Synthesis of Low Abundant Vitamin D Metabolites and Assaying Their Distribution in Human Serum by Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS) as a New Tool for Diagnosis and Risk Prediction of Vitamin DRelated Diseases

3.1. Linear synthesis of vitamin D metabolites

Following a linear synthesis (Scheme 2), the cis-diene moiety of vitamin D₂ (13) is protected by sulfur dioxide after silylation of the hydroxyl group at C-3 as a tert-butyldimethylsilyl ether, leading to the SO₂ adduct 14. Labeling with deuterium at C-6,19,19 can be carried out conveniently by treatment with D₂O or deuterated alcohol (e.g., isopropanol). Finally, thermal removal of SO₂ and photochemical isomerization leads to threefold deuterated vitamin D₂ (d-13). This simple and convenient method to obtain C-6,19,19 deuterated metabolites [31] can also be applied to a wide variety of other vitamin D₃ and D₂ derivatives. In order to obtain metabolites of the corresponding C-3-epi series of vitamin D₃ or D₂ metabolites, the configuration of the C-3 OH group of vitamin D₂ (13) can be epimerized under the so-called “Mitsunobu conditions” [32], by treatment with an aromatic acid and an azodicarboxylate, resulting in the corresponding ester 15 with concomitant inversion of the stereogenic center at C-3 (orientation of the substituent has changed from β to α). The ester is finally cleaved by saponification with potassium hydroxide or by reduction with lithium aluminum hydride to afford C-3-epi vitamin D₂ (16), which can be used to synthesize a wide variety of C-3 epimers, including 3-epi-25-hydroxyvitamin D₃ (3-epi-25(OH)D₃) (10, d-10) and 3-epi-1α,25-dihydroxyvitamin D₃ (3-epi-1,25(OH)₂D₃) (11, d-11). Yields are only moderate due to the favored formation of a C-3,4-5,6-7,8 all-trans triene system by elimination. To further proceed with the synthesis, the C-22/23 double bond in the side chain of SO₂ adduct 14 is cleaved by ozonolysis, leading to alcohol 17 after reductive workup with sodium boronhydride, which can be used to add a wide variety of modified side chains as appropriate. For instance, 17 can be converted to an iodide, which is coupled in a nickel-zinc-mediated reaction with ethyl acrylate after prior removal of the SO₂ protective group to afford 18. At this stage, an allylic oxidation by selenium dioxide can be carried out, leading to a 1α-hydroxy series of metabolites, such as 19 as appropriate precursor. Silylation of the newly generated hydroxyl group, followed by Grignard reaction with methylmagnesium bromide, leads to 1α,25-dihydroxy metabolites. If a deuterated Grignard reagent, labeled with 3 deuterium atoms at their methyl groups, is employed, the corresponding sixfold deuterium labeled metabolite is obtained, containing 3 deuterium atoms at C-26 and C-27, respectively. Finally, the C-5/6 double bond has to be isomerized photochemically, and the silyl protective groups have to be removed, ending up with nondeuterated or deuterated 25(OH)D₃ (3, d-3) or 1,25(OH)₂D₃ (4, d-4) [33], as shown in Scheme 2. If 3-epi-vitamin D₂ (16) is employed in this sequence, 3-epi-25(OH)₂D₃ (10, d-10) is obtained. 3-epi-1,25(OH)₂D₃ (11, d-11) is obtained, if 1α-hydroxylation by selenium dioxide was carried out, and the resulting C-1-OH-group is epimerized by oxidation and final reduction.

Vitamin D₂ metabolites are synthesized by application of analogous strategies as applied for their corresponding D₃ counterparts, although they are more challenging to synthesize, for two reasons: vitamin D₂ metabolites contain an olefinic double bond in C-22/23-position, which has to be arranged in trans geometry during the course of a suitable coupling reaction of a CD-ring precursor and an alternated side chain building block by olefination. Most synthetic olefination methods yield a mixture of cis and trans double bonds, leading to product mixtures that are difficult to separate due to their similar polarity. If the reaction conditions are too basic, C-20 epimerization can also occur, leading to even more complex product mixtures. Additionally, vitamin D₂ metabolites contain a chiral methyl group in C-24 position that has to be ranged in a defined configuration. When a suitable building block bearing this methyl group is employed in the synthesis, epimerization of the chiral center may occur under basic reaction conditions, along with the risk for epimerization at C-20 in the coupling reaction. Thus, alcohol 17 is converted to its corresponding phenylsulfone, which is coupled by olefination with aldehyde 20, a method originally invented by Julia [34]. Note that 20 can be synthesized in a few routine steps from commercially available (S)-(+)-3-hydroxyisobutyric acid methyl ester, which is submitted to a Grignard reaction with methyl-magnesium bromide. Analogously to the synthesis of the other sixfold deuterated metabolites labeled at C-26/27, the addition of deuterated Grignard reagent in the course of the reaction leads to sixfold deuterated vitamin D₂ metabolites. Finally, the double bond at C-5/6 of 21 has to be photochemically isomerized and the silyl protective groups have to be removed, leading to nondeuterated or deuterated 25(OH)D₂ (22, d-22) [35].

3.2. Convergent synthesis of vitamin D metabolites

Alternatively, a convergent approach can be applied (Scheme 3), which is more versatile than a linear synthesis and allows for more harsh reaction conditions and wider scope of suitable substrates and reagents. In a classical and widely applied strategy, vitamin D₂ (13) is submitted to ozonolysis, leading to Inhoffen-Lythgoe diol 23 and allylic alcohol 24 after reductive workup with sodium boronhydride. After appropriate chemical modification of the CD-ring precursor, this compound is coupled with an A-ring building block. Two strategies can be distinguished. In the most common approach, invented by Lythgoe and further developed by the Hoffmann-La Roche group [36], allylic alcohol 24 is converted via an allylic chloride 25 in diphenylphosphine oxide 26, which is coupled with a C-8 ketone by a Wittig-Horner olefination. The A-ring diphenylphosphine oxide 26 can be coupled before or after appropriate alteration of the side chain at the CD-ring building block. To employ an A-ring diphenylphosphine oxide such as 26 is favourable, if no further substituents except a C-3 OH group and a C-10 exo-methylene group in the A-ring are needed, as it is already present in vitamin D₂ (13). As example for an efficient application of this strategy, the synthesis of 24,25(OH)₂D₃ (6) and its deuterated counterpart (d-6), their 3-epi-analogs (9/d-9) and 1α-analogs (5/d-5), is shown in Scheme 3 [37]. Here, diol 23 is converted to its corresponding iodide, which is coupled with enone 27 (obtained in a few routine steps from (R)-(-)-lactic acid using a zinc-copper-catalyzed reaction mediated by ultrasound, leading to 28. By Grignard reaction of 28 with nondeuterated or deuterated methyl-magnesium bromide, 29 or d-29 are obtained. The 1,2-diol moiety is protected as isopropylidene ether, followed by oxidation with pyridinium chlorochromate to obtain ketone 30/d-30, which can be coupled with diphenylphosphine oxide 26 to obtain d-24,25(OH)₂D₃ (d-6) after removal of silyl- and isopropylidene protective groups. To access the corresponding C-3-epi-analog, vitamin D₂ is first bis-hydroxylated at C-7/8 using potassium permanganate. Inversion of the stereogenic center at C-3 under “Mitsunobu-conditions” using picolinic acid leads to the corresponding C-3-epi-triol 31. The yields of this reaction are much higher as compared to the synthesis of 16 used in the linear synthesis, because an elimination reaction leading to a favored C-3,4-5,6-7,8-all-trans system is avoided [38]. Silylation and cleavage of the diol moiety of triol 31 by lead tetraacetate and reductive workup with sodium boronhydride lead to the CD-ring fragment 32 and C-3-epi allylic alcohol 33, which can be used after conversion to the corresponding diphenylphosphine oxide 34. Note that 34 is coupled to d-30 to obtain d-3-epi-24,25(OH)₂D₃ (d-9) or its corresponding nondeuterated counterpart 9. 3-epi- diphenylphosphine oxide 34 can be used as a versatile A-ring building block to be employed in a Wittig-Horner reaction for the synthesis of a wide variety of 3-epi vitamin D₃ and D₂ metabolites, by connection with appropriate CD-ring building blocks with modified side chain, as shown for the synthesis of 3-epi-24,25(OH)₂D₃ (9, d-9) as representative example. Finally, permanganate oxidation of vitamin D₂ (13) leads to triol 35, which is converted in a few routine steps, including a stereoselective 1α-allylic oxidation with SeO₂ via 36 to diphenylphosphine oxide 37, which is then coupled with d-30 to afford 1,24,25(OH)₃D₃ (5/d-5).

Two other low abundant metabolites, 23(S),25(R)-dihydroxyvitamin D₃ 26,23-lactone 7 and 1α-1,23(S),25(R)-trihydroxyvitamin D₃ 26,23-lactone 8, are synthesized in an analogous manner [29] (Scheme 4), by coupling a CD-ring ketone carrying an appropriate side chain with diphenylphosphine oxide 26 or 37. Here CD-ring-building block 32 obtained from vitamin D₂ 13 is used, which is converted to aldehyde 38. In order to prepare an appropriate modified side chain, isopentenol 39 is converted using a few routine steps to sulfone 40, which is then coupled with 38 by a Julia olefination to afford 41. Removal of the isopropylidene protective group, oxidation to an unsaturated α-hydroxy carboxylic acid and subsequent iodolactonization leads to ketone 44 via 42 and 43, which is coupled with diphenylphosphine oxide 26 or 37, respectively, to give 7 and 8.

3.3. Convergent Palladium-catalyzed synthesis of vitamin D metabolites

The A-ring building block is preferably synthesized de novo, if further substituents apart from a C-3-OH group and a C-10 exo methylene group in the A-ring (already present in vitamin D₂) are needed. This is particularly valuable for synthetic analogs containing substituents at C-2, which were developed in the past as new drugs [30]. However, natural metabolites may also contain additional substituents in the A-ring, such as 4β,25-dihydroxy vitamin D₃ [39]. In these cases, Pd catalyzed tandem reactions can be applied (Scheme 5). Initially invented by Trost [40], an acyclic enyne 46 is coupled with a CD-ring vinyl bromide 47 obtained by Peterson olefination from the corresponding ketone. Closure of the A-ring and its connection with CD-ring is preferably carried out in one pot. Alternatively, the corresponding triflate 48 can be employed in a Zn-mediated reaction, which was invented by Negishi [41]. Additionally, CD-ring alkene boranates 49 can be used via an analogous coupling method originally invented by Suzuki [42]. Finally, a cyclic A-ring enynes, such as 50, can be coupled with CD-ring triflate 50, a coupling method invented by Sonogashira [43]. In all these cases, the CD-ring vinyl bromide 47, boranate 49, or triflate 51 are usually carrying the modified side chain already, before coupling with the acyclic or cyclic A-ring enyne. A major drawback of these methods is the fact that synthesis of acyclic or cyclic enynes or other A-ring building blocks not derived from vitamin D₂ involves many synthesis steps, and in some cases separation of diastereomeric mixtures is also necessary to obtain enantiomerically and diastereomerically pure product. However, the suitable starting materials can usually be obtained from the natural chiral pool (i.e., terpenes such as carvone [44], malic acid [45], quinic acid [43], or carbon hydrates such as D-glucose [46] or D-xylose [47]), which are usually readily available. A wide range of substrates are tolerated. Consequently, these advanced strategies involving Pd catalyzed reactions have widely been applied in recent years and are by now well established.

4.1. Liquid chromatographic separation of vitamin D metabolites

Due to the hydrophobic structure, vitamin D metabolites are generally easily separated on reversed-phase liquid chromatography stationary phases (e.g., octadecyl C₁₈) materials utilizing hydrophobic interactions. The vitamin D metabolites generally elute in the order trihydroxylated < dihydroxylated < monohydroxylated metabolites, that is, for vitamin D₃-related molecules, the order of chromatographic retention is 1,25(OH)₂D₃ < 25(OH)D₃ < D₃. Between corresponding D₂ and D₃ analogs, the D₂ metabolites elute marginally later than the D₃ versions. For the two important isomers of dihydroxylated vitamin D metabolites, 24,25(OH)₂D₃ and 1,25(OH)₂D₃, the order of retention is 24,25(OH)₂D₃ < 1,25(OH)₂D₃.

Possibly more interesting—from a chromatography point of view—are the biochemically formed isomers after stereochemical reversal (β→α) at C-3; for example, the epimers 25(OH)D₃ and 3-epi-25(OH)D₃. The subtle diastereomeric differences between these species can be distinguished using specialized columns, such as combined C18/chiral [50], pentafluorophenyl (PFP) [51–53], and cyano (CN) [54–57].

4.2. Ionization of vitamin D metabolites for mass spectrometric analysis

Mass spectrometric determination of vitamin D metabolites using liquid chromatography-mass spectrometry (LC-MS) is not trivial because of the structural limitations that the analytes provide with respect to attaching a charge to the molecules. Gas chromatography-mass spectrometry techniques have been used in the past for qualitative analysis and structure determinations, but these methods have been almost completely replaced by modern LC-MS methods. While analysis of transformation products such as vitamin D sulfates is relatively easy by LC-MS as the metabolites can be simply analyzed as deprotonated molecules, ionizing the relatively nonpolar vitamin D metabolites is not straightforward. For LC-MS, electrospray ionization (ESI) or atmospheric pressure chemical ionization (APCI) are the most common ionization techniques, which usually result in the formation of [M+H]⁺ ions for most biological molecules. APCI has been successfully applied to steroids, as gas-phase chemical ionization often efficiently transfers a proton to these types of molecules. ESI usually relies on a charging mechanism in the liquid phase and one would therefore expect ESI to be less efficient than APCI for vitamin D metabolites. In reality this is not the case, however, and both techniques have equally been applied to vitamin D analysis, with the analytical figures of merit being quite similar between the two ionization techniques. There appears to be a slight trend toward application of ESI rather than APCI in recent years.

An important, often neglected consideration in the choice of ionization source is the (usually detrimental) impact of coionized sample components that coelute with vitamin D metabolites from the LC column. Here, ionization suppression effects caused by these coeluting molecules may impact the outcome even stronger than the differences seen in ionization efficiencies between ESI and APCI [58, 59]. However, no systematic comparison of ionization efficiency/ionization suppression effects between ESI and APCI has been performed yet. Furthermore, coionization of coeluting components also leads to the formation of isobaric interferences when sample matrices such as plasma or serum are analyzed, which will be discussed in more detail below.

Another useful ionization technique is atmospheric pressure photoionization (APPI). This technique is very common in fields such as petroleum analysis, but surprisingly, it is virtually unknown in the clinical community. To our knowledge, only one study has systematically compared APPI with APCI and shown that APPI generated significantly higher ion currents for 25(OH)D₃ than APCI [60]. A second study applied APPI to quantification of 25(OH)D₃ without comparison to ESI or APCI [61].

Finally, another option to overcome problems of ionization efficiency and resulting problems with detection sensitivity is derivatization of vitamin D metabolites, to convert them into better responding transformation products. Such procedures are now quite common in the vitamin D analytical field; they comprise introduction of a chemical group that is readily ionized or is permanently charged. Cookson-type triazoline-diones and triazoline-dione-related reagents (e.g., 4-phenyl-1,2,4-triazoline-3,5-dione [PTAD]) are most often applied, which utilize the reaction of the reagent’s dienophile with the cis-diene group at C-5/C-6 and C-10/C-19 of vitamin D (Diels-Alder [4+2] reaction) (Scheme 6).

Several promising PTAD assays have been described in the literature [39, 49, 62–64]. The advantage of using the cis-diene moiety is that all vitamin D metabolites contained in the sample are converted when the reagent is present at sufficient concentrations (i.e., in large excess quantities) as all relevant vitamin D metabolites possess this structural motif. Ionization efficiency of 25(OH)D-PTAD has been reported to be 100-fold higher than the nonderivatized molecules in ESI mode [65]. A structural variation in PTAD includes a permanently charged quaternary ammonium moiety, which enhances detection sensitivity even further [66]. This reagent is commercially available. A recent chemotyping assay for multiple vitamin D metabolites has shown detection limits in the fg/mL range for vitamin D metabolites from human serum (see discussion below).

4.3 Mass analyzers for vitamin D analysis

Most clinical applications for vitamin D utilize quadrupole-based mass analyzers with low resolving powers. Because these mass spectrometers measure m/z signals at “unit mass resolution,” they cannot be operated in full scan mode for the accurate quantification of vitamin metabolites when complex sample matrices such as serum are analyzed. Serum contains multiple components of identical nominal molecular mass to vitamin D metabolites, several of which have been shown to coelute during High Performance Liquid Chromatography (HPLC) [67]. The coeluting metabolites then generate so-called “isobaric” ions in the mass spectra after ESI, which cannot be distinguished from the vitamin D target molecules using low resolution MS. Therefore, tandem mass spectrometry (MS/MS) is implemented, which activates the ionized species further and forces them to undergo collision-induced dissociation (CID) for generating structure-specific product ions. The metabolite-specific product ions are then measured and their response used for quantification of the target metabolites.

By far the most common low resolution tandem mass spectrometer for vitamin D analysis in clinical laboratories is the triple quadrupole (QqQ) instrument. In a QqQ mass spectrometer, the [M+H]⁺ ions of vitamin D metabolites are selected in the first quadrupole, fragmented in a quadrupole collision cell, and the products analyzed in the final quadrupole mass analyzer. The acquisition mode that is almost always applied is the so-called selected reaction monitoring (SRM) mode. In this mode, the [M+H]⁺ precursor ion is isolated in Q1 (e.g., m/z 401 for 25(OH)D₃), dissociated in q2, and one specific product ion selected in Q3 (e.g., m/z 383). The quality of the precursor/product ion selection determines the selectivity of analysis; that is, the product ion structure should ideally directly reflect the chemical structure of the chosen precursor ion for maximum specificity. For vitamin D metabolites, the selection of an appropriate product ion is surprisingly difficult because of some inherent structural limitation of the precursor structure, leading primarily to simple dehydration reactions at low collision energies, which turn into complex mass spectra at higher collision energies (Figure 1) [68]. Unfortunately, because of multiple, overlapping series of fragmentation reactions, leading to “picket fence” type product ion spectra, fragment ions for vitamin D metabolites in the diagnostic m/z range often resemble each other.

A few studies also report the use of a low resolution quadrupole ion trap (IT) [69, 70] in MS/MS mode for 25(OH)D₃ and 25(OH)D₂, but these instruments are rarely implemented in clinical environments. IT instruments are often not fit for purpose for quantitation of low abundant biological molecules, because of the large detrimental contributions of acquisition time overhead to the duty cycle, thus limiting detection sensitivity. There is also a general decline of this instrument type in modern mass spectrometry labs.

The most anticipated future development for vitamin D analysis is the introduction of high resolution mass spectrometry (HRMS), which is now firmly established in many other fields of modern mass spectrometry, in particular, in pharmaceutical applications and the proteomics field [71]. This trend is mostly due to the availability of robust quadrupole-quadrupole-time-of-flight (QqTOF) and orbitrap mass analyzers in recent years, which have transformed many analytical approaches to mass spectrometry. The use of mass defect as metabolite-specific property, for example, is now an integral part of many metabolite identification routines for drug metabolites [72]. While the number of applications of HRMS in the vitamin D field is still limited, the existing work has clearly demonstrated the potential of HRMS for vitamin D analysis. For example, orbitrap mass spectrometers in full scan [73, 74] and MS/MS [75, 76] modes have been applied successfully to the analysis of 25(OH)D₃ in human serum and analytical performance has been shown to be equivalent or better than triple quadrupole and immunoassays. The important role of HRMS in the separation of isobaric interferences will be shown below.

4.4. Interferences during LC-MS/MS

The LC-MS/MS analysis of vitamin D metabolites is affected by various sources of error, which can affect both precision and accuracy. As with any other LC-MS/MS analysis from biological samples, ion suppression by coeluting sample components or chemical modifiers from the sample preparation or chromatography can lead to reduced analyte signals. There are several options to assess whether or not ion suppression is present, which have been summarized in many review articles, e.g., by Matuszewski et al. [77]. Stable isotope standards of vitamin D metabolites (mostly deuterated analogs) can usually correct for accuracy errors from ion suppression effects, as long as it is guaranteed that protein binding for the isotope standard is the same as for the endogenous analyte, which requires implementation of a careful incubation routine [78]. Importantly, deuterated isotope standards are commercially available for most of the relevant vitamin D metabolites but unfortunately not for all.

A second important source of analytical error originates from isobaric noise; that is, endogenous or exogenous metabolites that coelute with the vitamin D analytes and erroneously contribute to the analytical signal, if unspecific ions are used for mass spectral analysis. This has recently been described in detail by Qi et al. [67], who clearly demonstrated the presence of multiple isobars of 25(OH)D₃ in human serum. The isobars have to be carefully removed in low resolution mass spectrometry, by application of appropriate MS/MS or ion mobility spectrometry routines [67]. A number of exogenous and endogenous molecules have been identified as relevant metabolites in serum samples [50, 67, 79].

Many of these interferences can be eliminated by application of high resolution mass spectrometry using sufficiently high resolving powers; for example, through implementation of orbitrap or Fourier-transform ion cyclotron resonance (FTICR) mass spectrometers. For example, Liebisch and Matysik demonstrated that the orbitrap MS instrument in their study was able to separate an isobaric interference of 25(OH)D₃ in the MS/MS domain; this interference was caused by fragmentation of the d6-25(OH)D₂ isotope standard [75].

Importantly, the issues relating to isobaric noise have only been studied for the 25(OH)D₃ analyte; other metabolites will likely be affected by similar interferences, the impact of which during quantification, in particular, in multimetabolite assays (see below), remains unknown.

4.5. Method accuracy and certified reference materials

The vitamin D analytical community is supported through the Vitamin D External Quality Assessment Scheme (DEQAS), a nonprofit organization that evaluates the performance of analytical assays of member laboratories for 25(OH)D₃ and 1,25(OH)₂D₃ [80] through round-robin analyses. DEQAS has clearly demonstrated that analytical performance of vitamin D analysis has greatly improved over the years between 1994 and 2009 [81].

The United States National Institute of Standards and Technology (NIST) provides certified standard solutions for 25(OH)D₃ and 25(OH)D₂. Furthermore, the NIST and the National Institutes of Health (NIH) Office of Dietary Supplements (ODS) have established the Vitamin D Metabolites Quality Assurance Program (VitDQAP), for interlaboratory comparison of measurement of 25(OH)D₂, 25(OH)D₃, and 3-epi-25(OH)D₃ in serum and plasma and provide method-appropriate control materials.

A number of commercial reference, calibration, and quality control materials are available from several companies that allow rapid implementation and validation of vitamin D analytical methodologies.

4.6. Vitamin D multimetabolite assays

One of the most important advantages of LC-MS/MS assays over clinical immunoassays is the ability to determine multiple vitamin D species independently and simultaneously. As a result, there are now several very capable LC-MS/MS assays described in the literature that provide the capability for profiling the most relevant vitamin D metabolites at the same time, within a single analytical run, and with sufficient dynamic range to allow measuring the required physiological levels, down to the picomolar range. The topic has recently been reviewed in detail and the interested reader is referred to Ref. [82]. Briefly, most of these multimetabolite assays utilize derivatization techniques that transform all vitamin D species into better responding analogs (vide supra). This procedure in turn permits analysis of both high and low abundant vitamin D species with similar analytical figures of merit [62, 83].

The simultaneous acquisition of all important metabolite levels then provides the possibility of using these vitamin D metabolite distributions (“chemotypes”) as complex diagnostic or prognostic biomarkers for correlation with disease phenotype or clinical outcome of treatment. A few examples for such correlations are summarized in the last section.