4-Octyl itaconate inhibits synovial fibroblast activation by targeting Nrf2 to exert anti-inflammatory effects in rheumatoid arthritis

Abstract

Objectives

Fibroblast-like synoviocytes (FLS) are pivotal for the expansive growth and invasiveness of rheumatoid synovitis, and patients with a fibroblastic-rich pathotype respond poorly to currently approved antirheumatic drugs. The aim of this project was to examine the anti-inflammatory effects of the nuclear factor erythroid 2–related factor 2 (Nrf2) activator 4-octyl itaconate (4-OI) on inflammatory rheumatoid arthritis (RA)-FLS–dominated pathology.

Methods

Rheumatoid arthritis fibroblast-like synoviocytes (RA-FLS) and autologous cocultures of RA-FLS and peripheral blood mononuclear cells were cultured with 4-OI, corticosteroids or antitumor necrosis factor alpha antibodies and analysed for inflammatory activity with Western blot and enzyme-linked immunosorbent assay. 4-OI–induced transcription factor Nrf2 was genetically ablated in immortalised FLS using Clustered Regurlarly Interspaced Short Repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) (CRISPR/Cas9) gene editing.

In vivo, 4-OI was administered intraarticularly in the monoarticular methylated bovine serum albumin antigen-induced arthritis model and intraperitoneally in the collagen-induced arthritis, with subsequent clinical evaluation.

Results

In RA-FLS in both monocultures and autolog cocultures, 4-OI treatment mediated an increase in gene expression of the Nrf2-regulated protein heme oxygenase-1 and NAD(P)H quinone dehydrogenase 1 with a concomitant decrease in monocyte chemoattractant protein-1 and specifically in monocultures also of interleukin 6 secretion. Silencing of Nrf2 reduced the effect mediated by 4-OI in RA-FLS. In the methylated bovine serum albumin mouse model, local administration of 4-OI significantly reduced joint swelling. In the collagen-induced arthritis model, a similar trend was seen accompanied by a significant reduction of interleukin 6 in serum.

Conclusions

4-OI ameliorates inflammation in models of RA, particularly in FLS-driven disease pathology. Overall, human in vitro data and murine in vivo data provide mechanistic insights into the role of 4-OI–induced activation of Nrf2 directly reducing the activation of inflammatory FLS in RA.

Graphical Abstract

Key message.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• •4-octyl itaconate is a known potent inducer of nuclear factor erythroid 2–related factor 2 (Nrf2) and serves as an immunoregulator in various disease models.

WHAT THIS STUDY ADDS

• •These findings add novel insight into the mechanistic effects of 4-octyl itaconate and Nrf2 activation in rheumatoid arthritis, directly targeting stromal fibroblast-like synoviocytes.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• •This study explores the possibility of therapeutically targeting the Nrf2 pathway also in rheumatoid arthritis stromal-dominated pathobiology.

Alt-text: Unlabelled box dummy alt text

INTRODUCTION

Rheumatoid arthritis (RA) is a common autoimmune disease characterised by joint inflammation that, if uncontrolled, can cause irreversible structural damage. In RA, the synovial membrane makes an invasive extracellular expansion, forming a pannus characterised by a deprived oxygen supply and an abundant cell infiltration of mainly fibroblasts, macrophages and lymphocytes [[1], [2], [3]]. The treatment and prognosis of RA have greatly improved by introducing biological and small-molecule immunosuppressive therapies predominantly targeting leukocytes and their cytokine production [4]. However, despite advances in treatment, a significant proportion of patients with RA fail to achieve sustained remission [5]. Thus, there is a continued need for elucidating the local modulators of the inflamed microenvironment to identify potential new treatment targets and modalities [6,7].

A large proportion of treatment nonresponders have a fibroblast-dominated local pathology [8]. This subtype of RA is characterised by being significantly driven by pathological fibroblasts in the synovial environment [9]. Currently, no available treatment specifically targets fibroblast-like synoviocytes (FLS) [8,9]. In the inflamed joint, a local imbalance between the cellular generation of reactive oxygen species and antioxidative capacity results in increased oxidative stress [6,10,11]. The synovial oxygen-deprived microenvironment further activates inflammatory rheumatoid arthritis fibroblast-like synoviocytes (RA-FLS) [12]. Inflammatory RA-FLS play a central role in the initiation and persistence of both synovitis and joint damage [13]. RA-FLS undergo phenotypic and metabolic changes leading to uncontrolled local proliferation, cytokine production and invasion of cartilage and bone [14]. Among the RA-FLS subsets, the PDPN+THY1+ sublining layer RA-FLS exhibits a robust pathogenic phenotype [9,15]. These specific inflammatory RA-FLS (PDPN+THY1+) are implicated in increased cytokine production, cellular infiltration and activation and subsequently accelerate joint destruction [[14], [15], [16]].

The altered metabolic pathways are considered a promising novel target for immune modulations in RA with a potential for direct effects in both leukocytes and stromal cells [17,18].

Targeting Nrf2 activation has gathered particular interest due to a growing number of reports supporting a significant role for Nrf2 in modulating inflammatory diseases [[19], [20], [21]]. Nrf2 is a transcription factor, which, under homeostatic settings, is inhibited by Kelch-like ECH-associated protein 1 (Keap1). During oxidative stress, Nrf2 is released from Keap1 binding and can translocate to the nucleus, where it mediates anti-inflammatory and antioxidative properties through multiple mechanisms, including the activation of heme oxygenase-1 (HO-1) and NAD(P)H-dehydrogenase [quinon] 1 (NQO1) [22]. Furthermore, Nrf2 activation mediates the downregulation of inflammatory cytokines such as monocyte chemoattractant protein-1 (MCP-1), interleukin 6 (IL-6) and tumour necrosis factor alpha (TNF-⍺) in immune cells like macrophages [[23], [24], [25], [26]]. Nrf2 activation can be achieved through increased levels of the endogenous tricarboxyclic acid (TCA) metabolite derivative, itaconate [27]. An esterified version of itaconate, 4-octyl itaconate (4-OI), has shown greater immunosuppressing potency than the endogenous molecule in inflammatory preclinical models [28].

It has previously been shown that Nrf2 deficiency can lead to aggravation of arthritis [29]. In RA, insufficient Nrf2 activity increases disease activity and accelerates progression [29,30]. Additionally, knockdown of Keap1 improved arthritis in mice [[30], [31], [32]]. Furthermore, intra-articular (i.a.) treatment with the Nrf2 activating tricarboxylic acid cycle derivative itaconate has recently been shown to reduce local arthritis in rats [33].

The effects of Nrf2 activation on macrophages in inflammatory diseases are well established [27,34]. Furthermore, it has been suggested that insufficient Nrf2 activation in inflammatory RA-FLS may enhance their capacity to proliferate and invade cartilage and bone [35]. Complementary treatment with itaconate reduces the activity of inflammatory RA-FLS in vivo and in vitro [33].

However, a better understanding of the possible effects of Nrf2 activation directly in inflammatory stromal cells like FLS in RA is needed. The aim of this study was to evaluate the capacity of Nrf2 activation following 4-OI treatment to suppress inflammation in stromal-dominated RA pathobiology.

METHODS

Patients and samples

A cross-sectional, paired set of peripheral blood mononuclear cells (PBMCs) and synovial fluid mononuclear cells (SFMCs), along with synovial fluid (SF) and plasma (P), was obtained from individual patients with active RA at the Department of Rheumatology, Aarhus University Hospital, from the INflammation in ARThritis biobank (n = 13). The mean Disease Activity Score 28CRP was 4.0 (95% CI [2.9, 5.0]) at the time of inclusion. Additional data on included patients can be found in Supplementary Table S1.

Treatment for the RA patients in this study adhered to the established treatment guidelines for RA [36,37]. All RA patients included in the study were over 18 years of age and met the European Alliance of Associations for Rheumatology/American College of Rheumatology 2010 classification criteria for RA [38]. SFMCs and PBMCs were isolated by conventional Ficoll-Paque (GE Healthcare) density-gradient centrifugation and cryopreserved at −135°C.

Immortalised FLS (^IMFLS) were kindly provided by Hermann et al [39] and cryopreserved at −135°C. Not all patient samples were used in all experiments. The exact number of patients in each experiment has been stated in the figure legends.

The pathobiology of early arthritis cohort

Ribonucleic acid (RNA)-sequencing data from the pathobiology of early arthritis cohort (PEAC), accessible on the website (https://peac.hpc.qmul.ac.uk/), were analysed. The data, deposited by Christopher John and Myles Lewis from Queen Mary University of London, can be found on ArrayExpress under Accession code E-MTAB-614 [8].

Culture conditions

Four-octyl-itaconate (4-OI) was chemically synthetised by Thomas B. Poulsen (Aarhus University, Denmark) as published before [40]. The compound was dissolved in dimethyl sulfoxide (DMSO) before addition to cell cultures. A concentration of 125 µM of 4-OI was utilised based on titrations, cell viability and previous reports [27]. Untreated cells and cells treated with DMSO served as controls. Adalimumab (Humira, AbbVie) at 5 µg/mL as previously described [41] was included as a leukocyte-targeted comparator, and hydrocortisone (Solu-Cortef, Pfizer Inc.) 1000 ng/mL was used as a control. The cells were incubated at 37°C in a humidified incubator containing 5% CO₂ in air for a duration of 48 hours. Afterwards, the cells and supernatants were harvested through a 2-step centrifugation process (1200 rpm/10 min/room temperature). The cells and the cell-free supernatant were separately stored at −80°C for subsequent analysis. DMSO was used as a vehicle control in vitro and cyclodextrin in vivo for the 4-OI experiments.

SFMC in vitro model of RA

The RA 48-hour model using SFMCs is characterised by the presence of all cells from the synovium, mainly lymphocytes and monocytes [42]. The SFMCs were thawed and then cultivated in Dulbecco’s modified Eagle’s medium (DMEM), which was further enriched with 10% foetal calf serum (FCS), penicillin, streptomycin, glutamine and 10 mM HEPES (Gibco). This medium is referred to as “DMEM complete,” and the cell density used for cultivation was 1 × 10⁶ cells per millilitre.

FLS in vitro model of RA

The FLS 48-hour monoculture of RA is a model of inflammatory RA fibroblast without leukocytes. RA-FLS were grown from SFMCs, changing medium every 3-4 days. When the cell layer was >70% confluent, the FLS were passaged by trypsin/EDTA treatment and used for analyses at passages 3 to 5. The FLS were then seeded as 6 × 10⁵ cells/mL in DMEM complete culture medium in a 12-well plate. When the cell layer was >70% confluent, the culture medium was refreshed, and the compounds of interest were introduced, allowing for a 48-hour incubation period [43].

FLS+PBMC in vitro model of RA

The FLS+PBMC in vitro model is a crosstalk model of pathological fibroblasts and autologous leukocytes. To establish this model, FLS monocultures were seeded at a concentration of 2 × 10⁴ cells per millilitre in DMEM complete culture medium in a 96-well plate in passages 3 to 5. When the FLS cell layer was >70% confluent, paired autologous PBMCs were added as 7.5 × 10⁵ cells/mL and cocultured for 48 hours in DMEM complete culture medium [42].

CRISPR-Cas9-mediated gene silencing of Nrf2 in immortalised human inflammatory environment–fibroblast-like synoviocyte cells

Immortalised RA-FLS (^IMRA-FLS) were derived from patients with RA as previously described [15]. ^IMRA-FLS Nrf2 gene-silenced cells were generated using electroporation of a single-guide RNA targeting Nrf2 together with recombinant Cas9 protein, following the overall guidelines from Synthego. Briefly, cells were detached using trypsin, and 2 × 10⁴ cells were combined with the RNP complex consisting of 6-µg Cas9 Nuclease V3 protein (ID technology) and 3.2-µg single-guide RNA (Synthego) in OptiMEM media (Gibco). Electroporation was performed using a 4D-Nucleofector (Lonza) program CM-138. After electroporation, cells were seeded in a 12-well plate in complete DMEM high-glucose DMEM supplemented with 10% heat-inactivated foetal bovine serum (Gibco), 2 mM l-glutamine (Gibco) and 1% penicillin/streptomycin (Gibco) and incubated in a cell incubator. The following day, the medium was replaced, and the cells were transferred to larger culture dishes as needed. Four days after electroporation, cells were harvested and used for further treatment as described below.

Sequences of the synthetic guides used are the following: Adeno-associated virus integration site 1 (the “neutral” editing site serving as control): G*G*G*GCCACUAGGGACAGGAU, NFE2L2: guide-1: A*U*U*UGAUUGACAUACUUUGG, guide-2: G*C*G*ACGGAAAGAGUAUGAGC. Treatment of the cells was performed for 20 hours, 4 days after electroporation with the CRISPR-Cas9 system. Treatment was stopped by harvesting the cells for RNA purification and Western blot analysis.

Real-time quantitative polymerase chain reaction analysis

Gene expression was investigated by real-time quantitative polymerase chain reaction (qPCR) analysis, using TaqMan detection systems (Applied Biosciences). RNA was isolated using the high-pure RNA isolation kit (Roche, 95211828665001) according to the manufacturer’s protocol. The quantity and quality were assessed using a NanoDrop One (Thermo Fisher Scientific). mRNA levels were analysed using the TaqMan RNA-to-Ct-1-Step kit, according to the manufacturer’s protocol (Applied Biosciences, Thermo Fisher Scientific). We used the following commercially available gene expression assays: TATA-box binding protein (TBP) Hs00427620, HMOX1 Hs01110250, AKR1B10 Hs00252524, NQO1 Hs01045993 and CCL2 Hs00234140. TBP was used as the housekeeping gene. Analysis was performed in biological duplicates, in 3 (NQO1, HMOX1, and AKR1B10) and 2 (CCL2) independent experiments.

SDS-PAGE and Western blotting

Immunoblotting analysis was performed as previously described [44]. In brief, cells were lysed in ice-cold Pierce RIPA lysis buffer (Thermo Fisher Scientific) supplemented with 10 mM NaF, 1 × protease inhibitors (Roche) and 5 IU/mL benzonase (Sigma Aldrich). The protein concentration was determined using the Pierce BCA protein assay kit (Thermo Fisher Scientific). Twenty micrograms of the cell lysate was denatured for 5 minutes at 95 °C in the presence of 1 × XT samples buffer and 1 × XT reducing agent (both BioRad). Separation was performed by SDS-PAGE on 4% to 20% Criterion Tris-Glycine eXtended (TGX) precast midi protein gels (BioRad). We used Precision Plus Dual Color Standards (BioRad) as a molecular weight marker. After separation, the proteins were transferred onto polyvinylidene diflouride (PVDF) membranes (BioRad) using the Trans-Blot-Turbo system (BioRad). After blocking the membranes using 5% skimmed milk in phosphate-buffered saline+0.05% Tween-20 (PBS-T), the membranes were probed overnight at 4 °C with 1:1000 dilutions (unless other dilution is stated) of the following antibodies (dilutions prepared in PBS-T): anti-Nrf2 (EP1808Y, abcam #62352), anti-TRXR1 (D1T3D, Cell Signaling Technology #15140), anti-NQO1 (D6H3A, Cell Signaling Technology #62262), anti-Beta-actin (D6A8, Cell Signaling Technology #8457), anti-HO-1 (D60G11, Cell Signaling Technology, #5853), anti-NQO1 (A180, Cell Signaling Technology, #3187) and anti-Vinculin (1:10,000 dilution, hVIN-1, Sigma, #V9131) or anti-Vinculin (E1E9V, Cell Signaling Technology, #13901). Membranes were washed and incubated for 1 hour at 4 °C with the horseradish peroxidase-conjugated antirabbit antibody (Jackson ImmunoResearch, 711-035-152) or antimouse antibody (Jackson ImmunoResearch, 715-036-150), diluted 1:10,000 in PBS-T. Blots were developed by the addition of SuperSignal West Pico PLUS chemiluminescent substrate (Thermo Fisher Scientific) and imaged using an iBright CL1500 Imager (Invitrogen) or an ImageQuant LAS4000mini (GE Healthcare).

Enzyme-linked immunosorbent assay

The concentrations of MCP-1 and IL-6 in the culture supernatants were determined using commercially available enzyme-linked immunosorbent assay (ELISA) kits, following the instructions provided by the manufacturer (Cat no. 438804, BioLegend, and Cat no. D6050, R&D). The optical density of the wells was measured at 450 nm using a Thermo Scientific Multiskan GO reader, with a reference reading taken at 570 nm.

Cell viability assay

Monocultures of FLS and cocultures of FLS and PBMCs were cultured as above for 48 hours in a 96-well plate with the addition of 4-OI and DMSO control. After 48 hours, the commercially available cell proliferation assay (MTT) (Cat. No. 11-465-007-001, Roche) was performed according to the manufacturer’s instructions. A Thermo Scientific Multiskan GO reader was used to acquire the optical density of the wells at 570 nm with a reference reading at 750 nm.

iCELLigence system

The real-time cell analysis (RTCA) setup included an RTCA impedance analyser, a computer with RTCA software for system control, a 16-well electronic plate (E-Plate) and the RTCA station containing E-Plates, which was kept in a tissue culture incubator. When cell culture medium is added and a low voltage is applied, an electric field forms between electrodes; cell presence within this field creates impedance.

Real-time cell–sensor impedance readings are displayed as the cell index (CI), a measure directly affected by cell attachment, spreading and proliferation. At each time point, CI is calculated as (Rn–Rb)/Rb, where Rn represents the impedance of the cell-containing well, and Rb represents the background impedance in a medium-only well.

For background measurement, 50 µL of cell culture medium was added. Following cell and 4-OI addition, the total volume reached 200 µL. E-Plates were then incubated at 37°C with 5% CO₂, with impedance readings taken by the RTCA system every 5 minutes for approximately 72 hours, either with or without 4-OI.

Flow cytometry

Supernatants were harvested, and the cell layer was treated with trypsin/EDTA. Cells were blocked with 0.1 mg/mL mouse IgG (Jackson ImmunoReseach) for 15 minutes and stained with HLA-DR BV 480 (BD Horizon #566113), CD90 BV 605 (BioLegend #328128), CD45RO FITC (BioLegend #304204), CD31 PE-Cy7 (BioLegend #303117), Podoplanin AF 647 (BioLegend #395003) and Live/dead nIR Dead Cell Stain Kit (Invitrogen #L10119) followed by fixation using 1% paraformaldehyde. Flow cytometry was performed the subsequent day using NovoCyte Quantoen 4025 (Agilent) with acquisition software NovoExpress (v. 1.6.2, Agilent) at the FACS Core Facility, Aarhus University, Denmark. Flow cytometry data were analysed through FlowJo 10.8.1. An example of a gating strategy is seen in Supplementary Figure S1.

Methylated bovine serum albumin -monoarthritis model

All mice used for methylated bovine serum albumin (mBSA) antigen-induced arthritis were male, wild-type C57Bl/6, aged 6 to 10 weeks. On day 14, mice were inoculated on either lateral side of the lower spine with two 50 µL subcutaneous injections containing a 1:1 emulsified ratio of mBSA (0.5 mg/mL) in complete Freund’s (incomplete Freund’s with 2 mg/mL Mycobacterium tuberculosis H37Ra (Difco; DF3114-33-8). For arthritis induction, on day 0, mice received an i.a. injection directly into a single knee joint containing 10 µL of mBSA (10 mg/mL) in PBS. Arthritis swelling was measured daily between days 0 and 9 using callipers. Therapeutic intervention occurred on day 2 (prior to peak of inflammation) via an i.a. injection of either 4-OI 1.6 mM, 4 mM or the highest dose cyclodextrin vehicle control. To ensure consistency, all clinical scoring was performed by the same investigator, who had been specifically trained in this assessment method. In flow cytometry, fibroblast engagement was evaluated using the gating strategy Live (BioLegend #423102), CD45-(BioLegend #103147), CD31-(BioLegend #102418) and PDPN+ (BioLegend #127408).

Collagen-induced arthritis model

Female DBA/1 (Dilute Brown Non-Agouti) mice were purchased from Jackson Laboratory. The collagen-induced arthritis (CIA) model was performed at the Animal Facility, Department of Biomedicine, Aarhus University. Eight-week-old female mice were intradermally (i.d.) injected in the tail base with an emulsion of type II chick collagen (Chondrex #20012) and incomplete Freund’s adjuvant (Chondrex #7002) at day 0, followed by an i.d. injection in the tail with chicken collagen and complete Freund’s adjuvant (Chondrex #7008) or incomplete Freund’s adjuvant at days 22 and 45, respectively. Mice were observed for swellings of the paws, and treatment was started at day 52, when >75% of the mice showed symptoms. Mice were treated intraperitoneally (i.p.) 3 times per week with 2 mg/kg dexamethasone dissolved in 20% propyleneglycol and 5% glucose (n = 7), 50 mg/kg 4-OI dissolved in 40% cyclodextrin (n = 7), 40% cyclodextrin (n = 7) or received no treatment (n = 6). In vivo study dosage of 50 mg/kg 4-OI was based on previous studies [45]. Mice were scored for inflammation and swelling 3 times per week, following the RA scoring system provided by Chondrex. On the final day of the study (day 71), blood was collected from the heart. Blood samples were centrifuged, and serum was collected and stored at −20 °C. Serum was analysed using the commercially available multiplex assay V plex proinflammatory panel 1 mouse kit (Mesoscale # K15048D).

Statistics

Statistical analyses and graphical representations were conducted using GraphPad Prism 7 and GraphPad Prism 9 for Mac (GraphPad Software). To account for donor variation, the secretion of MCP-1 and IL-6 from the cultures was calculated as ratios, comparing stimulated cells to untreated cells. Subsequently, the differences were compared with the corresponding controls and deemed statistically significant only if they differed from both the untreated cultures and the corresponding controls. Data were analysed using paired or unpaired t tests. Additionally, gene silencing data were assessed by calculating the 2^−dCt, followed by normalisation to the DMSO-treated Adeno-associated virus integration site 1 control in the respective experiment. Data are shown as means of biological replicates and standard error means. Statistical significances were calculated using 2-way analysis of variance and Holm-Sîdák test for multiple comparisons.

For mBSA mice, data were analysed using 2-way analysis of variance repeated measures with Sidak multiple comparisons tests comparing individual time point variances, treatment versus control. A 2-sided P value of<.05 was considered statistically significant in all tests. Graphs were presented as the mean with the standard error of the mean. A 2-sided P value < .05 was considered statistically significant.

Ethics

The Danish Data Protection Agency and the Ethics Committee at Central Region Denmark approved the collection of PBMCs and SFMCs (20-121-329). All human samples were obtained after informed written consent according to the Declaration of Helsinki.

The CIA model was conducted in accordance with the ethics in national guidelines for care and use of laboratory animals under the Danish animal experiment inspector license #2020-15-0201-00694 at the Department of Biomedicine, Aarhus University.

The mBSA-monoarthritis animal experiments were approved by the UK Home Office and conducted in accordance with the UK’s Animals Act 1986 and the UK Home Office Code of Practice under the project licence PP6493771. The project and experimental protocols were approved by the University of Birmingham Animal Ethics Review Committee.

RESULTS

An in vitro 4-OI dosage of 125 µM reduces inflammatory cytokines without affecting viability or proliferation

The cellular responses to 4-OI stimulation in ^IMRA-FLS, as well as THP-1 cell cultures, were used to establish optimal and nontoxic treatment dosage. The greatest reduction of MCP-1 secretion was achieved at dosages of 500 µM 4-OI (Fig 1A). However, when evaluating real-time cytotoxicity and cell proliferation (Fig 1B), dosages of 500 µM proved inhibitory to FLS proliferation. Dosages of 250 µM and below did not impact cell growth, and a suboptimum dosage of 125 µM of 4-OI was validated as tolerable using flow cytometry and the MTT assay in patient RA-FLS samples (Fig 1B-D). In addition, the effect of 125 µM 4-OI treatment on viability was evaluated using flow cytometry in patient sample-derived RA-FLS and showed no difference to the vehicle control DMSO (Fig 1D).

Figure 1.

4-OI dosage of 125 µM reduces MCP-1 and does not affect cell viability or growth. A, Quantification of MCP-1 secretion after 48 hours of stimulation and titration of 4-OI in human inflammatory environment–fibroblast-like synoviocytes (^IMRA-FLS) and THP-1 cell lines. B, Quantification of real-time monitoring of cytotoxicity and proliferation after titration of 4-OI using microelectronic plates in ^IMRA-FLS. C, Quantification of cell viability was determined using MTT assay in RA-FLS with 125 µM DMSO or 4-OI in DMSO after 48 hours (DMSO: n = 5; 4-OI: n = 4). D, Cell viability determination using flow cytometry in RA-FLS with 125 µM DMSO or 4-OI in DMSO after 48 hours (n = 7). 4-OI, 4-octyl itaconate; AAVS1, Adeno-associated virus integration site 1; ARE, antioxidant response element; DMSO, dimethyl sulfoxide; HIE-FLS, human inflammatory environment–fibroblast-like synoviocyte; HMOX1, heme oxygenase-1 gene; HO-1, heme oxygenase-1; IL-6, interleukin 6; ^IMRA-FLS, immortalised RA-FLS; Keap1, Kelch-like ECH-associated protein 1; MCP-1, monocyte chemoattractant protein-1; NQO1, NAD(P)H-dehydrogenase [quinon] 1; Nrf2, nuclear factor erythroid 2–related factor 2; RA-FLS, rheumatoid arthritis fibroblast-like synoviocyte; THP-1, immortilized human monocytic cell line; TRXR1, thioredoxin reductase 1.

Hence, 125 µM of 4-OI was chosen as the treatment dose for subsequent cultures containing RA patient samples.

Nrf2 is the target for 4-OI treatment in human inflammatory RA-FLS

To confirm Nrf2 as the target of 4-OI in RA-FLS, human ^IMRA-FLS were subjected to gene silencing of Nrf2 using CRISPR-Cas9 and the downstream mRNA and protein levels were evaluated (schematic pathway seen in Fig 2A). Hereafter, the ^IMRA-FLS were treated with 4-OI, and subsequently, Nrf2 downstream effector genes HMOX1 and NQO1 expression were quantified using real-time qPCR to determine the success of gene silencing. When control cells were treated with 4-OI, expression of NQO1and HMOX1 increased by 41% (P < .01) and 85% (P < .0001), respectively (Fig 2B,C). On the contrary, 4-OI treatment of cells that had undergone guide-1–mediated gene silencing of Nrf2 did not increase the expression of HMOX1 and NQO1; the same tendency applied to guide-2 (Fig 2B,C). Accordingly, the protein levels of the Nrf2-activator TRXR1 and the downstream effector molecule NQO1 were reduced after 4-OI treatment in CRISPR-Cas9 gene silencing ^IMRA-FLS compared with controls (Fig 2D). Quantification can be seen in Supplementary Figure S2. Collectively, these findings emphasise the role of 4-OI as an inducer of Nrf2 activity in human inflammatory RA-FLS.

Figure 2.

4-Octyl itaconate (4-OI) stimulation mediated a reduced expression of downstream effector molecules after gene silencing of nuclear factor E2-related factor 2 (Nrf2). A, Schematic presentation of 4-OI-mediated Nrf2 activation through inhibition of Kelch-like ECH-associated protein 1 (Keap1) and subsequent activation of antioxidant response element (ARE) with production of HMOX1/HO-1, NQO1, as well as thioredoxin reductase 1 (TRXR1) (created with BioRender under the agreement number MO27MFDC4A). B, Quantification of heme oxygenase-1 (HMOX1) mRNA expression with qPCR after gene silencing of Nrf2 using CRISPR-Cas9. Neutral editing site Adeno-associated virus integration site 1 (AAVS1) used as control. C, Quantification of NAD(P)H quinone dehydrogenase 1 (NQO1) mRNA expression with PCR after gene silencing of Nrf2 using CRISPR-Cas9. Neutral editing site AAVS1 used as control. D, Western blot of Nrf2 protein levels, the Nrf2-activator TRXR1 and downstream effector molecule NQO1 after CRISPR-Cas9 and subsequent 4-OI stimulation *P < .05, **P < .01, ***P < .001, ****P < .0001. 4-OI, 4-octyl itaconate; AAVS1, Adeno-associated virus integration site 1; ARE, antioxidant response element; CRISPR-Cas9, Clustered Regularly Interspaced Short Palindromic Repeats-CRISPR-associated protein 9; DMSO, dimethyl sulfoxide; HMOX1, heme oxygenase-1 gene; HO-1, heme oxygenase-1; Keap1, Kelch-like ECH-associated protein 1; MCP-1, monocyte chemoattractant protein-1; mRNA, messenger ribonucleic acid; NQO1, NAD(P)H-dehydrogenase [quinon] 1; NQO1, NAD(P)H-dehydrogenase [quinon] 1; Nrf2, nuclear factor erythroid 2–related factor 2; PBMC, peripheral blood mononuclear cell; PCR, polymerase chain reaction; RA-FLS, rheumatoid arthritis fibroblast-like synoviocyte; TNF, tumour necrosis factor; TRXR1, thioredoxin reductase 1.

4-OI stimulation reduced the production of proinflammatory cytokines in FLS-dominated in vitro models of RA

After establishing a well-tolerated treatment dosage of 125 µM 4-OI and confirming Nrf2 as a target of 4-OI treatment also in RA-FLS, RA-FLS were cocultured with autologous PBMCs to generate a FLS-dominated in vitro model of RA (n = 7). When these patient-derived RA-FLS and autologous PBMCs were cocultured with 4-OI, levels of secreted MCP-1 decreased by 84% (P < .01) compared with controls. This reduction was superior to that achieved by anti-TNF-⍺ antibodies and corticosteroid treatment in this FLS-dominated coculture model of RA. Anti-TNF-⍺ antibodies and corticosteroids reduced MCP-1 secretion by 41% and 54%, respectively (both P < .01) (Fig 3A). After coculturing, the inflammatory FLS were purified from the cocultures and examined separately by Western blotting. After 4-OI coculturing, the reduction in MCP-1 secretion from the RA-FLS and PBMC cocultures was accompanied by a concomitant increase in the presence of the Nrf2 downstream effector molecules NQO1 and HO-1 in the inflammatory RA-FLS cell subset (Fig 3B). A similar effect was seen in RA-FLS from monocultures (Supplementary Fig S3).

Figure 3.

Nrf2 activity is reduced in FLS-dominated pathology, and 4-OI-mediated Nrf2 activation reduces the production of proinflammatory cytokines in FLS-dominated in vitro models of RA. A, Ratio of MCP-1 secretion between stimulated and unstimulated RA cells in cocultures of RA-FLS and PBMCs after 48 hours (n = 7). B, Immunoblot analysis of whole cell lysate to evaluate mRNA levels of NQO1 and HO-1 after stimulation with steroid and 4-OI or vehicle control visualised using Western blot in RA-FLS from RA-FLS-PBMC cocultures supplemented by quantification of mRNA levels. C, Ratio of MCP-1 and IL-6 secretion between paired stimulated and unstimulated RA cells in monocultures of RA-FLS after 48 hours (IL-6 data: n = 6. MCP-1 data: n = 12, except anti-TNF-⍺ treatment: n = 7.) D, Ratio of the percentage of double-positive CD90 and podoplanin (PDPN) RA-FLS. E, HO-1 secretion in HIE-FLS (^IMRA-FLS) after 4-OI stimulation (n = 12). F. Synovium mRNA levels of HMOX1 in lymphoid, myeloid and fibroid RA (lymphoid vs myeloid P > .05; myeloid vs fibroid P < .05; lymphoid vs fibroid P < .0001). *P < .05, **P < .01, ***P < .001, ****P < .0001. 4-OI, 4-octyl itaconate; AIA, antigen-induced arthritis; DAS, disease activity score; FLS, fibroblast-like synoviocyte; HIE-FLS, human inflammatory environment–fibroblast-like synoviocyte; HMOX1, heme oxygenase-1 gene; HO-1, heme oxygenase-1; IL-6, interleukin 6; ^IMRA-FLS, immortalised RA-FLS; MCP-1, monocyte chemoattractant protein-1; mRNA, messenger ribonucleic acid; NQO1, NAD(P)H-dehydrogenase [quinon] 1; Nrf2, nuclear factor erythroid 2–related factor 2; PBMC, peripheral blood mononuclear cell; PDPN, podoplanin; RA, rheumatoid arthritis; RA-FLS, rheumatoid arthritis fibroblast-like synoviocyte; TNF-⍺, tumour necrosis factor alpha.

To build on the observed response in the FLS subset, we tested the effects of 4-OI in monocultures of inflammatory RA-FLS deprived of leukocytes. In this in vitro model consisting of only inflammatory patient-derived RA-FLS, 4-OI treatment also mediated a significant decrease in secretion of MCP-1 by 30% and IL-6 by 60% compared with controls (P < .05) (Fig 3C). This effect was comparable with that of corticosteroid treatment. Notably, anti-TNF–⍺ treatment showed no effect on MCP-1 levels in monocultures of inflammatory RA-FLS. The reduced secretion of MCP-1 and IL-6 was not a result of a reduction in the proportion of inflammatory FLS in the cultures (PDPN+THY+) (Fig 3D).

In addition, much like the increase in intracellular expression of HO-1 in purified RA-FLS from cocultures, secretion of HO-1 in monocultures of ^IMRA-FLS increased by 38% when stimulated with 4-OI compared with controls (P < .01) (Fig 3E).

Reduced Nrf2 activity selectively in fibroid-dominated synovial RA pathology

Moreover, we wanted to explore the Keap1-Nrf2 axis and examine whether a low expression of HMOX1 was indeed present in fibroblast-dominated synovial RA pathology. To approach this, we utilised the open-access database, PEAC [8]. In the open-access PEAC, RA samples were stratified after cellular, phenotypical synovial composition into either fibroid, myeloid or lymphoid.

NFE2L2, the gene encoding the transcription factor Nrf2, was expressed uniformly in the synovium across pathotypes of RA (Supplementary Fig S4). However, HMOX1 was expressed significantly less in stromal-dominated RA (fibroid subtype), when compared with both a lymphoid (P < .0001) and a myeloid (P < .05) dominated synovial pathology (Fig 3F). Indicating less activity of Nrf2 in fibroblast-dominated RA pathology compared with lymphoid and myeloid local pathology. NQO1 was similarly expressed between myeloid and stromal-dominated RA pathology (Supplementary Fig S4).

4-OI treatment ameliorates arthritis

For further examination and validation of our in vitro findings, 2 in vivo RA models were established. The mBSA-monoarthritis mouse model was established (Fig 4A). This model was specifically chosen for its distinct focus on localised inflammatory responses and FLS involvement (Fig 4C and Supplementary Fig S5). Similarly, there are previous reports demonstrating a significant role of inflammatory fibroblasts in the CIA model [46].

Figure 4.

Mouse model of RA with 4-OI treatment. A, Schematic representation of mBSA model. B, Knee swelling measured as percent increase from day 0 since i.a. injection in the mBSA model using both doses of 1.6 mM and 4 mM of 4-OI and area under the curve (AUC) (n = 5). The circle indicates the time of administration of active treatment. Components A and B represent data from the same model. C, Fibroblast engagement in the mBSA arthritis model. D, Progression using values of disease activity score (DAS) from days since arthritis onset and treatment start (day 0) to endpoint (day 19) in CIA model along with AUC (n = 7). E, Plasma IL-6 normalised to no treatment after euthanasia in the CIA model (n = 7). *P < .05, **P < .01, ***P < .001, and ****P < .0001. 4-OI, 4-octyl itaconate; AIA, antigen-induced arthritis; AUC, area under the curve; CIA, collagen-induced arthritis; DAS, disease activity score; IL-6, interleukin 6; mBSA, methylated bovine serum albumin; PDPN, podoplanin; RA, rheumatoid arthritis.

However, the mBSA model includes not only inflammatory FLS but also other cell types found in the arthritic joint, providing a more representative setting of RA-FLS involved pathology.

In the monoarthritis mBSA model, i.a. injections of a dosage of 4 mM 4-OI significantly decreased joint swelling after 2 days posttreatment compared with controls (Fig 4B) (P < .0001). This is particularly noticeable due to the fact that in this model, fibroblast engagement peaks at day 2 after arthritis induction (Fig 4C and Supplementary Fig S5).

A similar trend was seen in the 1.6 mM dose; however, not statistically significant. Despite the more complicated and diverse cellular composition, when applied locally, 4-OI reduced inflammation also in this more complex disease model, consistent with our in vitro findings.

In the CIA model, systemic i.p. 4-OI treatment did not result in a statistically significant improvement in disease activity. However, there was a noticeable trend toward improvement in disease activity score (DAS) score comparable to the effects of steroid treatment (Fig 4D). Notably, when 4-OI was administered systemically using i.p. injections in the CIA model, P IL-6 expression was significantly reduced by 78% when compared with no treatment (P < .05), whereas corticosteroid treatment did not affect systemic IL-6 levels (Fig 4E). No other cytokines measured were significantly affected by 4-OI in the CIA model (Supplementary Fig S6). In total, the mice were exposed to 300 mg/kg 4-OI and 32 mg/kg steroid over a 19-day period.

DISCUSSION

The synovial microenvironment plays a pivotal role in the pathology of RA. Low oxygen concentrations as well as unbalanced metabolic pathways exacerbate local inflammation highlighting the potential to target the metabolic pathways in RA [47]. In this study, we demonstrate that 4-OI treatment can induce the Nrf2-pathway in FLS-dominated in vitro models of RA and ameliorate arthritis both in in vitro and in vivo models of RA.

The therapeutic potential of Nrf2 targeting in inflammatory diseases has been promising [[23], [24], [25],27,31,45], and the anti-inflammatory effects of 4-OI and Nrf2 activation in leukocytes are well established [24,27,28,34]. However, data on the role of Nrf2 activation directly in inflammatory RA-FLS and FLS-dominated RA pathology are warranted.

In this study, we demonstrate that 4-OI can induce Nrf2 activity in FLS-dominated pathology, effectively dampening inflammatory cytokine production in vitro and in vivo. Local 4-OI application decreases joint swelling at the peak of FLS activation in this model. After systemic 4-OI treatment, IL-6 was selectively decreased, a cytokine produced primarily by inflammatory FLS [48]. Collectively indicating an effect of 4-OI on inflammatory FLS both in vitro and in vivo. This is particularly noteworthy given the current lack of treatments that directly target inflammatory stromal cells in RA [8].

Supportive hereof are the results that high levels of itaconate in patients with RA correlate with a reduction in disease activity [49]. Moreover, we have recently shown that P levels of HO-1, a downstream effector molecule of Nrf2, correlated with the number of swollen joints and that the level of HO-1 was markedly elevated in the SF compared with in P [50]. The local upregulation and correlation with swollen joints support the role of Nrf2 activity in RA pathology. Furthermore, the decreased expression of HMOX1 selectively in stromal-dominated RA pathology, despite comparable levels of NFE2L2 across pathotypes, indicates an increased proportion of inactive Nrf2 in stromal-dominated RA pathology. Together, this highlights the potential for therapeutically increasing Nrf2 activation in RA patients with stromal-dominated pathology, which constitutes a major part of the current treatment nonresponders [8]. The potential of Nrf2 as a relevant target also in FLS-dominated disease is supported by recent results showing an effect of endogenous itaconate directly on RA-FLS [33].

In this study, we chose the 4-OI dosage of 125 µM, after demonstrating that this dosage showed no toxicity in vitro. This dosage has previously been shown to be effective and nontoxic in other in vitro and in vivo studies [51]. However, in the local monoarthritis model, the dosage of 4-OI was increased to 4 mM due to limited systemic distribution when injected due to drug retention in the joint [52]. In addition, i.p. dosages of 6.6 mM have previously been used in rats [33]. The potential of i.a. treatment possesses multiple advances in terms of bioavailability. In the mBSA model, the clinical efficacy observed emphasises that the inflamed joint may be a target for Nrf2 activation in RA.

Four-OI is a known stable and strong inducer of Nrf2 activity in leukocytes [27]. Through gene silencing of Nrf2 in ^IMRA-FLS and subsequent 4-OI stimulation, this mechanism was reinforced and shown also to directly apply to inflammatory RA-FLS. Supported by the upregulation of downstream proteins HO-1 and NQO1 in patient-derived FLS-dominated models of RA after 4-OI treatment.

Direct measurement of Nrf2 protein levels is challenged as a marker of its signalling activity because it is kept at low abundance due to constant degradation and the functional status depends on complex posttranslational regulation, nuclear translocation and downstream gene induction rather than mere presence [53]. Hence, to avoid the caveats related to the direct measurement of Nrf2 levels, we looked at the modulation of Nrf2 activity by measuring multiple of the downstream proteins HO-1, NQO1 and TRXR1 as recently recommended [53]. HMOX1 is regarded as a reliable indicator of Nrf2 activation and becomes more informative when combined with additional downstream markers of Nrf2 activation such as NQO1 [51]. Our results demonstrate that 4-OI–mediated Nrf2 activation upregulates both mRNA expression of HMOX1 and accordingly secretion of HO-1 from inflammatory RA-FLS. HMOX1 and HO-1 are considered solid markers of Nrf2 activity despite also being regulated by other axes such as heme metabolism [50]. In the PEAC cohort, the Nrf2 downstream molecule NQO1 was inversely expressed compared with HMOX1 (Supplementary Fig S4), as also seen in other diseases [54]. One possible reason for this is the fact that NQO1 can also be activated through other pathways, i.e. the aryl hydrocarbon receptor, which does not affect HMOX1 [53,55].

IL-6 and MCP-1 are 2 inflammatory cytokines known to be produced in high levels by inflammatory RA-FLS and strongly linked to RA pathology why these cytokines were chosen as outcome measures for our in vitro setups [15]. Notably, we observed that the addition of 4-OI treatment decreased both MCP-1 and IL-6 in FLS monocultures as well as in autologous PBMC cocultures supportive of a role of Nrf2 targeting by 4-OI in both stromal and immune cells in RA. Moreover, 4-OI–mediated Nrf2 activation was associated with a robust reduction of both MCP-1 and IL-6 production in cultures of purely RA-FLS not seen after anti-TNF-⍺ antibody treatment.

The lack of effect of anti-TNF–⍺ antibody treatment in FLS monocultures stresses that anti-TNF-⍺ antibodies do not robustly downregulate inflammation in RA-FLS. However, anti-TNF-⍺ antibodies did reduce MCP-1 secretion in cocultures of FLS and PBMCs. This supports our previous findings that anti-TNF-⍺ treatment primarily targets classical immune cells but has limited efficacy directly on inflammatory FLS in RA [42].

In addition to ex vivo settings, we studied 2 in vivo models of RA to evaluate the effects of 4-OI treatment both locally and systemically. Specifically, the mBSA-monoarthritis model provides the possibility to examine 4-OI–mediated local effects in a model driven by both classical immune cells and characterised by upregulation of inflammatory FLS (Supplementary Fig S5) to mimic the physiological conditions of the FLS-rich joint inflammation.

Our results show the potential for 4-OI treatment to ameliorate joint inflammation, a result that is supported by the effects seen with the addition of unmodified itaconate i.a. in a recent study [33]. It is noteworthy that local i.a. 4-OI application introduced a faster reduction in joint swelling compared with systemic application. This emphasises the potency of targeting the Nrf2-pathway to reduce joint inflammation in RA affecting both leukocytes and fibroblasts.

Hence, these results may have clinical relevance in refractory RA where anti-TNF treatments have failed to bring patients into remission. Intriguingly, synovial biopsies from RA patients suggest that RA treatment nonresponders demonstrate a fibroblast-dominated pathotype [8,9]. In this study, we demonstrate that this particular pathotype shows decreased expression of Nrf2 activity. Thus, both local and systemic targeting of inflammatory fibroblasts by Nrf2 activation is an area of compelling interest and should be explored further to investigate clinical potential.

Furthermore, the prospect of establishing another locally applicable treatment in addition to glucocorticoids could prove valuable to optimise RA treatment strategies. The possibility to decrease disease activity through 4-OI application, especially in anti-TNF–⍺ refractory RA, is intriguing; however, it needs additional studies to confirm the clinical potential of 4-OI as a potential supplement to existing treatment in RA.

Collectively, these findings provide mechanistic insights into the role of Nrf2 targeting in FLS-dominated RA pathology and show the anti-inflammatory effects of 4-OI in 2 different in vivo RA models.

Limitations

A strength of this study is that gene silencing was performed in patient-derived immortalised RA-FLS; however, only 1 method was used and could be strengthened by other gene silencing methods such as siRNA.

To study the role of local Nrf2 and 4-OI treatment, we utilised 3 different in vitro RA-cultured systems. These cultures consisted of in situ preconditioned mononuclear cells and FLS grown from cells harvested from the SF of individual RA patients. A strength of this study is the inclusion of 13 unique SFMC-derived RA-FLS cultures enabling us to establish FLS models that are driven by the spontaneous endogenous production of multiple proinflammatory factors, without the addition of any exogenous stimulation. A caveat to working with patient-derived primary cell models is the degree of heterogeneous results obtained. The heterogeneous results previously seen [42] may to some extent reflect the differences seen in the included patients (Supplementary Table S1). However, this experimental design offers the possibility to isolate the different components of arthritis pathobiology and evaluate them separately. On the contrary, in our in vivo model, the reduction in joint swelling cannot be attributed solely to the effect of 4-OI directly on the inflammatory FLS but is likely influenced by the known effect on macrophages, despite the robust FLS activation shown in both models. However, although the effect of 4-OI was not shown directly on the FLS subset in vivo, multiple findings support the involvement of the inflammatory FLS subset in the effects seen after both systemic and local treatment. The fact that the local effects were shown after 2 days, where the activation of FLS activation peaks in the mBSA-monoarthritis model, and the findings that 4-OI treatment specifically reduces the IL-6 cytokine after systemic admission indicate a reduction of activity of the inflammatory FLS also in vivo. The general low disease activity and heterogeneity of the CIA model limit the clinical conclusions from this model.

CONCLUSION

In conclusion, these findings provide mechanistic evidence for the role of 4-OI as an Nrf2 activator in stromal-dominated local RA pathology in vitro. The 4-OI–nduced activation of Nrf2 decreased the secretion of inflammatory cytokines in fibroblast-dominated arthritis synovial cultures in vitro, and 4-OI application reduced joint swelling and disease activity in vivo. The clinical effect of the local treatment with 4-OI in vivo, together with reduced expression of HMOX1 seen in vitro selectively in FLS-dominated RA pathology, highlights the potential role of Nrf2 activation in FLS-dominated RA pathology. Collectively, these observations indicate that in RA, Nrf2 activation is implicated in regulating the activity of inflammatory FLS. Targeting Nrf2 with a small molecule, such as 4-OI, may emerge as a useful therapeutic adjunct in patients, especially in patients who do not respond adequately to current treatment strategies.

CRediT authorship contribution statement

Benedicte B. Andersen: Writing – review & editing, Writing – original draft, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Samuel Kemble: Writing – review & editing, Writing – original draft, Methodology, Formal analysis, Data curation. Stinne Greisen: Writing – review & editing, Methodology, Data curation, Conceptualization. Jacob Thyrsted: Writing – review & editing, Formal analysis, Data curation. Vivien R. Schack: Writing – review & editing, Writing – original draft, Formal analysis, Data curation. Elisabeth Fuchs: Data curation, Formal analysis, Writing – review & editing. Anne L. Hansen: Writing – review & editing, Methodology, Formal analysis. Thomas B. Poulsen: Writing – review & editing, Methodology, Data curation. Christian K. Holm: Writing – review & editing, Supervision, Conceptualization. David Olagnier: Writing – review & editing, Supervision, Methodology. Adam Croft: Writing – review & editing, Supervision, Conceptualization. Bent Deleuran: Writing – review & editing, Supervision, Methodology, Conceptualization. Morten A. Nielsen: Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition, Formal analysis, Conceptualization.