Stable isotope resolved metabolomics of lung cancer in a SCID mouse model

Abstract

We have determined the time course of [U-¹³C]-glucose utilization and transformations in SCID mice via bolus injection of the tracer in the tail vein. Incorporation of ¹³C into metabolites extracted from mouse blood plasma and several tissues (lung, heart, brain, liver, kidney, and skeletal muscle) were profiled by NMR and GC–MS, which helped ascertain optimal sampling times for different target tissues. We found that the time for overall optimal ¹³C incorporation into tissue was 15–20 min but with substantial differences in ¹³C labeling patterns of various organs that reflected their specific metabolism. Using this stable isotope resolved metabolomics (SIRM) approach, we have compared the ¹³C metabolite profile of the lungs in the same mouse with or without an orthotopic lung tumor xenograft established from human PC14PE6 lung adenocarcinoma cells. The ¹³C metabolite profile shows considerable differences in [U-¹³C]-glucose transformations between the two lung tissues, demonstrating the feasibility of applying SIRM to investigate metabolic networks of human cancer xenograft in the mouse model.

1 Introduction

Stable isotope tracers have been widely used to map metabolic pathways and quantify fluxes in cells, tissues and organisms (Bak et al. 2007; Beger et al. 2009; DeBerardinis et al. 2007; Delgado et al. 2004; Fan et al. 1986, 1988, 2003, 2008, 2009b; Lane et al. 2008; Mancuso et al. 2005; Mason et al. 2007; Mendes et al. 2006; Thornburg et al. 2008; Vizan et al. 2005; Zwingmann and Leibfritz 2003). Although studies with cell cultures are best defined in terms of experimental control and cell-specific metabolism, they lack the ability to probe microenvironmental influences on cellular metabolism in a tissue context. This is crucial to translating understanding of metabolic functions in model cell systems into elucidating human metabolism in situ.

Deciphering the intricate web of human metabolic networks and their perturbations due to disease development or drug treatment is essential to the functional annotations of relevant gene (including single nucleotide polymorphism) and protein level changes. This task has been a difficult challenge, which cannot be resolved fully by total metabolite profiling alone. This is because a given metabolite such as glutamate can participate in many different reactions (e.g. >200, M. Arita, personal communication) such that its total concentration cannot discern changes in all—or even a few—reactions. Even if the concentration of various metabolites in a given pathway is known, it is still difficult to deduce which are the perturbed reactions in the pathway without ambiguity.

To achieve such an understanding, the use of labeled tracers is required. For example, the time course changes in normalized total lactate concentration of neurons induced by LiCl were opposite to the normalized concentration of triply ¹³C labeled (¹³C₃) lactate (Fig. S1, from Fan et al. (2010)). Based on the total lactate concentration alone, one would conclude that glycolysis in neurons was inhibited by LiCl. However, with the use of uniformly ¹³C-labeled glucose (¹³C₆-Glc), the enhanced production of ¹³C₃-lactate by LiCl-treated neurons was evident, which requires an opposite interpretation, i.e. glycolysis was instead stimulated by Li. By combining labeled tracers with stable isotope-resolved metabolomic (SIRM) analysis, it is now possible to reconstruct multiple metabolic pathways based on numerous labeled metabolite patterns and to ascertain the impact of drug treatment or disease state on both targeted and unsuspected pathways (Fan et al. 2005, 2009b, 2010). In essence, the SIRM approach “encodes” each metabolite with pathway-specific dynamics.

Although tracer studies can—and have been—performed directly on human subjects (Fan et al. 2009b; Lane et al. 2009a; Mason et al. 2007), the degree of experimental control and sampling options are extremely limited due to confounding factors such as lifestyle and ethical concerns. In the past, radiotracers were extensively employed in metabolic research in animal models. However, they are unsuited for SIRM due to the general non-detectability of radiotracers by NMR, and their hazardous nature which places severe operational constraints for MS. Moreover, they are generally unsuited for in situ human studies. A few studies have explored the use of stable isotope tracers in investigating metabolic perturbations in animal models in response to drug or other treatments (Beger et al. 2009; Artemov et al. 1998; Bhujwalla et al. 1994). To translate detailed model cell findings to understanding of human cancers in situ, it is generally accepted to employ xenograft animal models that more closely approximates the complexity of human tissues, while permitting a wider range of experimental and sampling protocols. The severe combined immunodeficient (SCID) mouse is one such model that enables transformed human cells to be established as xenograft subcutaneously or orthotopically in compatible organs. They are commonly used to investigate the tumorigenicity of human tumor cells and their response to drugs in preclinical studies (PaineMurrieta et al. 1997; Richmond and Su 2008).

The small size (e.g. 20 g) and high metabolic activity of SCID mice hold certain advantages for SIRM, i.e. small amounts of tracers and shorter duration required for extensive labeling in tissues or biofluids. We have previously determined the optimal sampling times for human NSCLC patients following ¹³C₆-Glc infusion i.v. (Fan et al. 2009b). However, the metabolic rate of a 20 g mouse is much faster than a 70 kg human, as indicated by the approximately ¾ power scaling law of body mass (Downs et al. 2008; Glazier 2008; White et al. 2008) as well as the high heart (600 bpm) and respiration rate (200 breaths/min) of mice. Consequently, simply translating sampling times for human to mice based on differences in body mass is not warranted.

Here we report measurements of the time course of ¹³C₆-Glc utilization following a bolus injection through the tail vein and determined the incorporation of ¹³C labels into various metabolites extracted from different tissues and plasma to determine optimal sampling times for maximizing the extent of ¹³C labeling. We also applied the methodologies to an orthotopic xenograft of human lung cancer cells in SCID mice to demonstrate the ability in revealing distinct ¹³C labeling patterns of metabolites in the cancerous lung lobe.

2 Methods

2.1 Mouse handling

The murine tumor xenograft protocols were conducted in compliance with all guidelines and were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Louisville. Fox Chase ICR SCID mice were purchased from Taconic (Hudson, NY) and maintained in a barrier facility at the University of Louisville according to institutional guidelines.

2.1.1 Orthotopic lung tumor xenograft

Human PC14PE6 cells were grown in RPMI medium as previously described (Fan et al. 2005), mixed with matrigel (total 100 µl/10⁶ tumor cells) and injected into the left lung of ketamine-anesthesized mice. An equal volume of normal saline was injected into the right lung. Mice were fed ad libitum, and were monitored daily for signs of distress. Mice were sampled at 10 days post injection. Additional naïve SCID mice were used as controls (Onn et al. 2003).

2.2 ¹³C₆-glucose infusion and plasma sampling

A 20% solution of ¹³C₆-Glc in PBS was sterile-filtered and 100 µl of this solution (i.e. 20 mg of ¹³C₆-Glc) were injected into the tail vein of a restrained mouse without anesthesia. Lower doses (e.g. 2 mg) were also examined for a few mice, which resulted in less ¹³C labeling in tissue metabolites. Approximately 50 µl of blood samples were taken intraorbitally at timed intervals, chilled on ice 5 min after standing at room temperature and separated into plasma and blood cells by centrifugation at 4°C at 3,500×g for 15 min. Plasma was immediately flash-frozen in liquid N₂ for storage prior to metabolite extraction. The labeled glucose time course experiments were performed three times on separate dates, with two mice each for two of the experiments and five mice for the third experiment (see table below for lung cancer xenograft treatment and ¹³C₆-Glc labeling time). Qualitatively similar ¹³C metabolite profiles of individual organs were obtained from these experiments at comparable time points, as estimated from 1-D ¹H-{¹³C}-HSQC spectra (cf. Figs. 6, S3 (in Supplementary material), and data not shown). Due to some variability of mice age, handling, and tumor status, it was impractical to average metabolite values from different individuals.

Exp #	# Mice/xenograft/13C labeling time
1	1/Yes/15 min	1/Yes/30 min
2	1/Yes/15 min	2/No/30 min
3	1/Yes/5 min	1/Yes/15 min	1/Yes/25 min	1/No/15 min	1/No/25 min

Fig. 6.

1-D HSQC spectral comparison of six SCID mouse tissue extracts labeled with ¹³C₆-glucose. SCID mice were infused with ¹³C₆-glucose for 15 min before tissue dissection, extraction, and analysis by NMR as described in “Methods”. The 1-D HSQC spectra of all six tissues were normalized to tissue residue weight (remained after polar and lipid extractions) and spectral parameters so that the intensity or metabolite resonances reflected their tissue content

2.3 Tissue harvest

Mice were killed via decaptitation at different times post ¹³C₆-Glc injection, and the following organs were dissected sequentially: lung, heart, liver, kidney, brain, and thigh muscle. Dissected tissue was flash-frozen in liquid N₂ within 2–6 min of euthanasia.

2.4 Sample preparation

2.4.1 Plasma extraction

Twenty to thirty microliter of plasma was made to 10% trichloroacetic acid (TCA) concentration and centrifuged at 4°C, at ≥22,000×g for 20 min to remove denatured proteins. The polar supernatant was lyophilized to remove TCA before preparation for NMR and GC–MS analysis.

2.4.2 Tissue extraction

Frozen tissues were ground in liquid N₂ to <10 µm particles in a 6750 Freezer/Mill (Retsch, Inc., Newtown, PA) and extracted simultaneously for soluble and lipidic metabolites as follows. Up to 20 mg of frozen tissue powder in 15 ml polypropylene conical centrifuge tube (Sarstedt, Newton, NC) containing three 3 mm diameter glass beads was vigorously shaken in 2 ml of cold acetonitrile (mass spectrometry grade, stored at −20°C) to denature proteins, followed by addition of 1.5 ml nanopure water, and 1 ml HPLC-grade chloroform (Fisher Scientific). The mixture was shaken vigorously until achieving a milky consistency (ca. 5 min), followed by centrifugation at 3,000×g for 20 min at 4°C to separate the polar (top), lipidic (bottom), and tissue debris layers (interface). The polar and lipidic layers were recovered sequentially and the remaining tissue debris (mainly denatured proteins) was extracted again with 0.5 ml chloroform:methanol:butylated hydroxytoluene (BHT) (2:1:1 mM) and centrifuged at 4°C, 22,000×g for 20 min to separate the three phases again. The residual polar and lipid fractions were pooled with respective main fractions. All three fractions were vacuum-dried in a speedvac device (Vacufuge, Eppendorf, New York, NY) and/or by lyophylization. The dry weight of tissue debris was obtained for normalization of metabolite content. The polar extracts were redissolved in 100% D₂O containing 30 nmol perdeuterated DSS (2,2′-dimethyl-2-silapentane-5-sulfonate, Cambridge Isotope Laboratories, Andover, MA) as internal chemical shift and concentration reference for NMR measurement.

2.4.3 NMR spectroscopy

NMR spectra were recorded at 14.1 T on a Varian Inova spectrometer equipped with a 5 mm inverse triple resonance cold probe, at 20°C. 1-D ¹H NMR spectra were recorded with an acquisition time of 2 s and a recycle time of 5 s to minimize peak saturation. 1-D ¹H Spectra were typically processed with zero filling to 131k points, and apodized with an unshifted Gaussian and a 0.5 Hz line broadening exponential. Concentrations of metabolites and ¹³C incorporation were determined by peak integration of the ¹H NMR spectra referenced to the intensity of DSS methyl groups, with correction for differential relaxation, as previously described (Fan and Lane 2008; Fan et al. 2008; Lane et al. 2008).

¹³C profiling was achieved using 1-D ¹H-{¹³C} HSQC experiments recorded with a recycle time of 1.5 s and ¹³C GARP decoupling during the proton acquisition time of 0.15 s. 1-D HSQC spectra were processed with zero-filling to 16k points and apodized using an unshifted Gaussian function and 6 Hz line broadening.

TOCSY and HSQC-TOCSY spectra were recorded with an isotropic mixing time of 50 ms, a B₁ field strength of 8 kHz, and acquisition times of 0.341 s in t₂ and 0.05 s in t₁. The free induction decays were zero-filled once in t₂, and linear predicted and zero filled to 4096 points in t₁. The data were apodized using an unshifted Gaussian and a 1 Hz line broadening exponential in both dimensions. Positional ¹³C incorporation into labeled metabolites was quantified as previously described (Lane and Fan 2007; Lane et al. 2008).

2.4.4 GC–MS analysis

GC–MS analysis was conducted on the same samples as used for NMR, as described previously (Fan et al. 2005). Briefly, following NMR analysis, an aliquot (50–100 µl) of the same extract was re-equilibrated with H₂O and lyophilized to remove deuterium, then silylated with 50 µl 1:1 (v/v) acetonitrile:MTBSTFA (N-methyl-N-[tert-butyldimethylsilyl]trifluoroacetamide) (Regis Chemical, Morton Grove, IL) by 3 h of sonication followed by standing overnight. The solution was directly injected into a PolarisQ GC-ion trap MSn (ThermoFinnigan, Austin, TX) with a 0.5 µl injection volume, using instrument conditions stated previously (Fan et al. 2005). Metabolites were identified and quantified automatically using Xcalibur software (ThermoFinnigan), based on their retention times and mass fragmentation patterns matched against an inhouse database and external standards. Identities were extensively verified by manual inspection. The %RSD for all GC–MS quantification was generally <3% in triplicate analyses for abundant metabolites (≥5 µmol/g dry residue) such as lactate, and increased to 27% with metabolites ≤0.1 µmol/g dry residue.

3 Results and discussion

3.1 Time course changes of plasma ¹³C-glucose and ¹³C-lactate

To assess the optimal time for transforming the tracer ¹³C₆-glucose into downstream metabolites in mice, plasma samples were taken at 5 to 15-min intervals and processed for 1-D ¹H NMR analysis. Figure 1 shows a representative time course of the %¹³C enrichment in glucose (solid symbols) and lactate (open symbols) in the plasma of two SCID mice. The initial glucose enrichment ranged from 30 to 35% depending on the size of the mouse, suggesting that the plasma glucose concentration was increased by about 50% immediately after the bolus tracer injection. The % enrichment for glucose decreased rapidly and asymptotically within 1 h, presumably via the normal homeostatic mechanism.

Fig. 1.

Glucose consumption and lactate production in two SCID mice. Plasma processing and metabolite analysis by 1-D ¹H NMR were described in “Methods. %¹³C enrichment was calculated from the intensity of ¹³C satellites and central peaks of H-3 of lactate as described in Lane et al. (2008). Filled circle, filled square: Glucose; open circle, open square: Lactate

Once taken up by tissues, labeled glucose was metabolized principally via glycolysis, which generated labeled lactate and to a lesser extent labeled alanine. Lactate is an important metabolite involved in pH and energy balance (Lane et al. 2009b). Cellular lactate was then exported into the blood and ultimately to the liver to be reconverted into glucose by gluconeogenesis (i.e. the Cori Cycle). ¹³C-lactate in plasma therefore represents lactate newly synthesized since the ¹³C glucose bolus, and is an indicator of systemic metabolism. This is consistent with the observed ¹³C enrichment in lactate (Fig. 1), which started low, rapidly reached a maximum, and then depleted over longer periods, reflecting the decrease in the enrichment of the source glucose. The peak enrichment in lactate approached 12% during 15–20 min post injection in this experiment. A larger bolus in which the initial ¹³C glucose enrichment reached 65% was associated with a higher peak ¹³C lactate level (data not shown). These data determined on average the optimal duration of tissue harvest for SIRM analysis to be 15–20 min post tracer infusion.

3.2 Tissue-dependent metabolism in six mouse tissues

3.2.1 GC–MS analysis

Individual tissues may absorb and metabolize glucose at different rates, which will be reflected in the time-dependent distribution of ¹³C labeled metabolites in various tissues. Figure 2 shows the GC–MS analysis of ¹³C-lactate isotopologue content of six different tissues dissected from three SCID mice, each obtained 5, 15, and 25 min after injection of the ¹³C₆-glucose bolus. The ¹³C₃-lactate (lactate+3) isotopologue was produced at the highest level in brain, followed by lung and kidney after only 5 min of glucose metabolism. Since ¹³C₃-lactate is principally a product of glycolysis, it is expected that the highly glycolytic brain tissue would show the highest initial production of this isotopologue and maintenance of the initial level thereafter. In comparison, the ¹³C₃-lactate level declined in lung and kidney while it peaked in heart and liver after 15 min of labeled glucose injection (Fig. 2). The time course of ¹³C₃-lactate production in lung and kidney tracked closely with that of the plasma ¹³C₆-glucose level (cf. Fig. 1), which could reflect a high rate of glucose oxidation coupled with a high rate of lactate consumption and/or export in these two organs (Bartlett et al. 1984; Longmore and Mourning 1976). Also notable is the delayed but very high production of ¹³C₃-lactate in the heart, which could reflect its high-energy demand from contraction and preference for drawing energy from β-oxidation of fatty acids over glucose oxidation (Khairallah et al. 2004). Lactate has been shown to be both metabolized in the mitochondria and released by perfused mouse heart (Khairallah et al. 2004). Similar to the brain, muscle tissue maintained a constant but much lower production of ¹³C₃-lactate from ¹³C₆-glucose, which could be related to a significant contribution of unlabeled glycogen metabolism to lactate production in muscle (Woods and Krebs 1971).

Fig. 2.

Representative GC–MS analysis of ¹³C-isotopologue series of lactate in six different SCID mouse tissues. The six tissues were obtained from three SCID mice, injected with ¹³C₆-Glc and dissected 5, 15, and 25 min thereafter. Their polar extracts were prepared and analyzed by GC–MS as described in “Methods”. The values displayed are the ¹³C enrichment values for the tissue extracts in µmol/g dry residue weight. The dry residue weight was obtained from tissue residues remained after polar and lipid extractions. Lactate+1 to +3 are singly, doubly, and triply ¹³C-labeled isotopologues of lactate

Furthermore, readily detectable amounts of singly and doubly ¹³C labeled lactate (lactate+1 and +2 or ¹³C₁- and ¹³C₂-lactate) was observed in all six tissues (Fig. 2). ¹³C₁- and ¹³C₂-lactate was present at comparable levels within 5 min of tracer injection and the change in their levels thereafter was small in brain, heart, and kidney. In contrast, ¹³C₁-lactate level peaked in liver, lung and muscle after 15 min of tracer introduction and declined sharply thereafter. ¹³C₂-lactate level also peaked after 15 min of metabolism in liver but changed little with time in muscle and lung. Neither ¹³C₁-lactate nor ¹³C₂-lactate can be produced from ¹³C₆-glucose via glycolysis alone. Their production requires metabolic scrambling first through the non-oxidative branch of the pentose phosphate pathway (PPP) (Fig. 3a). Alternatively, in gluconeogenic tissues (i.e. liver and kidney), ¹³C₁- and ¹³C₂-lactate can also be produced from ¹³C₆-glucose via the sequence of glycolysis, Krebs cycle, gluconeogenesis, and glycolysis again (Fig. 3b). Thus, the initial buildup of the scrambled lactate isotopologues in non-gluconeogenic tissues (brain, heart, lung, and muscle) could reflect the carbon flow through the PPP while the subsequent depletion in lung and muscle could be attributed to a combination of labeled glucose depletion in the plasma and lactate export or metabolism. The transient buildup of scrambled lactate in liver could result from PPP and gluconeogenic activity, in part using labeled lactate imported from the blood (Cori cycle).

Fig. 3.

Tracking of ¹³C atoms from ¹³C₆-glucose to ¹³C-lactate through glycolysis, pentose phosphate pathway, Krebs cycle, and gluconeogenesis. a The ¹³C flow through the non-oxidative branch of PPP and glycolysis in the presence of ¹³C₆-glucose + unlabeled glucose, b tracks ¹³C atoms from glycolysis, Krebs cycle, gluconeogenesis, and glycolysis again, and c traces ¹³C flow in the 2nd turn of the Krebs cycle, where ¹³C-OAA is derived from (b) after the 1st turn and acetyl CoA is generated from unlabeled glucose via glycolysis. Black dots represent unlabeled carbons while color dots denote ¹³C. The pink and green dots in (b) after the SCS step illustrates scrambled ¹³C. Dashed black arrows denote multiple reactions steps, double-headed black arrows indicate reversible reactions, and dashed green arrows denote site of carbon–carbon bond break. PDH pyruvate dehydrogenase, SCS succinyl CoA synthetase, PEPCK phosphoenolpyruvate carboxykinase, OAA oxaloacetate, α-KG α-ketoglutarate, PEP phosphoenolpyruvate. (Color figure online)

In addition to lactate, ¹³C-labeled isotopologue series for a number of metabolites were quantified by GC–MS analysis. The time course changes in Ala, succinate, Asp, Glu, Gln are shown in Fig. 4, while those for fumarate, malate, citrate, GAB, Ser, Asn, and Gly are illustrated in Fig. S2 (Supplementary material). The time course of ¹³C labeling for Ala was generally similar to that of lactate for all tissues but lung, muscle, and kidney. Most notably, the level of the three ¹³C isotopologues of Ala in lung peaked at 15 min, in contrast to the ¹³C₃-lactate level in lung which began to decline 5 min after ¹³C₆-glucose injection. Both labeled Ala and lactate share the same precursor, i.e. labeled pyruvate derived from glycolysis and/or PPP. Yet they differed in their time course behavior in the lung and muscle. This implied the presence of two separate pools of pyruvate in the two tissues each leading to lactate and Ala synthesis. A similar observation was made for human lung cancer tissues (Fan et al. 2009b). As described for lactate, the scrambled ¹³C₁- and ¹³C₂-Ala reflected transformations of ¹³C₆-glucose via PPP and/or gluconeogenesis.

Fig. 4.

Representative GC–MS analysis of ¹³C-isotopologue series of selected metabolites in different SCID mouse tissues. The same polar extracts as in Fig. 2 were analyzed by GC–MS and quantified for ¹³C isotopologues for various metabolites. The values displayed are the ¹³C enrichment values for the tissue extracts in µmol/g dry residue weight, as in Fig. 2. Metabolite+1 to 5 are singly, doubly, triply, quadruply, and quintuply ¹³C-labeled isotopologues

The ¹³C-isotopologue series of succinate, Asp, Glu, and Gln were clearly present. Some of these reached relatively high levels, e.g. ¹³C₁/¹³C₂-Asp in brain, ¹³C₁-Asp in lung, ¹³C₁-/¹³C₂-/¹³C₃-Glu in brain, ¹³C₁-/¹³C₂-Glu in kidney, ¹³C₁-Glu in lung, ¹³C₁-/¹³C₂-Glu in liver, ¹³C₁-/¹³C₂-Glu in heart, ¹³C₁-/¹³C₂-Gln in brain heart, and liver, ¹³C₁-/¹³C₂-/¹³C₃/¹³C₄-succinate in liver, as well as ¹³C₁-/¹³C₂-succinate in kidney, brain, and heart. ¹³C₂-succinate, ¹³C₂-Asp, ¹³C₂-Glu, and ¹³C₂-Gln can be derived from ¹³C₆-glucose via glycolysis plus the 1st turn of the Krebs cycle (cf. Fig. 3b). These doubly labeled isotopologues along with the ¹³C₁-isotopologues could also be produced from a second turn of the Krebs cycle according to the scheme in Fig. 3c, where the starting ¹³C₂-oxaloacetate (OAA) is derived from ¹³C₆-Glc but acetyl CoA is produced from unlabeled glucose. Based on the dominance of the ¹³C₁-isotopologues of Asp, Glu, and Gln (branched metabolites of the Krebs cycle) in most tissues, a major fraction of the labeled glucose had undergone at least two turns of the cycle. In fact, some ¹³C₆-Glc may have passed through a third turn, as evidenced by the production of ¹³C₁-citrate (Fig. S2, Supplementary material), following the same scheme as in Fig. 3c. The unequal synthesis of ¹³C₁ and ¹³C₂ isotopologues of Asp, Glu, and Gln in most cases suggests additional source(s) of labeled carbon input into the Krebs cycle, e.g. ¹³C₁-acetyl CoA (derived from scrambled labeled lactate in Fig. 3a, b) plus unlabeled OAA will lead to the production of ¹³C₁-Glu, -Gln, and -Asp only.

The unusual abundance of ¹³C₁/¹³C₂-Asp, ¹³C₁/¹³C₂-Glu, and ¹³C₁/¹³C₂-Gln in the brain suggests a high flux through the Krebs cycle for the production of neurotransmitters (Glu, Gln in Fig. 4; GAB in Fig. S2, Supplementary material). GAB is produced from Glu via Glu decarboxylase activity. There was also an appreciable presence of ¹³C₃-/¹³C₄-Asp, ¹³C₃-/¹³C₄-Glu, and ¹³C₃-/¹³C₄-succinate in all six tissues (Fig. 4), which could be produced via the second and third turn of the Krebs cycle with input of both labeled acetyl CoA and OAA (cf. Fan et al. 2010), as opposed to Fig. 3c where only starting OAA is ¹³C-labeled. However, ¹³C₃-Asp could also be derived from the carboxylation of ¹³C₃-pyruvate (PC) via the anaplerotic pyruvate carboxylase activity (Fan et al. 2009b, 2010). PC coupled with the first turn of the Krebs cycle would lead to the production of ¹³C₃-citrate while ¹³C₄- and ¹³C₅-citrate would be the expected products from the second and third turn of the Krebs cycle activity, respectively (Fan et al. 2010). ¹³C₃-citrate was largely present at a higher level than those of ¹³C₄- and ¹³C₅-citrate in all six tissues (Fig. S2, Supplementary material), which suggests that PC contributed substantially to the production of ¹³C₃-Asp in these tissues. The relatively high PC activity in the brain, heart, and liver, where PC is strongly expressed (Jitrapakdee et al. 2008) is supported by the high level of ¹³C₃-Asp (Fig. 4) and abundance of ¹³C-2-Glu and ¹³C-3-Glu/Gln (cf. Fig. 6), which can be derived from ¹³C₆-glucose via glycolysis, PC, and the first turn of the Krebs cycle (Fan et al. 2010).

3.2.2 NMR analysis

The same set of extracts shown in Fig. 2 were analyzed by NMR to complement and verify the GC–MS analysis. The ¹³C enrichment in glucose/glucose-6-phosphate (not resolved) and lactate was determined by 1-D ¹H NMR. The enrichments from 5 to 25 min post infusion were substantial in all tissues, except for glucose and lactate in kidney where ¹³C enrichment was not determined after 25 min of metabolism due to interferences (Table 1). The peak % lactate enrichments differed presumably according to the metabolic activity and/or glucose uptake rates of the six tissues measured (Table 1).

Table 1.

%¹³C enrichment in tissue lactate and glucose

Tissue	Time (min)	%13C lactate	%13C glucose
Brain	5	47	81
	15	47	45
	25	47	54
Heart	5	25	48
	15	44	23
	25	47	36
Kidney	5	36	59
	15	18	38
	25	ND	ND
Liver	5	39	20
	15	40	8
	25	53	29
Lung	5	27	23
	15	28	29
	25	39	32
Muscle	5	33	47
	15	50	24
	25	61	37

%¹³C in lactate was calculated from the intensity of ¹³C satellites and central peak of H-3, while %¹³C in glucose was similarly calculated based on the intensity of ¹³C satellites and central peak of H-1α measured from 1-D ¹H NMR spectra, as described in Lane et al. (2008). ND not determined. Estimated errors were ±2 to 5%

More extensive ¹³C labeling patterns in metabolites was determined by two-dimensional NMR experiments. Figure 5 shows a high-resolution 2-D ¹H–¹³C HSQC-TOCSY (heteronuclear single quantum coherence-total correlation spectroscopy) spectrum (Fig. 5b) of a lung extract obtained from a SCID mouse 15 min after injecting ¹³C₆-glucose. The spectrum was acquired with sufficient digital resolution to resolve ¹³C coupling patterns of labeled resonances, which can be more clearly visualized in the 1-D projection spectrum onto the ¹³C dimension (Fig. 5a). The HSQC-TOCSY data not only confirmed the identity of ¹³C-labeled metabolites from their characteristic ¹H–¹³C and ¹H–¹H covalent linkages but also enabled determination of the labeled carbon position, i.e. positional isotopomers (Fan and Lane 2008). For example, lactate was confirmed by the covalent linkages represented by cross-peaks from H-3 to C-3, H-2 to C-2, H-3 to H-2, and H-2 to H-3 of lactate (traced by red rectangle in Fig. 5b). The ¹³C-coupling pattern of C-3 and C-2 of lactate was respectively doublet and triplet (Fig. 5a), indicating that lactate was largely labeled at all three carbon positions. This is consistent with the abundance of ¹³C₃-lactate by GC–MS analysis of the same extract (cf. Fig. 2). Similarly, the doublet pattern of C-1 and C-6 and triplet pattern of C-2, C-3, and C-4 of glucose-6-phosphate (Glc-6-P) indicate the presence of ¹³C₆-Glc-6-P in the extract, which was derived from ¹³C₆-glucose via hexose kinase or glucokinase activity. This information and the labeled patterns of glutathione, adenine nucleotides, and UDP-sugars were not obtainable by GC–MS. Moreover, the N-methyl carbons of phosphocholine and C-2 of Gly were singlets, which suggests that they were largely contributed from the natural abundance ¹³C, and therefore not derived from ¹³C₆-glucose.

Fig. 5.

2-D ¹H–¹³C HSQC-TOCSY analysis of SCID mouse lung extracts. ¹³C₆-glucose was infused into SCID mouse via tail vein and lung tissue was dissected, pulverized, and extracted as described in “Methods”. The 2-D spectrum, displayed as a contour plot (b), was acquired at 14.1 T, processed with linear prediction in the ¹³C dimension and zero-filling to 4k × 2k real digital points. Also displayed is the 1-D projection spectrum along the ¹³C dimension (a). The acquisition time in t₁ (50 ms) was sufficient to resolve ¹J_CC scalar couplings (40–50 Hz) in the ¹³C dimension. Rectangular boxes trace some of the covalent linkages represented by the 2-D cross-peaks. GSH reduced glutathione, Glc glucose, PCr-NMe, PC-NMe N-methyl carbon of phosphocreatine or phosphocholine, respectively, AXP adenine nucleotides

Once the labeled metabolites were assigned by 2-D HSQC methods, the 1-D HSQC analysis of ¹H directly attached to ¹³C provided a semi-quantitative overview of the abundance of various ¹³C-labeled metabolites in the six tissues (Fig. 6). It is clear that different tissues exhibited rather distinct patterns of labeled glucose metabolism, as reflected by the different 1-D HSQC profiles. Brain tissue built up a considerable pool of ¹³C-3-Asp, ¹³C-2, 3, or 4-Glu/Gln and ¹³C-2, 3, or 4-γ-aminobutyrate (GAB), which is consistent with its high Krebs cycle activity and specialized production of neurotransmitters Glu, Gln, and GAB, as evident also from the GC–MS data (Fig. 4). There was no detectable free ¹³C-labeled glucose in the brain tissue, which could be related to its high glycolytic activity (cf. Fig. 2) and high requirement for glucose in energy production (Owen et al. 1967). Heart tissue was also high in labeled Glu, Gln and succinate, which reflects a high Krebs cycle activity. It also had the highest labeled lactate level among the six tissues and an appreciable level of labeled Glc-6-P, which suggests a high glucose uptake and glycolytic rates. Liver tissue had the highest Glc-6-P but a moderate level of labeled lactate, which could be a result of fast glucose uptake and gluconeogenesis from lactate. Liver was the only tissue where labeled glycogen was observed, which could be attributed to its high capacity of glycogen synthesis (Hers 1976). Muscle tissue had the second highest labeled lactate level, which is consistent with a high rate of glycolysis. In addition, it was the only tissue where the ribose moiety of adenine nucleotides (AXP) was labeled, which suggests that muscle possesses high PPP activity. This is consistent with the high levels of scrambled labeled lactate in the muscle tissue as revealed by GC–MS analysis (cf. Fig. 2). Lung and kidney tissues had the lowest overall labeled metabolite content, which suggests a lower rate of glucose metabolism.

It should be noted that the above-described ¹³C-labeling patterns of metabolites were acquired from SCID mice after a brief period (<5 min) of restraint for ¹³C-glucose injection. Prolonged animal handling is known to cause tachycardia and release of stress hormones into the plasma of rodents (Kawashima et al. 1985; Roizen et al. 1978), which could in turn affect metabolic activity of various organs, such as glycogen degradation in liver (Gruetter et al. 1994), synaptic release of amino acids in brain (Timmerman et al. 1999), and energy metabolism in brain (Chance et al. 1978). Therefore, handling stress-induced alteration of ¹³C-glucose metabolism may contribute to the pattern of observed labeled products in Figs. 2, 4, 5, 6, 7, 8. Particularly notable is the lack of ¹³C-glucose and the considerable buildup of ¹³C-lactate in our mouse brain extracts (cf. Figs. 6, S3 (Supplementary material)), which is distinct from that observed in situ in human or rat brain (Gruetter et al. 2003). Nevertheless, it is reasonable to conclude that the glucose-dependent metabolic activities differed among the various tissue types of mice.

Fig. 7.

1-D HSQC spectral comparison of tumorous versus paired normal lung. A SCID mouse was injected with 1 million PC14PE6 lung cancer cells in matrigel in one of the lung lobes while the other lung lobe received saline only. After lung tumor establishment, the animal received ¹³C₆-glucose (Glc) 15 min prior to tissue dissection, extraction, and NMR analysis as described in “Methods”. The 1-D HSQC spectra were normalized to tissue residue weight (remained after polar and lipid extractions) and spectral parameters so that the intensity or metabolite resonances reflected their tissue content. NMe-PCholine N-methyl resonance of phosphocholine

Fig. 8.

Representative GC–MS quantification of ¹³C-isotopologue series of metabolites in tumorous and normal lung tissues of SCID mice. Tumorous and normal lung tissues were obtained from two SCID mice, similarly treated as in Fig. 7. Their polar extracts were analyzed by GC–MS as described in “Methods”. The values displayed are the difference in ¹³C enrichment between tumorous and normal lung tissue extracts in µmol/g dry residue weight. The dry residue weight was obtained from tissue remained after polar and lipid extractions. Metabolite+1 to 5 are singly, doubly, triply, quadruply, and quintuply ¹³C-labeled isotopologues

3.3 Metabolism in normal versus tumorous lung tissues

¹³C₆-glucose was used as a tracer to track changes in metabolic pathways associated with lung tumor development in SCID mice. Figure 7 compares 1-D HSQC NMR spectra of a tumorous lung with its paired normal lung dissected from the same mouse 15 min after the tracer infusion. It is clear that the ¹³C abundance of various positional isotopomers of metabolite (as reflected by the resonance intensity of their protons attached to ¹³C) were higher in tumorous lung than its normal counterpart. These included ¹³C-2 and 3-lactate, ¹³C-3-Ala, ¹³C-3 and 4-Pro, ¹³C-4-Glu, ¹³C-4-Gln, ¹³C-4-glutamyl moiety of GSH (¹³C-4-Glu-GSH), ¹³C-2,3-succinate, ¹³C-2-Gly, ¹³C-1 to 6-glucose (Glc), and N-methyl-phosphocholine (NMe-PCholine). The NMe-PCholine signal was attributed to natural abundance of ¹³C since the choline moiety is not expected to be synthesized from ¹³C₆-glucose.

A separate set of tumorous and normal lung tissue extracts similarly prepared as in Fig. 7 were quantified by GC–MS, as shown in Figs. 8 and S4 in the Supplementary material. The GC–MS quantification agreed with the NMR analysis in terms of the greater production of ¹³C labeled lactate, Ala, succinate, fumarate, Asp, Glu, Gly, and Ser in the tumorous than in normal lung tissues. Some of these metabolites, i.e. ¹³C-labeled malate, fumarate, and Ser were obscured by others in the 1-D HSQC spectra in Fig. 7 but were clearly quantified by GC–MS. Moreover, as described above, increased carbon flow through specific pathways in the tumorous lung could be deduced based on the combined ¹³C mass isotopologue series of metabolites acquired from GC–MS and the ¹³C positional isotopomer information obtained from the NMR analysis. For example, enhanced buildup of ¹³C-glucose in the tumorous lung (Fig. 7) may be related to increased glucose uptake. The increased production of ¹³C₃-lactate and -Ala is consistent with enhanced glycolysis while the larger buildup of ¹³C₁ and ¹³C₂-lactate and -Ala suggests more active PPP in the tumorous lung. Enhanced Krebs cycle activity was evident from the greater accumulation of ¹³C₁- and ¹³C₂-malate, -succinate, -fumarate, and -Glu in the tumorous lung (cf. Fig. 3b, c schemes). Activation of PC in the tumorous lung tissue could be inferred from the increased buildup of ¹³C₃-Asp (Fig. 8) and the presence of ¹³C₃-citrate isotopologue (data not shown), which is a unique product of PC. The higher synthesis of ¹³C₂-Gly in the tumorous lung suggests a more active one-carbon metabolism associated with tumor development.

The buildup of ¹³C-labeled glucose and lactate was previously observed by in vivo NMR in RIF-1 tumor xenograft in mice under relatively fast and slow infusion of ¹³C-1-Glc and anesthesia (Artemov et al. 1995, 1998). However, it was concluded in the RIF-1 tumor study (Artemov et al. 1998) that no intracellular glucose or glycolytic products were observable in vivo, which implies that the ¹³C-glucose and lactate NMR signals were originated from the blood. In the case of lung tumors resected from human patients (Fan et al. 2009a, b), glucose levels were also very low in tumors with poor vascularization. Therefore, it is possible that the accumulation of glucose in the lung tumor xenograft in the present study may have appreciable contribution from the blood. The same situation can be said for lactate accumulation. However, we note that the fractional ¹³C enrichment of lactate in the blood was considerably lower than that in the tissue (cf. Table 1, Fig. 1) from the same animals, suggesting that a substantial fraction of the observed metabolism was from the target tissue. As in vivo lactate level can increase significantly by reducing blood flow to the tumor (Bhujwalla et al. 1994), additional lactate buildup in our excised lung tumor could also be attributed to the time it took to dissect (i.e. 2–6 min) before liquid N₂ freezing. However, in contrast to the previous in vivo analysis of RIF-1 tumors (Artemov et al. 1998), the excised tumorous lung in this study contained an appreciable amount of ¹³C labeled Krebs cycle intermediates and Ala (a glycolytic product), suggesting a more active Krebs cycle and glycolytic activity. This distinction may be attributed to a much smaller (1–2 mm) and thus earlier stage of tumor, different tumor type (orthotopic adenocarcinoma instead of ectopic fibrosarcoma), mouse strain (SCID instead of C₃H/HeN), and/or animal handling (restraint during labeled glucose administration instead of anesthesia) in the present study. As the tumor burden was small, the true metabolic differences between the tumor-bearing lung and the paired nontumorous lung are presumably underestimated.

In conclusion, differential rates of glucose transformations through various metabolic pathways in six different tissues plus blood plasma from SCID mice can be discerned with tail-vein injection of a bolus of ¹³C₆-Glc. The ¹³C labeled patterns of metabolites in terms of isotopologues and isotopomers obtained from GC–MS and NMR analyses enabled reconstruction of multiple pathways including glycolysis, pentose phosphate pathway, multiple turns of the Krebs cycle, anaplerotic pyruvate carboxylation, Gly biosynthesis, nucleotide biosynthesis, and gluconeogenesis. When applied to human lung cancer cell xenograft in SCID mice, enhanced glucose uptake, glycolysis, PPP, Krebs cycle, pyruvate carboxylation, and Gly biosynthesis were found to be associated with tumor development, which is analogous to metabolic distinctions observed in human lung cancer tissues (Fan et al. 2009b).

Abbreviations

BHT

Butylated hydroxytoluene

DSS

2,2′Dimethylsilapentane-5-sulfonate

MTBSTFA

N-methyl-N-[tert-butyldimethylsilyl]trifluoroacetamide

NSCLC

Non small cell lung cancer

SIRM

Stable isotope resolved metabolomics