**1. Introduction**

Rheumatoid arthritis (RA) is an autoimmune disorder that particularly affects the synovial membranes. Affected synovial membranes produce an excess of pro-inflammatory cytokines and chemokines, such as tumor necrosis factor (TNF)-α, interleukin (IL)-6, IL-1b, and stromal cell-derived factor 1, which destroy the joints and cause pain and a restricted range of motion [1–3]. Other than the joints, RA also affects major internal organs, including the heart, lungs, liver, and brain [4]. The progress of these arthritic and systemic symptoms reduces patients' activities of daily living (ADL), and because it causes a variety of disabilities, RA is a disease of public health importance, and it must be managed as such.

In RA treatment, priority is given to pharmacotherapy. In recent years, a paradigm shift has occurred in RA pharmacotherapy due to the arrival of biological disease-modifying anti-rheumatic drugs (bDMARDs), and it has become possible to minimize joint destruction [5–7]. However, bDMARDs have the disadvantages of adverse drug reactions and high costs, and they can only be used in limited cases [8,9]. Exploring other modalities of treatment of RA is therefore becoming increasingly important. One of such methods is exercise therapy. Exercise therapy is a safe and economical treatment that is widely used for the treatment of several systemic diseases; not only joint diseases. Exercise therapy was conducted based on the experience of individual doctors or therapists. In recent years, its mechanisms are becoming clearer, and evidence is accumulating. Among the various available exercise options, exercise therapy using treadmills works to protect articular cartilage and subchondral bone [10]. In osteoarthritis (OA), it inhibits the formation of bone spurs and bone destruction [11], and it thereby reduces joint symptoms. Treadmill exercise has also been reported to improve cardio-pulmonary function and ADL for patients with heart or lung disease [12,13]. Exercise therapy for RA is strongly recommended by a Cochrane review [14], and it has been reported to improve bodily functions and ADL, clinical indicators of joint symptoms, muscle strength, and cardio-pulmonary function [15,16]. In recent years, treadmill running in rat RA models was reported to inhibit the production of connexin 43 (Cx43) and TNF-α in the synovial membranes, as well as prevent the degeneration of articular cartilage and subchondral bone [17]. It was also revealed that exercise therapy inhibits joint destruction through bio-molecular mechanisms by suppressing pro-inflammatory cytokines.

The natural course of RA is divided into an induction phase following immunological sensitization, a pre-arthritis phase during which autoantibodies are produced, and an established phase during which arthritis occurs and progresses [18]. The effectiveness of treatment may vary with the phase of disease. Dekkers et al. used animal models to study the differences in efficacy of pharmacotherapy for each phase, and they established that pharmacotherapy is more effective in the established phase than in the pre-arthritis phase [19]. It was found that compared with the pre-arthritis phase, in which cytokine production increases sharply, it is easier to get efficacy from anti-TNF-α antibodies in the established phase, during which cytokine quantities are stable. Thus, phase-related differences in cytokine production influence the efficacy of pharmacotherapy [20]. However, phase-related differences in the efficacy of exercise therapy are unknown. Since exercise therapy also demonstrates efficacy in suppressing cytokines, we posed the hypothesis that it would be more effective to suppress arthritis and joint destruction with treadmill running during the established phase than in the pre-arthritis phase.

Based on the above context, the objective of this study was to assess the differences in efficacy of treadmill running during each phase of RA, using rat RA models.

#### **2. Results**

#### *2.1. Kinetic Change in Body Weight and Paw Volume*

Body weight in each group (no intervention (control) group, pre-arthritis intervention short (PAS) group, pre-arthritis intervention long (PAL) group, and therapeutic intervention (T) group) gradually increased from day 0 to day 12, decreased from day 12 to day 20, and increased again starting on day 21. Paw volume in each group increased from day 14, reaching its maximum between day 20 and day 23, and gradually decreasing thereafter. No significant differences in either body weight or paw volume were observed among groups from day 0 to day 42 (Figure 1A–C).

**Figure 1.** Kinetic change in (**A**) body weight, (**B**) paw volume, and (**C**) clinical score after immunization. The parameters were measured once every three days until day 12 and every day thereafter. There were no significant differences among the four groups (no intervention group, control; pre-arthritis intervention short group, PAS; pre-arthritis intervention long group, PAL; therapeutic intervention group, T) on all days.

#### *2.2. E*ff*ect of Treadmill Running on Articular Cartilage*

To assess the histological effects of treadmill running on joints and their cartilage, we stained the articular cartilage from the rat paws with hematoxylin and eosin (H&E), and safranin O on day 42 (Figure 2A–D). In the control group and the PAS group, we observed intra-articular infiltration by inflammatory cells and pannus formation in hyperplastic synovial membranes. Joint structure destruction was also more severe in these groups than in the T group and the PAL group. Furthermore, safranin O staining in the control and PAS groups was less than in the T and PAL groups. The histological scores were significantly lower in the T and PAL groups than in the control and PAS groups. There was no difference in histological score between the PAS and control groups (Figure 2E,F).

#### *2.3. Influence of Treadmill Running on the Production of TNF-*α *and Cx43 in the Synovium*

Following past reports, we conducted immunostaining for TNF-α and Cx43 [17]. TNF-α expression was aggravated throughout the synovial membranes, but the level of staining was significantly lower for the T group than for the control group (Figure 3A,C). For Cx43, the level of staining was significantly lower for the T group than for the control and PAS groups (Figure 3B,D).

**Figure 2.** Representative micrographs of (**A**) hematoxylin and eosin, and (**B**) safranin O-stained sagittal sections. Representative micrographs of (**C**) hematoxylin and eosin, and (**D**) safranin O-stained sagittal sections in a normal rat without treadmill running. (**E**) The histological scores (mean ± standard deviation) and (**F**) only the cartilage evaluation scored based on the histological score (mean ± standard deviation) are shown. The PAL and T groups had suppressed destruction of the ankle joint more than the control and PAS groups. \*\* *p* < 0.01, \* *p* < 0.05. Scale bar = 200 μm. The black spot represents the position in the high-magnification figure.

**Figure 3.** Representative micrographs of immunohistochemical staining for (**A**) TNF-α and (**B**) Cx43 are shown. All images were evaluated semi-quantitatively using ImageJ for (**C**) TNF-α and (**D**) Cx43. T group had significantly suppressed TNF-α and Cx43 expression. \* *p* < 0.05. Scale bar = 200 μm. The black spot represents the position in the high-magnification figure.

#### *2.4. Prevention of Bone Loss by Treadmill Running*

We used micro-computed tomography (μ-CT) to assess periarticular skeletal composition changes (Figure 4). The T group had higher bone volume fraction (BV/TV), trabecular thickness (Tb.Th), and bone mineral content per tissue volume (BMC/TV) than the control group (Figure 4A,B,D). Furthermore, marrow star volume (MSV) values were significantly lower for the T group than for the control group (Figure 4C).

**Figure 4.** Representative three-dimensional reconstruction of **(A**) the sagittal sections of the talus architecture. Trabecular bone parameters such as (**B**) bone volume fraction (BV/TV), (**C**) trabecular thickness (Tb.Th), (**D**) bone mineral content per tissue volume (BMC/TV), and (**E**) marrow star volume (MSV) of the whole talus are shown. T group had improved bone loss. *n* = 4 in each group. \* *p* < 0.05.

#### *2.5. E*ff*ects of Treadmill Running on Bone Erosion*

We used μ-CT to assess the effects of treadmill running on bone erosion (Figure 5). The eroded bone surface per repaired bone surface (Es/Rps) value for T group was significantly lower than that of the control group (Figure 5B,C). To assess its effects on bone metabolism, we conducted immunostaining with cathepsin K, one of the osteoclast marker that serves as an indicator of the degree of bone resorption [21,22]. Many cathepsin K-positive cells were observed in the pannus areas of the control and PAS groups. Fewer cathepsin K-positive cells were in the T and PAL groups than in the control and PAS groups (Figure 5A).

**Figure 5.** (**A**) Representative micrographs of cathepsin K immunohistochemical staining. Cathepsin K positive cells were fewer in the PAL and T groups. (**B**) Representative 3D reconstruction of bone erosion area in the whole talus architecture. Red area is bone erosion area. (**C**) The eroded bone surface (Es) and repaired bone surface (Rps) was calculated using 3D-μ-CT, and Es/Rps values are shown for the four groups. Es/Rps was significantly lower in the T group compared to the control group. \* *p* < 0.05. Scale bar = 200 μm. The black spot represents the position in the high-magnification figure.
