A total of 50 male Sprague–Dawley rats (210 ± 10 g) were purchased from the National Institutes for Food and Drug Control [animal licence number: SCXK (Beijing) 2019-0017]. Rats were housed in the Experimental Animal Centre of the Institute of Clinical Medical Sciences, China–Japan Friendship Hospital [Experiment Animal Centre licence number: SYXK (Beijing) 2016-0043]. In a standard light/dark cycle (6:00–18:00 light phase), the rats were maintained at a constant temperature of 23°C ( ± 2°C) with free access to standard rodent feed and water in a specific pathogen-free animal laboratory. All experimental procedures were reviewed, approved, and directed by the Institute of Clinical Medical Science, China–Japan Friendship Hospital (Ethics No. 190104).

Decoction from Rheum officinale Baill. (Polygonaceae, Rhei radix et rhizoma) (Beijing Tongrentang Co., Ltd., Beijing, China) was used to induce the SD rat model. In detail, 1 kg of Rheum officinale Baill. was made into powder and dissolved in 5 L of pure water followed by ultrasonication for 1 h. Then, the solution was filtered with four layers of gauze and centrifuged at 2,500 rpm for 5 min. The supernatant was collected and concentrated to 500 ml by rotary evaporation and stored at 4°C. XFP is composed of the following herbs: Astragalus mongholicus Bunge. (15 g, Fabaceae, Astragali radix), Atractylodes macrocephala Koidz. (15 g, Asteraceae, Atractylodis macrocephalae rhizome), Saposhnikovia divaricata (Turcz. Ex Ledeb.) Schischk. (10 g, Apiaceae, Saposhnikoviae radix), Glycyrrhiza glabra L. (10 g, Fabaceae, Glycyrrhizae radix et rhizome), Rehmannia glutinosa (Gaertn.) DC. (15 g, Scrophulariaceae, Rehmanniae radix praeparata), Angelica sinensis (Oliv.) Diels. (10 g, Apiaceae, Angelicae sinensis radix), Ligusticum chuanxiong Hort. (10 g, Apiaceae, Chuanxiong rhizoma), Bupleurum chinense DC. (10 g, Apiaceae, Bupleuri radix), Aconitum carmichaeli Debeaux. (10 g, Ranunculaceae, Aconiti lateralis radix praeparaia), and Cinnamomum cassia Presl. (10 g, Lauraceae, Cinnamomi cortex). The herbs were purchased from Beijing Tongrentang Co., Ltd. and identified by professor Jun He, Department of Pharmacy, China–Japan Friendship Hospital. Because of the poor solubility, XFP was made into water extract to reach the gavage dose. The preparation process of XFP decoction was as follows: The medicinal materials were mixed and made into powder, and then the powder was soaked in water at a ratio of 1:10 (v/v) for 1 h, decocted twice for 1 h each time after boiling. Finally, the collected liquid was concentrated into 100% decoction by rotary evaporation and stored at 4°C. The chemical fingerprints and ingredients of Rheum officinale Baill. and the XFP decoction were assessed by ultrahigh-performance liquid chromatography -Q exactive hybrid quadrupole -Orbit high resolution accurate mass spectrometry and presented in Supplementary Material 1 .

After 3 days of acclimation, the rats were randomly divided into five groups with 10 rats in each group: the control group, SD group, CIA group, SD-CIA group, and XFP-treated SD-CIA group. Among them, the SD rat model and SD-CIA rat model were induced as previously described (Xiao et al., 2007). 1) SD rat model: Rats were administered a dose of Rheum officinale Baill. decoction at 10 g/kg/day irregular eating to cause SD in the first 2 weeks. Then, the rats were administered a dose of Rheum officinale Baill. decoction at 5 g/kg/day and remained irregular eating from the third week to the end of the experiment. 2). CIA rat model: Rats were immunized by intradermal injection of 50 µg of bovine type II collagen (Chondrex, Inc., Redmond, WA, USA) emulsified in isopyknic incomplete Freund’s adjuvant (IFA) (Chondrex) at the base of the tail. Seven days later, the same preparation was given on the other side of the tail for booster immunization. 3). SD-CIA rat model: This model is a combination of the SD rat model and CIA rat model. Briefly, rats were administered Rheum officinale Baill. decoction and irregular eating to establish an SD rat model. After 2 weeks, the rats were administered the first immunization to establish the CIA rat model.

After booster immunization in CIA and SD-CIA models, therapeutic intervention in rats began. The rats in the XFP group were administered XFP decoction at 23 g/kg/day (twice the human equivalent dose) once a day for 34 days. The other groups were intragastrically administered an equal volume of pure water (10 ml/kg/day). The entire experimental process is shown in Figure 1 .

According to the classical theory of TCM, research on standardization of TCM syndromes and the biological characteristics of rats (Du et al., 2012), the state of SD model rats primarily includes listlessness, narrow eyes, dry and dull hair, hunched posture, reduced physical activity, huddled posture, reduced food intake, weight loss, and loose stools.

The extent of hind limb swelling was assessed every 3 days from the day of administration. According to conventional criteria (Xiao et al., 2006), the severity of arthritis in each hind leg was expressed by the arthritis index (AI) scores, which ranged from 0 to 4 points: 0, no swelling or oedema; 1, detectable erythema and oedema limited to the digits; 2, slight erythema and oedema from digits to tarsal bone; 3, moderate erythema and oedema from foot to ankle; 4, maximal erythema and oedema at entire ankle. The maximum AI score was 8 points (4 points × 2 hind legs).

Bone erosion was evaluated by microcomputed tomography (micro-CT). After 34 days of drug treatment, the left ankles and claws were collected and fixed in formalin for 7 days. Then, the samples were scanned using a SkyScan 1174 micro-CT scanner (Bruker, Belgium) to evaluate bone damage. N-Recon software was used for three-dimensional (3D) image reconstruction, and matched CT-AN software was used to quantify 3D images. The bone volume (BV) and bone surface (BS) of the tarsus were then calculated to assess changes in bone structure, and the BS/BV ratio was used to evaluate surface density.

Haematoxylin-eosin (H&E) staining and tartrate-resistant acid phosphatase (TRACP) staining were used for histological scoring and bone damage evaluation of the ankle joints. After micro-CT, the claws and ankles were decalcified in 10% ethylene diamine tetraacetic acid (EDTA)–Na₂ for 1 month. After progressive dehydration and transparency, the samples were embedded in paraffin, prepared into 7-µm-thick slices, and finally stained with H&E. Histological scores were assessed by the destruction of cartilage damage and bone damage, cell infiltration, synovial hyperplasia, pannus respectively, and the mean value of the above indicators as the overall observation index, and were scored on a criterion of 0-3 (0, normal; 1, weak; 2, moderate; and 3, severe) (Li et al., 2020). TRACP (Sigma, St. Louis, MO, USA) staining assay was used to quantify the OCs of ankle joint slice samples, which were identified by the presence of TRACP-positive multinucleated cells that contained at least three nuclei. Typical images were captured using the Leica Qwin image analysis software (Leica Microsystem, Germany).

The levels of sIgA in the jejunum and ileum were determined by immunohistochemistry (IHC) assay. After dewaxing and hydration, paraffin-embedded slices (5-µm-thick) of the jejunum and ileum were repaired with antigen by high pressure in citrate buffer (pH = 6.0). Then, the sample was incubated in 3% hydrogen peroxide (H₂O₂) in darkness, followed by goat serum blocking buffer at room temperature. Afterward, the ileum was incubated with goat anti-rat sIgA (Abcam, Cambridge, MA, USA; 1:400) overnight at 4°C and then treated with horseradish peroxidase–related species-specific secondary antibody for 20 min. The sections were stained with 3,3′-diaminobenzidine (DAB) and then counterstained with haematoxylin. Finally, random images were captured using a ZEISS Axio Observer 3 microscope, and the optical density of the positively stained area was examined using Image Pro-Plus.

The levels of anti–collagen type II antibody (Col II Ab) in the serum and sIgA in the ileum were detected using enzyme-linked immunosorbent assay (ELISA). The blood, jejunum and ileum were collected after 34 days of treatment. After being obtained from the abdominal aorta, blood was centrifuged (3,000 rpm, 10 min, 4°C). The jejunum and ileum were homogenized in normal saline at a ratio of 1:10 (g/v) followed by centrifugation (3,000 rpm, 10 min, 4°C). Both serum and homogenate were collected and stored at −80°C. Col II Ab and sIgA were measured using ELISA kits following the manufacturer’s protocol. The sIgA ELISA kit was obtained from the Nanjing Jiancheng Bioengineering Institute (Nanjing, China). The Col II Ab ELISA kit was purchased from the Beijing Qisong Biotechnology Company (Beijing, China).

PPs in small intestines were isolated. Then, similar to our previous study (Li et al., 2020), Tregs, Th17 cells, and the Tregs/Th17 cells ratio in PPs were evaluated using a FACS Canto II flow cytometer (FACS Canto II, Becton, Dickinson, Co., Franklin Lakes, NJ, USA). Briefly, PPs from the five groups were ground into single cells and filtered through a cell sieve. Antibodies against CD4⁺ and IL-17⁺ were used to identify Th17 cells. Antibodies against CD4⁺, CD25⁺, and Foxp3⁺ were used to stain Tregs (BD Pharmingen, San Diego, CA, USA).

The quantitative real-time polymerase chain reaction (qRT-PCR) assay was used to detect the mRNA expression levels of tumor necrosis factor–α (TNF-α), IL-1β, IL-6, IL-10, and IL-17 in the ileum. According to the manufacturer’s recommended procedure, total RNA of the sample in rats was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). Then, the PrimeScript RT Reagent Kit with gDNA Eraser (TaKaRa, Tokyo, Japan) was used to conduct gDNA elimination and cDNA synthesis. SYBR premix Ex Taq (Tli RNaseH Plus) (TaKaRa) was used to perform qRT-PCR via a Quant Studio 5 real-time PCR system (Thermo Fisher Scientific, Waltham, MA, USA). The program was performed as follows: initial denaturation at 95°C for 30 s and amplification for 40 cycles at 95°C for 5 s and 60°C for 34 s. All experiments were repeated three times with three technical duplications, and the expression levels of relative mRNA were normalized to β-actin and determined using the 2^−△△Ct method. The primers used in this experiment are shown in Supplementary Material 2 .

The ileum intestinal contents were harvested, and microbial DNA was extracted using the Power Faecal DNA Kit (Qiagen, Valencia, CA, USA). After a series of pretreatments, the samples were amplified by PCR to form sequencing libraries. Then, the qualified libraries were sequenced using the Illumina sequencing platform. The original reads obtained by sequencing were quality-controlled and filtered to obtain clean reads. After splicing and assembly, the coding genes were predicted and annotated in the database. At the same time, clean reads were taxonomically analyzed, and the species composition, abundance of samples, and function based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database were statistically analyzed. In addition, association of pathways with dominant functions and the top 10 species in the XFP group was analyzed.

All data are shown as the mean ± SEM. One-way analysis of variance followed by the least significant difference test was used to determine differences between groups. The Kruskal–Wallis H and Mann–Whitney U tests were used when data were not normally distributed. Significance was defined as P < 0.05 using SPSS 20.0 software.

After 34 days of XFP treatment, the symptoms of SD and arthritis progression in SD-CIA were ameliorated. Rats in the SD and SD-CIA groups were listless with dull hair and loose stools. After treatment with XFP, the state of the rats was improved, their hair became glossy, and their stool was molded. Similarly, as another important index for evaluating SD, as shown in Figure 2A , the body weight of rats was much lower in the SD and SD-CIA groups than in the control group from the first week of Rheum officinale Baill. intervention (P < 0.01), whereas the body weight was increased in the XFP group after 1 week of treatment compared with the SD-CIA group (P < 0.05). These results indicated that XFP ameliorated arthritis with SD.

To determine the therapeutic effect of XFP in arthritis, the AI scores at the lesion location and serum Col II Ab in entirety were used. At the end of the experiment, more severe hind limb swelling occurred in the SD-CIA group than in the CIA group, and this swelling was distinctly relieved in the XFP group compared to the SD-CIA group ( Figure 2B ). The AI scores are shown in Figure 2C . Arthritis in the control and SD groups was as expected, with an AI score of 0. Seven days after modeling, the AI scores of the CIA model had nearly peaked. In general, the SD-CIA group displayed higher AI scores than the CIA group. After treatment, from day 28 onward, AI scores in the XFP group were significantly lower than those in the SD-CIA group (P < 0.05). The levels of Col II Ab in serum exhibited a similar trend to AI scores ( Figure 2D ). Col II Ab levels were significantly increased in the CIA group compared with the control group (P < 0.05), and Col II Ab levels in the SD-CIA group were significantly different from those in the CIA group (P < 0.05). After XFP treatment, the serum levels of Col II Ab were significantly lower than those of the SD-CIA group (P < 0.01) ( Figure 2D). These findings suggested that although SD did not directly cause the formation of CIA, it significantly aggravated CIA. Furthermore, XFP showed the potential to slow the progression of arthritis with SD.

Next, changes in ankle joint pathology were examined using H&E staining ( Figures 3A, D ). Compared with the control group, the conditions of cartilage destruction, bone destruction, cell infiltration, synovial hyperplasiaand pannus formation were significantly aggravated in CIA group (P < 0.05, P < 0.01); Compared with CIA group, the above indexes were heavier (P < 0.05); Compared with SD-CIA group, XFP group showed significant improvement in these indicators (P < 0.05, P < 0.01).Taking the mean value of the above indicators as the overall observation index, the trend is similar to the above trend.

In addition, micro-CT was used to evaluate the therapeutic effect of XFP on bone injury, and 3D reconstruction of the ankle joint is shown in Figure 3B . The BS, BV, and BS/BV values of the inflamed ankle were measured to quantify the degree of bone remodeling in the different groups ( Figures 3E–G ). The results demonstrated that the BS value of CIA was significantly different from that of the control group (P < 0.01), and the BS value in the SD-CIA group was higher than that in the CIA group (P < 0.01). After XFP treatment, the BS value had a decreasing tendency, but there was no significant difference compared to SD-CIA. The BV value was decreased in the SD-CIA group compared to the CIA group (P < 0.01), and XFP treatment increased it compared to the SD-CIA group (P < 0.05). With respect to the BS/BV value, we found that the CIA group still had bone destruction compared to the control group (P < 0.01), and the tendency of BS/BV was higher in the SD-CIA group than in the CIA group (P < 0.05). After XFP treatment, the BS/BV value was lower than that in the SD-CIA group (P < 0.05). On the one hand, the micro-CT results preliminarily confirmed our prediction that SD could aggravate the development of arthritis. These results also suggested that XFP inhibited joint damage.

Furthermore, the number of OCs in the ankle joints was measured using the TRACP assay to determine the severity of bone destruction. The results of TRACP ( Figures 3C, H ) showed that the number of OCs in the CIA group increased significantly compared to that in the control group (P < 0.01), and there was a significant difference between the SD-CIA and CIA groups (P < 0.01). In response to XFP intervention, the number of OCs in the XFP group decreased significantly compared to that in the SD-CIA group, which was also statistically significant (P < 0.01). The experiment also demonstrated that SD aggravated the bone destruction of joints, and XFP exerted a good therapeutic effect on SD-CIA.

In conclusion, SD might aggravate the development of CIA, whereas XFP exerted an obvious therapeutic benefit on SD-CIA.

To investigate the effect of XFP on the PPs of the intestine, the percentages of Tregs and Th17 cells and the ratio of Tregs/Th17 cells in the PPs were evaluated using flow cytometry. The percentage of Tregs in the SD-CIA group was lower than that in the CIA group (P < 0.05). In contrast, the Treg percentage was higher in the XFP group than in the SD-CIA group (P < 0.05) ( Figure 4A ). The proportion of Th17 cells in the CIA group was upregulated compared to that in the control group (P < 0.05), and the SD-CIA group displayed a higher tendency than the CIA group (P < 0.05). However, after treatment with XFP, the proportion of Th17 cells was decreased compared to that in the SD-CIA group (P < 0.05) ( Figure 4B ). The ratio of Tregs/Th17 cells was decreased in the CIA group compared with the control group (P < 0.05), and in the SD-CIA rats, there was a lower tendency compared to the CIA group (P < 0.05). After XFP treatment, the Tregs/Th17 cells ratio was significantly upregulated compared to that in the SD-CIA group (P < 0.01) ( Figure 4C ). These results demonstrated that SD had a detrimental effect on immune function and that XFP treatment exerted an extensive regulatory effect on intestinal mucosal immune function in SD-CIA rats.

The influence of XFP on ileal immune status was evaluated by detecting the expression of genes in the ileum using a qRT-PCR assay ( Figures 5A–E ). The CIA group exhibited upregulated expression of TNF-α, IL-1β, IL-6, and IL-17 and downregulated IL-10 levels in the ileum compared with the control group (P < 0.05 and P < 0.01, respectively), and the trend in the SD-CIA group was more significant than that in the CIA group (P < 0.05). Intervention with XFP significantly decreased TNF-α, IL-1β, IL-6, and IL-17 levels and increased levels of IL-10 compared to those in the SD-CIA group (P < 0.01) ( Figures 5A–E ). These results suggested that although SD had no significant effect on the immune state of the ileum, it aggravated the immune state disorder in CIA rats, and XFP treatment had an obvious therapeutic effect on the immune state of SD-CIA.

Because of the obvious variation in the contents of the intestinal tract, it is speculated that these effects may be caused by metabolites or antigens related to the intestinal microbiota. Therefore, we also examined the expression of sIgA using ELISA and IHC because sIgA is known to regulate intestinal microorganisms. ELISA suggested that sIgA levels in the CIA group were significantly lower than those in the control group (P < 0.05) and that sIgA levels in the SD-CIA group were lower than those in the CIA group (P < 0.05). XFP treatment upregulated sIgA levels compared to the SD-CIA group (P < 0.01) ( Figure 5F ). IHC experiments revealed that the protein expression density of sIgA in the CIA group was significantly decreased compared to that in the control group (P < 0.01). The density in the SD-CIA group was also significantly lower than that in the CIA group (P < 0.01). After XFP intervention, the density of sIgA protein expression increased significantly compared with that in the SD-CIA group (P < 0.01) ( Figures 5G, H ). These results suggested that CIA affected the intestinal tract mucosal immune barrier to a certain extent and reduced the pathogenic microorganism’s clearance rate, and the SD model might exacerbate this effect. In addition, XFP treatment affected sIgA levels, indicating that XFP treatments reshaped the immune barrier and reversed the regulatory function of the intestinal tract on intestinal bacteria.

The above results indicated that the immune barrier in CIA was impaired, and SD further aggravated the immune disorder in CIA, leading to a reduced clearance rate of pathogenic microorganisms and aggravating the occurrence of CIA. XFP treatment improved immune status and alleviated SD-CIA.

An increasing number of studies have shown that intestinal microorganisms play an important role in intestinal immune regulation (Xu C et al., 2020; Xu Z et al., 2020). Therefore, metagenomic sequencing of ileum contents was used to detect changes in the gut microbiota ( Figure 6 ). Among the top 10 phyla in terms of abundance, Firmicutes were dominant, followed by Chlamydiae. The abundance of Firmicutes in the XFP group and Chlamydiae in the control group and Proteobacteria, Bacteroidetes, and Actinobacteria in the SD-CIA group was highest among the three groups. In addition, other phyla accounted for a certain proportion ( Figure 6A ). Next, the analysis of similarities (Anosim) was used to assess bacterial community similarity. The results showed that compared with the control group, the SD-CIA group and the XFP group were significantly different in their distribution of microflora at the genus level. However, there were no differences between the SD-CIA group and the XFP group ( Figure 6B ). The distribution of samples in each group was relatively concentrated, but one sample in the XFP group overlapped with the region in the SD-CIA group as determined using principal coordinates analysis (PCoA) ( Figure 6C ). Furthermore, linear discriminant analysis effect size (LEfSe) was used to identify the biomarkers in each group ( Figure 6D ). The control group had the most types of biomarkers, whereas L. acidophilus and its different biological classifications dominated in the XFP group. Finally, the 10 genera and species with the highest abundance in each group are shown in Figures 6E,F . Consistent with LEfSe, the abundance of Lactobacillus and L. acidophilus was highest in the XFP group. Specifically, at the species level, the abundances of L. murinus, L. reuteri, B. pseudolongum, P. excrementihominis, and F. bacterium M10-2 in the SD-CIA group were upregulated compared with those in the control group. Compared with the SD-CIA group, the abundance of the above bacteria was decreased in the XFP group. However, L. johnsonii exhibited the opposite trend. These results illustrated that SD-CIA altered the distribution of bacterial flora and that XFP restored bacterial community disorder to a certain extent.

In addition, according to the literature, we found that among the 10 strains with the highest abundance, five strains, including L. acidophilus (Yanagihara et al., 2014; Chen et al., 2015; Li et al., 2019; Pan et al., 2019), L. murinus (Wilck et al., 2017; Bernard-Raichon et al., 2021), L. johnsonii (Lau et al., 2011; Zheng et al., 2021), L. reuteri (Mu et al., 2017), and B. seudolongum (Bromberg et al., 2020), showed the ability to regulate Tregs and Th17 cells. This suggested that the therapeutic effect of XFP on SD-CIA rats might be related to the regulation of the intestinal microenvironment in which bacteria participate.

Furthermore, we conducted a comprehensive analysis of the functional modules in the gut microbiome to explore the potential mechanism underlying the effects of XFP administration using the KEGG database. At KEGG pathway level 1, Metabolism, Genetic Information Processing, and Cellular Processes pathways were identified. At KEGG pathway level 2, 22 pathways were enriched, including pathways related to the maintenance of bacterial survival: carbohydrate metabolism, amino acid metabolism, nucleotide metabolism, and energy metabolism under Metabolism of KEGG pathway level 1; the gene-related pathways: translation, replication, and repair under Genetic Information Processing of KEGG pathway level 1; and the pathways including signal transduction and cell motility belonging to Environmental Information Processing and Cellular Processes in level 1 ( Figure 7A ). The abundance of pathways generally conforms to conventional understanding. Then, sample similarity was assessed using non-metric multidimensional scaling (NMDS) analysis ( Figure 7B ). Except for one sample in the XFP group, the sample distribution of each group was relatively concentrated, suggesting that the composition of functional genes was similar within the group and different among the groups. Furthermore, LEfSe was used to identify pathways with significant differences among the groups. Several dominant pathways were obtained in each group. As the smallest unit of the pathway, quorum sensing contributed the most to KEGG pathway level 3 in the XFP group, followed by ABC transporters ( Figure 7C ).

Finally, we analyzed the association between the intestinal bacteria with high abundance and the KEGG pathway in level 3 that exerted a functional advantage in the XFP group. The results showed that except for C. abortus, F. bacterium M10-2, and P. excrementihominis, the other seven species were associated with the dominant pathways in the top 10 most abundant species. The pathways related to the high abundance of strains in the XFP group were quorum sensing and ABC transporters ( Table 1 ). Recent reports have shown that the quorum sensing mechanism increases adherence to intestinal epithelial cells and that L. acidophilus can adhere to the intestinal epithelium and promote communication with the immune system (Buck et al., 2009). In addition, research has found that other strains can modulate immune responses via quorum sensing (Kariminik et al., 2017). Other pathways, such as those involving ABC transporters, are key processes to achieve self-immunity of bacteria (Smits et al., 2020).