All animal studies were approved by the Institutional Animal Care and Use Committee (IACUC) at Ponce Health Sciences University (PHSU). Female Sprague Dawley rats (Animal Research Facilities, PHSU, PR) of ~2 months old weighing ~180–200 g were randomized to one of three groups: Sham-No exercise (n = 9), Endo-No exercise (n = 10), or Endo-Exercise (n = 10). Animals were housed individually in a room maintained at 23°C with the light/dark cycle alternating every 12 h. Rats were fed with standard laboratory chow ad libitum. Weight was monitored twice a week and the experiments were performed at the same time of day to minimize the influence of circadian rhythms.

Endometriosis (Endo) was surgically induced as previously described by suturing autologous uterine horn tissue with the serosal surface next to the mesenteric vessels of the small intestine with the endometrial surface exposed to the peritoneum (17–19). Cytological analysis of smears was carried out before and after surgery (Figure 1A) to verify reproductive cyclicity and to allow for the interpretation of any effects of the stage of estrous cycle on the experimental results.

Endo-Exercise group animals had free access to a running wheel. These were unlocked for voluntary exercise from day -14 (2 weeks prior to endometriosis induction) until time of sacrifice at day 60 (Figure 1A). All animals were singly housed, where Endo-Exercise rats had running wheel cages (which are slightly larger to accommodate the wheels) with essentially the same free space (volume) as that found in Sham-No exercise and Endo-No Exercise groups which were housed in regular cages. All running wheels had a sensor (counter) connected to an interface (Scurry Interface) to record wheel rotations/min, # running bouts, speed and duration per bout. Cumulative distance (meters), count (revolution), average speed (m/min), and exercise intensity of all Endo-Exercise animals was collected, graphed and analyzed using the Scurry Activity System program to measure exercise parameters. The Scurry Activity program updated the collected data every 3 s (Lafayette Instruments, IN, USA). All animals met the minimum exclusion of 2 km/day (20).

An open field (OF) behavioral test was administered to the animals 7 days after surgery (day 0) and then again 1 day prior to sacrifice (day 59) to assess behavioral changes related to anxiety. The exploratory and locomotor activity of each rat in an open well-lit arena was measured using Anymaze software (Stoelting), a video tracking system designed to automate testing to track the center point of the rat body for 5 min. On each of the testing days this was followed by use of an elevated zero maze (ZM) 10 min after the open field test finished to assess anxiety related behavior. The elevated zero maze consisted of two open zones and two enclosed dark zones and it was situated in the center of an isolated room with dim illumination. Animals were placed in the elevated zero maze facing a closed zone. Exploratory and locomotor activity of each rat in the open and closed zones were recorded and measured for 5 min using Anymaze software (Stoelting). In each test increased time in the open area/arm indicates lower anxiety.

Blood samples were collected by cardia venipuncture at time of sacrifice. Serum corticosterone levels were determined using an ELISA kit (USA IB79175, IBL-America, Minnesota) according to the manufacturer's instructions. Serum was assayed for cytokine levels using MILLIPLEX MAP Rat Cytokine/Chemokine Magnetic Bead Panel (Cat# RECYTMAG-65K, Millipore). Median fluorescent intensity (MFI) was analyzed and the analyte concentrations were calculated according to standard curve.

All animals had a cytological smear taken, a laparotomy was performed, and after collecting peritoneal fluid by aseptic aspiration the peritoneal cavity was systematically examined for the presence of the implants and the original sutures. Peritoneal fluid was analyzed for lactoferrin levels by ELISA.

The vesicles were identified, and their longest length and width measured with a digital caliper to calculate vesicle area and assign them a score (0–5) as described previously (21). After weighing they were processed for paraffin embedding or frozen whole and stored at −80°C. Samples of uterus were frozen for MPO analysis. The mesenteric fat was removed from the abdominal cavity and weighed; samples were processed for paraffin embedding or frozen. Gastrocnemius and soleus muscles were dissected, removed, weighed, and stored at −80°C. Colon, liver, spleen, jejunum, and adrenal glands were also removed and stored for future follow up studies.

Tissue segments (vesicles and mesenteric fat) were fixed in 10% formalin, processed by routine histological techniques, and mounted on glass slides then stained with Toluidine blue to visualize mast cells as previously described (17). The numbers of mast cells in five randomly selected high-power fields were counted by two individual observers (JJ-G and RR-M) and the mean number of mast cells per field calculated.

Sections from mesenteric fat and vesicle tissue were double labeled with the macrophage marker CD68 and either FKN, CX₃CR1, or ERβ. Tissue sections were deparaffinized using Xylene Substitute and rehydrated in a series of steps, then boiled for 40 min in Citrate EDTA buffer with a pH = 6 at 95° C. Protein block was performed using Anti-Goat serum. PBS washes were done between steps. Tissue sections were double labeled and incubated overnight with CD68 (Bio Rad #MCA341R; 1:50) and FKN (Abcam #ab25088; 1:1,000), CXC3R1 (Abcam #ab8021; 1:100), or ERβ (Invitrogen #PA1311; 1:500). The next day sections were incubated with secondary antibodies (A11029, Alexa Fluor 488 goat anti-mouse) and (A21428, Alexa fluor 555 goat anti-rabbit) for 30 min, after washing with PBS. DAPI (Nuc Blue Fixed cells stain probe reagent #R37606, Invitrogen) was added and incubated for 5 min. After a PBS wash, the coverslips were mounted using the Prolong Gold Antifade Reagent (Cat#P36934, Invitrogen). Images were obtained using an Olympus BX-60 equipped with X-cite 120Q lamp and photographed with a Nikon DS-FI1 camera. Representative images were taken at 100x, at the same time using the same illumination parameters. FKN, CX₃CR1, ERβ, and CD-68 expressing cells per high power field (HPF) were analyzed by an observer using Image J software.

Vesicle and mesenteric fat tissue sections were deparaffinized in xylene substitute for 30 min and rehydrated in a descending series of alcohols to water. Tissue sections were exposed to 3% Hydrogen peroxide for 15 min. After washing with PBS, antigen retrieval was done in 0.01 M Citrate EDTA buffer (pH = 6) at 95–99°C for 40 min. Tissues were washed with distilled water in two intervals for 2 min each, followed by a PBS wash for 5 min. Protein block was performed using normal goat serum for 15 min (Cat#HK112-9KE, BioGenex), followed by an overnight incubation with ERα primary antibody (1:40; Cat#MA513304, Invitrogen). A negative control with PBS instead of primary antibody was run in each slide. Next day, slides were washed with PBS for 5 min. A Multi-Link was used as the secondary antibody for 20 min, followed by PBS wash for 5 min. The slides were placed in the Streptavidin Peroxidase for 20 min (Super Sensitive Link-Label IHC Detection System, Cat#LP000-UCLE, BioGenex, San Ramon, CA), followed by PBS wash for 5 min. For development, one drop of 3,3′ Diaminobencidine (DAB) (Cat#HK542-XAKE, BioGenex, San Ramon, CA) was used on each tissue and the exposure was monitored for 45 seconds under a light microscope. Then, the slides were dipped in distilled water, washed with running water for 5 min, dehydrated through graded alcohol, cleared with xylene and mounted with Cytoseal 60 (Cat#8310-4, Thermo Scientific). Two representative HPF per tissue were photographed using an Olympus BX-60 and a Nikon DS-FI1 camera and analyzed blindly for ERα expression by an observer using Image J software.

Hematoxylin and eosin staining was performed on collected white adipose tissues to measure size. Photographs from 2 high power fields were blindly assessed by MM-C using ImageJ software (NIH, USA) to measure the length and width of 10 randomly selected adipocytes per photo, to calculate and average μm² per animal.

Tissue myeloperoxidase (MPO) activity was determined in 100 mg of flash-frozen uterine samples as an index of granulocyte infiltration as previously described (21).

Thirty milligram of frozen mesenteric fat was placed into 700 μl QIAzol Lysis Reagent in 1.5 ml microcentrifuge tubes for disruption and homogenization. RNA was isolated using the AllPrep DNA/RNA/Protein Mini Kit (Cat# 800104, Qiagen) according to manufacturer's instructions. After mesenteric fat extraction, RNA concentration and integrity was verified using Nanodrop 200 (Thermofisher). mRNA was converted to complementary DNA using iScript cDNA synthesis kit (Cat#1708891, Bio-Rad). Realtime Polymerase reaction was performed using iQ SYBR Green Supermix (Cat#1708882, Bio-Rad) and primers: ERα (PPR44939B), ERβ (PPR48980A), and beta actin (PPR06570C as internal control, Qiagen) following the guidelines from the manufacturer. Data was reported as a fold change or mRNA levels using the equation 2^−ΔCT.

Data was analyzed by using GraphPad Instat version 3.0 (GraphPad Software, San Diego, CA) and GraphPad Prism. A p < 0.05 was considered to represent a statistical significance difference. The mean difference ± the standard error to the mean (S.E.M.) was used to assess the differences among treatment groups. Values more than 2 SD from the mean were excluded as outliers. In order to assess the statistical significance of the mean differences, a parametric one-way ANOVA was used for normally distributed variables, using the Student-Newman-Keuls. A non-parametric Kruskal-Wallis H-test was used for not normally distributed variables (development of vesicles), and the Mann-Whitney U-test was used for the post-hoc pairwise contrasts after taking into account the accumulation of type I error. A two-way ANOVA was used to analyse the differences in weight change and food consumption at different time points.

There were no significant differences in weight between treatment groups at the beginning of the protocol (average 196.62 ± 2.00 g). Animals in the exercise group tended to weigh less at time of surgery, but this did not reach significance. During the protocol all animals gained weight, however the endometriosis animals exposed to exercise gained significantly less weight at various time points than non-exercised (p < 0.01; Figure 1B), while having a significantly higher food consumption after surgery (p < 0.001; Figure 1C). By the time of sacrifice the Endo-Exercise animals weighed significantly less (253.00 ± 4.12 g) than the Endo-No exercise (271.20 ± 5.70 g; p < 0.05) but were comparable in weight to the Sham-No exercise (265.22 ± 3.90 g). No significant differences in anxiety behavior or corticosterone levels were noted among groups (Supplementary Figures 1A–E).

After the animals were sacrificed, classification of the vesicles was performed as described previously. Endo-No exercise animals developed vesicles in 85% of the implants (Figure 2A) with an average total area of 47.45 ± 5.70 mm² (Figure 2B) and average vesicle weight of 0.052 ± 0.004 g (Figures 2C,D). Exposure to exercise significantly decreased the number of vesicles that developed (62.5%; p < 0.01), as well as their size (21.31 ± 3.86 mm², p < 0.01) and weight (0.037 ± 0.004 g, p < 0.05). None of the exercise animals developed vesicles which were ≥6.0 mm (Figure 2E). As expected, no vesicles were found in the Sham-No exercise. We found no significant impact of the stress protocol on reproductive cycle length or stage (data not shown).

All animals adapted to the running wheels quickly and met the minimum exclusion of 2 km/day within 5 days. Distance and speed in Endo-Exercise group were inversely correlated with number of developed vesicles per rat (p < 0.05; Figures 3A,B). The muscle/fat ratio was increased in the Endo-exercise group when compared to the Endo-No exercise group (p < 0.05), indicating differences were due to the presence of less fat rather than differences in muscle weight following exercise (Figures 3C,D). There was no significant difference in weight of the adrenal glands when normalized to body weight (Supplementary Figure 2A). No significant differences in neutrophil infiltration as assessed by myeloperoxidase measurement was found between groups in the uterus (Supplementary Figure 2B).

Interestingly Endo-No exercise animals had significantly higher amounts of mesenteric fat when compared to the Sham-No exercise (p < 0.05) and this was mitigated in Endo-Exercise (p < 0.01; Figure 4A). Rats with a higher percent of fat developed more vesicles (p < 0.05; Figure 4B) which were of a greater size (p < 0.01; Figure 4C). Circulating serum levels of leptin, a hormone predominantly made by adipocytes, was significantly decreased by exercise (p < 0.001; Figure 4D). The size of the adipocytes in the mesenteric white adipose tissue from Endo-Exercise animals was approximately half of those found in either Sham-No exercise (p < 0.05) or Endo-No exercise (p < 0.01; Figures 4E,F).

Systemic levels of various cytokines, chemokines, and immunomodulators were assessed in serum collected at time of sacrifice (Table 1). Fractalkine, a chemokine known to be related with neuropathic pain, was downregulated by exercise (p < 0.01; Figure 5A). Lower levels of fractalkine in serum correlated with lower percentage of fat (Figure 5B) as well as decreased vesicle size (Figure 5C). Exercise also produced a significant reduction in levels of the chemokines RANTES (CCL5), which recruits leukocytes during inflammation, and LIX (CXCL5; Table 1).

Immunofluorescent staining for fractalkine expression in vesicles from Endo-Exercise animals was significantly less compared to Endo-No Exercise (p < 0.05; Figures 5D,F), while the expression of its receptor CX₃CR1 almost doubled but did not reach significance (Figures 5E,F). No significant differences in fractalkine or its receptor were found in the mesenteric fat (data not shown).

In the peritoneal fluid, levels of lactoferrin, a natural immunomodulator, was found to be higher in exercised animals, and tended to inversely correlate with lesion size (n = 8–9/group; r = −0.1093) suggesting that it might be released to counteract the inflammation in the endometriosis animals (Figures 6A,B).

The average mast cell count decreased significantly in vesicles in Endo-Exercise (p < 0.05; Figures 6C,E) and was also lower in mesenteric fat (p = 0.06; Figures 6D,F).

Although we observed no differences in serum levels of estradiol between treatment groups at time of sacrifice, nor variations with stage of cycle (data not shown), exercise resulted in a significant increase in the levels of ERα in the vesicles (p < 0.001: Figures 7A,B) concomitant with a reduction in macrophage infiltration (p < 0.05; Figure 7C) and ERβ immune-positive cells (p < 0.01: Figures 7D,E). This pattern was also noted in the adjacent mesenteric fat (Figures 7F–J). A positive correlation between ERβ and number of macrophages was found in vesicles (p < 0.01), with a tendency toward a similar pattern between ERβ and percent body fat observed in the mesenteric fat surrounding the vesicles (Supplementary Figure 3).

ERβ mRNA expression levels in mesenteric fat were also significantly decreased by exercise (p < 0.05; Figure 8A), in tandem with decreased levels of the inflammatory cytokines IL-6 (Figure 8C) and IL-1β (p < 0.01; Figure 8E). In each case these correlated with percent body fat reaching significance for ERβ (Figure 8B) and IL1β (Figure 8F) but not for IL-6 (Figure 8D). No significant differences were found in ERα expression.