Sexually dimorphic behaviors develop under the influence of sex steroids, which induce reversible changes in the underlying neural network of the brain. However, little is known about the proteins that mediate these activational effects of sex steroids. Here, we used a proteomics approach for large-scale identification of proteins involved in the development of a sexually dimorphic behavior, the electric organ discharge of brown ghost knifefish, Apteronotus leptorhynchus. In this weakly electric fish, the discharge frequency is controlled by the medullary pacemaker nucleus and is higher in males than in females. After lowering the discharge frequency by chronic administration of β-estradiol, 2-dimensional difference gel electrophoresis revealed 62 proteins spots in tissue samples from the pacemaker nucleus that exhibited significant changes in abundance of >1.5-fold. The 20 identified protein spots indicated, among others, a potential involvement of astrocytes in the establishment of the behavioral dimorphism. Indeed, immunohistochemical analysis demonstrated higher expression of the astrocytic marker protein GFAP and increased gap-junction coupling between astrocytes in females compared with males. We hypothesize that changes in the size of the glial syncytium, glial coupling, and/or number of glia-specific potassium channels lead to alterations in the firing frequency of the pacemaker nucleus via a mechanism mediating the uptake of extracellular potassium ions from the extracellular space.
- sexual dimorphism
- weakly electric fish
- neural oscillator
sex differences in behavior are widespread among animals and occur most commonly during social interactions, particularly in the context of courtship and parental care. It is well-established that sex steroids play an overarching role in the control of this sexual dimorphism of behavior by influencing the organization of brain structures during development and by inducing reversible changes in neural circuits of the adult brain (Baum 2003; Morris et al. 2004). However, whereas sex differences in brain structures that control specific sexually dimorphic behaviors have been well-characterized (Cooke et al. 1998; Kelley 1988), little is known about the multitude of genes and proteins that regulate these behaviors and mediate the activational effects of sex steroids (Xu et al. 2012). To address this issue, we employed an unbiased, large-scale approach to identify such proteins using the electric organ discharge (EOD) of the weakly electric brown ghost knifefish (Apteronotus leptorhynchus) as a behavioral model that exhibits a reversible sexual dimorphism.
In this species, the EOD is produced continuously by synchronous depolarization of modified axonal terminals of spinal electromotor neurons that constitute the electric organ (Bennett 1971; de Oliveira-Castro 1955; Waxman et al. 1972). The frequency of the electric discharge is determined by the firing frequency of pacemaker and relay neurons in the pacemaker nucleus (Pn) in the medulla oblongata (Dye and Meyer 1986). Within the species-specific range of 650–1,000 Hz in adult fish, males discharge at higher frequencies than females (Meyer et al. 1987). The EOD frequency, as well as the firing frequency of the Pn, is sensitive to the action of steroids in adult fish of both sexes, with estrogens inducing decreases in frequency (Meyer et al. 1987; Schaefer and Zakon 1996).
Taking advantage of this steroid-controlled behavioral and neural plasticity, we examined global changes in protein abundance in the Pn after lowering the EOD frequency through implantation of crystalline β-estradiol. Using 2-dimensional difference gel electrophoresis (2-D DIGE), 62 proteins spots that exhibited significant changes in abundance of >1.5-fold were detected. The proteins associated with 20 of these spots were identified via peptide mass fingerprinting (PMF) and/or tandem mass spectrometry (MS/MS). The abundance of one of these proteins, glial fibrillary acidic protein (GFAP), increased almost 2-fold after β-estradiol implantation. Further analysis through immunohistochemistry and confocal laser-scanning microscopy provided evidence that the spatial association of GFAP-expressing astrocytes with pacemaker and relay cells is significantly higher in females compared with males. We hypothesize that such changes in the size and the properties of this glial syncytium lead to alterations in the firing frequency of the Pn via a mechanism involving potassium buffering.
MATERIALS AND METHODS
Brown ghost knifefish (A. leptorhynchus; Gymnotiformes, Teleostei) were supplied by tropical fish importers and maintained in the laboratory as described previously (Gama Salgado and Zupanc 2011). A total of 139 fish (74 males, 64 females, and 1 fish that could not be sexed) were used. Their total length ranged from 80 to 189 mm, and their weight from 0.9 to 14.7 g. The gonadosomatic index ranged from 0.0006 to 0.0042 in males and from 0.0021 to 0.0599 in females. Animal care and procedures were conducted in accordance with the German Animal Welfare Act (Deutsches Tierschutzgesetz) of 1998 and approved by the local authorities, the Senate of the Free Hanseatic City of Bremen.
Differential recording of the fish's EOD and determination of the frequency of the signal were performed as described previously (Gama Salgado and Zupanc 2011). The EOD frequency was adjusted to an ambient temperature of 26°C using a Q10 of 1.56 (Zupanc et al. 2003).
Treatment with β-estradiol.
Under general anesthesia with 2% ethyl carbamate (urethane; Sigma-Aldrich) dissolved in aquarium water, fish were implanted with fine-bore nylon tubing (flexible grade; length = 6 mm; outer diameter = 0.94 mm; inner diameter = 0.75 mm; Smiths Medical International) packed with β-estradiol powder (Sigma-Aldrich). Implants were placed in the peritoneal cavity, and the wound was closed with Histoacryl Blue Topical Skin Adhesive (B. Braun Melsungen). Control fish received empty implants.
Isolation of Pn tissue.
Fish were killed by immersion into an overdose of a 1.5% solution of ethyl 3-aminobenzoate methanesulfonate (MS-222; Sigma-Aldrich) dissolved in aquarium water, and the heads were cooled with ice. The Pn was removed and immediately frozen in isopentane at −45°C, pooled with Pn of other fish, and stored at −80°C until further use.
Protein extraction, protein labeling with CyDye DIGE fluors, and separation of protein extract by 2-D DIGE were performed using the Ettan DIGE system (Amersham Biosciences/GE Healthcare) as described previously (Ilieş et al. 2012). Briefly, pooled tissue samples from β-estradiol-treated and control fish, respectively, were homogenized through sonication in lysis buffer (Amersham Biosciences). After centrifugation, supernatant containing 50 μg of protein from each sample were labeled with 400 pmol of Cy3 and Cy5, respectively. For the internal standard, 25 μg of protein from each sample were combined and then labeled with 400 pmol of Cy2. Finally, the two samples and the standard were pooled and used for 2-D-PAGE. Experiments were run in triplicate, separately for cytosolic and membrane fractions.
Analysis of protein spots.
The labeled gels were digitized using an Ettan DIGE Imager (GE Healthcare) at a resolution of 100 μm per pixel. In-gel multidye codetection of protein spots and quantification of protein abundance as well as between-gels matching of spots and calculation of average fold changes were performed using DeCyder 2D Software (Amersham Biosciences), separately for the cytosolic and membrane fractions. The matching of all protein spots exhibiting changes in abundance of >1.5-fold was verified manually. Statistical analysis was limited to protein spots found in all three replicates of either fraction.
Protein spots showing significant increases or decreases of >1.5-fold in either the cytosolic or the membrane fraction were selected for identification. Preparative gels were run as described above using 500 μg of total protein. After Coomassie staining, spot maps were matched against reference spot maps determined from the analytical gels. Matched spots of interest were excised, destained, and subjected to in-gel digestion with modified trypsin (Roche Diagnostics) overnight at 37°C. The extracted peptides were desalted using a C18 ZipTip (Millipore) and identified by PMF using a 4800 MALDI TOF/TOF Analyzer (Applied Biosystems) as described previously (Ilieş et al. 2012).
Fish were deeply anesthetized in a 2% solution of MS-222 in aquarium water and intracardially perfused with 2% freshly depolymerized paraformaldehyde (Fisher) in 0.1 M phosphate buffer, pH 7.4. The brain was cryosectioned coronally at a thickness of 16 μm. GFAP, HuC/D, and connexin-43 antigenic sites were labeled using rabbit anti-GFAP (Sigma-Aldrich) or chicken anti-GFAP (Abcam), mouse anti-HuC/D (clone 16A11; Invitrogen), and rabbit anti-connexin-43 (Cell Signaling) primary antibodies followed by goat anti-rabbit IgG conjugated to Alexa Fluor 488 or Alexa Fluor 546, goat anti-chicken IgG conjugated to Alexa Fluor 488, and goat anti-mouse conjugated to Alexa Fluor 635 secondary antibodies (all from Invitrogen). The sections were counterstained with 4′,6′-diamidino-2-phenylindole (DAPI).
Microscopy and image analysis.
Confocal microscopy was performed using Zeiss LSM 700 and Zeiss LSM 710 laser-scanning microscopes equipped with ×25 and ×63 objectives. Optical sections were taken at a resolution of 0.2–0.5 μm per pixel using ZEN (Carl Zeiss) software. Images were reconstructed in ImageJ (National Institutes of Health) using the Stitching plugin (Preibisch et al. 2009). Background subtraction through top-hat filtering, definition of regions of interest based on HuC/D labeling, and subsequent quantification of GFAP, HuC/D, and connexin-43 immunolabeling were performed in MATLAB (MathWorks) using built-in functions from the Image Processing Toolbox.
Sexual dimorphism in EOD frequency.
To confirm that brown ghost knifefish males and females occupy different frequency domains, the EOD of 83 individuals was recorded, and they were sexed through gonadal inspection. Analysis of the EOD frequencies adjusted to an ambient temperature of 26°C demonstrated a distinct sexual dimorphism in the frequency of the electric discharges (Fig. 1A). Whereas the mean EOD frequency of males was 879 Hz (median: 892 Hz; range: 704–987 Hz; n = 35 fish), the mean frequency of females was 737 Hz (median: 737 Hz; range: 671–809 Hz; n = 48 fish). This difference was highly significant (P < 0.0001, independent-samples t-test).
Effect of β-estradiol on EOD frequency.
Next, we experimentally manipulated the EOD frequency of both males and females by intraperitoneal implantation of tubings filled with β-estradiol, and we compared the changes in EOD frequency of these fish with the changes in frequency of control fish that had received empty implants. Administration of β-estradiol resulted in a gradual decrease of the EOD frequency in each of the fish examined (Fig. 1B). Eight days after the implantation, the frequency of β-estradiol-treated fish was on average 94 ± 23 Hz lower than the preimplantation baseline (P < 0.001, z-test, Bonferroni correction for multiple comparisons; n = 14 fish). In all treated fish, both male and female, the observed frequency decrease was proportional to the baseline EOD frequency with higher-frequency EODs showing larger reductions (Pearson ρ = 0.69, P < 0.01). By contrast, the EOD frequency of control fish did not exhibit any marked decrease beyond the initial effect of the implantation (Fig. 1B). Similar changes in the EOD frequency of control fish were reported previously (e.g., Meyer et al. 1987) and are likely due to the traumatic effects of the implant surgery. Eight days after implantation, the frequency of control fish was on average 23 ± 16 Hz lower than the baseline (n = 17 fish), a significantly smaller effect than that observed in β-estradiol-treated fish (P < 0.001, independent-samples t-test, Bonferroni correction for multiple comparisons).
Protein expression profiles in the Pn: differences between β-estradiol-treated fish and controls.
Eight days after the implantation of β-estradiol-filled or empty tubings, tissue was collected from whole Pn and processed for 2-D DIGE. An average number of 3,248 ± 199 and 3,006 ± 160 protein spots were detected in the gels run on the cytosolic and membrane fractions, respectively. A total of 1,145 and 819 spots were matched across the 3 2-D gels of the cytosolic and membrane fractions, respectively. In the cytosolic fraction, out of the 1,145 protein spots, the standardized protein abundances of 14 spots (1.2% of total) were significantly increased by a factor of at least 1.5 in the Pn of β-estradiol-treated fish relative to the Pn of control fish, whereas the abundances of 10 spots (0.9%) were significantly reduced by a factor of at least 1.5 (P < 0.05, independent-samples t-test). In the membrane fraction, out of the 819 protein spots, the standardized protein abundances of 38 spots (4.6%) were significantly reduced by a factor of at least 1.5 (P < 0.05) in the Pn of β-estradiol-treated fish relative to the Pn of control fish, whereas none of the protein spots showing significantly increased abundances reached the 1.5-fold threshold (Fig. 2A). The protein spots displaying these differences were located throughout the 2-D gels, thus covering both small and large molecular weight domains as well as the entire isoelectric point (pI) range analyzed.
Identification of differentially expressed proteins.
Out of the 62 protein spots that exhibited significant changes in abundance of >1.5-fold in either the cytosolic or the membrane fraction, the proteins associated with 20 spots could be identified via PMF and/or MS/MS (Table 1; Fig. 2, B and C). The following proteins showed an increase in abundance: fatty acid binding protein 11a (1 spot); GFAP (1 spot); NADH dehydrogenase (ubiquinone) Fe-S protein 1 (1 spot); pyruvate kinase isozymes M1/M2 (1 spot). The following identified proteins showed a decrease in abundance: actin-related protein 2-A (1 spot); beta-actin-1 (bactin1 protein; 2 spots); fascin (1 spot); glutamine synthetase (1 spot); internexin neuronal intermediate filament protein, alpha (1 spot); internexin neuronal intermediate filament protein, alpha b (gefiltin; 1 spot); intraflagellar transport protein 81 homolog (IFT81; 1 spot); isovaleryl-CoA dehydrogenase, mitochondrial (1 spot); 3-oxoacid CoA transferase 1a (1 spot); plastin 2 (1 spot); protein phosphatase 2 (formerly 2A) regulatory subunit A (PR 65) alpha isoform (1 spot); stathmin 1b (1 spot); succinate-CoA ligase, ADP-forming, beta subunit (1 spot); vesicle-fusing ATPase (2 spots).
Gross morphology of the Pn: absence of sex differences.
Among the proteins that exhibited significant alterations in abundance after β-estradiol implantation, several are known to be involved in structural plasticity. Therefore, we examined whether any sexual dimorphism exists in the gross morphology of the Pn and its two major types of neurons, pacemaker and relay cells. Alternate 30-μm transverse brain sections from three males and three females were immunostained against the neuronal marker HuC/D and analyzed (n = 14–18 sections per fish). No significant differences could be detected between males and females in the volume of the Pn, the numbers of relay and pacemaker neurons, and the average profile areas of these two neuronal cell types (Table 2; P > 0.50, independent-samples t-test).
Differences in GFAP expression between males and females.
As shown by 2-D DIGE, the expression of GFAP in the Pn increased almost twofold after β-estradiol implantation. Assuming that similar changes occur during normal sexual maturation and persist as a sexually dimorphic trait in adults, we hypothesized that the expression of GFAP in the Pn is higher in females than in males. In agreement with this hypothesis, confocal microscopy revealed numerous intensely labeled fibers in females (Fig. 3A) but fewer fibers and a lower labeling intensity in males (Fig. 3B). The total GFAP labeling (defined as the relative area immunostained multiplied by the average labeling intensity after background correction) in the whole Pn was almost twice as high in females compared with males (P < 0.05, independent-samples t-test; n = 3 males and 3 females; Fig. 3C, left).
Next, we examined possible differences in the association between GFAP-labeled fibers and neurons in the Pn identified by immunostaining against HuC/D. The total GFAP labeling associated with the area covered by pacemaker and relay cells was significantly higher in females than in males (P < 0.001; Fig. 3C, right). Similarly, the total GFAP labeling associated with the area covered by small interneurons (Turner and Moroz 1995) was higher in females than in males, but this difference was not significant (P > 0.10; Fig. 3C, middle).
Differences between males and females in connexin-43 expression associated with GFAP-labeled astrocytes.
To characterize further the glial meshwork in which the pacemaker and relay cells are embedded, sections through the Pn were immunostained against connexin-43, a member of the connexin family of transmembrane gap junction proteins. Glia, and in particular astrocytes, have been found to express high levels of connexins, which play a critical role in the formation of a glial syncytium (Giaume and Liu 2012). Indeed, an abundance of punctate, plaquelike connexin-43 immunofluorescence was found in close proximity to GFAP-labeled fibers (Fig. 4A). The relative area covered by connexin-43 immunoreactivity in a 20-μm-thick annular region around the HuC/D-immunolabeled pacemaker cells was almost 3-fold larger in females than in males (P < 0.01, independent-samples t-test; n = 4 males and 4 females; Fig. 4, B–C). The connexin-43-labeled area around relay cells was also increased in females, relative to males, but this difference was not significant (P > 0.10).
Proteomics approach to study the development of a sexually dimorphic behavior.
A better understanding of how sexually dimorphic behaviors develop demands an integrative strategy, combining behavioral studies with investigations at the cellular and molecular levels (Zupanc 2010). Whereas significant progress has been made over the past few decades at the behavioral, endocrinological, and, partly, cellular levels (Cooke et. al. 1998), the molecular basis of sexually dimorphic behaviors remains largely enigmatic. Here, we have, to our knowledge, for the first time, employed proteomics to carry out an unbiased large-scale identification of proteins potentially involved in the development of the neural correlates underlying sexual dimorphism in a specific behavioral pattern. The present study serves as proof-of-principle that proteomic analysis represents a powerful method for future investigations that aim to explore the neural basis of behavioral plasticity.
Potential candidates involved in the development of sex differences in EOD frequency, as revealed through proteomic analysis.
Proteomic analysis, combined with PMF and/or MS/MS, enabled us to identify the proteins associated with 20 spots for which abundance was significantly altered by a factor of at least 1.5 after implantation of β-estradiol. Several of these proteins, including protein phosphatase 2, plastin-2, stathmin 1b, internexin alpha and alpha b, fascin, and beta-actin-1, indicate potential changes in the morphology of cells. If such structural changes indeed occur, they should be accompanied by an increase in energy demand and higher metabolic activity after β-estradiol administration, proposed changes that appear to be reflected by the elevated levels of NADH dehydrogenase Fe-S protein 1, pyruvate kinase isoenzymes M1/M2, and fatty acid binding protein 11a. We examined whether the structural changes indicated by proteomics analysis result in sexual dimorphism in the size of the Pn or in the number and gross morphology of the pacemaker and relay cells, but we failed to detect any significant differences between males and females. We, therefore, hypothesize that, if structural changes indeed occur, they involve either a reorganization of neurons at the axonal and dendritic levels or structural changes of nonneuronal cells. The latter hypothesis is consistent with the observed increase in GFAP abundance after β-estradiol treatment, possibly reflecting the generation of new glial cells, or the outgrowth of the existing glia. The proposed functional significance of such a remodeling of glia for the firing pattern of the neuronal network of the Pn is discussed in the next section.
The decrease in the abundance of glutamine synthetase after β-estradiol administration, as also revealed through proteomic analysis, could indicate a decrease in synaptically released glutamate, an interpretation that is consistent with the observed reduced abundance of vesicle-fusing ATPase. These changes are likely related to a second sexual dimorphism in the EOD of A. leptorhynchus. When subjected to sensory stimulation with the EOD of a neighboring fish, whose discharge frequency is similar to their own, males respond with the production of transient amplitude and frequency modulations called chirps. By contrast, females do not exhibit such a behavior (Dulka and Maler 1994; Dunlap et al. 1998; Zupanc and Maler 1993). Chirps are controlled by a subpopulation of neurons in the central posterior/prepacemaker nucleus, the CP/PPn-C (Heiligenberg et al. 1981; Kawasaki and Heiligenberg 1988; Kawasaki et al. 1988; Metzner 1999; Zupanc 2002; Zupanc and Heiligenberg 1992; Zupanc and Maler 1997). The cells comprising this neuronal cluster project to relay cells in the Pn where they make glutamatergic synaptic contact involving non-NMDA glutamate receptors (Dye et al. 1989). This synaptic input from the CP/PPn-C results in rapid depolarization of the relay cells, which in turn leads to an acceleration of the firing frequency of both the pacemaker cells and the relay cells (Dye 1988), likely due to the extensive gap-junction coupling between these neurons (Bennett et al. 1967; Elekes and Szabo 1985; Moortgat et al. 2000; Tokunaga et al. 1980). Although never examined, it is plausible that the glutamatergic input from the CP/PPn-C to the relay cells of the Pn is much weaker in females than in males. If this is indeed the case, such a difference would explain the decrease revealed by proteomic analysis in abundance of glutamine synthetase and vesicle-fusing ATPase after β-estradiol administration.
Proposed role of GFAP-expressing astrocytes in the development of the sexual dimorphism in EOD frequency.
By combining chronic administration of β-estradiol with proteomic analysis, we showed that 8 days after the start of the treatment the abundance of GFAP was significantly increased in the Pn. Complementary immunohistochemical experiments demonstrated a corresponding sexual dimorphism in GFAP expression in the Pn. The surface area of pacemaker and relay cells covered by GFAP-immunopositive astrocytes was markedly higher in females than in males. Notably, this effect did not extend to the third neuronal type in the Pn, the small interneurons.
Confocal images showed that both pacemaker and relay cells are embedded in a dense meshwork of GFAP-expressing astrocytes, which form a gap-junction-coupled syncytium, as suggested by abundant expression of connexin-43 along GFAP-labeled processes. Notably, this coupling, particularly between those astrocytes that are associated with pacemaker cells, is markedly more pronounced in females than in males, as suggested by the quantitative analysis of connexin-43 labeling. The immunohistochemical demonstration of a dense astrocytic meshwork in the Pn is in agreement with prior ultrastructural studies that noted an intimate association of neurons and astrocytes in the Pn, particularly in the region of axon initial segments (Elekes and Szabo 1985; Tokunaga et al. 1980).
Physiological and immunohistochemical studies have demonstrated that three types of ions play important roles in the regulation of the electric activity of the Pn, including modulation of the firing frequencies, Na+, K+, and Ca2+ (Dye 1991; Smith and Zakon 2000). Glial processes surrounding the pacemaker and relay cells express voltage-gated potassium channels (Smith et al. 2006). Based on these physiological and morphological observations, we propose that the astrocytic syncytium plays a role in the regulation of the firing frequency of the Pn and that this control is mediated via uptake of K+ ions from the extracellular space through transporters or K+-sensitive channels.
Potassium buffering by glia is a well-documented phenomenon (Kofuji and Newman 2004), but its possible involvement in the regulation of the firing frequency in neuronal oscillatory networks has, to the best of our knowledge, never been examined. However, computer simulations of neuron-glia interactions mediated by ion flux have indicated that the firing pattern of tonically stimulated neurons can be altered by modifying the properties of adjacent astrocytes, including the density of specific types of membrane channels (Somjen et al. 2008). In the Pn, such a glia-mediated mechanism could accommodate the decrease in the oscillation frequency observed after chronic administration of β-estradiol, possibly by increasing the efficiency of removal of extracellular K+ ions through changes in one or several properties of the glial syncytium: 1) increase in the size of the glial syncytium; 2) enhancement of the glial coupling through gap junctions; and 3) increase in the number of glia-specific potassium channels and/or Na+/K+ ATPases on the glial membrane.
The present study provides evidence for alterations in the first two properties, as a sexual dimorphism has been found in both the size of the glial syncytium associated with the pacemaker and relay cells and the expression of the gap junction protein connexin-43 in glial cells. Assembly of these molecules into gap junctions and the maintenance of their stability are regulated by phosphorylation (Márques-Rosado et al. 2012). Phosphorylation might also play a prominent part in the modulation of the properties of membrane ion channels expressed in the astrocytic syncytium of the Pn, including potassium channels and/or Na+/K+ ATPases. Abundant evidence in other systems has demonstrated that a variety of ion channels, and perhaps all, are substrates for protein kinases and phosphoprotein phosphatases and that this phosphorylation has profound effects on channel activity (Chen and Roche 2007; Levitan 1994). In the Pn, calcium-dependent phosphorylation plays a critical role in the regulation of N-methyl-d-aspartate (NMDA) receptor-dependent plasticity (George et al. 2011). NMDA receptors expressed on relay cells are activated by glutamatergic input from the sublemniscal prepacemaker nucleus in the mesencephalon, mediating transient upward shifts by a few hertz of the pacemaker frequency.
The complexity of the discussed phenomenon clearly indicates that future investigations in this area will require a multidisciplinary approach combining molecular, cellular, physiological, behavioral, and computational modeling experiments. By identifying the essential players in this process and by dissecting the underlying mechanisms, the study of the Pn will provide an excellent opportunity to achieve a better understanding of the possible role of glia in the regulation of the activity of oscillatory neural networks with implications that could reach significantly beyond the attempt to gain a better understanding of the structure and function of the Pn.
This work was supported by a grant from the Tönjes Vagt Foundation (Germany) and by funds from Northeastern University.
No conflicts of interest, financial or otherwise, are declared by the author(s).
G.K.H.Z. conception and design of research; G.K.H.Z., R.F.S., and M.M.Z. performed experiments; G.K.H.Z., I.I., R.F.S., and M.M.Z. analyzed data; G.K.H.Z., I.I., R.F.S., and M.M.Z. interpreted results of experiments; I.I. and R.F.S. prepared figures; G.K.H.Z. drafted manuscript; G.K.H.Z., I.I., and R.F.S. edited and revised manuscript; G.K.H.Z., I.I., R.F.S., and M.M.Z. approved final version of manuscript.
- Copyright © 2014 the American Physiological Society