Twelve sedentary participants (11 women and 1 man) aged 64 ± 7 years and with confirmed and stable RA were recruited for the study. Participants were either seropositive or had radiographic joint erosions of finger joints, met the 1987 American College of Rheumatology criteria for RA , had no medication changes in the previous 3 months, and were using prednisone ≤ 5 mg/d. Exclusions were known diabetes mellitus or cardiovascular disease and an inability to walk unaided on a treadmill. All participants gave written informed consent, and the study was approved by the Duke University Medical Center Institutional Review Board (IRB no. Pro00064057).
Exercise training consisted of 10 weeks of 3 × 30-minute sessions per week of supervised treadmill walking. In aggregate, participants completed 99% of prescribed exercise sessions. Exercise intensities were determined from a cardiorespiratory fitness test. For exercise prescription, VO2 reserve was chosen and calculated as previously described . Participants were given between three and six sessions to become accustomed to the exercise (30- to 45-second intervals at target heart rates; total time, 20 minutes). Exercise consisted of a 5-minute warm-up and 5-minute cool-down as part of the total session. Intervals were designed to elicit a heart rate corresponding to 80–90% of VO2 reserve (high intensity = actual heart rate percentage of 85 ± 5%) and 50–60% VO2 reserve (active recovery). Speeds did not exceed walking pace (range, 1–4.6 mph), and if heart rate was not achieved by walking speed, gradient (range, 2–15%) was added to increase heart rate. High-intensity intervals were between 60 and 90 seconds, followed by active recovery intervals of a similar duration. Rather than controlling each session for energy expenditure, total intervals per session were adjusted so that the exercise duration per session was 30 minutes. Ratings of perceived exertion were detailed at the end of each high-intensity interval bout.
Assessment of disease activity
All participants were examined by a rheumatologist using previously described methods for determining disease activity using the Disease Activity Score in 28 joints (DAS28) . This score includes assessment of the number of swollen and tender joints, a visual analogue scale for general health, and C-reactive protein (CRP) concentration or erythrocyte sedimentation rate (ESR) (DAS28CRP and DAS28ESR, respectively).
Fitness, function, and body composition
Exercise treadmill testing was used to assess cardiorespiratory fitness. Aerobic capacity (VO2peak) was determined by a graded maximal treadmill test starting at 2 mph/0% grade and then increasing speed and/or grade such that the metabolic demand increased at approximately 3.5 ml/kg/min until volitional exhaustion. Twelve-lead electrocardiogram, ventilation, and gas exchange were continuously assessed and recorded as 15-second averages using a Parvo Metabolic Cart (Parvo Medics, Sandy, UT, USA). Highest 15-second values were used for determination of VO2peak, and a valid test was confirmed by either a respiratory exchange ratio (RER) > 1.1 (mean RER, 1.2 ± 0.1 at both times) or a rating of perceived exertion ≥ 17. Body composition was assessed according to Siri’s three-compartmental model . Body weight, fat mass, and lean mass were determined by air displacement plethysmography (BOD POD System; COSMED, Rome, Italy) . Resting blood pressure and heart rate were taken following 15 minutes of sitting quietly. Each participant also completed a short battery of physical functioning tests that are indicative of health and frailty in older adults [38,39,40]. Grip strength was assessed by dynamometry in both hands in triplicate; the best score was taken. Timed Up and Go (TUG), Berg Balance Scale, 30-second chair stands, and 400-m walk tests were all supervised by a trained exercise physiologist. All tests were completed before training and at least 24 hours after the last exercise session.
Physical activity and health questionnaires
During their fasting blood draw visits (see below) at baseline and following the intervention, participants completed standardized validated physical activity and health questionnaires. Physical activity questionnaires were used to determine that participants were not engaging in more physical activity than what was prescribed in the exercise program. Questionnaires used included the Incidental and Planned Exercise Questionnaire (IPEQ), the Stanford Brief Activity Survey (SBAS), the Hospital Anxiety and Depression Scale (HADS), a visual analogue scale for pain (VAS), and the Health Assessment Questionnaire Disability Index (HAQ-DI).
Fasting glucose and insulin
Blood was obtained following a 10-hour overnight fast both before training and between 16 and 24 hours after the last exercise bout. Glucose was measured with a YSI Biochemistry Analyzer (YSI Inc., Yellow Springs, OH, USA), and insulin was measured by enzyme-linked immunosorbent assay (ELISA) according to the manufacturer’s guidelines (Thermo Fisher Scientific, Waltham, MA, USA); both were measured in duplicate. Hemoglobin A1C (HbA1c) values were obtained through a commercial clinical analysis company (LabCorp, Burlington, NC, USA). The homeostasis model assessment of insulin resistance (HOMA-IR) was calculated as described previously using the HOMA2-IR online calculator [41, 42].
Immune cell isolation
Complete blood count differentials were measured using a commercial clinical analysis company (LabCorp). Neutrophils were isolated from heparin-treated blood following 2% dextran sedimentation and separation on a discontinuous Percoll gradient as previously described . Neutrophil purity and viability were determined by Giemsa staining (Diff-Quik; Gentaur Europe, Kampenhout, Belgium) and trypan blue exclusion, respectively. Neutrophil purity and viability were consistently ≥ 95%. Neutrophils were resuspended at 3 × 106 cells/ml in either RPMI 1640 (+calcium, +magnesium: RPMI) medium only (Sigma-Aldrich, St. Louis, MO, USA), RPMI 1640 containing 0.15% bovine serum albumin (BSA; RPMI-BSA; Sigma-Aldrich) or HBSS+/+ (+ calcium, + magnesium) medium, depending on the assay.
Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll-Paque (GE Healthcare Life Sciences, Marlborough, MA, USA) density gradient centrifugation. Briefly, blood was diluted 1:1 with RPMI 1640 and layered over the Ficoll before centrifugation at 400 × g for 30 minutes at room temperature. The mononuclear cell layer was removed and washed twice before being resuspended at 1 × 106 cells/ml in RPMI + 1% BSA.
Neutrophil migratory dynamics were assessed using an Insall chamber (Weber Scientific International Ltd., Teddington, UK) as previously described [4, 43]. Briefly, coverslips were coated with 7.5% culture-tested BSA (Sigma-Aldrich), and neutrophils were suspended in RMPI-BSA adhered to this surface for 20 minutes at room temperature. The use of this coating of BSA has previously been shown to mimic the ligand for CD11b and CD18, intercellular adhesion molecule (ICAM)-1 . The coverslip was then inverted on the Insall chamber before addition of buffer (RPMI) alone as a control or buffer containing 100 nM CXCL-8 (R&D Systems, Minneapolis, MN, USA).
Neutrophil migration was monitored using a Zeiss Axiovert 100 inverted microscope (Carl Zeiss Microscopy, Buffalo Grove, IL, USA) fitted with a Hamamatsu ORCA 100 digital camera (Hamamatsu, Japan). Time-lapse recordings and calculations of neutrophil migratory dynamics were performed as previously described . Briefly, the Insall chamber allows the formation of stable chemoattractant gradients, with defined, consistent direction in the y direction for each experiment . Only distance traveled in the y direction over time was included in calculations of chemotaxis. Migration was assessed using three parameters: average cell speed (μm/min) of movement toward the chemokine (termed chemokinesis), average velocity (μm/min) of cells (termed chemotaxis), and accuracy of movement (termed chemotactic index). Chemotactic index is expressed in a comparative scale and arbitrary units (a.u.) ranging from − 1 to + 1. Movement directly toward the chemoattractant is + 1, whereas movement directly away is − 1.
Recordings lasted 20 minutes per experiment, with 20 slides captured using OpenLab software (Improvision, Coventry, UK). The Java software ImageJ (Wayne Rasband, National Institutes of Health, Bethesda, MD, USA) was used to analyze cell tracks. All analyses were carried out by a single analyst blinded to subject group and cell conditions.
Phagocytosis of opsonized fluorescein isothiocyanate (FITC)-labeled Escherichia coli (Thermo Fisher Scientific) was assessed in whole blood as previously described . Briefly, phagocytosis was assessed in heparin-treated whole blood incubated for 10 minutes at 4 °C (control: no phagocytosis) or 37 °C (test) with FITC-labeled E. coli. Phagocytosis was halted by the addition of cold PBS, whereas cell surface-bound FITC was quenched by addition of 1% trypan blue solution. Unbound free bacteria were removed by washing in PBS, then erythrocytes were lysed and leukocytes were fixed using 1% fix/lyse solution (Thermo Fisher Scientific). Cell DNA was counterstained by addition of propidium iodide (PI) in order to gate on immune cells before flow cytometry was performed, and 10,000 neutrophils were acquired. Median fluorescence intensity (MFI) of the control was subtracted from the test to give a final phagocytic value. The MFI corresponds to a cell-by-cell “amount’ or phagocytic capacity of E. coli.
Neutrophil ROS production
ROS generation was assessed by luminol-amplified chemiluminescence as previously described . Briefly, resting neutrophils [1 × 105 in HBSS+/+ (+ calcium + magnesium)] were dispensed into a 96-well, white, flat-bottomed plate (Corning, Corning, NY, USA) containing 1 μM luminol (pH 7.3; Sigma-Aldrich). Cells were stimulated with 25 nM phorbol 12-myristate 13-acetate (PMA) (test) or HBSS+/+  and immediately assessed for ROS generation at 1-minute intervals for 60 minutes using an Infinite 200 PRO plate reader (Tecan Life Sciences, Männedorf, Switzerland). Experiments were performed in triplicate, with ROS production measured as relative light units and calculated as the AUC.
Measurement of immune cell surface receptor expression
Neutrophil surface receptor expression was assessed in whole blood. Briefly, 100 μl of heparin-treated blood was dispensed into 5-ml tubes and stored at 4 °C in the dark. Cells were stained with anti-CXCR2-phycoerythrin (anti-CXCR2-PE, clone 5E8-C7-F10; Thermo Fisher Scientific), anti-CD16-FITC (clone 3G8; BD Biosciences, San Jose, CA, USA), anti-CD11b-allophycocyanin (anti-CD11b-APC, clone ICRF44; BD Biosciences), anti-CD18-PE (clone 6.7; BD Biosciences), anti-Toll-like receptor 2 (TLR2)-Alexa Fluor 647 (clone 11G7; BD Biosciences), or anti-TLR4-APC (clone HTA-125; Thermo Fisher Scientific), or their relevant concentration-matched isotype controls for 60 minutes on ice in the dark. Following incubation, cells were washed twice in cold PBS, and erythrocytes were lysed and leukocytes were fixed using 1% fix/lyse solution (Thermo Fisher Scientific). Following fixing, cells were washed twice and resuspended in 300 μl of PBS for analysis by flow cytometry.
Monocyte surface receptor expression was assessed on freshly isolated PBMCs (1 × 105 cells/ml). Cells were stained with CD14-Pacific Blue (clone TuK4; Thermo Fisher Scientific), CD16-FITC, TLR2-Alexa Fluor 647, TLR4-PE, anti-HLA-DR-PE-CF594 (clone G46-6; BD Biosciences), or relevant isotype controls for 30 minutes at 4 °C in the dark. Postincubation cells were washed twice in PBS/1% BSA, resuspended in 300 μl of PBS/1% BSA, and transferred to polypropylene fluorescence-activated cell sorting tubes for analysis by flow cytometry.
All flow cytometric analyses were conducted on a BD FACSCanto II (BD Biosciences) flow cytometer equipped with three lasers using the Duke Cancer Institute Core Facility, which maintained daily quality controls of the machine. Ten thousand neutrophils and 5000 monocytes were acquired for analysis. Data were analyzed using FCS Express 6 (De Novo Software, Glendale, CA, USA).
Samples were processed immediately for plasma and immune cell isolation, and relevant samples were stored at − 80 °C until analysis. All plasma analyses were completed by the Core Facilities within the Duke Molecular Physiology Institute. Plasma samples obtained following an overnight fast were analyzed for five cytokines and one acute-phase protein: CXCL-8, IL-6, IL-10, TNF-α, IL-1β, and C-reactive protein (CRP). Concentrations of cytokines were determined in duplicate using a human proinflammatory 5-plex sandwich immunoassay according to the manufacturer’s instructions (Meso Scale Discovery, Rockville, MD, USA). High-sensitivity CRP was measured in duplicate using a commercially available ELISA (IBL International, Hamburg, Germany). The lower limits of detection (LLODs) were as follows: CXCL-8 (0.08 pg/ml), IL-6 (0.11 pg/ml), IL-10 (0.05 pg/ml), TNF-α (0.09 pg/ml), IL-1β (0.03 pg/ml), and CRP (0.02 mg/L). All samples had concentrations greater than the LLOD, with the exception of IL-1β with 77% of samples above the LLOD. Plasma concentrations of nonesterified fatty acid (NEFA) were assessed by an enzymatic colorimetric assay on a UniCel DxC600 Analyzer using the manufacturer’s guidelines (Beckman Coulter Life Sciences, Indianapolis, IN, USA).
All analyses were conducted using IBM SPSS Statistics version 23.0 software (IBM, Armonk, NY, USA), and all data are presented as mean ± SD unless otherwise stated. The primary outcome of the study was change in disease activity, and the secondary outcome was change in VO2peak. We initially powered the study to detect a 5% change in VO2peak based on preliminary data in diabetic adults as in our previous studies . The sample size for change in VO2peak was five participants. Following completion of our analysis of these five participants, we powered the remainder of the study on change in disease activity. Sample size calculations suggested a total of 12 participants would be needed to detect a significant 16% reduction (80% power) in disease activity. Normality was assessed using Kolmogorov-Smirnov analysis; natural logarithmic transformation of distributed variables violating normality was completed. Pairwise comparisons of variables were completed using paired t tests. Bivariate correlations were conducted between changes in fitness, body composition, inflammatory markers, and immune functions to tease out associations. Statistical significance was accepted as p ≤ 0.05.