Quantification of lactate from various metabolic pathways and quantification issues of lactate isotopologues and isotopmers

13C-labeled glucose combined with chromatography and mass spectrometry enables us to decipher the percentage of lactate generated from various metabolic pathways. We showed that lactate derived from glycolysis, pentose phosphate pathway, Krebs cycle, and other sources accounted for 82–90%, 6.0–11%, 0.67–1.8% and 1.5–7.9%, respectively, depending on different types of cells. When using glucose isotopomers ([1-13C]-, [3-13C]-, [4-13C]-, and [6-13C]glucose) or isotopologues ([1,2-13C2]- and [1,2,3-13C3]glucose) for tracing, the ratio of lactate derived from glucose carbon 1, 2, 3 over 4, 5, 6 via glycolysis varied significantly, ranging from 1.6 (traced with [1,2-13C2]glucose) to 0.85 (traced with [6-13C]glucose), but the theoretical ratio should be 1. The odd results might be caused by intramolecular 13C, which may significantly affect lactate fragmentation under tandem mass spectrometry condition, leading to erroneous quantification. Indeed, the fragmentation efficiency of [U-13C]lactate, [2,3-13C]lactate, and [3-13C]lactate were 1.4, 1.5 and 1.2 folds higher than lactate, respectively, but [1-13C]lactate was similar to lactate, suggesting that carbon-13 at different positions could differentially influence lactate fragmentation. This observed phenomenon was inconsistent with the data based on theoretical calculation, according to which activation energies for all lactate isotopomers and isotopologues are nearly identical. The inconsistency suggested a need for further investigation. Our study suggests that calibration is required for quantifying metabolite isotopolugues and isotopomers.

Cancer cells convert most incoming glucose to lactate, a metabolic hallmark called Warburg effect 1, 2 . Lactate and proton are important for cancer cells to survive through harsh conditions. We recently demonstrated that lactate and proton together switched cancer cells from Warburg effect to an economical metabolic mode with negligible or no net generation of lactate 3 and with 90% ATP from oxidative phosphorylation 4 . Moreover, lactate and proton together prevented cancer cells from glucose deprivation-induced death 5 . The findings suggested that targeting intratumoral lactic acidosis might be considered as a therapeutic target. Indeed, our clinical study demonstrated a remarkable effect of bicarbonate on local control of hepatocellular carcinoma 6 .
Many other investigators have independently reported the significance of intratumoral lactic acidosis in tumor biology. Clinical studies showed that high level of lactate was a strong prognostic indicator of increased metastasis and poor overall survival [7][8][9][10][11][12][13] . Gillies and Gatenby group demonstrated that systematic and tumor pHe alkalization could inhibit carcinogenesis, tumor invasion and metastasis, and they also provided integrated models that can predict the safety and efficacy of buffer therapy to raise tumor pHe [14][15][16] and related theoretical work 17,18 . Furthermore, lactic acidosis exhibited multifaceted roles in skewing macrophages 19 , inhibiting the function of cytotoxic T cells 20 , altering cancer cell metabolism 21,22 , inducing chromosomal instability 23 , and promoting tumor angiogenesis 7,24 .
Hence, lactate generation is an interesting topic in cancer metabolic research. Glucose 25 is the main sources of lactate generation in cancer cell metabolism. However, the percentage of lactate generated from glucose through

Analysis of Lactate by LC-MS/MS. Lactate isotopomers and isotopologues of the cell culture supernatant
were measured by liquid chromatography-tandem mass spectrometry (LC-MS/MS). LC was performed on a Waters ACQUITY UPLC system employing a ACQUITY BEH Amide column, 2.1 × 100 mm, 1.7 μm. Eluent A was 95% acetonitrile and 5% aqueous solution of 20 mM ammonium acetate (pH 9.0), whereas eluent B was 50% acetonitrile and 50% aqueous solution of 20 mM ammonium acetate (pH 9.0). The flow rate was set at 0.6 ml/min, injection volume is 5.0 μl, and the column was kept at 50 o C. The optimized gradient conditions were adopted from Waters ACQUITY UPLC BEH Amide column application notebook with a minor modification as follows: 0-0.4 min hold for 99.9% eluent A, 0.4-0.5 min from 99.9% −60% eluent A, 0.5-2 min from 60% −30% eluent A, 2-2.1 min from 30% −99.9% eluent A, and hold for 10 min. The retention time of lactate and lactate isotopomers and isotopologues was 1.29 min.
The MS detection was performed on a 4000 QTRAP mass spectrometer (AB SCIEX, Foster City, CA, USA) equipped with an ESI ion source (Turbospray) operated in negative ion mode. Instrument control, data acquisition, and processing were performed using the Analyst 1.5.2 software. Firstly, collision-induced dissociation (CID) experiment of lactic acid standard was performed in product ion scan mode and the spectrum was illustrated in Supplementary Fig. S1A. The dissociation mechanism was proposed as illustrated in Supplementary  Fig. S1B. Then MS/MS data were acquired in the multiple reaction monitoring (MRM) mode. The transition between precursor ion and the most abundant product ion (m/z 89.0 > 43.0) was monitored for quantitative determination. The ion transition m/z 89.0 > 71.0 was monitored for qualitative analysis to confirm the identity of lactic acid in samples. To determine the percentage of 13 C-labeled lactate derived from 13 C-labeled glucose through glycolysis and pentose phosphate pathway, the following ion transitions were monitored to determine the distribution of 13  To increase sensitivity, the ion source temperature (TEM) was set at 500 °C, and the ion spray voltage (IS) was set at −4.5 kV. Ion source gas 1 (GS1) and ion source gas 2 (GS2) used as the nebulizing and drying gases were set at 50 and 40 psi, respectively. Curtain gas (CUR) was set at 40 psi. The optimized MS conditions used for the analysis of the target analytes were shown in Supplementary Table S1.
Eliminate peak area of natural 13 C-labled lactate. Chemical structure of lactate and the numbering of its carbon, hydrogen and oxygen atom are shown in Supplementary Fig. S2, according to isotopic abundances of carbon, hydrogen, oxygen elements (Supplementary Table S2) and proposed fragmentation mechanism of lactate ( Supplementary Fig. S1B), the relative percentage of each natural 13 Tables S3-S4). We used formula (1) to exclude the peak area of lactate isotopomers and isotopologues that from nature. Let Χ be the peak area of 13 C-labled lactate produced by cells, P L be the measured peak area of 13 C-labled lactate which includes 13 C-labled lactate existed in nature and that produced by cells, P U be the peak area of unlabeled lactate, R be the corresponding relative percentage of each natural 13 C-labled lactate.
L U Theoretical calculations of the activation energy for fragmentation of lactate isotopomers and isotopologues. All theoretical calculations were performed by using the density functional theory (DFT) www.nature.com/scientificreports/ method at the B3LYP/6-31 G(d) level of theory in the Gaussian 03 program 27 . The optimized structures for the precursor ions, intermediates and products were identified as a true minimum in energy by the absence of imaginary frequencies. Transition states, on the other hand, were identified by the presence of one single imaginary vibration frequency with the normal vibrational mode, and further confirmed by the intrinsic reaction coordinates (IRC) analysis. The energies discussed here are the sum of electronic and thermal enthalpies. The DFT optimized structures were shown by Gauss View (version 3.07) software to give higher quality images of these structures.

Statistical analysis.
All the statistical analyses were performed using SPSS statistics 19.0 software (IBM, Armonk, NY, USA). T-test was applied to evaluate the differences of peak areas ratio of different 13 C-lableing lactate.

Results and Discussion
The rationale for quantification of lactate isotopomers and isotopologues. In  The percentage of each lactate isotopomers or isotopologues was quantified by tandem mass spectrometry. Under MRM mode of mass spectrometer with triple quadrupole linear ion trap (QTRAP) analyzer, the precursor ion m/z 89.0 was isolated in the first quadrupole (Q1), subsequently, the ion dissociated under collisional activation using collision gas such as nitrogen in Q2, and the fragment ions were isolated in Q3, finally, the ions were detected by the detector, as described in Supplementary Fig. S3.
As illustrated in Supplementary Fig. S1A, two fragment ions at m/z 71.0 and 43.0 were produced in the dissociation of unlabeled lactate at m/z 89.0. The transition between precursor ion and the most abundant product ion (m/z 89.0 > 43.0) was monitored for quantitative determination. And the ion transition m/z 89.0 > 71.0 was monitored for qualitative analysis to confirm the identity of lactate in samples. For 13 C-labeled lactate derived from 13 C-labeled glucose through glycolysis and pentose phosphate pathway, the following ion transitions were monitored to determine the distribution of 13  Representative chromatograms of lactate standard and major lactate species generated in cells were shown in Supplementary Fig. 4; the sum of their percentages was more than 98% of total lactate. And the retention time of lactate and each 13 C-labeled lactate generated in cells is identical to that of lactate standard. The relative standard deviation (RSD) in the peak area is ≤10% (n = 6) for lactate and each 13 C-labeled lactate mentioned in Supplementary Fig. 4. It can be clearly seen that lactate and each 13 C-labeled lactate generated in cells can be detected with good peak shapes, and the signal to noise ratios is above ten. And lactate isotopomers and isotopologues can be separated from each other based on m/z value.
According to the established principle, 13 C, which only adding a neutron into nucleus, should not affect chemical bonding, hence should not affect molecule fragmentation both quality and quantity under tandem mass spectrometry condition, i.e., the fragmentation of lactate with or without 13 C should be exactly the same.
A time course to determine the steady-state generation of lactate isotopologues or isotopomers. In order to exclude the interference of exogenous lactate, we used serum free RPMI-1640 to culture cells, as serum contains appreciable amount of lactate, which may interfere with the quantification of lactate isotopologues. Nevertheless, the condition of serum-free medium could be considered as a stress for cancer cells, which might interfere with the cell metabolism. We compared the percentage of lactate isotopologues generated by cells in the presence or absence of serum. The results were summarized in Supplementary Table S6, showing no significant difference between cultures with or without serum. Thus, we used serum free RPMI-1640 to culture cells in this study with the culture time within 12 hours.
We performed a time-course experiment and observed that a steady-state generation of lactate isotopologues was attained 4 hours after incubation and maintained thereafter ( Fig. 1; Supplementary Table S5). Based on the time-course experiment, all the experiments performed below were carried out using serum free RPMI-1640, with an incubation time of 12 hours.

Determination of lactate from glucose and other sources. We used [U-13 C]glucose to trace lactate.
Glucose-derived lactate constituted about 97% of total lactate (Table 1), other source-derived lactate constituted about 2%, and lactate derived from mixed carbon sources (glucose and other sources) was less than 1%, which was most probably derived from Krebs cycle (Fig. 2). The percentage of lactate isotopologues generated from 4T1, Hela, and K562 cancer cells were comparable with each other. In mouse thymocytes, lactate generated from glucose, mixed carbon sources, and other sources constituted about 90%, 2%, and 8%, respectively. Determination of lactate from glycolysis and pentose phosphate pathway. [ Thus, the percentage of lactate from glycolysis, pentose phosphate pathway, and other sources were about 90%, 6.6%, and 2.2%, respectively (the down part of Table 2), accounting for about 99% of total lactate. Using Hela and K562, we obtained nearly identical results (Table 2). For thymocytes, the percentage of lactate from glycolysis, pentose phosphate pathway, and other sources were about 85%, 6.0%, and 7.9%, respectively, accounting for about 99% of total lactate (Table 2).
Using 4T1, Hela, k562, and mouse thymocytes, we obtained nearly identical results (Tables 2-3), indicating that the results were highly reproducible and reliable.

C-labelling of lactate at different positions differentially influences its fragmentation efficiency under tandem mass spectrometry condition.
In theory, glucose labeling with 13 C should not alter the percentage ratio between lactate derived from glucose carbon 1, 2, 3 and lactate derived from glucose carbon 4, 5, 6, because there is no theoretical and experimental basis that enzymes responsible for glucose metabolism can distinguish 13 C from 12 C.
According to the methodology of mass spectrometry, isotopomers or isotopologues which differ only at 13 C or 12 C should not affect the ionization and fragmentation, so that the percentage of each isotopomer or isotopologue of a metabolite could be quantified. In fact, this is the principle used to measure isotopomers or isotopologues of metabolites in metabolomic studies [28][29][30][31] .
However, after generating so many confusing data as described above, we suspected that carbon-13 may significantly interfere with molecule fragmentation under tandem mass spectrometry condition and interfere the quantification of each lactate isotopomer and isotopologue.
Then we mixed [1-13 C]-, [3-13 C]-, or [U-13 C]lactate with lactate respectively, and their concentration ratios, analysis and calculation methods were similar as described above. While [1-13 C]lactate/lactate peak area ratio was comparable with their concentration ratio (the right panel of Fig. 5B), [3-13 C]lactate/lactate and [U-13 C]lactate/ lactate peak area ratio was 1.2 (the right panel of Fig. 5C) and 1.4 (the right panel of Fig. 5D) folds higher than    Table 3), and other sources (according to percentage listed in Table 1). Data are mean ± SD, n = 12, from 2 independent experiments.   Table 1). Data are mean ± SD, n = 12, from 2 independent experiments.   13 C affects ionization process of 13 C-labeled lactate. Data in Supplementary Tables S11-S14 point out that 13 C slightly affect ionization of 13 C-labeled lactate under MS condition.   Table 6. Relative energies of species involved in the fragmentation reaction routes of lactate isotopomers and isotopologues.
In order to investigate 13 C effect on lactate fragmentation under different MRM mode, the lactate and [2,3-13 C 2 ]lactate (or [1-13 C]-, [3-13 C]-, [U-13 C]lactate) mixed standard solutions mentioned above were analyzed by MRM mode in losing H 2 O. The results pointed out that 13 C enhanced fragmentation of [2,3-13 C 2 ]-, [3-13 C]-, or [U-13 C]lactate as well, but the influence was less than that under MRM mode in losing CO; and the fragmentation of [1-13 C]lactate was not affected by 13 C under MRM mode in losing H 2 O (Supplementary Tables S11-S14).
The results raised a theoretical issue: 13 C and 12 C differs in only one number of neutron, the chemical bond was formed by sharing the outer electrons and the strength of the bond was hardly influenced by the neutron. In tandem mass spectrometry, collision gas such as nitrogen was used and the fragmentation occurs when the collision energy was high enough to break the chemical bond. Under the same collision energy, the intensity of the fragment ion from the 12 C-and 13 C-labeled lactate was thought to be identical. Even if there is some difference due to the mass variability, it will be very little. However, in the present study, significant difference was observed.
Calibrations were performed on the data of Hela, K562, and thymocytes, the calibrated data indicated that the percentages of lactate derived from glucose carbon 1, 2, 3 and 4, 5, 6 were nearly equal as well (Supplementary Tables S15-S18).
Theoretical calculation of the activation energy of lactate fragmentation. To obtain insights into the mechanism of lactate fragmentation, we carried out theoretical calculations at the B3LYP/6-31 G(d) level of theory to quantitatively describe the energy requirements of these reactions 32-34 and a schematic potential energy surface is illustrated in Fig. 6. The energy of transition state 2 (TS2) for losing CO is higher than that of transition state 1 (TS1) for losing H 2 O, suggesting that the loss of CO process is the rate determining step for the formation of ion at m/z 43., the relative energies of lactate isotopomers and isotopologues in the fragmentation reaction routes were shown in Table 6. The energies of TS2 of all these compounds were nearly identical, suggesting that lactate isotopomers and isotopologues should have identical fragmentation behaviors in tandem mass spectrometry. The calculation was inconsistent with the experimental data and this inconsistency is worthy of further investigation.

Concluding remarks.
To the best of our knowledge, there is no report regarding the significant impact of carbon-13 labeling on fragmentation of molecules under tandem MS condition. In this study, we revealed that carbon-13 labeling could significantly interfere with lactate fragmentation. This observation is inconsistent with our theoretical calculation that the activation energy of fragmentation of lactate with or without carbon-13 labeling is nearly identical. Our study also points out that potential problems may exist in the previous studies involving quantification of isotopomers and isotopologues by tandem mass spectrometry technology. In the future, the proper calibrations for quantification of isotopomers and isotopologues would be required.