Non‐enzymatic reaction of carnosine and glyceraldehyde‐3‐phosphate accompanies metabolic changes of the pentose phosphate pathway

Abstract Objectives Carnosine (β‐alanyl‐l‐histidine) is a naturally occurring dipeptide that selectively inhibits cancer cell growth, possibly by influencing glucose metabolism. As its precise mode of action and its primary targets are unknown, we analysed carnosine's effect on metabolites and pathways in glioblastoma cells. Materials and methods Glioblastoma cells, U87, T98G and LN229, were treated with carnosine, and metabolites were analysed by gas chromatography coupled with mass spectrometry. Furthermore, mitochondrial ATP production was determined by extracellular flux analysis and reaction products of carnosine were investigated using mass spectrometry. Results Carnosine decreased the intracellular abundance of several metabolites indicating a reduced activity of the pentose phosphate pathway, the malate‐aspartate shuttle and the glycerol phosphate shuttle. Mitochondrial respiration was reduced in U87 and T98G but not in LN229 cells, independent of whether glucose or pyruvate was used as substrate. Finally, we demonstrate non‐enzymatic reaction of carnosine with dihydroxyacetone phosphate and glyceraldehyde‐3‐phosphate. However, glycolytic flux from glucose to l‐lactate appeared not to be affected by the reaction of carnosine with the metabolites. Conclusions Carnosine reacts non‐enzymatically with glycolytic intermediates reducing the activity of the pentose phosphate pathway which is required for cell proliferation. Although the activity of the malate‐aspartate and the glycerol phosphate shuttle appear to be affected, reduced mitochondrial ATP production under the influence of the dipeptide is cell‐specific and appears to be independent of the effect on the shuttles.

is found with higher concentrations (2.2 mmol per kg dry weight 4 ) in the olfactory bulb. In the human body, carnosine levels are controlled by three genes. Human serum carnosinase (CNDP1) and cytosolic non-specific dipeptidase (CNDP2) degrade the dipeptide, whereas carnosine synthase 1 (CARNS1), which is highly expressed in skeletal muscle, catalyses the formation of carnosine from β-alanine and l-histidine. 5 Surprisingly, although the dipeptide is known since more than 100 years, its physiological role has not completely been resolved. Among the physiological functions described are pH-buffering, chelation of heavy metal ions, scavenging of reactive oxygen species, protection from advanced glycation end-products and protection from lipid peroxidation-related cell damage (for a comprehensive review see Boldyrev et al 6 ). In 1986, Nagai and Suda demonstrated that carnosine aside from "normal" physiological functions inhibits tumour growth in a mouse in vivo model. 7 We and other groups confirmed this anti-neoplastic effect using different types of cancer models, including glioblastoma, in vivo 8,9 and in vitro. [10][11][12][13] Since the anti-neoplastic effect has been studied in more detail, a number of possibly involved signalling pathways have been suggested. Some reports point towards an inhibition of Akt phosphorylation 12,14 and HIF-1α signalling, 10 and G1 and G2/M cell cycle arrest has been detected. 15,16 Furthermore, we and others reported that carnosine inhibits glucose-dependent ATP production in glioblastoma cells. 11,17 However, the mechanisms responsible for carnosine's anti-neoplastic effect are far from being understood.
Additionally, a primary target has not been identified which especially pertains to carnosine's effect on glucose metabolism. Here, we analysed the metabolic profiles of tumour cells treated with or without carnosine in order to reveal whether primary targets of the dipeptide can be identified.
We decided to use glioblastoma cells in our study as this is the most frequent malignant tumour of the human brain 18 with an incident of 3.21 new cases per 100 000 inhabitants in the United States. Under standard therapy consisting of maximal safe resection of the tumour, radiotherapy and adjuvant chemotherapy with temozolomide, median survival of patients is only 12-15 months. 19 Thus, there is an urgent need for additional therapeutic options to treat this highly malignant tumour of the central nervous system. Understanding the mechanisms by which carnosine exerts its antineoplastic effect could well pave the way for the development of new therapeutic options.

| Chemicals and reagents
If not stated otherwise, all chemicals were purchased from Sigma-Aldrich. Carnosine was kindly provided by Flamma S.p.A.

| Cell culture
U87, T98G and LN229 cells were originally obtained from the ATCC and cultured in T-75 culture flasks (Sarstedt AG & Co.) in DMEM/25 mM glucose, without pyruvate (Thermo Fisher Scientific), supplemented with 10% foetal bovine serum (FBS superior, Biochrom), 2 mM l-alanyl-l-glutamine and antibiotics (Thermo Fisher Scientific) (further designated as "culture medium") at 37°C and 5% CO 2 in humidified air in an incubator. In order to confirm identity over long culture periods, cells were genotyped by STR analysis at the Genolytic GmbH using a PowerPlex ® 21 System (Promega) and cells were confirmed as the U87MG cell line from the ATCC. 20

| Growth rate determination
About 150 000 cells were seeded in culture medium (5 mL

| Determination of l-lactate production rate
About 10 6 cells were seeded in culture medium (2 mL) per well into 6-well plates. After 24 hours of incubation, cells were washed once with HBSS (1 mL) and then further incubated with DMEM N2 and supplemented with or without 50 mM carnosine. After 0, 3, 6, 12, 24 and 32 hours, medium (5 µL) was collected and stored at −80°C until further use. l-lactate was determined as described before. 21 Absolute amounts of l-lactate were calculated using a standard curve, and l-lactate production rate was determined by calculating the l-lactate production between three and six hours of incubation.
In order to validate the significance of the data obtained, repeated measures one-way ANOVA was performed using all time points.

| Determination of glucose uptake rate
A total of 10 000 cells were seeded in culture medium (200 µL) in 96well plates (µClear, Greiner Bio-One). After 24 hours of incubation, cells were washed once with HBSS (200 µL) and then further incubated with DMEM N2 and supplemented with or without 50 mM carnosine for 6 hours. Then, DMEM (30 µL) containing Hoechst 33 342 (10 ng/mL; Thermo Fisher Scientific) was added to each well and cells were counted using a Celigo Imaging Cytometer (Nexcelom Bioscience LLC) after an incubation for 30 minutes at 37°C. Afterwards, cells were washed once with HBSS (100 µL) and further incubated in the presence of 1 mM 2-desoxy-d-glucose for 10 minutes. Formed 2-desoxyd-glucose-6-phosphate was determined using the glucose uptake Glo assay (Promega) according to the manufactures recommendations.

| Metabolic profiling via GC-MS
Metabolic profiling was performed as described previously. 21 Briefly, Immediately after washing, ice-cold methanol (1 mL) was added to each well and metabolites were extracted for 24 hours on an orbital shaker at 8°C. Then, extracts were collected and wells were rinsed once with methanol. Afterwards, samples were evaporated to dryness using a speed vac (Maxi-Dry Lyo, Heto-Holten) and stored at −80°C until further use. Derivatization, gas chromatography coupled with mass spectrometry (GC-MS) analysis and data evaluation were performed as described before. 21,22 If not stated otherwise, the abundance of a metabolite is defined by the peak area determined from the selected ion chromatogram of an experiment normalized to total cellular protein (µg). or DHAP were quantified according to Bergmeyer. 23 Briefly, 1.5 µL of NADH solution (20 mM) was added to each reaction and the initial extinction at 340 nm was determined. Afterwards, enzyme mix (3.5 µL) was added to detect GA3P (1 U/µL triosephosphate isomerase; 0.085 U/µL glycerol-3-phosphate dehydrogenase diluted in triethanolamine buffer) or DHAP (0.085 U/µL glycerol-3-phosphate dehydrogenase, diluted in triethanolamine buffer). After an incubation of 30 minutes at 37°C, extinction at 340 nm was determined and the absolute amounts of GA3P and DHAP were calculated using a standard curve.

| MS-based analysis of carnosine reaction products
In order to identify a possible reaction product of carnosine and GA3P, GA3P (7.5 mM) was incubated in the presence or absence

| Determination of oxygen consumption rate (OCR) and extracellular acidification rates (ECAR)
For the determination of the OCR and the ECAR, 10 000 cells were

| Statistical analysis
Statistical analysis was carried out using SPSS (IBM; Version: 24.0.0.2 64-bit). False discovery rate (FDR) was calculated according to Benjamini and Hochberg. 24 For pairwise comparisons, a t test was performed. For multiple comparisons, a one-way ANOVA with the Games-Howell post hoc test was used. Principal component analysis (data were normalized to untreated control and auto-scaled) and metabolite pathway analysis (settings: Over Representation Analysis: Hypergeometric Test; Pathway Topology Analysis: Relative-betweenness Centrality; Pathway Database: Homo sapiens The Small Molecule Pathway Database) were performed using MetaboAnalyst. 25 If not stated otherwise, experiments were carried out in 6-tuplicate (independent experiments) and data are presented as mean ± standard deviation. Non-overlapping confidence intervals (CI 95%), P-value < .05 or FDR < 0.05 were presumed to be statistically significant.

| Carnosine affects metabolic pathways connected to glycolysis
Previous studies indicated that carnosine affects cancer cell metabolism by influencing the glycolytic pathway. 11,26 As neither its mode of action nor targets of carnosine were known, we employed a targeted method using GC-MS to study metabolites and pathways connected to glycolysis. Therefore, U87 glioblastoma  Figure 1F (for the complete results of the metabolic profiling see Figure S1).
The influence of carnosine on the PPP is highly interesting, as this pathway plays a critical role for proliferation, providing 5-phosphoribosyl diphosphate for DNA and RNA biosynthesis. 27 In accordance, treatment with 50 mM carnosine significantly increased the doubling time of U87, T98G and LN229 glioblastoma cells ( Figure 1E).
In addition, the influence of carnosine on the glycerol phosphate shuttle and the malate-aspartate shuttle suggests an inhibitory effect of carnosine on cellular redox homeostasis, as the malate-aspartate shuttle contributes ~20% to the total respiratory rate, and its inhibition leads to reduced production of glutamate from glucose. 28

| Carnosine inhibits mitochondrial respiration independent of the malate-aspartate and the glycerol phosphate shuttle
Our results suggest an inhibitory effect of carnosine on the glycerol phosphate shuttle and the malate-aspartate shuttle which are required for mitochondrial ATP production. Hence, we investigated the influence of carnosine on mitochondrial respiration using a Seahorse XF analyser. In this experiment ( idea. Therefore, we determined the OCR in HBSS with pyruvate as the only substrate, as mitochondrial ATP production from pyruvate is independent of the malate-aspartate shuttle and the glycerol phosphate shuttle. 29 As seen in Figure 2B,D,F, ATP-linked OCR was significantly higher in the presence of 5 mM pyruvate than in the presence of 25 mM glucose in all three cell lines. More important, we also observed reduced ATP-linked OCR under the influence of pyruvate in U87 and T98G but not in LN229 (for the complete analysis of data obtained from the Seahorse experiments see Figure S3). Therefore, we conclude that carnosine's impact on mitochondrial respiration is cell-specific and not necessarily dependent on its influence on the shuttle systems.

| Carnosine reduces the intracellular abundance of GA3P by a direct reaction with the metabolite
Next, we investigated whether reaction of carnosine with DHAP and GA3P may be responsible for their decreased abundance in the presence of the dipeptide. Therefore, both triose phosphates were incubated in vitro in the presence of the dipeptide ( Figure 4A).
We observed that equimolar concentrations of carnosine resulted

| D ISCUSS I ON
The naturally occurring dipeptide carnosine reduces tumour growth in vivo 8,9 and in vitro. [10][11][12][13] However, as pointed out in the introduction, neither the exact mechanisms behind this observation nor the primary targets of carnosine have been identified.
Using mass spectrometry, we detected that carnosine non-enzymatically reacts with GA3P in vitro similar to its previously described reaction with methylglyoxal. 30 As we also see reduced intracellular abundances of GA3P and DHAP in cultured cells, we presume this reaction to be responsible for the reduced abundances of these triose phosphates. Surprisingly, the reduced abundance of these metabolites was not unequivocally accompanied by a decreased glucose uptake rate or by a decreased lactate release rate. Although one has to note that a precise determination of the glycolytic flux would require more powerful methods, such as 13 Cbased metabolic flux analysis, this is an indication that the overall glycolytic flux is not decreased by the presence of the dipeptide.
More importantly, the reduced abundances of GA3P and DHAP are accompanied by reduced abundances of metabolites associated with the PPP, the glycerol phosphate shuttle and the malateaspartate shuttle. The PPP is especially important for tumour cells as it provides nucleotides for nucleic acid synthesis and NADPH required for fatty acid synthesis and scavenging of reactive oxygen species (for review see 31 ). Therefore, a decreased activity of this pathway will impair proliferation and defence against oxidative stress, and it has been discussed that the PPP could be in fact a potential target for cancer therapy. 32 In previous investigations, the anti-neoplastic of carnosine was shown to be associated with decreased ATP production. However, there have been different reports either pointing towards an effect on glycolysis, mitochondrial respiration or on both, depending on the cell type investigated. 11,15,17,18 As we observed a reduction of metabolites of the malate-aspartate shuttle and the glycerol phosphate shuttle in all three cell lines, it is straightforward to assume that this observation is responsible for reduced mitochondrial ATP production as these pathways deliver cytosolic NADH to mitochondrial oxidation. However, using pyruvate as substrate, only cells from the lines U87 and T98G responded to the presence of the dipeptide with reduced ATP-linked OCR but not cells from the line LN229. As LN229 cells do not respond with reduced ATP production even in the presence of glucose ( Figure S3), aside from a predicted impairment of the shuttles, we assume that the shuttles may also be of less importance for ATP production in U87 and T98G cells. In addition, the cell-specific response of mitochondrial ATP production to the presence of carnosine is highly interesting, pointing towards the possibility of a tumour-specific reactivity. A cell-specific response F I G U R E 3 Influence of carnosine on glucose uptake and l-lactate production. A, Cells were treated with or without 50 mM carnosine (Car) for 6 h. Then, 2-desoxy-d-glucose (2DG) uptake was determined as a measure for the glucose uptake rate (n = 6). B, Cells were treated with or without 50 mM Car and extracellular l-lactate was determined after 3 and 6 h to calculate l-lactate production rate (n = 6). Level of significance is indicated as: ***P < .0005; **P < .005; ns: P > .05  or mitochondrial respiration is difficult to resolve as these interactions are highly complex (for review see 34 ). At this point, it should also be emphasized that care has to be taken to compare the dipeptide's effect on different cancer cell types. Whereas in glioblastoma cells no effect on mTOR signalling was detected, 14 Zhang et al re- ported an inhibitory effect on mTOR/p70S6K-signalling in gastric carcinoma. 12 As increased GA3P levels lead to activation of mTOR, 35 a reduction of GA3P could be responsible for an inhibitory effect.
To solve the riddle why glioblastoma cells behave differently with regard to mTOR signalling would require further investigations on how phosphorylation/dephosphorylation of mTOR is controlled in different cells. However, both observations taken together argue against the notion that mTOR is a direct target of carnosine, pointing towards an upstream target such as GA3P.
As it is reported that carnosine suppresses HIF-1α signalling, 10,36 it is also tempting to speculate that inhibition of this pathway, which controls the expression of glycolytic enzymes, is responsible for the effects observed in this study. However, there are also reports that the dipeptide may increase HIF-1α activity accompanied by reduced tumour growth and extracellular acidification. 37 Figure 4A demonstrate that a concentration of >1 mM carnosine should be reached to reduce the available amount of GA3P to <50%. This is at least in the range of carnosine concentrations determined in kidney tissue of CN1 transgenic mice which were supplemented with a bolus of 200 mg carnosine. 42 In these experiments, carnosine concentrations reached ~0.5 mmol/kg 8 hours after supplementation. More interestingly, the authors could also demonstrate that in mice receiving a CN1 inhibitor, tissue concentrations increased ~7-fold, which is an interesting aspect considering carnosine as a therapeutic. In this context, one should also be aware that in the experiments of Qiu et al, 42  In conclusion, our investigation demonstrates that carnosine's anti-neoplastic effect is accompanied by reduced abundances of several metabolites associated with glycolysis and pathways dependent on it. We suggest that a non-enzymatic reaction of GA3P with the dipeptide is responsible for its reduced abundance that in turn leads to a decreased activity of the pentose phosphate pathway required for proliferation. However, we also observed an inhibitory effect on mitochondrial ATP production which is cell-specific and independent of carnosine's effect on glycolytic metabolites.

ACK N OWLED G EM ENTS
We like to thank Flamma (Flamma s.p.a. Chignolo di'Isola, Italy (http:// www.flamm agroup.com)) for the generous supply with very high quality carnosine (CarnoPure™) for all of our experiments. In addition, we like to thank Dr Hans-Heinrich Foerster from the Genolytic GmbH Alan Hipkiss for the fruitful discussions.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no potential conflict of interest.

AUTH O R CO NTR I B UTI O N S
HO performed the experiments and analysed the data. CB carried out the mass spectrometry measurements and identified the reaction products. JM revised the manuscript. HO and FG designed the study and wrote the manuscript. All authors reviewed and approved the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated in this study are available from the corresponding author on reasonable request.