Inhibitory effect of baricitinib on microglia and STAT3 in a region with a weak blood–brain barrier in a mouse model of rheumatoid arthritis

Abstract

Objectives

In patients with RA, baricitinib not only improves arthritis symptom severity, but also patients’ neuropsychological symptoms, such as depression and fatigue. However, the cellular mechanisms through which baricitinib can affect neural activity is unexplored. While the blood–brain barrier (BBB) permeability of this drug remains unclear, Janus kinase inhibitors (JAKi) might reach the area postrema, which is a unique brain region with a weak BBB function. Our recent study demonstrated microglial activation during experimental arthritis in the area postrema. Therefore, we sought to assess the effect of baricitinib on microglia in the area postrema using the CIA mouse model.

Methods

Microglia number and morphology in the area postrema were assessed by immunostaining for ionized calcium-binding adaptor molecule-1 (Iba-1). Data were collected on post-immunization day 35 (early phase) and 84 (late phase), and compared between baricitinib- and vehicle-treated mice. The effect on signal transducers and activators of transcription (STAT3) in the area postrema was also immunohistochemically examined. Behavioural outcomes were assessed by examining feeding behaviours and sucrose preference tests.

Results

In the early phase, activated microglial levels in the area postrema were decreased by baricitinib, accompanied by the inhibition of phosphorylated-STAT3 and recovery of food intake and sucrose preference. On the other hand, baricitinib did not affect microglial morphology in the late phase.

Conclusion

Our results demonstrate that baricitinib can affect brain cells, specifically microglia, in the brain region with a weak BBB and mitigate aberrant behaviours during autoimmune arthritis, pointing to the potential therapeutic effect of JAKi on brain pathologies underpinning RA.

Introduction

RA is primarily characterized by peripheral inflammation in joint tissue. In addition to joint symptoms, neuropsychological symptoms, such as depression, fatigue and sleep disorders, depending on aberrant functions in the brain, are also common in patients with RA [1–4]. Indeed, functional MRI (fMRI) studies have shown evidence demonstrating abnormal findings in the brains of individuals with RA [5, 6]. These aberrant brain findings correlate with the severity of neuropsychological symptoms [5–7]. As such, the aberrant brain activities might contribute to symptoms of patients with RA. Recent clinical studies have shown that Janus kinase (JAK) 1 and JAK2 inhibitors decrease the severity scores of neuropsychological symptoms [8–10]. However, it remains unexplored whether JAK inhibitors (JAKi) can affect not only the peripheral inflammation but also the activities of brain cells.

Animal studies using RA models are useful to assess the cellular mechanisms linked to the brain pathological mechanisms accompanying arthritis. Indeed, our previous study has confirmed the persistent activation of microglia, a resident macrophage in the brain, throughout the experimental arthritis condition [11]. It is widely accepted that the increase in activated microglia with morphologies characterized by swollen cell bodies, known as neuroinflammation, modulate neuronal excitation and synaptic plasticity [12, 13] and play critical roles in various pathological brain conditions [14, 15]. Therefore, molecules involved in microglial activation pathways have garnered attention as therapeutic targets, and the JAK/signal transducers and activators of transcription (STAT) pathway, especially JAK1, JAK2 and STAT3 signalling, is one of them [16–19]. As such, it is important to establish whether baricitinib, a JAK1/JAK2 inhibitor clinically used for RA treatment, affects microglial activation accompanied with arthritis.

Whether baricitinib can penetrate the blood–brain barrier (BBB) remains unclear, but it may nonetheless affect microglia in the area postrema with attenuated BBB function. The area postrema located in the medulla is a brain region with permeable fenestrated capillaries; blood substances can thus reach this region. Research addressing area postrema permeability reveals the molecular weight (MW)-dependent extravasation in this region [20, 21]. Additionally, our previous work has demonstrated marked microglial activation with morphological changes in the area postrema during CIA [11]. Hence, peripherally administered baricitinib with low MW (371.4 Da) may extravasate into the area postrema parenchyma and affect microglial activation in this region.

To shed light on the effect of baricitinib on the brain cells, this study determined whether microglial activation in the area postrema during CIA is affected by orally administered baricitinib. Additionally, we also assessed behaviours resembling human neuropsychological symptoms. Herein, we demonstrate the suppression effect of baricitinib on morphological activation of microglia and phosphorylated-STAT3 (pSTAT3) expression in the area postrema.

Methods

Animals

Six-week-old male DBA/1J mice (Sankyo Labo Service, Tokyo, Japan) were housed in groups of four to six and maintained on a light/dark cycle of 12/12 h (light on at 7:00 h and off at 19:00 h) with food and water available ad libitum.

IL-6 treatment

The dose of IL-6 was determined based on another study [22]. Mice were intravenously administered recombinant mouse IL-6 (406-ML-005/CF; R&D systems, MN, USA) diluted in saline (500 ng/100 µl) or an equivalent volume of saline.

CIA

In accordance with a previous report [11], two immunizations were carried out. Briefly, bovine type II collagen (CII; 200 µg/mouse; Collagen Research Center, Japan) emulsified in complete Freund’s adjuvant (FA) and in incomplete FA (Becton Dickinson and Company, NE, USA) was administered on post-immunization day (PID) 0 and 21, respectively. Mice immunized with emulsions without CII were used as controls for CIA mice (FA group). Arthritis severity for each limb was assessed on the following scale: 0, normal; 1, swelling of digits alone or localized swelling of the wrist and ankle; 2, swelling of both digits and wrist or ankle; and 3, swelling of a whole limb. Arthritis score was defined as the sum of four limb scores. The fore-limb arthritis score (FAS) and hind-limb arthritis score (HAS) were defined as the sum of the fore-limb scores and hind-limb scores, respectively. Brain analyses were performed on PID 35 and 84, corresponding to the early and late phases, respectively.

Baricitinib treatment

The dose of baricitinib was determined based on another study [23]. Baricitinib (HY-15315; MedChem Express, China) was suspended in 0.5% methylcellulose (133-17815; Wako Chemicals, Osaka, Japan). For experiments in the early phase, administration by oral gavage (10 mg/kg per single dose) was started from 19:00 h on PID 28 and given every 12 h until 7:00 h on PID 35. For experiments in the late phase, similar treatment was performed from PID 70 until PID 84. Control animals were treated with vehicle.

Behavioural tests

For acclimation, mice were moved to single-housed cages and administered a sham oral treatment every 12 h from 19:00 h on PID 21. Baricitinib or vehicle administrations were started from 19:00 h on PID 28.

Voluntary feeding was assessed from 19:00 h on PID 28 until 19:00 h on PID 33. The weights of leftover foods were measured every 12 h to calculate the amounts of consumed food. Following 12 h fasting from 19:00 h on PID33, food intake for 6 h was measured.

A sucrose preference test was performed following a previously report [11]. Briefly, animals were habituated to the environment with the two bottles filled with either 2% sucrose (w/v) or plain water. After fasting for 12 h, sucrose preference test for 24 h was initiated from 7:00 h on PID 34. The two bottle positions were switched every 6 h. The weights of bottles prior to and after the test were measured to calculate the liquid intake. A sucrose preference of each animal was defined as the ratio of consumed sucrose-containing water to the total liquid intake.

Tissue preparation

Under anaesthesia, animals were transcardially perfused with 0.1 M PBS followed by 4% paraformaldehyde in 0.1 M phosphate buffer. Brain blocks were embedded in Optimal Cutting Temperature compound and stored at −80°C. Sections containing the area postrema were acquired using a cryostat at a thickness of 20 µm.

Immunohistochemistry

Immunostaining for ionized calcium-binding adaptor molecule 1 (Iba-1), a microglial marker, was performed as previously described [11]. Briefly, sections incubated in blocking solution were then incubated for 21 h at 4°C with the primary antibody for Iba-1 (1:4000; Wako Chemicals, Osaka, Japan). Sections were then incubated for 2 h at room temperature (RT) with Alexa Fluor 568-conjugated goat anti-rabbit IgG (1:1000; Thermo Fisher Scientific, MA, USA).

For pSTAT3 immunostainings, sections were pre-treated in the following manner: (i) HistoVT one (NacalaiTesque, Kyoto, Japan) for 20 min at 67–70°C, (ii) 2 N HCl at 37°C for 10 min, (iii) 0.1 M boric acid (pH 8.5) at RT for 10 min, and (iv) 0.03% sodium dodecyl sulfate at RT for 10 min. Following incubation with blocking solution, sections were incubated with the following primary antibodies at 4°C for 72 h: rabbit anti-pSTAT3 (#9145, 1: 200; Cell Signalling), goat anti-SOX9 (AF3075, 1:1000; R&D systems) and goat anti-Iba-1 (NB100-1028, 1:500; Novus Biologicals, CO, USA). Rabbit non-specific IgG (5742S; Cell Signalling, MA, USA) was used as an isotype control for pSTAT3. They were then incubated with Alexa Fluor 488-conjugated donkey anti-rabbit IgG and 647-conjugated donkey anti-goat IgG (1:1000; Thermo Fisher Scientific) at RT for 2 h. All sections were incubated with 4′,6-diamidino-2-phenylindole (DAPI).

Quantitative analysis for immunohistochemistry

All fluorescence images were acquired using laser-scanning confocal microscopy (FV1200; Olympus, Japan). Image analyses were performed using ImageJ (National Institutes of Health, Bethesda, MD, USA).

Iba-1 analyses were carried out in accordance with our previous report [11]. Briefly, images were thresholded by triangle method to obtain binary images, and then each microglia image was extracted using the ‘analyseparticle’ function (Supplementary Fig. S1, available at Rheumatology online). The following morphological parameters were measured in each cell: perimeter, area, ferret length, minimum ferret length, maximum and minimum diameter, aspect ratio, circularity, roundness, and solidity (Supplementary Fig. S2, available at Rheumatology online).

For pSTAT3 analysis, each DAPI image was divided into two divisions, intranuclear and extranuclear areas, by thresholding using the Huang method (Supplementary Fig. S3, available at Rheumatology online). Mean intensity of pSTAT3 in each division was then measured.

The mean value of three to four sections from each animal was used for all analyses.

Ethics

Manipulation of animals was approved by the Institutional Animal Care and Use Committee of Jikei University (Approval No. 2020-011). All experiments conformed to the Guidelines for Proper Conduct of Animal Experiments of the Science Council of Japan (2006).

Statistical analysis

Data are expressed as mean (s.e.m.). Statistical analyses were performed using R (version 3.6.1; the R Foundation for Statistical Computing, Vienna, Austria). Unpaired t-test or Mann–Whitney U test were used to compare between two groups. Spearman’s rank correlation was used for correlation analysis. To classify cells according to morphological parameters, principal component analysis (PCA), and hierarchical clustering analysis (HCA) were used. Propensity score-matching was performed between the baricitinib and vehicle groups with respect to the parameters of arthritis score, HAS and FAS. Differences were considered significant when the P-value was <0.05.

Results

Inhibition of microglial activation in the area postrema in the CIA early phase by baricitinib

Our previous study had already demonstrated the microglial activation in the area postrema on PID 35 [11]. To identify whether baricitinib could modulate the morphological activation of microglia, we first compared Iba-1 immunoreactivity in the area postrema between vehicle- and baricitinib-treated mice in the early phase (PID 35) (Fig. 1A). Isotype-control images are shown in Supplementary Fig. S4, available at Rheumatology online. Iba-1-positive cells with highly activated morphologies characterized by swollen cell bodies were mainly observed in vehicle-treated mice (Fig. 1A, left panels). On the other hand, Iba-1-positive cells in baricitinib-treated mice primarily had a morphology characterized by a small cell body (Fig. 1A, right panels). To objectively analyze their morphologies, we obtained 10 morphological parameters (Supplementary Fig. S2, available at Rheumatology online). The mean value of 8 parameters out of the 10 showed significant differences between groups (Supplementary Fig. S5, available at Rheumatology online). These eight parameters corresponded to the parameters in our previous study, which were significantly different between CIA and control mice [11], emphasizing that baricitinib inhibited the morphological changes induced by CIA. PCA and HCA using 10 parameters classified these 1828 cells by their morphology. PCA provided us the first (PC1) and second principal component (PC2) (Supplementary Table S1, available at Rheumatology online). HCA using PC1 and PC2 automatically divided all microglia into two clusters, cluster-A and -B (Fig. 1B). Cluster-A cells with a high perimeter, area and minor diameter values were characterized by swollen cell bodies, which was one of the characteristics of activated microglia [24]. Thus, cluster-A cells were corresponded to activated microglia. Although the data from cluster-B cells did not show any differences (Fig. 1C and F), the number and proportion of cluster-A cells were significantly lower in the baricitinib group (Fig. 1D and E). These results demonstrated that baricitinib attenuated the CIA-induced microglial activation in the area postrema during the early phase.

Inhibition of pSTAT3 expression in the area postrema in the CIA early phase by baricitinib

Previous studies have revealed that STAT3 contributes to microglial activation [16–19, 25]. To assess STAT3 activity in the area postrema during CIA, we compared pSTAT3 immunoreactivity between CIA and FA mice in the early phase (Fig. 2). Despite weak pSTAT3 immunosignals in the FA group, marked nuclear pSTAT3 immunoreactivity was observed in the CIA group (Fig. 2A). Quantitative analyses revealed significant higher intensity in the intranuclear area in the CIA group (Fig. 2B), but no differences in extranuclear area (Fig. 2C). It is well-known that STAT3 shuttling between the nuclear inside and outside in resting state rapidly accelerates their nuclear transition if they are in activation with phosphorylation [26–28]. As such, to accurately evaluate their nuclear transition and correct experimental errors between slices, we calculated nuclear pSTAT3 intensity normalized by extranuclear intensity. The normalized intensity of nuclear pSTAT3 was significantly higher in the CIA group than in the FA group (Fig. 2D). Additionally, double immunolabelling using antibodies for Iba-1 and SOX9 detected pSTAT3 expression in microglia and astrocytes (Supplementary Fig. S6, available at Rheumatology online).

Next, to determine whether baricitinib affects STAT3 activity, we compared the pSTAT3 immunoreactivity between baricitinib- and vehicle-treated CIA mice (Fig. 3). Despite marked pSTAT3 expression in the vehicle group, the baricitinib-treated group showed weak immunosignals (Fig. 3A). Although pSTAT3 intensity in intra- and extranuclear area was not different (Fig. 3B and C), normalized intensity of nuclear pSTAT3 was significantly lower in the baricitinib group (Fig. 3D). To eliminate the impact of group differences in peripheral inflammation, we performed propensity score matching to balance the CIA severity of the two groups using a logistic regression model with adjustment for arthritis score, FAS and HAS. Even after propensity score matching, normalized intensity of nuclear pSTAT3 was still significantly lower in the baricitinib group (Supplementary Fig. S7, available at Rheumatology online). These results suggested that baricitinib inhibited CIA-induced pSTAT3 expression in the area postrema.

Inhibition of IL-6-induced pSTAT3 expression in the area postrema by baricitinib

Considering CIA improvement by baricitinib, there is a possibility that the inhibition of pSTAT3 expression in the area postrema was only a secondary result of arthritis improvement. Therefore, to assess the effects of baricitinib in non-arthritic mice, we analysed pSTAT3 immunoreactivity in the area postrema following the i.v. administration of IL-6. Marked STAT3 expression in the area postrema was already observed 15 min later (Supplementary Fig. S8, available at Rheumatology online). This rapid pSTAT3 expression was then strongly inhibited when baricitinib was administered 90 min prior (Supplementary Fig. S9, available at Rheumatology online).

Behavioural alterations of CIA mice by baricitinib

The area postrema may modulate bodyweight and feeding behavior [29]. Our previous study demonstrated a significant correlation of the number of activated microglia in the area postrema with bodyweight loss in CIA mice on PID 35 [11]. A similar correlation was observed in the vehicle group (Supplementary Fig. S10, available at Rheumatology online). Therefore, we examined the influence of baricitinib on bodyweight and feeding behaviour in the early phase. Baricitinib significantly attenuated not only the arthritis score but also bodyweight loss (Fig. 4A). The significant correlation of arthritis score with bodyweight changes was sustained in the vehicle group, but not in the baricitinib group (Supplementary Fig. S11, available at Rheumatology online). In feeding behavioural tests, cumulative consumed volume during voluntary feeding (Fig. 4B) and food consumption levels following fasting (Fig. 4C) were significantly higher in the baricitinib group.

Bodyweight and food intake could be modulated by depression [30]. Therefore, we also assessed the effect of baricitinib on depression-like behaviours by sucrose preference tests (Fig. 5A). Sucrose preference was significantly higher in the baricitinib group (Fig. 5B). Liquid intake revealed no differences (Fig. 5C).

Influence of baricitinib on microglial morphology in the late phase

Our previous study demonstrated sustained microglial activation in the area postrema and bodyweight loss even on PID 84 [11]. Thus, we examined effects of baricitinib in the late phase (PID 84) (Fig. 6). Despite CIA-severity improvement, baricitinib could not attenuate bodyweight loss (Supplementary Fig. S12, available at Rheumatology online). In both the vehicle- and baricitinib-treated groups, Iba-1-positive cells with swelling cell bodies were often observed (Fig. 6A). Ten morphological parameters of microglia did not show any differences between groups (Supplementary Fig. S13, available at Rheumatology online). Morphological classification by PCA (Supplementary Table S2, available at Rheumatology online) and HCA divided them two clusters, cluster-A′ and cluster-B′ (Fig. 6B). Cluster-A′ cells with high values for perimeter and Feret diameter were considered a cell group with highly activated forms. The number of total and cluster-B′ cells were significantly lower in the baricitinib group (Fig. 6C and F). However, unlike the results in the early phase, there were no differences in the number and proportion of cluster-A′ cells between groups (Fig. 6D and E).

Discussion

The novelty of this study lies in its demonstration that systemically administered JAKi can affect not only peripheral arthritis but also microglial activation, i.e. neuroinflammation, in the brain. Many animal studies point to the possibility of microglia as a therapeutic target in various neuropsychological symptoms [14, 15, 31, 32]. Clinical studies, meanwhile, demonstrate that baricitinib attenuates neuropsychological symptoms with RA [8, 10]. In our study, simultaneously with microglial activation, baricitinib treatment in CIA mice also modulated their aberrant behaviours. These findings point to the possibility of new mechanisms linked to the brain cells underlying improvement of the neuropsychological symptoms by this drug.

STAT3, known as a transcription factor participating in peripheral inflammation process, also contributes to neuroinflammation. Brain-STAT3 activation has been reported in some animal models of brain dysfunction, such as depression and cognitive disorder, which are known to be accompanied by microglial activation. Brain-specific interventions in these models have demonstrated the contribution of STAT3 in microglial activation [19, 33].

In RA animal models, despite reports showing pSTAT3 expression in peripheral organs [34, 35], STAT3 signalling in central tissues has remained unexplored. This study is the first to demonstrate STAT3 activation in the brain of a RA model. Additionally, we clarified that baricitinib, which suppresses STAT3 in the periphery [23, 34] could also inhibit pSTAT3 expression and neuroinflammation in the area postrema. Several animal studies have shown that neuroinflammation is suppressed by intervention in microglial and astrocytic JAK1/2-STAT3 activity [16–19]. For example, studies using cultured microglia cells show that JAK1/2 inhibitor suppresses their activation with phosphorylated JAK2 and STAT3 expression [16–18]. Genetic overexpression of the astrocytic endogenous inhibitor of STAT3 suppresses microglial activation in Alzheimer’s and Huntington’s disease models [19]. Although this study is limited in that we could not demonstrate the causality between microglial activation and STAT3 signalling, its expression could be detected within the nucleus of astrocyte and microglia, whose STAT3 signalling contributes to microglial activation. Our findings suggest that the JAK1/JAK2-STAT3 pathway might contribute to neuroinflammation. Our results are fruitful in that they point to the need to consider the effects on the brain during the treatment of arthritis. Determining the causal relationship between brain STAT3 and microglial activation during arthritis is an important future research study.

It remains an intriguing problem whether baricitinib can directly affect the area postrema. Research addressing the vascular permeability of the area postrema has revealed the MW-dependent extravasation in this region. For example, Morita et al. demonstrated signs of extensive leakage of FITC-dextran (300 Da) and the slow permeation of FITC-BSA (70 kDa) [20]. Baricitinib (371.4 Da) is small enough to reach the area postrema parenchyma.

Additionally, our results using exogenous IL-6 point to the direct actions of baricitinib in the area postrema. Research analysing STAT3 distribution in rat brains 60 min following an IL-6 i.v. administration shows marked STAT3 activation localized in the brain regions with weak BBB function, including the area postrema, despite no expression in the other regions protected by the BBB [22]. This evidence points to direct IL-6-induced downstream signals in the area postrema. The pSTAT3 suppression after an IL-6 injection by baricitinib in the present study suggested the area postrema reachability of this drug. Baricitinib permeation may likely contribute to the inhibitory effect of pSTAT3 expression during CIA.

In light of the influence of activated microglia on neural and synaptic activities [12, 13], neuroinflammation attenuation by baricitinib in the area postrema with neural projections to the nuclei which regulate physiological activities may be not limited to local changes, but may affect entire brain networks [36]. Therefore, it remains important to determine whether baricitinib treatments in CIA mice affect not only cellular activity but also behavioural outcomes reflecting changes in brain networks.

Depressive behaviour in CIA mice has already been demonstrated by the sucrose preference test [37], one of the most widely used tests for animal depression-like behaviour [38]. Although the roles of microglia in the area postrema remain almost entirely unexplored, many animal studies have pointed to the microglial regulation of depressive behaviours [39, 40]. Furthermore, our previous report in CIA mice has pointed to a negative correlation between mRNA expression of activated-microglia markers in the area postrema and sucrose preferences [11]. The present results showing sucrose preference recovery and microglial inactivation by baricitinib suggest that baricitinib may improve depression-like behaviour with CIA through the modulation of neuroinflammation.

The dramatic change in the correlation coefficient between arthritis score and bodyweight loss, with or without baricitinib, is interesting. Despite the significant correlation in vehicle-treated mice, bodyweight changes did not correlate with arthritis score in baricitinib-treated mice. If bodyweight changes are an entirely dependent variable on arthritis score, the correlation should be maintained. This difference indicates that not only peripheral inflammation but also other factors, including the brain, might underlay the mechanisms linked to bodyweight recovery by baricitinib. Based on the report showing contribution of the aberrant reward system in depressive mice to decreased food intake [30], it may be considered that depression mitigation contributes to the bodyweight regain in baricitinib-treated mice. Thus, it is estimated that baricitinib can affect feeding behaviours and bodyweight loss through the modulation of brain activity. On the other hand, there is a limitation in our behavioral tests. We observed only behavioral changes in drinking and feeding in this study; therefore, it remains possible that baricitinib altered sucrose preference by mediating taste and appetite. Future work remains necessary to examine whether baricitinib can mitigate other depressive behaviours.

In patients with RA, subjective symptoms may be dissociated from objective findings linked to arthritis. For instance, a mismatch in the perception of disease activity between the patients and physicians [41] and residual subjective symptoms in low disease activity [42] are reported. Although it remains rather unclear how such inexplicable symptoms are formed, fMRI analyses have shown interesting results. In this study, the abnormal brain connectivity of patients with RA did not correlate with objective findings, such as swollen joint count and serum CRP, but was found to significantly correlate with subjective parameters, such as tender joint count and the severity of pain and fatigue [6]. These findings point to the possibility of the brain contributing to the mismatched symptoms with peripheral inflammation.

Our previous study has shown that neuroinflammation in the area postrema during CIA is sustained in the late phase [11]. Interestingly, despite the therapeutic effect on arthritis in both early and late phases, the late phase treatment could not inhibit microglial activation in the area postrema. Additionally, baricitinib in the late phase did not attenuate bodyweight loss. The remains of neuroinflammation and bodyweight loss despite arthritis mitigation is reminiscent of the residual symptoms in RA. To further improve subjective symptoms in patients with RA, a new therapeutic strategy targeting the neuroinflammation might be fruitful.

The limitation of our study includes that we cannot verify the reachability of baricitinib via other possible pathways. The following possibilities should also be considered as pathways through which baricitinib might affect the brain: (i) a pathway via BBB transport, and (ii) a pathway by free diffusion from the regions with attenuated BBB function. The role of these other pathways needs to be examined in the future.

Our study demonstrated microglial activation with STAT3 activation in the area postrema, a brain region with attenuated BBB function, during CIA. Baricitinib could inhibit not only the peripheral arthritis but also neuroinflammation in the area postrema, which suggested that baricitinib may directly affect brain cells. Further investigations into the influence of this drug on the higher CNS are warranted.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Contributions statement

The research project was designed by all authors. T.M. performed the experiments and analysed the data. T.M. and D.K. wrote the manuscript. All authors approved the final version of the manuscript.

Funding

This work was supported by JSPS KAKENHI Grant Number 22K08551. D.K. was partly supported by Eisai Co., Ltd, Chugai Pharmaceutical Co., Ltd, Ayumi Pharmaceutical Co., Ltd, Astellas Pharma Inc., Mitsubishi Tanabe Pharma Co., Pfizer Inc., Daichi Sankyo Co., Ltd, Eli Lilly Japan K.K., Takeda Pharmaceutical Co., Ltd, AbbVie GK, Asahi Kasei Pharma Co. and Nihon Kayaku Co., Ltd.

Disclosure statement: D.K. received research grants from Eisai Co., Ltd, Chugai Pharmaceutical Co., Ltd, Ayumi Pharmaceutical Co., Ltd, Astellas Pharma Inc., Mitsubishi Tanabe Pharma Co., Pfizer Inc., Daichi Sankyo Co., Ltd, Eli Lilly Japan K.K., Takeda Pharmaceutical Co., Ltd, AbbVie GK, Asahi Kasei Pharma Co. and Nihon Kayaku Co., Ltd. None of these pharmaceutical industries had any role in the design of the study, the collection, analysis and interpretation of data, or in the writing of the manuscript.

Acknowledgements

We are grateful to Prof. Fusao Kato and Dr Yukari Takahashi for their invaluable advice. We also acknowledge the technical and blinding assistance of Ms Ying Kaku. We would like to thank Editage (www.editage.com) for English language editing.

References