Gut Microbiome and Bile Acid Metabolism Induced the Activation of CXCR5+ CD4+ T Follicular Helper Cells to Participate in Neuromyelitis Optica Spectrum Disorder Recurrence

From the perspective of the role of T follicular helper (Tfh) cells in the destruction of tolerance in disease progression, more attention has been paid to their role in autoimmunity. To address the role of Tfh cells in neuromyelitis optica spectrum disorder (NMOSD) recurrence, serum C-X-C motif ligand 13 (CXCL13) levels reflect the effects of the Tfh cells on B-cell-mediated humoral immunity. We evaluated the immunobiology of the CXCR5+CD4+ Tfh cells in 46 patients with NMOSD, including 37 patients with NMOSD with an annual recurrence rate (ARR) of<1 and 9 patients with NMOSD with an ARR of ≥1. Herein, we reported several key observations. First, there was a lower frequency of circulating Tfh cells in patients with an ARR of<1 than in those with an ARR of ≥1 (P< 0.05). Second, the serum CXCL13 levels were downregulated in individuals with an ARR<1 (P< 0.05), processing the ability to promote Tfh maturation and chemotaxis. Third, the level of the primary bile acid, glycoursodeoxycholic acid (GUDCA), was higher in patients with NMOSD with an ARR of<1 than in those with NMOSD with an ARR of ≥1, which was positively correlated with CXCL13. Lastly, the frequency of the Tfh precursor cells decreased in the spleen of keyhole limpet haemocyanin-stimulated animals following GUDCA intervention. These findings significantly broaden our understanding of Tfh cells and CXCL13 in NMOSD. Our data also reveal the potential mechanism of intestinal microbiota and metabolites involved in NMOSD recurrence.


INTRODUCTION
Neuromyelitis optica spectrum disorder (NMOSD) is an autoimmune disease of the central nervous system that repeatedly involves the optic nerve and spinal cord and is mediated by aquaporin-4 (AQP4) antibodies. The combination of AQP4-Ab with AQP4 expressed on astrocytes of the bloodbrain barrier causes complement-dependent cytotoxicity and neutrophil, eosinophil, and cytokine infiltration. Blood-brain barrier destruction results in oligodendrocyte death, myelin loss, and neuronal damage.
Recent data have indicated that the germinal centre (GC) may be a pathogenic hotspot for autoantibody production in autoimmune diseases. In particular, circulating Tfh cells that are derived from GC-Tfh cells shuffle between the peripheral blood and lymphoid tissues (1), and play an important role in the differentiation of B cells into memory B and plasma cells, promoting pathogenic autoantibody production, clinical symptom onset, continued immune responsiveness, and eventually irreversible tissue damage (2). Tfh cells function through chemokine CXC receptor 5 (CXCR5), programmed cell death protein-1 (PD-1), cytokine interleukin-21 (IL- 21), and C-X-C motif ligand 13 (CXCL13). Multiple studies have shown that Tfh cells expand in the peripheral blood of humans of systemic autoimmune diseases, including rheumatoid arthritis (3), primary Sjogren's syndrome (4), and immunoglobulin (IgG)-4 related disease (5). Recently, Tfh cells were demonstrated to be reportedly involved in the recurrence of neuroimmune diseases, such as multiple sclerosis (6) and NMOSD (7).
In general, the intestinal microbiota of patients with NMOSD are characterised by lower Clostridium, Parabacteroides, Oxalobacter, and Burkholderia and higher Streptococcus, Alistipes, Haemophilus, Veillonella, Butyricimonas, and Rothia abundance than in healthy controls (8,9). In a previous study, not only did the patients with NMOSD have significantly lower faecal butyrate levels, but the patients with lower Expanded Disability Status Scale (EDSS) scores also showed a major reduction in the butyrate levels (9). The anti-inflammatory effect of short-chain fatty acids is not limited to the intestinal tract; it also increases the Treg level and inhibits Th17 cell differentiation (10). However, limited information is available on the effects of the gut microbiota and metabolites on Tfh and NMOSD recurrence. Therefore, we assessed the Tfh cell number and frequency in patients with NMOSD with low and high annual recurrence rates (ARRs). We then investigated the effects of the metabolites on the immune system in keyhole limpet haemocyanin (KLH)stimulated animals. We aimed to provide valuable insights into the causal mechanisms underlying the possible clinical effects of intestinal microbiota and metabolites.

Clinical Study Design and Population
A total of 109 participants, including 59 patients and 50 age-and sex-matched healthy controls, were enrolled in this study from October 2019 to July 2021 at the Third Affiliated Hospital of Sun Yat-Sen University in southern China. The diagnosis of NMOSD was established based on the International Panel on NMO Diagnosis 2015 criteria (11) and AQP4-IgG seropositivity. We included patients according to the following inclusion criteria: (1) initiation of immunosuppressive treatment as a first-line therapy within 3 years following disease onset and (2) at least 3 months of follow-up. We excluded patients according to the following exclusion criteria: (1) EDSS score ≥ 6.0, (2) body mass index< 20 kg/m 2 , (3) a history of cardiovascular or renal disease and other autoimmune diseases, and (4) age< 18 years. The following data were collected from the participants' medical records at the baseline visit: demographics (including age, race/ ethnicity, sex, weight, and height), past and present diet attempts, serologic status, date of disease onset (first attack), EDSS scores, and modified Rankin Scale scores. NMOSD data collection included a history of recurrence, past/current immunosuppressive therapies, disease duration, medication history, and reasons for treatment discontinuation. Disease onset was defined as the first recurrence of NMOSD. Attack and recurrence were defined as new symptoms that occurred within at least 24 h and were associated with new magnetic resonance imaging lesions. The ethical committee of the Third Affiliated Hospital of Sun Yat-sen University of Medical Sciences approved the research proposal in 10-15-2019. We have registered our trial before the first participant was enrolled in the clinical trial with a project identification code NCT04101058 that are reported in manuscripts (More information about trial registration see https://register.clinicaltrials.gov/prs/app/ action/).
Of the 59 patients with NMOSD, serum samples were collected to measure the cytokine levels, and peripheral blood mononuclear cells (PBMCs) were collected for immunological analysis from the 46 patients, who also provided faecal samples for microbiome analysis and metabolomics. These 46 patients were further categorised into two groups based on the ARR: low ARR group (n = 37, ARR< 1) and high ARR group (n = 9, ARR ≥ 1) ( Table 1). All the 50 age-and sex-matched healthy controls only supplied stool samples for the microbiome analysis.

Stool Metabolomics
Faecal metabolites were extracted using 50% methanol buffer. Faecal samples (20 mL) were added to 120 mL of pre-cooled 50% methanol, vortexed for 60 s, incubated at 4°C for 10 min, incubated at -20°C for 60 min, centrifuged at 4,000 × g at 4°C for 15 min, transferred to another tube, and then analysed by liquid chromatography-mass spectrometry (LC-MS). Additionally, pooled quench cooled samples were prepared by mixing 10 mL of each extraction mixture. These samples were then processed using an LC-MS system according to the manufacturer's instructions. An LC-MS triple quadrupole mass spectrometer (Shimadzu, LCMS-8050) equipped with LabSolutions was used to collect the primary and secondary mass spectrometry data. The electrospray ionisation ion source parameters for the negative ion mode were set as follows: nebulising gas temperature, 300°C; nebulising gas flow, 3 L/min; heating gas flow, 13 L/min; sheath gas temperature, 350°C; DL temperature, 250°C; heating module temperature, 400°C; and sheath gas flow, 7 L/min. BAs were detected using LabSolutions LCMS software (Shimadzu) to perform peak extraction, peak integration, area calculation on the original file, and quantification using a standard curve.

Animals
Female BALB/c mice aged 5-6 weeks (15-18 g) were purchased from the Laboratory Animal Center. This study complied with all the relevant ethical regulations and was approved by the Ethics Committee of the South China Agricultural University (ethical number: 2021B002). Animals were housed under specific pathogen-free conditions and maintained over a 12-h light/dark cycle with free access to food and water. Mice received an injection of KLH in both the underarms and groin (100 µg/0.2 mL/site), and the control groups were treated with equal amounts of saline. Mice were then divided into three groups (n = 6 animals per group): control, KLH, and KLH +glycoursodeoxycholic acid (GUDCA) acid groups. Each experimental group was separately gavaged with 50 mg/kg body weight of GUDCA for 4 weeks. The control and KLH groups were treated with equal amounts of saline. After treatment, ileum, spleen, and lymph node samples were collected to analyse the Tfh cell phenotype by fluorescenceactivated cell sorting.

Statistical Analysis
All the statistical analyses were performed using the Statistical Package for the Social Sciences software version 19.0; the data are presented as mean values with standard deviation. The test of normal distribution was performed before the Student's t-test and analysis of variance (ANOVA). Multiple comparisons were performed using one-way ANOVA or Kruskal-Wallis tests between the different groups. The Shapiro-Wilk test was used to assess normality. Inflammatory cytokines levels were log 10transformed when analysed to meet the normal assumption. The correlation between the two variables was assessed using the Spearman rank test. For all analyses, a p-value of<0.05 was considered to indicate statistical significance. Correlations between each of the pair of datasets were computed using Pearson correlation coefficients, and visualisations were generated in R (v4.0.0).

CXCL13 Was Significantly Lower in Patients of NMOSD With a Low ARR
The potential value of serum inflammatory cytokine markers was investigated in the diagnosis of NMOSD ( Table 2). CXCL13 was significantly different between the two groups (51.82 vs. 78.88, p = 0.019, Figure 1A). The individual simplified signature could discriminate between the two groups based on the area under the curve (AUC) (CXCL13 [AUC = 0.656], Figure 1B). We also aimed to explore the correlations between these discriminative inflammatory factors and ARR to identify markers associated with disease recurrence. Spearman's rank correlation analysis showed that CXCL13 (r = 0.460, p = 0.001) was positively correlated with NMOSD recurrence ( Figure 1C).

Tfh Cells Were Positively Correlated With NMOSD Recurrence
The frequency of circulating CXCR5 + CD4 + Tfh, CXCR5 + CCR7 low Tfh precursor, and CXCR5+CCR7 hi resting Tfh cells were analysed using flow cytometry. As shown in Figure 2A, the percentage of CD4 + CXCR5 + T cells was significantly higher in the peripheral blood of patients with NMOSD with an ARR of ≥ 1 than in those with an ARR of <1 (24.6% vs. 14.5%, p < 0.01). Moreover, the percentage of CXCR5 + CCR7 hi Tfh cells among CD4 + T cells was higher in patients with NMOSD having an ARR of ≥ 1 than in those with NMOSD having an ARR of <1 (3.4.% vs. 1.6%, p = 0.07); however, owing to the large individual difference of patients, there was no significant difference between these patients. There were no significant differences in the frequency or number of CXCR5 + CCR7 low T cells between the two groups of patients, although both showed higher levels in the patients with an ARR of ≥ 1 than in the patients with an ARR of<1 (Figures 2A, B). Therefore, we speculated that more CXCR5 + CCR7 hi Tfh cells among CD4 + T cells in the patients with NMOSD having a higher ARR could explain the higher CD4 + CXCR5 + T cells in patients with higher ARR, although there was no significant difference between these patients.

Decreased CXCL13 Level Was Associated With GUDCA
The number of CXCR5+CD4+ T cells was significantly higher in the KLH group than in the GUDCA group (p < 0.001, Figures 3A, B). Further, there was an increasing trend of CXCR5 + CD4 + Tfh cell count in the GUDCA group (p = 0.02, Figures 3A, B). In addition, we evaluated the frequency of CD19 low CXCR5 + Tfh cells, the frequency of CCR7 hi PD-1 low resting Tfh and CCR7 hi PD-1 hi -activated Tfh cells. Tfh cell count did not change significantly, whereas CCR7 low PD-1 hi Tfh precursor cells in KLH mice tended to increase following GUDCA supplementation compared with those in the KLH group ( Figures 3A, C). In addition, a significant increase in the serum CXCL13 level was observed in the GUDCA group compared with that in the KLH group ( Figure 3D), which supports the immunomodulatory function of GUDCA. We then analysed the distribution of splenic (PD-1 + and CXCR5 + /CCR7 + double-positive) Tfh cells with immunohistochemical double staining. Splenic PD-1 + CCR7 + and PD-1 + CXCR5 + Tfh cells were localised in the T-B cell zone in the control group, whereas they were distributed in the GC-B-cell follicles in the KLH and GUDCA groups ( Figure 3E). There were barely PD-1 + CCR7 + and PD-1 + CXCR5 + positive Tfh cells in the intestinal Peyer's patch lymph node of the control group, which accumulated around the margin of Peyer's patch lymph node ( Figure 3F), supporting the notion that GUDCA supplied a permissive environment for Tfh cell generation.

Disease Relapse Was Associated With Significant Variations of the Gut Microbiome
Next, we analysed the changes in the stool microbiota of patients with NMOSD using 16S rRNA gene sequencing at the genus taxonomic rank levels ( Figure 4A). Patients with NMOSD having a high ARR were accompanied by a relatively higher abundance of Actinomyces and Sphingomonas but a low abundance of Veillomas, Atopobium, and Haemophilus ( Figure 4A). A higher stool relative abundance of Actinomycetaceae (such as Actinomyces) and Mitochondria predicted a high ARR in these 59 patients, whereas Bacteroidaceae (such as Bacteroides), Victivallaceae (such as Victivallis), or Pasteurellaceae family members were associated with an optimistic prognosis ( Figure 4B). Interestingly, Vagococcus, Anaerobiospirillum, Stenotrophomonas, Veillomas, Megasphaera, Atopobium, VadinCA11, Victivallis, Haemophilus and Bacteroides were negatively correlated with ARR in NMOSD ( Figure 4C). Hence, NMOSD recurrence is associated with the local microbiome.

GUDCA Increased the Circulating Tfh Cell and CXCL13 Levels in Mice
Subsequently, we concentrated on the association between CXCL13 and gut microbiota-derived metabolites. Pearson's correlation analysis revealed that GUDCA was positively correlated with the CXCL13 levels ( Figures 5A, B). A positive correlation between the Taurocholic acid (TCA), LCA ursodesoxycholic acid (UDCA), glycodesoxycholic acid (GDCA) and Taurodeoxycholic acid (TDCA) and the IL-10 levels (p < 0.001, p < 0.001, p < 0.001, p = 0.006 and p = 0.033, respectively) and a positive correlation between the DCA and LCA, and the IL-17 concentration (p = 0.005 and p =0.039, respectively) in NMOSD were also observed ( Figure 5A).
Anaerobiospirillum and Vagococcus were negatively correlated with GUDCA levels ( Figure 5B). Using the Virtual Metabolic Human database, we identified that reduced Veillomas and Haemophilus abundance in Spatients with NMOSD having a high ARR may contribute to decreased cholic acid (CA) and deoxycholic acid (DCA) concentration, which are the metabolites of GUDCA.
These data indicate that Anaerobiospirillum and Vagococcus were the most substantial gut microbiota influencing the levels of GUDCA in the hosts, whereas Veillomas and Haemophilus were the most substantial gut microbiota in converting GUDCA to DCA ( Figure 5B).

DISCUSSION
Our findings indicate that the gut microbiota play a crucial role in governing the Tfh immune response induced by CXCL13 in patients with NMOSD. In this study, we analysed the interplay between the stool metabolites and systemic immunomodulation and found significantly downregulated levels of CXCL13 in patients with NMOSD with low ARR. CXCL13 was negatively correlated with GUDCA, which was positively correlated with Vagococcus and Anaerobiospirillum abundance, which were negatively correlated with ARR in NMOSD, eliciting CXCL13induced Tfh cell immune responses possibly dependent on GUDCA and gut microbiota. In this study, we confirmed an increase in the CXCR5+CD4+ T cells in patients with NMOSD with a high ARR. Certainly, serum CXCL13 level reflects the priming and activation of Tfh cells in the spleen. In contrast, GC-Tfh is one of the main producers of plasma CXCL13. There is an increasing number of studies on the Tfh cells in NMO or NMOSD. The level of chemokine CXCL13 in the cerebrospinal fluid of patients with NMO is related to the degree of NMOrelated nerve defect (7), and patients with recurrence tend to have higher levels of CXCL13 in the serum and cerebrospinal fluid (CSF). The CXCR5 levels in the serum and CSF of patients with NMO are elevated (7). The administration of rituximab (a B-cell depletion therapy) can dramatically reduce the Tfh ratio and re-establish the Tfh subpopulation (12). In summary, the Tfh cells are associated with the disease activity of NMOSD, and could be potential immune markers for disease monitoring, and provide a new direction for NMOSD immunotherapy (13). Studies have also shown that memory Tfh cells in the circulation express the same surface molecules as effector Tfh cells and can effectively assist B cells in immune responses (14). Therefore, memory Tfh cells can be used as marker cells to monitor the immune activation status of autoimmune diseases. Previous study showed that circulating memory Tfh cells, especially CCR7 + ICOS + memory Tfh cells, may be associated with the relapse of MS and the numbers of circulating memory Tfh cells significantly decreased in the remitting stage (15). Among the circulating memory Tfh cells, CCR7 + memory Tfh cells can migrate into B cell follicles and promote humoral responses (16,17), and activated CCR7 -PD-1 + memory Tfh cells in the secondary lymphoid tissues also enhance the humoral responses, associated with the development of autoimmune diseases (18). Accordingly, in patients with systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA), increased levels of the CCR7 lo PD-1 hi subset correlated with elevated autoantibody profiles and more severe disease activities (18). Intriguingly, our results indicated significantly increased numbers of CCR7 -PD-1 + memory Tfh cells in GUDCA-treated KLH mice compared to KLH mice, and the CXCL13 levels were significantly increased in mice sero, Therefore, CXCL13 may participate in the migration of CCR7 + and CCR7 + ICOS + memory Tfh cells into the secondary lymphoid organs, and consequently, the activated CCR7 -PD-1 + memory Tfh cells could enhance the humoral responses associated with the relapse of NMOSD.
Another cytokine, IL-21, also plays a critical role in the survival of Tfh cells and the survival, proliferation, and differentiation of GC B cells. Studies have shown that the frequencies of circulating CD4 + CXCR5 + PD-1 + T cells and serum IL-21 were higher in patients with NMOSD than in those with MS (19). After glucocorticoid therapy, CD4 + CXCR5 + PD-1 + T cell count and serum IL-21 levels decreased, suggesting that the Tfh cells and IL-21 are associated with the disease activity of NMOSD (19). A recent study found that the circulating memory Tfh cells (ICOS + , CCR7 − , CCR7 − , ICOS + , CCR7 + , and CCR7 + ICOS + Tfh cells) and serum and cerebrospinal fluid IL-21 levels in patients with NMO/ NMOSD significantly increased (7). The proportion of CCR7and CCR7 -ICOS + memory Tfh cells is positively correlated with the ARR, plasma IL-21 level, and AQP-4 antibody levels (7). The proportion of CCR7 + and CCR7 + ICOS + memory Tfh cells is positively correlated with the number of white blood cells and IL-21 in CSF. After hormone therapy, the levels of CCR7 -ICOS + , CCR7 + ICOS + Tfh cells, and IL-21 reduced in patients with complete remission (7). Hence, IL-21 may participate in the development and relapse of NMO/NMOSD. Although significantly different levels of IL-21 between two groups were not observed in the present study, IL-21 could be utilised as a therapeutic target in NMOSD. We found that the increased faecal levels of GUDCA induced Tfh activation. Primary BAs in the gut are derived from cholesterol and are synthesised in the liver. BAs are primarily synthesised by the rate-limiting enzyme cholesterol 7ahydroxylase from the cholesterol in the liver, conjugated with either glycine or taurine (20). Bile salt deconjugation is catalysed by bile salt hydrolase (BSH) and generates free BAs (21), including mainly CDCA and trihydroxy BA CA. The gut microbiome is the only source of enzymes capable of generating hydrophobic SBAs. Clostridium and Desulfovibrio, which encode the bile acid-inducible operon, convert the host CA to DCA and CDCA and UDCA to LCA (22). In the present study, we found that the Veillomas and Haemophilus showed the highest increase in patients with NMOSD having an ARR of<1, which could explain the lower levels of GUDCA in the stool. In particular, Veillomas and Haemophilus spp. were downregulated in patients with NMOSD having a high ARR and negatively correlated with the serum GUDCA levels. The relative abundance of Veillomas increased in marathon runners, and inoculation of Veillonella into mice significantly increased the running time (23). Previous studies have reported that the abundance of Veillonella increased in patients with cirrhosis (24), whereas the levels of intestinal BAs in patients with liver cirrhosis are frequently insufficient (25). Haemophilus displayed a negative correlation with common indicators of dyslipidaemia, such as total cholesterol and high-density lipoprotein cholesterol (26). Haemophilus has also been reported to have a negative association with DCA (27). However, the levels of BAs were similar in both groups of NMOSD, ruling out that the discriminant bacteria observed between the two groups were caused by these metabolites. These investigations on the intestinal symbiotic bacteria provide evidence for a precise understanding of the complex interactions between the gut microbiota and host BA metabolites. Our data suggested that the decreased CXCL13 level was associated with GUDCA, CCR7 low PD-1 hi Tfh precursor cells in the KLH mice tended to increase following GUDCA supplementation, showing that the levels of CXCL13 and Tfh precursor cells were correlated with the metabolite GUDCA. Meanwhile, Veillomas and Haemophilus showed the highest increase in patients with NMOSD having an ARR of<1, which could explain the lower levels of GUDCA in the stool. Therefore, our data showed that the recurrence rate correlated with the gut microbiota and their metabolite in NMOSD, whereas the CXCL13-induced activation of the Tfh cells was the possible mechanism.
In addition, recent reports have confirmed that the intestinal microbiota is essential for the local activation of Tfh cells. Immunogenic commensals are distinguishable from tolerogenic bacteria and trigger migratory DCs to release IL-1b and IL-12p70, thereby increasing bacterial or self-antigen-specific Tfh cells or IgG2b responses (28). Various microbial contents can trigger Tfh cell responses, specifically bacterial RNA, which can induce IL-1bdependent differentiation of Tfh cells and GC B cells (29). After mucosal inoculation with inactivated enterotoxigenic Escherichia coli, activated ICOS + Tfh cells were recirculated in the blood and represented mucosal memory B cell responses (30). Whether antigen-specific Tfh cells are present in NMOSD remains to be defined. Further experiments are needed to address the gut microbiota to detect whether there is a Tfh-specific recognition motif.
In our study, we demonstrated that the gut microbiota and their metabolite correlated to the NMOSD relapse via CXCL13induced activation of Tfh cells. In addition, there was a significant correlation between the multiple microbe metabolites and IL-10 and IL-17A in our study. Previous studies have revealed that the interaction between hosts and their gut bacteria can regulate the host immunological IL10/ IL17A homeostasis via the BAs. Song et al. have reported that genetic abolition of the BA metabolic pathways in individual gut symbionts significantly decreases this Treg cell population. Restoration of the intestinal BA pool increases the colonic RORg+ Treg cell counts and ameliorates host susceptibility to inflammatory colitis via BA nuclear receptors (31). Campbell et al. found that the secondary bile acid 3b-hydroxydeoxycholic acid (isoDCA) increased the Foxp3 induction by acting on dendritic cells (DCs) to diminish their immunostimulatory properties. Ablating one receptor, the FXR, in DCs enhanced the generation of Treg cells and imposed a transcriptional profile similar to that induced by isoDCA (10). Moreover, IL10/IL17A homeostasis reportedly participates in the pathogenesis of NMOSD (32). Therefore, more research is warranted to better explain the role of gut microbiota and their metabolite in the pathogenesis of NMOSD.
This study has some limitations. Although we noted the accumulation of Tfh cells in the lymph follicle-like structure in the gut and spleen, the origin of Tfh cells remains unclear. We also noted that the number of participants in our crosssectional analysis was relatively small. We can only verify the role of GUDCA in Tfh activation in the animal experiments. However, there is no direct evidence of whether the gut microbiota affects Tfh activation. In fact, it is not clear if different immunosuppressive treatments and course of disease between the groups may affect Tfh activation.

CONCLUSIONS
In the present study, we demonstrated that the recurrence of NMOSD is correlated with Tfh cells and CXCL13, which is positively correlated with gut GUDCA. These data broaden our understanding of the mechanism of NMOSD and provide intestinal microbiota as a potential therapeutic target. Finally, we established a gut microbiome-metabolite-Tfh-CXCL13 system to predict the recurrence of NMOSD.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the medical ethics committee of the Third Affiliated Hospital of Sun Yat-sen University. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by Medical Ethics Committee of The Third Affiliated Hospital of Sun Yatsen University.

FUNDING
Financial support for the research was provided by the National Natural Science Foundation of China (grants. 82001284, 82071344, and 82101418).