Abstract

   Extracellular vesicle (EV)–based immunotherapeutics have emerged as
   promising strategy for treating diseases, and thus, a better
   understanding of the factors that regulate EV secretion and function
   can provide insights into developing advanced therapies. Here, we
   report that nutrient availability, even changes in individual nutrient
   components, may affect EV biogenesis and composition of immune cells
   [e.g., macrophages (Mφs)]. As a proof of concept, EVs from M1-Mφ under
   glutamine-depleted conditions (EV^GLN−) had higher yields, functional
   compositions, and immunostimulatory potential than EVs from
   conventional GLN-present medium (EV^GLN+). Mechanistically, the
   systemic metabolic rewiring (e.g., altered energy and redox metabolism)
   induced by GLN depletion resulted in up-regulated pathways related to
   EV biogenesis/cargo sorting (e.g., ESCRT) and immunostimulatory
   molecule production (e.g., NF-κB and STAT) in Mφs. This study
   highlights the importance of nutrient status in EV secretion and
   function, and optimizing metabolic states and/or integrating them with
   other engineering methods may advance the development of EV
   therapeutics.
     __________________________________________________________________

   Metabolic rewiring by fine-tuning nutrient condition is a promising
   strategy for bioengineering EVs as therapeutics in diseases.

INTRODUCTION

   Extracellular vesicles (EVs) are cell-secreted nanovesicles enclosed by
   lipid bilayer membranes that serve as key mediators of cell-cell
   communication ([44]1–[45]3). Immune cell–derived EVs play critical
   roles in regulating innate and adaptive immune processes (e.g., immune
   signaling and antigen presentation) by transferring diverse types of
   cargos (e.g., nucleic acids, proteins, and lipids) between cells. In
   recent years, the use of EV-based immunotherapies has emerged as a
   promising strategy for treating various immune-related diseases, such
   as autoimmune or inflammatory diseases, vaccinations, and cancers
   ([46]1, [47]4). For example, M1-like macrophage (Mφ)–derived EVs
   (M1-EVs) can enhance the effectiveness of cancer vaccines and
   chemotherapeutics ([48]5, [49]6). As cell-derived natural nanovesicles,
   EVs offer several advantages over other synthetic nanocarriers, such as
   lower toxicity, higher biocompatibility, and the ability to cross
   biological barriers. Currently, in vitro cell culture–based methods are
   commonly used to produce therapeutic EVs. However, the clinical
   translation of EV therapies is largely hindered by their low yield and
   unsatisfactory bioactivity ([50]3, [51]7). Although previous studies
   have explored methods (e.g., exogenous physical or drug stimuli) for
   increasing EV secretion or cargo levels to some extent ([52]7), the
   efficient production of highly bioactive EVs remains an enormous
   challenge in this field. A better understanding of the factors
   influencing EV secretion, cargo composition, and function can
   facilitate the manipulation EVs to improve their therapeutic potential.
   Thus, it is important to explore the specific factors that regulate EV
   biogenesis and the function of immune cells.

   Nutrients are indispensable substrates for supporting cell growth and
   various cellular processes both in vitro and in vivo. There is an
   abundant amount of studies indicating that metabolic status has a
   substantial impact on the phenotypes and functions of immune cells
   (e.g., Mφs and T cells). Immune cell behavior is influenced by changes
   in bioenergetic and material demands, including glucose (GLC), amino
   acids (AAs), and lipids ([53]8). For example, immunostimulatory M1-like
   Mφs (M1-Mφs) heavily rely on aerobic glycolysis while exhibiting a
   disrupted tricarboxylic acid (TCA) cycle to support rapid cell
   proliferation and cytokine production ([54]9). Immunosuppressive
   M2-like Mφs (M2-Mφs) exhibit an intact TCA cycle and predominantly use
   oxidative metabolism to regulate immune responses and promote tissue
   repair ([55]9). Additionally, α-ketoglutarate [αKG; mainly derived from
   glutamine (GLN)] deficiency can promote M1-Mφ polarization and enhance
   the production of inflammatory cytokines [e.g., interleukin-1β (IL-1β),
   IL-6, and tumor necrosis factor–α (TNF-α)] by activating the prolyl
   hydroxylase domain–mediated nuclear factor κB (NF-κB) pathway ([56]10).
   Several recent studies have shown that certain nutrient conditions,
   such as serum or GLC starvation, can induce EV release and/or alter EV
   compositions in different cells ([57]2, [58]7). For example, serum
   starvation for 20 hours can stimulate monocytes (THP-1 cells) to
   release more EVs containing TSG101 and active tissue factor than
   conventional culture conditions, and such EVs can promote endothelial
   thrombogenicity and apoptosis in human umbilical vein endothelial cells
   ([59]11). Additionally, cardiomyocytes subjected to GLC starvation for
   48 hours secreted larger amounts of EVs that can stimulate the
   proliferation and angiogenesis of endothelial cells, and these effects
   may be due to the enrichment of EV proteins and microRNAs (miRNAs)
   related to metabolic regulation ([60]12). These findings suggest that
   changes in nutrient availability may affect the biogenesis and
   composition of EVs. However, the nutrient conditions and underlying
   mechanism that can enhance the yield and bioactivity of EVs in immune
   cells have not been fully elucidated.

   Here, we report that the nutrient state regulates both the biogenesis
   and function of EVs in Mφs and that fine-tuning the nutrient state is a
   potent strategy for bioengineering immune cell–derived EVs. Briefly,
   changes in individual nutrient components [e.g., GLC, AAs, and fatty
   acids (FAs)] in the medium altered the secretion of EVs from Mφs to
   some extent. As a proof of concept, GLN-depleted (GLN^−) conditions
   increased the yield, functional content, and immunostimulatory potency
   of EV preparations from M1-Mφs. GLN^− conditions up-regulated pathways
   related to EV biogenesis/cargo sorting and immunostimulatory factor
   production in M1-Mφs, and these effects were largely dependent on
   systemic metabolic rewiring of donor cells in response to the GLN^−
   state. In summary, these results highlight the importance of metabolic
   regulation in EV biogenesis and immune cell function and suggest that
   fine-tuning metabolic states represents a powerful strategy for
   engineering immune cell–derived EVs with enhanced therapeutic
   potential.

RESULTS AND DISCUSSION

Changes in nutrient conditions affect EV secretion in Mφs

   Increasing evidence indicates that the availability of nutrients in the
   environment can affect the metabolism and function of immune cells, and
   the most important nutrients include GLC, AAs, FAs, and vitamins
   ([61]8). Mφs are primary innate immune cells involved in various immune
   actions, such as defense against pathogens/danger signals,
   inflammation, and tissue repair ([62]3). Additionally, the diverse
   functions of Mφ-derived EVs (Mφ-EVs) are involved in the pathology of
   many diseases, such as infections, autoimmune disorders, and cancers
   ([63]3). Here, Mφs were selected as an example to explore the impacts
   of nutrient conditions on EV production in immune cells. EVs were
   isolated from the conditioned medium of mouse Mφs (RAW 264.7) using an
   ultracentrifugation method as previously reported ([64]13) ([65]Fig.
   1A). The resulting Mφ-EV preparations exhibited a typical bilayer
   spherical structure with a size of approximately 150 nm, and they were
   positive for EV markers such as endosomal sorting complex needed for
   transport (ESCRT)–associated proteins (e.g., TSG101), chaperone
   proteins [e.g., heat shock protein 70 (HSP70)], and intraluminal
   vesicle (ILV) formation-related proteins (e.g., ALIX) ([66]Fig. 1B).
   Moreover, there were no obvious signals of the negative marker GM130, a
   protein from the Golgi apparatus, in these EV samples ([67]Fig. 1B).
   These results indicate that our EV preparations fulfill the fundamental
   criteria outlined in the Minimal Information Guidelines for
   Extracellular Vesicle Studies (MISEV2018), which advise confirming the
   presence of at least two EV-positive markers (e.g., TSG101 and ALIX)
   and the absence of at least one EV-negative marker (e.g., GM130) and
   conducting supplementary morphological analyses (e.g., membrane
   structure and size) ([68]14).

Fig. 1. Effect of nutrient availability on EV secretion in Mφs.

   [69]Fig. 1.
   [70]Open in a new tab

   (A) Schematic illustration of the EV isolation and characterization
   process. (B) Representative TEM images of EV preparations (left), size
   distributions of EVs measured by NTA (middle), and Western blot
   analysis of EV markers (ALIX, HSP70, TSG101, and GM130; right). (C)
   Schematic illustration of major cellular nutrient components. (D) EV
   sizes determined by NTA from Mφs cultured in control medium (nl) or
   modified medium supplemented with 2.5% FBS (lo), 20% FBS (hi), GLC (9
   mM, hi), OA (100 μM, hi), BCAA (5 mM, hi), or GLN (8 mM, hi) for 24
   hours. (E) Relative EV yields of the different groups were quantified
   by normalizing the EV protein mass to the cell protein mass. (F)
   Western blot analysis of the expression of EV markers (HSP70 and
   TSG101) in equal amounts of EV samples with equal protein amounts (10
   μg per panel). (G) Quantification of the expression of EV-positive
   markers (HSP70 and TSG101) in various groups. BSA was used as a carrier
   to prepare the OA solution, while BSA alone served as the vehicle
   control (n ≥ 3; **P < 0.01, *P < 0.05, ^NSP ≥ 0.05 versus the “nl”
   group).

   We then investigated the detailed effects of altered nutrient additions
   (GLC, AAs, or FAs) on the secretion of EVs from M0-Mφs ([71]Fig. 1C).
   M0-Mφs were initially cultured in control medium (complete RPMI 1640,
   widely used for Mφ culture) containing 10% fetal bovine serum (FBS), 11
   mM GLC, and 2 mM GLN, modified medium containing lower (2.5%) or higher
   (20%) FBS, higher GLC (20 mM), higher GLN (8 mM), branched-chain amino
   acid (BCAA; 5 mM) supplementation, or oleic acid (OA; 100 μM)
   supplementation for 24 hours. The supernatants from each culture
   condition were collected for EV preparation. Because of the base RPMI
   1640 (without FBS) already containing GLC (11 mM) and GLN (2 mM),
   medium with no or low GLC/GLN could not be included in this step. The
   impact of an individual nutrient on EV production will be assayed in
   the next steps using a specific medium containing no GLC or GLN. The
   results showed that the EV preparations from the control and altered
   nutrient conditions had similar mean sizes (~150 nm) ([72]Fig. 1D).
   Each method offers unique benefits and limitations in quantifying EVs
   based on size, concentration, and other characteristics ([73]15). Here,
   EV yield was assessed using the most common method, that is,
   determining the ratio of the protein concentration of EVs to the
   protein concentration of donor cells ([74]16). The results showed that
   changes in the individual nutrient contents altered the EV yield of
   M0-Mφs to some extent ([75]Fig. 1, E to G). The lower FBS, BCAA
   supplementation, and higher GLN conditions significantly reduced the
   secretion of EVs from M0-Mφs, compared to that in the control
   conditions, while OA supplementation slightly increased the EV
   secretion of these cells ([76]Fig. 1E). Neither higher FBS nor higher
   GLC concentrations significantly altered EV secretion compared to that
   in the control group ([77]Fig. 1E). Further validation tests also
   verified the elevated levels of EV yield in M0-Mφs from the no GLC (0
   mM) group or no GLN (0 mM) group compared to those from the GLC groups
   (2 and 10 mM) and GLN groups (2 and 10 mM), respectively (fig. S1).

   The levels of protein markers of EV preparations in the different
   groups were also examined using equal total protein amounts. Compared
   with the control group, the levels of HSP70 and TSG101 in the lower FBS
   group, higher GLN group, BCAA group, and OA group were lower, while the
   expression of HSP70/TSG101 in the higher GLC group was not
   significantly different ([78]Fig. 1, F and G). Rapamycin (Rapa) affects
   BCAA metabolism by inhibiting mechanistic target of rapamycin (mTOR)
   signaling, which in turn affects the activity of multiple enzymes
   (e.g., ribosomal protein S6 kinase-β1) involved in BCAA metabolism
   ([79]17), and thus, the effect of Rapa on EV secretion of M0-Mφ was
   also evaluated. The BCAA plus Rapa group showed lower levels of HSP70
   and TSG101 than the BCAA alone group (fig. S2A) but did not affect EV
   yield (fig. S2B), suggesting that BCAA metabolism may change the EV
   composition in M0-Mφs, but the detailed reasons are needed to be
   further researched in the future. Collectively, these results suggest
   that changes in nutrient status can affect EV secretion by Mφs.

   Because of the high heterogeneity of EVs, specific markers for
   distinguishing EV subtypes via different biogenetic routes (e.g.,
   exosomes and ecotosomes) are not yet available, making it difficult to
   identify EV subtypes in EV preparations ([80]1, [81]18). Instead,
   MISEV2018 and current reports suggest categorizing EV preparations by
   their size, such as small EVs (~50 to 150 nm), medium-sized EVs (~200
   to 800 nm), and large EVs (≥1 μm) ([82]18). Here, the prepared Mφ-EVs
   were found to be mainly small EVs, and the altered EV secretion is
   likely an overall effect induced by cell metabolic alterations in
   response to different nutrient conditions. Overall, these results
   suggest that nutrient states affect EV secretion by Mφs to some extent.
   Therefore, it is crucial to explore the optimal nutrient conditions for
   the efficient production of therapeutic EVs in diverse immune cells. We
   then aimed to examine the detailed effects of GLN availability on EV
   production, as GLN is one of the most abundant AAs in the blood and
   various tissues and is essential for immune cell functions ([83]19).

GLN-depleted conditions enhanced EV secretion in Mφs

   Since GLN availability markedly affects the EV yield of M0-Mφs
   ([84]Fig. 1, E to G, and fig. S1), as a proof of concept, we chose a
   challenging condition (GLN depletion, GLN^−) to assay the specific
   impact of GLN catabolism on Mφ-EV secretion and function, since this
   method has been used to assay the effect of GLN metabolism on Mφ
   phenotype switching ([85]10). GLN depletion can completely shut down
   intracellular GLN metabolic flux, which is essential for the carbon and
   nitrogen necessary for anabolic metabolism. Glutaminolysis converts GLN
   to glutamate [via glutaminase 1 (GLS1)] for the TCA cycle, and to
   glycine and other substrates crucial for nucleotide (e.g., DNA and RNA)
   synthesis ([86]20). Thus, the functions of diverse immune cells (e.g.,
   Mφs and T cells) heavily rely on GLN availability and consumption
   ([87]10). The effect of GLN^− conditions on Mφ cell growth and cell
   viability was first evaluated. The cell growth rates of the GLN^− group
   were not affected at 16 hours, were slightly reduced at 24 hours, and
   significantly declined at 48 hours compared to those of the GLN^+ group
   (control group) (fig. S3A). Similarly, compared with those in the GLN^+
   group, the percentage of live cells in the GLN^− group displayed a
   minor decrease (less than 5%) at 24 hours (fig. S3B). These results
   suggest that both the cell growth and cell viability of most Mφs are
   minimally affected by GLN depletion for 24 hours and that the increase
   in EV production in this state is primarily due to metabolic
   alteration; thus, these conditions were chosen for the following
   experiments.

   Next, M0-Mφs were cultured in medium supplemented with (GLN^+, 2 mM) or
   without GLN (GLN^−) for 24 hours, after which resulting EVs were
   isolated and assayed ([88]Fig. 2A). EV preparations from the GLN^− and
   GLN^+ conditions exhibited similar characteristics in terms of mean
   size (~150 nm) ([89]Fig. 2B) and bilayer membrane structure (fig. S4A).
   Moreover, the EV yield (as indicated by the ratio of EV protein
   mass/cell protein mass or EV number/cell protein mass) of the GLN^−
   group was higher (~1.5-fold) than that of the GLN^+ group ([90]Fig.
   2C). The levels of EV marker expression (HSP70, TSG101, and ALIX) were
   also higher in the GLN^− group than in the GLN^+ group ([91]Fig. 2, D
   and E), indicating that GLN^− conditions can promote the secretion of
   EV from M0-Mφs. To validate these results, M0-Mφs were treated with
   BPTES (a GLS1 inhibitor) in the presence of GLN ([92]Fig. 2F).
   Consistently, BPTES treatment also increased the EV yield (~1.4-fold
   increase in the EV number/cell protein mass ratio and ~2-fold increase
   in the EV protein mass/cell protein mass ratio) of M0-Mφ without
   altering the EV size distribution ([93]Fig. 2, G and H).

Fig. 2. Effect of GLN conditions on Mφ-derived EV secretion.

   [94]Fig. 2.
   [95]Open in a new tab

   (A and B) Schematic illustrating GLN^− experiments in M0-Mφs and the
   sizes of EVs from M0-Mφs cultured with GLN (GLN^+, 2 mM) or without GLN
   (GLN^−) for 24 hours. (C) EV yield of different groups determined by
   the EV number/cell protein ratio (left) or the EV protein/cell protein
   ratio (right). (D and E) Western blot analysis and the expression of EV
   markers (HSP70, TSG101, and ALIX) in equal amounts of EVs (10 μg per
   panel, n = 4). (F and G) Schematic map of the GLN catabolism inhibition
   experiment and EV sizes from M0-Mφs cultured in GLN^+ medium with or
   without BPTES (10 nM) treatment for 24 hours. (H) EV yield of different
   groups determined by the EV number/cell protein ratio (left) or the EV
   protein/cell protein ratio (right). (I and J) Schematic illustrating
   GLN^− experiments in M1-Mφs and EV sizes from M1-Mφs cultured with GLN
   (GLN^+, 2 mM) or without GLN (GLN^−) for 24 hours. (K) EV yield of
   different groups determined by the EV number/cell protein ratio (left)
   or the EV protein/cell protein ratio (right) (n = 3; relative to the
   “GLN^+” group; ***P < 0.001, **P < 0.01, *P < 0.05, ^NSP ≥ 0.05 versus
   the GLN^+ group).

   Depending on the different phenotypes of the donor Mφs, Mφ-EVs may have
   stimulatory or inhibitory effects on immune function. For example,
   M1-EVs serve as immune booster agents that enhance antitumor efficacy
   ([96]21), while M2-EVs play an immunosuppressive role in attenuating
   inflammation ([97]13). Thus, the impact of GLN^− conditions on EV
   secretion of M1-Mφ was assessed, as GLN metabolism can support Mφ
   activation and desirable immune responses ([98]10). Lipopolysaccharide
   (LPS)–induced M1-Mφs were cultured in GLN-free RPMI 1640 supplemented
   with (GLN^+, 2 mM) or without GLN (GLN^−) for 24 hours ([99]Fig. 2I).
   EVs from both conditions exhibited comparable size distributions (~150
   nm) and bilayer membrane structures (fig. S4B). The EV yield was
   significantly higher in the GLN^− group than in the GLN^+ group
   (~3-fold increase in the EV number/cell protein mass ratio, and ~2-fold
   increase in the EV protein mass/cell protein mass ratio) ([100]Fig. 2,
   J and K). These results collectively showed that limiting GLN
   availability (or inhibiting glutaminolysis) is sufficient to enhance EV
   production in both M0 and M1 Mφs. Previous studies have shown that
   M1-EVs exert superior therapeutic effects, such as cancer therapy,
   compared to M0-EVs, and this effect may be due to their distinct
   compositions ([101]3). For example, M1-EVs could move to lymph nodes
   and are mainly taken up by Mφs and dendritic cells, which subsequently
   enhances the cytotoxic T cell–mediated anticancer immune response in
   vivo ([102]5). Moreover, our results showed that there was no
   difference in EV yield between the M0-Mφ group and the M1-Mφ group
   under normal culture conditions (fig. S5). Considering the importance
   of future clinical translation, we therefore focused on M1-EVs due to
   their superior therapeutic effects (e.g., cancer therapy and
   anti-infections).

GLN^− conditions enriched the functional composition of Mφ-EVs

   Immune cell–derived EVs play crucial roles in modulating immune
   functions for possible therapeutic use, and they affect the functions
   of target cells by delivering diverse surface molecules (e.g., proteins
   and lipids) and/or encapsulated cargos (e.g., proteins and RNAs)
   ([103]1). EVs can inherit biological signals from donor cells; thus,
   the functions of immune cell–derived EVs can vary due to their diverse
   compositions from donor cells under different pathological conditions
   ([104]18). For instance, Mφ-EVs from the early stage of sepsis
   displayed immunostimulatory effects (e.g., promoting leukocyte
   chemotaxis). This effect is associated with enriched cytokines and
   damage-associated molecular patterns (DAMPs), such as histones,
   high-mobility group box 1 (HMGB1), and HSPs. However, Mφ-EVs from the
   later stage of sepsis can also exert immunosuppressive effects (e.g.,
   reducing serum proinflammatory cytokine levels and the expression of
   adhesion molecules on endothelial cells) due to the expression of CD14
   [glycosylphosphatidylinositol (GPI)–anchored protein] ([105]22). These
   studies suggest that EV compositions (e.g., proteins) are strongly
   associated with their function, and globally profiling EV contents can
   provide insights into their biological roles. Thus, in addition to the
   increase in EV yield, we investigated whether GLN^− conditions can
   alter the composition of Mφ-EVs using liquid chromatography–tandem mass
   spectrometry (LC-MS/MS)–based proteomics ([106]Fig. 3A). Following a
   strict data quality process, we identified a total of 2282 proteins in
   our EV preparations (table S1). The principal components analysis (PCA)
   scatterplot and heatmap showed clear separation and an overall
   differential protein expression pattern between the EV^GLN+ group and
   the EV^GLN− group ([107]Fig. 3, B and C). A volcano plot identified 692
   differentially expressed proteins (DEPs; P-adjusted < 0.05) in the
   EV^GLN− group compared to those in EV^GLN+ group ([108]Fig. 3D). The
   Gene Ontology (GO) enrichment analysis revealed that these DEPs were
   mainly enriched in pathways related to immune processes, such as
   regulation of the inflammatory response (e.g., Ddt, Tnfaip8l2, Lpl,
   IL-1β, and Ptgs2), cytokine production (e.g., Cdk9, Dek, Ddt, and
   Lmnb1), positive regulation of T cell activation (e.g., IL-1β, Cd81,
   and Igfbp2), and cell-cell adhesion (e.g., Tnfaip8l2, IL-1α, and Alox5)
   ([109]Fig. 3, E and F), suggesting that EV^GLN− may function in
   stimulating the immune response.

Fig. 3. Effect of GLN conditions on M1-EV composition.

   [110]Fig. 3.
   [111]Open in a new tab

   (A) Schematic illustrating the EV proteomics experiments. (B) PCA
   scatterplot of different groups based on proteomic data showing the
   differences between different groups (n = 3). (C) Heatmap displaying
   variations in protein composition between the EV^GLN+ group and the
   EV^GLN− group (n = 3). (D) Volcano plots showing the DEPs (FC > 1.5 and
   P-adjusted < 0.05) between the EV^GLN+ group and the EV^GLN− group (n =
   3). (E and F) GO enrichment analysis and interaction network analysis
   of DEPs in response to GLN conditions (FC > 1.5). (G) Heatmap
   illustrating specific immune processes related to DEPs in different
   groups (FC > 1.5). (H) qPCR analysis of cytokine (IL-1β, IL-6, and
   IL-10) gene expression in M1-Mφs or M1-EVs from the GLN^+ or GLN^−
   group (n = 3; ***P < 0.001, **P < 0.01, *P < 0.05, ^NSP ≥ 0.05 versus
   the GLN^+ group).

   To further uncover the potential immunoregulatory molecules enriched in
   the EV^GLN− preparations, the DEPs identified by GO analysis were
   subjected to hierarchical clustering analysis. As depicted in [112]Fig.
   3G, compared with those in the EV^GLN+ preparations, EV^GLN−
   preparations exhibited higher levels of several immunoregulatory
   proteins, such as IL-1α, IL-1β, CDK9, Alox5, and Tnfaip8l2. Multiple
   molecules known to participate in immune regulation, which are involved
   in various immune processes, such as influencing cytokine production,
   orchestrating inflammatory responses, and facilitating innate immunity,
   were identified in EV^GLN− preparations. For example, cytokines (e.g.,
   IL-1α and IL-1β) have been shown to induce M1-Mφ polarization and
   cytokine production by activating the NF-κB and mitogen-activated
   protein kinase (MAPK) signaling pathways ([113]23). IL-1 can boost T
   cell proliferation by up-regulating IL-2 receptor (IL-2R) expression
   and inducing the NF-κB and phosphoinositide 3-kinase pathways
   ([114]24). IL-1β was shown to prompt chemokine secretion in
   keratinocytes and the subsequent recruitment of proinflammatory
   neutrophils, thus initiating a robust innate immune response to
   pathogens ([115]25). In addition to proteins, several other types of EV
   cargos, such as RNAs, may also contribute to the immunoregulatory roles
   of EVs. Thus, the changes in mRNAs related to immune action were also
   analyzed in donor cells and their EVs. Similarly, the results showed
   that both M1-Mφs and EVs from the GLN^− group had higher mRNA levels of
   immunostimulatory cytokines (e.g., IL-1β and TNF-α) and lower levels of
   immunosuppressive cytokines (e.g., IL-10) than those from the GLN^+
   group ([116]Fig. 3H), suggesting that GLN^− conditions may alter the
   levels of diverse types of EV, such as proteins and RNAs. Overall,
   these results indicate that GLN^− conditions can promote the secretion
   and sorting of diverse immunostimulatory molecules into EVs, and that
   these EVs may serve as advanced therapeutics for immunoregulation.

GLN^− conditions enhanced the immunostimulatory effects of Mφ-EVs

   Increasing evidence indicates that immune cell–derived EVs can be
   applied to modulate immune responses and address various immune-related
   diseases due to their ability to either initiate or suppress the immune
   system. For example, M1-EVs modified with NF-κB p50 small interfering
   RNA (siRNA) reduced the levels of M2 markers and cytokines (Arg1,
   TFG-β, IL-10, and IL-4) while increasing the levels of M1-related
   cytokines (IL-12p40 and IFN-γ), and thus enhanced antitumor immunity in
   a breast tumor model ([117]21, [118]26). The previous proteomic results
   ([119]Fig. 3, E to H) indicated that the EV^GLN− protein profile was
   highly enriched in immunostimulatory pathways (e.g., cytokine
   production, positive regulation of T cell activation, and cell-cell
   adhesion), indicating the potential for enhanced immune activation via
   these EVs. Having identified the enrichment of immunostimulatory
   molecules in M1-EVs, we sought to determine whether such EVs can be
   efficiently taken up by targeted cells to shed their cargos ([120]1).
   To assess the cellular uptake of EVs, purified DiD (a lipophilic
   fluorescent dye)–labeled EV^GLN− (DiD-EVs) were incubated with adherent
   Mφs (RAW 264.7) or suspension monocytes (THP-1) for 4 hours. Compared
   with those in the phosphate-buffered saline (PBS) (solvent control) or
   the Dye control group, much higher levels of DiD-positive signals were
   observed in Mφs or monocytes using flow cytometry analysis (FCA) (fig.
   S6), which indicated that EVs can be efficiently taken up by diverse
   immune cells.

   Systematically injected (intravenous) EVs are primarily taken up by
   innate immune cells (e.g., Mφs and monocytes), which subsequently
   recruit and activate many other cells ([121]1, [122]13). To better
   mimic this situation in vitro, we first investigated the
   immunomodulatory effects of such EVs on innate immune cells (e.g., Mφs
   and monocytes). EV^GLN+ or EV^GLN− was added into M2-Mφs induced by
   IL-4/transforming growth factor–β (TGF-β), and the results showed that
   both treatments notably reduced the expression of M2-related genes
   (e.g., Arg1, Mrc1, and TGF-β) in M2-Mφs ([123]Fig. 4, A and B). EV^GLN+
   or EV^GLN− was also added to monocytes, and we found that EV^GLN− had
   higher potency to induce chemokine (e.g., Ccl2 and Cxcl2) expression
   than EV^GLN+ in monocytes ([124]Fig. 4, C and D). Consequently,
   supernatants (containing various chemokines and cytokines) from
   EV^GLN−-treated monocytes had a higher potential to induce chemotaxis
   in mouse splenocytes compared to those in the EV^GLN+ group ([125]Fig.
   4E). Mammalian splenic tissue is rich in functional immune cells. In
   addition, compared with the EV^GLN+ treatment, EV^GLN− treatment
   exhibited a higher capacity to induce splenocyte migration in a
   Transwell system ([126]Fig. 4F). These results suggest that EV^GLN− may
   enhance cytokine release and chemotaxis of innate immune cells and
   could be used as a promising therapeutic for diseases. Moreover, the
   effects of EV^GLN− on T cell activation were also evaluated in mouse
   splenocytes, which contain high numbers of T cells and other leukocytes
   ([127]27). Mouse splenocytes were stimulated with concanavalin A
   (ConA), cotreated with EV^GLN− or EV^GLN+ for 72 hours, and then
   detected using FCA (fig. S7A). The FCA data were visualized using a
   t-distributed stochastic neighbor embedding (t-SNE) scatterplot
   ([128]Fig. 4H, left panel) as previously reported ([129]28). EV^GLN−
   treatment enhanced the ability to induce CD8^+ T cell
   (CD3^+CD8^+CD69^+) or CD4^+ T cell (CD3^+CD4^+CD69^+) activation in
   ConA-primed splenocytes compared to EV^GLN+ treatment ([130]Fig. 4H),
   indicating that EV^GLN− might improve the potential to potentiate T
   cell activation in vitro.

Fig. 4. Immunostimulatory effects of M1-EVs from GLN^− conditions.

   [131]Fig. 4.
   [132]Open in a new tab

   (A and B) M2-Mφs were induced by IL-4/TGF-β (20 ng/ml each), and qPCR
   analysis was conducted to measure the expression of the M2 gene (Arg1,
   Mrc1, and TGF-β) expression in M2-Mφs treated with EV^GLN− (20 μg/ml)
   or EV^GLN+ (20 μg/ml) for 48 hours (n = 3; ***P < 0.001, **P < 0.01,
   ^##P < 0.01, ^###P < 0.01 versus the M2 group). (C and D) qPCR analysis
   of chemokine gene expression (Ccl2 and Cxcl2) in THP-1 monocytes
   treated with EV^GLN− preparations or EV^GLN+ preparations (20 μg/ml)
   for 24 hours (n = 3; ***P < 0.001, **P < 0.01 versus the “CON” group).
   (E and F) Chemotaxis evaluation of (E) conditioned culture medium from
   EV pretreated THP-1 cells and (F) EVs from splenocytes using a
   Transwell system, and the migrated cells in the lower chamber were
   counted using FCA (n = 3; **P < 0.01, *P < 0.05 versus the CON group).
   (G and H) Mouse splenocytes were treated with ConA or ConA plus EV^GLN−
   or EV^GLN+ (20 μg/ml) for 72 hours, and the populations of activated
   CD4^+ T cells and (CD3^+CD4^+CD69^+) activated CD8^+ T cells
   (CD3^+CD8^+CD69^+) were determined by FCA (n = 3). (I and J) Evaluation
   of immune responses in mice (n = 5) intravenously injected with EV^GLN−
   or EV^GLN+ (30 μg/mouse) for 4 hours and immune cell populations
   (F4/80^+ Mφs, Ly6C^+ monocytes, and Ly6G^+ neutrophils) in the spleen
   were analyzed using FCA (**P < 0.01, *P < 0.05 versus the CON group).

   To confirm these in vitro findings, we administered M1-EVs to normal
   mice and observed the effects of these M1-EVs on immune activation in
   vivo ([133]Fig. 4I). The in vivo distribution of M1-EVs was first
   assayed in normal mice by intravenously injecting Cy7 (a near-infrared
   dye)–labeled EV^GLN−. In line with the findings of previous reports
   ([134]13, [135]29), the imaging results revealed that EV^GLN−
   accumulated mainly in major organs, such as the liver, spleen, lung,
   kidney, and heart, of the mice at 4 hours after injection (fig. S8).
   Subsequently, the changes in populations of neutrophils, monocytes, and
   Mφs in response to EV injection were analyzed using FCA (fig. S9, A and
   B), since these cells are key innate cells that participate in various
   functions, such as host defense, inflammation, and tissue injury
   processes. Previous reports have indicated that systematically
   administered (intravenous) EVs are primarily taken up by innate immune
   cells, such as monocytes and Mφs ([136]30). Therefore, our in vitro
   results suggest that intravenously injected EV^GLN− may enhance the
   activation of the innate immune system in vivo. The spleen is a major
   immune organ that contains large numbers of innate immune cells, and
   populations of Ly6G^+ neutrophils, Ly6C^+ monocytes, and F4/80^+ Mφs in
   the spleen tissues of mice were detected at the acute phase (4 hours)
   after EV administration. As shown in [137]Fig. 4J, mice in the EV^GLN−
   group exhibited higher populations of Ly6C^+ monocytes, F4/80^+ Mφs,
   and Ly6G^+ neutrophils than those in the PBS or EV^GLN+ group, which
   corresponded with the findings obtained from the in vitro experiments.
   These findings were consistent with the in vitro findings. The enhanced
   ability of EV^GLN− treatment to boost innate immunity may provide
   additional beneficial outcomes in certain diseases, such as cancer and
   infectious diseases. For example, polarizing Mφs to a proinflammatory
   phenotype can reverse the immunosuppressive tumor environment and thus
   enhance the antitumor responses of Mφs in vivo ([138]31). Overall, our
   results suggest that, compared with conventional EV^GLN+, EV^GLN− has
   immunostimulatory effects in vivo, suggesting that it is a promising
   candidate for boosting immunostimulatory therapy.

GLN^− conditions up-regulated EV biogenesis and cytokine synthesis pathways

   Although GLN^− conditions can increase EV secretion and
   immunostimulatory potency in M1-Mφs, the underlying mechanisms are
   unclear and need to be explored. Thus, RNA sequencing (RNA-seq) was
   conducted to profile the global changes in the gene expression of Mφs
   (table S2). We observed a clear separation and distinct gene expression
   patterns between the GLN^+ and GLN^− groups, as shown in the PCA plot
   and heatmap ([139]Fig. 5, A and B). The differentially expressed genes
   (DEGs; 5206 up-regulated and 3439 down-regulated) that were induced by
   GLN^− conditions were identified ([140]Fig. 5C). GO enrichment analysis
   revealed that the DEGs were involved in multiple pathways related to
   vesicle formation, such as the membrane, plasma membrane, cytoplasmic
   vesicle, and external side of the plasma membrane ([141]Fig. 5D).
   Furthermore, we found that the expression of numerous genes involved in
   the ESCRT subcomplexes, such as ESCRT-0 (e.g., STAM, AP4S1, and
   PICALM), ESCRT-I (e.g., TSG101, VPS37C, and VPS37B), ESCRT-II (e.g.,
   VPS25 and SNF8), and ESCRT-III (e.g., CHMP4B, CHMP1A, and VPS4B), was
   up-regulated in M1-Mφs under GLN^− conditions compared to that in the
   GLN^+ group ([142]Fig. 5E). The gene profile of M0-Mφ under GLN^−
   conditions exhibited a similar pattern, with DEGs enriched in processes
   related to vesicle formation, including cytoplasmic vesicles,
   endosomes, and membranes (fig. S10, A and D, and table S3).
   Additionally, the expression levels of ESCRT genes, such as ESCRT-0
   (e.g., STAM, AP4S1, and CLBA1), ESCRT-I (e.g., TSG101, VPS37C, and
   MVB12B), and ESCRT-III (e.g., CHMP4B, CHMP1A, and VPS4B), were also
   higher in M0-Mφs of the GLN^− group than in those of the GLN^+ group
   (fig. S10E). Notably, the ESCRT pathway plays crucial roles in the
   formation of ILVs and multimembrane endosomes (MVEs) by cooperating
   with the adenosine triphosphatase (ATPase) VPS4 to drive membrane
   scission or sealing ([143]18, [144]32), while depletion of ESCRT
   components was shown to reduce exosome biogenesis in Mono-Mac6 cells
   ([145]33). ESCRT-related proteins, such as TSG101 and ALIX, have been
   widely used as positive markers for EV characterization. Thus, our
   results indicate that the increased secretion of EVs under GLN^−
   conditions is at least partially attributed to the up-regulation of
   ESCRT pathway activity.

Fig. 5. Effect of GLN^− conditions on EV biogenesis in M1-like Mφs.

   [146]Fig. 5.
   [147]Open in a new tab

   (A) PCA scatterplot based on RNA-seq data of M1-Mφs treated with GLN^+
   (2 mM) or without GLN (GLN^−) for 24 hours. (B) Heatmap showing the
   different gene expression profiles in the different groups. (C) Volcano
   plot showing DEGs (P-adjusted < 0.05) between groups. (D) GO enrichment
   analysis of DEGs (FC > 1.5, top 10) in response to GLN^− conditions.
   (E) Schematic illustrating the ESCRT machinery and heatmap showing
   ESCRT gene expression in the different groups. (F to H) Representative
   images and quantification of TSG101^+ early endosomes, CD63^+ MVBs, and
   Lamp2b^+ late endosomes in M1-Mφs under GLN^− conditions (scale bar, 50
   μm). (I) qPCR analysis of TSG101 and VPS4B gene expression in M1-Mφ
   under GLN^− conditions cotreated with GW4869 for 24 hours. (J) EV yield
   determined by the EV/cell protein ratio after GW4869 treatment. (K)
   qPCR analysis of cytokine gene expression (e.g., IL-6 and IL-1β) in
   EV^GLN− after GW4869 treatment (n = 3; **P < 0.01, *P < 0.05 versus the
   GLN^− group).

   To assess the changes in EV biogenesis processes under GLN^−
   conditions, we assessed endosome components during EV biogenesis in Mφs
   using immunofluorescence (IF) staining as previously reported
   ([148]32). The results showed that, compared with those from the GLN^+
   condition, the GLN^− condition increased the number of TSG101^+ early
   endosomes, the number of CD63^+ multivesicular bodies (MVBs), and the
   number of Lamp2b^+ late endosomes formed in Mφs ([149]Fig. 5, F to H),
   indicating an overall enhancement of EV biogenesis under GLN^−
   conditions. Additionally, M1-Mφs under GLN^− conditions were treated
   with GW4869, a neutral sphingomyelinase inhibitor known to block
   exosome generation by inhibiting ceramide-mediated pathways ([150]34).
   GW4869 treatment reduced the expression of several ESCRT-related genes
   (e.g., TSG101 and VPS4B) in M1-Mφs and increased the expression of
   cytokines (e.g., IL-6 and IL-1β) in M1-EVs. In addition, the GLN^−
   group had a slightly decreased EV yield compared with that of the GLN^−
   group ([151]Fig. 5, I to K). It has been proposed that the generation
   of different EV subtypes may involve diverse intracellular compartments
   and distinct EV biogenesis routes, and the ESCRT machinery is involved
   in both exosome and ecotosome biogenesis ([152]35). Moreover, studies
   have reported that GW4869 can reduce exosome secretion but increase
   ecotosome secretion ([153]35). Therefore, these results suggest that
   the enhanced EV biogenesis and cargo (e.g., cytokines) sorting in Mφs
   under GLN^− conditions partially rely on the ESCRT pathway. However,
   its impact on different EV subtypes needs further investigation in
   future studies.

   Having shown that GLN^− conditions increase the yield of EVs and
   facilitate the production of diverse immunostimulatory molecules, as
   well as their sorting into EVs, KEGG (Kyoto Encyclopedia of Genes and
   Genomes) pathway enrichment analysis was also conducted to elucidate
   the underlying mechanisms involved. The results showed that many DEGs
   of M1-Mφ under GLN^− conditions were also enriched in pathways related
   to cytokine production and immune responses, such as the Janus kinase
   (JAK)–signal transducer and activator of transcription (STAT) pathway
   (e.g., CREB3L4, IL6, IL11, and LTA), the NF-κB pathway (e.g., IL1R1,
   TNFRSF9, JAK2, PTEGS2, IL27, and AHR), TNF signaling pathway (e.g.,
   GBP3, IL1β, TNFSF1B, and IL33), cytokine–cytokine receptor interaction
   (e.g., GDF9, MMP13, IL13RA2, GBP7, and EDN1), and T[H]17 cell
   differentiation (e.g., IL-1α, LCN2, and FOS) ([154]Fig. 6, A to C).
   KEGG analysis also revealed that the DEGs in M0-Mφs under GLN^−
   conditions were enriched in pathways related to immune responses, such
   as cytokine–cytokine receptor interaction, the TNF signaling pathway,
   the JAK-STAT pathway, and the NF-κB pathway (fig. S10F). Several of
   these pathways (e.g., STAT and NF-κB) have been shown to play crucial
   roles in mediating various immune processes, such as cytokine and
   chemokine synthesis, inflammasome activation, and immune cell
   recruitment (e.g., monocytes, Mφs, and T cells) ([155]36, [156]37). For
   example, in endothelial cells treated with monocyte-derived EV
   treatment, the NF-κB, Toll-like receptor 4 (TLR4), and myeloid
   differentiation primary response gene 88 (MyD88) pathways were
   activated, leading to up-regulated expression of adhesion molecules
   (e.g., ICAM1 and VCAM1) and cytokines (e.g., IL1β and IL6) ([157]38).
   Further, quantitative real-time polymerase chain reaction (qPCR) assays
   confirmed the expression of genes associated with the STAT and NF-κB
   pathways (e.g., Stat4, Stat6, IL-12B, and Nlrp3) in the GLN^− group
   compared to that in the GLN^+ group ([158]Fig. 6D). Moreover, M1-Mφ in
   the GLN^− group showed higher protein expression levels of p-NF-κB and
   NLRP3 than those in the GLN^+ group ([159]Fig. 6E). Similarly, compared
   with those in the GLN^+ group, M1-Mφ in the GLN^− group released higher
   levels of immunostimulatory cytokines (e.g., TNF-α and IL-6) and
   chemokines (e.g., CCL2) and lower levels of immunosuppressive cytokines
   (e.g., IL-10) ([160]Fig. 6F). To verify the vital role of the
   identified cytokine synthesis pathway (e.g., NF-κB) in this process,
   M1-Mφs under GLN^− conditions were cotreated with
   dehydroxymethylepoxyquinomicin (DHMEQ; an NF-κB pathway inhibitor)
   ([161]39), and the changes in cytokine expression in parent cells and
   EVs were assayed. As shown in [162]Fig. 6 (G and H), compared with
   those in the GLN^− group, the levels of immunostimulatory cytokines
   (e.g., IL-6 and IL-1β) in M1-Mφ and cytokine mRNA (e.g., IL-6 and
   IL-1β) in M1-EVs in the GLN^− group were reduced by NF-κB inhibitor
   treatment, suggesting that the GLN^− state can induce cytokine
   synthesis and sorting process.

Fig. 6. Effect of GLN^− conditions on immunoregulatory cargo synthesis in
Mφs.

   [163]Fig. 6.
   [164]Open in a new tab

   (A) KEGG enrichment analysis of DEGs (FC > 1.5) in M1-Mφs under GLN^−
   conditions for 24 hours. (B) Heatmap showing the differential
   expression levels of immune-related genes in the different groups. (C)
   Interaction network analysis of DEGs in response to GLN^− conditions.
   (D) qPCR analysis of Stat4, Stat6, IL-12B, and Nlrp3 gene expression in
   Mφs from the different groups. (E) Western blot analysis and
   quantification of inflammatory pathway proteins (NF-κB and NLRP3) in
   Mφs from the different groups (n = 3). (F) Luminex assay measuring
   cytokine (TNF-α, IL-6, and IL-10) and chemokine (CCL2) levels in
   culture medium from M1-Mφs cultured under GLN^+ or GLN^− conditions for
   24 hours. (G) qPCR analysis of cytokine gene expression (IL-1β and
   IL-6) in M1-Mφs under GLN^− conditions cotreated with DHMEQ for 24
   hours. (H) qPCR analysis of cytokine gene expression (e.g., IL-6 and
   IL-1β) in EV^GLN− after DHMEQ treatment. (I and J) Representative
   images and quantification of the colocalization of cytokines (IL-1β)
   with HSP70^+ or TSG101^+ endosomes in M1-Mφs under GLN^− conditions
   (scale bar, 50 μm) (n = 3; ***P < 0.001, *P < 0.05 versus the GLN^+
   group).

   Recent reports have highlighted the involvement of both active and
   passive mechanisms in the EV cargo sorting process. The active sorting
   process relies on the EV biogenesis machinery (e.g., ESCRT), which
   captures and clusters cargos, while passive sorting occurs based on the
   cellular levels of cargos and the rate of EV production ([165]18). To
   determine the role of sorting mechanisms under GLN^− conditions, we
   detected the colocalization of EV biogenesis markers and cytokines in
   M1-Mφs. As shown in [166]Fig. 6 (I and J), we observed increased
   colocalization of cytokines (IL-1β, green) and EV biogenesis and
   trafficking markers (TSG101 and HSP70, red) in cells from the GLN^−
   group compared to those from the GLN^+ group. This finding suggested
   that the enrichment of immunostimulatory cytokines in EV^GLN− may be
   partly due to the enhanced activity of the sorting machinery and
   increased cytokine production in these Mφs. In addition to cargo
   sorting, the ESCRT machinery can also affect the rate of EV biogenesis,
   and thus, the passive sorting process might also contribute to the
   enrichment of cytokines in EV^GLN− ([167]32). Together, these results
   indicate that GLN^− conditions promote EV secretion and the synthesis
   and sorting of immunostimulatory cargos into EVs in M1-Mφs, and these
   effects are mediated by the up-regulation of pathways involved in EV
   biogenesis (e.g., ESCRT) and cytokine production (e.g., STAT and
   NF-κB).

GLN^− conditions improved EV yield and function through metabolic rewiring

   To further explore the detailed metabolic alterations contributing to
   EV secretion in the GLN^− state, the changes in the metabolic profiles
   of M1-Mφ were analyzed using LC-MS/MS–based targeted metabolomics
   (table S4). PCA and heatmap analysis revealed distinct metabolite
   profiles between the GLN^+ group and the GLN^− group ([168]Fig. 7, A
   and B), suggesting that GLN^− induces systemic metabolic rearrangement
   in Mφs. Different metabolites (12 up-regulated and 19 down-regulated)
   between the GLN^+ group and the GLN^− group were identified ([169]Fig.
   7C). Further KEGG enrichment analysis revealed that the differentially
   expressed metabolites were mainly involved in pathways related to
   cellular energy and redox metabolism, such as the TCA cycle, GSH
   [glutathione (reduced form)] metabolism, glycolysis, and glycine
   metabolism ([170]Fig. 7D). The GLN^− group exhibited lower levels of
   metabolites related to glutaminolysis (e.g., αKG, glutamate, and GSH)
   (fig. S11A), as well as the TCA cycle (e.g., fumarate, succinate, and
   citrate) (fig. S11B), and higher levels of glycolysis and pentose
   phosphate pathway (PPP) (e.g., G-6-P and F-6-P) (fig. S11C) and
   nucleotide metabolism [e.g., uridine diphosphate (UDP) and guanosine
   monophosphate (GMP)] (fig. S11D) than the GLN^+ group ([171]Fig. 7E and
   fig. S5E). GLN is an essential precursor of intracellular antioxidants,
   such as glutathione (GSH) and NADPH [reduced form of nicotinamide
   adenine dinucleotide phosphate (NADP^+)], while GLN^− can lead to
   impaired antioxidant capacity and elevated levels of reactive oxygen
   species (ROS) ([172]9). GLN can be converted to αKG through
   glutaminolysis and enter the TCA cycle and glutathione (GSH) synthesis
   to mitigate ROS ([173]20). Similarly, we observed that M1-Mφs in the
   GLN^− group exhibited a lower glutathione/oxidized glutathione
   (GSH/GSSG) ratio and nicotinamide adenine dinucleotide/reduced NAD^+
   (NAD^+/NADH) ratio and a greater reduction in the nicotinamide adenine
   dinucleotide phosphate/NADP^+ (NADPH/NADP^+) ratio than those in the
   GLN^− group ([174]Fig. 7F). NAD^+, NADPH, and GSH are essential
   cofactors for maintaining cellular energy metabolism and redox
   homeostasis. For example, NAD^+ can be reduced to NADH to regulate
   mitochondrial energy metabolism and redox homeostasis ([175]40). GSH
   can be oxidized to GSSG and act as a key scavenger of intracellular
   ROS. NADPH, an electron donor, is indispensable for converting GSSG to
   GSH ([176]40). Unlike GSH, NADPH is mainly generated from GLC via the
   PPP ([177]41), suggesting that the increase in glycolysis and the
   increase in the PPP in M1-Mφ under GLN^− conditions are likely adaptive
   effects that aim to maintain cellular energy and redox balance. These
   results collectively suggest that GLN^− conditions may induce systemic
   metabolic rewiring in M1-Mφs, leading to oxidative metabolism defects,
   redox imbalance, and enhanced glycolysis/PPP pathway activity (fig.
   S12).

Fig. 7. Systemic metabolic rewiring of Mφs under GLN^− conditions.

   [178]Fig. 7.
   [179]Open in a new tab

   (A) Cells were cultured in medium supplemented with GLN (2 mM, GLN^+)
   or without GLN (GLN^−) for 24 hours, and a PCA scatterplot was
   generated based on the metabolomic data of the different groups (n =
   5). (B) Heatmap showing different metabolites in different groups (n =
   5). (C) Volcano plots illustrating differentially abundant metabolites
   (P-adjusted < 0.05) between groups (n = 5). (D) Metabolic pathway
   enrichment of different metabolites under GLN conditions. (E) Heatmap
   showing detailed differentially abundant metabolites between groups (n
   = 5). (F) Measurements of cellular GSH/GSSG ratio (left), NADPH/NADP^+
   ratio (middle), and NAD^+/NADH (right) ratio in the different groups (n
   = 5). (G) Pie charts depicting the number (top) and sublocalization
   (bottom) of mitochondrial DEGs (mito-DEGs) revealed by RNA-seq analysis
   between the GLN^+ and GLN^− groups (n = 3). (H) KEGG enrichment
   analysis of mito-DEGs and the top 10 affected pathways (n = 3). (I)
   Heatmap showing mito-DEGs (FC > 2 and P-adjusted < 0.05) between groups
   (n = 3). (J) Measurements of mitochondrial ROS, mitochondrial membrane
   potential, and ATP levels in M1-Mφs under GLN^+ conditions or GLN^−
   conditions for 24 hours (**P < 0.01, *P < 0.05, ^NSP > 0.05 versus the
   GLN^+ group).

   Mitochondria play a central role in cellular energy production
   [adenosine triphosphate (ATP)] and redox metabolism and are also the
   primary source of intracellular ROS (~90%, mtROS) ([180]42). To better
   understand the impact of GLN depletion on mitochondrial function, we
   analyzed the changes in mitochondrial genes using MitoCarta2.0
   ([181]43). The results showed that 6.28% (543 of 8645) of the DEGs
   whose expression was induced by GLN^− conditions were mitochondrial
   genes, 48.25% of which were related to the mitochondrial matrix and
   30.57% to the mitochondrial inner membrane ([182]Fig. 7G). In line with
   the metabolomic results, GLN^− conditions affected multiple metabolic
   processes in M1-Mφ, such as metabolic pathways (e.g., SDHD, ACAT1,
   IDH2, COX7B, and GLUD1), ROS (e.g., ATP5, SDHC, COX7B, and SOD1),
   oxidative phosphorylation (e.g., ATP5G1, SDHD, SDHC, COX7B, and COX10),
   and the TCA cycle (e.g., PDHA1, CS, IDH3A, PCK2, and FH1) ([183]Fig. 7,
   H and I). Perturbations in energy and redox metabolism have been
   associated with cytokine production and EV secretion in both immune and
   nonimmune cells ([184]7), and oxidative stress can affect the
   production of EVs by redirecting their destinations (e.g., the
   lysosomal pathway). For instance, mitochondrial respiratory defects and
   mtROS overproduction under hyperglycemic conditions induce NF-κB/p38
   MAPK pathway and cytokine (e.g., TNF-α and IL-6) production in Mφs
   ([185]41). Likewise, mitochondrial stress caused by mitochondrial Lon
   overexpression promoted the secretion of EVs harboring mtDNA in cancer
   cells ([186]44). Consistently, M1-Mφs in the GLN^− group exhibited
   impaired mitochondrial oxidative metabolism, as indicated by increased
   mtROS and decreased mitochondrial membrane potential and ATP compared
   to those in the GLN^+ group ([187]Fig. 7J). These results suggest that
   the up-regulated pathways involved in EV biogenesis and cytokine
   production in M1-Mφ under GLN^− conditions may be linked to the
   rewiring of mitochondrial energy and redox metabolism.

   Next, we assessed whether restoring GLN metabolism through anaplerosis
   could partially reverse the increase in EV secretion and cytokine
   production in M1-Mφs ([188]Fig. 8, A and E). αKG is a metabolite of
   glutaminolysis and a vital intermediate of the TCA cycle, and
   N-acetylcysteine (NAC) is a vital precursor of GSH that can mitigate
   excessive intracellular ROS ([189]20, [190]45). Thus, M1-Mφs under
   GLN^− conditions were supplemented with αKG or NAC. The expression
   levels of immunostimulatory molecules (e.g., IL-1β, IL-6, and IL-12)
   were lower in the αKG supplementation group, compared to those in the
   GLN^− alone group, but the EV yield was not affected in M1-Mφs under
   GLN^− conditions ([191]Fig. 8, B to D). A possible explanation is that
   GLN regulates cellular functions via diverse metabolic pathways, and
   αKG may mainly function in certain aspects (e.g., energy metabolism),
   which could partially affect EV production (e.g., composition) in Mφs.
   On the other hand, NAC supplementation reduced the expression levels of
   immunostimulatory cytokines (e.g., IL-1β, IL-6, and IL-12B) and the
   yield of EVs in M1-Mφs under GLN^− conditions ([192]Fig. 8, E to G).
   Neither αKG nor NAC supplementation altered the EV size distribution of
   M1-Mφs ([193]Fig. 8, C and F). However, neither αKG nor NAC
   supplementation changed the EV production of M1-Mφs under GLN^+
   conditions, and cellular cytokine (IL-1β, IL-6) synthesis was not
   affected by NAC treatment (fig. S13). Notably, αKG treatment reduced
   cytokine synthesis in M1-Mφs under GLN^+ conditions (fig. S13),
   supporting the findings of previous reports, which indicated that
   excessive αKG can promote M2-Mφ activation ([194]10). These results
   confirmed that GLN^− conditions induced cellular metabolism rewiring
   and play vital roles in both promoting both EV secretion and
   stimulating the sorting of immunostimulatory cytokines into M1-EVs.
   GLN^− conditions induce mitochondrial stress (defects in the TCA cycle
   and GSH metabolism) accompanied by increased utilization of GLC and the
   PPP pathway in M1-Mφs (fig. S14), which subsequently up-regulates
   pathways related to cytokine production (e.g., NF-κB and STAT), EV
   biogenesis, and cargo sorting (e.g., ESCRT), ultimately resulting in an
   increase in the release of EVs containing abundant immunoregulatory
   contents from such cells.

Fig. 8. Effect of GLN metabolism anaplerosis on EV secretion and cytokine
synthesis in Mφs.

   [195]Fig. 8.
   [196]Open in a new tab

   (A and B) Schematic illustrating the αKG supplementation experiment.
   M1-Mφs were cultured in GLN medium supplemented with or without αKG
   (0.8 mM) for 24 hours, after which the gene expression levels of
   immunostimulatory molecules (IL-1β, IL-6, and IL-12B) were analyzed via
   qPCR analysis (n = 3). (C) EV sizes in the different groups (n = 3).
   (D) EV yields determined by the EV number/cell protein ratio (left) or
   the EV protein/cell protein ratio (right) in the different groups (n =
   3). (E and F) Schematic illustrating the NAC treatment experiment.
   M1-Mφs were cultured in GLN medium supplemented with or without NAC
   (0.5 μM) for 24 hours, and the gene expression of immunostimulatory
   molecules (IL-1β, IL-6, and IL-12B) was analyzed using qPCR analysis (n
   = 3). (G) EV sizes in the different groups (n = 5). (H) EV yields
   determined by EV number/cell protein ratio (left) or EV protein/cell
   protein ratio (right) in the different groups (n = 5) (**P < 0.01, *P <
   0.05, ^NSP > 0.05 versus the “GLN^−LPS^+” group).

   Although our results have elucidated the role of nutrient availability
   in regulating EV secretion, composition, and function, several
   questions remain elusive and need to be addressed in future studies.
   For example, we only investigated whether GLN^− conditions increased
   functional EV secretion in M1-Mφs, but whether modulating GLN
   metabolism could promote EV secretion in M2-Mφs has not been
   determined. It is unclear whether the increased EV secretion in M2-Mφs
   is dependent on metabolic alterations and signaling pathways similar to
   those observed in M1-Mφs. Mφs with different phenotypes exhibit diverse
   metabolic signatures and nutrient demands ([197]3); thus, it is
   necessary to evaluate the impacts of specific nutrient changes on EV
   secretion and function during M2-EV production. The metabolic states of
   immune cells are diverse, and various immune cells may respond
   differently to the same nutrient conditions ([198]19). Thus, the
   optimal conditions for EV production in Mφs may not be applicable to
   other immune cell types, necessitating careful assessment of the
   nutrient requirements for each specific immune cell. Additionally,
   while certain methods of engineered EV production (e.g.,
   preconditioning and genetic modification) have been shown to increase
   EV secretion or cargo levels ([199]46, [200]47), the potential
   enhancement of these effects through optimizing nutrient conditions has
   not been explored. Thus, further investigations are needed to develop
   advanced EV-based immunotherapeutics by integrating conventional
   approaches and metabolic (nutrient) modulation. Nevertheless, our study
   highlights the significance of nutrient availability in regulating
   functional EV secretion. Therefore, fine-tuning nutrient conditions
   should be carefully considered in future studies concerning therapeutic
   EV production for diverse disease treatments.

   In summary, the nutrient status of immune cells (e.g., Mφs) can
   regulate both the secretion and function of EVs from immune cells.
   Precisely modulating the metabolic states of donor cells is an
   effective strategy for enhancing the production of EVs with valuable
   bioactivities ([201]Fig. 9). Changes in the availability of an
   individual nutrient component (e.g., GLN depletion) in the medium are
   sufficient to induce systemic metabolic reprogramming and activate
   pathways related to EV biogenesis and cargo synthesis in donor cells.
   As a result, the yield, functional composition, and immunoregulatory
   potency of the produced EVs improved. These findings emphasize the
   importance of reconsidering the optimal metabolic state when generating
   therapeutic EVs. This topic has been overlooked in many previous
   studies. Furthermore, as indispensable substrates for cell growth,
   nutrients can be readily integrated with many other techniques to
   synergistically enhance the production of functional EVs. This study
   highlights the potential of fine-tuning nutrient conditions as a
   promising strategy for tailored EV engineering and provides valuable
   insights into the development of advanced EV-based therapeutics for
   various diseases.

Fig. 9. Nutrient availability can regulate EV secretion and function in
immune cells.

   [202]Fig. 9.
   [203]Open in a new tab

   Fine-tuning nutrient availability–induced metabolic rewiring is a basic
   and promising strategy for bioengineering EVs as advanced therapeutics
   for the treatment of diverse diseases. (A) GLN^− depletion efficiently
   promotes EV biogenesis and cargo packaging via systemic metabolic
   rewiring of Mφs. (B) EVs derived from Mφs under GLN^− conditions
   (EV^GLN−) exhibit favorable immunostimulatory potential.

MATERIALS AND METHODS

Cell culture

   Mouse Mφ RAW 264.7 cells and the human monocytic THP-1 cells were
   cultured in control RPMI 1640 (31800105, Gibco, La Quinta, CA, USA)
   supplemented with 10% heat-inactivated FBS (Gibco), penicillin (50
   U/ml), and streptomycin (50 U/ml). M0-Mφs were stimulated with LPS (40
   ng/ml, Sigma-Aldrich, St. Louis, MO, USA) for M1 activation and with
   mouse recombinant IL-4 and TGF-β (20 ng/ml each, Novoprotein, Shanghai,
   China) for M2 activation. To block glutaminolysis, M1-Mφs were treated
   with BPTES (10 nM, CSN Pharm, Coquitlam, BC, Canada) for 24 hours. For
   GLN metabolism anaplerosis, M1-Mφs cultured in GLN^+ or GLN^− medium
   were supplemented with dimethyl αKG (0.8 mM, Sigma-Aldrich) or NAC (0.5
   μM, Beyotime, Beijing, China) for 24 hours.

Isolation of cell-secreted EVs

   EV-depleted FBS was prepared using an ultracentrifugation (120,000g, 16
   hours, 4°C) method as previously reported ([204]48). M0-Mφs or M1-Mφs
   were cultured in RPMI 1640 supplemented with EV-depleted FBS and
   subjected to the indicated nutrient conditions for 24 hours. Then, the
   culture medium was collected and subjected to EV isolation via a
   differential centrifugation method ([205]16). In brief, the medium was
   centrifuged at 300g for 10 min and at 2000g for 30 min at 4°C to remove
   cells and other debris, followed by ultracentrifugation at 110,000g for
   90 min at 4°C in an SW32Ti rotor (Beckman Coulter, Brea, CA, USA) to
   obtain EVs. After washing with PBS, the EV pellets were recentrifuged
   at 110,000g for 90 min at 4°C. The purified EVs were resuspended in PBS
   and stored at −80°C until further use.

Transmission electron microscopy

   The morphology of the EVs was evaluated using TEM (HT7800, Hitachi
   Ltd., Tokyo, Japan) as previously described ([206]3). EV samples were
   diluted with PBS and dropped on a carbon-coated copper grid. To remove
   unattached EVs, the grid was washed with pure water. After staining
   with 2% phosphotungstic acid, the grids were air-dried and subjected to
   TEM.

Nanoparticle tracking analysis

   Nanoparticle tracking analysis (NTA) of the EV samples was carried out
   using a Zeta View PMX 120 (Particle Metrix, Meerbusch, Germany) as
   previously described ([207]3). Briefly, EVs were diluted (v/v, 1:2000)
   with pure water, and their size distributions and particle
   concentrations were analyzed using the following parameters (minimum
   area 5, maximum area 1000, and minimum brightness 20) at 25°C.

EV quantification

   The protein concentrations of the EV samples were quantified using
   Pierce BCA Protein Assay Kits (Thermo Fisher Scientific, Carlsbad, CA,
   USA). The EV yield of each group was assessed using the ratio of EV
   protein or particles relative to total protein in the donor cell as
   previously reported ([208]49).

Western blotting

   Protein extraction from the samples was carried out using
   radioimmunoprecipitation assay (RIPA) buffer containing protease and
   phosphatase inhibitors (Calbiochem, San Diego, CA, USA). The protein
   concentration of samples was quantified using a BCA kit. Protein
   electrophoresis was performed using 10% SDS–polyacrylamide gel
   electrophoresis (SDS-PAGE), followed by protein transfer from the gel
   to a nitrocellulose membrane (NC; Cytiva, Germany). The membranes were
   blocked in 5% nonfat milk and incubated with mouse anti-ALIX
   (12422-1-AP, Proteintech, Wuhan, China), mouse anti-HSP70 (A0284,
   ABclonal, Wuhan, China), mouse anti-TSG101 (14497-1-AP, Proteintech),
   mouse anti-GM130 (A11408, ABClonal), mouse anti–pNF-κB (WL02169,
   Wanleibio, China), mouse anti–NF-κB (10745-1-AP, Proteintech), mouse
   anti-NLRP3 (15101, Cell Signaling Technology), and mouse anti–β-actin
   (AC026, ABclonal) at 4°C overnight. The membranes were washed and
   incubated with a secondary antibody conjugated with horseradish
   peroxidase (HRP) at 37°C for 1 hour. Protein signals were detected with
   a chemiluminescence kit (MedChemExpress, Shanghai, China). Protein
   quantification was carried out using ImageJ software.

Modification of cellular nutrient conditions

   Mφs were cultured in control RPMI 1640 supplemented with 10% FBS, 11 mM
   GLC, and 2 mM GLN or in modified medium supplemented with altered
   nutrients, including FBS, GLC, AAs, and FAs. In brief, a stock solution
   of each nutrient was prepared and subsequently added to cells, after
   which the final concentration of each nutrient was considered lower
   (2.5%, v/v) FBS or higher (20%, v/v) FBS, higher concentrations of GLC
   (20 mM, Sigma-Aldrich), higher concentrations of GLN (8 mM, GlutaMAX,
   Gibco), BCAAs (5 mM), or OA (100 μM, Aladdin, Shanghai, China)
   dissolved in FA-free bovine serum albumin (BSA; Solarbio, Beijing,
   China). The BCAA stock solution was prepared as a mixture of leucine,
   isoleucine, and valine at 100 mM (Sigma-Aldrich), and BSA alone was
   used as a control in the OA supplementation experiments. In
   concentration course tests of GLC or GLN, cells were cultured in
   GLC-free RPMI (11879020, Gibco) supplemented with GLC (0, 2, and 10 mM;
   A2494001, Gibco) or GLN-free RPMI supplemented with GLN (0, 2, and 10
   mM) for 24 hours. In the case of GLN-depleted conditions, cells were
   cultured in GLN-free RPMI 1640 (21870076, Gibco) supplemented with
   (GLN^+, 2 mM) or without GLN for 24 hours (GLN^−). After incubation for
   the indicated times, the cells or conditioned medium from each group
   was collected for further tests.

LC-MS/MS–based proteomic analysis of EVs

   Total EV proteins were extracted from the EV samples and quantified
   using a BCA kit (Thermo Fisher Scientific). Protein samples (100 μg)
   were taken for reductive alkylation and digested with trypsin overnight
   at 37°C. The peptides were subjected to desalting and vacuum-dried, and
   the concentrations were determined by a peptide quantification kit
   (Thermo Fisher Scientific). Then, 30 μg of trypsin-digested peptides
   was analyzed by an EASY nLC-1200 system (Thermo Fisher Scientific, USA)
   coupled with a timsTOF Pro2 (Bruker, Germany) mass spectrometer at
   Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). Briefly, the
   tryptic peptides were separated by an ultrahigh-performance liquid
   phase system subjected to a capillary ion source and then analyzed by
   timsTOF Pro2 at a flow rate of 250 nl/min (the electrospray voltage was
   1.5 kV). The peptide parent ions and their secondary fragments were
   detected and analyzed using high-resolution time of flight. MS/MS
   spectra were searched using MaxQuant version 2.0.3.1 software.
   Bioinformatic analysis of the proteomic data was performed with a free
   online platform ([209]https://www.omicsolution.org/wkomics/main/). DEPs
   with fold change (FC) > 1.5 and P-adjusted < 0.05 were identified using
   a free online platform
   ([210]https://www.omicsolution.org/wkomics/main/) with SAMR analysis
   ([211]50). PCA plots, volcano plots, and heatmaps were generated using
   a free online platform
   ([212]https://www.omicsolution.org/wkomics/main/). GO analysis of the
   DEPs was performed using STRING ([213]https://cn.string-db.org/) and
   the Metascape tool (v3.5) ([214]51), and false discovery rate (FDR) <
   0.05 was considered to indicate statistical significance.

Quantitative real-time polymerase chain reaction

   Total RNA was extracted from cells using TRIzol reagent (Gibco), and
   cDNA was synthesized with a commercial kit (Vazyme, Nanjing, China).
   qPCR was performed with SYBR Green (Vazyme) on a CFX96 real-time PCR
   detection system (Bio-Rad, Hercules, CA, USA). The primers used in this
   study are listed in table S5. The data were analyzed by Bio-Rad CFX
   Manager software, and the relative FC in mRNA levels was determined
   using the ∆∆C[t] method with GAPDH or β-actin as the internal reference
   gene.

Impact of EV preparations on M2-Mφs

   M2-Mφs were induced by stimulating RAW 264.7 cells with mouse
   recombinant IL-4 and TGF-β (20 ng/ml each, Novoprotein, Shanghai,
   China) for 48 hours. M2-Mφs were treated with PBS, EV^GLN− preparations
   (20 μg/ml), or EV^GLN+ preparations (20 μg/ml) for 48 hours, after
   which the expression of M2 genes (Arg1, Mrc1, and TGF-β) in the cells
   was detected via a qPCR assay.

Impact of EV preparations on chemotaxis

   THP-1 monocytes were treated with PBS, EV^GLN− preparations (20 μg/ml),
   or EV^GLN+ preparations (20 μg/ml) for 24 hours, after which the gene
   expression of chemokines (Cxcl2 and Ccl2) in the cells was assayed via
   qPCR. The supernatants of each group were collected and concentrated
   using ultrafiltration tubes (3-kDa molecular weight cutoff,
   Sigma-Aldrich). Mouse splenocytes (mainly consisting of monocytes and
   lymphocytes) were isolated from 8-week-old male C57BL/6 mouse spleens
   using density gradient medium (Ficoll-Paque, Cytiva, Marlborough, USA)
   according to the manufacturer’s protocols. In a Transwell system (5-μm
   pore size, Labselect, Hefei, China), mouse mononuclear cells and
   concentrated medium (with equal protein amounts) were added to the
   upper and lower chambers, respectively. After 4 hours of incubation,
   the migrated cells in the lower chamber were collected and counted
   using flow cytometry (Beckman Coulter). To evaluate the direct
   chemotaxis effect of EVs, mouse splenocytes were cultured with PBS,
   EV^GLN+ preparations (10 μg/ml), or EV^GLN− preparations (10 μg/ml) in
   the upper chambers for 4 hours, and the migrated cells in the lower
   chamber were collected and counted using a flow cytometer (Cytoflex,
   Beckman Coulter).

Impact of EV preparations on T cells

   Mouse mononuclear cells were treated with ConA (2 μM, Thermo Fisher
   Scientific), ConA plus EV^GLN− preparations (10 μg/ml), or EV^GLN+
   preparations (10 μg/ml) for 72 hours (n = 3). After treatment, cells
   from each group were collected and stained with fluorescein
   isothiocyanate (FITC)–conjugated anti-mouse CD3e (553061, BD, Brea, CA,
   USA, USA), peridinin chlorophyll protein (PerCP)/Cyanine5.5-conjugated
   anti-mouse CD4 (100434, BioLegend), phycoerythrin (PE)–Cy7–conjugated
   anti-mouse CD8a (552887, BD), and allophycocyanin (APC)–conjugated
   anti-mouse CD69 (APC-65105, Proteintech, Wuhan, China) for 30 min.
   After washing with PBS, the stained cells were analyzed using a flow
   cytometer (LSRFortessa, BD Biosciences).

Immune responses to EV preparations in vivo

   All animal experiments were approved by the Animal Care and Use
   Committee of West China Hospital, Sichuan University (permit no.
   20230330008) and were performed according to the guidelines of the
   National Institutes of Health (NIH). BALB/c mice (male, 8 weeks) were
   purchased from Byrness Weil Biotech Ltd. (Chengdu, China) and housed in
   a pathogen-free facility. Normal mice were randomly divided into three
   groups (n = 5) and intravenously injected with 100 μl of PBS, EV^GLN−
   preparations (30 μg each mouse), or EV^GLN+ preparations (30 μg each
   mouse). At 4 hours after injection, the mice were anaesthetized, and
   the spleens were collected and prepared as single-cell suspensions to
   detect immune cells. In brief, mouse spleen tissues were ground in PBS
   and then passed through a 70-μm filter. The red blood cells in spleen
   samples were removed using a Mouse Erythrocyte Lysing Kit (Solarbio,
   Beijing, China). The single-cell suspension was stained with FVS
   (564997, BD), APC-Cy7 rat anti-mouse CD45 (557659, BD Pharmingen), PE
   rat anti-mouse F4/80 (565410, BD), APC-conjugated anti-mouse Ly6G
   (560599, BD), and BV421-conjugated anti-mouse Ly6C (562727, BD) for 30
   min. After washing with PBS, the stained cells were analyzed using a
   flow cytometer (LSRFortessa).

RNA-seq analysis of cells

   M1 or M0 Mφs cultured in GLN^+ medium or GLN^− medium for 24 hours were
   collected for RNA-seq analysis. In brief, total RNA was extracted from
   tissue samples using TRIzol, and genomic DNA was removed using
   deoxyribonuclease (DNase) I (Takara, Shiga, Japan). The RNA-seq
   transcriptome library was constructed using a TruSeq RNA Sample
   Preparation Kit (Illumina, San Diego, CA, USA), and the paired-end
   RNA-seq library was sequenced on an Illumina HiSeq Xten/NovaSeq 6000
   sequencer by Shanghai Majorbio Biopharm Technology Co. Ltd. (Shanghai,
   China). The expression levels of the transcripts were calculated using
   the transcripts per million reads (TPM) method. DEGs (FC > 1.5 and
   P-adjusted < 0.05) were considered significant. Venn plots, PCA plots,
   volcano plots, and heatmaps were generated using a free online platform
   ([215]https://www.omicsolution.org/wkomics/main/). GO and KEGG analyses
   of the DEGs were performed with DAVID Bioinformatics Resources
   ([216]https://david.ncifcrf.gov/home.jsp) and visualized using a free
   online platform ([217]https://www.omicsolution.org/wkomics/main/).

IF staining

   After different treatments, the cells were fixed (4% paraformaldehyde
   for 10 min), permeabilized (0.3% Triton X-100 for 10 min), and blocked
   (1% BSA for 1 hour). Then, the fixed cells were incubated overnight at
   4°C with mouse anti-TSG101 (14497-1-AP, Proteintech), mouse anti-CD63
   (25682-1-AP, Proteintech), mouse anti-Lamp2b (27823-1-AP, Proteintech),
   mouse HSP70 (A0284, ABclonal), and mouse anti–IL-1β (A11369, ABclonal),
   followed by incubation with donkey anti-rabbit immunoglobulin G (IgG)
   H&L (DyLight 550) (ab96892, Abcam), tetramethyl rhodamine
   isothiocyanate (TRITC)–conjugated secondary antibody (2294985, Thermo
   Fisher Scientific), or FITC-conjugated secondary antibody (abs20139,
   Absin, Shanghai, China) as needed at 37°C for 1 hour. DAPI
   (4′,6-diamidino-2-phenylindole) dyes were used to label the nuclei.
   After washing with PBS, the cells were observed under a fluorescence
   microscope (Zeiss, Axio Imager Z2, Oberkochen, Germany).

Luminex assay

   The concentrations of cytokines/chemokines, including TNF-α, CCL2,
   IL-6, and IL-10, in the culture medium were analyzed using a mouse
   magnetic Luminex assay (R&D Systems, Wiesbaden-Nordenstadt, Germany) as
   previously described ([218]48). Briefly, the diluted culture medium
   samples were incubated with a premixed cocktail of antibody-conjugated
   beads in a 96-well plate for 2 hours, followed by incubation with a
   cocktail of biotinylated antibodies for 1 hour and streptavidin-PE
   (SA-PE) for 30 min. The sample-loaded plate was washed with buffer, and
   the protein concentration was measured on a Luminex 200 analyzer
   (Luminex, Austin, TX, USA) according to the manufacturer’s
   instructions.

Pathway inhibition experiments

   M1-Mφs under GLN depletion conditions were treated with GW4869 (20 μM,
   HY-19363, MedChemExpress, Monmouth Junction, NJ, USA) or DHMEQ (2.5
   μg/ml, HY-14645, MedChemExpress) for 24 hours. After incubation for the
   indicated times, the cells or EVs from each group were collected for
   further tests.

LC-MS/MS–based targeted metabolomics

   The intracellular metabolites of Mφs were extracted using
   methanol/H[2]O solution as previously described ([219]52). After
   washing with PBS, the cells in the wells of a six-well plate were
   immediately extracted with buffer (1 ml, methanol/H[2]O = 8:2, v/v) and
   incubated at −20°C for 30 min. After centrifugation (13,000 rpm for 20
   min at 4°C), 0.7 ml of the supernatant was collected and dried in a
   drier (Eppendorf, Fisher Scientific, Pittsburgh, PA) at 30°C for 4
   hours. The dried samples were reconstituted in 100 μl of hydrophilic
   interaction chromatography (HILIC) solution. Targeted MS-based
   metabolomics was performed using an Agilent 1260 LC (Agilent
   Technologies, Santa Clara, CA) coupled with an AB Sciex Qtrap 5500 MS
   (AB Sciex, Toronto, Canada) system. The sample solution was injected
   into the LC-MS/MS system and analyzed under positive (2 μl of sample
   solution) and negative (10 μl of sample solution) ion modes. The mobile
   phase, gradient conditions, and MS parameters were set up as previously
   described ([220]52). Multiple reaction monitoring (MRM) mode was used
   to detect multiple metabolites. The metabolites were normalized to the
   cellular protein, and FC > 1.5 and P-adjusted < 0.05 were considered
   significant. PCA plots and heatmaps were generated using a free online
   platform ([221]https://www.omicsolution.org/wkomics/main/). The
   differentially expressed metabolites were analyzed and visualized using
   MetaboAnalyst (version 5.0,
   [222]https://www.metaboanalyst.ca/MetaboAnalyst/ModuleView.xhtml).

Mitochondrial membrane potential assay

   The mitochondrial membrane potential of the cell samples was analyzed
   using a fluorescent tetramethylrhodamine methyl ester probe (TMRM,
   Sigma-Aldrich). After different treatments, the cells were stained with
   20 nM TMRM for 15 min at 37°C. The stained cells were washed with PBS,
   and the fluorescence was measured using a flow cytometer (LSRFortessa).

Mitochondrial ROS measurement

   Intracellular levels of mitochondrial ROS were analyzed using the
   MitoSOX Red mitochondrial superoxide indicator (MitoSOX, Thermo Fisher
   Scientific). After different treatments, the cells were stained with 5
   μM MitoSOX for 20 min at 37°C. The stained cells were washed with PBS,
   and the fluorescence was measured using a flow cytometer (LSRFortessa).

Intracellular ATP measurement

   ATP was measured using a bioluminescence assay kit (Beyotime
   Biotechnology, Shanghai, China). After different treatments, the cells
   were lysed with 200 μl of lysis buffer provided with the kit. The
   supernatant was collected by centrifugation at 12,000g for 5 min at
   4°C. The concentration of ATP was determined by mixing 100 μl of
   luciferase reagent and 20 μl of supernatant to catalyze the production
   of luminescence from ATP and luciferin. The luminescence of each sample
   was measured using a microreader (Tecan Group Ltd., Switzerland).

Statistical analysis

   All the data are presented as the means ± SDs and were analyzed using
   GraphPad software (version 8.0.2, IBM Corporation, USA) with one-way
   analysis of variance (ANOVA) or Student’s t test; P < 0.05 was
   considered to indicate a significant difference. The data were obtained
   from at least three biological replicates, and “NS” represents the
   number of independent samples for each group.

Acknowledgments