Slow TCA flux implies low ATP production in tumors

The tricarboxylic acid (TCA) cycle oxidizes carbon substrates to carbon dioxide, with the resulting high energy electrons fed into the electron transport chain to produce ATP by oxidative phosphorylation. Healthy tissues derive most of their ATP from oxidative metabolism, and the remainder from glycolysis. The corresponding balance in tumors remains unclear. Tumors upregulate aerobic glycolysis (the Warburg effect), yet they also typically require an intact TCA cycle and electron transport chain1–6. Recent studies have measured which nutrients contribute carbon to the tumor TCA metabolites7,8, but not tumor TCA flux: how fast the cycle turns. Here, we develop and validate an in vivo dynamic isotope tracing-mass spectrometry strategy for TCA flux quantitation, which we apply to all major mouse organs and to five tumor models. We show that, compared to the tissue of origin, tumor TCA flux is markedly suppressed. Complementary glycolytic flux measurements confirm tumor glycolysis acceleration, but the majority of tumor ATP is nevertheless made aerobically, and total tumor ATP production is suppressed compared to healthy tissues. In murine pancreatic cancer, this is accommodated by downregulation of the major energy-using pathway in the healthy exocrine pancreas, protein synthesis. Thus, instead of being hypermetabolic as commonly assumed, tumors apparently make ATP at a lower than normal rate. We propose that, as cells de-differentiate into cancer, they eschew ATP-intensive processes characteristic of the host tissue, and that the resulting suppressed ATP demand contributes to the Warburg effect and facilitates cancer growth in the nutrient-poor tumor microenvironment.


35
Animals use ATP as their main energy currency, powering functions including muscle 36 contraction, ion pumping, and protein synthesis. ATP can be produced either from glycolysis or 37 by mitochondrial oxidative metabolism. In the latter pathway, the tricarboxylic acid (TCA) cycle 38 oxidizes fat and carbohydrates to make the high energy electron donors NADH and FADH2, 39 which then drive ATP production by the electron transport chain. Catabolism of a glucose 40 molecule in glycolysis yields two ATP, while the coupled TCA cycle and electron transport chain 41 make around 15 ATP per TCA turn while consuming one half of a glucose molecule (or 42 equivalently, one lactate) or one fatty-acid two-carbon unit. Accordingly, in the body as a whole, 43 ATP production by oxidative metabolism exceeds glycolytic ATP generation by more than 20-44 fold.

45
Tumors display metabolic alterations relative to healthy tissues, likely due both to

51
One potential trigger of glycolysis is impaired oxidative ATP production, due to a need to 52 fulfill ATP demand and relief of allosteric inhibition of glycolysis. Warburg hypothesized that 53 tumors are intrinsically respiration deficient. Initial experimental evidence was conflicting on 54 whether tumor TCA flux was higher or lower than other tissues [11][12][13][14][15] . Nevertheless, the concept 55 that tumors have defective mitochondria persisted until recent studies, which have shown that 56 the TCA cycle and electron transport chain are important for tumor cell growth and survival, with 57 active selection against certain mitochondrial DNA mutations in human tumors 1-6 . 58 Isotope tracing has been used to map fuel sources of tumor TCA metabolism. While 59 tumors often select similar substrates to their tissue of origin, human lung cancer showed 60 increased glucose contribution to the TCA cycle 7,8 . Comparable assessment of TCA turning rate 61 (flux) in tumors, however, has been lacking.

62
TCA flux is distinct from carbon substrate selection, similar to the distinction between 63 how fast a furnace burns fuel versus which type of fuel it takes. While the latter can be 64 measured by steady-state isotope tracing, measuring flux requires kinetic isotope labeling Here, we develop and validate an isotope-tracing mass spectrometry method to 68 measure TCA cycle flux in tissues and tumors in mice. We also quantify glucose usage flux with 69 2-deoxyglucose. Together, these methods show that healthy murine tissues make at least 90% 70 of their ATP using the TCA cycle and oxidative phosphorylation. Strikingly, tumors show 71 markedly suppressed TCA turning. Despite this decreased TCA turning, and elevated glucose 72 flux consistent with the Warburg effect, we calculate that tumors still make a majority of their 73 ATP oxidatively, with implied total ATP production rates in tumors significantly lower than their 74 tissues of origin.

88
Since lactate enters tissues quickly and feeds the TCA cycle in almost all tissues 7 , we 89 hypothesized that it could potentially be used for this purpose ( Figure 1A). We optimized a 90 primed infusion of [U-13 C] lactate (in which all three lactate carbon positions are carbon-13) so 91 that steady-state labeling of blood lactate was attained in less than 60 seconds ( Figure 1B-C).

92
This quick approach to steady state blood labeling was important: It made sure that the rate-93 limiting step in tissue TCA metabolite labeling was actual turning of the cycle within the tissue 94 ( Figure 1D, Extended Data Figure 1A-B

105
All TCA metabolites in a given tissue showed similar labeling kinetics (Extended Data 106 Figure 1C-D), so we used the mean m+2 labeling of three well-detected metabolites, malate, 107 succinate, and glutamate, for our calculations. We observed that different tissues accumulate 108 TCA labeling at different rates ( Figure 1D, Extended Data Figure 2A); for example, diaphragm 109 gains TCA labeling much more quickly than quadriceps muscle. Using these labeling timepoints,

110
we calculated the labeling rate constant k for all tissues using equation [1], with higher k 111 representing faster labeling rate ( Figure 1E).

112
Metabolite pool size is also required to calculate flux using kinetic measurements,  Figure 1F, Extended Data Figure 2B).

117
Oxaloacetate, succinyl-CoA, and fumarate are not well-detected by our LC-MS method, but are 118 present at much lower abundance than aspartate and glutamate and so would not materially 119 change the measured TCA pool sizes 18 .

120
By multiplying the labeling rate constant k ( Figure 1E) by the TCA pool size P ( Figure   121 1F) we calculated absolute TCA flux for each tissue ( Figure 1G). We found that heart and 122 diaphragm have the highest TCA flux per gram tissue, while skin has the lowest.  Figure 2C). This primed 132 infusion was then used to measure tissue TCA labeling (Figure 2A)

147
where JO2 is oxygen-consumption flux. We found that our measurements of TCA turning 148 correlated well with historical tissue slice data, with the exception of heart and diaphragm 149 ( Figure 2C, R 2 =0.85 excluding heart and diaphragm). In heart and diaphragm, we observed

154
To quantify the contribution of each tissue to whole body oxygen consumption, we

167
Imaging mass spectrometry of kidney TCA flux

168
The experiments above measured TCA flux averaged across each organ as a whole; 169 however, tissues are composed of different cell types with distinct metabolism. For example, the 170 kidney cortex is responsible for sodium and potassium pumping in order to recover metabolites 171 from the glomerular filtrate, and thus has a higher energy demand than the kidney medulla 34 . To 172 measure regional TCA flux across the kidney, we performed a primed infusion of carbon-13 173 lactate as above, then after 90 seconds isolated and froze the kidney, sectioned it, and 174 performed imaging mass spectrometry. This pre-steady state TCA metabolite labeling, which 175 correlates with flux, was much higher in the kidney cortex than the medulla ( Figure

247
Consistent with classical Warburg metabolism, we calculate that tumors net release 248 lactate, as they have higher estimated lactate production from glycolysis than lactate burning in 249 the TCA cycle (7-to 53-fold higher lactate production than lactate consumption in TCA 250 depending on tumor type, Extended Data Figure 4C). Note that tumors use circulating lactate as 251 a TCA fuel, however they release more lactate produced from glycolysis than they oxidize in the 252 TCA cycle.

253
Tumors make ATP slower than healthy tissues

254
A main goal of this study was to find how much the TCA cycle contributed to energy 255 production in tumors. Using our glucose and TCA fluxes, we estimated how much ATP tissues 256 and tumors derived from glycolysis (2 ATP produced per glucose consumed) versus from TCA 257 paired with the electron transport chain (approximately 14.5 ATP produced per two-carbon unit 258 consumed). Tumors derived a much higher fraction of their ATP from glycolysis than healthy 259 tissues, with an estimated 11-30% of ATP derived from glycolysis in tumors but a median of 260 1.8% derived from glycolysis in healthy tissues ( Figure 5E).

261
We further used our measured glucose and TCA flux values to calculate total ATP 262 production in healthy tissues and tumors. Remarkably, the implied ATP production rate of 263 tumors was much less than healthy tissues. This calculation assumes a similar efficiency of ATP       were allowed to recover from surgery for at least 5 days before tracer infusion.

382
At the desired timepoint, mice were euthanized quickly by cervical dislocation, and tissues were 383 collected as quickly as possible and freeze-clamped using a liquid nitrogen cooled Wollenberger 384 clamp. Tissues were stored at -80C until processed.

385
Glutamine primed infusion to measure TCA flux 386 [U-13 C] glutamine (99% purity, CLM-1822) was diluted to 100mM in sterile saline. This 387 tracer was made fresh for every experiment due to previously observed tracer instability.

388
Infusion was performed as in previous section, with these modifications: priming dose of 389 13+1.6*(mouse weight) microliters was provided in 60 seconds, then infusion rate was slowed 390 to 0.1 microliters*(mouse weight in grams) per minute. As above, at the desired timepoint, mice

408
As above, at the desired timepoint, mice were euthanized quickly by cervical dislocation, and 409 tissues were collected as quickly as possible and freeze-clamped using a liquid nitrogen cooled 410 clamp. Tissues were stored at -80C until processed.

464
To measure glucose concentration in arterial blood, standards were made by 465 resuspending 13 C-labeled glucose (Cambridge Isotope Laboratories CLM-1396) in water at a 466 known concentration, mixing standard into methanol extraction buffer, and blood serum samples 467 were extracted in extraction buffer on water ice over dry ice, then proceeded as above.

468
To measure protein enrichment from [U-13 C] valine infusion, tissues were ground using 469 the Retsch Cryo Mill, then 5mg of tissue powder was weighed into Eppendorf tubes. Free tissue 470 amino acids were removed using methanol-chloroform extraction as follows: 400ul methanol, 200ul chloroform, and 300ul water were added to each Eppendorf, vortexing after each addition.

472
Tubes were centrifuged 10min at 14000 RCF at 4C, then upper layer was discarded. Remaining   infused in trace amounts that do not inhibit endogenous glucose metabolism, and indeed the 589 peak 2-deoxyglucose concentration measured in blood was ~90 micromolar, or around 86-fold 590 lower than blood glucose concentration ( Figure 3B). We found a linear relationship between the 591 rate of 2-deoxyglucose infused and the tissue 2-deoxyglucose-phosphate concentration 592 (Extended Data Figure 3D) again suggesting the trace amount of 2-deoxyglucose infused was 593 not saturating tissue glucose use. Each healthy tissue was measured from at least n=2 mice at 594 5min, 7.5min, 10min, 15min, and one 0min timepoint; tumors were measured from at least n=2 595 mice at 5min, 10min, and 15min except for GEMM NSCLC, which included only n=1 for 5min. 2-596 deoxyglucose concentration in blood was measured in arterial blood using double catheterized deoxyglucose phosphate after 10 minutes, so the 15-minute timepoint was not used in further 608 analysis (Extended Data Figure 3B).

656
All t-tests used were two-sided; sample size for every experiment is recorded in Extended 657 Data Table 1.

658
To compute two-sided t tests for difference in TCA fluxes, the following procedure was used:

665
The p value of a two-sided t test comparing two TCA fluxes was then computed using a Welch's 666 t test using the Welch-Satterthwaite equation (not assuming equal variances).

667
To compute two-sided t tests for difference in glycolysis fluxes, a similar procedure was

676
To compute two-sided t tests for difference in ATP production fluxes, the standard 677 deviation was computed as the sum of squared standard deviations of glycolysis flux and TCA 678 flux. The p value of a two-sided t test comparing two ATP production fluxes was then computed 679 using a Welch's t test using the Welch-Satterthwaite equation (not assuming equal variances).

680
All R 2 values of linear correlation were computed using linear regression.

681
Data Availability

682
All raw data, analyzed data and materials will be provided upon request to the 683 corresponding author (joshr@princeton.edu). We plan to also make the data available through 684 an in-house data repository that is currently in development for easy retrieval and processing of 685 stable isotope tracing mass spectrometry data.