REFINING SIMULTANEOUS MEASUREMENT OF VITAMIN D AND OTHER SECOSTEROIDS IN HUMAN SERUM By Michael Kaven A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Biosystems Engineering – Master of Science 2023 ABSTRACT “Vitamin D” refers to a group of closely related secosteroid compounds with demonstrated roles in bone and immune health in humans. Its deficiency has been causally linked to a range of adverse human health outcomes including renal failure and rickets. A bioactive metabolite of the vitamin known as calcitriol (1,25-dihydroxycholecalciferol) is hypothesized to enhance human neurology and cognition but rarely quantified simultaneously with other vitamin D metabolites, steroids, or hormones. The aim of this study was to develop and validate a simple, effective, and reliable method to measure major secosteroids and vitamin D metabolites in human serum with precision, accuracy, and sensitivity. A dynamic extraction method was designed with focus on six common vitamin D metabolites circulating in serum: cholecalciferol, calcidiol, and calcitriol from the D 3 sub-family; and calciferol, ercalcidiol, and ercalcitriol from the D 2. Analysis began with a test run to validate the novel method. Then, fifty (50) human serum biospecimens were obtained from adult Ugandans with and without chronic HIV infection. Thirty-two (32) species of vitamin D metabolites, other steroids, phytosterols, oxysterols, and hormones were identified and annotated by pure external standards. Relationships including serum lathosterol to cholesterol and calcidiol to ercalcidiol ratios were measured for preliminary comparison of groups by HIV and established for future study with an enlarged sample size. A goal for future applications is to identify associations between calcitriol bioavailability and diagnosed cognitive disorders or chronic diseases in longitudinal sampling. Keywords: lipidomics, metabolite, calcitriol, human immunodeficiency virus (HIV), neurodegenerative disease, cytochrome P450 enzyme Copyright by MICHAEL KAVEN 2023 ACKNOWLEDGEMENTS I direct many thanks to my advisor Dr. Ilce Gabriella Medina-Meza 1,2, supportive colleagues at the Food and Health Engineering Laboratory, professor Dr. Yan Liu 1 and chairperson Dr. Bradley Marks of the Department of Biosystems and Agricultural Engineering at Michigan State University (MSU). My work was strengthened by conceptual and material resources of the International Neurologic and Psychiatric Epidemiology Program (INPEP) at MSU, especially those of advisor Dr. Amara Ezeamama3,4. I also give thanks for the very special people I am fortunate to call family. Their sincere support year-round, coupled with their soothing Upper Peninsula hospitality during times of rest, secured my balance in and visions for vitamin D research. 1 Department of Biosystems and Agricultural Engineering, Michigan State University, East Lansing, MI, USA. 2 Department of Chemical Engineering and Materials Science, Michigan State University, East Lansing, MI, USA. 3 Department of Epidemiology and Biostatistics, Michigan State University, East Lansing, MI, USA. 4 Department of Psychiatry, Michigan State University, East Lansing, MI, USA. iv TABLE OF CONTENTS LIST OF TABLES ....................................................................................................................... vi LIST OF FIGURES .................................................................................................................... vii LIST OF ABBREVIATIONS ................................................................................................... viii CHAPTER 1: INTRODUCTION ................................................................................................ 1 Vitamin D: The Nutrient ............................................................................................................. 1 Metabolite Bioavailability .......................................................................................................... 5 Vitamin D Deficiency ................................................................................................................. 7 Synapse Chemistry and Vitamin D ............................................................................................. 9 Metabolomics for Bioactive Form Separation .......................................................................... 10 CHAPTER 2: PROJECT BASIS .............................................................................................. 15 Serum Matrix and Mass Spectrometry ..................................................................................... 15 Neurodegenerative Diseases ..................................................................................................... 16 Research Gap and Aim ............................................................................................................. 17 CHAPTER 3: MATERIALS ..................................................................................................... 19 Biological Specimens................................................................................................................ 19 Solvents and Reagents .............................................................................................................. 21 Internal Standards ..................................................................................................................... 21 External Standards .................................................................................................................... 22 Preparing a Test Run................................................................................................................. 23 CHAPTER 4: METABOLOMICS PIPELINE ........................................................................ 25 (Step 1) Standard Dilution ........................................................................................................ 25 (Step 2) Sample Preparation ..................................................................................................... 25 (Step 3) Lipid Extraction .......................................................................................................... 25 (Step 4) Liquid Chromatography and Mass Spectrometry ....................................................... 28 (Step 5) Annotation and Integration ......................................................................................... 30 Statistical Techniques ............................................................................................................... 31 CHAPTER 5: RESULTS, DISCUSSION, AND OUTCOMES .............................................. 33 Method Validation using Breastmilk ........................................................................................ 33 Metabolites Observed in Serum by HIV Status ........................................................................ 37 Calcidiol to Ercalcidiol Ratio.................................................................................................... 39 Cholesterol Oxidation Ratio ..................................................................................................... 40 Challenge of Coelusion ............................................................................................................. 40 Bioactive Metabolites Challenge .............................................................................................. 45 Opportunity for Future Method Applications ........................................................................... 48 Conclusion ................................................................................................................................ 49 APPENDIX .................................................................................................................................. 52 REFERENCES ............................................................................................................................ 59 v LIST OF TABLES Table 4-1. Peak detection parameters used for analyte identification and annotation ………… 30 Table 5-1. Peak areas for each vitamin D metabolite of type D 3 and D2, followed by average, standard deviation, and CV across samples ……………………………………….. 36 Table 5-2. Sums of each class of metabolites found, by HIV infection status ………………… 39 Table A1. Production batch ID numbers for each vitamin D standard purchased …………….. 52 Table A2. Ratios calcidiol-to-ercalcidiol, etc. by HIV infection status; concentration basis ….. 53 Table A3a. Calcidiol, ercalcidiol, total 25-(OH) D, hormones, oxysterols, phytosterols, and cholesterol peak areas; concentration basis ratios for 25-(OH)D type, cholesterol oxidation status, and lathosterol-to-cholesterol in serum biospecimens #1-26 ……. 54 Table A3b. Calcidiol, ercalcidiol, total 25-(OH) D, hormones, oxysterols, phytosterols, and cholesterol peak areas; concentration-basis ratios for 25-(OH)D type, cholesterol oxidation status, and lathosterol-to-cholesterol in serum biospecimens #27-50 …... 55 Table A4a. Relative abundances of select analytes and classes, per individual sample (#1-26) . 56 Table A4b. Relative abundances of select analytes and classes, continued (#27-50), with averages by HIV variable ………………………………………………………….. 57 vi LIST OF FIGURES Figure 1-1. Molecular structure of vitamin D isoforms D3 and D2 (Tuckey et al., 2019) ………. 1 Figure 1-2. Process flow emphasizing vitamin D3 photosynthesis and metabolism in vivo (Ahmed et al., 2020) ………………………………………………………………... 2 Figure 1-3. Irradiation of ergosterol toward vitamin D2 or photoisomers (Sun et al., 2022) 4 Figure 1-4. Overview of metabolomics study according to Du, et al. (2022) …………………. 12 Figure 3-1. WordCloud™ depiction of many patient factors documented ……………………. 21 Figure 4-1. Polar fraction remains in 5 mL storage tube with white protein cake at bottom ….. 27 Figure 4-2. Original flowchart of DSHO extraction method ………………………………….. 28 Figure 5-1. Mass fragmentation of cholesterol-H2O in breastmilk test sample ……………….. 33 Figure 5-2. (a) Mass spectra observed for calcitriol-H2O in pure Calcitriol standard; (b) Mass spectra recorded for feature of calcitriol-H2O within breastmilk test sample …….. 34 Figure 5-3. Chromatogram view of calcitriol-H2O features in three breastmilk samples …….. 35 Figure 5-4. Chemical classification of all species annotated with the described pipeline …….. 37 Figure 5-5. Mass fragmentation (or “spectra”) recorded by run of pure 19-OH standard …….. 41 Figure 5-6. (a) Mass fragmentation (or “spectra”) recorded by run of pure 19-OH standard; (b, c) Fragmentation spectra observed in two distinct biospecimens’ data for 19- hydroxycholesterol derivative compound …………………………………………. 44 Figure 5-7. (a) Mass spectra for calcitriol in sample data, zoomed-in so precursor 417.34 Daltons is visible; (b) Full-view, zoomed-out of calcitriol mass fragmentation ….. 46 Figure 5-8. (a) Mass spectra for ercalcitriol in sample data, zoomed-in so precursor 429.34 Daltons is visible; (b) Full-view, zoomed-out of ercalcitriol mass fragmentation … 47 Figure A1. Illustrative workflow of metabolomics experimental activities (Chaleckis et al., 2019) ………………………………………………………………………………. 53 Figure A2. Cloud plot of all features from HPLC-MS/MS with expanded cohort (n = 148) …. 58 Figure A3. Principal Component Analysis (PCA) applied with preliminary application to pre- process, “balanced” sample selection (50 HIV+, 50 control) ……………………... 58 vii LIST OF ABBREVIATIONS 25-(OH)D cumulative 25-hydroxyvitamin D; sum of serum calcidiol and ercalcidiol CYP27A1 cytochrome P450 enzyme of family 27, subfamily A, member 1 (for example) DAS data acquisition system(s) DSHO vitamin D, secosteroids, hormones and oxysterols (a local acronym) GC-MS gas chromatography tandem mass spectrometry HPLC-MS/MS high-performance liquid chromatography tandem mass spectrometry, triple quadrupole viii CHAPTER 1: INTRODUCTION Vitamin D: The Nutrient The term “vitamin D” describes numerous secosteroids each identified as a byproduct or intermediate of the vitamin’s metabolism. These compounds can be grouped according to their two most common structural forms: D2 (calciferol) and D3 (cholecalciferol). Distinct biological and health functions of vitamin D3 as opposed to vitamin D2 (and vice versa) have not been well established to date (Bouillon & Carmeliet, 2018) (Mena-Bravo et al., 2015). Vitamin D2 poses an extra methyl group at carbon 24 and a double bond between carbons 22 and 23 not seen in the form D3. This is shown by Figure 1-1 (Tuckey, Cheng, & Slominski, 2019). This subtle change in molecular chemistry bears great consequence on nutrient bioavailability. Figure 1-1. Molecular structure of vitamin D isoforms D3 and D2 (Tuckey et al., 2019) 1 There exist numerous vitamin D metabolites distinguished by chemical group and level of bioactivity. These include conjugates forms like 25-(OH)D-3-sulfate, epimers like 3-epi- 25(OH)D, and catabolites such as 24,25-(OH)D (Abu Kassim, Shaw, & Hewavitharana, 2018). Each form results from unique, structurally altering processes and can be classified by D 3 or D2 origins, based on the molecular orientations at carbons 22, 23, and 24. Figure 1-2 (Ahmed et al., 2020) provides an illustrative schematic of metabolic activities and their increasingly bioactive vitamin D3 products. Figure 1-2. Process flow emphasizing vitamin D3 photosynthesis and metabolism in vivo (Ahmed et al., 2020) Vitamin D’s most bioactive form in the human body is calcitriol (1,25-dihydroxyvitamin D3), a twice-hydroxylated member of the D3 family that appears more hormonally active than any 2 other vitamin D metabolite (Tuckey et al., 2019). Widely accepted functions of calcitriol include regulations of biometals and phosphates in the blood, anti-proliferation, and cell signal transduction (Makin, Jones, Kaufmann, & Calverley, 2010). Vitamin D3 production involves photosynthesis in mammalian skin and some plants. This cascade begins when sunlight’s ultraviolet-B (UVB) fraction irradiates a precursor naturally occurring in skin, 7-dehydrocholesterol (Mayne & Burne, 2019). This photoreactive step requires UVB radiation with wavelength below 315 nanometers (Müller & Volmer, 2015). Earth’s atmospheric ozone layer prevents the passage of UVB light with wavelength shorter than 290 nanometers, which in function provides a lower bound on wavelengths that satisfy this step (Jäpelt & Jakobsen, 2013). This fraction of sunlight carries high-energy photons that absorb into the covalent bond between carbons 6 and 7 in the beta-ring of 7-dehydrocholesterol. By electron balance, these photons at 6 and 7, in effect, dissociate the bond between carbons 9 and 10 and leave the beta-ring open, constituting the form previtamin D 3 (Melmed, 2020). This intermediary form is the subject of a negative feedback loop that- in part- renders hypervitaminosis D very uncommon. Prolonged exposure to the same 290-315 nm fraction of ultraviolet radiation consumes previtamin D by photochemically converting it to lumisterol or tachysterol, stereoisomers of 7- dehydrocholesterol (Tuckey et al., 2019). Previtamin D3 is biologically inert and thermally sensitive. It undergoes a temperature- dependent reaction in which electrons shift and leave a trans-conformed 6,7-diene which distinguishes the product, cholecalciferol (Jäpelt & Jakobsen, 2013). Vitamin D 3, here called “in vivo,” diffuses into the blood from the skin. It is worth noting that before its metabolism in the liver and kidney, cholecalciferol is also biologically inactive (Holick, 2009). 3 Vitamin D2 as a human nutrient can only be obtained through diet. Its production begins when ultraviolet-B light irradiates the D2 ergosterol precursor within the membranes of yeast and other fungi (Sun, Nzekoue, Vittori, Sagratini, & Caprioli, 2022). Thus, edible mushrooms are one good source of vitamin D2 (Müller & Volmer, 2015). In the human body, the metabolized form ercalcitriol constitutes the most biologically active form of the D 2 sub-family and exhibits intracellular activity comparable to calcitriol (Bikle, 2014). Figure 1-3 (Sun et al., 2022) reveals that vitamin D2 photosynthesis in fungi has a similar feedback mechanism in place to up- and down-regulate the thermal isomerization of calciferol. Figure 1-3. Irradiation of ergosterol toward vitamin D2 or photoisomers (Sun et al., 2022) Vitamin D is naturally well regulated within the human body (Holick, 2009). As a patient becomes vitamin D insufficient, a decrease in epithelial calcium absorption prompts the 4 parathyroid glands to secrete parathyroid hormone (PTH). PTH then acts as a ligand at the kidney, hastening the bioactivation of intermediary vitamin D forms such as calcidiol [25-(OH)D 3] and increasing calcium reabsorption from urea (Jäpelt & Jakobsen, 2013). Thus, parathyroid hormone is responsible for a natural feedback loop regulating vitamin D bioactivation in the body. In many cases, D3 can be obtained in vivo sufficiently without use of an over-the-counter oral supplement or reliance on natural dietary sources of the vitamin. In the body, bioactive metabolites carry out health functions by binding to vitamin D receptors (VDR), intracellular proteins traditionally located at the membrane of cell nuclei (Tuckey et al., 2019). Metabolite Bioavailability For unprocessed vitamin D circulating in serum as cholecalciferol and calciferol, a preliminary metabolic step is 25-hydroxylation in the liver. This is achieved by activity of a cytochrome p450 enzyme in the endoplasmic reticulum of liver cells known as CYP2R1 and- in the case of D3 only- an additional oxidase known as CYP27A1 (Borel, Caillaud, & Cano, 2015) (Chun, 2012). One metabolite resulting from this step, calcidiol (25-hydroxyvitamin D 3), is the most abundant individual form of all vitamin D in serum circulation (Müller & Volmer, 2015). Many studies synonymize calcidiol level 5 with “total vitamin D” for purposes of quantification, but some qualitatively point out that there are more than fifty (50) human metabolites of vitamin D worth considering (Bouillon & Carmeliet, 2018). The 25-hydroxylated vitamin D intermediates then are transported to the kidney, where- independent of D2 or D3 origin- they are 1α-hydroxylated by the enzyme CYP27B1 (Bikle, 2014). In physiological nomenclature, this renal catalyst CYP27B1 may commonly be referred to as “1α- 5 In the most specific, refined experimental designs, reported 25-hydroxyvitamin D represents the sum of both 25- hydroxyvitamin D3 and 25-hydroxyvitamin D2 measured. 5 hydroxylase” because of its function. Fascinatingly, 1α-hydroxylase activity has been detected in other tissues such as bone (van Driel et al., 2006). Still unclear is whether or not vitamin D can be completely metabolized extra-renally. Human blood is equipped with Vitamin D Binding Protein (DBP), which attaches to and transports vitamin D metabolites between tissues (Müller & Volmer, 2015). DBP is widely bioavailable with a serum concentration twenty (20) times greater than that of all circulating vitamin D (Tuckey et al., 2019). At any given time, an individual’s serum DBP is only two to five percent saturated by vitamin D metabolites attached (Makin et al., 2010). Because the binding affinity of DBP for D3 produced in vivo is over 1,000 times greater than for D2 forms (Chun, 2012), locally synthesized vitamin D3 reserves biological transport priority over all vitamin D2. Within the D3 subfamily of metabolites, calcidiol [25-(OH)D3] binds to DBP with over ten (10) times the affinity with which calcitriol [1,25-(OH)2D3] does (Tuckey et al., 2019). The enhanced biological transport exhibited by cholecalciferol and calcidiol, especially those produced in vivo, helps in making hypervitaminosis of calcitriol very uncommon. As opposed to the consistent resource of DBP for D 3 in vivo transport, vitamin D obtained through diet (responsible for all D2 and some D3) relies primarily on chylomicron chemistry for transport from small intestine epithelia to the liver for ensuing metabolism, with some DBP activity as well (Jäpelt & Jakobsen, 2013). Even aside from its unique transport aid of DBP, vitamin D 3 in vivo retains enhanced bioavailability in the blood because all other vitamin D forms in serum must first survive traditional lipid absorption in the small intestine, a process dependent on several key nutrients and in which some amounts of vitamins and minerals are regularly lost by excretion (Makin et al., 2010). 6 Vitamin D Deficiency Vitamin D deficiency is defined- without consensus- by circulating serum concentration of less than 20 nanograms per milliliter of any 25-hydroxylated D metabolite(s) (Dror et al., 2022). Through hundreds of patient studies observing nearly eight million participants, an estimated forty-seven percent (47%) of the global population was classified as deficient (Cui et al., 2023). By the standard that 30 nanograms or more calcidiol plus ercalcidiol [collectively, 25- (OH)D] per milliliter of human serum is sufficient for bone and endocrine health functions of the vitamin, an estimated three quarters (75%) of the United States population falls below and is insufficient (Chun, 2012). By the early 2000s, bone fracturing in the elderly and rickets in children were among the few recognized risks associated with vitamin D deficiency. But since then, research has associated sufficient vitamin D with lower blood pressure, reduced rates of multiple sclerosis and cancer, and maintained muscle control among the elderly, among other health benefits (Lips, Bilezikian, & Bouillon, 2020). With the emergence of the COVID-19 pandemic in 2020, serum vitamin D was scrutinized yet again for its potential effect of reducing likelihood of infection with the virus SARS-CoV-2. A study in Israel demonstrated risk of COVID-19 infection to be fourteen (14) times greater for individuals with recent record of vitamin D deficiency than for those with serum 25-(OH)D of at least 40 nanograms per milliliter (Dror et al., 2022). Vitamin D deficiency has become increasingly prevalent and a few of its common causes are lack of routine exposure to sunlight, overuse of sunscreens, and malnutrition (Jäpelt & Jakobsen, 2013) (Saenger, Laha, Bremner, & Sadrzadeh, 2006). The health quandary of vitamin D deficiency concerns impoverished and developed societies alike, with D being one of the two most common vitamin deficiencies in the Unites States (CDC, 2012). Furthermore, the CDC 7 reveals African Americans are disproportionately affected, of whom 31% are D-deficient. A significant obstacle for this group results from increased skin pigmentation caused by the production of melanin in the epidermis, as shown by a study at Howard University and Northwestern University. In its cohort of self-identified African American men (n = 734) with mean age near 51 years residing in Washington, D.C, Chicago, IL, and Cincinnati, OH, past recorded serum 25-(OH)D levels were found to be deficient or severely deficient for more than half (50%) (Batai et al., 2021). A distinct risk for vitamin D deficiency has been observed among people of Middle Eastern cultures, as well, since regular use of skin-covering clothes physically restricts access to sunlight (Dror et al., 2022). Vitamin D3 production in vivo requires solar radiation of a specific wavelength range of 290 to 315 nanometers, which corresponds to incident energy to the epidermis no less than 20 kilojoules per square centimeter (Saenger et al., 2006). This nutritious resource, though, is often unavailable to humans based on geography of their residence. The angle of incidence of solar radiation tends to be large, or further from orthogonal to a given point on Earth’s surface, at any location during winter seasons. Increases in this angle, also known as the zenith, can be particularly extreme during winters at locations further from the equator. In these cases, incoming light is occupied by the ozone layer over a longer distance. This light undergoes greater filtration by ultraviolet wavelength, which reduces the survival of high-energy photons to persons and wildlife on the ground (Leal, Corrêa, Holick, Melo, & Lazaretti-Castro, 2021). Access to radiation necessary for vitamin D photosynthesis can be blocked, entirely, for months at a time, depending on distance from Earth’s equator and degree of planetary tilt. The population of Earth is unevenly split between the Northern and Southern hemispheres, with the former garnering 87.5% of all people as of 2005 (Kummu & Varis, 2011). 8 Of this vast majority, nearly half live northward of latitude 30°N. Yet, the solar UVB wavelengths required to irradiate 7-dehydrocholesterol and launch vitamin D synthesis in vivo are only consistently available year-round in locations below latitude 35°N, which in the United States represents the southern border of Tennessee (Jäpelt & Jakobsen, 2013). As a result, vitamin D nourishment without dietary intake is seasonally dependent for a substantial population across the world (Leal et al., 2021). This in part contextualizes the prevalence of vitamin D deficiency today. Synapse Chemistry and Vitamin D Vitamin D monitors calcium signaling and with it, nervous communication throughout the body. The ability to regenerate synapses and maintain their electrochemical and structural properties is called synaptic plasticity. Vitamin D has been implicated in the maintenance of synaptic plasticity, a key process in learning and memory functions and one dependent on frequent calcium transport (Mayne & Burne, 2019). Specifically, intraneuronal calcium is regulated by vitamin D at voltage-sensitive calcium ion channels (Banerjee et al., 2015). These proteins and the binding of calcitriol to the vitamin D receptor (VDR) thereby enable neuron development from stem cell proliferation to functional differentiation. Vitamin D receptors have been found colocalized in the plasma membrane of neurons with a precursor to amyloids, the same class of proteins identified as plaque-forming in neurodegenerative pathologies (Ouma et al., 2018). Some genes including those encoding cytochrome p450 proteins (CYP24A1) and calcium absorption (CaBP-D 9k, CaBP-D28k) are upregulated, or increased in rate of expression, as a direct result of vitamin D binding to VDR (Yutuc et al., 2020) (Wang, Zhu, & DeLuca, 2012). Other genes representing PTH release and 9 CYP27B1 are downregulated by the same activity. In summary, the biological cascade that results when vitamin D binds to its namesake nuclear receptors increases the expression of vitamin D-associated health effects and advances the homeostatic feedback loop regulating vitamin D, itself, in the body. Calcitriol takes on a critical role in maturation of nerve growth factor (NGF) in the central nervous system (Gezen-Ak, Dursun, & Yilmazer, 2014). NGF levels have been shown significantly reduced in the forebrain of patients with Alzheimer’s Disease. Vitamin D also has been shown in a rat experiment to exhibit properties protecting the subject from brain oxidative stress, enhancing brain energy balance and reducing tau hyperphosphorylation, a process associated with aging (Banerjee et al., 2015). While most vitamin D metabolites are biologically inert and while calcitriol generally serves one purpose in gene regulation toward calcium homeostasis, the human nervous system is remarkably dependent on its nutrition. Metabolomics for Bioactive Form Separation Metabolomics embodies high-throughput mapping of process-related small 6 molecules in biological or environmental matrices. Frequently applied to lipid analytes and coined “lipidomics,” it is closely related in science to proteomics, genomics, and other studies of the “Omics” field. Relative to health, for example, The Human Metabolome Project is among the most distinguished and ambitious goals of twenty-first century medical science. It attempts to catalog all metabolites in the human body and is motivated by several germane knowledge gaps (Wishart, 2007). 6 mass below 1,500 Daltons 10 Metabolomics represents a movement toward organization and enhanced accountability in biomolecular research. This is evidenced by the trend of its journals requiring “FAIR” data (Findable, Accessible, Interoperable, and Reusable) (Inau, Sack, Waltemath, & Zeleke, 2021). Figure 1-4 (Du et al., 2022) depicts a peer-reviewed workflow governing successful metabolomics projects. The first experimental task within the workflow is sample preparation. This step incorporates laboratory techniques such as solid phase extraction (SPE), enzymatic or antibody titer assays, and phase isolation. Depending on the equipment selected, such as GC-MS, derivatization steps like silanization may be necessary. All physical activities with samples are included in this first category of sample preparation. 11 Figure 1-4. Overview of metabolomics study according to Du, et al. (2022) Data acquisition follows sample preparation. Due to the high-throughput 7 nature of metabolomics work, a portion of this task is automated by a digital instrument. For vitamin D quantification, this is often a type of chromatography in tandem with a mass spectrometer. The experimenter sets specific conditions and parameters that allow a data acquisition system (DAS) to operate upon input sample(s) and generate meaningful data. The experimenter communicates sample load and instructions to a desktop computer through batch tables that govern setup, sample, and solvent injections as well as any programmed maintenance. Finally, the user 7 Suited for thousands of samples or more to be run with automated equipment, optimizing process time 12 receives data (traditionally in .RAW format) from the DAS and converts it to a digital type suited for their preferred data processing software. Data processing in vitamin D metabolomics may begin in one of many tools such as MS- DIAL™, El-MAVEN™, or MetaboAnalyst™ (Chen, Li, & Xu, 2022). These programs are distinguished by individual strengths and applications. Some are compatible with many instruments- LC-MS, GC-MS, and NMR, for example- while others boast high-resolution or color-enhanced displays. Some are open source, allowing cost-free access but lacking commercial infrastructure for consumer queries and support. Careful selection of data processing software affords the experimenter some ease in perhaps the most time-consuming lipidomics project step, overall. It may also improve their likelihood of high accuracy in manual annotations8. The choice of software does not, however, improve or alter the underlying quality of the data already recorded by the DAS. Bioactive and low-abundance metabolites can be integrated just as conveniently as the more dominant forms, assuming extraction efficacy. During data processing, annotation is said to be at Level 1 in targeted or semi-targeted applications in which the experimenter identifies a present metabolite by comparison to a previous run of a pure standard in the local equipment with the same method (Chaleckis, Meister, Zhang, & Wheelock, 2019). When all possible annotations at Level 1 are complete, the experimenter may proceed with remaining unidentified features to annotate at Levels 2 and 3. A Level 2 annotation takes place when a compound match is made by comparing order of elution, precursor, or spectral data to those of a third-party database or publication. Annotations at Level 3 incorporate forecasting of radical and ionization chemistry and is only exercised under the supervision of a professional analytical chemist. 8 The manual consideration of evidence from different DAS measurements toward qualitative and quantitative conclusions for each compound to be reported 13 Annotations convert the data from ambiguous features to numerical, integrated areas of each metabolite. Equation 1 (as well as Figure A1 in the Appendix) shows how an internal standard’s known concentration and peak area together enable concentration calculations for each identified compound. Vitamin D metabolites are most accurately quantified using a deuterated internal standard such as calcitriol-d6. For other secosteroids, oxysterols and hormones, a non- naturally occurring compound such as 19-hydroxycholesterol (19-OH) can prove effective internally. Internal standards represent just one mechanism of quantifying mass (or concentration). Another method is called normalization, in which the matrix permits all peak features to be annotated with clarity and figured as percents of a known total input mass (or concentration). Equation 1. Compound concentration (ng/mL) by Internal Standard 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝐶𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛 𝑈𝑠𝑒𝑑 ∗ 𝑆𝑎𝑚𝑝𝑙𝑒 𝑃𝑒𝑎𝑘 𝐴𝑟𝑒𝑎 𝑆𝑎𝑚𝑝𝑙𝑒 𝐶𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛 = 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝐴𝑟𝑒𝑎 Statistical analyses follow, concluding the data preparation step. Data are organized such that each identified compound is reported in amount by sample number. The statistical methods may consist of Student t-Tests, Analyses of Variance (ANOVA), and univariate analyses, among others, to demonstrate significance of differences between groups or factors. In the final group of tasks, collectively known as data interpretation, compounds and their respective data are grouped according to chemical classifications or for other literary purposes. Values are verified with previous studies when applicable. The experimenter obtains a command of the literature in preparation for insightful, evidence-based discussions upon the new data. The general workflow of metabolomics study is thereby complete. 14 CHAPTER 2: PROJECT BASIS Serum Matrix and Mass Spectrometry Serum as a biospecimen, as opposed to blood, constitutes the removal of coagulation factors such as fibrinogen. As such, working with isolated serum is a popular experimental design because it reduces risk of material clotting during extraction and increases the efficacy of protein precipitating steps. While calcitriol is considered the major bioactive form of D 3 and vitamin D in the human metabolome, it alone does not provide a sound basis for measurement and determination of an individual’s vitamin D levels. In human serum, the half-life of circulating calcitriol and its D 2 counterpart, ercalcitriol (1,25-dihydroxyvitamin D2), is just four to six hours (Holick, 2009). This compares to a much longer half-life of nineteen (19) days for the metabolites calcidiol and ercalcidiol [collectively “25-(OH)D”] used to define vitamin D deficiency today (Müller & Volmer, 2015). Despite being the consensus gold standard for measurement of vitamin D metabolites, HPLC-MS/MS techniques upon serum matrices do carry a few drawbacks (Saenger et al., 2006). For example, protein precipitation is common, which clouds samples and complicates extraction mechanics. In addition to its inconvenience to matrix handling, protein content creates a specific challenge of unbinding and freeing the vitamin D analytes intended to measure. Ninety percent (90%) of 25-hydroxyvitamin D in blood (or serum) is bound to a protein- most commonly DBP, albumin, or lipoprotein (Zelzer, Goessler, & Herrmann, 2018). Additionally, local conditions and performance of the mass spectrometry device limit the reproducibility between machines. 15 Neurodegenerative Diseases As of 2015, an estimated 37 million persons had dementia and this figure was predicted to double by year 2035 (Banerjee et al., 2015). Alzheimer’s Disease (AD) remains a leading cause of dementia-associated pathologies, defined in human health by progressive deficits of short- and long-term memories that lead to cognitive and motor incapacities and death within nine years of diagnosis (Ouma et al., 2018). Dyshomeostases of biometals such as calcium (Ca2+) have been linked to Alzheimer’s Disease (Elia et al., 2019) because they deteriorate synapse function, resulting in neuronal toxicity (Crouch, Barnham, Bush, & White, 2006). Similar dysregulations have been found in common among patients with autism spectrum disorders (ASD) (Pfaender & Grabrucker, 2014). A study at Fukuoka University in Japan measured serum 25-(OH)D3 in healthy, mild cognitively impaired (MCI), and Alzheimer’s Disease patients (in mild, moderate, and severe groups) by competitive radioimmunoassay with antibodies. It found serum calcidiol concentration significantly reduced among mild and severe Alzheimer’s Disease cases compared to healthy subjects with over ninety-nine percent (99%) statistical confidence (Ouma et al., 2018). Another study, for example, suggested a relationship between neurodegenerative diseases and categorized statuses of vitamin D levels- “sufficient,” “deficient,” and “severely deficient” (Littlejohns et al., 2014)- in which serum vitamin D was defined by 25-(OH)D. Because 25-(OH)D is generally higher in human serum concentration than precursor or bioactive metabolites of the vitamin, many experimental designs commonly exclude quantification and analyses of these numerous distinct metabolites as they pertain individually and collectively to health outcomes (Bouillon & Carmeliet, 2018). 16 Research Gap and Aim These and more members of the current literature make room for consequential discoveries at the nexus of nutrition and health. They also empower ample research questions… How do the human health roles of bioactive vitamin D2 compare to those of bioactive vitamin D3? What are the distinct levels of precursor D2 and D3 forms in vitamin D-deficient serum, currently categorized exclusively by 25-(OH)D? Amid enzymatic assays, antibody reactivity experiments, and mass spectrometry, which method enables the most precise measurement of vitamin D metabolites in human serum? Finally, do levels of oxysterols and hormones close in structure to serum vitamin D mimic the vitamin in circulating levels, and which conditions may cause them to shift? The project of focus was designed to answer a few of these questions and, most importantly, create a widely applicable method for future metabolomics research to resolve questions as they emerge. Liquid chromatography tandem mass spectrometry assays that report vitamin D in current literature predominantly observe only calcidiol and ercalcidiol [25-(OH)D] due to limitations in low-abundance metabolite detection (Müller & Volmer, 2015). As such, this project was unique by its attention to simultaneous mapping of multiple vitamin D metabolites and other secosteroids, hormones, and oxysterols (DSHO). Another distinction was made by the goal of reduction in minimum volume requirements for serum biospecimens. Traditionally, vitamin D analytes have been extracted from individual serum aliquots no less than three hundred (300) microliters, while many commercial laboratories have set a conservative one-milliliter minimum (Jenkinson et al., 2021). Designing the novel extraction pipeline for initial sample sizes of one hundred (100) microliters served to maximize use of the available cohort. Together, these considerations provided an enhanced foundation and precedent for future analyses of theorized associations between serum 17 vitamin D and human health outcomes. Criteria for achieving the project aim through the novel method are listed below.  Semi-targeted method  Profiling of analytes  Isolation of vitamin D bioactive forms calcitriol (D3) and ercalcitriol (D2)  Simultaneous quantification of secosteroids and vitamin D metabolites  Molecular classification of hormones, oxysterols, and other steroids present in human serum matrix The current experiment was motivated by the pressing global need for multidisciplinary investigation into public health obstacles like Alzheimer’s Disease and dementia. The project also sought to raise awareness to potential health inequities resulting from transethnic variation in skin composition as they relate to vitamin D deficiency. 18 CHAPTER 3: MATERIALS Biological Specimens Biological specimens were collected under the authority of the International Review Board (IRB: STUDY00000205) and supervision of Dr. Sarah Zalwango, Director of Medical Services at Kampala Capital City Authority (KCCA) in Kampala, Uganda. The KCCA facilitates research in pertinent disease topics including human immunodeficiency virus (HIV) and tuberculosis (TB). Dr. Zalwango is the site’s principal investigator and supervisor of the samples’ origin, collection, and early storage. Ms. Hellen Nansumba is a Laboratory Scientist at the Central Public Health Laboratories and works for the Ministry of Health in Uganda. By the diligence of these collaborators, human serum samples were collected in serum separator tubes (SST) at the Kawaala Health Center, a clinic under the administration of the KCCA. All specimens were immediately stored at -80°C and delivered to Central Public Health Laboratories within five hours of collection. Serum was obtained by centrifugation of whole blood in SST at 3400 RPM for 10 minutes, aliquoted and stored in 1.2 mL cryogenic tubes at -80°C. At constant temperature, samples were shipped from Kampala, Uganda to MSU in 2020. Upon receipt, samples were promptly sorted and kept at -80°C. The HIV and TB disease study from which sample collection and accessibility originated should not be overlooked. Three of many potential factors that distinguish the HIV-positive biospecimens from healthy serum controls with respect to vitamin and mineral nutrition are past HIV-infection, anti-retroviral therapy (ART) regimen, and lifestyle. Some of these factors have been studied and published upon. Isabirye et al. (2020) found frequent deficiencies in diet of iron, calcium, zinc, and vitamins A and B in a study of four hundred (400) HIV-infected adults beginning ART who were enrolled in a vitamin supplement plan. These deficiencies were more 19 common in female participants than males (Isabirye et al., 2020). This association with gender was justified by a survey of the participants which showed a tendency for higher micronutrient intake by the male participants. In summary, comments based on data produced by the novel project would be restricted to the scope of the biospecimens used, which came from a rural, landlocked geography near Earth’s equator and for which rates of HIV-affected lifestyles are above average per capita. Collectively, the group of fifty serum biospecimens obtained for this project are hereby named the “Uganda cohort.” Dr. Amara Ezeamama of the Department of Psychiatry at Michigan State University led the global project network and orchestrated the survey of background data from the subjects in the Uganda cohort. Each of these characteristics created a distinct opportunity for association with several laboratory test results. Figure 2-1 depicts many of the key background data collected in conjunction with this study. These included marital status, diet, mental health markers, and many more. Individual occupations included tailor, teacher, travel agent, brick layer, grocer, and houseworker, among others. Level of education ranged from none to university graduate. Primary home cooking method varied between charcoal stove and woodfire. Individuals with history of HIV-positive diagnosis (present or past) were noted by the collaborator but not provided to the experimenter until data interpretation. For quality control purposes, vitamin D (DSHO) research was conducted blind to all defining variables, most importantly HIV. The characteristics of the cohort have been previously published (Isabirye et al., 2020; Pobee et al., 2022; Yakah et al., 2019). 20 Figure 3-1. WordCloud™ depiction of many patient factors documented Solvents and Reagents Methyl tert-butyl ether (MTBE), methanol and butylated hydroxytoluene (BHT) were purchased from Sigma-Aldrich®. Isopropyl alcohol and hexane, used in standard dilution and storage only, were purchased from J.T. Baker® (Matyash, Liebisch, Kurzchalia, Shevchenko, & Schwudke, 2008). Water as a solvent was obtained by tap and purified with an Electron LED Micropure UV device manufactured by Thermo Scientific®. Internal Standards Two compounds were acquired and added for use as internal standard. Calcitriol-d6 (“calcitriol, deuterated with six”) was purchased from Cayman Chemical® in powder form. The batch number for this item (for verification or recall) is shown in the Appendix, Table A1. This standard arrived in a vial as powder with mass five hundred (500) micrograms. Addition of 1 21 milliliter of methanol as solvent resulted in dissolution of the deuterated standard and a concentration of 500 μg/mL. The use of calcitriol-d6 and its abundance in each sample’s chromatogram enables precise calculations of vitamin D metabolite masses within the samples (Saenger et al., 2006). 19-hydroxycholesterol (19-OH) was purchased from Steraloids® and diluted in 4:1 (v/v) hexane-isopropyl alcohol solution to a concentration of 0.1 mg/mL. The purpose of 19-OH within the sample was to quantify targeted and untargeted lipophiles outside of the vitamin D secosteroid family. External Standards Six vitamin D metabolites were prepared as external standards. Representing vitamin D 2, these were calciferol (vitamin D2), 25-hydroxyvitamin D2, and ercalcitriol (1α,25-hydroxyvitamin D2). For vitamin D3, these were cholecalciferol (vitamin D3), 25-hydroxyvitamin D3, and calcitriol (1α,25-hydroxyvitamin D3). Batch identification numbers for each are listed in Table A1 of the Appendix. These compounds were diluted in methanol to individual stocks of 10 μg/mL each. A working solution was prepared by combining 150 μL of each of these six diluted stocks and adding 14.1 mL methanol, leaving each of the six standards simultaneously present at concentration 100 ng/mL. Each external standard had a common purpose: to represent a specific vitamin D metabolite in a separate, controlled run with fixed, known concentration. Use of external standards allowed targeted metabolomic observations of DSHO within the matrix and justified method-specific annotation at Level 1. Unlike the internal standards calcitriol-d6 and 19-OH, all six external standards were prepared not only for individual injection with HPLC-MS/MS, but for gas chromatography tandem mass spectrometry (GC-MS), as well. The purpose of including GC-MS application was to observe 22 similar fragmentation behavior of the six vitamin D metabolite standards between the chromatography techniques. GC-MS requires an additional step of silanization for all sample and standard solutions prior to any injection. Standard silanization began by drying the standards with nitrogen flow. One-hundred (100) microliters of pyridine provided by Sigma-Aldrich® was added and the tube gently vortexed to remove any trace amounts of water, which could harm the data acquisition equipment. One- hundred (100) microliters of silanization mixture according to Sweeley (from Sigma-Aldrich®) was added, followed by another gentle tube vortex (Sweeley, Bentley, Makita, & Wells, 1963). Standard solutions were exposed to this reagent for sixty (60) minutes at the controlled temperature of 70°C. After this, they were dried again by nitrogen flow, resuspended in 100 uL pure hexane, and stored in amber vials to be injected with GC-MS. The machine used for acquisition in this verification step was a Shimadzu™ GC-MS QP2010 SE with an electrospray ionization detector. Preparing a Test Run The human serum biospecimens to be analyzed in the novel project were available in amounts ranging from 0.5 to 1.5 milliliters. These limited volumes were best preserved until repetition and validation of the extraction method were complete. A test run took place in December 2021. This followed a first-draft DSHO extraction method engineered for the complete isolation of secosteroid and hormone species from a biological matrix. Its chronology of laboratory activities reflected metabolomics customs to prepare each sample for efficient processing of mass spectrometry data. These are reinforced by Figure A1 (Chaleckis et al., 2019) in the Appendix, which illustrates the goals of each step in a general lipidomics pipeline. The experimental techniques practiced in this December 2021 test were comparable to the ensuing analysis of serum 23 in March 2022, for which a final, validated method version was used. First (in 2021), untreated human breastmilk was selected to assess and show the method’s performance with all desired compounds (DSHO) as a substitute biological matrix more available by volume than the specimens in the Uganda cohort. This choice was validated by literature including a Journal of Chromatography publication that demonstrated a universal mass spectrometry method to detect dopamine antagonists in both human serum and breastmilk (Zavitsanos, MacDonald, Bassoo, & Gopaul, 1999). Breastmilk was supplied by the Food and Health Engineering Laboratory from constant storage at -80°C. Standards from Cayman Chemical® included powdered forms of calciferol (vitamin D 2), ercalcidiol [25-(OH) D2], ercalcitriol [1,25-(OH)2 D2], cholecalciferol (vitamin D3), calcifediol [25-(OH) D3], and calcitriol [1,25-(OH)2 D2], all from the original containers listed in Table A1. With these vitamin D metabolites in mind and without guaranteed detectability of their natural abundances in human breastmilk substitutes, ten (10) micrograms of each of the six standards were added to three 0.25 mL aliquots of breastmilk. Lipid extractions via liquid-liquid equilibrium ensued for the three mixtures. By adding these standards to the breastmilk aliquots for the specific purposes of this test, concentrations of the six secosteroid members were spiked and made more readily identifiable in the resulting chromatogram displays. Test extracts were then run with HPLC-MS/MS and analyzed to critique the preliminary DSHO extraction method. 24 CHAPTER 4: METABOLOMICS PIPELINE (Step 1) Standard Dilution Standards ercalcidiol, ercalcitriol, calcitriol, calcitriol, and calcitriol-d6 were prepared at 500 μg/mL in pure methanol. Standards vitamin D 2 and vitamin D3 were prepared at 1 mg/mL. All seven vitamin D standard solutions were stored at -20°C until use. Standard 19-OH was diluted to a stock of 500 ng/mL and stored at 4°C until use. (Step 2) Sample Preparation Few serum vials presented as dark brown or with traces of whole blood. These biospecimens received an additional centrifugation step at the beginning of the extraction protocol. Then, a randomized identification number was assigned to each vial as follows: VD as per vitamin D, S as per Serum, and a counting number (e.g., VD-S01, etc.). As previously mentioned, some serum specimens required an initial centrifugation step at the beginning of the extraction day. These samples were VD-S18 and VD-S29. Pre- centrifugation consisted of a transfer to a 1.5 mL centrifuge-appropriate microtube (as needed) and 10,000 RPM for 2 minutes in the Eppendorf™. In each original sample centrifuged, the darker, more pigmented contents settled toward the bottom, allowing from the top a cream- colored aliquot comparable to those specimens not requiring this step. (Step 3) Lipid Extraction One-hundred (100) microliter aliquots of serum each were placed into a 5 mL plastic tube (approved for Eppendorf™ centrifuge) for extraction by liquid-liquid equilibria (Lipkie et al., 2013). The principles of this general approach include controlled, vigorous agitation and high 25 velocity centrifugation for the detachment and removal of bound proteins without physical damage to molecular analytes. Based on the high sensitivity of the receiving HPLC-MS/MS instrument and the estimated concentration of vitamin D in serum, a concentration of 0.1 nanogram per microliter (sample) was added of each internal standard. This was equivalent to 50 ng of calcitriol- d6 and 50 ng of 19-OH. To achieve this amount, 50 μL of calcitriol-d6 solution (at 1 μg/mL) and 100 μL of 19-hydroxycholesterol solution (at 500 ng/mL) were used. One milliliter of chilled 0.01% (v/v) butylated hydroxytoluene in methanol was added to each extraction tube, followed by a vigorous vortex to maximize dissolution and chemical contact with the BHT preservative. Then, 3 mL of MTBE were added to each tube before a second vigorous vortex for 30 seconds. Protein precipitation was given over one hour with simultaneous chilling, incubation, and orbital swirling (Belly Dancer™). After incubation, 1 mL of deionized water was added, and a third vigorous vortex ensued. Without delay, all samples were centrifuged at 3,900 RPM for 10 minutes while maintained at 4°C. Once complete, samples were covered to protect from light-induced oxidation and their lipid phases were isolated into 10 mL glass tubes by single-use, biological Pasteur pipets. All tubes- still protected from light- were placed in an ice storage bin and set aside. The remaining polar fractions received 2 mL of MTBE, followed by a vigorous vortex. Protein precipitation here was given fifteen (15) minutes in the device for simultaneous chilling, incubation, and swirling. Samples then re-entered the centrifuge for 10 minutes at 3,900 RPM and 4°C. The resulting lipid fraction, again settling to the top, was recovered and added to the corresponding reserve on ice. At this point, each tube contained about 5 mL of lipophilic solution from the corresponding biospecimen extraction. 26 Polar fractions (with protein precipitate remaining at the bottom) were kept and placed in -20°C for future analysis. A photo of this is given by Figure 4-1. The isolated lipid fractions were placed under nitrogen stream for evaporation (Organomation® Multivap™) while held at 50°C and protected from ambient light. After drying, 300 μL pure methanol was added to each before transferring to amber vials, each equipped with a glass insert, for HPLC-MS/MS analysis. Figure 4-2 illustrates the confluence of all routine extraction steps in this novel approach with liquid- liquid equilibrium. In the illustration, blue boxes represent collection of material. Purple ellipses portray the introduction of carefully chosen standards, reagents, or solvents. Green diamonds and red boxes represent agitation and other physical activities. Figure 4-1. Polar fraction remains in 5 mL storage tube with white protein cake at bottom 27 Figure 4-2. Original flowchart of DSHO extraction method (Step 4) Liquid Chromatography and Mass Spectrometry The method adapted was able to separate secosteroids, hormones and oxysterols including vitamin D and related metabolites. The HPLC-MS/MS equipment consisted of a Shimadzu Prominence HPLC coupled to a Thermo LTQ-Orbitrap Velos mass spectrometer. The choice of a detector featured a triple-quadrupole (MS/MS). In effect, this highly specific apparatus reduced spectral noise in the data acquisition process and reached below traditional limits of detection governing single-quadrupole (MS) performance. Though use of triple-quadrupole HPLC-MS/MS for lipidomic samples can generally incur higher cost, this choice proved an essential step toward achieving sensitivity for low-abundance, bioactive metabolites within reduced available (sample) volumes. The liquid chromatography apparatus featured two LC20AD pumps, a vacuum degassing system, an autosampler, and a column oven. The HPLC column was type Phenomenex 2.0 mm- by- 150 mm Synergi HydroRP-C18 (4 micron particle, 80 Angstrom pore size), complete with a 28 guard cartridge of the same column chemistry. The first solvent was water with 0.1% formic acid (v/v). The second was methanol with 0.1% formic acid. The flow rate through the column was held at 250 μL/minute and the column oven temperature was 50°C. The autosampler was kept at 4°C and 10 μL from each amber vial was injected. The gradient conditions had a duration of two minutes and used two percent (2%) of the methanolic solvent. The volume of methanol with 0.1% formic acid was increased linearly to 10% between 2.0 and 4.0 minutes. At 4.5 minutes it was increased to 30%, then ramped linearly from 30% to 60% between 4.5 minutes and 16.5 minutes. The solvent was held at 60% for four minutes, then increased linearly to 85% by 24.5 minutes. This methanolic solvent was then ramped linearly from 85% to 100% over 13 minutes and held there for an additional 10 minutes. All liquid chromatography solvents then were returned to starting conditions and the column was equilibrated for five minutes prior to the next injection. Column eluent was introduced to a Thermo® LTQ-Orbitrap Velos mass spectrometer via a heated electrospray ionization source. The mass spectrometer was operated in positive ion mode at 60,000 resolution with full scan MS data collected from 100-700 m/z. Data-dependent product ion spectra were collected on the 3 most abundant ions at 7,500 resolution using the FT analyzer. The electrospray ionization source was maintained at a spray voltage of 4.5 kilovolts with sheath gas at 30 (arbitrary units), auxiliary gas at 10 (arbitrary units), and sweep gas of 2 (arbitrary units). The inlet of the mass spectrometer was held at 350 degrees Celsius, and the S-lens was set to 50%. The heated ESI source was maintained at 350 degrees Celsius. Metabolites including vitamin D- related compounds, hormones and oxysterols were identified as their [M-H2O+H]+, [M- 2H2O+H]+, [M+H]+, and [M+Na]+ ions under the conditions employed. 29 (Step 5) Annotation and Integration Digital files output by the MS computer were received in .RAW format and converted to type .mzXML. For metabolomic analyses, El-MAVEN™ (Elucidata® version 12.0) was used. El- MAVEN™ is an open source software with cloud-based storage well suited for processing HPLC- MS/MS and GC-MS data (Chen et al., 2022). Chromatograms for all fifty (50) samples were analyzed in small batches of ten (10) each with preliminary steps of alignment and isotope correction. Subsequent steps were parameter setting and digitized peak suggestion. Table 4-1 elaborates on the specific parameters chosen for this annotation at Level 1. Importantly, each automatic match received rigorous manual annotation, as did any uncaught peaks including those slightly smaller than the detection limit or shortly outside of the anticipated retention time (RT) range. Mass-to-charge (m/z) ratios, fragmentation spectra, and retention times were considered in comparison with pure standards. Based on the spectral properties of each analyte, the precursor, or expected mass based on the atomic makeup of the pre-fragmented chemical, may or may not appear as a high-intensity fragment in the compound’s observed spectra in a given sample. In cases of spectra with many distinct fragments, comparison to external pure standards proved essential to confirm the identities of compounds. Table 4-1. Peak detection parameters used for analyte identification and annotation Parameter value (Unit) electron impact window (+/-) 10.00 ppm minimum peak width 3 scans limit number of groups per compound 10 best groups minimum peak intensity 5,000 relative area retention time range (+/-) 1.00 minutes minimum quality 0.50 minimum signal/blank ratio 2.00 minimum signal/baseline ratio 2.00 30 This process was executed with both samples and pure standards. Fragmentation data from the standards were first incorporated into the local metabolomics library to empower and inform all sample annotations at Level 1. Annotations at Level 2 were begun using external libraries and web databases such as The Human Metabolome Database® (HMDB) and LIPID MAPS®. Statistical Techniques In the SPSS™ v28 (IBM™) program, data were initially preserved as one group of fifty (50). A Shapiro-Wilk test for Normality was performed and descriptive statistics including mean, standard deviation, and coefficient of variation were collected. The significance level used was five percent (0.05). For sets determined by Shapiro-Wilk to be Normally distributed, a “Pooled Equal Variance” t-test would have been run on the binary variable: human immunodeficiency virus (HIV) status (Yakah et al., 2019). Dependent variables for which the Shapiro-Wilk test statistic was below five percent were concluded to have evidence against a Normality assumption. By Shapiro-Wilk distribution testing, each of the following dependent variables (with units of peak area or relative abundance) was labeled as non-Normal (p < 0.05): total calcidiol, total ercalcidiol, sum of 25-(OH)D, sum of hormones, sum of oxysterols, sum of phytosterols, and total cholesterol. Concentration ratios calcidiol to ercalcidiol, sum of oxysterols to cholesterol, and lathosterol to cholesterol each were found non-Normal, as well. As such, suitable statistical analyses for all dependent variables of current concern to the project were restricted to nonparametric tests for non-Normal distributions. The Mann-Whitney U-Test would provide the most appropriate demonstration of HIV infection as potentially related to any metabolite measure. This is because Mann-Whitney is designed for cases with binary groups, as seen here with “positive” (HIV) and “control” 31 classifications of all biospecimens in the cohort. This second statistical step was saved for future work due to the limitation in preliminary cohort size. Because the study was committed to satisfy a blind test with respect to HIV status of the individuals, only upon data interpretation was the preliminary cohort selection (of n = 50) revealed to comprise of uneven groups. More specifically, forty (40) biospecimens in the cohort belonged to HIV-positive subjects while the remaining ten (10) represented HIV-negative individuals. With just ten controls in this cohort, the purpose of this preliminary serum analysis was refined: to assess the quality of the method in extracting lipid metabolites from the matrix of focus: human serum. 32 CHAPTER 5: RESULTS, DISCUSSION, AND OUTCOMES Method Validation using Breastmilk The test run with human breastmilk as defined in Chapter 3 was completed with the adapted metabolomics pipeline. HPLC-MS/MS data was converted and viewed as chromatograms in El-MAVEN™ v12.0. Annotations at Level 1 were completed for comparisons of mass fragmentation and retention time data to those of external, pure standards. One of the species analyzed was cholesterol. Figure 5-1 depicts the observed mass fragmentation pattern of cholesterol in a feature called Cholesterol-H2O (“cholesterol minus one water”). The unique spectral properties of cholesterol made for straightforward identification, as the mass-to-charge ratio (m/z) of the dominant fragment (see 369.351746) happened to very closely reflect the molecular mass in Daltons of Cholesterol-H2O (the “precursor”). Figure 5-1. Mass fragmentation of cholesterol-H2O in breastmilk test sample Another DSHO metabolite verified by this test was calcitriol, the most bioactive form of vitamin D3 in serum. Figure 5-2(a) shows the observed spectra from pure calcitriol standard. It is 33 a near match of Figure 5-2(b), which reveals the observed spectra within a breastmilk sample for a feature identified accordingly as calcitriol. (a) (b) Figure 5-2. (a) Mass spectra observed for calcitriol-H2O in pure Calcitriol standard; (b) Mass spectra recorded for feature of calcitriol-H2O within breastmilk test sample Also noted as evidence in the above determination of calcitriol was the narrow retention time shift of just 0.20 minutes between standard and sample features. The minimization of shift in retention time for any analyte across multiple distinct runs proves reproducibility and reliability of the method. As shared previously by Table 4-1, a retention time delta of 1.00 minute was used. For any precursor ion found in the spectra of a sample feature farther than one minute from the retention time exhibited by the pure standard, the annotation was inconclusive at Level 1 and further levels were necessary to verify the identity of the compound. Final annotations yielded 34 similarly successful verifications for all six vitamin D metabolites run with internal standards and for all semi-targeted sterol and hormone species. Some metabolites required annotation of multiple distinct features. These features’ areas then were summed to define the total metabolite abundance that entered the equipment. Another measure of calcitriol was made by analysis of calcitriol-2H2O. Shown in Figure 5-3 is an abbreviated chromatogram view. Notably, the triplicates of breastmilk extract showed strong consistency between peak areas. Some variation is to be expected as the three samples, though collected from the same aliquot tube, were subject to biological variability due to heterogeneity of the original matrix. Figure 5-3. Chromatogram view of calcitriol-H2O features in three breastmilk samples By integrating the peak-like features produced by all three samples for each compound annotated at Level 1 in this test, dimensionless values of area were obtained and recorded in Table 35 5-1 below. Peak areas cannot be applied units of mass or concentration or directly define serum metabolite levels (or “deficiencies”). They can, however, be compared between areas of metabolites within each sample, as a dimensionless ratio of area (i.e. calcitriol divided by its precursor, calcidiol) equals the fraction that would be obtained by ratio of true concentrations of the same two species. Table 5-1. Peak areas for each vitamin D metabolite of type D 3 and D2, followed by average, standard deviation, and CV across samples Vitamin D3 25-(OH) D3 Calcitriol Vitamin D2 25-(OH) D2 Ercalcitriol VDBM 1 2.72E+06 3.51E+06 2.27E+07 1.82E+07 5.25E+06 1.22E+07 VDBM 2 3.10E+06 3.22E+06 2.09E+07 3.17E+06 6.20E+06 2.05E+07 VDBM 3 2.75E+06 3.07E+06 2.17E+07 2.86E+06 4.09E+06 1.21E+07 average 2.86E+06 3.27E+06 2.17E+07 8.08E+06 5.18E+06 1.50E+07 standard 2.09E+05 2.23E+05 8.82E+05 8.78E+06 1.06E+06 4.78E+06 deviation Coefficient 0.07 0.07 0.04 1.09 0.20 0.32 of Variation From these data, it was noted that the metabolites of vitamin D 2, the sub-family of the sunlight vitamin synthesized in fungal membranes, exhibited greater variation and perhaps less precision than the vitamin D3 forms capable of photosynthesis in vivo. One potential explanation is that the transport mechanisms differ slightly between D 3 and D2 metabolites in the body. Vitamin D3 has a much higher binding affinity to the DBP protein than vitamin D 2 exhibits. Such interaction with DBP, a macromolecular structure vastly larger than the individual vitamin D metabolites to which it binds, may affect the serum mechanics by differentially leading the DBP-bound D 3 content to settle toward the bottom of the original container. In this case, the depth of the pipet 36 insertion during aliquot isolation may contribute to the sample having more or less of either sub- family of the vitamin, depending on the binding protein’s identity and abundance. Throughout this test run, the purpose remained to confirm the efficacy of the method for vitamin D metabolites and other secosteroids, as well as sterols and hormones. By some minor revisions to internal standard and aliquot volumes, this was achieved, and the novel method was ready for DSHO extraction from human serum biospecimens of the Uganda cohort. Metabolites Observed in Serum by HIV Status Upon lipid extraction, HPLC-MS/MS injection, and data acquisition of the fifty serum biospecimens, a total of thirty-two species of phytosterols (5), oxysterols (7), other sterols (3) and hormones (17) were identified at Level 1. These metabolites are grouped by Figure 5-4. Phytosterols, Oxysterols, Hormones Steroids • β-sitosterol •17,β-Estradiol •Brassicasterol •24-Dihydrolanosterol •Campesterol •2-Hydroxyestradiol •Ergosterol •Aldosterone •Stigmasterol •Danazol •Dehydroepiandrosterone •22-Hydroxycholesterol •Epinephrine •24-Hydroxycholesterol •Estriol •27-Hydroxycholesterol •Estrone •5,6,β-Epoxycholesterol •Ethisterone •7-Hydroxycholesterol •Fucosterol •7β-Hydroxycholesterol •Hydrocortisone •7-Ketocholesterol •Lanosterol •Lathosterol •25-OH-D3 ("Calcidiol") •Norepinephrin •25-OH-D2 ("Ercalcidiol") •Progesterone •Cholesterol •Squalene (a precursor) Figure 5-4. Chemical classification of all species annotated with the described pipeline 37 Lathosterol is one metabolite of cholesterol that was identified, measured, and grouped according to subject HIV status. Known as the “L:C ratio,” lathosterol peak area to cholesterol peak area is functionally equivalent to lathosterol concentration divided by cholesterol concentration in the sample. It can be used on individual bases to forecast the magnitude of reduction in total serum cholesterol among patients with diets that include phytosterols (Mackay, Gebauer, Eck, Baer, & Jones, 2015). Serum lathosterol levels have long been scientifically regarded as one of the most effective indicators of cholesterol synthesis in vivo (Duane, 1995). Listed in the bottom entry of Table A2 in the Appendix and per sample in Tables S3a and S3b, L:C ratio seemed to produce unique values according to the HIV-positive and control groups. This preliminary hypothesis will be assessed with sufficient statistical power upon the data acquisition phase of an upcoming run with 148 specimens (89 HIV-positive, 59 control). It was hereby suggested that an individual’s history with HIV infection may impact their observed lathosterol to cholesterol ratio and, by extension, their ability to regulate serum cholesterol. Table 5-2 gives the summed areas for all metabolites found in each molecular class. Even with the limited sample size of fifty, data communicated clear distinctions by orders of magnitude between the sums of oxysterols and phytosterols in HIV-positive serum, for example. Worth noting was that cholesterol, a zoosterol with synthetic and regulatory pathways in vivo, remained verifiably dominant in relative abundance compared to plant sterols, oxidative products, and hormones across groups. 38 Table 5-2. Sums of each class of metabolites found, by HIV infection status HIV-Positive HIV-Negative (frequency, n) 40 10 Vitamin D 7.48E+05 6.53E+05 Cholesterol 4.74E+07 5.03E+07 Phytosterols 5.66E+05 5.40E+05 Oxysterols9 1.06E+07 1.54E+07 Hormones 3.38E+06 2.97E+06 This result was echoed by the computation of relative metabolite abundances for each serum biospecimen in this preliminary cohort, given by Tables A4a and A4b in the Appendix. Calcidiol to Ercalcidiol Ratio The chromatographic peak area measured for each compound in each sample is displayed in Tables A3a and A3b in the Appendix. Peak area represents a unique measure of molecular abundance that can be compared between annotated compounds similarly to theoretical serum concentrations in nanograms per milliliter. Area as a form of data quantification, though, should not be analyzed alone toward conclusions in nutrient deficiency for individuals. Calcidiol to ercalcidiol, constructed as a ratio of observed concentrations, can provide meaningful insight into sources of vitamin D nourishment. For the serum cohort from Uganda, a landlocked nation with a warm equatorial climate, a reasonable hypothesis may hold that vitamin D3 produced in vivo would be elevated, while diet diversity may be reduced and consequently limit exposure to the vitamin D2 forms of fungal origins. This ratio may one day lead to improved 9 “Sum of Oxysterols” is emphasized as distinct from “Total Oxysterols” in this case. 39 understanding of the contrast between human health functions of D 3 metabolites (such as bioactive calcitriol) and D2 forms (such as bioactive ercalcitriol) when applied to subject groups of contrasting health profiles. Cholesterol Oxidation Ratio Oxysterols comprise one dietary branch of cholesterol oxidative products (COP), byproducts of cholesterol proven to be elevated in serum concentration among cases of several chronic diseases (Poli, Biasi, & Leonarduzzi, 2013). As such, dividing the sum of measured oxysterol species by the abundance of cholesterol in the same matrix can provide insight on the rate or regulation of certain COP within the biospecimen or its subject. One limitation to the computations for sums of oxysterol species described by Tables S2, S3a, S3b, S4a, and S4b is that an ideal fraction of oxysterols to total cholesterol would specify and measure all known oxysterol species present in the sample matrix, whereas the current status of this report quantified the seven (7) commonly and independently occurring oxysterols observed in serum. The evidence of untargeted oxysterol metabolites present in these complex, disease- affected biospecimens reinforces that current data may not represent the true ratio of all serum oxysterols to cholesterol. Challenge of Coelusion All compounds run in this project externally as pure standards were resolved well by data acquisition. The internal standard calcitriol-d6, though, was one compound that could not be annotated with the present data processing approach. This is because the unique molecular formulae of deuterated compounds such as calcitriol-d6 are marked semi-formally by commercial 40 vendors with “D” to describe the number of deuterations 10. Data processing software conveniently store chemical formulae input by the user to create local databases and enable annotation at Level 1. Without the functionality to recognize deuterations (D) as listed by the user, the deuterated compound used in this project could not be precisely defined, searched for, or annotated in El- MAVEN™ v12.0. Alternative lipidomic software brands and MAVEN versions that may be able to comprehend deuterated compounds in mass spectroscopy data were considered as future steps for the project. Because it is strictly synthetic and does not occur in nature, calcitriol-d6 was the only standard whose measurement was affected by this limitation. The other internal standard used, 19-hydroxycholesterol, remained identifiable throughout the data processing step. Figure 5-5 shows the observed mass spectra for the dominant feature of a pure 19-OH standard injection with HPLC-MS/MS equipment. Figure 5-5. Mass fragmentation (or “spectra”) recorded by run of pure 19-OH standard The spectral data shown for 19-hydroxycholesterol by Figure 5-5 are of special import because, as hinted by Chapter 3, the peak area due to the internal standard in each sample can 10 atoms in an overall compound to which one neutron is added, allowing distinction from naturally-occurring, non- deuterated species 41 often be used to transform a specific analyte’s peak area into mass (or concentration) drawn from the original matrix. Thirty-two (32) species of phytosterols, oxysterols, cholesterol, secosteroids, and hormones were detected by the novel method with high resolution. Many untargeted features yet unidentified are hypothesized to be present in the sample data, as well. Their study is motivated by an analytical challenge called coelusion. The HPLC-MS/MS data acquisition method used in this project consisted of a series of distinct conditions and mechanical activities detailed in Chapter 4. This technological practice was not engineered from scratch during this project. Rather, it came as a close variation of a published method11 used regularly by the Food and Health Engineering Laboratory for the extraction of oxysterols and cholesterols from other biological matrices. The current project marked the first application of a variation of this method upon human serum. In its approximately 49 minutes of DAS run time, some hypothesized, yet untargeted features appeared to fragment at retention times that overlapped with targeted analytes in sample chromatograms. In more technical terms, the diversity of steroid metabolites that were successfully extracted and retained by the novel method saw some chemical members elute in mass spectrometer at the times similar or close to others’, increasing the overall peak areas observed at affected retention times and challenging the distinction and integration of features. This challenge of coelusion took place more frequently among oxysterols than other targeted classes. This was most likely due to the elevated abundances of cholesterol oxidative products (COP) expected in serum of subjects suffering from chronic disease like HIV/AIDS (Poli et al., 2013). The standard 19-hydroxycholesterol is not a naturally occurring member of the oxysterol family. Evidence was collected that suggested 19-OH ionization in the mass spectrometer took 11 Ketner, AB. In Vitro Cellular & Developmental Biology - Animal Volume: 56 Issue SUPPL 1 (2020) ISSN: 1071- 2690 Online ISSN: 1543-706X 42 place with coelusion of natural oxysterols from the serum matrix. This leads to a uniquely challenging case of coelusion, as distinct oxysterols often have equal molecular weights or precursor values across isomers (i.e. 7β-OH, 19-OH, 22-OH, etc.) Evidence includes Figure 5-6, which illustrates the recently shown spectra of 19-OH pure standard in its part (a). Spectra recorded in sample data at the retention time expected for 19-OH are shown by Figure 5-6(b)(c). 43 (a) (b) (c) Figure 5-6. (a) Mass fragmentation (or “spectra”) recorded by run of pure 19-OH standard; (b, c) Fragmentation spectra observed in two distinct biospecimens’ data for 19- hydroxycholesterol derivative compound 44 The mass spectra observed in Figure 5-6(a) by pure external standard became the expected pattern prior to comparison with features at comparable retention time (for 19-OH) within the samples. This was demonstrated first in the case of the validation run with human breastmilk. The serum biospecimen spectra shown in Figure 5-6(b) and Figure 5-6(c) show many similar ions and intensities between them. However, precursor and primary fragments expected for 19-OH appeared reduced in relative intensity compared to new ions that may represent untargeted, unannotated metabolites separate from 19-OH. Specifically, the ions or fragments occurring in the serum runs at m/z values such as 369.35 and 564.44 were uncharacteristic of 19- hydroxycholesterol as known by pure standard. As a direct consequence, some signature fragments of 19-OH like 425.34 appeared greatly reduced in intensity, if seen at all. Bioactive Metabolites Challenge Thirty-two metabolites of vitamin D, sterol and hormone classes were annotated sensitively and without coelusion of other species, rendering the metabolomics pipeline and data processing methods a strong success and worthy of maintenance and revision. Two of these compounds measured with high resolution were calcidiol and ercalcidiol, the two most abundant forms of the vitamin in human serum. The bioactive forms calcitriol and ercalcitriol were confirmed in detection, as well, and their method-specific retention times and spectral profiles obtained and stored for use with sample data. These bioactive forms within the serum specimens, however, were subject to coelusion with untargeted hormones. This is evidenced in part by the prevalence of molecular components that fragment relatively high in mass (exceeding 500 Da), which may contradict the notion of calcitriol identity by significantly reporting fragments larger in mass than 45 the pre-fragmented analyte, itself. With zoom-in and zoom-out views, Figure 5-7 and Figure 5-8 convey this phenomenon. (a) (b) Figure 5-7. (a) Mass spectra for calcitriol in sample data, zoomed-in so precursor 417.34 Daltons is visible; (b) Full-view, zoomed-out of calcitriol mass fragmentation 46 (a) (b) Figure 5-8. (a) Mass spectra for ercalcitriol in sample data, zoomed-in so precursor 429.34 Daltons is visible; (b) Full-view, zoomed-out of ercalcitriol mass fragmentation An associated challenge of coelusion was observed also with respect to cholecalciferol (D 3) and calciferol (D2) annotations. Because vitamin D is a steroid hormone, it follows that many other human hormone species may survive the described lipid extraction toward HPLC-MS/MS. This compatibility founded this method’s detection and quantification of seventeen (17) distinct, semi- targeted hormones not directly associated with vitamin D metabolism. However, this characteristic also enabled the extraction of an array of untargeted hormones that would elute with MS/MS in overlap of cholecalciferol and calciferol during the chromatography run. This would require a 47 specialization step to resolve their representative features in the data processing software. Ideas and applications for this were explored as future steps. Opportunity for Future Method Applications While the method validation established an applicable framework for associated projects, there remain opportunities for research development within the current Uganda cohort of biospecimens. For one, an experiment with increased sample size (n = 148) from the same reserves is currently underway to articulate with improved statistical power any associations between HIV- status and vitamin D concentration by metabolite. While this future-oriented phase of the project is early in data processing to date, a preliminary scan of all serum HPLC-MS/MS data from this cohort in an alternative data processing software known as XCMS™ revealed over eighteen thousand (18,000) features remaining to be annotated at Level 2 or beyond. This is demonstrated by Figure A2 in the Appendix. Principal Component Analyses (PCA) are valuable tools in the assessment of pre-annotated spectrometry data. In the Appendix, Figure A3 shows the difference in HPLC-MS/MS lipidomic profiles between HIV-positive and control groups as observed with statistical significance (p < 0.05) in a pre-process cohort of experimental balance (50 HIV+, 50 control). From this, multi- level annotations should commence to specify the identities of these diverse lipid metabolites and articulate which molecular nutrients may be most associated with the independent variable of HIV status. Recommended resources to incorporate annotation strategies at Level 2 are the MassBank™ of North America (MoNA) at University of California, Davis; LIPID MAPS®; and the Human Metabolome Database® (HMDB). Each provides academic access free of charge and compares input spectral characteristics observed with verified molecular indices. 48 It is suggested by the author to utilize MZmine 3™ (Schmid et al., 2023) in future annotations of HPLC-MS/MS data for serum vitamin D metabolomics. This software offers open source processing of mass spectrometry data and converts .RAW data automatically to compatible file types without the requirement of an external converter (Chen et al., 2022). It also has robust peak identification formulae that can distinguish features in the case of multi-analyte coelusion and, in effect, could overcome what has remained an annotative challenge in using other software. MZmine 3™ achieves this enhanced peak definition by measuring the expected fragmentation pattern for a given analyte as a fraction of the observed spectra in a single peak. For example, an experimenter may independently annotate a peak at an expected retention time for a targeted analyte but find the integrated area unrepresentatively high. By implementing the contemporary software, MZmine 3™, the experimenter can analyze automatically the measured fragmentation spectral intensity and conclude the relative abundance of the analyte’s expected pattern. This allows an individual GC-MS peak or HPLC-MS/MS feature to be spliced for the specific reporting of each metabolite in the sample. Conclusion Vitamin D is a critical nutrient in which much of the United States population is insufficient. Recent associations of vitamin D deficiency with mild cognitive impairment, COVID- 19 infection, and other conditions highlight the expanse of the nutrient’s roles, particularly beyond bone health. Calcitriol, a bioactive metabolite and low-abundance compound in serum, tends to be omitted from consideration by current studies. Distinction between bioactive D 3 and D2 functions and associations with chronic diseases remain to be explored, as well. 49 The viability of the novel extraction method was demonstrated soundly through an initial test run on human breastmilk and ultimately on the Uganda cohort of serum. Though small in sample size for the first run, the study with fifty (50) biospecimens revealed some key hypotheses, limitations, and opportunities. Serum lathosterol to cholesterol concentration ratio (L:C) is a meaningful modern indicator of cholesterol oxidation and nourishment in the body. The biosynthetic pathway for cholesterol in vivo may revert to synthesis of the metabolite lathosterol at a rate sensitive to chronic disease factors such as HIV infection. Coelusion of features in HPLC-MS/MS creates an obstacle between analyte identification and quantification steps. Some emerging software and tools may circumvent or otherwise resolve this occurrence in mass spectrometry data processing. One potential solution may entertain a “de-coeluting” code in MATLAB™, R™, or another interface by closely analyzing and fractioning the relative intensities of ions and fragmentation patterns in complex spectral events. Foremost, a cohort nearer to one hundred and fifty serum biospecimens from Uganda is being assembled and extracted for expansion of and evidence for the underlying hypothesis that infection with HIV may impact the vitamin D metabolome and balances of oxysterols, phytosterols, and hormones in serum. 50 Credit Author Statement Michael Kaven: Formal analysis, investigation, visualization, original writing. Ilce Medina-Meza: Project administration, funding acquisition, supervision, review. Conflict of interest: The authors declare no competing interest related to this work. 51 APPENDIX Table A1. Production batch ID numbers for each vitamin D standard purchased Compound Cayman Chem. Batch # Compound Cayman Chem. Batch # Vitamin D2 0604052-3 Vitamin D3 0594257-9 25-OH D2 0534531-25 25-OH D3 0487272-43 Ercalcitriol 0467683-25 Calcitriol 0601887-21 Calcitriol-d6 0509854-7 52 Figure A1. Illustrative workflow of metabolomics experimental activities (Chaleckis et al., 2019) Table A2. Ratios calcidiol-to-ercalcidiol, etc. by HIV infection status; concentration basis Analyte Group Ratios HIV-Positive HIV-Negative (frequency, n) 40 10 25-(OH) D3 25-(OH) D2 5.23 1 4.16 1 Oxysterols Cholesterol 0.31 1 0.32 1 Lathosterol Cholesterol 0.010 1 0.005 1 53 Table A3a. Calcidiol, ercalcidiol, total 25-(OH) D, hormones, oxysterols, phytosterols, and cholesterol peak areas; concentration basis ratios for 25-(OH)D type, cholesterol oxidation status, and lathosterol-to-cholesterol in serum biospecimens #1-26 HIV Sum Sum Sum Sum 25-(OH)D3 / Oxyst./ Lathosterol # +/- Calcidiol Ercalcidiol 25-(OH)D Hormones Oxysterols Phytosterols Cholesterol 25-(OH)D2 Cholest. /Cholest. 1 + 340172 139384 479556 2302918 5250620 323770 36800000 2.441 0.143 n/a 2 + 497490 97171 594661 2641831 7726400 542320 13400000 5.120 0.577 0.0042 3 + 392302 220297 612599 4888650 15832200 215710 16700000 1.781 0.948 0.0048 4 + 469523 119747 589270 3053800 10520300 475430 42600000 3.921 0.247 0.0011 5 + 370576 96241 466817 2193435 13145700 430380 52000000 3.851 0.253 0.0025 6 + 519026 199129 718155 4476065 11160900 325090 56100000 2.606 0.199 0.0027 7 + 626471 199862 826333 3216905 13948500 590000 50200000 3.135 0.278 0.0013 8 + 807912 143831 951743 3913766 12876200 606340 38300000 5.617 0.336 0.0182 9 + 526548 92325 618872 4711346 10234100 581850 38600000 5.703 0.265 0.0049 10 + 912381 159688 1072070 4463600 11402300 709510 20900000 5.714 0.546 0.0387 11 + 718049 141246 859295 2730031 8873500 501610 70500000 5.084 0.126 0.0053 12 + 1007546 78187 1085733 4593100 3294700 378870 44000000 12.886 0.075 0.0186 13 + 389837 77854 467691 2313409 2855200 167430 18300000 5.007 0.156 0.0249 14 + 892024 104546 996570 2099568 15754200 801150 61100000 8.532 0.258 0.0056 15 + 891796 50626 942422 3255392 6468800 582750 44700000 17.615 0.145 0.0099 16 + 424697 91619 516315 4651710 4662700 422790 64300000 4.635 0.073 0.0011 17 + 639767 87688 727455 3583860 14078200 951680 79100000 7.296 0.178 0.0126 18 + 873830 327378 1201208 4396554 11685300 371860 24000000 2.669 0.487 0.0336 19 + 504997 115423 620420 3146933 10165200 547060 17200000 4.375 0.591 0.0463 20 + 631463 114845 746308 3118450 5246500 1198740 89000000 5.498 0.059 0.0104 21 + 812490 198942 1011432 3843075 2921600 709360 59600000 4.084 0.049 0.0154 22 + 664370 116022 780391 2716520 10207400 439500 52700000 5.726 0.194 0.0127 23 + 657033 257146 914179 4351780 6449800 739380 87500000 2.555 0.074 0.0090 24 + 546807 23659 570466 3575140 7333500 291730 41000000 23.112 0.179 0.0045 25 + 491923 134312 626235 3141552 11344900 707480 39600000 3.663 0.286 0.0104 26 + 773009 206618 979627 3168740 20327700 2068420 81400000 3.741 0.250 0.0048 54 Table A3b. Calcidiol, ercalcidiol, total 25-(OH) D, hormones, oxysterols, phytosterols, and cholesterol peak areas; concentration-basis ratios for 25-(OH)D type, cholesterol oxidation status, and lathosterol-to-cholesterol in serum biospecimens #27-50 HIV Sum Sum Sum Sum 25-(OH)D3 / Oxyst./ Lathosterol/ # +/- Calcidiol Ercalcidiol 25-(OH)D Hormones Oxysterols Phytosterols Cholesterol 25-(OH)D2 Cholest. Cholest. 27 + 748125 437262 1185387 4133450 7734500 1381010 48400000 1.711 0.160 0.0052 28 + 371414 129708 501123 3410000 13181000 361310 61200000 2.863 0.215 0.0035 29 + 294716 161510 456226 2850686 10232900 507250 38500000 1.825 0.266 0.0073 30 + 455097 176017 631114 2670401 12063700 794640 82100000 2.586 0.147 0.0029 31 + 337477 64369 401847 1849926 16974900 472690 32600000 5.243 0.521 0.0252 32 + 926289 75105 1001394 5000450 13388600 668730 61500000 12.333 0.218 0.0081 33 + 896345 207271 1103616 3954260 27936200 380880 58200000 4.325 0.480 0.0208 34 + 814035 105021 919056 3969197 8730200 499160 65200000 7.751 0.134 0.0105 35 + 135142 32209 167351 2396634 9652000 206000 n/a12 4.196 n/a n/a 36 + 650958 189414 840372 2305773 6786800 185550 36400000 3.437 0.186 0.0072 37 + 293030 326417 619447 2628785 8150900 317491 44600000 0.898 0.183 0.0090 38 + 524626 521145 1045771 3744717 5122100 306910 42800000 1.007 0.120 0.0111 39 + 607639 163466 771105 2584742 18466100 202728 20500000 3.717 0.901 0.0086 40 + 148916 137339 286255 3018987 10776800 684760 58900000 1.084 0.183 0.0017 41 - 448221 113483 561704 1735692 3827800 1578910 50000000 3.950 0.077 0.0072 42 - 275329 130544 405873 2372157 9721500 422230 61700000 2.109 0.158 0.0023 43 - 451715 87644 539359 2250221 9925300 246900 20500000 5.154 0.484 0.0107 44 - 829378 153282 982660 5290760 51073800 469600 61500000 5.411 0.830 0.0036 45 - 807059 128653 935713 3016333 33447200 447130 59500000 6.273 0.562 0.0045 46 - 393249 254372 647621 1463880 4341200 302250 33900000 1.546 0.128 0.0034 47 - 898847 151421 1050268 3755270 14744700 235610 34100000 5.936 0.432 0.0121 48 - 464014 130229 594243 2992199 11457300 635030 57300000 3.563 0.200 0.0045 49 - 408837 79817 488655 3874770 5710700 736503 92400000 5.122 0.062 0.0019 50 - 231029 90973 322003 2904510 9843600 326300 31900000 2.540 0.309 0.0130 12 Sample area for cholesterol proved an outlier due to incomplete peak resolution that exempted #35 from analyses of cholesterol content and dependent ratios. 55 Table A4a. Relative abundances of select analytes and classes, per individual sample (#1-26) HIV Sum Sum Sum Sum # +/- Calcidiol Ercalcidiol 25-(OH)D Hormones Oxysterols Phytosterols Cholesterol 1 + 0.75% 0.31% 1.05% 5.05% 11.51% 0.71% 80.64% 2 + 1.95% 0.38% 2.33% 10.36% 30.30% 2.13% 52.55% 3 + 1.01% 0.57% 1.58% 12.58% 40.74% 0.56% 42.97% 4 + 0.81% 0.21% 1.02% 5.28% 18.19% 0.82% 73.67% 5 + 0.54% 0.14% 0.68% 3.19% 19.13% 0.63% 75.69% 6 + 0.71% 0.27% 0.98% 6.09% 15.19% 0.44% 76.33% 7 + 0.90% 0.29% 1.19% 4.62% 20.04% 0.85% 72.12% 8 + 1.40% 0.25% 1.65% 6.79% 22.35% 1.05% 66.49% 9 + 0.95% 0.17% 1.12% 8.51% 18.48% 1.05% 69.72% 10 + 2.30% 0.40% 2.71% 11.27% 28.78% 1.79% 52.75% 11 + 0.85% 0.17% 1.02% 3.24% 10.52% 0.59% 83.61% 12 + 1.85% 0.14% 1.99% 8.44% 6.05% 0.70% 80.83% 13 + 1.59% 0.32% 1.90% 9.42% 11.62% 0.68% 74.48% 14 + 1.09% 0.13% 1.22% 2.57% 19.27% 0.98% 74.74% 15 + 1.57% 0.09% 1.66% 5.72% 11.37% 1.02% 78.57% 16 + 0.57% 0.12% 0.69% 6.20% 6.21% 0.56% 85.65% 17 + 0.65% 0.09% 0.73% 3.61% 14.20% 0.96% 79.76% 18 + 2.04% 0.76% 2.80% 10.26% 27.27% 0.87% 56.00% 19 + 1.56% 0.36% 1.92% 9.74% 31.47% 1.69% 53.25% 20 + 0.63% 0.11% 0.75% 3.12% 5.24% 1.20% 88.95% 21 + 1.18% 0.29% 1.46% 5.56% 4.23% 1.03% 86.26% 22 + 0.98% 0.17% 1.15% 4.02% 15.09% 0.65% 77.93% 23 + 0.65% 0.25% 0.91% 4.31% 6.39% 0.73% 86.75% 24 + 1.03% 0.04% 1.07% 6.70% 13.75% 0.55% 76.86% 25 + 0.88% 0.24% 1.12% 5.61% 20.24% 1.26% 70.66% 26 + 0.71% 0.19% 0.90% 2.91% 18.66% 1.90% 74.73% 56 Table A4b. Relative abundances of select analytes and classes, continued (#27-50 13), with averages by HIV variable HIV Sum Sum Sum Sum # +/- Calcidiol Ercalcidiol 25-(OH)D Hormones Oxysterols Phytosterols Cholesterol 27 + 1.17% 0.68% 1.85% 6.46% 12.08% 2.16% 75.60% 28 + 0.47% 0.16% 0.63% 4.31% 16.65% 0.46% 77.32% 29 + 0.56% 0.30% 0.86% 5.38% 19.31% 0.96% 72.64% 30 + 0.46% 0.18% 0.64% 2.70% 12.20% 0.80% 83.02% 31 + 0.64% 0.12% 0.76% 3.51% 32.21% 0.90% 61.86% 32 + 1.12% 0.09% 1.21% 6.06% 16.22% 0.81% 74.49% 33 + 0.97% 0.22% 1.19% 4.27% 30.14% 0.41% 62.80% 34 + 1.01% 0.13% 1.15% 4.95% 10.88% 0.62% 81.26% 36 + 1.37% 0.40% 1.77% 4.87% 14.33% 0.39% 76.86% 37 + 0.51% 0.57% 1.09% 4.62% 14.32% 0.56% 78.33% 38 + 0.97% 0.96% 1.93% 6.93% 9.47% 0.57% 79.16% 39 + 1.40% 0.38% 1.78% 5.97% 42.65% 0.47% 47.35% 40 + 0.20% 0.19% 0.39% 4.08% 14.57% 0.93% 79.65% 41 - 0.77% 0.19% 0.96% 2.98% 6.57% 2.71% 85.81% 42 - 0.37% 0.17% 0.54% 3.16% 12.96% 0.56% 82.24% 43 - 1.33% 0.26% 1.59% 6.62% 29.19% 0.73% 60.29% 44 - 0.69% 0.13% 0.82% 4.40% 42.46% 0.39% 51.12% 45 - 0.82% 0.13% 0.95% 3.07% 34.03% 0.45% 60.54% 46 - 0.95% 0.62% 1.57% 3.54% 10.51% 0.73% 82.08% 47 - 1.64% 0.28% 1.91% 6.84% 26.84% 0.43% 62.07% 48 - 0.63% 0.18% 0.81% 4.07% 15.57% 0.86% 77.88% 49 - 0.39% 0.08% 0.47% 3.74% 5.51% 0.71% 89.10% 50 - 0.51% 0.20% 0.71% 6.37% 21.58% 0.72% 69.93% Average 1.03% 0.28% 1.30% 5.88% 17.73% 0.91% 72.88% HIV (+) Average 0.81% 0.22% 1.03% 4.48% 20.52% 0.83% 72.11% Control 13 Sample #35, noted previously as an outlier with respect to cholesterol detection, was thus exempt from relative abundance (%) computations. 57 Figure A2. Cloud plot of all features from HPLC-MS/MS with expanded cohort (n = 148) HIV-positive HIV-negative Figure A3. Principal Component Analysis (PCA) applied with preliminary application to pre-process, “balanced” sample selection (50 HIV+, 50 control) 58 REFERENCES Abu Kassim, N. S., Shaw, P. N., & Hewavitharana, A. K. (2018). Simultaneous determination of 12 vitamin D compounds in human serum using online sample preparation and liquid chromatography-tandem mass spectrometry. J Chromatogr A, 1533, 57-65. doi:10.1016/j.chroma.2017.12.012 Ahmed, L. H. M., Butler, A. E., Dargham, S. R., Latif, A., Chidiac, O. M., Atkin, S. L., & Abi Khalil, C. (2020). Vitamin D3 metabolite ratio as an indicator of vitamin D status and its association with diabetes complications. BMC Endocrine Disorders. Banerjee, A., Khemka, V. K., Ganguly, A., Roy, D., Ganguly, U., & Chakrabarti, S. (2015). Vitamin D and Alzheimer's Disease: Neurocognition to Therapeutics. Int J Alzheimers Dis, 2015, 192747. doi:10.1155/2015/192747 Batai, K., Cui, Z., Arora, A., Shah-Williams, E., Hernandez, W., Ruden, M., . . . Kittles, R. A. (2021). Genetic loci associated with skin pigmentation in African Americans and their effects on vitamin D deficiency. PLoS Genet, 17(2), e1009319. doi:10.1371/journal.pgen.1009319 Bikle, D. D. (2014). Vitamin D metabolism, mechanism of action, and clinical applications. Chem Biol, 21(3), 319-329. doi:10.1016/j.chembiol.2013.12.016 Borel, P., Caillaud, D., & Cano, N. J. (2015). Vitamin D bioavailability: state of the art. Crit Rev Food Sci Nutr, 55(9), 1193-1205. doi:10.1080/10408398.2012.688897 Bouillon, R., & Carmeliet, G. (2018). Vitamin D insufficiency: Definition, diagnosis and management. Best Pract Res Clin Endocrinol Metab, 32(5), 669-684. doi:10.1016/j.beem.2018.09.014 Chaleckis, R., Meister, I., Zhang, P., & Wheelock, C. E. (2019). Challenges, progress and promises of metabolite annotation for LC-MS-based metabolomics. Curr Opin Biotechnol, 55, 44-50. doi:10.1016/j.copbio.2018.07.010 Chen, Y., Li, E. M., & Xu, L. Y. (2022). Guide to Metabolomics Analysis: A Bioinformatics Workflow. Metabolites, 12(4). doi:10.3390/metabo12040357 Chun, R. F. (2012). New perspectives on the vitamin D binding protein. Cell Biochem Funct, 30(6), 445-456. doi:10.1002/cbf.2835 Crouch, P. J., Barnham, K. J., Bush, A. I., & White, A. R. (2006). Therapeutic treatments for Alzheimer's disease based on metal bioavailability. Drug News Perspect, 19(8), 469-474. doi:10.1358/dnp.2006.19.8.1021492 Cui, A., Zhang, T., Xiao, P., Fan, Z., Wang, H., & Zhuang, Y. (2023). Global and regional prevalence of vitamin D deficiency in population-based studies from 2000 to 2022: A 59 pooled analysis of 7.9 million participants. Frontiers in Nutrition, 10. doi:10.3389/fnut.2023.1070808 Dror, A. A., Morozov, N., Daoud, A., Namir, Y., Yakir, O., Shachar, Y., . . . Sela, E. (2022). Pre- infection 25-hydroxyvitamin D3 levels and association with severity of COVID-19 illness. PLoS One, 17(2), e0263069. doi:10.1371/journal.pone.0263069 Du, X., Aristizabal-Henao, J. J., Garrett, T. J., Brochhausen, M., Hogan, W. R., & Lemas, D. J. (2022). A Checklist for Reproducible Computational Analysis in Clinical Metabolomics Research. Metabolites, 12(1). doi:10.3390/metabo12010087 Duane, W. C. (1995). Serum lathosterol levels in human subjects reflect changes in whole body cholesterol synthesis induced by lovastatin but not dietary cholesterol. J Lipid Res, 36(2), 343-348. Elia, J., Carbonnelle, D., Logé, C., Ory, L., Huvelin, J. M., Tannoury, M., . . . Nazih, H. (2019). 4-cholesten-3-one decreases breast cancer cell viability and alters membrane raft- localized EGFR expression by reducing lipogenesis and enhancing LXR-dependent cholesterol transporters. Lipids Health Dis, 18(1), 168. doi:10.1186/s12944-019-1103-7 Gezen-Ak, D., Dursun, E., & Yilmazer, S. (2014). The Effect of Vitamin D Treatment On Nerve Growth Factor (NGF) Release From Hippocampal Neurons. Noro Psikiyatr Ars, 51(2), 157-162. doi:10.4274/npa.y7076 Holick, M. F. (2009). Vitamin D status: measurement, interpretation, and clinical application. Ann Epidemiol, 19(2), 73-78. doi:10.1016/j.annepidem.2007.12.001 Inau, E. T., Sack, J., Waltemath, D., & Zeleke, A. A. (2021). Initiatives, Concepts, and Implementation Practices of FAIR (Findable, Accessible, Interoperable, and Reusable) Data Principles in Health Data Stewardship Practice: Protocol for a Scoping Review. JMIR Res Protoc, 10(2), e22505. doi:10.2196/22505 Isabirye, N., Ezeamama, A. E., Kyeyune-Bakyayita, R., Bagenda, D., Fawzi, W. W., & Guwatudde, D. (2020). Dietary Micronutrients and Gender, Body Mass Index and Viral Suppression Among HIV-Infected Patients in Kampala, Uganda. Int J MCH AIDS, 9(3), 337-349. doi:10.21106/ijma.362 Jenkinson, C., Desai, R., Slominski, A. T., Tuckey, R. C., Hewison, M., & Handelsman, D. J. (2021). Simultaneous measurement of 13 circulating vitamin D3 and D2 mono and dihydroxy metabolites using liquid chromatography mass spectrometry. Clin Chem Lab Med, 59(10), 1642-1652. doi:10.1515/cclm-2021-0441 Jäpelt, R. B., & Jakobsen, J. (2013). Vitamin D in plants: a review of occurrence, analysis, and biosynthesis. Front Plant Sci, 4, 136. doi:10.3389/fpls.2013.00136 60 Kummu, M., & Varis, O. (2011). The world by latitudes: A global analysis of human population, development level and environment across the north-south axis over the past half century. Applied Geography, 31(2), 495-507. doi:10.1016/j.apgeog.2010.10.009 Leal, A. C. G. B., Corrêa, M. P., Holick, M. F., Melo, E. V., & Lazaretti-Castro, M. (2021). Sun- induced production of vitamin D. Photochem Photobiol Sci, 20(2), 265-274. doi:10.1007/s43630-021-00015-z Lipkie, T. E., Janasch, A., Cooper, B. R., Hohman, E. E., Weaver, C. M., & Ferruzzi, M. G. (2013). Quantification of vitamin D and 25-hydroxyvitamin D in soft tissues by liquid chromatography-tandem mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci, 932, 6-11. doi:10.1016/j.jchromb.2013.05.029 Lips, P., Bilezikian, J. P., & Bouillon, R. (2020). Vitamin D: Giveth to Those Who Needeth. JBMR Plus, 4(1), e10232. doi:https://doi.org/10.1002/jbm4.10232 Littlejohns, T. J., Henley, W. E., Lang, I. A., Annweiler, C., Beauchet, O., Chaves, P. H., . . . Llewellyn, D. J. (2014). Vitamin D and the risk of dementia and Alzheimer disease. Neurology, 83(10), 920-928. doi:10.1212/WNL.0000000000000755 Mackay, D. S., Gebauer, S. K., Eck, P. K., Baer, D. J., & Jones, P. J. (2015). Lathosterol-to- cholesterol ratio in serum predicts cholesterol-lowering response to plant sterol consumption in a dual-center, randomized, single-blind placebo-controlled trial. Am J Clin Nutr, 101(3), 432-439. doi:10.3945/ajcn.114.095356 Makin, H. L. J., Jones, G., Kaufmann, M., & Calverley, M. J. (2010). Analysis of Vitamins D, Their Metabolites and Analogues. In H. L. J. Makin & D. B. Gower (Eds.), Steroid Analysis (pp. 967-1096). Dordrecht: Springer Netherlands. Matyash, V., Liebisch, G., Kurzchalia, T. V., Shevchenko, A., & Schwudke, D. (2008). Lipid extraction by methyl-tert-butyl ether for high-throughput lipidomics. J Lipid Res, 49(5), 1137-1146. doi:10.1194/jlr.D700041-JLR200 Mayne, P. E., & Burne, T. H. J. (2019). Vitamin D in Synaptic Plasticity, Cognitive Function, and Neuropsychiatric Illness. Trends Neurosci, 42(4), 293-306. doi:10.1016/j.tins.2019.01.003 Melmed, S. (2020). Hormones and Disorders of Mineral Metabolism. In Williams Textbook of Endocrinology (14 ed.). Mena-Bravo, A., Ferreiro-Vera, C., Priego-Capote, F., Maestro, M. A., Mouriño, A., Quesada- Gómez, J. M., & Luque de Castro, M. D. (2015). Quantitative analytical method to evaluate the metabolism of vitamin D. Clin Chim Acta, 442, 6-12. doi:10.1016/j.cca.2014.12.039 61 Müller, M. J., & Volmer, D. A. (2015). Mass spectrometric profiling of vitamin D metabolites beyond 25-hydroxyvitamin D. Clin Chem, 61(8), 1033-1048. doi:10.1373/clinchem.2015.241430 Ouma, S., Suenaga, M., Hatip, F. F. B., Hatip-Al-Khatib, I., Tsuboi, Y., & Matsunaga, Y. (2018). Serum vitamin D in patients with mild cognitive impairment and Alzheimer's disease. Brain and Behavior, 8(3). doi:10.1002/brb3.936 Pfaender, S., & Grabrucker, A. M. (2014). Characterization of biometal profiles in neurological disorders†. Metallomics, 6(5), 960-977. doi:10.1039/c4mt00008k Pobee, R. A., Fenton, J. I., Sikorskii, A., Zalwango, S. K., Felzer-Kim, I., Medina, I. M., . . . Ezeamama, A. E. (2022). Association of serum PUFA and linear growth over 12 months among 6-10 years old Ugandan children with or without HIV. Public Health Nutr, 25(5), 1-11. doi:10.1017/S1368980022000611 Poli, G., Biasi, F., & Leonarduzzi, G. (2013). Oxysterols in the pathogenesis of major chronic diseases. Redox Biology, 1(1), 125-130. Saenger, A. K., Laha, T. J., Bremner, D. E., & Sadrzadeh, S. M. (2006). Quantification of serum 25-hydroxyvitamin D(2) and D(3) using HPLC-tandem mass spectrometry and examination of reference intervals for diagnosis of vitamin D deficiency. Am J Clin Pathol, 125(6), 914-920. doi:10.1309/J32U-F7GT-QPWN-25AP Schmid, R., Heuckeroth, S., Korf, A., Smirnov, A., Myers, O., Dyrlund, T. S., . . . Pluskal, T. (2023). Integrative analysis of multimodal mass spectrometry data in MZmine 3. Nat Biotechnol, 41(4), 447-449. doi:10.1038/s41587-023-01690-2 Sun, Y., Nzekoue, F. K., Vittori, S., Sagratini, G., & Caprioli, G. (2022). Conversion of ergosterol into vitamin D2 and other photoisomers in Agaricus bisporus mushrooms under UV-C irradiation. Food Bioscience, 50. Sweeley, C., Bentley, R., Makita, M., & Wells, W. (1963). Gas-liquid chromatography of trimethylsilyl derivatives of sugars and related substances. Journal of the American Chemical Society, 85(16), 2497-2507. Tuckey, R. C., Cheng, C. Y. S., & Slominski, A. T. (2019). The serum vitamin D metabolome: What we know and what is still to discover. J Steroid Biochem Mol Biol, 186, 4-21. doi:10.1016/j.jsbmb.2018.09.003 Van Driel, M., Koedam, M., Buurman, C. J., Hewison, M., Chiba, H., Uitterlinden, A. G., . . . van Leeuwen, J. P. (2006). Evidence for auto/paracrine actions of vitamin D in bone: 1alpha-hydroxylase expression and activity in human bone cells. FASEB J, 20(13), 2417- 2419. doi:10.1096/fj.06-6374fje 62 Wang, Y., Zhu, J., & DeLuca, H. F. (2012). Where is the vitamin D receptor? Arch Biochem Biophys, 523(1), 123-133. doi:10.1016/j.abb.2012.04.001 Wishart, D. S. (2007). Proteomics and the human metabolome project. Expert Rev Proteomics, 4(3), 333-335. doi:10.1586/14789450.4.3.333 Yakah, W., Fenton, J. I., Sikorskii, A., Zalwango, S. K., Tuke, R., Musoke, P., . . . Ezeamama, A. E. (2019). Serum Vitamin D is Differentially Associated with Socioemotional Adjustment in Early School-Aged Ugandan Children According to Perinatal HIV Status and In Utero/Peripartum Antiretroviral Exposure History. Nutrients, 11(7), 1570. Yutuc, E., Angelini, R., Baumert, M., Mast, N., Pikuleva, I., Newton, J., . . . Griffiths, W. J. (2020). Localization of sterols and oxysterols in mouse brain reveals distinct spatial cholesterol metabolism. Proceedings of the National Academy of Sciences, 117(11), 5749-5760. doi:doi:10.1073/pnas.1917421117 Zavitsanos, A. P., MacDonald, C., Bassoo, E., & Gopaul, D. (1999). Determination of domperidone in human serum and human breast milk by high-performance liquid chromatography-electrospray mass spectrometry. J Chromatogr B Biomed Sci Appl, 730(1), 9-24. doi:10.1016/s0378-4347(99)00163-2 Zelzer, S., Goessler, W., & Herrmann, M. (2018). Measurement of vitamin D metabolites by mass spectrometry, an analytical challenge. J. Lab. Precis. Med, 3, 99. 63