Metabolite Genome-Wide Association Study for Indoleamine 2,3-Dioxygenase Activity Associated with Chronic Kidney Disease

Chronic kidney disease (CKD) causes progressive damage to kidney function with increased inflammation. This process contributes to complex amino acid changes. Indoleamine 2,3-dioxygenase (IDO) has been proposed as a new biomarker of CKD in previous studies. In our research, we performed a metabolite genome-wide association study (mGWAS) to identify common and rare variants associated with IDO activity in a Korean population. In addition, single-nucleotide polymorphisms (SNPs) selected through mGWAS were further analyzed for associations with the estimated glomerular filtration rate (eGFR) and CKD. A total of seven rare variants achieved the genome-wide significance threshold (p < 1 × 10−8). Among them, four genes (TNFRSF19, LOC105377444, LOC101928535, and FSTL5) associated with IDO activity showed statistically significant associations with eGFR and CKD. Most of these rare variants appeared specifically in an Asian geographic region. Furthermore, 15 common variants associated with IDO activity were detected in this study and five novel genes (RSU1, PDGFD, SNX25, LOC107984031, and UBASH3B) associated with CKD and eGFR were identified. This study discovered several loci for IDO activity via mGWAS and provided insight into the underlying mechanisms of CKD through association analysis with CKD. To the best of our knowledge, this is the first study to suggest a genetic link between IDO activity and CKD through comparative and integrated analysis.


Introduction
Chronic kidney disease (CKD) is caused by several factors, including diabetes, high blood pressure, and glomerulonephritis [1]. CKD can lead to end-stage renal failure due to the gradual loss of kidney function and fibrosis caused by inflammation [2,3]. The Centers for Disease Control and Prevention (CDC) have reported that 90% of adults with CKD are unaware that they have CKD [4]. The early stages of CKD have few signs or symptoms. However, the progression of CKD increases cardiovascular morbidity and mortality, making treatment difficult [5]. Therefore, the management of patients with CKD should focus on delaying disease progression by identifying risk factors through early diagnosis [6].
Metabolomics is a field of interest in nephrology because many metabolites, which are small molecules, are freely filtered by the kidneys [7]. In addition, metabolites that play important roles in numerous biological pathways are also known as potential biomarkers of several diseases, including CKD [8][9][10]. Currently, the estimated glomerular filtration

Metabolite Measurements
To quantify kynurenine and tryptophan, serum samples collected from 2579 participants were analyzed using an AbsoluteIDQ p180 kit (BIOCRATES Life Science, Innsbruck, Austria) according to the manufacturer's instructions. Liquid chromatography/tandem mass spectrometry (LC-MS/MS) was conducted using an API 4000 QTRAP system (Applied Biosystems, Foster City, CA, USA) equipped with an Agilent 1200 HPLC system (Agilent Technologies, Santa Clara, CA, USA) to measure metabolites. The quality control (QC) process for analyzed metabolites has been described in detail elsewhere [29]. Briefly, both kynurenine and tryptophan used in this study met the following criteria: the coefficient of variance for each metabolite in the reference standards < 25%, 50% of the analyzed metabolite concentrations in the reference standards > limit of detection, and 50% of the analyzed metabolite concentrations in the experimental samples > limit of detection. Pooled human normal serums were used as reference standards. IDO activity was estimated as the ratio of kynurenine to tryptophan.

Genotyping and Imputation
Genotyping of the KARE dataset was performed using an Affymetrix Genome-Wide Human SNP array 5.0 (Affymetrix, Santa Clara, CA, USA). QC criteria of samples and variants have been previously described [30]. Briefly, samples with low call rates (<96%), DNA contamination, gender inconsistency, and serious concomitant illnesses were excluded. Exclusion criteria for variants were: Hardy-Weinberg equilibrium (HWE) p-values < 1 × 10 −6 , missing call rates > 5%, and minor allele frequency (MAF) < 0.01. After the QC process, imputation analysis of genetic variants was performed using an IMPUTE2 program with 1000 Genomes Phase I data as a reference panel [31]. A total of 6,461,358 SNPs were included in this study. Locations of variants were assigned using National Center for Biotechnology Information (NCBI) Human Genome Build 37 (hg19).

Statistical Analysis
All statistical analyses were conducted with PLINK version 1.90 β (https:// www.cog-genomics.org/plink2 (accessed on 26 July 2021)) [32]. Linear regression was used to assess associations of variants with IDO and eGFR. A case-control study was performed using logistic regression analysis. All regression analyses were based on an additive model and adjusted for age, gender, area, BMI, hemoglobin A1C (HbA1c), drinking, smoking, systolic blood pressure (SBP), and hs-CRP. The cutoff p-value was p < 5 × 10 −8 for rare variants and p < 1 × 10 −5 for common variants. Statistical significance between two groups was confirmed via Student's t-test. After performing mGWAS for IDO activity, linkage disequilibrium (LD) among variants was considered through clumping analysis. Variants were clumped through the following criteria: significance threshold of p < 0.05, LD threshold < 0.5, and physical distance threshold < 1000 kb. The variant with the lowest p-value among clumped variants was selected as the index variant. Manhattan plot and LD block were drawn using the Haploview version 4.1 program (Whitehead Institute for Biomedical Research, Cambridge, MA, USA). Geographical distribution maps for variants were generated based on the 1000 genome database via the Geography of Genetic Variants (GGV) browser (http://popgen.uchicago.edu/ggv/ (accessed on 20 August 2021)). LocusZoom browser (http://locuszoom.org/ (accessed on 16 September 2021)) was used to draw regional plots. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database (https://www.genome.jp/kegg/ (accessed on 2 September 2021)) was used to investigate biological processes involved in IDO activity and CKD.

Participant Characteristics
The baseline characteristics of the study population are shown in Table 1. A total of 2579 participants with metabolite and genotype data were included in this study. To investigate the genetic association of IDO activity and CKD, participants were divided into cases with CKD (n = 264) and controls (n = 1550). The mean age of all participants was 57.10 ± 9.05 years, with the case group being older (mean age: 65.72 ± 6.53 years) than the control group (mean age: 54.98 ± 8.64 years). In addition, values of kidney-related traits such as eGFR, creatinine, and BUN showed significant differences (p < 0.001) between cases and controls.

Associations between Common Variants and IDO Activity Related to CKD
This study performed mGWAS of IDO activity in the KARE dataset and summarized the association between SNPs and IDO activity through a Manhattan plot ( Figure 1).
As a result, 15 SNPs with MAF > 0.05 passed the significant threshold of 1 × 10 −5 ( Table 2). Of these, the strongest association with IDO activity was observed for rs59178336 in the RSU1 (Ras Suppressor Protein 1) gene located on chromosome 10. We further studied associations of variants related to IDO activity with eGFR and CKD. We found that rs59178336, a variant with the lowest p-value for IDO activity, showed a significant association with CKD (p < 0.05). Minor allele carriers of rs59178336 significantly increased both the IDO level (β = 0.26, p = 9.41 × 10 −8 ) and CKD risk (OR = 1.47, 95% CI: 1.02-2.14, p = 0.041). Although rs12226572 in the UBASH3B (Ubiquitin Associated and SH3 Domain Containing B) gene was significantly associated with CKD (p < 0.025), it did not exhibit a strong LD with surrounding SNPs at the 11q24.1 locus. Three SNPs (rs2513735, p = 0.012; rs78259836, p = 0.024; rs7237751, p = 0.043) in the PDGFD (Platelet-Derived Growth Factor D), SNX25 (Sorting Nexin 25), and LOC107984031 genes were also identified as significant variants for eGFR. Regional plots around the RSU1, PDGFD, SNX25, and LOC107984031 genes revealed several SNPs in the LD, with top SNPs involved in IDO activity ( Figure S1). On chromosome 11, rs2513735 near the PDGFD gene showed the most significant result for eGFR (β = −1.45, p = 0.012) among common variants. Additionally, the mitogenactivated protein kinase (MAPK) pathway known to regulate various cellular processes was identified through KEGG pathway analysis of the PDGFD gene ( Figure S2).

Geographical Distribution of Rare Variants
This study further analyzed the geographic distributions of rare variants associated with eGFR and CKD, as well as IDO activity using the GGV browser ( Figure 3). Rare variants (rs182145739, rs117150322, and rs146321869) were mostly seen in East Asia. For rs58332670, it was found in East Asia and America. In particular, rs117150322, located near the TNFRSF19 gene, was detected only in a Japanese population (MAF = 0.019). The MAF of rs117150322 in Korea was 0.009, which was lower than that in Japan.

Discussion
Several studies have performed metabolite profiling to evaluate physiological pathways for CKD [15,[33][34][35]. Interestingly, they have identified an association between CKD and the kynurenine pathway. According to another study, CKD is strongly associated with IDO activity, which degrades tryptophan to kynurenine in the kynurenine pathway [36]. Furthermore, IDO activity was positively correlated with CKD (OR = 12.65, 95% CI: 6.55-24.44) in a Korean population [21]. All of these studies were conducted during epidemiological investigations. Although CKD is a complex disease with high heritability, studies that perform a genetic analysis for the association between IDO activity and CKD have not been reported yet. Therefore, we performed mGWAS to identify genetic variants and potential loci affecting IDO activity associated with CKD in Koreans.
Our results revealed that 15 common variants had significant associations with IDO activity (p < 1 × 10 −5 ) ( Table 2) and that seven rare variants reached the GWAS threshold for IDO activity (Table 3). Additionally, SNPs related to IDO activity were analyzed for eGFR and CKD. Among genetic signals associated with eGFR and CKD, common variants were found at the RSU1, PDGFD, SNX25, LOC107984031, and UBASH3B genes and rare variants were identified at the LOC105377444, TNFRSF19, LOC101928535, FSTL5 genes. We focused on the RSU1, PDGFD, SNX25, and TNFRSF19 genes because evidence showing that other genes were associated with CKD was insufficient.
In the case of the RSU1, PDGFD, and TNFRSF19 genes, they were associated with the MAPK pathway [37][38][39]. The SNX25 gene was related to dopamine receptors. MAPK pathways include extracellular signal-regulated kinase (ERK), c-Jun N-terminal kinase (JNK), and p38 mitogen-activated protein kinase (p38MAPK) [40]. Although the MAPK pathway is generally known to regulate proliferation, many studies have suggested that it is an intracellular signaling pathway underlying kidney development [40][41][42]. Moreover, Fujigaki et al. reported that IDO activity is related to the MAPK pathways [43]. Dopamine receptors exist as D 1 -like (D 1 R and D 5 R) and D 2 -like (D 2 R, D 3 R, and D 4 R) subtypes according to their structure and pharmacology [44]. Among them, D 1 R is widely expressed in the kidney. It plays a central role in regulating blood pressure and sodium balance [45,46].
A recent study has reported that RSU1 is a critical mediator in downregulating ERK signaling through extracellular matrix (ECM) detachment [37]. The ERK pathway responsible for basic cellular processes is the most important signaling cascade of MAPK pathways [47]. It has also been reported that reduced ERK activity can improve antioxidant effects and kidney function [48]. In the present study, our results showed that, among common variants, rs59178336, located in the intron of the RSU1 gene, was significantly associated with IDO activity and CKD. Interestingly, Reznichenko et al. [49] have confirmed that the CUBN gene, located close to the RSU1 gene, is associated with end-stage renal disease. Therefore, the RSU1 gene might be associated with kidney disease.
Our results also identified associations of regions near the PDGFD gene involved in ERK signaling with IDO activity and eGFR (Table 2, Figure S2) [38]. Similar to our results, a previous study has reported that a genetic variant (rs7103465) in the PDGFD gene is associated with the ratio of urine albumin to creatinine (p = 3.7 × 10 −7 ) in Latin Americans [50]. Moreover, it has been reported that PDGFD is overexpressed in hepatic and renal fibrosis [51,52]. Charni et al. demonstrated that PDGFD is regulated by TGF-β, which activates MAPK pathways such as ERK, JNK, and p38MAPK [53].
Furthermore, it has been reported that SNX25 overexpression is associated with increased expression levels and signaling of D 1 R [54]. Another study has indicated that SNX5 depletion can result in hypertension in normotensive mice [55]. Therefore, the SNX25 gene encoding protein SNX25 might be associated with hypertension, a risk factor for kidney disease. In our data, minor carriers of rs78259836 belonging to the SNX25 gene increased IDO activity but decreased eGFR (Table 2).
Previous studies have reported that an understanding of progenitor cells involved in kidney damage and repair can provide insight into renal pathology and identify novel therapeutic targets [56,57]. Schutgens et al. [58] have demonstrated that TNFRSF19 is a marker gene for epithelial progenitor cells that contributes to adult kidney development in vivo. Moreover, previous studies have reported that overexpression of the TNFRSF19 gene can activate the JNK pathway and that its activation causes damage and fibrosis in the human kidney [39,59]. The JNK pathway involved in inflammation has also been reported as a mechanism regulating IDO [60,61]. Through mGWAS analysis, we discovered several SNPs located near the TNFRSF19 gene associated with IDO activity (Table 3, Figure 2). The results of our study suggest that the TNFRSF19 gene might regulate CKD by inducing IDO through JNK, which belongs to the MAPK family.
In summary, this study performed an mGWAS for IDO activity obtained from 2579 participants in the KARE cohort. A total of 22 novel SNPs (15 common and 7 rare variants) were found and further analyzed for genetic associations with eGFR and CKD. For variants selected based on metabolites, we investigated their associations with CKD compared to other study groups. As a result, four genes (RSU1, PDGFD, SNX25, and TNFRSF19) were associated with CKD by regulating IDO activity. In particular, our data highlight that the RSU1 and PDGFD genes are potential mediators of CKD associated with ERK, which belongs to the MAPK family. The results of this study also suggest that rare variants of the TNFRSF19 gene are associated with CKD, specifically in Asians, through the JNK pathway. The gene-metabolite associations identified in our study provide insight into the underlying mechanisms for CKD. However, functional analyses of mRNA expression levels and proteins should be performed to validate these findings. In addition, modern technological advances have made metabolite analysis possible, but the number of samples is still limited. Therefore, replication studies in other cohorts are needed to confirm the accuracy of this study.
Supplementary Materials: The following are available online athttps://www.mdpi.com/article/10 .3390/genes12121905/s1, Figure S1: Regional association plots of IDO activity at 10p13, 1p12, 4q35.1, and 18q21.1 loci. The statistical significances between the SNPs near RSU1 (A), PDGFD (B), SNX25 (C), and LOC107984031 (D) genes and IDO activity are plotted as −log 10 p values. The purple diamond represents the SNP most strongly associated with IDO activity. Levels of linkage disequilibrium (r 2 ) of top SNPs and surrounding SNPs are shown in different colors. The regional plots for SNPs were generated via LocusZoom browser (http://locuszoom.org/ (accessed on 16 September 2021)). Figure S2: The KEGG pathway for MAPK signaling pathway. A red box indicates the PDGFD gene found in growth factor (GF) according to KEGG. The image of a pathway was generated via KEGG browser (https://www.genome.jp/kegg/ (accessed on 2 September 2021)). Table S1: Rare variants associated with indoleamine 2,3-dioxygenase activity in Koreans.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to ethnical concerns.

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