Estradiol Feedback Alters Potassium Currents and Firing Properties of Gonadotropin-Releasing Hormone Neurons

Abstract

GnRH neurons are regulated by estradiol feedback through unknown mechanisms. Voltage-gated potassium channels determine the pattern of activity and response to synaptic inputs in many neurons. We used whole-cell patch-clamp to test whether estradiol feedback altered potassium currents in GnRH neurons. Adult mice were ovariectomized and some treated with estradiol implants to suppress reproductive neuroendocrine function; 1 wk later, brain slices were prepared for recording. Estradiol affected the amplitude, decay time, and the voltage dependence of both inactivation and activation of A-type potassium currents in these cells. Estradiol also altered a slowly inactivating current, I_K. The estradiol-induced changes in I_A contributed to marked changes in action potential properties. Estradiol increased excitability in GnRH neurons, decreasing both threshold and latency for action potential generation. To test whether estradiol altered phosphorylation of the channels or associated proteins, the broad-spectrum kinase inhibitor H7 was included in the recording pipette. H7 acutely reversed some but not all effects of estradiol on potassium currents. Estradiol did not affect I_A or I_K in paraventricular neurosecretory neurons, demonstrating a degree of specificity in these effects. Potassium channels are thus one target for estradiol regulation of GnRH neurons; this regulation involves changes in phosphorylation of potassium channel components.

GnRH NEURONS form the final common pathway determining fertility in all vertebrate species. GnRH induces synthesis and release of the pituitary gonadotropins LH and FSH, which subsequently activate the gonadal functions of gametogenesis and steroidogenesis; these steroids then engage feedback loops, acting centrally and at the pituitary to regulate the reproductive neuroendocrine axis. Before puberty, during the nonbreeding season and during parts of the reproductive cycle in breeding females, GnRH release is suppressed by estradiol (E)-negative feedback (1–3). In contrast, E-positive feedback induces the surge in GnRH release that is required for ovulation (4). Transitions between reproductive states are postulated to involve changes in sensitivity of GnRH neurons to these feedback actions, but mechanisms have rarely been evaluated at the level of the GnRH neuron itself (5, 6) and never in response to controlled in vivo treatment.

Because of the critical role potassium channels play in setting resting potential, sculpting action potentials and in determining firing rate and responsiveness to synaptic input (7), we hypothesized that a critical component of E feedback would be a change in the functional properties of these channels. To test this, voltage-gated potassium currents and firing properties of GnRH neurons were studied in acute brain slices prepared from mice treated with and without a constant physiological E feedback signal in vivo. Possible mechanisms of E action were assessed by determining the effects of kinase inhibition on the ability of E to alter these currents. To assess the specificity of E feedback actions, the response of GnRH neurons was compared with that of neuroendocrine neurons of the paraventricular nucleus of the hypothalamus.

RESULTS

Animal Model

Adult female mice expressing green fluorescent protein (GFP) under the control of the GnRH promoter (GnRH-GFP mice) (8) were ovariectomized (OVX) and half were treated at the time of surgery with a SILASTIC-brand capsule (Dow Corning, Midland, MI) providing constant physiological dose of E (OVX + E); mice were studied 6–9 d later to avoid acute effects of steroid manipulation. Serum E in OVX + E mice was 30.8 ± 6.1 pg/ml, similar to values reported during the estrous cycle of young adult mice (9). As expected, this E treatment suppressed LH (OVX 3.32 ± 0.29 ng/ml, n = 9; OVX + E 0.07 ± 0.01 ng/ml, n = 8, P < 0.01), implying negative feedback on GnRH release.

In Vivo E Alters Voltage-Gated Potassium Currents in GnRH Neurons

Two types of voltage-gated potassium channels were studied using whole-cell voltage clamp, noninactivating and inactivating (7). Both channels have gates that are sensitive to membrane potential and must be open for current to flow. Noninactivating channels have a single activation gate that is opened by depolarizing the membrane. Inactivating channels have a similar activation gate but also have an inactivation gate that is opened by hyperpolarizing the membrane. The different voltage dependencies of these channels makes it possible to isolate the types of currents present by manipulating membrane potential.

Recordings were obtained from bipolar GnRH neurons identified by the cell-specific expression of GFP in brain slices through the preoptic area and anterior hypothalamus. E was administered only to the animal and was not present during recordings. To isolate potassium currents, 0.5 μm tetrodotoxin and 100 μm CdCl₂ were used to block sodium and calcium channels, respectively. A three-step voltage-clamp protocol was used for identifying components of K⁺ currents (Fig. 1A). Cells were first stepped to −110 mV for 100 msec to remove inactivation from voltage-dependent channels (i.e. open inactivation gates). Next, membrane potential was varied during a prepulse of −110 to −20 mV (10 mV increments, 500 msec). Finally, current was measured during a −10-mV test pulse (500 msec). This protocol revealed voltage-gated potassium currents had two components in GnRH neurons from OVX mice (n = 13 cells from 9 animals; Fig. 1B, summary of all measured parameters in Table 1). The dominant component was a rapidly inactivating A-type current (I_A) that was blocked more than 80% by 5 mm 4-aminopyridine (4-AP, n = 4, not shown). I_A was evident during the test pulse after hyperpolarizing prepulses but was inactivated by depolarized prepulse potentials. The residual current remaining after depolarizing prepulse potentials was designated I_K. This current did not achieve steady state during the 500-msec test pulse; inactivation of this current to 20% of the peak occurred in the range of 5–10 sec when examined with extended test pulses (data not shown). I_K was blocked by bath application of tetraethylammonium (20 mm, n = 4, not shown).

Figure 1.

E Modulates Voltage-Gated Potassium Currents in GnRH Neurons A, Voltage-clamp protocol used for identifying components of K⁺ currents; for clarity, only three prepulse potentials are shown. B, Representative example of K⁺ currents in GnRH neurons from OVX mice illustrating two components: a fast transient A current (I_A) and a residual current (I_K). C, E treatment in vivo altered both I_A and I_K.. D and E, Representative recordings from paraventricular neuroendocrine neurons showing I_A and I_K and the lack of effect of E.

Table 1.

GnRH Neurons

Passive Properties	GnRH OVX (n = 13)	GnRH OVX + E (n = 13)	GnRH OVX + E + H7 (n = 5)
Input resistance (MΩ)	772 ± 69	866 ± 58	866 ± 77
Capacitance (pF)	16.1 ± 0.9	16.2 ± 1.2	13.9 ± 1.2
IA
Peak current (pA)	2387 ± 123	1816 ± 194a	2377 ± 387
Peak current density (pA/pF)	153 ± 11	113 ± 10a	168 ± 17
20–80% rise (msec)	0.52 ± 0.03	0.59 ± 0.05	0.51 ± 0.06
20–80% decay (msec)	27.0 ± 4.7	49.3 ± 7.1b	20.6 ± 3.4
Decay time constant (msec)	10.00 ± 1.32	21.64 ± 1.99b	11.88 ± 1.92
Imax (pA)	2401 ± 122	1807 ± 198a	2382 ± 395
V1/2 inactivation (mV)	−69.2 ± 1.8c	−58.4 ± 1.4	−61.0 ± 2.6
Steepness, s (pA/mV)	−4.80 ± 0.17	−5.27 ± 0.25	−5.85 ± 0.19d
IK
Peak current (pA)	1092 ± 123	697 ± 82a	1142 ± 154
Peak current density (pA/pF)	68 ± 7	42 ± 4a	82 ± 7
% Inactivation	28 ± 4	16 ± 3a	25 ± 2

Passive properties and potassium currents in GnRH neurons.

^a

P < 0.05 vs. OVX and OVX + E + H7.

^b

P < 0.01 vs. OVX + E and OVX + E + H7.

^c

P < 0.05 vs. OVX + E and OVX + E + H7.

^d

P < 0.05 vs. OVX and OVX + E by two-tailed t test.

E treatment in vivo markedly altered both of these currents (n = 13 cells from 8 animals, Fig. 1C). I_A current density (Fig. 2A) was reduced by approximately 25%, and the decay time constant (Fig. 2B) was more than doubled in GnRH neurons from mice treated with E. As in OVX mice, 5 mm 4-AP blocked

Figure 2.

Summary of E and H7 Effects on K⁺ Currents in GnRH Neurons A, E decreased the current density of I_A and I_K. B, I_A decay time constant obtained from single exponential fit of isolated IA (see Fig. 5). C, Percent inactivation of I_K during the 500 msec test pulse. *, P < 0.05; **, P < 0.01.

80% of I_A. This dose is similar to that reported for other hypothalamic neuroendocrine cells (10). E also altered the residual I_K. I_K current density was decreased by approximately 33% (Fig. 2A). Furthermore, the percent inactivation of I_K over the 500-msec test pulse was reduced by E (Fig. 2C). These changes in I_A and I_K reflect alterations in channel properties as E had no effect on the passive properties of these cells (Table 1).

E Feedback Does Not Alter I_A in Paraventricular Neuroendocrine Neurons

To address whether or not the effects of E feedback on GnRH neurons are specific or generalized throughout the central nervous system, I_A was measured in neurons of the paraventricular nucleus of the hypothalamus. These cells were selected because they have an I_A of similar magnitude to GnRH neurons and represent a hypothalamic neuroendocrine population that can be identified based upon location, morphology and electrophysiological properties (10). One possible drawback of this choice is that some of these neurons, specifically those that produce oxytocin, have been shown to have different properties in diestrous vs. lactating rats, and thus may be sensitive to reproductive state and possibly steroidal milieu (11). We felt, however, that this drawback was overcome by the ability to record from an identifiable population to obtain a consistent response, an important feature for this control group. In vivo E treatment did not alter any measured parameter of I_A or I_K, although there was a significant increase in capacitance induced by E in these neurons (Fig. 1, D and E; and summary of all measured parameters in Table 2), suggesting the alterations observed in potassium currents in GnRH neurons is a targeted action of E. The lack of effect of E on properties of these paraventricular nucleus neurons is of further interest as it suggests changes in GnRH neurons from E-treated animals is a specific effect of E-17β rather than a nonspecific change due to membrane fluidity.

Table 2.

PVN Neurons

Passive Properties	PVN OVX (n = 6)	PVN OVX + E (n = 5)
Input resistance (MΩ)	1042 ± 122	1129 ± 56
Capacitance (pF)	16.8 ± 0.6	21.0 ± 0.7
IA
Peak current (pA)	3187 ± 175	3711 ± 444
Peak current density (pA/pF)	179 ± 17	144 ± 19
20–80% rise (msec)	0.39 ± 0.03	0.37 ± 0.01
20–80% decay (msec)	16.8 ± 3.0	18.6 ± 1.4
Decay time constant (msec)	6.65 ± 1.10	7.17 ± 0.81
Imax (pA)	3118 ± 164	3690 ± 429
V1/2 inactivation (mV)	−69.1 ± 1.1	−70.6 ± 0.6
Steepness, s (pA/mV)	−5.31 ± 0.24	−5.02 ± 0.22
IK
Peak current (pA)	467 ± 41	512 ± 62
Peak current density (pA/pF)	28.1 ± 3.0	24.2 ± 2.7
% Inactivation	27.0 ± 3.9	23.9 ± 1.9

Passive properties and potassium currents in neurons from the PVN. There were no differences due to estradiol in any parameter except capacitance (^aP = 0.002).

I_A Sets the Firing Properties of GnRH Neurons

Hormone release from neuroendocrine neurons is tightly correlated with firing pattern (12, 13). This makes I_A a particularly interesting target for E regulation in the GnRH neuron because of its role in regulating the temporal firing patterns of neurons through effects on interspike interval, latency to firing and action potential repolarization, e.g. Refs. 10 and 14 . Because of its importance in setting firing properties in many neurons, I_A was isolated by subtraction for further analysis. The voltage at which one half of I_A was inactivated (V_1/2inact) was depolarized by E (P < 0.01, Table 1, Fig. 3A), but steepness (see Materials and Methods, Table 1) was not affected by this treatment. The voltage at which half of I_A was activated (V_1/2act) was also depolarized by E (−25.5 ± 0.6 OVX, n = 5 vs. −21.5 ± 1.1 mV OVX + E, n = 4, P < 0.05, Fig. 3A). In addition, E reduced (P < 0.05) the time constant of recovery from inactivation of I_A (Fig. 3B) from 26.5 ± 2.0 msec (OVX, n = 5) to 17.3 ± 2.7 msec (OVX + E, n = 4).

Figure 3.

E Modulated Activation, Inactivation, and Recovery from Inactivation of I_A A, E depolarized the voltage dependence of both inactivation (V_1/2inact, left as normalized current) and activation (V_1/2act, right as normalized conductance) as a function of membrane potential. B, The time dependence of recovery from inactivation was decreased by E. The normalized amplitude of I_A is plotted as a function of the duration of a step to −80 mV. *, P < 0.05.

The effect of I_A on membrane potential and firing properties of GnRH neurons was studied using current clamp. Injection of 5 pA hyperpolarizing current shifted the membrane potential of a representative GnRH neuron (Fig. 4A) from −75 mV at rest to −90 mV. This increased the contribution of I_A by removing inactivation and increased the latency to firing in response to a 10-pA depolarizing current injection by approximately 20% compared with controls. In contrast, injection of 5 pA of depolarizing current, which shifted membrane potential to −63 mV, halved spike latency compared with controls, consistent with less contribution from I_A due to voltage-dependent inactivation of this current. Blockade of I_A with 5 mm 4-AP also reduced spike latency (Fig. 4A).

Figure 4.

I_A Increases Latency to Action Potential Firing A, The membrane potential was varied before a test current pulse of 10 pA. At the resting potential of −75 mV (0 current injection), the action potential latency was about 100 msec (control). Hyperpolarization increased the latency to action potential firing, whereas depolarization decreased latency to firing. Pharmacological blockade of I_A with 4-AP decreased latency to firing from the resting potential. B and C, E-induced changes in I_A affect the firing properties of GnRH neurons. B, At the resting potential, 10 pA current injections give rise to action potentials with decreased latency in OVX + E mice. C, Latency was significantly decreased in neurons from OVX + E mice. Traces were truncated to emphasize differences in latency.

We hypothesized the properties of I_A in GnRH neurons from OVX animals would increase latency to firing compared with GnRH neurons from OVX + E mice. Results of current clamp experiments supported this hypothesis. Action potential generation was significantly delayed in GnRH neurons from OVX mice (Fig. 4B), whereas in cells from OVX + E mice, action potentials were initiated with short latency (Fig. 4C). Larger amplitude current injections decreased the latency in both groups; the delay to firing was still greater in cells from OVX mice (not shown). Analysis of firing properties (Table 3) revealed that, in addition to latency, threshold was significantly decreased by E, consistent with an increase in excitability in E-treated mice.

Table 3.

Firing Properties

Firing Properties	OVX (n = 7)	OVX + E (n = 8)
Mean threshold (mV, absolute)	−40.1 ± 1.2	−44.5 ± 1.1a
First spike latency (msec)	234 ± 46	141 ± 17b
Spike amplitude (mV)	80.4 ± 3.0	83.0 ± 5.1
Spike width (msec)	0.871 ± 0.057	0.99 ± 0.13
AHP depth (mV, absolute)	−66.9 ± 1.2	−69.0 ± 2.1
Mean interval (msec)	94.3 ± 6.4	87.2 ± 7.9

Active properties were measured from current clamp recordings of GnRH neurons in slices from OVX and OVX + E mice. Estradiol treatment significantly decreased the threshold and latency to action potential generation, but did not affect other properties.

^a

P < 0.01.

^b

P < 0.05.

E Action on I_A Is Due at Least in Part to Changes in Phosphorylation State

There is increasing evidence that E and other steroids work not only through genomic mechanisms to alter transcription but also at the cell membrane to initiate kinase signaling cascades (15). To test whether the effects of E on the properties of voltage-gated potassium currents in GnRH neurons are due to a change in phosphorylation state, the broad-spectrum kinase inhibitor H7 (20 μm, n = 5 cells) was included in the recording pipette. H7 restored the decay time constant and current density of I_A and current density of I_K in GnRH neurons from OVX + E mice to values observed in OVX mice (Figs. 2 and 5; and Table 1). In contrast, V_1/2inact of I_A was not altered by H7, although the steepness (see Materials and Methods) of the inactivation curve was significantly increased (Table 1) so that inactivation of I_A occurred over a smaller voltage range (Table 1). H7 did not alter I_A in OVX mice (n = 2 cells, 2 mice), nor did H7 affect the passive properties of GnRH neurons (Table 1).

Figure 5.

Inhibiting PKA, PKC, and PKG Activity by Including the Kinase Inhibitor H7 in the Patch Pipette Selectively Reversed Some Effects of E Normalization of isolated I_A illustrates the effects of E and kinase inhibition on the decay time constant.

DISCUSSION

To understand the role of voltage-gated potassium channels in the physiology of GnRH neurons and that of E feedback in modulating currents through these channels, a whole-cell patch-clamp approach was used. I_A is the major component of the voltage-gated potassium current in GnRH neurons, and both I_A and I_K are modulated by in vivo treatment with E. Taken together with the observation that neither I_A nor I_K was affected by E in neurosecretory neurons of the paraventricular hypothalamus, these data suggest potassium channels in GnRH neurons are physiological targets for homeostatic E feedback regulation that is, at least in part, specific to GnRH neurons. The difference between our observed results of changes in potassium conductance that lead to increased excitability and the logical hypothesis that negative feedback effects of E on voltage-gated potassium currents would reduce the excitability of these cells, serves to emphasize the importance of evaluating the cellular mechanisms of steroid action directly at the GnRH neuron.

I_A measured in the present study is an order of magnitude larger than was reported for immortalized GnRH neurons (16, 17). This is perhaps due to the immature or transformed nature of GT1 cells as the magnitude of I_A in GnRH neurons is similar to that in other neurons in slice preparations in hypothalamus (10, 18) and in neocortex (19). In support of this notion, I_A measured in cultured embryonic GnRH neurons from the olfactory placode was also severalfold smaller than I_A in the mature GnRH neurons in the present study (20). The robust nature of this current is striking given the large input resistance of these cells (∼800 MΩ), demonstrating I_A is poised to exert a major effect on action potential generation upon activation even at membrane potentials near the V_1/2inact and V_1/2act. Consistent with this, reducing the contribution of I_A by inactivation or pharmacological blockade increased excitability of GnRH neurons as it does in a variety of neuronal phenotypes (10, 18, 21, 22).

The physiological E regimen tested in the present study significantly reduced the contribution of I_A in GnRH neurons. When both the voltage dependence of I_A and the current density are considered, the net effect of E feedback on I_A is greater near threshold than at peak currents in these cells. At threshold, the availability of I_A in GnRH neurons from OVX mice is almost twice that of GnRH neurons from OVX + E mice; at the peak of the action potential, the availability of I_A in GnRH neurons from OVX animals is only approximately 25% greater. This difference between the influence of E on I_A at threshold vs. peak is reflected in the action potential properties. Specifically, E reduced latency to firing and hyperpolarized threshold in GnRH neurons, whereas a difference in spike width, which would be consistent with a different contribution of I_A at the peak of the action potential, was not detected.

The E-induced changes observed in I_A initially appear paradoxical with regard to the suppression of LH (and presumably GnRH) release observed in this animal model as the E-induced changes in voltage-gated potassium channels increased the excitability of GnRH neurons. In this regard, alterations in voltage-gated potassium channels are likely only one of several physiological changes in the GnRH neurosecretory system brought about by E. Other possibilities include changes in synaptic input and postsynaptic response. E has been associated with synaptic plasticity and increases spine density in hippocampal neurons (23). Consistent with this, gonadectomy reduces spineyness of GnRH neurons (24, 25). We have recently completed characterization of the effects of E on the long-term firing patterns of GnRH neurons in this same animal model (26). In those studies, E lengthened the interval between episodes of increased firing rate, consistent with a negative feedback effect and the observed reduction in LH levels. Of interest, blockade of ionotropic GABAergic and glutamatergic receptors acutely reversed the effects of E in half of the cells, suggesting GABAergic and/or glutamatergic neurons are involved in conveying some facets of E feedback to a substantial subset of GnRH neurons. These data are important to the interpretation of the present study for two reasons. First, they provide evidence that the integrated response of the GnRH neuron in terms of firing pattern reflects a negative feedback model. Second, they suggest alterations in synaptic inputs (in conjunction with changes in potassium currents) are important in producing the overall negative feedback state. With regard to the latter point, preliminary findings from this same animal model demonstrate an E-induced change in the frequency of GABAergic postsynaptic currents in GFP-identified GnRH neurons (27).

In addition to the outward potassium currents described in the present study and the synaptic influences suggested above, E feedback may also alter other conductance within GnRH neurons. GnRH neurons and immortalized GT1 cells have been shown to express a variety of ion channels, including cyclic nucleotide-gated channels (28), sodium (17, 29), calcium (20, 22, 30) and calcium-activated potassium channels (31), any or all of which may be modulated by E feedback. In further agreement with the notion that E may target multiple conductances in GnRH neurons, acute treatment of GnRH neurons in guinea pigs with E reduced firing rate (5), an observation later attributed to activation of a G protein-coupled inward rectifier (6).

In the present study, neither passive properties nor action potential properties aside from latency and threshold were changed. This suggests changes in I_A and I_K are the main factors that underlie differences in firing properties of GnRH neurons between OVX and OVX + E mice. While sodium channels help set threshold and latency, they are also important determinants of the rate of rise and amplitude of action potentials (30, 32)—neither parameter was affected by in vivo E treatment. Low voltage-activated calcium channels are active near threshold in many neurons. An increase in this current could decrease latency as observed in GnRH neurons from OVX + E mice. However, these currents are enhanced by hyperpolarization (via removal of inactivation), such that a hyperpolarizing prepulse in current clamp would shorten action potential latency (33). This is opposite to what is illustrated in Fig. 4. High-voltage activated calcium channels are activated by strong depolarization, such as action potentials. Influx of calcium during the action potential can permit activation of calcium-activated potassium channels and often result in a reduction in action potential width or depth of the posthyperpolarization; neither of these was observed (34). Our present results cannot rule out a role for these channels in the firing properties of GnRH neurons, but they are consistent with the hypothesis that I_A is the dominant factor in setting latency and threshold.

Despite the key role of potassium currents in neural function, direct comparison of these currents in the same type of neuron under different physiological conditions is very limited, and changes have not been found (35). The present study is the first demonstration that a central neural potassium current is a target of a physiological hormonal feedback signal in vivo, although pathophysiological examples of alterations in potassium currents in heart (36) and in different models of epilepsy (37, 38) have been described. The acute actions of E on potassium channels both in GnRH neurons, mentioned above, and also in other central neurons are somewhat better studied. For example, in the area postrema, estrogen increased a potassium current, an effect manifest within a minute of treatment and reversible within 3–4 min of wash out (39). Despite a time course consistent with nongenomic action, the pure estrogen receptor antagonist ICI 182,780 blocked this effect. In hypothalamic neurons, estrogen reduces the ability of μ-opioid and GABA_B agonists to activate an inward rectifying potassium channel (40). The effect occurs within 20 min of estrogen exposure and is blocked by ICI 182,780 or inhibitors of protein kinase A or C but not by a protein synthesis inhibitor. Estrogen has also been shown to depolarize unidentified hypothalamic neurons by decreasing a potassium conductance in a cAMP-mediated manner (41). Finally, E transiently increased a delayed rectifier potassium current in cultured ovine gonadotrope cells (42).

As illustrated by these acute actions of E to change potassium conductance, this steroid can act through nongenomic mechanisms such as activation of membrane-associated kinase cascades (15) as well as traditional genomic mechanisms to affect gene expression (43). With regard to the latter, the basic properties of I_A in GnRH neurons from both animal models in the current study resemble those of Kv4.x potassium channel subunits (44). Our results are thus not consistent with an E-mediated change in subunit expression outside of the Kv4.x family. At the molecular level, changes within the Kv4.x family or in the level of expression of channel or accessory subunits could account for our observations. In this regard, E has been reported to alter expression of potassium channels in the heart (45) and uterus (46), consistent with genomic actions.

The animal model used to produce a physiological E feedback precludes testing of rapid effects of E, but channels in the Kv4.x family have multiple sites for phosphorylation by various kinases and kinase inhibition has been reported to alter the V_1/2inact and V_1/2act of A-type currents (47–50). We thus examined the possibility that a shift in the phosphorylation state of I_A channel subunits or other proteins associated with A-type channels could account for the E-induced changes in I_A. Application of the broad-spectrum kinase inhibitor H7 via the patch pipette reversed some but not all effects of E, suggesting a role for phosphorylation in some of these effects. Specifically, the actions of E on current density and decay time of I_A and current density and percent inactivation of I_K were reversed by H7, whereas the V_1/2inact of I_A was unaffected. Interestingly, the steepness of inactivation was greater in GnRH neurons treated with H7. This shift in steepness of the inactivation curve implies the inactivation gate itself is sensitive to phosphorylation, but that phosphorylation of this site is not responsible for the shift in V_1/2inact induced by E. These shifts in phosphorylation of potassium channels or their associated proteins could be due to direct long-term activation of kinase cascades by E as has been shown in hippocampal neurons (51). Alternatively, E may induce a genomic change in the expression of kinases and/or phosphatases in GnRH neurons so that the change in phosphorylation revealed by H7 is secondary to this genomic action. The persistence of the E effect in the brain slice preparation in the absence of E suggests the latter is more likely. Whether this is a direct action of E on GnRH neurons or a synaptically mediated change must now be studied. Although GnRH neurons were long thought to be devoid of estrogen receptors, recent reports indicate these cells express the β isoform of the estrogen receptor (52, 53), suggesting direct action is possible.

The effects of kinase inhibition on voltage-gated potassium currents are of further interest to the function of GnRH neurons. Specifically, these neurons are episodically active (26) and the mechanisms underlying this intermittent activity are hypothesized to be posttranslational, such as changes in phosphorylation of key proteins (54, 55). The ability of a kinase inhibitor to alter the potassium currents exhibited by these cells is consistent with this hypothesis. If phosphorylation cycles play a role in generating the pattern of secretion, the potassium currents measured in a cell may depend upon whether that cell was at an active or quiescent phase of its cycle. This was not possible to monitor in the present study as potassium currents were isolated by inclusion of tetrodotoxin and cadmium. In this regard, however, although there were clear significant differences between GnRH neurons from OVX and OVX + E mice, there was within group variability that may be accounted for by endogenous changes in phosphorylation of potassium channels or associated proteins that are due to the phase of the pulse cycle.

E feedback upon the GnRH neurosecretory system has been classified as positive or negative based on hormone output. The present data at the cellular level suggest a new paradigm concerning the division of feedback actions of E on GnRH neurons. We propose a working model in which some feedback effects of E, such as those on voltage-gated potassium channels in the present study, make them more excitable. Concomitantly, some actions of E inhibit the activity of GnRH neurons. Under this model, shifts between negative and positive feedback actions of E on GnRH release are driven by changes in the predominance of these two modes of E action. More broadly, a similar balance of inhibitory and excitatory E actions may explain the sometimes conflicting reports of central actions of this steroid on learning, memory, and cell survival (56–58), in addition to its roles in reproductive behavior and physiology.

MATERIALS AND METHODS

Animals

The University of Virginia Animal Care and Use Committee approved all procedures. Adult female transgenic mice (8) were used for these studies. Because the goal was to examine the feedback effects of E in the absence of other variables, mice underwent OVX under surgical anesthesia and steroid levels manipulated using constant release sc SILASTIC-brand capsules. The main comparison was between OVX and OVX animals receiving capsules containing 0.625 μg E-17β (OVX + E) at the time of surgery. A study time point of 1 wk ± 2 d after steroid manipulation was chosen to balance sufficient time for steroid withdrawal, while still avoiding effects of long-term steroid deprivation.

Brain Slice Preparation

All chemicals were obtained from Sigma (St. Louis, MO). To prepare slices, the brain was rapidly removed and placed in cold (1–2 C) sucrose-substituted saline containing (in mm) sucrose 250, KCl 3.5, NaHCO₃ 26, MgCl₂ 2.5, MgSO₄ 1.2, NaPO₄ 1.25, and glucose 10. Two hundred-micrometer coronal slices were cut with a vibratome (Ted Pella, Inc., Redding, CA) and stored 15 min at room temperature in 50:50 sucrose saline:normal saline (normal saline was NaCl 125, KCl 3.5, NaHCO₃ 26, CaCl₂ 2.5, MgSO₄ 1.2, Na₂HPO₄ 1.25, and glucose 10), then transferred to normal saline at 32 C for at least 1 h before recording. All solutions were bubbled continuously with a gas mixture of 95% O₂ and 5% CO₂.

Electrophysiological Recordings and Analysis

During recording, slices were continuously superfused at 5 ml/min with oxygenated normal saline kept at 32 C with an inline-heating unit (Warner Instruments, Hamden, CT). GFP-positive cells were visualized with a combination of infrared differential interference contrast and fluorescence microscopy. Pipettes (2–4 MΩ) were filled with (in mm) K-gluconate 140, HEPES 10, EGTA 5, CaCl₂ 0.1, Mg-ATP 4, Na-GTP 0.4, and KCl 5. Voltage-clamp recordings were made with EPC-7 or EPC-8 patch clamp amplifiers (Instrutech, Port Washington, NY). The data acquisition system consisted of an ITC-18 acquisition interface (Instrutech) and Pulse control software (available from Instrutech) running within Igor Pro (Wavemetrics, Lake Oswego, OR) on a Macintosh G4 (Apple, Cupertino, CA). Series resistance was monitored throughout experiments and only stable recordings with less than 20 MΩ were included; series resistance was not compensated. All potentials were corrected for the calculated liquid junction potential of −10 mV [JPCalc (59)], and all traces were subjected to leak subtraction using a P/-4 protocol. To test the role of kinase activity, the broad spectrum kinase inhibitor H7 was included in the recording pipette at 20 μm, a concentration that would block protein kinase A, protein kinase C, and protein kinase G.

Series resistance, input resistance, and capacitance were measured from the averaged membrane response to sixteen 20-msec 5-mV hyperpolarizing voltage steps. Fast transients recorded after formation of the gigaohm seal in the cell-attached configuration were subtracted from the membrane response in the whole-cell configuration to correct for incomplete compensation of electrode capacitance. Series resistance was estimated from the peak of the response to the −5-mV step using Ohm’s law. The decay phase was not well fit by either a single or double exponential equation, consistent with the persistence of neuronal processes in the slice preparation. Capacitance was thus estimated by dividing the integral of the capacitive transient (after zeroing the steady-state component) by the voltage step. Input resistance was calculated from the steady-state current using Ohm’s law. Neither series resistance (OVX, OVX + E, OVX + E + H7; 8.4 ± 0.7, 8.9 ± 0.7, 7.3 ± 1.1 mΩ, respectively) nor passive properties (Table 1) differed among GnRH neurons from the different treatment groups. Of these properties, only capacitance was affected by E in neurons from the paraventricular hypothalamus (series resistance OVX, OVX + E; 12.8 ± 1.2, 11.6 ± 1.2 mΩ, respectively, passive properties in Table 2).

To study potassium currents, slices were superfused with normal saline supplemented with 0.5 μm tetrodotoxin and 100 μm cadmium. Cells were recorded in the whole-cell configuration and held at −70 mV in voltage-clamp mode. Distinct voltage-clamp protocols were used to determine inactivation (initial hyperpolarization to −110 mV for 100 msec, prepulse of −110 to −10 mV in 10-mV increments for 500 msec, test pulse of −10 mV for 500 msec, Fig. 1A), activation (prepulse of −110 mV or −40 mV, test potentials from −80 mV to + 40 mV, 10-mV increments) and recovery from inactivation of potassium currents (initial depolarization to −40 mV to inactivate I_A, prepulse to −80 mV for 1 to 500 msec, test pulse to −10 mV). To further characterize transient vs. steady-state currents, the duration of the test pulse in the inactivation protocol was extended to 2–10 sec in some recordings without leak subtraction. For analysis, I_A was mathematically isolated by making use of the voltage dependence of inactivation. Inactivation was complete at −40 mV; i.e. no fast transient current was present in current traces after the −40-mV prepulse. I_A was isolated by subtracting the current after the −40-mV prepulse from that after more hyperpolarized prepulses. Peak current was normalized by the average of the values after prepulses to −110 and −100 mV. For activation, I_A was isolated by subtracting the current after the −40-mV prepulse from that after the −110-mV prepulse for each test potential. Peak current was normalized and divided by the driving force calculated using the Goldman-Hodgkin-Katz-equation assuming a −110-mV reversal potential (60). Activation and inactivation curves were fit with equation 1 to obtain values for voltage of half activation or inactivation (V_1/2), maximum current (I_max), and steepness (s).

$$I(V){=}I_{\mathrm{max}}/1{+}\mathrm{exp}{[}({-}\mathrm{sF/RT})(\mathrm{V}{-}\mathrm{V}_{1/2}){]},$$

where V is step potential; F, Faraday constant; R, gas constant; and T, temperature in degrees Kelvin. The time constant of recovery was determined from an exponential fit to the peak of I_A during the test pulse plotted as a function of the duration of the prepulse. For the slowly inactivating component, designated I_K, both peak current and current at the end of the test pulse were measured. Percent inactivation over the 500 msec test pulse was determined from these values.

Current clamp recordings were obtained with the EPC-7 patch clamp amplifier. Holding current was adjusted to keep the cells near −70 mV. Cells were injected with current from 0–50 pA in 5- or 10-pA steps for 500 msec. Threshold for action potential generation was determined from the derivative of the voltage trace. The membrane potential at the time point the derivative exceeded 1 V/sec was defined as threshold and is reported as the mean threshold of all spikes during a given 500-msec current injection. First spike latency is defined as the time from start of the current injection to the peak of the first action potential. Spike amplitude is measured as the peak relative to threshold. Spike width is the full width at half amplitude. The depth of the posthyperpolarization was determined from the local minimum 20 msec after the peak. Mean interval refers to the time between peaks averaged for all spikes in a given trace. To compare cells across groups, traces with 4–7 spikes and 8–12 spikes were analyzed for spike latency, threshold and amplitude. Similar results were obtained for the two groups; data from the 4–7 spike groups are in Table 3. Action potential width was less than 1 msec in both groups. Although no significant difference was detected in action potential width, if such a difference were small it may not be detected due to the limitations of using a patch clamp amplifier for current clamp recordings. These limitations would not affect comparison of the other measured properties (61).

Although complete voltage clamp of cells with extensive processes is problematic, conclusions can be drawn if adequate precautions are taken. To assess the quality of voltage control over the currents recorded from GnRH neurons in the present study we compared the 20–80% rise time and decay time of the isolated A-current as a function of pre-pulse potential (similar to Ref. 10). All cells included in the analysis (all recordings that passed series resistance criterion) exhibited less than 20% difference in both these values when −100-mV and −70-mV prepulses were compared. An important control in the present study is the measure of cell capacitance as gonadal steroids have been reported to increase spineyness in GnRH (62) and other (63) neurons. Estimating membrane capacitance from the integral of the capacitive transients in response to a 5-mV voltage step avoids problems associated with fitting exponentials to the decay phase due to differing contributions from proximal and distal membrane compartments. These estimates show there was no difference in capacitance between the two groups, indicating E-induced differences in cell morphology cannot account for the differences in potassium currents. It is also important to point out that errors due to poor space clamp or series resistance result in the membrane being exposed to less than the command potential in the case of outward currents. Two of the main observations of this study (decreased amplitude and increased decay time constant in E-treated animals) go the opposite way than expected for these types of errors. Measurement of smaller currents results in less voltage error, and the decay time constant decreases with increasing voltage steps. It thus stands to reason that if the changes in current were due to inadequate voltage control, faster decay time constants would be expected for smaller currents, rather than the slower decay time constants of smaller currents observed in the present study. Finally, currents in both groups had similar rise times (Table 1), indicating that relocation of channels to a distant membrane compartment cannot account for the reduced amplitude observed in E-treated animals.

Immunoassasy

Trunk blood was collected from each mouse at the time of brain slice preparation. Serum LH concentration was determined by a modified supersensitive two-site sandwich immunoassay described previously (64, 65). Mouse LH reference preparation (AFP-5306A) provided by the National Hormone and Pituitary program was used as standard. The assay sensitivity was 0.07 ng/ml; intraassay and interassay coefficients of variation were 7.7% and 14%, respectively. Serum E was measured using a RIA kit (39100, Diagnostics Systems Laboratories, Inc., Webster, TX) according to manufacturer’s instructions. The small blood volume of mice precluded measuring E in all samples. We used two strategies to get around this difficulty, blood samples from animals treated in an identical manner were assayed for E and pools of remaining serum from the two treatment groups were also assayed. There was no difference between the levels of E detected for these two types of samples. All samples for E were measured in a single assay; the assay sensitivity was 1.5 pg/ml, and the intraassay coefficient of variation was 6.4%.

Analysis

All data are presented as mean ± sem and significance was set at P < 0.05. Potassium currents were compared between groups with two-tailed t test. The effects of changes in potassium currents on firing properties were compared using a one-tailed t test.

Acknowledgments

We thank Gene Block, Dearing Johns, Valerie Long, and Xu-Zhi Xu for technical assistance and Scott Baraban, Lou Byerly, Richard Day, Craig Nunemaker, Shannon Sullivan, and Pei-San Tsai for comments on the manuscript.

This work was supported by NIH HD-34860 and HD-41469 (to S.M.M.), and NICHD/NIH through cooperative agreement (U54-HD-28934) as part of the Specialized Cooperative Centers Program in Reproduction Research.

Abbreviations:

 
• E,

  Estradiol;

 
• GFP,

  green fluorescent protein;

 
• OVX,

  ovariectomized.