Six-week-old male BKS-db/db and db/m mice were purchased from Cavens Laboratory Animals Co., Ltd. (Changzhou, China. License Number: SCXK 2016-0010). They were housed at room temperature (22±2°C) under a 12-h dark/12-h light cycle and fed a standard chow diet ad libitum until 24weeks of age. Body weight was measured monthly, and overnight fasting abdominal blood glucose was tested using a tail nick and glucometer (Roche Accu-Chek Advantage Meter). A 24-h urine samples were collected from non-fasting mice placed in metabolic cages at 12 and 24weeks of age, respectively. After fasting for 12h, the mice were euthanized at 24weeks of age with 1% pentobarbital, the blood samples were collected through the angular vein and centrifuged at 3500g for 10min at 4°C to acquire serum. Their kidneys were immediately extirpated and flash-frozen in liquid nitrogen for 5s, then transferred to cryogenic vials and kept at −80°C until further analysis. The left kidneys were used for AFADESI–MSI analysis (db/m, n=2; db/db, n=2) or integrated metabolomics analysis (db/m, n=5; db/db, n=4), the contralateral kidneys were used for histopathological examination and real-time PCR experiments (db/m, n=7; db/db, n=6).

The experimental procedures were approved by the Animal Care and Use Committee of Shanghai Jiao Tong University (Shanghai, China).

Analyses of serum were performed by the Servicebio Technology Co., Ltd. (Wuhan, China). The concentration of urine albumin was detected using a mouse albumin ELISA kit (Abcam), then multiplied by the 24-h urine volume of each mouse to calculate the 24-h urine albumin excretion.

For histopathological examination, the right kidneys were split into two parts in an ice bath. One part was fixed in 10% formalin and sectioned at a thickness of 3μm using a Leica CM1860 cryostat (Leica Microsystem Ltd., Wetzlar, Germany) at −20°C for periodic acid–Schiff (PAS) staining and IHC examination. Portions of the renal cortex (1×mm³ in volume) were dissected and maintained in 2.5% glutaraldehyde for analysis using a transmission electron microscope. For IHC staining, the tissue sections were incubated overnight at 4°C with specific primary antibodies (Supplementary Table S1); then, a PV-9000 two-step immunohistochemical kit (Zhongshan Goldenbridge Biotechnology Ltd. Co., Beijing, China) was used, followed by counterstaining with Mayer hematoxylin and dehydration. For the quantification of glomerular areas, we took a pathological picture from the PAS staining image of each sample and measured 20 glomerular areas, respectively, using a NanoZoomer Digital Pathology Image Viewer (NanoZoomer 2.0, Hamamatsu Photonics, Hamamatsu City, Japan).

As previously mentioned, the integrated metabolomics analysis in this study was conducted by GC–MS and UPLC–MS. The methodology has previously been described (Cao et al., 2020). In brief, GC–MS analysis was conducted using an Agilent 7890B gas chromatography system coupled to an Agilent 5977A MSD system (Agilent Technologies Inc., CA, United States). UPLC–MS analysis was performed using a Dionex Ultimate 3,000 RS UHPLC system fitted with Q-Exactive quadrupole-Orbitrap mass spectrometer equipped with a heated electron spray ionization (ESI) source (Thermo Fisher Scientific, Waltham, MA, United States) to analyze metabolic profiling in both positive and negative ion modes. The detailed sample preparation, workflow of GC–MS, and UPLC–MS analysis of kidney homogenates are described in the Supplemental methods.

The kidney samples were dissected carefully to separate the cortex and medulla on ice. Total RNA was extracted from the samples using TRIzol (Invitrogen, Carlsbad, CA), and the cDNA was synthesized using the reverse transcript reagents (Takara, Otsu, Japan). Real-time PCR was conducted using an SYBR Green RT-qPCR Kit (ABclonal, Wuhan, China) and a 7,500 Real-Time PCR System (Applied Biosystems, Waltham, MA). The relative gene expression levels were calculated using the 2ΔΔCT method and analytical data were adjusted with the mRNA expression of β-actin as an internal control. Specific primers used are listed in Supplementary Table S2.

Body weight and overnight fasting abdominal blood glucose (FBG) levels were significantly higher in db/db mice than in db/m mice during the observation period (Figures 2A,B). In addition, 24-h urine albumin excretion (UAE) levels were significantly higher in db/db mice at 12 and 24weeks of age with glomerulomegaly compared with db/m mice (Figures 2C,D), revealing renal damage. Representative histopathological and electron micrographic data are shown (Figure 2E). PAS staining images (Figure 2E; upper) indicated prominent glomeruli hypertrophy accompanied by mesangial expansion in db/db mice, while the fusion of the podocyte foot processes (red arrows) was noticeable according to electron microscopy images (Figure 2E down) in db/db mice.

Serum levels of renal function and lipid indicators are listed in Table 1. The levels of serum glycated serum protein (GSP), serum uric acid (UA), triglyceride (TG), cholesterol (CHO), and high-density lipoprotein (HDL) were significantly higher in db/db mice compared with db/m mice, indicating substantial glucose and lipid dysregulation induced by diabetes. In addition, the level of malondialdehyde (MDA), one of the by-products of lipid peroxidase and thus an indicator of the extent of lipid peroxidation, was also significantly increased in db/db mice. However, there was no clear difference between the two groups in the levels of serum creatinine (SCr), possibly because of the limitations of specificity in the detection of type 2 DKD in this mouse model (Giralt-Lopez et al., 2020). Together, these pathological and biochemical parameter alternations indicate the progression of type 2 DKD (Brosius et al., 2009).

Kidney tissues consist of the cortex, medulla, and pelvis, all of which have distinctive main functions. In this study, we focused on the renal cortex and medulla portions. H&E staining was implemented on frozen tissue sections to distinguish histological features by a renal pathologist, and AFADESI–MSI experiments were performed in both the positive and negative ion mode on the adjacent tissue sections to acquire molecule profiling of the metabolites. Figure 3A illustrates one representative histology-specific MSI in the positive ion mode, in which the distribution of L-carnitine matches well with its histological features. Moreover, as depicted in the figure, the ion intensities were significantly different between db/m mice and db/db mice among the cortex and medulla, respectively. In total, 509 and 507 biologically informative peaks were observed in the renal cortex and medulla, respectively, for the negative ion mode, whereas in the positive ion mode, 724 and 728 peaks were probed in the renal cortex and medulla, respectively (not shown). In order to explore the overall differentiation of metabolic molecules in the DKD, OPLS–DA was used to choose histology-specific metabolic biomarkers based on the MSI pixel point in the negative ion mode (Figure 3B) and positive ion mode (Supplementary Figure 1A). As a result, both the renal cortex and medulla parts, respectively, could be clearly distinguished between db/db and db/m mice, indicating striking separations in the metabolite profiling of not only the cortex but also the medullary portion induced by T2DM. Next, we selected discriminating metabolites with clear identification potential from the two groups, using a variable importance in projection threshold (VIP)≥1.0 generated by OPLS–DA, and p<0.05 according to one-way analysis of variance (ANOVA). Volcano plots were drawn to visualize the dysregulated metabolites in the renal cortex and medulla between the two groups (Figure 3C for negative ion mode; Supplementary Figure 1B for positive ion mode). To further identify significantly perturbated metabolic pathways that may be involved in DKD, we performed metabolic pathway matching analysis by importing the discriminating metabolites into the kyoto encyclopedia of genes and genomes (KEGG) database (Sun et al., 2019; Chen et al., 2021b). We observed that taurine and hypotaurine metabolism, arginine and proline metabolism, histidine metabolism, biosynthesis of unsaturated fatty acids, and fatty acid degradation pathways were significantly dysregulated in DKD (Figure 3D for cortex portion, Supplementary Figure 1C for medulla portion). In total, the in situ molecular profiles of the kidneys in a T2DM mouse model based on a histology-specific analysis were first presented, suggesting that several metabolic pathways were remodeled in both the cortex and the medulla portion of a diabetic kidney.

To further investigate the disturbance of these metabolic pathways induced by DKD, 22 discriminating metabolites with a mass accuracy of <5ppm involved in these metabolic pathways were tentatively selected (Table 2; Figures 4, 5). The distribution of these metabolites clearly matches well with renal histology. In db/m mice, many of the detected metabolites were distributed throughout both the renal cortex and medulla, such as spermidine, spermine (Figure 4B), and L-carnitine (Figure 5). However, several were located within different parts of the kidney. In particular, glutathione disulfide (GSSG) was mainly detected within the cortex (Figure 4A), while L-histidine (Figure 4C) was abundant along the renal corticomedullary junction. Based on histology, many of these metabolites are specifically distributed in the kidney tissues, suggesting that they may perform different function in different regions.

The metabolites were evenly distributed in their particular regions of the kidney. However, compared with db/m mice, there were significant alternations in the diabetic kidney. Notably, diabetes-induced kidney metabolic reprogramming not only occurs in the cortex portion (which contains most of the glomerulus and proximal renal tubules that serve crucial filtration, reabsorption, and endocrine functions), but also in the medulla (which mainly performs reabsorption function). Compared with db/m mice, the relative abundances of GSSG were much greater in the renal cortex of db/db mice. Meanwhile, clear ion intensities are noted in the renal medulla of db/db mice, whereas very little was detected in the same portion of db/m mice (Figure 4A). In addition, taurine, spermidine, and spermine were observed at lower relative abundances in both the renal cortex and medulla of db/db mice (Figures 4A,B). In the renal cortex, diabetic mice presented relatively higher intensities of histamine (Figure 4C); however, the relative ion intensities of L-palmitoyl-carnitine and propionyl-carnitine (Figure 5) were lower in the diabetic renal medulla. Moreover, a list of fatty acids with different lengths of carbon chain and unsaturation levels was observed in our MSI study (Figure 5), many of which showed different distribution and abundance characteristics. Molecular distributions of saturated long-chain fatty acids, including FA (14:0), FA (16:0), and FA (18:0), presented higher intensities in the cortex and/or medulla of db/db mice, suggesting accumulation in these important regions when exposures to DKD. It was also observed that long-chain polyunsaturated fatty acids, including FA (20:2), FA (22:4), and FA (22:5), were significantly upregulated in both the cortex and the medulla portions of the diabetic kidney. However, two monounsaturated fatty acids detected in the present study showed reduced relative ion intensities in the kidney of db/db mice. The abundance of FA (16:1) and FA (18:1) was significantly reduced in the renal medulla and cortex, respectively. In summary, we discovered several significant metabolic pathway transformations in situ induced by DKD.

We also examined the whole renal tissue using traditional metabolic methods to detect the total amounts and types of metabolites. Here, we performed integrated metabolomics studies using GC–MS and UPLC–MS techniques together to improve the accuracy, integrity, and sensitivity of the identification of the phenotype-related metabolites (Zeki et al., 2020). A total of 1,544 peaks were detected by GC–MS and UPLC–MS (342 and 1,202 peaks were obtained from GC–MS and UPLC–MS, respectively, data not shown) from renal tissue. Scoring plots generated from OPLS–DA models presented a clear separation between db/db mice and db/m mice (Figure 6A). Next, we linked the metabolites with VIP score≥1.0 to metabolic pathways using the KEGG database (Figure 6B). Differences in taurine and hypotaurine metabolism, arginine metabolism pathways, histidine metabolism, and biosynthesis of unsaturated fatty acids were also observed, as in our MSI results. Moreover, purine metabolism and glucose metabolism processes, such as citrate cycle (TCA cycle), were enriched. A heat map for each metabolite involved in the metabolic pathways of interest and a bar graph for the fold changes of the metabolites of db/db compared with db/m mice is presented (Figure 6C). Interestingly, in the MSI experiments, the expression of putrescine, spermidine, L-carnitine, propionyl-carnitine, FA (18:1), and FA (20:2) was significantly different between the two groups, resulting in no statistical differences in metabolome analysis between db/db and db/m mice (although the trends were consistent to MSI data). This confirms that traditional metabolomics approaches using bulk tissue measurements result in the average measurements, which may lose key information from different tissue areas (Andersen et al., 2021).

Investigation into the metabolic pathway alternations involved in db/db mice was performed using the KEGG database (Figure 7). We verified that several glycolysis and TCA cycle intermediates were accumulated in the diabetic kidney, such as phosphoenolpyruvate (PEP), fumaric acid, and malate, suggesting an upregulation of glycolysis and the TCA cycle due to the excessive glucose inflow in DKD (Figure 7A). This finding is compatible with previous studies (Sas et al., 2016; Hasegawa et al., 2020). Moreover, several purine nucleoside concentrations were upregulated in db/db mice, indicating the upregulation of the purine metabolism. In contrast, the lower level of taurine indicated a reduced taurine synthetic metabolism (Figure 7A). The accumulation of long-chain saturated and polyunsaturated fatty acids was also shown (Figure 7B). Together, and consistent with our MSI data, the metabolic pathway alternations were confirmed according to metabolomic pathway analysis using the entire tissue bulk.

As crucial links in metabolic pathways, the rate-limiting enzymes control the rate and direction of metabolic processes. In the present study, we observed several remodeling metabolic pathways occurring in DKD, including taurine and hypotaurine metabolism, arginine and proline metabolism, histidine metabolism, biosynthesis of unsaturated fatty acids, and fatty acid degradation pathway. We next detected the expression level of rate-limiting enzymes in specific metabolic pathways. We employed real-time PCR analysis to detect the mRNA expression levels of rate-limiting enzymes involved in these metabolic pathways (Figure 8A). We further examined the in situ expression levels of these enzymes by performing IHC on kidney tissues (Figure 8B). The detailed rate-limiting enzymes are presented in Supplementary Table S3. Specifically, sulfinoalanine decarboxylase (CSAD) catalyzes the biosynthesis of taurine. Compared with db/m mice, there was a clear decrease in mRNA expression of CSAD in the renal cortex of db/db mice, but no statistical difference in medulla portions (p=0.06). In addition, IHC analysis indicated that CSAD was mainly expressed in the glomeruli with a small amount in the tubules in db/m mice, although there was a significant decrease in both the glomeruli and tubules of db/db mice. Carnitine O-palmitoyl-transferase 1 (CPT1) is the rate-limiting enzyme of fatty acid oxidation (FAO), allowing long-chain fatty acids shuttling from the mitochondrial outer membranes into the mitochondrial matrix. This was mainly located in the tubules of db/m mice and reduced in db/db mice according to IHC data. Spermidine/spermine N1-acetyl transferase 1 (SAT1) catalyzes the acetylation of spermidine and spermine to deplete them, histidine decarboxylase (HDC) induces the decarboxylation of histidine to form histamine, and fatty acid synthase (FAS) stimulates the formation of long-chain fatty acids. In db/db mice, the mRNA expression levels of SAT1 and FAS were substantially upregulated in both the cortex and medulla portions, whereas the significant mRNA accumulation of HDC was detected only in the cortex. Further, IHC staining of SAT1 indicated an increasing trend in tubulars in db/db mice kidney, and the subsequent IHC assay of HDC and FAS showed that they were upregulated in both glomeruli and tubulars of db/db mice. Together, using these approaches, we verified that taurine biosynthesis, arginine and proline metabolism, histidine metabolism, unsaturated fatty acid biosynthesis, and fatty acid degradation pathways were altered in DKD.

It is well known that, as an essential antioxidant, taurine (a sulfur-containing amino acid) participates in the prevention of oxidant generation (Schaffer et al., 2009) and lipid peroxidation (Parvez et al., 2008), and has increasingly captured attention due to its inhibitory effects of DKD (Koh et al., 2014; Zhang et al., 2020b). Moreover, taurine may protect mitochondria due to a reduction in mitochondrial reactive oxygen species (ROS) production and a partial recovery of mitochondrial Mn-superoxide dismutase (Chang et al., 2004). The MSI data based on histology in the present study indicate that the ion intensity of taurine in the renal cortex and medulla of db/db mice is significantly lower than that in db/m mice, as also confirmed by metabolome data. Based on the taurine metabolic pathway, CSAD is a crucial rate-limiting enzyme for the biosynthesis of taurine (Park et al., 2017), and its expression plays a pivotal role in the pathogenesis of T2DM in β-cells (Chen et al., 2021a). We assume that the decreased levels of taurine in the diabetic kidney may be attributed to the downregulated expression of CSAD. Real-time PCR and IHC staining data were consistent with our assumption. Interestingly, in db/m mice, the IHC analysis showed that CSAD was mainly expressed in the glomeruli with a small amount in the proximal tubules, which is consistent with the high abundance of taurine in the cortex portions according to our MSI data. To our knowledge, this is the first time that the molecular profiles of taurine have been detected with MSI in type-2 DKD mouse kidney tissue. The decreased abundance of taurine in the diabetic kidney is probably due to the low expression of CSAD.

In mammals, arginine can be catalyzed to polyamines, including putrescine, spermidine, and spermine. There is growing evidence suggesting that the dysregulation of polyamine metabolism can cause changes in high glucose-induced energy perturbation, including streptozotocin (STZ)-induced diabetic cardiomyopathy (Wang et al., 2020a), type 1 DKD (Zhang et al., 2021), and retinopathy (Liu et al., 2020), thus making it a promising target for therapeutic intervention. It has been reported that, as ROS scavengers, spermine and spermidine can protect DNA from free radical attacks, regulating cell proliferation, differentiation, and apoptosis (Pegg, 2009; Wang et al., 2020a). SAT1, which catalyzes the acetylation of spermidine and spermine, is the rate-limiting enzyme in the catabolism of polyamine, and overexpression of SAT1 results in an overall depletion of spermidine and spermine (Mandal et al., 2013). In this study, the abundance of spermidine and spermine was dramatically decreased in the diabetic kidney compared with db/m mice, especially in the cortex portions, although the metabolome data showed no significant difference in spermidine between the two groups, which may be due to the whole kidney tissue bulk analysis. According to these results, we speculated that the downregulated expression of spermidine and spermine may be the result of ascending SAT1. The subsequent real-time PCR and IHC validation were then performed to evaluate the mRNA and protein expression of SAT1 in the tissue, respectively. As predicted, db/db mice showed a higher mRNA expression of SAT1 in both the renal cortex and medulla, and also a stronger level of SAT1 expression in situ compared with db/m mice. These findings suggest a potential role of SAT1 in the development of type 2 DKD, as reported in a previous study (Zhang et al., 2021).

Histamine, decarboxylated from histidine induced by HDC, has four cognate G protein-coupled receptors, namely, H1R to H4R. Histamine exerts pro-fibrotic and pro-inflammatory effects through these receptors during the development of DKD, wherein the inhibition of H1R can maintain the integrated morphology of podocytes (Veglia et al., 2016), whereas inhibition of H4R can attenuate the reabsorptive dysfunction in proximal tubules in STZ-induced DKD (Pini et al., 2018). In particular, type-1 DKD rats showed significantly higher HDC activity in the kidney, and thus an increased abundance of histamine (Gill et al., 1990). Our MSI data suggested that histamine biosynthesis was severely upregulated in the renal cortex of db/db mice, and the subsequent real-time PCR performed to test the mRNA expression of HDC also showed a significantly higher level in the cortex portions. Interestingly, mast cells, which are well known as the major producer of histamine in tissues, could be scarcely found in the present study according to histological examinations. The IHC staining demonstrated an increased tendency of HDC mainly in the proximal tubules and to a lower extent in the glomeruli in db/db mice, suggesting the existence of a local histamine pool of intrinsic renal cells induced by type 2 DKD (Pini et al., 2019).

It is known that the kidney needs a large energy supply for its filtering and reabsorption functions. Renal tubules, especially the proximal convoluted tubule, thick ascending loop, and distal convoluted tubule that are mainly located at the outer medulla and corticomedullary junction, contain most of the renal mitochondria, in which ATP generation mainly depends on FAO (Bhargava and Schnellmann, 2017; Forbes and Thorburn, 2018). Dysfunctional FAO has been found in the development of DKD in several studies, although the mechanism remains controversial (Sas et al., 2016; Sha et al., 2020). On the other hand, the accumulation of long-chain fatty acids induced by DKD may negatively control FAO by creating a lipotoxicity environment inside the proximal tubules (Ruggiero et al., 2014; Simon and Hertig, 2015). By binding to albumin for the formation of fatty acid-binding proteins (FABPs), fatty acids can be filtered by glomeruli and reabsorbed by tubules (Bhargava and Schnellmann, 2017). As albuminuria excretion increases during the development of DKD, the filtration of FABPs is overloaded, resulting in a significantly increased reabsorption status, which induces severe damage in the tubules (Cobbs et al., 2018). Of note, according to our MSI and metabolome data, fatty acids with different lengths of carbon chains and numbers of double bonds showed differing abundance and location between db/db and db/m. The concentration of long-chain fatty acids, including both saturated [e.g., FA (16:0), FA (18:0)] and polyunsaturated [e.g., FA (20:4), FA (22:5)] acids, was significantly increased in the diabetic kidney, whereas monounsaturated fatty acids [e.g., FA (16:1), FA (18:1)] were less enriched compared with db/m mice. Similarly, glomerular and tubulointerstitial lipid accumulation has been found in both type 2 DKD patients (Herman-Edelstein et al., 2014) and db/db mice (Wang et al., 2005), and we further identified complex fatty acid composition on the spatial level to validate the accumulation. The higher abundance of FA (16:0) may induce inflammation and apoptosis in the renal proximal tubular cell (Soumura et al., 2010) and mitochondrial oxidative stress in podocytes (Xu et al., 2015), and the kidney is easily damaged by ROS due to the high level of polyunsaturated fatty acids (Zhang et al., 2020a). On the other hand, the administration of monounsaturated fatty acid FA (16:1) to KKAy mice (a T2DM model with low insulin sensitivity) can reduce fasting glycemia and insulin resistance in parallel with reduced relative mRNA expression of FAS (Yang et al., 2011). These results indicate the latent role of various types of fatty acids in progression of DKD, which may be involved in the progression of DKD.

Another significant finding for fatty acid metabolism in DKD is the dysregulation of FAO. As previously mentioned, proximal tubules utilize non-esterified fatty acids [e.g., FA (16:0)] via binding to carnitine, resulting in the formation of acyl-carnitines (e.g., palmitoyl-carnitine). Then, acyl-carnitines can be translocated by CPT1 into mitochondria for FAO to produce the main source of energy. In DKD, the dysfunction of FAO has been documented in several studies, although the results differ, partly because of the different DKD stages at which these studies were observed (Li et al., 2013; Sas et al., 2016; Afshinnia et al., 2019). In general, FAO fluxes in the renal tissues were higher in the early stage of DKD, with significant reduction in advanced DKD (Hasegawa and Inagi, 2021). According to the present study, the concentration of two acyl-carnitines was decreased in the diabetic kidney medulla. It has been reported that palmitoyl-carnitine can serve as a marker of FAO rate, and its low level may be attributed to the impaired β-oxidation in the diabetic kidney due to the decreased CPT1 (Bouchouirab et al., 2018; Afshinnia et al., 2019). In addition, the decreased abundance of propionyl-carnitine, a product of mitochondrial branched-chain amino acid (BCAAs) catabolism also reported to be an antioxidant agent, may also be the consequence of impaired FAO (Adams et al., 2009; Scioli et al., 2014).

Considering the above findings, we surmised that CPT1 would be weaker in the diabetic kidney, since it was demonstrated that the lower expression of CPT1 was related to tubulointerstitial fibrosis from DKD patients (Li et al., 2017). The real-time PCR and IHC data confirmed our assumption. Collectively, the upregulated biosynthesis of several saturated fatty acids by agitating the activity of FAS plus the impaired FAO through the suppressive activity of CPT1 was observed in the dysregulation of fatty acids metabolism, which may provide new insights into the potential lipid-treatment of DKD.

Further research is needed to fully elucidate the underlying mechanism of these metabolic pathways reprogramming. However, alternations of the distribution and abundance of most metabolites highlight the association with mitochondria dysfunction in DKD. The accumulation of fatty acids may damage podocytes and tubulars, which can be associated with ROS formation in mitochondria (Forbes and Thorburn, 2018). On the other hand, insufficient antioxidant capacity is another contributor to mitochondrial dysfunction in DKD. As a fundamental antioxidant, mitochondrial glutathione (GSH) helps to decrease excessive ROS by interacting with the superoxide anions and then being oxidized to GSSG (Lushchak, 2012). However, in db/db mice, it is worth noting that our results indicated a remarkably higher abundance of GSSG in the cortex and medulla accompanied by downregulated GSH in the medullar (Figure 4A), which suggested the high levels of ROS that exceed local antioxidant capacity. Furthermore, our metabolome data also exhibited a clear upregulation of 8-hydroxy-deoxyguanosine (8-OHdG) in db/db mice (Figure 6C), a well-known marker for measuring the effect of oxidative damage to mitochondria DNA (Kakimoto et al., 2002). In general, metabolic pathway reprogramming accompanied by dysfunctional mitochondria could be detrimental to energy production and renal function in DKD, whereas targeted activation or inhibition of the dysregulated metabolic enzymes to regulate the expression of metabolites may have a potential role in metabolism-based therapy. This area warrants further research.