Objectives Prevotella copri is considered to be a contributing factor in rheumatoid arthritis (RA). However, in some non-Westernised countries, healthy individuals also harbour an abundance of P. copri in the intestine. This study investigated the pathogenicity of RA patient-derived P. copri (P. copriRA) compared with healthy control-derived P. copri (P. copriHC).
Methods We obtained 13 P. copri strains from the faeces of patients with RA and healthy controls. Following whole genome sequencing, the sequences of P. copriRA and P. copriHC were compared. To analyse the arthritis-inducing ability of P. copri, we examined two arthritis models (1) a collagen-induced arthritis model harbouring P. copri under specific-pathogen-free conditions and (2) an SKG mouse arthritis model under P. copri-monocolonised conditions. Finally, to evaluate the ability of P. copri to activate innate immune cells, we performed in vitro stimulation of bone marrow-derived dendritic cells (BMDCs) by P. copriRA and P. copriHC.
Results Comparative genomic analysis revealed no apparent differences in the core gene contents between P. copriRA and P. copriHC, but pangenome analysis revealed the high genome plasticity of P. copri. We identified a P. copriRA-specific genomic region as a conjugative transposon. In both arthritis models, P. copriRA-induced more severe arthritis than P. copriHC. In vitro BMDC stimulation experiments revealed the upregulation of IL-17 and Th17-related cytokines (IL-6, IL-23) by P. copriRA.
Conclusion Our findings reveal the genetic diversity of P. copri, and the genomic signatures associated with strong arthritis-inducing ability of P. copriRA. Our study contributes towards elucidation of the complex pathogenesis of RA.
- rheumatoid arthritis
- arthritis, experimental
Data availability statement
Data are available on reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
WHAT IS ALREADY KNOWN ON THIS TOPIC
Rheumatoid arthritis (RA) has been associated with a relative expansion of faecal Prevotella copri, which is reportedly the dominant bacteria in the gut of healthy individuals with non-Westernised lifestyle.
P. copri exhibits a highly genetic and functional diversity in accordance with host’s lifestyle.
WHAT THIS STUDY ADDS
Comparative genomic analysis revealed that P. copri derived from patients with RA (P. copriRA) possesses a specific genomic region as a conjugative transposon (CTn). These unique genes within CTn were mediated by horizontal gene transfer.
P. copriRA induced more severe arthritis in mouse arthritis models than P. copri from healthy controls.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
The selective elimination of P. copriRA may be an effective therapeutic method for patients with RA who harbour P. copriRA in their intestine.
Rheumatoid arthritis (RA) is a systemic autoimmune disease that affects multiple joints.1 Although treatment has rapidly progressed, the aetiology of RA is still not fully understood.2 Both genetic and environmental factors are thought to contribute to the development of RA.3 Environmental factors causing RA include smoking, hormonal factors, infection and the composition of the gut microbiota.4–7 Among these factors, the gut microbiota plays a pivotal role in the development of murine models of arthritis.8 9 Many studies have demonstrated the altered composition of the intestinal microbiota (dysbiosis) in patients with RA.10–13 Several studies have shown that some patients with RA harbour an increased proportion of Prevotella species, especially Prevotella copri, in the intestine.14–16 Moreover, a recent study showed that Prevotella species are dominant in the intestine of some individuals at preclinical stages of RA.17 Immune responses against P. copri-derived antigen are shown to be associated with RA pathogenesis.18 19 However, in non-Westernised countries, healthy individuals with a plant-rich diet also harbour an abundance of P. copri possessing plant polysaccharide-metabolising enzymes in the intestine.20–25 Thus, it is controversial as to whether P. copri is harmful or beneficial to the host.
A study involving whole genome sequencing of P. copri isolated from the human intestine and large-scale metagenomic sequencing, showed that this bacterium is diverse, consisting of four genetically-distinct clades.26 27 Genomic variations in P. copri have been shown to be linked to the utilisation of polysaccharides.28 Strain-level diversity of P. copri was also investigated by metagenomic shotgun sequencing of the gut microbiota, and differences were shown to be dependent on the host’s diet.29 However, it remained to be determined whether P. copri strains present in the intestine of patients with RA were different from those of healthy individuals.
In this study, we isolated P. copri strains from patients with RA and healthy individuals from two different geographical regions, and characterised their genomic structures and functional potency, particularly in terms of their pathogenic effects on arthritis.
Materials and methods
Materials and methods are provided in online supplemental information.
Isolation of P. copri from patients with RA and healthy controls
To examine whether P. copri derived from patients with RA (P. copriRA) and P. copri from healthy controls (P. copriHC) possess differential effects on arthritis development, we isolated P. copri from the faeces of patients with RA and healthy individuals. For this purpose, we recruited 17 patients with RA and 14 healthy controls (cohort 1) who were the participants in the previous study15 and 55 patients with RA and 28 healthy individuals (cohort 2). There was no apparent difference of dietary habitants between patients with RA and healthy individuals (online supplemental table 1). Among them, faeces of two patients with RA from cohort 1, eight patients with RA and six healthy controls from cohort 2 with the high abundance of P. copri were used for isolation of P. copri (online supplemental figure 1A,B). It is possible that P. copri present in healthy individuals living in different geographical regions has distinct characteristics. Therefore, we also recruited healthy Indians, who showed the high abundance of P. copri, and two faeces were used for the isolation.21 Information on disease condition of the participants was summarised in online supplemental table 2. We successfully obtained 12 isolates from 5 patients with RA (P9, F3, N001, N016, N115), 9 isolates from three healthy Japanese controls (H012, H019, H105) and 7 isolates from two healthy Indian controls (Ind63, Ind117) (online supplemental figure 1C). These isolates were confirmed to be P. copri by 16S rRNA gene sequencing.
Differential genomic structures of P. copri from patients with RA and healthy controls
To investigate the genetic differences between P. copriRA and P. copriHC, we performed whole genome sequence analysis of 28 isolates of P. copri from 5 patients with RA (12 isolates), 5 healthy controls (16 isolates) and the reference strain JCM 13464. Next, we performed average nucleotide identity analysis and defined isolates with >99% similarity as identical (online supplemental table 3). As a result, 28 isolates we obtained and whole genome sequenced were found to be 13 different strains. We isolated five strains (RA-P9, RA-F3, RA-N001, RA-N016, RA-N115) of P. copri from patients with RA (P. copriRA), three strains (HC-H012, HC-H019, HC-H105) of P. copri derived from Japanese healthy controls (P. copriHC) and five strains (HC-Ind63-1, HC-Ind63-2, HC-Ind63-3, HC-Ind117-1, HC-Ind63-2) from Indian healthy controls (online supplemental table 4). The predicted protein coding sequences (CDSs) were obtained in all the isolates by homology based in silico annotation. To define key genomic components, we constructed the pangenome of P. copri from 13 strains. This analysis showed that each P. copri possessed 2817–3661 genes (online supplemental figure 2A), a core genome of 605 genes and a pangenome consisting of 13 482 genes (online supplemental figure 2B).
Phylogenetic analysis was performed based on 284 conserved genes using genome sequences obtained from 15 strains in a previous study,26 our 13 strains and the reference strain JCM 13464. Consistent with previous studies,24 26 28 the 13 P. copri strains were genetically categorised into three distinct clades (figure 1A). The majority of the strains (11 out of 13 strains) belonged to clade A. Among the 11 strains belonging to clade A, 4 strains were derived from patients with RA (4 out of 5 strains) and 7 were from healthy controls (7 out of 8 strains). Thus, there were no apparent differences in conserved gene contents between P. copriRA and P. copriHC. These results suggested that the pangenome of P. copri could be classified as an open pangenome with high genome plasticity.
We next focused on the species-specific regions. Overviews of genome annotation for the 13 P. copri strains revealed that P. copriRA possessed a unique genomic region outside of the core region (figure 1B). We then analysed whether this genomic region was present only in microbiota of patients with RA. Using the data of faecal metagenome shotgun sequencing of patients with RA and healthy controls including participants of other previous study (cohort 3),30 we assessed the number of reads for this genomic region. The relative number of this genomic region in faecal samples from patients with RA was higher than those of healthy controls (figure 1C, online supplemental figure 2C). We further analysed whether this genomic region was present in other bacterial species by performing homology search of this sequence in blastn pipeline. Among registered bacterial strains, only one bacterium, P. copri strain YF2 chromosome (Accession number: CP0424641), was found to possess this region. Thus, this genomic region was present only in P. copri.
To analyse whether this additive genomic region of P. copriRA was associated with the pathogenesis of RA, we analysed protein CDSs in this region by a nucleotide blast search (figure 2). The size of the P. copriRA-specific region was approximately 100 kilobase pair (kbp) and this region encoded approximately 90 to 100 CDSs. Almost 70% of the CDSs in this region were hypothetical proteins with no characterised function. Genetic plasticity of the bacterial pangenome is triggered by genetic rearrangement including inversions, duplications and horizontal gene transfer (HGT) mediated by a variety of mobile and mobilisable elements such as a conjugative transposon (CTn).31 32 CTns, which are integrated into the microbial genome, are reported to transfer genes involved in antibiotic resistance,33–35 the use of alternative carbon sources36 37 and virulence factors.38 39 The P. copriRA-specific region of each strain contained sequences sharing a set of backbone genes for integration-related and conjugal transfer-related functions, which were originally identified in Porphyromonas, Prevotella and Bacteroides species.40–42 An integrase gene was localised at the start of the P. copriRA-specific genomic region. In addition, CTn-related genes, such as the cluster of transfer genes including genes encoding DNA primase and mobilisation protein/coupling protein, were observed. Moreover, the sequences of these genes in this specific genome region were highly conserved among P. copriRA strains. These results indicated that the P. copriRA-specific region was a CTn, which we named CTnPc, on the pangenome of P. copriRA. To examine the insertion site for CTnPc, we analysed the region adjacent to CTnPc on the genome. In strains from the same patient, CTn was present at the same position on the genome. However, the insertion site was different for each patient, with the exception of patients P9 and N001. These results suggested that CTnPc was inserted into different sites on the genome of each strain driven by HGT events.
Development of severe collagen-induced arthritis in mice harbouring P. copriRA
We next analysed the arthritis-inducing ability of P. copriRA and P. copriHC using mouse models. We orally administered 6.0×1010 of P. copriRA (RA-N001) and P. copriHC (HC-H012) to DBA/1 J mice reared under specific-pathogen-free (SPF) conditions; however, both P. copriRA and P. copriHC failed to colonise the intestine of SPF mice (online supplemental figure 3A,B). Therefore, in a subsequent experiment, SPF-DBA/1J mice were first orally treated with combinations of antibiotics (ampicillin, neomycin, metronidazole and vancomycin) for 5 days to reduce the intestinal microbiota, and then orally inoculated with 6.0×1010 of P. copriRA (RA-N001) or P. copriHC (HC-H012) for five consecutive days (online supplemental figure 3C). Under these conditions, both P. copriRA (RA-N001) and P. copriHC (HC-H012) successfully colonised the intestines to a similar extent (online supplemental figure 3D). Using these DBA/1J mice harbouring P. copriRA (RA-N001) or P. copriHC (HC-H012), we induced collagen-induced arthritis by challenging on two occasions (days 0 and 21) with type II collagen (online supplemental figure 3E). On day 21, at the time point of the second challenge, monitoring of the arthritis score and the incidence of arthritis was initiated. DBA/1J mice harbouring P. copriRA (RA-N001) developed more severe arthritis, as evidenced by a higher arthritis score (figure 3A) and an earlier onset and higher incidence of arthritis (figure 3B), compared with mice harbouring P. copriHC (HC-H012). More serious arthritis development in DBA/1J mice harbouring P. copriRA (RA-N001) was confirmed by more severe swelling of ankle joints (figure 3C) and more severe histological changes in ankle joints, such as proliferation of synovial lining cells, infiltration of inflammatory cells and bone destruction associated with pannus formation (figure 3D,E). Moreover, the serum concentration of type II collagen-specific IgG at day 63 was higher in DBA/1J mice harbouring P. copriRA (RA-N001) than in mice harbouring P. copriHC (HC-H012) (figure 3F). The more severe arthritis observed in DBA/1J mice harbouring P. copriRA (RA-N001) was not due to the preservation of a higher number of P. copriRA (RA-N001) in the intestine because there were almost equal proportions and numbers of P. copriRA (RA-N001) and P. copriHC (HC-H012) in the intestines of DBA/1J mice during the course of the experiments (online supplemental figure 4A–C).
Because effector T helper (Th) cells, particularly Th1 and Th17 cells, have been shown to be involved in the pathogenesis of collagen-induced arthritis,9 43 44 we next analysed the number of Th1 and Th17 cells in popliteal lymph nodes (PLN). At 77 days after the first immunisation, cells were isolated from PLNs, stimulated with phorbol 12-myristate 13-acetate and ionomycin, and analysed the production of IFN-γ and IL-17A from CD4+ T cells by flow cytometry. Although the proportion of Th1 and Th17 cells remained unaltered, the number of Th1 and Th17 cells was increased in DBA/1J mice harbouring P. copriRA compared with mice harbouring P. copriHC (figure 3G,H).
We then analysed whether other P. copriRA strains possess the arthritis-inducing ability and CTnPc is implicated in the pathogenesis. For this purpose, each of P. copriRA strains with CTnPc (RA-P9, RA-N001, RA-N016: RA+), a P. copriRA strain without CTnPc (RA-N115: RA−), P. copriHC strains (HC-H012, HC-H019, HC-H105: HC) was colonised in the intestine of DBA/1J mice (online supplemental figure 5A). For control, DBA/1J mice with oral administration of phosphate buffered saline (PBS) was also prepared. Severities of collagen-induced arthritis in DBA/1J mice harbouring a P. copriRA strain without CTnPc (RA-N115) or P. copriHC strains (HC-H012, HC-H019, or HC-H105) were similar to those with DBA/1J mice with PBS. In contrast, DBA/1J mice harbouring P. copriRA strains with CTnPc (RA-P9, RA-N001, or RA-N016) developed severe arthritis, as evidenced by higher arthritis score and incidence (online supplemental figure 5B,C). We analysed the proportion and number of immune cell subsets including CD19+ B cells, CD4+ and CD8+ T cells in spleen (online supplemental figure 5D,E) and mesenteric lymph node (online supplemental figure 5F,G), CD11b+CD11c+, CD11b+CD11c−, and CD11b−CD11c+ myeloid cells in lamina propria of large intestine (online supplemental figure 5H,I) in addition to Th1 and Th17 CD4+ T cells in PLN (online supplemental figure 5J,K). Although the proportion of all of these immune cells were similar among all the mice, the number of only Th17 cells in PLN was increased in DBA/1J mice harbouring P. copriRA strains with CTnPc (RA-P9, RA-N001, or RA-N016) (online supplemental figure 5K). Taken together, these results suggested that the colonisation of P. copriRA in the intestines induced severe inflammation in a collagen-induced arthritis model with slight increases in the Th1 and Th17 responses.
Development of severe arthritis in SKG mice harbouring P. copriRA
We next analysed whether P. copriRA alone can induce severe arthritis. We orally inoculated germ-free SKG mice with 1.0×108P. copriRA (RA-N001) and P. copriHC (HC-H012) once (online supplemental figure 6A). Both P. copriRA (RA-N001) and P. copriHC (HC-H012) similarly colonised the intestines of SKG mice, as evidenced by the similar number of P. copri detected in the faeces by quantitative (Q)-PCR (online supplemental figure 6B). We also performed fluorescence in situ hybridisation of the large intestine of SKG mice inoculated with P. copriRA (RA-N001) and P. copriHC (HC-H012) using a bacteria-specific probe (online supplemental figure 6C). Both P. copriRA (RA-N001) and P. copriHC (HC-H012) were found to be present in the lumen of the large intestine.
Then, SKG mice monocolonised with P. copriRA (RA-N001) or P. copriHC (HC-H012) were injected with zymosan to induce arthritis (online supplemental figure 6D).45 SKG mice monocolonised with P. copriRA (RA-N001) developed more severe arthritis after zymosan treatment compared with SKG mice monocolonised with P. copriHC (HC-H012), as shown by the higher arthritis score (figure 4A) as well as the higher incidence of arthritis (figure 4B). The more severe development of arthritis in SKG mice monocolonised with P. copriRA (RA-N001) was affirmed by the more severe swelling of ankle joints (figure 4C) and the more severe synovial inflammation and bone destruction in the ankle joints (figure 4D,E). These results suggested that P. copriRA has strong arthritis-inducing ability.
Aberrant Th17 responses of autoimmune T cells are implicated in the development of arthritis in SKG mice.46 47 Therefore, the number of Th1 and Th17 cells in PLNs was analysed in SKG mice monocolonised with P. copriRA (RA-N001) and P. copriHC (HC-H012). The proportion and number of Th17 cells was increased in SKG mice monocolonised with P. copriRA (RA-N001) compared with those monocolonised with P. copriHC (HC-H012) (figure 4F,G). Taken together, these findings may indicate that the colonisation of P. copriRA in the intestines induced severe arthritis via the activation of Th17 cells in SKG mice.
P. copriRA-mediated activation of innate immune cells
Based on these findings, we hypothesised that P. copriRA is a potent inducer of Th17 cells that induce inflammation in the joints, and analysed the mechanisms by which Th17 cells were induced in P. copriRA-colonised mice. There is a previous report demonstrating that the heat-stable components of P. copri activate innate immune cells (antigen-presenting cells) and increase expression of major histocompatibility complex (MHC) class II and production of Th17-related cytokines such as IL-6 and IL-23.15 Although there were no differences in the proportion or number of CD11b+CD11c+, CD11b+CD11c−, CD11b−CD11c+ cells in the large intestine (online supplemental figure 5J), the intensity of MHC class II expression in the intestinal CD11b+CD11c+ cells was upregulated in DBA/1J mice harbouring P. copriRA with CTnPc (RA-P9, RA-N001, or RA-N016) compared with mice harbouring P. copriRA without CTnPc (RA-N115) or P. copriHC strains (HC-H012, HC-H019, or HC-H105) (online supplemental figure 7A). These results suggest that intestinal colonisation with P. copriRA resulted in activation of innate immune cells in the intestinal lamina propria.
To determine whether P. copriRA is more effective at activating innate immune cells, we performed in vitro experiments using bone marrow-derived dendritic cells (BMDCs). BMDCs were stimulated with heat-killed P. copriRA (RA-N001) or P. copriHC (HC-H012) and analysed for the expression of MHC class II, CD86 and CD80 (figure 5A). Compared with P. copriHC, P. copriRA induced higher expression of MHC class II, CD86 and CD80, as evidenced by a higher mean fluorescence intensity. We also analysed the expression of mRNA of Th17-related cytokine genes, Il6 and Il23a, in BMDCs stimulated with P. copriRA or P. copriHC by Q-PCR (figure 5B). P. copriRA induced higher expression of both genes than P. copriHC. The production of IL-6 and IL-23 proteins from P. copriRA-stimulated BMDCs was also analysed by ELISA (figure 5C,D). IL-23 production was higher in P. copriRA-stimulated BMDCs (DC: P. copri ratio=1: 2.5) than in P. copriHC-stimulated BMDCs. IL-6 production was similar in BMDCs stimulated with P. copriRA and P. copriHC (DC: P. copri ratio=1: 2.5). However, when the number of P. copri used for stimulation was reduced, IL-6-inducing capacity was sustained in P. copriRA, but not in P. copriHC. We next examined whether P. copriRA possesses a higher ability to induce antigen-specific Th17 cells from naïve CD4+ T cells. BMDCs of DBA/1J mice were first treated with P. copriRA or P. copriHC. Then, naïve CD4+ T cells from DBA/1J mice were co-cultured with the P. copri-treated dendritic cells (DC: P. copri ratio=1: 2.5) for 5 days. Type II collagen was then added and cells were co-cultured for an additional 2 days. After a total of 7 days of co-culturing with BMDCs, T cells were stimulated with type II collagen and analysed for the production of IL-17A (figure 5E). T cells co-cultured with P. copriRA-stimulated BMDCs produced higher amounts of IL-17A. We also analysed the responses to other strains: P. copriRA with CTnPc (RA-P9, or RA-N016: RA+), P. copriRA without CTnPc (RA-N115: RA−) or P. copriHC (HC-H019, or HC-H105: HC). BMDCs stimulated with P. copriRA with CTnPc expressed the higher level of MHC class II (online supplemental figure 7B) and produced higher amounts of IL-6, IL-23 and TNF-a (online supplemental figure 7C) compared with those stimulated with P. copriRA without CTnPc or P. copriHC. Induction of T cell IL-17 response was induced potently by stimulation with P. copriRA with CTnPc (online supplemental figure 7D). These results indicated that P. copriRA strongly induced a Th17 response via the activation of dendritic cells.
In this study, we isolated P. copri strains from the faeces of patients with RA (P. copriRA) and healthy individuals (P. copriHC), and found that P. copriRA strains possess a unique region in their genome. Furthermore, colonisation of P. copriRA in mouse intestines caused severe pathologies in two different mouse arthritis models.
RA is an immune disorder caused by the complex interaction of genetic and environmental factors.3 Considering environmental factors, mucosal surfaces are potentially an important site for triggering RA development.3 Indeed, smoking, which affects the airway mucosa, is a prominent environmental factor contributing to the development of RA.48 Periodontitis, mainly caused by Porphyromonas gingivalis in the oral mucosa, is also implicated in RA pathogenesis.49 Furthermore, many studies in recent years have shown that dysbiosis in the intestinal mucosa is associated with susceptibility to RA.10–13 Among studies focusing on dysbiosis of the intestinal microbiota, Prevotella species, particularly P. copri, are often increased in patients with RA.14–17 However, P. copri is also predominant among the intestinal microbiota of healthy individuals with a diet rich in plant-derived fibre.20–25 Based on the findings of this study, we propose that colonisation of pathogenic P. copri, rather than Prevotella-predominant dysbiosis, in the intestines is one of the causes of RA pathogenesis.
We found that P. copriRA possesses a distinct genomic structure from that of P. copriHC. P. copri is genetically highly diverse and can be classified into at least four different clades.26 27 Among these clades, P. copri strains belonging to clade A are the most abundant.24 26 Many of the P. copri strains isolated in this study also belonged to clade A, including both P. copriRA and P. copriHC strains. This indicated that the different genomic structures between P. copriRA and P. copriHC strains did not result from distinct clades. However, we found that unique sequences were conserved among P. copriRA strains and contained CDSs characteristic to CTn-related genes that mediate HGT. Thus, P. copriRA is thought to have acquired a unique genome (CTnPc) by HGT.
A recent study isolated the antigen, Pc-p27, from P. copri from a patient with RA who was presented on HLA-DR and a triggered immune response.18 19 The Pc-p27 gene exists in four out of five P. copriRA strains and six out of eight P. copriHC strains. Thus, the Pc-p27 gene is not selectively present in P. copriRA (online supplemental table 5) and therefore does not correlate with the pathogenicity of P. copriRA. P. copriRA was more potent than P. copriHC in activating innate immune cells, as evidenced by the enhanced surface expression of MHC class II and the increased cytokine production from mouse myeloid cells. Thus, the pathogenicity of P. copriRA might correlate with the capacity to induce a strong innate immune response. By activating innate immune cells and thereby enhancing Pc-p27 antigen presentation by MHC class II, P. copriRA may induce more severe arthritis.
Within the P. copriRA-specific genome region (CTnPc), many genes encoding hypothetical proteins reside (online supplemental table 6). Genes present in the CTnPc region are expected to exert pathogenic ability to activate innate immunity, ultimately leading to the development of arthritis. To identify a pathogenic factor in the CTnPc region of P. copriRA, we attempted to generate mutant P. copriRA strains. However, we encountered difficulty in introducing gene modifications into P. copri. Recently, a novel gene tool to introduce gene insertions and deletions into P. copri has been developed,24 which may aid identification of the pathogenic factor responsible for arthritis development in a future study.
In this study, we successfully isolated pathogenic strains of P. copri from the faeces of patients with RA. Development of a system to detect P. copriRA would be a good strategy to predict the future development of RA. In addition, the development of a method to selectively eliminate P. copriRA may be an effective therapeutic measure for patients with RA who harbour P. copriRA in their intestine.
Data availability statement
Data are available on reasonable request.
Patient consent for publication
This study was approved by the ethical committee of Osaka University. Participants gave informed consent to participate in the study before taking part.
We thank Y. Magota for technical assistance, Kate Fox from Edanz (https://jp.edanz.com/ac) for editing a draft of this manuscript, and C. Hidaka for secretarial assistance.
TN and YM are joint first authors.
Handling editor Josef S Smolen
Contributors TN performed the experiments, analysed the data and wrote the manuscript; YMae performed the experiments and analyzed the data; DM, MN, YMat, TO, EO-I, TKi, MY, SK, TKu, RO, HK, MM and TS assisted with experiments; BD, SN, YO and AK supervised the study; KT planned and directed the overall project and wrote the manuscript and is the guarantor.
Funding This study was supported by grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan (JP18H04028 and JP20K20595) and the Japan Agency for Medical Research and Development (JP20gm1010004h0005).
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.