Study Design
Plasma samples were obtained from a previously conducted study of a single bout of prolonged fasting in healthy human volunteers 37. Full clinical study protocols and inclusion and exclusion criteria can be found at clinicaltrials.gov; NCT03487679. Briefly, the study involved 20 participants, 10 males and 10 females, aged 20–40 with no reported health conditions or extreme dietary and exercise patterns. Study participants first came in on Day 1 after an overnight fast (12 hours) at around 8am (Baseline: 12h fasted, Timepoint A) for a baseline blood draw, and two hours after their last meal at around 8pm for their baseline postprandial blood draw (Fed: 2h postprandial prior to 36h of fasting, Timepoint B). Participants tracked every food and beverage consumed over the entire course of Day 1. After the evening meal on Day 1 they began their 36-hour water-only fast until the morning of Day 3, when they came in for a blood draw (Fasted: 36h fasted, Timepoint C) at around 8AM. After this blood draw they were instructed to consume the exact same foods and beverages as those they consumed on Day 1, and then came in for a final blood draw two hours after their last meal on Day 3 (Refed: 2h postprandial after 36h of fasting, Timepoint D), again at around 8PM. Participants were monitored for fasting compliance with glucose monitoring every 2 hours during the study, and ketone body concentrations were later also confirmed.
HDL Isolation
HDL particles were isolated from plasma using a validated, optimized method 13. Briefly, 500 µL starting plasma was thawed and placed under 4.1 ml density solution 1.006 g/ml of potassium bromide (KBr) in 4.7 ml OptiSeal tube (Beckman-Coulter, IN, USA). Then, samples were ultracentrifuged for 30 min at 110,000 RPM using a fixed angle rotor (TLA-110, k factor = 13, Beckman-Coulter). The top 4ml were removed from the OptiSeal and the remaining 700 µl of sample was mixed with 1.1 ml 1.34 g/ml KBr density solution to create 1.8 ml of a 1.21 g/ml KBr-Lipoprotein and plasma protein fraction. This fraction was pipetted under 2.8 ml of 1.21 g/ml KBr density solution in the OptiSeal tube, and topped off with 100 µl of 1.21 density solution for a second ultracentrifuge spin of 3 hours and 30 minutes at 110,000 RPM. After the second spin, the first 2 ml of the tube was isolated and filtered through 50 KDa Amicon Ultra-4 for 8 minutes at 4500 RPM to 250 µl, of which 200 µl was injected into a fast protein liquid chromatograph (FPLC, AKTA P-920) with a size exclusion chromatograph (SEC) column (Superdex 200 GI 10/300), and the system was run at a flow rate of 0.5 ml/min. The HDL fraction was selected based on elution time collecting between the troughs of the LDL-HDL peak and the HDL-albumin peak. The 4 mL of total eluted volume representing the HDL fraction was concentrated using Amicon 50 KDa filters, 2% sucrose as a cryoprotectant, 61 was added and samples were aliquoted into 20 µL fractions before storage at -80°C.
Proteomic and Glycoproteomic Analysis
Isolated HDL fractions were analyzed using a previously published method 62 which utilizes dynamic multiple reaction monitoring (DMRM) coupled to a Fusion Lumos MS/MS Orbitrap (Thermo Fisher Scientific) for quantifying HDL-associated proteins. Targeted glycoproteomic analysis was conducted using Agilent 1290 Infinity Liquid Chromatography coupled to an Agilent 6495B Triple Quadrupole Mass Spectrometry. A total of 47 peptides and 170 glycopeptides from 33 HDL-associated proteins were monitored, and for each protein a non-glycosylated synthetic peptide was spiked in as a reference standard for quantification. Absolute protein concentrations for Apolipoprotein-AI, Apolipoprotein-CI, Apolipoprotein-D, Apolipoprotein-E, and Clusterin were determined using a calibration curve. Relative abundances of other peptides were determined by measuring indicator peptides from each protein 63 and relative abundances of glycopeptides were quantified by the ion count ratio of glycopeptide to indicator peptide of each protein 63. Total protein was quantified using the micro BCA kit (Thermo Fisher Scientific, C23235).
NMR Lipoprofile Analysis
Lipoprotein particle sizes and concentrations were analyzed using proton nuclear magnetic resonance (NMR) spectroscopy from whole plasma. NMR Lipoprofile is an FDA approved and Clinical Laboratory Standards Institute (CLSI) certified according to guideline EP5-A2 for estimating the concentration of lipoprotein subclasses, which includes LDL particles (LDL-P), HDL particles (HDLP), and VLDL particles (VLDL-P) from small to large, as well as insulin resistance markers 64,65. Additional parameters, including ketone bodies and amino acid concentrations were also measured and previously reported 37.
Microglia Experiments
Human microglia clone 3(HMC3) were cultured using EMEM (ATCC 30-2003), penicillin-streptomycin (10,000 U/mL, Thermo Fisher Scientific, 15140122) and 10% fetal bovine serum (FBS, ATCC 30-2020). HMC3 were seeded on 96-well microplates (Corning; Costar; 3916) 40,000 cells per 200 µL, per well. Seeded cells were allowed to adhere for at least 7 hours. Following adherence, media was replaced with EMEM without phenol red, to reduce background signal (Thermo Fisher Scientific, C837K00), penicillin-streptomycin (10,000 U/mL, Thermo Fisher Scientific, 15140122) and 10% fetal bovine serum (FBS, ATCC 30-2020), along with the corresponding treatment.
Dose treatment for both cholesterol and amyloid beta oligomer (AβO) was determined in a previously published paper 42. Briefly, microglia were loaded with water soluble cholesterol at various concentrations (5, 10, 20, 30, 40, 50 µg/mL) then assessed using a Green Cytotoxicity Assay (Cat.# G8741) (Results were shown in Supplementary Fig. 2a). Further incubation testing was done on AβO treatment at various concentrations (0.1, 0.25, 0.5 µM) for 24 hours. Cytotoxicity of AβO following cholesterol loading was conducted after as well (Results were shown in Supplementary Fig. 2b, and c) 42.
Microglia were treated with AβO at 2 µM for 24 hours. At the 18h time-point, the HMC3 were loaded with water soluble cholesterol at 20 µg/mL for 6 hours. At the 21h time-point, the cells were supplemented with HDL at 0.1 mg/mL for 3 hours. At 24h, the supernatant in all wells was removed and stored at -80°C. HDL from 18 out of 20 of the individuals who participated in the study were included due to lack of sufficient material from 2 of the participants. HDL pools from the baseline 2h postprandial state and the 36h fasted state were prepared to generate adequate HDL for the in vitro experiments. HDL samples from each time point were thawed on ice and 10µL of each isolate from each participant were taken and combined to generate a pooled HDL sample from each time point with protein concentration determined by micro BCA assay (Thermo Fisher Scientific, C23235). HMC3 cells were lysed using chloroform (Sigma Aldrich, C2432), isopropyl alcohol (Fisher Scientific, 67-63-0) and Np40/ Igepal Ca 630 (Sigma Aldrich, 9002-93-1), at 7:11:0.1, respectively. The cellular content of unesterified and esterified cholesterol (CE) was measured using the Total Cholesterol Assay (Cell Biolabs, Inc, STA-390) kit following the manufacturer’s instructions. A portion of the culture supernatant was probed for ApoE content using the Human ApoE ELISA (Cell Biolabs, Inc, STA-367) following the manufacturer's instructions. The remaining portion of the supernatant was used for analysis of the lipoprotein content, size and structure via electron microscopy.
Electron Microscopy and particle analysis
Transmission electron microscopy (TEM) analysis was performed as reported 13,35 with minor modifications. Briefly, 4 microliter samples at 0.1 mg/ml were loaded on a glow-discharged 200-mesh carbon-coated TEM grid (TedPella Inc., CA, USA) for 1 minute. The grid was then blot-dried with filter paper and loaded with 2% w/v uranyl formate negative stain solution (pH 4.6), which was quickly blot-dried with a filter paper. The loading and drying of the negative stain solution was repeated 4 more times. After the last negative stain solution was removed and the grid was completely dried, it was stored in a dark environment until sample imaging. Sample grids were imaged using TEM (Talos L120C, Thermo Fisher Scientific) combined with a bottom-mounted CCD camera (Ceta, Thermo Fisher Scientific) at 36,000x magnification and 80 kV electron beam voltage.
Micrographs were processed and particle diameters were obtained using FIJI 26. Images were processed with the bandpass filter function to filter structures down to 40 pixels and up to 10 pixels. The threshold of the images was then set by the “Moments” mode. The Analyze Particles function was used to obtain particle size in area (nm2) and shape information (e.g. circularity, aspect ratio, roundness), using the following parameters: Size: 19.625–706.5 nm2; Circularity = 0.15–1.00; Exclude on edges; Include holes. Selected particle data were then filtered using additional geometrical parameters (aspect ratio < 1.5, roundness > 0.5) to remove non-spherical particles. For data analysis, particle size for each selected particle was reported as nm in diameter, which was calculated by the equation Area/ in RStudio (RStudio Team, 2020).
Statistical Analysis
Differential abundances of glycopeptides, peptides, and NMR lipoprofile features were performed as linear models (features ~ Timepoint + Subject (+ batch)) using the limma package in R (R version 4.1.0) 66. Subjects in the study were used as a fixed effect in the linear models. A one-way ANOVA-like test was used to detect differences between the means of time points. Then the differences between pairs of time points were determined by post hoc pairwise comparison. Correction for multiple hypothesis testing was performed using the Benjamini–Hochberg method. The error distribution was tested with the Shapiro-Wilk test. For data violating normality, a log transformation was applied to approach the normal distribution.
Cholesterol concentration, ApoE secretion, and particle sizes were analyzed using linear mixed effect models with the lmerTest and emmeans packages. Two-way ANOVA tests were applied to determine the changes in cholesterol concentration and ApoE secretion according to the treatments and HDL conditions. And a one-way ANOVA test was applied to detect differences in particle sizes in treatments across different HDL conditions. P-values of host hoc pairwise comparisons were adjusted using the Tukey’s method.