LC-MS/MS-Based Serum Metabolomics and Transcriptome Analyses for the Mechanism of Augmented Renal Clearance

Augmented Renal Clearance (ARC) refers to the increased renal clearance of circulating solute in critically ill patients. In this study, the analytical research method of transcriptomics combined with metabolomics was used to study the pathogenesis of ARC at the transcriptional and metabolic levels. In transcriptomics, 534 samples from 5 datasets in the Gene Expression Omnibus database were analyzed and 834 differential genes associated with ARC were obtained. In metabolomics, we used Ultra-Performance Liquid Chromatography-Quadrupole Time-of-Flight Mass Spectrometry to determine the non-targeted metabolites of 102 samples after matching propensity scores, and obtained 45 differential metabolites associated with ARC. The results of the combined analysis showed that purine metabolism, arginine biosynthesis, and arachidonic acid metabolism were changed in patients with ARC. We speculate that the occurrence of ARC may be related to the alteration of renal blood perfusion by LTB4R, ARG1, ALOX5, arginine and prostaglandins E2 through inflammatory response, as well as the effects of CA4, PFKFB2, PFKFB3, PRKACB, NMDAR, glutamate and cAMP on renal capillary wall permeability.

There have been many studies on the diagnostic criteria, pharmacokinetics, and drug regimens of ARC. However, there are few reports on how ARC occurs. There are three hypotheses about the mechanisms of ARC: The theory of systemic inflammatory response syndrome, the theory of renal function reserve, and the theory of brain-kidney crosstalk [34]. The systemic inflammatory response syndrome theory suggests that inflammatory responses lead to a highly dynamic state of ARC, while targeted fluid loading therapy increases the peripheral circulatory blood flow, ultimately improving blood perfusion in the kidneys, leading to increased GFR [35][36][37]. The theory of renal function reserve suggests that nephrons are not fully dispatched under normal physiological conditions. When ARC occurs, its critical state mobilizes more nephrons, resulting in elevated GFR [38][39][40]. The theory of brain-kidney crosstalk is mainly used to explain the occurrence of ARC in brain-related diseases such as brain trauma and subarachnoid hemorrhaging. It is believed that diseases of the brain change the regulation of the central nervous system to the peripheral nerves and also alter the relevant receptors that regulate the glomerulus, resulting in the occurrence of ARC [41][42][43]. At present, these three theories lack evidence, so further research is needed.
In this study, we retrospectively collected clinical data and serum samples from ARC patients and non-ARC (NARC) patients, and analyzed the metabolites of the two groups using Ultra-Performance Liquid Chromatography Quadrupole Time-of-Flight Mass Spectrometry (UPLC-QTOF MS)-based non-targeted metabolomics methods to obtain the metabolic characteristics of the ARC patients. Transcriptomics data in the open-source Gene Expression Omnibus (GEO) database were used to screen the possible differential expression genes of ARC patients, and the metabolic pathways related to ARC were analyzed through a joint analysis of non-targeted metabolomics and the transcriptome in order to provide evidence and reasonable explanations for the pathogenesis of ARC.

Clinical Evaluation of ARC
To study the clinical features of patients with ARC, we selected relevant samples and then analyzed them. In addition, to eliminate the interference of age and body mass index (BMI) in the analysis results between the two groups, propensity score matching was performed between the ARC group and the NARC group. After the initial screening of samples, 157 samples that met the criteria were obtained, and their demographic and clinical features, and biochemical indicators are shown in Table 1. The propensity score matching tolerance was determined to be 0.16, and 102 samples were included after matching, with no significant differences in age and BMI indicators. The demographic and clinical features, and the biochemical parameters of these samples are shown in Table 2. The samples before and after matching showed significant differences in renal function indicators such as urea, creatinine, and cystatin C.

Quality Control of Untargeted Metabolic Profiling
To ensure the reliability of the metabolomic data, we methodologically validated quality controller (QC) samples and examined their stability, repeatability, and replicability during data acquisition. After viewing the distribution information of the original results, a total of six parent ions were selected and named with their "mass-to-charge ratio (m/z) @ retention time (RT)": 837.8305@0.851, 165.0791@3.634, 563.3502@5.235, 431.3025@7.675, 1485.992@10.744, and 328.1513@13.637. These six fragments were chosen because they are evenly distributed on the mass and time axes, and the peak shape is good, which can reflect the performance of the instrument at different masses and RTs. Table 3 shows the final verification results. The total ion chromatography/count (TIC) pattern expressions of the QC samples during the injection sequence were found to strongly overlap ( Figure 1). These results together indicated that the instrument system offered good reproducibility stability during metabolomics analysis, and that the metabolomics data were stable and reliable.

Screening of Differential Metabolites
Differential metabolite screening was performed using a combination of univariate analysis, multiplier analysis, and multivariate analysis. The results of unsupervised principal component analysis (PCA) clustering ( Figure 2A) showed that all samples were within a 95% confidence interval in positive ion mode. The metabolites in the ARC group and the NARC group showed a certain extent of separation tendency, and a difference in metabolic behavior was observed between the two groups. The separation trend shown in the score plot of the orthogonal partial least squares discriminant analysis (OPLS-DA) model ( Figure 2B) is more pronounced, which demonstrates that the OPLS-DA model is effective in amplifying the differences between the two groups. The performance parameters of the OPLS-DA model R 2 Y were 99.7%, and those of Q 2 were 96.0%, indicating a good fit and predictive power. The permutation test results showed that the model was

Screening of Differential Metabolites
Differential metabolite screening was performed using a combination of univariate analysis, multiplier analysis, and multivariate analysis. The results of unsupervised principal component analysis (PCA) clustering ( Figure 2A) showed that all samples were within a 95% confidence interval in positive ion mode. The metabolites in the ARC group and the NARC group showed a certain extent of separation tendency, and a difference in metabolic behavior was observed between the two groups. The separation trend shown in the score plot of the orthogonal partial least squares discriminant analysis (OPLS-DA) model ( Figure 2B) is more pronounced, which demonstrates that the OPLS-DA model is effective in amplifying the differences between the two groups. The performance parameters of the OPLS-DA model R 2 Y were 99.7%, and those of Q 2 were 96.0%, indicating a good fit and predictive power. The permutation test results showed that the model was not overfitted (Figure 3).

Screening of Differential Metabolites
Differential metabolite screening was performed using a combination of univariate analysis, multiplier analysis, and multivariate analysis. The results of unsupervised principal component analysis (PCA) clustering ( Figure 2A) showed that all samples were within a 95% confidence interval in positive ion mode. The metabolites in the ARC group and the NARC group showed a certain extent of separation tendency, and a difference in metabolic behavior was observed between the two groups. The separation trend shown in the score plot of the orthogonal partial least squares discriminant analysis (OPLS-DA) model ( Figure 2B) is more pronounced, which demonstrates that the OPLS-DA model is effective in amplifying the differences between the two groups. The performance parameters of the OPLS-DA model R 2 Y were 99.7%, and those of Q 2 were 96.0%, indicating a good fit and predictive power. The permutation test results showed that the model was not overfitted (Figure 3).   The screening results of the differential compounds are shown in Figure 4. The combined screening results of univariate analysis and fold change (FC) analysis were visualized using volcano plots ( Figure 4A), and 232 total candidate differential compounds were obtained. In total, 671 metabolites satisfied the screening conditions for multivariate analysis ( Figure 4B), and 105 significantly different compounds were obtained after intersection ( Figure 4C). The screening results of the differential compounds are shown in Figure 4. The combined screening results of univariate analysis and fold change (FC) analysis were visualized using volcano plots ( Figure 4A), and 232 total candidate differential compounds were obtained. In total, 671 metabolites satisfied the screening conditions for multivariate analysis ( Figure 4B), and 105 significantly different compounds were obtained after intersection ( Figure 4C).
The screening results of the differential compounds are shown in Figure 4. The combined screening results of univariate analysis and fold change (FC) analysis were visualized using volcano plots ( Figure 4A), and 232 total candidate differential compounds were obtained. In total, 671 metabolites satisfied the screening conditions for multivariate analysis ( Figure 4B), and 105 significantly different compounds were obtained after intersection ( Figure 4C).

Annotation Results of Differential Metabolites
In addition to 105 differential compounds screened using the above analysis methods, 45 differential compounds such as hypoxanthine, glutamic acid, and arginine were

Annotation Results of Differential Metabolites
In addition to 105 differential compounds screened using the above analysis methods, 45 differential compounds such as hypoxanthine, glutamic acid, and arginine were ultimately annotated. Their m/z, RT, molecular formula, HMDB ID, KEGG ID, UP/DOWN, and p-value are given in Table 4. "Up/Down" refers to the upregulation or downregulation of metabolites obtained via the FC analysis: "Up" means that the FC value of the metabolite is greater than 1.5 and that its content in the ARC group is upregulated relative to the NARC group; "Down" means that the FC value of the metabolite is less than −1.5, and that its content in ARC group is downregulated relative to the NARC group. "p-value" is the corrected p-value obtained via the moderated t-test, corrected with the Benjamini-Hochberg method. We then plotted the expression matrix to show the expression of each metabolite in each sample and its overall trend ( Figure 5).  ultimately annotated. Their m/z, RT, molecular formula, HMDB ID, KEGG ID, UP/DOWN, and p-value are given in Table 4. "Up/Down" refers to the upregulation or downregulation of metabolites obtained via the FC analysis: "Up" means that the FC value of the metabolite is greater than 1.5 and that its content in the ARC group is upregulated relative to the NARC group; "Down" means that the FC value of the metabolite is less than −1.5, and that its content in ARC group is downregulated relative to the NARC group. "p-value" is the corrected p-value obtained via the moderated t-test, corrected with the Benjamini-Hochberg method. We then plotted the expression matrix to show the expression of each metabolite in each sample and its overall trend ( Figure 5).

Figure 5.
Heat map of differential metabolite expression. Each horizontal row represents a differential metabolite; each column represents a sample; the color indicates the relative expression of the differential metabolites in the individual samples. The clustering methods for horizontal rows and vertical columns are complete linkage agglomerative clustering.  Figure 5. Heat map of differential metabolite expression. Each horizontal row represents a differential metabolite; each column represents a sample; the color indicates the relative expression of the differential metabolites in the individual samples. The clustering methods for horizontal rows and vertical columns are complete linkage agglomerative clustering.

Included Transcriptome Datasets and Their Basic Characteristics
In transcriptomics analysis, we screened the dataset based on risk factors for the occurrence of ARC. Ultimately, five GEO datasets, GSE11374, GSE37069, GSE57065, GSE19743, and GSE28750, were included in the study. The sample information in the dataset is detailed in Table 5; 398 total samples were included in the experimental group, and 136 samples were included in the healthy control group. Before differential gene screening, samples were normalized and standardized to eliminate systematic errors such as batch effects. The results of sample normalization and standardized pretreatment are shown in Figure 6. In the boxplot ( Figure 6A), the median and interquartile spacing levels of each sample were found to be consistent, indicating that normalization led to a consistent overall gene expression level across samples. The sample cluster plots before and after treatment ( Figure 6B,C) show that normalization eliminated systematic errors between batches.  (B) sample cluster plot before processing; (C) sample cluster plot after processing.

Screening of DEGs
After processing the samples, we first performed differential genetic screening on each dataset. The screening results for the differentially expressed genes (DEGs) are shown in Figure 7. The DEGs obtained from each dataset were quantified as follows: (1) GSE11357: 1022 upregulated genes and 1815 downregulated genes; (2) GSE19743: 3420 upregulated genes and 1466 downregulated genes; (3) GSE37069: 1831 upregulated genes and 1354 downregulated genes; (4) GSE57065: 1042 upregulated genes and 1172 downregulated genes; and (5) GSE28750: 1094 upregulated genes and 1104 downregulated genes. These differential genes are thought to be associated with risk factors for ARC. In order to further target the differential genes associated with ARC, we intersected the differential genes obtained from each dataset. After taking the intersection, we finally obtained 366 upregulated genes and 468 downregulated genes related to ARC (as shown in Figure 8).

Screening of DEGs
After processing the samples, we first performed differential genetic screening on each dataset. The screening results for the differentially expressed genes (DEGs) are shown in Figure 7. The DEGs obtained from each dataset were quantified as follows: (1) GSE11357: 1022 upregulated genes and 1815 downregulated genes; (2) GSE19743: 3420 upregulated genes and 1466 downregulated genes; (3) GSE37069: 1831 upregulated genes and 1354 downregulated genes; (4) GSE57065: 1042 upregulated genes and 1172 downregulated genes; and (5) GSE28750: 1094 upregulated genes and 1104 downregulated genes. These differential genes are thought to be associated with risk factors for ARC. In order to further target the differential genes associated with ARC, we intersected the differential genes obtained from each dataset. After taking the intersection, we finally obtained 366 upregulated genes and 468 downregulated genes related to ARC (as shown in Figure 8).

The Results of WGCNA
Analyzing 834 DEGs remains cumbersome, so we chose to use the weighted correlation network analysis (WGCNA) for further processing of the DEGs. In total, 14 outlier samples were removed according to the Z-score value, and 520 samples were retained for the construction of the subsequent weighted gene co-expression network. The network topology analysis plot (Figure 9) determined that the optimal soft threshold β should be 12, under which the comprehensive performance of the correlation coefficient and average number of connections were good. The average number of connections between genes was 50.

The Results of WGCNA
Analyzing 834 DEGs remains cumbersome, so we chose to use the weighted correlation network analysis (WGCNA) for further processing of the DEGs. In total, 14 outlier samples were removed according to the Z-score value, and 520 samples were retained for the construction of the subsequent weighted gene co-expression network. The network topology analysis plot (Figure 9) determined that the optimal soft threshold β should be 12, under which the comprehensive performance of the correlation coefficient and average number of connections were good. The average number of connections between genes was 50. The dynamic cut tree ultimately divided the 834 DEGs into 3 modules ( Figure 10); the blue module contained 257 DEGs, the turquoise module contained 468 DEGs, and the brown module contained 109 DEGs. The correlation heat map showed a strong correlation between the three modules, with a certain interaction in function ( Figure 11A). The correlation coefficients between the blue, green, and brown modules and ARC were 0.816, −0.767, and 0.775, respectively ( Figure 11B), with the best correlation being found between the blue module and ARC. The three module-ARC scatterplots ( Figure 12) depict the correlation between genes and ARC in different modules. Here, the genes in the blue module and the brown module have a good linear positive correlation with ARC, while those in the turquoise module have a good linear positive correlation with ARC. The correlation curve in the subplot has a large slope and is linear, indicating linear correlation with the occurrence of ARC. Gene scattering was evenly scattered near the module curve, indicating that module eigengene (ME) could better represent the expression patterns of genes in the module when analyzing the correlation between the module and ARC. The dynamic cut tree ultimately divided the 834 DEGs into 3 modules ( Figure 10); the blue module contained 257 DEGs, the turquoise module contained 468 DEGs, and the brown module contained 109 DEGs. The correlation heat map showed a strong correlation between the three modules, with a certain interaction in function ( Figure 11A). The correlation coefficients between the blue, green, and brown modules and ARC were 0.816, −0.767, and 0.775, respectively ( Figure 11B), with the best correlation being found between the blue module and ARC. The three module-ARC scatterplots (Figure 12) depict the correlation between genes and ARC in different modules. Here, the genes in the blue module and the brown module have a good linear positive correlation with ARC, while those in the turquoise module have a good linear positive correlation with ARC. The correlation curve in the subplot has a large slope and is linear, indicating linear correlation with the occurrence of ARC. Gene scattering was evenly scattered near the module curve, indicating that module eigengene (ME) could better represent the expression patterns of genes in the module when analyzing the correlation between the module and ARC.          Lastly, the first 20 genes were obtained in each module as the hub genes of the corresponding modules, and a visual network (see Figure 13) was drawn to represent the interactions between the hub genes. Lastly, the first 20 genes were obtained in each module as the hub genes of the corresponding modules, and a visual network (see Figure 13) was drawn to represent the interactions between the hub genes.

The Result of Network Analysis
We performed a metabolite-metabolite network analysis to annotate the interaction relationship between metabolites. Ultimately, a metabolic network with node number 41, connection number 145, and seed number 26 was obtained (as shown in Figure 14). The seed information in the network is shown in Table 6.  (Table 4), but is closely related to other differential metabolites. The size of the dot represents the magnitude of the FC.

The Result of Network Analysis
We performed a metabolite-metabolite network analysis to annotate the interaction relationship between metabolites. Ultimately, a metabolic network with node number 41, connection number 145, and seed number 26 was obtained (as shown in Figure 14). The seed information in the network is shown in Table 6. Lastly, the first 20 genes were obtained in each module as the hub genes of the corresponding modules, and a visual network (see Figure 13) was drawn to represent the interactions between the hub genes.

The Result of Network Analysis
We performed a metabolite-metabolite network analysis to annotate the interaction relationship between metabolites. Ultimately, a metabolic network with node number 41, connection number 145, and seed number 26 was obtained (as shown in Figure 14). The seed information in the network is shown in Table 6.  (Table 4), but is closely related to other differential metabolites. The size of the dot represents the magnitude of the FC.  (Table 4), but is closely related to other differential metabolites. The size of the dot represents the magnitude of the FC. The gene-metabolite network analysis yielded a gene metabolite network with a node number of 95, an edge number of 120, and a seed number of 95, as shown in Figure 15). In total, 19 genes were found to be involved in L-glutamate regulation:

Pathway Analysis
The results of the pathway analysis of differential metabolites using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database are shown in Figure 16A, including purine metabolism, arginine biosynthesis, sphingolipid metabolism, folate biosynthesis, and arachidonic acid metabolism. The pathway analysis of differential metabolites using

Pathway Analysis
The results of the pathway analysis of differential metabolites using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database are shown in Figure 16A, including purine metabolism, arginine biosynthesis, sphingolipid metabolism, folate biosynthesis, and arachidonic acid metabolism. The pathway analysis of differential metabolites using The Small Molecule Pathway Database (SMPDB) database is shown in Figure 16B, and the enrichment results feature a total of 26 metabolic pathways. The results of the joint-pathway analysis are shown in Table 7 and Figure 16C.

Pathway Analysis
The results of the pathway analysis of differential metabolites using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database are shown in Figure 16A, including purine metabolism, arginine biosynthesis, sphingolipid metabolism, folate biosynthesis, and arachidonic acid metabolism. The pathway analysis of differential metabolites using The Small Molecule Pathway Database (SMPDB) database is shown in Figure 16B, and the enrichment results feature a total of 26 metabolic pathways. The results of the joint-pathway analysis are shown in Table 7 and Figure 16C. Figure 16. Pathway analysis enrichment bubble diagram: (A) Only differential metabolites were used for the results of the pathway analysis in KEGG; (B) Only differential metabolites were used for the results of pathway analysis in the SMPDB; (C) Joint-pathway analysis results of differential genes and differential metabolites in combination. The size of the bubbles is determined by the pathway impact values from the pathway topology analysis, and the larger the value, the larger the Figure 16. Pathway analysis enrichment bubble diagram: (A) Only differential metabolites were used for the results of the pathway analysis in KEGG; (B) Only differential metabolites were used for the results of pathway analysis in the SMPDB; (C) Joint-pathway analysis results of differential genes and differential metabolites in combination. The size of the bubbles is determined by the pathway impact values from the pathway topology analysis, and the larger the value, the larger the bubble. The color of the bubbles is determined by the p-values from the pathway enrichment analysis. The larger the p -value, the darker the color (reddish), while the smaller the p-value, the lighter the color (yellow or even white). Table 7. Results for the joint-pathway analysis of differential metabolites and genes.

Mechanism Analysis
We plot the enrichment results of the ARC pathway, as well as significant changes in differential metabolites and DEG nodes in Figure 17, to demonstrate which differential metabolites and genes are included in the pathway and their interactions. We draw the overall metabolic spectrum changes of ARC in Figure 18 to demonstrate the relationships between various metabolic pathways. Finally, the pathogenesis of ARC was speculated and plotted in Figure 19. The detailed inference process can be found in the Discussion section. We plot the enrichment results of the ARC pathway, as well as significant changes in differential metabolites and DEG nodes in Figure 17, to demonstrate which differential metabolites and genes are included in the pathway and their interactions. We draw the overall metabolic spectrum changes of ARC in Figure 18 to demonstrate the relationships between various metabolic pathways. Finally, the pathogenesis of ARC was speculated and plotted in Figure 19. The detailed inference process can be found in the Discussion section.       Figure 19. ARC pathogenesis diagram.

Discussion
Genes exist upstream of metabolic regulation, and after the transcription stage, subsequent translation, post-translational modification, environmental factors, etc., will affect the final metabolic outcome. Therefore, in this study, we chose to start with the differential metabolites of ARC. First, we sought to determine the roles of differential metabolites in the occurrence of ARC, and then we explored the regulatory role of DEGs in ARC. It can be seen from the seeds with a large number of nodes in the metabolite-metabolite network analysis results (Figure 14 and Table 6) and the gene-metabolite network analysis results ( Figure 15) that L-glutamic acid, L-arginine, cAMP, and prostaglandin E2 were associated with the occurrence of ARC. In addition, the effects of upregulated CA4 on L-glutamate, upregulated LTB4R on L-arginine, downregulated PRKACB, upregulated PFKFB3, and PFKFB2 on cAMP, and upregulated LTB4R and ALOX5 on prostaglandin E2 may represent common changes in ARC patients. These differential metabolites and pivotal genes are of high value in studying the pathogenesis of ARC. We attributed the effects of these differential metabolites and differential genes on ARC to improving renal perfusion and improving glomerular capillary wall permeability, and summarized them in Figure 17.

Improving Renal Perfusion
The change of renal blood perfusion in the ARC state is mainly realized through the regulation of the inflammatory response by L-arginine, L-glutamic acid, and prostaglandin E2.
L-Arginine is a substrate for the production of nitric oxide (NO) via nitric oxide synthase (NOS). NO is a key signaling molecule in cardiovascular physiology and pathology, with negative chronotropic, negative, or positive inotropic effects, as well as positive gonadotropin effects [44,45]. NOS is a double-plus oxidase consisting of reductase and oxidase domains. NOS and O 2 oxidize guanidine nitrogen from L-arginine to produce NO and Lcitrulline [46][47][48]. This reaction requires the involvement of flavin mononucleotide (FMN), flavin adenine dinucleotide (FAD), and tetrahydrobiopterin. The depletion of L-arginine and/or tetrahydrobiopterin leads to the uncoupling of NOS, resulting in O 2 substituting L-arginine as an end electron acceptor and leading to the formation of superoxide [49]. The further binding of superoxide to NO leads to the production of peroxynitrite [50], which can lead to the development of congestive heart failure through cell damage and reduced myocardial contractility [51]. The production of peroxynitrite by NOS exposed to low L-arginine concentrations in renal cell lines has also been reported [52]. In addition, studies have shown that patients with sepsis exist in a state of impaired arginine, and mice with sepsis were observed to experience a decrease in arginine within 90 min of lipopolysaccharide (LPS) infusion [53]. Based on the above evidence, patients suffering from sepsis with impaired arginine may be at risk of renal hypoperfusion, resulting in a decreased GFR due to decreased myocardial contractility and decreased cardiac output, resulting in the obstruction of peripheral circulation.
In this study, the upregulation of arginase ARG1 in transcriptomic results was consistent with the increased arginine consumption observed in patients with sepsis. However, in the metabolomic results, L-arginine in the ARC group showed upregulation, suggesting the presence of other unknown pathways that increased the source of arginine to counter the increased arginine catabolism caused by ARG1 upregulation. Upregulated L-arginine provides sufficient conditions for the production of NO, which can reduce cell damage and improve cardiac output, ultimately increasing renal blood flow and the GFR. Arginine and tetrahydrobiopterin supplementation in rats with salt-induced elevated blood pressure was previously shown to improve renal hemodynamics with increased renal perfusion [54], consistent with the results of this study.
Glutamate was downregulated in the experimental group. It is naturally produced in the body and performs a variety of functions in the body, with circulating glutamate being filtered in the glomerulus and reabsorbed in the proximal tubule [55]. In the fasted state, the kidneys metabolize glutamine absorbed from the blood to produce ammonia and glutamate [56]. The glutamate-regulated N-methyl-D-aspartate receptor (NMDAR) is a glutamate-gated nonselective cation channel that is highly expressed in the kidneys, in addition to being widely expressed in the central nervous system [57,58]. Multiple reports have suggested that NMDAR plays an important role in maintaining renal homeostasis [58,59], and that its overexpression or inhibition may cause renal homeostasis imbalance and lead to the occurrence of renal disease. Studies have shown that the regulation of NMDAR agonists can change the GFR and the GFR of mononephron units [60,61], so it is speculated that the change of glutamate on NMDAR regulation may play an important role in ARC genesis.
The interaction between L-glutamate and L-arginine is of particular interest. This interaction indirectly improves renal blood perfusion through L-arginine. In addition, NO has a positive gonadotropin effect, and the enrichment results for the metabolic pathway in the SMPDB database also showed changes in androgen and estrogen metabolism, which may be the reason for why ARC exhibited more of this epidemiological feature in male patients.
Prostaglandin E2 is produced by all kidney cells and is the most abundant prostaglandin that plays an important physiological and pathological role in the kidneys. For example, this prostaglandin is involved in regulating the reabsorption of osmotic water through vasopressin, regulating water metabolism in the body [62], and improving renal function through a variety of antioxidant, anti-apoptotic, and inflammatory inhibitory effects. In the inflammatory response, prostaglandins, lipid signaling molecules involved in pain and inflammation, enhance tissue regeneration processes after injury in different organ systems, and regulate hematopoietic stem cell/progenitor cell homeostasis to regulate hematopoietic function [63]. LTB4R, ALOX5, and prostaglandin E2 are involved in the arachidonic acid metabolism pathway. The potent chemoattractant and proinflammatory mediator LTB4 is synthesized from arachidonic acid by ALOX5 [64].Multiple risk factors for ARC such as trauma [64], sepsis [65], cancer [66], etc., also found differences in LTB4R expression. Additionally, studies have found that ALOX5 can improve kidney function by weakening the inflammatory response [67], which is also consistent with the upregulation of ALOX5 in the ARC population in this study. In this study, prostaglandin E2 was found to be downregulated, which reduced the inflammatory inhibition effects on ARC patients and also regulated tubular water and sodium transport, glomerular filtration, and vascular resistance through receptors, ultimately leading to the occurrence of ARC.

Improving Glomerular Capillary Wall Permeability
The glomerular capillary wall is a dynamic system that forms the glomerular filtration barrier. The permeability of this wall depends on three layers: endothelial cells, basement membranes, and podocytes [68]. Podocytes are terminally differentiated and highly specialized cells covering the outer surfaces of glomerular capillaries, and the dynamic structures formed by their foot processes and fissure septa determine the final size of the glomerular filtration barrier, which is essential for maintaining the integrity of the glomerular filtration barrier.
Purinergic signaling is involved in a variety of physiological processes in the renal system, including the regulation of glomerular filtration. cAMP is a universal second messenger that has multiple effects on kidney function, in addition to regulating vasoactivity and regulating downstream protein targets (including ion channels, enzymes, and transcription factors) through cAMP-dependent protein kinase A (PKA). The findings suggest that the cAMP-PKA signaling pathway in podocytes may regulate actin cytoskeletal organization for glomerular protection [69], as well as the GFR [69]. At the same time, cAMP levels were found to be reduced in fibrotic kidney tissue, and cAMP supplementation improved tubular atrophy and extracellular matrix deposition [70].
In this study, the metabolomics results showed that cAMP in the purine metabolic pathway exhibited upregulation in the ARC group, with upregulated cAMP enhancing the size of the glomerular filtration barrier and reducing extracellular matrix precipitation, which ultimately manifested as an increase in the GFR. The upregulation of cAMP may be the result of a combination of upregulating the pivot genes PRKFB2 and PRKFB3, and downregulating PRKACB, among which PRKFB2 and PRKFB3 are involved in the AMPK signaling pathway.
AMPK is an important cellular energy sensor that is activated in response to intracellular ATP depletion and plays a role in restoring energy homeostasis by activating ATP production and inhibiting ATP consumption pathways [69]. In addition, carbohydrate metabolism is an important source of energy production. In our transcriptomics analyses, we found changes to the insulin signaling pathways in ARC. Moreover, previous studies have shown that glucagon increases the concentration of cAMP in the blood [71], suggesting that the regulatory effect of cAMP on the GFR may be related to carbohydrate homeostasis and energy stability. Glucagon increases the concentration of cAMP in the blood, affecting proximal tubular reabsorption. The authors in [71] showed that low-dose glucagon increases GFR only when combined with an infusion of cAMP (mimicking glucagon-induced hepatic release). The above evidence suggests that patients with critical illnesses may exist in an imbalanced state of energy metabolism and carbohydrate metabolism. The resulting changes in the insulin signaling pathway and AMPK signaling pathway work together to upregulate cAMP, and upregulated cAMP causes an increase in the GFR in the ARC state by changing the permeability of the glomerular capillary wall.

Limitations of This Study
The results for the ARC and NARC plasma metabolomesin in this study indicated that there are indeed differences in metabolic patterns between the ARC and NARC populations, providing meaningful information for exploring the pathogenesis of ARC. However, this study also has the following limitations to be considered. (1) The serum metabolomics samples in this study were collected retrospectively, while transcriptomics uses a public database. Thus, the correlation between the two was insufficient. Therefore, a prospective study on the collection of serum samples and transcriptome samples should be carried out to obtain more comprehensive validation. (2) The method used in this study was semiquantitative non-targeted metabolomics analysis. In addition, our metabolite annotation was mainly based on consistency with the m/z and MS/MS fragments in the database. To validate the above results, a targeted approach with authentic standards should be used in a future validation study. (3) During the metabolomics analysis of data, the results of OPLS-DA and PCA showed a split within the same phenotypic group. We only excluded the possibility of categorical variables such as gender, department, and clinical outcome, but we have not yet found the specific reasons for this phenomenon. (4) This study found the potential mechanism of ARC, but no in-depth study was carried out. Future studies should be performed in the cell line or by using animal models.

Information Collection and Sample Screening
The present study used a retrospective collection of vancomycin blood concentration monitoring samples from patients suspected of or diagnosed with Gram-positive bacterial infections in the First Affiliated Hospital of China Medical University, from 1 January 2017 to 31 October 2020. According to the creatinine clearance rate (CrCL) and vancomycin blood trough concentration results (C trough ), the inclusion criteria of the study were established as follows: (1) experimental group (ARC group): CrCL > 130 mL/min and Ctrough < 10 mg/L, and (2) control group (NARC group): 80 mL/min ≤ CrCL ≤ 130 mL/min and 10 mg/L ≤ Ctrough ≤ 20 mg/L. Exclusion criteria were as follows: (1) age < 18 years or age > 80 years; (2) pregnant or lactating women; (3) missing samples; (4) non-valley concentration of blood sample; (5) serum sample collection time < 48 h after first administration. We completed the preliminary screening of samples according to the above criteria. CrCL was calculated using the Cockcroft-Gault formula [72]. SPSS v22.0 software was used to match the propensity scores of the primary screening samples, with age and BMIas the predictive variables. Finally, two groups of samples were obtained for subsequent statistical and metabolomics analyses.

Sample Preparation
Serum samples stored in an ultra-low temperature freezer at −80 • C were taken and dissolved in a 4 • C freezer. Next, 30 µL of each serum sample was placed in a 1.5 mL centrifuge tube, and 150 µL of pre-refrigerated methanol was added. The solution was then vortexed for 2 min to achieve the purpose of mixing homogeneous and precipitating proteins. The mixture was centrifuged at 13,000 rpm (centrifugal force = 15,115× g) for 10 min at 4 • C, and then the clear upper liquid was aspirated for detection.

Quality Control
In total, 20 µL of each sample was mixed to prepare (QC) samples. The reliability of the metabolomic data was confirmed primarily using these samples. Before the formal analysis of the sample, the method was verified by calculating the RSD indicators of the corresponding peak area and RT of the reference ions. (1) System repeatability: The same QC sample was injected five times in a row to evaluate the suitability of the instrument system. (2) Method replicability: Five QC samples were processed in parallel to evaluate the applicability of the sample treatment method. (3) Sample stability: The dissolved serum sample was placed at room temperature to investigate the stability of the sample after 4 h, 8 h, and 12 h. Deviations throughout the injection process were evaluated by interspersing one QC sample every 10 samples. The RSD of the peak area of the corresponding sample was less than 5.0%, and the RSD of its RT was less than 1.5, indicating good data quality. After screening out differential metabolites, we further annotated these metabolites by collecting data from the QC samples in targeted MS/MS mode. In this mode, the daughter ion fragments were obtained through the targeted collision of specific m/z parent ions in a certain time range. The collision energy was optimized as 10 v, 20 v, and 40 v for each target. The fragmentation information of the metabolites obtained above at different collision energies was used to annotate the metabolites.

Data Processing
The entire data analysis process is shown in Figure 20. We used a Mass Hunter Profinder (B.08.00, Agilent, Technologies, Inc., Santa Clara, CA, USA) and Mass Profiler Professional (MPP, v14.9.1, Agilent, Technologies, Inc., Santa Clara, CA, USA) to preprocess the raw data. The specific steps are: (1) Molecular feature extraction: Molecular information is extracted from the original data features, and the molecular characteristics are initially integrated through the processes of noise filtering, classification and alignment, and post-processing filtering; (2) Recursive processing of compound mass spectrometry information: Based on the ion information in the original data, the ion peak of the compound is attributed, peak integration and filtering are performed, the extraction spectrum is extracted, and post-processing and filtering are carried out again, based on the accurate mass and mass spectrometry characteristic data. After the above steps, each sample is given a list of compounds with molecular weight, RT, m/z, and abundance, and is displayed in ".CEF" format output. Then, the experimental group and the control group samples of the corresponding ".CEF" file are imported into MPP software for normalization and standardization, and a "metabolite-abundance" data matrix is obtained for differential compound screening. normalization and standardization, and a "metabolite-abundance" data matrix is obtained for differential compound screening.
The univariate analysis method was a moderated t-test corrected with the Benjamini-Hochberg method. Multivariate analysis used SIMCA (v17.0.2, Biotree Biomedical Technology Co., Shanghai, China). Unsupervised PCA was used to describe within-group similarities and between-group differences. OPLS-DA was used as a supervised principal component analysis method to zoom in on differences and to further screen for differential metabolites. The final screened differential metabolites met all three conditions at the same time: (1)    The univariate analysis method was a moderated t-test corrected with the Benjamini-Hochberg method. Multivariate analysis used SIMCA (v17.0.2, Biotree Biomedical Technology Co., Shanghai, China). Unsupervised PCA was used to describe within-group similarities and between-group differences. OPLS-DA was used as a supervised principal component analysis method to zoom in on differences and to further screen for differential metabolites. The final screened differential metabolites met all three conditions at the same time: (1) a False Discovery Rate (FDR) in the univariate analysis of <0.05; (2) FC analysis results of |log 2 (FC)| ≥ 0.58; and (3) a corresponding variable importance in the projection (VIP) of >1.5 in the OPLS-DA model.

Metabolite Annotation
According to the MS information for the metabolites (monoisotopic peaks and isotope peaks of different adducts) and MSMS information (characteristic fragment peaks after fragmentation of the parent ions), an annotation of the differential metabolites was completed through comparison with the Human Metabolome Database (HMDB) (http://www.hmdb. ca/), KEGG (http://www.kegg.com/), ChemSpider (http://www.chemspider.com/), Metlin (http://metlin.scripps.edu/), and MassBank (https://massbank.eu/). We visited these websites between 1 November 2021 and 15 January 2022. Successfully annotated metabolites satisfy the condition that the matching tolerance of the parent ion is less than 50 ppm, while the m/z and relative response of more than 2 fragment ions are consistent with the database.

Dataset Filtering and Preparation
In the GEO database (https://www.ncbi.nlm.nih.gov/geo/, accessed between 4 July 2021 and 23 August 2021) of the National Biotechnology Center of the United States, 'Burn', 'Sepsis', and 'Trauma' were used as search terms, and the species was limited to 'Homo sapiens'. The inclusion criteria were as follows: (1) the sample type was a human serum sample; (2) the sample groups included a disease group and a healthy control group; (3) the study included the determination of the messenger ribonucleic acid (mRNA) expression profile; and (4) the disease duration was ≤120 h. Exclusion criteria were as follows: (1) age > 80 years and age < 18 years. (2) pregnant or lactating women.
We downloaded the raw gene expression data as MINIML files from the GEO database. The probe treatment scheme was as follows: (1) remove probes without the corresponding genes; (2) when multiple probes corresponded to the same gene, the average value of all corresponding probes was selected. The data preprocessing process was as follows: (1) for sample normalization, Log2 transform was used; (2) sample standardization was used for the R v3.4.1 platform (the normalize.quantiles function in the preprocessCore package from http://www.r-project.org/, accessed on 8 September 2021); (3) we removed the batch effect via the remove BatchEffect function in the limma package of the R software; (4) for probe annotation, according to the platform annotation file, the probe ID was converted into the gene name. After the above probe processing and data preprocessing, the corresponding gene expression matrix was obtained.

DEGs Screening and Annotation
DEGs were screened using the limma package in R software, including the construction of a standard data matrix, the creation of a comparative model, linear fitting, difference calculation, and statistical testing. The final screening was based on the significance of the statistical test and the FC. The methods and parameters were as follows. (1) Univariate analysis: Student's t-test was used to calculate the p-value of the gene expression difference between the disease group and the healthy control group in each dataset, and the p-value of the t-test was corrected using the Benjamini-Hochberg method. The screening condition was FDR < 0.05. (2) For differential expression fold screening, the threshold used was |log 2 (FC)| ≥ 0.58. After obtaining the DEGs of each dataset through the above threshold, intersection processing was performed, and the obtained intersecting DEGs were identified as genes that are related to ARC.

WGCNA Analysis
The WGCNA of the DEGs obtained in Section 2.2.2. was performed using the WGCNA package in the R software. The specific steps included the following steps. (1) Data preparation, which involved preparing the DEG expression profiles and clinical feature information table for the specimens. (2) Eliminating outliers, which used −2.5 as the Z-score cutoff value to eliminate outliers. (3) Construction of the gene correlation matrix: The correlation between genes was calculated using the Bicor function. (4) Converting to the adjacency matrix and constructing the weighted gene co-expression network: The key parameter in this process was determined as the soft threshold β with the network topology analysis diagram, and the maximum traversal value of the soft threshold was 30. (5) Dynamic tree cutting to generate the co-expression modules: We set the tree branch depth to 2, the minimum module gene size to 25, and merged related modules. (6) Calculating the correlation between module-module and module-clinical features: Calculation of the module was completed with ME using Pearson correlation analysis. Through the above steps, we identified the association between the module and the occurrence of ARC, and selected the first 20 genes with the largest weight values in the relevant module as the hub genes for ARC occurrence.

Gene Function Enrichment Analysis
ID conversion used the org.Hs.eg.db package (v3.10.0) and enrichment analysis used the clusterProfiler package (v3.14.3) in the R software [73]. The Gene Ontology (GO) and the KEGG databases were selected to enrich the hub genes. (1) GO was used to enrich cellular component (CC), molecular function (MF), and biological process (BP) hub genes.
(3) The KEGG PATHWAY library and the KEGG ORTHOLOGY library were used to enrich the metabolic pathways.

Method of Pathway Analysis
The action network was drawn to understand the relationship between metabolitesmetabolites and gene-metabolites in the DEGs of transcriptomics and differential compounds of metabolomics. Using the network analysis function of the online analysis tool Metaboanalyst 5.0 (https://www.metaboanalyst.ca/, accessed between 20 January 2022 and 27 January 2022), the input node information and corresponding change multiples were mapped to an interactive network, and the network algorithm was based on the desparse graphical lasso modeling process [74,75]. The association between nodes was extracted from STITCH, and the final output contained a subnet of at least three nodes, with only the shortest path being retained between the various sub-networks. The network results included (1) the gene-metabolite interaction network describing the interactions between genes and metabolism, and (2) the metabolite-metabolite interaction network describing the interactions between metabolites.
Pathway changes for DEGs and differential metabolites were analyzed using the jointpathway analysis function in the online analysis tool Metaboanalyst 5.0. Analysis. The visualization method was a scatter diagram, the enrichment method was a hypergeometric test, the topological measurement method was degree centrality, the integration method was combinatorial query, and the selected pathway database was the metabolic pathway (synthesis), which excluded regulatory pathways containing only genes.
The obtained information was integrated to analyze the mechanisms of ARC. The main nodes of the changes in each pathway were extracted, and the changed metabolic pathways were drawn in the KEGG global metabolic network (KO01100) to holistically visualize the role of each pathway in the occurrence of ARC and to draw a mechanism map of ARC occurrences.

Conclusions
(1) ARC patients exhibit the upregulation of arginine under the combined effects of LTB4R and ARG1. Although the upregulation of ARG1 indicates an increase in arginine consumption in critically ill patients, LTB4R may have a more positive improvement effect, ultimately improving renal blood flow perfusion and leading to the occurrence of ARC through NO-related inflammatory reactions. (2) Glutamic acid not only directly regulated the GFR through NMDAR, but also indirectly improved renal blood perfusion through the glutamine pathway and arginine synthesis pathway, resulting in the occurrence of ARC.
(3) Under the joint regulation of LTB4R and ALOX5, downregulated prostaglandin E2 led to ARC by indirectly regulating inflammatory pathways and directly adjusting tubular water and sodium transport, glomerular filtration, and vascular resistance. (4) PRKFB2, PRKFB3, and PRKACB affected cAMP through the AMPK signaling pathway and insulin signaling pathway, and changed the permeability of the glomerular capillary wall and extra stromal precipitation, resulting in ARC.  Informed Consent Statement: Patient consent was waived due to the retrospective nature of the study.

Data Availability Statement:
The data of the transcriptomics presented in this study are openly available in Gene Expression Omnibus (GEO); the GEO accession numbers are GSE11357, GSE19743, GSE28750, and GSE37069. The data of the metabolomics presented in this study are available upon request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.