2-fluoro-1-methylpyridinium p-toluene sulfonate: a new LC-MS/MS derivatization reagent for vitamin D metabolites

Vitamin D analysis by MS faces several analytical challenges, including inefficient ionization, nonspecific fragmentation, interferences from epimers, isomers, and isobars, as well as very low concentration levels. In this study, we used 2-fluoro-1-methylpyridinium (FMP) p-toluene sulfonate for derivatization of vitamin D3 metabolites to increase detection sensitivity and allow for full chromatographic separation of vitamin D isomers and epimers. UHPLC-MS/MS was used for measurement of five vitamin D3 metabolites in human serum. Compared with Amplifex and 4-phenyl-1,2,4-triazolin-3,5-dion, the FMP p-toluene sulfonate reaction required less time to be performed. The method was optimized and validated to ensure accuracy, precision, and reliability. In-house and commercial quality control samples were used to assure the quality of the results for 25-hydroxyvitamin D3. The method showed very good linearity and intraday and interday accuracy and precision; coefficients of determination (r2) ranged between 0.9977 and 0.9992, relative recovery from 95 to 111%, and coefficient of variation from 0.9 to 11.3. Stability tests showed that the extracted derivatized serum samples were stable for 24 h after storage at −20°C; 24,25-dihydroxyvitamin D3 and 1,25-dihydroxyvitamin D3-FMP derivatives were stable for 1 week at −80°C. The method was applied to samples of healthy individuals for quantitative determination of vitamin D3, the two epimers of 25-hydroxyvitamin D3 and 24,25-dihydroxyvitamin D3.

Determining the vitamin D status of individuals requires accurate and precise quantification of 25hydroxyvitamin D (25(OH)D) concentration levels (1). Currently, LC-MS/MS is considered the "gold standard" technique for this purpose (2,3). As other vitamin D metabolites have also been proposed as markers for different pathological conditions (4), it is important to develop methods for simultaneous measurement of multiple vitamin D 3 metabolites with adequate precision, accuracy, and sensitivity.
Analytical method development for vitamin D metabolites using MS is challenging, however, because of the poor ionization efficiencies of these molecules and the low concentration levels in biological samples (5). To overcome these difficulties, optimized sample preparation methods, including chemical derivatization, can improve analyte recovery from the matrix, eliminate coexisting interferences, and increase ionization efficiency (6). In addition, derivatization shifts masses to higher m/z values, which generally exhibit less isobaric noise and often generate more specific fragmentation patterns during MS/MS analysis. Moreover, derivatization can improve chromatographic separation of analytes, which is particularly important for vitamin D metabolites, as separation of isomers and epimers can be challenging. Derivatization of vitamin D compounds often yields multiple products for each metabolite, however, which will then generate several peaks during chromatography (7).
In contrast, reagents such as isonicotinoyl chloride and Girard P (18,19) target the much less specific hydroxyl groups. Vitamin D metabolites possess one or more -OH groups and, depending on stereochemical hindrances and reagent concentration, usually lead to between one and three products for each molecule. Diels-Alder and hydroxyl group derivatizations can also be combined, as described by Higashi et al. (20). Depending on the polarization of the C-OH bond, different synthetic strategies are required; for example, by nucleophilic substitution or via initial alkene formation and subsequent derivatization. Dehydration is required for the latter, requiring the -OH group to be located next to a quaternary carbon (21).
In this study, 2-fluoro-1-methylpyridinium ptoluene sulfonate (FMP-TS) was used for hydroxyl group derivatization. FMP-TS is a 2-halopyridinium salt used for synthesis of esters, amides, and thiol esters (22,23). As hydroxyl groups are poor leaving groups, reactions via S N 2 are not very effective (24), but reactions employing oxygen as nucleophile can be used to produce ionizable groups or add preionized groups via nucleophilic substitution. It was shown that primary and secondary hydroxyl groups can be converted to N-methylpyridyl ether salts by nucleophilic displacement of FMP-TS by oxygen (21). Bald (25) demonstrated that the reaction takes place in various solvents in the presence of triethylamine (TEA).
In this study, we present a new derivatization method for LC-MS/MS quantification of five major vitamin D 3 metabolites in serum samples using FMP-TS after liquid-liquid extraction (LLE). The derivatization reaction was very quick, easy to perform, and inexpensive. We describe optimization of various parameters of the derivatization reaction as well as validation of the entire assay. To the best of our knowledge, this is the first time that FMP-TS has been used for measurement of vitamin D 3 , 3β-25-dihydroxyvitamin D 3 (3β-25(OH)D 3 ), 3α-25-dihydroxyvitamin D 3 (3α-25(OH)D 3 ), 1,25-dihydroxyvitamin D 3 (1,25(OH) 2 D 3 ), and 24,25-dihydroxyvitamin D 3 (24,25(OH) 2 D 3 ) in serum samples. It has been previously reported as derivatization reagent for estrogen in water, human plasma, and serum (26)(27)(28).

Preparation of standard solutions and QC samples
Stock standard solutions of vitamin D 3 metabolites were prepared in methanol at 1 mg/ml and their respective isotope IS at 1 μg/ml. All solutions were stored at −20 • C.
Human vitamin D 3 free serum was spiked with vitamin D 3 standard solutions to produce QCs and samples for calibration curves for each analyte ( Table 1). Concentration of QCs was prepared at three levels representing samples with high, medium, and low concentration levels of the five investigated vitamin D 3 metabolites.

Biological samples
The method was applied to 12 human plasma samples from healthy individuals and approved by the Local Research Ethics Committee (Charité Universit¨atsmedizin Berlin, ethics number EA2/176/19). Written informed consent was provided by all participants. The human study reported in this article abided by the Declaration of Helsinki principle.
The reagent was dissolved in regular ACN and dry ACN (5 mg/ml) to investigate if the presence of water affected the reaction. Dry ACN was prepared by using a bead-type molecular sieve (0.3 nm, sodium aluminum silicate). Furthermore, two bases (TEA and TMA) were examined in their suitability to promote the derivatization reaction. Base concentration was optimized by addition of 1, 2, and 4% v/v. The concentration of the reagent was studied at 1, 5, 10, and 20 mg/ml. After optimization of the added components, incubation parameters were examined; viz., incubation temperatures at 22 • C (room temperature), 30

Extraction
Prior to sample preparation, 10 μl of a mixture of the five ISs were added to 100 μl of sample to give the following concentrations: 3α-25(OH)D 3 -d6 at 6 ng/ml, 3β-25(OH)D 3 -d 6 at 20 ng/ml, 1,25(OH) 2 D 3 -d 6 at 6 ng/ml, 24,25(OH) 2 D 3 -d6 at 6 ng/ml, and vitamin D 3 -d 3 at 10 ng/ml. Protein precipitation, two-step LLE, and derivatization were used, based on a previous method reported by us (8) and a modified version described by Alexandridou et al. (29) for 100 μl of serum sample. Briefly, protein precipitation was achieved by adding 250 μl of ACN to 100 μl of sample, followed by 1 min of vortexing and 15 min of centrifugation at 10,000 rpm. A Concentrator plus/Vacufuge® plus (Eppendorf, Hamburg, Germany) was used to evaporate the supernatant to dryness after its transfer to a new vial. For LLE, 100 μl of water and 200 μl of ethyl acetate were added to the dry residue followed by 30 s of vortexing and 5 min of centrifugation at 10,000 rpm. The organic phase was transferred to a new vial, and the aqueous phase was re-extracted by adding 200 μl of ethyl acetate. The two organic fractions were combined and evaporated to dryness prior derivatization.

LC-MS/MS
Chromatographic separations were performed on a 1290 Infinity II LC system (Agilent, Santa Clara, CA) after injection of 5 μl of sample. The autosampler was set to 10 • C. A Phenomenex (Torrance, CA) Kinetex® 2.6 μm F5 (pentafluorophenyl) 100 Å column (100 × 2.1 mm) was used in combination with a mobile phase consisting of (A) water (+0.1% formic acid) and (B) ACN (+0.1% formic acid). The column temperature was set to 35 • C, and the flow rate was 0.4 ml/min. An isocratic step of 44% B was held for the first 4 min, followed by the following gradient steps: increase to 49% B in 2.5 min, steep increase to 70% in 0.1 min, and increase from 70% to 82% B in 3.5 min. Finally, B was held at 100% for 2 min, before 2 min re-equilibration at the initial conditions. The total time of analysis was 14 min.
The UHPLC system was coupled to a Sciex (Concord, ON, Canada) QTRAP 6500+ quadrupole-linear ion trap MS. Electrospray ionization was conducted in positive ion mode using a Turbo-V source using multiple reaction monitoring. Ion source and multiple reaction monitoring parameters were optimized to achieve the highest detection sensitivity (supplemental Table S1).
Dwell times were chosen to obtain 12 to 20 data points across chromatographic peaks. Analyst software (Sciex), version 1.7, was used for data acquisition, whereas data analysis was performed using MultiQuant (Sciex), version 3.0.3.

Recovery
Extraction recoveries were determined for the five metabolites at three concentration levels; low, medium, and high; results are presented in the Results and Discussion section.

Linearity
Samples for calibration curves were prepared using vitamin D-free serum, spanning the expected concentration range in the blood samples. Eight calibrants were prepared for each metabolite; each concentration was measured in triplicate. Moreover, a blank sample without the investigated analyte or the respective IS, and a "zero" sample containing no analyte, but IS was prepared for every calibration curve. The coefficient of determination (r 2 ) was used to evaluate the linearity for every analyte.

Accuracy and precision
Accuracy and precision were determined for all investigated metabolites using spiked serum samples (QCs) at three different concentration levels (low, medium, and high) in the presence of IS. QCs were prepared independently from the samples used for calibration curves. Three independent samples were prepared per concentration level and were measured in triplicate to evaluate intraday or within-run accuracy and precision. Interday or between-run accuracy and precision were validated by three independent runs within 1 week of the QCs above (three samples analyzed in triplicate). Relative recovery (R%) (Equation 1) was used to express accuracy, whereas precision was expressed as coefficient of variation (%CV) (Equation 2).

R% =
Mean of experimental concentrations Nominal concentration × 100 (Eq. 1) ClinChek® Serum Control-certified reference standards were used to evaluate the accuracy and precision for 25(OH)D 3 at two different concentration levels (low and high). Three samples of the different concentration levels were prepared and analyzed three times each.
Accuracy was considered acceptable when the mean concentration of the QCs was within 15% of the nominal concentration. An acceptable precision required CV % values not exceeding 15% for the QC samples.

Limit of detection and lower limit of quantitation
To determine limit of detection (LOD) and lower limit of quantitation (LLOQ), vitamin D-free serum was spiked at decreasing concentration levels; these samples were measured in triplicate. The lowest concentration of analyte with signal-to-noise (S/N) ratio >3 determined LOD. LLOQ was defined as the lowest calibrant concentration, which could be quantified with acceptable accuracy and precision according to the "Guideline on Bioanalytical Method Validation" by the European Medicines Agency (https://www.ema.europa.eu/en/documents/scientificguideline/guideline-bioanalytical-method-validation_en. pdf) and the guidelines of the Food and Drug Administration for bioanalytical method validation (https://www. fda.gov/files/drugs/published/Bioanalytical-Method-Validation-Guidance-for-Industry.pdf).

Stability of the derivatization products
The stability of the derivatized analytes was examined in order to investigate if processed samples could still be measured after remaining in the autosampler for 24 h at 10 • C and after storage for 24 h at −20 • C. Long-term stability of the derivatization products was also investigated, including storage for 1 week at −80 • C. In every experiment, three concentration levels were measured (low, medium, and high), and three independent samples were prepared and measured in triplicate. Freshly prepared samples of each concentration level were immediately measured to establish time 0 responses (t = 0). Stability was considered acceptable if the mean concentration of each level was between ±15% of the nominal concentration.

Chromatography and MS of FMP-TS products
All FMP-TS products for the investigated vitamin D metabolites were readily resolved by LC using the instrumental conditions described in the Materials and methods section. Figure 1A shows extracted ion chromatograms of derivatized standards and ISs. For the epimers, baseline separation was achieved for derivatized 25(OH)D 3 with calculated resolution R s = 2.0 (Fig. 1A, green traces). Figure 1B illustrates extracted ion chromatograms, showing that all derivatized analytes can be readily separated from interferences in human serum. Total chromatographic run time was 14 min; total injection-to-injection cycle time was 14.2 min.
The common vitamin D metabolites offer up to three -OH groups for reaction with FMP-TS via nucleophilic substitution; viz., the native -OH group of vitamin D at C-3 plus two metabolic oxidations. However, only single addition of FMP was observed for 1,25(OH) 2 Table S1).
The sensitivity gain of the method was investigated from spiked serum samples after measuring derivatized and nonderivatized samples. Serum samples were preferred over standards in solvent because matrix effects were included in this comparison. The results are summarized in Table 2. Gains were expressed as relative peak areas, calculated as peak area ratio of derivatized to nonderivatized analyte.

Derivatization optimization
The reaction conditions were optimized using spiked serum samples containing the different metabolites at concentration levels comparable to human samples (see Materials and methods section). Also, blank serum samples were used to ensure the absence of coeluting isomers or isobars for analytes and ISs. Several experimental conditions were investigated consecutively to enhance the detection sensitivity of the method: 1) type of solvent and the presence of moisture in the reaction   3. The experiments above were conducted at FMP-TS concentration of 5 mg/ml. We also systematically investigated the impact of FMP-TS concentration on the reaction outcome. The results are shown in Figure. 2. Increasing reagent concentration from 1 to 5 mg/ml increased peak areas for all metabolites, whereas further increases provided only detrimental effects for 1,25(OH) 2  Finally, Fig. 3 demonstrates the effects of incubation time on reaction yields in our experiments. Increasing incubation time from 5 to 15 min strongly improved yields for all metabolites, which then declined again for longer reaction times (up to 30 min), most likely as a result of degradation reactions. We chose the optimum yield at 15 min for our assay. Any variability of yields as a result of small variations of the incubation time is compensated by the equally affected IS, thus maintaining the accuracy. Equally, as LLOQ values in this study were determined based on the precision of the assay and thus gave much higher values than LLOQs based on signal-tonoise ratios (e.g., 10× S/N), absolute ion currents for all metabolites were always high, even at LLOQ, regardless of the incubation time.

Method validation
The method was validated by means of guidelines of the Food and Drug Administration for bioanalytical method validation (https://www.fda.gov/files/drugs/ published/Bioanalytical-Method-Validation-Guidancefor-Industry.pdf) and the European Medicines Agency (https://www.ema.europa.eu/en/documents/scientificguideline/guideline-bioanalytical-method-validation_ en.pdf). Here, recovery, linearity, accuracy, precision, LOD, LLOQ, as well as stability were determined according to these guidelines.

Recovery
The extraction recovery was investigated for every metabolite at three different concentration levels (low, medium, and high) to ensure that extraction is efficient and reproducible. Percentage extraction recovery (R%) was calculated as follows: E BE : peak area of investigated metabolite spiked in the sample prior to extraction. E AE : peak area of investigated metabolite spiked after extraction.
E IS : peak area of the IS of each metabolite. Vitamin D-free serum samples were spiked with a mixture of the analytes in methanol preextraction and postextraction in triplicate, followed by derivatization and analysis. Mean values of the three independent samples were used (Table 3). Satisfactory R% values between 81% and 99% were achieved, demonstrating that the five metabolites were adequately extracted across the investigated concentration levels.

Linearity
Simple linear regression was chosen as calibration model only for 24,25(OH) 2 D 3 using the least squares method. For the other metabolites, weighted factor 1/x using least squares was applied. The coefficient of determination (r 2 ) was used to evaluate linearity for each analyte. Table 3 shows r 2 values, calibrant concentrations, and working range for the investigated metabolites.
Commercial ClinChek® Serum Control was used as additional measure to evaluate accuracy and precision for 3β-25(OH)D 3 at two different concentration levels. These samples were analyzed during the intraday evaluation phase for accuracy and precision as well as during the analyses of human samples ( Table 4). The resulting R% (97-100%) and %CV (<3%) values demonstrate that the method's performance was excellent for the vitamin D status marker.

LOD and LLOQ
For the native vitamin D 3 species, LLOQ was defined by the blank concentration contained in the commercial vitamin D-free serum used in this study. By standard addition, we found that VD-DDC Mass Spect Gold® serum contained vitamin D 3 at 0.7 ng/ml, which is within specifications for this product (vitamin D 3 , ≤1 ng/ml). For this reason, no LOD was determined for vitamin D 3 . To determine LLOQ of the other metabolites, three samples of the lowest calibrator were analyzed and measured in triplicate to demonstrate that precision (CV% <20%) and accuracy (80-120%) criteria were met. Table 3 summarizes LOD and LLOQ values determined for the investigated compounds.

Stability of derivatization products
Short-term stability (24 h at −20 • C) and long-term stability (1 week at −80 • C) were investigated for the derivatized metabolites in the processed matrix. The obtained concentrations were compared with the nominal concentrations using a freshly prepared calibration curve for each analyte. The results are summarized in Table 5. The stabilities of the extracts in the autosampler were also investigated (supplemental Table S2) and demonstrate that the processed samples can safely remain in the autosampler for 24 h at 10 • C until measurement. All five derivatized vitamin D 3 metabolites were accurately measured at all concentration levels (R% ranged between 90 and 107%) with good precision (% CV ≤7.4) after storage for 24 h at −20 • C. However, only 24,25(OH) 2 D 3 and 1,25(OH) 2 D 3 products meet acceptable criteria after storage for 1 week at −80 • C (R% ranged between 85 and 97%). For the other metabolites, the mean concentration ranged between −15% and −26% of the nominal concentration. Further information on longer storage periods (1 and 3 month storage at −20 • C) was reported previously and is in agreement with the results presented here (29).

Application to serum samples from healthy individuals
The newly developed method was applied to 12 plasma samples of healthy individuals. Vitamin D 3 , 24,25(OH) 2 D 3 , 3α-25(OH)D 3 , and 3β-25(OH)D 3 were successfully quantified in all samples with very good precision (Table 6). For one individual, 24,25(OH) 2 D 3 levels were below LLOQ. The same person also exhibited the lowest 3β-25(OH)D 3 concentrations. For most samples, levels of 1,25(OH) 2 D 3 were below LLOQ.

DISCUSSION
It is shown in the Results section above that all products of the FMP-TS reaction were fully resolved for the investigated vitamin D metabolites under the chosen chromatographic conditions, including isomers and epimers.  only at one -OH group since the second -OH group is a tertiary alcohol. Vitamin D 3 was derivatized at the single hydroxyl group, which is connected to a secondary carbon atom.
Most of the metabolites' derivatization products exhibited the expected enhanced responses because of the permanent charge, which is generally beneficial for electrospray ionization response. The strongest gains were seen for the native vitamin D 3 , exhibiting significant 285-fold improvement of the peak areas. The smallest gains were observed for 1,25(OH) 2 D 3 (3×). A possible explanation for the discrepancy between monohydroxylated and dihydroxylated metabolites (e.g., 24,25(OH) 2 D 3 showed lower increases in relative peak areas than 25(OH)D 3 epimers) is the formation of more than one product because of the two hydroxyl groups at C-3 and C-24. Similarly, Lin et al. reported that signal intensities of some of their investigated compounds were positively affected, whereas others were not. Specifically, signal intensities of 17β-estradiol-FMP, estrone-FMP, and 17α-ethinyl estradiol-FMP were 2.19-12.1 higher than the nonderivatized molecules, whereas estriol-FMP produced smaller signals than the nonderivatized metabolite (28). Again, the formation of multiple products after reaction of more than one of the three hydroxyl groups of the estriol molecule was the likely reason for the reduced signals. Furthermore, the authors observed that additives such as formic acid and the reconstitution procedure before measurement influenced ion suppression and thus final signal intensities.
We optimized the derivatization reaction further by studying the type of solvent and the presence of moisture in the reaction solvent; type of base and concentration, FMP-TS concentration, as well as incubation temperature and incubation time (see Results section). Recovery of the method was very satisfactory for every metabolite. For the calibration curves, simple linear regression was chosen with or without weighted factor as the calibration model. Moreover, the method was accurate and precise both within-day as well as between-day measurements. In addition, extra care was taken to ensure quality of the measurement of 3β-25(OH)D 3 , thus commercial QC samples were used with satisfactory results. LODs and LLOQs demonstrated the suitability of the method for the application human serum samples, except for 1,25(OH) 2 D 3 , for which the LLOQ was too high, and thus, the method was not fit for purpose for 1,25(OH) 2 D 3 .
We also designed stability experiments for FMP-TS products, as there is only limited information in the literature. This is important, as in daily practice, there is always the possibility that the analysis of processed samples (including derivatization) cannot be completed within the planned time frame, for example, in the case of instrument failure. It may then be necessary to store these processed samples at low temperatures and run them the following day or even later. It is therefore important to determine the maximum length of time, for which processed samples can be preserved, before they undergo unacceptable degradation. This is particularly important for products of the chemical derivatization, as they may be less stable than the native compounds. Some stability data on FMP derivatization products have been previously reported for other analytes. Bald (25) investigated the stability of thiol-FMP products after a period of 2 weeks and reported no changes during electrophoretic separation, with no detailed information on potential quantitative losses. Fagehi et al. (27) found that FMP products of estrone and estradiol from plasma extracts were stable for over 24 h in the autosampler at 10 • C for 24 h at −20 • C (which agrees with our observations) and −80 • C as well as for 28 days at −80 • C. Significant degradation (−28% to −42%) of estrogen-FMP products was observed after 48 h at −20 • C. Moreover, Beinhauer et al. (26) reported that estrone, α-estradiol and β-estradiol, and estriol FMP products in serum were stable for 8 h at room temperature and for 48 h at −20 • C.
In our experiments, the FMP products in our study were stable at −20 • C for short periods (24 h). For longer periods, lower temperature storage (−80 • C) is required.
Finally, we successfully demonstrated the applicability of the developed method for human serum measurements of healthy individuals.
In conclusion, a simple, fast, and sensitive LC-MS/MS method was developed for comprehensive measurement of vitamin D 3 metabolites after chemical derivatization with FMP-TS, which has not previously been used for vitamin D analysis. The derivatization reaction was initially optimized, and the assay was validated with excellent results. FMP-TS derivatization strongly enhanced the detection sensitivity of the investigated analytes because of the permanent positive charge, which was added to the molecule. At the same time, chromatographic separation of all investigated species was achieved, including full separation of the isomeric dihydroxylated species as well as the epimers of 25(OH)D 3 . Importantly, the described protocol is simpler, less expensive, and faster (approximately 15 min) than many derivatization protocols in the literature, which usually require shaking in a water bath, 1-3 h of incubation (22,25,28), or additional clean-up steps after the derivatization reaction (30).
Unfortunately, the developed method was not fit for the purpose for 1,25(OH) 2 D 3 as levels of this metabolite in serum are usually lower than the LLOQ of our assay. To overcome this limitation for 1,25(OH) 2 D 3 , Amplifex reagent can be readily used (15,31,32); however, at much higher cost as compared with FMP-TS and longer assay run times.

Data availability
The data that support the findings of this study are available from the corresponding author (Prof Dr Dietrich Volmer, dietrich.volmer@hu-berlin.de) upon reasonable request.

Supplemental data
This article contains supplemental data.