Analysis of Vitamin D₃-3-Sulfate and 25-Hydroxyvitamin D₃-3-Sulfate in Breastmilk by LC-MS/MS

Abstract

Sulfated metabolites of vitamin D have been suggested to be in breastmilk, although current methods to measure sulfated vitamin D compounds in breastmilk by liquid chromatography-tandem mass spectrometry (LC-MS/MS) have not adequately accounted for increased aqueous solubility of these sulfated metabolites. The purpose of this study was to generate a method of LC-MS/MS for measuring vitamin D₃-3-sulfate (VitD₃-S) and 25-hydroxyvitamin D₃-3-sulfate (25OHD₃-S) specifically in human breastmilk. The resulting method uses methanol to precipitate protein and solid phase extraction to prepare the samples for LC-MS/MS. The limits of quantification for analytes in solvent were 0.23 ng/mL VitD₃-S and 0.2 ng/mL 25OHD₃-S. Various experiments observed concentrations ranging 0.53 to 1.7 ng/mL VitD₃-S and ≤0.29 ng/mL 25OHD₃-S. Both analytes were present in aqueous skim milk, demonstrating the enhanced aqueous solubility of these vitamin D sulfates. In conclusion, we describe an effective method for measuring VitD₃-S and 25OHD₃-S in breastmilk by LC-MS/MS.

1. Introduction

In the first months of life, an infant’s diet is entirely dependent on the nutritional content of breastmilk or formula. Vitamin D metabolites foster normal calcium absorption and bone growth in the infant, but breastmilk is not recognized as a source of vitamin D (VitD, including D₂ and D₃) [1, 2]. Vitamin D deficiency during infancy and childhood impairs calcium homeostasis, hindering bone formation and mineralization to cause rickets can lead to hypocalcemic seizures, respiratory distress, and cardiac arrest [3, 4]. On a global scale, clinical deficiency (serum 25-hydroxyvitamin D <12 ng/mL) in infants remains prevalent, even in regions of abundant sunshine where endogenous vitamin D₃ (VitD₃) can be generated year-round and in societies that provide infant supplementation [5-12].

Vitamin D and its hydroxylated metabolites are present in breastmilk, but the concentrations are generally low. The vitamin D content measured in breastmilk is often less than 0.4 ng/mL (<16 IU/L) and known to increase upon maternal high-dose supplementation of ≥4000 IU VitD/day [6, 9, 13-24]. Concentrations of 25-hydroxyvitamin D (25OHD, including D₂ and D₃) are similar to VitD and contribute more to the total vitamin D activity of the milk than VitD [13-15, 17, 18, 20-26]. 1,25-Dihydroxyvitamin D₂/D₃ is at even lower concentrations in breastmilk and is rarely reported [13, 14, 27]. The combination of VitD and 25OHD in breastmilk provide 5 to 80 IU/L of vitamin D activity to the infant, yet this is too low to meet the recommended intake of 400 IU/day for infants [1, 28-30]. Despite the insufficient vitamin D activity of the milk, the serum 25OHD of a nursing infant often correlates with its mother’s serum 25OHD [9, 31-33]. This correlation suggests that additional vitamin D metabolites may be transferred to the infant via the mother’s milk.

Vitamin D metabolites can undergo conjugation with polar substrates to create modified compounds with enhanced solubility and shielded activity [34-38]. 25-Hydroxyvitamin D₃-3-sulfate (25OHD₃-S) is notably abundant in serum, at concentrations of 14-38 ng/mL or approximately 40% of serum 25OHD concentrations [39-41]. Older reports identified sulfated vitamin D compounds in fresh human milk with concentrations of vitamin D₃-3-sulfate (VitD₃-S) estimated at 10-24 ng/ml (400-950 IU/L) by thin layer chromatography and HPLC [42-44]. At the same time, other groups were unsuccessful at measuring the vitamin D-sulfates and concluded that sulfate conjugates were not present in milk [13, 14, 34, 45, 46]. More recently, a method for LC-MS/MS analysis corroborated the notion that unsubstantial amounts of sulfated vitamin D compounds were in breastmilk, but analyzed only the lipid extract [22]. Here, we present an improved method of LC-MS/MS for measuring sulfated vitamin D metabolites in human breastmilk that adequately reflects the enhanced solubility of these modified products and opens the possibility of additional vitamin D metabolites contributing to the milk composition.

2. Methods

2.1. Breastmilk samples

Breastmilk was collected at Mayo Clinic (Rochester, MN, USA) from participants involved in an ongoing clinical study. This study included lactating mothers, ≥ 2 weeks after delivery. Exclusions were taking daily supplement ≥ 600 IU of vitamin D in the past 30 days, premature birth (< 37 weeks of gestation), and history of sarcoidosis or renal disease. The study was approved by the Mayo Clinic Institutional Review Board and in adherence to the Declaration of Helsinki. Breastmilk was expressed at Mayo Clinic and prepared for analysis or frozen at −80°C within 1 hr. Two pooled samples (<4 participants per pooled sample) and two individual participant samples were collected from a random subset of participants, November 2022 to May 2023, and used for method development and validation.

2.2. Reagents

Vitamin D₃-3-sulfate (MW: 464.3 g/mol) and 25OHD₃-S (MW: 480.3 g/mol; Toronto Research Chemicals, Toronto, CAN), shown in Figure 1, were suspended in methanol (MeOH). 1,25-Dihydroxyvitamin D₃-25-glucuronide (1,25(OH)₂D₃-G; MW= 592 g/mol; GlycoMyr, Inc., Ames, IA) was suspended in MeOH and used as an internal standard (IS). Analyte standards were quantified by ultraviolet absorbance at 264 nm with ε = 18,300 M⁻¹cm⁻¹.

Figure 1:

Chemical structures of vitamin D₃-3-sulfate (A), and 25-hydroxyvitamin D₃-3-sulfate (B).

2.3. Sample preparation

Breastmilk samples were diluted with 4 volumes of MeOH, vortexed, and incubated overnight at −20°C to precipitate proteins. Precipitated protein was removed by centrifugation (1500 x g, 15 min) and the supernatant was collected, purged with nitrogen, and stored at −20°C. Prior to analysis, 5 ng of IS (1,25(OH)₂D₃-G) was added to the sample supernatant and to standards (250 μL) for quantification curve. Samples and standards were evaporated to dryness under nitrogen in a warm water bath (45-60°C), resuspended in 0.4 mL MeOH, and diluted to 2 mL, 20% MeOH solution for loading onto 500 mg C18 solid phase extraction (SPE) cartridges (Agilent, Santa Clara, CA) with positive pressure. After loading, the cartridges were successively washed with 1 mL 20% MeOH, twice, and the analytes were eluted with two successive additions of 1 mL of 100% MeOH. SPE elutions were evaporated to dryness, and then resuspended in 125 μL 60% MeOH for LC-MS/MS.

2.4. Assay conditions for LC-MS/MS

2.4.1. Liquid chromatography

Analytes were separated by Cohesive Technologies LC system (Leap Technologies, Carrboro, NC) in line with mass spectrometry. Aqueous mobile phase (A) contained water modified with 5 mM ammonium acetate and organic mobile phase (B) contained MeOH with the same concentration of modifier. Samples were loaded with 60% B for 1.5 min, eluted with a 5 min gradient to 100% B, and followed by a 3 min wash at 100% B with equilibration at 60% B for 3 min. Analytes were separated by Luna C-18(2), 150 mm x 2 mm, 3 μm (Phenomenex, Torrance, CA), at 60°C with 30 μL injection volume and 0.4 mL/min flow rate.

2.4.2. Mass spectrometry

Liquid chromatography elutions were detected by a SCIEX API 5000 tandem quadripole mass spectrometer (AB Sciex, Framingham, MA). Electrospray ionization (ESI) in negative ion mode was used with a source temperature of 500°C, ion spray voltage of −4500 V, collision gas pressure of 4 bar, curtain gas pressure of 10 bar, and ion source gas pressure of 10 bar. Analytes, 25OHD₃-S and VitD₃-S, and internal standard, 1,25(OH)₂D₃-G, were monitored for quantitation under the parameters listed in Table 1. MRM chromatogram peaks were selected and quantified by ratiometric comparisons of peak areas relative to internal standard in Analyst 1.7.2 quantitation software (AB Sciex, Framingham, MA). Additional analyses and statistics for validation were calculated in GraphPad Prism 9.5.1 (GraphPad Software, Boston, MA).

Table 1:

Mass spectrometry parameters used for analyte detection.

Analyte	Precursor ion (m/z)	Product ions (m/z)	Declustering Potential (V)	Entrance Potential (V)	Collision Energy (V)	Collision Exit Potential (V)
25OHD3-S	479	80, 97*	−340	−12	−95	−14
VitD3-S	463	97*	−365	−12	−88	−31
1,25(OH)2D3-G (Internal Standard)	591	75, 85, 113*	−220	−4	−50	−10

^*Transition used for analysis and quantitation.

2.5. Methods of validation

2.5.1. Quantification analysis

Analytical parameters were determined based on limit of blank (LOB) calculations and calibration standards in 60% MeOH solvent. Concentrations of standards used for the calibration curve were verified by UV absorption spectroscopy before dilution. The calibration curve ranged 0.2-50 ng/mL in 60% MeOH solvent for each analyte and were subjected to SPE as described for sample preparation. The LOB is the average concentration of the solvent matrix plus 1.65 standard deviations, [blank_conc.+(1.65*SD_blank)]. The limit of detection (LOD) is defined as 1.65 standard deviations of the lowest reportable calibrator (0.2 ng/mL) above the LOB, [LOB_conc.+(1.65*SD_cal)]. Limit of quantification (LOQ) is the lowest concentration that is greater than the LOD and maintains a coefficient of variation (CV) of < 20%. Relative error (RE) is determined as the percent ratio of the difference of the measurement from the actual value relative to expected [(measured – expected) /expected *100%]. Extraction efficiency (EE) evaluates the effect of SPE on the recovery of the analytes (in 60% MeOH) by comparing preparations with SPE to those without SPE. EE was calculated as [(with SPE) /(without SPE) *100%] by using measurements for analytes or peak areas for internal standard. Matrix effect (ME) compares analyte quantification in milk to analytes in solvent after SPE. ME was calculated for analytes by determining the observed spike concentration, [(spiked milk)-(unspiked milk)], then comparing the observed spike in milk to the spike in solvent, [(spike in milk) /(spike in solvent) *100%]. For the IS, ME was calculated from the IS peak areas in milk and in solvent, [(IS in milk) /(IS in solvent) *100%].

2.5.2. Accuracy and precision

Accuracy and precision were determined from repeated analyses of a pooled breastmilk specimen with and without spiked analyte standards. Pooled breastmilk was spiked with 25OHD₃-S and VitD₃-S for final concentrations of 4.7 ng/mL with each analyte and compared with baseline aliquots. The breastmilk was vortexed and aliquoted (250 μL) for repetition of methods. CV values were calculated in GraphPad Prism 9.5.1 (GraphPad Software, Boston, MA).

2.6. Solubility experiments

2.6.1. Liquid-liquid extraction

Liquid-liquid extractions (LLE) were completed for breastmilk spiked with 4.7 ng/mL of 25OHD₃-S and VitD₃-S each. Three protocols were compared: (1) 1 mL milk with 6 mL MeOH, 2.5 mL hexane and 1.25 mL water, (2) protein precipitation with 3 volumes acetonitrile (ACN; −20°C, 24hr), followed by 1:1 addition of hexane:methylene chloride (4:1, v/v) solution, and (3) protein precipitation with 4 volumes MeOH (−20°C, 24hr), followed by addition of an equal volume of hexane:methylene chloride (4:1, v/v). After solvents were added, all samples were shaken vigorously, then centrifuged at 1500 x g, 4°C for 15 min. Both organic and aqueous layers were collected separately, IS added to each phase, and then prepared for LC-MS/MS following the protocol described in 2.3. A dilution factor of 0.25x was applied in quantitation for each phase.

2.6.2. Cream separation of breastmilk

Freshly collected breastmilk was separated by gravity (0.5 mL, incubated at 4°C for 2 hrs) or by centrifugation (1 mL, 1500 x g for 15 min at room temperature) to allow milk fat globules to rise and accumulate. A half volume of the lower skim milk layer was compared with the upper half of milk containing cream (cream milk). A needle with syringe was used to aspirate skim milk and transfer to a clean tube. Both skim and cream milk layers were prepared as described above and analyzed by LC-MS/MS following identical methods. A dilution factor of 0.5x was used for quantitation of centrifuged samples (0.5 mL from each layer).

3. Results and Discussion

3.1. Method for measuring sulfated vitamin D compounds in breastmilk

3.1.1. Optimized sample preparation method

Protein precipitation is often the first step in preparing samples for vitamin D metabolite analysis [47]. With breastmilk, a 4:1 ratio of MeOH to milk formed a solid protein precipitate that was easily centrifuged. Numerous standardized vitamin D assays use ACN in 1:1 to 3:1 ratios to serum for effective and quick protein precipitation [22, 39, 40, 48-50], but this method was not effective in protein precipitation with the breastmilk matrix.

Theoretically, the supernatant collected after protein precipitation could be injected into the LC-MS/MS system. Attempts at injecting this supernatant, however, caused accumulation of debris on the inlet cone of the mass spectrometer, thereby disrupting analysis. The debris was suspected to be milk oligosaccharide or other carbohydrates that may be co-extracted in the methanol supernatant. To maintain integrity of the instrument, a C18 SPE (Agilent) was included in the sample preparation protocol. Waters hydrophilic-lipophilic balanced (HLB) and weak-anion exchange (WAX) SPE cartridges were tested but were not used because the HLB cartridge preferentially bound the 25OHD₃-S and the WAX cartridge did not bind analytes.

3.1.2. LC-MS/MS method

Vitamin D₃-sulfate and 25OHD₃-S were separated through a C-18 LC column with an increasing gradient of MeOH organic mobile phase and then injected into the mass spectrometer by electrospray ionization in negative ion mode. Transitions monitored were based on fragmentation of the sulfate group (m/z 97) for both 25OHD₃-S (m/z 479) and VitD₃-S (m/z 463) with retention times at 3.4 and 5.4 min, respectively. The IS eluted before analytes, at 1.9 min. The method presented here agrees with previous LC-MS/MS methods for the same analytes described [22, 39, 40, 50].

3.1.3. Limits of analysis

Analytical limits were determined from standards ranging 0.2 to 50 ng/mL of 25OHD₃-S and VitD₃-S in 60% MeOH solvent and are presented in Table 2. Both milk and charcoal-stripped serum contain analytes of interest, and neither matrices were eligible for use in LOB calculation [50]. The IS used for quantification was 1,25(OH)₂D₃-G, and the absence of 1,25(OH)₂D₃-G in breastmilk was confirmed (not shown). Quantification was optimized to a linear regression weighted by 1/x. Determination coefficients of the linear regression fit were acceptable for both analytes (r² > 0.995). The LODs in 60% MeOH were determined to be 0.056 ng/mL 25OHD₃-S and 0.11 ng/mL VitD₃-S. LOQs in solvent matrix were 0.2 ng/mL 25OHD₃-S and 0.23 ng/mL VitD₃-S.

Table 2:

Analytical limits of calibration curve in solvent matrix^a.

	25OHD3-S	VitD3-S
Linearityb (r2)	≥ 0.9958	≥ 0.9953
LOB (n=16)	0.008 ng/mL	0.041 ng/mL
LOD (n=9)	0.056 ng/mL	0.11 ng/mL
LOQ (n=9)	0.2 ng/mL	0.23 ng/mL
LOQ %CV (n=9)	13.5%	19%

Abbreviations: LOB, limit of blank; LOD, limit of detection; LOQ, lower limit of quantification.

^aAnalytes were suspended in 60% MeOH for measurement.

^bCalibration curve was fit to 1/x weighted linear regression.

Accuracy and extraction efficiency for 25OHD₃-S and VitD₃-S in 60% MeOH solvent matrix are described in Table 3. Concentrations included low (1 ng/mL), medium (5 ng/mL), and high (20 ng/mL) values for each analyte. CVs were <13% and relative errors deviated from −12% to 7%. SPE affected analyte abundance such that extractions for 25OHD₃-S (66-77%) were more efficient than for VitD₃-S (47-59%).

Table 3:

Accuracy and extraction efficiency of analyte standards in solvent matrix^a.

Analytea	Measured (ng/mL)	%CVb	REc (%)	EEd (%)
25OHD3-S
1 ng/mL	0.90 ± 0.12	13	−12	66 ± 16
5 ng/mL	4.84 ± 0.4	8.2	−3.2	67 ± 12
20 ng/mL	20.6 ± 1.5	7.5	3	77 ± 15
VitD3-S
1 ng/mL	0.95 ± 0.05	5.7	−5	47 ± 15
5 ng/mL	4.78 ± 0.47	9.8	−4.4	48 ± 12
20 ng/mL	21.35 ± 0.93	4.3	7	59 ± 19

Statistical data are mean ± SD, n=4.

^aAnalytes were suspended in 60% MeOH for measurement.

^bInter-assay coefficient of variation (CV), n=4.

^cRelative error (RE) is [(measured - theoretical) /theoretical]*100%.

^dExtraction efficiency (EE) is [(with extraction) /( without extraction)]*100%.

3.1.4. Precision and accuracy of measurements in breastmilk

The LC-MS/MS method was assessed by repeated measurements of 250 μL for (1) spike analytes in blank solvent, (2) baseline breastmilk, and (3) spiked breastmilk. All breastmilk preparations were derived from the same pooled collection of breastmilk. LC-MS/MS chromatograms depict 25OHD₃-S and VitD₃-S peaks in breastmilk consistently overlayed with the respective analyte standard at retention times of approximately 3.4 and 5.4 minutes, respectively (Figure 2).

Figure 2:

MRM chromatograms of 25OHD₃-S (left) and VitD₃-S (right) from standards (A,B) and breastmilk (C,D). Chromatograms depict standards containing 5 ng/mL 25OHD₃-S (A) and VitD₃-S (B), and 250 μL unspiked human breastmilk (C,D). Transitions include 25OHD₃-S, m/z 479-97, and VitD₃-S, m/z 463-97.

Accuracy and precision of the analyte quantification in solvent and in breastmilk are presented in Table 4. In the solvent matrix, the spike had low variability for intra-assay (CV ≤ 5%) and inter-assay (CV ≤ 7.2%) analyses and the quantification of the 25OHD₃-S spike had low relative error (RE = −0.4%) with extraction efficiency of 67 ± 11%. Analysis of the VitD₃-S spike in solvent resulted in relative error of 14% and extraction efficiency of 59 ± 8%. The IS in solvent had similar extraction efficiency of 60 ± 6% with slightly higher %CVs (≤ 20%) that did not affect the %CV of the analytes of interest.

Table 4:

Recovery of Analytes in 250 uL Breastmilk

Analysis	Theoreticala ng/mL	Measured ng/mL (n)	Intra-assay %CVb (n)	Inter-assay %CVb (n)	REc (%)	EEd(%)	MEe (%)
25OHD3-S
Spike, solvent	4.7	4.68 ± 0.34 (6)	3 (3)	7.2 (6)	−0.4	67 ± 11	-
Baseline Breastmilk	-	<0.2	<LOQ	<LOQ	-	-	-
Spiked Breastmilk	4.84	6.81 ± 0.75 (15)	6.4 (9)	11.0 (15)	41	-	139 ± 17
VitD3-S
Spike, solvent	4.7	5.35 ± 0.3 (6)	5.0 (3)	5.4 (6)	14	59 ± 8	-
Baseline Breastmilk	-	1.70 ± 0.23 (15)	8.8 (9)	13.5 (15)	-	-	-
Spiked Breastmilk	6.40	6.87 ± 0.92 (15)	9.3 (9)	13.5 (15)	7.3	-	94 ± 20
IS [1,25(OH)2D3-G]
Solvent	-	-	15.7 (n=4)	20.4 (12)	-	60 ± 6	-
Breastmilk	-	-	7.5 (n=9)	8.4 (17)	-	-	169 ± 22

Statistical data are means ± SD. Replicates are derived from a common pooled sample.

^aTheoretical concentration is baseline measurement + 4.7 ng/mL of analyte.

^bCoefficient of variation (CV).

^cRelative error (RE) is [(measured - theoretical) /theoretical]*100%.

^dExtraction efficiency (EE) is [(with extraction) /(without extraction)]*100%.

^eMatrix effect (ME) is ([spike]_breastmilk / [spike]_solvent)*100%.

The pooled breastmilk analysis, without spike and without dilution factors, had quantifiable VitD₃-S and concentrations of 25OHD₃-S near the LOD. Breastmilk contained 1.70 ± 0.23 ng/mL VitD₃-S with CVs ≤ 13.5%. 25OHD₃-S was detected in 7 of 15 repeated measurements, but all remained below the LOQ. When spiked with 4.7 ng/mL of each analyte, a matrix effect of 139 ± 17% was observed for the quantification of 25OHD₃-S, but only 94 ± 20% for VitD₃-S. As a result, 25OHD₃-S was quantified with high relative error (RE = 41%), despite acceptable variability (CVs ≤ 11%). Quantification of VitD₃-S in the spiked breastmilk had low error (RE =7.3%) with acceptable reproducibility for intra-assay (CV = 9.3%) and inter-assay (CV = 13.5%) measurements. A matrix effect was also observed with the IS, such that the peak areas in breastmilk samples were approximately 169 ± 22% of the solvent standards. Moreover, the peak areas of the IS in milk were comparable to the peak areas of analytes in solvent that did not undergo SPE. This observed matrix effect has been previously described by Gao and colleagues when measuring derivatized sulfated vitamin D compounds in serum [39]. Here, the prominent matrix effect with the IS may shift the quantitation by inadvertently decreasing the calculated concentration. Inclusion of deuterated standards would eliminate the matrix effect and provide a more straight-forward quantification of the vitamin D sulfated analytes using the methods described.

3.2. Solubility of sulfated vitamin D compounds

3.2.1. Comparison of organic vs. aqueous extracts

As cholesterol-derived molecules, the unmodified vitamin D metabolites are lipid-soluble, and, in blood, require a carrier protein such as vitamin D binding protein, albumin, or lipoproteins. Protocols for measuring VitD or 25OHD in milk often use LLE to isolate the lipid-soluble analytes and analyze only the organic phase [15, 20, 24, 27, 51-53]. We tested three methods of LLE on a pooled breastmilk sample spiked with 4.7 ng/mL of each analyte and analyzed both organic and aqueous extraction layers to determine the phase partitioning of the analytes. MRM chromatograms revealed an abundance of 25OHD₃-S and VitD₃-S in the aqueous extract, but little in the organic extract (Figure S1). Internal standards (added to each extract layer after LLE) were recovered, confirming equal sample preparation after phase separation. In quantification, it was determined that over 90% of the analytes were collected from the aqueous phase with 3-5% appearing in the organic extract (Table 5). VitD₃-S measurements were more consistent across methods than 25OHD₃-S, yielding a mean concentration of 6.1 ± 0.6 ng/mL with a calculated contribution from the breastmilk of 1.4 ± 0.6 ng/mL VitD₃-S.

Table 5:

Liquid-liquid extraction of spiked breastmilk

Extraction Method	Aqueous Extract ng/mL	REe (%)	Organic Extract ng/mL	REe (%)
Method 1a
25OHD3-S	3.35	−7.4	Not Detected	−100
VitD3-S	6.3	9.6	0.44	−92
Percentd	97%	-	3%	-
Method 2b
25OHD3-S	10.2	182	0.13	−96
VitD3-S	6.5	13	0.53	−91
Percentd	96%	-	4%	-
Method 3c
25OHD3-S	3.6	−0.5	<0.2	−99
VitD3-S	5.4	−6	0.14	−98
Percentd	97%	-	3%	-

^aSeparation by hexane:methanol:water.

^bAcetonitrile precipitation and separation by hexane:methylene chloride.

^cMethanol precipitation and separation by hexane:methylene chloride.

^dPercent abundance is averaged from both analytes.

^eRelative error (RE) is [(measured - theoretical) /theoretical]*100% with theoretical values of 3.62 ng/mL 25OHD₃-S and 5.75 ng/mL VitD₃-S

The presence of sulfated vitamin D compounds in the aqueous extract exposes a major caveat in the analysis of sulfated metabolites compared to the unmodified metabolites. In a similar study measuring breastmilk sulfated vitamin D compounds, Gomes, et al., followed the method for unmodified metabolites by using ACN precipitation and LLE, and then analyzing the organic layer [22]. Even with sample volumes of 4 mL breastmilk, only 90 pmol (36 pg/mL) 25OHD₃-S and 260 pmol (104 pg/mL) VitD₃-S were measured. In the present work, LLE Method 2 repeats the protocol used by Gomes, et al., with 1 mL of milk. The concentrations observed in the organic extract are similar to their assessment; however, the organic extract accounted for only 3% of the total spiked analyte in the sample (Table 5). Interestingly, this method had a high rate of quantification error which may be an artifact of matrix effects. These data imply that approaches to preparation of these sulfate-modified molecules differ from the unmodified 25OHD and VitD metabolites and are important for optimized analyses.

3.2.2. Comparison of skim milk vs. cream milk

Breastmilk stored at room temperature or in refrigeration naturally forms a cream layer as the milk fat globules rise to the top of the milk. Here, gravity and centrifugation methods were used to allow the milk fats to accumulate. In order to minimize disproportions of the milk layers, 50% of the sample was aspirated from the skim milk layer, leaving the remaining half volume containing both cream and milk for comparison. Chromatograms of the milk detected peaks for both 25OHD₃-S and VitD₃-S in the cream milk and revealed peaks of similar size in the skim milk (Figure S2). In the sample separated by gravity, the skim milk portion contained 38% of the analytes with 0.5 ng/mL VitD₃-S and <0.2 ng/mL 25OHD₃-S (approximated at 0.1 ng/mL; Table 6). The centrifuged sample yielded higher concentrations in the skim milk with 0.33 ng/mL 25OHD₃-S and 0.76 ng/mL VitD₃-S compared to mean compositions of 0.29 ± 0.06 ng/mL 25OHD₃-S and 0.53 ± 0.33 ng/mL VitD₃-S, and accounted for 65% of the total analytes in the sample. These data indicate that the 25OHD₃-S and VitD₃-S are soluble in the skim milk and not confined to milk fat globules.

Table 6:

Separation of Milk from individuals

Separation Method	Skim Milk	Cream Milk	Sample Meana
Gravity
Volume (μL)	250	250	-
25OHD3-S (ng/mL)	<LOQ (0.10b)	<LOQ (0.14b)	<LOQ (0.12 ± 0.02b))
VitD3-S (ng/mL)	0.5	0.97	0.74 ± 0.33
Percentc	38%	62%	-
Centrifuge
Volume (μL)	500	500	-
25OHD3-S (ng/mL)	0.33	0.24	0.29 ± 0.06
VitD3-S (ng/mL)	0.76	0.29	0.53 ± 0.33
Percentc	65%	35%	-

^aSample means ± SD combine skim and cream milk measurements.

^bEstimated concentration.

^cPercent abundance averaged of both analytes.

Modification of vitamin D metabolites with the highly polar sulfate increases solubility for enhanced distribution in the mixed matrix of breastmilk. This distribution poses an important consideration for digestion as it becomes plausible that absorption may occur via alternate pathways rather than from chylomicron transport. Additional research on digestion and absorption of these soluble vitamin D conjugates in infants is needed to understand their role in nutrition.

3.3. VitD₃-S and 25OHD₃-S in breastmilk

In human breastmilk, VitD₃-S is more abundant than 25OHD₃-S and can be measured consistently in 250 μL of milk sample. In our assessments we found that (1) method validation of a pooled milk sample contained 1.7 ± 0.23 ng/mL VitD₃-S, (2) liquid extractions of a second pooled sample contained 1.4 ± 0.6 ng/mL VitD₃-S in the aqueous phase, and (3) cream separations of two individual milk samples contained 0.74 ± 0.33 ng/mL and 0.53 ± 0.33 ng/mL VitD₃-S. Taken together, the concentration range of VitD₃-S in milk was observed at 0.53-1.7 ng/mL VitD₃-S.

25OHD₃-S was also present in pooled and individual breastmilk samples at concentrations up to 0.29 ng/mL. Quantification was achievable in experiments that used sample volumes ≥500 μL; however, experiments that used 250 μL yielded detectable concentrations with estimated concentrations ranging 0.05-0.2 ng/mL 25OHD₃-S. From these experiments, it is recommended that if 25OHD₃-S in breastmilk is an analyte of interest, then a sample volume of ≥500 μL will provide more reliable quantification.

The abundances of the sulfated metabolites resemble concentrations previously described for unmodified metabolites. Human breastmilk contains <0.4 ng/mL VitD and 0.1-0.6 ng/mL 25OHD [6, 9, 13-24, 26, 54]. In the samples assessed, VitD₃-S exceeded the known concentrations for VitD, and 25OHD₃-S was near the expected range for 25OHD. These comparisons reveal that sulfated vitamin D metabolites indeed contribute to the total vitamin D content of the breastmilk.

While similarly abundant to VitD and 25OHD, the activity and effectiveness of sulfated metabolites in the infant have yet to be determined. In past studies, the role of conjugated vitamin D metabolites were unknown and considered inactive waste products [45, 46, 55]. Since then, conjugate hydrolysis has been found to enable normal activity [37, 38, 56-58]. Currently, it is uncertain where, in an infant, sulfatase reactions occur to liberate the vitamin D parent metabolite. If 25OHD-S is cleaved within the digesta, for example, then the resulting 25OHD would be in a free form and may act as a hormonal agonist [58-60]. Internally, the sulfated metabolites may provide a reservoir for vitamin D and 25OHD or target specific cells that express sulfatase enzymes [61-64]. Understanding the metabolic activity of these sulfate conjugates is needed to determine the total nutritive value of sulfated vitamin D metabolites in human breastmilk.

3.4. Limitations

The method described here is subject to various limitations. First, the internal standard differs from the analytes of interest. While previously commercially available, deuterated standards for the sulfated compounds were not readily accessible at the time of this study, therefore, we identified a similar conjugated metabolite that is unlikely to be present in milk. Use of deuterated standards will provide more detail as to the recovery of analytes and will enhance quantitation, particularly because we observe a larger matrix effect with the IS than with either of the analytes. Second, a blank matrix has not been established for vitamin D sulfated metabolites in milk or for serum, as charcoal-stripped serum contains abundant 25OHD₃-S. Using a similar matrix rather than solvent may improve precision and accuracy of the assay quantification. Finally, the number of samples analyzed is low and, therefore, we have not made generalized statements of population-based concentrations for 25OHD₃-S and VitD₃-S in breastmilk.

4. Conclusions

The present study describes an improved LC-MS/MS method for measuring VitD₃-S and 25OHD₃-S in human breastmilk and highlights the enhanced aqueous solubility of the sulfated metabolites. This method uses MeOH for protein precipitation followed by SPE in the sample preparation protocol. In an evaluation of LLE sample preparations, both sulfated analytes were found to partition to the aqueous phase of the extraction rather than the organic phase where unmodified vitamin D metabolites are known to appear. Chromatography and mass spectrometry conditions agree with previous reports for fragmentation of both metabolites at m/z 97 [22, 40, 50]. The LOQs for the method in solvent matrix are 0.2 ng/mL 25OHD₃-S and 0.23 ng/mL VitD₃-S. VitD₃-S was readily quantified in 250 μL sample volumes, whereas 25OHD₃-S is detectable in 250 μL and quantifiable with ≥500 μL sample volumes. Cream and skim milk separation experiments further revealed that the sulfated analytes appear in skim milk and suggest that the analytes are not entirely contained in the milk fat layer.

The sample preparation optimized for breastmilk has enabled more robust assessments of VitD₃-S and 25OHD₃-S by LC-MS/MS than previous efforts. The concentrations of VitD₃-S observed in various experiments with pooled and individual milk samples ranged from 0.53 ng/mL to 1.7 ng/mL VitD₃-S. 25OHD₃-S was less abundant than VitD₃-S, ranging from detectable (0.056-0.2 ng/mL 25OHD₃-S) to 0.29 ng/mL 25OHD₃-S. These findings are approximately 10-times greater than previously described LC-MS/MS methods and correspond to the concentration ranges described for unmodified vitamin D metabolites in human breastmilk [9, 13-15, 17, 18, 20-24, 26, 54]. The contribution of sulfated vitamin D compounds to human breastmilk provides insight for improving our understanding of infant vitamin D nutrition through maternal breastfeeding.

Highlights:

• LC-MS/MS methods for vitamin D₃-sulfate and 25OHD₃-sulfate in breastmilk.

• Vitamin D₃-sulfate is more abundant in breastmilk than 25OHD₃-sulfate.

• Vitamin D sulfates are soluble in the aqueous portion of breastmilk.

Abbreviations:

1,25(OH)₂D-G

1,25-dihydroxyvitamin D₃-25-glucuronide

25OHD

25-hydroxyvitamin D₂ and 25-hydroxyvitamin D₃

25OHD₃-S

25-hydroxyvitamin D₃-3-sulfate

ACN

acetonitrile

CV

coefficient of variation

IS

internal standard

LOB

limit of blank

LOD

limit of detection

LOQ

limit of quantification

LC-MS/MS

liquid chromatography tandem mass spectrometry

LLE

liquid-liquid extraction

MeOH

methanol

MRM

multiple reaction monitoring

SPE

solid phase extraction

VitD

vitamins D₂ and D₃

VitD₃-S

vitamin D₃-3-sulfate

Data Availability:

Data will be made available on request.