Analysis of 25 underivatized amino acids in human plasma using ion-pairing reversed-phase liquid chromatography/time-of-flight mass spectrometry

Transcription

1 RAPID COMMUNICATIONS IN MASS SPECTROMETRY Rapid Commun. Mass Spectrom. 2007; 21: Published online in Wiley InterScience ( Analysis of 25 underivatized amino acids in human plasma using ion-pairing reversed-phase liquid chromatography/time-of-flight mass spectrometry Michael Armstrong 1, Karen Jonscher 2 and Nichole A. Reisdorph 1 * 1 Department of Immunology, National Jewish Medical and Research Center, Denver, CO 80206, USA 2 University of Colorado at Denver and Health Science Center, Clinical Nutrition Research Unit, Department of Anesthesiology, Denver, CO 80262, USA Received 1 December 2006; Revised 25 May 2007; Accepted 26 May 2007 Amino acids in biological fluids have previously been shown to be detectable using liquid chromatography/electrospray ionization mass spectrometry (LC/ESI-MS) with perfluorinated acids as ion-pairing agents. To date, these studies have used precursor mass, retention time and tandem mass spectrometry (MS/MS) to identify and quantify amino acids. While this is a potentially powerful technique, we sought to adapt the method to time-of-flight (TOF)MS. A new application of a recently described liquid chromatographic separation method was coupled with TOFMS to employ accurate mass for qualitative identification; resulting in additional qualitative data not available with standard single quadrupole data. In the current study, we evaluated 25 physiological amino acids and one dipeptide that are routinely quantified in human plasma. Accuracy and precision of the method was evaluated by spiking human plasma with a mix of the 25 amino acids; in addition, the inclusion of a cation-exchange cleanup step was evaluated. The calibration curves were linear over a range from 1.56 to 400 mm. The dynamic range was found to be within physiological levels for all amino acids analyzed. Accuracy and precision for most of the amino acids was between % spike recovery and <10% relative standard deviation (RSD). The LC/MS technique described in this study relies on mass accuracy and is suitable for the quantitation of free amino acids in plasma. Copyright # 2007 John Wiley & Sons, Ltd. *Correspondence to: N. A. Reisdorph, Department of Immunology, National Jewish Medical and Research Center, 1400 Jackson Street K924, Denver, CO 80206, USA. ReisdorphN@njc.org Contract/grant sponsor: Colorado Clinical Nutrition Research Unit; contract/grant number: NIH/NIDDK P30 DK Free amino acid analysis has applications in a variety of areas, including the diagnosis of inherited metabolic disorders, 1 4 and nutritional studies of neonates. 5 8 Traditionally, free amino acids in plasma have been analyzed by ion chromatography (IC) using ninhydrin post-column derivatization, 9 or by cation-exchange solid-phase extraction followed by derivatization and analysis by gas chromatography/mass spectrometry (GC/MS) Both of these methods have disadvantages, including long run times and extensive sample preparation, respectively. Although useful for a broad range of compounds, neither high-performance liquid chromatography (HPLC) nor MS techniques were generally employed for amino acid analysis due to the inability to separate more polar amino acids using reversed-phase (RP)-HPLC, 16 and amino acid signal suppression in electrospray ionization (ESI)-MS caused by co-elution of components in complex biological matrices. 17,18 Some approaches to enable the application of LC/MS to amino acid analysis include dedicated amino acid analysis kits such as Waters AccQtag, 19 and tandem mass spectrometry (MS/MS) utilizing flow injection analysis, 4 each of which has certain disadvantages. For example, buffers included in kits may be incompatible with ESI. Recently published methods have described the use of perfluorinated acids as ion-pairing agents to improve the separation of amino acids on C18 columns without the requirement for specialty columns or pre-/post-column derivatization. 16,20 23 Piraud et al. 23 utilized HPLC and tandem mass spectrometry (LC/MS/MS) with tridecafluoroheptanoic acid (TDFHA) as an ion-pairing agent with a C18 column. While TDFHA improved separation, all amino acids of interest were not completely resolved by HPLC. Therefore, multiple reaction monitoring (MRM) was used to improve the specificity of the method by monitoring specific transitions of precursor to product ions (e.g. glutamine ¼147>84). The method of Piraud et al. was used to quantitate 76 amino acids of biological interest and the quantitation of 16 amino acids was validated. 23 In the current study, we have adapted Piraud s method of ion-pairing reversed-phase liquid chromatography (IPRP- LC) for use with an electrospray ionization time-of-flight (ESI-TOF) mass spectrometer equipped with an analogto-digital converter (ADC) for signal processing. Sample Copyright # 2007 John Wiley & Sons, Ltd.

2 2718 M. Armstrong, K. Jonscher and N. A. Reisdorph preparation involves precipitation of proteins from plasma using methanol fortified with stable isotope labeled internal standards followed, in some cases, by cation exchange. Extremely accurate mass measurements (approximately 1 10 ppm) obtained with a TOF were used to deduce the identity of an analyte with a much higher degree of certainty than a standard, high-resolution single quadrupole mass spectrometer. With the exception of isobaric molecules such as leucine and isoleucine, the mass-to-charge (m/z) of amino acids can be verified to within 1 5 ppm, reducing misidentification of target amino acids in the presence of co-eluting matrix components of similar molecular weight. We found that IPRP-LC/ESI-TOF provides a quick, simple, reproducible alternative to MS/MS analysis and used this technique for the analysis of 25 amino acids in human plasma. The method described in this study uses minimal amounts of standards, reagents, and sample, can be applied to any amino acid that ionizes by ESI, and can easily be adapted to high-throughput sample analysis. EXPERIMENTAL Reagents Nanopure water (18.2 VOhms) was used for sample preparation. Water (HPLC grade) and acetonitrile (UV) used for HPLC mobile phases was obtained from Burdick and Jackson (Morristown, NJ, USA). HPLC-grade methanol was obtained from Fisher Scientific (Hampton, NH, USA). Tridecafluoroheptanoic acid (TDFHA) was obtained from Aldrich Chemicals (St. Louis, MO, USA). Hydrochloric acid was obtained from Sigma (St. Louis, MO, USA). The primary amino acid calibration standard at 2.5 mm (standard H ) was obtained from Pierce (Rockford, IL, USA). Hyp, Gly, Glu, Ala, Trp, Tau, Asn, Gln, Cit, Ala-Glu, Nor and Orn were obtained from Sigma. Stable isotope labeled analogs of amino acids used as internal standards (glutamine-d5, glutamic acid-d3, methionine-d3, leucine-d10 and tryptophan-d5) were obtained from Cambridge Isotope Laboratories (Andover, MA, USA). Outdated human blood plasma was provided by Bonfil s Blood Center (Denver, CO, USA). The use of outdated plasma samples for method validation and quality control purposes was considered exempt by the Colorado Multiple Institutional Review Board (COMIRB). Standards preparation procedure Amino acid calibration and spike standards were prepared at physiological concentration ranges from pure powder or commercially available standards. Amino acid mix #1 contained amino acids which are stable in 0.1% hydrochloric acid solution such as the branched chain amino acids and the hydroxyl group containing amino acids, including Asp, Hyp, Ser, Gly, Thr, Glu, Ala, (Cys) 2, Pro, Cys, Val, Met, Tyr, Ile, Leu, Phe, His, Trp, Arg, and Lys. Amino acid mix #2 contained the amino acids which are not stable in an acid solution, such as Tau, Gln, Asn, Cit, Ala-Gln and Orn. All calibration stocks and working standards were stored at 208C until use. In addition to calibration standards, two separate internal standard (IS) mixes were used to quantitate. IS mix #1 contained glutamine-d5 and methionine-d3 in water. IS mix #2 contained leucine-d10, glutamic acid-d3 and tryptophan-d5 in 0.1% hydrochloric acid. The internal standard working solution was prepared immediately prior to sample preparation, by adding equal parts of IS #1 and IS #2 to eight parts methanol (1:1:8). All internal standard stocks and IS mix #1 and #2 were stored at 208C until use. Pooled human plasma samples, used for accuracy and precision measurements, were spiked with amino acid calibration mixes 1 and 2 (100 mm) and frozen at 808C for 1 2 days prior to thawing for extraction and analysis. Sample preparation procedure Calibration standards were prepared by combining 10 ml each of amino acid mixes 1 and 2 and 100 ml of IS working solution. Standards were vortexed briefly and then centrifuged at g for 5 min at 48C. An aliquot (70 ml) of supernatant was transferred to a 96-well plate or HPLC vial containing 30 ml of 1.7 mm TDFHA in water, providing a final concentration of 0.5 mm TDFHA. Samples were prepared by adding 20 ml of plasma to 100 ml of IS working solution and briefly vortexing. Samples were then centrifuged at g for 5 min at 48C, resulting in a protein precipitate that was subsequently discarded. An aliquot (70 ml) of supernatant was transferred to a 96-well plate or HPLC vial containing 30 ml of 1.7 mm TDFHA in water. Solid-phase extraction For some samples, cleanup was performed via solid-phase extraction (SPE) using a cation-exchange cartridge. Strata X-C cartridges with a capacity of 30 mg (Phenomenex, Torrance, CA, USA) were placed on a vacuum SPE manifold, conditioned with 1 ml of methanol, then equilibrated with 1 ml of 0.1 N HCl in water, as per the manufacturer s protocol. Subsequently, 100 ml of plasma was mixed by vortexing with 100 ml of the IS working solution prepared in 0.2 M HCl. The entire sample was then loaded onto the SPE cartridge and drawn through by vacuum. Afterwards, the cartridge was washed with 1 ml of methanol, and sample was eluted into a new test tube using 5% ammonium hydroxide in methanol. The eluate ph was neutralized by vacuum evaporation of the ammonium hydroxide. Samples were then lyophilized to dryness and reconstituted with 100 ml of 50 mm TDFHA in 1:1 methanol/water prior to analysis. The final volume results in a 5-fold increase in sample over the samples not extracted by SPE. High-performance liquid chromatography Liquid chromatography was carried out using an Agilent 1100 series HPLC system equipped with a binary pump and a micro wellplate autosampler (Agilent Technologies, Palo Alto, CA, USA). Amino acids were separated using an XDB-C18 column ( mm) with a 1.8 mm particle size (Agilent Technologies) operated at ambient temperature. Buffer A was 0.5 mm TDFHA in HPLC-grade water, and buffer B was 100% acetonitrile. The initial flow rate was 0.2 ml/min. Separation was accomplished using a gradient as follows: 0% B for 2 min, then 0% to 15% B from 2 to 3 min, hold at 15% B from 3 to 8 min, then 15% to 25% B from 8 to 11 min. The column was held at 25% B from 11 to 18 min, and

3 Analysis of 25 underivatized amino acids in human plasma 2719 then returned to 0% B from 18 to 19 minutes. The flow rate was then increased to 0.4 ml/min from to 29 min to recondition the column. The flow rate was then returned to 0.2 ml/min at min, and allowed to equilibrate for 3 min. A long reconditioning time and re-equilibration time is required to obtain consistent retention times with this column. The column was flushed with 100% acetonitrile for 1 h after every 30 injections to wash off accumulated non-target analytes. Failing to flush the column with acetonitrile after 30 injections will result in a degradation in chromatography and retention time drift. ESI-TOFMS Detection of amino acids was accomplished using an Agilent 1969 orthogonal TOF mass spectrometer coupled to a positive ESI source with dual spray needles for continuous infusion of reference mass solution. Heated (3508C) drying gas flowing at 9.0 L/min, with a nebulizer pressure of 40 psig, was used for droplet desolvation. Spray was induced with a capillary voltage of 3000 V and the fragmentor voltage was 100 V. The TOF was tuned and calibrated using Agilent ESI-TOF calibration and tuning mix (Agilent Technologies). The data acquisition mass range was m/z at transients/scan and 0.93 scans/s. Reference mass correction on each sample was performed with a continuous infusion of Agilent TOF biopolymer analysis mix containing purine (m/z ) and hexamethoxyphosphazine (m/z ) (Agilent Technologies) at 20 ml/min. Ions monitored for quantitation Ions monitored for quantitation (see Table 1) were extracted using Analyst QS software (Applied Biosystems, Foster City, CA). Signals from internal standards were extracted with a window ranging from 0.05 to 0.15 Da, while target amino acids were provided a 0.02 Da extraction window. Calibration curves Calibration curves for each amino acid were constructed using Analyst QS software and prepared so all amino acids would be within expected physiological concentrations. Most amino acids were calibrated from 1.56 to 400 mm. The more abundant amino acids (Gln, Glu, Gly and Ala) were calibrated from 25 to 3200 mm (see Table 2). Analyst QS was used to choose the best fit for the calibration curve. Either a quadratic or linear fit was applied to quantify most amino acids. Method accuracy and precision To test method accuracy and precision, pooled human plasma was analyzed unspiked and spiked at 100 mm (nominal) of each amino acid. Intra-day accuracy and precision (n ¼ 5) and inter-day accuracy and precision (n ¼ 3) were calculated. Table 1. List of experimental parameters for amino acids and signal-to-noise (S/N) ratios obtained from a 125pm (nominal) injection Compound Molecular formula Exact mass [MþH] Extracted ion window Low cal std (nm/ml) High cal std (nm/ml) S/N ratio (pm injected) IS used for quantitation Taurine C2H7NO3S (125) Glutamine-d5 Aspartic acid C4H7NO (125) Glutamic acid-d3 Hydroxyproline C5H9NO (125) Glutamine-d5 Serine C3H7NO (125) Glutamic acid-d3 Glycine C2H5NO (3125) Glutamic acid-d3 Glutamine-d5 C5H5D5N2O NA NA 637 (1000) NA Glutamine C5H10N2O (3125) Glutamine-d5 Asparagine C4H8N2O (125) Glutamine-d5 Threonine C4H9NO (125) Glutamic acid-d3 Glutamic acid-d3 C5H6D3NO NA NA 15.2 (1000) NA Glutamic acid C5H9NO (1562) Glutamic acid-d3 Alanine C3H7N (125) Leucine-d10 (Cysteine)2 C6H12N2O4S (125) Methionine-d3 Citrulline C6H13N3O (125) Glutamine-d5 Proline C5H9NO (125) Glutamine-d5 Gly-Gln C7H13N3O NA NA 2560 (1000) NA Ala-Gln C8H15N3O (125) Gly-Gln Valine C5H11NO (125) Leucine-d10 Methionine-d3 C5H8D3NO2S NA NA 1810 (1000) NA Methionine C5H11NO2S (125) Methionine-d3 Tyrosine C9H11NO (125) Leucine-d10 Isoleucine C6H13NO (125) Leucine-d10 Leucine-d10 C6H3D10NO NA NA 1450 (1000) NA Leucine C6H13NO (125) Leucine-d10 Phenylalanine C9H11NO (125) Leucine-d10 Histidine C6H9N3O (125) Tryptophan-d5 Tryptophan C11H12N2O (125) Tryptophan-d5 Tryptophan-d5 C11H7D5N2O NA NA 1110 (1000) NA Arginine C6H14N4O (125) Tryptophan-d5 Ornithine C5H12N2O (125) Tryptophan-d5 Lysine C6H14N2O (125) Tryptophan-d5 Internal standard.

4 2720 M. Armstrong, K. Jonscher and N. A. Reisdorph Table 2. Amino acids and internal standards used for quantitation, including the curve fit used and the correlation coefficient obtained Without cation exchange With cation exchange Amino acid IS Curve fit R 2 IS Curve fit R 2 Taurine (TAU) Glutamine-d5 Quadratic 1 Glutamine-d5 Linear Aspartic acid (ASP) Glutamic acid-d3 Quadratic Glutamine-d5 Quadratic Hydroxyproline (HYP) Glutamine-d5 Quadratic 1 Glutamine-d5 Linear Serine (SER) Glutamic acid-d3 Linear Glutamine-d5 Quadratic Glycine (GLY) Glutamic acid-d3 Quadratic Glutamine-d5 Linear Glutamine (GLN) Glutamine-d5 Quadratic 1 Glutamine-d5 Linear Asparagine (ASN) Glutamine-d5 Quadratic 1 Glutamine-d5 Linear Threonine (THR) Glutamic acid-d3 Quadratic Glutamine-d5 Quadratic Glutamic acid (GLU) Glutamic acid-d3 Quadratic Glutamate-d3 Linear Alanine (ALA) Leucine-d10 Quadratic Leucine-d10 Linear (Cysteine)2 Methionine-d3 Quadratic Glutamine-d5 Linear Citrulline (CIT) Glutamine-d5 Quadratic 1 Glutamine-d5 Linear Proline (PRO) Glutamine-d5 Quadratic Glutamine-d5 Linear Valine (VAL) Leucine-d10 Quadratic Leucine-d10 Linear Methionine (MET) Methionine-d3 Linear Methionine-d3 Linear Tyrosine (TYR) Leucine-d10 Linear Leucine-d10 Linear Isoleucine (ISO) Leucine-d10 Linear Leucine-d10 Linear Leucine (LEU) Leucine-d10 Quadratic 1 Leucine-d10 Linear Phenylalanine (PHE) Leucine-d10 Quadratic Leucine-d10 Linear Histidine (HIS) Leucine-d10 Linear Tryptophan-d5 Linear Tryptophan (TRP) Tryptophan-d5 Quadratic 1 Tryptophan-d5 Linear Arginine (ARG) Leucine-d10 Quadratic Tryptophan-d5 Linear Ornithine (ORN) Leucine-d10 Quadratic Tryptophan-d5 Linear Lysine (LYS) Leucine-d10 Quadratic Tryptophan-d5 Linear Intra-day accuracy and precision (n ¼ 5) was also measured on samples that were prepared using SPE prior to analysis. RESULTS Chromatographic separation of amino acids A brief comparison was conducted using Agilent XDB-C mm columns with either 1.8 or 3.5 mm solid-phase particle sizes. Under identical gradient conditions the 1.8 mm column showed superior resolution of the early eluting polar amino acids (data not shown). An attempt was made to improve the separation efficiency of the 3.5 mm column using different gradient profiles and flow rates; however, the 1.8 mm column still appeared to provide the best separation (data not shown). Although the ESI-TOF provides excellent specificity via its high mass accuracy, there are still instances where complete or partial chromatographic resolution must be obtained for accurate quantitation. One example is when an isotopomer, or m þ n (where n ¼ the number of Daltons the ion is shifted from the m þ 0 ion), in the mass spectrum of a compound adds to the m þ 0 area of another compound (e.g. Gln and Glu) as a result of co-elution. Another is the differentiation of isobaric compounds such as Leu and Ile. The chromatographic resolution obtained using the 1.8 mm column was >90%, separating Gln from Glu, and allowing for complete resolution of Ile and Leu isobars. When the final gradient was optimized and established, all of the amino acids eluted within 16.5 min (Fig. 1). A relatively long column re-equilibration time resulted in a total cycle time of 32 min. While this method is an improvement in throughput and specificity over traditional amino acid analysis methods, the throughput could be almost doubled by using a quaternary or additional HPLC pump, a column-switching module, and an additional column to alternate column regeneration and sample analysis. Integration reproducibility and signal-to-noise ratio The ability of the ESI-TOF to maintain consistent mass accuracy and peak integration over time was assessed. The extracted ion chromatograms for glutamine in five replicate spiked plasma samples were integrated and compared (Fig. 2). The mass window for each replicate sample was m/z The relative standard deviation (RSD) over the five replicates was 7.29, showing good sampleto-sample integration reproducibility. Signal-to-noise (S/N) ratios were also calculated for all amino acids (See Table 1). Most S/N ratios were greater than 200:1, with Gly being the lowest at 22:1 and Hyp being the highest at 1750:1. Matrix interference in plasma samples There were significant differences in retention times for amino acids from extracted standards versus plasma samples, particularly for the later eluting compounds. Retention times for amino acids eluting after 4 min were shifted to as much as 1.5 min earlier in plasma (e.g. Ile and Leu). The retention time shift in plasma samples did not typically result in a decreased chromatographic resolution except for the peak shape of Orn, which was significantly

5 Analysis of 25 underivatized amino acids in human plasma 2721 Figure 1. Extracted ion chromatograms for 25 amino acids. Amino acid calibration and spike standards were prepared as described in the Experimental section and analyzed according to the parameters listed in Table 1. Overlaid extracted ion chromatograms of all 25 amino acids in a 400 nm/ml (nominal concentration) standard are shown. Note the separation between the isobaric amino acids leucine and isoleucine. All peaks are displayed to scale. broadened. The peak shape for Orn was significantly worse in plasma samples (data not shown). In order to determine if this retention time shift was due to column overloading, 10, 5 and 2 ml of a spiked plasma sample were loaded onto the column, and the retention time for leucine-d10 was compared to that of the calibration standard. The data show that a reduction in the amount of sample loaded decreased the shift in retention times, suggesting that the column is indeed overloaded. However, decreasing the sample load significantly raised the lower limit of quantitation for several compounds. Ion suppression from TDFHA adducts While examining data acquired in a wider mass range to determine the major source of column overload, a very large peak at a retention time of 8 9 min was observed that could not be detected when using the normal acquisition parameters. This peak was almost non-existent in calibration standards, but appeared at an extremely high abundance in plasma samples. Using the accurate mass obtained from the ESI-TOF data, the empirical formula for the most abundant ion in the spectrum was calculated to be C 7 O 2 F 13 Na 2. This empirical formula corresponds to a sodium adduct of tridecafluoroheptanoate, a product of the binding of sodium salt in plasma with the ion-pairing agent in the aqueous buffer. The exact reference corrected mass of this ion was m/z ¼ , with a theoretical mass of m/z ¼ (see Fig. 3(A)). The less abundant ions in the spectrum corresponded to clusters of this compound with additional C 7 O 2 F 13 Na (m/z ¼ ) subunits. Alanyl-glutamine (Ala-Gln) dipeptide and Val co-eluted with this peak, which significantly suppressed the signal of Figure 2. Glutamine integration reproduciblity. Overlay of glutamine extracted ion chromatograms (m/z ) of five replicates of spiked plasma, showing sample-to-sample integration reproducibility.

6 2722 M. Armstrong, K. Jonscher and N. A. Reisdorph Figure 3. A tridecafluoroheptanoate (TDFHA) adduct elutes between 8 and 9 min and interferes with the analysis of Ala-Gln and Val. Plasma samples were spiked with internal standards and analyzed as described in the Experimental section. The mass spectrum of a large peak eluting between 8 and 9 min (A) and extracted ion chromatograms of TDFHA adduct, Ala-Gln and Val (B) are shown. The exact mass of m/z corresponds to the molecular formula of disodium tridecafluoroheptanoate, an adduct formed between the TDHFA ion-pairing agent and sodium in plasma. The Ala-Glu dipeptide and Val are almost completely obscured by the adduct (B). both analytes (Fig. 3(B)). In an effort to separate Ala-Gln and Val from this peak, we experimented with (a) increasing the hold at 15% B from 5 to 8 min, (b) changing the hold from 15% B to 12 % B, and (c) changing the hold from 12% B to 10% B. None of these changes improved the separation of Ala-Gln and Val from the TDHFA adduct peak enough to reduce ion suppression. SPE cleanup To eliminate the TDFHA adduct and additional non-target compounds from the sample extract, a cation-exchange cleanup step was performed on a series of calibration standards, unspiked plasma, and spiked plasma. Although slightly more time-consuming, the improvement in chromatographic performance provided by the cation-exchange cleanup was significant. Retention time shift between the calibration standards and plasma samples was virtually eliminated. The abundance of the TDFHA adduct was also decreased significantly enough to dramatically improve both the accuracy and the precision of Val in plasma and spiked plasma. The accuracy and precision of Ala-Gln was not improved significantly after cation-exchange cleanup (data not shown). Calibration linearity Calibration linearity was compared between the samples analyzed with and without cation-exchange cleanup. Overall, much better results were obtained with cation-exchange cleanup. In the calibration without cation-exchange cleanup, a quadratic regression was selected as the preferred fit for most of the amino acids whilst, with cation exchange, a linear regression fit was determined to be optimal. This improvement in performance could also be due to a 5-fold increase in sample amount used for the preparation with cationexchange cleanup. In the cation-exchange cleanup, glutamine-d5 was used with much better quantitative results as an internal standard for several of the early eluting amino acids when compared to the samples that had not been cleaned up. Also for the later eluting amino acids, tryptophan-d5 produced much better quantitative results in the cation-exchange cleanup. Amino acids which used an isotopically labeled analog for quantitation (e.g. glutamine/glutamine-d5, methionine/ methionine-d3) produced excellent calibration curves as expected. Much more accurate results could be obtained with this method if stable isotope labeled analogs were utilized for all target amino acids. This would be prohibitively expensive

7 Analysis of 25 underivatized amino acids in human plasma 2723 Table 3. Inter-day accuracy and precision of amino acids in human plasma without cation-exchange cleanup Avg. conc. (mm) Avg. conc. (mm) Avg. conc. (mm) Inter-day Inter-day Precision Spike conc. Inter-day Sample name Analyte peak name (n ¼ 5) (day 1) (n ¼ 5) (day 2) (n ¼ 5) (day 3) Average Std Dev. (%RSD) (mm) Spike recovery (%) PLASMA 0-5 (Cysteine) PLASMA (Cysteine) PLASMA 0-5 Alanine (ALA) PLASMA Alanine (ALA) PLASMA 0-5 Alanyl-Glutamine (ALA-GLN) PLASMA Alanyl-Glutamine (ALA-GLN) PLASMA 0-5 Arginine (ARG) PLASMA Arginine (ARG) PLASMA 0-5 Asparagine (ASN) PLASMA Asparagine (ASN) PLASMA 0-5 Aspartic acid (ASP) PLASMA Aspartic acid (ASP) PLASMA 0-5 Citrulline (CIT) PLASMA Citrulline (CIT) PLASMA 0-5 Glutamic acid (GLU) PLASMA Glutamic acid (GLU) PLASMA 0-5 Glutamine (GLN) PLASMA Glutamine (GLN) PLASMA 0-5 Glycine (GLY) PLASMA Glycine (GLY) PLASMA 0-5 Histidine (HIS) PLASMA Histidine (HIS) PLASMA 0-5 Hydroxyproline (HYP) PLASMA Hydroxyproline (HYP) PLASMA 0-5 Isoleucine (ISO) PLASMA Isoleucine (ISO) PLASMA 0-5 Leucine (LEU) PLASMA Leucine (LEU) PLASMA 0-5 Lysine (LYS) PLASMA Lysine (LYS) PLASMA 0-5 Methionine (MET) PLASMA Methionine (MET) PLASMA 0-5 Ornithine (ORN) PLASMA Ornithine (ORN) PLASMA 0-5 Phenylalanine (PHE) PLASMA Phenylalanine (PHE) PLASMA 0-5 Proline (PRO) PLASMA Proline (PRO) PLASMA 0-5 Serine (SER) PLASMA Serine (SER) PLASMA 0-5 Taurine (TAU) PLASMA Taurine (TAU) PLASMA 0-5 Threonine (THR) (Continues)

8 2724 M. Armstrong, K. Jonscher and N. A. Reisdorph Table 3. (Continued) Avg. conc. (mm) Avg. conc. (mm) Avg. conc. (mm) Inter-day Inter-day Precision Spike conc. Inter-day (n ¼ 5) (day 1) (n ¼ 5) (day 2) (n ¼ 5) (day 3) Average Std Dev. (%RSD) (mm) Spike recovery (%) Sample name Analyte peak name PLASMA Threonine (THR) PLASMA 0-5 Tryptophan (TRP) PLASMA Tryptophan (TRP) PLASMA 0-5 Tyrosine (TYR) PLASMA Tyrosine (TYR) PLASMA 0-5 Valine (VAL) PLASMA Valine (VAL) to do for all amino acids, but if only a few amino acids are targeted, the benefits of improved accuracy would conceivably outweigh the costs. Accuracy and precision of method To measure accuracy and precision, human plasma was prepared and analyzed without amino acids spiked, and with amino acids spiked at levels equivalent to the level 4 calibration standard (100 mm nominal concentrations). The plasma samples were spiked and frozen at 808C for at least 24 h before thawing, preparing and analyzing. Three separate aliquots of the unspiked/spiked plasma were prepared and analyzed on three different days. Results from the first day were used to calculate the intra-day accuracy and precision and results from all three days were used to calculate inter-day accuracy and precision (Table 3). The same unspiked/spiked samples were prepared using cation-exchange cleanup, and intra-day precision was calculated. Intra-day precision of all amino acids without cationexchange cleanup was <20 RSD (n ¼ 5) for all amino acids measured. Only Cit, Gly, His, and Val had RSDs >10. Intra-day spike recoveries for most amino acids was within % with the exception of Ala-Gln (21.4%), Asn (67.4%), His (209%), Hyp (66.6%), Orn (440%), Pro (126%) and Tau (30.1%). Due to extremely poor chromatography, His and Orn peaks were poorly integrated, resulting in aberrantly high recoveries. Ion suppression from plasma co-extractives eluting in the void volume reduced the recovery of Tau, while Ala-Gln recovery was affected by ion suppression from TDFHA adducts. Inter-day precision of most amino acids was <20 RSD (see Table 3). However, some amino acids had very high inter-day RSD due to low endogenous concentration (Cys 2 ), poor chromatography (Orn), or ion suppression (Ala-Gln, Val). Inter-day spike recoveries for most amino acids were within % with the exception of Cys 2 (154%), Ala-Gln (26.7%), Asn (63.4%), His (160%), Hyp (66.7%), Orn (274%), Tau (32%) and Val (67.7%). Reproducibility of Val quantitiation was good on the first two days of the study; however, due to progressively degrading chromatography, Val eventually co-eluted with the TDFHA adduct peak and its signal was suppressed. This degradation in chromatography can be improved through more frequent washes with 100% acetonitrile, as was noted by Piraud et al. 23 Intra-day precision of amino acids following cationexchange cleanup was <20 RSD (n ¼ 5) for all amino acids measured. Intra-day accuracy following cation-exchange cleanup was vastly improved over analysis without cation-exchange cleanup. Recoveries for all amino acids were between %, with only Ala (127%), Asn (78.3%), Cit (78.3%) and Glu (79.2%) outside of %. The most dramatic improvements occurred with Tau (118%), His (93.4%) and Orn (84.8%). We attribute the improvement in the results to the removal of co-extracted non-target analytes and reduction of the TDFHA adduct that resulted in significantly diminished ion suppression. The degradation in chromatography over time was also much less pronounced. While similar intra-day precision was achieved without cation-exchange cleanup,

9 Analysis of 25 underivatized amino acids in human plasma 2725 Figure 4. Mass accuracy of the ESI-TOF is under 1 ppm for His and under 3 ppm for Lys and Arg. Plasma samples were prepared and analyzed using RP-LC/ESI-TOF as described in the Experimental section. The mass spectra for Lys, His and Arg are shown and mass accuracy was calculated using experimental mass and actual mass values as shown. Overall the mass accuracy is well below 5 ppm, thereby reducing the number of possible empirical formulas for these peaks and improving the qualitative spectral data. the few additional steps involved with the cleanup greatly improve the overall performance of the method. DISCUSSION Analysis of amino acids by ion-pairing RP-LC/ESI-TOF is a viable alternative to traditional amino acid analysis methods, such as GC/MS and ninhydrin methods, both of which require derivatization. While the IC/ninhydrin method requires very little sample preparation, relatively long (approximately 1 2 h per sample) analysis times are needed in order to achieve baseline separation suitable for quantitation. The method also generally requires a dedicated system for online derivatization of samples in order to obtain consistent results. Conversely, GC/MS methods traditionally have shorter run times, excellent chromatography, and increased specificity; however, GC/MS methods require extensive sample preparation and derivatization. Increased handling of the sample due to numerous steps is not only time-consuming, but can potentially lead to increased error and variability in the results. Although the preparation of samples for LC/MS analysis using an amino acid kit is relatively simple compared to derivatization, the kits are designed to be used with the specific derivative chemistry they were developed for and may not be optimal for every amino acid, nor for every detection technique. Analysis of the resultant samples requires very high flow rates ( ml/min) and the use of non-volatile salt buffers which are not readily compatible with ESI-MS. Tandem mass spectrometry with flow injection analysis can be used to identify amino acids without chromatographic separation through the use of monitoring the transition of precursor ions to product ions or multiple reaction monitoring (MRM). While this technique is extremely specific and sensitive, some analytes may be subject to ion suppression due to the complex nature of the sample matrix, 23 resulting in anomalous quantitation levels. While there have been some promising advances in amino acid analysis utilizing sample introduction and ionization methods such as matrix-assisted laser desorption/ionization (MALDI) 24 and high-field asymmetric waveform ion mobility spectroscopy (FAIMS), 25 as of the time of writing neither of these techniques has been investigated for analysis of amino acids in biological fluids such as plasma or urine. In spite of advantages such as high mass accuracy, TOFMS has not been used extensively for quantitative analysis due to limitations imposed by time-to-digital converters (TDC), which have poor dynamic range and can have considerable dead times when measuring high concentrations of analyte. An analog-to-digital converter (ADC) can more accurately measure signal intensity than a TDC. ADC technology allows TOF mass spectrometers to be used for quantitative analysis while retaining a high degree of mass accuracy. The mass accuracy for amino acids obtained by this method of correction is well below 5 ppm (Fig. 4). CONCLUSIONS Ion-pairing reversed-phase chromatography coupled with the current generation of small particle size columns makes separation of amino acids possible without derivatization, allowing for quick and reproducible sample preparation. Mass accuracy obtained through time-of-flight mass spectrometry can be used in addition to retention time to provide qualitative data that is not available with single quadrupole or triple quadrupole mass spectrometers. The described method can be utilized to provide quick and accurate results for amino acids in human plasma.

10 2726 M. Armstrong, K. Jonscher and N. A. Reisdorph Acknowledgements Support for this work was generously provided through the Colorado Clinical Nutrition Research Unit (Funding through NIH/NIDDK P30 DK , PI Dr. James Hill). The authors would like to thank Dr. Patti Thureen for her helpful comments. REFERENCES 1. Jellum E, Kvittingen EA, Stokke O. Biomed. Environ. Mass Spectrom. 1988; 16: Deng C, Deng Y. J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 2003; 792: Magni F, Arnoldi L, Galati G, Galli Kienle M. Anal. Biochem. 1994; 220: Pitt JJ, Eggingtin E, Kahler SG. Clin. Chem. 2002; 48: Shen X, Deng C, Wang B, Dong L. Anal. Bioanal. Chem. 2006; 384: Deng C, Li N, Zhang X. Rapid Commun. Mass Spectrom. 2004; 18: des Robert C, Le Bacquer O, Piloquet H, Roze JC, Darmaun D. Pediatr. Res. 2002; 51: Miller RG, Jahoor F, Reeds PJ, Heird WC, Jaksic T. J. Pediatr. Surg. 1995; 30: Spackman DH, Stein WH, Moore S. Anal. Chem. 1958; 30: Mahwhinney TP, Robinett RS, Atalay A, Madson MA. J. Chromatogr. 1986; 358: Calder AG, Garden KE, Andersen SE, Lobley GE. Rapid Commun. Mass Spectrom. 1999; 13: Wood PL, Amin Khan M, Moskal JR. J. Chromatogr. B 2006; 831: Husek JP. J. Chromatogr. 1991; 552: Husek JP. J. Chromatogr. B 1998; 717: Namera A, Yashiki M, Nishida M, Kojima T. J. Chromatogr. B 2002; 776: Petritis K, Chaimbault P, Elfakir C, Dreux M. J. Chromatogr. A 1999; 833: Muller C, Schafer P, Stortzel M, Vogt S, Weinmann W. J. Chromatogr. B 2002; 773: Matuszewski BK, Constanzer ML, Chavez-Eng CM. Anal. Chem. 1998; 70: Cohen SA, Michaud DP. Anal. Biochem. 1993; 211: Petritis K, Chaimbault P, Elfakir C, Dreux M. J. Chromatogr. A 2000; 896: Chaimbault P, Petritis K, Elfakir C, Dreux M. J. Chromatogr. A 1999; 855: Piraud M, Vianey-Saban C, Petritis K, Elfakir C, Steghens JP, Morla A, Bouchu D. Rapid commun. Mass Spectrom. 2003; 17: Piraud M, Vianey-Saban C, Petritis K, Elfakir C, Steghens JP, Bouchu D. Rapid Commun. Mass Spectrom. 2005; 17: Alterman M, Gogichayeva N, Kornilayev B. Anal. Biochem. 2004; 335: McCooeye M, Mester Z. Rapid Commun. Mass Spectrom. 2006; 20: 1801.