Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Cancer Res. Author manuscript; available in PMC 2012 August 1.
Published in final edited form as:
PMCID: PMC3148325

18F-fluorodeoxy-glucose positron emission tomography (18FDG-PET) marks MYC-overexpressing human basal-like breast cancers


In contrast to normal cells, cancer cells avidly take up glucose and metabolize it to lactate even when oxygen is abundant, a phenomenon referred to as the Warburg effect. This fundamental alteration in glucose metabolism in cancer cells enables their specific detection by Positron Emission Tomography (PET) following intravenous injection of the glucose analogue 18F-fluorodeoxy-glucose (18FDG). However, this useful imaging technique is limited by the fact that not all cancers avidly take up FDG. To identify molecular determinants of 18FDG-retention, we interogated the transcriptomes of human cancer cell lines and primary tumors for metabolic pathways associated with 18FDG radiotracer uptake. From 95 metabolic pathways that were interrogated, the glycolysis and several glycolysis-related pathways (pentose-phosphate, carbon fixation, aminoacyl-tRNA biosynthesis, one-carbon-pool by folate) showed the greatest transcriptional enrichment. This “FDG signature” predicted FDG-uptake in breast cancer cell lines and overlapped with established gene expression signatures for the “basal-like” breast cancer subtype and MYC-induced tumorigenesis in mice. Human breast cancers with nuclear MYC staining and high RNA expression of MYC target genes showed high 18FDG-PET uptake (p < 0.005). Presence of the FDG signature was similarly associated with MYC gene copy gain, increased MYC transcript levels, and elevated expression of metabolic MYC target genes in a human breast cancer genomic dataset. Together, our findings link clinical observations of glucose uptake with a pathologic and molecular subtype of human breast cancer. Further, they suggest related approaches to derive molecular determinants of radiotracer retention for other PET-imaging probes.

Keywords: FDG-PET, breast cancer, MYC, basal-like, metabolism, imaging


Glycolysis and oxidative phosphorylation represent the main metabolic pathways that fuel energy dependent processes in cells by generating adenosine triphosphate (ATP). Compared to normal differentiated cells, cancer cells show increased glycolytic rates and high lactate production in-vitro even when oxygen levels are sufficient to support oxidative phosphorylation, a process called aerobic glycolysis or “Warburg effect” (1). Why cancer cells favor an energetically less efficient path of glucose utilization and how they avoid oxidative phosphorylation has remained unclear for decades. A number of recent observations have begun to shed light on the latter question. Knockdown of LDH-A in neu-initiated mammary epithelial tumor cell lines stimulated mitochondrial respiration, indicating that the glycolytic phenotype of cancer cells is not necessarily due to intrinsic mitochondrial defects in oxidative phosphorylation. Furthermore, activation of tyrosine kinase signaling, frequently caused by mutations in growth factor signaling pathways, has been found to inhibit the entry of pyruvate into the mitochondrial TCA cycle through hypoxia-inducible factor 1 (HIF1) mediated induction of pyruvate dehydrogenase kinase 1 (PDK1) and posttranslational modification of the M2 splice isoform of pyruvate kinase (2).

The altered glucose metabolism of tumor cells can be observed in cancer patients by positron emission tomography (PET) following intravenous injection of the glucose analogue 18F-fluorodeoxy-glucose (18FDG)(3). Compared to normal surrounding tissue, tumors often show an increase in the FDG-PET signal which reflects their high rate of radiotracer uptake through membrane glucose transporters, phosphorylation by one of several hexokinase enzymes, and the resultant intracellular trapping of the radiotracer which is not further metabolized in the cell. Not all cancers, however, avidly take up FDG. Human breast cancers, for example, show up to 20-fold differences in their FDG-PET signal. This heterogeneity has been attributed, with substantial variation between studies, to differences in histopathologic subtypes, tumor size, microvasculature, tumor cell proliferation, hormone receptor status, and expression levels of hexokinase or glucose transporters (4).

Despite the widespread use of FDG-PET imaging in oncology, the relationship between FDG uptake of primary human tumors and their metabolic and genetic alterations is largely unknown. To address this question, we performed genome wide transcriptomal analysis of cell lines and primary human tumors after determining their FDG uptake. We found that 18FDG retention is associated with coordinated transcriptional upregulation of multiple metabolic pathways, including the core glycolysis, pentose-phosphate, and carbon fixation pathways. This “FDG signature” predicted radiotracer uptake in breast cancer cell lines and was closely linked to the “basal-like” intrinsic breast cancer subtype and activation of the MYC oncogene.


Cell lines

LAPC4 and HCT116 PTEN knockout cells were derived and provided by Dr. Sawyers and Dr. Waldmann, respectively. All other cell lines were obtained from ATCC and the NCI.


Eighteen breast cancer patients who presented for operative management of primary breast carcinoma were imaged with FDG-PET within 4 weeks prior to surgery; excluding patients with secondary breast cancers and recurrent disease. None of the patients received systemic therapy or radiation prior to imaging. All breast tumor samples were collected surgically. Our study included one patient with anaplastic astrocytoma. This study was approved by the IRB of Memorial Sloan-Kettering Cancer Center and all participating patients signed informed consent.

Gene expression analysis

Gene expression signature-based predictions of FDG uptake were made using weighted gene voting (WGV)(5). The “rank-rank hypergeometric overlap” (RRHO) algorithm (6) was used to examine the statistical significance of similarity between our FDG signature and other gene expression signatures. Details of the bioinformatic approaches are described under Supplementary Methods.


Upregulation of glycolysis and glycolysis branch pathways in 18FDG-avid cancer cells

Our strategy to identify determinants of FDG-retention consisted of measurements of 18FDG retention in cancer cell lines and primary human tumors, RNA expression profiling of these samples, comparison of 18FDG “high” versus “low” samples using gene set enrichment analysis (GSEA) (7), and mining of genomic datasets for this “FDG signature” (Fig.1A).

Figure 1
Deriving an “FDG-uptake” metabolic gene expression signature

We first measured 18FDG radiotracer uptake of sixteen cancer cell lines representing prostate cancer (CaP), glioblastoma (GBM), and melanoma (MEL). We observed up to 5-fold differences in FDG-uptake between cell lines of the same histologic type (Fig.1B). The clinical sample set included tumors from eighteen breast cancer patients. 18FDG-tumor uptake was quantified as standardized uptake values (SUVs) and showed the expected wide dynamic range (0 to 22.1)(Fig.1C, top). Breast cancers with the highest 18FDG-PET SUVs frequently lacked expression of the estrogen receptor (ER) and the progesterone receptor (PR), but hormone receptor negative tumors were also represented amongst the tumors with the lowest FDG-uptake (Table 1). Our clinical sample set also included a patient with anaplastic astrocytoma whose tumor showed areas of distinct FDG-uptake. Both an FDG-high and FDG-low region were amenable to stereotactic biopsy (Fig.1C, bottom) and represented viable tumor.

Table 1
FDG-PET tumor uptake in 18 patients with locally advanced breast cancer

For each tumor type represented in our panel, we selected samples with particularly high and low FDG-uptake for gene set enrichment analysis (GSEA) (7) using 95 metabolic pathways annotated by the Kyoto Encyclopedia of Genes and Genomes (KEGG)(8). We hypothesized that a comparison of samples at the extremes of the FDG-uptake spectrum would facilitate the identification of FDG-uptake associated metabolic pathways. The following FDG-high versus FDG-low sample sets were used: i.) LNCaP vs. LAPC4 cells (prostate cancer), ii.) U87 vs. SF268 (glioblastoma), and iii.) SKMel28 vs. CHL cells (melanoma)(marked with asterisk in Fig.1B), iv.) FDG-high vs. FDG-low region of the anaplastic astrocytoma, and v.) breast cancers with SUVs above 10 vs. breast cancers with SUVs below 5 (marked with asterisk in Table 1). We initially excluded lobular breast carcinomas, because they have been shown to take up less FDG than ductal carcinomas (9). We initially also excluded large breast carcinomas (>5 cm) and breast carcinomas with multifocal FDG-uptake because our protocol did not include tissue autoradiography to direct the molecular tissue analysis to areas of distinct radiotracer retention. There was no significant difference in patient age, tumor volume, and lymph node involvement between the group of FDG-high and FDG low breast cancers.

For our combined GSEA analysis using the 5 FDG-high vs. FDG-low sample sets and the 95 KEGG metabolic pathways, we first defined a rank-based gene expression signature for each histology type (breast and astrocytoma tumors; and prostate, glioblastoma and melanoma cell lines). Then the average rank for each gene was determined to define an average-rank signature that was interrogated using GSEA. The glycolysis/gluconeogenesis pathway scored as the most highly enriched metabolic pathway. The related carbon fixation and pentose-phosphate pathways also showed significant enrichment in the FDG-high samples, as well as the pathways for aminoacyl-tRNA biosynthesis and one-carbon-pool by folate (Fig.1D). Results were consistent between this average-rank approach and enrichment analysis of the individual signatures (Table S1).

To exclude the possibility that our selection of breast cancer samples had introduced experimental bias, we repeated the GSEA analysis with all 18 primary breast cancer samples using a continuum SUV correlation-based ranking approach. Enrichment of the glycolysis, pentose phosphate, and carbon fixation remained significantly associated with the 18FDG-PET signal (Table S1, Column “Breast Cancer, SUV Continuum).

Gene-expression based “FDG signature” predicts FDG-uptake in-vitro

We next examined whether pathway enrichment was driven by modest differences in the levels of many pathway members or more dramatic effects on only 1–2 key enzymes. As illustrated in Fig. 2A for the glycolysis core pathway in the breast cancer samples, pathway enrichment was due to moderate (less than 2-fold), but highly concordant differences in the transcript levels of many pathway members. Similarly modest differences in transcript levels of functionally related genes have been shown to regulate metabolic flow in other biological and disease systems (1012).

Figure 2
FDG signature score predicts in-vitro FDG-uptake

We also determined the contribution of individual enzymes to overall pathway enrichment based on their rank in each separate GSEA analysis (Fig.2B and Fig. S1). While there were differences between tissue types (Fig. S1B), enzymes that direct metabolic flow towards glycolysis (e.g., phosphofructokinase, hexokinase, pyruvate kinase) showed the most consistent enrichment within the glycolysis/ gluconeogenesis KEGG pathway (Fig.2B and Fig. S1B–C). The mean rank in the combined analysis for members of the top three enriched metabolic pathways (glycolysis, pentose-phosphate, carbon fixation) is shown in Figure 2B. The “carbon fixation” KEGG pathway is functionally complete only in plants and scored in our analysis due to the overlap of its enzymes with the glycolysis/gluconeogenesis and pentose-phosphate pathways.

We next tested whether transcript levels of the most highly ranked members of the top three enriched metabolic pathways, could serve as an “FDG signature” and predict FDG-uptake. We explored this question in a panel of seven human breast cancer cell lines which were not included in our initial FDG-uptake studies. We selected breast cancer cell lines for the validation of our FDG signature because the majority of human tumor samples used for the derivation of this signature were breast carcinomas. Predictions were made using the weighted gene voting approach (13) and using the primary breast tumors as the FDG signature training set (Table S2, Fig. S1C). FDG-uptake assays were performed blinded to our computational analysis and showed a wide range of FDG-uptake, as has been reported for breast cancer cell lines (14). We found a strong correlation between measured FDG-uptake and predicted FDG-uptake (r = 0.92, permutation p value = 0.03)(Fig.2C).

We next tested weighted gene voting predictions using individual genes with the greatest differential expression between FDG high and FDG low breast cancers (top 100 – top 2000 genes). Predictions of FDG uptake using the top differentially expressed individual genes showed less correlation with the measured FDG-uptake (Fig.2D), suggesting that our metabolism-oriented bioinformatic approach uncovered a shared metabolic state in FDG high samples that would be more difficult to detect using “gene-centric” data analysis approaches.

FDG signature is associated with the “basal-like” breast cancer subtype

The derivation of a gene expression-based FDG signature enabled us to search published genomic datasets for the presence of this signature with the goal to identify tumor types, genetic lesions, or signaling pathways that might be associated with FDG uptake. We focused this analysis on human breast cancer because of the wealth of validated RNA expression signatures in this disease (15). We first developed a method for quantitating the degree of overlap between two signatures defined by differential gene expression (6)(Fig. S2). We then applied this “rank-rank hypergeometric overlap” (RRHO) method to a dataset of 295 primary human breast cancers (16) to determine the overlap between the FDG signature and the main intrinsic breast cancer subtypes, i.e. “basal-like”, “luminal”, “HER2/ ErbB2”, and “normal-like”. We found significant overlap with the signature for the “basal-like” subtype, an inverse relationship with the signature for the luminal and normal subtypes (all with p-values < 10−4), and no overlap with the ErbB2 subtype (Fig.3A).

Figure 3
FDG signature overlaps with “basal-like” breast cancer subtype

Our results indicated that the expression of genes in basal-like breast cancers resembled FDG-high cancer cells more closely than any other breast cancer subtype. If true, a direct GSEA comparison between “basal-like” and “other” breast cancers should identify similar metabolic pathways as our previous GSEA comparison between “FDG-PET high versus FDG-PET low” breast cancers. We tested this hypothesis in a second dataset of 286 primary human breast cancers (17). Only two metabolic pathways were significantly enriched in this analysis (glycolysis/gluconeogenesis and aminoacyl-tRNA biosynthesis)(Table S3), both of which had demonstrated strong enrichment in our prior GSEA comparison of FDG high versus FDG low human primary breast cancers (Table S1).

We further tested whether our “FDG signature score”, which had predicted FDG-uptake in breast cancer cell lines (Fig.2C), would preferentially identify basal-like breast cancers within a sample set representing all breast cancer subtypes. To test this hypothesis, we used metabolic pathway-based weighted gene voting to predict the FDG-uptake of tumors in a third independent dataset of 80 locally advanced primary human breast cancers (18). Consistent with our prior analysis, 14/18 (77.8 %) tumors with the highest FDG-signature score were of the “basal-like” subtype, compared to 0/18 (0 %) tumors with negative FDG signature score (multivariate hypergeometric p-value, 10−8). A Kolmogorov-Smirnov sliding threshold-based analysis using all 80 tumors also yielded statistical significance with a permutation p-value of 10−7 (Fig.3B)(Table S4).

Basal-like breast cancer has been shown to harbor a greater number of low level gene copy number alterations than other breast cancer subtypes (19) and breast carcinomas with highest FDG-signature score harbored a significantly greater number of gene copy number alterations than tumors lacking the FDG signature (Fig.3C, top). Based on these results, we performed array-cGH profiling of breast cancer samples for which we had residual frozen tissue (11 of the original 18). We found significantly more gene copy number alterations in the group of FDG-high tumors (SUV>10) compared to the group of FDG low tumors (SUV<5)(Fig. 3C, bottom), providing further evidence that FDG-avid breast carcinomas exhibit genetic properties of the “basal-like” breast cancer subtype.

Activation of MYC in FDG-avid breast cancers

Mutant alleles of ras and activation of the phosphatidylinositol 3-kinase (PI3-K) pathway have been shown to increase glucose uptake in experimental models (20). None of the breast tumors in our collection harbored an activating mutation in a ras-family member (data not shown). We found activating mutations in the catalytic subunit of PI(3)K in three breast carcinomas (BT07: C420R; BT09: H1047R; BT04: H1047R), all three estrogen receptor positive tumors, consistent with the reported association between PIK3CA mutations and ER-positivity in breast cancer (21). We next mined our breast carcinomas for transcriptional evidence of increased PI(3)K pathway activity, based on similarity with signatures derived from an Akt-driven murine tumor model (22) and human breast cancer samples lacking expression of the PTEN tumor suppressor (23). As with our FDG uptake signature (Fig. 1D), these PI(3)K-activation signatures showed enrichment of the glycolysis, pentose phosphate and carbon fixation pathways (Fig. S3A), and this enrichment was driven by similar glycolysis promoting enzymes (Fig. S3B).

We next examined protein levels of PTEN by immunohistochemistry (Fig. S3C) as diminished protein levels of this tumor suppressor have been reported for 15–25 % of all breast carcinomas and more commonly in basal-like breast cancer (24). Compared with adjacent non-neoplastic cells, we observed reduced PTEN staining of tumor cells in 6/16 (37.5 %) of tumors; five of the PTEN-deficient breast carcinomas were in the group of tumors with highest FDG-PET uptake. We next examined the effects of PTEN inactivation on cellular FDG-uptake in cancer cell lines. In HCT116 colon cancer cells (25), which harbor mutations in PIK3CA and KRAS, PTEN knockout raised the FDG uptake by about 50% (p<0.05) (Fig. S3D). PTEN knockdown in three other cancer cell lines (A431, HCC827, SKBR3), on the other hand, only raised FDG-uptake in one of the lines (SKBR3) and this increase was not statistically significant (Fig. S3E). These results suggest that the effects of PTEN on glucose metabolism are cell context-specific.

To identify additional signaling pathways that are associated with FDG-PET uptake in breast cancer, we searched a Molecular Signatures Database (MSigDB) for our FDG signature. This database is comprised of 1822 gene sets representing canonical signaling pathways, cellular processes, chemical and genetic perturbations, and human disease states (7). 110 of the 1822 gene sets, extracted directly from MSigDB without modifying their contents, were positively enriched in the transcriptome of FDG-high samples (Table S5). The top gene sets included gene sets related to poor prognosis (rank 6, 9) and high tumor grade (rank 25) in breast carcinoma. The top gene sets also included multiple gene sets both directly and indirectly related to the transcription factor MYC (Fig. 4A)(Table S5). The direct MYC group was comprised of gene signatures that are upregulated in transgenic mouse models of c-myc induced cancer (rank 12, 28, 49) (2628); MYC-related gene signatures included the serum fibroblast response/ wound healing signature (rank 2, 17, 39) linked to MYC activation in breast cancer (2931), and a signature linked to MYC activation in lymphoma (rank 38)(32). We found an inverse relationship (i.e., negative enrichment score) between our FDG signature and genes repressed by MYC (rank 1628, 1643, 1751) (26, 33, 34). The overall association between our FDG signature and gene sets directly linked to MYC was statistically significant (p=0.002)(Fig.4A).

Figure 4
MYC activation in FDG-high primary human breast cancers

We next stained all breast carcinomas for which we had remaining tissue (16/18) with an antibody against the MYC protein. 8/16 (50 %) tumors showed nuclear staining of tumor cells, similar to the reported frequency of MYC immunoreactivity (40–50 %) in human breast cancer (35, 36)(Fig.4B, left). MYC immunopositivity was significantly associated with high18FDG PET SUV-values (p=0.002)(Fig. 4B, right). Nuclear localization of MYC was associated with increased MYC transcriptional activity based on the overexpression of genes under direct transcriptional control of MYC in the MYC IHC-positive group, including the glutamine transporter SCL7A5 (37)(P<0.001), serine hydroxymethyl-transferase (SHMT)(38)(p<0.05), lactate dehydrogenase LDH-A (39)(p < 0.05), and transferrin receptor 1 (TFRC1)(40)(p<0.01)(Fig. 4C).

We also examined the relationship between the FDG-signature score, MYC levels and MYC target gene expression in the breast cancer dataset (18) used in our prior analysis (Fig. 3B). The frequency of c-myc copy gain (log2 ratio ≥ 0.4) in the top half of the FDG-signature score ranked tumors (19/40= 47.5%) significantly exceeded the frequency of c-myc copy gain in the bottom half (6/40 = 15 %) and in the entire cohort of patients (25/80= 31.3 %)(hypergeometric p-value=0.002)(Fig. 4D), demonstrating that the FDG signature significantly enriches for tumors harboring this molecular alteration. When we focused on the subgroup of tumors with the highest versus lowest FDG signature scores, we found elevated c-myc gene dosage in 10/18 (55.5 %) of breast cancers with high FDG signature score, but none (0/18) of the breast cancers with low FDG signature score (t-test p=0.0002). MYC transcript levels and the expression of MYC target genes were similarly statistically associated with the FDG signature score (Fig. 4D).

As genes induced by hypoxia also emerged from our pathway analysis (rank 8, 79)(Fig. 4A), we also stained all breast tumors with an antibody against hypoxia-induced factor (HIF-1). HIF1α is overexpressed in human cancers as a result of intratumoral hypoxia and genetic alterations in tumor cells (41). 12/16 (75 %) tumors showed intense nuclear staining for HIF1α (Fig. S4A), including all breast carcinomas with nuclear MYC-staining and highest FDG-PET signal. 7/12 HIF-1 positive tumors, but none of the HIF-1 negative tumors, also showed cytoplasmic staining with an antibody against phosphorylated proline-rich AKT1 substrate 1 (PRAS40)(Fig. S4B/C), a target of Akt and readout for PI(3)K pathway activity (42).


FDG-PET is widely used in the clinic for the detection of cancer. Despite a wealth of data linking glucose uptake to mutations in oncogenes and tumor suppressor genes in vitro (20), most studies of primary human tumors have focused on expression levels of hexokinase and glucose transporters. Our study sought to define the broader context of metabolic and genetic alterations in FDG-avid cancers. We show that FDG-avid tumors share a transcriptional program that involves not only members of the core glycolysis pathway, but also several glycolysis branch pathways critical for nucleotide and amino acid synthesis. These findings support the model that cancer cells favor aerobic glycolysis, despite the “penalty” of inefficient ATP production, because its metabolic intermediates can be used by the proliferating cancer cell for the replenishment of NADPH and the synthesis of highly needed macromolecules (43).

We identified overexpression of the transcription factor MYC as the molecular alteration most highly associated with FDG-uptake in human breast cancer. MYC is a plausible candidate to orchestrate the metabolic program of FDG-avid cancers. NMR studies have shown that MYC regulates the flux of glucose carbon not only through the core glycolysis pathway, but also through glycolysis branch pathways which were consistently upregulated in our analysis (i.e., pentose-phosphate pathway, amino-acid metabolism, and C1/folate metabolism)(44). Furthermore, MYC directly regulates RNA levels of several members of the glycolysis and glutamine pathway which showed increased transcript levels in FDG-avid (Fig. 4C) and FDG-signature positive tumors (Fig. 4D). These include PDK1 and LDH-A, which attenuate entry of pyruvate into the TCA cycle, the glutamine transporters SLC7A5 and SLC1A5, and – with some differences between published experimental models (45, 46) – glutaminase (GLS).

FDG-uptake in breast cancer did not correlate with Akt activation, a finding previously reported in short-term human breast cancer cultures (14). Our further examination of PI(3)K “pathway output” showed that the PI(3)K pathway is nonetheless activated in the majority of FDG-avid breast cancers, perhaps through alterations parallel or downstream of Akt (47). In peripheral nerve sheath tumors induced by monoallelic PTEN inactivation and mutant K-ras in mice, loss of the second PTEN allele coincides with a marked increase in tumor FDG-uptake (48), suggesting that the strength of PI(3)K pathway activation may be an important determinant of the glycolytic state. PI(3)K pathway activation may also cooperate with other oncogenic events, such as MYC, to induce a maximally glycolytic state. Several genes which were significantly upregulated in FDG-avid breast cancers in our study have previously been shown to be regulated by both MYC and HIF-1, including LDH-A, PDK1, and transferrin receptor I. Studies in a larger panel of primary human tumors are warranted to define the relationship between FDG-uptake, the PI(3)K-HIF1 axis, and other cancer genes.

The role of FDG-PET imaging in the management of human breast cancer remains to be defined (49). One of the challenges is the detection of small tumors (< 2.0 cm) as partial volume effects (PVEs) can result in underestimation of true radiotracer retention (50). This may have affected our estimation of the SUVs and should be addressed in future studies using different PVE correction schemes (51). Our study connects the clinical observation of altered glucose metabolism with a molecular subtype of human breast cancer, namely basal-like breast cancer with MYC activation. This conclusion, reached through a genome-wide approach, links prior observations that i.) the basal-like breast cancer subtype is enriched for tumors with MYC copy gain (52) and a MYC gene expression signature (30, 31) and that ii.) breast cancers which lack expression of estrogen receptors, progesterone receptors, and HER2 gene amplification (“triple-negative”), as is true for the majority of basal-like breast cancers, have shown increased FDG-uptake in larger clinical studies (53) (54). However, basal-like breast cancers are defined by their gene expression profile, express ER or overexpress HER2 in up to 20 % of cases, and represent a disease subgroup that is distinct from triple-negative breast cancer (55). Our findings suggest that FDG-PET may be particularly useful as biomarker for therapies that target the basal-like breast cancer subtype or the “addiction” of MYC-induced tumors to the glycolysis and glutamine pathway (45, 46).

Supplementary Material


FINANCIAL SUPPORT: This work was supported through grants 5P50 CA086306-07 (TGG and IKM), U54CA143798 (IKM), R21-CA137896 (IKM), 5 R25 CA 098010 (NP), P50 CA086438-10 (SML), and 2PO1CA094060 (SML) from the NCI. Further support was provided by the Leon Levy foundation (JTH, IKM), the Sontag Foundation, and the Doris Duke Charitable Foundation (IKM).

This work is dedicated to the memory of William Gerald (MSKCC). We thank members of the Mellinghoff and Graeber Laboratories for helpful discussions and Drs. Jim Fagin, Neal Rosen, and Charles Sawyers for reviewing the manuscript. TGG is an Alfred P. Sloan Research Fellow. IKM is the recipient of an Advanced Clinical Research Award from the American Society of Clinical Oncology and a Forbeck Scholar.


1. Warburg O. On the origin of cancer cells. Science. 1956;123:309–14. [PubMed]
2. Kaelin WG, Jr, Thompson CB. Q&A: Cancer: clues from cell metabolism. Nature. 2010;465:562–4. [PubMed]
3. Fletcher JW, Djulbegovic B, Soares HP, Siegel BA, Lowe VJ, Lyman GH, et al. Recommendations on the use of 18F-FDG PET in oncology. J Nucl Med. 2008;49:480–508. [PubMed]
4. Jadvar H, Alavi A, Gambhir SS. 18F-FDG uptake in lung, breast, and colon cancers: molecular biology correlates and disease characterization. J Nucl Med. 2009;50:1820–7. [PMC free article] [PubMed]
5. Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP, et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science. 1999;286:531–7. [PubMed]
6. Plaisier SB, Taschereau R, Wong JA, Graeber TG. Rank-rank hypergeometric overlap: identification of statistically significant overlap between gene-expression signatures. Nucleic Acids Res. 2010 [PMC free article] [PubMed]
7. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102:15545–50. [PubMed]
8. Kanehisa M, Araki M, Goto S, Hattori M, Hirakawa M, Itoh M, et al. KEGG for linking genomes to life and the environment. Nucleic Acids Res. 2008;36:D480–4. [PMC free article] [PubMed]
9. Avril N, Menzel M, Dose J, Schelling M, Weber W, Janicke F, et al. Glucose metabolism of breast cancer assessed by 18F-FDG PET: histologic and immunohistochemical tissue analysis. J Nucl Med. 2001;42:9–16. [PubMed]
10. Zaslaver A, Mayo AE, Rosenberg R, Bashkin P, Sberro H, Tsalyuk M, et al. Just-in-time transcription program in metabolic pathways. Nat Genet. 2004;36:486–91. [PubMed]
11. Ihmels J, Levy R, Barkai N. Principles of transcriptional control in the metabolic network of Saccharomyces cerevisiae. Nat Biotechnol. 2004;22:86–92. [PubMed]
12. Mootha VK, Lindgren CM, Eriksson KF, Subramanian A, Sihag S, Lehar J, et al. PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat Genet. 2003;34:267–73. [PubMed]
13. Matei D, Graeber TG, Baldwin RL, Karlan BY, Rao J, Chang DD. Gene expression in epithelial ovarian carcinoma. Oncogene. 2002;21:6289–98. [PubMed]
14. Robey IF, Stephen RM, Brown KS, Baggett BK, Gatenby RA, Gillies RJ. Regulation of the Warburg effect in early-passage breast cancer cells. Neoplasia. 2008;10:745–56. [PMC free article] [PubMed]
15. Fan C, Oh DS, Wessels L, Weigelt B, Nuyten DS, Nobel AB, et al. Concordance among gene-expression-based predictors for breast cancer. N Engl J Med. 2006;355:560–9. [PubMed]
16. Chang HY, Nuyten DS, Sneddon JB, Hastie T, Tibshirani R, Sorlie T, et al. Robustness, scalability, and integration of a wound-response gene expression signature in predicting breast cancer survival. Proc Natl Acad Sci U S A. 2005;102:3738–43. [PubMed]
17. Wang Y, Klijn JG, Zhang Y, Sieuwerts AM, Look MP, Yang F, et al. Gene-expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer. Lancet. 2005;365:671–9. [PubMed]
18. Chin K, DeVries S, Fridlyand J, Spellman PT, Roydasgupta R, Kuo WL, et al. Genomic and transcriptional aberrations linked to breast cancer pathophysiologies. Cancer Cell. 2006;10:529–41. [PubMed]
19. Bergamaschi A, Kim YH, Wang P, Sorlie T, Hernandez-Boussard T, Lonning PE, et al. Distinct patterns of DNA copy number alteration are associated with different clinicopathological features and gene-expression subtypes of breast cancer. Genes Chromosomes Cancer. 2006;45:1033–40. [PubMed]
20. Kroemer G, Pouyssegur J. Tumor cell metabolism: cancer’s Achilles’ heel. Cancer Cell. 2008;13:472–82. [PubMed]
21. Saal LH, Holm K, Maurer M, Memeo L, Su T, Wang X, et al. PIK3CA mutations correlate with hormone receptors, node metastasis, and ERBB2, and are mutually exclusive with PTEN loss in human breast carcinoma. Cancer Res. 2005;65:2554–9. [PubMed]
22. Majumder PK, Febbo PG, Bikoff R, Berger R, Xue Q, McMahon LM, et al. mTOR inhibition reverses Akt-dependent prostate intraepithelial neoplasia through regulation of apoptotic and HIF-1-dependent pathways. Nat Med. 2004;10:594–601. [PubMed]
23. Saal LH, Johansson P, Holm K, Gruvberger-Saal SK, She QB, Maurer M, et al. Poor prognosis in carcinoma is associated with a gene expression signature of aberrant PTEN tumor suppressor pathway activity. Proc Natl Acad Sci U S A. 2007;104:7564–9. [PubMed]
24. Saal LH, Gruvberger-Saal SK, Persson C, Lovgren K, Jumppanen M, Staaf J, et al. Recurrent gross mutations of the PTEN tumor suppressor gene in breast cancers with deficient DSB repair. Nat Genet. 2008;40:102–7. [PMC free article] [PubMed]
25. Lee C, Kim JS, Waldman T. PTEN gene targeting reveals a radiation-induced size checkpoint in human cancer cells. Cancer Res. 2004;64:6906–14. [PubMed]
26. Yu D, Cozma D, Park A, Thomas-Tikhonenko A. Functional validation of genes implicated in lymphomagenesis: an in vivo selection assay using a Myc-induced B-cell tumor. Ann N Y Acad Sci. 2005;1059:145–59. [PMC free article] [PubMed]
27. Schuhmacher M, Kohlhuber F, Holzel M, Kaiser C, Burtscher H, Jarsch M, et al. The transcriptional program of a human B cell line in response to Myc. Nucleic Acids Res. 2001;29:397–406. [PMC free article] [PubMed]
28. Coller HA, Grandori C, Tamayo P, Colbert T, Lander ES, Eisenman RN, et al. Expression analysis with oligonucleotide microarrays reveals that MYC regulates genes involved in growth, cell cycle, signaling, and adhesion. Proc Natl Acad Sci U S A. 2000;97:3260–5. [PubMed]
29. Adler AS, Lin M, Horlings H, Nuyten DS, van de Vijver MJ, Chang HY. Genetic regulators of large-scale transcriptional signatures in cancer. Nat Genet. 2006;38:421–30. [PMC free article] [PubMed]
30. Alles MC, Gardiner-Garden M, Nott DJ, Wang Y, Foekens JA, Sutherland RL, et al. Meta-analysis and gene set enrichment relative to er status reveal elevated activity of MYC and E2F in the “basal” breast cancer subgroup. PLoS One. 2009;4:e4710. [PMC free article] [PubMed]
31. Chandriani S, Frengen E, Cowling VH, Pendergrass SA, Perou CM, Whitfield ML, et al. A core MYC gene expression signature is prominent in basal-like breast cancer but only partially overlaps the core serum response. PLoS One. 2009;4:e6693. [PMC free article] [PubMed]
32. Shipp MA, Ross KN, Tamayo P, Weng AP, Kutok JL, Aguiar RC, et al. Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning. Nat Med. 2002;8:68–74. [PubMed]
33. Chang HY, Sneddon JB, Alizadeh AA, Sood R, West RB, Montgomery K, et al. Gene expression signature of fibroblast serum response predicts human cancer progression: similarities between tumors and wounds. PLoS Biol. 2004;2:E7. [PMC free article] [PubMed]
34. Ellwood-Yen K, Graeber TG, Wongvipat J, Iruela-Arispe ML, Zhang J, Matusik R, et al. Myc-driven murine prostate cancer shares molecular features with human prostate tumors. Cancer Cell. 2003;4:223–38. [PubMed]
35. Chrzan P, Skokowski J, Karmolinski A, Pawelczyk T. Amplification of c-myc gene and overexpression of c-Myc protein in breast cancer and adjacent non-neoplastic tissue. Clin Biochem. 2001;34:557–62. [PubMed]
36. Naidu R, Wahab NA, Yadav M, Kutty MK. Protein expression and molecular analysis of c-myc gene in primary breast carcinomas using immunohistochemistry and differential polymerase chain reaction. Int J Mol Med. 2002;9:189–96. [PubMed]
37. Gao P, Tchernyshyov I, Chang TC, Lee YS, Kita K, Ochi T, et al. c-Myc suppression of miR-23a/b enhances mitochondrial glutaminase expression and glutamine metabolism. Nature. 2009;458:762–5. [PMC free article] [PubMed]
38. Nikiforov MA, Chandriani S, O’Connell B, Petrenko O, Kotenko I, Beavis A, et al. A functional screen for Myc-responsive genes reveals serine hydroxymethyltransferase, a major source of the one-carbon unit for cell metabolism. Mol Cell Biol. 2002;22:5793–800. [PMC free article] [PubMed]
39. Shim H, Dolde C, Lewis BC, Wu CS, Dang G, Jungmann RA, et al. c-Myc transactivation of LDH-A: implications for tumor metabolism and growth. Proc Natl Acad Sci U S A. 1997;94:6658–63. [PubMed]
40. O’Donnell KA, Yu D, Zeller KI, Kim JW, Racke F, Thomas-Tikhonenko A, et al. Activation of transferrin receptor 1 by c-Myc enhances cellular proliferation and tumorigenesis. Mol Cell Biol. 2006;26:2373–86. [PMC free article] [PubMed]
41. Semenza GL. Targeting HIF-1 for cancer therapy. Nat Rev Cancer. 2003;3:721–32. [PubMed]
42. Solit DB, Mellinghoff IK. Tracing cancer networks with phosphoproteomics. Nat Biotechnol. 2010;28:1028–9. [PubMed]
43. Vander Heiden MG, Cantley LC, Thompson CB. Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science. 2009;324:1029–33. [PMC free article] [PubMed]
44. Morrish F, Isern N, Sadilek M, Jeffrey M, Hockenbery DM. c-Myc activates multiple metabolic networks to generate substrates for cell-cycle entry. Oncogene. 2009;28:2485–91. [PMC free article] [PubMed]
45. Wise DR, DeBerardinis RJ, Mancuso A, Sayed N, Zhang XY, Pfeiffer HK, et al. Myc regulates a transcriptional program that stimulates mitochondrial glutaminolysis and leads to glutamine addiction. Proc Natl Acad Sci U S A. 2008;105:18782–7. [PubMed]
46. Dang CV, Le A, Gao P. MYC-induced cancer cell energy metabolism and therapeutic opportunities. Clin Cancer Res. 2009;15:6479–83. [PMC free article] [PubMed]
47. Shaw RJ, Cantley LC. Ras, PI(3)K and mTOR signalling controls tumour cell growth. Nature. 2006;441:424–30. [PubMed]
48. Gregorian C, Nakashima J, Dry SM, Nghiemphu PL, Smith KB, Ao Y, et al. PTEN dosage is essential for neurofibroma development and malignant transformation. Proc Natl Acad Sci U S A. 2009;106:19479–84. [PubMed]
49. Hodgson NC, Gulenchyn KY. Is there a role for positron emission tomography in breast cancer staging? J Clin Oncol. 2008;26:712–20. [PubMed]
50. Avril N, Rose CA, Schelling M, Dose J, Kuhn W, Bense S, et al. Breast imaging with positron emission tomography and fluorine-18 fluorodeoxyglucose: use and limitations. J Clin Oncol. 2000;18:3495–502. [PubMed]
51. Soret M, Bacharach SL, Buvat I. Partial-volume effect in PET tumor imaging. J Nucl Med. 2007;48:932–45. [PubMed]
52. Hynes NE, Stoelzle T. Key signalling nodes in mammary gland development and cancer: Myc. Breast Cancer Res. 2009;11:210. [PMC free article] [PubMed]
53. Basu S, Chen W, Tchou J, Mavi A, Cermik T, Czerniecki B, et al. Comparison of triple-negative and estrogen receptor-positive/progesterone receptor-positive/HER2-negative breast carcinoma using quantitative fluorine-18 fluorodeoxyglucose/positron emission tomography imaging parameters: a potentially useful method for disease characterization. Cancer. 2008;112:995–1000. [PubMed]
54. Specht JM, Kurland BF, Montgomery SK, Dunnwald LK, Doot RK, Gralow JR, et al. Tumor metabolism and blood flow as assessed by positron emission tomography varies by tumor subtype in locally advanced breast cancer. Clin Cancer Res. 2010;16:2803–10. [PMC free article] [PubMed]
55. Foulkes WD, Smith IE, Reis-Filho JS. Triple-negative breast cancer. N Engl J Med. 2010;363:1938–48. [PubMed]