In order to determine the optimal time to sample tissue after infusing [U-¹³C]-Glc into lung cancer patients, we used ¹H NMR to analyze the profiles of ¹³C-labeled products in plasma samples at several time points. Other than the administered tracer itself ([U-¹³C]-Glc), the major ¹³C-labeled metabolite in plasma was lactate. The time course changes in glucose and lactate concentrations and their % ¹³C enrichment are shown in Figure 1A. Plasma glucose was maximally enriched (up to 49%) in ¹³C immediately (0.5 hr) following the tracer infusion. Three hours later, a significant fraction (8.3–32%) of the plasma glucose remained ¹³C-labeled but by 12 hrs, the % ¹³C enrichment dropped to 2–5%, as illustrated for four patients. The % ¹³C-labeled lactate showed a similar time course as the % ¹³C-labeled glucose, except that the % enrichment reached a maximum (5–22%) after 3 hrs of infusion (Figure 1A). These data show that once taken up by tissues, ¹³C-glucose was metabolized quickly to ¹³C-lactate and secreted back into the blood. In addition, a large fraction of the plasma ¹³C-lactate was uniformly labeled in ¹³C, as evidenced by the fine splitting structure of the ¹³C satellites of the 3-methyl protons of lactate (Figures 1B and 2) 1920 and by GC-MS analysis (Table 1). GC-MS analysis also revealed a significant presence of mass isotopomers of lactate with one or two ¹³C labels for patients #8–10, which is consistent with an active Cori cycle 9. Based on this time-course analysis of ¹³C-isotopomers of metabolites in human plasma samples, we chose a duration of 3–4 h between ¹³C-glucose infusion and surgical resection for patients #6–10 in order to optimize ¹³C incorporation from [U-¹³C]-Glc into various metabolites.

Of the twelve subjects undergoing ¹³C glucose infusion, the median age was 63 (range 52–76), 67% were male, 67% had squamous cell carcinoma, and 33% adeno- or adenosquamous carcinoma, all of grade II or III. All subjects had a history of heavy smoking.

An example of the assignment of metabolites and their ¹³C-isotopomers in the TCA extracts by 2-D ¹H TOCSY and high-resolution 1-D NMR spectra is illustrated in Figure 2. Assignments were made by matching the chemical shifts, spin-spin coupling patterns, and available covalent connectivities (traced by solid red rectangles in the TOCSY contour maps) of individual resonances against those of the standard compounds 2021. Metabolites that were commonly observed in all lung tissue extracts include isoleucine (Ile), leucine (Leu), valine (Val), lactate, alanine (Ala), arginine (Arg), proline (Pro), glutamate (Glu), oxidized glutathione (GSSG), glutamine (Gln), succinate, citrate, aspartate (Asp), creatine (Cr), phosphocholine (P-choline), taurine, glycine (Gly), phenylalanine (Phe), tyrosine (Tyr), myo-inositol, α- and β-glucose, NAD⁺, cytosine nucleotides (CXP), uracil nucleotides (UXP), guanine nucleotides (GXP), and adenine nucleotides (AXP). The ¹H TOCSY assignment of metabolites was complemented by the 2-D ¹H-¹³C HSQC analysis of the same extract, as shown in Figure 3, where the ¹H-¹³C covalent bonding patterns were observed. The HSQC spectrum provided better resolution for some metabolites such as Glu, Gln, and GSSG (Fig 3C, inset), thereby confirming their assignment.

In addition to metabolite identification, the TOCSY and HSQC analyses provided ¹³C positional isotopomer information for several metabolites in the lung TCA extracts. The ¹H TOCSY data (cf. Fig. 2B) unambiguously revealed the presence of uniformly ¹³C labeled lactate ([U-¹³C]-lactate) and Ala (([U-¹³C]-Ala) by the ¹³C satellite cross-peak pattern (patterns 1–4 ) of 3-methyl and 2-methine protons of lactate (lactate-H3 and H2) and Ala (Ala-H3 and H2) (traced by dashed green rectangles in Figure 2B) 20. This pattern is consistent with the enrichment of multiply labeled lactate and alanine of 200–300 fold over the natural abundance levels. The presence of ¹³C-3-Glu, ¹³C-3-Gln, and ¹³C-3-glutamyl residue of oxidized glutathione (Glu-GSSG) was evidenced by the ¹³C satellite cross-peak patterns 9, 10, 14 () and 11–13 () (traced by dashed green lines in Fig. 2B). Patterns 5–8 and 15 denote the presence of ¹³C-2-Glu-GSSG and ¹³C-2-Glu while patterns 16 and 17 indicate the presence of ¹³C-2-Asp.

The ¹³C-positional isotopomer information obtained from the TOCSY analysis was confirmed by the HSQC analysis, including the positional isotopomers of ¹³C-3-Ala, ¹³C-3-lactate, ¹³C-3-Glu, ¹³C-3-Gln ¹³C-3-Glu-GSSG, ¹³C-2-Glu-GSSG, ¹³C-2-Glu, and ¹³C-2-Asp (cf. respectively Ala-C3, Lactate-C3, Glu-C3, Gln-C3, GSSG-Glu-C3, Glu-GSSG-C2, Glu-C2, and Asp-C2 in Fig. 3B and 4B). In addition, the HSQC data also provided complementary information on isotopomers whose ¹³C satellite cross-peaks were too weak to observe or masked by other cross-peaks in the crowded regions of the TOCSY spectrum [see text and Figure S1 in Additional file 1]. These included selective ¹³C enrichment in C-3-Asp, C-2,3-succinate, C-2,4-citrate, C-1'-ribose-5'AXP, and C-1-α- and -β-glucose (cf. respectively Asp-C3, succinate-C2,3, citrate-C2,4, 5'AXP-C1', and α- and β-Glc-C1, Fig. 3B and 4B).

Based on the metabolite assignment in Figs. 2 and 3, the 1-D ¹H NMR and ¹³C HSQC projection spectra allowed comparison of metabolite and ¹³C isotopomer profiles in the TCA extracts of paired non-cancerous and cancerous lung tissues. This is illustrated in Figure 4 for patient #6. It is clear in Fig. 4A that the majority of the metabolites (except for glucose) were present at a higher level in the cancerous than in the non-cancerous lung tissue. These also included the ¹³C-labeled isotopomers, [U-¹³C]-lactate and [U-¹³C]-Ala (as denoted respectively by the Lactate_sat and Ala_sat resonance in Fig. 4A). The extent of ¹³C enrichment in lactate, Ala (as uniformly labeled species) and Glu (at the C-2 position) was quantified from the respective 2-D ¹H TOCSY cross-peak patterns (cf. Figure 2B), as previously described 18. They were consistently higher in the cancer than in the non-cancerous tissue from the same patient, as shown in Table 2. The extent of ¹³C enrichment in glucose (as [U-¹³C]-Glc) for non-cancerous tissues was determined from the 1-D ¹H spectra (cf. Fig. 4A), which was considerably lower in the cancer than in its non-cancerous counterpart (cf. Table 2). Moreover, increased ¹³C abundance (or ¹³C peak intensity) of most metabolites in cancer relative to non-cancerous tissues was evident in the 1-D HSQC projection spectra in Fig. 4B. Part of the increase in ¹³C abundance from non-cancerous to cancer tissues reflected the difference in total metabolite concentration, which contains 1.1% ¹³C at natural abundance. However, as reasoned in Additional file 1, selective enrichment in ¹³C for a number of metabolite carbons also contributed to the increase in their ¹³C peak intensity. These include C-3-Ala, C-2,3-lactate, C-3-Gln+GSSG, C-2 to 4-Glu, C-4-Gln, C-4-GSSG, C-2,4-citrate, C-2,3-Asp, C-1',4',5'-5'-AXP, and C-1',4'-5'-UXP.

To quantify the ¹³C abundance of metabolites at specific carbon positions, the 2-D HSQC spectra were utilized for the better resolution than the 1-D ¹³C projection spectra (cf. Fig. 4B). This was performed for ¹³C-2,3-succinate of patients #6–10 by integrating the volume of its HSQC cross-peak (cf. Fig. 3C). The relative ¹³C abundance (a measure of selective enrichment) of these two carbons was calculated by normalizing the HSQC peak volume to the total succinate concentration. The relative ¹³C abundance of C-2,3-succinate for tumor tissues (3.6 ± 1.2) was significantly greater than that for non-cancerous tissues (0.6 ± 0.7) with a p value of < 0.01.

Parallel analysis of the lung TCA extracts by GC-MS served to verify key findings in the metabolite profile obtained by NMR while providing absolute quantification of a subset of metabolites and their ¹³C mass isotopomers. Table 3 shows the quantification of selected metabolites and their total ¹³C enrichment (in excess of natural abundance) by GC-MS. Also shown is the quantification of the m+3 mass isotopomer of Asp (¹³C₃-Asp or Asp with three of its carbons labeled in ¹³C). For patient #6, the GC-MS data revealed excess total ¹³C enrichment in tumor over non-cancerous tissues for Ala, Asp, Glu, Gln, lactate, citrate and succinate. An enhanced production of ¹³C₃-Asp in the tumor compared with the paired non-cancerous tissue was also evident. This is consistent with the NMR observation (cf. Fig. 4B) of the buildup of various ¹³C positional isotopomers of metabolites including [¹³C-3]-Ala, [¹³C-2,3]-lactate, [¹³C-2,3,4]-Glu, [¹³C-4]-Gln, [¹³C-2,3]-Asp, [¹³C-2,4]-citrate and [¹³C2,3]-succinate for the tumor, compared with the paired non-cancerous tissues of #6. A similar ¹³C enrichment pattern of Ala, lactate, succinate and citrate in lung tumor tissues was observed in all five patients where ¹³C labeling in these metabolites was sufficient to be quantified by GC-MS. In addition, the enhanced synthesis of ¹³C₃-Asp in tumor tissues was evident in four of the five patients (Table 3). It should be noted that the fraction of ¹³C₃-Asp in tumor tissues (0.5 to 1.4%) was far above the natural abundance background (5 × 10^-4%) (cf. Table 3).

Utilizing the GC-MS data, pairs of biosynthetically-related ¹³C-labeled metabolites in tumor and non-cancerous tissues were tested for precursor-product relationships, as illustrated in Figure 5 and Table 4. ¹³C-succinate, ¹³C-citrate and ¹³C-Glu are products of the Krebs cycle while ¹³C-Ala is derived from pyruvate, the end product of glycolysis (cf. Figure 6).

A linear correlation in the concentration of ¹³C-succinate with that of ¹³C-Ala, ¹³C-Glu, or ¹³C-citrate (Fig. 5A–C, Table 4) was discernable for lung cancer tissues but not for non-cancerous lung tissues. Also noted was the less significant correlation between ¹³C-succinate and ¹³C-lactate for the lung tumor tissues (Table 4). When total concentrations were plotted, the correlation was even less apparent, particularly for the non-cancerous tissues (Fig. 5D–F, Table 4). This observation underscores the need for acquiring ¹³C-isotopomer data, instead of just steady-state concentrations, to deduce meaningful relationships between transformed products in related pathways. Moreover, in all six plots of Fig. 5, a separation of non-cancerous and tumor tissues was evident, i.e. the tumor and non-cancerous tissues clustered in the high and low concentration quadrants, respectively.

The distinct ¹³C labeling patterns in the Krebs cycle metabolites in tumor tissues described above indicates the possibility of altered gene expression in relevant enzymes. This was examined by real-time PCR analysis of key mitochondrial dehydrogenases (DH) along with the anaplerotic pyruvate carboxylase (PC) and glutaminase for tumor and surrounding non-tumorous tissues, as shown in Table 5. Increased expression of the two isoforms of PC gene was evident in patients #6–10 with an average fold change of 3.36 ± 1.13 i.e. significantly higher in tumors (p < 0.01). In contrast, the expression of the other important anaplerotic enzyme gene, glutaminase (GLS) was lower in the tumors relative to the surrounding non-cancerous tissues, with an average fold change of 0.52 ± 0.16. For Krebs cycle DH, there was a substantial decrease of isocitrate DH (IDH) and α-ketoglutarate DH (OGDH) in contrast to a modest activation of malate DH (MDH) expression, while succinate DH (SDH) and fumarate hydratase (FH) showed no statistically significant changes in expression in tumor tissues.

These data differ from a recent report on glioblastoma cells in culture, where glutaminolysis was shown to be significant while pyruvate carboxylation was undetected 13, even though the parent astrocytes have high PC activity 1232. Our findings in human NSCLC patients indicate that the activation of either anaplerotic pathway may depend on the tumor cell type. This is consistent with a low glutaminolysis (Gln to lactate) capacity that we found in human NSCLC A549 cells (A.N. Lane, T. W-M. Fan, M.M. McKinney and J.L. Tan, unpublished data), in contrast to the findings for glioblastoma cells.

The in vivo ¹³C isotopomer profile and gene expression data (see above) indicate increased PC activity in the NSCLC tumors compared with non-tumorous lung tissue. To determine whether PC gene activation leads to an enhanced expression of the enzyme, Western blotting was performed on paired tumor and non-cancerous tissues from the five patients. The PC response normalized to that of α-tubulin is shown in Figure 7B. Lung tumor tissues from patients #6, 9, and 10 exhibited a significant enrichment of PC protein over the non-cancerous counterpart. Patient #8 had a high interfering background over the PC band region for the non-cancerous tissue, which made it difficult to quantify the PC response. The normalized PC response for patient #7 was comparable between the paired non-cancerous and tumor tissues.

Based on the ¹³C isotopomer analysis by NMR (e.g. Fig. 4, Table 2) and GC-MS (Tables 3, 4), it is clear that human lung tumor tissues exhibited an increased capacity for carbon incorporation from glucose into lactate, Ala, citrate, Glu, succinate, ribosyl moiety of nucleotides, and Asp relative to the surrounding "non-cancerous" lung tissues. The transformation of [U-¹³C]-Glc into [U-¹³C]-lactate and [U-¹³C]-Ala can only occur via glycolysis and to a much lesser extent, the pentose phosphate pathways (PPP), whereas ¹³C-ribose of nucleotides can only be derived from PPP 33. The enhanced production of these metabolites in tumor tissues provided direct metabolic evidence for the enhancement of the glycolytic capacity 13435. However, it is unclear whether the increased production of ¹³C-ribose of nucleotides resulted from an enhancement in oxidative and/or non-oxidative branches of the PPP in lung tumor tissues. Further studies with additional tracers (e.g. [¹³C-1,2]-glucose 29) will be needed to resolve this issue.

The increased conversion of ¹³C carbons from glucose into the Krebs cycle intermediates (citrate and succinate) or related metabolites (Glu and Asp) in tumor compared to non-cancerous tissues (cf. Fig. 4 and Tables 2, 3) is unexpected. We anticipated a reduced and/or disrupted transformation of glucose-carbon into the intermediates of the Krebs cycle, based on the commonly recognized concept of mitochondrial "dysfunction" in cancer 363738. Instead, the cycle capacity was not reduced in any of the five lung tumor tissues analyzed. This finding is supported by the enhanced ¹³C incorporation into citrate, succinate, Glu (a surrogate marker of α-ketoglutarate or αKG) and Asp (a surrogate marker of oxaloacetate or OAA) in tumor tissues (Tables 2, 3, 4 and Fig. 4B). The production of these metabolites reflects the operation of the entire cycle (cf. Figure 6), which is also consistent with the significant correlations between ¹³C-labeled citrate, succinate, and Glu in the lung tumor tissues (cf. Fig. 5B, C). It is also interesting to note a stronger relationship between ¹³C-labeled Ala and succinate (Fig. 5A) than that between ¹³C-labeled lactate and succinate in lung tumor tissues. This could imply a separate pool of pyruvate derived from Ala for entry into the Krebs cycle (cf. Fig. 6B). Although unexpected, the combination of activated glycolysis and Krebs cycle rationally explains how the excess demand for energy, NADPH reducing equivalents, and biosynthetic precursors (e.g. Asp for nucleotides, citrate for fatty acyl chains, Glu/Gln for proteins) is fulfilled for tumor growth and proliferation.

The display of lung tumor tissues in the enhanced production of the [¹³C₃]-Asp mass isotopomer (cf. Table 3) is intriguing. Given the enhanced expression of PC, this is best explained with the activation of pyruvate carboxylation, i.e. carboxylation of [U-¹³C]-pyruvate to generate [¹³C₃]-oxaloacetate, which is transaminated to produce [¹³C₃]-Asp (Fig. 6B). Alternative paths to Asp synthesis from [U-¹³C]-glucose via glycolysis and the first turn of the Krebs cycle leads to two ¹³C labels in Asp, not the three ¹³C-labeled carbons in Asp that was observed (cf. Table 3). For the second turn of the Krebs cycle, [¹³C₃]-Asp can be produced, provided that [¹³C₂]-acetyl CoA (derived from [U-¹³C]-pyruvate via PDH) is condensed with [¹³C₂]-OAA from the first turn. However, the % enrichment of acetyl CoA or [U-¹³C]-pyruvate and [¹³C₂]-OAA was low, as evidenced by the = 15% enrichment in pyruvate surrogate Ala and lactate and < 8% enrichment in OAA surrogate Asp. Thus, the probability for the condensation of [¹³C₂]-acetyl CoA in the second turn with [¹³C₂]-OAA from the first turn is very low (< 1%). In essence, labeled acetyl CoA will always condense with unlabelled OAA, and the labeling pattern of the Krebs cycle intermediates will have the appearance of a single turn, regardless of the actual turn numbers.

Under the low enrichment conditions, the PDH activity produces ¹³C-4,5-αKG (and thus ¹³C-4,5-Glu) and [¹³C₂]-Asp (m0+2) through the Krebs cycle (cf. Fig. 6A). In contrast, pyruvate carboxylation leads to the production of ¹³C-2,3-Glu via the Krebs cycle, which is distinct from the PDH pathway (cf. Fig. 6B). This distinction in the labeled pattern of Glu provided further metabolic evidence for PC activation in lung tumor tissues. Namely, the ¹³C enrichment of Glu at the C-2 and C-3 positions (Fig. 4B and Table 2) was higher in tumor tissues than its non-cancerous counterpart. Taken together, the ¹³C isotopomer analysis by both NMR and GC-MS revealed the activation of anaplerotic pyruvate carboxylation pathways in NSCLC. It is unclear whether the other anaplerotic pathway, glutaminolysis, is also activated, from the present metabolic data, although the gene expression data indicated otherwise (cf. Table 5).

The metabolomic data were also corroborated by measurements of enhanced gene and protein expression of PC in lung tumors relative to their non-cancerous counterparts. In this case, the ¹³C-isotopomer data was pivotal in directing the interrogations of PC gene expression by RT-PCR and protein analysis by Western blotting. Thus, such metabolomics-edited gene and protein expression analysis (which we have demonstrated for cultured lung cancer cells 22) is equally useful for uncovering PC up regulation in human patients.

Although previously unknown in human lung cancer, PC activation was recently reported in hepatic tumors in rat 39 and in estradiol-stimulated breast cancer cells 40. PC expression or suppression has also been respectively associated with enhanced metabolism/proliferation of mammalian cells 41 or inhibition of tumor cell growth 42. Most recently, PC-induced anaplerotic flux into the Krebs cycle and pyruvate cycling between the cytoplasm and mitochondria was shown to be important in the regulation of glucose-stimulated insulin secretion 30. These effects of PC are presumably mediated through its ability to replenish the Krebs cycle intermediates, thereby enhancing energy metabolism and fulfilling biosynthetic demands from proliferating cells 41. From this perspective, it is reasonable to postulate that PC activation may be important for the transformation of lung primary cells into a more highly proliferative state.

Lung cancer patients were recruited based on the criteria of surgical eligibility and no history of diabetes. Each patient was consented in accordance with the U.S. HIPAA regulations. The following patient protocol was approved by the Internal Review Board at the University of Louisville. Ten grams uniformly ¹³C labeled glucose ([U-¹³C]-Glc) in sterile saline solution were infused i.v. over a 30 min time period into each patient in the pre-op room approximately 3 or 12 hr prior to resection. Whole blood samples were collected into a vacutainer containing the anticoagulant K₃EDTA before and after [U]-¹³C-Glc infusion as well as after the surgery. Additional blood samples were collected 3 and 12 hr after the ¹³C-Glc infusion for the 12 hr treatment cases. To minimize metabolic changes, all tissue and blood samples were collected at the operating table with a comparable delay before liquid N₂ freezing. The blood samples were placed on ice immediately after collection and centrifuged at 4°C at 3,500 × g for 15 min to recover the plasma fraction. All blood samples were aliquoted and flash-frozen in liquid N₂ within 30 min of collection. Potassium EDTA was used as an anti-coagulant to minimize metabolic artifacts associated with anti-coagulation; it also served to remove the influence of interfering cations such as paramagnetic Fe³⁺ and Cu²⁺ for NMR analysis.

All timings from first incision to cutting arteries and veins to the lung were recorded so that the period of ischemia could be determined. Immediately after tissue resection, excess blood was blotted from the tissue, and small pieces of non-cancerous and tumor tissue were cut by the surgeon and freeze-clamped in liquid N₂ within 5 minutes of removal from the chest cavity. The freezing process arrested metabolic changes almost instantaneously. In all cases, the tumors were well defined so that the extent of the tumor was assessed visually and by palpation. Non-cancerous tissue was removed at least 2 cm from the tumor margin and certified to be tumor-free by trained pathologists. Subsequent pathological evaluation also confirmed tumor status and provided the tumor stage. All samples were stored at -80°C until further processing for analysis.

Frozen tissue samples were pulverized into < 10 μm particles in liquid N₂ using a Spex freezer mill (Spex CertiPrep, Inc., Metuchen, NJ) to maximize efficiency for subsequent extraction while maintaining biochemical integrity. An aliquot of the frozen powder was lyophilized before extraction for polar metabolites.

Water-soluble or polar metabolites were extracted from lyophilized tissue powders (4–48 mg) in ice-cold 10% trichloroacetic acid (TCA) (v/w minimum 40/1) while leaving proteins, nucleic acids, and polysaccharides behind, as described previously 43. The extraction was performed twice for quantitative recovery. An aliquot (150 μl) of plasma samples were made to final 10% TCA concentration to precipitate proteins and recover polar metabolites in the supernatant. TCA was then removed from tissue or plasma extracts by lyophilization. The dry pellet was dissolved in nanopure water and two small aliquots were lyophilized for silylation and GC-MS analysis while the remaining bulk was passed through a Chelex 100 resin column (Bio-Rad Laboratories, Inc., Hercules, CA) to neutralize and remove interfering multi-valent cations for NMR analysis.

The ¹H reference standard, DSS (2,2-Dimethyl-2-silapentane-5-sulfonate sodium salt) was added (30 or 50 nmoles) to the TCA extracts of tissue or plasma samples, respectively. NMR analysis of the TCA extracts was performed at 20°C on a Varian Inova 14.1 T NMR spectrometer (Varian, Inc., Palo Alto, CA) equipped with a 5-mm HCN inverse triple resonance pfg cold probe. The following NMR experiments were conducted for the determination of metabolite and ¹³C positional isotopomers: 1-D ¹H, 2-D ¹H TOCSY, 2-D ¹H-¹³C HSQC and HSQC-TOCSY 2021. ¹H and ¹³C chemical shifts of the TCA extracts were referenced to DSS at 0.00 ppm and indirectly to the ¹H shift, respectively. Various metabolites were identified based on their ¹H and ¹³C chemical shifts, ¹H coupling patterns, as well as ¹H-¹H and ¹H-¹³C covalent linkage patterns (acquired from the TOCSY and HSQC experiments, respectively), in comparison with those in our in-house standard database http://research.louisville.edu/metabolomics.

For metabolite and ¹³C-isotopomer quantification, selected ¹H peaks in the 1-D NMR spectra were deconvoluted and integrated using MacNuts software (Acorn NMR, Inc., Livermore, CA). The resulting intensity of peaks of interest was calibrated by the peak intensity of DSS for absolute quantification. Percentage ¹³C abundance of labeled metabolites at specific carbon positions was quantified by integrating appropriate ¹³C satellite peaks in 1-D ¹H or 2-D TOCSY spectra, as previously described 1820.

The same extracts from NMR analysis were subjected to GC-MS analysis for quantifying total and ¹³C-labeled mass isotopomers, as described in full previously 22. Briefly, the lyophilized extract was derivatized in MTBSTFA (N-methyl-N-[tert-butyldimethylsilyl]trifluoroacetamide) (Regis Chemical, Morton Grove, IL) and the tert-butyldimethylslylyl derivatives were separated and quantified on a PolarisQ GC-ion trap MSn (ThermoFinnigan, Austin, TX) equipped with a 50 m × 0.15 mm i.d. open tubular column with 0.4 μm coat BPX-5 (5% phenyl/methyl equivalent) (SGE, Austin, TX). Metabolites were identified based on their GC retention times and mass fragmentation patterns by comparison with those of the standards. Absolute quantification of metabolites was done by calibrating the response of selected ions characteristic of a given metabolite from sample runs with that from standard runs 22. Relative metabolite abundances were calculated using Xcalibur (ThermoFinnigan, San Jose, CA) or Met-IDEA software 44 to extract peak areas of individual ions characteristic of each component. For Met-IDEA, default program settings for ion trap mass spectrometer were used in the data analysis except for a mass accuracy m/z set at 0.001; and a mass range set on either side of the target m/z at ± 0.6. For Xcalibur, peak detection, identification, background subtraction, and quantification was performed using parameters custom-tuned to each analyte. Quantification of mass isotopomers was conducted by subtraction of the isotopic profile at natural-abundance (determined empirically from the analyses of standards) from that of the sample. This procedure was repeated for each series of mass isotopomers to arrive at the ¹³C-enriched profiles 22. In all cases, the pseudo-molecular ion cluster, characteristic of MTBSTFA-derivatives, was used to ensure that true mass isotopomers were quantified.

From an initial gene microarray analysis of a separate set of six paired tissue samples, a number of statistically significant gene expression differences were discerned between lung cancer tissues and their non-cancerous counterparts. This included an over expression of the pyruvate carboxylase (PC) gene in lung cancer versus paired non-cancerous tissues. Due to sample limitation, the array analysis was not performed for patients #6–10, for which metabolic evidence for PC activation was obtained.

Real time (RT)-PCR was used instead to probe PC gene expression in patients #6–10. Non-cancerous and cancer tissue RNA was isolated using RNeasy minikit (Qiagen) essentially following the manufacturer's instruction. The only modification was that Trizol Reagent (invitrogen), instead of Buffer RLT provided by the kit, was used to disrupt and homogenize the pulverized frozen tissues at the first step of RNA isolation. The modified procedure was found to provide a higher RNA yield and better purity. The integrity of RNA was confirmed by 1% agarose gel electrophoresis. RNA was reverse transcribed into first strand cDNA using oligo(dT)₁₈ and SuperScript II reverse transcriptase (Invitrogen). Specifically, 1 μg RNA was added into a 40 μl reaction mixture containing 2 μl 500 μg/ml Oligo(dT)₁₈, 2 μl dNTP mix (10 mM each), 8 μl 5× first-strand buffer, 4 μl 0.1 M DTT, 2 μl RNaseOUT™ (40 units/μl) and 2 μl SuperScript II reverse transcriptase (200 units/μl). The reaction mixture was incubated at 42°C for 50 min and then heated at 70°C for 15 min to terminate the reaction.

RT-PCR amplification was performed with SYBR green dye using a Mastercycler ep Realplex 4S (Eppendorf). For each run, 20 μl 2.5 × Real Master Mix (Qiagen), 0.3 μM of forward and reverse primers along with 2 μl first strand cDNA were mixed. The thermal cycling conditions included an initial denaturation step at 95°C for 2 min, 50 cycles at 95°C for 15 s, 55°C for 15 s and 72°C for 20 s. Each reaction was performed in duplicates. The efficiency of the amplification was close to 2.0 (i.e. 100%) for all primer pairs. Relative expression level of each gene was calculated using the Livak method as described previously 45 with 18S ribosomal RNA as the internal control gene. The primer sequences used were designed by Beacon Designer 5.0 (Premier Biosoft International, Palo Alto, CA) as shown in Table 6.

Pulverized and lyophilized lung tissues were extracted twice in chloroform:methanol (2:1) plus 1 mM butylated hydroxytoluene to remove lipids, which interfered with the extraction of PC. The delipidated tissue powder was then extracted in 62.5 mM Tris-HCl plus 2% sodium dodecyl sulfate (SDS) and 5 mM dithiothreitol and heated at 95°C for 10 min. to denature proteins. The protein extract was analyzed by SDS-PAGE using a 10% polyacrylamide gel and separated proteins were transferred to a PVDF membrane (Immobilon™-P, Millipore, Bedford, MA), and blotted against an anti-PC rabbit polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C. PC was visualized with incubation in a secondary anti-rabbit antibody linked to horseradish peroxidase (HRP) (Thermo Scientific, Rockford, IL), followed by reaction with chemiluminescent HRP substrates (Supersignal^® West Dura Extended Duration substrate, Thermo Scientific), and exposure to X-ray film. The film was digitized using a high-resolution scanner and the image density of appropriate bands (130 kDa for PC and 50 kDa for α-tubulin) was analyzed using Image J (NIH, Bethesda, MD). The image density of the PC protein band was normalized to that of the α-tubulin protein band.

GC-MS analysis was performed in triplicate and % RSD (relative standard deviation) of analysis was calculated for each reported metabolite (Table 3). Student's paired t-test was performed on each pair of tumor and non-cancerous tissues for the real-time PCR data (Table 5) and for the ¹H-TOCSY data (Table 2). Linear regression calculations were done for pairs of metabolites in Table 4.