Mitochondrial Fragmentation Promotes Inflammation Resolution Responses in Macrophages via Histone Lactylation

Macrophage polarization influences mitochondrial fission and fusion

To understand how macrophage polarization affects mitochondrial morphology, we first characterized the mitochondrial network of bone marrow-derived macrophages (BMDMs) polarized to traditional M1 or M2 phenotypes. We observed that macrophages stimulated with LPS and IFNγ for 24 h (pro-inflammatory or M1) have a significantly longer mitochondrial network compared to both unpolarized macrophages (resting, M0), and IL-4 stimulated macrophages (proresolving or M2) (Fig. 1a, b, d, Supplemental Movies a-c). The increase in mitochondrial length was observed as early as 2 h after stimulation and prior to the establishment of an M1 identity (Fig. 1e and f). Mitochondrial morphology and changes in mitochondrial length between M0, M1 and M2 macrophages was confirmed by transmission electron microscopy showing longer mitochondria in M1, and a higher ratio of mitochondria undergoing dynamic events (either fission or fusion) (Fig. 1g to i). Furthermore, we saw an increase in cristae width in M1 macrophages compared to M0 and M2 (Fig. 1j). In human THP-1 macrophages polarized to M1 with LPS + IFNγ, there was a similar increase in mitochondrial length compared to both M0 and M2 (Fig. 1a and c). These data demonstrate that pro-inflammatory and M1 macrophages adopt an elongated mitochondrial network, whereas resting and M2 macrophages adopt a shorter mitochondrial network that corresponds with their polarized phenotype.

Promoting mitochondrial fusion does not alter pro-inflammatory IL-1β responses in macrophages

Previous studies have demonstrated that mitochondrial fission and fusion control cell phenotype commitment (e.g. neural and muscle stem cells, T cells).^13,14,19 Therefore, we next sought to determine whether mitochondrial fission and/or fusion directs macrophage M1 or M2 polarization. To phenocopy the hyperfused mitochondrial network in M1 macrophages, we developed a model wherein the mitochondrial fission machinery was silenced using a combination of siRNAs targeting Drp1, Mff and Fis1 to prevent mitochondrial fission (fission KD, Fig. 2a). Fission knockdown resulted in a significantly elongated mitochondrial network compared to control siRNA treated macrophages (Fig. 2b and c). We confirmed knockdown for each siRNA combination at both the mRNA and protein levels (Fig. 2d and e). We assessed pro- and anti-inflammatory gene expression in macrophages from fission KD cells both at baseline (unstimulated) and in response to inflammatory stimulation to LPS- a prototypical model of inflammatory activation. In unstimulated macrophages, phenocopying a hyperfused mitochondrial network resulted in a decreased expression of the M1 markers Il1b & Tnfa, whereas Nos2 expression was increased which corresponds with decreased M2 marker Arg1 (Fig. 2f and g). To determine whether macrophages with hyperfused mitochondria respond normally to an inflammatory stimulus, we assessed Il1b gene expression after LPS stimulation. Macrophages with elongated mitochondria had decreased Il1b gene expression after 24 h of LPS stimulation compared to controls (Fig. 2h). To assess how changes in gene expression may impact protein function, we examined the expression of the precursor, mature and secreted forms of IL-1β protein in response to LPS stimulation over time. The loss of mitochondrial fission had no effect on either pro-IL-1β or IL-1β secretion (Fig. 2i to k). To determine whether elongated mitochondria influenced the full activation of IL-1β protein secretion, which requires the engagement of the NLPR3 inflammasome via a priming and an activation signal, we tested an in vitro model of inflammasome activation with LPS priming followed by ATP activation.^20–23 Priming with LPS and subsequent activation by ATP in macrophages with hyperfused mitochondria resulted in equivalent IL-1β release compared to controls, indicating a normal inflammasome priming and activation process (Fig. 2l). Together, these data indicate that while mitochondrial fusion appears to control the basal expression of macrophage Il1b gene expression, it has little influence on IL-1β protein secretion upon inflammatory stimulation. Therefore, although M1 macrophages adopt a lengthened mitochondrial network in parallel with their pro-inflammatory gene expression state, forcing an elongated mitochondrial network does not directly cause a heightened inflammatory response to LPS.

A fragmented mitochondrial network alters pro-inflammatory gene expression in response to LPS via noncanonical inflammatory pathways

Because we observed that M2-polarized macrophages have a more fragmented mitochondrial network compared to M1 (Fig. 1a and b), we sought to phenocopy this state by silencing mitochondrial fusion machinery using siRNA targeting Opa1, Mfn1 and Mfn2 (fusion KD, Fig. 3a). As expected, silencing the fusion machinery resulted in a fragmented mitochondrial network (Fig. 3b to e). Interestingly, similar to what we observed upon knocking down of the mitochondrial fission machinery, fusion KD resulted in a decrease in basal Il1b gene expression in unstimulated macrophages (Fig. 3f). However, unlike what happened on the loss of mitochondrial fission, loss of mitochondrial fusion resulted in a basal increase in the M2 polarization marker Arg1 (Fig. 3g). To determine whether preventing mitochondrial fusion would result in an impaired inflammatory response, we stimulated control and fusion KD macrophages with LPS and found that impaired mitochondrial fusion results in increased Il1b gene expression after 6 and 24 h (Fig. 3h). Activation of IL-1β protein was assessed after LPS stimulation where loss of mitochondrial fusion results in an increase in pro-IL-1β at 6 and 24 h (Fig. 3i) and a corresponding small but not statistically significant increase in IL-1β secretion at 24 h (Fig. 3j). Upon inflammasome activation with LPS and ATP (as described above), loss of mitochondrial fusion led to a significant increase in IL-1β secretion (Fig. 3j). These data are similar to what is observed in a parallel model of impaired mitochondrial fusion where genetic loss of Opa1 led to an increase in IL-1β secretion upon inflammasome activation,¹⁵ and support that this is likely due to functional losses in mitochondrial fusion.

To investigate how mitochondrial fusion alters the expression of IL-1β upon LPS stimulation, we investigated transcriptional pathways known to promote cytokine production. We first assessed NFκB activity, which is a potent inducer of Il1b expression and is quickly activated upon LPS stimulation, where the p65 subunit becomes phosphorylated and translocates to the nucleus to activate inflammatory gene expression.^24–26 We assessed NFκB activity via measuring the phosphorylation of the p65 subunit by Western blot. Surprisingly, despite increases in Il1b gene expression upon loss of mitochondrial fusion, we find that p65 phosphorylation is significantly decreased after 24 h of LPS stimulation (Fig. 3k, Western blot not shown). These data indicate that the observed increase in IL-1β expression and secretion upon loss of mitochondrial fusion may be caused by an alternative NFκB-independent mechanism. HIF1α nuclear translocation, which is stimulated in response to LPS and drives inflammatory cytokine production downstream of NFκB was next assessed by microscopy.²⁷ Upon LPS stimulation, the loss of mitochondrial fusion caused a decrease in HIF1α nuclear translocation (Fig. 3l). This observed decrease in HIF1α translocation would not be expected to result in increased IL-1β secretion upon LPS stimulation; therefore we looked into alternative pathways that are known to be downstream of inflammatory stimulation and may activate cytokine expression. We assessed the activity of the ATF4/c-Jun pathway, which can promote Il1b expression in response to LPS.²⁸ When mitochondrial fusion is impaired, both ATF4 and phosphorylated c-Jun are significantly increased after 6 h of LPS stimulation (Fig. 3m and n). To confirm the increase in ATF4 activity upon knockdown of mitochondrial fusion, another ATF4-responsive gene, Slc7a11 was measured after both 6 h and 24 h of LPS stimulation,²⁹ and indeed was increased in macrophages with fusion KD (Fig. 3o). Together these data indicate that, in response to inflammatory stimulation, mitochondrial fusion contributes to the macrophage inflammatory IL-1β response possibly via activation of the ATF4/c-Jun axis.

A fragmented mitochondrial network enhances the proresolving macrophage phenotype

Given that M2 macrophages had a more fragmented mitochondrial network compared to M1 macrophages (Fig. 1a and b) and that loss of mitochondrial fusion led to an increase in baseline expression of Arg1 (Fig. 3g), we sought to determine whether mitochondrial fusion alters the anti-inflammatory macrophage phenotype during the resolution phase of inflammation. Arginase 1 is a critical enzyme in macrophages that promotes the resolution of inflammation by reducing nitric oxide signaling, promoting L-arginine metabolism to assist in tissue repair.³⁰ In normal macrophages, Arg1 expression is temporally activated by LPS, with expression appearing late following LPS stimulation, following the peak of the pro-inflammatory response.³¹ We therefore assessed whether a fragmented mitochondrial network influenced the late activation of Arg1. Upon LPS stimulation, Arg1 gene expression was maximally increased in mitochondrial fusion-impaired cells after 24 h, translating to an increased protein level at 48 h (Fig. 4a to c). IL-4 is a potent inducer of M2 macrophage responses and is produced by Th2 cells to trigger the anti-inflammatory program in macrophages.^32,33 We therefore assessed the impact of impaired mitochondrial fusion upon direct anti-inflammatory stimulation using IL-4. Arg1 gene and protein expression were both increased upon fusion KD after 24 h of IL-4 treatment (Fig. 4d to f). To test how mitochondrial fusion impacts macrophage plasticity, we first stimulated cells with LPS for 6 h, after which we stimulated them with IL-4, to accelerate the proresolving phase. Loss of mitochondrial fusion increased Arg1 gene expression and protein compared to control cells, indicating a heightened proresolving response (Fig. 4g to i). A key function of M2 proresolving macrophages is phagocytic capacity; therefore, we measured phagocytosis of latex beads. Loss of mitochondrial fusion resulted in an increase in phagocytosis compared to controls basally and under LPS, IL-4 or combined stimulation. (Fig. 4j to m). Together these data indicate that inducing a fragmented mitochondrial phenotype promotes inflammation resolution following inflammatory or anti-inflammatory activation.

Mitochondrial fragmentation increases histone lactylation and promotes an M2 phenotype

Pro-inflammatory macrophages are highly glycolytic, and as a result, produce high levels of lactate upon inflammatory stimulation.^34,35 Lactate is a powerful signaling molecule and is known to promote M2-like responses in tumor macrophages.³⁶ It was recently shown that once sufficiently elevated, lactate causes a change in gene expression via lactylation of lysine residue on histones, turning on proresolving machinery such as Arg1.^37–39 To understand whether mitochondrial fragmentation promotes changes in lactate metabolism, we assessed lactate levels in control and fusion KD cells upon LPS stimulation. After 24 h of LPS treatment, lactate levels were increased in fusion KD compared to controls (Fig. 5a). In agreement with the increase in lactate levels, histone lactylation (marked by the pan-histone lactylation antibody, KLA) was increased with knockdown of mitochondrial fusion after 24 h of LPS stimulation (Fig. 5b and c). Both lactate and histone lactylation were equivalent to that of control macrophages after 48 h of LPS stimulation (Fig. 5a to c). To better understand the increase in lactate and KLA upon loss of mitochondrial fusion, we first measured the expression of pyruvate dehydrogenase (PDH), which converts pyruvate to acetyl-CoA to enter the TCA cycle.³⁵ Upon loss of mitochondrial fusion, we observe a decrease in PDH expression after 24 h of LPS stimulation (Fig. 5d and e). It would be expected that lower PDH expression in cells with fragmented mitochondria would result in reduced conversion of pyruvate to acetyl-coA, and an increase in conversion of pyruvate to lactate.⁴⁰ To test this hypothesis, we inhibited lactate accumulation by inhibiting the activity of lactate dehydrogenase (LDH) using sodium oxamate (Ox).³⁶ Co-treatment of macrophages with LPS and sodium oxamate for 48 h prevented the induction of ARG1 protein in fusion KD macrophages (Fig. 5f and g) and induction of KLA in control cells (Fig. 5h and i). In summary, the above data indicate that increased lactate accumulation in cells with fragmented mitochondria is driving the increase in Arg1 expression in macrophages following inflammatory activation.

The observation that siRNA-induced mitochondrial fragmentation resulted in an increase in histone lactylation and Arg1 expression prompted us to examine whether this occurs also during the physiological macrophage resolution response. Having observed a fragmented mitochondrial network in IL-4 treated M2 polarized macrophages, so we next sought to characterize the mitochondrial phenotype of macrophages as they enter the proresolving phase after LPS exposure, during which the switch to histone lactylation occurs and Arg1 expression is induced. Analysis of mitochondrial morphology after 24, 48, and 72 h following LPS stimulation revealed maximal mitochondrial elongation at 24 h, and a return to near baseline length after 48 and 72 h (Fig. 5j to l). There was a corresponding increase in lactate (Fig. 5m) and histone lactylation, as measured by KLA, between 24 and 72 h (Fig. 5n). Furthermore, co-treatment with LDH inhibitor sodium oxamate and LPS for 72 h prevented the induction of both ARG1 and KLA (Fig. 5o and p). Together, these data indicate that a fragmented mitochondrial network is associated with an increase in histone lactylation during the resolution phase of inflammation.

Animal procedures

C57BL/6 (wild-type, Charles River Laboratories) mice were maintained and housed in the University of Ottawa Heart Institute (UOHI) Animal Care and Veterinary facility. Mice ranged in age between 4 and 10 months with equal distribution of males and females.

Cell culture

Murine bone marrow-derived macrophages (BMDM) were obtained by flushing the femur and tibia in DMEM high glucose media (Gibco, 11965092) supplemented with 20% L929 conditioned media (a source of MCSF produced in-house from L929 cell culture), 10% heat-inactivated FBS (Gibco, 12483020) and 1% Antibiotic Antimycotic Solution (Gibco, 15240062). Monocytes were left to differentiate for 6 days in an incubator at 37 °C with 5% CO₂. On day 6 the differentiated macrophages were lifted using 50 mM EDTA, counted and plated according to experimental conditions. All experimental procedures were started on day 7. Human monocytic THP-1 cell line were maintained in culture in RPMI medium (Gibco, 21870076) supplemented with 10% of heat-inactivated FBS (Gibco, 12483020), 10 mM Hepes (Gibco, 15630080), 10 mM L-Glutamine (Gibco, 25030149), 1 mM sodium pyruvate (Gibco, 11360070), 2.5 g/l D-glucose (Merck), 1% Penicillin-Streptomycin (Gibco, 15140122) and 50 pM β-mercaptoethanol (Gibco; 31350–010). THP-1 monocytes were differentiated into macrophages through incubation with 100 nM PMA (Sigma, P1585) in full media for 72 h.

Macrophage treatments

BMDM macrophages were polarized to M1 or M2 through treatment with 100 ng/mL LPS (Sigma, L4391) and 100 ng/mL IFN-γ (R&D, 485-MI-100) or 10 ng/mL IL-4 (R&D, 404-ML-010) for 24 h. THP-1 macrophages were polarized to M1 or M2 by incubation with 20 ng/mL IFN-γ (PeproTech, 300-02) and 100 ng/mL LPS (Sigma, L4391) or 20 ng/mL IL-4 (PeproTech, #200-04) and 20 ng/mL IL-13 (PeproTech, 200-13) for 24 h. LPS time courses were performed in full media with 100 ng/mL LPS (Sigma, L4391) with or without 10 mM sodium oxamate (SelleckChem, S6871). IL-4 time courses were performed in full media with 10 ng/mL IL-4 (R&D, 404-ML-010).

Knockdown

Fission knockdown macrophages were obtained by 72 h siRNA cotransfection with 40 nM Drp1 siRNA (Qiagen, SI0098226), 10 nM Mff siRNA (Qiagen, SI00855918) and 10 nM Fis1 siRNA (Qiagen, SI0145779) in OPTIMEM media (Gibco, 11058021) with Lipofectamine RNA iMAX transfection reagent (Life Technologies, 13778150) used according to manufacturers’ instructions. Fusion knockdown cells were obtained using the same method with 20 nM Opa1 siRNA (Qiagen, SI01365707), 20 nm Mfn1 siRNA (Qiagen, SI01304387), 20 nM Mfn2 siRNA (Qiagen, SI04392010). Knockdown efficiency was determined in comparison to control cells transfected with 60 nM negative control siRNA (Qiagen, 1027310).

RNA isolation & RT-qPCR

Total RNA was isolated using TRIzol (Invitrogen, 15596018) according to manufacturers instruction. The cDNA reactions were prepared from 750 ng of total RNA using iScript reverse transcription super mix (Bio-Rad, 1708841) and a 1:5 dilution was used for qPCR. Real-time quantitative PCR reactions were carried out using primer sequences obtained from the Harvard Medical School PrimerBank database (Table of primer sequences in Supplemental table 1) and SYBR green dye (Bio-Rad, 1725275). Data were normalized to either SRP14, TFAM or HPRT gene expression using the comparative ΔΔCt method and calculated fold changes relative to the controls.

Western blots

Total protein was extracted on ice in radioimmunoprecipitation assay buffer (RIPA) containing protease (Roche, 4693132001) and phosphatase (Roche, 4906837001) inhibitors. Proteins were separated on acrylamide gels of varying percentages and transferred to 0.22 μM polyvinylidene difluoride (PVDF) membranes. Membranes were blocked in 5% skim milk or 5% bovine serum albumin (BSA), depending on the antibody manufacturers’ specifications. Primary antibodies were incubated on the membrane overnight at 4 °C in 1% blocking. Membranes were incubated with LI-COR Secondaries (IRDye 800CW α-Rabbit and IRDye 680RD α-Mouse) at room temperature for an hour and imaged on the LI-COR Odyssey imaging system. Protein bands were quantified on ImageJ and normalized to housekeeping bands, either HSP90 or β-Actin. Proteins ATF4, Phos C-Jun, KLA and ARG1 required the use of HRP secondaries and chemiluminescence to be detected. Full list of antibodies and their catalogue numbers can be found in Supplemental Table 2.

Immunofluorescence microscopy

For all immunofluorescence microscopy, the macrophages were plated on glass coverslips within a 24-well plate. At the end of treatment (either polarization or siRNA transfection), the macrophages were fixed with 4% paraformaldehyde for 15 min at 37 °C, followed by 10 min of quenching with 10 mM NH₄Cl. The cells were permeabilized with 0.1% Triton X-100 for 10 min. 10% FBS was used as a blocking agent and incubated on the coverslips for 30 min. Primary antibodies were then incubated for an hour in 5% FBS followed by secondary antibodies also in 5% FBS for an hour. Primary antibodies used are detailed in the associated figure legends. Secondaries used were Alexa Fluor 555 or 488 in the associated species. 1 μg/mL DAPI (Becton Dickinson, 564907) was used as a nuclear dye and incubated on the coverslips for 5 min. Coverslips were then washed and mounted using Dako Fluorescence mounting media (Cedarlane, S3022380-2) on microscopy slides (Fisher Scientific, 12-552). Cells were imaged within 2 days at 63× on a Zeiss LSM 880 microscope. Mitochondrial length was measured manually using the ImageJ software and presented as mean length (μM) per cell. HIF-1α nuclear translocation was also measured on ImageJ, whereby we quantified the total mean fluorescence intensity (MFI) per cell and the MFI within the nucleus. Data is presented as a ratio of nuclear/cytoplasmic MFI per cell.

Transmission electron microscopy

Cells were fixed for 60 min in 2% PFA and 2.5% glutaraldehyde. The samples were processed and imaged at the University of Ottawa Heart Institute Cell Imaging and Histology Core Facility according to standard protocols. Those images were quantified for mitochondrial length and cristae width using the ImageJ software. All dynamic events (either fission or fusion) were quantified as a ratio between static vs dynamic mitochondria per cell.

Phagocytosis assay

Cells plated on glass coverslips within a 24-well plate were incubated with equal concentrations of polystyrene fluoresbrite YG microspheres (Polysciences, 17154-10) for 2 h at 37 °C. The wells were washed with PBS to remove excess beads, then fixed and stained for Tom20 according to the above IF protocol. The cells were imaged within 2 days at 63× on a Zeiss LSM 880 microscope. Images were manually quantified for the number of internalized fluorescent beads per cell on ImageJ.

Il-1β ELISA

Fission and fusion knockdown cells were assessed for changes in IL-1β secretion upon LPS (100 ng/μL) stimulation with and without 5 mM ATP (Bio-Rad, 1725275). Cell media was collected at end of LPS treatment centrifuged at 1200 g for 5 min to discard any cell debris. Media was quantified for secreted IL-1β by mouse IL-1 beta Quantikine ELISA Kit (R&D, MLB00C) according to the manufacturer’s instruction.

Lactate levels

Fusion knockdown cell lysate lactate levels were quantified using the L-Lactate Assay Kit (ab65331) according to manufacturers’ instructions.

Statistical analysis

All data involving statistics are presented as mean ± SEM. The same control was used for individual comparisons to fission KD or fusion KD. The statistical significance of the differences between groups was determined on Prism V9.5.0 software (GraphPad Software Inc.) using unpaired Student’s t tests, multiple unpaired t tests or one-way ANOVA. Normal distribution was determined by the Shapiro–Wilk test. The number of replicates and the statistical test used are described in the figure legends.