Cross-sectional analyses of metabolites across biological samples mediating dietary acid load and chronic kidney disease

Summary Chronic kidney disease (CKD) is a major public health burden, with dietary acid load (DAL) and gut microbiota playing crucial roles. As DAL can affect the host metabolome, potentially via the gut microbiota, we cross-sectionally investigated the interplay between DAL, host metabolome, gut microbiota, and early-stage CKD (TwinsUK, n = 1,453). DAL was positively associated with CKD stage G1-G2 (Beta (95% confidence interval) = 0.34 (0.007; 0.7), p = 0.046). After adjusting for covariates and multiple testing, we identified 15 serum, 14 urine, 8 stool, and 7 saliva metabolites, primarily lipids and amino acids, associated with both DAL and CKD progression. Of these, 8 serum, 2 urine, and one stool metabolites were found to mediate the DAL-CKD association. Furthermore, the stool metabolite 5-methylhexanoate (i7:0) correlated with 26 gut microbial species. Our findings emphasize the gut microbiota’s therapeutic potential in countering DAL’s impact on CKD through the host metabolome. Interventional and longitudinal studies are needed to establish causality.


Highlights
Multi-biological sample metabolites associate with dietary acid load and kidney function

Multiple associated metabolites mediated between DAL and mild kidney function decline
The stool metabolite, 5-methylhexanoate(i7:0), correlated with several gut microbial species

INTRODUCTION
Chronic kidney disease (CKD) is a major cause of morbidity and mortality, with incident rates increasing globally. 1Multiple modifiable risk factors for CKD development have been identified, including diet and gut microbiota composition and function. 2,3Recently, there has been an evidence-based shift toward manipulating dietary acid load (DAL), when managing CKD progression. 4DAL is defined as the difference (in mEq H + /day) between endogenously produced acid and base, originating from diet. 4 Modern Western diets generally contain large amounts of acid-forming protein, processed foods, and limited amounts of base-producing fruits and vegetables, thereby inducing a state of chronic metabolic acidosis. 5An almost linear relationship exists between acidosis and poorer clinical outcomes in individuals with CKD, including CKD progression and mortality. 6][9] Indeed, a systematic review and meta-analysis of 31 observational studies reported that a higher DAL is associated with increased systolic and diastolic blood pressure. 10Increased DAL calculated via potential renal acid load (PRAL) methods was also associated with impaired fasting glucose and increased levels of HbA1c in this study.
However, the underlying metabolic pathways resulting in this dysregulation remain unclear.Metabolomics, a high-throughput technology that provides a snapshot of an individual's metabolic profile at a particular time point, can be used to provide insights into these pathways and has been used successfully to identify novel markers of CKD risk. 11,12Recently, Tariq and colleagues (2022) identified circulating levels of the amino acid N-methylproline to be inversely associated with both DAL and CKD (stage G3-G4), suggesting a protective effect of diet. 13evertheless, the effect of DAL on the host metabolome of individuals with early-stage CKD remains largely unexplored.Moreover, the role of the gut microbiome in this relationship is unknown.It is paramount to identify molecular pathways and gut microbiota signatures regulating the DAL-CKD association, as this could lead to effective strategies for mitigating kidney function decline.

OPEN ACCESS
Therefore, the aims of this large population-based study were to (i) explore the relationship between DAL and early-stage CKD (stage G1-G2), (ii) investigate the biological pathways underlying this association using metabolomics profiling from four biological samples (serum, urine, stool, and saliva) with mediation analyses to estimate effects, and (iii) assess the interaction between DAL and gut microbiota composition in kidney function decline.

RESULTS
The descriptive characteristics of the study population are presented in Table 1.This study included 1,453 individuals from TwinsUK with metabolomics profiling available in serum (fasting), urine (spot), stool, and saliva, as well as estimated glomerular filtration rate (eGFR) and dietary data available.Briefly, the predominantly female (89.7%) sample included 560 individuals with CKD stage G1 and 893 CKD stage G2, and on average they were 61.3 (G12) years of age with an average body mass index (BMI) of 25.9 (G4.7) kg/m 2 .A flowchart of the study design is presented in Figure 1.

DAL and renal function exhibit a coordinated signature on the serum, urine, and stool metabolome
To explore the serum, urine, stool, and saliva metabolite profiles of individuals in CKD stage G1 and G2 in relation to their DAL, we performed principal-component analysis and Permutational multivariate analysis of variance (PERMANOVA).We found that the metabolome profile of all biological samples, except for saliva, clustered differently according to their DAL on a global scale (p < 0.05; Figure S1).
We then performed random forest (RF) models on the residuals-adjusted metabolites and identified 41 metabolites in serum that overlap between CKD and DAL, of which 15 passed FDR correction, 41 in urine, of which 14 passed FDR correction, 27 in stool, of which 8 passed FDR correction, and 22 in saliva, of which 7 passed FDR correction (Figure 2; Table S6).

Effect of DAL on renal function is mediated through serum, urine, and stool metabolites
We next tested if the overlapping metabolites associated with both DAL and early-stage CKD potentially mediated the association between DAL and CKD stage G1-G2 progression through mediation analysis, correcting for confounding factors such as age, sex, and BMI.We tested both the DAL/metabolite of interest/CKD stage G1-G2 as well as the DAL/CKD stage G1-G2/ metabolite of interest order (Figures 3  and S2).
When considering CKD stage G1-G2 as a potential mediator in the relationship between DAL/ metabolite of interest, we found that 7/8 serum metabolites, 1 urine metabolite, and 1 stool metabolite were mediated by CKD stage (Figure S2).The VAF was smaller in this model compared to the model using metabolites of interest as potential mediators (Figure 3).The VAF had a range from 8.38% (tiglylcarnitine (C5:1ÀDC)) to 18.88% (N-acetylalanine) in serum metabolites, 5.94% in urine alpha-ketoglutorate, and 4.14% in stool 5-methylhexanoate (i7:0).CKD stage G1-G2 did not affect saliva metabolites.A step-by-step report of the mediation models used is reported in Tables S2-S5.
The major super pathways involved with metabolites that mediated the association between DAL/CKD stage G1-G2 were mainly amino acid (n = 6) and lipid metabolism (n = 2) (Table S7) across all biological samples.Spearman's correlation showed that the metabolites strongly correlate with each other, even across biological fluids (Figure 3D).

DAL and the gut microbiota
Next, we explored if the metabolites mediating the DAL-CKD relation were associated with gut microbiota composition (alpha diversity, represented as Shannon index) and microbial species (Figure 4).We found that the gut microbiota could predict 13% of the variance in stool abundances of the metabolite, 5-methylhexanoate (i7:0), and we detected 26 bacterial species that were significantly associated with it.Out of these 3 were belonging to different Clostridia bacterium species genomic bins (SGBs).
Furthermore, we found no association between the mediating serum and urine metabolites and gut microbiota composition.

DISCUSSION
In this large cross-sectional study integrating serum, urine, stool, and saliva metabolites, we report, for the first time, that the effect of earlystage kidney function declines (CKD stage G1-G2) due to DAL being possibly mediated by serum, urine, and stool, but not by saliva metabolites.The metabolites mediating the association are predominantly related to amino acid (n = 6) and lipid (n = 2) metabolism and strongly correlate with each other, even across biological sample types.Moreover, RF machine learning models identified that 13% of the variance of the stool metabolite, 5-methylhexanoate (i7:0) that potentially mediates the DAL-CKD association, could be explained via the gut microbiota composition.However, it is important to note that longitudinal, as well as interventional studies, are warranted to further confirm the potential mediating effects of the found metabolites.This study also revealed multiple microbial species linked with the mediating stool metabolite, thereby providing potential for novel treatment options.
Our finding that DAL contributes to CKD progression supports previous reports and meta-analysis. 14,15The underlying pathophysiology of this relationship is complex and poorly understood, with the metabolome likely to play an important role. 11o date, only one study has investigated the role of the serum metabolome in the DAL-CKD relationship. 13In this study performed by Tariq et al. (2022), the authors identified 12 metabolites to be inversely associated with DAL, of which only one (N-methylproline) was also inversely associated with incident CKD.Our study identified, through machine learning models, 3 serum metabolites to be inversely associated with both DAL and CKD, and 8 positively associated.Our study confirms that indolepropionate and tartronate (hydroxymalonate) are negatively associated with DAL, as previously reported, 16 highlighting the robustness of our result.However, we also found these metabolites to be negatively associated with CKD.The contrast in our findings likely originates from the fact that Tariq et al. (2022) studied these associations using a population with overt CKD (stage 2-4) that also used medications known to affect the serum metabolome, such as antihypertensive drugs, 17 as well as different metabolomic platforms to those used in our study.
Moreover, our study identified 8 serum, 2 urine, and 1 stool metabolite to potentially mediate the positive association between DAL and CKD stage G1-G2.The majority of the mediating metabolites identified play a role in amino acid metabolism, which is to be expected as DAL is largely driven by protein intake. 4,18Interestingly, one metabolite, 3-MH, was identified as a mediating metabolite in both serum and urine.3-MH is a product of histidine metabolism and produced after actin and myosin degradation. 193-MH has been proposed as a marker for protein turnover but is influenced by meat intake. 203-MH, therefore, has gained attention as a metabolite that can potentially serve as a positive biomarker for meat intake, especially poultry. 21,22However, the effects of 3-MH on kidney function remain inconclusive.One study using rat models showed that serum 3-MH is a candidate biomarker for acute renal injury, as its levels increased when drug-induced nephrotoxicity was induced. 23Another human observational study showed higher levels of 3-MH in subjects with moderate kidney failure compared to healthy individuals. 24However, another study in individuals on hemodialysis identified low levels of 3-MH to be predictive of cardiovascular events. 25his study also identified the stool metabolite 5-methylhexanoate (i7:0) to potentially mediate the effects of DAL on CKD and identified several gut microbiota species that correlate with this metabolite.5-methylhexanoate belongs to the group of medium-chain fatty acids (MCFAs) but is not well investigated as an individual metabolite.However, the effect of MCFA in general on human health has been extensively reviewed elsewhere, 26 and was generally found to be positive for human health, but deleterious effects have also been reported. 27 itself did not associate with gut microbiota composition, which aligns with previous studies reporting that protein intake, a major source of DAL, does not affect gut microbiota diversity. 28,29s kidney function can also affect host metabolome, 11,30 we also performed mediation analyses by using CKD stage as potential mediator, i.e., DAL/ CKD stage G1-G2 (potential mediator)/ metabolite of interest.Though we found that early-stage CKD partially mediates the DAL/ metabolome relationship in selected metabolites, the VAF was lower when using metabolites as the mediator.This could be explained by the fact that we only used early-stage CKD and the association may be stronger as kidney function declines toward end-stage CKD. 30 Multiple studies have reported an intricate relationship between the gut microbiota and kidney disease, [31][32][33][34] with the gut microbiota correlating with end-stage CKD 34 In our data, we did not find an association between gut microbiota composition and CKD; however, our study participants were early-stage CKD (stage G1-G2). 34We did however find 26 microbial species to be associated with fecal abundances of the stool metabolite 5-methylhexanoate (i7:0) that mediated the association between DAL and CKD, thus linking the gut microbiome's involvement in CKD.
Of the 26 microbial species, Alistipes spp.had the strongest positive association and has been consistently reported to be increased in CKD, and to be associated with an increased intake of a Western diet, high in animal protein and fat but low in fiber. 2,35Our study also identified Intestimonas butyriciproducens, a known producer of the beneficial metabolite butyrate, a short-chain fatty acid (SCFA), 36 to be negatively associated with 5-methylhexanoate (i7:0).This finding further emphasizes a role for SCFA in line with a previous study reporting that supplementation with the probiotic L. casei Zhang increased SCFA levels and reduced CKD progression in mice. 31n conclusion, we identified serum, stool, and urinary metabolites to possibly mediate the positive relationship between DAL and CKD stage G1-G2.Moreover, this study identified several gut microbiota species, such as Alistipes spp and Intestimonas butyriciproducens, that can be used as novel targets to mitigate this relationship, suggesting that the gut microbiota may be a therapeutic option to combat the effects of DAL on kidney function decline by altering host metabolome.Longitudinal, as well as interventional, studies are warranted to investigate the exact relationship of the identified metabolites and microbial species and to elucidate potential causality in the relationship between DAL and kidney function decline.

Limitations of the study
This study has limitations and strengths.Strengths of this study include the use of a large population-based cohort with accurate phenotyping, dietary information, and concurrent metabolomics in multiple biological fluids.Another strength of our study is that we explored the host metabolome, using the Metabolon, Inc, platform, that covers a wide range of metabolites from multiple biological pathways leading to a more detailed understanding of the underlying biology.
However, our findings should also be appreciated in the context of some limitations.First, DAL was determined using self-reported dietary intake.Although a previously validated formula was used to calculate DAL, 4,37 this approach is prone to several biases, including misreporting bias, which may cause misclassification bias.However, there is no gold standard to measure DAL and each method has its own drawbacks.Second, our study sample was limited to individuals with normal-to-mild renal impairment and was predominantly females of Caucasian ancestry.Accordingly, our results cannot be generalized to those with moderate-to-end-stage renal disease, and future research should explore the relationships identified here in those with later-stage CKD.
Third, the mediation analysis has some limitations.Although our study identified multiple metabolites that potentially mediate the relationship between DAL and CKD progression, our data do not permit comments on causality due to the lack of a temporal relationship.Moreover, our study cannot rule out a different directionality between exposure, mediator, and outcomes.Hence, we tested both directions.These analyses indicate a potential mediating effect of CKD stage on the metabolite of interest and, to a larger extent, of the metabolite of interest on CKD.
Finally, we were unable to replicate our results in an independent sample as, to our knowledge, there are currently no cohorts with such a comprehensive assessment of serum, urine, stool, and saliva metabolomics as well as clinical creatinine, dietary data, and shotgun metagenome profiling of the gut microbiome.
Longitudinal metabolomics data, as well as dietary and kidney function, are needed to confirm causality and understand directionality.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Figure 1 .
Figure 1.Flow chart of the study design

Figure 2 .
Figure 2. Metabolites associated with CKD and acid intake Metabolites identified through random forest machine learning that pass multivariable regression analysis (p < 0.05, FDR corrected [Benjamini & Hochberg]) and are associated with CKD (red) and acid intake (blue) in (A) serum, (B) urine, (C) stool, and (D) saliva.All analyses were corrected for age, sex, and BMI.

Figure 3 .
Figure 3. Mediation analysis between DAL and CKD stage G1-G2 Mediation analysis of the association between DAL and CKD stage G1-G2, using metabolite of interest as potential mediator in (A) serum,(B) urine, and (C) stool.Path coefficients are illustrated beside each path and variance accounted for (VAF) score is denoted below the mediator.All associations are statistically significant (p < 0.05).(D) Spearman correlation between mediating metabolites across biological samples (*p < 0.05; **p < 0.01 and ***p < 0.001).All analyses are corrected for age, sex, and BMI.Abbreviations: DAL, dietary acid load; CKD, chronic kidney disease; VAF, variance accounted for.

Table 1 .
Demographic characteristics of the study population overall and by CKD status

TABLE
d RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability d EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS B Study population and ethics approval d METHOD DETAILS B Study design B Metabolite profiling B Metabolite measurement and standardization B Quality control B Dietary intake B Gut microbiota fecal collection, DNA extraction and metagenome profiling d QUANTIFICATION AND STATISTICAL ANALYSIS