Targeted Metabolomic Profiling of Total Fatty Acids in Human Plasma by Liquid Chromatography-Tandem Mass Spectrometry

This article reports a targeted metabolomic method for total plasma fatty acids (FAs) of clinical or nutritional relevance. Thirty-six saturated, unsaturated, or branched-chain FAs with a chain length of C8-C28 were quantified using reversed-phase liquid chromatography-tandem mass spectrometry. FAs in plasma (10 μL) were acid-hydrolyzed, extracted, and derivatized with DAABD-AE (4-[2-(N,N-Dimethylamino)ethylaminosulfonyl]-7-(2-aminoethylamino)-2,1,3-benzoxadiazole) at 60 °C for 1 h. Derivatization resulted in a staggering nine orders of magnitude higher sensitivity compared to underivatized analytes. FAs were measured by multiple-reaction monitoring using stable isotope internal standards. With physiological and pathological analyte levels in mind, linearity was established using spiked plasma. Intra-day (n = 15) and inter-day (n = 20) imprecisions expressed as variation coefficient were ≤10.2% with recovery ranging between 94.5–106.4%. Limits of detection and limit of quantitation ranged between 4.2–14.0 and 15.1–51.3 pmol per injection, respectively. Age-stratified reference intervals were established in four categories: <1 month, 1–12 month, 1–18 year, and >18 year. This method was assessed using samples from patients with disorders affecting FAs metabolism. For the first time, C28:0 and C28:0/C22:0 ratio were evaluated as novel disease biomarkers. This method can potentially be utilized in diagnosing patients with inborn errors of metabolism, chronic disease risk estimation, or nutritional applications.

Zellweger syndrome, reported by Chen et al. of 12.0 ± 5.7 µmol/L is concerning [32]. This value is significantly higher than the established reference interval in the literature and in clinical laboratories of ≤1.31 µmol/L and seems to be inaccurate, suggesting an unrecognized interference [14,25,34]. Further, reference intervals of other FAs, such as C8:0 and C10:0, reported by Chen et al., are orders of magnitude lower than known literature values [14] and should be reevaluated for potential analytical issues.
In the present study, we aimed at developing a high throughput quantitative method for FAs analysis for diagnostic and nutritional investigations using commonly available LC-MS/MS instrumentation. For this purpose, saturated, unsaturated, and branched-chain FAs with a chain length between C8 to C28 were analyzed after DAABD-AE derivatization (Figure 1). Where available, stable isotope-labeled analogs were used as IS. The method was optimized, validated, and applied to the analysis of total plasma FAs of healthy individuals and patients with established inborn errors of metabolisms.
Metabolites 2020, 10, x FOR PEER REVIEW 3 of 16 primary marker of Zellweger syndrome, reported by Chen et al. of 12.0 ± 5.7 μmol/L is concerning [32]. This value is significantly higher than the established reference interval in the literature and in clinical laboratories of ≤1.31 μmol/L and seems to be inaccurate, suggesting an unrecognized interference [14,25,34]. Further, reference intervals of other FAs, such as C8:0 and C10:0, reported by Chen et al., are orders of magnitude lower than known literature values [14] and should be reevaluated for potential analytical issues.
In the present study, we aimed at developing a high throughput quantitative method for FAs analysis for diagnostic and nutritional investigations using commonly available LC-MS/MS instrumentation. For this purpose, saturated, unsaturated, and branched-chain FAs with a chain length between C8 to C28 were analyzed after DAABD-AE derivatization (Figure 1). Where available, stable isotope-labeled analogs were used as IS. The method was optimized, validated, and applied to the analysis of total plasma FAs of healthy individuals and patients with established inborn errors of metabolisms.

Study Samples
This study was approved by the Al Ain Medical District Human Research Ethics Committee (ERH-2017-555917-3). All experiments were carried out according to applicable local rules and regulations. Informed consent was obtained from participants or their parents and/or legal guardian for study participation.
The reference intervals of plasma total FAs were generated using samples collected from control subjects (n = 282). A commercially available software package (MedCalc version 19.4.1) was used to calculate double-sided 95 percentile reference intervals using the non-parametric percentile method. Plasma samples from patients with genetically confirmed inborn errors of metabolism were also analyzed (n = 18). Commercially available human plasma used for method development and optimization was purchased from BioIVT (Westbury, NY, USA). Except during use, samples were stored at −20 • C.
To assess sensitivity improvement obtained with DAABD-AE derivatization, we compared the signal to noise (S/N) ratio (n = 3) of C8:0, C12:0, C16:0, C20:0, and C24:0 with and without derivatization. These analytes were measured on the same LC-MS/MS system with optimized mass-to-charge (m/z) transitions, and identical mobile phase composition and injection volumes.

Sample Preparation
FAs were extracted from plasma as previously described [25] with slight modification. Briefly, 10 µL aliquots of plasma were transferred into 100 × 13 mm screw-capped borosilicate tubes and mixed with HCl (60 µL, 5.0 mol/L) and 400 µL of the working IS mixture (See Table 1 footnote for individual IS concentrations). The sealed tubes were then incubated at 100 • C for 1 h to release the bound FAs. After cooling to room temperature, the total FAs content was extracted by 1.0 mL of n-hexane through 3 min of vigorous shaking followed by centrifugation at 3800 rpm for 5 min at 4 • C. The hexane phase was transferred to a new borosilicate test tube and evaporated to dryness under a flow of N 2 gas at room temperature.
DAABD-AE derivatization was achieved by reconstituting the extraction residue in 200 µL of a mixture (1:1:2 v/v/v) of EDC (25 mmol/L in water), DMAP (25 mmol/L in acetonitrile), and DAABD-AE (2 mmol/L in acetonitrile), followed by vortex mixing for 30 sec and incubation at 60 • C. After 60 min, 2.0 mL of 10% acetonitrile in water containing 0.5 g/L PFOA (mobile phase A) were added to stop the reaction. Aliquots of the resultant mixture (1 µL) were analyzed by LC-MS/MS.

LC-MS/MS System and Operating Conditions
Analyses were conducted on Shimadzu ultra-high-performance liquid chromatography (Nexera X2) consisting of two solvent delivery pumps, thermostated autosampler, column oven, degasser, and system controller (Shimadzu, Kyoto, Japan). An LC-MS 8060 triple quadrupole mass spectrometer equipped with ESI source operating in the positive mode was used for detection (Shimadzu). LabSolutions software (v 5.86; Shimadzu) running under Microsoft Windows 7 Professional environment was used to control the system and for data acquisition.

Method Validation
The linear relationship of analyte concentration versus detector response was assessed using plasma spiked with standard FAs to produce the concentration ranges shown in Table 1. Intra-day (n = 15) and inter-day (n = 20) imprecisions expressed as variation coefficient (CV%) were determined by repeated analysis of spiked plasma samples at two different levels. Analyte recovery was calculated using the following formula: Analyte recovery (%) = 100 × (measured concentration-endogenous concentration)/added concentration.
Limits of detection (LOD) were determined by recording the minimum concentrations that reliably produced S/N of 3. The limits of quantitation (LOQ) were calculated by establishing the analyte levels that produced S/N ratio of 10. Post-processing stability of DAABD-FA derivatives at 4 • C was examined by repeatedly analyzing the reaction mixture of a plasma sample that was stored in the autosampler tray for 168 h (7 days) after sample preparation.

Derivatization of FAs with DAABD-AE
In principle, analysis of unaltered FAs by LC-MS/MS can be achieved in the negative ESI mode using anion transitions generated from the elimination of water or carbon dioxide. In practice, neither of these transitions is adequately useful for reliable quantitation in complex matrices. This study aims to develop a simple, sensitive, and selective LC-MS/MS method to routinely quantify a broad range of FAs in small plasma volume for clinical evaluations. As shown in Figure 1, FAs were reacted with DAABD-AE to form stable amides with high proton affinity, ionization efficiency, and improved chromatographic properties. Collision-induced fragmentation produced a positively chargeable tertiary amine moiety with a mass-to-charge (m/z) ratio of 151originating from the derivatization reagent and was common to all studied analytes [25]. This m/z transition is detectable by positive ion ESI-MS/MS and was used conveniently to detect the studied FAs. In comparison with negative ion ESI-MS/MS detection of native FAs anions, the positive ion modification achieved through DAABD-AE derivatization resulted in significant improvement in detection sensitivity. To demonstrate the effect of derivatization on analytical sensitivity, we compared underivatized FAs with their DAABD-FA amides counterparts using the same LC-MS/MS system. Native FAs were analyzed under optimized conditions in the negative ESI mode, whereas DAABD-derivatives were analyzed by positive ESI. By comparing the S/N ratios normalized to the amount injected (µg), the sensitivity of DAABD-FA amides was a staggering nine orders of magnitude higher compared to native analytes irrespective of the FA chain length. This superior improvement of sensitivity determined a large number of FAs in a relatively small sample volume of 10 µL, an important consideration in the pediatric population.

Derivatization of FAs with DAABD-AE
The extraction of total FAs from a diminutive plasma volume (10 µL) was done as previously described [25]. Coupling of DAABD-AE with FAs was achieved using published conditions with minor modifications to accommodate the qualitative and quantitative diversity of analytes in this study [20,21,25,35,36]. This modification involved a facile single-step derivatization protocol that involves the use of premixed reagents added directly to the residual plasma extract followed by incubation at 60 • C for 1 h. When various DAABD-AE concentrations were tested, we confirmed that 2.0 mmol/L is adequate to achieve the desired derivatization yield. Figure 3A shows the derivatization yield of DAABD-FAs as a function of time, and Figure 3B illustrates the effect of DAABD-AE concentration on the derivatization yield. Predictably, derivatization with DAABD-AE imparted superb chromatographic, ionization, and fragmentation characteristic that allowed for multiplexed sensitive determination of a wide variety of clinically and nutritionally relevant FAs, including species at the extreme ends of the high and low abundance using 10 µL of plasma. In a recent work, Volpato et al. described that the derivatization of FAs with DAABD-AE can be achieved if the reaction mixture is incubated for 24 h at room temperature [31]. Our 1 h reaction conditions protocol is more practical than that of Volpato et al., as it allows for processing and reporting clinical samples without delay [31].

Derivatization of FAs with DAABD-AE
The extraction of total FAs from a diminutive plasma volume (10 μL) was done as previously described [25]. Coupling of DAABD-AE with FAs was achieved using published conditions with minor modifications to accommodate the qualitative and quantitative diversity of analytes in this study [20,21,25,35,36]. This modification involved a facile single-step derivatization protocol that involves the use of premixed reagents added directly to the residual plasma extract followed by incubation at 60 °C for 1 h. When various DAABD-AE concentrations were tested, we confirmed that 2.0 mmol/L is adequate to achieve the desired derivatization yield. Figure 3A shows the including species at the extreme ends of the high and low abundance using 10 μL of plasma. In a recent work, Volpato et al. described that the derivatization of FAs with DAABD-AE can be achieved if the reaction mixture is incubated for 24 h at room temperature [31]. Our 1 h reaction conditions protocol is more practical than that of Volpato et al., as it allows for processing and reporting clinical samples without delay [31].

Chromatographic Separation
Separation of DAABD-FAs by reversed-phase chromatography was achieved using a gradient program that increases the organic percentage of the mobile phase while maintaining constant ionic strength of the ion-pairing agent PFOA. Chemical standards and stable isotope IS were used for positive compound confirmation. FAs with shorter, branched, or unsaturated chains eluted faster

Chromatographic Separation
Separation of DAABD-FAs by reversed-phase chromatography was achieved using a gradient program that increases the organic percentage of the mobile phase while maintaining constant ionic strength of the ion-pairing agent PFOA. Chemical standards and stable isotope IS were used for positive compound confirmation. FAs with shorter, branched, or unsaturated chains eluted faster than the longer, linear, or saturated FA compounds. Under the conditions used in this work, DAABD-C8:0 eluted first at 1.6 min, whereas that of DAABD-C28:0 eluted last at 9.1 min. Retention times for the studied FAs are shown in Table 1. With a column conditioning step, the injection-to-injection time was 15 min. This relatively short analysis time is an important consideration in high volume service labs, where competition on instrument time is high, and shorter analysis time is desirable. Figure 4 shows a representative multiple reaction monitoring LC-MS/MS, overlaid with chromatograms obtained by the current method. was 15 min. This relatively short analysis time is an important consideration in high volume service labs, where competition on instrument time is high, and shorter analysis time is desirable. Figure 4 shows a representative multiple reaction monitoring LC-MS/MS, overlaid with chromatograms obtained by the current method.

Linearity, LOD, and LOQ
Linearity was assessed using plasma samples spiked with commercially available standard FAs. The studied concentration ranges of FAs were selected to encompass physiological and pathological circumstances (Table 1). Compensation for potential analytical flaws was achieved by using appropriate stable isotope IS analogs. For compounds with no commercially available IS, the stable isotope analog with the nearest chain length was used (Table 1). Regression analysis by plotting the detector response of the analyte to IS ratio against the spiked concentration confirmed linear relationships (r ≥ 0.995) in the studied concentration ranges (Table 1). LOD and LOQ were established for analytes for which standard material is available commercially. As shown in Table 1, LOD (LOQ) expressed as pmol per injection ranged between 4.2 (14.0) for C16:0 and 15.1 (50.3) for C26:0. Despite that high sensitivity achieved in this work, the lower and higher limits of the dynamic range were selected to allow for reliable determination of normal and abnormal levels regardless of endogenous analyte abundance being at the high or the low end of the concentration spectrum.

Imprecision and Recovery
Imprecision was evaluated by calculating the CV% of intra-day (n = 15), and inter-day (n = 20) studies using plasma spiked at two different FAs levels. With intra-day and inter-day CV% of less than 10.2% and 10.0%, respectively, the method described here is adequately reproducible ( Table 2). The recovery of FAs calculated from spiked samples ranged from 94.5 to 106.4% (Table 2). DAABD-FA derivatives were stable for 72 h post-processing when kept in capped vials at 4 °C in the dark.

Linearity, LOD, and LOQ
Linearity was assessed using plasma samples spiked with commercially available standard FAs. The studied concentration ranges of FAs were selected to encompass physiological and pathological circumstances (Table 1). Compensation for potential analytical flaws was achieved by using appropriate stable isotope IS analogs. For compounds with no commercially available IS, the stable isotope analog with the nearest chain length was used (Table 1). Regression analysis by plotting the detector response of the analyte to IS ratio against the spiked concentration confirmed linear relationships (r ≥ 0.995) in the studied concentration ranges (Table 1). LOD and LOQ were established for analytes for which standard material is available commercially. As shown in Table 1, LOD (LOQ) expressed as pmol per injection ranged between 4.2 (14.0) for C16:0 and 15.1 (50.3) for C26:0. Despite that high sensitivity achieved in this work, the lower and higher limits of the dynamic range were selected to allow for reliable determination of normal and abnormal levels regardless of endogenous analyte abundance being at the high or the low end of the concentration spectrum.

Imprecision and Recovery
Imprecision was evaluated by calculating the CV% of intra-day (n = 15), and inter-day (n = 20) studies using plasma spiked at two different FAs levels. With intra-day and inter-day CV% of less than 10.2% and 10.0%, respectively, the method described here is adequately reproducible ( Table 2). The recovery of FAs calculated from spiked samples ranged from 94.5 to 106.4% (Table 2). DAABD-FA derivatives were stable for 72 h post-processing when kept in capped vials at 4 • C in the dark.

Determination of FAs Reference Intervals
In the present study, a total of 282 samples from control individuals were analyzed. Non-parametric double-sided 95 percentile reference intervals stratified according to age were established in four categories: Less than 1 month (n = 59), 1 to 12 months (n = 30), 1 to 18 years (n = 71), and more than 18 years (n = 122). Table 3 provides a summary of the reference intervals of total plasma FAs in µmol/L units obtained in this study. Shown also are the reference intervals of the sum in mmol/L units of total FAs, saturated FAs, monounsaturated FAs (MUFA), and polyunsaturated FAs (PUFA). The reference intervals obtained in this study are comparable with those published in the literature [14].

Diagnostic Application on Samples from Patients with Inborn Errors of Metabolism
The diagnostic utility of the current method was evaluated using samples from patients (n = 18) with the following inborn errors of metabolism: Peroxisome biogenesis defect (PBD), X-linked adrenoleukodystrophy (X-ALD), adrenomyeloneuropathy (AMN), and RD. Results from five representative patients are shown in Table 4. In clinical laboratories, patients with PBD, X-ALD, and AMN are routinely diagnosed based on elevated plasma C26:0 and C26:0/C22:0 ratio. In this work, for the first time, we evaluated C28:0 and the C28:0/C22:0 ratio in these patients and observed significant elevations compared to controls (p < 0.0001). While C26:0 and its ratio to C22:0 are widely accepted as reliable diagnostic markers for peroxisomal disorders, C28:0 and its ratio to C22:0 described in this work are additional biomarkers with the potential to discriminate patients with PBDs from healthy individuals. This is of special importance in patients with subtle biochemical disruptions, such as patients 2, 4, and 5, shown in Table 4. Nonetheless, to establish C28:0 and its ratio to C22:0 as biomarkers of PBDs, additional studies are required to assess the diagnostic utility using a larger patients sample size that takes into account the clinical and genetic heterogeneity of PBDs. Interestingly, C28:0 and its C22:0 ratio was within the respective reference intervals in the patient with RD. This is not unexpected as this disorder is characterized by isolated PHA elevation due to deficiency of phytanoyl CoA hydroxylase, an enzyme not known to disrupt the peroxisomal β-oxidation pathway.

Method Comparison
A group of FAs, namely, C22:0, C24:0, and C26:0 for which standard GC-MS methods are available, were used to demonstrate method comparison. These compounds are valued diagnostic markers for inborn errors of metabolism associated with peroxisomal dysfunctions. Plasma samples from patients with an established diagnosis of peroxisomal disease (n=18) and samples from unaffected individuals (n = 63) were used for comparison. Bland-Altman analysis suggests that the results obtained by the current method, which fall within the 95% confidence interval, are accurate and comparable to those obtained by gold-standard GC-MS ( Figure 5).  Compared with other published LC-MS/MS methods for FAs [23,[25][26][27][28][29][30][31][32][33], our method is superior because of the following: (1) Simultaneous analysis of 36 clinically relevant saturated, unsaturated, and branched-chain FAs species between C8-C28, (2) differentiation between diagnostically significant branched-chain FAs (i.e., PRA and PHA) and their linear-chain antipodes (C19:0 and C20:0), (3) establishment of age-specific reference intervals that are in agreement with the literature [14,34] technical simplicity (i.e., single-step derivatization with no need for derivatives clean up after reaction) that allows for high throughput routine analysis suitable for large volume service Compared with other published LC-MS/MS methods for FAs [23,[25][26][27][28][29][30][31][32][33], our method is superior because of the following: (1) Simultaneous analysis of 36 clinically relevant saturated, unsaturated, and branched-chain FAs species between C8-C28, (2) differentiation between diagnostically significant branched-chain FAs (i.e., PRA and PHA) and their linear-chain antipodes (C19:0 and C20:0), (3) establishment of age-specific reference intervals that are in agreement with the literature [14,34] technical simplicity (i.e., single-step derivatization with no need for derivatives clean up after reaction) that allows for high throughput routine analysis suitable for large volume service laboratories, and (4) utilization of standard LC-MS/MS instrumentation commonly found in clinical laboratories.

Conclusions
We have reported a new LC-MS/MS approach for the quantification of 36 FAs that range in chain length between C8 and C28. This approach utilizes the superior LC-MS/MS characteristics that DAABD-AE, as a derivatization reagent, imparts onto carboxylic acid compounds. Compared to native FAs analysis, DAABD-FA derivatization improved the detection sensitivity by nine orders of magnitude. This superb sensitivity allowed for carrying out this assay using as little as 10 µL of plasma with adequate precision and accuracy, as shown by method comparison with GC-MS. Our method offers equally high coverage for medium-, long-, and very-long-chain FAs that are clinically or nutritionally significant, including MUFA, PUFA, saturated, and branched-chain FAs. As such, it can potentially be utilized in the diagnosis and monitoring of patients with various inborn errors of metabolism, such as peroxisomal and mitochondrial FA oxidation, as well as defects involving arachidonic acid metabolism. In addition, circulatory FAs measured by our method may provide estimates of chronic disease risk (e.g., cardiovascular diseases and cancer), as well as providing guidance of appropriate dietary recommendations. Given the important clues on diagnostic hallmarks and dietary biomarkers it provides, we anticipate this method to find widespread utilization in clinical and nutritional applications.