Validation and Application of Biocrates AbsoluteIDQ® p180 Targeted Metabolomics Kit Using Human Milk

Human-milk-targeted metabolomics analysis offers novel insights into milk composition and relationships with maternal and infant phenotypes and nutritional status. The Biocrates AbsoluteIDQ® p180 kit, targeting 40 acylcarnitines, 42 amino acids/biogenic amines, 91 phospholipids, 15 sphingolipids, and sum of hexoses, was evaluated for human milk using the AB Sciex 5500 QTRAP mass-spectrometer in liquid chromatography-tandem mass-spectrometry (LC-MS/MS) and flow-injection analysis (FIA) mode. Milk (<6 months lactation) from (A) Bangladeshi apparently healthy mothers (body mass index (BMI) > 18.5; n = 12) and (B) Bangladeshi mothers of stunted infants (height-for-age Z (HAZ)-score <−2; n = 13) was analyzed. Overall, 123 of the possible 188 metabolites were detected in milk. New internal standards and adjusted calibrator levels were used for improved precision and concentration ranges for milk metabolites. Recoveries ranged between 43% and 120% (coefficient of variation (CV): 2.4%–24.1%, 6 replicates). Milk consumed by stunted infants vs. that from mothers with BMI > 18.5 was lower in 6 amino acids/biogenic amines but higher in isovalerylcarnitine, two phospholipids, and one sphingomyelin (p < 0.05 for all). Associations between milk metabolites differed between groups. The AbsoluteIDQ® p180 kit is a rapid analysis tool suitable for human milk analysis and reduces analytical bias by allowing the same technique for different specimens. More research is needed to examine milk metabolite relationships with maternal and infant phenotypes.


Introduction
The postnatal period is a critical stage for infant's physiology, accompanied by rapid changes in brain development, and in metabolic, immunologic, intestinal, and physiological systems [1]. Human milk undergoes dynamic changes in composition during lactation to provide all nutrients and other bioactive substrates for optimal infant growth and development, and exclusive breastfeeding (EBF) is recommended during the first 6 months of life [2]. Even beyond the EBF period, it remains a vital dietary source of nutrients; however, the mechanisms behind its ability to drive and affect infant metabolism, health, development, and long-term outcomes are not fully understood. The recent resurgent interest in human milk composition triggered the development of state-of-the-art methods

AbsoluteIDQ ® p180 Assay and Sample Preparation
The AbsoluteIDQ ® p180 kit is a fully automated assay based on phenylisothiocyanate (PITC) derivatization of the target analytes in bodily fluids (e.g., plasma, serum, urine) using internal standards for quantitation. Amino acids and biogenic amines are determined in LC-MS mode, acylcarnitines, phospholipids (lyso-phosphatidylcholines with acyl residue at CXX:X, phosphatidylcholine with diacyl residue sum CXX:X (PC aa), and phosphatidylcholine with acyl-alkyl residue sum CXX:X (PC ae)), sphingomyelins, and the sum of hexoses are analyzed in flow injection analysis (FIA). Human milk sample preparation was carried out according to the manufacturer's protocol. Briefly, 2 to 10 µL of human milk was transferred to the upper 96-well plate and dried under a nitrogen stream. Thereafter, 50 µL of a 5% PITC solution was added to derivatize amino acids and biogenic amines. After incubation, the filter spots were dried again before the metabolites were extracted using 5 mM ammonium acetate in methanol (300 µL) into the lower 96-well plate for analysis after further dilution using the MS running solvent A. Quantification was carried out using internal standards and a calibration curve (Cal 1 to Cal 7). The experimental metabolomics measurement technique is described in detail by Nutrients 2019, 11, 1733 3 of 10 patents EP1897014B1 and EP1875401B1 [25,26]. See Supplementary Materials (Table S1) for the full list of metabolites and their abbreviations.

LC-MS
The LC-MS/MS system was comprised of an ACQUITY UPLC-system (Waters, Milford, MA, USA) coupled to a QTRAP 5500 mass spectrometer (AB Sciex, Redwood City, CA, USA) in electrospray ionization (ESI) mode. Amino acids and biogenic amines were analyzed via LC-MS in positive mode. Two microliters of the sample extract were injected onto an ACQUITY UPLC BEH C18 column, 2.1 × 7.5 mm, 1.7 µm protected by an ACQUITY BEH C18, 1.7 µm VanGuard pre-column (Waters, Milford, MA, USA) at 50 • C using a 7.3 min solvent gradient employing 0.2% formic acid in water (solvent A) and 0.2% formic acid in acetonitrile (solvent B).
Twenty microliters of the sample extract were used in the flow injection analysis (FIA) in positive mode to capture acylcarnitines, glycerophospholipids, and sphingolipids, while hexoses were monitored in a subsequent run in negative mode. All FIA injections were carried out using the Biocrates MS Running Solvent. Additional LC and MS settings for LC-MS and FIA mode are described in Table 1. All metabolites were identified and quantified using isotopically-labeled internal standards and multiple reaction monitoring (MRM) as optimized and provided by Biocrates Life Sciences AG (Innsbruck, Austria).

Method Validation
Pooled milk was used to evaluate matrix effects and metabolite recovery. The samples were diluted in phosphate buffered saline solution (PBS) 1:2 and 1:5 to assess ion suppression due to the matrix. A 3-level analyte recovery was carried out using Cal 2, 3, and 5 in half, normal, and double concentrations, covering up to 230-fold of the endogenous milk concentrations. All experiments were conducted in replicates of six. Individual milk samples (n = 25) from Bangladeshi women were Analyte recovery was calculated using the following equation: where C measured is the measured concentrations of the spiked sample, C endogenous the measured concentration of the non-spiked sample, and C added the theoretically added concentration by spiking. Relative recovery in the diluted samples was calculated as follows: Relative recovery [%] = C X/dF × dF × 100/C X (2) where C X/dF is the concentration measured in the diluted sample, dF the dilution factor (2 or 5), and C X the concentrations measured in the non-diluted milk samples.

Human Milk Samples
To determine recovery rates and evaluate matrix effects, a human milk pool, obtained from milk samples provided by apparently healthy women in the Vancouver, BC, Canada area, was used. Individual milk samples, available in the laboratory but not collected for the purpose of this validation, were analyzed in the feasibility plate. These convenient samples were collected during the first 6 months of lactation from apparently healthy Bangladeshi lactating women ("BMI > 18.5 mothers", body mass index (BMI) > 18.5, n = 12) whose infants' anthropometry was unknown, and Bangladeshi mothers with stunted infants ("stunted infants", HAZ-score < −2, n = 13), to try to obtain a wide range of metabolite concentrations in the milk. All milk samples were shipped from the United States Department of Agriculture -Agricultural Research Service (USDA/ARS) Western Human Nutrition Research Center in Davis, CA, USA on dry ice to Biocrates Life Science AG in Innsbruck, Austria for testing. Upon arrival, samples were stored at −80 • C until analysis.

Statistical Analysis
Raw data was computed in MetIDQ TM version Carbon (Biocrates Life Science AG, Innsbruck, Austria). Mean, standard deviation, and coefficient of variation (CV) for the validation were calculated in Excel 2016 (Microsoft Corporation, Redmond, WA, USA). R statistical software (3.5.2, R Foundation for Statistical Computing) was used for statistical analysis and visualization of the results for the feasibility study. The Wilcoxon rank sum test was used to explore differences in analyte concentrations based on group assignment (BMI > 18.5 mothers vs. stunted infants), and metabolite associations were assessed using Spearman's rank correlation. p-values < 0.05 were considered significant.

Feasibility Plate
Across all analyzed milk samples (BMI > 18.5 mothers and stunted infants), all 21 amino acids, 15 sphingomyelins, and the sum of hexoses were quantifiable. In addition, 10 biogenic amines, 11 acylcarnitines, and 54 phospholipids were observed, as shown in Supplementary Table S6.

Feasibility Plate
Across all analyzed milk samples (BMI > 18.5 mothers and stunted infants), all 21 amino acids, 15 sphingomyelins, and the sum of hexoses were quantifiable. In addition, 10 biogenic amines, 11 acylcarnitines, and 54 phospholipids were observed, as shown in Supplementary Table S6.
Milk from BMI > 18.5 mothers had significantly higher concentrations of the amino acids and biogenic amines citrulline, glutamate, glycine, phenylalanine, serine, and the biogenic amine sarcosine (all p < 0.045), and lower concentrations of isovalerylcarnitine, phosphatidylcholine with diacyl residue sum C36:6 (PC aa), phosphatidylcholine with acyl-alkyl residue sum C30:2 (PC ae), and sphingomyelin with acyl residue sum C22:3 (all p < 0.048), as shown in Table 2, when compared to milk fed to stunted infants. While milk metabolites were significantly associated (p < 0.05) in both groups, the correlation profile differed considerably. These significant relationships were found not only among metabolites that were significantly different in concentration between the two groups but also among metabolites that had similar concentrations in all milk samples across groups, as shown in Figure 2.  While milk metabolites were significantly associated (p < 0.05) in both groups, the correlation profile differed considerably. These significant relationships were found not only among metabolites that were significantly different in concentration between the two groups but also among metabolites that had similar concentrations in all milk samples across groups, as shown in Figure 2.

Validation Using Pooled Human Milk
LC-MS: The new internal standards for amino acids and biogenic amines, and the extension of the calibrator curve allowed a more accurate analysis of the milk metabolites. While proline, tryptophan, and valine revealed saturation effects at mid calibrator levels, the measured milk concentrations of these amino acids, however, were well within the linear calibrator curve and the high level calibrators were not needed. While the low range of the calibrator curve was also extended to Cal 0.5 and 0.25, alanine, spermidine, spermine, and ADMA were not quantifiable at the lowest calibrator. Given that these metabolites were present in milk above Cal 0.5, the low accuracy at Cal 0.25 concentrations was not problematic. Some of the amino acids and biogenic amines, including ornithine, were found already in very low concentrations in the non-diluted pooled milk sample; consequently, their concentrations in the diluted samples were around or below the limit of detection (LOD) and reliable data could not be obtained. For metabolites within the calibrator curve, the diluted samples were mostly within 80%-120%, indicating some but not serious effects for quantification due to the matrix. The lower ion spray voltage (IS) intensity found for milk samples when compared to plasma further confirmed effects due to the milk matrix. To reduce matrix interferences due to early eluting matrix components (e.g., lactose), results for creatinine, the first eluting metabolite in LC-MS mode could be sacrificed to maintain MS-integrity and accuracy over time.
FIA: Phospholipids showed the greatest variation in metabolite recovery, and also represented the largest compound class of metabolites analyzed. Some of the variation could be explained by low abundance of metabolites at the edges of the calibrator range, or standard additions possibly to levels of saturation and therefore loss of linearity and accuracy. However, selectivity issues by milk-specific interferences cannot be excluded. No extended calibrator curve was necessary in FIA mode. The relative concentration of the analytes in the diluted samples compared to the non-diluted samples of generally above 100%, and often above 150% at higher dilutions, suggested some ion suppression due to the matrix which can be overcome by sample dilution. However, no metabolite was affected by the matrix to a degree that would not allow quantitation. However, given the abundance of lactose in the sample (~7%), a greater dilution would be preferable to reduce contamination of the MS by matrix constituents in FIA mode.
We found all detectable amino acids, sphingomyelins, and the sum of hexoses in human milk, but only 25% of the acylcarnitines, 50% of the biogenic amines, and 68% of the phospholipids. A recent study on plasma metabolites of healthy men and women (18-80 years) using the AbsoluteIDQ ® p180 kit reported fewer of the amino acids, biogenic amines, acylcarnitines, and sphingomyelins in plasma, but more phospholipids [27], illustrating the variations in metabolite profile depending on specimen analyzed. Moreover, that study reported that the metabolic profile of plasma and urine can predict sex with >90% accuracy, as well as age (based on gender), and within females the menopausal status could be predicted with >80% accuracy. Other studies showed that differences in plasma metabolite profiles could be a useful diagnostic tool for Alzheimer's disease and cognitive impairment, or type 2 diabetes [28,29], further indicating the importance and relevance of the information than can be derived from the milk metabolome to the phenotype and status [27].

Differences by Group Assignment
Overall, the metabolites detected in the feasibility plate using milk from Bangladeshi women were similar to those in the pooled human milk sample. The observed higher concentrations in milk from BMI > 18.5 mothers for selected amino amides and biogenic amines and lower concentrations of phospholipids, acylcarnitines, and sphingomyelins are consistent with alterations found in the serum metabolome of stunted compared to non-stunted children (1 to 5 years of age) in Malawi by Semba et al., using the same AbsoluteIDQ ® p180 kit [30]. Stunting can be caused by intrauterine growth retardation, inadequate nutrition for optimal infant growth and development, or repeated infections early in life, and usually originates in utero [31]. Thus, the reported low circulating amino acids and other metabolome alterations in the blood of stunted children in Malawi may stem from challenges in utero and early infancy, including possible effects derived from the lower supply of amino acids in human milk as reported here. This hypothesis is supported by the fact that human growth is regulated by the mechanistic target of rapamycin complex C1 (mTORC1), a major growth regulating pathway, which suppresses protein and lipid synthesis and cellular growth when intake of specific amino acids is insufficient [30,32]. However, since we do not have the full information about maternal and infant characteristics, including exact stage of lactation for each sample, we cannot exclude that additional maternal or infant factors possibly affect the milk metabolome. Nonetheless, correlation patterns changed in milk fed to stunted infants, indicating alterations of the milk metabolome and its pattern could be related to the infant phenotype. Since the alterations in the metabolome correlation profile were not limited to metabolites showing significant differences in concentrations, these relationships may be also affected by maternal or infant status. However, our sample size was very small, the anthropometric status of infants in the group of BMI > 18.5 mothers was unknown, and further research is needed to confirm and explore the results found in this feasibility study.

Conclusions
This is the first study to validate the commercially available AbsoluteIDQ ® p180 assay for targeted metabolomics in the human milk matrix. A greater range of calibrators is needed to accommodate the wide range of milk metabolite concentrations; new available internal standards improved accuracy of the results. While the primary purpose of this study was the validation of the AbsoluteIDQ ® p180 assay for the human milk matrix, and it was not our intention to draw conclusions about group differences in metabolites from the feasibility plate, the observed differences in concentrations and metabolome profiles in the two groups warrant further investigation of the mother-infant relationship and the importance of milk metabolome for infant optimal growth and development, which is now possible using identical targeted metabolomics assays for blood and milk samples.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2072-6643/11/8/1733/s1, Table S1: List and abbreviations of metabolites, Table S2: Metabolite concentrations (mean ± SD) and recovery rates (CV) of pooled milk in LC-MS mode, Table S3: Relative recovery (mean ± SD) of milk metabolites in diluted pooled human milk based on non-diluted samples in LC-MS mode, Table S4: Metabolite concentrations (mean ± SD) and recovery rates (CV) of pooled milk in FIA mode, Table S5: Relative recovery (mean ± SD) of milk metabolites in diluted pooled human milk based on non-diluted samples in FIA mode, Table S6: Median concentrations and interquartile ranges (IQR) of milk metabolites in the BMI > 18.5 mothers and the stunted infants groups.