C and fig. S11). Thus, the selection for growth in low-glucose conditions resulted in a permanent up-regulation of GLUT1 expression in the majority of clones that survived, and this up-regulation persisted after normoglycemia was reinstituted, indicating a heritable change. Control clones derived analogously but, with 25 mM glucose substituted for 0.5 mM glucose during the selection period, did not show elevated GLUT1 expression (Fig. 3C and fig. S11). When the clones derived from DLD1 KRAS (–/+) cells were assessed for mutations, 4.4% of the clones arising under hypoglycemic conditions had mutations in KRAS (73.5% of these had G12D, 25.2% had G13D, 1.3% had G13C, and 0% had BRAF V600E or other mutations in KRAS at codon 12 or 13). No KRAS or BRAF mutations were identified in 2000 DLD1 KRAS (–/+) clones generated in the presence of standard concentrations of glucose (P < 0.000001, χ2). In the clones derived from RKO BRAF(–/–/+) cells, 0.8% of the clones surviving low-glucose exposure had a G12D KRAS mutation, whereas none of 2000 clones grown in the presence of standard glucose concentrations had such mutations (P < 0.01, χ2).

We next attempted to exploit this phenotype to specifically target cancer cells with KRAS or BRAF mutations. We reasoned that cells with KRAS or BRAF mutations had stably reprogrammed their metabolic pathways and might be dependent on glycolysis for growth. Accordingly, an agent such as 3-bromopyruvate (3-BrPA), which inhibits glucose metabolism through inhibition of hexokinase (19), might be selectively toxic to cells with KRAS or BRAF mutations. When we tested this hypothesis using the paired isogenic cell lines, we found that 3-BrPA was highly toxic to HCT116, DLD1, VACO432, and RKO cells with KRAS or BRAF mutations but was much less toxic to the matched cell lines lacking KRAS or BRAF mutant alleles (Fig. 4A).

Fig. 4

The glycolysis inhibitor 3-BrPA is selectively toxic to cells with mutant KRAS or BRAF alleles. (A) Colony formation was assessed after 3-BrPA treatment (110 μM) for 3 days. Colony counts were normalized to those obtained from cells subjected to the same procedure without exposure to 3-BrPA. The differences between MUT and WT clones were statistically significant in all cases (P < 0.008, Student’s t test). Error bars indicate SD. (B) Mice with subcutaneous tumors established from HCT116 (KRAS: G13D/+) or VACO432 (BRAF: V600E/+) cells were injected intraperitoneally with 3-BrPA or phosphate buffered saline (PBS) daily for 2 weeks. n represents the number of mice used in each group. Data points and error bars represent the means and SD for each group of mice. Asterisks denote times when there were significant differences between the tumor sizes in the PBS versus 3-BrPA groups (P < 0.05, Student’s t test).

We next explored whether this approach might be applicable to experimental tumors in mice. As a prelude, we found that cells with disrupted mutant KRAS (20) or BRAF alleles grew poorly as xenografts in nude mice compared with their isogenic counterparts with mutant alleles (fig. S12). DLD1 and RKO cells in which the GLUT1 gene was disrupted also grew poorly in nude mice, even though these cells contained mutant KRAS and BRAF alleles, respectively (fig. S12). These results suggest that the microenvironment in xenografts mimics the low-glucose environment in vitro and provides a reasonable system in which to test the effects of glycolytic inhibitors. 3-BrPA significantly inhibited the growth of established xenografts derived from HCT116 and VACO432 cells (Fig. 4B). These results provide proof of principle that glycolytic inhibitors can retard tumor growth at doses that are nontoxic to normal tissues in vivo.

Our results led us to investigate glucose metabolism in a completely unbiased way, complementing previous work by other investigators. A role for metabolic abnormalities in cancer has become increasingly recognized (21, 22). These metabolic abnormalities often appear to involve abnormal glycolysis, as first demonstrated decades ago by Warburg (23). Insightful hypotheses about the manifold ways in which such metabolic abnormalities can promote tumor progression have been described (2426). It has also been demonstrated that transformation of rodent fibroblasts by several oncogenes, including HRAS, can up-regulate glucose transporter expression (16, 2729). However, because transformation by overexpressed oncogenes affects the expression of hundreds of genes and dramatically alters the phenotype of rodent fibroblasts, the relation between increased glucose transporter expression and tumorigenesis was not clear. In human tumor cells, no obvious relation between GLUT1 and RAS mutations has been identified (30, 31). Moreover, in many previous experimental studies in rodent cells, the increased GLUT1 expression was ascribed to induction of HIF1A and linked to hypoxia (15, 16, 32, 33). Our results show that the increased GLUT1 transcription was unrelated to HIF1A, because genetic disruption of the HIF1A gene did not affect the expression of GLUT1, nor did it affect survival under hypoglycemic conditions. Cells without mutant KRAS or BRAF alleles were remarkably sensitive to hypoglycemia, but not to hypoxia (Fig. 3 and fig. S9). Furthermore, the changes in GLUT1 expression and resultant metabolic changes in human CRC cells were stable phenotypes rather than transient responses to low-glucose levels, as they persisted under normoglycemic conditions. This stability is consistent with them being the consequence of specific genetic mutations, such as those in KRAS or BRAF. In aggregate, our results suggest that low-glucose environments are a driving force underlying the development of KRAS and BRAF mutations during tumorigenesis.

F-18-Fluoro-deoxyglucose (FDG)–positron emission tomography (PET) scans are routinely used to image cancers in the clinic. Positive signals in cancers are the result of increased glucose transporter expression or glucose uptake (34). Our data showed that in four different human cancer cell lines, an increase in GLUT1 expression and glucose uptake was critically dependent on KRAS or BRAF mutations. It is interesting that abnormal FDG-PET signals can be observed in progressing pre-malignant colorectal neoplasms (adenomas) congruent with the time during tumorigenesis in which KRAS or BRAF mutations appear (35, 36).

Our results raise a variety of questions. One concerns the relation between hypoxia and hypoglycemia. Though both of these deficiencies are likely to be encountered in tumor microenvironments, it is possible that each condition sets the stage for the selection of particular genetic abnormalities (24). For example, hypoglycemic conditions favor the selection of cells with KRAS or BRAF mutations, whereas hypoxic conditions may favor the selection of cells with PIK3CA, CMYC, or TP53 mutations (21, 37). Another question addresses why 90% of CRCs exhibit high FDG-PET signals and GLUT1 expression (35, 38, 39), whereas KRAS or BRAF mutations are only observed in ~50% of such cancers (4). One possible explanation is that other genetic alterations that affect the same pathway can substitute for KRAS and BRAF mutations in up-regulating GLUT1. This idea is consistent with recent data indicating that the same pathway can be mutationally activated through disparate mutations in numerous genes (1, 40, 41). It is also consistent with our in vitro selection experiments. Though the majority of clones that survived hypoglycemia up-regulated GLUT1, only a minority of these clones had acquired KRAS or BRAF mutations.

Supporting Online Material

Materials and Methods

Figs. S1 to S12

Table S1


  • * Present address: Novartis Institute for Biomedical Research, Cambridge, MA 02139, USA.

  • Present address: Cardiovascular Medicine, Brigham and Women’s Hospital, Boston, MA 02115, USA.

  • Present address: Sanofi-Aventis, 13 Quai Jules Guesde, 94400 Vitry-sur-Seine, France.

References and Notes

  1. See supporting material on Science Online.
  2. We thank Y. He for helpful discussions, W. Yu for help with the microarray experiments, and E. Watson for expert technical assistance. This work was supported by the Virginia and D. K. Ludwig Fund for Cancer Research and NIH grants CA43460 and CA62924. Under agreements between the Johns Hopkins University, Genzyme Molecular Oncology, Novartis, Wyeth, Amgen, Glaxo-Smith-Kline, and Horizon, J.Y., V.E.V., H.R., R.P., C.L., K.W.K., B.V., and N.P. are entitled to a share of the royalties and licensing fees received by the university on cell lines described in this paper, some of which are the subject of patent applications. The Johns Hopkins University and V.E.V., K.W.K., and B.V. also own stock in Genzyme, which is subject to certain restrictions under university policy. The terms of these arrangements are being managed by the university in accordance with its conflict of interest policies.
View Abstract

Navigate This Article