Development and optimisation of a gas chromatography time-of-flight mass spectrometry method for the quantification of amino acids in infant formula

Abstract

Improved infant food protein testing methods have become mandatory for testing laboratories around the world to ensure food safety and to curb infant food adulteration such as the melamine adulteration incident that occurred in China, 2008. In this study a speed optimised flow rate (SOF) gas chromatography time-of-flight mass spectrometry (GC-TOFMS) method for quantifying 14 N-(tert-butyldimethylsilyl)-N-methyltrifluoroacetamide (MTBSTFA) derivatised amino acids (AAs) viz. alanine, glycine, valine, leucine, isoleucine, proline, serine, threonine, phenylalanine, aspartic acid, glutamic acid, lysine, histidine, and tyrosine in infant formula was developed. Using this method, 14 target compounds together with additional analytes, namely, cysteic acid, methionine sulfone, taurine, ornithine, and tryptophan, were resolved in 12.5 minutes. Using the GC-TOFMS method developed in this study, the above-mentioned analytes were quantified using two approaches, the external calibration approach, and the isotope dilution approach. An internal standard stock solution comprised of 13C valine, 13C isoleucine, 13C proline and 13C phenylalanine was used for the isotope dilution quantification method. Limits of detection (LODs) of between 0.0111 g/100g and 0.1064 g/100g were obtained by external calibration while LODs of between 0.01950 g/100g and 0.2456 g/100g were obtained by isotope dilution. Limits of quantification (LOQs) of between 0.0371 and 0.3548 g/100g were obtained by external calibration while LOQs of between 0.06510 and 0.8186 g/100g were obtained by isotope dilution. Linear regression correlation coefficients (r2) of between 0.9988 and 1.0000 were obtained from the calibration curves generated by external calibration while r2 values of between and 0.9959 and 0.9999 were obtained from the calibration curves generated using the isotope dilution approach. The GC-TOFMS (external calibration and the isotope dilution) methods developed in this study were validated using the National Institute of Standards and Technology (NIST) infant/adult nutritional formula standard reference material (SRM 1849-a) that had been hydrolysed with hydrochloric acid to obtain protein hydrolysates. On analysis of the NIST SRM (1849-a) protein hydrolysates, analyte recoveries (accuracy) of between 61.13% and 103.99% were obtained by external calibration while analyte recoveries of between 73.31% and 104.76% were obtained using the isotope dilution method. With the external calibration approach, coefficients of variation (precision) ranging from 7.32% to 25.76% were obtained while coefficients of variation of between 2.99% and 41.53% were obtained by isotope dilution. Method ruggedness was assessed by comparing the results obtained using the GC-TOFMS methods with the results obtained using the Waters Corporation’s AccQ·Tag method on an ultra-performance liquid chromatography (UPLC) system with ultraviolet (UV) detection. Method transferability was assessed by comparing the results obtained with a GC-TOFMS (Pegasus III) system with the results obtained on an alternate GC-TOFMS (Pegasus IV) system. Additionally, the results obtained from the GC-TOFMS method using the HCl hydrolysis method were compared with the results obtained from the same instrument using the trifluoroacetic acid (TFA) hydrolysis method. The main purpose of using an additional hydrolysis method (the TFA hydrolysis method) and applying two independent analytical techniques (UPLC and GC-TOFMS technique), was to develop and validate two independent analytical methods for value assigning the amino acid content of infant formula reference material to be produced by the National Metrology Institute of South Africa (NMISA). Using the Pegasus IV GC-TOFMS system, recoveries of between 50.85% and 101.62% were obtained through the isotope dilution method while recoveries ranging from 73.18% to 133.29% were obtained by external calibration. Additionally, using the Pegasus IV GCTOFMS method, coefficients of variation ranging from 0.36% to 7.39% were obtained through the isotope dilution method while coefficients of variation ranging from 1.45% to 12.69% were obtained through the external calibration method. Although there were differences between the recoveries and the coefficients of variation obtained using the Pegasus III and the Pegasus IV GC-TOFMS systems, using the student's t-test, significant differences between the results obtained by these methods were only found between the experimental means of proline, threonine, phenylalanine, and histidine. Therefore, based on the t-test results both the external calibration and the isotope dilution methods were readily transferable between the Pegasus III GC-TOFMS system and Pegasus IV GC-TOFMS system with significant differences only found between the abovementioned analytes. The Pegasus III GC-TOFMS results obtained by external calibration were comparable with the UPLC AccQ·Tag method results obtained by a similar calibration approach with significant differences found between alanine, leucine, isoleucine, proline, phenylalanine, and tyrosine. Most of the differences were observed between the results of the isotope dilution quantification method on the GC-TOMS system and the results of the internal standard method on the UPLC system. These include the experimental means of alanine, lysine, valine, leucine, isoleucine, proline, serine, histidine and tyrosine. Furthermore, the UPLC system yielded better precision compared to the GC-TOFMS methods. Using the UPLC method, coefficients of variation ranging from 5.30% to 13.15% were obtained by the internal standard method while coefficients of variation ranging from 3.86% to 20.21% were obtained by external calibration. Analyte recoveries ranging from 73.01% to 142.90% were obtained by the internal standard method while analyte recoveries of between 59.51% and 104.49% were obtained by external calibration. During method development, the guidelines provided in the Guide to Expression of Uncertainty in Measurement (GUM) were used to develop a cause and effect diagram which was subsequently used to identify experimental variables that may affect the accuracy and the uncertainty of measurements. Where possible, uncertainty contributions of the experimental variables identified through the cause and effect diagram were quantified mathematically using the GUM approach excluding the uncertainty contributions due to (1) the derivatisation temperature, (2) derivatisation period, (3) analyte reconstitution solvent type and (4) the stability of MTBSTFA derivatised amino acids. The uncertainty contributions due to the abovementioned variables could not be quantified mathematically due to complexity hence these variables were optimised experimentally to eliminate the need for their inclusion in the assessment of the uncertainty budget. For the optimisation process, a two-way or one-way ANOVA in conjunction with a Tukey honest significant difference (HSD) post hoc test were used to statistically assess the significance of the differences of the optimisation results. From the derivatisation time and derivatisation temperature results, it was found that all amino acids (AA) of interest were completely derivatised after incubation at 100 ˚C for 4 hours. Furthermore, acetonitrile was identified as a better reconstitution (injection) solvent for the analysis of MTBSTFA derivatised amino acids compared to isooctane. Additionally, MTBSTFA derivatised AAs showed varying stability under the storage conditions (ambient temperature and 3 ˚C) tested in this study. Alanine, glycine, valine, leucine, lysine and tyrosine derivatives were stable under both storage conditions. In contrast, isoleucine, phenylalanine, aspartic acid and glutamic acid were only stable at room temperature while proline, serine, and threonine derivatives were only stable at 3 ˚C. Analysis of MTBSTFA derivatised amino acids in infant formula by GC-TOFMS using both the external calibration and isotope dilution method gave results that were comparable to the results obtained through the routinely employed AccQ·Tag method as determined by (Bosch et al., 2006a). The advantages