Journal of Applied Physiology

Neuroplastic adaptations to exercise: neuronal remodeling in cardiorespiratory and locomotor areas

Amanda J. Nelson, Janice M. Juraska, Timothy I. Musch, Gary A. Iwamoto


Neuronal activity has been shown to be attenuated in cardiorespiratory and locomotor centers of the brain in response to a single bout of exercise in trained (TR) vs. untrained (UN) animals, but the mechanisms remain obscure. Based on this finding, dendritic branching patterns of seven brain areas associated with cardiorespiratory and locomotor activity were examined in TR and UN animals. Twenty-eight male Sprague-Dawley rats were kept in individual cages and divided into TR and UN. TR were provided with a running wheel and exercised spontaneously. After 85 or 120 days, exercise training indexes were obtained, including maximal oxygen consumption, percent body fat, resting heart rate, and heart weight-to-body weight ratios. The brain was removed and processed according to a modified Golgi-Cox procedure. Impregnated neurons from seven brain areas were examined in coronal sections: the periaqueductal gray, posterior hypothalamic area, nucleus of the tractus solitarius, rostral ventrolateral medulla, cuneiform nucleus, nucleus cuneatus, and cerebral cortex. Neurons were traced using a camera lucida technique and analyzed using the Sholl analysis of dendritic branching. t-Tests were conducted to compare the mean number of intersections per neuron by grouping inner rings and outer rings and also comparing the total number of intersections per animal. There were significant differences between groups in the posterior hypothalamic area, periaqueductal gray, cuneiform nucleus, and nucleus of the tractus solitarius in the inner rings, outer rings, and the total number of intersections per animal. Our results show that dendritic fields of neurons in important cardiorespiratory and locomotor centers of the brain are attenuated in TR animals.

  • Golgi staining
  • dendritic branching
  • exercise training indexes
  • locomotion

although a single bout of exercise can induce a variety of physiological responses, chronic exercise stimuli cause long-term adaptations, including structural remodeling and biochemical adjustments at many levels. This includes increased capacity to oxidize carbohydrates and fats (42), as well as increases in mitochondrial and capillary numbers (71) in skeletal muscle, decreased blood lactate levels (27), and profound changes in the myocardium (71). Cardiorespiratory function is similarly affected. It is well known that exercise training induces resting bradycardia (24), which may be attributed to increased parasympathetic activity, decreased sympathetic activity, and possibly a reduction in the intrinsic heart rate (HR) (10, 15, 16, 36, 50). In addition, lower blood pressure (66) and changes in baroreceptor reflex function (16, 50) occur. Many of these changes are held to be beneficial in humans. The net effects of exercise training are often expressed as changes in aerobic capacity.

Data collected from animal models suggest exercise training has a direct impact on many other types of neurological functioning. Voluntary wheel running has been shown to increase levels of brain-derived neurotrophic factor (62), enhance neurogenesis (78), increase the number of synapses (56), and increase long-term potentiation (30) in various areas of the central nervous system (CNS).

These studies of chronically induced autonomic adjustments and general brain function demonstrate that the nervous system is adaptive and plastic in response to exercise training. A part of this may be related to structural neuroplasticity, which has been shown in response to other types of chronic stimuli. McEwen and colleagues (57) demonstrated that chronic (21 days), but not acute, restraint stress caused attenuation of apical dendrites in pyramidal neurons of the hippocampus. Structural neuroplastic change due to chronic stimulation by the environment is a foundation of the learning and memory literature (38). Greenough and colleagues (39, 80) have extensively studied the effects of enriched or complex vs. impoverished environments on structural plasticity in the cerebral cortex (CTX) and cerebellum, finding enriched environments led to dendritic proliferation. Given that chronic stimulation leads to extensive changes in structure and function of other tissues (e.g., skeletal and cardiac muscle) and given the involvement of the CNS in controlling cardiorespiratory and locomotor function, it seemed probable that exercise training might lead to similar neuroplastic changes in the cardiorespiratory and locomotor centers (CRLC) of the brain. A previous report from our laboratory (43) demonstrated a decrease in neuronal activity, measured by Fos activity, in CRLC of the CNS in response to a single bout of exercise with a fixed, absolute workload in trained (TR) rats. Although this might represent an intrinsic change in the CNS, it was also possible that the Fos expression changes were simply a result of well-known peripheral adaptations whether in the efficiency of muscle contraction or in mechanically or metabolically evoked afferent activity. Because exercise-induced changes were known to occur in the CNS, it was possible that intrinsic structural changes occurred within the CRLC due to exercise training. In the present study, we used a Golgi-Cox staining procedure to examine the dendritic branching pattern of six areas of the brain associated with cardiorespiratory-locomotor activity and a single area of the CTX in TR and untrained (UN) animals. We hypothesized that moderate levels of exercise training would be sufficient to induce structural neuroplastic changes within the CRLC of the brain. Based on the work of Greenough and colleagues (37–39, 80) with systems involving increased activity, these changes were likely to be in the direction of dendritic proliferation.


All animal use was approved by the University of Illinois' Institutional Animal Care and Use Committee under protocol no. 03151.

Animal preparation.

Forty-five male Sprague-Dawley rats (Sasco, Charles River) were kept in individual cages and randomly divided into two groups: TR (n = 24) and UN (n = 21), beginning as 21-day-old weanlings. For the present study, TR (n = 14) and UN (n = 16) were selected at random from these groups and evaluated for indexes of exercise training and structural neuroplastic changes. The remaining animals from the original group (15 of 45) were used in another study. Males were selected for this investigation because they demonstrate body composition changes similar to humans (23). Animals were maintained in a temperature-controlled environment, fed ad libitum, and kept on a 12-h light-dark cycle. TR rats were provided with a running wheel of 106 cm in circumference (Nalgene) that was placed inside their cage. All rats were allowed to exercise spontaneously. The total number of wheel rotations was recorded daily for the duration of the experiment. The total distance run by each animal was then calculated from the total number of wheel rotations. The UN rats were not provided with a running wheel.

Treadmill familiarization.

After 85 or 120 days of exercise training or control conditions and 4–6 days before final aerobic testing, all animals were given two familiarization trials on the treadmill apparatus (12) to adapt them to the testing environment. Each familiarization trial lasted 7 min, and the trials were conducted on nonconsecutive days. The speed during the first familiarization trial increased progressively from 10 to 15 m/min throughout the 7-min duration. The speed during the second familiarization trial increased progressively from 15 to 20 m/min.

Maximal oxygen uptake testing protocol.

Maximal oxygen uptake (V̇o2 max) was determined for all rats according to previously established methods (40). This method uses a metabolic chamber designed to fit into a stall of a four-lane rodent treadmill and utilizes the techniques described by Brooks and White (12) for determining oxygen uptake (V̇o2) and carbon dioxide production.

o2 max was determined by having each rat perform a maximal exercise test. Briefly, this test (40) began with a 3-min warm-up at a treadmill grade and speed of 0% and 15 m/min, respectively. The treadmill speed and/or grade was increased every 3 min. V̇o2 max was defined as the point at which the V̇o2 did not increase with further increases in workload or when the rat was unable or unwilling to continue running. Confirmation that V̇o2 max was truly attained in each animal was demonstrated by having each rat perform a subsequent maximal exercise test after 48 h of recovery from the initial test. With the second test, each rat was given a 3-min warm-up at a treadmill grade and speed of 0% and 15 m/min. The treadmill grade and speed were then increased to the highest workload each animal was able to sustain during the initial maximal test. V̇o2 and carbon dioxide production were recorded. The treadmill speed was then increased by 3–5 m/min, and V̇o2 and carbon dioxide production were recorded. If the measured V̇o2 was similar between the two workloads, the animal was considered to be at V̇o2 max, and the exercise test was terminated. If the rat demonstrated an increase in V̇o2 during the second exercise test, the test was terminated, and the same procedure was repeated after 48 h of recovery.

No electrical stimulation was used to induce exercise. This was partly because we did not wish to induce stress, which has been shown to cause neuroplastic changes (58). When necessary, a nudge with a piece of a common windshield snow-brush (connected to an external lever) was applied to induce exercise (animals wished to avoid the texture). After each V̇o2 max test, the chamber was flushed and recalibrated for the next test.

Gross anatomical and physiological adaptations to wheel running.

Two days after the second V̇o2 max test, the rats were sedated briefly by halothane inhalation. Immediately after the animals became manageable, they were anesthetized with a combination of α-chloralose (65 mg/kg) and urethane (800 mg/kg) administered intraperitoneally. The resting HR of each animal was measured using the amplified EKG signal and processed using a Gould Biotach. Each animal was weighed. Body composition analysis using dual-energy X-ray absorptiometry (Hologic 4500A) was performed on each animal. All animals were killed with an overdose of sodium pentobarbital. Their hearts were removed and weighed. The entire brain was removed and used for analysis of neuroplastic adaptations.

Tissue preparation.

After data regarding training effects was acquired, as previously described, the entire brain was removed and immersed in Golgi-Cox impregnation solution (32, 34). The brains were stored individually in glass jars and kept in a light-safe box. Forty-eight hours later, the Golgi-Cox solution was renewed and the brains were returned to the light-safe box for 2 wk.

The tissue was frozen, and 200-μm coronal sections of the entire brain were obtained from a sliding microtome (68). The coronal sections were mounted on 2% gelatin-coated slides and pressed to the slides using moistened bibulous paper. Each slide was kept in a humidity chamber for a minimum of 3 h.

Slides were placed in glass staining dishes and processed according to Gibb and Kolb's (32) Golgi-Cox procedure. Briefly, the slides were placed through distilled H2O (dH2O; 1 × 1 min); ammonium hydroxide in the dark (1 × 30 min); dH2O (1 × 1 min); Kodak Fix diluted 1:1 with dH2O (1 × 30 min); dH2O (1 × 1 min); 70% ethanol (1 × 1 min); 95% ethanol (1 × 1 min); 100% ethanol (2 × 5 min); HemoDe (1 × 15 min). After dehydration, slides were coverslipped using Eukitt mounting medium (Calibrated Instruments).

Visualization of neurons.

Not all animals processed for Golgi staining yielded useable tissue in every area of interest because of inadvertent mechanical damage either during dissection or tissue sectioning. Therefore, tissue from 14 animals (UN = 7, TR = 7) was used to evaluate each area. We examined Golgi-impregnated neurons from seven areas of the brain including periaqueductal gray (PAG), posterior hypothalamic area (PH), nucleus of the tractus solitarius (NTS), rostral ventrolateral medulla (RVL), cuneiform nucleus (CfN), nucleus cuneatus (Cu), and the frontal area I/hindlimb area of the CTX. The areas were selected from sections containing these regions. The sections included +7.20, +4.70, +0.28, −3.30, and −4.30 in reference to the interaural line, according to Paxinos and Watson (65). Figure 1 illustrates the CRLC examined in this study. Each area was further delineated by superimposing a 5 × 5 rectangular grid over known anatomical landmarks using an eyepiece reticle. The photomicrographs shown in this report were obtained using a multiplane technique. This involved superimposing three or four photomicrographs per neuron, each taken from a different plane of focus. This allowed for a more comprehensive depiction of the dendritic arborization but is still only an approximation, which does not reveal the resolution attained during microscopic examination.

Fig. 1.

Localization of cardiorespiratory and locomotor centers (CRLC) [periaqueductal gray (PAG), posterior hypothalamic area (PH), nucleus of the tractus solitarius (NTS), rostral ventrolateral medulla (RVL), cuneiform nucleus (CfN), nucleus cuneatus (Cu)] and control [cerebral cortex (CTX)] areas investigated. Figures, including abbreviations, are adapted from Paxinos and Watson (65).

Selected sections of the neuraxis were analyzed quantitatively. In this study, 20 Golgi-Cox stained neurons from each of the seven areas were traced at ×400 total magnification using a Reichert (American Optical) microscope equipped with a drawing tube. The neurons traced were chosen by an independent investigator blind to the experimental conditions to minimize any bias. Neurons were only used in these analyses if the whole neuron and its processes were visible and traceable without intervening obstructions.

Localization and rationale of CRLC and control areas investigated.

The cardiovascular controlling network within the PAG is organized in columns (3). The dorsal column of the PAG is involved in pressor, and the ventrolateral column mediates depressor responses (55). According to Behbehani (8), stimulation of the dorsomedial and dorsolateral columns produces an increase in blood pressure and causes defensive behavior in rats. The dorsomedial column extends through the dorsal PAG. The dorsolateral column originates in the caudal PAG at the level of the dorsal raphe and extends to the rostral PAG at the level of Edinger Westphal nuclei (8). We chose a 1.00 × 1.50 mm sampling area in the caudal-most PAG within the dorsomedial and dorsolateral columns for analysis located on section +0.28 of the Paxinos and Watson rat brain atlas (65).

The PH has been identified as a pressor and locomotor region and is located within the anatomical boundaries of the fornix, mamillothalamic tract, and the third ventricle (70, 79, 82). Electrical stimulation of this region of the brain has been shown to elevate sympathetic nerve activity and cardiovascular responses (5, 28, 31, 81). The PH also contains the hypothalamic locomotor region, which evokes coordinated locomotion when electrically stimulated or exposed to GABA antagonists (28, 81). Paxinos and Watson's (65) depiction of this area includes the posterior hypothalamic nucleus. Section +4.70 of their rat brain atlas was chosen for the purposes of this study. A single 1.50 × 1.50 mm sampling area was superimposed over the PH and is illustrated in Fig. 1.

The NTS plays a vital role in coordinating autonomic function by relaying visceral afferent input, which controls HR and blood pressure (4, 67). Significantly, hindlimb somatic input, including that of muscle afferents, has a major contribution to this area (76). The NTS is located in the dorsal medial medulla and can be identified on section −4.30 of Fig. 1, which is adapted from Paxinos and Watson (65). A single 0.50 × 3.00 mm sampling area was superimposed bilaterally over the medial NTS for sampling.

The anatomical RVL is located within the reticular formation and can be identified on section −3.30 of Fig. 1. Bilateral 1.00 × 0.50 mm sampling areas were used for RVL cell sampling. The anatomical RVL is a major part of the principal pathway from the medulla to the intermediallateral cell column, receives projections from the NTS and/or projects to other autonomic areas (18), and mediates the somatic pressor reflex (74).

The CfN is a readily discerned part of the mesencephalic locomotor region (73) and is located on section +0.28 of Fig. 1. A 0.50 × 1.00 mm sampling area was placed bilaterally over the area of interest for cell selection.

The Cu is located on section −4.30, as shown in Fig. 1. Bilateral 0.50 × 1.00 mm sampling areas were superimposed over the Cu for sampling. The Cu composes part of the dorsal column nuclei where afferent fibers relay in the dorsal column lemniscal sensory system (84).

Pyramidal cells from layers II and III in the hindlimb area and adjacent frontal area 1 of the CTX, which could be compared with previous studies (46, 85), were also examined to ensure the neurons were properly stained. Apical and basilar branches were analyzed together for this study. The CTX area is located on section +7.20 of Fig. 1. Bilateral 1.50 × 1.50 mm sampling areas were placed over the CTX area of interest for sampling.

Analysis of neurons.

For the purposes of analysis, the data from both 85- and 120-day animals were considered to be a single group of TR and a single group of UN. This is further justified by the finding that the number of dendritic branches did not differ between the 85-or 120-day groups, except as will be described between TR and UN. After the neurons were traced, the maximum and minimum diameters of each soma was recorded. The average of these measures was taken as the average soma “diameter” after the method of Burke et al. (14). Each neuron was then analyzed using the Sholl analysis (72) of dendritic branching, which assumes that dendritic arborization is an indirect measure of available postsynaptic space. As shown in Fig. 2, a series of concentric rings calibrated at 20-μm intervals were superimposed on each neuron and centered on the cell body. Intersections between dendrites and each concentric ring were then counted. The location and number of intersections were then plotted, as shown in Fig. 6, and also used for statistical comparisons.

Fig. 2.

Sholl analysis of dendritic branching. Concentric rings superimposed on a multiplane photomicrograph of a neuron from the PAG in an untrained (UN) animal. Concentric rings are calibrated at intervals of 20 μm. The number of dendritic intersections within each concentric ring are determined and used as the basis for constructing Figs. 610.

Data analysis.

All data are reported as means ± SE. Individual means were compared between groups using a Student's t-test for each of the training indexes under investigation. For a group difference to be statistically significant, P ≤ 0.05 had to be achieved. When evaluating the training effect, the one-tailed approach was used if the training adaptation had been previously documented (71).

Although descriptive comparisons were made between TR and UN at each concentric ring, t-tests were conducted for comparisons of the mean number of intersections per neuron between TR and UN by grouping rings 1–6 (inner rings) and rings ≥7 (outer rings) (44). The inner rings encompass 1–120 μm, and the outer rings encompass ≥ 121 μm. t-Tests were also performed for comparisons of the total number of intersections per animal between TR and UN in each CRLC investigated, which is representative of the dendritic field. The average diameter of each soma was determined, and t-tests were performed between TR and UN to compare soma size. All analyses were conducted using the GBSTAT software package.


Wheel running.

On average, TR rats (n = 14) ran 7,225 ± 735 m/wk. A peak in the total amount of running performed per week occurred at week 8 of the conditioning and steadily decreased before reaching a plateau at week 14 (Fig. 3). The 120-day TR animals (n = 7) ran fewer meters per week (6,115 ± 439 m/wk) than the 85-day TR animals (n = 7) (9,486 ± 1,319 m/wk). There were, however, no differences between the 85-day TR and 120-day TR animals in V̇o2 max, HR, heart weight-to-body weight ratio, and percent fat-free mass (%FFM). Based on this information, we felt justified in pooling the data from these two groups.

Fig. 3.

Average weekly distance ran by the trained animals (TR) throughout the duration of the experiment. Seven animals ran for 18 wk, and 7 additional animals ran for 11 wk. The mean ± SE from week 1 to week 11 includes all 14 animals.

Absolute body and heart mass data.

Voluntary wheel running resulted in a significant difference in both absolute body mass and absolute heart weight between TR and UN animals (Table 1). Mean body weight differed significantly between the 85-day group (376 ± 10 g) and 120-day group (425 ± 10 g). Heart weight-to-body weight ratio is often used to assess training effects in animal models (1, 17) and was also significantly different between TR and UN in the present study, regardless of the 85-day (TR: 0.39 ± 0.02, n = 7; UN: 0.31 ± 0.01, n = 6) or 120-day (TR: 0.41 ± 0.01, n = 7; UN: 0.32 ± 0.01, n = 10) period (Table 1). A heart weight-to-FFM ratio was also calculated for each animal and is included in Table 1.

View this table:
Table 1.

Gross anatomical and physiological adaptations to spontaneous running

Body composition.

Body weight ranged from 322 to 488 g in the present investigation, which was well above weight minimums (>200 g) needed for dual-energy X-ray absorptiometry scanning to be feasible (9). As shown in Fig. 4A, TR rats had a significantly higher %FFM than UN rats (TR: 95 ± 1%; UN: 92 ± 1%). Despite this overall difference, there was no significant difference in %FFM between TR (94 ± 1%, n = 7) and UN (95 ± 1%, n = 6) in the 85-day group. There was a significant difference in %FFM between TR (95 ± 1%, n = 7) and UN (92 ± 1%, n = 10) in the 120-day group with the TR having a greater %FFM.

Fig. 4.

A: comparison of percentage of fat-free mass between TR and UN. B: comparison of the maximal oxygen consumption between TR and UN. **Significance at P < 0.01.

HR training effect.

After being anesthetized, the resting HR of each animal was obtained using an EKG. Spontaneous wheel running resulted in a significantly decreased resting HR in TR compared with UN rats (Table 1). There were also significant differences in resting HR when comparing TR and UN rats in both the 85-day (TR: 364 ± 10, n = 7; UN: 332 ± 8, n = 6) or 120-day period (TR: 351 ± 13, n = 7; UN: 309 ± 8, n = 10).

o2 max.

o2 max values were obtained at average speeds of 48 ± 2 m/min for TR and at 28 ± 1 m/min for UN rats. Figure 4B shows that V̇o2 max was significantly higher for TR (70 ± 4 ml·kg−1·min−1) compared with UN (49 ± 2 ml·kg−1·min−1). When taking FFM into consideration, V̇o2 max is significantly higher in TR (75 ± 5 ml·kg−1·min−1) compared with UN (53 ± 2 ml·kg−1·min−1). However, there was no significant difference between using FFM and absolute body mass as the standard for body mass.

Analysis of neurons.

Photomicrographs of representative examples of Golgi-Cox staining in TR and UN animals from PAG and PH appear in Fig. 5. As described in methods, these cells were drawn using camera lucida and analyzed using the concentric ring or Sholl analysis. These data are depicted in Figs. 6 and 7, which show the mean number of intersections per neuron between each ring in a series of concentric rings of all seven anatomical areas studied. Note that the same general trend was illustrated for each of the CRLC investigated in that the mean number of intersections per neuron decreased as the distance from the cell body increased. Although it may be clear from simple inspection of Fig. 5, when analyzed, UN had a higher number of intersections per neuron than TR at each concentric ring in many of the CRLC.

Fig. 5.

Examples of multiplane photomicrographs of representative neurons from the PAG and PH in TR and UN animals. A: PAG neuron from TR animal. B: PAG neuron from UN animal. C: PH neuron from TR animal. D: PH neuron from UN animal.

Fig. 6.

Sholl analysis of the number of Golgi-stained dendritic intersections per concentric ring in the NTS (A), PH (B), PAG (C), and Cu (D).

Fig. 7.

Sholl analysis of the number of Golgi-stained dendritic intersections per concentric ring in the RVL (A), Cu (B), and CTX (C) (frontal area I and hindlimb area).

Statistical comparisons were made on data incorporating populations of concentric rings. Figure 8 illustrates that significant differences were found between groups in the PH, PAG, CfN, and NTS when comparing the inner rings. There were also significant differences between groups in the PH, PAG, CfN, and NTS when comparing the outer rings. Figure 9 shows that no significant differences of the inner or outer rings were found between groups in the CTX, RVL, and Cu. Figure 10 shows the total number of intersections per animal and represents the total dendritic field. There were significant differences between TR and UN in the PH, PAG, CfN, and NTS. The results of the total number of intersections per animal did not change when the 85- or 120-day TR animals were compared with their matched UN controls, despite one exception. There was no significant difference in the total dendritic field of the PAG when only the 85-day TR and UN animals were compared.

Fig. 8.

Comparison of the mean number of intersections per neuron between TR and UN by grouping rings 1–6 (inner rings) and rings ≥7 (outer rings) in the NTS (A), PH (B), PAG (C), and Cu (D). *Significance at P < 0.05. **Significance at P < 0.01.

Fig. 9.

Comparison of the mean number of intersections per neuron between TR and UN by grouping rings 1–6 (inner rings) and rings ≥7 (outer rings) in the RVL (A), Cu (B), and CTX (C) (frontal area I and hindlimb area). There were no significant differences.

Fig. 10.

Comparison of the total number of intersections per animal between TR and UN of all CRLC and control areas investigated. *Significance at P < 0.05. **Significance at P < 0.01.

Soma size.

The average diameter of cell bodies was significantly different between groups in the PH and CfN, as shown in Table 2. No significant differences were found in the CTX, PAG, RVL, Cu, and NTS.

View this table:
Table 2.

Average diameter of 20 cell bodies per animal from each of the cardiorespiratory and locomotor centers and control areas examined


In this study, we demonstrate the first evidence of structural neuroplasticity in the CRLC of the brain accompanying the effects of exercise training.

Efficacy of training.

It should be noted that the rats in this report were more active than those in our laboratory's previous studies (43). The reason for this can, most likely, be attributed to the difference in wheel circumference. The wheel circumference in Ichiyama et al. (43) was considerably smaller (83 cm) and with fewer cross members for the rats to catch during the exercise task.

Although rats can be conditioned to run for long durations and high intensities on a treadmill, there are some inherent disadvantages. This method of training is labor intensive (52). Furthermore, 10–15% of rats will typically refuse to run on a treadmill (51, 69). We intended to characterize a training environment that mimics the variable daily energy expenditure in human situations, which can be best represented with the opportunity for voluntary physical activity. In addition, we intended to characterize the general population of Sprague-Dawley rats. Thus we did not restrict our investigations to highly active runners whether by strain (e.g., Long-Evans hooded) (48, 53), selection (33, 75), or breeding (11).

Thus, although treadmill exercise has been used by several investigators as a training method for rats (2, 13, 25, 40, 59, 61), voluntary exercise has also been shown to produce training adaptations (16, 52, 75, 86). Chen et al. (17) showed a 20% increase in heart weight-to-body weight ratio and a 7% decrease in resting HR with 8–13 wk of spontaneous running. These cardiovascular trends are consistent with the present study. In addition to these training adaptations, DiCarlo and colleagues (16, 17) have identified a number of cardiorespiratory parameters that are altered with wheel running (e.g., reduction in arterial baroreflex function). Changes in resting HR are also used to assess the effects of exercise training. After 8–9 wk of daily spontaneous running, resting HR decreased 21% compared with controls in male Sprague-Dawley rats (16). Although our HR results were not as profound, the same pattern existed. The smaller adaptation may be due to the method of analysis. In the present study, EKGs were obtained while the animal was under an anesthesic, albeit one that is held to have minimal cardiorespiratory effects. In Chen and DiCarlo's study (16), the animals were fully conscious during HR data collection. Based on previous studies and the results from the present investigation, spontaneous wheel running is efficacious in producing a moderate training effect (7, 52).

Exercise-induced dendritic remodeling.

Although CRLC have been shown to change their activity in response to a fixed, absolute exercise task with exercise training (43) and functional indexes, such as the baroreflex (16), are altered, our results suggest that intrinsic changes in the CNS occur with exercise training. Although it remains possible that structural changes in these neurons may not be directly related to actual changes in cardiorespiratory and locomotor function, the known roles of the CRLC of the brain in these processes provide a strong inference that these structural modifications may be of great significance explaining the effects of exercise training.

The present data demonstrate that neurons in the CRLC of the brain are profoundly remodeled with exercise training. The total dendritic fields in the PH, PAG, CfN, and NTS were much smaller for TR compared with UN animals. Although cell body size may tend to correspond with changes in the dendritic field, the cell body size apparently only decreased in the PH and CfN. Interestingly, it is in these areas that the strongest case may be made for combined cardiorespiratory and locomotor function (81).

The only area with cardiorespiratory significance which did not show readily demonstrable changes was RVL. This is consistent with our laboratory's previous observation (43) that RVL is less likely to show activity changes with training. Our previous study using Fos-expression showed that changes in RVL, although statistically significant, were less pronounced than in other areas of the brain with cardiorespiratory function. This is perhaps not surprising, since this area may subserve a multiplicity of functions in addition to sympathoexcitation (18). In addition, since RVL neurons are continuously involved in tonic maintenance of cardiorespiratory functions, such as resting blood pressure (18), it might be possible that they have comparatively stable behavior. Upstream areas may be more involved in integration with other systems/behavior and more subject to plastic changes.

Because body weight is significantly lower in TR vs. UN and %FFM is significantly higher in TR vs. UN, we cannot rule out that some of the effects may be due to exercised-induced anorexia (45). This process appears to be a normal occurrence with voluntary wheel running (45).

There is ample evidence that dendrites and synapses undergo morphological changes in response to altered, chronic patterns of synaptic activity (22, 35, 60, 80). Numerous studies have investigated the experience-driven paradigm of morphological plasticity (21, 29, 37, 38, 60, 83). Among the more noteworthy examples, animals reared in enriched environments showed increased dendritic branching in the occipital cortex (38) and altered forepaw representation in the primary somatosensory cortex (22). Furthermore, exposure to a complex environment leads to increased synaptic size (83) and an increased number of synapses per neuron (77) in the visual cortex. As has been discussed, complex motor tasks also lead to an increase in the number of synapses in the motor cortex (47). Experience-driven neuroplasticity does not occur in an all-inclusive manner. Different areas of the brain are involved in the processing of specific components of experience (39). Conversely, attenuation of hippocampal dendrites has been observed in connection with stress and is believed to be linked to glucocorticoid hormones working in concert with excitatory amino acids and N-methyl-d-aspartate (58).

In addition, activity-induced neuroplasticity appears to be regional as well as task specific. For example, training-induced neuroplasticity did not occur in the cerebellar cortex with simple exercise but did occur with complex motor-learning tasks (48). However, van Praag and colleagues (78) used bromodeoxyuridine-labeling to demonstrate that voluntary exercise, but not an enriched environment, doubled the number of surviving newborn cells in the dentate gyrus. In the present study, dendrites showed neuroplastic changes in the CRLC of the brain with simple exercise training.

In contrast to these substantial structural changes in the CRLC with exercise training, the layer II and III pyramidal cells of the CTX in the present study did not show significant changes. This result is consistent with the findings of Kleim et al. (46) who studied the same cell population in animals similarly exposed to wheel running, with respect to synaptogenesis and also the topography of movement representations within the motor cortex. These investigators stated, “Simple increases in motor activity, in the absence of motor skill learning, does not alter neuron morphology.”

Resting bradycardia is a common result of exercise training in normotensive subjects (10) as is reduced blood pressure in hypertensive subjects (66), both of which may be tied to a reduction in sympathetic activity (19, 50). There seems to be some inconsistency in studies of resting norepinephrine levels and/or spillover rates with exercise training (63). Although no training-associated differences have been observed for directly measured muscle sympathetic nerve activity (63), at least one study has shown a correlation between reduced resting renal sympathetic activity and exercise training (50). There is also a reduced sympathetic component of baroreflex function with chronic exercise training (19). Daily exercise has also been shown to enhance the influence of cardiac afferents in normal rabbits (24) and attenuate the sympathetic nerve response to exercise in spontaneously hypertensive rats (20).

Exercise training is also known to ameliorate conditions associated with cardiovascular disease. The role of exercise as a therapy for prevention, treatment, and control of hypertension is well known (66). Exercise training has been shown to blunt the rise in resting arterial pressure and improve cardiovascular regulation in the spontaneously hypertensive rat (41, 64). Furthermore, Beatty et al. showed that spontaneous neuronal activity is decreased in the PH of exercise-trained spontaneously hypertensive rats (6). This may be due to an enhancement of GABAergic inhibition. Chronic exercise leads to upregulation of glutamic acid decarboxylase gene transcription in the caudal hypothalamus of spontaneously hypertensive rats (54), which may increase tonic GABAergic inhibition of cardiovascular outflow (49). Exercise is also often part of the treatment of human heart failure. Zucker and colleagues used a chronic heart failure animal model to demonstrate that exercise training decreases resting sympathetic outflow and normalizes baroreflex and other abnormal cardiovascular reflex sensitivity found in this condition (87).

We found that chronic exercise resulted in attenuated dendritic arborizations in the CRLC of the brain. Our data may support the notion that reduced synaptic activity correlates with structural neuroplasticity. Certainly, our laboratory's previous study using Fos activation (43) is consistent with this idea of a lowered level of CRLC excitation for a given absolute workload. If one assumes that most areas we have studied are reflective of potential sympathoexcitation, then it would not be unreasonable to speculate that a reduction of sympathetic activity might occur with a loss of excitatory synapses. It has long been held that excitatory synapses are located on the distal dendrites, whereas the inhibitory synapses are located on the soma and proximal dendrites (26). It is also possible that with this attenuation the locomotor system is simply preserving the most critical parts related to function in the presence of an enhanced locomotor periphery.

In conclusion, neurons in the critical areas of the CNS responsible for cardiorespiratory and locomotor function demonstrate a profound attenuation of their dendritic fields with exercise training. Because it is likely that this represents changes in potential synaptic activity, it may explain many of the changes observed in neural-based cardiorespiratory function in an exercise-trained organism.


This study was supported in part by National Heart, Lung, and Blood Institute Grant HL-37400 and the University of Illinois at Urbana-Champaign Campus Research Board.


The authors thank Dr. Ellen Evans for making the dual-energy X-ray absorptiometry facility available to us and providing expert assistance.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract