An fMRI study on the acute effects of exercise on pain processing in trained athletes

Ò

PAIN 153 (2012) 1702–1714

www.elsevier.com/locate/pain

An fMRI study on the acute effects of exercise on pain processing in trained athletes Lukas Scheef a, Jakob Jankowski a, Marcel Daamen a, Gunther Weyer a, Markus Klingenberg a, Julia Renner a, Sara Mueckter a, Britta Schürmann b,c, Frank Musshoff d, Michael Wagner b, Hans H. Schild e, Andreas Zimmer c, Henning Boecker a,⇑ a

Functional Neuroimaging Group, Department of Radiology, University of Bonn, Germany Department of Psychiatry, University of Bonn, Germany Institute of Molecular Psychiatry, University of Bonn, Germany d Institute of Forensic Medicine, University of Bonn, Germany e Department of Radiology, University of Bonn, Germany b c

Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.

a r t i c l e

i n f o

Article history: Received 22 November 2011 Received in revised form 28 March 2012 Accepted 2 May 2012

Keywords: Affect Endorphin fMRI Opioid Pain Periaqueductal gray Physical exercise

a b s t r a c t Endurance exercise is known to promote sustained antinociceptive effects, and there is evidence that the reduction of pain perception mediated by exercise is driven by central opioidergic neurotransmission. To directly investigate the involved brain areas and the underlying neural mechanisms in humans, thermal heat-pain challenges were applied to 20 athletes during 4 separate functional magnetic resonance imaging (fMRI) scans, i.e., before and after 2 hours of running (exercise condition) and walking (control condition), respectively. Imaging revealed a reproducible pattern of distributed pain-related activation in all 4 conditions, including the mesial and lateral pain systems, and the periaqueductal gray (PAG) as a key region of the descending antinociceptive pathway. At the behavioral level, running as compared with walking decreased affective pain ratings. The influence of exercise on pain-related activation was reflected in a significant time  treatment interaction in the PAG, along with similar trends in the pregenual anterior cingulate cortex and the middle insular cortex, where pain-induced activation levels were elevated after walking, but decreased or unchanged after running. Our findings indicate that enhanced reactive recruitment of endogenous antinociceptive mechanisms after aversive repeated pain exposure is attenuated by exercise. The fact that running, but not walking, reproducibly elevated b-endorphin levels in plasma indicates involvement of the opioidergic system in exercise. This may argue for an elevated opioidergic tone in the brain of athletes, mediating antinociceptive mechanisms. Our findings provide the first evidence using functional imaging to support the role of endurance exercise in pain modulation. Ó 2012 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

1. Introduction Endurance exercise has been shown to promote various psychophysiological effects, including enhanced attentional and memory capacity, mood induction, and antinociception [12,24,25,44]. Pertaining to pain perception, elevated pain thresholds were demonstrated in athletes as compared with untrained individuals [13,26]. Antinociceptive effects were attributed to opioidergic mechanisms [41] and may account for beneficial effects of physical activity in chronic pain states [31]. Also, the raised incidence of silent myocardial ischemia in athletes [52] may be explained by raised opioidergic tone, thereby blunting the alerting pain symptoms of ischemia. Pain threshold elevations induced by regular ⇑ Corresponding author. Address: Functional Neuroimaging Group, Department of Radiology, University of Bonn, Sigmund-Freud-Strasse 25, 53105 Bonn, Germany. Tel.: +49 (0)228 287 15970/80; fax: +49 (0)228 287 14457. E-mail address: [email protected] (H. Boecker).

exercise are reversed by the opioid receptor antagonist naloxone [53] and the antinociceptive effects of exogenous opiates are attenuated by exercise, a phenomenon referred to as cross-tolerance [36]. Similar cross-tolerance effects can also be elicited directly by local morphine administration into the periaqueductal gray (PAG) [37], a key area of the descending antinociceptive pathway that receives input from the pregenual anterior cingulate cortex (pgACC) to mediate opioidergic antinociceptive effects [45]. Human exercise studies indicate endogenous endorphin release via the hypophyseal gland: after submaximal exercise, plasma levels of b-endorphin immunoreactive material (IRM) are increased in the order of 2- to 5-fold [17]. After sustained anaerobic exercise, levels of b-endorphin IRM increase to an even larger extent [43,46] and stay elevated for a longer time period [18,23]. However, the quantitative relationship between peripheral (plasma) and central (brain) opioidergic effects remains unknown because released endorphins in plasma hardly reenter the brain via the blood-brain barrier. Hence, positron emission tomography (PET)

0304-3959/$36.00 Ó 2012 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.pain.2012.05.008

Ò

L. Scheef et al. / PAIN 153 (2012) 1702–1714

ligand activation studies have been proposed recently as a more direct approach for studying opioidergic effects of exercise [7,8], and previous work in athletes revealed opioidergic activation upon a 2hours endurance exercise challenge in the pgACC and the anterior insular lobule [9]. Based on these PET findings, we hypothesized endurance exercise of a similar intensity and length to specifically modulate pain responses in regions with high opioid receptor expression, notably the mesial affective (e.g., insula, anterior cingulate) and the descending antinociceptive pathway including pgACC and PAG. Peripheral b-endorphin IRM was expected to increase after exercise, but according to the literature [8] no hypothesis was generated regarding the magnitude of individual plasma level changes and their link to behavioral and/or systemic indices of the individual pain experience. 2. Materials and methods 2.1. Participants Twenty-two right-handed healthy male athletes (average training distance 59.25 ± 29.17 km/week) without a history of neurological, psychiatric, or cardiovascular disease were enrolled after written informed consent. Two subjects were excluded due to imaging artifacts or technical problems, resulting in a final sample size of N = 20 subjects (mean age 39.0 ± 9.2 years). In 1 of these, the measurement of plasma b-endorphin IRM failed, resulting in N = 19 subjects available for endorphin analysis. The mean laterality quotient of the Edinburgh Handedness Inventory [40] was 85.7 ± 19.3. The mean estimated verbal intelligence quotient, using a German vocabulary test (Wortschatztest [47]) was 113.45 ± 8.87. At study entrance, the volunteers’ mean score for the Beck Depression Inventory (German version) was 3.16 ± 3.57 [22]. The study was approved by the local ethics committee of the University Hospital Bonn (l.f.d.Nr. 132/06) and is in compliance with national legislation and the Code of Ethical Principles for Medical Research Involving Human Subjects of the World Medical Association (Declaration of Helsinki).

1703

pain stimulation and rating of the perceived pain inside the scanner to familiarize the participants with the subsequent procedures; 5) functional magnetic resonance imaging (fMRI) pain experiment; 6) rating of the perceived pain immediately after the pain fMRI experiment; 7) measurement of body weight, lactate, and blood glucose levels; 8) 2-hours outdoor Walk or Run and rating of exertion levels after 1 and 2 hours; 9) measurement of body weight, lactate, and blood glucose levels; and 10) the identical schedule as described above (2 through 6). The time delay between the end of the exercise challenge/control condition and the pain fMRI experiment was approximately 45 minutes. The duration of the entire procedure per subject and day was approximately 6 hours (Fig. 1). To ensure that the relative intensity of exercise (Run condition) was comparable across all subjects, the maximal heart rate as well as the anaerobic lactate threshold was individually determined by an exercise physiologist (M.K.) performing a lactate endurance treadmill test about 1 week before the MRI examinations took place. For the Run condition, each athlete was instructed to stay within a given heart rate that targeted the individual aerobic-anaerobic transition zone, but remained below the lactate threshold. During the interventions, we monitored covered distance, running speed, and heart rate using a Polar RS 300x running computer (Polar Electro Oy, Kempele, Finland). Subjective exertion levels were assessed during (i.e., after 1 h) and immediately after (i.e., after 2 h) the intervention using the Borg Scale [[10], German version:[32]]. Immediately after the intervention, blood lactate (Lactate Pro LT 1710 analyzer, Arkray, Kyoto, Japan) and blood glucose levels (Ascensia Contour analyzer, Bayer AG, Leverkusen, Germany) were measured. These blood samples were taken from the participants’ earlobes. Before and after each treatment condition, blood samples were collected on each examination day for later quantification of plasma b-endorphin IRM levels, and urine samples were collected for drug screening. Levels of b-endorphin IRM were determined from blood plasma using a radioactive immunosorbent assay [35]. 2.3. Pain and sensory threshold recordings

2.2. General procedures Each subject participated in 2 different treatment conditions, walking (Walk, control condition) and running (Run, endurance exercise as experimental condition), in a randomized, counterbalanced within-subject design (group I: Walk/Run, N = 10; group II: Run/Walk, N = 10). The experimental and control conditions were conducted on separate days, with a mean interval between examination days of 55 ± 49 days. On each examination day, participants completed the following schedule: 1) blood and urine samples; 2) mood ratings; 3) individual heat and pain thresholds outside the scanner; 4) probes of heat

Because it may be argued that an opioid entry into the brain would not be required for an antinociceptive effect due to potential effects of endogenous opioids on peripheral pain afferents and/or spinal neurons, we performed measurements of warmth perception thresholds (WPT) and thermal heat pain thresholds (HPT) before and after the interventions. In fact, demonstration of identical pain threshold recordings over all 4 conditions is considered an important prerequisite for interpreting any given changes of evoked pain responses in the brain by assuring that the sensory and nociceptive input is constant across conditions and interventions. WPT and HPT were acquired with the MEDOC-TSA II thermal

Fig. 1. Study design. Each subject participated in 2 different treatment conditions on 2 separate days, Walk (control condition) and Run (experimental condition), in a randomized, counterbalanced within-subject design (group A: Run-Walk; group B: Walk-Run).

1704

Ò

L. Scheef et al. / PAIN 153 (2012) 1702–1714

pain stimulation device (Medoc, Ramat-Yishai, Israel) using the ‘‘ascending methods of limits’’ option (rise time: 0.5°C/s, starting temperature 32°C). The 9 cm2 contact Peltier thermode was placed on the right volar forearm 2 cm below the arm bend. To assure the identical position before and after intervention and over treatment days, the position of the Peltier element was marked on the skin using a customized positioning device. For the WPT, the participants were instructed to immediately press a button when experiencing a first-time temperature change. For the HPT, they were instructed to press a button when the thermal heat perception was accompanied for the first time by a painful sensation. According to the WPT/HPT protocol provided by Bär et al. [3,57], 3 test trials were followed by 5 main trials for the WPT and the HPT assessments. The test trials allowed the participants to familiarize with the protocol and were not included in the data analysis. The main trials were averaged to determine the respective thresholds.

separately in duplicate assays. As standard curve, peptide concentrations of 1 to 640 pg/mL were used, enabling us to measure a range of approximately 0.03 to 85 pmol/mL plasma. The antibody used showed a reactivity of 100% to human b-endorphin, b-lipotrophin, and to a lower extent, to other fragments of b-endorphin (S2013 RIA, see earlier). 2.6. Drug testing for opioids To ensure that the pain examinations during fMRI were not affected by external opioid intake, urine samples were acquired in each participant during each of the 2 treatments and analyzed by an opiate immunoassay (CEDIA, Microgenics, Germany). The ‘‘Mandatory Guidelines for Federal Workplace Drug Testing’’ (http://www.workplace.samhsa.gov) were considered using a cutoff level of 2000 ng/mL. Drug testing revealed negative results in all samples.

2.4. Mood ratings 2.7. Analysis of behavioral data Mood ratings were collected using the Profile of Mood States questionnaire (POMS) [38] and the Visual Analogue Mood Scale (VAMS) [1]. An additional item ‘‘euphoria’’ was added to the 8 items of the VAMS scales (sadness, tension, fear, anger, confusion, fatigue, happiness, energy), as in Boecker et al. [9]. All 9 VAMS items were permutated from trial to trial in order to prevent repetition effects. The verbal descriptors at the end of the 100-mmlength visual analogue scales were ‘‘not at all’’ on the left side versus ‘‘maximum imaginable’’ on the right side. 2.5. Determination of plasma levels of b-endorphin IRM For all participants, blood samples were collected twice for each treatment condition (see general procedures). Each time, 9 mL of blood was taken from the cubital vein of the right arm using a cooled syringe, containing 14.4 mg of EDTA (1.6 mg/mL blood) and 800 lg of the protease inhibitor aprotinin (85 lg/mL equivalent to 500 KIU/mL blood; Roche, Mannheim, Germany). Samples were placed on ice immediately, centrifuged at 1500g at 4°C for 15 minutes, plasma transferred into a new tube, directly frozen on dry ice, and stored at 70°C until further processing. To ensure maximum comparability of measurements, samples of a particular subject were always processed in parallel at identical conditions (i.e., same extraction and radioimmunoassay (RIA)). Using reversed-phase Sep-Pak plus cartridges (containing 360 mg of C18 sepharose; Waters, Milford, MA), extraction of plasma peptides was performed at 4°C and at flow rates of approximately 1 mL/min. Each cartridge was activated using 10 mL of 100% methanol, followed by 10 mL of urea (6 mol/L in water) and washed with 20 mL of water. Plasma samples were thawed, acidified with 0.1 mL of 1 N HCl per 1 mL plasma, and centrifuged at 18,600g and 4°C for 20 minutes. Next, 4.4 mL of the supernatant was passed through the activated cartridges, which were subsequently washed with 10 mL of acidic acid (4% in water) and 20 mL of water. Elution of peptides from the cartridges was achieved with 4 mL of 1-propanol/acetic acid (96:4 vol/vol) and the eluate divided into 2 samples of 2 mL each. Volatile components of the eluate were removed at room temperature by using a Speed-Vac concentrator (Savant Instruments Inc., Holbrook, NY). The remaining aqueous phase was dried in a lyophilizer (Lyovac GT2, AMSCO/Finn-Aqua, Hürth, Germany), and the extracts were stored at 70°C. The b-endorphin IRM levels in the plasma extracts were determined using a commercially available peptide radioimmunoassay (S-2013 RIA; Penninsula Laboratories, San Carlos, CA) following the instructions provided by the supplier. For the RIA, the dried extract of 2 mL of plasma was dissolved in 230 lL of RIA buffer, and two 100-lL aliquots of this solution were analyzed

Behavioral data with single measurements on each examination day were analyzed using a 2-factorial repeated-measures ANOVA, with a within-subject factor treatment, and order as an additional between-subject factor. Behavioral data with repeated measurements on each examination day were analyzed using a 3-factorial repeated-measures ANOVA, with treatment and time as withinsubject factors, and order as between-subject factors. Post hoc comparisons were performed using paired Student t tests. All statistical analyses were performed using PASW Statistics 18.0 (Chicago, IL). 2.8. fMRI paradigm The fMRI paradigm consisted of 2 main conditions: nonpainful Heat (average temperature 43.0°C, oscillating around ±1.5°C, return slope 10°/s) and Pain (average temperature 46.5°C ± 1.5°C, return slope 10°/s). Stimuli were delivered by the MEDOC-TSA II system. The conditions were presented as blocks of approximately 30 seconds in duration and interleaved by a constant baseline temperature of 35°C, which was also held for approximately 30 seconds. The temperature profile is given in Fig. 2. One period consisted of 2 baseline conditions and 2 main conditions (Baseline,

Fig. 2. Temperature profile during functional magnetic resonance imaging pain experiment. Representative temperature profile of the thermal stimuli over 1 period, showing the undulating temperature time course. The nonpainful heat stimuli are colored in blue, and the painful heat stimuli are colored in red.

Ò

1705

L. Scheef et al. / PAIN 153 (2012) 1702–1714

Heat, Baseline, Pain) and was repeated 5 times in fixed order. Each Heat and Pain block was preceded by a short acoustic signal (0.5 seconds) announcing an upcoming temperature change. The start points of the temperature profiles (nonpainful Heat/Pain conditions) were synchronized with the fMRI data acquisition and recorded for each session. This paradigm (block order, timing, and intensity profiles) was used in all 4 sessions (preRun, postRun, preWalk, postWalk). Immediately after the experiment, participants judged the unpleasantness and the intensity of both stimuli classes on a Likert scale (ranging from 0 (no pain/no unpleasantness) to 10 (strongest pain intensity imaginable/strongest pain unpleasantness imaginable). We also collected the affective and evaluative subscales of the McGill Pain Questionnaire (MPQ) [39], directly after the pain experiment, while subjects were still inside the MRI. 2.9. Image acquisition, preprocessing, and analysis All MRI examinations were performed on a 3T Philips Achieva equipped with an 8-channel SENSE head coil. A single pain fMRI run consisted of 205 continuously acquired image volumes, with a total scan duration of about 8.5 minutes. The sequence parameters for the functional scans were as follows: T2*-weighted echo planar imaging (EPI) sequence, repetition time (TR) = 2595 ms, echo time (TE) = 35 ms, flip angle = 90°; sense factor 2, field of view (FOV) = 230  230  147.6 mm3, voxel size = 3.6  3.6  3.6 mm3; 41 axial slices were acquired per image volume in an interleaved fashion covering the entire brain. To minimize between-session effects due to intraopertor and interoperator variability of the image orientation, an automatic scan examination planning routine (SmartScan, Philips Medical Systems, Best, NL) was used to orient the image volumes parallel to an axial plane intersecting the genu and splenium of the corpus callosum. For anatomical reference and spatial normalization, 2 T1weighted 3-dimensional MPRAGE datasets (TI = 1300 ms, TR = 7.7 ms, TE = 3.9 ms, flip angle = 15°) were acquired with an isotropic voxel resolution of 1 mm3 and averaged post hoc. Preprocessing of the MRI data was performed with SPM5 including motion correction with bias field correction, spatial normalization, and spatial smoothing (8  8  8 mm3). The fMRI data were analyzed on an individual basis using the general linear model (GLM) [19]. Within 1 design matrix, the 4 sessions (PreWalk,

PostWalk, PreRun, PostRun) were modeled separately and contained the 2 main conditions: Heat and Pain. Motion parameters were included as covariates of no interest. To account for the temperature profile of the paradigm (Fig. 2), we modeled these conditions by convolving the temperature peaks with the canonical hemodynamic response function (HRF). Based on this model, the contrast ‘‘Pain–Heat’’ was calculated for each of the 4 sessions and carried over to the second level to analyze the data on a group statistical level. At the second level, a 2  2 factorial repeated-measure randomeffect model was used with treatment (Walk versus Run) and time (before versus after) as factors, and treatment order as covariate of no interest. Significance was considered for the main effects at a statistical threshold of P < .05 (corrected for familywise errors). For the differential contrasts (i.e., effects of treatment, time, and time  treatment interaction) significance was considered at P 6 .001 (uncorrected), when concordant with our anatomical a priori hypotheses (insula, anterior cingulate, pgACC, PAG) as outlined in the Introduction. Anatomical localizations of the activation peaks were determined using the statistical parametric mapping (SPM) Anatomy Toolbox 1.5 [64], and for regions not covered by the toolbox, the AAL-template implemented in MRIcron (http://www.cabiatl.com/ mricro/, Atlanta, GA). 3. Results 3.1. Exercise performance During the walk condition, the average covered distance was approximately 10 km, during the Run condition 23 km. The covered distance during the Run condition significantly exceeded the distance for the Walk condition, (F(1,16) = 470.81; P < .001; Table 1). On the Borg scale, the rated levels of exertion were significantly higher after 1 hours (F(1,18) = 98.94, P < .001; Table 1) and 2 hours (F(1,18) = 104.68, P < .001; Table 1) of running, as compared to the Walk condition. This was corroborated by differences in a number of physiological parameters. The average heart rate during running was significantly higher (F(1,18) = 564.36, P < .001; Table 1). Blood glucose levels showed a significant main effect of time (F(1,19) = 5.63, P = .029; Table 1). Unexpectedly, this effect was mainly driven by a significant decrease in the Walk

Table 1 Descriptive statistics of the exercise and physiological parameters. N* Covered distance, km Exertion level (Borg Scale) Average heart rate, beats/min Body weight, kg

During treatment After 1 h After 2 h During treatment Before After

18 20 20 19

Walk Mean ± SD

Run Mean ± SD

Statistics Main effects and interactions

9.82 ± 2.20 8.50 ± 1.15 8.80 ± 1.28 83.70 ± 8.90 72.50 ± 9.14 72.22 ± 9.1

22.63 ± 2.69 13.25 ± 1.65 14.25 ± 2.20 148.30 ± 11.70 72.52 ± 9.16 71.58 ± 9.22 86.75 ± 9.16 83.2 ± 12.32

Treatment: F(1,16) = 470.807, P < .001 Treatment: F(1,18) = 88.577, P < .001 Treatment: F(1,18) = 80.458, P < .001 Treatment: F(1,18) = 564.364, P < .001 Treatment: F(1,17) = .436, P = .518 Time: F(1,17) = 14.213, P = .002 Treatment  time: F(1,17) = 5.202, P = .036 Treatment: F(1,18) = 1.491, P = .238 Time: F(1,18) = 5.632, P = .029 Treatment  time: F(1,18) = .398, P = .536 Treatment: F(1,18) =.488, P = .499 Time: F(1,18) = 15.216, P = .001 Treatment  time: F(1,18) = 5.652, P = .029 Treatment: F(1,17) = 13.17, P = .002 Time: F(1,17) = 12.87, P = .002 Treatment  time: F(1,17) = 15.406, P = .001

Blood glucose level, mg/dL

Before After

20

85.15 ± 9.09 79.85 ± 8.34

Blood lactate level,  mmol/L

Before After

20

0.64 ± 0.33 0.83 ± 0.32

0.55 ± 0.31 1.0 ± 0.43

b-endorphin IRM plasma level, pmol/L

Before After

19

3.1 ± 4.2 3.6 ± 6.7

3.1 ± 5.0 15.5 ± 18.0

IRM = immunoreactive material. For some comparisons, data for single time points were missing for technical reasons, leaving N < 20 subjects for statistical analysis.   None of the participants showed blood lactate level elevations P2.0 mmol/L after 2 hours running, indicating that the level of exercise remained in the aerobic exercise range. *

Ò

1706

L. Scheef et al. / PAIN 153 (2012) 1702–1714

Fig. 3. Plasma levels of b-endorphin immunoreactive material. Levels of b-endorphin immunoreactive material in pmol/L are shown for each individual subject before and after control (preWalk and postWalk) and before and after exercise (preRun and postRun) conditions.

condition (t(19) = 3.01, P = .007), whereas the decrease remained nonsignificant for the Run condition (t(19) = 1.27, P = .221). This is most likely due to a higher caloric substitution (energy bars and water on demand) in the Run condition, but noteworthy, did not result in a significant difference of posttreatment blood glucose levels between the 2 conditions (t(19)= 1.20, P = .245; Table 1). None of the participants showed blood lactate level elevations P2.0 mmol/L after 2 hours of running, indicating that the level of endurance exercise remained in the intended aerobic exercise range. Body weight showed a main effect of time (F(1,17) = 14.21, P = .002; Table 1), as well as a time  treatment interaction (F(1,17) = 5.20, P = .036). Post hoc comparisons indicated significant decreases in both the Walk condition (t(19) = 2.25, P = .037) and particularly the Run condition (t(18) = 3.63, P = .002). The weight loss was significantly larger after running (t(19) = 2.48, P = .023). The plasma b-endorphin IRM (Fig. 3, Table 1) showed main effects of time (F(1,17) = 12.87, P = .002) and treatment (F(1,17) = 13.17, P = .002), as well as a significant time  treatment interaction (F(1,17) = 15.41, P = .001). Although the mean plasma levels before versus after walking remained stable (t(18) = .57,

P = .576), there was a significant, approximately 5-fold increase of mean plasma levels after as compared to before running (t(18) = 3.57, P = .002). In an exploratory analysis, the absolute b-endorphin IRM changes did not correlate with exercise performance measures (covered distance, speed, heart rate, blood glucose), changes of mood (POMS, VAMS), or pain perception scores (WPT, HPT, MPQ). 3.2. Mood ratings The POMS showed significant effects only for fatigue and depression and anxiety (Table 2). For fatigue, there was a significant main effect of treatment (F(1,18) = 4.93, P = .039) and time (F(1,18) = 6.99, P = .017), which were qualified by a significant time  treatment interaction (F(1,18) = 11.14, P = .004). Although the main effects indicated that fatigue scores were higher for the running than for the walking condition, and higher after than before treatment, the effects were primarily driven by a highly significant increase of fatigue levels from pre-exercise to postexercise in the running condition (t(19)= 3.629, P = .002), whereas there was no significant change from pre-exercise to postexercise in the walk

Table 2 Descriptive statistics for the subjective mood ratings collected before and after both treatments. Walk

Run

Main effect

Interaction

Pretreatment Mean ± Std.

Posttreatment Mean ± Std.

Pretreatment Mean ± Std.

Posttreatment Mean ± Std.

Time

Treatment

Time  Treatment

POMS Fatigue Depression/ Anxiety Hostility Vigor

14.00 ± 5.69 18.60 ± 8.31

12.85 ± 7.12 17.15 ± 9.33

12.55 ± 5.30 17.30 ± 6.53

18.75 ± 6.50 14.60 ± 2.48

F(1,18) = 6.99 p = .017 F(1,18) = 7.99 p = .011

F(1,18) = 4.93 p = .039 F(1,18) = 1.73 p = .205

F(1,18) = 11.14 p = .004 F(1,18) = .57 p = .461

8.95 ± 3.93 29.65 ± 7.73

9.35 ± 5.98 27.50 ± 9.35

8.20 ± 2.29 30.20 ± 8.15

7.65 ± 1.39 29.20 ± 7.96

F(1,18) = 0,05 p = .823 F(1,18) = 2.15, p = .160

F(1,18) = 2.03 p = .171 F(1,18) = 1.09 p = .310

F(1,18) = .91 p = .352 F(1,18) = .25 p = .621

VAMS Anger Confusion Fear Happiness Sadness Tension Energy Fatigue Euphoria

4.05 ± 4.31 7.35 ± 7.64 5.40 ± 8.09 54.60 ± 14.72 9.30 ± 15.01 20.55 ± 18.50 51.85 ± 20.20 30.70 ± 20.50 35.45 ± 24.57

5.55 ± 12.30 6.25 ± 11.67 4.45 ± 4.65 54.05 ± 21.32 7.65 ± 11.96 15.50 ± 16.30 48.90 ± 24.36 22.05 ± 18.56 37.15 ± 25.11

4.55 ± 6.83 9.05 ± 12.74 8.55 ± 13.41 60.55 ± 18.30 8.00 ± 12.93 25.75 ± 21.14 60.20 ± 18.03 29.35 ± 19.72 44.55 ± 25.91

4.75 ± 5.84 6.50 ± 10.62 3.30 ± 3.01 60.20 ± 19.80 3.85 ± 3.76 16.85 ± 18.31 55.35 ± 23.08 29.70 ± 21.29 57.55 ± 21.99

F(1,18) = .74, p = .400 F(1,18) = 3.55, p = .076 F(1,18) = 2.31, p = .145 F(1,18) = .025, p = .875 F(1,18) = 2.09, p = .166 F(1,18) = 5.36, p = .033 F(1,18) = 1.37, p = .258 F(1,18) = 1.29, p = .271 F(1,18) = 6.54, p = .020

F(1,18) = .01 p = .910 F(1,18) = .19 p = .664 F(1,18) = 1.90 p = .184 F(1,18) = 5.17 p = .035 F(1,18) = 1.53 p = .232 F(1,18) = .73 p = .405 F(1,18) = 8.13 p = .011 F(1,18) = .63 p = .437 F(1,18) = 20.65 p = .000

F(1,18) = .13 p = .723 F(1,18) = .14 p = .713 F(1,18) = 2.92 p = .105 F(1,18) = .00 p = .969 F(1,18) = .92 p = .351 F(1,18) = .29 p = .595 F(1,18) = .20 p = .663 F(1,18) = 1.36 p = .259 F(1,18) = 5.24 p = .034

POMS = Profile of Mood States, VAMS = Visual Analogue Mood Scale.

Ò

1707

L. Scheef et al. / PAIN 153 (2012) 1702–1714 Table 3 Descriptive statistics for pain-related parameters. Walk Pretreatment Quantitative sensory testing WPT [°C]* 34.5 ± 1.3 HPT [°C] 47.1 ± 2.1 Numerical rating scales Nonpainful heat Pain intensity Pain unpleasantness Painful heat Pain intensity Pain unpleasantness

Run

Main effect

Interaction

Posttreatment

Pretreatment

Posttreatment

Time

Treatment

35.2 ± 1.7 46.8 ± 2.1

34.8 ± 1.2 47.1 ± 1.5

35.5 ± 1.3 46.9 ± 1.6

F(1,17) = 11.1; P = .004 F(1,18) = 1.9; P = .185

F(1,17) = 1.3; P = .27 F(1,18) < 1

F(1,17) < 1 F(1,18) < 1

of pain perception 0.30 ± 1.27 0.30 ± 0.80

0.10 ± 0.45 0.10 ± 0.45

0.05 ± 0.22 0.05 ± 0.22

0.2 ± 0.70 0±0

F(1,18) < 1 F(1,18) = 2.42; P > .1

F(1,18) < 1 F(1,18) = 1.8; P > .1

F(1,18) = 2.15; P = .11 F(1,18) < 1

7.20 ± 1.51 7.35 ± 1.57

7.50 ± 1.70 7.25 ± 1.71

7.40 ± 1.14 7.45 ± 1.40

7.55 ± 1.15 7.35 ± 1.60

F(1,18) = 2.98; P > .1 F(1,18) < 1

F(1,18) < 1 F(1,18) < 1

F(1,18) < 1 F(1,18) < 1

McGill Pain Questionnaire Affective subscale  2.50 ± 2.24 Evaluative subscale 1.50 ± 1.05

2.95 ± 2.31 1.70 ± 1.13

2.60 ± 2.14 1.40 ± 1.05

2.10 ± 1.55 1.15 ± 0.67

F(1,18) < 1 F(1,18) < 1

F(1,18) = 1.3; P = .27 F(1,18) = 2.531; P = .13

F(1,18) = 3.574; P = .075 F(1,18) = 2.24; P = .152

Quantitative sensory testing heat and pain threshold data were collected immediately before the scanning session. Numerical rating scale and McGill Pain Questionnaire ratings were collected immediately after the scanning session. HPT = thermal heat pain threshold; WPT = warmth perception threshold. * Corrupted WPT logfile data for 1 subject (i.e., N = 19).   Significant decrease in affective pain ratings postRun compared with postWalk (t(19) = 1.78, P = .046, 1-sided t test, post hoc).

condition (P > .3). For depression and anxiety, there was only a significant main effect of time (F(1,18) = 7.99, P = .011). Subjects tended to show lower posttreatment levels of negative mood, irrespective of treatment condition. In the VAMS (Table 2), there was a significant main effect of treatment for the items happy (F(1,18) = 5.17, P = .035) and powerful (F(1,18) = 8.13, P = .011), indicating that participants generally reported to be happier and more powerful, respectively, on the day of running. Regarding tension, there was a significant main effect of time, i.e., participants felt significantly less tense after the intervention (F(1,18) = 5.36, P = .033). The supplementary item euphoria showed a significant main effect of treatment (F(1,18) = 20.65, P < .001), time (F(1,18) = 6.54, P = .02), and a time  treatment interaction (F(1,18) = 5.24, P = .034). Participants already gave higher baseline euphoria ratings before running, as compared with the baseline ratings before walking (t(19 = 2.70, P = .014). Nevertheless, euphoria ratings showed a further increase after running (t(19)= 3.15, P = .005), whereas ratings after walking remained stable (t(19) = .47, P = .648). 3.3. Warmth perception and pain threshold ratings The average WPT across subjects and conditions was 35.0 °C ± 1.4 °C, the average HPT 47.2 °C ± 1.8 °C. Neither of these thresholds showed significant exercise-related changes. See Table 3 for further details. 3.4. Pain ratings The ratings of the perceived nonpainful heat stimuli delivered during the fMRI experiment were generally not perceived as painful (Likert Scale, pain intensity = 0.2 ± 0.3). The ratings for the heat pain stimuli delivered during the fMRI experiment were perceived as strongly painful (Likert Scale, pain intensity = 7.4 ± 1.2). The rating scales for pain intensity and unpleasantness were highly correlated (R P .75, P < .001, for all 4 conditions) and did not show significant effects of treatment or time or time  treatment interactions (all P > .05, see Table 3). On the MPQ Affective Subscale, there was no significant main effect of treatment or time; however, the time  treatment interaction approached significance (F(1,18) = 3.57, P = .075). As post hoc paired t tests indicated, there was no baseline difference before treatment, but a significant difference in the affective pain ratings after treatment, which were significantly lower after running

Fig. 4. Effect of treatment condition on affective subscale of the McGill Pain Questionnaire. Comparison of the means for the McGill Pain Questionnaire– Affective Subscale. The affective ratings for the postRun condition were significantly lower than for the postWalk condition (⁄P < .05, 1-tailed).

as compared with after walking (t(19) = 1.78, P = .046, 1-sided; Fig. 4). The MPQ Evaluative Subscale, on the other hand, did not show significant main effects or interactions (Table 3). Because it could be suggested that the modulation of the affective response to the pain stimuli was related to general affective changes that were observed in the running condition, we conducted an exploratory correlation analysis to examine whether the changes in the MPQ Affective Subscale after running were systematically related to the changes in the other affective variables that had shown significant effects after running (i.e., POMS Fatigue, VAMS Euphoria). None of these 3 variables showed a significant correlation (P > .1). 3.5. Functional MRI The heat pain stimuli, as compared with the nonpainful heat stimuli, induced pain-specific activation patterns when tested in each of the 4 conditions, involving sensory-discriminative (S1, S2, posterior insula, thalamus), affective-motivational (ACC, middle insular cortex extending into anterior insula), and attentional (dorsolateral prefrontal cortex) components of the human pain matrix. Moreover, there was consistent activation of regions belonging to the descending antinociceptive system, including the pgACC and the PAG (Fig. 5; Table 4A to D).

1708

Ò

L. Scheef et al. / PAIN 153 (2012) 1702–1714

The 2  2 ANOVA allowed us to assess differential main effects of treatment (i.e., running versus walking days, independent of prepost assessments), time (i.e., pre versus post assessments, independent of treatment type), and time  treatment interaction (i.e., differential effects of pre-post running versus pre-post walking) on pain processing. Neither treatment nor time contrasts showed significant main effects of pain-related activation, but the time  treatment interaction contrasts, which are specific for the differential effects of running versus walking upon pain processing, showed a significant interaction in the PAG (Fig. 6A, Table 5). Changes in the pgACC (Fig. 6B) and in the middle insular cortex (Fig. 6C) did not survive the defined significance criteria (Table 5), but showed a strong trend (P = .002, uncorrected). No other brain regions except the posterior cingulate gyrus surpassed this more liberal statistical threshold of P < .005, uncorrected. Therefore, the interaction effect, which is specific to the modulation of pain responses to exercise, was identified nearly exclusively in a priori defined brain regions, based on independent imaging evidence. Notably, in all 3 regions concurring with our a priori hypothesis, the pain-evoked blood oxygen level

dependent (BOLD) responses showed an divergent effect after running, as compared to after walking (i.e., increased postWalk, decreased postRun; see Fig. 6). This pattern also applies to the pgACC (Fig. 6B), although negative activation levels were seen in all 4 conditions. Here, the blood oxygen level dependent (BOLD) response became less negative after walking, whereas it was further deactivated after running. 4. Discussion Our results provide evidence that endurance exercise modulates pain-evoked activity in a key region of the human antinociceptive pathway (PAG; see later) [11,29,48]. Notably, the BOLD response was elevated after 2 hours of walking (control condition), whereas it was decreased after 2 hours of running (experimental condition). Other relevant regions of the descending and affective pain systems [11,29,48], namely the pgACC and the insula, showed clear trends toward decreased activation after the intervention. Based on opioid ligand PET data in athletes [9], it was anticipated that the modulation of central pain processing by endurance

Fig. 5. SPM analysis of main effect of pain > heat in 4 treatment conditions separately. Regions of pain-specific (pain > heat) activation in the 4 treatment conditions superimposed onto 3 orthogonal T1-weighted magnetic resonance imaging sections. Note the involvement of sensory-discriminative, affective-motivational, attentional, and antinociceptive components of the human pain matrix. To demonstrate the full extent of activation/deactivation, the data are displayed at P < .005, false discovery rate (FDR)-corrected, extent threshold 20 voxels. The familywise errors-corrected data are given in Table 4A to D).

Ò

1709

L. Scheef et al. / PAIN 153 (2012) 1702–1714 Table 4 Main effects (pain–heat) per condition. A. Pain–heat (preWalk) Brain region

Z scores

X, Y, Z (mm)

Pain-specific activations Pain-associated regions (discriminative/affective/antinociceptive/attentional) Right thalamus, ventral lateral nucleus 168 Right precentral gyrus, BA 6 14 Left postcentral gyrus, BA 40 60 Left postcentral gyrus, BA 40 Right secondary sensory cortex, BA 40 316 Right secondary sensory cortex, BA 40 Right superior temporal gyrus, BA 22 894 Right anterior insula/right claustrum Right superior temporal gyrus, BA 22 Left superior temporal gyrus, BA 22 201 Left middle insula, BA 13 Right posterior insula, BA 13 176 Right putamen Right middle cingulate gyrus, BA 32 651 Right midbrain, PAG 98 Left midbrain, PAG 78

K

5.50 5.00 5.50 5.19 6.26 5.86 6.99 6.98 6.56 5.90 5.68 5.65 5.39 6.34 5.78 5.76

18, 13, 12 48, 2, 44 61, 20, 19 61, 28, 22 55, 24, 25 61, 18, 19 55, 12, 2 34, 4, 9 57, 4, 5 53, 4, 3 40, 8, 0 38, 17, 12 32, 17, 6 8, 12, 36 10, 20, 4 10, 16, 6

Sensory and sensorimotor regions Left cerebellum, posterior lobe, declive Left cerebellum, posterior inferior lobe Left cerebellum, posterior inferior lobe Left cerebellum, anterior lobe, culmen Left cerebellum, anterior lobe Right cerebellum, anterior lobe, culmen Right cerebellum, posterior lobe, declive Right cerebellum, anterior lobe, culmen Right cerebellum, anterior lobe, culmen Right middle cingulate gyrus, BA 32 Right middle frontal gyrus, BA 6 Right superior frontal gyrus, BA 6

6.89 5.78 5.64 5.23 5.11 4.81 5.69 5.50 5.43 6.34 6.11 5.39

26, 61, 19 22, 66, 40 8, 68, 35 12, 45, 15 2, 49, 16 0, 61, 19 24, 61, 19 28, 56, 22 34, 50, 26 8, 12, 36 10, 5, 57 6, 14, 51

695

14 13 122

651

Pain-specific deactivations Default network Left posterior cingulate gyrus, BA 31

15

B. Pain–heat (postWalk) Brain region

T values

5.15

6,

41, 33

Z scores

X, Y, Z (mm)

Pain-specific activations Pain-associated regions (discriminative/affective/antinociceptive/attentional) Right postcentral gyrus, BA 40 Right secondary sensory cortex 106 Right mid/anterior insula, BA 13 371 Right superior temporal gyrus, BA 22 Right superior temporal gyrus, BA 22 Left middle insula, BA 13 118 Left superior temporal gyrus, BA 22 Left superior temporal gyrus, BA 22 Right thalamus, medial dorsal nucleus 438 Right midbrain, PAG Left thalamus, medial dorsal nucleus Right putamen 283 Right thalamus, ventral lateral nucleus Right middle cingulate gyrus, BA 24 111 Right middle cingulate gyrus, BA 32 Left middle cingulate gyrus, BA 24 Right midbrain, PAG 17

5.14 5.86 6.22 6.21 5.80 5.52 5.25 4.77 5.85 5.45 5.18 5.69 5.63 5.62 5.23 4.91 5.00

61, 18, 19 51, 22, 21 36, 6, 9 57, 2, 5 50, 6, 2 38, 10, 0 53, 2, 7 57, 4, 2 8, 17, 6 10, 16, 8 12, 19, 12 30, 21, 7 16, 13, 14 6, 13, 32 4, 6, 40 6, 13, 31 4, 26, 16

Sensory and sensorimotor regions Left cerebellum, posterior lobe, declive Left occipital lingual gyrus, BA 18 Left cerebellum, posterior lobe, declive Left cerebellum, posterior lobe Right cerebellum, posterior lobe, pyramis Right pons, cerebellar peduncle Left midbrain, subthalamic nucleus Right middle frontal gyrus, BA 6

9.80 7.51 7.41 6.23 5.68 5.63 5.40 5.32

Pain-specific deactivations Default network –

2383

31 10 18 12 20

–

–

7.81 6.46 6.40 5.57 5.16 5.13 4.95 4.88

–

28, 59, 21 0, 72, 10 10, 67, 13 18, 62, 41 36, 79, 33 8, 39, 33 10, 16, 6 4, 1, 59

– (continued on next page)

1710

Ò

L. Scheef et al. / PAIN 153 (2012) 1702–1714

Table 4 (continued) C. Pain–heat (preRun) Brain region

T values

Z scores

X, Y, Z (mm)

Pain-specific activations Pain-associated regions (discriminative/affective/antinociceptive/attentional) Right precentral gyrus, BA 6 61 Right postcentral gyrus, BA 5 106 Left postcentral gyrus, BA 40/S2 107 Left postcentral gyrus, BA 43 Right secondary sensory cortex, BA 40 824 Right insula, BA 13 Right insula, BA 13 Right anterior insula, BA 13 1389 Right superior temporal gyrus, BA 22 Right superior temporal gyrus, BA 22 Right thalamus 303 Right thalamus, ventral lateral nucleus Left lateral globus pallidus 116 Left thalamus, ventral lateral nucleus Right midbrain, PAG 233 Right middle frontal gyrus, BA 32 1763 Right anterior cingulate gyrus, BA 32 Right middle frontal gyrus, BA 10 21 Left middle frontal gyrus, BA 10 41 Right middle frontal gyrus, BA 9 23 Right posterior cingulate gyrus, BA 31 20

6.69 6.61 7.42 5.35 9.21 7.30 6.74 9.40 9.32 8.59 6.55 6.35 5.60 5.60 6.98 8.53 7.89 5.53 6.04 5.97 5.50

5.90 5.85 6.40 4.91 7.51 6.32 5.94 7.61 7.57 7.15 5.81 5.66 5.11 5.10 6.10 7.11 6.71 5.05 5.44 5.38 5.02

28, 19, 54 24, 38, 59 61, 22, 21 53, 15, 14 57, 26, 25 48, 20, 19 40, 17, 14 36, 6, 9 55, 12, 1 55, 4, 7 22, 19, 12 16, 13, 14 20, 8, 2 12, 10, 4 10, 20, 4 6, 6, 44 6, 13, 36 32, 42, 24 32, 46, 22 53, 6, 37 14, 21, 40

Sensory and sensorimotor regions Left cerebellum, posterior lobe, declive Left cerebellum, anterior lobe, culmen Left cerebellum, posterior lobe Left precentral gyrus, BA 44 Left precentral gyrus, BA 44 Left midbrain, subthalamic nucleus Left brainstem, cerebellar peduncle Right putamen Left claustrum

47 12 15 11

9.63 8.52 7.77 8.78 7.47 5.94 5.70 5.47 5.45

7.73 7.10 6.63 7.26 6.44 5.36 5.18 5.00 4.98

26, 30, 28, 42, 49, 10, 4, 30, 34,

Pain-specific deactivations Default network –

–

–

–

–

D. Pain–heat (postRun) Brain region

K

T values

Z scores

X, Y, Z (mm)

Pain-specific activations Pain-associated regions (discriminative/affective/antinociceptive/attentional) Right secondary sensory cortex, BA 40 311 Right inferior parietal lobule, BA 40 Left middle cingulate gyrus, BA 32 635 Right medial frontal gyrus, BA 32 Right anterior cingulate gyrus, BA 32 Right claustrum 143 Right anterior insula, BA 13 Right thalamus, ventral lateral nucleus 153 Right thalamus Left superior temporal gyrus, BA 22 47 Left middle insula, BA 13 Left secondary sensory cortex, BA 40 10 Right thalamus, ventral lateral nucleus 15

7.29 5.84 7.25 6.99 6.84 6.86 6.30 6.14 5.37 5.64 5.46 5.60 5.30

6.32 5.28 6.29 6.11 6.01 6.02 5.62 5.51 4.92 5.14 4.99 5.10 4.87

55, 24, 25 57, 37, 39 2, 10, 38 4, 6, 44 6, 13, 34 34, 4, 9 34, 18, 6 16, 13, 14 22, 17, 10 51, 0, 6 44, 6, 2 63, 20, 19 12, 15, 4

Sensory and sensorimotor regions Left cerebellum, posterior lobe, declive Left cerebellum, posterior lobe, declive Left cerebellum, posterior lobe, declive Left cerebellum, posterior lobe Right precentral gyrus, BA 44 Right superior temporal gyrus, BA 22

9.11 7.32 6.16 5.64 7.70 7.45

7.45 6.34 5.53 5.14 6.59 6.42

28, 59, 21 8, 73, 17 10, 55, 14 26, 60, 42 51, 0, 7 51, 8, 1

–

–

–

Pain-specific deactivations Default network –

K

2897

571

1163

20 256

–

59, 52, 52, 6, 5 0, 6 16, 39, 17, 5 20,

19 26 39

6 35 2

Indicated are the activated and deactivated clusters for the main contrast (pain–heat). P < .05, familywise errors corrected. BA = brodmann area; K = cluster size, X, Y, Z = coordinates in Talairach space; PAG = periaqueductal gray.

exercise would involve opioidergic mechanisms. Areas of the mesial affective pain system have a higher opioid receptor distribution compared with the lateral pain system [27], and hence more available opioid binding sites [30,42,50,54,58,63–65]. It was also hypothesized that running would modulate neural activity in the

rostral pgACC and the PAG, as opioid-based analgesia is propagated via the descending antinociceptive pathway [4,45], i.e., through the rostral pgACC [5,6] and the PAG [33] down to the spinal cord [15]. Both regions, the PAG and the ACC, are anatomically connected [2,34], have high opioid receptor distributions [20,28,59,62], and

Ò

L. Scheef et al. / PAIN 153 (2012) 1702–1714

1711

Fig. 6. SPM analysis of the time  treatment interaction contrasts. Regions of pain-specific (pain > heat) activation changes in the time  treatment interaction contrasts superimposed onto 3 orthogonal T1-weighted magnetic resonance imaging sections. The contrast (preRun – postRun) > (preWalk – postWalk), which assesses relative decreases of pain-related activation after treatment, is associated with a modulatory effect of running versus walking in the periaqueductal gray (significant effect, top row), the pregenual anterior cingulate cortex (trend finding, middle row), and the middle insular cortex (trend finding, bottom row). For display purposes, a threshold of P < .01, uncorrected, was chosen. The boxplots show the averaged individual effect sizes [a.u.] across all above-threshold voxels of the given regions. To account for the fact that the experimental and control conditions were conducted on separate days, with a mean interval between examination days of 55 ± 49 days, they were normalized to 1 for each preintervention state.

produce naloxone-reversible analgesia after local morphine administration [30]. Inversely, mu opioid receptor antagonists administered into the PAG produce pronociceptive effects [60]. Human studies have demonstrated enhanced coupling between the pgACC and the PAG as a neural mediator of placebo-analgesia

[5]. This coupling is abolished by naloxone, indicating that the effects are driven by endogenous opioidergic mechanisms [14]. In line with the hypotheses derived thereof, our data demonstrate a significant modulation of pain-related BOLD activity in the PAG, and on a trend level in the pgACC and the insula. As shown in

1712

Ò

L. Scheef et al. / PAIN 153 (2012) 1702–1714

Table 5 Imaging data—interaction effects (pain–heat). Brain region

K

T values

Z scores

X, Y, Z (mm)

(PreRun–postRun) > (preWalk–postWalk) Right midbrain, periaqueductal gray* Left posterior cingulate gyrus, BA 24  Left perigenual anterior cingulate, BA 24  Left middle insular cortex 

85 32 10 16

3.67 3.19 3.05 2.93

3.51 3.07 2.95 2.84

2, 26, 17 12, 10, 32 0, 25, 1 36, 7, 14

(PreRun–postRun) < (preWalk–postWalk) –

–

–

–

–

Indicated are the regions showing an interaction effect for the main contrast (pain–heat). K = cluster size, X, Y, Z = coordinates in Talairach space, BA = brodmann area. * Significant at P < .001, uncorrected.   P = .002, uncorrected (trend).

Fig. 6, there is a consistent difference in pain-related responses in these regions, namely an increase of the BOLD signal at the second pain exposure in the control condition, which is suppressed at the second pain exposure in the exercise condition. Our findings need to be discussed in the context of the published literature, in which acute antinociceptive effects elicited by placebo treatments [5,42] or attentional distraction [56] were shown to be associated with increased subgenual rACC and PAG activity. In our experiment, relative activation increases were observed in these regions during second pain exposure (postWalk), particularly in the PAG. Although involvement of the PAG was found in all 4 conditions, this relative activation increase in the PAG postWalk is thought to reflect an acute upregulation of the descending (antinociceptive) system in response to the second pain exposure on the same day, which accords with reports of enhanced aversiveness of identical pain stimuli delivered, repetitively [55]. Contrary to this PAG upregulation, the state change caused by exercise suppressed this effect. Thus, our data likely indicate a tonic opioidergic component induced by endurance exercise, blunting the reactive upregulation of the PAG observed in the control condition. Exercise of the same duration has been shown to elicit long-lasting endogenous opioidergic effects in the brain, with some of the brain regions (insula, rACC) overlapping with the current study [9]. Also, the rACC findings overlap with observations in placebo analgesia in which a deactivation was reported under placebo conditions that was influenced by naloxone [14,49]. Therefore, we would like to speculate that the protracted endogenous opioidergic effects after exercise lead to a prolonged modulation of pain processing: when the subjective pain experience increases due to repetitive pain exposure, there is an enhanced reactive recruitment of the descending antinociceptive pathway (as shown in the control condition). On the other hand, when the subjective pain experience decreases despite repetitive pain exposure due to a tonic opioidergic modulation (as shown in the postRun condition), the necessity to recruit the endogenous antinociceptive system by additional acute endogenous opioidergic release upon pain stimulation is not required. This is reflected in the differential MPQ affective subscale in running, as compared to walking, as revealed by the post hoc t test (Fig. 4). However, we have to acknowledge that this effect was not identified as significant in the analysis of the VAS pain ratings. Differences between the 2 rating scales might be attributed to the difficulty of volunteers in differentiating between pain intensity and unpleasantness. Although pain intensity and unpleasantness are conceptually different aspects of the subjective pain experience, both items were highly correlated with each other. Although the imaging effects were significant for the PAG, indicating a predominant effect of endurance exercise in reducing activation of the descending antinociceptive system, the trend effects in the mesial affective pain system (insula) further support our a priori hypotheses, if validated in future experiments. It is

noteworthy, however, that no effects were observed in the lateral pain system. Accordingly, the administration of morphine in animals produces a selective attenuation of pain affect, which is represented in the mesial pain system [30]. Likewise, exogenous opioid administration causes increased cerebral blood flow in the ACC and the insula [42,58,63,64], regions belonging to the mesial pain system, where endogenous opioidergic release has been demonstrated during acute experimental pain states [50,54,65], and after exercise [9]. Indirect evidence for opioidergic effects is given (1) by the significant changes of peripheral b-endorphin IRM after endurance exercise, and (2) by the significant treatment effect and the time  treatment interaction for euphoria ratings. The peripheral b-endorphin IRM levels were reproducible across subjects and corroborate previous findings [17,26,61]. As the large molecule size of endorphins hinders a reentry via the blood-brain barrier, it was argued that the hypophyseal excretion into blood does not reflect locoregional opioidergic transmitter effects in the brain [8,51]. This may explain why we did not observe a significant correlation between b-endorphin plasma levels and psychological changes in mood or pain assessments. Nevertheless, the increased plasma bendorphin IRM levels after endurance exercise can be taken as an indirect measure of opioidergic involvement in exercise, carefully suggesting central endorphinergic effects affecting pain and mood processing, which were the major behavioral effects of running, along with postexercise fatigue. Indeed, a significant effect of exercise was observed on euphoria ratings, corroborating our previous findings in which central opioidergic effects were linked to exercise-induced euphoria [9]. However, it must be pointed out that pain fMRI in combination with b-endorphin measurements are necessarily indirect measures for studying opioidergic modulation of brain function. 4.1. Methodological considerations Exercise did not modulate peripheral pain and sensory thresholds. The unchanged pain and sensory thresholds thus strongly argue for underlying central effects. It must be acknowledged that temporary elevations of pain thresholds by exercise have been observed in rats [53] and in humans [21,26]. Presumably, the latency between the end of exercise and the pain threshold assessments was too long (approximately 20 minutes) to pick up such effects in our sample. Also, the type of sport challenge, an individually adapted medium-intensity exercise, may have been too low to induce sustained pain threshold elevations. However, this does not preclude longer lasting psychophysiological effects of exercise, as indicated by PET [9]. Elevated plasma b-endorphin levels have been observed up to 5 hours after exercise [18]. When reasoning why the effects of 2 hours of endurance exercise upon pain appraisal were rather weak in this cohort, timing may be a crucial issue: it is important to consider that the

Ò

L. Scheef et al. / PAIN 153 (2012) 1702–1714

interaction analysis was based on pain ratings that took place 55 minutes after the end of the exercise intervention, i.e., immediately after the 10-minutes-duration pain fMRI experiment. Thus, one may speculate that the antinociceptive effects induced by exercise might have already been reduced at this time point. In our opinion, future studies will have to be conducted to investigate time effects of exercise-induced antinociception in further detail. Another issue for future studies is the interdependency of the duration and/or intensity of exercise challenges, as related to the magnitude and/or duration of potential antinociceptive effects. As discussed for the affective responses to acute exercise regimens [e.g., [16]], this dose-response relationship may be complex and determined by multiple mechanisms. However, it is worth mentioning that a plethora of investigations, as highlighted in a recent review, indicates that psychophysical effects of exercise are enhanced in strenuous exercise conditions [8]. 4.2. Conclusions In conclusion, our complementary behavioral, neuroimaging, and biochemical data indicate that endurance running has a modulatory effect on affective pain processing, counteracting the sensitizing effect of repeated pain exposure after walking. Our imaging findings corroborate the assumption that the brain is in a less pain-sensitive state after endurance exercise and suggest that these effects are mediated via interference with the PAG. Although the demonstration of elevated b-endorphin plasma levels after running are considered as indirect measures, they allow speculating that the tonic antinociceptive effects of exercise are mediated by central opioidergic mechanisms [7,9]. Our current investigation, derived from moderate exertion levels (e.g., aerobic exercise), may provide a neurobiological basis for more concerted use of physical exercise and as a therapeutic option in patients with chronic pain states. Conflict of interest statement None of the authors have a conflict of interest related to the presented work. Acknowledgements We thank Dr. C. Voelker and Prof. V. Gieselmann of the Institute for Biochemistry and Molecular Biology for their technical support in the plasma extraction of b-endorphin, as well as Prof. J. Oldenburg of the Institute for Experimental Hematology and Transfusion Medicine and Prof. K. Schilling of the Institute of Anatomy at the University of Bonn for assistance in the processing of the blood samples. References [1] Aitken RC. Measurement of feelings using visual analogue scales. Proc R Soc Med 1969;62:989–93. [2] An X, Bandler R, Ongur D, Price JL. Prefrontal cortical projections to longitudinal columns in the midbrain periaqueductal gray in macaque monkeys. J Comp Neurol 1998;401:455–79. [3] Bar KJ, Greiner W, Letsch A, Kobele R, Sauer H. Influence of gender and hemispheric lateralization on heat pain perception in major depression. J Psychiatr Res 2003;37:345–53. [4] Basbaum AI, Fields HL. Endogenous pain control mechanisms: review and hypothesis. Ann Neurol 1978;4:451–62. [5] Bingel U, Lorenz J, Schoell E, Weiller C, Buchel C. Mechanisms of placebo analgesia: rACC recruitment of a subcortical antinociceptive network. Pain 2006;120:8–15. [6] Bingel U, Schoell E, Herken W, Buchel C, May A. Habituation to painful stimulation involves the antinociceptive system. Pain 2007;131:21–30. [7] Boecker H, Henriksen G, Sprenger T, Miederer I, Willoch F, Valet M, Berthele A, Tolle TR. Positron emission tomography ligand activation studies in the sports sciences: measuring neurochemistry in vivo. Methods 2008;45: 307–18.

1713

[8] Boecker H, Othman A, Mückter S, Scheef L, Pensel M, Daamen M, Jankowski J, Schild H, Tölle T, Schreckenberger M. Advocating neuroimaging studies of transmitter release in human physical exercise studies. Open Access J Sports Med 2010;1:167–75. [9] Boecker H, Sprenger T, Spilker ME, Henriksen G, Koppenhoefer M, Wagner KJ, Valet M, Berthele A, Tolle TR. The runner’s high: opioidergic mechanisms in the human brain. Cereb Cortex 2008;18:2523–31. [10] Borg G. Borg’s perceived exertion and pain scales. Champaign, IL: Human Kinetics; 1998. [11] Casey KL. Forebrain mechanisms of nociception and pain: analysis through imaging. Proc Natl Acad Sci USA 1999;96:7668–74. [12] Dishman RK, Berthoud HR, Booth FW, Cotman CW, Edgerton VR, Fleshner MR, Gandevia SC, Gomez-Pinilla F, Greenwood BN, Hillman CH, Kramer AF, Levin BE, Moran TH, Russo-Neustadt AA, Salamone JD, Van Hoomissen JD, Wade CE, York DA, Zigmond MJ. Neurobiology of exercise. Obesity (Silver Spring) 2006;14:345–56. [13] Droste C, Greenlee MW, Schreck M, Roskamm H. Experimental pain thresholds and plasma beta-endorphin levels during exercise. Med Sci Sports Exerc 1991;23:334–42. [14] Eippert F, Bingel U, Schoell ED, Yacubian J, Klinger R, Lorenz J, Buchel C. Activation of the opioidergic descending pain control system underlies placebo analgesia. Neuron 2009;63:533–43. [15] Eippert F, Finsterbusch J, Bingel U, Buchel C. Direct evidence for spinal cord involvement in placebo analgesia. Science 2009;326:404. [16] Ekkekakis P, Hall EE, Petruzzello SJ. Variation and homogeneity in affective responses to physical activity of varying intensities: an alternative perspective on dose-response based on evolutionary considerations. J Sports Sci 2005;23:477–500. [17] Farrell PA, Gates WK, Maksud MG, Morgan WP. Increases in plasma betaendorphin/beta-lipotropin immunoreactivity after treadmill running in humans. J Appl Physiol 1982;52:1245–9. [18] Farrell PA, Gustafson AB, Garthwaite TL, Kalkhoff RK, Cowley AW Jr AW, Morgan WP. Influence of endogenous opioids on the response of selected hormones to exercise in humans. J Appl Physiol 1986;61:1051–7. [19] Friston KJ, Holmes AP, Poline JB, Grasby PJ, Williams SC, Frackowiak RS, Turner R. Analysis of fMRI time-series revisited. Neuroimage 1995;2:45–53. [20] Gutstein HB, Mansour A, Watson SJ, Akil H, Fields HL. Mu and kappa opioid receptors in periaqueductal gray and rostral ventromedial medulla. Neuroreport 1998;9:1777–81. [21] Haier RJ, Quaid K, Mills JC. Naloxone alters pain perception after jogging. Psychiatry Res 1981;5:231–2. [22] Hautzinger M, Bailer M, Worall H, Keller F. Beck-Depressions-Inventar (BDI). Bern: Huber; 1994. [23] Heitkamp HC, Schmid K, Scheib K. Beta-endorphin and adrenocorticotropic hormone production during marathon and incremental exercise. Eur J Appl Physiol Occup Physiol 1993;66:269–74. [24] Hillman CH, Erickson KI, Kramer AF. Be smart, exercise your heart: exercise effects on brain and cognition. Nat Rev Neurosci 2008;9:58–65. [25] Hoffman MD, Hoffman DR. Does aerobic exercise improve pain perception and mood? A review of the evidence related to healthy and chronic pain subjects. Curr Pain Headache Rep 2007;11:93–7. [26] Janal MN, Colt EW, Clark WC, Glusman M. Pain sensitivity, mood and plasma endocrine levels in man following long-distance running: effects of naloxone. Pain 1984;19:13–25. [27] Jones AK, Qi LY, Fujirawa T, Luthra SK, Ashburner J, Bloomfield P, Cunningham VJ, Itoh M, Fukuda H, Jones T. In vivo distribution of opioid receptors in man in relation to the cortical projections of the medial and lateral pain systems measured with positron emission tomography. Neurosci Lett 1991;126: 25–8. [28] Kalyuzhny AE, Arvidsson U, Wu W, Wessendorf MW. mu-Opioid and deltaopioid receptors are expressed in brainstem antinociceptive circuits: studies using immunocytochemistry and retrograde tract-tracing. J Neurosci 1996;16:6490–503. [29] Kurth F, Zilles K, Fox PT, Laird AR, Eickhoff SB. A link between the systems: functional differentiation and integration within the human insula revealed by meta-analysis. Brain Struct Funct 2010;214:519–34. [30] LaGraize SC, Borzan J, Peng YB, Fuchs PN. Selective regulation of pain affect following activation of the opioid anterior cingulate cortex system. Exp Neurol 2006;197:22–30. [31] Lockett DM, Campbell JF. The effects of aerobic exercise on migraine. Headache 1992;32:50–4. [32] Löllgen H. Das Anstrengungsempfinden (RPE, Borgskala). Deutsche Zeitschrift für Sportmedizin 2004;55:299–300. [33] Loyd DR, Murphy AZ. The role of the periaqueductal gray in the modulation of pain in males and females: are the anatomy and physiology really that different? Neural Plast 2009;2009:462879. [34] Mantyh PW. Forebrain projections to the periaqueductal gray in the monkey, with observations in the cat and rat. J Comp Neurol 1982;206: 146–58. [35] Matejec R, Ruwoldt R, Bodeker RH, Hempelmann G, Teschemacher H. Release of beta-endorphin immunoreactive material under perioperative conditions into blood or cerebrospinal fluid: significance for postoperative pain? Anesth Analg 2003;96:481–6. [36] Mathes WF, Kanarek RB. Wheel running attenuates the antinociceptive properties of morphine and its metabolite, morphine-6-glucuronide, in rats. Physiol Behav 2001;74:245–51.

1714

Ò

L. Scheef et al. / PAIN 153 (2012) 1702–1714

[37] Mathes WF, Kanarek RB. Chronic running wheel activity attenuates the antinociceptive actions of morphine and morphine-6-glucouronide administration into the periaqueductal gray in rats. Pharmacol Biochem Behav 2006;83:578–84. [38] McNair DM, Lorr M, Doppleman LF. Profile of mood states (Deutsche Bearbeitung von B. Biehl, S. Dangel und A. Reiser). In: Collegium Internationale Psychiatriae Scalarum, editor. Internationale Skalen für Psychiatrie (2 Auflage). Weinheim: Beltz; 1981. [39] Melzack R. The McGill pain questionnaire: major properties and scoring methods. Pain 1975;1:277–99. [40] Oldfield RC. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 1971;9:97–113. [41] Paulev PE, Thorboll JE, Nielsen U, Kruse P, Jordal R, Bach FW, Fenger M, Pokorski M. Opioid involvement in the perception of pain due to endurance exercise in trained man. Jpn J Physiol 1989;39:67–74. [42] Petrovic P, Kalso E, Petersson KM, Ingvar M. Placebo and opioid analgesia— imaging a shared neuronal network. Science 2002;295:1737–40. [43] Rahkila P, Hakala E, Alen M, Salminen K, Laatikainen T. Beta-endorphin and corticotropin release is dependent on a threshold intensity of running exercise in male endurance athletes. Life Sci 1988;43:551–8. [44] Reed J, Ones DS. The effect of acute aerobic exercise on positive activated affect: s meta-analysis. Psychol Sport Exerc 2006;7:477–514. [45] Sandkuhler J. The organization and function of endogenous antinociceptive systems. Prog Neurobiol 1996;50:49–81. [46] Scavo D, Barletta C, Vagiri D, Letizia C. Adrenocorticotropic hormone, betaendorphin, cortisol, growth hormone and prolactin circulating levels in nineteen athletes before and after half-marathon and marathon. J Sports Med Phys Fitness 1991;31:401–6. [47] Schmidt K-H, Metzler P. Wortschatztest (WST). Weinheim; 1992. [48] Schnitzler A, Ploner M. Neurophysiology and functional neuroanatomy of pain perception. J Clin Neurophysiol 2000;17:592–603. [49] Schoell ED, Bingel U, Eippert F, Yacubian J, Christiansen K, Andresen H, May A, Buechel C. The effect of opioid receptor blockade on the neural processing of thermal stimuli. PLoS One 2010;5:e12344. [50] Scott DJ, Stohler CS, Egnatuk CM, Wang H, Koeppe RA, Zubieta JK. Placebo and nocebo effects are defined by opposite opioid and dopaminergic responses. Arch Gen Psychiatry 2008;65:220–31. [51] Sforzo GA, Seeger TF, Pert CB, Pert A, Dotson CO. In vivo opioid receptor occupation in the rat brain following exercise. Med Sci Sports Exerc 1986;18:380–4. [52] Sheps DS, Adams KF, Hinderliter A, Price C, Bissette J, Orlando G, Margolis B, Koch G. Endorphins are related to pain perception in coronary artery disease. Am J Cardiol 1987;59:523–7.

[53] Shyu BC, Andersson SA, Thoren P. Endorphin mediated increase in pain threshold induced by long-lasting exercise in rats. Life Sci 1982;30:833–40. [54] Sprenger T, Valet M, Boecker H, Henriksen G, Spilker ME, Willoch F, Wagner KJ, Wester HJ, Tolle TR. Opioidergic activation in the medial pain system after heat pain. Pain 2006;122:63–7. [55] Valet M, Sprenger T, Boecker H, Rummeny E, Tölle TR. Repetitive pain exposure: neuronal correlates in the human brain. In: H Flor, E Kalso, JO Dostrovsky, editors. Proceedings of the 11th world congress on pain. Seattle: IASP Press; 2006. p. 431–8. [56] Valet M, Sprenger T, Boecker H, Willoch F, Rummeny E, Conrad B, Erhard P, Tolle TR. Distraction modulates connectivity of the cingulo-frontal cortex and the midbrain during pain—an fMRI analysis. Pain 2004;109:399–408. [57] Wagner G, Koschke M, Leuf T, Schlosser R, Bar KJ. Reduced heat pain thresholds after sad-mood induction are associated with changes in thalamic activity. Neuropsychologia 2009;47:980–7. [58] Wagner KJ, Willoch F, Kochs EF, Siessmeier T, Tolle TR, Schwaiger M, Bartenstein P. Dose-dependent regional cerebral blood flow changes during remifentanil infusion in humans: a positron emission tomography study. Anesthesiology 2001;94:732–9. [59] Wang H, Wessendorf MW. Mu- and delta-opioid receptor mRNAs are expressed in periaqueductal gray neurons projecting to the rostral ventromedial medulla. Neuroscience 2002;109:619–34. [60] Wilcox RE, Mikula JA, Levitt RA. Periaqueductal gray naloxone microinjections in morphine-dependent rats: hyperalgesia without ‘‘classical’’ withdrawal. Neuropharmacology 1979;18:639–41. [61] Wildmann J, Kruger A, Schmole M, Niemann J, Matthaei H. Increase of circulating beta-endorphin-like immunoreactivity correlates with the change in feeling of pleasantness after running. Life Sci 1986;38:997–1003. [62] Willoch F, Tolle TR, Wester HJ, Munz F, Petzold A, Schwaiger M, Conrad B, Bartenstein P. Central pain after pontine infarction is associated with changes in opioid receptor binding: a PET study with 11C-diprenorphine. AJNR Am J Neuroradiol 1999;20:686–90. [63] Wise RG, Rogers R, Painter D, Bantick S, Ploghaus A, Williams P, Rapeport G, Tracey I. Combining fMRI with a pharmacokinetic model to determine which brain areas activated by painful stimulation are specifically modulated by remifentanil. Neuroimage 2002;16:999–1014. [64] Wise RG, Williams P, Tracey I. Using fMRI to quantify the time dependence of remifentanil analgesia in the human brain. Neuropsychopharmacology 2004;29:626–35. [65] Zubieta JK, Bueller JA, Jackson LR, Scott DJ, Xu Y, Koeppe RA, Nichols TE, Stohler CS. Placebo effects mediated by endogenous opioid activity on mu-opioid receptors. J Neurosci 2005;25:7754–62.