Introduction

Rheumatoid arthritis (RA) is a chronic inflammatory joint disease that affects up to 1% of the population.1 The key feature of RA is synovial inflammation, which is characterised by the activation of fibroblast-like synoviocytes, neovascularisation and immune cell infiltration, resulting in an overall expansion of the synovial membrane.2 The phenotypic traits that are linked to synovial hyperplasia, such as cellular dedifferentiation, sustained proliferation and tissue invasion impose extensive nutritional demands on immune and stromal cells.3 The microenvironment of inflammatory joint lesions is furthermore defined by low oxygen and nutrient availability, enforcing additional bioenergetic changes on resident cells.4 5 To understand the implications of metabolic dysregulation on disease development, extensive research has focused on identifying inflammation-related metabolic adaptations in RA-associated cells over the last decade. These studies revealed that although RA-specific metabolic profiles vary across different cell types, they are consistently linked to enhanced cell pathogenicity, signifying metabolic reprogramming as an important contributing factor for RA progression rather than a mere consequence of the hypoxic and nutrient-deprived microenvironment of the joint.6–8

Another important hallmark of RA, besides synovial inflammation, is enhanced bone degradation. It can manifest as local erosion of bone tissue within the inflamed joint, periarticular bone loss that occurs in close proximity to the inflamed tissue area and systemic osteoporosis of the axial and appendicular skeleton.9–11 Local damage in the articular tissue limits joint mobility and can cause joint destruction while the generalised bone loss puts patients with RA at a higher risk of bone fracture.10 Osteoclasts, which are the only cells in the body capable of bone degradation, are primarily responsible for all types of RA-associated bone erosion.12 13 Similar to inflammation, osteoclast mediated bone resorption is a highly energy-demanding process that is sensitive to metabolic modification. Accordingly, osteoclast differentiation is associated with active metabolic reprogramming that involves enhanced mitochondrial biogenesis as well as increased oxygen consumption and glycolytic activity.14 15 Although the decisive role of metabolic reprogramming for osteoclast development and function has been established, is it still unclear how these dynamic changes are coordinated and translated into specific cellular functions. Moreover, the metabolic profile of osteoclasts and osteoclast precursors (OCPs) in the presence of RA is largely uncharacterised.

The mitochondrial enzyme aconitate decarboxylase 1 (Acod1, also known as immune responsive gene 1 (Irg1)) serves as a mediator between the metabolic condition and the functional state of different types of cells (eg, macrophages and monocytes). It catalyses the production of the tricarboxylic acid cycle intermediate itaconate and is induced in response to numerous stimuli (eg, lipopolysaccharide (LPS) or CpG DNA) that are related to stress, inflammation, infection or tissue damage.16–18 Itaconate has emerged as a multifunctional immunoregulatory metabolite with primarily anti-inflammatory activity. Its functional mechanisms include the stabilisation of the immunosuppressive transcription factor nuclear factor erythroid 2-like 2 (Nfe2l2, also known as Nrf2), the suppression of aerobic glycolysis through inactivation of glyceraldehyde 3-phosphate dehydrogenase (Gapdh) as well as competitive inhibition of the enzyme succinate dehydrogenase (Sdh).19–22 More recently, itaconate was also suggested to act as a negative regulator of osteoclast development. 4-octyl-itaconate (4-OI), a cell permeable derivative of itaconate, enhanced Nrf2 activity by inhibition of the E3 ubiquitin ligase Hrd1 in bone marrow-derived macrophages (BMDMs) and was able to suppress osteoclast differentiation in vitro and ovariectomy-induced bone loss in vivo.23 However, the link between the Acod1-itaconate axis and the metabolic state of osteoclasts as well as their influence on bone erosion in inflammatory arthritis is unknown. The aim of this study was, therefore, the analysis of Acod1-dependent reprogramming of osteoclast metabolism in the context of bone loss in inflammatory arthritis.

MethodsHuman samples

Peripheral blood mononuclear cells (PBMCs) were taken from patients with RA that fulfil the 2010 defined RA classification criteria of the American College of Rheumatology/European League Against Rheumatism.24 The following analyses were carried out: (1) osteoclast-differentiation assays with tartrate-resistant acid phosphatase (TRAP) staining in 11 patients with RA (8 females, 3 males; mean±SD age: 55±10 years) and 10 healthy controls (8 females, 2 males; mean±SD age: 51±14 years); (2) RNA extraction and qPCR in 21 patients with RA (all females; mean±SD age: 57±10 years) and 15 healthy subjects (13 females, 2 males; mean±SD age: 48±6 years); (3) extracellular flux assays in 3 patients with RA (1 female, 2 males; mean±SD age: 49±7 years) and 4 healthy subjects (3 females, 1 male; mean±SD age: 32±22 years); (4) itaconate measurement using mass spectrometry in 21 patients with RA (donors were female; mean±SD age: 57±11 years) and 10 healthy donors (donors were female; mean±SD age: 47±5 years). Disease activity of RA was measured by DAS28. Healthy controls and RA patients provided written informed consent for the study.

Human osteoclast in vitro differentiation

PBMCs were isolated from EDTA blood of healthy donors and patients with RA using a Ficoll gradient (Lymphoflot, Bio-Rad). Cells were plated at a cell density of 3×106/mL in adhesion medium, composed of αMEM+GlutaMAX with 1% fetal calf serum (FCS) and 1% penicillin/streptomycin (PS) and incubated at 37°C and 5.5% CO2 for 1.5 hours to purify monocytes by plastic adhesion. Non-adherent cells were subsequently removed together with the supernatant and the adherent cells were washed and supplemented with OC-medium (αMEM+GlutaMax with 10% FCS and 1% PS), containing 30 ng/mL macrophage colony-stimulating factor (M-CSF), 2 ng/mL receptor activator of nuclear factor kB ligand (RANKL) and 1 ng/mL transforming growth factor-β (TGF-β) (all Peprotech) to a cell density of 1.5×106/mL. Cells were cultured at 37°C and 5.5% CO2 for 7–10 days until maturation. The assay was terminated on the same day for each pair of matched donors. Medium was exchanged on day 3, day 5 and day 7 of cell culture. Osteoclast differentiation was evaluated by TRAP staining using the acid phosphatase leucocyte kit (Sigma-Aldrich). Stimulation with indicated doses of 4-OI (Cayman Chemical) started on day 3 of osteoclast cell culture and was repeated with each medium exchange. Images were acquired with the All-in-One Fluorescence Microscope BZ-X710 (KEYENCE) and quantification of osteoclast number was performed with ImageJ.

Extracellular flux analyses with human cells

4×105 PBMCs were plated in Seahorse XF96 Cell Culture Microplates (Agilent Technologies) in 200 µL of adhesion medium and incubated at 37°C and 5.5% CO2 for 1.5 hours to purify monocytes by plastic adhesion. Non-adherent cells were subsequently removed and the adherent cells were washed and supplemented with 200 µL of OC-medium for osteoclast differentiation. For 4-OI stimulation, 50 µM 4-OI was added on day 2 of cell culture for 24 hours. Glycolysis and Cell Mito Stress Tests were performed on day 3 of osteoclast in vitro differentiation and adenosine triphosphate (ATP) Rate Assay was performed on days 0–10 of cell culture according to manufacturer’s instructions. Seahorse XF RPMI Medium (Agilent Technologies) was supplemented with 2 mM L-glutamine (Agilent Technologies) for the Glycolysis Stress Test and with 10 mM glucose, 1 mM pyruvate and 2 mM L-glutamine (Agilent Technologies) for the ATP Rate Assay as well as the Cell Mito Stress Test. The final concentrations of the utilised metabolic compounds were as follows: 10 mM glucose, 2 µM oligomycin A, 50 mM 2-deoxy-D-glucose (2-DG), 2 µM carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP), 1 µM rotenone and 1 µM antimycin A (all from Sigma-Aldrich). Extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) were measured in a 96-well XF Extracellular Flux Analyzer (Agilent Technologies) and data were obtained using the Seahorse Wave Desktop Software (Agilent Technologies). Changes in ECAR and in OCR values were used to calculate glycolytic and mitochondrial parameters via Microsoft Excel.

Mice

C57BL/6NJ-Acod1em1(IMPC)J/J mice (stock #029340) were initially purchased from Jackson Laboratories, bred with C57BL/6NRj mice (from Janvier Labs) and maintained in the local animal facility. Wild-type (WT) C57BL/6NRj mice were co-housed with C57BL/6NJ-Acod1em1(IMPC)J/J mice for at least 1 week prior to the start of experiments. Human TNF transgenic (hTNFtg) mice (strain Tg197 on C57BL/6 background) were previously described.25 Arthritis evaluation was performed twice a week. Littermates were used as controls for the hTNFtg experiments. All mice were housed in a temperature-controlled and humidity-controlled facility with free access to food and water.

K/BxN serum-induced arthritis

Mice aged 7–9 weeks were injected intraperitoneally with 150 µL pooled serum from adult, arthritic K/BxN mice as previously described.26 Development and progression of arthritis were monitored using a semiquantitative scoring system (0–4 per paw; maximum score of 16).27 Mice were sacrificed for ex vivo analyses on day 9 post K/BxN serum transfer.

4-OI in vivo treatment

For the K/BxN SIA model, 4-OI treatment was conducted on days 1, 3, 5 and 7 after K/BxN serum transfer via intraperitoneal injection of 1 mg of 4-OI. For the hTNFtg arthritis model, 1 mg of 4-OI was administered every third day over a total period of 21 days, starting at 7 weeks of age.

Histological analysis

Hind paws and long bones of mice were fixed overnight in 4% PFA at 4°C followed by decalcification in 14% EDTA for 14 days until bones were pliable. H&E and TRAP staining via the acid phosphatase leucocyte kit (Sigma-Aldrich) were performed on serial paraffin sections (2 µm) of the paw for the quantification of inflammation, bone erosion and osteoclast number. TRAP staining was additionally performed on serial paraffin sections (2 µm) of tibial bones for osteoclast detection. Histomorphometric analysis of inflammation area (I.Ar), erosion area (E.Ar), number of osteoclasts per bone perimeter (N.Oc/B.Pm), bone volume per total volume (BV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N), trabecular space (Tb.Sp) and osteoclast surface per bone surface (Oc.S/BS) was performed using an Axio Lab.A1 microscope (Carl Zeiss), equipped with a digital camera and image analysis system (OsteoMeasure, OsteoMetrics).

Micro-CT analysis

Long bones were fixed in 4% PFA overnight before the analyses. All micro-CT (µCT) imaging was performed using the cone-beam Desktop Micro Computer Tomograph µCT 40 (SCANCO Medical). The settings were optimised for calcified tissue visualisation at 55 kVp with a current of 145 µA and 200 ms integration time for 500 projections per 180° and an isotropic voxel size of 6.0 µm. The three-dimensional modelling of the bone was performed with optimised greyscale thresholds of the operating system Open VMS (SCANCO Medical).

Flow cytometry analysis

Ankles were cut into small pieces with scissors and digested with 1 mg/mL Collagenase A and 0.1 mg/mL DNaseI in RPMI medium (+10% FCS and 1% PS) at 37°C for 1 hour. Digested ankles were put through 40 µm cell strainers to obtain a single-cell suspension. For surface marker staining, cells were first incubated with anti-mouse CD16/32 antibody (Biolegend, clone 93) in 1×PBS for 5 min in the dark at 4°C, and then stained with the surface markers: CD45 PerCP Cy5.5 (Biolegend, 30-F11), CD45 EF780 (eBioscience, 30-F11), CD11b BV605 (Biolegend, M1/70), Ly6C PE-Cy7 (Biolegend, HK1.4), CD115 BV421 (Biolegend, AFS98), RANK PE (Biolegend, R12-31) in 1×PBS in the dark at 4°C for 20 min. For intracellular staining, cells were washed with 1×PBS and incubated in fixation/permeabilisation buffer (Foxp3/Transcription Factor Staining Buffer Set, eBioscience) for 20 min in the dark at room temperature (RT), followed by two washing steps in 1×permeabilisation buffer and a 10 min incubation in blocking solution (1×permeabilisation buffer+20% FCS) at RT for 10 min. Cells were then stained with antibodies for the intracellular markers: HIF1α PE (R&D, IC1935P) or Puromycin Alexa Fluor 488 (Sigma-Aldrich, 12D10) in blocking solution for 1 hour at 4°C in the dark. After washing with 1×permeabilisation buffer, cells were resuspended in FACS buffer (1×PBS with 2% FCS and 5 mM EDTA) for flow cytometric analyses. Flow cytometry was performed on the Cytoflex S flow cytometer (Beckman Coulter) and data were analysed by Kaluza V.2.1 (Beckman Coulter).

Single cell energetic metabolism by profiling translation inhibition

Cells isolated from ankle tissue were resuspended in HBSS and stimulated either with 100 mM 2-DG (Sigma-Aldrich) for 10 min at 37°C, with 20 µM oligomycin A (Sigma-Aldrich) for 5 min at 37°C or successively with both compounds. All cells were subsequently incubated with 10 µg/mL puromycin dihydrochloride (Sigma-Aldrich) in HBSS for 1 hour at 37°C. Cells were then washed in ice-cold 1×PBS and stained with antibodies for surface markers, followed by intracellular staining with an anti-puromycin antibody as described in the flow cytometry section.

Murine osteoclast in vitro differentiation

Total bone marrow cells from mice were isolated by flushing femur and tibia. The cells were plated overnight at 37°C, 5.5% CO2 in a 100×20 mm dish in OC medium (αMEM+GlutaMax with 10% FCS and 1% PS), containing 5 ng/mL M-CSF (Peprotech). The next day, non-adherent bone marrow-derived monocytes (BMMs) were collected, washed and plated at a cell density of 1×106 cells/mL and 37°C, 5.5% CO2 in OC medium with 20 ng/mL M-CSF and 10 ng/mL RANKL (Peprotech). Medium was exchanged every 2 days. On day 3, when osteoclasts were fully differentiated, cells were washed with PBS, fixed and stained using the acid phosphatase leucocyte kit (Sigma-Aldrich). 4-OI (Cayman Chemical) stimulation occurred on day 1 and 2 of osteoclast cell culture using the concentration indicated in the respective experiment. Stimulation with 0.5 mM N-acetyl-l-cysteine (NAC) (Abcam) was performed on days 1 and 2 of cell culture. Stimulation with 2 mM 2-DG (Sigma-Aldrich) or 5 µM cyanide m-chlorophenyl hydrazone (CCCP) (Sigma-Aldrich) was performed on day 1 of cell culture for 3 hours, followed by medium exchange. Stimulation with 20 mM dimethyl-malonate (DMM) (Sigma-Aldrich) was performed on day 1 and 2 of cell culture for 3 hours each day, followed by medium exchange. Images were acquired with the All-in-One Fluorescence Microscope BZ-X710 (KEYENCE) and quantification of osteoclast number was performed with ImageJ.

Bulk RNA sequencing

Total RNA was extracted on day 3 of osteoclast cell culture and purified with the RNeasy Mini Kit (QIAGEN) according to the manufacturer’s instructions. Whole transcriptome RNA sequencing (RNA-seq) was carried out by Novogene (UK) with a total amount of 1 µg RNA per sample. Raw paired-end reads were aligned to the reference genome (mm10) using STAR (V.2.5). HTSeq V.0.6.1 was used to count the read numbers mapped to each gene. Differential expression analysis between two groups (three biological replicates per group) was performed using the DESeq2 R package.

Transmission electron microscopy

BMM-derived in vitro differentiated osteoclasts were washed with 1×PBS and fixed with 2.5% glutaraldehyde (Carl Roth) in 0.1 M phosphate buffer for at least 48 hours. Thereupon, cells were post-fixed in 2% buffered osmium tetroxide (Carl Roth) dissolved in 0.1 M phosphate buffer (pH 7.4) for 2 hours and dehydrated in graded alcohol concentrations/propylene oxide and embedded in epoxy resin (Sigma-Aldrich) according to standard protocol. For orientation, 1 µm semithin sections (ultramicrotome, Reichert Ultracut E) were stained with toluidine blue. Ultrathin sections were stained with uranyl acetate and lead citrate and examined with a transmission electron microscope (EM 906E; Carl Zeiss Microscopy). Size and number of mitochondria were determined using ImageJ.

Extracellular flux analyses with murine cells

2×105 murine BMMs were plated in Seahorse XF96 Cell Culture Microplates (Agilent Technologies) in a volume of 200 µL and differentiated into osteoclasts as described in the respective section. Extracellular flux assays were performed on days 0–3 of murine in vitro cell culture according to manufacturer’s instructions and similarly to the description of the procedure in human cells.

Immunofluorescence staining

BMMs were plated in 24-well plates containing 12 mm circle cover slips (Thermo Fisher Scientific) at a cell density of 1×106 cells/mL and cultured under osteoclast differentiating conditions as described above. For detection of mitochondrial reactive oxygen species (ROS), cells were stained with MitoSOX Red (Invitrogen) on days 2 and 3 of cell differentiation, for measurement of mitochondrial membrane potential, cells were stained using the MitoProbe JC-1 Assay Kit (Thermo Fisher Scientific) on day 3 of cell culture and to visualise the F-actin ring formation, cells were stained using the F-Actin Visualisation Biochem Kit (Cytoskeleton) according to the respective manufacturer’s protocol. For Hif1α immunofluorescence staining, cells were fixed with 4% PFA, washed with 1×PBS and permeabilised using 0.1% TritonX-100 in 1×PBS. After washing with 1×PBS, cells were incubated with blocking buffer (PBS+1% BSA+0.3 M Glycine+2% goat serum) for 1 hour at RT followed by incubation with the primary antibody solution (HIF1α polyclonal antibody, Cayman Chemical) in blocking buffer at 4°C overnight. After washing, cells were incubated with the secondary antibody solution (goat-anti-rabbit IgG DyLight 650, Invitrogen) and Phalloidin-iFluor 555 (Cytoskeleton) in blocking buffer for 45 min at RT in the dark. After the staining, the circle coverslips were transferred onto microscope slides, mounted with Fluoroshield with DAPI (Sigma-Aldrich) and covered with coverslips. Images were acquired with the All-in-One Fluorescence Microscope BZ-X710 (KEYENCE) for the mitochondrial ROS, mitochondrial membrane potential and F-Actin staining and using the THUNDER Imager (Leica Microsystems) and the LAS X Software for the Hif1α staining. The generated images were processed using the Imaris (Oxford Instruments) and ImageJ software, respectively.

Sdh activity assay

BMMs were plated in 24-well plates at a cell density of 1×106 cells/well in 1 mL and cultured under osteoclast differentiating conditions as described above. On day 3 of cell differentiation, cells were harvested and homogenised using the Sdh Assay Buffer followed by the measurement of Sdh activity via the Succinate Dehydrogenase Activity Colorimetric Assay Kit according to manufacturer’s instructions (BioVision).

CRISPR/Cas9-mediated Hif1a gene editing

Bone marrow cells from murine femur and tibia were plated overnight at 37°C, 5.5% CO2 in a 100×20 mm dish in OC medium (αMEM+GlutaMAX with 10% FCS and 1% PS), supplemented with 5 ng/mL M-CSF (Peprotech). The next day, non-adherent BMMs were collected and resuspended in primary nucleofection buffer P3 (Lonza) at a concentration of 1×106 cells/ 20 µL. The cell suspension was combined with a reaction mixture containing 12.5 µg of Alt-R S.p. Cas9 Nuclease V3 (IDT) and 0.04 nmol predesigned Alt-R CRISPR-Cas9 single guide RNA for Hif1a (Mm.Cas9HIF1A.1AB, IDT) or Alt-R CRISPR-Cas9 Negative Control crRNA (IDT) in a total volume of 5 µL per 1×106 cells. The mixture was then transferred into cuvettes (P3 primary Cell 4D-Nucleofector X Kit (Lonza)) and nucleofected via the Amaxa 4D Nucleofector (Lonza) using the code CM-137. Immediately after the nucleofection, cells were supplemented with pre-warmed, antibiotic-free αMEM medium and incubated at 37°C for 5 min. The BMMs were then adjusted to a cell density of 2×106 cells/mL using OC medium with 20 ng/mL M-CSF and 10 ng/mL RANKL (Peprotech) for osteoclast differentiation and plated into 96-well plates at 200 µL/well for TRAP staining and extracellular flux analysis as described in previous sections and into 48-well plates at 500 µL/well for genomic DNA and RNA extraction.

RNA extraction and real-time PCR for human and murine cells

RNA from in vitro cultivated cells was extracted using RNA-Solv Reagenz (VWR Peqlab) according to the manufacturer’s instructions. Extracted RNA was freed from genomic DNA using DNase I Kit (Thermo Scientific) and reversely transcribed into cDNA using high capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time PCR was performed using Takyon ROX SYBR 2X MasterMix dTTP blue (Eurogentec) on CFX96TM Real-Time System (Bio-Rad) with primers listed in online supplemental tables 1 and 2. Gene expression was normalised against B2M for human cells and Actb1 for murine cells. All primers were purchased from Invitrogen.

Statistical analysis

All statistical analyses were performed using Graph-Pad Prism Software V.9. Data were presented as mean±SEM. Statistical significance was calculated by two-tailed Student’s t-test for two-group comparison and one-way analysis of variance (ANOVA) or two-way ANOVA for multiple comparisons. Statistical details (eg, number of samples/group, number of independent experiments) can be found in the figure legends. P values less than 0.05 were considered statistically significant: *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001.

ResultsItaconate suppresses glycolytic activity and in vitro differentiation of osteoclasts from patients with RA

To examine the connection between Acod1 and RA in human osteoclasts, we collected PBMCs from a group of patients with RA as well as from healthy control donors (HD) and differentiated them into mature osteoclasts. We noticed that the amount of polynucleated TRAP-positive cells, obtained from HD donors was inversely correlated with the mRNA expression level of ACOD1 in freshly isolated PBMCs before the onset of osteoclast differentiation, supporting the idea that Acod1 acts as a negative regulator of osteoclast development under healthy conditions (figure 1A). Notably, ACOD1 mRNA expression was reduced in the PBMCs of patients with active RA as compared with healthy donors while no such downregulation was observed in patients with RA, who were in remission (online supplemental figure 1A). The analysis of itaconate levels in PBMC samples by mass spectrometry revealed no significant difference between healthy donors and active RA patients or RA patients in remission (online supplemental figure 1B). However, itaconate levels were correlated with ACOD1 mRNA expression (online supplemental figure 1C) and inversely correlated with RA disease activity (online supplemental figure 1D).

Figure 1

Itaconate inhibits glycolytic activity and differentiation of human osteoclasts. (A) Correlation analysis between the mRNA expression level of aconitate decarboxylase 1 (ACOD1) in peripheral blood mononuclear cells (PBMCs) from healthy donors (HD), measured before the onset of osteoclast in vitro differentiation (day 0), and the numbers of polynucleated osteoclasts (≥3 nuclei) from the same donors at the end of osteoclast in vitro differentiation (days 7–10) (n=8). (B) Representative pictures and (C) fold change quantification of tartrate-resistant acid phosphatase (TRAP)-positive, polynucleated osteoclasts (≥3 nuclei), differentiated from HD or rheumatoid arthritis (RA) patient-derived PBMCs in the presence or absence of 50 µM of 4-octyl-itaconate (4-OI) (n=10–11). Scale bar: 200 µm. (D) Percentage of mitochondrial versus glycolytic adenosine triphosphate (ATP) production, determined via ATP rate assay and measured on days 0, 3, 6 and 10 of osteoclast in vitro differentiation in cells derived from HD and RA patients (n=4–6 cell replicates from one RA patient and HD each). (E) Percentage of mitochondrial and glycolytic ATP production in cells derived from HD versus RA patients, determined via ATP rate assay and measured on days 0, 3, 6 and 10 of osteoclast cell culture (n=4–6 cell replicates from one RA patient and one HD). (F) Glycolytic activity parameters and (G) extracellular acidification rate (ECAR) profile plot, determined by glycolytic stress test assay on day 3 of HD and RA patient-derived human osteoclast cell culture after the stimulation with 50 µM 4-OI for 24 hours as compared with unstimulated controls (representative of two experiments, each with five cell replicates from one donor per group). (H) Mitochondrial activity parameters and (I) oxygen consumption rate (OCR) profile plot, determined by mitochondrial stress test assay on day 3 of human osteoclast cell culture with HD and RA patient cells after the stimulation with 50 µM 4-OI for 24 hours as compared with unstimulated controls (representative of two experiments, each with five cell replicates from one donor per group). (J) Percentage of mitochondrial vs glycolytic ATP production, determined via ATP rate assay on day 3 of human osteoclast in vitro differentiation using HD cells, stimulated with 50 µM 4-OI for 24 hours as compared with unstimulated cells (representative of two experiments, each with five cell replicates from one donor). (K) Fold change mRNA expression level of glycolytic genes measured on day 4 of HD osteoclast in vitro differentiation after the stimulation with 50 µM 4-OI for 24 hours in comparison to unstimulated cells (n=4 cell replicates). Data are shown as mean±SEM. Symbols represent individual donors. P values were determined by two-tailed Student’s t-test for single comparisons and two-way ANOVA for multiple comparisons. Correlations were tested using the linear regression F test. ANOVA, analysis of variance.

We then stimulated PBMCs from RA and healthy donors with the cell-permeable itaconate derivative 4-OI during osteoclast cell culture, observing that RA-derived PBMCs show higher osteoclast differentiation potential than PBMCs from HD controls, as illustrated by increased numbers of polynucleated TRAP-positive cells (figure 1B,C). 4-OI administration, on the other hand, suppressed osteoclast differentiation from RA-derived and HD-derived PBMCs in a dose-dependent manner (figure 1B,C, online supplemental figure 1E,F) and reduced mRNA expression levels of the osteoclast differentiation markers nuclear factor of activated T-cells cytoplasmic 1 (NFATC1), TRAP 5 (ACP5), matrix metallopeptidase 9 (MMP9) and cathepsin K (CTSK) (online supplemental figure 1G).

Subsequently, we aimed to evaluate the connection between the antiosteoclastogenic properties of 4-OI and a potential modulation of osteoclast metabolism. We, therefore, performed ATP rate assay analyses with PBMCs from patients with RA and HD controls at different time points throughout osteoclast in vitro differentiation. Here, we discovered that osteoclast development was associated with a gradual shift towards glycolytic ATP production. While osteoclast progenitor cells (day 0) predominantly produced ATP via oxidative phosphorylation (OXPHOS), mature osteoclasts (day 10) derived the majority of their ATP from glycolysis. This metabolic transition occurred earlier and was more pronounced in RA-derived osteoclasts as compared with osteoclasts from HD controls (figure 1D,E). To test the effect of 4-OI on the metabolic profile of developing osteoclasts, we treated the cells with 50 µM of 4-OI for 24 hours, followed by glycolytic and mitochondrial stress test assays on day 3 of osteoclast cell culture. RA-derived OCPs showed an increased glycolytic activity and reduced mitochondrial respiration as compared with HD-derived cells in the absence of 4-OI while 4-OI stimulation inhibited glycolytic activity in RA-derived cells (figure 1F–I). Using an ATP rate assay as well as qPCR analysis of the glycolytic genes solute carrier family 2 member 1 (SLC2A1), aldolase, fructose-bisphosphate C (ALDOC), phosphoglycerate kinase 1 (PGK1), enolase 1 (ENO1), lactate dehydrogenase A (LDHA) and pyruvate dehydrogenase kinase 1 (PDK1), we confirmed the ability of 4-OI to inhibit glycolytic ATP production in HD OCPs (figure 1J,K). Together, our data indicate that RA-associated osteoclasts strongly rely on glycolytic energy production while itaconate suppresses the metabolic shift towards glycolysis in human OCPs and inhibits osteoclast maturation.

Acod1-deficiency exacerbates arthritis and arthritis-associated bone loss in mice

To further analyse the importance of Acod1 for osteoclast metabolism in the context of inflammatory arthritis, we subjected WT and Acod1-deficient mice to the K/BxN serum-induced arthritis model (SIA).26 Acod1−/− mice developed stronger inflammation, as illustrated by increased arthritis score, paw swelling and H&E-stained inflammation area (figure 2A–C; online supplemental figure 2A). Along with that, we observed a local increase in osteoclast numbers and bone erosion within the affected joints of Acod1−/− mice (figure 2A,C) as well as upregulated expression of the osteoclast-associated genes tumour necrosis factor receptor superfamily, member 11 a, NFKB activator (Tnfrsf11a), Nfatc1, Acp5, Ctsk and Mmp9 (online supplemental figure 2B). To find out, if the increase in osteoclast numbers was due to enhanced infiltration by OCPs, we quantified the amount of CD45+CD11b+Ly6ChiCD115+ cells within the synovial tissue of arthritic WT and Acod1−/− mice via flow cytometry. While the proportion of OCPs was significantly higher in WT mice with SIA as compared with naïve WT mice, Acod1-deficiency did not cause any further changes in OCP number (online supplemental figure 2C,D). However, we discovered that OCPs from the inflamed synovium of Acod1−/− mice expressed higher levels of receptor activator of NF-κB (Rank) protein than WT OCPs (figure 2D), suggesting that Acod1 controls inflammatory bone erosion through the regulation of osteoclast differentiation, rather than the recruitment of their progenitors. In order to examine the impact of Acod1-deficiency on systemic bone loss in SIA, we performed µCT on tibial bone as well as histomorphometry analyses on TRAP-stained tibial sections. Here, we found that arthritic Acod1−/− mice are characterised by systemic reduction of bone mass and an increased number of TRAP-positive osteoclasts (figure 2E,F). Naïve Acod1−/− mice also showed increased osteoclast numbers and osteoclast-covered bone surface when compared with naïve WT mice, despite a lack of difference in bone volume (online supplemental figure 2E,F). These findings demonstrate that Acod1 inhibits osteoclast development in vivo and protects against both local as well as systemic bone loss in experimental arthritis.

Figure 2

Acod1-deficiency exacerbates serum-induced arthritis in mice. (A) Representative images of hematoxylin and eosin (H&E) and tartrate-resistant acid phosphatase (TRAP)-stained paw sections from wild-type (WT) and aconitate decarboxylase 1-deficient (Acod1−/−) mice with K/BxN serum-induced arthritis (SIA). Black arrowheads (H&E staining) indicate inflammation, blue arrowheads (TRAP staining) indicate eroded bone tissue with osteoclasts. Scale bar: 200 µm. (B) Arthritis score and proportional paw swelling in WT and Acod1−/− mice during the course of SIA (n=21–23). (C) Quantification of inflammation area (I.Ar), eroded bone area (E.Ar) and osteoclast number per bone perimeter (N.Oc/B.Pm) in H&E and TRAP-stained paw tissue sections of WT and Acod1−/− mice with SIA (n=9–13). (D) Flow cytometry analysis of receptor activator of NF-κB (Rank) expression in CD45+CD11b+Ly6ChiCD115+ osteoclast precursors (OCPs) in the synovial ankle tissue of WT and Acod1−/− mice with SIA, depicted as fold changes of the mean fluorescence intensity (MFI) in Acod1−/− as compared with WT cells (n=9–10). (E) Representative images of TRAP staining and micro-CT (µCT) analysis of tibiae from naïve and SIA affected WT as well as SIA affected Acod1−/− mice. White arrowheads indicate osteoclasts. Scale bar TRAP: 50 µm. Scale bar µCT: 200 µm. (F) Quantification of bone volume per total volume (BV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N), trabecular space (Tb.Sp), number of osteoclasts per bone perimeter (N.Oc/B.Pm) and osteoclast surface per bone surface (Oc.S/BS) (n=11–23). (G) SCENITH analysis of metabolic parameters in OCPs from WT mice with and without SIA and Acod1−/− mice with SIA (n=4–5). Data are shown as mean±SEM. Symbols represent individual mice. P values were determined by two-tailed Student’s t-test for single comparisons and one-way or two-way ANOVA for multiple comparisons. ANOVA, analysis of variance.

Our human in vitro analyses indicated that mature osteoclasts, derived from the PBMCs of patients with RA, exhibit a greater dependence on glycolysis for ATP production when compared with osteoclasts from healthy donors. In order to investigate, if similar metabolic deviations are present in our experimental arthritis model and whether these changes are linked to Acod1, we analysed the metabolic profile of synovial OCPs, obtained from naïve or arthritic WT and Acod1−/− mice using the SCENITH technique in combination with an antibody panel for OCP surface markers (CD45+CD11b+Ly6ChiCD115+).28 In accordance with our human in vitro data, we discovered that arthritis induction results in increased glycolytic capacity and reduced mitochondrial dependence in WT OCPs. Acod1−/− OCPs from SIA mice exhibited an even higher level of glycolytic capacity and further reduction of mitochondrial dependence as compared with WT SIA OCPs (figure 2G, online supplemental figure 2G,H). The levels of glucose dependence as well as the capacity for fatty acid and amino acid oxidation (FAO and AAO) on the other hand remained mostly similar in WT naïve, WT SIA and Acod1−/− SIA OCPs, indicating that arthritis induction and Acod1-deficiency have little influence on the preferred energy source that is used for ATP generation in synovial OCPs (figure 2G). Our results, therefore, suggest that SIA causes a metabolic reprogramming of OCPs towards glycolytic ATP production while Acod1 acts as a negative regulator of this process.

Acod1-dependent suppression of osteoclast differentiation is accompanied by an alteration of the osteoclast transcriptional network

To delineate the mechanism behind the influence of Acod1 on osteoclast development and metabolic activity, we performed osteoclast in vitro differentiation assays with BMMs, isolated from WT and Acod1−/− mice. Acod1-deficiency resulted in enhanced osteoclast differentiation, as demonstrated by the elevated number of TRAP-positive polynucleated cells, increased actin-ring formation as well as an upregulated mRNA expression of osteoclast-related genes (figure 3A–C, online supplemental figure 3A). We presumed that the cellular phenotype of Acod1−/− osteoclasts is caused by itaconate-deficiency (online supplemental figure 3B). In accordance, the addition of 4-OI to the cell culture abolished the pro-osteoclastogenic effect of Acod1-deficiency and inhibited the ability of WT BMMs to differentiate into mature osteoclasts in a dose-dependent manner, without affecting cell viability (figure 3A–C, online supplemental figure 3C–F). Furthermore, through mass spectrometry analysis, we detected 4-OI within the cell lysates of 4-OI stimulated osteoclasts, thus confirming its capacity to enter the cells (online supplemental figure 3G).

Figure 3

Acod1-deficiency alters metabolic gene expression and enhances murine osteoclastogenesis. (A) Representative images and (B) quantification of tartrate-resistant acid phosphatase (TRAP)-stained polynucleated (≥5 nuclei) osteoclasts that were differentiated from bone marrow monocytes of wild-type (WT) and aconitate decarboxylase 1-deficient (Acod1−/−) mice that were either unstimulated or stimulated with 100 µM 4-octyl-itaconate (4-OI) for 48 hours (n=10–11). Scale bar: 100 µm. (C) Fold change expression levels of osteoclast-related genes in Acod1−/− osteoclasts (day 3 of cell culture) that were either unstimulated or stimulated with 100 µM 4-OI for 48 hours as compared with unstimulated WT osteoclasts (n=3–13). (D) Venn diagram of differentially expressed genes in unstimulated Acod1−/− osteoclasts as compared with unstimulated WT osteoclasts (Acod1−/− vs WT) and in 4-OI stimulated Acod1−/− osteoclasts as compared with unstimulated Acod1−/− osteoclasts (Acod1−/− 4-OI vs Acod1−/−) following an RNA sequencing and differential gene expression analysis using the DESeq2 method with a p value cut-off of 0.05. (E) Over-representation analysis (ORA) of gene ontology (GO) biological processes using the overlapping set of genes that were differentially expressed in the comparison Acod1−/− vs WT as well as in the comparison Acod1−/− 4-OI vs Acod1−/−. The graph shows a lollipop chart of overrepresented GO terms with a p value cut-off of 0.05. The position on the x-axis gives the enrichment factor and the dot size gives the number of genes present in each ontology. A list of the enriched pathways with their related genes for the GO ORA analysis is attached as online supplemental file 2 (ard-2023–2 24 774_RNA-Seq_Enriched pathways). (F) Gene set enrichment analysis (GSEA) for GO and Kyoto Encyclopaedia of Genes and Genomes (KEGG) enrichment in Acod1−/− osteoclasts as compared with WT osteoclasts with a p value cut-off of 0.05 and visualisation of the hallmark terms ‘osteoclast differentiation’, ‘hypoxia’ and ‘reactive oxygen species (ROS) metabolic process’. (G) Heatmap illustrating the H-clusters of differentially expressed genes that were assigned to the GO terms ‘osteoclast differentiation’, ‘response to hypoxia’ and ‘response to ROS’ within the groups WT, Acod1−/− and Acod1−/− 4-OI. A gene list with the corresponding fold change data and statistical values is attached as online supplemental file 3 (ard-2023–2 24 774_RNA-Seq_Heatmap_Genelist). Data are shown as mean±SEM. Symbols represent individual mice. P values were determined by one-way ANOVA for multiple comparisons. ANOVA, analysis of variance.

To ascertain the influence of the Acod1-itaconate axis on the genomic transcriptional network of osteoclasts, we performed a whole transcriptome RNAseq analysis with fully differentiated osteoclasts fr