ATP citrate lyase links increases in glycolysis to diminished release of vesicular suppressor of cytokine signaling 3 by alveolar macrophages

Abstract

Extracellular vesicles (EVs) are important vectors for intercellular communication. Lung-resident alveolar macrophages (AMs) tonically secrete EVs containing suppressor of cytokine signaling 3 (SOCS3), a cytosolic protein that promotes homeostasis in the distal lung via its actions in recipient neighboring epithelial cells. AMs are metabolically distinct and exhibit low levels of glycolysis at steady state. To our knowledge, whether cellular metabolism influences the packaging and release of an EV cargo molecule has never been explored in any cellular context. Here, we report that increases in glycolysis following in vitro exposure of AMs to the growth and activating factor granulocyte-macrophage colony-stimulating factor inhibit the release of vesicular SOCS3 by primary AMs. Glycolytically diminished SOCS3 secretion requires export of citrate from the mitochondria to the cytosol and its subsequent conversion to acetyl-CoA by ATP citrate lyase. Our data for the first time implicate perturbations in intracellular metabolites in the regulation of vesicular cargo packaging and secretion.

1. Introduction

Intercellular communication is central to the maintenance of tissue homeostasis. Cell-cell contact and the release of soluble factors have long been recognized as means of cellular crosstalk. More recently, extracellular vesicles (EVs) have become increasingly appreciated for their role as vehicles for the transfer of information amongst cells [1, 2]. EVs are small, membrane-delimited structures that encapsulate a diverse array of molecular cargoes (e.g., lipids, nucleic acids, and proteins) and which alter phenotype and function when internalized by recipient cells [3]. Historically, EVs have been classified into two major populations, exosomes (Exos) and microvesicles (MVs), distinguishable by mode of biogenesis and once thought to categorically differ in size and molecular composition. However, recent observations have revealed substantial overlap in the biochemical properties of Exos and MVs [4, 5], thus complicating their discrimination.

The lung is a unique milieu that is continuously exposed to exogenous insults including microbes, antigens, and toxins. The protective inflammatory responses directed towards these inhaled environmental stimuli must be appropriately restrained to ensure preserved physiologic function. Achieving this delicate balance of response vs. restraint requires crosstalk between the two predominant cellular constituents of the alveoli – 1) the alveolar epithelial cells (AECs) that comprise the respiratory surface and 2) the alveolar macrophages (AMs) that are the most numerous resident immune cells. Our laboratory has recently shown that AMs secrete EVs enriched with the anti-inflammatory cytosolic protein suppressor of cytokine signaling 3 (SOCS3), which are taken up by neighboring AECs [6]. This acquisition of AM-derived, SOCS3-containing EVs by AECs serves as a homeostatic mechanism for dampening pro-inflammatory Janus kinase (JAK)-signal transducer and activator of transcription 3 (STAT3) signaling responses within the lung in vitro and in vivo [6–9]. Although we have shown the release of vesicular SOCS3 by AMs to be dynamically tuned by various endogenous and exogenous cues [6–8, 10, 11], illuminating the cellular and molecular mechanisms that control vesicular SOCS3 secretion has remained elusive. More broadly, the cellular processes mediating the modulated packaging and release of vesicular cargo molecules of cytosolic origin are still poorly understood.

In addition to fulfilling the bioenergetic demands of the cell, metabolic processes are now recognized to control basic immune cell function. For example, metabolic rewiring is associated with functional changes that support the phenotypic characteristics of pro-inflammatory or alternatively activated macrophages, depending on the stimulus [12]. AMs, however, are metabolically distinct among macrophage populations in that they exhibit remarkably low levels of glycolysis at steady state [13–19], a metabolic phenotype programmed by the lung microenvironment [13] that supports their quiescent functional status [13, 16–18, 20, 21]. In cancer cells, changes in cellular metabolism have been shown to influence the subtype [22–24] or number [25] of EVs secreted. To our knowledge, however, the ability of metabolic rewiring to dynamically alter the packaging of an EV cargo molecule has never been studied.

Granulocyte-macrophage colony-stimulating factor (GM-CSF) is a pleiotropic cytokine that promotes the survival and/or differentiation/activation of myeloid cells [26], including in the lung where it shapes AM development [27, 28]. Additionally, GM-CSF is known to metabolically remodel macrophages through glycolytic activation [29, 30]. Here, using GM-CSF as a biologically relevant in vitro driver of metabolic reprogramming, we show that stimulation of glycolytic activity in AMs inhibits the packaging and release of vesicular SOCS3. Increases in AM glycolytic flux were linked to diminished SOCS3 release via the activity of ATP citrate lyase (ACLY), the cytosolic enzyme responsible for converting glucose-derived citrate to acetyl-CoA [31] and whose activity augments the inflammatory function of macrophages [32, 33]. Our studies implicate a critical role for mitochondrial metabolites in the regulation of vesicular cargo packaging, thereby significantly advancing a mechanistic understanding of metabolic control over EV content and biology.

2. Methods

2.1. Sources of primary AMs

Animals were maintained at the University of Michigan Unit for Laboratory Animal Medicine and experiments were conducted with approval by the University of Michigan Institutional Animal Care and Use Committee. All animal experiments complied with the ARRIVE guidelines and in accordance with the National Research Council’s Guide for the Care and Use of Laboratory Animals. Taking advantage of the ~20-fold greater yield of AMs collected by lung lavage of rats vs. mice, primary rat AMs were used for experiments requiring a large number of cells to yield a reliably robust signal (i.e., Western blot of conditioned media and lysates, nanoparticle tracking analysis, and isocitrate dehydrogenase activity measurements) whereas primary mouse AMs were used for experiments in which cell number was not a limiting factor (i.e., quantitative reverse transcription PCR and L-lactate assay).

2.2. Isolation and in vitro treatment of primary AMs for detection of secreted vesicular SOCS3

As previously described [34], primary AMs were collected from the lavage fluid of pathogen-free female Wistar rats (Charles River Laboratories, Wilmington, MA, USA). Isolated AMs (2 – 3 × 10⁶ per well) were plated in serum-free RPMI 1640 culture medium (Thermo Fisher Scientific, Waltham, MA, USA) in 6-well, polystyrene plates and adhered for at least 1 h. Cells were then washed prior to treatment to remove EVs and other secreted products generated during adherence to plastic. AMs were treated continuously with recombinant GM-CSF (50 ng/mL, BioLegend, San Diego, CA, USA) or the α-ketoglutarate analog dimethyloxalylglycine (DMOG, 1 mM, Sigma-Aldrich, St. Louis, MO, USA) in serum-free medium for 20 – 24 h. For glucose dose-response experiments, AMs were treated with 50 ng/mL GM-CSF in glucose-free RPMI 1640 culture medium supplemented with the specified concentrations of D-(+)-glucose (Sigma-Aldrich). AMs were also treated exogenously with sodium citrate or sodium acetate (both 1 mM and from Sigma-Aldrich) in serum-free medium for 20 – 24 h. Additionally, the following small-molecule inhibitors were used for in vitro studies: 2-deoxy-D-glucose (2-DG), UK-5099, citrate transport protein (CTP) inhibitor, SB-204990, BMS-303141 (all from Sigma-Aldrich), and acyl-CoA synthetase short chain family member 2 (ACSS2) inhibitor (Selleck Chemicals, Houston, TX, USA). Inhibitors were added to AMs for 20 – 24 h during continuous treatment with GM-CSF or sodium acetate.

2.3. EV isolation

Conditioned medium (CM) from rat AMs treated with GM-CSF or DMOG was collected after 20 – 24 h culture and centrifuged at 500 x g for 5 min and 2,500 x g for 12 min to remove dead cells, cell debris, and apoptotic bodies. EVs were then isolated using two approaches, as previously described [10]. Briefly, for rapid concentration of a total vesicular fraction, CM was centrifuged at 4,000 x g for 20 min in 100-kDa centrifugal filter units (MilliporeSigma, Burlington, MA, USA). The resulting >100-kDa concentrate was used for analysis of vesicular SOCS3 secretion. Alternatively, EVs were fractionated using ultracentrifugation by spinning CM at 17,000 x g for 160 min to pellet 17,000 x g EVs, from which the supernatant was then spun at 100,000 x g for 90 min to pellet 100,000 x g EVs. The resulting 17,000 x g and 100,000 x g EV fractions were used for analysis of SOCS3 and EV secretion.

2.4. Western blot

Given the diversity of proteins packaged into distinct vesicular fractions, no loading control for EVs is universally accepted [4]. Therefore, vesicular SOCS3 was evaluated by immunoblot analysis of EV fractions comprising entire >100-kDa, 17,000 x g, or 100,000 x g EV samples, which were collected from equal numbers of cultured rat AMs. Potential differences in the number of EVs secreted across conditions were then accounted for by quantifying EVs, as described below. This approach to standardizing Western blot of total EV fractions has been previously published [10]. For immunoblotting of cell lysates, protein concentrations were determined by DC protein assay (Bio-Rad, Hercules, CA, USA), and aliquots containing 10 μg were used to detect SOCS3. All samples were separated by SDS-PAGE using 12.5% gels to probe for SOCS3. Transfer of samples to nitrocellulose membranes was performed using Trans-Blot Mini Nitrocellulose Transfer Packs (Bio-Rad). Membranes were then blocked for 1 h with 4% BSA and incubated overnight with monoclonal antibodies recognizing SOCS3 (mouse, SO1), glucose transporter 1 (GLUT1) (rabbit, EPR3915), hexokinase 2 (HK2) (rabbit, EPR20839) (all from Abcam, Cambridge, UK) or α-tubulin (mouse, B-5–1-2) (Sigma-Aldrich). After washing and incubation (1 h) with peroxidase-conjugated anti-mouse or anti-rabbit antibodies, the film was developed using ECL detection for lysate samples or ECL Prime detection (both from GE Healthcare, Chicago, IL, USA) for vesicular SOCS3 samples. Immunoblotting with the SOCS3 (SO1) antibody routinely identifies a single band at the predicted molecular weight of 27 kDa [6], as demonstrated in Fig. 1c.

Figure 1: GM-CSF treatment of AMs inhibits SOCS3 packaging and release in EVs.

(a-c) Adherent AMs collected by lung lavage of rats were treated with GM-CSF (50 ng/mL) for 20 – 24 h. (a) EVs in CM were concentrated by 100-kDa centrifugal filtration and vesicular SOCS3 secretion was determined via Western blot of the total vesicular fraction (>100-kDa) samples. (b) Lysates were collected and subjected to Western blot for determination of SOCS3 intracellular content, with α-tubulin as a loading control. (c) EVs were isolated by sequential ultracentrifugation of CM at 17,000 x g and 100,000 x g, which were then probed for SOCS3 via Western blot (bottom left panel) and enumerated by NTA (bottom right panel). Data (mean ± SEM) are from ≥3 independent experiments, and significance was determined using paired sample t-test. Western blot images are cropped to encapsulate the band for SOCS3 visible at 27 kDa (a and b) or to demonstrate the presence of a single band at 27 kDa (c). Full-length Western blots are presented in Supplementary Fig. S5. Ctrl. = control and n.s. = not significant. *, p < 0.05.

2.5. Nanoparticle tracking analysis (NTA)

The ZetaView Twin 405/488 (Particle Metrix GmbH, Ammersee, DEU) was used to determine the concentration of EVs secreted by rat AMs. Entire 17,000 x g or 100,000 x g EV fractions collected from AMs were diluted in 3 mL RPMI 1640, and 3 captures of 1 mL were collected per sample. Injection with RPMI 1640 culture medium was used to wash tubing in between samples. Data are presented as 3 independent experiments with 3 captures performed per experiment.

2.6. Isolation and in vitro treatment of primary AMs for quantitative PCR analysis and lactate assay

Primary AMs were lavaged from the lungs of 6 – 8 week-old C57Bl/6 female mice (Jackson Laboratory, Bar Harbor, ME, USA) as previously described [6]. AMs (0.5 – 1 × 10⁶ cells per well of a 6-well tissue culture plate) were adhered overnight in RPMI 1640 supplemented with 2% FBS, washed, and then treated with GM-CSF (50 ng/mL) for 6, 12, or 24 h.

2.7. RNA isolation and quantitative reverse transcription PCR (RT-qPCR)

Total RNA was extracted from mouse AMs using QIAGEN columns (QIAGEN, Hilden, DEU) per manufacturer’s instructions and converted to cDNA via reverse transcription. Relative gene expression of mouse Glut1, Hk2, Pfkp, Idh1, Idh2, Idh3a, Idh3b, and Idh3g was determined by ΔCt method using SYBR Green dye (Applied Biosystems, Foster City, CA) with β-actin (Actb) as a reference gene. The following primer sequences were used: mGlut1 Forward: CATGGTTCATTGTGGCCGAG; mGlut1 Reverse: GAAGAAGAGCACGAGGAGCA; mHk2 Forward: TCATTGTGGGTACTGGCAGT; mHk2 Reverse: TGAGCCCATGTCGATCTCTC; mPfkp Forward: GAAGCCAAATGGGACTGTGT; mPfkp Reverse: CACGCACAGATTGGTTATGC; mIdh1 Forward: ATCATCATTGGCCGACATGC; mIdh1 Reverse: TCCTGGTTGTACATGCCCAT; mIdh2 Forward: CTATGACGGGCGTTTCAAGG; mIdh2 Reverse: TGAGCCAGGATGTCAGACTG; mIdh3a Forward: CATCACCGAAGAAGCAAGCA; mIdh3a Reverse: GAGCCCATCTGACATCCTCA; mIdh3b Forward: GCTTCTGAGGAGAAGCTGGA; mIdh3b Reverse: CTTCACGTGGACTACGTTGG; mIdh3g Forward: CACCTCCATCCGAAAAGCTG; mIdh3g Reverse: CCACAGCCCGTCCATTAATG; mActb Forward: GACGGCCAGGTCATCACTAT; mActb Reverse: GCACTGTGTTGGCATAGAGG.

2.8. L-Lactate assay

Primary mouse AMs were isolated and cultured as described above. Cell culture supernatants were collected after 3, 6, 12, and 24 h treatment with GM-CSF (50 ng/mL) and centrifuged at 500 x g for 10 min to remove dead cells and debris. Cell-free supernatants were then deproteinized using an Abcam deproteinization kit per manufacturer’s instructions. Lactic acid content was determined in the deproteinized cell culture supernatants using an L-Lactate Assay Kit (Abcam) per manufacturer’s instructions.

2.9. Isocitrate dehydrogenase (IDH) activity assay

Primary rat AMs were isolated and cultured as described above, and total IDH activity was measured using a colorimetric assay kit (Abcam) after 6 and 24 h treatment with GM-CSF (50 ng/mL) per manufacturer’s instructions. Briefly, 1 × 10⁶ AMs were lysed in 200 μL lysis buffer, which was then sonicated and centrifuged at 13,000 x g for 10 min to remove insoluble material. Supernatants were then used to measure total IDH activity using NAD⁺/NADP⁺ as substrate with subtraction of background NADPH. Measurements were performed with absorbance read at 450 nm over 30 – 45 min on a SpectraMax iD5 plate reader (Molecular Devices, San Jose, CA, USA), and IDH activity (nmol of NADH/NADPH generated per min at pH 8 at 37°C) was calculated. Data are presented as 3 independent experiments with experimental duplicates analyzed in technical triplicates.

2.10. Data collection and analysis

Results were from at least 3 independent experiments containing single samples per condition unless otherwise specified. When AMs were cultured with GM-CSF in the presence of various inhibitors or their vehicle controls, % SOCS3 secretion was calculated by dividing the densitometrically determined optical density (OD) of vesicular SOCS3 released from GM-CSF-treated AMs by the OD of vesicular SOCS3 released from untreated AMs. Pooled data were expressed as mean ± SEM and analyzed using the Prism 5.0 statistical program (GraphPad Software, San Diego, CA, USA). Significance was determined using a paired student’s t-test and was inferred at a p < 0.05. Asterisks (*) were used to label significant values, as specified in the figure legends.

3. Results

3.1. GM-CSF inhibits the vesicular packaging and release of SOCS3 by AMs

We have reported that AM secretion of SOCS3 within EVs is diminished in various inflammatory disease states and following exposure to various pro-inflammatory mediators [6, 7, 11]. GM-CSF promotes inflammation within the lung [35–37] as well as the proliferation [38–41] and functional activation [28, 42] of AMs. To test its effects on AM SOCS3 secretion, equal numbers of AMs were treated with GM-CSF or its vehicle control for 20 – 24 h to allow for elaboration of EVs into CM and subsequent detection of total vesicular (>100-kDa) SOCS3 by Western blot. Indeed, treatment with GM-CSF inhibited the AM release of vesicular SOCS3 (Fig. 1a). Notably, this inhibitory effect was unaccompanied by any change in SOCS3 levels detected in AM lysates (Fig. 1b). These data suggest that GM-CSF inhibited the secretion of SOCS3 by AMs without affecting intracellular SOCS3 content. Then, to specifically localize changes in secreted SOCS3 to EVs, we performed differential ultracentrifugation and collected two vesicular fractions: EVs pelleted at 17,000 x g and EVs pelleted at 100,000 x g. In agreement with our prior reports [6, 10], we only detected tonic SOCS3 secretion in the 17,000 x g but not 100,000 x g pellet, and secretion of SOCS3 within 17,000 x g EVs was similarly inhibited by GM-CSF treatment of AMs (Fig. 1c). However, the inhibition of SOCS3 secretion by AMs was associated with only a modest and non-significant reduction in the number of total EVs detected by NTA in the 17,000 x g pellet (Fig. 1c). GM-CSF also had no effect on the number of EVs quantified in the 100,000 x g pellet (Supplementary Fig. S1). Therefore, as diminished SOCS3 secretion was unaccompanied by any substantial changes in SOCS3 intracellular expression or AM vesiculation, we conclude that GM-CSF specifically attenuated the active sorting of SOCS3 into AM-derived 17,000 x g EVs (i.e., the amount of SOCS3 packaged per individual EV).

3.2. GM-CSF inhibition of AM vesicular SOCS3 secretion is glycolysis-dependent

As noted previously, AMs at baseline exhibit markedly lower levels of glycolysis than do other tissue macrophage populations [13–19]. Because such glycolytic restraint has been implicated in the homeostatic functions of AMs [13, 16–18, 20, 21] – to which SOCS3 secretion contributes [6] – and because GM-CSF has been shown to promote glycolysis in macrophages [29, 30], we hypothesized that increases in glycolysis were causally involved in the loss of vesicular SOCS3 secretion observed following treatment of AMs with GM-CSF. To evaluate the plausibility of this hypothesis, we first measured the expression of genes involved in the regulation of glucose metabolism by RT-qPCR. As shown in Fig. 2a, and consistent with a previous report [29], GM-CSF enhanced the expression of the key rate-limiting glycolytic genes glucose transporter 1 (Glut1), hexokinase 2 (Hk2), and phosphofructokinase (Pfkp). Accordingly, expression of GLUT1 and HK2 proteins was also augmented in GM-CSF-treated AMs (Supplementary Fig. 2a). Resulting increases in glycolytic activity were then functionally confirmed by the detection of significant increases in secreted lactate caused by GM-CSF (Fig. 2b). We next employed a variety of approaches to determine whether the observed increases in glycolysis were implicated in the diminished release of vesicular SOCS3. The inhibition of SOCS3 secretion by GM-CSF was dose-dependently mitigated in AM culture by glucose restriction (Fig. 2c). Additionally, culture of AMs in the presence of 2-DG, an inhibitor of glycolysis, completely abrogated the inhibitory effect of GM-CSF on vesicular SOCS3 secretion (Fig. 2d). Finally, treatment with DMOG, an analog of α-ketoglutarate that prevents degradation of the transcription factor hypoxia-inducible factor-1α (HIF-1α) and which robustly augments glycolysis in AMs [14], phenocopied GM-CSF in inhibiting the release of vesicular SOCS3 without affecting intracellular SOCS3 content (Supplementary Fig. S2b–c). Taken together, these data conclusively demonstrate that GM-CSF-induced inhibition of SOCS3 packaging and release within EVs depends on glucose utilization and glycolytic flux.

Figure 2: GM-CSF inhibits vesicular SOCS3 release by AMs in a glycolysis-dependent manner.

(a-b) Adherent AMs collected by lung lavage of mice were treated with GM-CSF (50 ng/mL) for 3, 6, 12, or 24 h. (a) RNA was collected and reverse transcribed to cDNA, and relative gene expression for Glut1, Hk2, and Pfkp was determined by qPCR. (b) Cell-free supernatants from murine AMs cultured for X h as described in (a) were deproteinized, and extracellular lactate concentrations were measured by L-Lactate assay. (c-e) Adherent AMs collected by lung lavage of rats were treated with GM-CSF (50 ng/mL) for 20 – 24 h in varying concentrations of glucose (c) or in the presence or absence of 2-DG (5 mM) (d) or MPCi (20 μM) (e). EVs in CM were concentrated by 100-kDa centrifugal filtration and vesicular SOCS3 secretion was determined via Western blot of the total vesicular fraction (>100-kDa) samples. Data, expressed as (mean ± SEM) (a-b) or as values from individual experiments (c-e), are from 3 independent experiments, and significance was determined using paired sample t-test. Dashed line represents SOCS3 released by GM-CSF-treated AMs during culture in glucose-free medium (c) or by untreated AMs (i.e., no GM-CSF) during culture with inhibitor or its vehicle control (d and e). Western blot images are cropped to encapsulate the band for SOCS3 visible at 27 kDa (c-e). Full-length Western blots are presented in Supplementary Fig. S6. Ctrl. = control and Glc = glucose. *, **, ***, and ****, p < 0.05, p < 0.01, p < 0.001, and p < 0.0001, respectively.

3.3. ACLY mediates glycolytic inhibition of vesicular SOCS3 secretion

During aerobic metabolism, the product of glycolysis – pyruvate – undergoes transport from the cytosol into mitochondria where it is converted to acetyl-CoA to support Krebs cycle flux. To determine whether this process was required for GM-CSF’s inhibitory effect on vesicular SOCS3 secretion, we evaluated the effects of UK-5099 (MPCi), a potent inhibitor of the mitochondrial pyruvate carrier (MPC). Indeed, co-treatment of AMs with MPCi negated GM-CSF’s inhibitory effect on SOCS3 release by AMs (Fig. 2e). Given these data, we then hypothesized that a glucose-derived Krebs cycle metabolite mediated the inhibition of AM vesicular SOCS3 release caused by GM-CSF. We first considered a potential contribution of citrate given its role in metabolic reprogramming in inflammatory myeloid cells [43–46]. To evaluate this possibility, we measured the activity of IDH, the enzyme comprising three different isoforms (i.e., IDH1 [cytosolic] as well as IDH2 and IDH3 [both mitochondrial]) responsible for converting isocitrate to oxoglutarate and whose downregulation causes a “broken” Krebs cycle in glycolytically remodeled macrophages to promote increased levels of citrate [47, 48]. As shown in Fig. 3a, treatment of AMs with GM-CSF markedly reduced total (i.e., NAD⁺- and NADP⁺-dependent) IDH activity within 6 h, an effect that was sustained throughout the 24 h duration of AM culture with GM-CSF. This effect involved substantial downregulation of Idh1 expression coinciding with modest increases in expression of the Idh2, Idh3a, Idh3b, and Idh3g isoforms (Supplementary Fig. S3). Such a result is consistent with prior reports describing diminished IDH activity parallel to divergent regulation of IDH isoform expression in macrophages [48]. We therefore postulated that GM-CSF might glycolytically inhibit vesicular SOCS3 release by downregulating IDH activity to augment mitochondrial citrate levels, thus culminating in citrate export to the cytosol where it serves as a substrate for the generation of acetyl-CoA [31]. Accordingly, to evaluate the role of mitochondrial citrate export in the inhibition of AM secretion of SOCS3 by GM-CSF, we used a competitive inhibitor of the citrate transport protein CTP (CTPi). Indeed, culture of AMs with the CTPi reversed the GM-CSF-induced loss of vesicular SOCS3 release (Fig. 3b). Also consistent with an inhibitory role for citrate on SOCS3 release, treatment of AMs with exogenous sodium citrate (1 mM), an approach previously used to artificially augment intracellular acetyl-CoA levels in myeloid cells [49], robustly inhibited vesicular SOCS3 secretion (Fig. 3c). Furthermore, as GM-CSF has been previously shown to augment the expression of ACLY [50], the cytosolic enzyme responsible for converting citrate to acetyl-CoA [31], and is known to enhance the generation of acetyl-CoA in macrophages [30], we next tested whether ACLY was also involved. Indeed, ACLY inhibition with either of two structurally distinct inhibitors, SB-204990 (ACLYi[1]) and BMS-303141 (ACLYi[2]), overcame the attenuation of SOCS3 secretion by GM-CSF (Fig. 3d and Supplementary Fig. S4a). Recovery of SOCS3 release by ACLY inhibition was not due to changes in AM intracellular SOCS3 content (Supplementary Fig. S4b), further suggesting that GM-CSF’s inhibitory effect on SOCS3 secretion involved control over its packaging in EVs. Together, these data strongly suggest that the activity of ACLY links increases in glycolytic flux to diminished release of vesicular SOCS3 by way of citrate transport to the cytosol and its subsequent conversion to acetyl-CoA.

Figure 3: Inhibition of vesicular SOCS3 secretion by GM-CSF requires the cytosolic conversion of citrate to acetyl-CoA by ACLY.

(a) Adherent AMs collected by lung lavage of rats were treated with GM-CSF (50 ng/mL) for 6 or 24 h, and IDH activity was measured by colorimetric assay. (b-e) Adherent AMs collected by lung lavage of rats were treated with GM-CSF (50 ng/mL) (b and d), sodium citrate (1 mM) (c), or sodium acetate (1 mM) (e) for 20 – 24 h in the presence or absence of CTPi (1 mM) (b), ACLYi(1) (12.5 μM) (d), or ACSS2i (2 μM) (e). EVs in CM were concentrated by 100-kDa centrifugal filtration and vesicular SOCS3 secretion was determined via Western blot of the total vesicular fraction (>100-kDa) samples. (f) Proposed model of regulation of vesicular SOCS3 packaging by glycolytic flux and ACLY. Data, expressed as (mean ± SEM) (a and c) or as values from individual experiments (b, d, and e), are from ≥3 independent experiments, and significance was determined using paired sample t-test. Dashed line represents SOCS3 released by untreated AMs (i.e., no GM-CSF b and d] or sodium acetate [e]) during culture with inhibitor or its vehicle control. Western blot images are cropped to encapsulate the band for SOCS3 visible at 27 kDa (b-e). Full-length Western blots are presented in Supplementary Fig. S7. Ctrl. = control. *, **, ***, and ****, p < 0.05, p < 0.01, p < 0.001, and p < 0.0001, respectively.

Although ACLY is generally regarded as the primary enzyme responsible for generating cytosolic acetyl-CoA [31, 51, 52], the nucleocytosolic enzyme ACSS2, which produces acetyl-CoA from exogenous acetate, can also contribute in certain biologic scenarios [53–57]. We therefore explored the possible role of this alternative route of cytosolic acetyl-CoA generation in the inhibition of secretion of SOCS3 by AMs. Treatment of AMs with sodium acetate (1 mM) robustly inhibited SOCS3 secretion in a manner reversible by addition of an ACSS2 inhibitor (ACSS2i) (Fig. 3e). However, treatment of AMs with the ACSS2i failed to substantially or significantly abrogate the inhibitory effect of GM-CSF on vesicular SOCS3 release (Supplementary Fig. S4c). Together, these data strongly suggest that increased acetyl-CoA production plays a critical role in mediating the diminished release of vesicular SOCS3 in response to GM-CSF. Furthermore, although acetyl-CoA can be generated from either mitochondrial-derived citrate or exogenous acetate, GM-CSF predominantly utilized citrate and the ACLY pathway to generate the acetyl-CoA responsible for inhibition of SOCS3 secretion (summarized in Fig. 3f).

4. Discussion

Using GM-CSF treatment as a biologically relevant in vitro driver of metabolic remodeling within the lung mucosal microenvironment, we demonstrate here that the resulting glycolytic flux inhibits vesicular SOCS3 release by primary AMs, a phenomenon attributable to ACLY-mediated conversion of mitochondrial-derived citrate to cytosolic acetyl-CoA. These results were also recapitulated in vitro by augmentation of glycolytic flux following AM treatment with DMOG. To our knowledge, our studies are the first to suggest that cellular metabolites control the dynamic packaging and release of a cargo molecule within EVs. Further, our findings expand on the limited body of literature exploring how lung microenvironmental perturbations alter the metabolism of resident AMs. Although we anticipate that this interplay between metabolic remodeling and EV composition extends to other cellular contexts, the importance of enhanced glycolytic flux as a determinant of these processes may be especially meaningful in the lower respiratory tract where glucose levels – measured to be 3 – 20 times lower than plasma concentrations at steady state – increase during inflammation [58–62]. Additionally, given the magnitude by which increases in glycolysis diminish the release of SOCS3, our findings predict that glycolytically reprogrammed AMs lose the ability to dampen STAT3 signaling in AECs. Therefore, we anticipate that diminished SOCS3 release by AMs underlies inflammatory dysregulation characteristic of the glycolytic alveolar milieu.

A major outstanding question concerns the unique microenvironmental factors that drive metabolic remodeling in tissue-resident vs. recruited macrophages. Here, we have used GM-CSF as a biologically relevant driver of glycolysis in the lung given its prominent role in developmentally and functionally regulating AMs, both at steady state and during inflammation [27, 28]. The extent to which these findings are generalizable to other endogenous and exogenous factors similarly known to promote metabolic remodeling is of great interest. For example, an obvious line of inquiry is whether pathogen-associated molecular patterns (PAMPs) that promote glycolytic remodeling in bone marrow-derived macrophages (BMDMs) (e.g., lipopolysaccharide [LPS]) [32] also inhibit vesicular SOCS3 release by AMs. Prior work from our laboratory showed an attenuation of vesicular SOCS3 packaging in LPS-treated AMs [6], suggesting a potential contribution by glycolysis. However, recent evidence has demonstrated disparate regulation of metabolism in BMDMs vs. AMs, as the latter do not exhibit glycolytic remodeling following LPS treatment [14]. Indeed, our preliminary findings confirm that LPS-treated AMs do not exhibit increases in glucose uptake (data not shown). Therefore, it is highly unlikely that diminished vesicular SOCS3 secretion is an effect generalizable to all PAMPs that cause glycolytic remodeling in BMDMs, though thorough interrogation of this question will require additional experimentation.

Despite the fundamental importance of metabolism for regulating cellular phenotype and function, metabolic control over the release and composition of EVs remains poorly understood. Predictably, its investigation has been restricted to the study of tumor cells where metabolic reprogramming is a hallmark of cellular transformation and an adaptive response to nutrient availability within the tumor microenvironment. For example, glutamine metabolism in breast cancer cells has been shown to drive EV formation, thus demonstrating metabolic control over the number of secreted EVs [25]. Furthemore, recent work from Fan et al. proposed an “exosome switch” model where deprivation of glutamine or inhibition of mechanistic target of rapamycin complex 1 caused a switch from the production of CD63-containing Exos to Rab11a-containing Exos [22]. Glutamine metabolism has also been shown to uniquely drive the production of large EVs by tumor cells [23, 24]. These studies thus established precedent for the control of cellular metabolism over the number and subtype of EVs secreted. In contrast, we propose a novel model of metabolic regulation in which glycolytic flux dynamically inhibits the sorting of SOCS3 into AM-derived EVs. This conclusion is strongly supported by our observation that GM-CSF glycolytically inhibited the release of SOCS3 in EVs pelleted at 17,000 x g without significantly altering EV number (Fig. 1c). Although these findings support a mechanism of diminished vesicular SOCS3 packaging, we cannot, however, eliminate the possibility of a similar “EV switch” effect where enhanced glycolysis in AMs perhaps supplants release of SOCS3-containing EVs with an alternative EV subtype. Given the lack of SOCS3 detected in a classical Exo (i.e., 100,000 x g) fraction (Fig. 1c) [4], it is highly unlikely that any such mechanism would be analogous to what was previously described for CD63- and Rab11a-containing Exos [22]. Nonetheless, exclusion of this scenario will require a more nuanced understanding of SOCS3 sorting into EV subtypes, which we are actively investigating.

A key mechanistic conclusion from our studies is that glycolytic inhibition of vesicular SOCS3 secretion by GM-CSF requires mitochondrial pyruvate transport (Fig. 2e). This finding pointed us toward a mechanism involving suppression of SOCS3 release by a mitochondrial metabolite, which we experimentally attributed to citrate and its subsequent conversion to cytosolic acetyl-CoA. Mitochondrial metabolites are recognized to assume diverse, non-metabolic signaling roles, such as in regulating chromatin modifications, DNA methylation, stem cell function, thermogenesis, tumorigenesis, and immune modulation [63]. Importantly, our work reveals another arm of metabolite signaling in the control over nonclassical protein export, and we anticipate that it is relevant for the secretion of cargoes beyond SOCS3. Indeed, it will be important to determine the breadth of EV molecules whose packaging is controlled by acetyl-CoA, which is especially pertinent given its role as a central metabolic intermediate [64]. Proteomic analysis of Exos derived from epithelial ovarian cancer cell lines revealed an abundance of acetylated proteins [65], suggesting positive regulation of cargo packaging by acetyl-CoA. Conversely, acetylation has been shown to inhibit the sorting of glucose-regulated protein 78 into Exos by promoting its retention in the endoplasmic reticulum [66]. It is therefore likely that the effects of acetyl-CoA on EV composition are both cargo- and context-dependent, and understanding this interplay will require further investigation. Additionally, although our results implicate a causal link between enahnced glycolysis and ACLY activity in the regulation of vesicular SOCS3 secretion, it will be informative to see how alternative acetyl-CoA-generating pathways influence cargo packaging. For example, under low-glucose, nutrient-restricted conditions or during metabolic stress, cells use exogenous acetate as a substrate for production of acetyl-CoA by ACSS2 [53–57]. We demonstrated that this pathway too could diminish the release of vesicular SOCS3 (Fig. 3e), but was not the predominant means of inhibition involved in the actions of GM-CSF (Supplementary Fig. S4c). Furthermore, tumor cells with defective mitochondria use reductive carboxylation of glutamine to fuel acetyl-CoA production and cell growth [67, 68], which is particularly intruiging given our data showing that HIF-1α stabilization with DMOG inhibited vesicular SOCS3 release by AMs (Supplementary Fig. S2b). In addition to augmenting glycolysis, HIF-1α is also known to activate pyruvate dehyrogenase kinase 1 to suppress Krebs cycle flux supported by pyruvate [69]. As a result, during AM treatment with DMOG, the citrate implicated in inhibiting vesicular SOCS3 release could be derived independently of glucose, for example, by glutamine that has undergone reductive carboxylation. The extent to which these mechanisms alter vesicular cargo composition remains unknown, but they may represent attractive biologic targets for therapeutically tuning the bioactivity of EVs. Lastly, it will also be critical to determine whether metabolites other than acetyl-CoA also affect the sorting of vesicular cargoes. For example, acyl-CoA metabolites linked to major metabolic processes with known biologic functions also include succinyl-CoA, propionyl-CoA, butyryl-CoA, and crotonyl-CoA [70]. Future experiments will need to assess the extent to which these diverse mitochondrial metabolites control EV composition and function.

One of the key unanswered questions that arises from the in vitro observations described herein is whethter GM-CSF-induced glycolytic remodeling in AMs mediates the loss of vesicular SOCS3 secretion observed in vivo with inflammatory conditions such as cigarette smoking [6]. The plausibility of such a mechanism is based on the fact that GM-CSF is produced during exposure to cigarette smoke [71] and that AMs from smokers exhibit increased rates of basal glycolysis [72]. Future studies will be required to test this possibility, as well as the potential for ACLY inhibition to restore homeostatic SOCS3 secretion.

5. Conclusions

In summary, we have illuminated a previously unknown mechanism in which vesicular secretion of the cytosolically localized protein SOCS3 is tuned in AMs by glycolytic regulation of ACLY activity and the production of acetyl-CoA. Although revealed in the context of changes to the lung microenvironment, we anticipate these findings to be broadly relevant to other cellular and tissue contexts where glycolytic remodeling occurs. Future studies will be required to determine the extent to which metabolic reprogramming and metabolite flux control EV composition and bioactivity. Furthermore, future investigation is needed to illuminate control of metabolite signaling over EV bioactivity in vivo and the potential targeting of these signaling pathways for therapeutic benefit.

Data availability

Raw data can be made available from the corresponding author on reasonable request.