Optimization of Ultra-High-Performance Liquid Chromatography-Electrospray Ionization-Mass Spectrometry Detection of Glutamine-FMOC Ad-Hoc Derivative by Central Composite Design

Glutamine (Gln) is converted to excitatory (glutamate, aspartate) and inhibitory (γ-amino butyric acid) amino acid neurotransmitters in brain, and is a source of energy during glucose deprivation. Current research utilized an Analytical Quality by Design approach to optimize levels and combinations of critical gas pressure (sheath, auxiliary, sweep) and temperature (ion transfer tube, vaporizer) parameters for high-sensitivity mass spectrometric quantification of brain tissue glutamine. A Design of Experiments (DOE) matrix for evaluation of relationships between these multiple independent variables and a singular response variable, e.g. glutamine chromatogram area, was developed by statistical response surface methodology using central composite design. A second-order polynomial equation was generated to identify and predict singular versus combinatory effects of synergistic and antagonistic factors on chromatograph area. Predicted versus found outcomes overlapped, with enhanced area associated with the latter. DOE methodology was subsequently used to evaluate liquid chromatographic variable effects, e.g. flow rate, column temperature, and mobile phase composition on the response variable. Results demonstrate that combinatory AQbD-guided mass spectrometric/liquid chromatographic optimization significantly enhanced analytical sensitivity for Gln, thus enabling down-sized brain tissue sample volume procurement for quantification of this critical amino acid.

The amino acid glutamine (Gln) is critical for optimal neurological function as it is processed to yield excitatory (glutamate, aspartate) and inhibitory (γ-amino butyric acid) amino acid neurotransmitters, and is an alternative source of energy during glucose deficiency 13 . The structural heterogeneity of the brain necessitates capabilities for accurate quantification of Gln in small volume brain tissue samples. In order to avoid disadvantages of available colorimetric, amperometric, and fluorescence Gln detection methods [14][15][16][17][18][19] , including issues arising from matrix interference, prolonged analysis duration, and analyte instability, we developed a combinatory high-resolution micropunch dissection/UHPLC-electrospray ionization mass spectrometric (LC-ESI-MS) approach for quantification of the fluorenylmethyloxycarbonyl (FMOC) derivative of Gln, e.g. Gln-FMOC in discrete brain structures. FMOC was used as a derivatizing agent as it is superior to other compounds such as benzoyl chloride and dansyl chloride for amine detection [20][21][22] . The initial phase of this research utilized CCD methodology, involving performance of DOE-recommended experiments, DOE-generated quadratic equation-based validation of predicted responses, and statistical comparison of design versus desirability in brain tissue samples, to assess five critical mass spectrometric process variables, e.g. sheath gas pressure (SGP), auxiliary gas pressure (AGP), sweep gas pressure (SWGP), ion transfer tube temperature (ITT), and vaporizer temperature (VT), on Gln-FMOC chromatographic area. CCD methodology was next employed to evaluate effects of critical liquid chromatographic process variables, such as column temperature, mobile phase flow rate, and mobile phase composition, in order to further optimization of analytical sensitivity for Gln quantification in neural tissue.
Liquid chromatography-mass spectrometry parameters. Samples analysis was performed using a system comprised of a UHPLC Vanquish pump, Vanquish Autosampler, Vanquish UHPLC + column compartment, and ISQ EC mass spectrometer (ThermoFisherSci.), in conjunction with ThermoFisherScientificDionexChromeleon 7 Chromatography Data System software. A C18 column (4.6 mm ID X 100 mm L, 5 µm, 120 Å; Acclaim 120; ThermoFisherSci.) was used with a 0.25 mL/min. flow rate and 0.5 µL injection volume; the autosampler needle was washed with 10% (v/v) methanol (10 s) after each sample injection. Mobile phases A and B consisted of 10 mM ammonium acetate and LC-MS grade acetonitrile, respectively. The linear gradient mobile phase flow was characterized by an increase in acetonitrile from 50% to 80% between 0 to 4 min, followed by a decrease to 50% between 4 to 8 or 15 min. Preliminary results guided selection of the following mass spectrometric parameters: vaporizer temperature (VT) = 250 °C; ion transfer tube temperature (ITT) = 200 °C; optimum sheath gas pressure (SGP) = 25 psig; auxiliary gas pressure (AGP) = 2 psig; sweep gas pressure (SWGP) = 0.5 psig; and negative mode, which resulted in an extracted Gln-FMOC chromatogram at m/z 367.1. Column and autosampler temperatures were maintained at 35 °C and 15 °C, respectively. This method represented here as Gen-MS-I.  Table 1), chosen to meet minimum and maximum safe operative limits of the instrument, were utilized by Design Expert (Version: 9.0.6.2, Serial Number: 1997-5434-3543-9229) to create a series of 50 experimental runs for LC-ESI-MS analysis of Gln-FMOC (Suppl. Data; Table 2). Gln-FMOC chromatograph area measurements were analyzed by software to yield plot data, quadratic equation, and statistical results (Suppl. Data; Table 3). CCD methodology was similarly employed to assess effects of liquid chromatography column temperature, mobile phase flow rate, and mobile phase % acetonitrile (% B)−3min., % B-4.1 min., and % B-6 min. on Gln-FMOC chromatographic area, and to determine if and how optimization of these additional variables affects mass spectrometry performance.

Results and Discussion
Mass spectrometric optimization.  Fig. 1A). Plots of predicted vs. internally studentized residuals and run number vs. internally studentized residuals showed random residual distribution around the horizontal axis, with one outlier violation, demonstrating nonlinear dependency between residuals and fitted values and therefore suitability of a regression model or non-linear model for data analysis (Supp. Data Fig. 1B,C). Plots of predicted vs. actual (area in counts) exhibited one experimental outlier (Suppl. Data, Fig. 2A). Cook's distance (D i ) was used to identify experimental runs (runs 13 and 46) that do not correspond to others (Suppl. Data 2B). The current model yielded leverage measures of less than 0.6, with design points distributed in three different unique leverage lanes (Suppl. Data, Fig. 2C). Very few experimental runs produced a DFFITS value that exceeded 1.0 (Suppl Data, Fig. 2D  Analysis of combinatory mass spectrometric parameter effects on chromatographic area shows that this response was increased at medium AGP or SWGP levels, and decreased at higher AGP or SWGP levels under SGP influence (Suppl. Data, Fig. 3A,B). Improved chromatographic area occurred as a result of augmentation of either VT or ITT alongside increased SGP (Suppl. Data, Fig. 3C,D). Maximum chromatographic response was observed at medium AGP levels in combination with medium SWGP and higher VT and ITT levels, respectively (Suppl. Data, Fig. 4A-C). Medium SWGP and higher VT or ITT increased chromatographic response (Suppl. Data, Fig. 5A,B). Greatest chromatographic response was observed at maximum VT or ITT (Suppl. Data, Fig. 6). A maximum response of design can be achieved through desirability (97.5%), indicative of higher SGP = 50, and SWGP = 1.3 (Fig. 3). These data indicate that chromatographic response is maximized by increasing VT or ITT alongside SGP.
The Fresh Gln-FMOC matching stock exhibited large differences in chromatographic response versus prediction analyses. A matching concentration of Gln-FMOC was prepared for comparison of chromatographic response with design-predicted responses. Differences in response were approximately 3000 counts × min (Suppl Data, Fig. 7). The matching stock was used to evaluate chromatographic response under mass spectrometric parameter variation (Suppl Data, Table 6). Predictions of overlap between predicted versus experimental values were verified, but overlap was accompanied by differences in response area ( Fig. 4; Suppl. Data Table 6).

Selection of highest mass spectrometric area based on design runs and desirability approach.
Mass spectrometric optimization was performed using the above-described CCD technique, resulting in selection of maximum response parameters as VT = 270 °C, ITT = 301 °C, SGP = 36 °C, AGP = 7.5 °C, and SWGP = 1 °C (Design run #13 in parameters and response in Suppl. Data, Tables 2 and 3); optimized Gln-FMOC extracted mass chromatogram area-under-the-curve is denoted as CCD-MS-II. Cortical neural tissue samples were analyzed for Gln-FMOC response by utilizing Gen-MS-I, CCD-MS-II and desirability parameters and compared their difference in responses (Fig. 5). There was a significant difference between Gen-MS-I and CCD-MS-II, and Gen-MS-I and Desirability (P < 0.0001), but there was no significant difference between CCD-MS-II and Desirability. These observations led to selection of CCD-MS-II owing to lesser SGP, AGP, SWGP and VT over Desirability parameters, i.e. reduction in temperature requirement. Parameters of CCD-MS-II were kept constant during subsequent optimization of LC parameters toward further improvement of Gln-FMOC chromatographic response.
Similar prediction study was performed for LC optimization, and the predicted responses pattern were in alignment with experimental response (Fig. 7; Supp. Data Table 12). LC parameters of Prediction run-4 were similar to the design parameters run 19 (Supp. Data Tables 8 and 12), which was compared with run-1 with a view to observe the response maximum between the design and prediction run. Prediction run-1 showed significantly  Table 6).  Table 12). This indicates that quadratic equation is potential in producing the higher responses over the design set parameters. The Prediction run-1 parameters were denoted here as CCD-LC-III.
Comparison of MS and LC optimization for maximum chromatographic response. Improvement in analytical method sensitivity was verified by comparison of Gen-MS-I, CCD-MS-II, and CCD-LC-III parameters on the response variable. Microwaved cortical neural tissue was derivatized with FMOC for detection of Gln-FMOC at m/z 367.1, and chromatographic responses was compared between Gen-MS-I, CCD-MS-II, and CCD-LC-III parameters (Fig. 8). Data show that CCD-MS-II resulted in significantly higher area compared to Gen-MS-I;   Table 12).

conclusions
In the current studies, application of CCD for optimization of mass spectrometric detection of Gln-FMOC revealed significant individual and interactive effects of the critical variables AGP, SGP, SWGP, VT, and ITT on chromatographic area. CCD significantly augmented Gln-FMOC response compared to original mass spectrometric parameters. This outcome affords higher analytical selectivity with least sample injection volume, which  is beneficial for reduction in biological sample volume requirement for discriminative quantification of Gln. Efficiency of improvement of analytical sensitivity is indicated by equivalence of highest design response in the design with desirability parameters. Observations of differential yields of experimental response compared to area predicted by quadratic equation owing to matching stock of Gln supported the necessity of performance of Gln analyses on an ad-hoc basis.
CCD-LC-III optimization revealed that column temperature and mobile phase composition is indispensible for improving the mass chromatographic response over the optimization of CCD-MS-II alone. Quadratic equation allowed generating significantly higher chromatographic response when compared to design parameters in LC optimization. Higher acetonitrile concentration is essential at peak elevation besides higher column temperature; this maximum acetonitrile concentration is not essential at peak terminals. The present CCD-LC method favors further decrease in the brain tissue sample for analysis of Gln and brain microstructures.

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary  Information files).