Determination of the concentration range of inhibitorsIn order to determine the appropriate concentration range of a glycolysis inhibitor, we investigated the relationship between the concentration of 2-DG and cell survival. The experimental results are presented in Fig. 1. Following treatment with varying concentrations of 2-DG, the survival rate of four different cell types (A549, PC-9, NCI-H460, BEAS-2B) exhibited a dose-dependent decrease with increasing inhibitor concentration. Specifically, as the concentration of 2-DG increased from 0 to 40 mmol/L, the survival rate decreased from 100 to 40% for A549 cells, from 100 to 63% for PC-9 cells, from 100 to 47% for NCI-H460 cells, and from 100 to 47% for BEAS-2B cells. Notably, BEAS-2B cells demonstrated higher sensitivity towards low concentrations of 2-DG compared to other cell types; conversely, A549 cells displayed greater sensitivity towards high concentrations of 2-DG.Figure 1Survival rate of four cell types (A549, PC-9, NCI-H460, BEAS-2B) were assessed under varying concentrations of 2-DG.To ensure that the characteristic VOCs are from living cells, according to the principle that the cell survival rate is nearly 80%, we finally chose the concentration of the experimental inhibitor as 2-DG of 10 mmol/L.Differences in VOCs between lung cancer cells and normal cells in resting stateIn the resting state, the headspace VOCs of A549, PC-9, NCI-H460, and BEAS-2B were analyzed using the SPME–GC–MS technique. OPLS-DA was conducted utilizing the VOCs released by three distinct types of lung cancer cells, as well as normal cells during resting state. The resulting score plots are presented in Fig. 2a–c, where it is evident that BEAS-2B cells exhibit a clear separation from the three lung cancer cell types. Notably, these separations exhibit high goodness of fitting (R2Y) and prediction (Q2), indicating the robustness of the model. To further validate the efficacy of the OPLS-DA approach, 200 permutation tests were performed, and the results are depicted in Fig. 2d–f. The R2 and Q2 values obtained from the actual model exceed all corresponding values in the permutation models, unequivocally demonstrating the validity and statistical significance of the OPLS-DA analysis conducted in this study.Figure 2A549 (a)/PC-9 (b)/NCI-H460 (c) and BEAS-2B cells headspace VOCs OPLS-DA score plots in resting state (a–c), and the results of their corresponding 200 permutation tests (d–f).Following OPLS-DA, Mann–Whitney U test, and FC analysis, the results revealed distinct VOC profiles between each type of lung cancer cell and normal cells. Initially, based on the U-test analysis and FC a Volcano plot (Fig. S5a–c) was generated to visualize the changes in VOCs between each lung cell lines and normal cell line during the resting state. Subsequently, with reference to the OPLS-DA VIP > 1 and NIST database RSI > 800 differential VOCs were summarized in Table 1. As depicted in Venn diagram of difference VOCs (Fig. S6a), three common substances were found: VOC19 (ethyl propionate), VOC23 (acetoin), and VOC80 (3-decen-5-one). Ethyl propionate and 3-decen-5-one exhibited lower levels in the headspace of each lung cancer cell type compared to normal cells; conversely, acetoin showed higher levels. In conclusion, these three aforementioned VOCs may serve as potential biomarkers for distinguishing lung cancer cells from normal cells.Table 1 Differences in VOCs between lung cancer cells (A549, PC-9, NCI-H460) and normal cells (BEAS-2B) in resting state.Differences in VOCs between lung cancer cells and normal cells after glycolysis inhibitionAfter the glycolysis inhibition by 2-DG, headspace VOCs of A549, PC-9, NCI-H460, and BEAS-2B cells were analyzed using SPME–GC–MS technique and subjected to statistical analysis. OPLS-DA scores plots and corresponding 200 permutation tests are presented in Fig. 3. All three OPLS-DA models have high goodness of fitting and goodness of prediction. Additionally, the simulated values of 200 replacement tests of the corresponding models are greater than all their original values, indicating the effectiveness of the OPLS-DA model.Figure 3A549 (a)/PC-9 (b)/NCI-H460 (c) and BEAS-2B cells headspace VOCs OPLS-DA score plots after the inhibition by 2-DG (a–c), and the results of their corresponding 200 permutation tests (d–f).The Volcano plot (Fig. S5d–f) serves as a visualization tool to illustrate the alterations in VOCs between each lung cancer cell and normal cells after 2-DG regulation. Table 2 presents the differential VOCs information. Furthermore, the Venn diagram (Fig. S6b) provides a concise representation of the overlap among the three sets of differential VOCs, revealing that only one substance, VOC23 (acetoin), was consistently detected across all three groups. All three types of lung cancer cells exhibited higher levels of acetoin in the headspace compared to normal cells.Table 2 Differences in VOCs between lung cancer cells (A549, PC-9, NCI-H460) and normal cells (BEAS-2B) after the glycolysis inhibition by 2-DG.Identification of characteristic VOC for lung cancer cells under glycolysis regulationIt is worth emphasizing that in comparison to the resting state, the glycolysis inhibition by 2-DG significantly augmented the disparity in acetoin levels in the headspace of lung cancer cells (A549, PC-9, NCI-H460) and normal cells (BEAS-2B). The FC values increased from 5.67, 6.29, 8.10 times to 12.93, 18.74, and 20.56 times respectively (Fig. 4). To further validate the association between increased acetoin levels and glycolysis inhibition, we conducted a repetitive experiment using another glycolysis inhibitor (3-BrPA). Initially, an appropriate concentration range of 100 μmol/L was selected for 3-BrPA (Fig. S7). Subsequently, SPME–GC–MS detection and statistical analysis were performed on the headspace VOCs of A549, PC-9, NCI-H460, and BEAS-2B under glycolysis regulation with 3-BrPA treatment. Relevant results are presented in Fig. S8, Fig. S5h–j, Table S2, and Fig. S6c. Importantly, acetoin was still detected among all three groups of differential VOCs. The FC values for lung cancer cells (A549, PC-9, NCI-H460) and normal cells (BEAS-2B) increased from 5.67, 6.29, 8.10 times to 13.85, 16.76, 21.97 times, respectively (Fig. 4). Overall, glycolysis inhibition can significantly amplify the disparity in acetoin levels between lung cancer cells and normal cells. This suggests that acetoin could serve as the characteristic VOC for distinguishing between lung cancer cells and normal cells under glycolysis regulation.Figure 4Differences in the peak area of acetoin in the headspace before and after glycolysis inhibition between lung cancer cells and normal cells.We also noticed that Feinberg et al.22 mentioned that the substance with m/z of 49 experienced the most significant reduction in A549 cells after the action of inhibitor 3-BrPA, which speculated to be methanethiol. In our experiment, no characteristic VOC with m/z of 49 was detected, including methanethiol, and the reason for the two differences is not clear.Biochemical sources of characteristic VOC for lung cancer cells under glycolysis regulationSeveral studies have observed the presence of acetoin23,24,25 in the breath of individuals with lung cancer, Sanni et al.25 have identified that this compound may originated from the glucose metabolism of oral bacteria. To our knowledge, there is currently no available research on the biochemical origins of acetoin in human cells. However, it has been confirmed that three distinct pathways exist for the production of acetoin in bacteria and yeast, which are schematically represented in Fig. S926,27,28. Pyruvate serves as the starting substrate to produce acetoin based on these three pathways. Glycolysis is a process by which glucose is broken down into pyruvate in the cytoplasm, releasing energy. When glycolysis is inhibited, mitochondria increase their utilization of glutamine to produce organic molecules such as pyruvate and acetyl CoA to maintain energy and biosynthesis needs29, as depicted in Fig. 5.Figure 5Metabolism pathway of glucose and glutamine in lung cancer cells. The bold arrow indicates that the process is dominant. Glut glucose transporter, HK hexokinase, G6P glucose-6-phosphate, PEP phosphoenol pyruvate, PK pyruvate kinase, LDH lactate dehydrogenase, MCT monocarboxylate transporter, PC pyruvate carboxylase, PDH pyruvate dehydrogenase, PDK pyruvate dehydrogenase kinase, OAA oxaloacetate, Mal malate, α-KG α-ketoglutarate, ACLY ATP citrate lyase, TCA tricarboxylic acid cycle.Therefore, we hypothesize that the glycolysis inhibition leads to an upregulation of the glutamine degradation pathway, resulting in a compensatory increase in acetoin production, as illustrated in Fig. 6a. Consequently, it is reasonable to consider whether concurrent restriction of the glutamine breakdown pathway and glycolysis inhibition would lead to a reduction in acetoin (Fig. 6b).Figure 6(a) Mechanism underlying the upregulation of acetoin resulting from glycolysis inhibition. (b) Mechanism explaining the downregulation of acetoin due to combined inhibition of glycolysis and blockade of glutaminolysis. The thickness of the arrows represents pathway strength or weakness.To investigate the potential of glycolysis inhibition in enhancing glutaminolysis for pyruvate production in lung cancer cells, we used 2-DG/3-BrPA to inhibit glycolysis while utilizing PA to block glutaminolysis30. As depicted in Fig. 7, compared to the resting state, treatment with 2-DG/3-BrPA significantly increased acetoin levels in the headspace of all three types of lung cancer cells. However, there was no significant alteration observed when treated with PA alone. Moreover, concurrent administration of 2-DG/3-BrPA and PA effectively suppressed the upward trend of acetoin release. These findings further validate our hypothesis (Fig. 6b). It is evident that the elevation in acetoin is associated with compensatory enhancement of glutaminolysis.Figure 7Alteration in the peak area of acetoin within the headspace of lung cancer cells when 2-DG/3-BrPA inhibiting glycolysis while PA blocking glutaminolysis.
