Efficacy of Dynamics-based Features for Machine Learning Classification of Renal Hemodynamics.

Different machine learning approaches for analyzing renal hemodynamics using time series of arterial blood pressure and renal blood flow rate measurements in conscious rats are developed and compared. Particular emphasis is placed on features used for machine learning. The test scenario involves binary classification of Sprague-Dawley rats obtained from two different suppliers, with the suppliers' rat colonies having drifted slightly apart in hemodynamic characteristics. Models used for the classification include deep neural network (DNN), random forest, support vector machine, multilayer perceptron. While the DNN uses raw pressure/flow measurements as features, the latter three use a feature vector of parameters of a nonlinear dynamic system fitted to the pressure/flow data, thereby restricting the classification basis to the hemodynamics. Although the performance in these cases is slightly reduced in comparison to that of the DNN, they still show promise for machine learning (ML) application. The pioneering contribution of this work is the establishment that even with features limited to hemodynamics-based information, the ML models can successfully achieve classification with reasonably high accuracy.

Salt-Sensitive Hypertension, Renal Injury, and Renal Vasodysfunction Associated With Dahl Salt-Sensitive Rats Are Abolished in Consomic SS.BN1 Rats.

Background Abnormal renal hemodynamic responses to salt-loading are thought to contribute to salt-sensitive (SS) hypertension. However, this is based largely on studies in anesthetized animals, and little data are available in conscious SS and salt-resistant rats. Methods and Results We assessed arterial blood pressure, renal function, and renal blood flow during administration of a 0.4% NaCl and a high-salt (4.0% NaCl) diet in conscious, chronically instrumented 10- to 14-week-old Dahl SS and consomic SS rats in which chromosome 1 from the salt-resistant Brown-Norway strain was introgressed into the genome of the SS strain (SS.BN1). Three weeks of high salt intake significantly increased blood pressure (20%) and exacerbated renal injury in SS rats. In contrast, the increase in blood pressure (5%) was similarly attenuated in Brown-Norway and SS.BN1 rats, and both strains were completely protected against renal injury. In SS.BN1 rats, 1 week of high salt intake was associated with a significant decrease in renal vascular resistance (-8%) and increase in renal blood flow (15%). In contrast, renal vascular resistance failed to decrease, and renal blood flow remained unchanged in SS rats during high salt intake. Finally, urinary sodium excretion and glomerular filtration rate were similar between SS and SS.BN1 rats during 0.4% NaCl and high salt intake. Conclusions Our data support the concept that renal vasodysfunction contributes to blood pressure salt sensitivity in Dahl SS rats, and that genes on rat chromosome 1 play a major role in modulating renal hemodynamic responses to salt loading and salt-induced hypertension.

Autonomic and cholinergic mechanisms mediating cardiovascular and temperature effects of donepezil in conscious mice.

Donepezil is a centrally acting acetylcholinesterase (AChE) inhibitor with therapeutic potential in inflammatory diseases; however, the underlying autonomic and cholinergic mechanisms remain unclear. Here, we assessed effects of donepezil on mean arterial pressure (MAP), heart rate (HR), HR variability, and body temperature in conscious adult male C57BL/6 mice to investigate the autonomic pathways involved. Central versus peripheral cholinergic effects of donepezil were assessed using pharmacological approaches including comparison with the peripherally acting AChE inhibitor, neostigmine. Drug treatments included donepezil (2.5 or 5 mg/kg sc), neostigmine methyl sulfate (80 or 240 µg/kg ip), atropine sulfate (5 mg/kg ip), atropine methyl bromide (5 mg/kg ip), or saline. Donepezil, at 2.5 and 5 mg/kg, decreased HR by 36 ± 4% and 44 ± 3% compared with saline (n = 10, P < 0.001). Donepezil, at 2.5 and 5 mg/kg, decreased temperature by 13 ± 2% and 22 ± 2% compared with saline (n = 6, P < 0.001). Modest (P < 0.001) increases in MAP were observed with donepezil after peak bradycardia occurred. Atropine sulfate and atropine methyl bromide blocked bradycardic responses to donepezil, but only atropine sulfate attenuated hypothermia. The pressor response to donepezil was similar in mice coadministered atropine sulfate; however, coadministration of atropine methyl bromide potentiated the increase in MAP. Neostigmine did not alter HR or temperature, but did result in early increases in MAP. Despite the marked bradycardia, donepezil did not increase normalized high-frequency HR variability. We conclude that donepezil causes marked bradycardia and hypothermia in conscious mice via the activation of muscarinic receptors while concurrently increasing MAP via autonomic and cholinergic pathways that remain to be elucidated.

Pathophysiology of unilateral ischemia-reperfusion injury: importance of renal counterbalance and implications for the AKI-CKD transition.

Unilateral ischemia-reperfusion (UIR) injury leads to progressive renal atrophy and tubulointerstitial fibrosis (TIF) and is commonly used to investigate the pathogenesis of the acute kidney injury-chronic kidney disease transition. Although it is well known that contralateral nephrectomy (CNX), even 2 wk post-UIR injury, can improve recovery, the physiological mechanisms and tubular signaling pathways mediating such improved recovery remain poorly defined. Here, we examined the renal hemodynamic and tubular signaling pathways associated with UIR injury and its reversal by CNX. Male Sprague-Dawley rats underwent left UIR or sham UIR and 2 wk later CNX or sham CNX. Blood pressure, left renal blood flow (RBF), and total glomerular filtration rate were assessed in conscious rats for 3 days before and over 2 wk after CNX or sham CNX. In the presence of a contralateral uninjured kidney, left RBF was lower (P < 0.05) from 2 to 4 wk following UIR (3.6 ± 0.3 mL/min) versus sham UIR (9.6 ± 0.3 mL/min). Without CNX, extensive renal atrophy, TIF, and tubule dedifferentiation, but minimal pimonidazole and hypoxia-inducible factor-1α positivity in tubules, were present at 4 wk post-UIR injury. Conversely, CNX led (P < 0.05) to sustained increases in left RBF (6.2 ± 0.6 mL/min) that preceded the increases in glomerular filtration rate. The CNX-induced improvement in renal function was associated with renal hypertrophy, more redifferentiated tubules, less TIF, and robust pimonidazole and hypoxia-inducible factor-1α staining in UIR injured kidneys. Thus, contrary to expectations, indexes of hypoxia are not observed with the extensive TIF at 4 wk post-UIR injury in the absence of CNX but are rather associated with the improved recovery of renal function and structure following CNX.

BP Fluctuations and the Real-Time Dynamics of Renal Blood Flow Responses in Conscious Rats.

BACKGROUND: Renal autoregulation maintains stable renal function despite BP fluctuations and protects glomerular capillaries from hypertensive injury. However, real-time dynamics of renal autoregulation in conscious animals have not been characterized. METHODS: To develop novel analytic methods for assessing renal autoregulation, we recorded concurrent BP and renal blood flow in conscious rats, comparing animals with renal autoregulation that was intact versus impaired (from 3/4 nephrectomy), before and after additional impairment (from the calcium channel blocker amlodipine). We calculated autoregulatory indices for adjacent short segments of increasing length (0.5, 1, 2.5, 5, 10, and 20 seconds) that exhibited a mean BP difference of at least 5 mm Hg. RESULTS: Autoregulatory restoration of renal blood flow to baseline after BP changes in conscious rats occurs rapidly, in 5-10 seconds. The response is significantly slower in states of impaired renal autoregulation, enhancing glomerular pressure exposure. However, in rats with severe renal autoregulation impairment (3/4 nephrectomy plus amlodipine), renal blood flow in conscious animals (but not anesthetized animals) was still restored to baseline, but took longer (15-20 seconds). Consequently, the ability to maintain overall renal blood flow stability is not compromised in conscious rats with impaired renal autoregulation. CONCLUSIONS: These novel findings show the feasibility of renal autoregulation assessment in conscious animals with spontaneous BP fluctuations and indicate that transient increases in glomerular pressure may play a greater role in the pathogenesis of hypertensive glomerulosclerosis than previously thought. These data also show that unidentified mechanosensitive mechanisms independent of known renal autoregulation mechanisms and voltage-gated calcium channels can maintain overall renal blood flow and GFR stability despite severely impaired renal autoregulation.

Autoregulatory Efficiency Assessment in Kidneys Using Deep Learning.

A convolutional deep neural network is employed to assess renal autoregulation using time series of arterial blood pressure and blood flow rate measurements in conscious rats. The network is trained using representative data samples from rats with intact autoregulation and rats whose autoregulation is impaired by the calcium channel blocker amlodipine. Network performance is evaluated using test data of the types used for training, but also with data from other models for autoregulatory impairment, including different calcium channel blockers and also renal mass reduction. The network is shown to provide effective classification for impairments from calcium channel blockers. However, the assessment of autoregulation when impaired by renal mass reduction was not as clear, evidencing a different signature in the hemodynamic data for that impairment model. When calcium channel blockers were given to those animals, however, the classification again was effective.

Progression of Chronic Kidney Disease After Acute Kidney Injury: Role of Self-Perpetuating Versus Hemodynamic-Induced Fibrosis.

The relative contribution of self-perpetuating versus hemodynamic-induced fibrosis to the progression of chronic kidney disease (CKD) after acute kidney injury (AKI) is unclear. In the present study, male Sprague-Dawley rats underwent right uninephrectomy and were instrumented with a blood pressure radiotelemeter. Two weeks later, separate groups of rats were subjected to 40 minutes renal ischemia-reperfusion or sham surgery and followed up for 4 or 16 weeks to determine the extent to which glomerulosclerosis and tubulointerstitial fibrosis as a result of the AKI-CKD transition (ie, at 4 weeks post AKI) change over time during the progression of CKD (ie, at 16 weeks post AKI). On average, tubulointerstitial fibrosis was ≈3-fold lower (P<0.05), whereas glomerulosclerosis was ≈6-fold higher (P<0.05) at 16 versus 4 weeks post AKI. At 16 weeks post AKI, marked tubulointerstitial fibrosis was only observed in rats exhibiting marked glomerulosclerosis, proteinuria, and kidney hypertrophy consistent with a hemodynamic pathogenesis of renal injury. Moreover, quantitative analysis between blood pressure and renal injury revealed a clear and modest blood pressure threshold (average 16-week systolic blood pressure of ≈127 mm Hg) for the development of glomerulosclerosis. In summary, modest levels of blood pressure may be playing a substantial role in the progression of renal disease after AKI in settings of preexisting CKD associated with 50% loss of renal mass. In contrast, these data do not support a major role of self-perpetuating tubulointerstitial fibrosis in the progression CKD after AKI in such settings.

Glomerulosclerosis in the diet-induced obesity model correlates with sensitivity to nitric oxide inhibition but not glomerular hyperfiltration or hypertrophy.

The diet-induced obesity (DIO) model is frequently used to examine the pathogenesis of obesity-related pathologies; however, only minimal glomerulosclerosis (GS) has been reported after 3 mo. We investigated if GS develops over longer periods of DIO and examined the potential role of hemodynamic mechanisms in its pathogenesis. Eight-week-old male obesity-prone (OP) and obesity-resistant (OR) rats (Charles River) were administered a moderately high-fat diet for 5 mo. Radiotelemetrically measured blood pressure, proteinuria, and GS were assessed. OP (n=10) rats developed modest hypertension (142±3 vs. 128±2 mmHg, P<0.05) and substantial levels of proteinuria (63±12 vs. 12±1 mg/day, P<0.05) and GS (7.7±1.4% vs. 0.4±0.2%) compared with OR rats (n=8). Potential hemodynamic mechanisms of renal injury were assessed in additional groups of OP and OR rats fed a moderately high-fat diet for 3 mo. Kidney weight (4.3±0.2 vs. 4.3±0.1 g), glomerular filtration rate (3.3±0.3 vs. 3.1±0.1 ml/min), and glomerular volume (1.9±0.1 vs. 2.0±0.1 µm3×10(-6)) were similar between OP (n=6) and OR (n=9) rats. Renal blood flow autoregulation was preserved in both OP (n=7) and OR (n=7) rats. In contrast, Nω-nitro-L-arginine methyl ester (L-NAME) administration in conscious, chronically instrumented OP (n=11) rats resulted in 15% and 39% increases in blood pressure and renal vascular resistance, respectively, and a 16% decrease in renal blood flow. Minimal effects of L-NAME were seen in OR (n=9) rats. In summary, DIO-associated GS is preceded by an increased hemodynamic sensitivity to L-NAME but not renal hypertrophy or hyperfiltration.

Hemodynamic basis for the limited renal injury in rats with angiotensin II-induced hypertension.

ANG II is thought to increase the susceptibility to hypertension-induced renal disease (HIRD) via blood pressure (BP)-dependent and BP-independent pathways; however, the quantitative relationships between BP and HIRD have not been examined in ANG II-infused hypertensive rats. We compared the relationship between radiotelemetrically measured BP and HIRD in Sprague-Dawley rats (Harlan) chronically administered ANG II (300-500 ng·kg(-1)·min(-1), n = 19) for 4 wk versus another commonly employed pharmacological model of hypertension induced by the chronic administration of N(ω)-nitro-l-arginine methyl ester (l-NAME, 50 mg·kg(-1)·day(-1), n = 23). [DOSAGE ERROR CORRECTED]. Despite the significantly higher average systolic BP associated with ANG II (191.1 ± 3.2 mmHg) versus l-NAME (179.9 ± 2.5 mmHg) administration, the level of HIRD was very modest in the ANG II versus l-NAME model as evidenced by significantly less glomerular injury (6.6 ± 1.3% vs. 11.3 ± 1.5%, respectively), tubulointerstitial injury (0.3 ± 0.1 vs. 0.7 ± 0.1 injury score, respectively), proteinuria (66.3 ± 10.0 vs. 117.5 ± 10.1 mg/day, respectively), and serum creatinine levels (0.5 ± 0.04 vs. 0.9 ± 0.07 mg/dl, respectively). Given that HIRD severity is expected to be a function of renal microvascular BP transmission, BP-renal blood flow (RBF) relationships were examined in additional conscious rats administered ANG II (n = 7) or l-NAME (n = 8). Greater renal vasoconstriction was observed during ANG II versus l-NAME administration (41% vs. 23% decrease in RBF from baseline). Moreover, administration of ANG II, but not l-NAME, led to a unique BP-RBF pattern in which the most substantial decreases in RBF were observed during spontaneous increases in BP. We conclude that the hemodynamic effects of ANG II may mediate the strikingly low susceptibility to HIRD in the ANG II-infused model of hypertension in rats.

Blood pressure-renal blood flow relationships in conscious angiotensin II- and phenylephrine-infused rats.

Chronic ANG II infusion in rodents is widely used as an experimental model of hypertension, yet very limited data are available describing the resulting blood pressure-renal blood flow (BP-RBF) relationships in conscious rats. Accordingly, male Sprague-Dawley rats (n = 19) were instrumented for chronic measurements of BP (radiotelemetry) and RBF (Transonic Systems, Ithaca, NY). One week later, two or three separate 2-h recordings of BP and RBF were obtained in conscious rats at 24-h intervals, in addition to separate 24-h BP recordings. Rats were then administered either ANG II (n = 11, 125 ng·kg(-1)·min(-1)) or phenylephrine (PE; n = 8, 50 mg·kg(-1)·day(-1)) as a control, ANG II-independent, pressor agent. Three days later the BP-RBF and 24-h BP recordings were repeated over several days. Despite similar increases in BP, PE led to significantly greater BP lability at the heart beat and very low frequency bandwidths. Conversely, ANG II, but not PE, caused significant renal vasoconstriction (a 62% increase in renal vascular resistance and a 21% decrease in RBF) and increased variability in BP-RBF relationships. Transfer function analysis of BP (input) and RBF (output) were consistent with a significant potentiation of the renal myogenic mechanism during ANG II administration, likely contributing, in part, to the exaggerated reductions in RBF during periods of BP elevations. We conclude that relatively equipressor doses of ANG II and PE lead to greatly different ambient BP profiles and effects on the renal vasculature when assessed in conscious rats. These data may have important implications regarding the pathogenesis of hypertension-induced injury in these models of hypertension.

Systolic and mean blood pressures and afferent arteriolar myogenic response dynamics: a modeling approach.

The afferent arteriolar myogenic response contributes to the autoregulation of renal blood flow (RBF) and glomerular filtration rate (GFR), and plays an essential role in protecting the kidney against hypertensive injury. Systolic blood pressure (SBP) is most closely linked to renal injury, and a myogenic response coupled to this signal would facilitate renal protection, whereas mean blood pressure (MBP) influences RBF and GFR. The relative role of SBP vs. MBP as the primary determinant of myogenic tone is an area of current controversy. Here, we describe two mathematical models, Model-Avg and Model-Sys, that replicate the different delays and time constants of vasoconstrictor and vasodilator phases of the myogenic responses of the afferent arteriole. When oscillating pressures are applied, the MBP determines the magnitude of the myogenic response of Model-Avg, and the SBP determines the response of Model-Sys. Simulations evaluating the responses of both models to square-wave pressure oscillations and to narrow pressure pulses show decidedly better agreement between Model-Sys and afferent arteriolar responses observed in cortical nephrons in the in vitro hydronephrotic kidney model. Analysis showing that the difference in delay times of the vasoconstrictor and vasodilator phases determines the frequency range over which SBP triggers Model-Sys's response was confirmed with simulations using authentic blood pressure waveforms. These observations support the postulate that SBP is the primary determinant of the afferent arteriole's myogenic response and indicate that differences in the delays in initiation vs. termination of the response, rather than in time constants, are integral to this phenomenon.

Dynamic blood pressure load and nephropathy in the ZSF1 (fa/fa cp) model of type 2 diabetes.

Diabetes and increased blood pressure (BP) are believed to interact synergistically in the pathogenesis and progression of diabetic nephropathy. The present studies were performed to examine if there were differences in BP load and/or protective renal autoregulatory capacity between the obese diabetic Zucker fatty /spontaneously hypertensive heart failure F1 hybrid (ZSF1) (fa/fa cp) rats and their lean controls. By approximately 26 wk of age, ZSF1 (n = 13) but not their lean controls (n = 16) had developed substantial proteinuria (180 +/- 19 vs. 16 +/- 1.4 mg/24 h) and glomerulosclerosis (19 +/- 2.4 vs. 0.6 +/- 0.2%; P < 0.001). However, average ambient systolic BP by radiotelemetry (12-26 wk of age) was modestly lower in ZSF1 than in lean controls (130 +/- 1.4 vs. 137 +/- 1.7 mmHg, P < 0.002), although the 24-h BP power spectra showed a mild increase at frequencies <0.1 Hz in the ZSF1. Autoregulatory capacity under anesthesia in response to step changes in perfusion pressure between 100 and 140 mmHg was similarly well preserved in both ZSF1 and lean controls at 16-18 wk of age [autoregulatory indexes (AI) <0.1]. Similarly, differences were not observed for dynamic autoregulation in conscious rats [transfer functions between BP (input) and renal blood flow (output) using chronic Transonic flow probes]. Collectively, these data indicate that the pathogenesis of nephropathy in the ZSF1 model of type 2 diabetic nephropathy is largely independent of differences in systemic BP and/or its potential renal transmission. However, these data do not exclude the possibility that the diabetic milieu may alter the glomerular capillaries in the ZSF1, such that there is an enhanced local susceptibility to injury with even normal glomerular pressures.

Characterization of dynamics in renal autoregulation using volterra models.

The dynamics of renal autoregulation are modeled using a modified Volterra representation called the fixed pole expansion technique (FPET). A data dependent procedure is proposed for selecting the pole locations in this expansion that enables a reduction in model complexity compared to standard Volterra models. Furthermore, a quantitative characterization of frequency dependent features of the renal autoregulatory response is enabled via the model's pole locations. The utility of this approach is demonstrated by applying the modeling technique to renal blood pressure and renal blood flow measurements in conscious rats. The model is used to characterize the myogenic autoregulatory response in control rats and rats whose renal autoregulation has been impaired by calcium channel blockers.

Differential effects of salt on renal hemodynamics and potential pressure transmission in stroke-prone and stroke-resistant spontaneously hypertensive rats.

Salt-supplemented stroke-prone spontaneously hypertensive rats (SHRsp) develop more severe hypertension-induced renal damage (HIRD) compared with their progenitor SHR. The present studies were performed to examine whether in addition to increasing the severity of hypertension salt also enhanced the transmission of such hypertension to the renal vascular bed in the SHRsp. "Step" and "dynamic" renal blood flow (RBF) autoregulation (AR) were examined in approximately 12-wk-old SHR and SHRsp after 3-5 days of an 8% NaCl diet. During step AR under anesthesia (n = 8-11), RBF was significantly higher in the SHRsp at all perfusion pressures (P < 0.01), but AR capacity was not different. Similarly, in separate conscious chronically instrumented rats (n = 8 each), both blood pressure (BP) and RBF were modestly but significantly higher at baseline before salt in the SHRsp (P < 0.05). However, transfer function analysis did not show significant differences in the admittance gain parameters. However, after 3-5 days of salt, although average BP was not significantly altered in either strain, RBF increased further in the SHRsp and there was a significantly greater transfer of BP into RBF power in the SHRsp. This was reflected in the significantly higher admittance gain parameters at most frequencies including the heartbeat frequency (P < 0.05 maximum). These differential hemodynamic effects of salt have the potential to enhance BP transmission to the renal vascular bed and also contribute to the more severe HIRD observed in the salt-supplemented SHRsp.

Effects of calcium channel blockers on "dynamic" and "steady-state step" renal autoregulation.

Renal autoregulation (AR) mechanisms provide the primary protection against transmission of systemic pressures and hypertensive renal damage. However, the relative merits of the "step" change vs. "dynamic" methods for the assessment of AR capacity remain controversial. The effects of 48-72 h of orally administered amlodipine (L-type) and mibefradil (T-type) calcium channel blockers (CCBs) on step and dynamic AR in Sprague-Dawley rats were compared. Both CCBs significantly impaired "steady-state step" AR (autoregulatory indexes = approximately 0.5 vs. approximately 0.1 in controls, P < 0.05; n = 9-10/group). By contrast, dynamic AR compensation in separate conscious rats (n = 12) was not significantly altered by either amlodipine (n = 10) or mibefradil (n = 6; fractional gain in admittance approximately 0.4-0.5 in all groups at frequencies in the range of 0.0025-0.025 Hz). However, both CCBs tended to attenuate the myogenic resonance peak along with shifting it to a significantly slower frequency (P < 0.001) during dynamic AR, but no consistent effects were observed on the tubuloglomerular feedback resonance peak. While the reasons for the insensitivity of dynamic vs. steady-state step AR capacity estimates to CCBs remain to be established, the present data indicate that dynamic AR methods may have a limited utility for assessing AR capacity but may provide potentially important insights into the operational characteristics of AR control mechanisms. A strong correlation was also observed between the average conductance and the admittance gain at the heart beat frequency (r = 0.77, P < 0.001), suggesting that such parameters may provide additional and possibly more meaningful indexes of BP transmission in conscious animals during dynamic AR.

"Step" vs. "dynamic" autoregulation: implications for susceptibility to hypertensive injury.

Renal autoregulatory (AR) mechanisms provide the primary protection against transmission of systemic pressures, and their impairment is believed to be responsible for the enhanced susceptibility to hypertensive renal damage in renal mass reduction (RMR) models. Assessment of AR capacity by the "step" change methodology under anesthesia was compared with that by "dynamic" methods in separate conscious control Sprague-Dawley rats and after uninephrectomy (UNX) and (3/4) RMR (RK-NX) (n = 7-10/group). Substantially less AR capacity was seen by the dynamic vs. the step methodology in control rats. Moreover, dynamic AR capacity did not differ among controls, UNX, and RK-NX rats (fractional gain in admittance approximately 0.4-0.5 in all groups at frequencies in the range of 0.0025-0.025 Hz). By contrast, significant impairment of step AR was seen in RK-NX vs. control or UNX rats (AR indexes 0.7 +/- 0.1 vs. 0.1 +/- 0.02 and 0.2 +/- 0.04, respectively, P < 0.01). We propose that the step and dynamic methods evaluate the renal AR responses to different components of blood pressure (BP) power with the step AR assessing the ability to buffer large changes in average BP (DC power), whereas the present "dynamic" methods assess the AR ability to buffer slow BP fluctuations (<0.25 Hz) superimposed on the average BP (AC power), a substantially smaller component of total BP power. We further suggest that step but not dynamic AR methods as presently performed provide a valid index of the underlying susceptibility to hypertensive glomerular damage after RMR.