We investigated whether the relative contribution of the baroreflex and central commands to the control of heart period differs between spontaneously hypertensive rats (SHR) and Wistar-Kyoto normotensive rats (WKY) during physiological behavior. Rats were instrumented with an arterial catheter and with electrodes for discriminating wakefulness, nonrapid eye movement sleep (NREMS), and rapid eye movement sleep (REMS). The cross-correlation function (CCF) between spontaneous fluctuations of heart period and mean arterial pressure was computed at frequencies <0.2 Hz. The baroreflex determines a positive correlation between heart period and previous pressure values. This pattern was observed in the CCF during quiet wakefulness (QW) and NREMS, and in QW, it was accompanied by a pronounced negative correlation between heart period and subsequent pressure values. The relative baroreflex contribution to the control of heart period, estimated from the positive peak value of the CCF, was lower in SHR than in WKY during QW but not during NREMS. During REMS, the CCF showed a negative correlation between heart period and both previous and subsequent pressure values, reflecting the prevalence of central autonomic commands. The relative contribution of central commands to the control of heart period, estimated from the negative peak value of the CCF, was lower in SHR than in WKY during REMS. These results suggest that during QW and REMS, the control of heart period exerted by the baroreflex and central commands, respectively, is less effective in SHR than in WKY. This difference is not apparent in a behavioral state of autonomic stability such as NREMS.
- arterial blood pressure
- spontaneous fluctuations
- central autonomic commands
- cardiac baroreflex sensitivity
- cross-correlation analysis
the spontaneous fluctuations of heart period (HP) and arterial pressure carry information on cardiovascular control mechanisms (10, 30). Among these, the baroreflex is of particular interest because low cardiac baroreflex sensitivity (BRS) is a negative prognostic factor after myocardial infarction (22) or heart failure (33) and characterizes arterial hypertension (6, 19).
Several BRS indexes have been developed based on spontaneous cardiovascular fluctuations (23). However, these indexes yield estimates that agree partially (36) if at all (26) with those obtained by applying artificial cardiovascular perturbations. This may be expected because multiple control mechanisms, including central autonomic commands (15) and reflexes from cardiac and aortic walls (29), impinge on spontaneous cardiovascular fluctuations in addition to the baroreflex. For the same reason, however, spontaneous cardiovascular fluctuations allow the unique opportunity to evaluate the relative contribution of different mechanisms to cardiovascular control during real-life behavior. The contributions of the baroreflex and central commands to the control of HP change predictably among wake-sleep states (41, 42, 47). If these contributions also change in disease conditions, they will be of potential relevance as novel prognostic indexes based on spontaneous cardiovascular variability.
In the present study, we aimed at assessing whether abnormalities in the baroreflex and central contributions to the control of HP may be involved in the pathophysiology of arterial hypertension, which is a disease of dramatic worldwide prevalence (46). To this aim, we tested the hypothesis that the cross-correlation function (CCF) (5) computed on spontaneous fluctuations of HP and mean arterial pressure (MAP) differs between spontaneously hypertensive rats (SHR) and their Wistar-Kyoto normotensive controls (WKY) during physiological wake-sleep behavioral states. SHR are an experimental model widely studied to understand the pathophysiology of arterial hypertension, as they allow the control of environmental and genetic confounders.
The cardiac baroreflex operates as a delayed negative feedback control (45), causing a positive correlation between HP and previous MAP values. Central autonomic commands (15) and positive-feedback reflexes (29) cause opposite changes in HP and MAP and thereby a negative correlation between these variables. The CCF, which yields the linear correlation between HP and MAP as a function of the underlying temporal relationship, allows to discriminate these patterns. This technique has been successfully utilized to study the autonomic dysfunction due to spinal cord injury (1, 3) and the sleep-dependent changes of cardiac control in newborn life (41). For the purpose of integrating the information provided by the CCF analysis, BRS was also estimated from spontaneous fluctuations of HP and MAP with a validated technique (35).
MATERIALS AND METHODS
Experiments were performed on male SHR and WKY (Charles River Italy, n = 6 per group), which were kept on a light-dark cycle of 12:12-h period (light on at 0900), ambient temperature at 23°C, and free access to food and water. The study protocol was approved by the Bologna University ethical committee on animal experimentation.
At the age of 9 wk, rats were implanted under general anesthesia (1–2% halothane, 70% N2O, O2 balance) and sterile conditions with electrodes for electroencephalographic (EEG) and electromyographic (EMG) recordings and a catheter in the abdominal aorta. The catheter was tunneled subcutaneously and fixed to the skull together with the electrodes by dental acrylic. Details of the surgical procedure are described elsewhere (42). Ketoprofen (1 mg/100 g body wt; Aventis), benzilpenicillin benzatinic (15,000 IU/100 g body wt; Fournier), and streptomycin sulphate (20,000 IU/100 g body wt; Bristol-Myers Squibb) were administered subcutaneously for postoperative analgesia and antibiotic prophylaxis.
After 1-wk recovery, EEG, EMG, and arterial pressure (P23 transducer, Statham) were continuously recorded for 4 consecutive days from 1000 to 1800 with the animals undisturbed and freely moving in their own cages. EEG, EMG, and arterial pressure signals were filtered (0.3–60, 100–1,000, and <100 Hz, respectively). EEG and EMG were digitized at 128 Hz. Arterial pressure was digitized at 1,024 Hz to improve the resolution of HP data, which were obtained as pulse intervals. For data acquisition and analysis, custom software was developed in C and MATLAB (The MathWorks) languages. On day 5 of recording, a polyethylene tube (20 μl in volume) was connected to the arterial catheter to obtain three blood samples (150 μl) during uninterrupted wake-sleep states (40). The samples were immediately analyzed (Gem 3000, Instrumentation Laboratory). The pH, the partial pressures of oxygen and carbon dioxide, the concentrations of sodium and potassium ions, glucose, lactate, and the hematocrit were measured to evaluate the health status of the animals during the recording period.
Discrimination of wake-sleep states.
Wake-sleep states were visually scored following the criteria that were adopted in previous reports by our group (40, 42, 47). The scoring was performed on 2-s epochs based on EEG and EMG recordings by four trained investigators. The recordings of each rat were scored on a consensual basis by at least two scorers. Wakefulness was scored when EEG activity was at low voltage and devoid of slow waves (δ frequency band), K complexes, and sleep spindles. Nonrapid eye movement sleep (NREMS) was scored when EEG slow waves were predominant or mixed with K complexes and sleep spindles. Rapid eye movement sleep (REMS) was scored when EEG θ waves were predominant and the EMG displayed atonia with occasional muscle twitches. Care was taken to exclude cortical microarousals (i.e., temporary desynchronization and reduction in EEG voltage) and epochs of intermediate sleep (i.e., when EEG displayed prominent sleep spindles on a background of θ wave activity).
Time series of cardiovascular signals.
Beat-to-beat time series of HP and MAP were obtained from the arterial pressure signal within uninterrupted wake-sleep episodes of duration ≥60 s. MAP, which was computed as the average arterial pressure during each cardiac cycle, is more reliable than systolic pressure in long-term recordings in rats (35). HP was computed as the time interval between the onsets of successive systolic upstrokes. In wakefulness, the occurrence of movement artifacts distorted the morphology of the pulse waves. Although these alterations rarely affected the quality of the MAP signal, they impeded the accurate determination of HP from the pulse wave. Thus, in case movement artifacts occurred during an episode of wakefulness, only the episode's portions with duration ≥60 s and devoid of artifacts were retained for analysis. Analysis was thus restricted to quiet wakefulness (QW), when the animals were not involved in locomotion or major movement, to optimize the accuracy of the HP determination. In the whole study group, HP and MAP data were analyzed on 662 episodes in QW, 3,263 episodes in NREMS, and 917 episodes in REMS.
Time series of cardiovascular signals were resampled at 16 Hz by piecewise cubic Hermite interpolation and low-pass filtered (<0.2 Hz, 10 pole Butterworth filter), to focus the analysis on spontaneous fluctuations at frequencies at which the baroreflex is effective in buffering arterial pressure variability. In fact, blood pressure variability at 0.4 Hz is dampened rather than enhanced by sino-aortic denervation in rats (7, 9, 13). The cut-off frequency of the filter was well below that of cardiovascular variability at the breathing rate in rats (7).
For each wake-sleep episode, the CCF between HP and MAP (41), HP variance, and MAP variance were averaged over consecutive data subsets of duration 60 s overlapped for 58 s. The CCF was normalized so that the auto-correlations at time 0 shift were identically one.
The CCF analysis yields the linear correlation coefficient between HP and MAP as a function of the time shift between these variables (5). The absolute value of the correlation coefficient indicates the degree of linear dependence between HP and MAP, while its sign indicates whether the relationship between HP and MAP is direct (positive sign, with both HP and MAP increasing or decreasing) or inverse. The absolute value of the time shift indicates the time interval considered between HP and MAP, while its sign indicates whether HP fluctuations precede (positive sign) or are preceded by (negative sign) those of MAP. For example, a time shift of −1 s corresponds to the correlation coefficient between the HP values and the MAP values that occurred 1 s before them.
The HP vs. MAP correlation at time 0 shift is positive in rats during wakefulness and NREMS and becomes negative during REMS (42, 47). Based on such evidence, the analysis was focused on the positive CCF peak in QW and NREMS and on the negative CCF peak in REMS. To determine the correlation coefficients at the CCF peaks, the analysis was performed in two steps (41). Each step was repeated over all episodes analyzed for each wake-sleep state in each subject, yielding a total of 36 (subject, state) combinations. Step 1 determined the median value τ of the time shifts corresponding to the positive (in QW and NREMS) or the negative (in REMS) CCF peak. Step 2 determined the median value of the correlation coefficient corresponding to the value of τ in the same (subject, state) combination.
To assess the role of random factors on the correlation between HP and MAP, the CCF was also computed on an isospectral set of surrogate data with randomized phase (44).
Cardiac baroreflex sensitivity.
BRS was estimated by adapting a technique validated in rats (35). Time intervals were identified, in which MAP spontaneously underwent a monotonic variation (ramp) of amplitude >1 mmHg and duration ≥0.75 s. For each MAP ramp, a set of sequences of HP values were considered, each having the same duration as the MAP ramp and starting from 0.625 to 1.25 s (step 1/16 s) after its onset. The regression coefficients between each HP sequence and the MAP ramp were computed and averaged. The results were further averaged over different MAP ramps to estimate BRS in each wake-sleep episode.
Mean values within rat and state were retained for analysis except for variables (variances, peak values of the CCF) with nonnormal distributions within rat and wake-sleep state, of which median values were retained instead. Data were analyzed by ANOVA (GLM procedure with mixed-model design and significance at P < 0.05, SPSS software, SPSS). In case of significant interaction between the rat strain factor and wake-sleep state factor, simple effects of the rat strain were tested in each state with independent-sample t-tests. Data are presented as means ± SE in the text, tables, and figures, with n = 6 animals for each rat strain.
Total duration of the wake-sleep episodes recorded and analyzed.
Table 1 reports the total duration of the wake-sleep episodes recorded in SHR and WKY as well as the total duration of the episodes, on which the analysis of cardiovascular variables was performed in the two strains.
Analysis of arterial blood samples.
Results of the arterial blood analysis did not display significant state or state × strain interaction effects. The effect of the strain was significant only for pH, whose lower value in SHR may be due to a reduced tubular bicarbonate reuptake (27). Table 2 shows the results averaged within rat and strain, which were similar to those previously obtained in healthy rats (40).
Mean value and variance of HP and MAP.
The mean value of HP and MAP and the variances of their low-pass filtered time series varied significantly among wake-sleep states (Table 3). The mean value of MAP was significantly higher in SHR than in WKY in each wake-sleep state. In SHR, HP variance was significantly lower than in WKY during QW and significantly higher than in WKY during REMS. MAP variance was higher in SHR than in WKY, the difference being significant during NREMS and REMS.
Low-frequency fluctuations of HP displayed a pattern similar and parallel to that of previous changes in MAP during QW and particularly during NREMS. During QW, increases of MAP were also evident after episodes of HP reduction. During REMS, conversely, HP and MAP tended to fluctuate in opposite directions (Fig. 1).
In agreement with the visual analysis of the time series, the average CCF between HP and MAP showed a positive peak value at negative time shifts (i.e., for HP following MAP fluctuations) during QW and NREMS. During QW, this pattern coexisted with a prominent negative correlation between HP and subsequent pressure values. Conversely, in REMS, the average CCF showed a negative correlation between HP and both previous and subsequent pressure values, with a negative peak at positive time shifts (Fig. 2).
The positive peak value of the CCF in QW and the negative peak value of the CCF in REMS had significantly lower magnitude in SHR than in WKY. During NREMS, conversely, the magnitude of the positive peak value of the CCF differed negligibly and nonsignificantly between SHR and WKY (Fig. 3).
On isospectral surrogate data in SHR, the positive peak value of the CCF in QW, the positive peak value of the CCF in NREMS, and the negative peak value of the CCF in REMS were −0.01 ± 0.01, 0.01 ± 0.01, and −0.03 ± 0.02, respectively. Corresponding values in WKY were 0.01 ± 0.01, 0.00 ± 0.01, and −0.02 ± 0.01. Neither the wake-sleep state and rat strain nor their interaction significantly affected the peak values of the CCF computed on isospectral surrogate data.
BRS differed significantly among wake-sleep states and was significantly lower in SHR than in WKY, without significant interaction between the wake-sleep state and the rat strain (Fig. 4).
We tested the hypothesis that the relative contribution of the baroreflex and the central commands to the control of HP differs between SHR and WKY during physiological behavior. To this aim, we analyzed the CCF between spontaneous fluctuations of HP and MAP during QW, NREMS, and REMS. During QW and REMS, the CCF significantly differed between SHR and WKY, supporting our hypothesis.
In QW and NREMS, the positive peak value of the CCF occurred at negative time shifts (Fig. 2), which refer to the correlation between HP and previous MAP values. This pattern is consistent with the baroreflex contribution to cardiac control (41) because the baroreflex operates as a delayed negative feedback system (45). The positive peak value of the CCF had lower magnitude in SHR than in WKY during QW, but not during NREMS (Fig. 3). This result thus indicates that the relative contribution of the baroreflex to the modulation of HP was lower in SHR than in WKY during QW, but not during NREMS. The difference observed in QW is intriguing and prompts its confirmation in the episodes of wakefulness as a whole. On the other hand, it is worth noting that NREMS is a state of autonomic stability (37), as confirmed by values of HP and MAP variance (Table 3).
In REMS, HP fluctuations were negatively correlated with subsequent and previous MAP values (Fig. 2). Sequences of HP and arterial pressure that vary in opposite directions, and hence are negatively correlated, reflect nonbaroreflex autonomic control (24). This control may be exerted by central autonomic commands (15) and enhanced by positive-feedback reflexes from cardiac and aortic walls (29). In addition, the mechanical effect of HP on cardiac output, and hence on MAP, contributes to the negative correlation between HP and subsequent MAP values.
In REMS, the negative peak value of the CCF had lower magnitude in SHR than in WKY (Fig. 3). Since HP is positively related with stroke volume in conscious rats (17), differences in this relationship between SHR and WKY might underlie our result. However, low-frequency MAP fluctuations do not arise from fluctuations in cardiac output, but rather from those in vascular resistance (17). Moreover, the mechanical effect of HP on cardiac output may actually be strengthened in SHR by their lower arterial compliance (38). Finally, surges of arterial pressure in REMS, although associated with tachycardia (39, 41), are actually driven by increases in peripheral vascular resistance (14, 31) that are due to an increased sympathetic efferent activity (2, 43). Therefore, in SHR, the weaker negative correlation between HP and MAP during REMS (Fig. 3) suggests that the relative contribution of central commands to the modulation of HP was lower than in WKY in this state. This is a novel finding, which is intriguing because REMS is a condition of disengagement from the external environment. Thus our finding supports the view that the increased cardiovascular reactivity to environmental stress, which characterizes SHR with respect to WKY, is due to behavioral trait differences between these strains (20).
Relationship between the cross-correlation analysis and the estimate of cardiac BRS.
With the exception of a recent report during active wakefulness (21), a reduced BRS has been consistently reported in SHR with respect to WKY with a variety of techniques (16, 32, 34). Accordingly, we observed that BRS was lower in SHR than in WKY in all wake-sleep states (Fig. 4). Clearly, the differences between SHR and WKY disclosed by the CCF analysis do not mirror those found in BRS. The different relative contribution of the baroreflex and central commands to cardiac control, which is detected by the CCF analysis in SHR, provides novel insight on cardiac control in this model of arterial hypertension.
The baroreflex was effective in all wake-sleep states, as demonstrated by positive BRS estimates (Fig. 4). However, during QW, the CCF did not only show a positive correlation between HP and previous values of MAP, but also a prominent negative correlation between HP and subsequent values of MAP (Fig. 2). Such negative correlation indicates that the baroreflex and the central commands coexisted during QW, as it has been previously shown in rats by Cerutti et al. (8) with an original statistical approach. On the other hand, during REMS, the lack of a positive correlation between HP and previous MAP values (Fig. 2) indicates that the relative contribution of the baroreflex to cardiac control was negligible with respect to that of central commands. However, consistently with the coexistence of the baroreflex and the central commands, a hump of the CCF was evident in REMS at negative time shifts, corresponding to the positive CCF peaks in QW and NREMS (Fig. 2). The comparison between the CCF analysis and the BRS thus highlights the concept that the baroreflex interacts dynamically with the central commands in physiological behavior (12, 24, 47).
The analysis of the CCF quantifies the strength of the linear relationship between HP and MAP as a function of the time shift between the variables (5). Although multiple nonlinearities are known to characterize cardiovascular regulation (45), simple linear processes provide a conservative starting point to understand cardiovascular variabilities (10).
Our CCF analysis, which is based on the whole time series of HP and MAP, yields information similar to other linear techniques, which focus on spontaneous ramps of arterial pressure and are based on the direction of the associated HP changes (12, 24, 25). However, the applicability of these techniques is unclear in rats, because in this species the only validated sequence technique for BRS estimation takes into account MAP ramps regardless of the direction of the associated HP changes (35).
In NREMS and particularly in QW, the time shifts at the positive CCF peaks (Fig. 2) were longer than the estimated latency of the earliest cardiac baroreflex response (less than 1 s) (4). These time shifts may thus reflect the latency of the maximal cardiac baroreflex response (see Ref. 11) and in QW suggest a relevant sympathetic contribution to cardiac control. Accordingly, in dogs, slow (0.05 Hz) pressure fluctuations elicit opposite changes in heart rate with a delay of 2–5 s, a pattern attributed to the baroreflex sympathetic cardiac response (28). On the other hand, the time shifts at the negative CCF peak in REMS (Fig. 2) were also longer than the latency of the pure mechanical effect of HP on MAP, which shows up within the same cardiac cycle. This observation supports the view that the negative CCF peak does not merely depend on the mechanical effect of HP on MAP, but rather reflects the activity and the relative time courses of central autonomic commands to the heart and resistance vessels.
In conclusion, the results of the CCF analysis suggest that the relative contribution of the baroreflex to the control of HP is lower in SHR than in WKY during QW but not during NREMS, which is a state of autonomic stability. On the other hand, the results suggest that the control of HP exerted by central autonomic commands is less effective in SHR than in WKY during REMS. These results are thus consistent with a novel abnormality in the control of HP, which may be involved in the pathophysiology of arterial hypertension in SHR and strongly depend on the wake-sleep behavioral state.
Our study in SHR provides some indications for translational research in human patients. The development of better prognostic indexes in hypertensive patients is an active area of research (46). In patients with chronic renal failure and hypertension, a low effectiveness of cardiac baroreflex control (baroreflex effectiveness index) (12) is a predictor of mortality (18). Our data indicate that abnormalities in the relative contribution of the baroreflex and central autonomic commands to the control of HP may be involved in the pathophysiology of arterial hypertension independently of renal failure. Moreover, we observed that sleep states strongly modulate such abnormalities, suggesting that the behavioral state should be taken into account when assessing autonomic cardiac control in human patients. Finally, the CCF analysis that we applied may represent a powerful tool to detect abnormalities in the control of HP and investigate their prognostic significance in hypertensive patients. Interestingly, a CCF analysis of heart rate and blood pressure fluctuations has successfully been applied to study autonomic dysfunction due to spinal cord injury in human subjects (1).
The study was supported by the Italian Ministry of Education, University and Research (PRIN Grant 2004050080) and by the University of Bologna (RFO A.A5.A52004238; A.A5.A52005263).
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 © 2007 the American Physiological Society