Whole Body Distribution of Labile Coenzymes and Antioxidants in a Mouse Model as Visualized Using ¹H NMR spectroscopy

Abstract

Coenzyme A, acetyl-coenzyme A, coenzymes of cellular energy, coenzymes of redox reactions, and antioxidants mediate biochemical reactions fundamental to the functioning of all living cells. There is an immense interest to measure them routinely in biological specimens to gain insights into their roles in cellular functions and to help characterize the biological status. However, it is challenging to measure them ex vivo as they are sensitive to specimen harvesting, extraction, and measurement conditions. This challenge is largely underappreciated and carries the risk of grossly inaccurate measurements that lead to incorrect inferences. To date, several efforts have been focused on alleviating this challenge using NMR spectroscopy. However, a comprehensive solution for measurement of the compounds in a wide variety of biological specimens is still lacking. As a part of addressing this challenge, we demonstrate here that the total pool of each group of unstable metabolites offers a starting place for the representation of labile metabolites in biological specimens. Based on this approach, in this proof-of-concept study, we determine the distribution of the labile compounds in different organs including heart, kidney, liver, brain, and skeletal muscle of a mouse model. The results were independently validated using different specimens and a different metabolite extraction protocol. Further, we show that both stable and unstable metabolites distributed differentially in different organs, which signify their differential functional roles, the knowledge of which is currently lacking for many metabolites. Intriguingly, the concentration of taurine, an amino sulfonic acid, in skeletal muscle is > 30 mM, which is the highest for any metabolite in a mammalian tissue known to date. To the best of our knowledge this is the first study to profile the whole-body distribution of the labile and other high concentration metabolites using NMR spectroscopy. The results may pave ways for gaining new insights into cellular functions in health and diseases.

Graphical Abstract

INTRODUCTION

Coenzyme A (CoA), acetyl-coenzyme A (acetyl-CoA), coenzymes of cellular energy, coenzymes of redox reactions, and antioxidants mediate biochemical reactions fundamental to the functioning of all living cells, and their measurement provides insights into many cellular functions as well as biological status and dynamics. CoA and acetyl-CoA are ubiquitous cellular molecules, which mediate hundreds of anabolic and catabolic reactions including energy metabolism. Adenosine triphosphate (ATP) fuels a vast number of energy dependent biochemical processes and hence it is regarded as the energy currency of living cells. Adenosine diphosphate (ADP) and adenosine monophosphate (ADP) are precursors/products of ATP synthesis or hydrolysis. The levels of ATP, ADP, and AMP thus represent a measure of the energetics of the functioning cells and mitochondria.¹ Major redox coenzymes, NAD⁺ (nicotinamide adenine dinucleotide, oxidized), NADH (nicotinamide adenine dinucleotide, reduced), NADP⁺ (nicotinamide adenine dinucleotide phosphate, oxidized), and NADPH (nicotinamide adenine dinucleotide phosphate, reduced) mediate a vast number of biochemical reactions and are important indicators of normal and pathological conditions including heart disease, diabetes and cancer.^2-4 They are associated with a multitude of biological processes, including energy metabolism, mitochondrial functions, calcium homeostasis, regulation of oxidative stress, aging, and cell death. Investigations into the metabolism and function of these coenzymes are therefore of immense interest for uncovering fundamental cellular properties and treatment of diseases.^5-7 Glutathione is a ubiquitous and powerful cellular antioxidant that is critically required to protect cells from the deleterious effects of oxidative damage and free radical injury. It exists in two forms, reduced glutathione (GSH) and oxidized glutathione (GSSG). The magnitude and the ratio of GSH and GSSG are useful indicators of the status of health as well as numerous diseases associated with the oxidative stress and free radical injury. Hence, there is immense interest to measure the coenzymes and antioxidants reliably in biological specimens.

Currently, there are a limited number of methods for reliable and one-step measurement of energy and redox coenzymes, and antioxidants. The most often-used methods for analysis of redox coenzymes involve enzymatic assays, which are suboptimal as they necessitate separate protocols for analysis of each coenzyme or their ratios.^8-11 There were also efforts to analyze the coenzymes based on chromatographic separation and UV-vis absorption¹² or detection using targeted mass spectrometry (MS).^13,14 However, ion suppression, peak interference between oxidized and reduced forms, or in-source fragmentation, for example, of ATP and ADP to AMP pose a major challenge for their analysis using MS.¹⁴ Similarly, for the analysis of GSH and GSSG, numerous analytical methods have been developed over the past more than half century, which include spectrophotometry,^15,16 fluorometry^17-19 and mass spectrometry-based methods.^20-25 However, a major challenge, unconnected with any analytical method, is the instability of the coenzymes and antioxidants to specimen harvesting, extraction, and analysis conditions. This is largely an underappreciated challenge and it carries the risk of inaccurate measurements that lead to grossly incorrect inferences.

In the last several years, we have focused our efforts on addressing factors that affect the stability of the compounds. Using ¹H NMR spectroscopy, specimen harvesting, extraction, and analysis conditions were optimized for mouse heart tissue for analysis of ATP, ADP, AMP, NAD⁺, NADH, NADP⁺, NADPH, CoA, and acetyl CoA, all in one step.^26,27 Separately, a method was also developed for analysis of energy and redox coenzymes in human blood.²⁸ More recently, this method was extended for analysis of the antioxidants, GSH and GSSG using chemical derivatization step.^29,30 These methods promise new avenues for analysis of the labile metabolites. However, methods for their comprehensive analysis in a wide variety of tissue/biological specimens are still not available. Owing to the notoriously unstable nature of the co-enzymes, a method developed for one bio-specimen is not suitable for another. As a part of addressing this challenge, in this study, we demonstrate that total pool of each group of unstable metabolites offers a reliable method to measure the coenzymes and antioxidants in different organs and thus offers an opportunity for biological insights. Based on this approach, in this proof-of-concept study, we have determined the distribution of important labile metabolites in different organs including heart, kidney, liver, brain, and skeletal muscle from a C57BL/6J mouse model. Results obtained were validated using different specimens and a different metabolite extraction protocol, independently. The ability to obtain whole body distributions of the labile metabolites based on data from the conventional analysis methods represents a new dimension for profiling the labile metabolites in biological systems. We also show a differential distribution of many stable metabolites (obtained in the same NMR experiment), in different organs, the functional knowledge of which is currently lacking in many cases. Finally, this study shows for the first time that the taurine, an amino sulfonic acid, is the highest concentration metabolite found in a mammalian tissue known to date.

MATERIALS AND METHODS

Methanol, chloroform, monosodium phosphate (NaH₂PO₄), disodium phosphate (Na₂HPO₄), 3-(trimethylsilyl)propionic acid-2,2,3,3-d₄, sodium salt (TSP), and sodium azide were obtained from Sigma-Aldrich (St. Louis, MO). Deuterium oxide (D₂O) was obtained from Cambridge Isotope laboratories, Inc. (Andover, MA). Deionized (DI) water was purified using an in-house Synergy Ultrapure Water System from Millipore (Billerica, MA). All chemicals were used without further purification.

Mouse tissue harvesting and metabolites extraction:

The investigations using mouse tissue were performed with the approval of the Institutional Animal Care and Use Committee of the University of Washington. A total of twenty C57BL/6J wild type (WT) mice, one of the widely used mouse model strains, aged 3.5 to 6 months were used for method evaluation and metabolite analysis (Table S1). After each mouse was anesthetized, tissue specimens from heart, kidney, brain, liver, and skeletal muscle were separated quickly, and snap frozen in liquid nitrogen. We employed two commonly used metabolite extraction protocols,^26,28,31,32 one using methanol and the other using a mixture of methanol and chloroform.

For methanol extraction, the tissue specimens (~80 to 100 mg) were mixed with a solution of DI water and methanol (200 μL; 1:5 v/v, 4 °C) using lockable Eppendorf vials and then homogenized using disposable tissue pestles. A further 800 μL of cold water and methanol (1:5 v/v) was added, and each mixture was then vortexed and incubated on dry ice (−75 °C) for 30 min. Subsequently, the mixtures were sonicated in an ice bath for 10 min and centrifuged for 5 min at 2000 rcf and cold temperature (4 °C). The soluble extracts were separated, frozen using dry ice and lyophilized to dryness. The dried extracts were then mixed separately with a cold phosphate buffer (0.1 M; pH = 7.45; 4 °C) in D₂O containing 50 μM TSP (210 μL for 3mm NMR tube, 600 μL for 5mm NMR tube) and transferred to NMR tubes.

For the methanol and chloroform extraction, weighed tissue specimens were mixed with a 1.0 mL mixture of cold methanol and chloroform (1:2 v/v; 4 °C) in 2 mL Eppendorf vials and homogenized using a Tissue-Tearor handheld homogenizer and sonicated for 20 s. A further 800 μL cold chloroform/distilled water mixture (1:1 v/v) was added, and the sample was then vortexed and set aside for 30 min on ice to separate the solvent layers. Next, after centrifugation at 2000 rcf, the aqueous (top) layer was separated and filtered using 1.5 mL 0.2 μm syringe filters, frozen using dry ice and lyophilized to dryness. The dried extracts were mixed with phosphate buffer and TSP as described in the preceding paragraph.

NMR Spectroscopy:

NMR experiments for tissue extracts were performed at 298 K on a Bruker Avance III 800 MHz spectrometer equipped with a cryogenically cooled probe and z-gradients suitable for inverse detection. One dimensional NOESY pulse sequence with residual water suppression using presaturation, 10204 Hz spectral width, 5 s recycle delay, 128 transients and 32,768 time-domain points were used for ¹H 1D NMR experiments. NMR experiments were performed immediately after preparing the solutions and, for a portion of the samples, a second time 24 h after preparation to assess the stability of the compounds. Two-dimensional (2D) NMR experiments, including ¹H-¹H double quantum filtered correlation spectroscopy (DQF-COSY) and ¹H-¹H total correlation spectroscopy (TOCSY) experiments, were performed for representative samples from each type of tissue as described previously.²⁶ Chemical shifts were referenced to the internal TSP signal for both the ¹H 1D and 2D NMR spectra. The Bruker software package TopSpin version 3.5pl6 or 4.1.4 was used for NMR data acquisition processing, and analyses.

Peak assignments, metabolite identification and quantitation:

Assignment of peaks from the unstable and stable metabolites were based on the established literature and our own NMR spectral library of authentic compounds.^26-28,32 2D NMR experimental data were used for further confirming the assignments. Bruker AMIX software was used for peak integration. Concentrations of metabolites in different tissues were derived using the peak from the internal standard, TSP, and taking into account the weight of tissue used for the analysis. Absolute concentrations (in μM) were calculated based on the wet tissue weights and considering the wet tissue density as 1.00 g/cm³ for all tissue type; the density difference between different tissue types is less than 1%.^33-35 The labile metabolites were categorized into five different groups: energy coenzymes (ATP, ADP, and AMP), redox coenzymes (NAD⁺, NADH), phosphorylated redox coenzymes (NADP⁺, NADPH), antioxidants (GSH, GSSG), and CoAs (CoA and acetyl-CoA). The high energy metabolite, phosphocreatine, was grouped with creatine. Concentrations of metabolites were compared among different tissue type individually or by combining them in their respective groups. Statistical analyses were performed using two-sided Student’s t-test.

RESULTS AND DISCUSSION

NMR spectra show that the labile coenzymes and antioxidants are major components in all types of tissue specimens investigated (Figure 1 and S1). The spectra show distinct peaks for each compound, all of which can be quantitated individually. The limit of quantitation for the measured metabolites using the 800 MHz 5 mm cryoprobe was 1 to 2 μM after signal averaging of 128 transients. However, depending on the method used for metabolite extraction or condition used for analysis, different spectral patterns were obtained for labile metabolites in the same specimens. For example, the ATP, ADP, and AMP levels were altered drastically due to the undesired ATP hydrolysis, which yielded ADP and AMP (Figure 2). Similarly, NADH, NADPH, and GSH levels were attenuated drastically due to their oxidation to NAD⁺, NADP⁺, and GSSG, respectively (Figure S2). Because of such lability, the individually derived concentrations that are based on methods currently in widespread use are often poorly reproducible, which leads to challenges in deriving useful biological insights.

Figure 1.

Portions of 800 MHz ¹H NMR spectra of a mouse heart, kidney, skeletal muscle, brain, and liver tissue extracts with highlighting of characteristic peaks for the major coenzymes. NAD⁺: oxidized nicotinamide adenine dinucleotide; NADP⁺: oxidized nicotinamide adenine dinucleotide phosphate; NADH: reduced nicotinamide adenine dinucleotide; NADPH: reduced nicotinamide adenine dinucleotide phosphate; ATP: adenosine triphosphate; ADP: adenosine diphosphate; AMP: adenosine monophosphate; CoA: Coenzyme A; and IMP: Inosine monophosphate.

Figure 2.

Measured concentrations of ATP, ADP and AMP in heart, kidney, skeletal muscle, brain, and liver tissue from mice. The data were obtained using ¹H NMR spectroscopy based on tissue metabolite extraction using (a) methanol or (b) methanol and chloroform.

On the other hand, we found that pooling the metabolites within each group considers the instability of the metabolites and hence provided results that show distinct distribution of the metabolites in different organs (Figure 3). The results were independently validated based on the analysis using separate biospecimens and a different extraction protocol, which show practically identical distribution of the labile compounds for different organs (Figure S3). The differences for the labile metabolite groups between different mouse tissues for both the test and validation sets of biospecimens were comparable, statistically (Table S2). This is in stark contrast to the individual metabolites that alter depending on the method used for metabolite extraction or analysis owing to their instability (Figures 2 and S2). These results provide a new route for visualization of the unstable metabolite distribution in different organs. The knowledge of the individual metabolite levels is desirable as they offer a greater insight into biological processes; the ratios between oxidized and reduced forms represent a measure of redox status and oxidative stress in cells. However, currently, there is a lack of simple and one step method for their comprehensive analysis in different organs. Even a most recent patent describes the need for using several different protocols for measuring redox coenzymes and antioxidants, individually.³⁶ Therefore, the whole-body distribution of the labile metabolites obtainable using the new approach potentially helps to gain insights into their roles in cellular functions and to help characterize the biological status.

Figure 3.

Distribution of (a) energy coenzymes; (b, c) redox coenzymes, (d) antioxidants, and (e) CoA+ Acetyl-CoA in heart, kidney, skeletal muscle, brain, and liver of mice. The data were obtained using ¹H NMR spectroscopy based on metabolite extraction using methanol. GSH: Glutathione, reduced; and GSSG: Glutathione, oxidized. See Figure S3 for independent validation of the results obtained using different mice and a different extraction procedure.

Heart and skeletal muscle exhibited the highest levels of the energy coenzymes (ATP/ADP/AMP), with no significant difference between the two organs (Figures 3(a) and S3(a), and Table S2). These results are in agreement with the results obtained using in vivo ³¹P NMR, where ATP levels in human heart and calf muscle were shown to be comparable.³⁷ On the other hand, kidney and brain tissues exhibited significantly lower amounts compared to the heart and skeletal muscle. Liver showed the least amount of the energy co-enzymes, which agrees with the fact that hepatocytes contain lower number of mitochondria (1000 to 2000 in each cell)³⁸ compared to heart muscle cells (5000 to 8000 in each cell).³⁹ NAD/NADH were highest in kidney and heart, followed by the liver, while skeletal muscle and brain had the least amount (Figures 3(b) and S3(b), and Table S2). The high NAD/NADH levels found in heart, kidney and liver tissues agreed with results reported earlier.⁴⁰ The levels of NADP⁺/NADPH followed a similar trend; however, they were lower by a factor of about 5 when compared to the actual concentrations/g of NAD⁺/NADH (Figures 3(c) and S3(c), and Table S2). Further, interestingly, the distribution trend for CoA/Acetyl-CoA was similar to NAD⁺/NADH and NADP⁺/NADPH, with the highest levels found in kidney, heart, and liver, and the least in brain and skeletal muscle (Figures 3(e) and S3(e), and Table S2). The antioxidants GSH/GSSG were remarkably higher in the liver, whereas they were lowest in skeletal muscle (Figures 3(d) and S3(d), and Table S2). The distribution trend for glutathione in different organs agreed somewhat with that reported previously.⁴¹ The levels of CoA and acetyl CoA were low or not detected in skeletal muscle and brain compared to other organs (Figures 3 and S3). The reason for the absence or low levels in skeletal muscle and brain is not known; however, it may potentially arise from the increased mitochondrial activity and higher flux through TCA cycle.

In addition to the labile metabolites, several stable metabolites were highly concentrated and were characteristic of specific organs; they were either absent or extremely low in concentration in other organs. For example, N-acetyl aspartate (NAA), γ-aminobutyric acid (GABA), and myoinositol were characteristic of brain, hippurate and myoinositol were characteristic of kidney, and carnosine and anserine were characteristic of skeletal muscle (Figures 4-6 and S4). Interestingly, while the roles of some of these metabolites are well-known, for others the roles are still not understood clearly. For example, NAA is a well-known neuronal marker, which is synthesized in neuronal mitochondria⁴² and GABA is the major inhibitory neurotransmitter in the brain. Hippurate is the major metabolite excreted in urine and its biosynthesis occurs within the mitochondrial matrix.⁴³ In humans, hippurate is thought to be synthesized in the liver as well as in the kidney.^44-46 However, to date, there is a lack of clarity as to which tissues is its major source. In accordance with the literature, our NMR results show that hippurate is detected in both liver and kidney; however, the amount of hippurate is extremely low in the liver compared to kidney (Figure 4). These results clearly indicate that kidney is the major source of hippurate synthesis, at least in mice. Similarly, the distribution and physiological roles of carnosine and anserine have not been fully understood to date,⁴⁷ although some studies report that they are highly concentrated in muscle and brain,⁴⁸ and exhibit antioxidant activity and stabilize pH in skeletal muscle,^48,49 However, contrary to earlier reports, both carnosine and anserine were detected in a remarkably high concentration only in skeletal muscle (Figure 4). Myoinositol, although less specific to a particular organ is present at a far higher concentration in kidney and brain.

Figure 4.

Metabolites that are characteristic of a specific organ in mice obtained using ¹H NMR spectroscopy based on tissue metabolite extraction using methanol. See Figure S4 for an independent validation of these results obtained using different mice and using a different extraction procedure.

Figure 6.

Portions of 800 MHz ¹H NMR spectra of mouse heart, kidney, skeletal muscle, brain, and liver tissue extracts with highlighting of characteristic peaks for a few high concentration metabolites.

In addition, there were many other high concentration metabolites that were invariably found in all the organs investigated. As an example, Figures 5-7 and S5 describe some of these metabolites, including 3-hydroxyburyrate, aspartate, creatine/phosphocreatine, glucose, glutamate, glutamine, glycine, mannose, phosphocholine, and taurine. Their distribution in different organs was unique for each metabolite and the results were validated, independently, using different specimens and a different extraction protocol as was performed for the labile metabolites. One important example is taurine, an amino sulfonic acid, which is one among many such compounds detected in all organs. Taurine is a natural amino acid known to be present abundantly in mammalian tissue. Intriguingly, our NMR results indicate that the concentration of taurine in skeletal muscle is >30 mM. This is the highest concentration for a small molecule metabolite in mammalian tissue known to date. Further, the results demonstrate that its concentration in different organs varies systematically and such distribution in different organs had not been reported (Figure 7; Table S3). These results were consistent across all the samples investigated. It is known that the taurine concentration is species dependent and is found in high concentration in most mammalian tissues.⁵⁰ It is particularly high in cardiac and skeletal muscle, and is considered as the most abundant free amino acid in skeletal muscle.^50,51 Its concentration is reported to be in a range as high as 1 to 20 micromol/g tissue in vertebrate animals, which corresponds to about 1-20 mM.⁵² Hence, our results on taurine concentrations are not inconsistent with the literature but explicitly highlight the underappreciated fact about its concentration and provide a more systematic distribution in different organs. Since its discovery nearly a century ago, taurine has been attributed to several cellular functions, the mechanisms of which, however, are still not clearly understood to date.^53,54 Phosphocreatine is central to cellular energetics. Under ex vivo measurement conditions, often it gets converted to creatine (Figure S5). Hence, we have used the sum of creatine and phosphocreatine for comparison between different tissues. Interestingly, the ratio for creatine/phosphocreatine between skeletal muscle and heart (Figure 7 and S5; ratio ~2.4) is identical to the ratio for phosphocreatine between calf muscle and heart in humans measured under in vivo conditions.³⁷ Similarly, many other highly concentrated metabolites such as 3-hydroxybutyrate, aspartate, glucose, glutamate, glutamine, glycine and mannose were distributed differentially across the various organs (Figures 5-7, S5). Such a systematic distribution of the metabolites potentially indicates their specialized functions in the different organs.

Figure 5.

Portions of 800 MHz ¹H NMR spectra of mouse heart, kidney, skeletal muscle, brain, and liver tissue extracts with highlighting characteristic peaks for a few high concentration metabolites. The singlet peak around 3.35 ppm is from the residual methanol solvent and the singlet peak around 3.7 ppm in brain spectrum is due to an impurity."

Figure 7.

Distribution of metabolites in heart, kidney, skeletal muscle, brain, and liver tissue in mice obtained using ¹H NMR spectroscopy based on tissue metabolite extraction using methanol. See Figure S5 for an independent validation of these results obtained using different mice and a different extraction procedure.

The mouse model is a widely used experimental model for investigating human health and diseases. The knowledge of the whole-body distribution of metabolites in mouse tissues therefore provides an important baseline for investigations of cellular metabolism. To date, a few efforts have been made to profile metabolites in different organs in mice. However, there are no studies so far that provide the distribution of the labile coenzymes and antioxidants that are fundamental to cellular functions, even though they are detected abundantly by the conventional metabolite analysis methods. The notoriously unstable nature of these compounds is the major reason for the lack of such knowledge. Even worse, the significantly varying peak intensities in the NMR spectra, from one sample preparation method to another (Figure 2), has resulted in a gross misrepresentation of the levels and sometimes even the identities of labile metabolites in numerous published studies. Efforts during the last several years in our laboratory have established their identity in NMR spectra and has resulted in solutions to address the challenges associated with their quantitation.^26-30 Combining this knowledge, in this study, we have established the distribution of labile metabolites in different organs of one of the widely used strains of the mouse model (C57BL/6). This study is the first of its kind and provides an useful baseline values of whole-body distribution of the labile metabolites.

In conclusion, the labile metabolites, which include energy and redox coenzymes as well as antioxidants, are fundamental to many cellular functions. To date, their reliable measurement ex vivo is hampered by their notoriously unstable nature. NMR spectroscopy is extremely sensitive to assess the hydrolysis or oxidation of the coenzymes and antioxidants and can simultaneously and quantitatively detect the intact as well as hydrolyzed/oxidized forms. Using NMR and careful sample preparation has enabled the unambiguous identification and quantitation of the major labile cellular compounds. Here, the use of these methods enabled the visualization of whole-body distribution of the coenzymes and antioxidants based on the sum of the labile metabolites within each group or class. The distribution of the compounds was validated, independently, using different samples and sample processing protocols; these results indicate that the distribution of metabolites is reproducible and the distribution outweighs errors, if any, caused by differences in sample processing.

Apart from the labile metabolites, many other high concentration metabolites that were either unique or detected in all the organs were also investigated; those detected in all organs were uniquely distributed. In particular, we found that the concentration of taurine in skeletal muscle is > 30 mM. While it was known that taurine is an abundant free amino sulfonic acid in mammalian tissue and most abundant in skeletal muscle, it is a somewhat surprising discovery that the taurine concentration is the highest for any metabolite in a mammalian tissue ever known to date. Along the same lines, hippurate is one of the very few high concentration metabolites excreted in urine, though clear evidence for the major source of its synthesis was lacking. Our results that show extremely high concentrations of hippurate in kidney tissue, indicate that the kidney is the major source for its synthesis.

Overall, this is the first study to provide the whole-body distribution of stable and unstable metabolites in a mouse model. Numerous findings reported in this study are in agreement with several prior reports, which stress the robustness of the new approach to obtain the labile metabolite distribution. Considering that there is a lack of understanding of the functions of many metabolites, their unique distribution in different organs determined in this study should potentially serve as a baseline data for hypothesis generation and further investigations of their cellular functions. In this proof-of-concept study, we have used one of the widely used mouse strain (C57BL/6). Further studies using different mouse strains will strengthen generalizability of the results described herein.

Supplementary Material

Supplementary Material

Table S1: Mice used for this study.

Table S2. P-values for comparison of labile metabolite groups.

Table S3. P-values for comparison of stable metabolites.

Figure S1. ¹H NMR spectra of mouse heart, kidney, brain, liver, and skeletal muscle.

Figure S2. Portions of ¹H NMR spectra highlighting peaks from the coenzymes and antioxidants.

Figure S3. Distribution of labile metabolites in different organs.

Figure S4. Distribution of characteristic metabolites in different organs.

Figure S5. Distribution of a few stable metabolites in different organs.