The synchronization of cardiac and locomotor rhythms has been suggested to enhance the efficiency of arterial delivery to active muscles during rhythmic exercise, but direct evidence showing such a functional role has not been provided. In this study, we tested the hypothesis that the heartbeat is coupled with intramuscular pressure (IMP) changes so as to time the delivery of blood through peripheral tissues when the IMP is lower. To this end, we developed a computer-controlled, dynamic, thigh cuff occlusion device that enables bilateral thigh cuffs to repeatedly inflate and deflate, one side after the other, to simulate rhythmic IMP changes during bipedal locomotion. Nine healthy subjects were examined, and three different occlusion pressures (50, 80, and 120 mmHg) were applied separately to the thigh cuffs of normal subjects while they were sitting. Alternate occlusions of the bilateral thigh cuffs administered at the frequency of the mean heart rate produced significant phase synchronization between the cardiac and cuff-occlusion rhythms when 120 mmHg pressure was applied. However, synchronization was not observed when the occlusion pressure was 50 or 80 mmHg. During synchronization, heartbeats were most likely to occur in phases that did not include overlap between the peak arterial flow velocity in the thigh and elevated cuff pressure. We believe that phase synchronization occurs so that the cardiac cycle is timed to deliver blood through the lower legs when IMP is not maximal. If this can be extrapolated to natural locomotion, synchronization between cardiac and locomotor activities may be associated with the improved perfusion of exercising muscles.
- phase synchronization
- blood flow
biological rhythms interact with external rhythmic stimuli and other internal oscillatory processes. Two oscillators with different periodicities are often led to oscillate in coincidence, thereby giving rise to synchronization (12). During exercise, synchronization between breathing and locomotor activities occurs in humans during running, in galloping horses, and in trotting dogs (4–6, 17). Synchronization between heart rate and ventilation (8, 11, 24, 26) and between heartbeats and locomotor activity also occurs in birds and mammals, including humans (1, 2, 16, 18, 19, 21, 22). The physiological significance of such coordination has been suggested to be a system of economical coaction and of energetic advantage to the organism (6, 17). Despite its appeal, experimental evidence showing functional significance of synchronization between biological rhythms has rarely been presented.
We previously discovered the occurrence of phase-locked synchronizations between heartbeat and locomotor cycle (cardiac-locomotor synchronization; CLS) during treadmill walking and running in humans (18, 19). We therefore suggested that CLS probably occurs because of the presence of a neuronal circuit that modulates cardiac pacemaker activity, depending on the timing of muscle contraction in the cardiac cycle (20). However, it remains unclear whether this synchronization has some functional role or whether it is an incidental consequence of coupled oscillators. One possible explanation for the functional significance of CLS is that it enhances cardiovascular efficiency (7, 15, 16). During locomotion, muscle contraction produces an elevation in intramuscular pressure (IMP) and compresses the vascular bed, which results in closure of the microcirculation and the cessation of perfusion (3, 28). Therefore, it is likely that locomotor activity periodically occludes blood flow through active muscles, such as during each stride. As has been previously suggested (16), if during locomotor activity peak IMP occurs at a time other than when arterial pressure is highest, then a reduction in cardiac afterload and consequent optimization of blood flow to active muscle can be expected.
The present investigation was conducted to test this hypothesis by simulating rhythmic IMP changes using thigh cuff occlusion. It was designed to determine whether spontaneous phase synchronization occurs between cardiac and thigh cuff occlusion rhythms (simulating bipedal locomotion) and, if so, whether the synchronized phase is appropriate to ensure perfusion to peripheral tissue. We used dynamic synchronization analysis to examine the relationship between cardiovascular variables and simulated IMP changes (30, 31). Herein we report that each heartbeat shows in-phase coupling with occlusion rhythm such that the peak arterial flow in the legs is not overlapped by the elevated phase of IMP.
Nine healthy men with no history of cardiopulmonary diseases volunteered to participate in this investigation. Their age, height, and body weight were 22.5 ± 2.5 (SE) yr, 169 ± 7 cm, and 62.3 ± 8.0 kg, respectively. Each subject gave informed consent after a verbal explanation of the experimental procedures was given, and the experimental protocol was approved by the institutional ethics committee.
First, a cuff-inflation device was developed that allowed intrathigh cuff pressure to be rapidly elevated above the systolic pressure at a desired rate. The cuff pressure was regulated by a computer-controlled solenoid valve (FGG-31; CKD Japan) connected to a high-pressure gas cylinder (Fig. 1). The computer (II) calculated the mean heart rate (mHR) from a subject's electrocardiograph (ECG) every five beats by detecting the QRS complex of the ECG and transmitted an analog signal proportional to the mHR to the other computer (I) through a digital-to-analog converter. The computer (I) alternately drove two solenoid valves at the mHR with an opening duration of 250 ms. Thus the device enabled bilateral thigh cuffs to repeatedly inflate and deflate, one side after the other (simulating IMP changes during bipedal locomotion) at the mHR frequency. The duration of deflating was dependent on cuff compliance, and its time constant (63% deflation) was ∼0.38 s.
Procedures for occlusion.
Each subject wore contoured thigh cuffs 19 cm in width on both upper thighs (CC-17; Hokanson). They were in the sitting position during the experiment and were instructed to remain relaxed. After instrumentation, the subjects were asked to breathe in time with a 0.2-Hz metronomic buzzer signal to control the respiration frequency, which affects the amplitude of heart rate variability (respiratory sinus arrhythmia). The administration of thigh cuff occlusion was applied to the subject when resting. During the measurement, occlusion pressures were continuously and alternately applied to the bilateral thigh cuffs at the frequency of the mHR. The right and left intracuff pressures were monitored by a semiconductor transducer (COPAL; P-2000) and were displayed on a real-time basis on a computer screen to confirm whether the occlusion pressure was appropriately applied. The peak cuff pressure was set to 50, 80, or 120 mmHg and manually regulated by adjusting a secondary valve attached to a high-pressure cylinder. The pressure transducers were calibrated with a mercury manometer.
The R-R interval (RRI) was continuously measured from a surface ECG using standard bipolar leads. The skin was abraded with polish gel (SkinPure; Nihon Kohden) and cleaned with alcohol to reduce skin electrode impedance. The ECG signal was amplified and filtered (10–300 Hz) to distinguish the R waves of the QRS complex and was digitized with a sampling frequency of 1 kHz by a personal computer-based system equipped with a 12-bit analog-to-digital converter (AD12–16A; Contec). Systolic blood pressure (SBP) data were also measured using a volume compensation finger cuff (Finapres, Ohmeda 2300) placed on the index or middle finger of the subject's left hand. During data collection, the left arm was supported at heart level. Blood velocity (BV) measurements were made using the right common femoral artery ∼5 cm distal to the inguinal ligament using Doppler ultrasound velocimetry (SD-50EX; Hadeco). A flat probe with an operating frequency of 5 MHz was used and was fixed to the skin over the femoral artery with a beam angle of 45 degrees with respect to the skin. Respiratory flow was also monitored by a hot wire-type flowmeter (model RF-2; Minato) attached to the expiratory port of a breathing valve (model 7930; Hans Rudolph). Cardiovascular parameters, including ECG, SBP, and arterial BV as well as thigh cuff pressures and respiration signals, were simultaneously recorded online using a FM DAT tape recorder (RD-130TE; Teac). The duration of each record was 20 min.
The cuff pressure onset was defined as the time when the rising phase of the cuff pressure crossed a preset threshold close to the peak of the cuff pressure.
To obtain the relative phase relationship between the cardiac and occlusion rhythms (φC-P), the timing of each successive QRS spike and that of each successive peak of the intracuff pressure was measured using a customized software data-acquisition system interfaced with a personal computer. Each successive marked event is equivalent to one oscillatory cycle. A 2π increase in the phase is attributed to the interval between the subsequent marked events. Within one oscillatory cycle, the instantaneous phase is φ(t) = 2π(t − tk)/(tk+1 − tk) + 2πk, where tk is the time of the kth marked event. The relative phase for the occurrence of the QRS complex with respect to the cuff-pressure interval was calculated as φC-P = [φP(tk) mod 2π]/2π, where φP(tk) is the instantaneous phase of the peak cuff-pressure at time of kth heartbeat (i.e., instantaneous cardiac phase, φc). Because of the experimental conditions in which the right and left thigh cuff pressures were alternately increased at the frequency of the mHR, 2:1 (cardiac cycle-IMP cycle) phase synchronization could be detected, if present, by plotting the relative phase of φC-P vs. time (synchrogram) and looking for horizontal plateaus (14, 30, 31).
To characterize the strength of synchronization, the deviation of the actual distribution of the relative phase from a uniform one should be quantified. For this purpose, a synchronization index, ρ(tk), based on the Shannon entropy was calculated (31). ρ(tk) was defined as: where S=− pn ln pn is the entropy of the distribution of φC-P, Smax = ln(M), M is the total number of bins (20 in this case), and pn is the probability in each bin (n) at the kth time window. The sliding ρ(tk) values were computed using 60 data point intervals (N), incremented in 10-point steps. The mean ρ value (ρ̂) was calculated by averaging over the whole experimental period. The index is restricted to the unit interval 0 ≤ ρ ≤ 1 and is minimal for a uniform probability distribution and maximal in the case of a δ-function like probability distribution. Because we exclusively consider 2:1 (cardiac cycle-IMP cycle) phase synchronization, the maximum value of ρ is [ln(20) − ln(2)]/ln(20) = 0.769 when perfect phase locking occurs. The entropy S varies depending on the number of bins M. There is no theoretical rule for determining which M is sufficient or best for estimating S. In addition, since pn is represented by maximum likelihood estimates |Axp̂n (i.e., frequency count) with finite samples of N, the estimator S leads to a systematic underestimation of true entropy. It is suggested that this bias should be corrected by the term (M − 1)/2N by taking the first order of Taylor expansion around the probability pn to the ln function (25). In our analysis, this correction was ∼5.3%. Moreover, the bias and variance of the estimator ρ̂ with the use of the identical bins and window length were evaluated by creating independent identically distributed time series originating from a pseudorandom generator. Those values are <0.06 for bias and 0.02 for variance, respectively.
Furthermore, to determine the probability of entrainment between cardiac and simulated IMP rhythms, a surrogate technique was applied (21, 26). The surrogate data were generated by randomly shuffling the original instantaneous phase of the pressure rhythm φP(t) to secure the statistical properties of the original data. The sequence of heartbeats remained the same as for the actual data. Next, a new relative phase between two rhythms was created using the instantaneous cardiac phase, φc(t). Because of the random order of φP(t), deterministic phase relationships between the heartbeat and occlusion rhythm are destroyed. If two rhythms have an exact phase relationship, similar structures to the original relative phase structures occur in the surrogate data. By computing multiple trial-shuffled estimates (100 shuffles), a distribution of values under the null hypothesis was obtained. Next, a group-averaged ρ̂ value of the original data was compared with that of the surrogate data calculated from an identical choice of the observation window.
If significant phase synchronization was detected, the timing between the onset of the peak arterial flow in the thigh and the peak of the cuff pressure was examined and displayed as normalized values to infer the functional significance of the phases that favored coupling with the occlusion rhythm.
The values were represented as means ± SD. The synchronization index for the original and surrogate data was compared using paired t-tests, and a Chi Square test was performed to test whether the observed frequencies of the histogram of the phase difference significantly differed from the uniform one. Values of P < 0.05 were considered to be statistically significant.
A representative synchrogram between the cardiac and intracuff pressure rhythms from subject 7, for which 50-, 80-, and 120-mmHg cuff pressures were applied separately, is shown in Fig. 2A. The histogram of φC-P distribution shows two distinct peaks when the cuff pressure was elevated to a level comparable to a systolic pressure of 120 mmHg. More heartbeats occurred in phase at around 0.17 and 0.67 of φC-P in this case. The average ρ value during the 20-min measurement period for the 120-mmHg cuff pressure administration was 0.225 (Fig. 2B), and the histogram of the phase distribution was significantly different from the uniform one (P < 0.01). However, relatively uniform distributions were seen when the peak cuff pressure was low, such as for 50 and 80 mmHg administration. The average ρ values for 50- and 80-mmHg pressure administration were 0.121 and 0.167, respectively, and the distribution profiles of these histograms were not significant from the uniform one. To determine whether the phase synchronization that occurred was of incidental consequence or not, a surrogate technique was employed. An example of the ρ(t) value for 100 surrogate series for 120 mmHg administration is depicted as thin (mean) and dotted (±SD) lines in Fig. 2B. The mean value of the surrogate ρ (ρ̂) averaged over the whole time period was 0.092 ± 0.008 (SD) in this case.
Because significant phase synchronization could not be observed when the peak cuff pressure was lower than the systolic arterial pressure (ρ̂ values for 50 and 80 mmHg administration were 0.083 ± 0.028 and 0.099 ± 0.042, respectively), data from the experiments for 120 mmHg administration are summarized in Table 1. There were no significant differences in the mean values of the RRI and cuff-pressure interval, indicating that the cuff pressure administration was successfully followed by the mean frequency for the heart rate for each subject (also see Fig. 2C). The group mean values for SBP were 123 mmHg. For all recordings, the distribution of the heartbeats was significantly different from the surrogate data [group mean ρ̂ values: 0.192 (original) vs. 0.088 (surrogate), P < 0.01] when the peak cuff pressure was elevated to 120 mmHg, indicating that phase synchronization was the result of the entrainment of the heartbeat by the simulated IMP changes.
Figure 3 compares the timing of cuff pressure changes with the ECG signal, finger blood pressure, and femoral artery flow-velocity tracing for the same subject as presented in Fig. 2 to show the phase relationship for which heartbeats were most likely to occur. The timing of peak cuff pressure occurred during the relaxation phase of the cardiac cycle; this phase was not overlapped by the inflow phase of the femoral artery.
Figure 4A shows histograms of φC-P distributions for all subjects for 120-mmHg peak pressure administration. For each subject, there were two distinct peaks, and the distribution profiles of these histograms were significant from the uniform one. To obtain a quantitative assessment of the phase that favored coupling with the occlusion rhythm, the phases were expressed as a percentage of the timing of the peak arterial inflow for the thigh (%BV). Because of difficulties with taking the BV measurements during pneumatic thigh cuff occlusion, accurate BV profiles for the thigh were obtained for only five of the subjects during cuff occlusion. For subjects for whom BV profiles could not be obtained, the peak arterial flow timing was estimated from the relationship between the timing of SBP and that of peak arterial flow for the thigh obtained without applying thigh cuff pressure. As shown in Fig. 3, SBP and peak arterial inflow appeared to almost have the same timing, since the distance from heart to finger, measurement site for blood pressure, and the inguinal ligament was not very different. Figure 4B is a plot of the peak cuff pressure timing relative to a recording of femoral artery BV that occurred most frequently for each subject. For all subjects, the synchronized phases of the peak cuff pressure were clustered at on average ∼30 or 80% with respect to the peak BV signal for the thigh.
The present investigation has revealed that alternating occlusion of the bilateral femoral arteries causes human heartbeats to become phase locked with the occlusion rhythm, provided that the occlusion pressure is comparable to the systolic pressure and that the frequency of occlusion is close to that of the cardiac rhythm. The strength of the phase synchronization assessed by the Shannon entropy index was significantly diminished after surrogation, indicating that phase synchronization was the result of the entrainment of the heartbeat by the simulated IMP changes.
The present investigation was designed to determine whether spontaneous phase synchronization occurs between the cardiac and thigh cuff occlusion rhythms when both rhythms are almost synchronous. Because previous studies suggested that the development of CLS during walking and running is speed specific (16, 18) and therefore that the synchronization region has a narrow range of differences in frequencies between the driven oscillator (i.e., cardiac rhythm) and external force, the frequency of cuff pressure administration was set at the mHR frequency.
The group-averaged synchronization index for 120 mmHg administration was 0.192 ± 0.029. This value was only ∼25% of the strength of perfect synchronization but was statistically significant compared with that for the surrogate data (0.088 ± 0.015). The lack of strong synchronization between the heartbeat and occlusion rhythm may have been because of the influence of the respiratory modulation of the heartbeat. As has been suggested (24), the synchronization of heartbeats to other intrinsic oscillatory processes (e.g., respiratory rhythm) is considered to depend on the strength of respiratory sinus arrhythmia. Furthermore, although the maximum thigh cuff pressure was limited to an SBP level of 120 mmHg in the present study to prevent the unpleasantness of ischemic pain resulting from thigh compression, studies increasing the cuff-occlusion pressure beyond the SBP level are expected to cause a stronger synchronization between cardiac and thigh cuff occlusion rhythms.
The mechanisms responsible for phase synchronization between the cardiac and occlusion rhythms remain controversial. Our previous study suggested that the synchronization between cardiac and muscle contraction rhythms occurs because of the phase dependence of heartbeat modulation on muscle contraction (20). We speculated that the phase-dependent property involves an intrinsic characteristic of the heart and as interactions of afferent signals arising from the stimulation of mechanoreceptors in contracting muscles. In the present study, afferent signals arising from active muscle, such as mechano- and metaboreceptors, did not contribute to phase synchronization, since the experiments were carried out under the absence of muscle contractions. For a similar reason, central command, which has been suggested to modulate cardiac rhythm by acting on a central cardiovascular control area, can be eliminated as a possible mechanism responsible for phase synchronization.
Phase synchronization was abolished when the occlusion pressure was below the systolic pressure, suggesting that a venous mechanism, such as that of the muscle pump, was not important. Toska et al. (32) have shown that rapid inflation to suprasystolic pressure of bilateral thigh cuffs does not change central venous pressure significantly. This observation supports the hypothesis that the involvement of a venous mechanism in the synchronization was unlikely. Rather, the existence of dynamic interaction in the baroreflex system or an intrinsic property of heart might have been involved. In a “high-pressure” system, arterial baroreceptors increase their rate of discharge during the systolic phase of the cardiac cycle on a beat-by-beat basis, resulting in a decrease in sympathetic outflow and an increase in parasympathetic outflow to the heart (9, 13). Impulses that travel along afferent nerves to the cardiovascular medullary center occur in the aortic nerve in synchrony with the rising phase of ventricular systole (9, 27). These impulses inhibit the heart rate by stimulating the vagus nerve, resulting in prolongation of the RRI. By discharging this burst in synchrony with the afferent input arising from some receptors triggered by occlusion, the efferent parasympathetic effect on the heart might be abolished, resulting in a shortening of the RRI. Such a mechanism could operate to modulate coupling between heartbeat and IMP development. Greater muscle compression by thigh cuff occlusion might act as a mechanical stimulus. Thus IMP-sensitive mechanoreceptors, if present, may contribute to some degree to producing phase synchronization.
Alternatively, forceful rhythmic occlusion could affect the rate of ventricular filling by altering afterload and therefore may alter the duration of the cardiac cycle. Simmons et al. (29) observed a brief period of locomotor-cardiac coupling in conscious trotting dogs and hypothesized that the transient coupling is a function of locomotor and ventilatory influences on venous return and/or ventricular ejection. They stressed the importance of a reflex mechanism, such as the baroreceptor reflex, Bainbridge reflex, and Frank-Starling reflex, for modulating venous return and ventricular ejection. These reflexes interact and permit modulation of heart rate and stroke volume so that the cardiac pump maintains homeostasis. It has been suggested that synchronization results from some intrinsic property of cardiac cells and/or tissue. Evidence exists that volume loading to the isolated whole heart induces the entrainment of electrical heart rhythm (10). This suggests that cardiac tissues/cells act as nonlinear oscillators capable of forcing the pacemaker to beat at cycle lengths different from its intrinsic cycle length.
As shown in Fig. 4B, for all subjects investigated, the heartbeats were most likely to occur during phases when the peak arterial flow velocity of the thigh and SBP of the finger were not overlapped by the elevated phase of the cuff pressure. It was previously found that muscle blood flow is occluded or even moves in reverse during the contraction phase of rhythmic exercise (33), probably because IMP rises beyond the level of SBP during contraction. Thus blood delivery is increased, and therefore more oxygen is supplied during the relaxation phase of rhythmic muscular exercise. Because it is important to ensure adequate blood flow to contracting muscle for prolonged exercise, it might be meaningful if the cardiac rhythm is coordinated so that the period of muscle contraction (period of increase in IMP) does not override the systolic phase of the cardiac cycle in such a way that the heart expels blood during the relaxation phase of rhythmic exercise, such as for the lower IMP phase. This would require the coupling of cardiac and muscle contraction rhythms with appropriate phase differences for blood flow to reach active muscle. Thus the present results indicate that the heart coordinates the timing of contraction so that the peak arterial flow in the thigh is not overlapped by the elevated phase of the IMP.
As shown in Fig. 4B, the synchronized phases were preferentially clustered in two regions with respect to the relative phases of the cardiac cycle. According to the theoretical analysis (23, 34), stable phase synchronization can occur in phase-advanced regions where the slope of the phase-response curve is greater than −2 and <0. Therefore, this result suggests that the heart exhibits a phase-dependent property in response to the increase in IMP and that its phase-response curve displays a negative slope for the corresponding two regions. This needs to be confirmed in a future study.
Finally, there were some methodological limitations in the present study. First, the direct measurement of IMP change during thigh cuff occlusion was not made but was considered to be the same as the intracuff pressure change. Also, the pattern of simulated IMP change may not have been exactly the same as for actual IMP development during locomotion. Although there are no known reports on the thigh IMP profile during locomotion in humans, IMP can vary in different parts of activated muscle. Ballard et al. (3) found that IMP in the soleus muscle exhibits a peak during the late-stance phase of gait, whereas the tibialis anterior shows a biphasic response during walking. However, the period of increase in soleus muscle IMP exceeding 50 mmHg is comparable with that of the intracuff pressure pattern of the present study; they were estimated to be 35% of gait cycle during walking and 25% during running in the study by Ballard et al., whereas they were ∼28% for the pressure cycle in the present study. Our interpretations assume that alternative thigh cuff occlusion produces similar IMP oscillations in the thigh during bipedal locomotion. Second, the experiments were carried out while the subjects were sitting, where respiratory modulation of heartbeats is greater than that during exercise. Modulation of heartbeats by respiration might interfere with the development of phase synchronization between heartbeats and occlusion rhythm. Therefore, studies utilizing an altered autonomic condition, such as with the subject standing upright or pharmacological blockade of the autonomic nervous system, may yield different synchronization strengths.
In summary, the phase locking of heartbeats to rhythmic thigh cuff occlusion as a simulation of IMP change during bipedal locomotion was observed when the occlusion rhythm was close to the cardiac rhythm. A transient increase in IMP comparable to the SBP induced the inhibition of cardiac contractions. If the present experimental results can be extrapolated to natural locomotion, synchronization between cardiac and locomotor activities may be associated with the improved perfusion of exercising muscles. Further studies are needed to confirm whether the synchronized phases seen in the present study are consistent with those during actual locomotion in a natural setting.
This work was partly supported by Grant-in-Aid for Scientific Research 16500390 from the Japan Society for the Promotion of Science to K. Niizeki.
I am thankful for the participation of the subjects in this research.
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.
- Copyright © 2005 the American Physiological Society