1. Robust Measurement of Branched Chain Amino Acids on the Vantera Clinical Analyzer and the Clinical Association of NMR-Measured Valine with Type 2 Diabetes
Justyna Wolak-Dinsmore, Irina Shalaurova, Steven P. Matyus, Margery A. Connelly, and James D. Otvos
LipoScience Inc., Raleigh, NC
AACC 2014
Poster # A-149
Abstract
Results
Background: Metabolomic studies have shown that branched chain amino acid (BCAA) levels are independently associated with insulin resistance and type 2 diabetes (T2D).1-5 Nuclear magnetic resonance (NMR) technology has been employed for years to measure lipoprotein particle concentrations in a clinical laboratory setting. However, the information-rich nature of the NMR spectrum lends itself to the measurement of other clinically useful metabolites. Therefore, assays were developed to quantify the BCAAs, valine, leucine and isoleucine from the lipoprotein spectra.
Methods: Proton NMR spectra were collected on fasting serum samples using the Vantera Clinical Analyzer, a 400MHz NMR platform with automated fluidics sample handling, data processing and analysis. The NMR spectra were deconvoluted using proprietary software with models containing reference spectra from serum proteins and lipoproteins. For method comparison purposes, NMR-measured BCAAs were compared with concentrations determined by mass spectrometry at Mayo Medical Laboratories. Valine concentrations were quantified from NMR spectra previously obtained from participants at baseline in the Multi-Ethnic Study of Atherosclerosis (MESA) and multivariable logistic regression analyses were performed to interrogate the association of valine with the development of T2D.
Results: The levels of valine, leucine and isoleucine quantified in serum samples using NMR and mass spectrometry, were highly correlated (e.g. valine r=0.98). For the valine NMR assay, the coefficients of variation (CVs) for the inter-assay and intra-assay precision data were %. NMR-measured valine concentrations in MESA subjects (n=3309), who were non-diabetic and whose fasting plasma glucose was <110 mg/dL, were strongly associated with incident diabetes and made a statistically significant contribution to a logistic regression model containing age, gender, race and glucose
Conclusions: Levels of the BCAAs valine, leucine and isoleucine can be obtained from the same NMR spectra acquired for lipoprotein particle concentrations. Similar to published BCAA literature, NMR-measured valine is strongly associated with incident diabetes.
Figure 2: Method Comparison: BCAA Measured by Mass Spectrometry vs. NMR
Table 2: NMR-Measured Valine is Strongly Associated With Incident Diabetes
MESA
(n=352/3309)
(incident diabetes/total # subjects)
Wald Х2 P
Age
11.4
0.0007
Gender
20.8
<0.0001
Race
20.1
0.0002
Glucose
218.8
Valine
32.9
r = 0.98
r = 0.97
r = 0.95
Serum samples (n=25) were tested in singlicate on the reference (NMR Spectrometer) and comparator (Mass Spectrometer) instruments.
MESA subjects with fasting plasma glucose <110 mg/dL who were non-diabetic at baseline. 352 subjects developed diabetes during the mean 9.4 years of follow-up
Figure 3: Linearity: Expected vs. Measured Valine Concentrations
Table 1: Valine Precision Data
Valine
Low
Intermediate
High
Within-lab
Mean
156.5
283.1
473.7 SD
6.4
9.3
11.4 CV
4.1
3.3
2.4
Repeatability
5.9
8.4
7.3
3.8
3.0
1.5
Within-run
163.8
288.6
481.4
7.0
6.6
8.9
4.3
2.3
1.9
Conclusions
Levels of the BCAAs valine, isoleucine and leucine can be obtained from the same NMR spectra acquired for lipoprotein particle quantification.
Serum levels of valine, isoleucine and leucine quantified by NMR and mass spectroscopy were highly correlated.
Linearity and precision data reveal the NMR valine assay to be robust and suitable for clinical testing.
Similar to published BCAA literature, NMR-measured valine is strongly associated with incident diabetes.
Figure 1: NMR Signals for the Branched Chain Amino Acids Arise Downfield of the Lipoprotein MethylSignals
R2 = 0.99
References
1Newgard, C.B., Interplay between lipids and branched-chain amino acids in development of insulin resistance. (2012) Cell Metabolism, 15:
2Wang, T.J. et. al., Metabolite profiles and the risk of developing diabetes (2011) Nature Medicine, 17:
3Suhre, K., et. al., Metabolic Footprint of Diabetes: A Multiplatform Metabolomics Study in an Epidemiological Setting. (2010) PLoS ONE, 5(11):E13953
4Wurtz, P., et. al., Metabolic Signatures of Insulin Resistance in 7,098 Young Adults. (2012) Diabetes, 61(6):
5Wurtz, P., et. al., Branched-Chain and Aromatic Amino Acids are Predictors of Insulin Resistance in Adults. (2013) Diabetes Care, 36:
Assay linearity was evaluated based on CLSI guidelines by comparing linear regression to higher order polynomial regression for results from a series of mixtures of high and low serum pools. Serum pools with elevated valine were obtained by enriching serum samples with valine. Samples with intermediate levels were obtained by mixing serum pools with elevated and low levels of the valine in varying proportions. Four measurements were made for each mixture.
Within-laboratory imprecision and repeatability were determined based on CLSI guidelines using serum pools targeted at low, intermediate and high ranges. Testing consisted of duplicate tests run twice per day for 20 days (n=80) on one Vantera, while within-run imprecision was calculated from 20 replicates.
For the NMR isoleucine and leucine assays, the CVs for the intra-assay (within-run) imprecision data were 12.1, 18.5, 23.0% and 4.1, 7.8, 13.4% for the high, intermediate and low serum pools, respectively. Due to the fact that the valine NMR signal was higher and the assay precision was better for valine than that for isoleucine and leucine, the valine values were interrogated for their association with incident diabetes in the MESA population.