Circulating amino acids and the risk of macrovascular, microvascular and mortality outcomes in individuals with type 2 diabetes: results from the ADVANCE trial

Circulating amino acids and the risk of macrovascular, microvascular and mortality outcomes in... Aims/hypotheses We aimed to quantify the association of individual circulating amino acids with macrovascular disease, microvascular disease and all-cause mortality in individuals with type 2 diabetes. Methods We performed a case-cohort study (N = 3587), including 655 macrovascular events, 342 microvascular events (new or worsening nephropathy or retinopathy) and 632 all-cause mortality events during follow-up, in a secondary analysis of the Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation (ADVANCE) study. For this study, phenylalanine, isoleucine, glutamine, leucine, alanine, tyrosine, histidine and valine were measured in stored plasma samples by proton NMR metabolomics. Hazard ratios were modelled per SD increase in each amino acid. Results In models investigating associations and potential mechanisms, after adjusting for age, sex and randomised treatment, phenylalanine was positively, and histidine inversely, associated with macrovascular disease risk. These associations were attenuated to the null on further adjustment for extended classical risk factors (including eGFR and urinary albumin/creatinine ratio). After adjustment for extended classical risk factors, higher tyrosine and alanine levels were associated with decreased risk of microvascular disease (HR 0.78; 95% CI 0.67, 0.91 and HR 0.86; 95% CI 0.76, 0.98, respectively). Higher leucine (HR 0.79; 95% CI 0.69, 0.90), histidine (HR 0.89; 95% CI 0.81, 0.99) and valine (HR 0.79; 95% CI 0.70, 0.88) levels were associated with lower risk of mortality. Investigating the predictive ability of amino acids, addition of all amino acids to a risk score modestly improved classification of participants for macrovascular (continuous net reclassification index [NRI] +35.5%, p <0.001) and microvascular events (continuous NRI +14.4%, p =0.012). Conclusions/interpretation We report distinct associations between circulating amino acids and risk of different major compli- cations of diabetes. Low tyrosine appears to be a marker of microvascular risk in individuals with type 2 diabetes independently of fundamental markers of kidney function. . . . . Keywords Amino acid Diabetes complications Metabolomics Risk factors Type 2 diabetes Abbreviations Paul Welsh and Naomi Rankin are joint first authors. Mark Woodward AAA Aromatic amino acid and Naveed Sattar are joint senior authors. ACR Albumin/creatinine ratio ADVANCE Action in Diabetes and Vascular Disease: Electronic supplementary material The online version of this article (https://doi.org/10.1007/s00125-018-4619-x) contains peer-reviewed but Preterax and Diamicron Modified Release unedited supplementary material, which is available to authorised users. Controlled Evaluation BCAA Branched-chain amino acid * Paul Welsh CRP C-reactive protein Paul.Welsh@glasgow.ac.uk CVD Cardiovascular disease hsTnT High-sensitivity troponin T Extended author information available on the last page of the article. NRI Net reclassification index NT-proBNP Amino-terminal pro B-type natriuretic peptide 1582 Diabetologia (2018) 61:1581–1591 Introduction recently reported in a randomised placebo-controlled trial that metformin treatment (for 18 months in men with CHD Prior to an individual developing overt type 2 diabetes, but without type 2 diabetes) led to improved insulin sensi- there appears to be a period of subclinical metabolic tivity and was associated with increases in alanine and his- abnormality, manifesting in the altered circulating levels tidine and reductions in phenylalanine and tyrosine concen- of many metabolites [1, 2]. Specifically, several studies trations,withnoeffectonBCAAs [16]. have now reported that circulating concentrations of Therefore, the existing literature highlights inconsistent amino acids predict the development of type 2 diabetes. associations of amino acids with different outcomes in diffe- Anestedcase–control study from the Framingham rent studies, apparently contrary to the observations made in Offspring study showed that branched-chain amino acids general population studies wherein elevated levels of BCAAs (BCAAs) isoleucine, leucine and valine and aromatic and AAAs are an adverse signal. This raises the possibility amino acids (AAAs) tyrosine and phenylalanine showed that the mechanisms that influence circulating amino acids positive associations with insulin resistance and risk of might be more subtle than previously thought and as such it type 2 diabetes [3]. The European Investigation into is worth investigating and contrasting the associations of mea- Cancer and Nutrition (EPIC) Potsdam study, the surable circulating amino acids with different adverse out- Metabolic Syndrome in Men (METSIM) study, the comes in people with type 2 diabetes. Cardiovascular Risk in Young Finns (CRY) study and Avery small (N = 80) nested case–control study did the Southall and Brent Revisited (SABRE) study report- not find that amino acids were associated with diabetic ed similar findings [4–7]. Glycine and glutamine have retinopathy [17]. However, we are aware of no large also been reported to be consistently inversely associat- studies investigating the association of circulating amino ed with risk of type 2 diabetes in a meta-analysis [8]. acids with outcomes in individuals with type 2 diabetes. In general population studies, elevated levels of BCAAs Developing an understanding of any relationship be- and AAAs also appear to be associated with increased risk tween amino acids and a range of adverse outcomes in of cardiovascular disease [9–12], although these associations diabetes is important from an aetiological perspective, to have not been entirely consistent [13]. In the large Estonian develop hypotheses for intervention studies and poten- Biobank study, inverse associations between the concentra- tially to develop clinical risk scores. We thus aimed to tion of several amino acids (including BCAAs) and all- simultaneously investigate the association of circulating cause mortality were observed [14]. An inverse association amino acids with the following outcomes in people with between BCAAs and clinical dementia or Alzheimer’s type 2 diabetes: (1) macrovascular disease; (2) micro- vascular disease and (3) all-cause mortality. disease has also been observed [15]. Finally, we have Diabetologia (2018) 61:1581–1591 1583 Methods elsewhere, vitreous haemorrhage, pre-retinal haemor- rhage and fibrous proliferations on the disc or elsewhere Participants The Action in Diabetes and Vascular in a participant found not to have this condition at entry); Disease: Preterax and Diamicron Modified Release (5) development of macular oedema (characterised by a Controlled Evaluation (ADVANCE) study retinal thickening within one disc diameter of the macu- (ClinicalTrials.gov registration no. NCT00145925) lar centre in a participant not found to have this condition recruited 11,140 participants with type 2 diabetes at entry); between June 2001 and March 2003 [18]. Primary (6) occurrence of diabetes-related blindness (corrected visual outcomes of the trial have been published [19, 20]. acuity 3/60 or worse, persisting for ≥3 months and known Participants were ≥55 years of age and had been to not be due to non-diabetes-related causes in a partici- diagnosed with type 2 diabetes after the age of pant found not to have this condition at entry); 30 years. In addition, they were required to have a (7) use of retinal photocoagulation therapy. history of cardiovascular disease (CVD) or one or more additional cardiovascular risk factors. The trial included Blood samples were available from 17 out of 20 countries two randomised interventions: (1) a double-blind assess- participating in the ADVANCE study (the exceptions were ment of the efficacy of perindopril/indapamide (2 mg/0. China, India and the Philippines), giving a total potential 625mgfor 3months, increasing to4 mg/1.25 mg if source cohort size for the study of 7376 individuals (66.2% tolerated) vs placebo and (2) an open-label evaluation of of the overall study cohort). an intensive glucose-lowering regimen using modified- To improve efficiency of the biomarker studies in the release gliclazide (with a target HbA of ≤48 mmol/ ADVANCE trial a case-cohort study has been established 1c mol [6.5%]) vs standard care. Participants had their se- [21, 22]. In case-cohort studies, a random sample (called the rum creatinine levels measured as part of the study pro- ‘subcohort’) is drawn and phenotyped from the full cohort; tocol at baseline, 4 months and 1 year and annually this is very likely to contain both individuals who are ‘cases’ thereafter until completion of the study, with further and ‘non-cases’. Cases (generally for multiple case defini- tests at the discretion of clinicians. Urinary albumin/ tions, such as microvascular disease and macrovascular creatinine ratio (ACR) was measured as part of the disease) who were not included in the subcohort are then study protocol at baseline, 2 years, 4 years and comple- identified from the remainder of the cohort and were also tion of the study. GFR was estimated using the phenotyped. The case-cohort study has several advantages Modification of Diet in Renal Disease formula. over the nested case–control design, including the ability to Participants underwent formal eye examination and vi- investigate multiple endpoints simultaneously. For this case- cohort study, a random subcohort (n = 3500) was selected sual acuity testing at baseline, 2 years, 4 years and completion of the study. Each participating centre ob- from the base population, which was enriched by the addition tained ethical approval, and all participants provided of individuals who had a cardiovascular event, a microvascu- written informed consent. lar event or died during follow-up, giving a total study size of The primary trial outcomes were composites of major 4197 (Fig. 1)[21, 22]. macrovascular and microvascular events that occurred during a median of 5 years of follow-up. An independent adjudication Proton NMR analysis Plasma samples were obtained at baseline committee validated all outcomes. Major macrovascular events from all study participants when they were in an unfasted state, were cardiovascular death, non-fatal myocardial infarction or given that these were people with type 2 diabetes at risk of non-fatal stroke. Major microvascular events were defined as a hypoglycaemic episodes. Samples were collected across sites in composite of new or worsening nephropathy or retinopathy, in a pragmatic fashion (commensurate with a multinational RCT) turn defined as any of the following: according to local facilities. Plasma samples were separated and stored centrally at −80°C until measurement. The present study (1) development of macroalbuminuria (urinary ACR used a previously unthawed aliquot of plasma for H-NMR >33.9 mg/mmol, confirmed by two results); analysis. H-NMRspectroscopywasperformedonall available (2) doubling of serum creatinine level to ≥200 μmol/l (with EDTA plasma samples from the ADVANCE case-cohort study non-qualifying exceptions of terminal illness or acute at baseline using a low-volume (100 μl) variation of the quanti- illness and subsequent recovery of renal function); tative H-NMR method (Nightingale Health, Helsinki, Finland) (3) the need for renal replacement therapy due to kidney described previously [23, 24] and reviewed [25]. Sample spectra disease (in the absence of other medical causes requiring were analysed on a Bruker AVANCE III HD spectrometer to transient dialysis), or death due to renal disease; quantify a targeted list of metabolites, lipids and lipoproteins, (4) development of proliferative retinopathy (identified by as described previously [25]. This list included eight amino acids the incidence of new blood vessels on the disc or (alanine, glutamine, histidine, isoleucine, leucine, valine, 1584 Diabetologia (2018) 61:1581–1591 Fig. 1 Flow diagram for design 11,140 recruited in ADVANCE study and sample analysis in the 3764 not included in biobank ADVANCE study of amino acids (including all Chinese and (note microvascular, Indian participants) macrovascular and all-cause 7376 included in biobank mortality not mutually exclusive) 3500 randomly selected participants including Further selection of all other participants with 374 macrovascular events, 320 microvascular macrovascular events (396), microvascular events and 367 all-cause deaths during events (180) or all-cause mortality (410) during follow-up follow-up 4197 included in the case-cohort study 610 insufficient or unsuitable for NMR analysis for all metabolites Phenylalanine(3539 samples) Isoleucine (3587 samples) 648 macrovascular events 655 macrovascular events 337 microvascular events 342 microvascular events 624 all-cause deaths 632 all-cause deaths Alanine (3586 samples) Tyrosine (3579 samples) 655 macrovascular events 655 macrovascular events 342 microvascular events 341 microvascular events 632 all-cause deaths 632 all-cause deaths Glutamine (2228 samples) Leucine (3583 samples) 389 macrovascular events 654 macrovascular events 198 microvascular events 341 microvascular events 376 all-cause deaths 632 all-cause deaths Histidine (3566 samples) Valine (3587 samples) 653 macrovascular events 655 macrovascular events 339 microvascular events 342 microvascular events 631 all-cause deaths 632 all-cause deaths phenylalanine and tyrosine), which are detectable using the Cox regression models were fitted using the STSELPRE method, and are not in ‘congested’ regions of the NMR spectrum procedure for case-cohort analyses (StataCorp, College where multiple metabolites overlap. Metabolomic analyses of Station, TX, USA). Models estimated HRs for a 1 SD increase plasma samples tend to yield lower analyte concentrations than in each amino acid with each of the endpoints. Two models, serum, both by NMR spectroscopy and other methods, although with different potential confounding variables, were fitted for plasma demonstrates better stability and reproducibility [26]. each amino acid/outcome combination: model 1 with age, sex, Samples with a low glutamine/glutamate ratio were excluded region and randomised treatment; model 2 with, additionally, a from analyses of glutamine associations. Levels of all other ami- prior macrovascular complication of diabetes (myocardial in- no acids were consistent with published data. farction, stroke, hospital admission for a transient ischaemic attack or for unstable angina, coronary or peripheral Statistical analysis Continuous data with approximately nor- revascularisation, or amputation secondary to peripheral vas- mal distributions (including all amino acids) are presented as cular disease), duration of diabetes, current smoking, systolic mean ± SD; those with skewed distributions are presented as blood pressure, BMI, urinary ACR, eGFR, HbA , plasma 1c median (with interquartile range). Categorical data are pre- glucose, total cholesterol, HDL-cholesterol, triacylglycerols, sented as n (%). Pearson correlations were used to explore aspirin or other antiplatelet agent, statin or other lipid- associations of the amino acids with each other. Associations lowering agent, β-blocker, ACE inhibitor or angiotensin re- of amino acids with classical risk factors were investigated ceptor blocker, metformin use, history of heart failure, partici- across quarters of the distribution of each amino acid. pation in moderate and/or vigorous exercise for >15 min at Diabetologia (2018) 61:1581–1591 1585 Table 1 Pearson correlations (r) Amino acid Phenylalanine Isoleucine Glutamine Leucine Alanine Tyrosine Histidine of the amino acids with each other Isoleucine 0.32 – Glutamine −0.05 0.05 – Leucine 0.34 0.89 0.03 – Alanine 0.2 0.44 0.13 0.44 – Tyrosine 0.4 0.25 0.13 0.29 0.26 – Histidine 0.16 0.24 0.43 0.2 0.27 0.17 – Valine 0.27 0.67 0.13 0.75 0.34 0.29 0.25 All correlations have p values of <0.001, except for phenylalanine vs glutamine (p = 0.02), isoleucine vs gluta- mine (p = 0.03) and glutamine vs leucine (p =0.23) least once weekly, and high-sensitivity C-reactive protein sensitivity troponin T (hsTnT) and amino-terminal pro B- (CRP). A third adjustment model, attempting to include all type natriuretic peptide (NT-proBNP). In contrast, the other amino acids in the same model, resulted in collinearity and AAA, tyrosine, showed inverse associations with HbA and 1c estimates were thus not available. Non-linearity was tested ACR and a positive association with eGFR. Histidine, alanine by comparing the deviances of linear and categorical models and glutamine showed inconsistent associations with classical and by the inclusion of polynomial components (quadratic and risk factors. The BCAAs leucine, isoleucine and valine were cubic terms). Other analyses were performed using SAS v9.2 inversely associated with age, HDL-cholesterol and NT- (SAS Institute, Cary, NC, USA). All p values reported are two- proBNP but were positively associated with CVD, male sex, sided, with the 5% threshold used to determine significance. BMI, triacylglycerols and HbA . 1c For the random subcohort, the ability of amino acids to discriminate between those who will and those who will not Macrovascular disease, microvascular disease and all-cause go on to suffer each of the three adverse outcomes were esti- mortality Baseline risk factors associated with all three end- mated, in the context of model 2, using c statistics for 5 year points included male sex, increased duration of diabetes, his- risk, accounting for censoring. In addition, the ability of amino tory of macrovascular disease, higher systolic blood pressure, acids to reclassify participants according to 5 year risk, using lower HDL-cholesterol, higher HbA , higher ACR and 1c the continuous net reclassification index (NRI), was assessed higher hsTnT and NT-proBNP (Table 2). by methods suitable for survival data, using bootstrapping [27]. Among the amino acids, after adjustment for age, sex, re- Primary results came from use of all available data; sensi- gion and randomised treatment (model 1), higher phenylala- tivity analyses using only participants with complete data nine and lower glutamine and histidine concentrations were were also performed. associated with increased macrovascular risk (HR per 1 SD increase was 1.22 [95% CI 1.12, 1.32], 0.88 [95% CI 0.79, 0.98] and 0.86 [95% CI 0.79, 0.94], respectively) but these Results associations were attenuated to the null on further adjustment for classical risk factors (model 2) (Fig. 2aand ESMTable 10). Baseline associations A maximum of 3587 samples had avai- Higher tyrosine alone was associated with decreased risk of lable data for at least one amino acid (Fig. 1). Due to the design microvascular events (HR 0.74 [95% CI 0.64, 0.86]) in model of the multicentre study, there was some variability in sample 1 and this was only slightly attenuated on adjustment for a full processing time, leading to some samples having low range of classical risk factors in model 2 (HR 0.78 [95% CI glutamine/glutamate ratios. As such, fewer samples had a result 0.67, 0.91]) (Fig. 2b and ESM Table 10). A higher alanine for glutamine [28]. The detected absolute concentrations of level was also associated with decreased risk of microvascular amino acids were generally comparable with data from other events after further adjustment (HR 0.86 [95% CI 0.76, 0.98]). studies (see electronic supplementary material [ESM] Table 1). The association between tyrosine and renal impairment was In general, the amino acids showed a broad range of corre- further investigated by assessing the HRs across tertiles of lations with each other. Taking extreme examples, leucine and eGFR and ACR. There was no evidence of interaction by glutamine were not correlated (r =0.03, p = 0.23) but BCAAs eGFR or ACR (data not shown). leucine, isoleucine and valine were highly intercorrelated (r ≥ In contrast, several amino acids were associated with all- 0.67, p < 0.001) (Table 1). The associations of amino acids cause mortality. Phenylalanine was positively associated with with classical CVD risk factors are shown in ESM Tables 2– risk of mortality, while glutamine, leucine, alanine, histidine and 9. Phenylalanine was positively associated with older age, valine were all inversely associated with risk of mortality in baseline CVD, higher CRP and higher baseline high- model 1 (ESM Table 10). After adjustment for classical risk 1586 Diabetologia (2018) 61:1581–1591 Table 2 Baseline characteristics of the cohort classified by outcome status Characteristic Macrovascular disease Microvascular disease All-cause mortality Yes No p value Yes No p value Yes No p value N 655 2932 342 3245 632 2955 Age, years 68.92 ± 6.52 66.36 ± 6.51 <0.001 65.85 ± 6.38 66.93 ± 6.60 0.004 69.94 ± 6.56 66.17 ± 6.40 <0.001 Male sex 451 (68.9) 1719 (58.6) <0.001 227 (66.4) 1943 (59.9) 0.020 439 (69.5) 1731 (58.6) <0.001 Region ANZ/SEA 155 (23.7) 714 (24.4) 0.375 123 (36.0) 746 (23.0) <0.001 120 (19.0) 749 (25.3) <0.001 Canada 33 (5.0) 185 (6.3) 28 (8.2) 190 (5.9) 34 (5.4) 184 (6.2) Continental Europe 262 (40.0) 1157 (39.5) 91 (26.6) 1328 (40.9) 264 (41.8) 1155 (39.1) Northern Europe 205 (31.3) 876 (29.9) 100 (29.2) 981 (30.2) 214 (33.9) 867 (29.3) Duration of diabetes, years 9.19 ± 7.10 7.61 ± 6.29 <0.001 9.74 ± 6.89 7.71 ± 6.40 <0.001 9.24 ± 7.60 7.62 ± 6.17 <0.001 Current smoker 94 (14.4) 439 (15.0) 0.686 48 (14.0) 485 (14.9) 0.652 103 (16.3) 430 (14.6) 0.263 History of macrovascular disease 323 (49.3) 929 (31.7) <0.001 77 (22.5) 276 (8.5) <0.001 283 (44.8) 969 (32.8) <0.001 History of heart failure 56 (8.5) 110 (3.8) <0.001 13 (3.8) 153 (4.7) 0.445 61 (9.7) 105 (3.6) <0.001 Participation in moderate or vigorous 266 (40.6) 1470 (50.1) <0.001 164 (48.0) 1572 (48.4) 0.863 262 (41.5) 1474 (49.9) <0.001 activity Diastolic BP, mmHg 81.55 ± 11.41 81.63 ± 10.74 0.863 81.73 ± 11.30 81.60 ± 10.82 0.831 80.55 ± 11.70 81.84 ± 10.67 0.007 Total cholesterol, mmol/l 5.11 ± 1.18 5.15 ± 1.17 0.500 5.16 ± 1.08 5.14 ± 1.18 0.737 5.06 ± 1.10 5.16 ± 1.18 0.058 HDL-cholesterol, mmol/l 1.17 ± 0.31 1.24 ± 0.33 <0.001 1.18 ± 0.31 1.23 ± 0.33 0.005 1.18 ± 0.31 1.23 ± 0.33 <0.001 Triacylglycerol, mmol/l 1.63 (1.20, 2.30) 1.70 (1.20, 2.36) 0.436 1.80 (1.27, 2.60) 1.69 (1.20, 2.31) 0.01 1.61 (1.20, 2.30) 1.70 (1.20, 2.36) 0.4912 HbA , mmol/mol 59.5 ± 17.2 56.9 ± 14.8 61.3 ± 17.5 56.9 ± 15.4 59.3 ± 16.8 56.9 ± 15.2 1c HbA , % 7.59 ± 1.60 7.36 ± 1.39 <0.001 7.76 ± 1.60 7.36 ± 1.41 <0.001 7.58 ± 1.58 7.36 ± 1.40 <0.001 1c Glucose, mmol/l 8.61 ± 2.85 8.43 ± 2.68 0.115 9.04 ± 3.37 8.40 ± 2.62 <0.001 8.53 ± 2.88 8.45 ± 2.67 0.453 Urinary ACR, mg/mmol 2.4 (1.0, 8.0) 1.5 (0.7, 4.0) <0.001 5.6 (1.6, 14.2) 1.5 (0.7, 3.8) <0.001 2.4 (0.9, 7.6) 1.5 (0.7, 4.0) <0.001 −1 −2 eGFR, ml min 1.73 m 67.68 ± 17.61 72.70 ± 16.37 <0.001 69.97 ± 18.74 71.98 ± 16.48 0.034 66.58 ± 17.57 72.89 ± 16.32 <0.001 CRP, nmol/l 19.33 (9.05, 42.38) 16.67 (8.10, 38.48) 0.012 15.71 (8.48, 32.95) 17.43 (8.29, 39.33) 0.249 19.90 (9.71, 45.81) 16.86 (8.10, 37.71) <0.001 hsTnT, pg/ml 9 (4, 17) 5 (1.50, 10) <0.001 7 (1.50, 13) 5 (1.50, 11) <0.001 10 (4, 18) 5 (1.50, 9) <0.001 NT-proBNP, pg/ml 198 (74, 479) 74 (30, 169) <0.001 107 (38, 260) 87 (34, 212) 0.010 201 (80, 506) 74 (30, 171) <0.001 Medication use Aspirin or other antiplatelet agent 387 (59.1) 1380 (47.1) <0.001 171 (50.0) 1596 (49.2) 0.774 352 (55.7) 1415 (47.9) <0.001 Statin or other lipid-lowering agent 283 (43.2) 1311 (44.7) 0.484 158 (46.2) 1436 (44.3) 0.490 260 (41.1) 1334 (45.1) 0.066 β blocker 211 (32.2) 881 (30.0) 0.276 96 (28.1) 996 (30.7) 0.316 196 (31.0) 896 (30.3) 0.731 ACE inhibitor or angiotensin 418 (63.8) 1671 (57.0) 0.0014 232 (67.8) 1857 (57.2) <0.001 395 (62.5) 1694 (57.3) 0.017 receptor blocker Aminoacidlevel,mmol/l Phenylalanine 0.063 ± 0.010 0.061 ± 0.009 <0.001 0.062 ± 0.009 0.062 ± 0.009 0.1327 0.063 ± 0.010 0.061 ± 0.009 <0.001 Isoleucine 0.063 ± 0.017 0.062 ± 0.017 0.7429 0.065 ± 0.017 0.062 ± 0.017 0.0059 0.061 ± 0.017 0.063 ± 0.017 0.0643 Diabetologia (2018) 61:1581–1591 1587 Table 2 (continued) Characteristic Macrovascular disease Microvascular disease All-cause mortality Yes No p value Yes No p value Yes No p value Glutamine 0.373 ± 0.112 0.380 ± 0.109 0.2634 0.369 ± 0.118 0.380 ± 0.109 0.1696 0.367 ± 0.112 0.381 ± 0.109 0.0247 Leucine 0.081 ± 0.019 0.082 ± 0.020 0.0679 0.084 ± 0.020 0.082 ± 0.020 0.0518 0.079 ± 0.020 0.082 ± 0.020 <0.001 Alanine 0.366 ± 0.064 0.371 ± 0.065 0.0591 0.370 ± 0.066 0.370 ± 0.064 0.9372 0.361 ± 0.063 0.372 ± 0.065 <0.001 Tyrosine 0.053 ± 0.012 0.053 ± 0.011 0.9293 0.050 ± 0.012 0.053 ± 0.011 <0.001 0.052 ± 0.012 0.053 ± 0.011 0.1873 Histidine 0.049 ± 0.010 0.050 ± 0.009 0.0032 0.050 ± 0.009 0.050 ± 0.010 0.4117 0.048 ± 0.010 0.050 ± 0.009 <0.001 Valine 0.172 ± 0.035 0.175 ± 0.035 0.1355 0.177 ± 0.037 0.174 ± 0.035 0.1729 0.167 ± 0.036 0.176 ± 0.035 <0.001 Values are mean ±SD, median (interquartile range) or n (%) ANZ, Australia and New Zealand; SEA, south-east Asia 1588 Diabetologia (2018) 61:1581–1591 factors (model 2), the inverse association with risk remained for Phenylalanine leucine (HR 0.79 [95% CI 0.69, 0.90]), histidine (HR 0.89 Isoleucine [95% CI 0.81, 0.99]) and valine (HR 0.79 [95% CI 0.70, Glutamine 0.88]) but the positive association of phenylalanine was attenu- Leucine ated to the null (Fig. 2c and ESM Table 10). A sensitivity analysis using samples from individuals with complete data Alanine gave similar results (ESM Table 11). There was no evidence Tyrosine of a randomised treatment interaction in any model (data not Histidine shown). Valine A model including all the classical CVD risk factors in model 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 2 yielded a c statistic of 0.716 for macrovascular events, 0.728 HR (95% CI) for microvascular events and 0.747 for all-cause mortality (Table 3). Addition of the amino acids in combination did not Phenylalanine improve the c statistic for any endpoint but did improve the Isoleucine continuous NRI for macrovascular events (+35.5%, p<0.001) Glutamine and microvascular events (+14.4%, p = 0.012). The improve- Leucine ment in prediction of microvascular events was driven by the Alanine addition of tyrosine alone. Tyrosine Histidine Valine Discussion 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 HR (95% CI) Although previous observational studies have reported asso- ciations of circulating BCAAs and AAAs with adverse out- Phenylalanine comes in healthy people, the present report contrasts for the Isoleucine first time the associations of multiple circulating amino acids Glutamine with the major vascular complications of diabetes. Rather than Leucine one (or more) amino acids being a consistent signal for adverse Alanine outcomes of any kind, we report that their associations with Tyrosine risk of macrovascular events, microvascular events and all- Histidine cause mortality are strikingly different from each other. A key finding is the inverse association of tyrosine with risk of Valine microvascular events, independent of eGFR and urinary ACR. 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 Although the evidence from the present study suggests that HR (95% CI) these might only be very moderately useful biomarkers in in- Fig. 2 Adjusted associations (model 2, log scale HR) of amino acids individually (per 1 SD increase) with macrovascular outcomes (a), mi- cremental prediction of adverse events in individuals with type crovascular outcomes (b) and all-cause mortality (c) 2 diabetes, the pathophysiology underlying these associations Table 3 Prediction of endpoints using amino acids in combination or individually Model Macrovascular events Microvascular events All-cause mortality c statistic Continuous NRI (%) c statistic Continuous NRI (%) c statistic Continuous NRI (%) Basic model 0.716 – 0.728 – 0.747 – Plus all amino acids +0.010 +35.5 +0.010 +14.4 +0.009 +8.0 p value 0.17 <0.001 0.23 0.012 0.26 0.09 Plus tyrosine only –– 0.007 +14.9 –– p value –– 0.30 0.01 –– Adjusted for age, sex, region, randomised treatment, previous macrovascular event, duration of diabetes, current smoking, systolic blood pressure, BMI, urinary ACR, eGFR, HbA , plasma glucose, total and HDL-cholesterol, triacylglycerols, use of aspirin or other antiplatelet agent, statin or other 1c lipid-lowering agent, β-blocker, ACE inhibitor or angiotensin receptor blocker, or metformin, history of heart failure, participation in moderate and/or vigorous exercise for >15 min at least once weekly, and CRP Adjusted associations Adjusted associations Adjusted associations with all-cause with microvascular with macrovascular mortality outcomes outcomes Diabetologia (2018) 61:1581–1591 1589 and the possibility of intervention studies are intriguing and have reported that metformin in fact lowers, not raises, tyrosine worthy of further investigation. levels in individuals with CVD and at high risk of diabetes [16]. The association of high circulating concentrations of BCAAs The effect of other glucose-lowering drugs on amino acid profiles and AAAs with obesity has been known since the 1960s [29]and in individuals with diabetes would now be of interest. Further has been proposed to be at least partially mediated by insulin studies are now needed to validate our novel observations and to resistance. Insulin is thought to be a regulator of branched- examine whether our findings may represent causal pathways. chain α-keto acid dehydrogenase complex [30]. Insulin resis- Strengths of the study include the use of a well-characterised tance may hence suppress BCAA catabolism, as suggested by clinical trial cohort, an efficiently designed case-cohort study to associations noted in observational epidemiology studies yield a powerful study for a range of endpoints, which were [31–33]. The causal pathway may not be unidirectional; a recent independently adjudicated according to pre-defined criteria. Mendelian randomisation study suggests that genetically elevat- Like other RCT populations, ADVANCE study participants rep- ed BCAAs (via impaired catabolism) are associated with in- resent a selected cohort. For instance, ADVANCE study partic- creased risk of type 2 diabetes [34], although a better understand- ipants were required to have a history of CVD or CVD risk ing of the underlying pathway is required to increase confidence factors. Therefore, our results may not be generalisable to all in this observation [35]. There is also the possibility that amino individuals with diabetes, although other risk factors we mea- acids themselves (particularly BCAAs) may affect metabolism sured are generally associated with risk of major endpoints in by suppressing postprandial glucose levels [36]. Increased pro- the expected directions. Amino acids were measured in pragmat- tein turnover in people with central obesity may result in higher ically collected plasma samples in the context of a multinational circulating levels of amino acids [37] and might therefore cause RCT and we cannot rule out the potential for differential pre- elevations in amino acids in people who are overweight and have analytical sample handling or sample degradation during storage, type 2 diabetes. There are hence a variety of potential mecha- which may have biased our results [44], although these samples nisms related to type 2 diabetes pathologies that might influence were analysed at first thaw. We also present data suggesting circulating amino acids in individuals with type 2 diabetes. broadly comparable concentrations of amino acids relative to Given this background, and prior findings in general popula- other cohorts. Another potential limitation is the analysis of sam- tions of associations of specific amino acids with CVD [9–12], ples from non-fasted participants, although in clinical practice, we wished to examine whether amino acid levels associated with fasting is rarely required among individuals with type 2 diabetes. adverse outcomes in individuals with type 2 diabetes. In the NMR spectroscopy has been used to investigate changes in ami- ADVANCE study, the BCAAs leucine, valine and isoleucine no acids 30 min after a standardised liquid meal [45] and effects showed no association with macrovascular events, but low levels sizes were generally relatively small, although the immediate of leucine and valine were associated with increased all-cause postprandial state is likely to give larger effect estimates than are atplayinthisstudy. mortality. However, the positive, albeit not independently predic- tive, association of phenylalanine with CVD and all-cause mor- In conclusion, we report distinct associations of different ami- tality we observed is broadly in line with other published data. no acids with risk of major adverse endpoints in individuals with There are limited intervention studies investigating the effect of type 2 diabetes. Most notably, the identification of tyrosine as a amino acid supplements on health outcomes, with most research potential marker of microvascular risk requires further study. coming from short-term trials examining surrogate health markers in the sports science area [38]. Our data strongly support Acknowledgements We thank E. Butler, University of Glasgow, UK for the need for further studies to determine why higher phenylala- technical assistance in conducting the study. nine appears to be a consistently adverse signal for CVD out- comes. Our study provides observations that are the basis for Data availability Summaries of the ADVANCE trial data can be found at testable hypotheses investigating the effect of genetic variants, http://www.advance-trial.com. Restrictions apply to the availability of these data, which were used by agreement of the ADVANCE steering which are instrumental variables for circulating amino acids, on committee for the current study, and so are not publicly available. health outcomes [39, 40]. The inverse association of tyrosine with risk of microvascular Funding The biomarker work in the present study was funded by the events is perhaps the most intriguing individual finding from this Chest Heart and Stroke Association Scotland (R13/A149) and by the Glasgow Molecular Pathology NODE, which is funded by The Medical study. Tyrosine itself was positively associated with baseline Research Council and The Engineering and Physical Sciences Research eGFR and inversely associated with baseline HbA and urinary 1c Council (MR/N005813/1). The ADVANCE trial (ClinicalTrials.gov ACR. Impaired conversion of phenylalanine to tyrosine has been registration no. NCT00145925) was funded by the National Health and reported in renal disease [41, 42]. Low tyrosine levels might Medical Research Council (NHMRC) of Australia (project grant ID 211086 and program grant IDs 358395 and 571281) and by Servier. therefore simply reflect impaired kidney function, which itself PWu is supported by the Academy of Finland (312476 and 312477) predicts future microvascular events. Tyrosine is also linked to and the Novo Nordisk Foundation. MAK was supported by the Sigrid catecholamine synthesis, which, also speculatively, might be rel- Juselius Foundation, Finland. MAK works in a Unit that is supported by evant to our findings [43]. That noted, counter-intuitively, we 1590 Diabetologia (2018) 61:1581–1591 the University of Bristol and UK Medical Research Council (MC_UU_ risk of subsequent cardiovascular events. Circ Cardiovasc Genet 3: 12013/1). The study sponsors were not involved in the design of the 207–214 study, the collection, analysis, and interpretation of data, writing the report 11. Magnusson M, Lewis GD, Ericson U et al (2013) A diabetes- or the decision to submit the report for publication. predictive amino acid score and future cardiovascular disease. Eur Heart J 34:1982–1989 Duality of interest JC has received research grants from Servier as 12. Ruiz-Canela M, Toledo E, Clish CB et al (2016) Plasma branched- Principal investigator for ADVANCE and for the ADVANCE-ON post trial chain amino acids and incident cardiovascular disease in the follow-up study and honoraria from Servier for speaking about these studies PREDIMED trial. Clin Chem 62:582–592 at scientific meetings. MW reports receiving consulting fees from Amgen. 13. Floegel A, Kühn T, Sookthai D et al (2018) Serum metabolites and PWu is an employee and shareholder of Nightingale Health Ltd, which risk of myocardial infarction and ischemic stroke: a targeted conducted the biomarker quantification. All other authors declare that there metabolomic approach in two German prospective cohorts. Eur J is no duality of interest associated with their contribution to this paper. Epidemiol 33:55–66 14. Fischer K, Kettunen J, Würtz P et al (2014) Biomarker profiling by Contribution statement MM, NP, PH, MW and JC conceived, designed nuclear magnetic resonance spectroscopy for the prediction of all- and acquired the ADVANCE trial data. PWe, MW, NR, PM and NS con- cause mortality: an observational study of 17,345 persons. PLoS ceived this secondary study, and PWe and NS obtained grant funding. PWu Med 11:e1001606 and MA-K acquired biomarker data. MW and QL undertook the statistical 15. Tynkkynen J, Chouraki V, Van der Lee S et al (2018) Association of analyses. All authors were involved in data interpretation. PWe and NR branched-chain amino acids and other circulating metabolites with wrote the initial drafts of the manuscript. These drafts were revised for risk of incident dementia and Alzheimer s disease: a prospective important scientific content by MW, NS, PM, PWu, MA-K, QL, MM, study in eight cohorts. Alzheimers Dement. https://doi.org/10. NP, PH and JC. All authors gave final approval of the version to be pub- 1016/j.jalz.2018.01.003 lished. MW is the guarantor of this work. 16. Preiss D, Rankin N, Welsh P et al (2016) Effect of metformin therapy on circulating amino acids in a randomized trial: the Open Access This article is distributed under the terms of the Creative CAMERA study. Diabet Med 33:1569–1574 Commons Attribution 4.0 International License (http:// 17. Chen L, Cheng C, Choi H et al (2016) Plasma metabonomic pro- creativecommons.org/licenses/by/4.0/), which permits unrestricted use, filing of diabetic retinopathy. Diabetes 65:1099–1108 distribution, and reproduction in any medium, provided you give appro- 18. ADVANCE Management Committee (2001) Study rationale and priate credit to the original author(s) and the source, provide a link to the design of ADVANCE: Action in Diabetes and Vascular disease – Creative Commons license, and indicate if changes were made. Preterax and Diamicron MR Controlled Evaluation. Diabetologia 44:1118–1120 19. ADVANCE Collaborative Group, Patel A, MacMahon S et al (2008) Intensive blood glucose control and vascular outcomes in References patients with type 2 diabetes. N Engl J Med 358:2560–2572 20. Patel A, ADVANCE Collaborative Group, MacMahon S et al 1. Gaillard T, Osei K (2016) Ethnic differences in serum lipids and lipo- (2007) Effects of a fixed combination of perindopril and proteins in overweight/obese African-American and white American indapamide on macrovascular and microvascular outcomes in pa- women with pre-diabetes: significance of NMR-derived lipoprotein tients with type 2 diabetes mellitus (the ADVANCE trial): a particle concentrations and sizes. BMJ Open Diabetes Res Care 4: randomised controlled trial. Lancet 370:829–840 e000246 21. Hillis GS, Welsh P, Chalmers J et al (2014) The relative and com- 2. Liggi S, Griffin JL (2017) Metabolomics applied to diabetes-lessons bined ability of high-sensitivity cardiac troponin T and N-terminal from human population studies. Int J Biochem Cell Biol 93:136–147 pro-B-type natriuretic peptide to predict cardiovascular events and 3. Wang TJ, Larson MG, Vasan RS et al (2011) Metabolite profiles death in patients with type 2 diabetes. Diabetes Care 37:295–303 and the risk of developing diabetes. Nat Med 17:448–453 22. Welsh P, Woodward M, Hillis GS et al (2014) Do cardiac bio- 4. Floegel A, Stefan N, Yu Z, Mühlenbruch K et al (2013) markers NT-proBNP and hsTnT predict microvascular events in Identification of serum metabolites associated with risk of type patients with type 2 diabetes? Results from the ADVANCE trial. 2 diabetes using a targeted metabolomic approach. Diabetes Diabetes Care 37:2202–2210 62:639–648 23. Würtz P, Kangas AJ, Soininen P, Lawlor DA, Davey Smith G, Ala- 5. Tillin T, Hughes AD, Wang Q et al (2015) Diabetes risk and amino Korpela M (2017) Quantitative serum nuclear magnetic resonance acid profiles: cross-sectional and prospective analyses of ethnicity, metabolomics in large-scale epidemiology: a primer on -omic tech- amino acids and diabetes in a South Asian and European cohort nology. Am J Epidemiol 186:1084–1096 from the SABRE (Southall And Brent REvisited) Study. 24. Soininen P, Kangas AJ, Würtz P et al (2009) High-throughput se- Diabetologia 58:968–979 rum NMR metabonomics for cost-effective holistic studies on sys- 6. Würtz P, Soininen P, Kangas AJ et al (2013) Branched-chain and temic metabolism. Analyst 134:1781–1785 aromatic amino acids are predictors of insulin resistance in young 25. Soininen P, Kangas AJ, Würtz P, Suna T, Ala-Korpela M (2015) adults. Diabetes Care 36:648–655 Quantitative serum nuclear magnetic resonance metabolomics in 7. Stancakova A, Civelek M, Saleem NK et al (2012) Hyperglycemia cardiovascular epidemiology and genetics. Circ Cardiovasc Genet and a common variant of GCKR are associated with the levels of 8:192–206 eight amino acids in 9,369 Finnish Men. Diabetes 61:1895–1902 26. Yu Z, Kastenmüller G, He Y et al (2011) Differences between hu- 8. Guasch-Ferré M, Hruby A, Toledo E et al (2016) Metabolomics in man plasma and serum metabolite profiles. PLoS One 6:e21230 prediabetes and diabetes: a systematic review and meta-analysis. 27. Pencina MJ, D’Agostino RB, Steyerberg EW (2011) Extensions of Diabetes Care 39:833–846 9. Würtz P, Havulinna AS, Soininen P et al (2015) Metabolite profiling net reclassification improvement calculations to measure usefulness and cardiovascular event risk: a prospective study of 3 population- of new biomarkers. Stat Med 30:11–21 based cohorts. Circulation 131:774–785 28. Grossie VB, Yick J, Alpeter M, Welbourne TC, Ota DM (1993) 10. Shah SH, Bain JR, Muehlbauer MJ et al (2010) Association of a Glutamine stability in biological tissues evaluated by fluorometric peripheral blood metabolic profile with coronary artery disease and analysis. Clin Chem 39:1059–1063 Diabetologia (2018) 61:1581–1591 1591 29. Felig P, Marliss E, Cahill GF (1969) Plasma amino acid levels and 37. Tai ES, Tan MLS, Stevens RD et al (2010) Insulin resistance is associated with a metabolic profile of altered protein metabolism insulin secretion in obesity. N Engl J Med 281:811–816 in Chinese and Asian-Indian men. Diabetologia 53:757–767 30. Costeas PA, Chinsky JM (1996) Effects of insulin on the regulation 38. Xu Z, Tan Z, Zhang Q, Gui Q, Yang Y (2014) Clinical effectiveness of branched-chain α-ketoaciddehydrogenase E1α subunit gene of protein and amino acid supplementation on building muscle expression. Biochem J 318:85–92 mass in elderly people: a meta-analysis. PLoS One e10:9141 31. Shah SH, Crosslin DR, Haynes CS et al (2012) Branched-chain 39. Yu B, de Vries PS, Metcalf GA et al (2016) Whole genome se- amino acid levels are associated with improvement in insulin resis- quence analysis of serum amino acid levels. Genome Biol 17:237 tance with weight loss. Diabetologia 55:321–330 40. Teslovich TM, Kim DS, Yin X et al (2018) Identification of seven 32. McCormack SE, Shaham O, McCarthy MA et al (2013) Circulating novel loci associated with amino acid levels using single variant and branched-chain amino acid concentrations are associated with obe- gene-based tests in 8545 Finnish men from the METSIM study. sity and future insulin resistance in children and adolescents. Pediatr Hum Mol Genet. 27:1664–1674 Obes 8:52–61 41. Kopple JD (2007) Phenylalanine and tyrosine metabolism in chronic 33. Würtz P, Wang Q, Kangas AJ et al (2014) Metabolic signatures of kidney failure. J Nutr 137:1586S–1590S adiposity in young adults: Mendelian randomization analysis and 42. Druml W, Roth E, Lenz K, Lochs H, Kopsa H (1989) Phenylalanine effects of weight change. PLoS Med 11:e1001765 and tyrosine metabolism in renal failure: dipeptides as tyrosine 34. Lotta LA, Scott RA, Sharp SJ et al (2016) Genetic predisposition to source. Kidney Int Suppl 27:s282–s286 an impaired metabolism of the branched-chain amino acids and risk 43. Fernstrom JD, Fernstrom MH (2007) Tyrosine, phenylalanine, and of type 2 diabetes: a Mendelian randomisation analysis. PLoS Med catecholamine synthesis and function in the brain. J Nutr 137: 13:e1002179 1539S–1547S 35. Holmes MV, Ala-Korpela M, Smith GD (2017) Mendelian random- 44. Anton G, Wilson R, Yu Z et al (2015) Pre-analytical sample ization in cardiometabolic disease: challenges in evaluating causali- quality: metabolite ratios as an intrinsic marker for prolonged ty. Nat Rev Cardiol 14:577–590 room temperature exposure of serum samples. PLoS One 10: 36. Manders RJ, Little JP, Forbes SC, Candow DG (2012) e0121495 Insulinotropic and muscle protein synthetic effects of branched- 45. Schutte B, Van den Akker EB, Deelen J et al (2016) The effect of chain amino acids: potential therapy for type 2 diabetes and standardized food intake on the association between BMI and 1H- sarcopenia. Nutrients 4:1664–1678 NMR metabolites. Sci Rep 6:38980–38985 Affiliations 1 1 2 1 3,4 5,6,7,8,9 Paul Welsh & Naomi Rankin & Qiang Li & Patrick B. Mark & Peter Würtz & Mika Ala-Korpela & 10,11,12 13 14,15,16 2 2,17,18 1 Michel Marre & Neil Poulter & Pavel Hamet & John Chalmers & Mark Woodward & Naveed Sattar 1 10 BHF Glasgow Cardiovascular Research Centre, Institute of Inserm, UMRS 1138, Centre de Recherche des Cordeliers, Cardiovascular & Medical Sciences, University of Glasgow, 126 Paris, France University Place, Glasgow G12 8TA, UK Assistance Publique Hôpitaux de Paris, Bichat Hospital, DHU The George Institute for Global Health, University of New South FIRE, Department of Diabetology, Endocrinology and Wales, Sydney, NSW, Australia Nutrition, Paris, France 3 12 Research Programs Unit, Diabetes and Obesity, University of University Paris Diderot, Sorbonne Paris Cité, UFR de Helsinki, Helsinki, Finland Médecine, Paris, France 4 13 Nightingale Health Ltd, Helsinki, Finland International Centre for Circulatory Health, Imperial College, London, UK Computational Medicine, Faculty of Medicine, University of Oulu and Biocenter Oulu, Oulu, Finland Department of Experimental Medicine, McGill University, Montreal, QC, Canada NMR Metabolomics Laboratory, School of Pharmacy, University of Eastern Finland, Kuopio, Finland Department of Medicine, CRCHUM, Université de Montréal, Montreal, QC, Canada Population Health Science, Bristol Medical School, University of Bristol and Medical Research Council Integrative Epidemiology Department of Medicine, Gene Medicine Services, CRCHUM, Unit at the University of Bristol, Bristol, UK Université de Montréal, Montreal, QC, Canada 8 17 Systems Epidemiology, Baker Heart and Diabetes Institute, The George Institute for Global Health, University of Oxford, Melbourne, VIC, Australia Oxford, UK 9 18 Department of Epidemiology and Preventive Medicine, School Department of Epidemiology, Johns Hopkins University, of Public Health and Preventive Medicine, Faculty of Medicine, Baltimore, MD, USA Nursing and Health Sciences, The Alfred Hospital, Monash University, Melbourne, VIC, Australia http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diabetologia Springer Journals

Circulating amino acids and the risk of macrovascular, microvascular and mortality outcomes in individuals with type 2 diabetes: results from the ADVANCE trial

Free
11 pages

Loading next page...
 
/lp/springer_journal/circulating-amino-acids-and-the-risk-of-macrovascular-microvascular-3nZ7twli5f
Publisher
Springer Berlin Heidelberg
Copyright
Copyright © 2018 by The Author(s)
Subject
Medicine & Public Health; Internal Medicine; Metabolic Diseases; Human Physiology
ISSN
0012-186X
eISSN
1432-0428
D.O.I.
10.1007/s00125-018-4619-x
Publisher site
See Article on Publisher Site

Abstract

Aims/hypotheses We aimed to quantify the association of individual circulating amino acids with macrovascular disease, microvascular disease and all-cause mortality in individuals with type 2 diabetes. Methods We performed a case-cohort study (N = 3587), including 655 macrovascular events, 342 microvascular events (new or worsening nephropathy or retinopathy) and 632 all-cause mortality events during follow-up, in a secondary analysis of the Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation (ADVANCE) study. For this study, phenylalanine, isoleucine, glutamine, leucine, alanine, tyrosine, histidine and valine were measured in stored plasma samples by proton NMR metabolomics. Hazard ratios were modelled per SD increase in each amino acid. Results In models investigating associations and potential mechanisms, after adjusting for age, sex and randomised treatment, phenylalanine was positively, and histidine inversely, associated with macrovascular disease risk. These associations were attenuated to the null on further adjustment for extended classical risk factors (including eGFR and urinary albumin/creatinine ratio). After adjustment for extended classical risk factors, higher tyrosine and alanine levels were associated with decreased risk of microvascular disease (HR 0.78; 95% CI 0.67, 0.91 and HR 0.86; 95% CI 0.76, 0.98, respectively). Higher leucine (HR 0.79; 95% CI 0.69, 0.90), histidine (HR 0.89; 95% CI 0.81, 0.99) and valine (HR 0.79; 95% CI 0.70, 0.88) levels were associated with lower risk of mortality. Investigating the predictive ability of amino acids, addition of all amino acids to a risk score modestly improved classification of participants for macrovascular (continuous net reclassification index [NRI] +35.5%, p <0.001) and microvascular events (continuous NRI +14.4%, p =0.012). Conclusions/interpretation We report distinct associations between circulating amino acids and risk of different major compli- cations of diabetes. Low tyrosine appears to be a marker of microvascular risk in individuals with type 2 diabetes independently of fundamental markers of kidney function. . . . . Keywords Amino acid Diabetes complications Metabolomics Risk factors Type 2 diabetes Abbreviations Paul Welsh and Naomi Rankin are joint first authors. Mark Woodward AAA Aromatic amino acid and Naveed Sattar are joint senior authors. ACR Albumin/creatinine ratio ADVANCE Action in Diabetes and Vascular Disease: Electronic supplementary material The online version of this article (https://doi.org/10.1007/s00125-018-4619-x) contains peer-reviewed but Preterax and Diamicron Modified Release unedited supplementary material, which is available to authorised users. Controlled Evaluation BCAA Branched-chain amino acid * Paul Welsh CRP C-reactive protein Paul.Welsh@glasgow.ac.uk CVD Cardiovascular disease hsTnT High-sensitivity troponin T Extended author information available on the last page of the article. NRI Net reclassification index NT-proBNP Amino-terminal pro B-type natriuretic peptide 1582 Diabetologia (2018) 61:1581–1591 Introduction recently reported in a randomised placebo-controlled trial that metformin treatment (for 18 months in men with CHD Prior to an individual developing overt type 2 diabetes, but without type 2 diabetes) led to improved insulin sensi- there appears to be a period of subclinical metabolic tivity and was associated with increases in alanine and his- abnormality, manifesting in the altered circulating levels tidine and reductions in phenylalanine and tyrosine concen- of many metabolites [1, 2]. Specifically, several studies trations,withnoeffectonBCAAs [16]. have now reported that circulating concentrations of Therefore, the existing literature highlights inconsistent amino acids predict the development of type 2 diabetes. associations of amino acids with different outcomes in diffe- Anestedcase–control study from the Framingham rent studies, apparently contrary to the observations made in Offspring study showed that branched-chain amino acids general population studies wherein elevated levels of BCAAs (BCAAs) isoleucine, leucine and valine and aromatic and AAAs are an adverse signal. This raises the possibility amino acids (AAAs) tyrosine and phenylalanine showed that the mechanisms that influence circulating amino acids positive associations with insulin resistance and risk of might be more subtle than previously thought and as such it type 2 diabetes [3]. The European Investigation into is worth investigating and contrasting the associations of mea- Cancer and Nutrition (EPIC) Potsdam study, the surable circulating amino acids with different adverse out- Metabolic Syndrome in Men (METSIM) study, the comes in people with type 2 diabetes. Cardiovascular Risk in Young Finns (CRY) study and Avery small (N = 80) nested case–control study did the Southall and Brent Revisited (SABRE) study report- not find that amino acids were associated with diabetic ed similar findings [4–7]. Glycine and glutamine have retinopathy [17]. However, we are aware of no large also been reported to be consistently inversely associat- studies investigating the association of circulating amino ed with risk of type 2 diabetes in a meta-analysis [8]. acids with outcomes in individuals with type 2 diabetes. In general population studies, elevated levels of BCAAs Developing an understanding of any relationship be- and AAAs also appear to be associated with increased risk tween amino acids and a range of adverse outcomes in of cardiovascular disease [9–12], although these associations diabetes is important from an aetiological perspective, to have not been entirely consistent [13]. In the large Estonian develop hypotheses for intervention studies and poten- Biobank study, inverse associations between the concentra- tially to develop clinical risk scores. We thus aimed to tion of several amino acids (including BCAAs) and all- simultaneously investigate the association of circulating cause mortality were observed [14]. An inverse association amino acids with the following outcomes in people with between BCAAs and clinical dementia or Alzheimer’s type 2 diabetes: (1) macrovascular disease; (2) micro- vascular disease and (3) all-cause mortality. disease has also been observed [15]. Finally, we have Diabetologia (2018) 61:1581–1591 1583 Methods elsewhere, vitreous haemorrhage, pre-retinal haemor- rhage and fibrous proliferations on the disc or elsewhere Participants The Action in Diabetes and Vascular in a participant found not to have this condition at entry); Disease: Preterax and Diamicron Modified Release (5) development of macular oedema (characterised by a Controlled Evaluation (ADVANCE) study retinal thickening within one disc diameter of the macu- (ClinicalTrials.gov registration no. NCT00145925) lar centre in a participant not found to have this condition recruited 11,140 participants with type 2 diabetes at entry); between June 2001 and March 2003 [18]. Primary (6) occurrence of diabetes-related blindness (corrected visual outcomes of the trial have been published [19, 20]. acuity 3/60 or worse, persisting for ≥3 months and known Participants were ≥55 years of age and had been to not be due to non-diabetes-related causes in a partici- diagnosed with type 2 diabetes after the age of pant found not to have this condition at entry); 30 years. In addition, they were required to have a (7) use of retinal photocoagulation therapy. history of cardiovascular disease (CVD) or one or more additional cardiovascular risk factors. The trial included Blood samples were available from 17 out of 20 countries two randomised interventions: (1) a double-blind assess- participating in the ADVANCE study (the exceptions were ment of the efficacy of perindopril/indapamide (2 mg/0. China, India and the Philippines), giving a total potential 625mgfor 3months, increasing to4 mg/1.25 mg if source cohort size for the study of 7376 individuals (66.2% tolerated) vs placebo and (2) an open-label evaluation of of the overall study cohort). an intensive glucose-lowering regimen using modified- To improve efficiency of the biomarker studies in the release gliclazide (with a target HbA of ≤48 mmol/ ADVANCE trial a case-cohort study has been established 1c mol [6.5%]) vs standard care. Participants had their se- [21, 22]. In case-cohort studies, a random sample (called the rum creatinine levels measured as part of the study pro- ‘subcohort’) is drawn and phenotyped from the full cohort; tocol at baseline, 4 months and 1 year and annually this is very likely to contain both individuals who are ‘cases’ thereafter until completion of the study, with further and ‘non-cases’. Cases (generally for multiple case defini- tests at the discretion of clinicians. Urinary albumin/ tions, such as microvascular disease and macrovascular creatinine ratio (ACR) was measured as part of the disease) who were not included in the subcohort are then study protocol at baseline, 2 years, 4 years and comple- identified from the remainder of the cohort and were also tion of the study. GFR was estimated using the phenotyped. The case-cohort study has several advantages Modification of Diet in Renal Disease formula. over the nested case–control design, including the ability to Participants underwent formal eye examination and vi- investigate multiple endpoints simultaneously. For this case- cohort study, a random subcohort (n = 3500) was selected sual acuity testing at baseline, 2 years, 4 years and completion of the study. Each participating centre ob- from the base population, which was enriched by the addition tained ethical approval, and all participants provided of individuals who had a cardiovascular event, a microvascu- written informed consent. lar event or died during follow-up, giving a total study size of The primary trial outcomes were composites of major 4197 (Fig. 1)[21, 22]. macrovascular and microvascular events that occurred during a median of 5 years of follow-up. An independent adjudication Proton NMR analysis Plasma samples were obtained at baseline committee validated all outcomes. Major macrovascular events from all study participants when they were in an unfasted state, were cardiovascular death, non-fatal myocardial infarction or given that these were people with type 2 diabetes at risk of non-fatal stroke. Major microvascular events were defined as a hypoglycaemic episodes. Samples were collected across sites in composite of new or worsening nephropathy or retinopathy, in a pragmatic fashion (commensurate with a multinational RCT) turn defined as any of the following: according to local facilities. Plasma samples were separated and stored centrally at −80°C until measurement. The present study (1) development of macroalbuminuria (urinary ACR used a previously unthawed aliquot of plasma for H-NMR >33.9 mg/mmol, confirmed by two results); analysis. H-NMRspectroscopywasperformedonall available (2) doubling of serum creatinine level to ≥200 μmol/l (with EDTA plasma samples from the ADVANCE case-cohort study non-qualifying exceptions of terminal illness or acute at baseline using a low-volume (100 μl) variation of the quanti- illness and subsequent recovery of renal function); tative H-NMR method (Nightingale Health, Helsinki, Finland) (3) the need for renal replacement therapy due to kidney described previously [23, 24] and reviewed [25]. Sample spectra disease (in the absence of other medical causes requiring were analysed on a Bruker AVANCE III HD spectrometer to transient dialysis), or death due to renal disease; quantify a targeted list of metabolites, lipids and lipoproteins, (4) development of proliferative retinopathy (identified by as described previously [25]. This list included eight amino acids the incidence of new blood vessels on the disc or (alanine, glutamine, histidine, isoleucine, leucine, valine, 1584 Diabetologia (2018) 61:1581–1591 Fig. 1 Flow diagram for design 11,140 recruited in ADVANCE study and sample analysis in the 3764 not included in biobank ADVANCE study of amino acids (including all Chinese and (note microvascular, Indian participants) macrovascular and all-cause 7376 included in biobank mortality not mutually exclusive) 3500 randomly selected participants including Further selection of all other participants with 374 macrovascular events, 320 microvascular macrovascular events (396), microvascular events and 367 all-cause deaths during events (180) or all-cause mortality (410) during follow-up follow-up 4197 included in the case-cohort study 610 insufficient or unsuitable for NMR analysis for all metabolites Phenylalanine(3539 samples) Isoleucine (3587 samples) 648 macrovascular events 655 macrovascular events 337 microvascular events 342 microvascular events 624 all-cause deaths 632 all-cause deaths Alanine (3586 samples) Tyrosine (3579 samples) 655 macrovascular events 655 macrovascular events 342 microvascular events 341 microvascular events 632 all-cause deaths 632 all-cause deaths Glutamine (2228 samples) Leucine (3583 samples) 389 macrovascular events 654 macrovascular events 198 microvascular events 341 microvascular events 376 all-cause deaths 632 all-cause deaths Histidine (3566 samples) Valine (3587 samples) 653 macrovascular events 655 macrovascular events 339 microvascular events 342 microvascular events 631 all-cause deaths 632 all-cause deaths phenylalanine and tyrosine), which are detectable using the Cox regression models were fitted using the STSELPRE method, and are not in ‘congested’ regions of the NMR spectrum procedure for case-cohort analyses (StataCorp, College where multiple metabolites overlap. Metabolomic analyses of Station, TX, USA). Models estimated HRs for a 1 SD increase plasma samples tend to yield lower analyte concentrations than in each amino acid with each of the endpoints. Two models, serum, both by NMR spectroscopy and other methods, although with different potential confounding variables, were fitted for plasma demonstrates better stability and reproducibility [26]. each amino acid/outcome combination: model 1 with age, sex, Samples with a low glutamine/glutamate ratio were excluded region and randomised treatment; model 2 with, additionally, a from analyses of glutamine associations. Levels of all other ami- prior macrovascular complication of diabetes (myocardial in- no acids were consistent with published data. farction, stroke, hospital admission for a transient ischaemic attack or for unstable angina, coronary or peripheral Statistical analysis Continuous data with approximately nor- revascularisation, or amputation secondary to peripheral vas- mal distributions (including all amino acids) are presented as cular disease), duration of diabetes, current smoking, systolic mean ± SD; those with skewed distributions are presented as blood pressure, BMI, urinary ACR, eGFR, HbA , plasma 1c median (with interquartile range). Categorical data are pre- glucose, total cholesterol, HDL-cholesterol, triacylglycerols, sented as n (%). Pearson correlations were used to explore aspirin or other antiplatelet agent, statin or other lipid- associations of the amino acids with each other. Associations lowering agent, β-blocker, ACE inhibitor or angiotensin re- of amino acids with classical risk factors were investigated ceptor blocker, metformin use, history of heart failure, partici- across quarters of the distribution of each amino acid. pation in moderate and/or vigorous exercise for >15 min at Diabetologia (2018) 61:1581–1591 1585 Table 1 Pearson correlations (r) Amino acid Phenylalanine Isoleucine Glutamine Leucine Alanine Tyrosine Histidine of the amino acids with each other Isoleucine 0.32 – Glutamine −0.05 0.05 – Leucine 0.34 0.89 0.03 – Alanine 0.2 0.44 0.13 0.44 – Tyrosine 0.4 0.25 0.13 0.29 0.26 – Histidine 0.16 0.24 0.43 0.2 0.27 0.17 – Valine 0.27 0.67 0.13 0.75 0.34 0.29 0.25 All correlations have p values of <0.001, except for phenylalanine vs glutamine (p = 0.02), isoleucine vs gluta- mine (p = 0.03) and glutamine vs leucine (p =0.23) least once weekly, and high-sensitivity C-reactive protein sensitivity troponin T (hsTnT) and amino-terminal pro B- (CRP). A third adjustment model, attempting to include all type natriuretic peptide (NT-proBNP). In contrast, the other amino acids in the same model, resulted in collinearity and AAA, tyrosine, showed inverse associations with HbA and 1c estimates were thus not available. Non-linearity was tested ACR and a positive association with eGFR. Histidine, alanine by comparing the deviances of linear and categorical models and glutamine showed inconsistent associations with classical and by the inclusion of polynomial components (quadratic and risk factors. The BCAAs leucine, isoleucine and valine were cubic terms). Other analyses were performed using SAS v9.2 inversely associated with age, HDL-cholesterol and NT- (SAS Institute, Cary, NC, USA). All p values reported are two- proBNP but were positively associated with CVD, male sex, sided, with the 5% threshold used to determine significance. BMI, triacylglycerols and HbA . 1c For the random subcohort, the ability of amino acids to discriminate between those who will and those who will not Macrovascular disease, microvascular disease and all-cause go on to suffer each of the three adverse outcomes were esti- mortality Baseline risk factors associated with all three end- mated, in the context of model 2, using c statistics for 5 year points included male sex, increased duration of diabetes, his- risk, accounting for censoring. In addition, the ability of amino tory of macrovascular disease, higher systolic blood pressure, acids to reclassify participants according to 5 year risk, using lower HDL-cholesterol, higher HbA , higher ACR and 1c the continuous net reclassification index (NRI), was assessed higher hsTnT and NT-proBNP (Table 2). by methods suitable for survival data, using bootstrapping [27]. Among the amino acids, after adjustment for age, sex, re- Primary results came from use of all available data; sensi- gion and randomised treatment (model 1), higher phenylala- tivity analyses using only participants with complete data nine and lower glutamine and histidine concentrations were were also performed. associated with increased macrovascular risk (HR per 1 SD increase was 1.22 [95% CI 1.12, 1.32], 0.88 [95% CI 0.79, 0.98] and 0.86 [95% CI 0.79, 0.94], respectively) but these Results associations were attenuated to the null on further adjustment for classical risk factors (model 2) (Fig. 2aand ESMTable 10). Baseline associations A maximum of 3587 samples had avai- Higher tyrosine alone was associated with decreased risk of lable data for at least one amino acid (Fig. 1). Due to the design microvascular events (HR 0.74 [95% CI 0.64, 0.86]) in model of the multicentre study, there was some variability in sample 1 and this was only slightly attenuated on adjustment for a full processing time, leading to some samples having low range of classical risk factors in model 2 (HR 0.78 [95% CI glutamine/glutamate ratios. As such, fewer samples had a result 0.67, 0.91]) (Fig. 2b and ESM Table 10). A higher alanine for glutamine [28]. The detected absolute concentrations of level was also associated with decreased risk of microvascular amino acids were generally comparable with data from other events after further adjustment (HR 0.86 [95% CI 0.76, 0.98]). studies (see electronic supplementary material [ESM] Table 1). The association between tyrosine and renal impairment was In general, the amino acids showed a broad range of corre- further investigated by assessing the HRs across tertiles of lations with each other. Taking extreme examples, leucine and eGFR and ACR. There was no evidence of interaction by glutamine were not correlated (r =0.03, p = 0.23) but BCAAs eGFR or ACR (data not shown). leucine, isoleucine and valine were highly intercorrelated (r ≥ In contrast, several amino acids were associated with all- 0.67, p < 0.001) (Table 1). The associations of amino acids cause mortality. Phenylalanine was positively associated with with classical CVD risk factors are shown in ESM Tables 2– risk of mortality, while glutamine, leucine, alanine, histidine and 9. Phenylalanine was positively associated with older age, valine were all inversely associated with risk of mortality in baseline CVD, higher CRP and higher baseline high- model 1 (ESM Table 10). After adjustment for classical risk 1586 Diabetologia (2018) 61:1581–1591 Table 2 Baseline characteristics of the cohort classified by outcome status Characteristic Macrovascular disease Microvascular disease All-cause mortality Yes No p value Yes No p value Yes No p value N 655 2932 342 3245 632 2955 Age, years 68.92 ± 6.52 66.36 ± 6.51 <0.001 65.85 ± 6.38 66.93 ± 6.60 0.004 69.94 ± 6.56 66.17 ± 6.40 <0.001 Male sex 451 (68.9) 1719 (58.6) <0.001 227 (66.4) 1943 (59.9) 0.020 439 (69.5) 1731 (58.6) <0.001 Region ANZ/SEA 155 (23.7) 714 (24.4) 0.375 123 (36.0) 746 (23.0) <0.001 120 (19.0) 749 (25.3) <0.001 Canada 33 (5.0) 185 (6.3) 28 (8.2) 190 (5.9) 34 (5.4) 184 (6.2) Continental Europe 262 (40.0) 1157 (39.5) 91 (26.6) 1328 (40.9) 264 (41.8) 1155 (39.1) Northern Europe 205 (31.3) 876 (29.9) 100 (29.2) 981 (30.2) 214 (33.9) 867 (29.3) Duration of diabetes, years 9.19 ± 7.10 7.61 ± 6.29 <0.001 9.74 ± 6.89 7.71 ± 6.40 <0.001 9.24 ± 7.60 7.62 ± 6.17 <0.001 Current smoker 94 (14.4) 439 (15.0) 0.686 48 (14.0) 485 (14.9) 0.652 103 (16.3) 430 (14.6) 0.263 History of macrovascular disease 323 (49.3) 929 (31.7) <0.001 77 (22.5) 276 (8.5) <0.001 283 (44.8) 969 (32.8) <0.001 History of heart failure 56 (8.5) 110 (3.8) <0.001 13 (3.8) 153 (4.7) 0.445 61 (9.7) 105 (3.6) <0.001 Participation in moderate or vigorous 266 (40.6) 1470 (50.1) <0.001 164 (48.0) 1572 (48.4) 0.863 262 (41.5) 1474 (49.9) <0.001 activity Diastolic BP, mmHg 81.55 ± 11.41 81.63 ± 10.74 0.863 81.73 ± 11.30 81.60 ± 10.82 0.831 80.55 ± 11.70 81.84 ± 10.67 0.007 Total cholesterol, mmol/l 5.11 ± 1.18 5.15 ± 1.17 0.500 5.16 ± 1.08 5.14 ± 1.18 0.737 5.06 ± 1.10 5.16 ± 1.18 0.058 HDL-cholesterol, mmol/l 1.17 ± 0.31 1.24 ± 0.33 <0.001 1.18 ± 0.31 1.23 ± 0.33 0.005 1.18 ± 0.31 1.23 ± 0.33 <0.001 Triacylglycerol, mmol/l 1.63 (1.20, 2.30) 1.70 (1.20, 2.36) 0.436 1.80 (1.27, 2.60) 1.69 (1.20, 2.31) 0.01 1.61 (1.20, 2.30) 1.70 (1.20, 2.36) 0.4912 HbA , mmol/mol 59.5 ± 17.2 56.9 ± 14.8 61.3 ± 17.5 56.9 ± 15.4 59.3 ± 16.8 56.9 ± 15.2 1c HbA , % 7.59 ± 1.60 7.36 ± 1.39 <0.001 7.76 ± 1.60 7.36 ± 1.41 <0.001 7.58 ± 1.58 7.36 ± 1.40 <0.001 1c Glucose, mmol/l 8.61 ± 2.85 8.43 ± 2.68 0.115 9.04 ± 3.37 8.40 ± 2.62 <0.001 8.53 ± 2.88 8.45 ± 2.67 0.453 Urinary ACR, mg/mmol 2.4 (1.0, 8.0) 1.5 (0.7, 4.0) <0.001 5.6 (1.6, 14.2) 1.5 (0.7, 3.8) <0.001 2.4 (0.9, 7.6) 1.5 (0.7, 4.0) <0.001 −1 −2 eGFR, ml min 1.73 m 67.68 ± 17.61 72.70 ± 16.37 <0.001 69.97 ± 18.74 71.98 ± 16.48 0.034 66.58 ± 17.57 72.89 ± 16.32 <0.001 CRP, nmol/l 19.33 (9.05, 42.38) 16.67 (8.10, 38.48) 0.012 15.71 (8.48, 32.95) 17.43 (8.29, 39.33) 0.249 19.90 (9.71, 45.81) 16.86 (8.10, 37.71) <0.001 hsTnT, pg/ml 9 (4, 17) 5 (1.50, 10) <0.001 7 (1.50, 13) 5 (1.50, 11) <0.001 10 (4, 18) 5 (1.50, 9) <0.001 NT-proBNP, pg/ml 198 (74, 479) 74 (30, 169) <0.001 107 (38, 260) 87 (34, 212) 0.010 201 (80, 506) 74 (30, 171) <0.001 Medication use Aspirin or other antiplatelet agent 387 (59.1) 1380 (47.1) <0.001 171 (50.0) 1596 (49.2) 0.774 352 (55.7) 1415 (47.9) <0.001 Statin or other lipid-lowering agent 283 (43.2) 1311 (44.7) 0.484 158 (46.2) 1436 (44.3) 0.490 260 (41.1) 1334 (45.1) 0.066 β blocker 211 (32.2) 881 (30.0) 0.276 96 (28.1) 996 (30.7) 0.316 196 (31.0) 896 (30.3) 0.731 ACE inhibitor or angiotensin 418 (63.8) 1671 (57.0) 0.0014 232 (67.8) 1857 (57.2) <0.001 395 (62.5) 1694 (57.3) 0.017 receptor blocker Aminoacidlevel,mmol/l Phenylalanine 0.063 ± 0.010 0.061 ± 0.009 <0.001 0.062 ± 0.009 0.062 ± 0.009 0.1327 0.063 ± 0.010 0.061 ± 0.009 <0.001 Isoleucine 0.063 ± 0.017 0.062 ± 0.017 0.7429 0.065 ± 0.017 0.062 ± 0.017 0.0059 0.061 ± 0.017 0.063 ± 0.017 0.0643 Diabetologia (2018) 61:1581–1591 1587 Table 2 (continued) Characteristic Macrovascular disease Microvascular disease All-cause mortality Yes No p value Yes No p value Yes No p value Glutamine 0.373 ± 0.112 0.380 ± 0.109 0.2634 0.369 ± 0.118 0.380 ± 0.109 0.1696 0.367 ± 0.112 0.381 ± 0.109 0.0247 Leucine 0.081 ± 0.019 0.082 ± 0.020 0.0679 0.084 ± 0.020 0.082 ± 0.020 0.0518 0.079 ± 0.020 0.082 ± 0.020 <0.001 Alanine 0.366 ± 0.064 0.371 ± 0.065 0.0591 0.370 ± 0.066 0.370 ± 0.064 0.9372 0.361 ± 0.063 0.372 ± 0.065 <0.001 Tyrosine 0.053 ± 0.012 0.053 ± 0.011 0.9293 0.050 ± 0.012 0.053 ± 0.011 <0.001 0.052 ± 0.012 0.053 ± 0.011 0.1873 Histidine 0.049 ± 0.010 0.050 ± 0.009 0.0032 0.050 ± 0.009 0.050 ± 0.010 0.4117 0.048 ± 0.010 0.050 ± 0.009 <0.001 Valine 0.172 ± 0.035 0.175 ± 0.035 0.1355 0.177 ± 0.037 0.174 ± 0.035 0.1729 0.167 ± 0.036 0.176 ± 0.035 <0.001 Values are mean ±SD, median (interquartile range) or n (%) ANZ, Australia and New Zealand; SEA, south-east Asia 1588 Diabetologia (2018) 61:1581–1591 factors (model 2), the inverse association with risk remained for Phenylalanine leucine (HR 0.79 [95% CI 0.69, 0.90]), histidine (HR 0.89 Isoleucine [95% CI 0.81, 0.99]) and valine (HR 0.79 [95% CI 0.70, Glutamine 0.88]) but the positive association of phenylalanine was attenu- Leucine ated to the null (Fig. 2c and ESM Table 10). A sensitivity analysis using samples from individuals with complete data Alanine gave similar results (ESM Table 11). There was no evidence Tyrosine of a randomised treatment interaction in any model (data not Histidine shown). Valine A model including all the classical CVD risk factors in model 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 2 yielded a c statistic of 0.716 for macrovascular events, 0.728 HR (95% CI) for microvascular events and 0.747 for all-cause mortality (Table 3). Addition of the amino acids in combination did not Phenylalanine improve the c statistic for any endpoint but did improve the Isoleucine continuous NRI for macrovascular events (+35.5%, p<0.001) Glutamine and microvascular events (+14.4%, p = 0.012). The improve- Leucine ment in prediction of microvascular events was driven by the Alanine addition of tyrosine alone. Tyrosine Histidine Valine Discussion 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 HR (95% CI) Although previous observational studies have reported asso- ciations of circulating BCAAs and AAAs with adverse out- Phenylalanine comes in healthy people, the present report contrasts for the Isoleucine first time the associations of multiple circulating amino acids Glutamine with the major vascular complications of diabetes. Rather than Leucine one (or more) amino acids being a consistent signal for adverse Alanine outcomes of any kind, we report that their associations with Tyrosine risk of macrovascular events, microvascular events and all- Histidine cause mortality are strikingly different from each other. A key finding is the inverse association of tyrosine with risk of Valine microvascular events, independent of eGFR and urinary ACR. 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 Although the evidence from the present study suggests that HR (95% CI) these might only be very moderately useful biomarkers in in- Fig. 2 Adjusted associations (model 2, log scale HR) of amino acids individually (per 1 SD increase) with macrovascular outcomes (a), mi- cremental prediction of adverse events in individuals with type crovascular outcomes (b) and all-cause mortality (c) 2 diabetes, the pathophysiology underlying these associations Table 3 Prediction of endpoints using amino acids in combination or individually Model Macrovascular events Microvascular events All-cause mortality c statistic Continuous NRI (%) c statistic Continuous NRI (%) c statistic Continuous NRI (%) Basic model 0.716 – 0.728 – 0.747 – Plus all amino acids +0.010 +35.5 +0.010 +14.4 +0.009 +8.0 p value 0.17 <0.001 0.23 0.012 0.26 0.09 Plus tyrosine only –– 0.007 +14.9 –– p value –– 0.30 0.01 –– Adjusted for age, sex, region, randomised treatment, previous macrovascular event, duration of diabetes, current smoking, systolic blood pressure, BMI, urinary ACR, eGFR, HbA , plasma glucose, total and HDL-cholesterol, triacylglycerols, use of aspirin or other antiplatelet agent, statin or other 1c lipid-lowering agent, β-blocker, ACE inhibitor or angiotensin receptor blocker, or metformin, history of heart failure, participation in moderate and/or vigorous exercise for >15 min at least once weekly, and CRP Adjusted associations Adjusted associations Adjusted associations with all-cause with microvascular with macrovascular mortality outcomes outcomes Diabetologia (2018) 61:1581–1591 1589 and the possibility of intervention studies are intriguing and have reported that metformin in fact lowers, not raises, tyrosine worthy of further investigation. levels in individuals with CVD and at high risk of diabetes [16]. The association of high circulating concentrations of BCAAs The effect of other glucose-lowering drugs on amino acid profiles and AAAs with obesity has been known since the 1960s [29]and in individuals with diabetes would now be of interest. Further has been proposed to be at least partially mediated by insulin studies are now needed to validate our novel observations and to resistance. Insulin is thought to be a regulator of branched- examine whether our findings may represent causal pathways. chain α-keto acid dehydrogenase complex [30]. Insulin resis- Strengths of the study include the use of a well-characterised tance may hence suppress BCAA catabolism, as suggested by clinical trial cohort, an efficiently designed case-cohort study to associations noted in observational epidemiology studies yield a powerful study for a range of endpoints, which were [31–33]. The causal pathway may not be unidirectional; a recent independently adjudicated according to pre-defined criteria. Mendelian randomisation study suggests that genetically elevat- Like other RCT populations, ADVANCE study participants rep- ed BCAAs (via impaired catabolism) are associated with in- resent a selected cohort. For instance, ADVANCE study partic- creased risk of type 2 diabetes [34], although a better understand- ipants were required to have a history of CVD or CVD risk ing of the underlying pathway is required to increase confidence factors. Therefore, our results may not be generalisable to all in this observation [35]. There is also the possibility that amino individuals with diabetes, although other risk factors we mea- acids themselves (particularly BCAAs) may affect metabolism sured are generally associated with risk of major endpoints in by suppressing postprandial glucose levels [36]. Increased pro- the expected directions. Amino acids were measured in pragmat- tein turnover in people with central obesity may result in higher ically collected plasma samples in the context of a multinational circulating levels of amino acids [37] and might therefore cause RCT and we cannot rule out the potential for differential pre- elevations in amino acids in people who are overweight and have analytical sample handling or sample degradation during storage, type 2 diabetes. There are hence a variety of potential mecha- which may have biased our results [44], although these samples nisms related to type 2 diabetes pathologies that might influence were analysed at first thaw. We also present data suggesting circulating amino acids in individuals with type 2 diabetes. broadly comparable concentrations of amino acids relative to Given this background, and prior findings in general popula- other cohorts. Another potential limitation is the analysis of sam- tions of associations of specific amino acids with CVD [9–12], ples from non-fasted participants, although in clinical practice, we wished to examine whether amino acid levels associated with fasting is rarely required among individuals with type 2 diabetes. adverse outcomes in individuals with type 2 diabetes. In the NMR spectroscopy has been used to investigate changes in ami- ADVANCE study, the BCAAs leucine, valine and isoleucine no acids 30 min after a standardised liquid meal [45] and effects showed no association with macrovascular events, but low levels sizes were generally relatively small, although the immediate of leucine and valine were associated with increased all-cause postprandial state is likely to give larger effect estimates than are atplayinthisstudy. mortality. However, the positive, albeit not independently predic- tive, association of phenylalanine with CVD and all-cause mor- In conclusion, we report distinct associations of different ami- tality we observed is broadly in line with other published data. no acids with risk of major adverse endpoints in individuals with There are limited intervention studies investigating the effect of type 2 diabetes. Most notably, the identification of tyrosine as a amino acid supplements on health outcomes, with most research potential marker of microvascular risk requires further study. coming from short-term trials examining surrogate health markers in the sports science area [38]. Our data strongly support Acknowledgements We thank E. Butler, University of Glasgow, UK for the need for further studies to determine why higher phenylala- technical assistance in conducting the study. nine appears to be a consistently adverse signal for CVD out- comes. Our study provides observations that are the basis for Data availability Summaries of the ADVANCE trial data can be found at testable hypotheses investigating the effect of genetic variants, http://www.advance-trial.com. Restrictions apply to the availability of these data, which were used by agreement of the ADVANCE steering which are instrumental variables for circulating amino acids, on committee for the current study, and so are not publicly available. health outcomes [39, 40]. The inverse association of tyrosine with risk of microvascular Funding The biomarker work in the present study was funded by the events is perhaps the most intriguing individual finding from this Chest Heart and Stroke Association Scotland (R13/A149) and by the Glasgow Molecular Pathology NODE, which is funded by The Medical study. Tyrosine itself was positively associated with baseline Research Council and The Engineering and Physical Sciences Research eGFR and inversely associated with baseline HbA and urinary 1c Council (MR/N005813/1). The ADVANCE trial (ClinicalTrials.gov ACR. Impaired conversion of phenylalanine to tyrosine has been registration no. NCT00145925) was funded by the National Health and reported in renal disease [41, 42]. Low tyrosine levels might Medical Research Council (NHMRC) of Australia (project grant ID 211086 and program grant IDs 358395 and 571281) and by Servier. therefore simply reflect impaired kidney function, which itself PWu is supported by the Academy of Finland (312476 and 312477) predicts future microvascular events. Tyrosine is also linked to and the Novo Nordisk Foundation. MAK was supported by the Sigrid catecholamine synthesis, which, also speculatively, might be rel- Juselius Foundation, Finland. MAK works in a Unit that is supported by evant to our findings [43]. That noted, counter-intuitively, we 1590 Diabetologia (2018) 61:1581–1591 the University of Bristol and UK Medical Research Council (MC_UU_ risk of subsequent cardiovascular events. Circ Cardiovasc Genet 3: 12013/1). The study sponsors were not involved in the design of the 207–214 study, the collection, analysis, and interpretation of data, writing the report 11. Magnusson M, Lewis GD, Ericson U et al (2013) A diabetes- or the decision to submit the report for publication. predictive amino acid score and future cardiovascular disease. Eur Heart J 34:1982–1989 Duality of interest JC has received research grants from Servier as 12. Ruiz-Canela M, Toledo E, Clish CB et al (2016) Plasma branched- Principal investigator for ADVANCE and for the ADVANCE-ON post trial chain amino acids and incident cardiovascular disease in the follow-up study and honoraria from Servier for speaking about these studies PREDIMED trial. Clin Chem 62:582–592 at scientific meetings. MW reports receiving consulting fees from Amgen. 13. Floegel A, Kühn T, Sookthai D et al (2018) Serum metabolites and PWu is an employee and shareholder of Nightingale Health Ltd, which risk of myocardial infarction and ischemic stroke: a targeted conducted the biomarker quantification. All other authors declare that there metabolomic approach in two German prospective cohorts. Eur J is no duality of interest associated with their contribution to this paper. Epidemiol 33:55–66 14. Fischer K, Kettunen J, Würtz P et al (2014) Biomarker profiling by Contribution statement MM, NP, PH, MW and JC conceived, designed nuclear magnetic resonance spectroscopy for the prediction of all- and acquired the ADVANCE trial data. PWe, MW, NR, PM and NS con- cause mortality: an observational study of 17,345 persons. PLoS ceived this secondary study, and PWe and NS obtained grant funding. PWu Med 11:e1001606 and MA-K acquired biomarker data. MW and QL undertook the statistical 15. Tynkkynen J, Chouraki V, Van der Lee S et al (2018) Association of analyses. All authors were involved in data interpretation. PWe and NR branched-chain amino acids and other circulating metabolites with wrote the initial drafts of the manuscript. These drafts were revised for risk of incident dementia and Alzheimer s disease: a prospective important scientific content by MW, NS, PM, PWu, MA-K, QL, MM, study in eight cohorts. Alzheimers Dement. https://doi.org/10. NP, PH and JC. All authors gave final approval of the version to be pub- 1016/j.jalz.2018.01.003 lished. MW is the guarantor of this work. 16. Preiss D, Rankin N, Welsh P et al (2016) Effect of metformin therapy on circulating amino acids in a randomized trial: the Open Access This article is distributed under the terms of the Creative CAMERA study. Diabet Med 33:1569–1574 Commons Attribution 4.0 International License (http:// 17. Chen L, Cheng C, Choi H et al (2016) Plasma metabonomic pro- creativecommons.org/licenses/by/4.0/), which permits unrestricted use, filing of diabetic retinopathy. Diabetes 65:1099–1108 distribution, and reproduction in any medium, provided you give appro- 18. ADVANCE Management Committee (2001) Study rationale and priate credit to the original author(s) and the source, provide a link to the design of ADVANCE: Action in Diabetes and Vascular disease – Creative Commons license, and indicate if changes were made. Preterax and Diamicron MR Controlled Evaluation. Diabetologia 44:1118–1120 19. ADVANCE Collaborative Group, Patel A, MacMahon S et al (2008) Intensive blood glucose control and vascular outcomes in References patients with type 2 diabetes. N Engl J Med 358:2560–2572 20. Patel A, ADVANCE Collaborative Group, MacMahon S et al 1. Gaillard T, Osei K (2016) Ethnic differences in serum lipids and lipo- (2007) Effects of a fixed combination of perindopril and proteins in overweight/obese African-American and white American indapamide on macrovascular and microvascular outcomes in pa- women with pre-diabetes: significance of NMR-derived lipoprotein tients with type 2 diabetes mellitus (the ADVANCE trial): a particle concentrations and sizes. BMJ Open Diabetes Res Care 4: randomised controlled trial. Lancet 370:829–840 e000246 21. Hillis GS, Welsh P, Chalmers J et al (2014) The relative and com- 2. Liggi S, Griffin JL (2017) Metabolomics applied to diabetes-lessons bined ability of high-sensitivity cardiac troponin T and N-terminal from human population studies. Int J Biochem Cell Biol 93:136–147 pro-B-type natriuretic peptide to predict cardiovascular events and 3. Wang TJ, Larson MG, Vasan RS et al (2011) Metabolite profiles death in patients with type 2 diabetes. Diabetes Care 37:295–303 and the risk of developing diabetes. Nat Med 17:448–453 22. Welsh P, Woodward M, Hillis GS et al (2014) Do cardiac bio- 4. Floegel A, Stefan N, Yu Z, Mühlenbruch K et al (2013) markers NT-proBNP and hsTnT predict microvascular events in Identification of serum metabolites associated with risk of type patients with type 2 diabetes? Results from the ADVANCE trial. 2 diabetes using a targeted metabolomic approach. Diabetes Diabetes Care 37:2202–2210 62:639–648 23. Würtz P, Kangas AJ, Soininen P, Lawlor DA, Davey Smith G, Ala- 5. Tillin T, Hughes AD, Wang Q et al (2015) Diabetes risk and amino Korpela M (2017) Quantitative serum nuclear magnetic resonance acid profiles: cross-sectional and prospective analyses of ethnicity, metabolomics in large-scale epidemiology: a primer on -omic tech- amino acids and diabetes in a South Asian and European cohort nology. Am J Epidemiol 186:1084–1096 from the SABRE (Southall And Brent REvisited) Study. 24. Soininen P, Kangas AJ, Würtz P et al (2009) High-throughput se- Diabetologia 58:968–979 rum NMR metabonomics for cost-effective holistic studies on sys- 6. Würtz P, Soininen P, Kangas AJ et al (2013) Branched-chain and temic metabolism. Analyst 134:1781–1785 aromatic amino acids are predictors of insulin resistance in young 25. Soininen P, Kangas AJ, Würtz P, Suna T, Ala-Korpela M (2015) adults. Diabetes Care 36:648–655 Quantitative serum nuclear magnetic resonance metabolomics in 7. Stancakova A, Civelek M, Saleem NK et al (2012) Hyperglycemia cardiovascular epidemiology and genetics. Circ Cardiovasc Genet and a common variant of GCKR are associated with the levels of 8:192–206 eight amino acids in 9,369 Finnish Men. Diabetes 61:1895–1902 26. Yu Z, Kastenmüller G, He Y et al (2011) Differences between hu- 8. Guasch-Ferré M, Hruby A, Toledo E et al (2016) Metabolomics in man plasma and serum metabolite profiles. PLoS One 6:e21230 prediabetes and diabetes: a systematic review and meta-analysis. 27. Pencina MJ, D’Agostino RB, Steyerberg EW (2011) Extensions of Diabetes Care 39:833–846 9. Würtz P, Havulinna AS, Soininen P et al (2015) Metabolite profiling net reclassification improvement calculations to measure usefulness and cardiovascular event risk: a prospective study of 3 population- of new biomarkers. Stat Med 30:11–21 based cohorts. Circulation 131:774–785 28. Grossie VB, Yick J, Alpeter M, Welbourne TC, Ota DM (1993) 10. Shah SH, Bain JR, Muehlbauer MJ et al (2010) Association of a Glutamine stability in biological tissues evaluated by fluorometric peripheral blood metabolic profile with coronary artery disease and analysis. Clin Chem 39:1059–1063 Diabetologia (2018) 61:1581–1591 1591 29. Felig P, Marliss E, Cahill GF (1969) Plasma amino acid levels and 37. Tai ES, Tan MLS, Stevens RD et al (2010) Insulin resistance is associated with a metabolic profile of altered protein metabolism insulin secretion in obesity. N Engl J Med 281:811–816 in Chinese and Asian-Indian men. Diabetologia 53:757–767 30. Costeas PA, Chinsky JM (1996) Effects of insulin on the regulation 38. Xu Z, Tan Z, Zhang Q, Gui Q, Yang Y (2014) Clinical effectiveness of branched-chain α-ketoaciddehydrogenase E1α subunit gene of protein and amino acid supplementation on building muscle expression. Biochem J 318:85–92 mass in elderly people: a meta-analysis. PLoS One e10:9141 31. Shah SH, Crosslin DR, Haynes CS et al (2012) Branched-chain 39. Yu B, de Vries PS, Metcalf GA et al (2016) Whole genome se- amino acid levels are associated with improvement in insulin resis- quence analysis of serum amino acid levels. Genome Biol 17:237 tance with weight loss. Diabetologia 55:321–330 40. Teslovich TM, Kim DS, Yin X et al (2018) Identification of seven 32. McCormack SE, Shaham O, McCarthy MA et al (2013) Circulating novel loci associated with amino acid levels using single variant and branched-chain amino acid concentrations are associated with obe- gene-based tests in 8545 Finnish men from the METSIM study. sity and future insulin resistance in children and adolescents. Pediatr Hum Mol Genet. 27:1664–1674 Obes 8:52–61 41. Kopple JD (2007) Phenylalanine and tyrosine metabolism in chronic 33. Würtz P, Wang Q, Kangas AJ et al (2014) Metabolic signatures of kidney failure. J Nutr 137:1586S–1590S adiposity in young adults: Mendelian randomization analysis and 42. Druml W, Roth E, Lenz K, Lochs H, Kopsa H (1989) Phenylalanine effects of weight change. PLoS Med 11:e1001765 and tyrosine metabolism in renal failure: dipeptides as tyrosine 34. Lotta LA, Scott RA, Sharp SJ et al (2016) Genetic predisposition to source. Kidney Int Suppl 27:s282–s286 an impaired metabolism of the branched-chain amino acids and risk 43. Fernstrom JD, Fernstrom MH (2007) Tyrosine, phenylalanine, and of type 2 diabetes: a Mendelian randomisation analysis. PLoS Med catecholamine synthesis and function in the brain. J Nutr 137: 13:e1002179 1539S–1547S 35. Holmes MV, Ala-Korpela M, Smith GD (2017) Mendelian random- 44. Anton G, Wilson R, Yu Z et al (2015) Pre-analytical sample ization in cardiometabolic disease: challenges in evaluating causali- quality: metabolite ratios as an intrinsic marker for prolonged ty. Nat Rev Cardiol 14:577–590 room temperature exposure of serum samples. PLoS One 10: 36. Manders RJ, Little JP, Forbes SC, Candow DG (2012) e0121495 Insulinotropic and muscle protein synthetic effects of branched- 45. Schutte B, Van den Akker EB, Deelen J et al (2016) The effect of chain amino acids: potential therapy for type 2 diabetes and standardized food intake on the association between BMI and 1H- sarcopenia. Nutrients 4:1664–1678 NMR metabolites. Sci Rep 6:38980–38985 Affiliations 1 1 2 1 3,4 5,6,7,8,9 Paul Welsh & Naomi Rankin & Qiang Li & Patrick B. Mark & Peter Würtz & Mika Ala-Korpela & 10,11,12 13 14,15,16 2 2,17,18 1 Michel Marre & Neil Poulter & Pavel Hamet & John Chalmers & Mark Woodward & Naveed Sattar 1 10 BHF Glasgow Cardiovascular Research Centre, Institute of Inserm, UMRS 1138, Centre de Recherche des Cordeliers, Cardiovascular & Medical Sciences, University of Glasgow, 126 Paris, France University Place, Glasgow G12 8TA, UK Assistance Publique Hôpitaux de Paris, Bichat Hospital, DHU The George Institute for Global Health, University of New South FIRE, Department of Diabetology, Endocrinology and Wales, Sydney, NSW, Australia Nutrition, Paris, France 3 12 Research Programs Unit, Diabetes and Obesity, University of University Paris Diderot, Sorbonne Paris Cité, UFR de Helsinki, Helsinki, Finland Médecine, Paris, France 4 13 Nightingale Health Ltd, Helsinki, Finland International Centre for Circulatory Health, Imperial College, London, UK Computational Medicine, Faculty of Medicine, University of Oulu and Biocenter Oulu, Oulu, Finland Department of Experimental Medicine, McGill University, Montreal, QC, Canada NMR Metabolomics Laboratory, School of Pharmacy, University of Eastern Finland, Kuopio, Finland Department of Medicine, CRCHUM, Université de Montréal, Montreal, QC, Canada Population Health Science, Bristol Medical School, University of Bristol and Medical Research Council Integrative Epidemiology Department of Medicine, Gene Medicine Services, CRCHUM, Unit at the University of Bristol, Bristol, UK Université de Montréal, Montreal, QC, Canada 8 17 Systems Epidemiology, Baker Heart and Diabetes Institute, The George Institute for Global Health, University of Oxford, Melbourne, VIC, Australia Oxford, UK 9 18 Department of Epidemiology and Preventive Medicine, School Department of Epidemiology, Johns Hopkins University, of Public Health and Preventive Medicine, Faculty of Medicine, Baltimore, MD, USA Nursing and Health Sciences, The Alfred Hospital, Monash University, Melbourne, VIC, Australia

Journal

DiabetologiaSpringer Journals

Published: May 4, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off