This study was carried out according the Ethical Principles in Animal Research of the Brazilian Council for Control of Animal Experimentation (CONCEA), in compliance with the ARRIVE guidelines. Surgical procedures and experimental protocols were reviewed and approved by the Institutional Animal Care and Use Committee of the Biomedical Sciences Institute, University of Sao Paulo (CEUA, protocol number 93/2017).

Chronically hypertensive male SHR (12-weeks old) and normotensive controls (Wistar) were housed in the Animal Facilities of Department of Physiology and Biophysics, University of São Paulo under controlled temperature/humidity, 12/12-h light/dark cycle with free access to standard chow and water. Rats body weight was measured throughout the experiment. During a 2-week adaptation period rats were pre-selected for their ability to walk/run on a treadmill (0.4–0.8 km/h, 0% grade, 10 min/day, Millenium, Inbramed, Porto Alegre, Brazil). Only active rats were included in this study. Rats were then submitted to progressive exercise tests (MET (Cavalleri et al., 2011; Fragas et al., 2021) to determine maximal individual aerobic capacity, to allocate rats with identical capacities to trained (T) and sedentary (S) groups and to set the intensity of aerobic training (T = 50–60% of maximal capacity, performed 1 h/day, 5 days/week for 4 week). Previous studies confirmed that 4 weeks of exercise training were enough to restore BBB permeability and induce cardiovascular adjustments in SHR (Masson et al., 2014; Buttler et al., 2017; Fragas et al., 2021). S groups were handled every day to approximate their conditions to those experienced by T groups. MET was performed again at the end of experimental protocols to evaluate the aerobic capacity of T and S groups (Fragas et al., 2021). SHR and Wistar rats were then pre-anesthetized with acepromazine (2.5 mg/kg ip, Syntec, Santana de Parnaiba, São Paulo, Brazil) followed by ketamine (80 mg/kg ip, Fort Dodge IA, United States of America) + xylazine (12 mg/kg ip, Alcon, Fort Worth TX, United States of America) for catheterization of the femoral artery. Rats were treated with penicillin (24,000IU/kg im, Pentabiotico Veterinario, Fontoura Wyeth, Brazil) and ketoprofen (2 mg/kg sc, Biofarm, Jaboticabal, Brazil) and return to their home cages for recovery, which extended for at least 24–30 h.

Baseline arterial pressure (AP) and heart rate (HR) were continuously acquired on a beat-to-beat basis (50–60 min, LabChart Pro, ADInstruments, sampling frequency of 2,000Hz) in conscious unrestrained rats resting in their home cages (Ichige et al., 2016). Time series of systolic AP (SAP) and pulse interval (PI) were used to evaluate pressure and HR variabilities at the frequency domain (Buttler et al., 2017; Fragas et al., 2021). Power spectral density for the low frequency (LF, 0.20–0.75 Hz, indicative of sympathetic vasomotor activity and sympathetic + parasympathetic activity to the heart), the high frequency (HF, 0.75–3.00 Hz, indicative of cardiac vagal modulation) and very low frequency (VLF, <0.20 Hz, suggestive of hormonal modulation) were evaluated (Masson et al., 2014). LF/HF ratio to the heart and spontaneous baroreflex sensitivity were also calculated.

After hemodynamic recordings a subgroup of SHR-S, SHR-T, Wistar-S, and Wistar-T was again anesthetized for catheterization of right carotid artery. A mixture of dyes [rhodamine isothiocyanate dextran, 70 kDa (RHO) and fluorescein isothiocyanate dextran 10 kDa (FITC], Sigma-Aldrich) was slowly administered and allowed to recirculate as previously described (Biancardi et al., 2014). Rats received then and overdose of anesthesia (300 mg/kg ketamine +60 mg/kg xylazine ip) for brain harvesting immediately after the respiratory arrest. Brains were post-fixed (4% phosphate-buffered paraformaldehyde, 48 h), cryoprotected (30% sucrose in PBS for 72 h) and stored until processing (Buttler et al., 2017; Rocha-Santos et al., 2020).

Sequential coronal PVN, NTS and RVLM slices (30 μm, Leica CM1850 cryostat, Germany) were collected and mounted in gelatinized slides as previously described (Buttler et al., 2017). The BBB permeability was analyzed by the quantitative assessment of intravascular and extravascular dyes according the technique developed by Biancardi et al. (2014). With an intact BBB both dyes are colocalized within brain capillaries; in the presence of compromised barrier integrity the large-size dye are still contained by the capillaries whereas the small-size dye partially leaks into the brain parenchyma (Biancardi et al., 2014). Tissues were examined by a blind observer on a fluorescent microscope (Leica BMLB, Nussloch, Germany) attached to an Exiblue camera (Imaging, Canada). Selected images were acquired by Image-Pro Plus software (Media Cybernetics, United States) and quantified by the ImageJ software (NIH, United States).

Another subgroup of SHR-S, SHR-T, Wistar-S, and Wistar-T received, after the functional measurements, an overdose of ketamine + xylazine. Immediately after the respiratory arrest, the thorax was opened and the left ventricle cannulated for sterile saline perfusion (∼30 mL/min, Daigger pump, Vernon Hills IL United States) followed by modified Karnovsky solution (2.5% glutaraldehyde +2% paraformaldehyde in 0.1 M PBS, pH 7.3). Brain was removed and placed on a coronal brain matrix (72–5029, Harvard Apparatus) to obtain hypothalamic and brainstem slices. PVN, NTS, and RVLM nuclei were microdissected with the aid of a magnifying lens, using as anatomic markers the third ventricle and optic chiasma, the central canal and 4th ventricle, and, the nucleus ambiguous, raphe obscurus and inferior olive, respectively. The nuclei were immersed in a 2.5% glutaraldehyde solution for 2 h, washed in PBS and post-fixed in a 2% osmium tetroxide solution for 2 h at 4°C. Tissues were then stained overnight with uranyl acetate, dehydrated in 60% up to 100% ethanol series and immersed in pure resin. Semi-thin slices (400 nm, ultra-microtome Leica EMUC6) were obtained, placed in glass slides and stained with Toluidine Blue in order to select adequate areas for further processing. Ultra-thin slices (60 nm) were obtained with diamond knife, contrasted with 4% uranyl acetate and 0.4% lead acetate and disposed in 200 copper mesh screens.

Transverse sections of PVN, NTS, and RVLM capillaries of the 4 experimental groups were acquired in a transmission electron microscope (FEI Tecnai G20, 200 KV) and analyzed by a blind observer using the ImageJ software. The following parameters were analyzed in 9–11 capillaries/area/rat, 3 rats/experimental group: luminal and abluminal perimeter, lumen diameter, area of the endothelial cell, thickness of the basement membrane, pericytes’ coverage of capillaries, extension of capillary border between adjacent endothelial cells, the occurrence/extension of tight junctions, and, the counting of transcellular vesicles/capillary. To avoid the inclusion of non-transcytotic vesicles such as lysosomes, endosomes, peroxisomes, only the vesicles being formed at the luminal, and abluminal membranes were counted. Vesicle counting was expressed as number/capillary. Using the zoom to expand acquired images, the whole extension of capillaries was analyzed.

Some rats of each experimental group were used to confirm our counting of transcytotic vesicles by means of caveolin-1 expression. Caveolin-1, the main component of transcellular vesicle membranes (Predescu et al., 1997; Zhao et al., 2014) was used as a marker of transcytosis. After the functional measurements rats received an overdose of ketamine + xylazine. Immediately after the respiratory arrest brains were harvested, post-fixed, cryoprotected and stored as previously described (Fragas et al., 2021).

For caveolin-1 immunofluorescence assay coronal PVN sections (30 μm) were collected (Fragas et al., 2021). Briefly, free-floating sections were incubated with a mixture of primary antibodies (rabbit anti-caveolin-1, Cell Signaling 1:100 dilution + mouse-anti-endothelial cell antibody RECA-1, Abcam, 1:800 dilution for 48 h at 4°C) followed by incubation with secondary antibodies (anti-rabbit Alexa Fluor 488 + anti-mouse Alexa Fluor 549, Jackson ImmunoResearch, 1:500 dilution each, at room temperature for 1 h). Sections, mounted in gelatinized slides, were examined by a blind observer in a fluorescence microscope (Axioimager AI, Zeiss, Munchen, Germany attached to a Zeiss Axiocam 512 camera). Caveolin-1 immunofluorescence was normalized by RECA-1 immunoreactivity, both signals being acquired in the same ROI. Images were acquired with identical acquisition settings and analyzed as previously described (Rocha-Santos et al., 2020). Values of several PVN slices (expressed as integrated density) were averaged to yield a mean value/rat.

Some rats of each group were, at the end of functional experiments, euthanized with an overdose of ketamine + xylazine. Immediately after the respiratory arrest the thorax was opened and the heart removed to isolate the left ventricle which was weighed in a semi-analytical scale (Micronal B400, SP, Brazil). The ratio of left ventricle/body weight was used as an index of cardiac hypertrophy.

Data are presented as means ± SEM and submitted to Shapiro-Wilk homogeneity variance test. Treadmill performance and body weight in S and T SHR and Wistar rats were analyzed by three-way ANOVA with repeated measurements (time). Differences in hemodynamic and autonomic parameters, BBB permeability, ultrastructural changes in BBB constituents, and caveolin-1 expression between groups and conditions were analyzed by two-way factorial ANOVA. Tukey was the post hoc test. Correlations analyses used the Pearson statistics. All statistical analyses were performed by the GraphPad Prism 8 software. Differences were considered significant at p < 0.05.

SHR had smaller body weight and better treadmill performance than respective age-matched controls during the experimental protocols (Table 1). SHR-S, SHR-T, Wistar-S and Wistar-T exhibited similar body weight gain during the 4-week protocols. Both trained groups showed a significant and similar performance gain after 4 weeks of daily exercise while sedentary groups exhibited a small decrease (SHR-S) or no change (Wistar-S) (Table 1).

At the end of protocols MAP (+56%), SAP variability (+95%), vasomotor and cardiac sympathetic activities (+59% and +81% for LF-SAP and LF-PI, respectively) and hormonal modulation (VLF-SAP, +134%) were markedly increased in SHR-S vs. Wistar-S (Table 1). In contrast, there were significant decreases in PI variability (−43%) and parasympathetic modulation of the heart (HF-PI, −36%) in such a way that cardiac sympathovagal balance was augmented and spontaneous baroreflex sensitivity reduced in SHR-S vs. Wistar-S (Table 1). These changes were also accompanied by significant cardiac hypertrophy (2.74 ± 0.13 and 1.84 ± 0.04 mg/g for SHR-S and Wistar-S, respectively). On the other hand, SHR-T exhibited reduced MAP and resting HR (−9% and −8%) which were accompanied by large reductions in SAP variability and hormonal modulation and an almost normalized autonomic control of the circulation (data on Table 1). Spontaneous baroreflex sensitivity was increased in SHR-T vs. SHR-S, even in the presence of maintained cardiac hypertrophy (SHR-T = 2.89 ± 0.16 mg/g). Wistar rats responded to training with resting bradycardia and a mild increase in spontaneous baroreflex sensitivity (Table 1).

BBB permeability was evaluated in anesthetized rats by the amount of FITC leakage into the brain parenchyma after hemodynamic and autonomic recordings. Hypertension was accompanied by marked BBB dysfunction, characterized by a huge leakage into the three autonomic areas (11.44 ± 0.61, 10.52% ± 1.02% and 8.31% ± 1.09% area for the PVN, NTS, and RVLM, respectively, corresponding to increases of 3.3-, 1.8- and 1.5-fold when compared with respective Wistar-S controls, Figure 1). Notice that exercise training not only reduced, but completely normalized BBB leakage within the three autonomic areas. Mild non-significant FITC leakage reductions were observed in Wistar-T vs. Wistar-S rats (Figure 1).

To uncover the mechanisms conditioning hypertension- and training-induced BBB permeability changes we analyzed, in the experimental groups, the ultrastructure of brain capillaries within autonomic areas (Table 2). Hypertension was accompanied by significant increases in capillaries’ lumen diameter within the three areas analyzed (+31%, +15% and +17%, SHR-S vs. Wistar-S, for PVN, NTS, and RVLM, respectively, p < 0.05), a change that was corrected by exercise training in the NTS and RVLM, but not in the PVN whose capillaries’ lumen were still distended in the SHR-T (Table 2). Except for the RVLM that exhibited a reduction (−19% in SHR-S vs. Wistar-S, p < 0.05), the basement membrane thickness was not changed by hypertension. Exercise training increased capillaries’ membrane thickness only in the NTS of Wistar rats, with no changes within autonomic areas of the other groups (Table 2).

Luminal and abluminal vesicle number within the endothelium of PVN capillaries was increased by hypertension, but reduced after exercise training in both groups, with a large effect in hypertensive rats (Figure 2A). Quantitative data confirmed this observation showing a 71% augmentation in vesicles/capillary in SHR-S vs. Wistar-S (Figure 2B). Interestingly, the vesicle number was greatly reduced by training in SHR-T, showing a value similar to those exhibited by normotensive rats. The smaller reduction in vesicle number observed in Wistar-T vs. Wistar-S did not attain significance (Figure 2B). Similar hypertension- and training-induced effects on vesicle number were observed in both NTS and RVLM capillaries (Figures 2C, D; Supplementary Figure S1). It was also observed that the increased transcytosis within autonomic brain areas of the SHR-S was accompanied by a larger endothelial cell area, an effect that was reversed by exercise training (Table 2).

To corroborate the electron microscopy measurements of transcellular, not other cytoplasmatic vesicles, we analyzed the effects of hypertension and exercise on caveolin-1 expression, the main component of caveolae-mediated transcytosis. Within PVN capillaries (marked by RECA-1 antibody) caveolin-1 density was 61% higher in SHR-S (vs. Wistar-S), but largely reduced in SHR-T, attaining a value similar to those exhibited by normotensive groups (Figures 3A, B). Exercise training caused only a slight not significant reduction of caveolin-1 density in Wistar-T vs. Wistar-S.

The occurrence, density and extension of TJs within capillary borders were also quantified as indexes of possible hypertension- and exercise-induced changes in the paracellular transport. Within the three areas analyzed, the number of TJ/capillary (on average 1.04 ± 0.09/capillary) was not changed by hypertension or exercise training. There were slight changes in capillary border extension and TJs extension within the three areas analyzed (values on Table 2). To analyze the effects of hypertension and exercise on TJ it is important to consider both, that is TJ extension to capillary border extension ratio. Figure 4A depicts representative photomicrographs of PVN capillaries of the four experimental groups comparing the percentage of occupancy of the capillary border by the TJ. Quantitative data (Figure 4B) confirmed that TJ extension/capillary border extension was not changed by hypertension within the 3 autonomic areas. In contrast, trained exhibited significant increases in TJ occupancy of the capillary border within the PVN (Wistar-T = +78%, SHR-T = +43%, Figures 4A, B) and NTS (Wistar-T = +44%. SHR-T = +72%, Figure 4C; Supplementary Figure S2), with a slight not significant increase in the RVLM (Wistar-T = +37%, SHR-T = +36%, Figure 4D; Supplementary Figure S2).

Consequently, TJ extension/capillary border was not changed by hypertension, but significantly increased in both trained groups Quantitative data (Figure 4B) confirmed that exercise training greatly increased TJ/capillary border by 78% and 43% (Wistar-T and SHR-T, vs. Wistar-S and Wistar-S, respectively), without significant change in SHR-S compared to Wistar-S. Similar hypertension- and exercise-induced effects were observed in the NTS and RVLM (Table 2; Figures 4C, D).

Ultrastructural analyses of pericytes length in relation to capillary abluminal perimeter revealed that except for the RVLM (Figure 5C), exercise training increased pericytes’ capillary coverage within the PVN and NTS of normotensive rats, an effect that was abrogated by hypertension (Table 2; Figures 5B, C).

It is important to note that, within the three autonomic areas, BBB permeability changes showed strong positive correlations with vesicle number/capillary. While hypertension-induced augmentation of vesicles counting was accompanied by increased BBB leakage, training-induced reduction was accompanied by decreased BBB permeability within the PVN, NTS, and RVLM (Figures 6A, C, E, respectively). Notice that, except for TJ extension, exercise training did not change BBB leakage, vesicle counting and autonomic function in normotensive groups. On the other hand, the mild negative correlations between tight junctions’ extension/capillary border (exercise dependent but hypertension-independent) and BBB leakage observed in the PVN, NTS, and RVLM did not attain significance (Figures 6B, D, F).

Since the increased vesicles ‘counting, the augmented BBB leakage and the autonomic dysfunction were both hypertension- and exercise-dependent, we also correlated the observed BBB permeability changes within the PVN, NTS, and RVLM with recorded changes in SAP and PI variabilities and their spectral components. In all autonomic areas, BBB leakage changes exhibited strong positive correlations with both SAP variability and sympathetic vasomotor activity changes (regression equations, r and p values on Table 3). On the other hand, BBB permeability was negatively correlated with both PI variability and parasympathetic modulation of the heart (Table 3).