Abstract

   Variations in serum amino acid levels are linked to a multitude of
   complex disorders. We report the largest genome-wide association study
   (GWAS) on nine serum amino acids in the UK Biobank participants
   (117 944, European descent). We identified 34 genomic loci for
   circulatory levels of alanine, 48 loci for glutamine, 44 loci for
   glycine, 16 loci for histidine, 11 loci for isoleucine, 19 loci for
   leucine, 9 loci for phenylalanine, 32 loci for tyrosine and 20 loci for
   valine. Our gene-based analysis mapped 46–293 genes associated with
   serum amino acids, including MIP, GLS2, SLC gene family, GCKR, LMO1,
   CPS1 and COBLL1.The gene–property analysis across 30 tissues
   highlighted enriched expression of the identified genes in liver
   tissues for all studied amino acids, except for isoleucine and valine,
   in muscle tissues for serum alanine and glycine, in adrenal gland
   tissues for serum isoleucine and leucine, and in pancreatic tissues for
   serum phenylalanine. Mendelian randomization (MR) phenome-wide
   association study analysis and subsequent two-sample MR analysis
   provided evidence that every standard deviation increase in valine is
   associated with 35% higher risk of type 2 diabetes and elevated levels
   of serum alanine and branched-chain amino acids with higher levels of
   total cholesterol, triglyceride and low-density lipoprotein, and lower
   levels of high-density lipoprotein. In contrast to reports by
   observational studies, MR analysis did not support a causal association
   between studied amino acids and coronary artery disease, Alzheimer’s
   disease, breast cancer or prostate cancer. In conclusion, we explored
   the genetic architecture of serum amino acids and provided evidence
   supporting a causal role of amino acids in cardiometabolic health.

Introduction

   Circulating levels of amino acids are associated with the development
   of a variety of diseases, including type 2 diabetes (T2D) ([32]1–4),
   cardiovascular disease ([33]5–7), inflammatory disease ([34]8,[35]9),
   non-alcoholic fatty liver disease ([36]10) and Alzheimer’s disease (AD)
   ([37]11,[38]12), and further with several cardiometabolic risk factors,
   including insulin resistance, blood pressure and lipid profile
   ([39]4,[40]7,[41]13). In addition, serum amino acid levels are
   suggested to be a useful marker for disease status in patients with
   head and neck cancer ([42]14) as well as for plasma-free amino acid
   (PFAA) profile, which serves as one of the biological markers for
   cancer patients ([43]15,[44]16). Nonetheless, most of these observed
   associations originated from observational studies, and can be
   influenced by biases and confounding factors. Furthermore,
   observational studies cannot ascertain causality or determine the
   direction of the effect. Therefore, identification of the genetic
   variants associated with the circulating serum amino acid
   concentrations can be utilized to investigate their impact on chronic
   conditions. The genetic markers are inherited randomly during
   conception and thus could be used as a proxy to the exposures in
   Mendelian randomization (MR) analysis to prevent the biases afflicted
   in traditional epidemiological studies ([45]17). Genome-wide
   association studies (GWAS) unravel the genetic determinants of serum
   amino acids, which, on one hand, could elucidate the role of serum
   amino acid levels in the pathophysiology of complex disorders and, on
   the other hand, could be used to conduct MR studies.

   Here, we present large-scale GWAS on nine serum levels of amino acids
   [alanine, isoleucine, leucine and valine (branched-chain amino
   acids, BCAAs), glycine, glutamine, histidine, phenylalanine and
   tyrosine] using data from the UK Biobank (UKB) study participants to
   identify genetic determinants of serum amino acid levels and a
   post-GWAS analysis to explore the related biological pathways.
   Moreover, we utilize our GWAS findings and rich phenotypic data from
   the UKB study to elucidate the potential relationship between serum
   amino acid concentrations and a broad range of diseases (n = 617
   clinical diagnoses with >200 cases) through the MR phenome-wide
   association study (MR-PheWAS) approach and assess the causality.
   Moreover, we used the largest studies available that use genome-wide
   association study (GWAS) to look at common complex disorders. These
   included coronary artery disease (CAD), AD, breast cancer and prostate
   cancer. We did this to increase the statistical power when looking for
   associations with these outcomes. This was necessary because the
   prevalence of these conditions is limited in the UKB.

Materials and methods

   Our research complies with all relevant ethical regulations, as the UKB
   has ethics approval from the North-West Multi-Centre Research Ethics
   Committee (11/NW/0382). We utilized data from the UKB under the
   approved application ID 52569. Additional ethical approval was not
   required for the present study.

Study design and study sample

   We used data from the UKB study, a large study including over 500 000
   participants aged 40–69 years at recruitment from 2006 to 2010. The UKB
   has used a purpose-designed genotyping array to assay genome-wide
   genetic data. This array comprised 847 441 genetic polymorphisms. The
   genotyped data are imputed to 1000 Genomes reference panels ([46]18).
   High-throughput nuclear magnetic resonance (NMR) metabolic measurements
   were done using the Nightingale metabolomics platform for a random
   subset of 117 994 participants at baseline assessment, including
   measurements of nine amino acids: alanine, glycine, glutamine,
   histidine, isoleucine, leucine, phenylalanine, tyrosine and valine.

GWAS on serum amino acid levels in the UKB population

   We performed a linear mixed model regression using BOLT-LMM version 2.3
   to run GWAS on serum amino acid levels of individuals from European
   ancestry adjusting for age, sex and the first 10 genetic principal
   components. The mixed model addresses population stratification and
   relatedness and therefore minimizes the sample exclusions ([47]19). All
   serum amino acid levels of individuals from European ancestry of the
   UKB data were standardized and normalized using rank-based inverse
   normal transformation. We filtered variants with minor allele
   frequency <0.01 and Hardy–Weinberg equilibrium P-value <0.0001.

Replication of previously reported SNPs in our UKB GWAS analysis

   We sought replication for our lead single-nucleotide polymorphisms
   (SNPs) in each genomic locus using data from a previously published
   study by Kettunen et al. ([48]20). The study combined data from four
   large-scale cohort studies, totalling over 24 000 individuals. They
   employed NMR spectroscopy as the metabolomics platform to measure
   circulating metabolites. We did a lookup of our lead SNPs in this
   dataset. The locus was considered replicated if the P-value was
   significant after Bonferroni correction and a concordant effect
   direction.

Functional annotation analysis in the UKB population

   We applied a range of functional annotation analyses using the FUMA web
   server ([49]http://fuma.ctglab.nl) ([50]21). All significant SNPs
   (P < 5e−08) and SNPs in linkage disequilibrium (LD) with them
   (r^2 ≥ 0.6) from the UKB GWAS summary statistics in the previous step
   were annotated for functional consequences on gene functions, based on
   Ensemble genes (build 85) using ANNOVAR ([51]21,[52]22). Independent
   significant SNPs and correlated SNPs were linked to the GWAS catalogue
   to postulate known associations of the SNPs in the risk loci that were
   reported previously with several phenotypes ([53]21). Gene mapping was
   conducted following the acquisition of annotated SNPs. This mapping
   process employed positional mapping, which relies on the physical
   distance between the SNPs and the genes ([54]19). For genome-wide
   gene-based association analysis, FUMA implements MAGMA gene-based
   analysis ([55]21,[56]23) by mapping the input SNPs to protein-coding
   genes attained from Ensemble build 92. Same as for gene-based analysis,
   FUMA implements MAGMA ([57]21) for gene-set enrichment analysis, which
   is conducted for curated gene sets and gene ontology terms attained
   from Molecular Signatures Database v5.2. Gene-set enrichment analysis
   assesses whether genes within the gene set are more strongly associated
   with any of the phenotypes than other genes.

   To test the association between tissue-specific gene expression and
   serum amino acids, MAGMA gene–property analysis was conducted using
   FUMA ([58]21). This analysis is based on the computed gene-based
   P-values in the gene-based analysis.

One-sample MR-PheWAS in the UKB population

   To explore the effects of nine serum amino acids on a wide range of
   diseases, we conducted PheWAS analyses. We extracted the phenotypic
   data of the UKB participants, including patient hospital records,
   cancer registry data and death registry data defined as ICD-9/10 codes,
   which were linked to Hospital Episode Statistics, as well as the
   genotypic data (48–138 amino acid-associated SNPs identified in our
   GWAS analysis) of the UKB participants (n = 117 944). We selected SNPs
   with P-value <1 × 10^−5 identified in our UKB GWAS with an F-statistic
   >10 (calculated by regressing serum levels per each SNP) as genetic
   instruments. Multiallelic and palindromic SNPs were excluded, and the
   remaining SNPs were clumped with r^2 < 0.001. We conducted a one-sample
   MR analysis approach using the two-stage least-squares method by
   regressing each serum amino acid on their genetic instruments using a
   linear regression model and saved the residual values. Subsequently, we
   used a logistic regression model to regress the saved residual values
   on each phecode comprising a wide range of clinical diagnoses in the
   UKB study (n = 617 clinical diagnoses with >200 cases). To avoid
   overlap with the sample that was used for GWAS, the PheWAS analysis was
   merely done in the ∼385 000 individuals who were not used for GWAS. In
   addition, the PheWAS analysis was restricted to participants from
   European ancestry and a kinship coefficient of >0.088 was used to
   exclude the related individuals. Our PheWAS analysis was adjusted for
   age, sex and the first 10 genetic principal components. We used the
   false discovery rate (FDR) method to account for multiple testing as
   the Bonferroni procedure is suggested to be too conservative due to the
   correlation between clinical diagnoses.

Two-sample MR

   To further validate the associations identified through our MR-PheWAS
   analysis, we employed a two-sample MR analysis on five traits,
   including diabetes, hypertension, disorders of lipid metabolism,
   hypercholesterolaemia and hyperlipidaemia. We used the same genetic
   instruments as we selected for MR-PheWAS; however, for their
   association with outcomes, we used data from the largest GWAS datasets
   available for each trait and excluded UKB data. We applied the
   two-sample MR analysis using the inverse-variance weighted (IVW)
   method. The estimates were calculated per one standard deviation (SD)
   increase in each serum amino acid level. To further investigate the
   robustness of the findings to possible pleiotropy, we used weighted
   median MR ([59]24) and MR-Egger regression ([60]25) as sensitivity
   analysis. The weighted median MR method assumes that at least half of
   the instruments are valid instruments. In MR-Egger regression, the
   intercept indicates the presence of directional pleiotropy, and the
   slope of that regression will be a consistent estimate of the causal
   effect of exposure on the outcome of interest even in the presence of
   invalid instruments ([61]26). Additionally, the MR-PRESSO method was
   applied to detect and remove outlying genetic variants that
   substantially affect the IVW estimates ([62]27).

   We further conducted a two-sample MR analysis for traits of a number of
   complex disorders, including CAD ([63]28), late-onset AD ([64]29),
   breast cancer ([65]30) and prostate cancer ([66]31), using genetic
   estimate associations reported in the largest GWAS summary statistics
   in independent populations ([67]Supplementary Table S1).

Statistical software

   All analyses were conducted using R version 4.0.2. Phenome-wide
   associations were conducted using the R package ‘PheWAS’, and
   two-sample MR and multivariable MR were performed using ‘TwoSampleMR’
   and ‘MRPRESSO’ R packages.

Results

   The study design is illustrated in Figure [68]1. We performed a GWAS of
   nine serum amino acids (alanine, glutamine, glycine, histidine,
   isoleucine, leucine, phenylalanine, tyrosine and valine) in ∼117 000
   population cohort of the UKB study, measured by the Nightingale
   platform as previously described. Of the 9 933 468 genetic variants
   included in the analysis, 3225 variants exhibited genome-wide
   significant associations (P < 5 × 10^−8) with serum alanine, 5619
   variants with serum glutamine, 9965 variants with serum glycine, 2008
   variants with serum histidine, 2046 variants with serum isoleucine,
   1212 variants with serum leucine, 934 variants with serum
   phenylalanine, 5250 variants with serum tyrosine and 1625 variants with
   serum valine ([69]Supplementary Tables S2–[70]S11).

Figure 1.

   [71]Figure 1.
   [72]Open in a new tab

   Schematic overview of the study.

Functional annotation and pathway enrichment

   We applied a range of functional annotation analyses to leverage the
   amino acid GWAS results using FUMA–MAGMA (see Figure [73]2 for
   Manhattan plot of GWAS summary statistics). Our FUMA ANNOVAR findings
   showed that most of the significant SNPs and SNPs in LD with the
   significant SNPs lie within intronic region followed by intergenic gene
   regions for all serum amino acids ([74]Supplementary Figures
   S1–[75]S9).

Figure 2.

   [76]Figure 2.
   [77]Open in a new tab

   Manhattan plot of GWAS summary statistics for nine serum amino acids.

   The findings were mapped to 34 genomic loci for circulatory levels of
   alanine, 48 loci for glutamine, 44 loci for glycine, 16 loci for
   histidine, 11 loci for isoleucine, 19 loci for leucine, 9 loci for
   phenylalanine, 32 loci for tyrosine and 20 loci for valine, by using
   the FUMA platform ([78]Supplementary Table S2). We identified 20–132
   significant lead SNPs (P < 5e−8) for each serum amino acid, including
   58 for alanine, 131 for glutamine, 132 for glycine, 49 for histidine,
   20 for isoleucine, 39 for leucine, 39 for phenylalanine, 109 for
   tyrosine and 60 for valine ([79]Supplementary Table S2).

   Using gene-based analysis implemented in MAGMA, we were able to
   identify 176 genes for alanine, 293 genes for glutamine, 212 genes
   for glycine, 95 genes for histidine, 85 genes for isoleucine, 134 genes
   for leucine, 46 genes for phenylalanine, 152 genes for tyrosine and 145
   genes for valine ([80]Supplementary Table S2). In the gene-based
   analysis, a significant number of genes were found to be shared among
   all nine serum amino acids. Notably, some of these shared genes include
   MIP, GLS2, SLC gene family, GCKR, LMO1, CPS1 and COBLL1
   ([81]Supplementary Table S12).

   Our gene-set MAGMA analysis revealed that prioritized gene sets are
   involved in the regulation of metabolites (small molecules and protein
   measures) and PFAA levels (adjusted for 20 other PFAAs), except for
   serum aromatic amino acids (phenylalanine and tyrosine). Moreover, the
   pathway enrichment analysis demonstrated the involvement of serum amino
   acids in the regulation of fasting blood glucose levels, blood lipid
   profile including triglycerides, cholesterol, high-density lipoprotein
   (HDL) and low-density lipoprotein (LDL), blood pressure (both systolic
   and diastolic), as well as mean arterial pressure, coronary heart
   disease, C-reactive protein levels and many more ([82]Supplementary
   Figures S10–[83]S18).

   The gene–property analysis across 30 tissues highlighted the expression
   of most of serum amino acids in liver tissues, except for serum
   isoleucine and valine, in muscle tissues for serum alanine and glycine,
   in adrenal gland tissues for serum isoleucine and leucine, and in
   pancreatic tissues for serum phenylalanine (Figure [84]3). This
   analysis did not highlight any further significant associations for the
   other amino acid levels, including histidine and valine. The total
   proportion of variance explained (R²) for serum amino acids, based on
   selected instruments, ranged from 1% for isoleucine with a combined
   F-statistic of 24.35 for 60 SNPs to 6% for serum glycine with a
   combined F-statistic of 53.58 for 120 SNPs. The proportion of variance
   explained for serum BCAAs (isoleucine, leucine and valine) ranged from
   1% to 3% with an F-statistic of 24–34 ([85]Supplementary Table S13).

Figure 3.

   [86]Figure 3.
   [87]Open in a new tab

   Gene–property analysis across 30 tissues for serum amino acids.

Replication of significant identified loci

   We looked up our lead SNPs for associations with serum amino acids in a
   previously published independent study (N = 24 925) ([88]20). We
   replicated (P < 0.01) 4 variants for leucine, 3 variants for
   phenylalanine, 6 variants for isoleucine and valine, 9 variants for
   histidine and glycine, 14 variants for tyrosine and 17 variants for
   alanine (see the ‘Materials and methods’ section and [89]Supplementary
   Tables S14–[90]S22 for full replication results).

One-sample MR-PheWAS in the UKB population

   We performed PheWAS utilizing genetic variants for all serum amino
   acids. After clumping (r^2 < 0.001) and excluding the multiallelic
   SNPs, we had 101 independent (r² <0.001) genome-wide significant SNPs
   (P < 5e−08) for serum alanine, 120 for glutamine, 138 for glycine, 62
   for histidine, 60 for isoleucine, 67 for leucine, 48 for tyrosine, 110
   for phenylalanine and 91 for valine ([91]Supplementary Tables
   S3–[92]S11). Out of nine serum amino acids, we found clinical diagnoses
   associated with only four of them, which included alanine and three
   BCAAs (leucine, isoleucine and valine). Figure [93]4 shows the
   Manhattan plot for these analyses. Serum alanine, leucine and valine
   were significantly associated with the risk of diabetes [odds
   ratio (OR) = 1.80 per SD and P= 1.61e−11 for serum alanine; OR = 1.58
   and P= 0.003 for serum leucine; OR = 1.62 and P= 0.02 for isoleucine]
   ([94]Supplementary Table S23).

Figure 4.

   [95]Figure 4.
   [96]Open in a new tab

   MR-PheWAS Manhattan plot for serum alanine and BCAAs.

   Serum isoleucine and valine were associated with disorders of lipid
   metabolism (OR = 1.53 per SD and P= 0.001 for serum isoleucine;
   OR = 1.38 per SD and P= 3.60e−05 for serum valine), hyperlipidaemia
   (OR = 1.53 per SD and P= 0.001 for serum isoleucine; OR = 1.38 per SD
   and P= 3.60e−05 for serum valine) and hypercholesterolaemia (OR = 1.54
   per SD and P= 0.001 for serum isoleucine; OR = 1.40 per SD
   and P= 3.60e−05 for serum valine). Serum valine was also significantly
   associated with hypertension (OR = 1.28 per SD and P= 3.60e−05) (Figure
   [97]3 and [98]Supplementary Table S23).

   No significant association was observed for serum histidine, glycine,
   glutamine, tyrosine and phenylalanine across phenome.

Validation of PheWAS results using two-sample MR in the UKB population

   We conducted a two-sample MR analysis on findings from PheWAS using
   data from the UKB. Figure [99]5 depicts significant associations
   between genetically predicted serum levels of alanine [OR for
   IVW = 1.43; 95% confidence interval (CI) = 1.18–1.74] and leucine (OR
   for IVW = 1.22; 95% CI = 1.02–1.46) and an increased risk of diabetes.
   Moreover, we found a significant association of serum valine levels
   with T2D and an increased risk of diabetes (OR for IVW = 1.35; 95%
   CI = 1.10–1.64) and hypertension (OR for IVW = 1.25; 95%
   CI = 1.15–1.35), with consistency across both IVW and weighted median
   estimates. After removing the outliers by the MR-PRESSO method, results
   were also consistent with the IVW estimate ([100]Supplementary Table
   S24).

Figure 5.

   [101]Figure 5.
   [102]Open in a new tab

   Two-sample MR analysis on studied amino acids and clinical diagnoses
   using data from the UKB.

   We did not find a significant association between genetically elevated
   levels of any serum amino acids and the risk of CAD (Figure [103]6 and
   [104]Supplementary Table S25). Among all amino acids, genetically
   elevated levels of serum isoleucine were associated with the risk of
   AD; however, the association did not remain significant after
   correcting for multiple testing using FDR (Figure [105]6 and
   [106]Supplementary Table S26). No association was observed between
   genetically elevated levels of serum amino acids and risk of breast
   cancer (Figure [107]6 and [108]Supplementary Table S27) and prostate
   cancer (Figure [109]6 and [110]Supplementary Table S28).

Figure 6.

   [111]Figure 6.
   [112]Open in a new tab

   MR results for nine serum amino acids and complex disorders.

Discussion

   We have conducted the most extensive GWAS of nine serum amino acid
   levels, involving ∼117 000 participants from the UKB study. Our
   research significantly enhances the current understanding of the
   genetic architecture of serum amino acid levels. Furthermore, we
   provided a systematic evaluation of potential causal associations
   between serum amino acids and a range of diseases and clinical traits.
   In our study, we uncovered a spectrum of genomic risk loci, varying in
   number from 9 for serum phenylalanine to 48 for serum glutamine. We
   have created a diverse array of genetic variants that can be used as
   instrumental variables to explore the underlying biological mechanisms
   and to facilitate further research into potential links between serum
   amino acids and the risk of complex disorders. These genetic variants
   accounted for different proportions of variance, ranging from 1–3% for
   serum BCAAs to 6% for serum glycine and glutamine, respectively.

   Our investigation unveiled that genetically predicted higher levels of
   serum valine were associated with an increased risk of T2D.
   Furthermore, higher levels of genetically predicted serum alanine and
   BCAAs were linked to elevated cholesterol, triglyceride and LDL levels,
   while being associated with lower levels of HDL.

   We report a potential causal association between BCAAs and
   dyslipidaemia, which is supported by evidence linking higher BCAA
   levels with dyslipidaemia and higher triglyceride levels
   ([113]5,[114]6). However, the connection between serum BCAAs and LDL
   cholesterol levels remains unclear, requiring further exploration to
   disentangle the complexities linking BCAAs, lipid metabolism and
   cardiovascular disease risk. We further report a potential causal
   association between serum BCAA levels and diabetes, as supported by
   previous observational ([115]32,[116]33) and MR studies
   ([117]2,[118]3). Unlike prior studies, our analysis leverages data from
   over 117 000 individuals, offering a comprehensive examination of the
   genetic determinants of serum BCAA levels and their correlation with
   diabetes.

   Our findings are in agreement with observational studies associating
   higher BCAA levels with cardiometabolic risk factors and CAD
   ([119]7,[120]34,[121]35), but also expand on these insights by
   suggesting a possible increase in cardiovascular disease risk with
   raised isoleucine levels supported by observational evidence ([122]36).
   Our results are in good agreement with a nested case-control study on
   BCAAs and risk of breast cancer ([123]37), and may explain why studies
   on dietary BCAA intake and risk of breast cancer are inconclusive
   ([124]38,[125]39). Despite these findings, no significant associations
   were found between increased serum amino acids and the risk of CAD,
   AD or cancers such as breast cancer or prostate cancer. This could be
   due to the limited number of cases in the UKB, potentially leading to
   insufficient statistical power for detecting such associations and
   highlighting the need for larger studies to further explore these
   associations.

   The association of elevated serum alanine levels with T2D and
   dyslipidaemia identified in the study is an interesting finding that
   fits very well with alanine’s role in both glucogenic and proteogenic
   pathways, especially during conditions of physiological stress such as
   fasting ([126]40). The breakdown of muscle proteins releases alanine
   into the bloodstream, which is then utilized by the liver for
   gluconeogenesis, essential for maintaining blood glucose levels during
   fasting ([127]40). The gluconeogenic pathway, deeply entangled with
   lipid metabolism, could potentially influence serum lipid profiles,
   thus linking alanine with dyslipidaemia. Additionally, alanine might
   affect the activity of AMP-activated protein kinase, which is key for
   cellular energy homeostasis, and in turn could enhance fatty acid
   oxidation and suppress lipid synthesis ([128]41). The resulting
   increase in gluconeogenic flux may disrupt the energy balance within
   hepatocytes, leading to upregulation of lipogenesis and elevation of
   triglyceride levels. The altered hepatic conversion of alanine to
   pyruvate and subsequent changes in acetyl-CoA levels could further
   exacerbate these dyslipidaemic profiles ([129]42).

   We also found a potential causal link between blood levels of BCAAs and
   lipid profiles, proposing that BCAAs, such as leucine, could activate
   the mTOR signalling pathway in a manner distinct from insulin,
   suggesting that BCAAs may directly affect lipid metabolism and
   contribute to dyslipidaemia ([130]43,[131]44). Moreover, MAGMA
   gene-based analysis further identified genes such as GCKR, a highly
   pleiotropic gene associated with glucose metabolism and fasting plasma
   glucose levels, and the SLC superfamily, which is involved in the
   transmembrane transport of essential compounds, including amino acids.
   These genes are implicated in a variety of conditions, such as
   dyslipidaemia and neurodegenerative disorders, establishing a genetic
   link to the observed metabolic alterations ([132]45–50). The
   gene–property analysis also contributed to understanding the expression
   patterns of serum amino acids, indicating the significant expression of
   alanine in muscle tissues and its potential as a biological marker for
   cardiometabolic health ([133]51–53).

   Our MAGMA gene-set analysis further indicated that serum amino acids
   are integral in pathways regulating serum metabolites, PFAA profiles
   and fasting glucose levels. These pathways are closely linked with
   cardiometabolic risk factors such as blood pressure, lipid levels and
   diabetes. PFAA profiles have been associated with the future risk of
   diabetes and cardiovascular disease in both non-diabetic and diabetic
   populations, as well as with Crohn’s disease, underlining their
   importance in disease prognosis and the potential for targeted
   therapeutic interventions ([134]54–56).

   BCAA levels prior to the onset of T2D are altered, suggesting a role in
   disease progression. Epidemiological data indicate that changes in BCAA
   metabolism are associated with cardiometabolic health detriments such
   as insulin resistance, elevated blood pressure and dyslipidaemia, which
   are indicators of CAD ([135]7,[136]13,[137]57). Insulin is a key
   regulator of BCAA catabolism, acting through the BCKD (branched-chain
   alpha-keto acid dehydrogenase) complex. It is suggested that insulin
   resistance, common in obesity and T2D, may elevate BCAA levels due to
   reduced activity of the BCKD enzyme, leading to altered glucose
   homeostasis and an increased risk of developing T2D ([138]58).

   The findings from our study offer significant insights into the
   clinical implications of serum amino acid levels on cardiometabolic
   health. The genetic associations identified with serum valine levels
   suggest a potential causal relationship with the risk of developing
   T2D, underscoring the potential of monitoring and managing these amino
   acid levels to prevent or treat T2D. Furthermore, the study establishes
   a link between elevated serum alanine and BCAAs with dyslipidaemia,
   characterized by increased cholesterol and triglyceride levels, and
   decreased HDL. Further research is needed to examine whether serum
   amino acid profiling could be added as a part of risk assessment for
   patients predisposed to cardiometabolic disorders, particularly T2D.
   The study’s implications suggest that dietary modifications or targeted
   supplementation could be explored as potential strategies to mitigate
   risk.

Strengths and limitations

   The strength of this investigation includes the agnostic nature of the
   PheWAS and the large sample size of the UKB study for both GWAS
   (∼117 000 individuals) and PheWAS analyses (n = 323 223). The PheWAS
   design enabled us to agnostically investigate the association of serum
   amino acid status with a wide range of phenotypes (617 phenotypes).
   Finally, we validated the MR-PheWAS results using GWAS in independent
   samples as well as summary statistics from published GWAS to avoid
   false positive findings.

   This study has some limitations. Our analysis merely included
   individuals of European ancestry and the included population of the UKB
   study is young and health-oriented; thus, caution is needed when
   generalizing the results to other populations ([139]45). Some of the
   included diseases in this study have a very low prevalence in the UKB;
   thus, some of the non-significant results may be due to a lack of
   power.

Conclusion

   In conclusion, this large-scale study provided a comprehensive
   understanding of the genetic architecture of nine serum amino acids.
   The genetic risk loci we identified have provided valuable insights
   into the underlying biology of these serum amino acids. Furthermore,
   this study supports a potential causal role of genetically elevated
   levels of serum valine and increased risk of T2D, and elevated levels
   of serum alanine and BCAAs with higher levels of blood lipids. These
   findings enhance our understanding of the genetic factors influencing
   amino acid metabolism and their potential implications for various
   health conditions.

Supplementary Material

   lqae046_Supplemental_Files
   [140]lqae046_supplemental_files.zip^ (2.5MB, zip)

Acknowledgements