Exosomal lipid PI4P regulates small extracellular vesicle secretion by modulating intraluminal vesicle formation

Abstract

Membrane lipids play vital roles in small extracellular vesicle (sEV) biogenesis. However, the function of various lipids in the biogenesis of sEVs is still poorly understood. Phosphoinositolphosphates (PIPs), a group of the most critical lipids in vesicle transport, can undergo rapid conversion in response to a variety of cell signals, which in turn influence the generation of vesicles. Due to the challenge in detecting the low amount of PIP content in biological samples, the function of PIPs in sEVs has been insufficiently investigated. Here, we employed an LC‐MS/MS method to detect the levels of PIPs in sEVs. We revealed phosphatidylinositol‐4‐phosphate (PI4P) was the main PI‐monophosphate in macrophage‐derived sEVs. The release of sEVs was regulated in a time‐dependent manner and correlated with the PI4P level during the lipopolysaccharide (LPS) stimulation. In terms of mechanism, within 10 h of LPS treatment, the LPS‐induced production of type I interferon inhibited the expression of PIP‐5‐kinase‐1‐gamma, which increased the PI4P content on multivesicular bodies (MVBs) and recruited RAB10, member RAS oncogene family, to promote sEV generation. When LPS stimulation was extended to 24 h, the heat shock protein family A member 5 (HSPA5) expression level was elevated. PI4P interacted with HSPA5 on the Golgi or endoplasmic reticulum away from MVBs, which disrupted the continuous fast sEV release. In conclusion, the present study demonstrated an inducible sEV release model response to LPS treatment. The inducible release may be due to PI4P regulating the generation of intraluminal vesicles secreted as sEVs.

1. INTRODUCTION

Various lipids are essential for the biophysical features of membranes, including their biolayer curvature, electrostatics, and lipid packing, which affect protein recruitment and signal transduction (van Meer et al., 2008). Extracellular vesicles (EVs) are nanosized particles with lipid bilayers secreted by cells. Lipids are important for ensuring the proper biological properties of EVs. Lipid species comprise over 47,000 molecules and are unequally distributed among cellular membranes (Egea‐Jimenez & Zimmermann, 2020; Shevchenko & Simons, 2010). The biolayers of EVs mainly comprise sphingomyelin, phospholipids, ganglioside, and cholesterol (Donoso‐Quezada et al., 2021). Nonetheless, the relative abundance of lipids in EVs varies by cell type and state. Many studies have suggested that lipids play crucial roles in EV formation and recognition. For example, ceramide promotes domain‐induced budding to generate EVs (Trajkovic et al., 2008). However, the functions of vital lipids at low levels, such as PIPs, in EVs are unclear.

Phosphoinositolphosphates (PIPs) are a group of unique lipids that play a guiding role in vesicle formation and transport (De Craene et al., 2017; Posor et al., 2022). PIPs exist as seven isomers, depending on the number and positions of phosphates (Figure 1a), and they can be converted into each other by phosphoinositide kinases and phosphatases (Hammond & Balla, 2015). Different PIPs are hallmarks of various organelle membranes (Di Paolo & De Camilli, 2006). PIP conversion is an essential step in protein recruitment (Wallroth & Haucke, 2018). It is widely accepted that the roles of lipids in EVs are essential for EV biogenesis. However, the evidence to identify the PIPs in EVs is insufficient due to limited enrichment methods and analytical techniques (Skotland et al., 2020). In addition to one methodological study reporting the detection of PIPs in exosomes from cancer cell lines (Morioka et al., 2022), few studies report the function of PIPs in EVs.

FIGURE 1.

Profiling of phosphoinositide regioisomers in macrophage‐derived sEVs. (a) Schematic diagram of the PIP structure. Mono‐PIPs can be present as three isomers after the addition of a phosphate group at the −3, −4, and −5 sites, respectively. (b) Chromatograms of the three regioisomers of PIP^Me 18:1(9Z)/18:1(9Z) standard (top panel) and separation of PIP^Me 36:2 in RAW264.7‐cell‐derived sEVs (bottom panel). (c) MRM spectra of PI3P(purple) and PI4P(yellow) in RAW264.7‐cell‐derived sEVs. (d) Pie chart of the ratios of PI3P and PI4P in RAW264.7‐cell‐derived sEVs. (e) SIM imaging of isolated RAW264.7‐cell‐derived sEVs. Green represents sEVs carrying GFP or GFP fused PIP biosensor or CD63, and red represents PKH26 stained sEV particles. Scale bar, 2 μm. (f) Quantitative results from Figure 1e. The bar shows the mean with SEM. The bar plot represents eight independent images from four biological replicates. Unpaired two‐tailed t‐test (^*** P < 0.001; ^**** P < 0.0001; ns, not significant). (g) The TEM results of anti‐PI4P and anti‐GFP immunogold staining of RAW264.7‐cell‐derived sEVs. The negative control consists of only the secondary antibody. The white arrow indicates the immune gold particles. Scale bar, 200 nm.

Due to the current limitation to purify a particular subpopulation of EVs, we specifically analyzed a small EV (sEV) cohort with a diameter of less than 0.22 microns precipitated at 100,000×g (Théry et al., 2018). sEVs are generated as intraluminal vesicles (ILVs) with the lumen of endosome invagination to form multivesicular bodies (MVBs) that consequently transport and fuse with plasma membranes (Colombo et al., 2014). For ILV formation, endosomal sorting complex required for transport (ESCRT)‐dependent and ESCRT‐independent pathways are employed. In the process of vesicle formation and transport, the association of organelles with the cytoskeleton, molecular motors, and small GTPase are involved (Bonifacino & Glick, 2004). The lipid and protein composition and types of formation distinguish subpopulations of MVBs. The different types of MVBs cause the heterogeneity of sEVs (Colombo et al., 2013). PIPs are not only a part of the membrane lipid composition but also interact with RAB GTPase to assist vesicle transport (De Craene et al., 2017). Systematic research on the role of PIPs in sEV generation is fundamental for expanding the biogenesis mechanism of sEVs and is beneficial to the classification of EVs.

sEVs release is correlated with different pathological and physiological conditions (Shah et al., 2018). Previous studies have shown that the release of sEVs is augmented under various biological and physiological processes (Steinbichler et al., 2017; Yin et al., 2021; Zhang & Wang, 2015). For example, sEV release increases in response to the accumulation of cytotoxic material or immune stimulation (van Niel et al., 2022). Previous studies have also shown that the activation of Toll‐like receptor (TLR) nine promotes the release of sEVs and contributes to expanding immune activation (Zhang et al., 2019). An increase in the release of sEVs in response to many stimuli has been reported. However, the decrease in the release of sEVs after initial‐stimulated aggravation has rarely been demonstrated. Lipopolysaccharide (LPS) in gram‐negative bacteria (Poltorak et al., 1998)‐induced TLR4 activation (Medzhitov et al., 1997) is an immune response model. This model has an acute inflammation phase at 0.5–10 h and is followed by a secondary response phase from 10 to 24 h. The expression of the LPS‐affected genes shows two different waves with a strict time‐dependent pattern. (Ma et al., 2013; Ramirez‐Carrozzi et al., 2009; Toshchakov et al., 2002). Therefore, LPS‐stimulated macrophages are an appropriate model to investigate the kinetics of sEV release. In this dynamic release model of sEVs, we explored the mechanism involved in PIPs guiding the biogenesis and release of sEVs.

2. MATERIALS AND METHODS

2.1. Antibodies and inhibitors

Anti‐EEA1 (#3288), anti‐GFP (#2956), anti‐GAPDH (#2118), and anti‐HSPA5 (#3177) aantibodies were purchased from Cell Signaling Technology. Anti‐VPS16 (#ab206326), anti‐RAB10 (#ab237703) and anti‐CD9 (#ab92726) antibodies were purchased from Abcam. Anti‐CD63 (#sc‐5275) antibody was purchased from Santa Cruz Biotechnology. The anti‐rabbit IgG Gold antibody (#G7402) was purchased from Sigma Aldrich. A purified anti‐PI4P IgM antibody was purchased from Echelon Biosciences. Anti‐ALIX antibody (#12422‐1‐AP) was obtained from the Proteintech group. HRP‐conjugated goat anti‐mouse IgG(H+L) (#HX2032), HRP‐conjugated goat anti‐rabbit IgG(H+L) (#HX2031), Dylight 488‐conjugated donkey anti‐goat IgG(H+L) (#HX2086), Dylight 649‐conjugated rabbit anti‐goat IgG(H+L) (#HX2097), and DyLight 649‐goat anti‐rabbit IgG(H+L) (#HX2095) antibodies were obtained from Huaxing Bio. Alexa Fluor 647‐donkey anti‐rabbit IgG (H+L) (#34213ES60) antibody was purchased from Yeasen Biotechnology. Goat anti‐mouse IgG (H+L) cross‐adsorbed secondary antibody, Alexa Fluor 488 (#A‐11001), rabbit anti‐goat IgG (H+L) cross‐adsorbed secondary antibody, Alexa Fluor 568 (#A‐11079), goat anti‐rabbit IgG (H+L) cross‐adsorbed secondary antibody, and Alexa Fluor 568 (#A‐11011) were purchased from Invitrogen. Cy3‐conjugated AffiniPure donkey anti‐mouse IgM, μ chain‐specific (#JAC715165020) antibody was purchased from Jackson ImmunoResearch. The PIK‐93 inhibitor (#CSN17170) was purchased from CSN Pharm. Autophinib (#HY‐101920) and UNC3230 (#HY‐110150) were purchased from MedChemExpress. HA‐15 recombinant human TNF‐α (#300‐01A) and recombinant human IFN‐β (#300‐02BC) were purchased from PeproTech.

2.2. Plasmid construction

The pCMV3.1, pLKO.1, eGFP C1, and eGFP N1 backbone plasmids were purchased from Addgene. The pTY backbone plasmid was a kind gift from Dr. Conggang Zhang. The blue fluorescent protein (BFP) was cloned from the pLVX‐BFP plasmid. The PH domain of oxysterol‐binding protein (OSBP) and human heat shock protein family A member 5 (HSPA5) gene were cloned from the THP‐1‐cell cDNA library. The FYVE domain of hepatocyte growth factor 1‐regulated tyrosine kinase substrate (HRS) was purchased from Addgene. All shRNA plasmids were purchased from Thermo Fisher.

2.3. Cell culture and transfection

All cell lines were purchased from ATCC. RAW264.7 cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium with 10% foetal bovine serum (FBS) and 1% penicillin–streptomycin. HEK 293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% FBS and 1% penicillin‐streptomycin. All cells were cultured at 37°C in a sterile, constant‐temperature incubator containing 5% CO₂.

Bone marrow‐derived macrophages (BMDMs) were isolated from C57BL/6J mouse bones according to a Cold Spring Harbor protocol (Weischenfeldt & Porse, 2008). Under sterile conditions, intact bones were cut at both ends, and marrow was released by using HEPES (N‐2‐hydroxyet hylpiperazine‐N'−2‐ethane sulfonic acid) buffer (pH = 7.5) with needles. BMDMs were cultured in DMEM with 10% FBS and 1% penicillin‐streptomycin supplied with L929 cell‐conditioned medium for 6 days. The BMDMs were then scraped from 15‐cm dishes and seeded into cell culture plates for the western bot (WB), immunofluorescence, or transmission electron microscopy (TEM) analyses within 24 h. Cells were cultured in DMEM with 10% FBS and 1% penicillin‐streptomycin.

2.4. Lipid nomenclature

Annotation of lipids followed the classification system recommended by LIPIDMAPS (http://www.lipidmaps.org) (Liebisch et al., 2013). For example, 2‐(9Z‐oleoyl)−1‐palmitoyl‐sn‐glycero‐3‐phosphoinositol was denoted as PI 16:0/18:1 (9Z). Briefly, the lipid class was represented by the first two letters, PI. Fatty acyl chain information was indicated as the number of carbon‐atoms as follows: the number of double bonds (16:0 or 18:1). Fatty acyls linked to the glycerol was donated as sn‐1/sn‐2 with known sn‐locations or separated by an underscore without differentiating sn‐position.

2.5. Lipid extraction

sEV samples were extracted using a modified methyl tert‐butyl ether (MTBE) protocol (Matyash et al., 2008). In brief, MTBE (1000 μl), methanol (300 μl) and an aqueous 0.1 M hydrochloride (HCl) solution (250 μl) were added to a 2‐ml centrifuge tube. The mixture was then vortexed for 5 min. To separate the organic and aqueous phases,the mixture was centrifuged at 10,000×g for 10 min. The upper phase was collected, and the lower phase was re‐extracted with 500 μl of the upper phase of the MTBE/methanol/0.1 M HCl (10:3:2.5, v/v/v) system. Finally, the combined organic phases were collected and dried under N₂ flow for further use. The extracted samples were immediately subjected to methylation and LC‐MS analysis to avoid degradation.

2.6. Trimethylsilyl diazomethane (TMSD) methylation

A modified TMSD methylation method was employed (Clark et al., 2011). Briefly, a solution of TMSD (2.5 M) in hexane (50 μl) was added to the lipid extracts dissolved in methanol (150 μl) to obtain a yellow solution. After 10 min, the reaction was quenched with 10 μl of glacial acetic acid, and the sample was then subjected to chiral column chromatography separation and tandem mass spectrometry analysis.

2.7. Chiral liquid chromatography (LC) and high‐resolution time‐of‐flight (TOF) mass spectrometry

Chiral LC‐multiple reaction monitoring (MRM) was conducted on a Shimadzu LC‐20AD system (Kyoto, Japan) coupled with an X500R Quadrupole‐TOF mass spectrometer (Sciex, Toronto, Canada). The injection volume was 10 μl per run, and a CHIRALPAK IC‐U column [100 mm × 3.0 mm, 1.6 μm, DAICEL] was used. Mobile phase A consisted of methanol supplemented with 10 mM ammonium acetate, and mobile phase B consisted of acetonitrile supplemented with 10 mM ammonium acetate. The flow rate was set to 0.2 ml/min. The chromatographic gradient was as follows: 40% A at 0‐0.4 min, 40%−85% A at 0.4‐1.2 min, 85% A at 1.2‐5.6 min, 85%−40% A at 5.6‐5.8 min, and 40% A at 5.8‐8 min. The MS parameters were optimized as follows: ESI voltage, 5000 V; curtain gas, 20 psi; interface heater temperature, 400°C; ion source gas 1 and gas 2, 30 psi; declustering potential, 80 V; and CID energy for MRM, 35 eV. The peak area was analyzed to represent the content of PIPs, normalized to per 10⁸ cells or 10¹⁰ sEVs.

2.8. WB analysis

To determine protein expression levels, WB analysis was performed according to a Cold Spring Harbor protocol. Briefly, cells were washed with phosphate‐buffered saline (PBS; pH = 7.4) and then lysed in immunoprecipitation (IP) lysis buffer (50 mM Tris‐HCl, 150 mM NaCl, 10% glycerol, and 0.2% Triton X‐100; pH = 7.5) supplemented with protease inhibitor (#78430, Thermo Fisher) and phosphatase inhibitor (#78420, Thermo Fisher) for 30 min at 4°C. After centrifugation at 12,000×g for 10 min, whole‐cell proteins were harvested. A Pierce™ bicinchoninic acid protein assay kit (#23225, Thermo Fisher) was used to detect the protein concentration according to the manufacturer's instructions. The concentration‐normalized cell lysate was mixed thoroughly with 6×protein loading buffer (#DL101‐02, Transgen Tech) and heated at 95°C for 10 min. Samples were loaded on sodium dodecyl sulfate‐polyacrylamide gel electrophoresis (SDS‐PAGE) and electrophoresed, and the proteins were transferred to Immobilon®‐E PVDF transfer membranes via the wet transfer membrane method. The membranes were incubated with primary antibody for 3 h at room temperature (RT) or 4°C overnight, followed by incubation with secondary antibody for 1 h at RT. A wash step with Tris‐buffered saline with 2% Tween (pH = 7.5) for 5 min followed each antibody incubation step. Thermo Fisher Pierce™ SuperSignal West Pico Plus substrate was added to the membranes, and images were acquired using an Invitrogen iBright 1500 instrument.

2.9. Real‐time quantitative polymerase chain reaction (RT‐qPCR)

Total RNA was extracted from cells and sEVs with TRIzol reagent (#15596026, Invitrogen) according to the manufacturer's protocol. mRNA was reverse transcribed into cDNA using the iScript cDNA synthesis kit (#1708891, Bio‐Rad) according to the manufacturer's instructions. RT‐qPCR was conducted with an iTaq Universal SYBR Green Supermix kit (#172‐5121, Bio‐Rad). Bio‐Rad CFX96 instruments were programmed with the appropriate thermal cycling protocol to detect fluorescent signals for quantification. mRNA expression levels were analyzed according to the delta‐delta Ct method.

2.10. sEV isolation

sEVs were isolated according to the differential centrifugation method. The cell culture medium was centrifuged at 500×g at 4°C for 10 min twice to remove live cells. The supernatant was centrifuged at 2000×g at 4°C for 10 min twice to remove cell debris and large EVs. The supernatant was centrifuged at 12,000×g at 4°C for 20 min twice to discard organelles, including the endoplasmic reticulum (ER) and Golgi. The supernatant was filtered with a 0.22‐μm filter and then centrifuged at 100,000×g at 4°C for 70 min. The sEV pellets were washed with 50 ml of PBS to discard the soluble proteins. A final centrifugation step was performed at 100,000×g at 4°C for 70 min to harvest sEVs. sEVs were suspended in 100 μL of PBS for subsequent experiments.

2.11. TEM

For negative staining, sEV samples were coated on carbon membrane‐supported copper grids for 10 min. The grids were washed with double‐distilled water (ddH₂O) three times. The samples were stained with 1% uranyl acetate for 10 min at RT. sEVs were visualized on a TEM at 80 kV.

For sEVs immunogold staining, sEVs were fixed with 2.5% glutaraldehyde for 10 min at RT and then coated on carbon membrane‐supported copper grids for 10 min. The grids were blocked with PBS buffer (pH = 7.4) containing 2% bovine albumin (BSA) for 5 min. The grids were incubated with primary antibody for 1 h at RT, followed by incubation with 10‐nm gold‐conjugated anti‐mouse secondary antibody for 1 h at RT. A wash step followed each incubation with PBS buffer (pH = 7.4) for 1 min of five times. The grids were incubated with 1% glutaraldehyde for 10 min and washed with ddH₂O of five times for 1 min. The grids were incubated with methylcellulose/uranyl acetate for 5 min. After drying, the grids were visualized with a TEM at 80 kV.

For MVB observation, cultured cells were fixed with 2% paraformaldehyde (PFA)+2.5% glutaraldehyde (pH = 4.0∼5.0) at 37°C for 10 min. The samples were washed three times with 0.1 M phosphate buffer (pH = 7.5) for 15 min. The samples were immobilized with 1% osmic acid+1.5% potassium ferrocyanide for 1 h on ice. The samples were then washed three times with ddH₂O for 15 min. The samples were washed with 1% uranyl acetate for 1 h at RT and then washed three times with ddH₂O for 15 min each time. The samples were then subjected to gradient ethanol dehydration on ice and propylene oxide incubation for 15 min; this procedure was performed two times. Finally, the samples were embedded in gradient resin 812/propylene oxide and heated at 60°C. The cells were cut into ultra‐thin sections measuring ∼80 nm, and about 30 sections were observed.

2.12. Intracellular crosslinking with disuccinimidyl suberate (DSS)

Cultured cells were scraped into PBS (pH = 7.4) and washed three times. DSS at 5 mM (#21655, Thermo Fisher) was added to the cell samples and incubated for 30 min at RT. The reaction was quenched by adding 1 M Tris·HCl (pH = 7.5) for 15 min at RT, followed by centrifugation at 800×g for 3 min at 4°C. Pellets were used for the following experiment.

2.13. Nanoparticle tracking analysis (NTA)

sEVs were diluted 50 or 100 times to 500 μl of ddH₂O. A NanoSight LM14 (Malvern Instruments, UK) instrument was used to detect the concentration and size of sEVs. Each sample was measured three times, and the data were averaged.

2.14. Structured illumination microscopy (SIM) sample preparation and statistics

sEVs were immobilized on 0.1% poly‐L‐lysine‐pretreated coverslips. Before staining, the coverslips were blocked with 3% BSA in PBS buffer (pH = 7.4). PKH26 (#MINI26, Sigma–Aldrich) was used to label sEVs on coverslips. Then, the coverslips were washed with PBS buffer (pH = 7.4) containing 0.1% tween three times. 20 μl of Fluoromount (#F4680, Thermo Fisher) was applied to the sEV‐coated area, which was gently covered with glass slides. After drying, the samples were observed on a Nikon N‐SIM microscope. Fluorescent dots on the imaging field represent sEVs, counted by ImageJ software (Fiji, ver: 2.0.0‐rc‐69/1.52p).

2.15. Total internal reflection fluorescence (TIRF) imaging and calculation of the sEV secretion rate

Cells transfected with CD63‐PHluorin were cultured in a glass cell‐culture dish. A Nikon TIRF microscope was used to observe cell membrane fluorescent signals. The flash of a fluorescent dot on the membrane indicated the secretion of a single sEV event. Videos were analyzed with Imaris software (ver: 9.7, Oxford Instruments).

2.16. Immunofluorescent staining

Cells were cultured in a glass cell‐culture dish and immobilized with 4% PFA. Cells were then permeabilized with 0.1% Triton X‐100 (pH = 7.5) for 10 min and blocked with 3% BSA/PBS‐0.1%Tween (pH = 7.5) for 1 h. Cells were then incubated with primary antibody for 2 h at RT, followed by incubation with secondary antibody for 1 h. A wash step with 0.1% PBS‐0.1%Tween (pH = 7.5) was applied after each incubation step. Finally, the cells were imaged under a confocal microscope.

2.17. Proteomics

Flag bead‐tagged proteins were loaded and run on SDS–PAGE gels. The gels were stained with Coomassie brilliant blue. Proteins in gels were pretreated with 25 mM dithiothreitol to break disulfide bonds, and free sulfhydryl groups were blocked with 55 mM iodoacetamide. Proteins were digested with trypsin, and the reaction was terminated with 10% trifluoroacetic acid. Proteins were then extracted with 50% acetonitrile and 1% formic acid and vacuum‐concentrated. Finally, the concentrated solution was dissolved in 1% formic acid. High‐performance liquid chromatography–mass spectrometry (HPLC‐MS/MS) was conducted on an Orbitrap Fusion LUMOS, Synapt G2 instrument.

2.18. Bioinformatics and statistical analysis

The raw mass spectral data were analyzed with MaxQuant Perseus 2.0.3.1. Differential expression analysis was performed with the ProVision webpage (https://provision.shinyapps.io/provision/). Gene Ontology (GO) analysis was performed with DAVID 6.8. Visualization was performed using R version 3.6.2 and the ggplot2 library. Statistical analyses (unpaired t‐test and one‐way ANOVA) were performed using GraphPad Prism software.

2.19. Early and late endosome isolation

Continuous density gradient centrifugation was performed following Cold Spring Harbour protocol. Briefly, a 10%−40% continuous sucrose gradient was prepared according to the instruction manual of a Gradient Master. The gradient solutions were overlaid with post‐nuclear supernatant samples. Centrifugation at 210,000×g was applied for 16 h at 4°C. After centrifugation, the continuous gradient was collected in 24 fractions.

3. RESULTS

3.1. Profiling of PIP regioisomers in macrophage‐derived sEVs via chiral LC‐MS/MS

We adopted a chiral LC‐MS/MS workflow similar to that reported by Morioka et al. (2022) to profile the regioisomers of PIP. To test the separation capability, the equal molar mixture of the PI3P 18:1(9Z)/18:1(9Z), PI4P 18:1(9Z)/18:1(9Z), and PI5P 18:1(9Z)/18:1(9Z) standards were methylated by TMSD. The methylated products (PIP^Me) were subjected to separation by a chiral selector cellulose tris (3,5‐dichlorophenylcarbamate) column hyphenated with a TOF‐MS (resolving power > 25000) via MRM mode (Figure S1 and Table S1). The ion chromatogram (top panel in Figure 1b) showed that PI3P^Me 18:1/18:1 eluted earliest, followed by PI4P^Me 18:1/18:1 and PI5P^Me 18:1/18:1. Of note, at the same concentration, the peak area ratio between PI4P^Me 18:1/18:1 and PI3P^Me 18:1/18:1 was 1.1, indicating that the difference in ionization efficiency between PI3P^Me and PI4P^Me was not significant. Although the separation was not perfect, PI3P^Me 18:1/18:1 and PI4P^Me 18:1/18:1 were baseline separated. Given that PI5P typically exists at less than 5% of PI4P in mammalian cells (Hammond & Balla, 2015; Sarkes & Rameh, 2010; Zolov et al., 2012), its interference with the profiling of PI4P and PI3P in sEVs and cells can presumably be neglected.

To detect PIP in sEVs, we isolated sEVs (Figure S2a) from the conditioned medium of RAW264.7 cells by differential centrifugation (Théry et al., 2018). As measured by NTA, the particle sizes of the sEVs ranged between 30 and 200 nm, and the median was approximately 120 nm (Figure S2b). The sample concentration was determined by NTA and used to normalize the final dose for mass spectrometry detection at 10¹⁰ particles. WB detection results of marker proteins in the sEVs showed that CD9 and ALIX were significantly enriched in sEVs compared to cells, while GAPDH in the sEVs was below the detection limit (Figure S2c). The TEM results of isolated sEVs showed a prominent lipid bilayer structure (Figure S2d).

PIPs extracted from approximately 10¹⁰ sEV particles were subjected to the chiral LC‐MS/MS procedure described above. We found that all PIP species in the sEVs were separated into two peaks. Using PIP 36:2 as an example, by aligning with the retention time of the standards, we identified that the early eluted peak was PI3P 36:2 and the later eluted peak was PI4P 36:2 (Figure 1b bottom panel). Figure 1c shows the profiles of PI3P and PI4P in the EVs. PIPs containing C38:4 and C38:3 were the most abundant species, followed by C36:2 and C36:1. We did not detect PIPs in the FBS‐containing medium, which might be due to the low concentrations and thus below the detection limit for PIP^Me (i.e., ∼1 nM) used in our method (Figure S2e). By summing up the chromatographic peak areas of all PI4Ps and PI3Ps, relative quantitation of PI3P and PI4P was achieved. PI4P was the dominant components, accounting for 70 ± 0.9% of the detected PI‐monophosphate (PIP) in sEVs (Figure 1d).

To study the distribution of PIP in sEVs in situ, super‐resolution SIM imaging was used to observe sEVs. Specific PIP biosensors fused with a green fluorescent protein (GFP) were employed to label PIPs (Hammond & Balla, 2015; Sankaran et al., 2001), while sEVs were labelled with PKH26, a proprietary membrane dye. By recording the fluorescence on the sEVs through SIM, sEVs with the PI4P sensor were significantly more abundant than sEVs with the PI3P sensor. CD63‐GFP was used as a positive assay control because most sEVs were CD63‐positive (Figure 1e,f). As a negative control, we detected the non‐sEV PKH26‐only staining, which showed no signal (Figure S2f). In addition, WB analysis showed that the 2×FYVE_HRS biosensor for PI3P expressed equally to the PH_OSBP biosensor for PI4P in cells (Figure S2g) and sEVs carried more of the PH_OSBP‐GFP protein (Figure S2h).

Previous in vitro studies demonstrated that the PIP domains could not be generated at acidic pH < 4. Meanwhile, other lipids, such as phosphatidylcholine, significantly affect microdomain stability (Graber et al., 2012; Redfern & Gericke, 2004). In the exposure of an acidic MVB environment at pH 5.5, the stability of the PIP microdomain may be interfered during the sEV production. We thus examined the stability of sEVs produced by RAW264.7 cells in a pH 5.5 buffer. The results showed that sEVs remained intact after 2.5 hours (Figure S2i) for a duration sufficient to produce sEVs (Chiu et al., 2016). Therefore, the in vivo environment of the PI4P‐microdomain of cell membrane may contribute to maintain the stability of the microdomain due to other lipids. These evidences showed that PI4P microdomain could remain stable in acidic environment.

Subsequently, the topology of PI4P on the phospholipid bilayer was assessed. Immunogold staining was employed to detect the PI4P on sEVs. The results showed that the PI4P antibody labels were observed on the sEVs. The GFP labels of the GFP‐tagged PI4P sensor were also observed on sEVs, while the secondary antibody negative control was not labelled (Figure 1g). The antibody cannot penetrate the cell membrane, thus suggesting that PI4P might be on the outer leaflet of sEVs. To further verify the possible presence of PI4P, digitonin permeabilization was conducted in PH_OSBP‐GFP expressing sEVs. After permeabilization, the sEVs were immune‐stained with PI4P antibody. The results showed that the colocalization of PH_OSBP‐GFP‐marked sEVs with PI4P antibody‐stained sEVs was improved by permeabilization (Figure S2i). These results revealed that PI4P may exist on the inner and outer leaflets. PI4P is usually expected to be on the cytosolic leaflet of membranes. Although the formation of PI4P may occur on the outer leaflet, we could not rule out the possibility that some parts of the membranes had changed structure during the treatments by flipping parts inside out. Moreover, synthetic vesicles can be reassembled in vitro after freeze‐thaw rupture (Sou et al., 2003). sEVs may be formed when the sEV membrane ruptured during separation and the membrane was closed in an inverted manner. Alternatively, it is possible that PI4P kinase is encapsulated in the MVB lumen, synthesizing PI4P on the outer leaflet of ILVs. In addition, PI4P may be transferred to the outer leaflet by exchanger proteins, such as OSH4P (de Saint‐Jean et al., 2011), in response to immune stimulation during MVB formation.

3.2. The PI4P content positively correlates with sEVs release upon TLR4 activation

To monitor the release rate of sEVs, we stably expressed pHluorin‐tagged CD63 in RAW264.7 cells (Sung et al., 2020). The pHluorin tag is a pH‐sensitive fluorescent protein that fluoresces under neutral conditions and no fluoresces under acidic conditions (Figure S3a, Video S1). CD63 is highly enriched in MVBs, which are acidic environments. The pHluorin that contacts the neutral extracellular environment emits light when MVBs fuse with the cell membrane. We used TIRF imaging to record flashes on the cell membrane to monitor sEV release events (Figure 2a). We monitored the instantaneous sEV release rate at 0, 1, 2, 4, 10, and 24 h of LPS treatment. The rate was significantly increased at 1, 2, and 4 h after LPS stimulation but began to decline after 10 h of treatment and had returned to the basal level by 24 h (Figure 2b). These findings suggested that sEV release was inducible by LPS stimulation. In addition to real‐time detection, we detected the total sEV counts upon LPS stimulation for 8 and 24 h. The number of sEVs released at 8 h of LPS stimulation increased compared to that of the 8 h control, whereas sEV release at 24 h of LPS treatment showed a decrease to 24 h control (Figure 2c). The sizes of the sEVs barely changed (Figure S3b).

FIGURE 2.

The PI4P content positively correlates with sEV release upon TLR4 activation. (a) TIRF imaging monitoring sEV release from the cell membrane surface. Each curve shows an MVB‐PM fusion event. (b) The instantaneous release rate of sEVs in RAW264.7 cells after LPS stimulation. The number of fluorescent flashes on the cell surface, as shown in Figure 2a, was recorded. Each dot represents an MVB‐PM fusion event (fluorescent spots) over a time‐lapse recording of 3‐min onto a single cell. The separated scatter plot shows the mean with SEM, and represents three biological replicates. Unpaired two‐tailed t‐test (^*** P < 0.001; ^**** P < 0.0001). (c) NTA determining the cumulative number of sEVs in RAW264.7 cells after different LPS stimulation periods. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^*** P < 0.001; ^**** P < 0.0001). The bar plot represents three independent biological replicates. (d) The PIP content in sEVs detecting by LC‐MS/MS. The bar shows the mean with SEM. Unpaired two‐tailed t‐test (^**** P < 0.0001). The bar plot represents three independent biological replicates. (e) sEVs derived from RAW264.7 cells stably expressing GFP‐tagged PI4P or PI3P sensor were observed by SIM imaging before and after LPS stimulation. Green indicates the PIP sensor‐GFP, and red indicates PKH26‐stained particles. Scale bar, 2 μm. (f) Quantitative results from Figure 2e. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^**** P < 0.0001). The bar plot represents four independent biological replicates. (g) The PIP species in RAW264.7‐cell‐derived sEVs before and after LPS stimulation. The concentration was equivalent to the internal standard. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^**** P < 0.0001). The bar plot represents three independent biological replicates. (h) PIP species in RAW264.7 cells before and after LPS stimulation. The concentration was equivalent to the internal standard. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^**** P < 0.0001). Bar plot represents three independent biological replicates. (i) WB detection of the distribution of PIP sensors in early endosomes and MVBs isolated from RAW264.7 cells by density gradient centrifugation. Left panel shows untreated cells, and right panel shows cells treated with LPS for 24 h. EEA1, early endosome marker; VPS16, MVB marker; GFP, PI4P sensor; FLOT1, cell membrane marker; P5, the pellets from 5000×g centrifugation; and P17, the pellets from 17,000×g centrifugation.

Subsequently, the variation of the PIP content in sEVs was investigated at different release rates. We examined the PIP content in sEVs isolated from cells without LPS treatment and cells treated with LPS for 8 and 24 h. The PIP content was normalized to the number of sEVs. It showed that PIP levels per 10¹⁰ sEVs were significantly increased in the group treated for 8 h and had recovered in the group treated for 24 h (Figure 2d). SIM imaging was performed to determine the distribution of PIPs on sEVs. After LPS stimulation, the number of sEVs with the PI4P sensor increased significantly, but the change in the PI3P content was not obvious (Figure 2e,f). PI‐bisphosphates (PIP2) and PI‐trisphosphate (PIP3) were also examined to exclude these effects on sEV release. The content of PI(4,5)P₂ was the most abundant and was similar to previous reports (Morioka et al., 2022). Nonetheless, both PIP2 and PIP3 were unrelated to the release of sEVs (Figure S3c,d). These results indicated that the PI4P content positively correlated with the sEV release rate. Further analysis of the PI4P species showed that the levels of almost all species remained unchanged or decreased (Figure 2g), indicating that the sEVs did not exhibit selectivity among PI4P species containing a different number of carbon atoms for encapsulation.

To explore the source of the changes in PI4P in the sEVs, we analyzed the PI4P content in cells. It showed that the cellular PI4P content increased significantly after 24 h of LPS treatment (Figure 2h). Compared to the tendency of sEVs to decrease, the changes in the abundance of PI4P in sEVs may be derived from the aggregation or transfer of specific organelles. Subsequently, the expression levels of PH_OSBP‐GFP in the subcellular organelles separated by density gradient centrifugation were detected. The data showed that PI4P was located on the endosomal, cellular organelle and plasma membranes (Figure 2i). The PI4P in separated early endosomes and MVBs was measured by chiral LC‐MS/MS, the PI4P increased in MVBs (Figure S3e,f). These findings indicated that the distribution of PI4P in early endosomes is slightly reduced after LPS stimulation, suggesting that LPS stimulation may slightly enrich intracellular PI4P in organelles associated with the sEV secretion pathway, such as early endosomes.

3.3. The PI4P‐interacting protein HSPA5 decreases the sEVs release rate

To investigate why the amount of PI4P in sEVs decreased and why the sEV release rate slowed even though the intracellular production of PI4P increased after long‐term LPS stimulation. Proteomic analysis was performed to identify the proteins that interact with PI4P. RAW264.7 cells were stably expressed Flag‐OSBP_PH and intracellular proteins crosslinking using DSS to stabilize proteins close to PI4P (Figure S4a). The interacting proteins were captured by Flag beads and identified by mass spectrometry. We compared the changes in PI4P‐interacting proteins in untreated RAW264.7 cells and RAW264.7 cells treated with LPS for 24 h (Figure 3a). The results showed that HSPA5 was significantly enriched in the 24‐h LPS treatment group (Figure 3b). HSPA5 is a member of the heat shock protein 70 (HSP70) family and localizes to the ER lumen where it functions as a typical HSP70 chaperone involved in protein folding and is a master regulator of ER homeostasis. Elevated expression of this protein and its atypical translocation to the cell surface during PAMP stimulation have been reported (Abdel‐Nour et al., 2019). Considering the effects of LPS stimulation time on the sEV release rate, we detected the relationship between LPS stimulation time and HSPA5 expression. The mRNA expression of HSPA5 increased rapidly at the late stage of LPS stimulation. The HSPA5 mRNA content increased by four to eight times beginning at 18 h after LPS stimulation (Figure 3c).

FIGURE 3.

HSPA5 interacts with PI4P to reduce sEV release. (a) Volcano plot showing differentially enriched immunoprecipitated proteins between untreated RAW264.7 cells and cells treated with LPS for 24 h. Coloured dots indicate proteins with a P‐value < 0.05 and log2(fold change) > 1.5. (b) Heatmap showing the expression pattern of significantly differentially expressed proteins in sEV‐related proteins. (c) The mRNA expression level of HSPA5 in RAW264.7 cells induced by different treatment periods of LPS. The bar plot shows the mean with SEM. (d) TIRF monitoring changes in the release rate of sEVs after HSPA5 inhibition or downregulation. shRNA mock: PLKO.1 empty vector control. HA‐15, HSPA5 ATPase inhibitor. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^** P < 0.01; ^**** P < 0.0001; ns, not significant). The bar plot represents three independent biological replicates. (e) The release rate of sEVs after knocking down HSPA5 or re‐expressing HSPA5‐BFP. OE mock, overexpression empty vector control. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^** P < 0.001; ns, not significant). The bar plot represented at least three independent biological replicates. (f) Confocal images showing the localization of MVBs and PI4P after HSPA5 knockdown. Red, green, and blue indicated anti‐PI4P, anti‐VPS16 and DAPI, respectively. Scale bar, 2 μm. (g) Confocal images showing the localization of MVBs and Golgi apparatus distribution with HA‐15 or LPS treatment. Green, pink, blue, and orange indicate VPS16, CD63, GM130, and PI4P, respectively. Scale bar, 2 μm. The white dashed lines indicate the Golgi apparatus. (h) Pearson's coefficient evaluating the colocalization rate from Figure 3g. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (*P < 0.05; ^** P < 0.01; ns, not significant). The bar plot represents three independent biological replicates.

Subsequently, we examined the regulatory effect of HSPA5 on the sEV release rate. HEK293T cells stably expressing pHluorin‐CD63 were treated with HA‐15, an ATPase inhibitor of HSPA5. When HA‐15 inhibited the ATPase activity of HSPA5, the sEV release rate increased. To further genetically confirm the direct role of HSPA5, HSPA5 was knocked down in HEK293T cells stably expressing pHluorin‐CD63 (Figure S4b), which significantly increased the sEV release rate (Figure 3d). To determine the function of HSPA5, we re‐expressed of HSPA5‐BFP in HSPA5 knockdown cells (Figure S4c). The results showed that rescue of HSPA5 expression reduced the sEVs release, which indicated that HSPA5 was required for shutting down sEV release (Figure 3e). Considering that the HSPA5 response to PAMP stimulation is usually realized through changes in localization, we hypothesized that the localization of PI4P and MVB is related to HSPA5 and sEV generation. To elucidate the function of HSPA5, we knocked down its expression in macrophages. HSPA5 deficiency led to MVB aggregation, and the number of MVBs increased (Figure 3f). Consistent with this hypothesis, the LPS‐regulated localization of PI4P and MVB was detected by imaging, which indicated that LPS did not interfere with Golgi apparatus. PI4P was disseminated from the Golgi to MVBs after 8 h of LPS treatment and relocated to the Golgi after 24 h of treatment (Figure 3g,h). HSPA5 inhibition caused the same depletion phenotype in which MVBs clustered near the PI4P‐localized Golgi apparatus. These results indicated that HSPA5 may regulate the release of sEVs by changing the localization of MVBs.

3.4. HSPA5 recruits RAB10 to restore sEV release

HSPA5 knockdown cells were subsequently treated with the PIP5K1C inhibitor, UNC3230, which increased the PI4P content in cells, and a synergistic effect with HSPA5 knockdown was found to increase the sEV release rate (Figure 4a). We hypothesized that PI4P may affect other proteins during short‐term LPS stimulation to affect the rate of sEV release. To identify the proteins that are regulated by HSPA5 to achieve a reduction in sEV release in early‐stage LPS stimulation, we explored specific protein expression in the 4 h LPS treatment group through proteomic analysis (Figure 4b). Comparing the enrichment of proteins in two groups, we found 17 differentially enriched proteins, of which four proteins were enriched in the 4 h group, and 13 proteins were enriched in the 24 h group (Figure 4c). GO analysis showed that “phagocytic vesicle membrane” was the term with the highest enrichment score (Figure 4d). Among the differentially enriched proteins were two proteins related to this term, namely, RAB10 member RAS oncogene family (RAB10) and vesicle transport through interaction with t‐SNAREs 1B (VTI1B). RAB10 was enriched in the 4 h group. We verified the interaction between PI4P and RAB10 by co‐IP. The interaction between RAB10 and Flag‐OSBP_PH was increased at 8 h, and the effect was decreased at 24 h (Figure 4e). We observed changes in the intracellular localization of RAB10 and found a significant increase in RAB10 aggregation on MVBs after LPS stimulation, and the colocalization of PI4P and the aggregation site increased (Figure 4f). After the knockdown of RAB10 in pHluorin‐CD63‐expressing HEK293T cells (Figure S4d), the sEV release rate was decreased (Figure 4g). From the above results, we concluded that the increase in PI4P on intracellular MVBs promoted the release of sEVs by recruiting RAB10 after short‐term LPS stimulation.

FIGURE 4.

HSPA5 recruits RAB10 to the ER, decreasing sEV release. (a) TIRF monitoring changes in the sEV release rate after HSPA5 knockdown or combined UNC3230 treatment. UNC3230, a PIP5K1C inhibitor; shRNA mock, PLKO.1 empty vector control. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^* P < 0.05; ^**** P < 0.0001). Bar plot represents three independent biological replicates. (b) Volcano plot showing differentially enriched immunoprecipitated proteins between RAW264.7 cells treated with LPS for 4 and those treated for 24 h. Colored dots represent proteins with a P‐value < 0.05 and log2(fold change) > 1.5. (c) Heatmap showing the expression pattern of significantly differentially expressed proteins in Figure 4b. CTL, untreated; LPS‐4, treated with LPS for 4 h; LPS‐24, treated with LPS for 24 h. (d) GO analysis of the protein‐protein interaction network made up of a group of proteins in Figure 4c with significantly differential enrichment. (e) Co‐IP results showing the interaction between RAB10 and PI4P after 8 h or 24 h LPS stimulation. RAW264.7 cells were crosslinked by DSS and pulled down by anti‐Flag (PI4P sensor), followed by WB analysis. (f) Confocal imaging showing changes in the localization of RAB10 and PI4P before and after LPS stimulation. Red indicates anti‐PI4P, and green indicates anti‐RAB10. Scale bar, 1 μm. (g) TIRF imaging quantified the sEV release rate after RAB10 knockdown in HEK293T cells. shRNA mock, PLKO.1 empty vector control. Unpaired two‐tailed t‐test (**P < 0.01). Bar plot represents three independent biological replicates. (h) Co‐IP analysis of the interaction between RAB10 and HSPA5 in HEK293T cells.

Since both RAB10 and HSPA5 caused changes in MVBs, we investigated whether inhibition of the sEV release rate by HSPA5 regulated RAB10. We performed co‐IP of HSPA5 and RAB10 and found that HSPA5 interacted with RAB10 (Figure 4h). The results indicated that LPS promoted the expression of HSPA5 at the late stage of LPS stimulation, induced the interaction of HSPA5 and RAB10, abolished the increase in sEV release caused by RAB10, and restored sEV release to basal levels.

3.5. TLR4 activation regulates the PI4P content on MVBs and the expression of HSPA5 through IFN‐β

There are four main mechanisms involved in PI4P synthesis and metabolism as follows: (1) PI4P is synthesized by PI phosphorylation catalyzed by four kinases (PI4KA, PI4KB, PI4K2A, and PI4K2B), increasing in PI4P levels (N. Y. Hsu et al., 2010); (2) PI(4,5)P₂ is dephosphorylated to PI4P, which is catalyzed by SHIP2, SYNJ1, and OCRL, thereby increasing PI4P levels. (3) The dephosphorylation of PI4P to PI catalyzed by SAC1 decreases PI4P levels (Altan‐Bonnet & Balla, 2012); 4) the phosphorylation of PI4P to PI(4,5)P₂ catalyzed by PIP5K1C causes a decline in PI4P levels (Posor et al., 2022). We next comprehensively analyzed the kinases and phosphatases mRNA expression levels influenced by LPS stimulation. The data indicated that PI4K and SAC1 expression increased at 24 h, thus it cannot estimate whether these increases led to changes in PI4P content (Figure 5a). PIP5K1C decreased sharply at 1 h, and OCRL, SYNJ, and SHIP2 expression was changed less than 50%; these changes led to a PI4P increase (Figure 5b). The competitive reaction by which PI3P is synthesized from PI, along with the expression of PIK3C3 (an enzyme related to the conversion of PI to PI3P), was downregulated, and expression of the MTM1 enzyme, which is related to the reverse reaction (Ketel et al., 2016), upregulated the conversion of PI3P to PI (Figure 5c). These data suggested that LPS triggered PI4P content changes in sEVs might through a PIP5K1C decrease and that a PI4K increase might promote the conversion from PI to PI4P.

FIGURE 5.

The LPS/IRF3/IFN‐β axis regulates PI4K2B and PIP5K1C expression to affect PI4P abundance on MVBs. (a) The relative mRNA levels of OCRL, PIP5K1C, SYNJ, and SHIP2 induced by different periods of LPS stimulation. (b) The relative mRNA levels of PI4K2B, PI4KA, PI4KB, PI4K2A, and SAC1 induced by different periods of LPS stimulation. (c) The relative mRNA levels of MTM1 and PIK3C3 induced by different periods of LPS stimulation. (d) TIRF monitoring changes in sEV release rate after inhibition of PIP5K1C, PI4K, and PIK3C3. UNC3230, a PIP5K1C inhibitor; Autophinib, a PIK3C3 inhibitor; PIK93, a PI4K inhibitor. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^* P < 0.05; ^**** P < 0.0001; and ns, not significant). The bar plot represents three independent biological replicates. (e) The relative mRNA expression levels of PIP5K1C after stimulation with LPS, IFN‐β and TNF‐α. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^*** P < 0.001; ^**** P < 0.0001; and ns, not significant). The bar plot represents three independent biological replicates. (f) The relative mRNA expression levels of PIP5K1C before and after the IFN‐β blocking antibody treatment. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^**** P < 0.0001). The bar plot represents three independent biological replicates. (g) Confocal images showing changes in the localization of MVBs and PI4P before and after LPS stimulation. MVB, anti‐VPS16; PI4P, anti‐PI4P. Scale bar, 1 μm. (h) Quantitative analysis of the results in Figure 5g showing colocalization efficiency. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^**** P < 0.0001). The bar plot represents three independent biological replicates. (i) WB detection of the variation in the expression of HSPA5 after IFN‐β stimulation in RAW264.7 cells. (j) ELISA measuring TNF‐α levels in recipient RAW264.7 cells triggered by LPS‐stimulated cell‐derived sEVs. The bar plot shows the mean with SEM. Unpaired two‐tailed t‐test (^**** P < 0.0001). The bar plot represents three independent biological replicates. (k) WB analysis of p‐IκBα activation in recipient RAW264.7 cells triggered by LPS stimulated cell‐derived sEVs.

To address whether the sEV release rate was modulated following PI4P synthesis, we monitored the sEV release rate influenced by the inhibition of kinases and phosphatases. PIP5K1C and PIK3C3 inhibition increased sEV release, similar to LPS‐triggered PIP5K1C and PIK3C3 downregulation. PI4K inhibition showed a decrease of sEV release consisting of the PI4K upregulating results (Figure 5d). These data indicated that LPS‐triggered changes in sEV release may be caused by modulation of the expression of kinases and phosphatases. Based on the temporal data, PIP5K1C levels changed within 0.5‐1 h after LPS stimulation, and the rapid decrease in the PIP5K1C level led to an increase in the PI4P level. We then explored which branch of LPS/TLR4 signalling caused the changes in the expression of PIP5K1C. After LPS stimulation, two downstream targets of TLR4, the LPS/NF‐κB/TNF‐α and LPS/IRF3/IFN‐β pathways, were activated. These two pathways function through different adaptor proteins and the mechanism by which TLR4 activation stimulates sEV generation can be determined by distinguishing the roles of the two pathways in regulating sEV generation. We stimulated RAW264.7 cells with TNF‐α and IFN‐β and found that IFN‐β had a significant effect on the expression of these two kinases (Figure 5e), and LPS stimulation after pretreatment with IFN‐β antibody blocked LPS‐induced changes in PIP5K1C expression (Figure 5f). These results suggested that LPS affects the expression of PI4P kinase through the IRF3/IFN‐β axis, thereby changing the intracellular PI4P content. Next, the localization of PI4P was observed by confocal imaging, and the colocalization of PI4P with MVBs was found to be increased after LPS stimulation (Figure 5g,h). Other innate immune stimuli were used to treat RAW264.7 CD63‐pHluorin cells, and the sEV release rates were upregulated by IFN‐β‐related ligand treatment (Figure S5a). The distributions of PI4P and MVBs changed in a manner similar to the change observed after LPS treatment (Figure S5b).

To establish associations between IFN‐β stimulation and sEV formation, we examined the effect of IFN‐β stimulation on HSPA5 protein expression levels. The results showed a timeline consistent with the effect of LPS on HSPA5 (Figure 3c). HSPA5 expression was significantly increased after 8 h IFN‐β stimulation (Figure 5i). Subsequently, we separately added untreated or LPS‐treated cell‐derived sEVs to untreated RAW264.7 cells. The LPS‐treated cell‐derived sEVs significantly increased TNF‐α (Figure 5j) and p‐IκBα (Figure 5k) production. To evaluate the sEV secretion rate under physiological conditions, we constructed a Cre/LoxP reporter system. In this system, the recipient cells that expressed the LoxP‐DsRed‐LoxP‐GFP exhibited red fluorescence, and the donor cells expressed the Cre mRNA. When Cre mRNA was carried into recipient cells by sEVs, the recipient cells changed to express GFP as a green signal (Zomer et al., 2015). In the transwell cell‐culture system, we cultured donor cells in the bottom well and recipient cells in the upper well (Figure S5c). The donor cells were pretreated with IFN‐β or TNF‐α and then cocultured for 24 h. IFN‐β pretreatment produced more significant GFP signals than TNF‐α pretreatment (Figure S5d). These results suggested that IFN‐β promoted the release of sEVs, and that the temporal and spatial distribution of PI4P in cells changed after LPS stimulation. In addition, an increase in the PI4P content on MVBs was caused by LPS activation‐induced IFN‐β production, which inhibited the expression of PIP5K1C. This PI4P distribution caused sEV release to promote intercellular TLR4 signal transduction.

3.6. TLR4 activation modulates the sEV release rate in primary mouse macrophages

In the macrophage cell line RAW264.7, LPS modulated sEV release in a time‐dependent manner. Because it remained uncertain whether this state is consistent with physiological immune responses, we validated the results in primary BMDMs from C57BL/6J mice. BMDMs were separated and cultured with growth factors for maturation. The MVBs in the cells were observed by TEM after negative staining, and the MVBs counts in each ultrathin section cell ranged between 4 and 10. The ILVs indicated the precursor of sEVs, and the amounts of ILVs in each MVBs of a single ultrathin section were counted. After 4 h of LPS stimulation, the number of ILVs in the intracellular MVBs was significantly increased, and the number of ILVs in the MVBs was significantly decreased after the treatment extended to 12 h (Figure 6a,b). BMDMs were then stimulated with LPS for different durations, and the distributions of PI4P, MVBs, and HSPA5 were observed to be similar to those in RAW264.7 cells (Figure 6c). After stimulation with IFN‐β, the number of ILVs significantly increased in the MVBs, but the change due to TNF‐α stimulation was not obvious (Figure 6d,e). Similarly, the regulation of ILV counts by PI4P‐generated kinases was verified as BMDMs treated with the PIP5K1C inhibitor significantly increased the number of ILVs (Figure 6f,g). The distributions of PI4P, MVBs, and HSPA5 were also the same (Figure S6). These results suggested that under physiological immune conditions, LPS regulated PIP5K1C via IFN‐β, affected PI4P formation, and regulated ILV production in MVBs.

FIGURE 6.

TLR4 activation modulates the sEV release rate in mouse primary macrophages. (a) Characterization of MVB morphology and ILV numbers in BMDMs from C57BL/6 mice using TEM. BMDMs were treated with LPS for the indicated period. CTRL samples were treated with DMSO only. Scale bar as shown. (b) The number of ILVs per MVB shown in Figure 6a. The MVBs in 15–20 ultrathin section cells were counted per treatment from three independent experiments. Each dot represents an individual MVB. The result shows the mean with SEM. One‐way ANOVA test (^**** P < 0.0001; ns, not significant). (c) Confocal imaging of the PI4P, MVB, and HSPA5 localization in BMDMs treated with LPS for different durations. Green, pink, blue, and orange indicate VPS16,HSPA5, CD63, and PI4P. Scale bar, 5 μm. The white dashed lines indicate the PI4P enriched location. (d) Characterization of MVB morphology and ILV numbers in BMDMs treated with 100 Units of TNF‐α, or IFN‐β for 2 h. CTRL samples were treated with DMSO only. Scale bar as shown. (e) The number of ILVs per MVB shown in Figure 6d. The MVBs in 15–20 ultrathin section cells were counted per treatment from three independent experiments. Each dot represents an individual MVB. The results show the mean with SEM. One‐way ANOVA test (ns, not significant; ^**** P < 0.0001). (f) The number of ILVs per MVB shown in Figure 6g. The MVBs in 15–20 ultrathin section cells were counted per treatment from three independent experiments. Each dot represents an individual MVB. The results show mean with SEM. Unpaired t‐test (^*** P < 0.001). (g) Characterization of MVB morphology and ILV numbers in BMDMs treated with or without 1 mM UNC3230 for 1 h. Scale bar as shown.

In conclusion, as shown in both cell lines and primary mouse cells, the LPS‐triggered innate immune response regulates sEV release to modulate signal transduction. Mechanistically, after short‐term treatment, LPS activates TLR4 to produce IFN‐β, altering PIP5K1C to change the abundance of PI4P on MVBs. PI4P recruits RAB10 to MVBs and promotes sEV release. With long‐term treatment, IFN‐β‐induced HSPA5 expression caused PI4P to migrate from MVBs to the ER or Golgi, thereby decreasing sEV release (Figure 7).

FIGURE 7.

Model of the mechanism by which PI4P regulates sEV generation during early and late LPS induction. The right side of the cell shows the early response in which PI4P kinases and phosphatases are regulated via the LPS/IFN‐β axis. RAB10 acts as a PI4P effector protein, increasing ILV formation and fusion with the plasma membrane and increasing sEVs release. The left side of the cell shows that at the late stage, an increase in HSPA5 expression is induced by LPS, and HSPA5 recruits PI4P away from MVBs to the ER to reduce sEVs release.

4. DISCUSSION

The biogenesis of sEVs involves the protein sorting machinery of the MVB membrane, including the ESCRT and various lipidic molecules (Record et al., 2018). Previously reported lipids were involved in the biogenesis of sEVs, especially abundant lipids, including ceramides, sphingomyelin (Trajkovic et al., 2008), lysobisphosphatidic acid (Matsuo et al., 2004), and phosphatidic acid (Egea‐Jimenez & Zimmermann, 2018). However, they cannot response to signal rapidly. The present study first identified PIPs in macrophage‐derived sEVs and clarified the content of PI4P changed in sEVs along with different release rates in a time‐dependent manner. These imply that low levels of lipids in sEVs may be required for the mechanism of sEV release. The abundance of PIPs was similar in sEVs derived from macrophages and the cancer cell line (Morioka et al., 2022). The pattern of PIP species in sEVs, serum, lipoproteins, and cells were similar, with C38:4 and C38:3 being the most, followed by C36:2 and C36:1 (Gardner et al., 2019; Morioka et al., 2022). Previous work revealed that 50–80% of the PIP molecules in primary mammalian cells to be C38:4 (Traynor‐Kaplan et al., 2017). Our results showed a similar pattern in cells and sEVs. However, the distribution of acyl chains is different among various tissues. Further, this distribution can change with disease mutations or with the knockout of lipid metabolizing enzymes (D'Souza & Epand, 2014). Previous studies reported that PIP species with less unsaturation would not be as readily converted to PI(3,4,5)P₃, a lipid required for the activation of AKT, with effects on cell proliferation (Epand, 2017). Moreover, a reduction in the percentage of C38:4 in PIPs accompanied by prolonged Ca²⁺ oscillations upon stimulation (Takemasu et al., 2019). In the present study, LPS treatment caused detectable changes in each acyl chain of PIPs in cells, with no apparent selectivity. It may illustrate that LPS treatment has little effect on lipid metabolizing enzymes. In sEVs, C38:4 was the most abundant species. The C38:4 also decreased most obviously upon the LPS treatment, while C36:1 did not change. It may suggest that sEVs with different generation rates might enrich different acyl chains.

PI4P achieves its functions by regulating the cellular localization of its effector proteins (Boura & Nencka, 2015). Endosomal PI4P within pigment cells was required for cargo delivery to melanosomes in coordination with the biogenesis of lysosome‐related organelles complex 1 (BLOC‐1) (Jani et al., 2022). Small GTPases as effector proteins, are recruited by PIP kinases and phosphatases to regulate sEV secretion. Many studies have shown that RAB27A, RAB27B (Ostrowski et al., 2010), and RAB7 (Baietti et al., 2012) are employed in different control steps of the sEV secretion pathway. The inhibition of RAB35 function leads to intracellular accumulation of endosomal vesicles and impairs sEV secretion (C. Hsu et al., 2010). We identified the effector proteins of PI4P through IP‐based proteomics. Comparing the known RABs with our candidate HSPA5, we tested the RAB GTPases found in our proteomic results, namely, RAB2A and RAB5A. Knocking down of RAB2A and RAB5A showed the opposite behaviour to HSPA5 in sEV secretion. These results suggest that PI4P‐mediated sEV secretion in response to LPS stimulation is a new pathway independent of known secretion routes. Thus, HSPA5 may be a potent target in inhibiting sEV biogenesis.

PI4P conversion is a precise process in response to signal transduction. Regulating PI4P kinases and phosphatases is a vital method of modulating the biogenesis of EVs. The generation controlled by PI4K2A of PI4P at endosomes is necessary for the biogenesis of sEVs (Crivelli et al., 2022). ARF6 and its effector, PIP5K (specifically the gamma i5 isoform), are required for EGFR sorting into ILVs of the MVBs upon EGF stimulation in MDA‐MB‐231 cells (Sun et al., 2013). Depletion of PIP5K1C does not impact syntenin endosomal budding, but the effect of PIP5K on the release of sEVs was not investigated (Ghossoub et al., 2014). The present study provided evidence that PIP5K1C depletion increases sEV release in macrophages. Therefore, PI4P modulated by PIP5K1C may be a possible way to regulate the generation of ILVs.

Dysregulation of sEV release is implicated in various pathological conditions. A variety of pathogens, such as influenza A virus, hijack the host circulating endosome system and exploit PI4P/BLOC‐1 module during the building of functional recycling endosomal tubules to their benefit (Jani et al., 2022). Modulating sEV secretion is a potential therapeutic approach in the clinic (Murphy et al., 2019). Previous studies have reported many cascades leading to increased sEV secretion, including physical contact with neighbouring cells to provoke sEVs, cytokines, and calcium that initiate signalling (van Niel et al., 2022). In THP‐1 cells, macrophages secreted cytokines shortly after stimulation with IFN‐β, leading to increased sEV release (Cai et al., 2018). The present study investigated the release of sEVs regulated by IFN‐β in a time‐dependent manner. The kinetics of IFN‐β induced the release of sEVs, which was increased by acute stimulation at 1 h and decreased to basal levels at 24 h. This result shows an inducible process that cytokines control the release of sEVs, thereby protecting the immune system from overactivation and providing new insight into the drug discovery of immunomodulators.

Taken together, the present findings provide evidence for the inducible release of sEVs, which is triggered by the LPS treatment. This inducible release may be regulated by PI4P in sEVs produced by PI5K1C depletion, through the LPS/IFN‐β cascade. Mechanistically, PI4P interacts with the effector proteins, HSPA5 and RAB10, to change the generation of ILVs, which are precursors of sEVs. PI4P in sEVs is used to generate ILVs secreted into the extracellular space.

AUTHOR CONTRIBUTIONS

Hang Yin and Yu Xia conceptualized the project and supervised all the experiments. Xue Jin designed and performed the experiments and wrote the manuscript. T.X. developed and performed the mass spectrometry experiments. Shuchen Luo performed the sEVs isolation and SIM imaging experiments. Ying Zhang supervised the biological experiments and revised the manuscript.

COMPETING INTEREST STATEMENT

The authors declare that they have no competing interests.

Supporting information

Figure S1 The fragmentation map of [M+NH₄] ⁺ for M = PI3P 18:1/18:1+ 5Me.

Figure S2 Isolation and characterization of EVs.

Figure S3 LPS stimulation modulated EV secretion and PIP content.

Figure S4 Regulation and interaction ofHSPA5 and RAB10

Figure S5 IFN‐β modulated MVB localization and sEV release

Figure S6 LPS/TLR4 activation regulated PI4P, MVB and HSPA5 distribution in BMDMs.

Table S1 MRM transitions and collision energy values used to measure each phosphoinositide acyl variant.

Table S2 Primers used for quantitative PCR and constructing shRNA cell lines

Supporting Information

Jin, X. , Xia, T. , Luo, S. , Zhang, Y. , Xia, Y. , & Yin, H. (2023). Exosomal lipid PI4P regulates small extracellular vesicle secretion by modulating intraluminal vesicle formation. Journal of Extracellular Vesicles, 12, e12319. 10.1002/jev2.12319

DATA AVAILABILITY STATEMENT

All data associated with this study are presented in the paper or the Supplementary Materials. Materials that support the findings of this study are available from the corresponding author upon request.