Infection-induced peroxisome biogenesis is a metabolic strategy for herpesvirus replication

2.1. Cell lines and primary cultures

MRC5 primary human fibroblasts (HFs) (ATCC CCL-171, passage number 18 - 28) were cultured under standard conditions (37°C and 5% CO₂) in complete growth medium (DMEM supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin antibiotics). The original MRC5 cell line, as well as the CRISPR KO cell lines produced, were authenticated by morphology check by microscope and growth curve analysis, and subject to mycoplasma detection tests. Peroxisome biogenesis disorder (PBD) dermal fibroblasts GM16513, GM16866, and GM07371 were obtained from the Coriell Institute and used within 5 passages.

2.2. Virus strains and viral infection

Viral stocks of wild-type HCMV strain AD169 and AD169-GFP, a modified strain encoding green fluorescent protein (GFP), were produced from bacteria artificial chromosomes in HFs (Wang et al., 2004; Yu et al., 2002). Adenovirus type 5 (ADV5) encoding a GFP cassette (ADV5-GFP) was a gift from Jane Flint (Chahal and Flint, 2012). Wild-type HSV-1 (strain 17+) was propagated on Vero cells. Virus stocks were kept at −80C for up to 6 months, and virus titers were determined using tissue culture infectious dose (TCID₅₀) for HCMV and plaque assays for HSV-1 and ADV5. To determine relative infectious units (IUs) produced under experimental conditions, the supernatant was used to infect a reporter plate and asses the number of infected cells with anti-IEl (HCMV), anti-ICPO (HSV-1), or anti-ICP4 (HSV-1). Cell-associated virus was collected by washing cells with PBS in the dish and scrapping in complete growth medium. The cells were subject to a single freeze/thaw cycle and sonication to release intracellular virus. To inhibit late gene expression, HCMV infections were performed with complete media supplemented with 300 μg/ml phosphonoformate (PFA). For infection with UV-irradiated virus, virus stock was irradiated using a Stratalink UV crosslinker in auto setting for five cycles. This setting was the minimum to achieve less than 99.9% infected cells as measured by anti-IE1 immunofluorescence. Infections were performed at a multiplicity of infection of 3 for HCMV, 10 for HSV-1, and 30 for ADV5.

3.1. 4-Phenylbutyric acid treatment

To induce peroxisome biogenesis, cells were treated with 5mM 4-Phenylbutyric acid (4PBA, Sigma-Aldrich) for 10 days with media changes every third day, as previously described (Kemp et al., 1998). A parallel cell culture of untreated cells was used as a control.

3.2. Cell death assays

Cell death was monitored with trypan blue or by terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay and was observed to remain at a rate below 5%. TUNEL assay was performed using the in situ Cell Death Detection Kit (Roche # 11684795910) as per manufacturer recommendations. Cells were imaged by fluorescence microscopy, and number of apoptotic cells were determined by fluorescein-dUTP signal in the nucleus.

3.3. HCMV genomes quantification

Cell-associated virus or virus from supernatant were treated with DNAseI to remove capsid-free DNA followed by DNAseI inactivation and Proteinase K digestion in 0.5% SDS for 3 hours at 37C. Viral DNA was recovered by phenol:chloroform extraction and ethanol precipitation. Viral genomes were quantified by qPCR using the SYBR green PCR master mix (Life Technologies) with primers specific to the IE1 gene (Forward primer: 5′-TCGTTGCAATCCTCGGTCA-3′; Reverse primer: 5′-ACAGTCAGCTGAGTCTGGGA-3′). CT values were normalized to a standard curve generated from 1:5 dilutions of sorbitol-purified HCMV viral particles.

3.4. Selection of target proteins and peptides for parallel reaction monitoring (PRM) analysis

A list of 83 candidate peroxisome proteins with experimental validation of peroxisome localization in humans was collected from various sources, including the Uniprot database (Bateman et al., 2017), the PeroxisomeDB 2.0 database (Schlüter et al., 2010), and experimental localization to peroxisomes from previous proteomic studies (Gronemeyer et al., 2013; Jean Beltran et al., 2016). This set of proteins includes peroxisome proteins from major peroxisome functions: peroxisome biogenesis, lipid metabolism, antioxidant activity, isoprenoid metabolism, and other minor functions. A peroxisome peptide spectral library was generated using Skyline (v4.0) (MacLean et al., 2010) from data-dependent acquisition (DDA) MS analyses of whole cell lysates and organelle enriched fractions from previous DDA experiments (Jean Beltran et al., 2016) and additional DDA experiments performed in this study. For proteins with less than 5 peptides containing tandem mass spectra information, additional spectra were collected from the National Institute of Standards and Technology (NIST) HCD human peptide spectral library. Up to 10 unique tryptic peptides per proteins were selected for a first-pass PRM analysis in biological triplicate to determine retention times and validate the reliable detection of transition ions. The list of peptides was further filtered to include up to 3 per protein based on ion intensity and peak area reproducibility across replicates. Peptides with residues prone to modifications were generally avoided. A final list of 160 peptides across 60 peroxisome proteins was used for peroxisome proteome quantification by PRM.

3.5. Protein sample preparations for mass spectrometry (MS) analysis

For whole cell analysis, cell pellets were washed twice with PBS, scraped in PBS, and pelleted by centrifugation. Cell pellets were stored at −80°C until ready for analysis. Organelle-enriched cell fractions were collected as described previously (Jean Beltran et al., 2016). Briefly, cells were incubated in hypotonic solution (0.25M sucrose, 6mM EDTA, 120mM HEPES-NaOH; pH 7.4) and mechanically lysed through 14 μM polycarbonate membrane filters (Sterlitech Corporation). Organelle enrichment by differential centrifugation was achieved by pelleting the nucleus (1400xg, 10min), collecting the supernatant, and pelleting the cytosolic organelles by a second centrifugation (20,000xg for 30 min). Peroxisomes were further enriched by density-gradient ultracentrifugation using a discontinuous Optiprep (Axis-shield) stepped gradient in 5% intervals from 0 to 30% (v/v) Optiprep. Ultracentrifugation was performed in an SW60 Beckman rotor at 129,000xg for 3hrs at 4°C.

Whole cells and organelle-enriched fractions derived from differential centrifugation or density gradient ultracentrifugation were lysed in lysis buffer (4% SDS, 0.1M NaCl, 50mM TrisHCl, 0.5mM EDTA, pH8.0) and reduced/alkylated using 25mM TCEP solution (Thermo Fisher #77720) and 50mM 2-chloroacetamide (MP Biomedicals #ICN15495580) at 70°C for 20 min. Proteins were recovered by methanol-chloroform precipitation (Wessel and Flügge, 1984), and resuspended in 25mM HEPES (pH8.2) with MS-grade Pierce trypsin (Thermo Fisher #90057) at a 1:50 trypsin:protein weight ratio for 16hrs at 37°C. Peptides were adjusted to 1% trifluoroacetic acid (TFA) and desalted using the StageTip method (Rappsilber et al., 2007) and SDBP-RPS material (3M Analytical Biotechnologies). Peptides were washed with 0.2% TFA, and eluted (5% ammonium hydroxide, 80% acetonitrile), followed by drying completely in a SpeedVac (Thermo Fisher) and resuspension in 0.1% formic acid (FA) and 2% acetonitrile (ACN).

3.6. Peptide MS analysis

Samples were analyzed by nLC-MS in a Dionex Ultimate nRSLC coupled to a Thermo Fisher Q-Exactive HF mass spectrometer using DDA or PRM modes. Peptides were separated in a 120 minutes (DDA) or 60 minutes (PRM) reverse-phase linear gradient (2 to 27% solvent B; solvent A: 0.1% FA, solvent B: 0.1% FA/ 97% ACN) in a PepMap RSLC C19 2μm particle size, 75μm × 50cm EASYSpray column (ThermoFisher ES803) set to 50°C. For DDA mode, full scan range was set to 350-1800 m/z with a 120,000 resolution. The top 20 most intense precursors with min signal of 2.5E3 were subject to fragmentation for MS² analysis at 15,000 resolution. Isolation window set to 1.2 m/z, normalized collision energy (NCE) to 28, and target values of 3E5 and 1E5 for full-scan MS and MS², respectively. For PRM mode, the instrument was set to 30,000 resolution, 100ms maximum injection time, 5E5 acquisition target, 0.7 m/z isolation window, NCE to 28, and 150m/z fixed first mass.

3.7. Informatic analysis of MS data

Tandem MS spectra collected from DDA mode were analyzed by Proteome Discoverer v2.1 (ThermoFisher). A processing workflow containing the SEQUEST algorithm was used to search tandem MS spectra against a UniProt-SwissProt human database and common contaminants, generating peptide spectrum matches. Percolator was used to calculate matches q-values based on reverse sequence database searches. The resulting MSF files were imported into Skyline with a cut-off score of 0.95 and filtered for spectra matching the list of peroxisome proteins and peptide subset. RAW files containing MS² spectra collected in PRM mode were imported into Skyline to extract product ion chromatograms (XICs) and calculate peak areas. The top three most intense coeluting product XICs were used for peak area quantification. Peak areas were normalized to peptides used as loading controls from H2B1A (LLLPGELAK) and H2A1A (AGLQFPVGR). A third peptide from α-tubulin (ISVYYNEATGGK), a protein commonly used for loading control in HCMV time course experiments, was monitored as an extra control for loading.

Normalized peak areas were exported for data analysis in R (v3.4.3) and RStudio (v1.1.423). Normalized peptide peak areas were used to determine coefficient of variation across samples. Relative protein-level abundances were determined as follows. Normalized peak areas were divided by the value in mock, and these ratios were averaged across peptides from each single protein. Heatmaps were created using the package pheatmap (v1.0.8). Hierarchical clustering for these heatmaps was done using euclidean distances and complete linkage method. To determine clusters from temporal profiles, first protein abundances relative to mock were converted to z-scores (mean-centered with unit standard deviation). The package ppclust (v.0.1.1) was used to cluster the z-score values with the fuzzy C-means algorithm (number of clusters = 4).

3.8. Microscopy

An inverted fluorescence confocal microscope (Nikon Ti-E) equipped with a Yokogawa spinning disc (CSU-21), digital CMOS camera (Hamamatsu ORCA-Flash TuCam), and precision microscope stage (Piezo) was used for all microscopy work. An environmental control chamber was used for live-cell imaging experiments to maintain a 37° C and 5% CO₂ environment. Z-stacks were acquired with 0.2μm steps throughout the cell depth using a Nikon 100X Plan Apo objective for analysis of peroxisome morphology and colocalization and a Nikon 60X Plan Apo objective for all other experiments. For long-term peroxisome imaging, Z-stacks were acquired at 50ms exposure per step to limit laser exposure to <10 seconds per cell. Image acquisition was automated for sequential imaging of 20 individual cells over an 8-hour period. Time points in which cells moved out of the field of view were discarded.

3.9. Transfection for expression of fluorescent tags

MRC5 cells were seeded in 35mm glass-bottom microscope dishes (MatTek) at 1.6×10⁵ cells/dish and transfected 24 hours later. A 1:3 ratio of μg plasmid DNA to μl X-tremeGENE HP DNA transfection reagent (Sigma-Aldrich) were mixed in 200μl Opti-MEM (LifeTechnologies) and incubated at room temperature in for 20 minutes. Transfection mix was added dropwise to cells in 1.8mL of Opti-MEM and allowed to incubate at 37° C for 6 hours. Cells were then rinsed in PBS, placed in 2mL complete medium, and allowed to recover for 18 hours prior to imaging.

The following amounts of plasmid DNA were used: 200ng eGFP-Peroxisomes-0 (gift from Michael Davidson, AddGene #54501) and 700ng mito-BFP (gift from Gia Voeltz, AddGene #49151).

3.10. Western blotting

Cells collected for western blot analysis were lysed in lysis buffer. Protein concentration was measured by Pierce BCA assay (ThermoFisher). Ten μg were reduced with 2-mercaptoethanol and subjected to SDS-PAGE. Proteins were transferred to a PVDF membrane, and the membranes were blocked in blocking buffer (5% milk in TBS), primary antibody in blocking buffer with 0.2% Tween, and secondary antibody in blocking with 0.2% Tween and 0.01% SDS. Fluorescent signal from secondary antibodies was detected using the Odyssey CLX system (Li-cor). ImageStudio v4.0 (Li-cor) was used for band intensity quantification.

The following antibody dilutions were used for primary incubation: 1:200 for mouse anti-pp28 (clone 10B4 gift from Tom Shenk (Silva et al., 2003)), 1:100 for mouse anti-IE1 (clone 1B12 gift from Tom Shenk (Zhu et al., 1995)), 1:100 for mouse anti-pUL26 (clone 7H1-5 gift from Tom Shenk (Munger et al., 2006)), 1:100 for mouse anti-pp65 (clone 8F5 (Nowak et al., 1984)), 1:5000 for mouse anti-α-tubulin (ThermoFisher T6199), 1:2000 for rabbit anti-PEX14 (Abcam ab183885), 1:500 rabbit anti-GNPAT (Proteintech 14931-1-AP). The following antibodies at 1:10000 dilutions were used for secondary incubations: Alexa Fluor 680 goat anti-rabbit IgG (ThermoFisher A-21109) and IRDye 800CW goat anti-mouse (Li-cor 926-32210).

3.11. Immunofluorescence

For all immunofluorescence experiments, cells were fixed in 4% paraformaldehyde (PFA) for 15 minutes, then washed with PBS and permeabilized in cold methanol at −20° C for 20 minutes. Uninfected cells were blocked in 10% goat serum in PBS + 0.01% Triton (PBT) for 30 minutes, and infected cells were blocked in 10% goat serum with 5% human serum in PBT for 30 minutes. Cells were incubated in primary antibodies for 2 hours, washed with PBT, incubated in secondary antibodies for 45 minutes, washed with PBT, and rinsed in PBS. For peroxisome counting experiments, MRC5 cells were grown on glass coverslips in 2mL wells and mounted with ProLong Diamond antifade (ThermoFisher) on glass slides after staining. For CRISPR, morphology, and MFF/FIS1 imaging, cells were grown in glass-bottom microscope dishes (MatTek), then stored and imaged in 2mL TBS after staining. For MFF/FIS1 colocalization experiments, cells were first transfected with eGFP-Peroxisomes-0 and mito-BFP, as described above.

Antibodies were diluted in blocking buffer as follows: 1:500 for rabbit anti-Pex14 (Abcam ab183885), 1:40 for mouse anti-pp28 (clone 10B4 gift from Tom Shenk (Silva et al., 2003)), 1:40 for mouse anti-IE1 (clone 1B12 gift from Tom Shenk (Zhu et al., 1995)), 1:5000 for rabbit anti-ICP0 (Virusys Corporation H1A027-100), 1:300 for mouse anti-ICP4 (Abcam ab6514), 1:1500 for mouse anti-MFF (SantaCruz SC-398617), 1:2000 for rabbit anti-FIS1 (ThermoFisher PA5-22142), 1:1000 anti-NBR1 (Cell Signaling D2E6), 1:2000 for goat anti-rabbit IgG highly cross-adsorbed AlexaFluor 568 (ThermoFisher A-11036), 1:2000 for goat anti-mouse highly cross-adsorbed AlexaFluor 568 (ThermoFisher A-11019), and 1:2000 for goat anti-mouse IgG highly cross-adsorbed AlexaFluor 633 (ThermoFisher A-21052). Sequential staining was used for Mff co-visualization with IE1 or UL99.

3.12. Peroxisome counting, morphology analysis, and colocalization

Peroxisomes in a cell were counted with the Nikon NIS-Elements AR v5.0 imaging software. Image Z-stacks were deconvolved with the 3D Deconvolution tool, and the 3D Spot Detection tool was used to detect peroxisomes (50-300 contrast, 1:2 Z-axis elongation, 0.5 μm diameter spot, and variable size spots). For analysis of counts to be used for inference, cells were transfected with EGFP-Peroxisome-0 as described above and imaged for 8 hours at 5-minute intervals. Counts were excluded from the final dataset if cells moved out of frame.

To analyze peroxisome morphology, the ImageJ (National Institutes of Health) analysis tool 3D Objects Counter was used to threshold and measure surface area, volume, and mean length of each peroxisome, with a minimum size requirement of 10 consecutive voxels (at a resolution of 0.065 μm/pixel).

To analyze MFF and FIS1 colocalization with peroxisomes and mitochondria, maximum intensity projections were generated from z-stack images and the ImageJ analysis tool Coloc 2 (J Schindelin) was used to threshold and analyze colocalization between channels via Spearman’s correlation analysis with 100 Costes randomizations and PSF adjustment of 3.29.

To analyze the number of NBR1-positive peroxisomes, images were manually inspected and peroxisomes that were in contact with NBR1 puncta were marked as NBR1-positive.

3.13. Modeling of peroxisome biogenesis from count time series data

Our model is based on a stochastic model for yeast organelle abundance proposed by (Mukherji and O’Shea, 2014), which models the temporal evolution of the number of organelles n, or more specifically the probability that the number equals n at time t, p(n, t), in terms of the rates of four processes governing n (de novo generation k_d, fission k_f, fusion k_fus and degradation γ). Here, in the case of human peroxisomes, fusion is neglected as this process does not occur, so that the following three processes are considered:

• a)
  de novo peroxisome generation from ER vesicles, k_d:

  $$\varnothing \stackrel{k_d}{\to }\text{Peroxisome}$$

• b)
  growth and fission of pre-existing peroxisomes, k_f:

  $$\text{Peroxisome}\stackrel{k_f}{\to }\text{Peroxisome}+\text{Peroxisome}$$

• c)
  degradation through pexophagy, γ:

  $$\text{Peroxisome}\stackrel{\gamma }{\to }\varnothing$$

These stochastic processes are combined to generate the evolution equation for p(n, t):

The relative importance of the de novo/fission/degradation processes in a cell with n peroxisomes is given by k_d : k_f × n : γ × n. This is because the number of fission and degradation events depend on the number of peroxisomes present, as opposed to de novo events. In the following we designate the rates of the model by θ=[k_d, k_f, γ].

In this work, we considered temporal count data acquired for K cell replicates via fluorescent microscopy. We denote the measured peroxisome count in cell replicate 1 ≤ k ≤ K at time ${\mathrm{t}}_{\mathrm{1}}^{\mathrm{k}}\le {\mathrm{t}}_{\mathrm{t}}^{\mathrm{k}}\le {\mathrm{t}}_{{\mathrm{T}}_{\mathrm{k}}}^{\mathrm{k}}$ by ${\mathrm{y}}_{\mathrm{t}}^{\mathrm{k}}$ and all the measured data (i.e. all time points and replicates) by $\mathrm{D}=\left\{{\left({\mathrm{t}}_{\mathrm{t}}^{\mathrm{k}},y_t^k\right)}_{1\le t\le t_K}^{1\le k\le K}\right\}$. We assumed a Normal error model with a replicate specific standard deviation so that the measured count is obtained from the true (hidden) count, ${\mathrm{x}}_{\mathrm{t}}^{\mathrm{k}}$, following $y_t^k\sim \mathcal{N}\left(x_t^k,{\sigma }^{\mathrm{k}}\right)$, i.e.

We modelled the cell to cell variations of the rates by assuming that the rates each follow a Gamma distribution between cells. Each rate has its own rate parameter and shape parameter: k_d ~ Gamma (α_kd, β_kd), k_f ~ Gamma(α_kf, β_kf) and γ ~ Gamma(α_γ, β_γ). This ensures that the rates are always positive and can provide a good approximation of normally distributed rates.

For each rate, we can equivalently reformulate the parameters of the gamma distribution in terms of the rates mean and standard deviations, e.g. in the case of $k_d:{\alpha }_{k_d}={\left(\frac{{\mu }_{k_d}}{{\sigma }_{k_d}}\right)}^2,{\beta }_{k_d}=\frac{{\sigma }_{k_d}^2}{{\mu }_{k_d}}$ which are easier to more useful and easier to interpret.

3.14. Bayesian inference of peroxisome biogenesis rates

We described the method for inference of rates that regulate peroxisome number in (Galitzine et al., 2018), to which the reader is referred for more details. We want to obtain the distribution of the rate means and standard deviations, e.g. μ_kd, σ_kd, as well as the error standard deviation deviations. We denote all the parameters to be inferred by α = [μ_kd, σ_kd, μ_kf, σ_kf, μ_γ, σ_γ, σ¹, σ², …, σ^K]. A Bayesian formulation of the inference problem is thus:

Since no prior information about the rates is known besides their positivity, we use a flat uninformative prior p(α) = 1 for all parameters.

We sample α from p(α|D) in Eq. (3) with a Metropolis Hasting method (Wilkinson, 2011). The distribution functions of the rate means and standard deviations are then directly obtained from the samples.

Sampling Eq. (3) necessitates the calculation of the likelihood of the data p(D|α). The overall likelihood of the data is calculated by assuming that both the replicates and the consecutive increments of each replicate are statistically independent:

where $\mathrm{p}(x_{t+1}^k\mid x_t^k)$ is obtained by solving Eq. (1), $\mathrm{p}(y_t^k\mid x_t^k,{\sigma }^{\mathrm{k}})$ with the error model of Eq. (2) and p(θ|α) is the product of the three rate gamma distributions.

We obtain the likelihood of each individual replicate in Eq. (4) with a particle filter method (Doucet and Johansen, 2009). In such a method, computational particles are propagated following Eq. (1) along each replicate time trace. At each measurement time, they are resampled according to the normal error model of Eq. (2). Finally, the likelihood of a replicate is obtained by taking the product of particle weight sums at each measurement time.

3.15. CRISPR-mediated knockouts

CRISPR guide RNA sequences designed for knockout of human PEX3, PEX19, PEX11B, PEX7, or non-targeting negative control (from Origene pCas-Scramble) were cloned into and delivered by the LentiCRISPR v2 vector (Addgene plasmid 52961) (Table S4) from Feng Zhang (Sanjana et al., 2014). A total of 1 × 10⁵ MRC5 cells were transduced with lentivirus (2 days) and selected with puromycin (1 μg/ml) for 4 days. Cells were cultured for an additional week to allow depletion of protein targets before performing experiments.

3.16. siRNA-mediated knockdowns

siRNA duplexes sequences were purchased from Sigma-Aldrich for GNPAT (#1, 5′-CAAGGUACCUCUCAAUGUU[dT][dT]-3′; and #2, 5′-GGCUUAUGCUCCAGCACAU[dT][dT]-3′) or negative control (MISSION siRNA Universal Negative Control #1; Sigma-Aldrich #SIC001). Transfections were performed using Lipofectamine RNAiMAX Transfection Reagent (ThermoFisher Scientific) as per manufacturer protocol. A total of 0.5 × 10⁵ MRC5 cells were transfected with 10pg siRNA duplex and 1.5μl transfection reagent in Opti-MEM. Cells were recovered after 16 hours using full medium and either immediately infected or processed for WB analysis at 48 hours post transfection.

3.17. Plasmalogen quantification by MS

Phospholipids were isolated for MS analysis as previously described(Ivanova et al., 2007). A total of 1×10⁶ cells were incubated in serum-free DMEM for 24 hours prior to processing. Cells were recovered by scrapping in 0.8ml of ice-cold 1:1 0.1N HCl:MetOH (v/v) and transferred to a microcentrifuge tube. A total of 200ng odd-carbon chain internal standard PE 17:0/17:0 (Avanti Polar Lipids, Inc.; No. 830756) and 0.4ml ice-cold chloroform were added. The sample was vortexed and centrifuged (5 min, 18,000xg). The organic phase was collected, solvent evaporated (Speedvac SC110A, ThermoSavant), and dissolved in 50μl methanol prior to injection of 20μl for LC-MS analysis.

A Shimadzu UFLC system coupled to a Thermo LTQ XL mass spectrometer was used for LC-MS analysis in positive-ion mode. Phospholipids were separated in a 60-minute stepped linear reverse-phase gradient (solvent B set to 0% 5min, 30% 15 min, 50% 60min) using an ACE C18 colum(μm, 75mm × 1mm) at 45°C. Solvent composition for buffer A (2:2:1 acetonitrile:methanol:water (v/v/v) supplemented with 0.1% formic acid and 0.028% ammonia) and buffer B (isopropanol supplemented with 0.1% formic acid and 0.028% ammonia). Scan range was set to 680-820 m/z with a 60,000 resolution for MS¹ in the FT and 200-800 m/z for MS² in the IT. Instrument was set to perform a full MS scan followed by isolation and collision induced dissociation (CID) for MS² acquisition from an inclusion list containing m/z for phosphatidylethanolamine (PE) plasmalogen precursor [M+H]+ (Table S3) and for PE 17:0/17:0 internal standard (m/z = 720.5). Isolation window set to 1.9 m/z, normalized collision energy (NCE) to 32, acquisition Q of 0.250, and activation time of 30ms. Raw data files were imported into skyline for analysis. Two characteristic fragment ions were used to confirm the plasmalogen species (Table S3), as described in (Nakanishi et al., 2009; Zemski Berry and Murphy, 2004). If an isomeric plasmalogen species was observed to co-elute, the major species is reported, and minor species are mentioned in Table S3. The 579.4 m/z fragment ion was used to identify PE 17:0/17:0 internal standard. Precursor XIC areas under the curve were used for quantification and were normalized to 17:0 PE.