WHAT IS KNOWN ON THIS TOPIC

WHAT THIS STUDY ADDS

This study shows that the death of intestinal epithelial cells via necroptosis is an important factor that promotes the onset of arthritis. Inhibition of intestinal epithelial cell necroptosis by pharmacologically stabilising hypoxia-inducible factor 1α or by blocking receptor-interacting protein kinase-3 inhibits the development of arthritis

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE AND POLICY

Introduction

The intestinal barrier coordinates nutrient absorption, bacterial symbiosis and the immune system.1 The integrity of the intestinal barrier is considered to influence the development of inflammatory arthritis, suggesting a link between intestinal pathology and arthritis.2 Intestinal pathology characterised by abnormal immune cell infiltration and dysbiosis has been observed in the preclinical phase of rheumatoid arthritis (RA).3 4 The monolayer of intestinal epithelial cells (IECs), connected by tight junctions and encapsulated within the mucus, constitutes the crucial part of this mechanical barrier. If this barrier is destabilised (‘leaky gut’), autoimmunity and systemic inflammation can arise.5 In accordance, the maintenance of the intestinal barrier integrity, for example, by targeting tight junction proteins, alleviates inflammatory arthritis.6 7 However, the role of IEC survival in the development of inflammatory arthritis is currently unknown.

Aberrant and excessive IEC death has been reported as a sign of progression in chronic inflammatory bowel disease and necrotising enterocolitis.8 9 Necroptosis of the IEC disconnects tight junctions and lowers intestinal barrier integrity, thereby shifting it to a proinflammatory milieu both in experimental and human colitis.10–12 This phenotype is strongly linked to microbial dysbiosis and chronic intestinal inflammation, characterised by mucosal and systemic Th17 responses.10–12 To date, the role of IEC death in the development of arthritis is, however, unclear.

Intestinal inflammation consumes large amounts of oxygen, rendering the inflamed tissue hypoxic.13 14 Hypoxia activates a highly conserved family of transcription factors, among which hypoxia-inducible factor (HIF) 1α and HIF2α are controlling these metabolic processes.15 By transcriptionally enhancing claudin-15 expression, which impairs tight junctions in the intestinal barrier, HIF2α allows the opening of the intercellular tight junction and promotes arthritis.7 Conversely, in the context of chronic intestinal inflammation, HIF1α increases barrier function, elicits protective innate immune responses and activates antimicrobial responses.16–18 Therefore, HIF1α expression and stabilisation in IECs might also influence arthritis. Here, we observed that HIF1α had a critical role in IEC survival as it inhibited necroptosis of IEC. This effect was mediated by the control of receptor-interacting protein kinase-3 (RIPK3), a central mediator for necroptosis. In the absence of HIF1α, IEC death shifted from physiological apoptosis to pathological necroptosis and triggered mucosal barrier dysfunction and arthritis. Stabilisation of HIF1α by roxadustat or inhibition of RIPK3 by GSK’872 maintained intestinal barrier function and ameliorated the development of arthritis. These data suggest that RIPK3-mediated epithelial cell necroptosis is a critical factor in linking gut pathology to the development of arthritis.

MethodsHuman tissue

Sample information was described previously.7 Briefly, ileal biopsies from healthy individuals and patients with new-onset RA were used for the present study. Paraffin-embedded sections were subjected to immunofluorescence staining for HIF1α (Cayman). Nuclei were visualised by 4′,6-diamidino-2-phenylindole (DAPI) staining (Life Technologies). Paraffin-embedded sections were subjected to immunohistochemistry staining for RIPK1 (CellSignal), RIPK3 (CellSignal) and MLKL (Biorbyt). Confocal microscopy was used to screen the slides. HIF1α-positive cells were counted from three random high-power microscopic fields (×400 magnification). Immunohistochemistry staining of RIPK1, RIPK3 and MLKL was quantitatively scored, ranging from 0 (no staining) to 5 fully stained from three random high-power microscopic fields (×400 magnification).

Mice

C57BL/6 wild-type (WT) mice (6–8 weeks old) and DBA1/J (8 weeks old) were purchased from Janvier Laboratories. Epithelial-specific HIF1α-deficient mice (Hif1aΔI EC) were generated by crossing Hif1aflox/flox mice with Villin-cre mice. Hif1aflox/flox mice and Villin-cre mice were previously reported.19 20 Ripk3−/−, Mlkl−/−, Ripk3−/−Mlkl−/− mice were previously described and acquired from Professor Claudia Günther’s laboratory (FAU Germany).21 22 The genetically modified mice were maintained on a C57BL/6 background. Hif1aflox/flox cre-negative or Hif1a+/+ cre-positive littermates were considered as controls. The experiments were approved by the local ethics committee of the Regierung von Mittelfranken.

Collagen-induced arthritis (CIA) model

Mice were injected twice (days 1 and 21) with 100 μL of 0.25 mg chicken type II collagen emulsified in complete Freunds’ adjuvant (Sigma) at the base of the tail. The clinical score of each paw was assessed every other day by the same person. Scores were recorded on a scale of 0–4, as previously described.7

Treatments

Mice were orally given a vehicle or prolyl-hydroxylase domain (PHD) inhibitor (FG-4592, Medchemexpress, 20 mg/kg) or RIPK3 inhibitor, GSK'872 (Medchemexpress, 7.5 mg/kg) every other day from day 6 post-first immunisation (pfi) until day 36 pfi.

Flow cytometry

Single-cell suspensions were harvested from Peyer’s patches, mesenteric lymph nodes and spleen. Following Fc-blocking (CD16/CD32), cells were stained with antibodies (Supplementary table 1). The expression of cell surface molecules was analysed by a Cytoflex (Beckman Coulter) flow cytometer. LIVE/DEAD Fixable Orange (602) Viability Kit (L34983, Invitrogen) was used to exclude dead populations before analysis. For Th1 and Th17 cell analysis, phorbol-12-myristate-13-acetate (20 ng/mL), ionomycin (1 μg/mL) and monensin (L+PIM) were added to the culture medium 5 hours before fixation. Foxp3/Transcription Factor Staining Buffer Set (00/5523/00, eBioscience) was used for membrane permeabilisation. The analyses were performed with the CytExpert V.2.4 software.

IEC isolation

Small intestines were dissected, washed with ice-cold phosphate-buffered saline (PBS) containing 0.154 M sodium chloride (NaCl) and 1 mM dithiothreitol (DTT) and cut into 0.5-cm fragments, which were incubated with 1.5 mM EDTA and 0.5 mM DTT for 30 min. After incubation, the intestinal fragments were washed in the same buffer and vortexed to detach the intestinal epithelium. Acquired supernatants were filtered through a 70-μm cell strainer twice to isolate single cells.

RNA isolation and quantitative real-time (qRT) PCR analysis

The total RNA of a cell or tissue was extracted with Trizol (Invitrogen), and complementary DNA (cDNA) was synthesised with a HighCapacity cDNA Reverse Transcription Kit (4368814, Thermo Scientific). qRT-PCR was performed using SYBR Green I-dTTP (Eurogentec). Primers were listed in online supplementary table 2. The relative expression of a target gene was calculated according to the ∆∆Ct method.

Organoids culture

Small intestine organoids were cultured as previously described.23 Tissue was dissected and washed, and intestinal villi were scraped out of the epithelium. After incubation in dissociation buffer (PBS with 2 mM EDTA), intestinal crypts were isolated and seeded in matrigel (BD Bioscience). Culture medium (advanced Dulbecco's modified Eagle medium (DMEM)/F12 (Invitrogen), 10 mM (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 2 mM GlutaMax (Invitrogen), 100 U/mL penicillin–streptomycin (Gibco), 50 ng/mL murine epidermal growth factor (Immunotools), 1 mg/mL recombinant human R-spondin (R&D Systems), 1 mg/mL B27 (Invitrogen), 1 mM N-acetylcystein (Sigma-Aldrich) and 50 ng/mL recombinant murine Noggin (Peprotech) were changed every 3 days.

RNA-seq analysis

cDNA library and RNA-seq were conducted by NovogeneEurope (Cambridge Science Park, UK) on the Illumina HiSeq 2500 platform. The raw data were first quality-controlled (fastqc v0.11.8) and then the adapter and low-quality reads were removed (fastqc v0.11.8). The cleaned data were mapped to the reference genome GRCm38 (star v2.6.1c). Then transcripts were assembled to obtain count numbers for each gene (samtools v1.8, subread v1.6.1). Count data were normalised and subjected to downstream analysis and visualisation in RStudio (R V.4.1.1.) software. The RNA-seq data used in this paper have been uploaded to the GEO databases (GSE175907, GSE176266 and GSE225597).

Micro-CT (μCT)

Mice’s hind paws were fixed overnight in 4% paraformaldehyde and measured by the cone-beam system (SCANCO Medical AG, Bruettisellen, Switzerland) at 55 Kv, 177 μA and 200 ms integration time. The isotropic size was 8.6 μm.

Histology

Paws were fixed overnight in 4% paraformaldehyde, decalcified in 14% EDTA for 2 weeks and embedded in paraffin. Sections of 5-μm thickness were generated from the paraffin tissue blocks, and then H&E and tartrate-resistant acid phosphatase (TRAP) staining were performed. Intestines were cut into 5-μm sections. After 4% paraformaldehyde fixation and paraffin embedment, H&E or periodic acid Schiff (PAS) stains were performed. Intestinal histological scores were assessed on H&E slides following published criteria.24

Fluorescence staining

Sections were deparaffinised and epitopes were retrieved by the heat-induced method at 95°C for 10 min in sodium citrate buffer (10 mM sodium citrate, pH 6.0). Sections were blocked with 5% bovine serum albumin and 2% horse serum at room temperature for 1 hour. Sections were incubated with the primary antibodies listed in supplementary table 1 . Immunofluorescence was realised following published protocols.25 BZ-X700-All-in-One Fluorescence Microscope (Keyence) or Carl Zeiss LSM 700 Laser Scan Confocal Microscope (confocal fluorescence microscopy) were used for capturing images.

Fluorescence in situ hybridisation (FISH)

Ileal sections were deparaffined, rehydrated and incubated in hybridisation buffer (20 mM Tris–hydrochloric acid, 0.9 M NaCl and 0.1% sodium dodecyl sulphate) for 10 min at 50°C. Then, sections were incubated with 100 nM 16S rRNA-targeted bacterial probes (5′−3′: GCTGCCTCCCGTAGGAGT, fluorescein isothiocyanate (FITC)-conjugated; Sigma) in the hybridisation buffer for 4 hours at 50°C and mounted with DAPI. A keyence microscope was used for imaging acquisition. Bacterial probe signal intensity was quantified by measuring the positive area for FITC fluorescence.

Chromatin immunoprecipitation sequencing and qRT-PCR (ChIP-Seq and ChIP-qPCR)

IECs were isolated from Hif1a ∆IEC mice and littermate control mice at day 36 pfi. HIF1α ChIP experiments were performed with the ChIP-IT Express kit (53040, Active Motif) using IgG as a control. The enriched DNA was sequenced in NovogeneEurope. ChIP-Seq raw data were processed on the EU server of the Galaxy Community Hub. Quality control was performed using FastQC; short reads were mapped using Bowtie2, and peak calling was performed using MACS2. The peak annotation was performed using the R package ChIPseeker. Genome distribution and peak visualisation were processed by the IGV V.2.13.1 software. Primers targeting Ripk1, Ripk3 and Mlkl promoter hormone response elements (HREs) were listed (online table 2). HIF1α or HIF2α were predicted to bind on the same core element (A/GCGTG) in the JASPAR database, and ChIP primers were designed by Primer-BLAST. The ChIP-Seq data used in this paper have been uploaded to the GEO database (GSE225594).

Cell line culture and transfection

MC38 cells (C57BL6 murine colon adenocarcinoma cell line) were cultured in DMEM, supplemented with 10% (v/v) fetal bovine serum (GIBCO, Invitrogen) and 1% penicillin/streptomycin at 37°C under a 5% carbon dioxide humidified atmosphere. A lipofectamine 3000 transfection kit (Invitrogen, L3000001) was used for transfections according to the manufacturer’s recommendations. HIF1α short hairpin RNA (shRNA) plasmid (pLV(shRNA)-U6>mHif1 a, Plasmid#VB900122-2421ngn, Biozol/Vectorbuilder) was used for HIF1α knockdown. To assess knockdown efficiency, transfected cells were checked by western blot after 1% hypoxic treatment for 8 hours.

Dual luciferase reporter assay

Ripk1, Ripk3 and Mlkl promoter HREs (online supplementary table 2) were amplified by PCR from genomic DNA extracted from splenocytes in C57BL/6 WT mice and cloned using HINDIII and NheI (New England Biolabs) into the pGL4.23 firefly reporter vector, respectively (Promega). The recombinant vectors were sequenced in Microsynth Seqlab (Göttingen, Germany) to validate the presence of inserted HREs. MC38 cells were co-transfected with luciferase reporter plasmids, HIF1α or HIF2α TM plasmids (Addgene) and Renilla plasmids (Addgene) using the Lipofectamine 3000 kit (Invitrogen). After 48 hours of culture, cells were lysed, and luciferase activity was measured and normalised to Renilla activity.

ELISA

Interleukin (IL)-17A and interferon-gamma (IFN-γ) levels were measured by commercial ELISA (88-7371-88 and 88-7314-88 Invitrogen, respectively).

Statistical analysis

Data are expressed as mean±SD. The data distribution was tested by the Shapiro-Wilk or Kolmogorov-Smirnov tests. A two-tailed unpaired Student’s t-test or Mann-Whitney U test was used for a single comparison. One-way or two-way analysis of variance (ANOVA) followed by Tukey’s or Bonferroni’s test, multiple t-tests and the Kruskal-Wallis H test followed by Dunn’s test were used for multiple comparisons. For multiple comparisons, non-normal distributed data were transformed by a square root or logarithm. Kaplan-Meier analysis with a log-rank test and χ2 test were used for survival curve or incidence comparisons. All experiments were repeated at least two times. P values of 0.05 were considered statistically significant and are shown as *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. Graph generation and statistical analyses were performed using the Prism V.9 software (GraphPad, La Jolla, CA, USA).

ResultsIEC death at the onset of experimental arthritis

To determine IECs’ integrity during the course of CIA, we delineated the translocation of intestinal bacteria as macroscopic evidence of barrier dysfunction. Using a bacterial-specific probe for FISH, bacterial translocation could be detected in the ileum shortly before (day 29 pfi) and at the onset of CIA (day 36 pfi) (figure 1A and C). The death of ileum IECs was detected by terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) staining (figure 1B and D) and electron microscopy (figure 1E). However, no significant increase in bacterial translocation or cell death was shown in the colon (online supplemental figure 1A–D). To elucidate the cell death machinery at play, IECs from the small intestine were isolated at the onset of arthritis (day 36 pfi) and analysed by bulk-RNA sequencing. The volcano plot (figure 1F) and Gene Ontology (GO) enrichment analyses (figure 1G) indicated a significant alteration of the apoptotic pathway at the onset of arthritis. Among the differentially expressed genes, Caspase-3, Caspase-8, Ripk1, Ripk3 and the Mlkl gene were significantly increased in IECs from arthritic mice compared with non-arthritic controls (figure 1H). These results were confirmed by qRT-PCR analysis (figure 1I) and immune fluorescence, indicating upregulation of cleaved caspase-3 and caspase-8 (online supplemental figure 2A–C) as well as key proteins involved in necroptosis, such as p-RIPK1, p-RIPK3 and p-MLKL in IEC at the onset of arthritis (day 36 pfi; figure 1, online supplemental figure 2D–G), suggesting that substantial death of IECs occurs at the onset of arthritis.

Figure 1

IEC death at the onset of experimental arthritis. Representative pictures and quantification of fluorescence in situ hybridisation using the bacteria-specific probe EUB 338-fluorescein isothiocyanate (green) (A and C) and TUNEL staining for cell death (green) (B and D) on ileal sections in DBA/1 mice subjected to CIA at indicated days pfi (n=5); nuclei visualisation by DAPI (blue); scale bar 50 µm. (E) EM showing the intestinal epithelial barrier in the ileum of mice induced for CIA or non-induced C57BL/6 mice at day 36 pfi; scale bar 1 µm. (F) Volcano plot of RNA bulk sequencing showing the DEGs in IECs from mice induced for CIA or non-induced C57BL/6 mice at day 36 pfi (n=3). (G) GO enrichment analysis of the DEGs. Enriched pathways are presented in a dot plot. (H) Heatmap of the DEGs. (I and J) Quantitative real-time PCR analysis of Caspase-3, Caspase-8, Ripk1, Ripk3 and Mlkl mRNA expression in IECs from the ileum of DBA/1 mice induced for CIA (n=5). (K) KEGG pathway enrichment analysis of the DEGs; enriched pathways are presented in a dot plot; symbols represent individual C57BL/6 mice; data are shown as mean±SD. Statistical significance was determined by a one-way analysis of variance (C, D, I and J) followed by Tukey’s test for multiple comparisons. CIA, collagen-induced arthritis; DAPI, 4′,6-diamidino-2-phenylindole; DEGs, differentially expressed genes; EM, electron microscopy; GO, Gene Ontology; HIF, hypoxia-inducible factor; IEC, intestinal epithelial cell; KEGG, Kyoto Encyclopedia of Genes and Genomes; pfi, post-first immunisation; TUNEL, terminal deoxynucleotidyl transferase dUTP nick end labelling.

HIF1α expression in gut epithelial cells controls the development of arthritis

To identify a putative molecular correlation between the increased necroptosis markers and arthritis, we further analysed the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment results obtained from IEC bulk-RNA sequencing. The T cell differentiation pathways and the HIF1α pathway were significantly enhanced in IEC at the onset of arthritis (figure 1K). Immunofluorescence staining of HIF1α in ileum biopsies from patients diagnosed with new-onset RA showed increased HIF1α expression in IECs compared with healthy controls (figure 2A and B). Similarly, Hif1α mRNA expression was increased at the onset of CIA (day 36 pfi), while protein expression was elevated even earlier in the murine model (day 20 pfi; figure 2C–E). To dissect the function of HIF1α in the intestinal epithelium, we generated IEC-specific HIF1α knockout mice (Hif1a ΔIEC). Selective expression of Villin gene expression in IECs was reported using Villin-cre;Rosa26 tdTomato naive mice (online supplemental figure 3A). Subsequently, 90% HIF1α knockout efficiency in IECs was confirmed by qRT-PCR and western blot in IECs from Hif1a ΔIEC mice (online supplemental figure 3B–E). Interestingly, Hif1a ΔIEC mice subjected to CIA showed more severe (figure 2F) and earlier onset (figure 2G) arthritis than littermate controls. Hif1a ΔIEC mice presented increased paw thickness (figure 2H), bone erosion, synovial inflammation and osteoclast numbers (figure 2I and J). IFN-γ and IL-17A serum levels were increased in Hif1a ΔIEC mice (figure 2K) as well as Th1 and Th17 cell numbers in the Payer’s patches (PPs; figure 2L), mesenteric lymph nodes (MLN; figure 2M) and the spleen (figure 2N). IgA+B220 cells were also increased in PP and MLN of arthritic Hif1a ΔIEC mice (online supplemental figure 3F). Together, these results point to a protective role of HIF1α expression in IECs against the development of arthritis.

Figure 2

HIF1α expression in gut epithelial cells controls the development of arthritis. (A) IF and (B) quantification of HIF1α (green) and DAPI (blue) in the ileum of HC and patients with new-onset RA; scale bar 25 µm. (C) Quantitative real-time PCR analysis of Hif1α mRNA expression in ileal IECs from DBA/1 mice induced for CIA (n=5). (D) Quantifications and (E) representative pictures of HIF1α IF staining from ileal sections from DBA/1 mice induced for CIA; scale bar 50 µm . (F) Clinical score and (G) arthritis incidence in C57BL/6 mice with HIF1α-deletion in IECs (HIF1αΔIEC) and their littermate control subjected to CIA (n=7 and 8, respectively). (H) Representative hind paw pictures from CIA-induced HIF1αΔIEC and littermate controls C57BL/6 mice at day 36 pfi. (I) Representative images of µCT scans, H&E staining and TRAP staining and (J) quantification of erosion area, inflammation area and osteoclast numbers in the hind paws of CIA-induced HIF1αΔIEC C57BL/6 mice and littermate controls at day 36 pfi (n=7 and 8, respectively). White arrowheads indicate inflammation (H&E) and bone erosion (TRAP, µCT); histology scale bar 50 μm; µCT scale bar 1 mm. (K) IL-17A and IFN-γ levels measured by ELISA in serum of CIA-induced HIF1αΔIEC C57BL/6 mice (n=7 and 6, respectively) and littermate controls at day 36 pfi (n=7). Representative fluorescence-activated cell sorting plots and quantification of Th17 cells (IL-17A+CD4+) and Th1 cells (IFN-γ+CD4+) in (L) PP, (M) mesenteric lymph nodes and (N) the spleen in CIA-induced HIF1αΔIEC C57BL/6 mice (n=7) and littermate controls (n=8) at day 36 pfi. Symbols represent individual mice. Data are shown as mean±SD. Statistical significance was determined by two-tailed unpaired (B, J–N) Student’s t-test or Mann-Whitney U test, (C and D) one-way analysis of variance followed by (F) Tukey’s test, Kaplan-Meier analysis with log-rank test and (G) χ2 test. CIA, collagen-induced arthritis; DAPI, 4′,6-diamidino-2-phenylindole; HIF, hypoxia-inducible factor; IEC, intestinal epithelial cells; IF, immunofluorescence; HC, healthy individuals; IL, interleukin; IFN-γ, interferon-gamma; pfi, post-first immunisation; PP, Peyer’s patches; µCT, micro-CT; RA, rheumatoid arthritis; TRAP, tartrate-resistant acid phosphatase.

Figure 3

Pharmacological stabilisation of HIF1α inhibits arthritis and intestinal epithelial cell death. (A) Clinical score, (B) arthritis incidence and (C) representative paw picture from C57BL/6 mice induced for CIA and orally treated with PHD inhibitor RXD (n=12) or vehicle (n=13). (D) µCT scans, H&E staining and TRAP staining of the paws from C57BL/6 mice induced for CIA treated with RXD or vehicle at day 36 pfi. White arrows indicate inflammation (H&E) and bone erosion (TRAP, µCT); histology scale bar 50 µm. µCT scale bar 1 mm. (E) H&E, PAS and FISH staining of ileal sections of C57BL/6 mice induced for CIA and treated with RXD or vehicle at day 36 pfi (n=5); scale bar 50 µm. (F–H) Representative fluorescence-activated cell sorting plots and quantification of Th17 cells (IL-17A+CD4+) and Th1 cells (IFN-γ+CD4+) in the (F) PPs, (G) MLN and (H) spleen from C57BL/6 mice induced for CIA and treated with RXD or vehicle at day 36 pfi (n=5). (I) Serum IL-17A and IFN-γ levels in C57BL/6 mice induced for CIA and treated with RXD or vehicle at day 36 pfi (n=5). (J) Representative pictures of IF staining for HIF1α (red, upper panel), HIF2α (red, lower panel) and DAPI (blue); scale bar 50 µm. (K) Representative pictures of TUNEL staining (green) and nuclei visualisation by DAPI (blue) and EM pictures of ileal sections from C57BL/6 mice induced for CIA and treated with RXD or vehicle at day 36 pfi; TUNEL scale bar 50 µm. EM scale bar 1 µm. Quantitative-PCR analyses of (L) Caspase-3, Caspase-8, (M) Ripk1, Ripk3 and Mlkl mRNA expression in IECs from C57BL/6 mice induced for CIA and treated with RXD or vehicle. (N) Western blot analysis of HIF1α, HIF2α cleaved caspases-3 and caspases-8, RIPK1, RIPK3 and MLKL of IECs from C57BL/6 mice induced for CIA and treated with RXD or vehicle at day 36 pfi. Actin was used as aPe loading control. Data are shown as mean±SD. Statistical significance was determined by Kaplan-Meier analysis with (A) log-rank test, (B) χ2 test, two-tailed unpaired Student’s t-test or Mann-Whitney U test. CIA, collagen-induced arthritis; DAPI, 4′,6-diamidino-2-phenylindole; EM, electron microscopy; FISH, fluorescence in situ hybridisation; HIF, hypoxia-inducible factor; IEC, intestinal epithelial cells; IF, immunofluorescence; HC, healthy individuals; IL, interleukin; IFN-γ, interferon-gamma; pfi, post-first immunisation; PAS, periodic acid Schiff; PP, Peyer’s patches; PHD, prolyl-hydroxylase domain; µCT, micro-CT; RA, rheumatoid arthritis; MLN, mesenteric lymph nodes; RXD, roxadustat; TRAP, tartrate-resistant acid phosphatase; TUNEL, terminal deoxynucleotidyl transferase dUTP nick end labelling.

Induction of HIF1α alleviates arthritis and inhibits intestinal epithelial cell death

To further investigate the role of HIF1α in arthritis, we treated mice with roxadustat, which stabilises HIF expression by inhibiting PHD protein. Roxadustat treatment (oral gavage every other day) reduced the clinical score (figure 3A) and the incidence of arthritis (figure 3B) compared with vehicle treatment. Paw swelling, synovial inflammation, bone erosion and osteoclast numbers tend to reduce (figure 3C and D, online supplemental figure 4A–C). Intestinal inflammation was suppressed in roxadustat-treated CIA mice, as shown by reduced inflammatory areas and intestinal bacterial translocation (figure 3, online supplemental figure 4D–F). Except in the spleen, flow cytometry analysis of lymphocytes from PPs and MLN showed decreased Th1 and Th17 cell infiltration in roxadustat-treated CIA mice (figure 3), which were associated with lower serum IFN-γ and IL-17A levels (figure 3I). IgA+B220 cells were decreased in the PPs but not in the MLN or spleen (online supplemental figure 4G and H), while Treg cells were increased in the PPs and MLN, indicating the induction of an immune regulatory environment in the gut of roxadustat-treated CIA mice (online supplemental figure 4I and J).

Next, we tested whether roxadustat alters HIF1α and HIF2α expression in IECs. While there was no significant effect on HIF2α, HIF1α protein level was significantly increased by the treatment (figure 3J). TUNEL staining and electron microscopy scans confirmed that roxadustat protected IEC viability (figure 3K, online supplemental figure 4K). mRNA levels of caspase-3, caspase-8, ripk3 and mlkl, but not Ripk1, were decreased in IECs from roxadustat-treated CIA mice (figure 3L and M). Protein levels of cleaved caspase-3, cleaved caspase-8, RIPK3, MLKL and RIPK1 were also decreased in the IECs of roxadustat-treated CIA mice (figure 3N). qRT-PCR analyses showed that Caspase3, Caspase8, Ripk3, Mlkl, Hmgb1, Tnfrsf1a and Tnf gene expression were decreased, while Ripk1, Tnfrsf1b, Bax and Bcl-2 expression remained unchanged in organoid culture exposed to 1% oxygen after roxadustat treatment (online supplemental figure 4L). To investigate the involvement of HIF1α in cell death pathways, we stimulated WT and HIF1α-deficient organoids with tumor necrosis factor alpha (TNFα) and various Toll-like receptor agonists for 24 hours, with and without 20 μM roxadustat treatment. HIF1α-deficient organoids showed reduced viability after TNFα stimulation compared with WT organoids. Interestingly, roxadustat treatment only demonstrated a mitigating effect on TNFα-induced cell death in WT organoids (online supplemental figure 4M). Taking together, these data suggest that stabilisation of HIF1α in IECs supports their survival, maintains intestinal barrier function and ameliorates arthritis.

HIF1α expression preserves the gut epithelial cells from death

Of note, HIF1α deficiency did not affect gut integrity at a steady state, as characterised by the intestinal histology (online supplemental figure 5A–D) and the quantification of the Th1, Th17 and IgA+B220 populations in PPs, MLN and spleen (online supplemental figure 5E–H). However, in the case of CIA, histological sections of the ileum showed increased inflammation (figure 4, online supplemental figure 5I and J) and bacterial invasion (figure 4, online supplemental figure 5K) in Hif1a ΔIEC compared with littermate control mice, while there was no significant change in mucus secretion (figure 4D, online supplemental figure 5L). TUNEL staining (figure 4, online supplemental figure 5M) and electron microscopy depicted increased epithelial cell death in Hif1a ΔIEC CIA mice (figure 4F). We next isolated IECs from non-induced and CIA-induced WT and Hif1a ΔIEC mice at the onset of arthritis (day 36 pfi) and subjected cells to RNA bulk sequencing. While IECs from non-induced WT and Hif1a ΔIEC mice showed a very similar mRNA expression profile (figure 4G), 1837 genes were differentially expressed between WT and Hif1a ΔIEC mice during CIA (figure 4H). GO and KEGG enrichment analyses suggested that HIF1α deletion affected genes controlling T cell activation and proliferation (figure 4I), confirming our previous fluorescence-activated cell sorting analysis (figure 2L–N). With respect to cell death, Caspase-3 mRNA expression was downregulated, while the necroptosis gene Ripk3 was upregulated in Hif1a ΔIEC IEC from CIA-induced mice (figure 4J). Both changes were confirmed by western blot (figure 4K) and immunofluorescence of cleaved caspase-3 and phosphoryllated RIPK3 (p-RIPK3) (figure 4L). To further corroborate the interdependence between HIF1α and necroptosis markers, Hif1α was silenced by shRNA in the intestinal epithelial cell line MC38 under hypoxic conditions. Increased necroptotic markers (RIPK1, RIPK3 and MLKL) and decreased apoptotic markers (cleaved caspase-3 and cleaved caspase-8) were observed when HIF1α was knocked down in MC38 cells (online supplemental figure 5N). As a consequence, MC38 cells showed increased necrosis and less apoptosis after the knockdown of HIF1α (online supplemental figure 5O). Collectively, those data showed that HIF1α is a negative regulator of necroptosis in IECs.

Figure 4

HIF1α expression protects from intestinal epithelial cell death. Representative images of (A) H&E staining, (B) MPO, (C) PAS, (D) FISH, (E) TUNEL staining and (F) EM from the ileum of HIF1αΔIEC mice and littermate controls with CIA at day 36 pfi; scale bar 50 μm. (G–H) Volcano plot of bulk RNA sequencing showing DEGs of ileal IECs from HIF1αΔIEC mice and littermate controls, which were induced or not induced for CIA (day 36 pfi; n=each 3). (I) Dot plot of GO and KEGG pathway enrichment analysis of the DEGs of ileal IECs from CIA-induced HIF1αΔIEC mice and littermate controls at day 36 pfi. (J) Heatmap of cell death-related DEGs in IECs from HIF1αΔIEC mice and littermate controls at day 36 pfi. (K) Western blot analysis of cleaved caspase-3 and caspase-8, RIPK1, RIPK3 and MLKL in IEC lysates from HIF1αΔIEC mice and littermate controls at day 36 pfi. (L) IF staining for cleaved caspase-3 (red, upper panel), p-RIPK3 (red,lower panel) and DAPI (blue) on ileal sections from HIF1αΔIEC mice and littermate controls at day 36 pfi. Data are shown as mean±SD. Statistical significance was determined by a two-tailed unpaired Student’s t-test or Mann-Whitney U test (J). CIA, collagen-induced arthritis; DAPI, 4′,6-diamidino-2-phenylindole; DEGs, differentially expressed genes; EM, electron microscopy; FISH, fluorescence in situ hybridisation; GO, Gene Ontology; HIF, hypoxia-inducible factor; IEC, intestinal epithelial cells; IF, immunofluorescence; IL, interleukin; IFN-γ, interferon-gamma; KEGG, Kyoto Encyclopedia of Genes and Genomes; pfi, post-first immunisation; PAS, periodic acid Schiff; PP, Peyer’s patches; PHD, prolyl-hydroxylase domain; µCT, micro-CT; MPO, myeloperoxidase; RA, rheumatoid arthritis; TRAP, tartrate-resistant acid phosphatase; TUNEL, terminal deoxynucleotidyl transferase dUTP nick end labelling.

HIF1α transcriptionally represses RIPK3

To elucidate how HIF1α inhibits necroptosis, small intestine organoids from Hif1a ΔIEC and littermate control mice were exposed to hypoxic conditions. While WT organoids survived, Hif1a ΔIEC organoids showed a deteriorated morphology (figure 5A and B). Immunofluorescence staining and qRT-PCR analyses showed increased mRNA levels of necroptotic genes Ripk1, Ripk3 and Mlkl, as well as their phosphorylated protein forms, but reduced expression of caspase-3 and caspase-8 mRNA or cleaved caspase-3 and caspase-8 proteins in Hif1a ΔIEC organoids (figure 5C and D). To illustrate the function of HIF1α at the transcriptional level, we performed HIF1α ChIP-seq analysis on IECs from CIA WT and Hif1a ΔIEC mice. HIF1α-regulated genes were defined after ChIP-seq data annotation (figure 5E) and comparison (figure 5F). GO and KEGG enrichment analyses showed that direct HIF1α target genes were associated with epithelial cell death and proliferation (figure 5G). Genomic data were visualised for Ripk1, Ripk3 and Mlkl (figure 5H). Strong binding activity of HIF1α was observed at the Ripk1, Ripk3 and Mlkl transcription start sites. Ripk1, Ripk3 and Mlkl HREs binding activities within 5000 bp upstream of the promoter region were then confirmed by HIF1α ChIP-qPCR (figure 5I). Each of the validated HRE fragments was cloned into a luciferase reporter plasmid for subsequent functional analyses in MC38 cells. Luciferase activity displayed a dose-dependent decrease for HRE2 and HRE3 from the Ripk3 promoter (figure 5J). To explore possible regulatory functions