Many hormonal pathways contribute to the regulation of renal epithelial sodium channel (ENaC) function, a key process for maintaining blood volume and controlling blood pressure. In the present study, we examined whether the peptide hormone prolactin (PRL) regulates ENaC function in renal epithelial cells (A6). Basolateral application of several different concentrations of PRL dramatically stimulated the transepithelial current in A6 cells, increasing both amiloride-sensitive (ENaC) and amiloride-insensitive currents. Using cell-attached patch clamp, we determined that PRL increased both the number (N) and open probability (Po) of ENaC present in the apical membrane. Inhibition of PKA with H-89 abolished the effect of PRL on amiloride-sensitive and insensitive transepithelial currents and eliminated the increase in ENaC NPo with PRL exposure. PRL also increased cAMP in A6 cells, consistent with signaling through the cAMP-dependent PKA pathway. We also identified that PRL induced activity of a 2-pS anion channel with outward rectification, electrophysiological properties consistent with ClC4 or ClC5. RT-PCR only detected ClC4, but not ClC5 transcripts. Here, we show for the first time that PRL activates sodium and chloride transport in renal epithelial cells via ENaC and ClC4.
- ClC channels
- protein kinase A
renal control of total body fluid balance is an essential process in the maintenance of blood pressure. One key mechanism for regulating fluid balance is the epithelial sodium channel (ENaC), present in the connecting tubule and principal cells of the cortical collecting duct (4). ENaC selectively reabsorbs sodium ions from the tubular lumen back into the bloodstream, providing a significant driving force for water reabsorption. Accordingly, gain-of-function mutations in ENaC produce hypertension in human patients. These mutations were first recognized in families with a condition called Liddle's syndrome (14). However, due to human genome-sequencing efforts, investigators have now found several ENaC gene variants present in individuals with salt-sensitive hypertension (49). One of these variants was recently shown to greatly increase ENaC activity, indicating that gain-of-function ENaC mutations may also be present in those with salt-sensitive hypertension (7).
Hormonal control of renal ENaC function is an area of intense interest given the known contributions of ENaC and its effector hormones (i.e., aldosterone, angiotensin II, and vasopressin) to hypertension (25, 34, 40, 47, 48). The steroid hormone aldosterone, via binding and activation of the mineralocorticoid receptor, potently stimulates ENaC activity and subunit expression in the aldosterone-sensitive nephron (connecting tubule and collecting duct), contributing to renal retention of salt and water that lead to elevations in blood pressure. Disruptions in circulating levels of aldosterone thus adversely affect blood pressure control, and considerable effort is underway to develop pharmacological treatments directed at this pathway (15).
Interestingly, investigators have recently shown that ENaC is the chief contributing factor in blood volume expansion during the latter stages of pregnancy in rats (44, 45). Inhibition of the mineralocorticoid receptor largely eliminated this effect, suggesting that aldosterone-mediated activation of ENaC is the primary mechanism (44). However, blood volume expansion was not completely abolished after mineralocorticoid receptor antagonism, indicating the involvement of another hormonal pathway. Given its elevation in pregnancy and known osmoregulatory effects, prolactin (PRL) may well be the other hormonal pathway regulating ENaC activity during pregnancy (19).
A couple of studies have shown that PRL stimulates amiloride-sensitive sodium transport that is probably mediated by ENaC in toad skin, suggesting that PRL might directly affect channel function (38, 39). Besides amiloride-sensitive ENaC activity, there have also been several reports of an amiloride-insensitive current (2, 6, 35). To determine whether PRL activates ENaC in renal epithelial cells, we 1) examined changes in the amiloride-sensitive transepithelial current with exposure to PRL, 2) identified the intracellular signaling pathways involved in PRL's effects, and 3) determined alterations in ENaC single-channel properties [number of channels (N) and open probability (Po)] in A6 renal epithelial cells after PRL exposure.
Deachapunya and colleagues (10) recently reported that PRL (1 μg/ml) profoundly stimulated cAMP-dependent K+ secretion in the distal colon epithelium possibly via KCNQ1/KCNE3 as well as Cl− secretion in the proximal and transverse colon via the CFTR (10). Although our laboratory has been unable to detect K+ channel activity in A6 cells, we have previously reported that A6 cells express CFTR message and protein and an 8-pS cAMP-stimulated Cl− channel consistent with CFTR's electrophysiological characteristics (23). Thus the last objective of this study was to determine whether PRL activated CFTR in the A6 renal epithelial cells, since CFTR plays an important role in fluid absorption and secretion in a variety of epithelia. In the absence of CFTR activation, PRL could also activate some alternative chloride channel.
Herein, we demonstrate that PRL, at the concentrations detected during pregnancy-induced hypertension, produces a remarkable surge in ENaC-mediated sodium reabsorption and a smaller, but significant, increase in Cl− secretion likely via ClC4.
MATERIALS AND METHODS
We obtained ovine PRL, amiloride hydrochloride, N-[2-(p-bromocinnamylamino) ethyl]-5-isoquinolinesulfonamide dihydrochloride (H-89 dihydrochloride hydrate), and 3-isobutyl-1-methylxanthine (IBMX) from Sigma-Aldrich. We obtained 2-cyano-3-(3,4-dihydroxyphenyl)-N-(phenylmethyl)-2-propenamide (AG-490), 2-(4-morpholinyl)-8-phenyl-4H-1-benzopyran-4-one (LY294002), and the cAMP enzyme immunoassay (EIA) from Cayman Chemical. We obtained 4-amino-5-(4-chloro-phenyl)-7-(t-butyl)pyrazolo[3,4-d]pyrimidine (PP2) from BioMol and 3-[(3-trifluoromethyl)phenyl]-5-[(4-carboxyphenyl) methylene]-2-thioxo-4-thiazolidinone (CFTR Inhibitor-172) from Santa Cruz Biotechnology.
We dissolved powdered reagents in an appropriate vehicle (water for PRL, ethanol for amiloride, AG-490, LY294002, and IBMX, or dimethylsulfoxide for H-89, PP2, and CFTR Inhibitor-172) with a final concentration of the vehicle not to exceed 0.1% of total volume. We included vehicle-treated controls in each experiment to account for any effects of the vehicle alone.
We used an antibody generated to a KLH-conjugated synthetic peptide between 663 and 689 amino acids from the C-terminal region of human CLC4 (Pierce Antibody PA5-13348). The epitope has 87% homology to Xenopus laevis ClC4.
We grew A6 cells (subclone 2F3; passages 96–106) on either collagen-coated permeable supports (Costar Transwell inserts) or patch-clamp rings as previously described (46). We used the cells several days past confluence to ensure a tight epithelial layer was present, a requirement for electrophysiological measurements. To remove other hormones from the media before exposure to PRL, we replaced the growth media with phenol red and serum-free media for 24–48 h before use of the cells in the experiments.
We used a World Precision Instruments Epithelial Voltohmmeter (EVOM) to measure transepithelial voltage and resistance in cells grown on permeable supports. We then calculated transepithelial current using Ohm's law, I = V/R, and corrected for surface area differences. Tissue resistances and voltages were large enough that we did not correct for the resistance of the permeable supports by themselves.
For single-channel analyses, bath and electrode solutions contained 96 mM NaCl, 3.4 mM KCl, 0.8 mM CaCl2, 0.8 mM MgCl2, and 10 mM HEPES titrated to a pH of 7.4 with NaOH. We used a two-stage vertical puller to produce electrodes (TW-150F, World Precision Instruments) with a resistance of 5–10 MΩ. After creating gigaohm seals on individual cells, we visualized channel activity for ∼5 min at a pipette holding potential of 0 mV via a Dagan PC-One patch-clamp amplifier with a low-pass three-pole Bessel filter set at either 100 or 1,000 Hz and digitally recorded the data at a 5-kHz sampling frequency (Digidata 1440a and pCLAMP10, Axon Instruments). If necessary to resolve single channels, we used additional digital filtering with corner frequencies of 10–30 Hz for single-channel analyses. We calculated channel number (N) in a given patch by counting the number of individual current transitions (levels) observed in the recording. Single-channel analysis using pCLAMP 10 (Molecular Devices) provided a calculated NPo by utilizing the following equation where i signifies the number of channels open, ti the amount of time channels spent open, and T the total amount of recording time. We divided NPo by N to determine open probability (Po). To assess channel conductance, we recorded single-channel events at different holding potentials and measured open channel current amplitudes.
used a modified constant field approach first described by Goldman (13) and later by Hodgkin and Katz (16) to describe the permeability of anions through ClC2 and CFTR channels. In general, their approach gave the ionic flux (as current) and ion permeability across a cell membrane as described by the following equation where IS is the current of ion S, [S]i and [S]o are the internal and external concentrations of ion S, z is the valence, and E the potential difference across the cell membrane.
We modeled the voltage dependence of the chloride channel Po as a Boltzmann function (8) where Po(V), Po(max), and Po(min) are the open probability at a specific voltage, maximum open probability, and minimum open probability, respectively; α is the slope of the function (and magnitude of change in Po for an e-fold change in voltage); V is the apical membrane potential, and V0 is the voltage at which channels are at half of maximum Po.
We collected total RNA from A6 cells grown on permeable supports using TRIzol reagent (Invitrogen) and converted RNA into cDNA using a QuantiTect Reverse Transcription Kit (Qiagen) according to the manufacturer's instructions. We next performed PCR (32 cycles) with 10 μg cDNA and primers specific for X. laevis ClCn4 (forward primer, 5′ CTAGTCACCGAGCTGTGAGG 3′, and reverse primer, 5′ CCGTTCAAACCATTGCCGTT 3′) and ClCn5 (forward primer, 5′ GCACGAGTGTGAGGATCTG 3′, and reverse primer, 5′ CCGTTCAAACCATTGCCGTT 3′) transcripts using Qiagen HotStarTaq DNA Polymerase.
We produced graphs and performed statistical analyses using SigmaPlot 12. For transepithelial current measurements, we used two-way repeated-measures ANOVA with an appropriate post hoc test. For comparison of only two groups, we used an unpaired Student's t-test or, if the data was not normal, the nonparametric Mann-Whitney Rank Sum test. For I–V curve analysis, we used a nonlinear least-squares fit of the Goldman-Hodgkin-Katz equation to predict apical membrane potential, intracellular chloride activity, apical membrane chloride permeability, and the approximate slope (G, conductance) of the line. We show all data as means ± SE and considered differences significant at the P < 0.05 level.
PRL increases amiloride-sensitive current in A6 cells.
To determine whether PRL affected ENaC activity, we used a renal epithelial cell line (A6) known to express the three subunits required for ENaC function and to produce large amiloride-sensitive transepithelial currents characteristic of ENaC. In addition, ENaC activity in these cells is sensitive to aldosterone, ADH, and ANF as one would expect for renal cortical collecting duct principal cells. We measured the transepithelial current in A6 cells immediately before and at several time points following basolateral exposure to vehicle (0.1% water) or PRL (1 μg/ml). The concentration of PRL used is consistent with median levels detected in urine of patients with severe preeclampsia, a life-threatening hypertensive disorder of pregnancy (20, 21). Within 30 min after treatment, transepithelial current was significantly greater in PRL-treated vs. vehicle-treated cells, an effect sustained for at least 24 h after initial exposure to PRL (Fig. 1A). To determine whether this increase in transepithelial current was due to ENaC, we performed a similar experiment, but added 2 μM amiloride immediately following the 60-min time point measurement. We observed that amiloride significantly reduced the PRL-induced current (Fig. 1B). However, a small but significant component of the PRL-induced transepithelial current was amiloride insensitive, suggesting activation of another ion transport mechanism. A lower concentration of PRL (0.1 μg/ml) consistent with levels detected in pregnant and lactating women also stimulated the amiloride-sensitive current about twofold, similar to the effect of the larger concentration (Fig. 1C).
PRL increases amiloride-sensitive and -insensitive currents via a PKA-dependent pathway.
To investigate the molecular signaling mechanism involved in the effect of PRL on ENaC, we measured whether the effect of PRL on the transepithelial current differed in the presence of inhibitors of known PRL signaling pathways. Inhibition of Jak2 with AG-490 (25 μM) did not affect the initial increase (30-min time point) in the transepithelial current after PRL administration (P > 0.05 for PRL vs. PRL+AG-490) (Fig. 2A). However, at 60 min post-PRL exposure, the transepithelial current in the presence of AG-490 decreased 15% while it rose nearly 20% in the absence of AG-490. Neither PP2 (a SRC inhibitor) nor LY294002 [a phosphatidylinositol 3-kinase (PI3K) inhibitor] affected PRL's stimulation of the amiloride-sensitive current (Fig. 2, B and C). Surprisingly, LY294002 enhanced the effect of PRL to stimulate an amiloride-insensitive transepithelial current. The amiloride-insensitive current comprised 45% of the current in PRL+LY294002-treated cells compared with only 24% in PRL-only treated cells (Fig. 2C). Pretreatment with the PKA inhibitor H-89 (15 μM) inhibited the transepithelial current and eliminated the stimulatory effect of PRL on the transepithelial current (Fig. 2D).
Both ENaC Po and channel density increase with PRL exposure.
To further investigate the mechanisms behind PRL's stimulatory effect on ENaC, we used cell-attached patch clamp to examine differences in single-channel events from A6 cells exposed to vehicle or PRL for 30–90 min. Representative tracings from single-channel events recorded in A6 cells demonstrate greater channel activity from patches of cells treated with PRL (Fig. 3A). Consistent with the observed recordings, calculated ENaC single-channel activity (NPo) was significantly greater in patches from PRL-treated cells (Fig. 3B) compared with vehicle-treated cells. Separation of the data into N and Po revealed both significantly greater ENaC N and Po in patches from PRL-treated cells (Fig. 3, C and D). These results are consistent with our transepithelial current measurements showing that PRL stimulates the amiloride-sensitive transepithelial current.
Effect of PRL on ENaC is dependent on cAMP-dependent PKA pathway.
We had already observed that the PKA inhibitor H-89 dramatically inhibited PRL's stimulation of both the amiloride-sensitive and -insensitive currents in A6 cells. Since PKA is cAMP dependent, we measured cAMP accumulation in A6 cells treated with vehicle or PRL. PRL significantly increased cAMP abundance in A6 cells (Fig. 4A). To ensure that the PKA pathway specifically affected ENaC stimulation by PRL and was not a nonspecific effect of amiloride, we also examined ENaC NPo in patches from cells pretreated with the PKA inhibitor H-89 (10 μM), followed by either vehicle or PRL. As expected, we observed no significant difference in NPo between the H-89 only and H-89+PRL groups (Fig. 4B), revealing that PRL was no longer able to increase ENaC N or Po if PKA was not functional. When used at a 10-fold lower concentration (1 μM), H-89 did not inhibit PRL's stimulatory effect on the amiloride-sensitive transepithelial current. However, it did reduce the baseline amiloride-sensitive current in this cell line (Fig. 4C).
PRL stimulates chloride channel activity in A6 cells.
In all transepithelial current analyses, we observed an amiloride-insensitive current stimulated by PRL. Since our laboratory has previously shown the presence of chloride channels CFTR and ClC-2 in A6 epithelia, we sought to determine whether these anion channels were responsible for the amiloride-insensitive current induced by PRL (1, 23). We first measured the transepithelial current in vehicle- and PRL-treated cells in the presence and absence of 10 μM inhibitor-172, traditionally considered an inhibitor of CFTR. This inhibitor almost completely eliminated the remaining amiloride-insensitive current induced by PRL and significantly decreased the basal transepithelial current in A6 cells (Fig. 5A).
To investigate this further, we performed cell-attached patch clamp in A6 cells and examined single channels present before and after addition of PRL to the basolateral media. After PRL treatment, we detected a small-conductance channel often barely visible due to the presence of the ENaC. We analyzed the I–V relationship of the observed channel to determine its electrophysiological characteristics (Fig. 5B). We plotted the mean current over a range of holding potentials for the chloride channel (mean currents and standard errors for 5 patches). The outward rectification and the reversal potential near zero (in these cell-attached patches, the reversal potential is approximately the same as the apical membrane potential) are characteristic of ClC4/5. Analysis of five patches with distinct anion channel activity in the absence of ENaC revealed a channel with a conductance of 2 pS exhibiting outward rectification at positive potentials (Fig. 5C). The best nonlinear least-squares fit to the Goldman-Hodgkin-Katz equation predicts that the apical membrane potential is −45 ± 9.4 mV, intracellular chloride is 20 ± 7.5 mM, and apical membrane chloride permeability is (0.58 ± 0.014) × 10−7 cm/s. The Po of the chloride channels depends upon voltage as shown in Fig. 5D [normalized to Po(max)]. The line through the points is the best nonlinear least-squares fit to the Boltzmann voltage equation and predicts that at −46 ± 2.6 mV a single channel is at half Po(max) and that the channels respond to voltage with an e-fold change in Po for every 40 ± 6.0-mV change in membrane potential (implying that multiple ions can occupy a channel).
Surprisingly, we did not observe a channel consistent with the electrophysiological characteristics of CFTR (8-pS conductance with outward rectification) as was suggested based on a previous report and our own observation that a CFTR inhibitor greatly reduced the amiloride-insensitive currents activated by PRL. Instead, the anion channel detected with application of PRL had electrophysiological characteristics typical of endosomal chloride channels (ClC) which are known to also reside in the plasma membrane (18, 32, 43). To better define the exact identity of the channel, we used RT-PCR to determine whether either ClCn4 or ClCn5 transcripts were present in the frog (X. laevis) kidney and A6 cells (Fig. 5E). We only detected the presence of ClCn4 (not ClCn5) transcripts in the frog kidney and A6 cells. We then detected ClC4 in Western blots using a commercially available antibody whose epitope has high homology to X. laevis ClC4. Figure 5F shows that ClC4 can be detected in A6 cells at the expected molecular mass of 85 kDa. Exposure to PRL does not alter the amount of total cellular ClC4, suggesting that PRL increases the activity of ClC4 by increasing channel Po or by insertion of preexisting channels (Fig. 5B).
To investigate the mechanism behind PRL's stimulation of the ClC channel, we pretreated A6 cells with 2 μM amiloride to remove the amiloride-sensitive current component and either applied an inhibitor of PKA (1 μM H-89) or PI3K (10 μM LY294002) for 30 min before application of PRL (Fig. 6). H-89 marginally amplified PRL's stimulation of the amiloride-insensitive current whereas LY294002 to some extent reduced this effect.
Here, we show for the first time that the peptide hormone PRL stimulates both sodium and chloride transport in A6 renal epithelial cells. PRL specifically induced activity of the ENaC and a 2-pS chloride channel predominantly via the cAMP-dependent protein kinase PKA. These results suggest a role for PRL in fluid homeostasis and blood pressure control in conditions of hyperprolactinemia including pregnancy, lactation, and prolactinomas.
Although most well known for its role in mammalian reproduction, PRL also plays a critical role in osmoregulation in freshwater fish (36). Although very little is known about the osmoregulatory function of PRL in mammals, given its great importance to milk secretion in lactation, it is likely that PRL also regulates fluid movement in mammals. However, the exact mechanisms remain to be fully defined. Our results demonstrate that PRL directly stimulates sodium reabsorption through ENaC, the principal mechanism for fluid reabsorption in most mammalian epithelial cells, providing evidence that PRL probably contributes to fluid balance in mammals as well. The importance of PRL's regulation of ENaC function likely extends to other conditions and disorders including pregnancy wherein PRL levels are high and blood volume increases considerably to accommodate fetal/placental demand. Several studies now indicate that PRL dysregulation contributes to hypertension in both male and female patients with essential and pulmonary hypertension (12, 17, 22, 37). This effect may well be due to PRL's induction of sodium reabsorption through renal ENaC.
In addition to ENaC, our laboratory has previously identified two chloride channels, CFTR and ClC2, in the apical membrane of A6 epithelia (1, 23). Thus we first considered the possibility that the identity of the amiloride-insensitive current induced by PRL might be one of these channels. Indeed, the CFTR inhibitor CFTR-Inh-172 (10 μM) largely eliminated the amiloride-insensitive current present in PRL-treated cells. Nonetheless, the electrophysiological characteristics of single channels activated by PRL were not consistent with either the CFTR channel or ClC2 previously detected by our laboratory (1). The observed anion channel instead displayed intermediate characteristics, with a 2-pS conductance similar to ClC2 and outward rectification similar to CFTR. These properties are most consistent with either ClC4 or ClC5 channels (18, 43). Inhibition of the amiloride-insensitive portion of the PRL-induced current in the presence of CFTR-Inh-172 is not surprising as our previous work on A6 cells and that of other investigators have observed inhibition of volume-sensitive outwardly rectifying chloride channels such as ClC2 when using this inhibitor (27).
Interestingly, we detected expression of ClCn4 transcripts (ClC4) but not ClCn5 (ClC5) in cDNA samples from frog kidney and A6 cells, suggesting that the identity of the PRL-induced chloride channel is likely ClC4. However, the mRNA sequence homology for X. laevis ClCn4 and ClCn5 is nearly identical with only a short (<100 bp) region in the 5′-untranslated region that differs. Thus both have identical amino acid sequences. This sequence similarity suggests a gene duplication event in X. laevis with subsequent sequence divergence during the evolution of mammalian species. In turn, it is possible that ClCn4 detected in X. laevis may encode either ClCn4 or ClCn5 in mammals today. However, in Western blots, we can detect ClC4 at the expected molecular weight of 85 kDa. There is no change in total ClC4 protein after overnight addition of PRL. Therefore, we assume that the increase in channel activity is not due to an increase in translation, but rather to an increase in Po or membrane insertion of preexisting intracellular channels.
It will be most informative in follow-up studies to identify whether the effect of PRL to stimulate sodium and chloride transport occurs in epithelial tissues known to express ENaC and ClC. Of interest is the fact that ClC5 localizes to epithelial tissues where ENaC is highly abundant, including the kidney, colon, inner ear, and lung. Indeed, defects in ClC5 cause Dent's disease, symptoms of which include proteinuria and renal failure (11, 32, 41, 42). The possible implications of ClC regulation by PRL are as yet unclear but may be important in states of proteinuria in pregnancy (e.g., preeclampsia). Actually, several publications provide striking evidence that PRL is one of the most reliable biomarkers for preeclampsia (20, 21, 30, 31, 33). However, the role of PRL in producing the renal pathologies associated with preeclampsia is still undefined. Future studies should investigate whether PRL-induced changes in renal tubular ion transport contribute to the proteinuria and hypertension characteristic of preeclampsia.
Our results indicate that the predominant signaling mechanism mediating the effect of PRL on both ENaC and ClC is the cAMP-dependent PKA pathway. We originally hypothesized that the canonical signaling pathways for PRL action which include Jak2/Stat and SRC/PI3K would likely mediate this effect (9). Although inhibition of Jak2 signaling with AG-490 eliminated a small portion of the PRL-induced current, the pathway did not contribute to the majority of the current, suggesting the involvement of another unidentified signaling mechanism. Complicating the issue is that AG-490 is a potent inhibitor of the EGF receptor, and our laboratory has previously shown that EGF through the EGF receptor alters ENaC activity in A6 cells (24).
In patch-clamp studies, we showed that, in addition to Po, ENaC N increased with exposure to PRL. Previous studies from several laboratories have shown that the cAMP-PKA pathway largely mediates movement of new channels to the apical plasma membrane but also has more immediate effects on channel Po (3, 5, 26, 28, 29). Consistent with our patch-clamp data showing an increase in both ENaC N and Po with PRL exposure, an inhibitor of PKA (H-89) completely abolished the effect of PRL on ENaC activity and also suppressed the effect of PRL on ClC. Together with the observed greater cAMP abundance in the presence of PRL, our results imply that PRL mobilizes cAMP to affect ion transport in renal epithelial cells.
In conclusion, we demonstrate herein that PRL activates ENaC and ClC in A6 renal epithelial cells via the cAMP-dependent PKA signaling pathway. Future studies will address the physiological role of PRL-mediated regulation of ENaC and ClC to affect fluid homeostasis in a variety of tissues and to possibly regulate blood pressure.
This work was supported by T32 HL076118-08 and K12 GM000680-14 to M. M. Greenlee and R37 DK037963 to D. C. Eaton.
No conflicts of interest, financial or otherwise, are declared by the authors.
Author contributions: M.M.G., O.K.A.-K., H.-F.B., and D.C.E. provided conception and design of research; M.M.G., J.D.M., B.J.D., O.K.A.-K., and H.-F.B. performed experiments; M.M.G., J.D.M., O.K.A.-K., H.-F.B., and D.C.E. analyzed data; M.M.G., H.-F.B., and D.C.E. interpreted results of experiments; M.M.G., H.-F.B., and D.C.E. prepared figures; M.M.G. drafted manuscript; M.M.G., H.-F.B., and D.C.E. edited and revised manuscript; M.M.G., J.D.M., B.J.D., O.K.A.-K., H.-F.B., and D.C.E. approved final version of manuscript.
- Copyright © 2015 the American Physiological Society