[U-¹³C]glucose infused ccRCC tumors show differential labeling relative to adjacent kidney

A schematic of labeling from [U-¹³C]glucose is shown in Figure 2A, and a representative ¹H decoupled ¹³C NMR spectrum of metabolites extracted from a ccRCC tumor is shown in Figure 2B. The ccRCC spectrum is dominated by the carbon 2–3 doublet at the lactate C3 resonance (Figure 2B), as expected for glycolytic tissue. The prominence of lactate signal in the ccRCC spectrum is consistent with previously published reports of high lactate levels in ccRCC tumors (Hakimi et al., 2013; Sudarshan et al., 2013). The prominent 2–3 doublet in alanine C3 is also consistent with anaerobic glucose metabolism, when pyruvate is converted to alanine through a transamination reaction catalyzed by alanine aminotransferase (ALT) (Figure 2B). Other features in the ccRCC spectrum include a small C4–5 doublet in glutamate C4 (i.e. Glu4D45; Figure 2B inset). This doublet arises from the PDH-dependent production of [1,2-¹³C]acetyl-CoA followed by entry of this acetyl-CoA into the TCA cycle and conversion to [4,5-¹³C]α-ketoglutarate and [4,5-¹³C]glutamate (Figure 2A). Some spectra revealed small quartets in glutamate C4 (Glu4Q); this feature arises when [4,5-¹³C]α-ketoglutarate proceeds around the TCA cycle to oxaloacetate, followed by condensation with [1,2-¹³C]acetyl-CoA and conversion to [3,4,5-¹³C]α-ketoglutarate and [3,4,5-¹³C]glutamate. The ratio of the GluQ to GluD45 provides a surrogate for enrichment in the acetyl-CoA pool used to supply the TCA cycle. The estimated acetyl-CoA enrichment in tumors infused with [U-¹³C]glucose was less than 5% (Figure 2C), consistent with low levels of PDH activity in the tumors. In contrast, glioblastoma tumors had an average fractional enrichment of 11% in the acetyl CoA pool when infused with [U-¹³C]glucose (Maher et al., 2012). Despite detectable multiplets in glutamate C3 and C4, essentially no label was detected in glutamine, indicating a paucity of glutamine synthetase activity in the tumors, another marked difference from human tumors in the brain (Maher et al., 2012).

Open in a separate window

Figure 2:

In vivo labeling from ccRCC tumors and adjacent kidney infused with [U-¹³C]glucose.

(A) Schematic for metabolism of [U-¹³C]glucose. “Labeled” ¹³C carbons are colored circles and “unlabeled” ¹²C carbons are white circles. Labeling in the first turn of the TCA cycle are blue circles. Labeling in the second turn of the TCA cycle are light blue circles. The positional carbons notated on the NMR spectrum are labeled in the metabolites glutamate and lactate. (B) Representative ¹H-decoupled ¹³C NMR tumor spectrum from patient RCC2. The inset is a magnification of the peaks from glutamate (Glu) at carbon 4 (C4). The number after the indicated metabolite refers to the carbon position in the metabolite. D45 refers to the doublet formed from glutamate labeled at both C4 and C5. Q refers to the quartet formed from glutamate labeled at C3, C4, and C5. Glu = Glutamate, Gln = Glutamine, Glu = Glutamate, Lac = Lactate, Ala = Alanine. (C) Fractional enrichment of labeled acetyl CoA from each patient calculated from glutamate labeling on NMR. The Acetyl CoA fractional enrichment from glioblastoma (GBM) patients published by Maher et al 2012., is shown for comparison. (D) Total labeling of indicated metabolite as analyzed by mass spectrometry and normalized to ¹³C Glucose enrichment in patient plasma. A full isotopologue distribution is provided in Supplemental Table 1. P values: ns = P > 0.05; * =P ≤ 0.05; ** = P ≤ 0.01; *** = P ≤ 0.001; **** = P ≤ 0.0001. Unpaired t tests were used in panel D, and the data is shown as mean ± standard deviation

Mass spectrometry analysis of the same samples provided a broader view of metabolite labeling. Mass spectrometry was used to analyze metabolism in both the tumors and adjacent kidney. For this analysis, we considered all labeled forms of each metabolite together (i.e. [1.0-(m+0)]), and normalized this overall enrichment value to plasma glucose enrichment from the corresponding patient at the time of resection. Complete analysis of all isotopologues from each tissue fragment is in Supplemental Table 1. Labeling of all measured glycolytic intermediates in the tumors exceeded labeling in adjacent kidney, whereas labeling of metabolites related to the TCA cycle was generally suppressed (Figure 2D). This analysis indicates that human primary ccRCCs have prominent glycolytic metabolism in vivo, but that labeling of TCA cycle intermediates is suppressed, consistent with low activity of PDH and glucose oxidation.

Primary human ccRCC tumors are metabolically distinct from human tumors in the brain and lung

Previous studies published by our group have shown that primary brain tumors (anaplastic ependymoma, anaplastic astrocytomas and glioblastomas), brain metastases and primary non-small cell lung cancers (NSCLC) (mostly adenocarcinoma), have significant levels of glucose oxidation (Hensley et al., 2016; Maher et al., 2012). These patients were studied using a similar [U-¹³C]glucose infusion protocol to the one employed here in ccRCC. As a result, they provide an opportunity to compare metabolism among these three anatomic locations. We used previously reported mass spectrometry analysis from 31 non-small cell lung cancer patients (Faubert et al., 2017; Hensley et al., 2016), and performed new mass spectrometry analysis on samples from five brain tumor patients (primary and metastatic lesions; see Supplemental Tables 2 and 3) previously analyzed only by NMR (Maher et al., 2012). The ccRCC tumors contained markedly enhanced relative enrichment of 3-phosphoglyerate, pyruvate and lactate relative to both the brain and lung tumors (Figure 3A, 3B). For 3-phosphoglycerate and pyruvate, the enrichment normalized to plasma glucose was approximately twice as high in ccRCC (Figure 3A). Lung and brain tumors were similar to each other by this labeling metric. In terms of relative labeling of TCA cycle intermediates from [U-¹³C]glucose, tumors in the brain had the highest enrichment and primary ccRCC had the lowest, while lung tumors had an intermediate labeling phenotype (Figure 3A). To better visualize labeling differences between glycolytic and TCA cycle intermediates among these tumors, labeling in all metabolites was normalized to 3-phosphoglycerate (Figure 3B). In ccRCC, but not in brain or lung tumors, this analysis revealed a sharp decline in labeling between pyruvate and citrate, consistent with reduced contribution of glucose to the TCA cycle in ccRCC.

Open in a separate window

Figure 3:

Glucose labeling profiles of tumors from the brain, kidney, and lung.

(A) Total labeling of each metabolite normalized to ¹³C-Glucose enrichment in plasma. (B) Total labeling mean of metabolites indicated in panel A divided by the 3-PG total labeling mean. 3-PG means are 33.94% for brain, 40.56% for lung, 73.96% for kidney tumors. (C) Schematic illustrating labeling in citrate from [U-¹³C]glucose via pyruvate dehydrogenase (PDH) and pyruvate carboxylase (PC) in the first turn of the TCA cycle. (D) PDH and PC activity. Citrate m+2/pyruvate m+3 ratio can be a surrogate for PDH while the citrate m+3/pyruvate m+3 ratio can be a surrogate for PC. (E) Lactate/3-phosphoglycerate (Lactate/3-PG) ratios for adjacent kidney, ccRCC tumors, brain tumors and lung tumors. Lung tumors shown to utilize lactate had lactate/3-PG ratios above 1.5 (Faubert et al., 2017). P values: ns = P > 0.05; * =P ≤ 0.05; ** = P ≤ 0.01; *** = P ≤ 0.001; **** = P ≤ 0.0001. Statistical tests of two way ANOVA with Tukey posthoc test was used in 3A, unpaired t tests were used in 3D, and ordinary one way ANOVA was used in 3E. The data is shown as mean ± standard deviation

We next examined the isotopologue distribution in citrate to obtain a more granular view of TCA cycle activity in the tumors. A schematic showing the contribution of [U-¹³C]glucose derived pyruvate to the TCA cycle through either pyruvate carboxylase (PC) or pyruvate dehydrogenase (PDH) is shown in Figure 3C. The citrate m+2/pyruvate m+3 ratio can serve as a surrogate for PDH while the citrate m+3/pyruvate m+3 ratio can be a surrogate for PC (Comte et al., 1997; Panchal et al., 2000). Both of these ratios are reduced in the ccRCC tumors compared to adjacent kidney (Figure 3D), indicating reduced contribution of glucose-derived carbon to the TCA cycle via both PDH and PC.

Our findings support the hypothesis that decreased PDH activity in ccRCC leads to suppressed glucose oxidation. Carbon flow into the TCA cycle is determined in large part by activity of the PDH complex, which is regulated by inhibitory phosphorylation via PDH kinases (PDKs). In cell culture models, PDK1 is a transcriptional target of HIF1α and its expression is enhanced upon VHL loss (Kim et al., 2006; Papandreou et al., 2006). Expression of PDK1 and suppression of PDH is predicted to result in reduced enrichment of the citrate m+2 isotopologue compared to pyruvate m+3 as shown in Figure 3D, and subsequent decrease in glucose derived intermediates contributing to the TCA cycle.

Several papers have recently reported the use of lactate from the plasma as a fuel for tumors (Faubert et al., 2017; Hensley et al., 2016; Hui et al., 2017; Kennedy et al., 2013). Direct evidence for lactate utilization by human lung tumors was observed after intra-operative infusions with ¹³C-lactate (Faubert et al., 2017). Although we did not infuse ¹³C-lactate into patients with ccRCC or brain tumors, we did examine the ratio of labeling in lactate to 3-phosphoglycerate (lactate/3-PG) during infusions with [U-¹³C]glucose, because this ratio may provide a surrogate for entry of plasma lactate. If [U-¹³C]glucose is converted to [U-¹³C]lactate outside the tumor, this labeled lactate can enter the circulation and eventually contribute to ¹³C labeling in tumor lactate and other metabolites. As a result, lactate/3-PG labeling ratios above 1.0 are consistent with entry of ¹³C-lactate from the circulation. While lactate/3-PG ratios are an imperfect marker of the ability of tumors and tissues to use lactate, and ratios of ~1.0 do not rule out lactate utilization (Faubert et al., 2017; Hui et al., 2017), a subset of human lung tumors have ratios exceeding 1.5. In brain tumors, the ratio was heterogeneous but had an average lactate/3-PG ratio of approximately 2.0, strongly suggesting that these tumors metabolize lactate as a carbon source. ccRCC tumors had an average ratio lower than the normal kidney tissue, 1.13 and 1.31 respectively (Figure 3E), arguing against robust import of extracellular lactate by ccRCC tumors. These data indicate that lactate may provide a fuel for human tumors in the brain, but is unlikely to be a prominent fuel in ccRCC.

[1,2-¹³C]Acetate infusion in ccRCC shows evidence for low TCA cycle turnover

We also infused one ccRCC patient with [1,2-¹³C]acetate (Figure 4A). The rationale for this infusion was two-fold. First, acetate can be used as a fuel when glucose contribution to acetyl-CoA is sub-maximal in human tumors (Mashimo et al., 2014). Acetate is converted to acetyl-CoA by acetyl-CoA synthetases, and then may contribute to the TCA cycle, lipogenesis, and other metabolic processes (Comerford et al., 2014; Mashimo et al., 2014; Schug et al., 2015). Second, any delivery of acetate-derived acetyl-CoA to the TCA cycle would provide a complementary view of cycle turnover independent of reduced PDH activity (Figure 4B). [1,2-¹³C]acetate infusion of the single patient achieved high enrichment in the acetyl-CoA pool and excellent signal of [4,5-¹³C]glutamate by NMR. This indicates that this tumor took up acetate, converted it to acetyl-CoA, and used it to supply the first segment of the TCA cycle. Areas from these multiplets allowed us to estimate that 34% of acetyl-CoA was labeled in this tumor (Malloy et al., 1990). However, production of [3,4,5-¹³C]glutamate, which results from retention of ¹³C in TCA cycle intermediates past the first complete turn of the TCA cycle, was still small (Figure 4C). Combining the stable isotope labeling data presented here with the ccRCC metabolite levels reported in a previous study (Hakimi et al., 2016) suggests that overall TCA cycle turnover is low in this tumor, and/or that the cycle includes a large anaplerotic flux from an unlabeled carbon source which would dilute labeling in TCA cycle intermediates. Either of these hypotheses is interesting in the context of ccRCC. The human ccRCC TCGA study (KIRC) indicates that reduced expression of TCA cycle enzymes correlated with poor clinical outcomes, also suggesting that aggressive tumors down-regulate mechanisms for fuel oxidation (Cancer Genome Atlas Research, 2013). It is important to emphasize that only one patient with ccRCC was infused with [1,2-¹³C]acetate. Additional infusions are needed to assess the generalizability of these findings.

Open in a separate window

Figure 4:

In vivo labeling of a ccRCC tumor from [1,2-¹³C]acetate.

(A) Demographic and clinical characteristics of a ccRCC patient infused with [1,2-¹³C]acetate. (B) Schematic showing how acetate infusion bypasses PDH. (C) ¹H-decoupled ¹³C NMR spectrum from the acetate infused tumor. The inset is a magnification of the peaks from glutamate (Glu) at carbon 4 (C4). D45 refers to the doublet formed from glutamate labeled at both C4 and C5. Q refers to the quartet formed from glutamate labeled at C3, C4, and C5. The quartet forms after a turn of the TCA cycle. The number after the indicated metabolite refers to the carbon position in the metabolite. Glu C4 = Glutamate C4, Gln C4 = Glutamine C4, Glu C3 = Glutamate C3, Lac C3 = Lactate C3

ccRCC is predominantly driven by loss of the tumor suppressor VHL. VHL inactivation through mutation, epigenetic silencing, or loss of heterozygosity occurs in up to 90% of ccRCCs (Moore et al., 2011; Rathmell and Chen, 2008; Turajlic et al., 2018). This results in impaired regulation of HIFα protein stability and chronic transcriptional activation of HIF target genes, many of which regulate metabolism. In cultured cells, the pseudohypoxic state of VHL loss is characterized by heightened glycolysis, suppressed glucose oxidation and increased reductive metabolism of glutamine (Hu et al., 2003; Kim et al., 2006; Metallo et al., 2011; Mullen et al., 2011; Papandreou et al., 2006). Metabolomics in vivo has provided a snapshot of the metabolic phenotype of human ccRCC and revealed broad alterations in metabolite abundance relative to adjacent kidney (Hakimi et al., 2016; Li et al., 2014). However, understanding the dynamic utilization of fuels and metabolite turnover in these tumors requires the use of isotope tracers to assess pathway activity. The work presented here with ¹³C-glucose and ¹³C-acetate indicates perturbed glycolysis and TCA cycle activity in primary ccRCC tumors compared to adjacent kidney and to other human tumors in vivo. Specifically, human ccRCC displays enhanced enrichment of glycolytic intermediates and lower labeling of TCA cycle intermediates relative to the adjacent kidney.

To our knowledge, this is the first report of isotope tracing in human ccRCC in vivo. ccRCC tumors exhibit enhanced glycolysis and suppressed PDH activity compared to their matched adjacent kidney tissue, and primary ccRCC tumors seem to have minimal glucose oxidation and turnover of the TCA cycle in vivo. This phenotype is consistent with the classical description of the Warburg effect, once thought to characterize the majority of aggressive tumors. However, in vivo analysis of human primary NSCLCs, primary brain tumors, and brain metastases revealed a surprisingly high degree of oxidation of glucose and other fuels in vivo; in NSCLC, rates of glucose oxidation actually exceeded rates in the adjacent lung. Therefore, these new findings in ccRCC tumors are unique among human tumors analyzed to this point. Primary ccRCC tumors have a distinctive metabolic profile, and these findings are the first to provide strong in vivo evidence for the Warburg effect in human tumors.

EXPERIMENTAL MODEL AND SUBJECT DETAILS:

The project was approved by the Institutional Review Board of UT Southwestern Medical Center (“An Investigation of Tumor Metabolism in Patients Undergoing Surgical Resection” (STU 062010-157). Patients were selected for inclusion based on imaging and clinical features consistent with renal cell carcinoma. Patients were referred for surgery based on standard clinical practice. Patients were excluded if they had a Karnofsky Performance status <70%, NYHA class III or IV congestive heart failure, psychiatric or addictive disorders that precluded obtaining informed consent, unstable angina, or were pregnant or lactating. Clinical details about patients participating in the study and their tumors are summarized in Figure 1A and Supplemental Table 3.

METHOD DETAILS:

This research was supported by ACS-IRG-02-196, NCI SPORE P50CA196516, NCI R01CA154475, NCI R35CA220449, NIH P41EB015908, and NIH NCATS KL2TR001103. K.D.C. was supported by NIH NCATS KL2TR001103. K.D.C., P.K., I.P., R.J.D., and E.A.M. are supported by NCI SPORE P50CA196516. I.P. is supported by R01CA154475. D.B. is supported by NIH NCATS TL1TR001104. R.J.D. is supported by CPRIT grant RP160089 and by the Howard Hughes Medical Institute Faculty Scholars Program. The authors would like to thank James Brugarolas, PI of NCI SPORE P50CA196516, for scientific discussion. We thank Jiyeon Kim for scientific discussion and critical reading of the manuscript. Graphical abstract was created using BioRender. The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.