Lower dynamic cerebral autoregulation following acute bout of low-volume high-intensity interval exercise in chronic stroke compared to healthy adults

Abstract

Fluctuating arterial blood pressure during high-intensity interval exercise (HIIE) may challenge dynamic cerebral autoregulation (dCA), specifically after stroke after an injury to the cerebrovasculature. We hypothesized that dCA would be attenuated at rest and during a sit-to-stand transition immediately after and 30 min after HIIE in individuals poststroke compared with age- and sex-matched control subjects (CON). HIIE switched every minute between 70% and 10% estimated maximal watts for 10 min. Mean arterial pressure (MAP) and middle cerebral artery blood velocity (MCAv) were recorded. dCA was quantified during spontaneous fluctuations in MAP and MCAv via transfer function analysis. For sit-to-stand, time delay before an increase in cerebrovascular conductance index (CVCi = MCAv/MAP), rate of regulation, and % change in MCAv and MAP were measured. Twenty-two individuals poststroke (age 60 ± 12 yr, 31 ± 16 mo) and twenty-four CON (age 60 ± 13 yr) completed the study. Very low frequency (VLF) gain (P = 0.02, η² = 0.18) and normalized gain (P = 0.01, η² = 0.43) had a group × time interaction, with CON improving after HIIE whereas individuals poststroke did not. Individuals poststroke had lower VLF phase (P = 0.03, η² = 0.22) after HIIE compared with CON. We found no differences in the sit-to-stand measurement of dCA. Our study showed lower dCA during spontaneous fluctuations in MCAv and MAP following HIIE in individuals poststroke compared with CON, whereas the sit-to-stand response was maintained.

NEW & NOTEWORTHY This study provides novel insights into poststroke dynamic cerebral autoregulation (dCA) following an acute bout of high-intensity interval exercise (HIIE). In people after stroke, dCA appears attenuated during spontaneous fluctuations in mean arterial pressure (MAP) and middle cerebral artery blood velocity (MCAv) following HIIE. However, the dCA response during a single sit-to-stand transition after HIIE showed no significant difference from controls. These findings suggest that HIIE may temporarily challenge dCA after exercise in individuals with stroke.

INTRODUCTION

Dynamic cerebral autoregulation (dCA), or the cerebral vasculature’s ability to react to rapid changes in mean arterial pressure (MAP) (1–3), is impaired after stroke (4–12). One type of aerobic exercise, which may influence dCA and has gained in popularity for individuals after stroke, is high-intensity interval exercise (HIIE) (13–15). Despite multiple narrative reviews suggesting that HIIE may attenuate dCA to a larger degree in individuals with a cerebrovascular injury such as stroke compared with healthy adults (16–18), HIIE is starting to be implemented within stroke rehabilitation to improve motor outcomes (13–15). However, it is unknown whether HIIE may challenge dCA in individuals after stroke because of the rapid fluctuations in MAP when switching between high-intensity exercise and recovery (16, 17, 19).

Evaluating the cerebrovascular system’s capacity to respond to alterations in MAP following aerobic exercise (17, 18) holds significance, given that poststroke dCA might exhibit reduced efficiency in mitigating MAP variations, which could potentially result in diminished cerebral protection (13). A narrative review by Lucas et al. (17) detailed the current gaps in knowledge of the influence of HIIE on the cerebrovasculature in individuals after stroke and stated “unless countered by the neuroprotective influences of sympathetic activation or cerebral autoregulation, [HIIE] potentially increases the risk of hyperperfusion injury predisposing to stroke or blood-brain barrier breakthrough.” Poor dCA regulation after a challenging HIIE stimulus in individuals poststroke may allow for the deleterious transference of arterial blood pressure changes into the smaller downstream cerebrovascular vessels. In individuals with a preexisting injury to the cerebrovascular system such as stroke, the dCA response should be an important factor in prescribing aerobic exercise and determining postexercise recovery care (18).

To our knowledge, no studies have examined dCA after an acute bout of HIIE in individuals poststroke (20). Previous research conducted in healthy adults reported that dCA was less effective during repeated squat-stands following HIIE (21). During exercise recovery after HIIE, MAP decreases (19, 22), which could challenge the dCA response, as dCA seems less effective at buffering decreases compared with increases in MAP (23–28). If dCA decreases after HIIE in healthy adults, one might speculate that greater attenuation of dCA in individuals poststroke compared with healthy adults might increase the cerebrovascular vulnerability of these patients during postexercise recovery following HIIE (18). In healthy adults, spontaneous fluctuation of dCA during recovery after moderate- to vigorous-intensity continuous exercise has also been shown to be affected immediately after exercise and to recover by 30 min (29, 30). However, the dCA response may take longer than 30 min to recover in individuals poststroke potentially because of postexercise hypotension and prolonged recovery of MAP (18). Recovery of end-tidal pressure of carbon dioxide ($\text{P}{\text{ET}}_{{\text{CO}}_2}$) also influences cerebral blood velocity after HIIE (31) in individuals poststroke compared with healthy adults (22) and could play a role in the dCA response during recovery (19, 32, 33). Previous studies show that changes in $\text{P}{\text{ET}}_{{\text{CO}}_2}$ cause increases or decreases in arteriole vascular tone, which influences dCA (34, 35). In young healthy adults, MCAv is reduced immediately after HIIE because of continued hyperventilation causing reduced $\text{P}{\text{ET}}_{{\text{CO}}_2}$ and arteriole vasoconstriction (19). Increases in arteriole vascular tone due to hypocapnia have been shown to improve the dCA response in healthy adults (34, 36). However, individuals poststroke have impaired cerebrovascular vasomotor reactivity to changes in $\text{P}{\text{ET}}_{{\text{CO}}_2}$ (37, 38), which could also contribute to a reduced dCA response to HIIE. Therefore, when examining the dCA response between groups, we need to adjust for differences in breathing recovery or $\text{P}{\text{ET}}_{{\text{CO}}_2}$ following HIIE.

We have addressed the current gaps in knowledge by examining the effects of low-volume HIIE on the dCA response in individuals with chronic stroke (18). At rest, although multiple time- and frequency-domain methods are available and no gold standard exists to quantify dCA (39), transfer function analysis (TFA) using spontaneous fluctuations in MAP and cerebral blood velocity to obtain gain (amplitude of cerebral blood velocity change for a given oscillation in MAP) and phase (difference in the timing of the MAP and cerebral blood velocity waveforms) metrics represents a popular analytical tool. dCA also responds to changes in MAP during everyday activities, such as standing up from a seated position (3). As an individual transitions from sitting to standing, a momentary decrease in MAP occurs because of gravitational effects and venous pooling, and dCA responds by elevating the cerebrovascular conductance index (CVCi) before the arterial baroreflex activation (3).

In the present study, we characterized the dCA response using spontaneous fluctuations in MAP and middle cerebral artery blood velocity (MCAv) as well as during a single sit-to-stand after an acute bout of low-volume HIIE in individuals with chronic stroke compared with age- and sex-matched adult control subjects (CON). We hypothesized that individuals with chronic stroke would have attenuated dCA initially after HIIE and 30 min after HIIE, compared with CON. In addition, we hypothesized that $\text{P}{\text{ET}}_{{\text{CO}}_2}$ may act as a covariate during recovery after HIIE and could influence the between-group differences in the dCA response. Therefore, we statistically adjusted for $\text{P}{\text{ET}}_{{\text{CO}}_2}$ when comparing the dCA response between individuals poststroke and CON.

MATERIALS AND METHODS

The University of Kansas Medical Center Human Subjects Committee approved all study procedures, and the Declaration of Helsinki and local statutory requirements were followed. This study was registered on clinicaltrials.gov (NCT04673994) and STROBE guidelines reported.

Our methods have been described previously (19, 22, 40). Individuals with stroke were included within this study if they were 1) between 40 and 85 yr old, 2) treated for stroke 6 mo to 5 yr ago, 3) classified as inactive by performing <150 min of moderate-intensity exercise per week, and 4) able to answer consenting questions and follow a two-step command. Individuals within the CON group were matched to the sex and age ± 5 yr of an individual poststroke and also classified as inactive. Individuals were excluded from the study if 1) stroke etiology was due to COVID-19, 2) they were unable to stand from a chair without physical assistance from another person, 3) they were unable to perform exercise on a recumbent stepper, 4) they were diagnosed with insulin-dependent diabetes, 5) they were diagnosed with another neurological disease, or 6) we were unable to find a transcranial Doppler ultrasound (TCD) signal on the left or right side.

Study Visit 1

All participants were informed of study procedures, benefits, and risks before providing voluntary written consent. We then collected participant demographics, current medications, prior medical history, and self-reported disability via the modified Rankin scale (mRS) (41). Lower extremity function in individuals poststroke was assessed with the Fugl–Meyer Lower Extremity Subscale and the 5 Times Sit-to-Stand Assessment (42, 43). We then screened for an MCAv signal on both the left and right side with TCD probes (2 MHz; Multigon Industries Inc, Yonkers, New York). TCD signal acquisition was performed following standard depth, gain, and power (44). The position of the probes was recorded in study visit 1 to ensure accurate signal acquisition in study visit 2 when recordings were performed.

Submaximal exercise test.

As previously reported (19, 22), the total body recumbent stepper (TBRS) (T5XR; NuStep, Inc. Ann Arbor, MI) submaximal exercise test was used to determine the estimated maximal watts (estimatedWatt_max) for the acute HIIE bout (45, 46). If individuals were on a beta-blocker medication, maximum heart rate (HR) was determined with the beta-blocker equation [164 – (0.7 × age)] (47).

Exercise familiarization.

The low-volume HIIE exercise familiarization was performed at least 15 min after the TBRS submaximal exercise test or once HR and MAP returned to near-baseline values. Based on previously published low-volume HIIE protocols in older adults and clinical populations, low-volume HIIE was dosed at 70% estimatedWatt_max and active recovery at 10% estimatedWatt_max (14, 48–53). Participants practiced switching between high-intensity exercise and active recovery bouts every minute for ∼10 min.

Study Visit 2

Following our laboratory protocols, the room was dimly lit and kept at a constant temperature (21–23°C) (19, 54). All participants were asked to refrain from caffeine for 8 h (55), food for 2 h (56), vigorous exercise for 24 h (21), and alcohol for 24 h (57).

Equipment setup.

Participants were seated on a recumbent stepper and fitted with the following equipment: 1) bilateral TCD probes secured to an adjustable headset to ensure a stable position and angle to measure MCAv throughout testing; 2) a five-lead electrocardiogram (ECG) (Cardiocard; Nasiff Associates, Central Square, New York) to collect HR; 3) a left middle finger Finometer cuff (Finapres Medical Systems, Amsterdam, The Netherlands) to measure beat-to-beat MAP; 4) a right arm brachial automated sphygmomanometer with a microphone (Tango M2; Suntech, Morrisville, NC) for calibration of the Finometer; and 5) a nasal cannula attached to a capnograph (BCI Capnocheck Sleep 9004; Smiths Medical, Dublin, OH) for $\text{P}{\text{ET}}_{{\text{CO}}_2}$ monitoring. In individuals poststroke with left arm spasticity, the Finometer was placed on the right arm.

Cerebral autoregulation measures.

To quantify spontaneous dCA metrics, participants performed a 5-min seated rest recording 1) at baseline (BL), 2) immediately after HIIE, and 3) 30 min after HIIE while MCAv, MAP, HR, and $\text{P}{\text{ET}}_{{\text{CO}}_2}$ were captured (Fig. 1).

Figure 1.

Protocol measuring dynamic cerebral autoregulation after an acute bout of high-intensity interval exercise (HIIE). BL, baseline; H, high intensity; Post, recovery immediately after HIIE; Post30, recovery at 30 min after HIIE; R, active recovery; TFA, transfer function analysis (conducted during seated rest).

Using a sit-to-stand procedure, we recorded MCAv, MAP, HR, and $\text{P}{\text{ET}}_{{\text{CO}}_2}$ and quantified dCA, based on our previously published methodology, 1) at BL, 2) after 5 min of passive seated recovery after HIIE, and 3) 35 min after HIIE (58). Participants placed the hand with the Finometer across their chest, and a sling was used to secure the arm at heart level during the sit-to-stand recording (58). Participants placed their feet flat on the ground, sat in an upright posture, and had a force-sensitive resistor (SEN-09376; Mouser Electronics Inc., Texas) to detect the exact moment of stance from the chair or the moment of arise-and-off (AO) (58). The sit-to-stand recording lasted a total of 3 min (see Fig. 1). Individuals began in a seated position for 1 min and then were given a 3-s countdown and asked to quickly transition to standing at 60 s and remained standing for 2 min for hemodynamic stability. Study personnel stood next to the participants during the sit-to-stand transition. Initial orthostatic hypotension (within 15 s of standing) was determined by a drop of systolic blood pressure > 40 mmHg and a drop in diastolic blood pressure > 20 mmHg as well as symptoms of dizziness, blurry vision, nausea, or weakness with standing (59–61).

Low-volume HIIE protocol.

Repetitive 1-min intervals of high-intensity exercise (70% estimatedWatt_max) separated by 1-min active recovery bouts (10% estimatedWatt_max) were performed for 10 min (19, 22). The first minute of HIIE was an active recovery bout, to avoid a Valsalva maneuver, which would cause a transient decrease in MCAv (62). After HIIE, resistance and step rate were decreased for a 2-min cooldown.

Data Acquisition

Data with collected at 500 Hz via an analog-to-digital unit (NI-USB-6212; National Instruments) and custom-written code within MATLAB (v2014a; The MathWorks Inc, Natick, MA) (54). Beat-to-beat data were processed offline using the QRS complex of the ECG (54). The left MCAv signal for CON was used for analysis. However, if left MCAv was not obtainable or had noise, the right MCAv signal was used (54). Consistent with previous work in individuals after stroke, the ipsilesional hemisphere’s MCAv signal was used to compare to CON (63, 64).

Transfer Function Analysis

Quantification of spontaneous TFA metrics followed the recommendations of the Cerebrovascular Research Network (CARNet) (65, 66). The 5-min seated rest recordings were computed to cross-spectral analysis via fast Fourier transform utilizing Welch’s method, 100-s Hanning windows, and 50% superposition. Waveforms of MAP and MCAv were analyzed via TFA and described through gain (amplitude), phase (shift in degrees), and coherence (association). Because of dCA acting as a high-pass filter, measures of gain, phase, coherence, and power spectral density (PSD) of the integrate area were reported within very low frequency (0.02–0.07 Hz) and low frequency (0.07–0.2 Hz) bands. Following CARNet recommendations, the critical threshold of coherence was set at 0.34 (65). Coherence determined whether estimates of gain and phase were reliable and could be compared across time (65). We visually inspected our data for phase wraparound, and if detected these data points were removed (65). Attenuated dCA was interpreted as higher gain (greater differences in amplitude) and lower phase (smaller shift) after HIIE compared with baseline.

Sit-to-Stand Analysis

Metrics of dCA included the time delay (TD) before the onset of the regulatory response, rate of regulation (ROR), percent change in MCAv (%ΔMCAv), percent change in MAP (%ΔMAP), and the ratio of %ΔMCAv to %ΔMAP. To measure the TD before the onset of the regulatory response, beat-to-beat CVCi was calculated (CVCi = MCAv/MAP). Based on previously published methods (2, 58), two trained researchers selected the physiological heart beat when CVCi began to increase (without immediate transient reduction) (2). The TD before the onset of the regulatory response was calculated as the time in seconds from AO until the continuous increase in CVCi. The ROR was used to determine the efficiency of the regulatory response, by calculating the slope of the three heartbeats following TD (ROR = ΔCVCi/Δt/ΔMAP) (2). The ROR is taken within phase 1 of the sit-to-stand response (1–7 s after standing), before the onset of the arterial baroreflex response (3, 59). The %ΔMCAv and %ΔMAP were calculated from the seated baseline (15 s before AO) to nadir (the lowest value of MCAv and MAP after standing).

$$\%\Delta =\left[\left(\text{seated\hspace{0.17em}baseline}-\text{nadir}\right)/\text{seated\hspace{0.17em}baseline}\right]\mathrm{\hspace{0.17em}\times \hspace{0.17em}}100$$

An attenuated dCA response was interpreted as a longer TD before the onset of the regulatory response and smaller ROR, greater %ΔMCAv, and a higher %ΔMCAv/%ΔMAP (2, 59).

Statistical Analysis

Statistical analysis was performed with SPSS Statistics Software (IBM Corp, released 2021, IBM SPSS Statistics for Windows, version 28.0. Armonk, NY). A priori α was set at 0.05. Participant characteristics were compared by Mann–Whitney U tests for continuous variables and chi-square tests for categorical variables.

The primary aim of this study was to examine dCA during spontaneous fluctuations and during a sit-to-stand 1) at BL, 2) after HIIE, and 3) at a follow-up 30 min after HIIE in individuals poststroke compared with CON. To analyze the primary aim, we used mixed-model ANOVAs with fixed effects for time (BL/immediately after HIIE/30 min after HIIE), group (stroke/CON), and group × time interaction to examine spontaneous dCA metrics such as VLF and LF gain, normalized gain, phase, and coherence as well as sit-to-stand dCA metrics such as TD before the onset of the regulatory response, ROR, %ΔMCAv, %ΔMAP, and %ΔMCAv/%ΔMAP. Partial eta squared (${\eta }_{\mathrm{p}}^2$) effect sizes were reported for the group effects and group × time interaction effects for each mixed-model ANOVA. For ${\eta }_{\mathrm{p}}^2$ coefficients, thresholds for small, medium, and large effect sizes were 0.01, 0.06, and 0.14, respectively (67, 68). If a significant group effect was found, post hoc comparisons used one-sided Mann–Whitney U tests to determine whether dCA was significantly attenuated in individuals poststroke compared with CON immediately after HIIE and 30 min after HIIE. If a significant group × time interaction effect was found, post hoc comparisons used one-sided Mann–Whitney U tests for between-group differences and Wilcoxon signed-rank tests for within-group differences immediately after HIIE and 30 min after HIIE compared with BL. To control for multiple post hoc comparisons (6 group comparisons and 8 time comparisons), a Bonferroni correction was applied and α was adjusted to 0.0035. As a sensitivity analysis, we also adjusted for $\text{P}{\text{ET}}_{{\text{CO}}_2}$ after HIIE within the mixed-model ANOVAs.

For secondary outcomes such as average MCAv, MAP, CVCi, $\text{P}{\text{ET}}_{{\text{CO}}_2}$, MCAv PSD, and MAP PSD, collected during the measures of dCA, we performed the same statistics as the primary aim. We used a mixed-model ANOVA with fixed effects for time and group and group × time interactions. Assumptions of all models were assessed with residual plots.

RESULTS

Sixty individuals were enrolled within the study. Forty-six individuals were included in the analysis: individuals poststroke (n = 22) and CON (n = 24), as shown in Fig. 2.

Figure 2.

Flow diagram. HIIE, high-intensity interval exercise; TCD, transcranial Doppler ultrasound.

Participant characteristics are described in Table 1 and have been published in part within our previous work (19, 22, 40, 58). Individuals poststroke showed a significantly slower 5 times sit-to-stand, lower average workload at high intensity, and lower workload during active recovery compared with CON. Individuals poststroke exhibited slight to moderate disability as indicated by the mRS and lower extremity Fugl–Meyer subscore (69).

Table 1.

Participant characteristics

	Individuals Poststroke (n = 22)	CON (n = 24)	P Value
Age, yr	60 ± 12	60 ± 13	0.80
Female, n (%)	9 (41%)	8 (33%)	0.76
BMI, kg/m2	30.7 ± 6.0	28.9 ± 6.6	0.37
Race, n (%)			0.34
Asian	0	1 (4%)
Black/African American	4 (18%)	2 (8%)
White/Caucasian	17 (77%)	21 (88%)
White and Native American	1 (5%)	0
Blood pressure medications
Beta-blocker	9 (41%)	1 (4%)	0.004*
ACE inhibitor	6 (27%)	2 (8%)	0.13
Angiotensin II receptor blocker	6 (27%)	5 (21%)	0.73
Calcium channel blocker	8 (36%)	3 (13%)	0.09
Diuretic	2 (9%)	3 (13%)	1.00
Vasodilator	1 (5%)	0	0.48
5 Times sit-to-stand, s	21.9 ± 13.5	8.6 ± 2.7	<0.001*
TBRS estimated V̇O2max, mL·kg−1·min−1	29.2 ± 9.4	32.3 ± 10.2	0.26
Workload during high intensity, W	79 ± 27	117 ± 36	<0.001*
Workload during active recovery, W	16 ± 2	18 ± 4	0.04*
Stroke characteristics
Months after stroke	31 ± 16
Right-sided stroke, n (%)	11 (50%)
Type of stroke (ischemic, hemorrhagic), n (%)	18 (82%), 4 (18%)
Stroke location, n (%)
MCA and territories	9 (41%)
ACA and territories	2 (9%)
PCA and territories	6 (27%)
Vertebrobasilar	2 (9%)
Combined MCA and ACA	3 (14%)
mRS, n (%)
No significant disability (1)	9 (41%)
Slight disability (2)	6 (27%)
Moderate disability (3)	7 (32%)
Fugl–Meyer Lower Extremity Score	23 ± 6

Values are means ± SD. ACA, anterior cerebral artery; BMI, body mass index; CON, age- and sex-matched adults; estimatedWatt_max, estimated maximal watts; MCA, middle cerebral artery; mRS, modified Rankin scale; PCA, posterior cerebral artery; TBRS, total body recumbent stepper submaximal exercise test; V̇O_2max, maximal oxygen consumption.*Significantly different between groups (Mann –Whitney U and chi-square tests).

Dynamic Cerebral Autoregulation during Spontaneous Fluctuations in MAP and MCAv

Twenty-two individuals had complete TFA data sets across all three time points, shown in Table 2. The reduction in TFA sample size is common and occurs because of reduced coherence within the VLF range (65), which could be attributed to nonlinearity (70).

Table 2.

Transfer function analysis of spontaneous fluctuations at rest

	Individuals Poststroke (n = 8)			CON (n = 14)			P Values
	BL	Immediately after HIIE	30 min after HIIE	BL	Immediately after HIIE	30 min after HIIE	Group	Time	Interaction
MCAv, cm/s	47 ± 5	45 ± 5	47 ± 6	48 ± 10	47 ± 8	50 ± 11	0.20	0.003*	0.73
MAP, mmHg	75 ± 12	70 ± 6	73 ± 7	81 ± 11	77 ± 9	82 ± 11	0.06	0.02*	0.74
$\text{P}{\text{ET}}_{{\text{CO}}_2}$	37 ± 5	36 ± 5	35 ± 5	36 ± 3	35 ± 3	34 ± 3	0.53	<0.001*	0.90
Very low frequency (0.02–0.07 Hz)
MAP PSD, mmHg2/Hz	5.58 ± 3.99	3.71 ± 1.68	6.46 ± 2.94	4.49 ± 2.28	4.97 ± 2.67	5.47 ± 2.21	0.72	0.11	0.23
MCAv PSD, cm2/s2/Hz	2.93 ± 1.75	3.13 ± 1.64	3.99 ± 1.63	3.60 ± 5.09	2.73 ± 2.50	3.04 ± 3.19	0.86	0.43	0.23
Coherence	0.62 ± 0.17	0.55 ± 0.09	0.67 ± 0.18	0.58 ± 0.14	0.52 ± 0.13	0.55 ± 0.18	0.26	0.12	0.33
Gain, cm/s/mmHg	0.66 ± 0.31	0.74 ± 0.25	0.73 ± 0.32	0.71 ± 0.34	0.62 ± 0.29	0.56 ± 0.24^	0.53	0.51	0.02*
Normalized gain, %/mmHg	1.38 ± 0.54	1.65 ± 0.58	1.52 ± 0.62	1.44 ± 0.42	1.30 ± 0.40	1.11 ± 0.36	0.23	0.14	0.01*
Phase,°	49.7 ± 14.8	43.3 ± 20.9	44.1 ± 18.1	59.9 ± 14.8	65.8 ± 18.5	55.8 ± 15.2	0.03*	0.31	0.17
Low frequency (0.07–0.20 Hz)
MAP PSD, mmHg2/Hz	2.61 ± 2.10	2.36 ± 2.79	3.36 ± 3.37	2.28 ± 2.33	3.91 ± 4.04	3.70 ± 3.72	0.67	0.26	0.36
MCAv PSD, cm2/s2/Hz	2.37 ± 2.00	2.92 ± 4.32	2.93 ± 3.97	1.91 ± 1.92	3.18 ± 3.57	2.34 ± 2.47	0.83	0.25	0.62
Coherence	0.75 ± 0.20	0.66 ± 0.20	0.76 ± 0.16	0.71 ± 0.18	0.73 ± 0.12	0.74 ± 0.17	0.96	0.32	0.35
Gain, cm/s/mmHg	0.91 ± 0.27	0.97 ± 0.34	0.91 ± 0.32	0.88 ± 0.28	0.90 ± 0.22	0.83 ± 0.26	0.61	0.21	0.76
Normalized gain, %/mmHg	1.92 ± 0.42	2.12 ± 0.57	1.88 ± 0.52	1.85 ± 0.49	1.93 ± 0.39	1.64 ± 0.35	0.36	0.01*	0.55
Phase,°	25.8 ± 5.5	26.7 ± 7.7	30.4 ± 9.2	32.4 ± 9.7	33.5 ± 10.4	33.5 ± 9.3	0.10	0.40	0.63

Values are means ± SD. BL, baseline; CON, age- and sex-matched adults; HIIE, high-intensity interval exercise; MAP, mean arterial pressure; MCAv, middle cerebral artery blood velocity; $\text{P}{\text{ET}}_{{\text{CO}}_2}$, end-tidal pressure of carbon dioxide; PSD, power spectral density. *Significantly different at P < 0.05; ^significantly different from BL (Wilcoxon signed-rank test).

Our primary analysis showed a significant and large group × time crossover interaction effect for VLF gain (P = 0.02, ${\eta }_{\mathrm{p}}^2$ = 0.18) and VLF normalized gain (P = 0.01, ${\eta }_{\mathrm{p}}^2$ = 0.43), with VLF gain and normalized gain increasing after HIIE in individuals poststroke and decreasing in CON, shown in Fig. 3. Our Bonferroni-corrected post hoc analysis showed significantly decreased VLF normalized gain 30 min after HIIE compared with BL (P = 0.001) in the CON group. However, there were no significant differences immediately after HIIE in VLF gain (P = 0.01) or VLF normalized gain (P = 0.02) or VLF gain 30 min after HIIE (P = 0.004) compared with BL in the CON group. Although average VLF gain and VLF normalized gain in individuals poststroke changed in the opposite direction of CON after HIIE, creating a crossover interaction, there were no significant differences between groups or within-group differences across time in individuals poststroke.

Figure 3.

Spontaneous dynamic cerebral autoregulation response following high-intensity interval exercise (HIIE) in individuals poststroke compared with control subjects (CON). ●, Individuals with chronic stroke; ○, age- and sex- matched control subjects (CON). A mixed-model ANOVA was used to determine group, time, and group × time interaction effects. BL, baseline; Post, immediately after HIIE; Post30, 30 min after HIIE; VLF, very low frequency. *Significantly different at P < 0.05.

Our primary analysis also showed there was a significant and large group effect for VLF phase (P = 0.03, ${\eta }_{\mathrm{p}}^2$ = 0.22), with VLF phase decreasing after HIIE in individuals poststroke and increasing in CON, shown in Fig. 3. However, post hoc Bonferroni-adjusted comparisons between groups showed no significant difference in VLF phase between groups immediately after HIIE (P = 0.01) and 30 min after HIIE (P = 0.17). There was also a linear time effect for LF normalized gain (P = 0.01). MAP PSD and MCAv PSD within the VLF and LF range were not different between groups or across time. Average MCAv, MAP, and $\text{P}{\text{ET}}_{{\text{CO}}_2}$ during spontaneous dCA measurements had a significant effect across time but not between groups.

Our sensitivity analysis showed that when adjusting for reductions in $\text{P}{\text{ET}}_{{\text{CO}}_2}$ after HIIE the group × time interaction remained for VLF gain (P = 0.01) and VLF normalized gain (P = 0.004). The group effect for VLF phase (P = 0.046) also remained when adjusting for $\text{P}{\text{ET}}_{{\text{CO}}_2}$. However, the time effect for LF normalized gain (P = 0.56) was no longer significant when adjusting for $\text{P}{\text{ET}}_{{\text{CO}}_2}$. The TFA across the frequency ranges (0.02–5 Hz) in individuals with chronic stroke compared with CON are shown in Fig. 4.

Figure 4.

Transfer function analysis in individuals poststroke compared with control subjects (CON) after high-intensity interval exercise (HIIE). Solid black line, individuals with chronic stroke; dotted line, age- and sex-matched adult control subjects (CON). a.u., Arbitrary units.

Sit-to-Stand Measurements of Dynamic Cerebral Autoregulation

There were no significant group or group by time effects on the sit-to-stand measurements of dCA after HIIE in individuals poststroke compared with CON, shown in Table 3. When adjusting for seated $\text{P}{\text{ET}}_{{\text{CO}}_2}$ after HIIE, there were still no group or group × time effects for TD onset, ROR, %ΔMCAv, %ΔMAP, or %ΔMCAv/%ΔMAP. There were no significant differences between groups in seated and standing MCAv, MAP, CVCi, or $\text{P}{\text{ET}}_{{\text{CO}}_2}$, shown in Table 3. However, there was a significant effect of time for seated and standing resting CVCi (P = 0.03). Standing MCAv also had a significant effect of time (P < 0.001), with both groups having an average decrease in standing MCAv of 1–2 cm/s after HIIE compared with BL. Seated and standing $\text{P}{\text{ET}}_{{\text{CO}}_2}$ also decreased after HIIE in both groups.

Table 3.

Sit-to-stand dynamic cerebral autoregulation metrics

	Individuals Poststroke (n = 22)			CON (n = 23)			P Values
	BL	After HIIE	35 min after HIIE	BL	After HIIE	35 min after HIIE	Group	Time	Interaction
Seated MCAv, cm/s	42 ± 11	42 ± 11	42 ± 11	50 ± 15	49 ± 14	49 ± 13	0.052	0.07	0.66
Seated MAP, mmHg	84 ± 12	82 ± 10	82 ± 8	82 ± 11	86 ± 11	85 ± 11	0.52	0.49	0.08
Seated CVCi, cm/s/mmHg	0.52 ± 0.16	0.51 ± 0.14	0.52 ± 0.15	0.61 ± 0.16	0.57 ± 0.17	0.59 ± 0.16	0.10	0.03*	0.21
Seated $\text{P}{\text{ET}}_{{\text{CO}}_2}$, mmHg	36 ± 5	34 ± 6	35 ± 4	36 ± 4	34 ± 4	34 ± 4	0.71	<0.001*	0.31
Standing MCAv, cm/s	41 ± 11	39 ± 11	41 ± 11	48 ± 14	47 ± 13	48 ± 13	0.60	<0.001*	0.53
Standing MAP, mmHg	84 ± 14	84 ± 14	84 ± 13	83 ± 13	85 ± 12	87 ± 12	0.74	0.32	0.26
Standing CVCi, cm/s/mmHg	0.50 ± 0.17	0.47 ± 0.14	0.49 ± 0.15	0.58 ± 0.15	0.55 ± 0.14	0.56 ± 0.16	0.11	0.03*	0.88
Standing $\text{P}{\text{ET}}_{{\text{CO}}_2}$, mmHg	36 ± 4	34 ± 4	35 ± 4	35 ± 3	33 ± 4	33 ± 4	0.27	<0.001*	0.51
TD onset, s	2.99 ± 2.05	2.88 ± 1.97	2.66 ± 2.00	3.00 ± 2.14	2.83 ± 1.86	2.99 ± 2.77	0.83	0.90	0.87
ROR	0.005 ± 0.009	0.026 ± 0.080	0.005 ± 0.006	0.003 ± 0.003	0.003 ± 0.003	0.003 ± 0.003	0.11	0.21	0.21
%ΔMCAv	12.1% ± 10.0%	16.9% ± 8.2%	14.9% ± 9.7%	15.3% ± 10.1%	15.9% ± 11.6%	16.0% ± 10.1%	0.66	0.17	0.32
%ΔMAP	16.5% ± 8.5%	18.0% ± 7.7%	15.5% ± 9.9%	18.7% ± 8.9%	19.8% ± 10.8%	18.5% ± 9.6%	0.35	0.22	0.88
%ΔMCAv/%ΔMAP	0.64 ± 0.61	1.06 ± 1.12	0.87 ± 1.13	0.77 ± 0.47	0.80 ± 0.63	0.85 ± 0.50	0.80	0.19	0.30

Values are means ± SD. BL = baseline; CON, age- and sex-matched adults; HIIE, high-intensity interval exercise; MAP, mean arterial pressure; MCAv, middle cerebral artery blood velocity; $\text{P}{\text{ET}}_{{\text{CO}}_2}$, end-tidal pressure of carbon dioxide, ROR, rate of regulation; TD, time delay; %Δ, percent change from baseline to nadir upon standing. *Significantly different at P < 0.05.

Figure 5 shows the CVCi response during the sit-to-stand at BL in individuals poststroke compared with CON. Monitoring of beat-to-beat MAP during the sit-to-stand allows for the detection of initial orthostatic hypotension (61), which differs from the clinical thresholds of delayed orthostatic hypotension (systolic >20 mmHg and/or diastolic >10 mmHg) taken after 3 min of standing (71). We report that no individuals met the criteria of initial orthostatic hypotension (having both decreases in arterial blood pressure and symptoms of cerebral hypoperfusion) (61) during our study. However, after HIIE three CON individuals met the criteria for initial drops in arterial blood pressure upon standing but reported no symptoms of dizziness. At 35 min after HIIE, two different CON individuals met the criteria for an initial drop in arterial blood pressure upon standing. Although no individuals poststroke met the arterial blood pressure thresholds of drops of >40 mmHg systolic or >20 mmHg diastolic upon standing, one individual poststroke reported slight dizziness upon standing at baseline (lasting <3 s) and another individual poststroke reported slight dizziness upon standing (lasting <3 s) after HIIE.

Figure 5.

Baseline sit-to-stand measure of dynamic cerebral autoregulation: mean and SD lines. Gray line, individuals poststroke (n = 22); black line, age- and sex-matched control subjects (CON, n = 23). CVCi, cerebrovascular conductance index [middle cerebral artery blood velocity(MCAv)/mean arterial pressure (MAP)]. Vertical black line marks the moment of arise-and-off (AO) detected by the custom force sensor as the moment individuals stood up. Arrows indicate time delay (TD) of the regulation response before an increase in CVCi. Square box shows the slope of the rate of regulation (ROR). Resampled to 1 Hz for group averages.

DISCUSSION

Our findings collectively show that dCA was lower during spontaneous fluctuations in MAP and MCAv following HIIE in individuals poststroke. Although various reviews have outlined the importance of examining the dCA response to HIIE in individuals after stroke (16–18), our study is the first to show that 1) VLF gain and normalized gain had a crossover interaction effect in which CON individuals had a decrease and individuals poststroke had an increase in VLF gain and VLF normalized gain after HIIE compared with BL and 2) VLF phase had a group effect in which individuals poststroke had lower VLF phase after HIIE compared with CON. However, with a statistically rigorous Bonferroni correction applied to the post hoc analysis, all significant differences were lost except for decreased VLF gain at 30 min after HIIE compared with BL in the CON group. Reductions in $\text{P}{\text{ET}}_{{\text{CO}}_2}$, due to continued hyperventilation after HIIE (22), only accounted for the significant effect of time on LF normalized gain. Therefore, differences in breathing recovery (i.e., $\text{P}{\text{ET}}_{{\text{CO}}_2}$) (22) that could influence vascular tone (34, 35) between individuals poststroke and CON did not account for the change in VLF phase and gain following HIIE. Interestingly, although dCA was lower after HIIE during spontaneous MAP and MCAv fluctuations in individuals poststroke, our study did not find differences in the sit-to-stand dCA response between individuals poststroke and CON, even when adjusting for $\text{P}{\text{ET}}_{{\text{CO}}_2}$. Therefore, dCA in individuals poststroke may maintain the ability to respond to a single drop in MAP upon standing, after 6 min of seated rest recovery following HIIE.

Spontaneous Dynamic Cerebral Autoregulation

Dynamic cerebral autoregulation at baseline was not different between individuals with chronic stroke and healthy adults. This is consistent with previous literature reporting spontaneous fluctuation of dCA being impaired within 48 h after stroke and recovering within 3 mo (72). Therefore in chronic stroke (>6 mo), the cerebrovascular system may need to be hemodynamically challenged (63) to show differences in dCA compared with healthy control subjects, such as with an acute bout of HIIE. We show that after HIIE dCA during spontaneous MAP and MCAv fluctuations was lower in individuals poststroke compared to CON, with lower VLF phase and higher VLF gain and VLF normalized gain. Lower VLF phase following HIIE in individuals after stroke can be interpreted as dCA being less efficient at buffering the cerebrovascular system from changes in peripheral blood pressure (73). The crossover interaction in VLF gain and normalized gain shows that whereas CON individuals had improvements in the ability of the cerebrovasculature to dampen the amplitude changes in MAP fluctuations (73) following HIIE, individuals poststroke did not. Our findings of a reduced VLF gain and VLF normalized gain in CON following HIIE (during decreases in $\text{P}{\text{ET}}_{{\text{CO}}_2}$) are supported by prior studies showing improvements in the dCA response in healthy adults during decreases in $\text{P}{\text{ET}}_{{\text{CO}}_2}$, due to increased vascular arteriole tone (34, 36). Even when adjusting for changes in $\text{P}{\text{ET}}_{{\text{CO}}_2}$, the differences in dCA between groups after HIIE were maintained within the VLF range.

Lower dCA in individuals poststroke, quantified via TFA after HIIE, could be due to dCA’s sensitivity to decreasing MAP (22, 26, 74). Previous studies have shown that dCA at rest (23) and during repetitive squat-stands (25, 26, 74) was less efficient at regulating decreases in MAP in comparison with MAP increases, and HIIE may exacerbate this directional sensitivity in healthy adults (75). Whether individuals poststroke have lesser dCA sensitivity to decreasing MAP compared with CON is unknown; however, this could potentially explain the differences between groups found after HIIE. However, when forcing arterial blood pressure to decrease with a single sit-to-stand after HIIE, we did not find any differences in the cerebral pressure-flow relationship in individuals poststroke.

Previous studies have also shown that arterial blood pressure medications such as beta-blockers and calcium channel blockers may attenuate dCA (76–79). Our study reports that individuals poststroke had greater beta-blocker use and trended toward greater calcium channel blocker use compared with CON. We did not directly measure the effects of arterial blood pressure medications on spontaneous dCA. Within the TFA sample, only two individuals poststroke were taking beta-blockers, potentially contributing to the lower dCA response in individuals poststroke after HIIE. Absolute workload during HIIE was also not statistically accounted for, as it was unlikely to have contributed to the dCA differences between groups. One would hypothesize that a greater absolute workload would reduce dCA, but we found that individuals poststroke who worked at a lower workload had reduced dCA.

Although our study is the first to show a potentially attenuated dCA response during spontaneous fluctuations of MAP and MCAv following HIIE, our previous work reported that individuals poststroke had a lower absolute MCAv response during HIIE and a reduced MCAv responsiveness [calculated with MCAv coefficient of variation (CoV)] when switching between high-intensity and active recovery compared with CON (22). Our present analysis showing attenuated spontaneous dCA following HIIE in individuals poststroke may be due to differences in regulation during passive recovery (e.g., spontaneous MAP oscillations) compared with during exercise (e.g., large transient MAP oscillations) (22, 80). Although regulatory mechanisms are not mutually exclusive, individuals poststroke had both a blunted MCAv responsiveness during HIIE (22) and attenuated dCA during recovery, based on the predominating regulatory system at the time (80, 81). During exercise, there are many overlapping regulatory mechanisms that may play a larger role in maintaining cerebral blood flow compared with rest, including cerebrovascular reactivity to carbon dioxide, neurovascular coupling with increased cerebral metabolism, increased sympathetic activity, and greater cardiac output (81). A previous study hypothesized that during exercise “fluctuations in arterial pressure with each muscle contraction can be too rapid to be countered by [cerebral autoregulation]” (82), and the 1-min intervals of HIIE limit the ability to examine dCA with TFA [recommended to have >5-min recording for TFA (65, 66)]. However, during exercise recovery, dCA may play a larger role in the spontaneous fluctuations in MAP and MCAv (within the VLF range), through myogenic vasomotor activity (83, 84).

Previous studies reporting impaired dCA at rest in individuals after stroke (85) were conducted during the acute stage (1–7 days) and subacute stage (7 days to 6 mo) after stroke (8, 86, 87). Reviews postulating the timeline of dCA after stroke suggest recovery within 1–3 mo (8, 12). Our study is supported by these reviews and found no difference in resting BL dCA between individuals 31 ± 16 mo after stroke and their age- and sex-matched peers. To our knowledge, only one other study has examined dCA in the chronic stage (≥6 mo) after stroke and reported decreased resting phase (within 0.06–0.12 Hz) at exactly 6 mo after stroke compared with healthy control subjects (88). The discrepancy between their findings and ours could be due to many differences including their inclusion criteria of more heavily involved lacunar infarct strokes, supine versus seated rest recordings, TFA frequency ranges utilized, or the chronicity of stroke (88). Although our study included individuals with mild to moderate stroke severity, it is possible that dCA at rest during spontaneous arterial blood pressure oscillations may continue to recover within the chronic stage with greater stroke severity and involvement (85).

Sit-to-Stand Dynamic Cerebral Autoregulation Response

There were no differences between individuals poststroke and CON in the sit-to-stand dCA response at BL and after HIIE. Despite decreased seated and standing CVCi after HIIE (potentially due to continued hyperventilation and decreased $\text{P}{\text{ET}}_{{\text{CO}}_2}$ after HIIE causing arteriole vasoconstriction), the dCA sit-to-stand response did not change across time. We did not find a significant change in the TD before the onset of dCA or the ROR response in individuals poststroke after HIIE compared with CON. Therefore, the temporal response of CVCi during a sit-to-stand was maintained after HIIE in individuals poststroke and CON, even when adjusting for changes in $\text{P}{\text{ET}}_{{\text{CO}}_2}$. To our knowledge no prior studies have examined the effects of hyperventilation and changes in arteriole tone on the temporal dCA response to a single sit-to-stand. A previous study in healthy adults has shown that the initial dCA response during a sit-to-stand (occurring 2 ± 2 heartbeats after standing) may be due in large to compliance of the cerebrovasculature, which occurs before cerebral vasodilation (occurring 11 ± 3 heartbeats after standing) (89). Although not measured within the study, cerebrovascular compliance may play a role in the preserved dCA sit-to-stand response after HIIE in individuals after stroke and aging adults. We also did not show a significant difference in %ΔMCAv/%ΔMAP after HIIE. These findings may suggest that the cerebrovascular system in individuals poststroke did not become more pressure passive after HIIE, and changes in MAP when standing up from a seated position resulted in similar concordant changes in MCAv (90). However, when interpreting the dCA sit-to-stand findings, one must also consider that the participants had a notable amount of seated recovery (∼6 min) before the dCA sit-to-stand measure. Therefore, future studies should examine the sit-to-stand response immediately after aerobic exercise, without seated rest, to determine the stability of the dCA response and how long someone should wait before standing up.

Clinical Importance

Within the clinic, traditionally only peripheral blood pressure is used to determine hemodynamic stability during exercise recovery. However, our study shows the need to also examine the cerebrovascular system in individuals after stroke. Although the clinical implications for a lower dCA response during passive recovery following HIIE are unknown, clinicians may want to continue practicing the recommended postexercise precautions within the first 5 min after HIIE (18). Although our study included a 2-min cooldown after HIIE, a longer cooldown ≥ 5 min may allow for a more gradual ramping down of arterial blood pressure and may minimize the effect on dCA (18). Similarly, the dCA response to passive recovery intervals during HIIE, which separate bouts of high-intensity and cause an abrupt stop in exercise, should also be examined.

When waiting ≥6 min after HIIE to stand up from the seated position, individuals poststroke did not differ in the dCA response compared with BL or CON. However, two individuals poststroke did report slight dizziness upon standing, despite not meeting the peripheral blood pressure criteria for initial orthostatic hypotension. Symptomatic thresholds associated with drops in MCAv during sit-to-stands following exercise should be established (91) in individuals after stroke to better understand the cerebrovascular response rather than making inferences of “hypoperfusion” from peripheral blood pressure.

Methodological Considerations

The results of this study cannot be generalized to the entire population of individuals after stroke, such as individuals with more severe impairments. Our findings within the chronic stage of stroke also cannot be generalized to individuals within the acute and subacute stages, where dCA has been shown to be impaired at baseline (85). Although TCD is currently the best method to capture beat-to-beat MCAv for TFA, one must assume a constant MCA diameter to use MCAv as a measure of MCA blood flow. Although we could not directly measure MCA diameter during the study, a previous study using four-dimensional (4-D) flow magnetic resonance imaging in older adults showed no significant change in diameter during an 8-Torr change in $\text{P}{\text{ET}}_{{\text{CO}}_2}$ (92). Our sample size to quantify TFA metrics was also reduced because of low coherence, potentially due to nonlinearity of spontaneous fluctuations (70). Although we followed the CARNet recommendation of a 5-min recording (65), to increase coherence via nonlinear modeling requires longer recordings (∼40 min) of spontaneous fluctuations in MAP and MCAv (70). The limited amplitude of spontaneous arterial blood pressure oscillations (i.e., low signal-to-noise ratio) could also lead to less reproducible estimations of TFA metrics (93). dCA is also a nonstationary phenomenon (i.e., not constant over time), which could also affect the reproducibility of TFA metrics (94). This protocol could be repeated using TFA on forced blood pressure oscillations (e.g., using repeated squat-stands or sit-stands), which would increase the input (e.g., arterial blood pressure) power and improve the linear interpretability of TFA metrics (93). The ability to perform repeated squat-stands for at least 3–5 min (95) to measure the dCA response in individuals after stroke with decreased lower extremity strength and greater fatigue is still unknown. However, repeated squat-stands to measure the dCA response have been shown to be feasible in some clinical populations such as pulmonary arterial hypertension (96). Measuring the dCA response via a single sit-to-stand was chosen because of its real-life applicability as well as accounting for neuromuscular fatigue after HIIE in individuals poststroke. Although our sit-to-stand methodology using a force sensor accounted for the exact time in which individuals stood from the chair, we did not use an accelerometer to account for the speed of standing. During the sit-to-stand, we were also unable to account for other potential confounding factors including neurovascular coupling, sympathetic activity, and cardiac output. Therefore, the preservation of the hemodynamic response during a single sit-to-stand after HIIE could be due to dCA as well as additional overlapping regulatory mechanisms. Since accumulating evidence suggests directional sensitivity of the cerebral pressure-flow relationships (23–27, 74, 75), further research quantifying dCA while considering the direction of MAP changes will be necessary to better understand whether the reported attenuated dCA after HIIE is in response to reductions in MAP, increases in MAP, or both. Our primary statistical analysis performed unadjusted mixed-model ANOVAs; therefore, our post hoc statistical analysis adjusted for multiple comparisons with a rigorous Bonferroni correction. The small sample size, compounded by the rigorous application of a Bonferroni correction, led to a notable attenuation of statistical significance in the findings. Future studies with a larger sample should also take into consideration stroke hemispheric differences, large- versus small-vessel occlusion, severity of stroke, and differences between ischemic and hemorrhagic stroke.

Conclusions

We show that dCA during spontaneous fluctuations in MAP and MCAv was lower after HIIE in individuals poststroke compared with CON. However, the dCA response during a sit-to-stand after HIIE (after ∼6-min seated rest) did not show a significant difference compared with CON. These findings provide the groundwork for further characterization of the dCA response following aerobic exercise of varying intensities in individuals across the stages of stroke recovery. The dCA response to aerobic exercise and recovery is “an essential physiological consideration to protect the brain when progressing exercise intensity [poststroke]” (18). Future research should develop therapeutic strategies to improve dCA after exercise in individuals after stroke.

DATA AVAILABILITY

Data are available upon request to the corresponding author.

GRANTS

A.A.W. was supported by the National Heart, Lung, and Blood Institute (T32HL134643), Cardiovascular Center’s A.O. Smith Fellowship Scholars Program, Eunice Kennedy Shriver National Institute of Child Health and Human Development (T32HD057850), American Heart Association (898190), and Kansas Partners in Progress Inc. S.E.A. was supported by National Center for Advancing Translational Sciences (TL1TR002368). S.A.B. and E.D.V. were supported in part by the National Institute on Aging (P30 AG072973). REDCap at University of Kansas Medical Center was supported by National Center for Research Resources (ULTR000001). The REACH laboratory was supported by Georgia Holland Endowment.

DISCLAIMERS

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

DISCLOSURES

P. Brassard is an editor of Journal of Applied Physiology and was not involved and did not have access to information regarding the peer-review process or final disposition of this article. An alternate editor oversaw the peer-review and decision-making process for this article. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

A.A.W., M.C., E.D.V., S.M.E., and S.A.B. conceived and designed research; A.A.W., S.E.A., J.B., K.N., and S.W. performed experiments; A.A.W. and R.N.M. analyzed data; A.A.W., S.E.A., M.C., P.B., E.D.V., R.N.M., and S.A.B. interpreted results of experiments; A.A.W. prepared figures; A.A.W., P.B., R.N.M., and S.A.B. drafted manuscript; A.A.W., S.E.A., M.C., P.B., J.B., K.N., E.D.V., S.W., S.M.E., R.N.M., and S.A.B. edited and revised manuscript; A.A.W., S.E.A., M.C., P.B., J.B., K.N., E.D.V., S.W., S.M.E., R.N.M., and S.A.B. approved final version of manuscript.