HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) All Versions of this Article: 576/3/833    most recent jphysiol.2006.118505v1 Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Schiess, A. R. B. Articles by Partridge, L. D. Search for Related Content PubMed PubMed Citation Articles by Schiess, A. R. B. Articles by Partridge, L. D. Related Collections Neuroscience

¹ Department of Neurosciences, University of New Mexico School of Medicine, Albuquerque, NM 87131, USA

	Abstract

Top Abstract Introduction Methods Results Discussion Appendix References

 
We used Magnesium Green AM to measure Ca²⁺ transients in Schaffer collateral presynaptic terminals simultaneously with postsynaptic field potentials (fEPSPs) to investigate the mechanism of neurosteroid enhancement of short-term synaptic facilitation. Measurement of [Ca²⁺]_i, isolated to presynaptic events, using the fluorescence ratio (F/F₀) demonstrated that at a constant stimulus intensity there was no change in the excitability of presynaptic fibres between paired stimuli or between ACSF and 1 µM pregnenolone sulphate (PREGS). Paired-pulse facilitation (PPF) was correlated with residual Ca²⁺ ([Ca²⁺]_res), and there was an additional increase in the F/F₀ for the [Ca²⁺]_res-subtracted response to the second of paired stimuli, resulting primarily from a slowing of the decay time constant. In addition to the role of presynaptic [Ca²⁺]_res in PPF, we observed a decrease in EC₅₀ and a greater maximum for Hill function fits to fEPSP versus F/F₀ during the second of paired responses. The enhancement of fEPSP PPF by PREGS did not result from an increase of F/F₀. The data presented here support a PREGS-induced increase in presynaptic glutamate release from the second, but not the first, of a pair of stimuli for a given presynaptic [Ca²⁺] because: (a) there is actually a decrease in the F/F₀ of the [Ca²⁺]_res-subtracted second response over that seen in ACSF; (b) PREGS causes no change in presynaptic Ca²⁺ buffering; and (c) there is a decrease in EC₅₀ and an increase of y_max in the Hill function fits to F/F₀ versus fEPSP data. We hypothesize that PREGS enhances short-term facilitation by acting on the Ca²⁺-dependent vesicle release machinery and that this mechanism plays a role in the cognitive effects of this sulphated neurosteroid.

(Received 2 August 2006; accepted after revision 18 August 2006; first published online 24 August 2006)
Corresponding author L. D. Partridge: Department of Neurosciences, University of New Mexico School of Medicine, Albuquerque, NM 87131, USA. Email: dpartridge{at}salud.unm.edu

	Introduction

Top Abstract Introduction Methods Results Discussion Appendix References

 
Neurosteroids are produced in the brain independently of peripheral endocrine glands and act locally in the nervous system. Pregnenolone sulphate (PREGS) is generally reported to be one of the most common neurosteroids in the brain (Mathur et al. 1993); however, see Ebner et al. (2006). Preclinical studies have shown that low hippocampal PREGS levels are associated with memory deficits that can be improved by PREGS injection (Vallee et al. 1997, 2001), prolonged intracerebroventricular injection of PREGS enhances memory performance (Ladurelle et al. 2000), and PREGS and other neurosteroids attenuate amyloid peptide-induced amnesia (Maurice et al. 1998). Although less conclusive, clinical studies have demonstrated a negative correlation between ß-amyloid proteins and PREGS in the brains of Alzheimer's patients (Weill-Engerer et al. 2002) and have demonstrated low plasma PREGS levels in patients with generalized anxiety disorders (Heydari & Le Melledo, 2002).

Paired-pulse facilitation (PPF) is an augmentation of synaptic transmission lasting from tens to hundreds of milliseconds following a single presynaptic stimulus. It is generally accepted that this facilitation is dependent upon residual Ca²⁺ ([Ca²⁺]_res) in the presynaptic terminal (Katz & Miledi, 1968; Cabezas & Buno, 2006). This form of synaptic plasticity is important in the frequency dependency of information processing in the nervous system (Zucker & Regehr, 2002; Thomas et al. 2005), and PPF abnormalities are correlated with impairments in learning and memory (Matilla et al. 1998).

The effects of neurosteroids on postsynaptic GABA_A and NMDA receptors are well documented (Baulieu, 1998). In addition, neurosteroids have been shown to affect the release of neurotransmitters including noradrenaline (Monnet et al. 1995), dopamine (Barrot et al. 1999), acetylcholine (Darnaudery et al. 2000) and glutamate (Partridge & Valenzuela, 2001). Both exogenous and endogenous PREGS were found to act at a presynaptic site to dramatically enhance glutamatergic PPF in mature neurons (Partridge & Valenzuela, 2001; Thomas et al. 2005), but not in immature neurons (Meyer et al. 2002; Mameli et al. 2005). A similar enhancement of PPF occurs with agonists of the ₁ receptor and antagonists of this receptor block the enhancement produced by PREGS; furthermore, prolonged treatment of slices in pertussis toxin blocks the enhancement produced by PREGS (Schiess & Partridge, 2005). These observations indicate a presynaptic site of action of PREGS at a ₁-like receptor and subsequent interaction with a G_i/o signalling pathway. The downstream effects of this second messenger pathway could affect facilitation through [Ca²⁺]_res by altering Ca²⁺ influx or release from stores (Carter et al. 2002; Schneggenburger & Neher, 2005; Cabezas & Buno, 2006), modulation of cytoplasmic Ca²⁺ buffers or extrusion processes (Blatow et al. 2003; Felmy et al. 2003; Matveev et al. 2004), or by modulation of a distinct facilitatory site downstream from [Ca²⁺]_res that is associated with the vesicle release protein complex (Tang et al. 2000; Zucker & Regehr, 2002; Bark et al. 2004).

We have measured simultaneously presynaptic [Ca²⁺]_i in Schaffer collateral terminals and postsynaptic field potentials in CA1 stratum pyramidale in an effort to elucidate mechanisms of facilitation during PPF and its enhancement by PREGS. We conclude that in the adult rat hippocampus, a significant component of the action of PREGS is downstream from [Ca²⁺]_res.

	Methods

Top Abstract Introduction Methods Results Discussion Appendix References

 
Slice preparation and field potential recordings

Experiments were performed in coronal hippocampal slices prepared from approximately 50-day-old Sprague–Dawley rats. Animals were deeply anaesthetized by I.P. injection of 250 mg kg^–1 ketamine, brains were rapidly removed, and slices were cut at 300 µm with a vibroslicer (Pelco 101, St Louis, MO, USA) in an ice bath with a cutting solution containing (mM): 220 sucrose, 3 KCl, 1.2 NaH₂PO₄, 26 NaHCO₃, 12 MgSO₄, 0.2 CaCl₂, 10 glucose and 0.01 mg ml^–1 ketamine equilibrated with 95%O₂–5%CO₂. Slices were then transferred to artificial cerebrospinal fluid (ACSF) containing (mM): 126 NaCl, 3 KCl, 1.25 NaH₂PO₄, 1 MgSO₄, 26 NaHCO₃, 2.5 CaCl₂ and 10 glucose equilibrated with 95%O₂–5%CO₂ at 30°C for 1 h and then maintained at room temperature until recording in a chamber (Warner Instruments, Hamden, CT, USA or Scientific Systems Design, Mercerville, NJ, USA) maintained at 32°C and continuously perfused at 2 ml min^–1 with ACSF saturated with 95%O₂–5%CO₂. All experiments were performed in accordance with the University of New Mexico animal care and use guidelines.

We used standard electrophysiological techniques that are well established in our laboratory for slice field potential recordings in the Schaffer collateral to CA1 pyramidal neuron synapse in the hippocampus (Molnar et al. 2002). Briefly, field potentials (fEPSP) were recorded with an Axoclamp 2B, Multiclamp 700B (both from Axon Instruments, Union City, CA, USA), or Electro 705 (WPI, Sarasota, FL, USA) amplifier and a Digidata 1322A interface using pCLAMP 8.2 or 9.2 software (Axon Instruments) for experimental control and data analysis. Field potentials, recorded in the stratum pyramidale, were digitized at 500 kHz and filtered at 2 kHz. Presynaptic constant current pulses (100 or 150 µs duration) were applied to Schaffer collateral fibres with an Iso-Flex constant current stimulator (API Instruments, Jerusalem, Israel) through a concentric bipolar electrode (FHC, Bowdoinham, ME, USA) at an amplitude adjusted to produce 50% of the maximum fEPSP population spike amplitude. We used two experimental protocols: In the first, ‘ interpulse interval’, protocol, we varied the interpulse interval between 50 and 300 ms in 50 ms increments and kept the stimulus current (I_stim) constant at a level necessary to produce approximately a half-maximal response. In the second, ‘ stimulus intensity’, protocol, we increased the stimulus intensity incrementally, usually from 0.1 to 1.0 mA, until a maximal response was obtained while keeping the interpulse interval at 50 ms.

Field potential population spike amplitudes were measured as the mean of both positive components minus the negative tip of the spike. For simplicity, we have used the abbreviation R1 for the first response to a pair of stimuli and R2 for the second response of the pair. Paired-pulse facilitation of fEPSP population spikes (pop) was calculated as 100 x [pop(R2) – pop(R1)]/pop(R1). In experiments where we altered the extracellular [Ca²⁺] ([Ca²⁺]_o), the paired-pulse ratio (PPR) was measured as pop(R2)/pop(R1). Note that PPF = (PPR – 1) x 100. It was necessary to use PPR rather than PPF in this later series of experiments because, with high [Ca²⁺]_o, PPF was consistently close to zero and therefore inappropriate for use in normalization. The paired-pulse ratio, however, allowed for consistent normalization throughout the range of [Ca²⁺]_o studied.

The paired-pulse ratio was compared at different values of [Ca²⁺]_o to the control PPR at [Ca²⁺]_o = 2.5 mM before and after 1 µM PREGS exposure. The divalent ion concentration was kept constant at 4 mM by adjusting the extracellular [Mg²⁺]_o ([Mg²⁺]_o). As we will consider further in the Discussion, a major contribution to facilitation from buffer occupancy would be expected to yield PPR [Ca²⁺]_o, whereas a minimal contribution from buffer occupancy would be expected to yield PPR 1/[Ca²⁺]_o (Granseth et al. 2002; Blatow et al. 2003; Mori-Kawakami et al. 2003; Wasling et al. 2004).

We have defined as a ‘recording site’ the location of simultaneous presynaptic optical and postsynaptic electrical recordings. We observed variability in the robustness of the PREGS enhancement of facilitation among individual recording sites. In order to assess the effects of PREGS in these studies clearly, we selected recording sites (approximately 2/3 of the total) at which PREGS clearly produced a robust enhancement of fEPSP population spike PPF or PPR. Some of this variability probably resulted from inhomogeniety in synaptic properties of recording sites owing to the location within the CA1 field, the depth below the surface of the slice, or the rostral–caudal level in the hippocampus of the slice.

Presynaptic Ca²⁺ imaging

Presynaptic fibres were filled with the Ca²⁺ fluorophore, Magnesium Green AM (Molecular Probes, Eugene, OR, USA) using a well-established technique (e.g. Regehr & Tank, 1991). Briefly, an ejection electrode (tip diameter, 5–10 µm) containing Magnesium Green AM (0.9 mM Magnesium Green AM, 10% DMSO, 1% pluronic acid in ACSF) was lowered into the fibre pathway between the stimulating electrode and the presynaptic terminal field to be observed. While observing the emission image following 490 nm excitation, an air pressure pulse was applied with a syringe to the ejection electrode until a small bright spot ( 1 µl) was observed in the fibre pathway. The slice was then maintained with a 2 ml min^–1 flow of oxygenated ACSF at 32°C for 1 h to allow intracellular diffusion of the dye to the presynaptic imaging site 500 µm away from the ejection site (Regehr & Tank, 1991; Atluri & Regehr, 1996). The excitation light was then reduced to a 100–200 µm diameter spot with a diaphragm in the epi-illumination path, and the emitted light was measured with a photomultiplier tube (PMT). A single stimulus or pairs of stimuli were delivered orthodromically at 0.05 or 0.067 Hz by a Master 8 pulse generator (API Instruments) under control of the imaging system (TILL Photonics, Pleasanton, CA, USA). Fluorescence responses are reported as the ratio of the change in fluorescence to the prestimulus fluorescence (F/F₀). The F/F₀ signals were corrected for bleaching by subtraction of a linear baseline slope and were inverted so that increasing presynaptic [Ca²⁺]_i produced an upward deflection. Saturation of the fluorophore was assessed in experiments in which paired F/F₀ responses (50 ms interpulse interval) were measured first in normal ACSF and then after increasing [Ca²⁺]_o.

Raising [Ca²⁺]_o from 2.5 to 3.5 mM significantly increased R1 F/F₀ by 14% and R2 F/F₀ by 17% (n = 4, P < 0.0005). This yielded a similar proportion for R2 F/F₀ to R1 F/F₀ (1.31) compared to that in 2.5 mM [Ca²⁺]_o (1.28; n = 4, P = 0.180). Raising [Ca²⁺]_o from 2.5 to 5 mM increased R1 F/F₀ by 39% and R2 F/F₀ by 31% (n = 4, P < 0.05). This also yielded a similar proportion for R2 F/F₀ to R1 F/F₀ (1.25) compared to that in 2.5 mM [Ca²⁺]_o (1.30; n = 4, P = 0.073). The larger effect of 5 mM [Ca²⁺]_o when compared with 3.5 mM [Ca²⁺]_o and the similar percentage increases in R1 and R2 F/F₀ are strong indications that the Magnesium Green AM was not saturated under the conditions of our experiments.

In the presence of 10 µM CNQX, 25 µM D-(1)-2-amino-5-phosphonopentanoic acid (D-AP5) and 20 µM bicuculline, the fEPSP, but not the presynaptic fibre volley, was blocked while the F/F₀ signal was left essentially unchanged; however, subsequent addition of 600 nM TTX blocked both the F/F₀ signal and the presynaptic fibre volley. This is consistent with the measured F/F₀ signal representing predominately [Ca²⁺]_i in the presynaptic Schaffer collateral axons and axon terminals (Wu & Saggau, 1994; Atluri & Regehr, 1996; Sinha et al. 1997; Kamiya & Ozawa, 1999). Furthermore, the larger volume of presynaptic terminals and their higher density of calcium channels relative to that of axons (Regehr & Atluri, 1995) suggest that a major portion of the F/F₀ signal reflects changes in [Ca²⁺]_i of presynaptic boutons ([Ca²⁺]_pre).

Magnesium Green AM, with a K_D of 6 µM for Ca²⁺, minimizes the effect of exogenous buffer on the time course of F/F₀ decay (Regehr & Atluri, 1995). To diminish noise inherent with this low-affinity indicator, it was necessary to average three fluorescence responses and to filter the PMT signal at 1 kHz. We performed two experiments to check that we could accurately measure the derivative of F/F₀ ((F/F₀)/t) under these recording conditions (Wu & Saggau, 1994; Sabatini & Regehr, 1998). First, the derivative of a LED light pulse with a 10–90% rise time of 500 µs was two to 10 times greater than the amplitude of (F/F₀)/t determined from presynaptic F/F₀ signals. Second, increasing [Ca²⁺]_o produced the expected increase in (F/F₀)/t.

Presynaptic fibre excitability

Changes in F/F₀ may reflect either changes in the number of presynaptic fibres recruited or in the [Ca²⁺]_pre within individual boutons of these fibres. When the stimulus electrode position and I_stim are unchanged, however, only a change in fibre excitability should lead to a change in the number of fibres recruited. We used two tests to rule out significant contributions to our data from changes in fibre excitability. First, we calculated (F/F₀)/t as an index of the number of voltage-gated Ca²⁺ channels activated and found less than a 5% change in this measure of excitability either between the R1 and R2 or between ACSF and PREGS at a fixed stimulus electrode position and I_stim. Second, since a change in the presynaptic fibre volley amplitude indicates a direct change in axonal excitability (Lante et al. 2006; Winegar & MacIver, 2006), we measured the presynaptic fibre volley amplitude as the difference between the positive and negative excursion in traces where there was a clear separation of fibre volley from the stimulus artifact. We generated input–output curves for the presynaptic fibre volley versus I_stim and found no change in the shape or magnitude of these curves between R1 and R2 or between ACSF and PREGS (data not shown). Similar measurements of presynaptic excitability for this synapse have been used to show that K⁺ channel block enhances presynaptic Ca²⁺ influx and not the number of fibres recruited (Qian & Saggau, 1999).

Data analysis

Input-output relationships were fit using a least squares regression routine to Hill function of the form:

(1)

where y is the fEPSP population spike amplitude at a given F/F₀, y_max is the maximum fEPSP population spike amplitude, EC₅₀ is a measure of [Ca²⁺]_pre at half-maximal response as determined by F/F₀, and n_H is the Hill coefficient. Average data are presented as means ± S.E.M., and statistical significance was determined at P < 0.05. Goodness of fit for least squares regression fits to data is given by the coefficient of determination (c.d.).

Drugs

Drugs were stored frozen in aliquots and diluted to the appropriate concentration in ACSF on the day of the experiment. 6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX), D-(1)-2-amino-5-phosphonopentanoic acid (D-AP5), bicuculline and tetrodotoxin (TTX) were obtained from Tocris (Ellisville, MO, USA), and 5-pregnen-3ß-ol-20-one sulphate (PREGS) was obtained from Steraloids, Inc. (Newport, RI, USA). Aliquots of CNQX and PREGS were made in DMSO. Drugs were applied through a bath perfusion system, and slices were maintained in PREGS for a minimum of 15 min after complete bath exchange before recording commenced.

	Results

Top Abstract Introduction Methods Results Discussion Appendix References

 
Simultaneously recorded [Ca²⁺]_pre and fEPSP population spikes both show facilitation

Simultaneously measured F/F₀ and fEPSP population spikes both showed facilitation at interpulse intervals between 50 and 300 ms (Fig. 1Aa and b). At a 50 ms interpulse interval, there was a very consistent 30% increase in amplitude of R2 F/F₀ when compared to that of R1 (c.d. = 0.9860 for linear regression) using the stimulus intensity protocol (Fig. 1B). The simplest explanation for this consistent increase in R2 F/F₀ is that it represents a contribution from residual [Ca²⁺]_pre ([Ca²⁺]_res), which, at a fixed interval of 50 ms, is proportional to the amplitude of R1. A similar comparison of fEPSP population spikes (Fig. 1C) indicates that, while facilitation was always observed at this interpulse interval even at the largest stimulus intensities, there was considerably more variability in the facilitation of the amplitude of the R2 fEPSP population spike than in the simultaneously recorded F/F₀. On closer inspection, it was apparent that the fEPSP population spike data fell into two groups, and we used c.d. = 0.9500 of a linear regression fit to the data from each recording site as a criterion for distinguishing between them. Those from some recording sites were well fitted with a straight line (Fig. 1C, , c.d. = 0.9723) and those from the remaining recording sites were distinctly non-linear (Fig. 1C, •, c.d. = 0.7786). We attempted to fit these latter data using a model that considers both [Ca²⁺]_res and depletion of the readily releasable pool of vesicles. The results of these fits are shown as a continuous black curve in Fig. 1C (c.d. = 0.9508) and will be further considered in the Discussion. Thus, while there is a strong correlation between presynaptic [Ca²⁺]_res and synaptic facilitation, the non-linearity and variability in the facilitation of some of the fEPSP population spikes strongly suggest contributions from mechanisms in addition to [Ca²⁺]_res (Wu & Saggau, 1994; Zucker & Regehr, 2002). We thus proceeded to investigate these additional mechanisms so that we could assess their contributions to the enhancement of facilitation caused by PREGS.

View larger version (36K): [in this window] [in a new window]   Figure 1.  Simultaneously recorded F/F0 and fEPSP population spikes A, F/F0 and fEPSP responses to a single stimulus are shown in black, and successive R2 responses at 50 ms increments in interpulse interval are shown in grey. Note the difference in time scales. B and C, relationship of R2 to R1 at 11 recording sites at stimulus intensities up to that necessary for maximum responses using the stimulus intensity protocol. Data in C were divided into two groups as described in the Results (• and continuous black line, 8 recording sites; and dotted black line, 4 recording sites), and these same symbols are used in B for simultaneously recorded F/F0. Dotted grey lines in B and C indicate unity slopes. B, continuous black line and • respresent least squares regression fit with slope = 1.30, c.d. = 0.9869; dotted black line and represent least squares regression fit with slope = 1.34, c.d. = 0.9827. C, points from individual recording sites are connected with thin grey lines. Open squares were fitted with a least squares linear regression with slope = 1.18, c.d. = 0.9723. See Discussion and Appendix for descriptions of fit to filled circles.

Decay time constants were determined from exponential least squares regression fits to the falling phase of the F/F₀ signal. Using a single exponential to fit the F/F₀ decay following a single stimulus led to the prediction of a 53% increase in the amplitude of R2 F/F₀ at a 50 ms interpulse interval rather than the consistently observed 30% increase (Fig. 1B). However, when we fitted the F/F₀ decay with the sum of two exponentials with a fast time constant (_f) that differed from the slow time constant (_s) by more than 10-fold, the fit of the data, especially over the first 100 ms, was considerably improved, and there was now a calculated 35% increase at 50 ms.

We anticipated that, if [Ca²⁺]_res decay were the result of Ca²⁺ removal from invariant presynaptic compartments, the time constants of [Ca²⁺]_res decay for R1 and R2 would be equal. The best double exponential fit to the F/F₀ decay for a single stimulus yielded a _f = 13.6 ± 1.4 and a _s = 142.7 ± 12.0 ms (Fig. 2C), while a similar fit to R2 F/F₀ decay at a 50 ms interpulse interval yielded a _f = 23.7 ± 3.3 ms and a _s = 365.1 ± 61.6 ms (data not shown). There were significant increases in both time constants (Student's paired t test: _f, P < 0.01; _s, P < 0.005; n = 15), suggesting that there is a change in the compartments underlying [Ca²⁺]_res decay between the R1 and R2.

View larger version (30K): [in this window] [in a new window]   Figure 2.  [Ca2+]res-subtracted R2 F/F0 Aa, superimposed F/F0 responses at a representative recording site to a single stimulus (grey) and to two stimuli at a 50 ms interpulse interval (black). Ab, superimposed F/F0 responses to a single stimulus (grey) and R2 following subtraction of this single stimulus response ([Ca2+]res-subtracted; black). Dotted lines are the time integrals of the respective F/F0 traces. Ac, falling phases of F/F0 responses in Ab fitted with a double exponential for: single stimulus (left, f = 12.1, s = 97.3, c.d. = 0.9380); [Ca2+]res-subtracted R2 (right, f = 21.6, s = 212.7, c.d. = 0.9303) B, [Ca2+]res-subtracted R2 F/F0 normalized to R1 F/F0 using the interpulse interval protocol. One-way ANOVA with Bonferroni post hoc test against control: 50 ms interpulse interval, ***P < 0.005; 100 ms interpulse interval, ****P < 0.001; 150 ms interpulse interval, **P < 0.01; n = 19. C, summed data for single stimulus F/F0 responses (R1, grey) and [Ca2+]res-subtracted R2 F/F0 responses (R2, black). Ca, the falling phase of F/F0 responses were fitted with double exponentials and there was a significant increase of R2 over R1 in both time constants (Student's paired t tests: f, *P < 0.05; s, ***P < 0.0005; n = 15). Cb, there was a significant difference in the initial amplitude of the fit to the F/F0 signal over that for R1 (Student's paired t test: **P < 0.01, n = 15).

In experiments using the interpulse interval protocol, there was not a significant change (one-way ANOVA with repeated measure, P = 0.2693, n = 17, data not shown) in the rising phase (F/F₀)/t, which is proportional to I_Ca (Wu & Saggau, 1994; Sabatini & Regehr, 1998), between the R1 and the subsequent [Ca²⁺]_res-subtracted R2 F/F₀ signals. This implies that there is not a major contribution to PPF from an increase in presynaptic Ca²⁺ influx. We further determined the F/F₀ for the [Ca²⁺]_res-subtracted R2 (Fig. 2A) as an indication of the total change in [Ca²⁺]_pre that accompanies PPF. We found a significant increase in the F/F₀ during the [Ca²⁺]_res-subtracted R2 for interpulse intervals between 50 and 150 ms (Fig. 2B), indicating that there was an increase in [Ca²⁺]_pre during R2 in addition to that resulting simply from [Ca²⁺]_res. One explanation for an increase in [Ca²⁺]_pre without an increase in I_Ca is a contribution from Ca²⁺ buffer occupancy, which could also be a factor in PPF (Blatow et al. 2003; Felmy et al. 2003). We therefore assessed the Ca²⁺ buffer contribution to PPF, and this will be discussed below (Fig. 6).

View larger version (27K): [in this window] [in a new window]   Figure 6.  Effects of [Ca2+]o and 1 µM PREGS on the PPR Aa, summary of all data shown in B (n = 18). ACSF PPR: grey solid circles and continuous line. Pregnenolone sulphate PPR: and solid line. One micromolar PREGS PPR at all values of [Ca2+]o multiplied by the ratio at 2.5 mM [Ca2+]o of ACSF PPR/PREGS PPR: and dashed line. Ab, PREGS PPR normalized to ACSF PPR at the same [Ca2+]o. Student's paired t tests: 2.5 mM [Ca2+]o and 1.5 mM [Ca2+]o, P = 0.556, n = 10 (Ba); 2.5 mM [Ca2+]o and 2.0 mM [Ca2+]o, P = 0.140, n = 4 (Bb); 2.5 mM [Ca2+]o and 3.0 mM [Ca2+]o, P = 0.602, n = 4 (Bc). B, PPR enhancement by PREGS at different values of [Ca2+]o with 2.5 mM [Ca2+]o as a paired control (Ba, n = 10; Bb, n = 4; Bc, n = 4). ACSF, grey bars; PREGS, black bars. One-way ANOVA with repeated measures followed by Tukey's post hoc test for 95% confidence: *P < 0.05, **P < 0.01, and ***P < 0.001. Student's paired t tests: P < 0.05 and P < 0.001. There was a large reduction in the amplitude of the R1 fEPSP population spike in 1.5 mM [Ca2+]o (–73.9 ± 6.4%, n = 10, Ba), which resulted not only in an increase in the PPR, but also in an increase in the underlying variability of the PPR.

Paired-pulse facilitation is accompanied by changes in the synaptic input–output relationship

In addition to the observed changes in [Ca²⁺]_pre that accompany PPF, it was of interest to assess synaptic input–output changes that might also contribute to facilitation. Importantly, our measurements of (F/F₀)/t and fibre volley amplitude showed no change in presynaptic excitability between R1 and R2 (see Methods). We thus used the stimulus intensity protocol to generate synaptic input–output curves using presynaptic F/F₀ as the input and the resultant fEPSP population spike amplitude as the output (Fig. 3). These input–output curves were accurately fitted with a Hill function for both R1 and R2. Comparison of these fits indicated that there was a significant increase in y_max and decrease in EC₅₀, but no change in n_H for [Ca²⁺]_res-subtracted R2 F/F₀ when compared with R1 (Fig. 3B and C). This suggests that during R2, in addition to the presence of [Ca²⁺]_res (Fig. 1) and independent of the larger F/F₀ (Fig. 2), there are factors that lead to a larger postsynaptic response for a given presynaptic F/F₀.

View larger version (20K): [in this window] [in a new window]   Figure 3.  Hill function fits to the input–output relationship A, R1 (grey) and [Ca2+]res-subtracted R2 (black) for one representative experiment using the stimulus intensity protocol. B and C, mean and individual values for ymax and EC50 for least squares regression fits to the Hill function. Student's paired t test: ymax, ****P < 0.0005, n = 14 (B) and EC50, *P < 0.05, n = 14 (C). There was not a significant difference in the Hill coefficient, nH (Student's paired t test: P = 0.54).

We have shown previously that 1 µM PREGS acts presynaptically to enhance PPF by selectively increasing the second of a pair of postsynaptic responses. We therefore asked which presynaptic site was the primary locus of action for this low concentration of PREGS. One site at which PREGS might enhance PPF could be through an increase in [Ca²⁺]_pre. Qualitatively, this did not appear to be the case because the PREGS fEPSP population spike facilitation was consistently enhanced over ACSF while F/F₀ facilitation was not altered (Fig. 4A). However, to assess the possible contribution of PREGS enhancement of [Ca²⁺]_res to fEPSP PPF quantitatively, we measured [Ca²⁺]_res and simultaneously determined PPF of fEPSP population spikes using the interpulse interval protocol. We then compared the relationship between [Ca²⁺]_res and PPF in PREGS to previously measured PPF at the same recording site in ACSF. Pregnenolone sulphate caused a consistent upward shift of the amount of fEPSP population spike PPF at each [Ca²⁺]_res (Fig. 4B). Since there was no rightward shift of this relationship to larger values of [Ca²⁺]_res, it is clear that PREGS does not enhance fEPSP PPF by simply increasing [Ca²⁺]_res. Furthermore, there was no change in the linear relationship between R1 F/F₀ and R2 F/F₀ as a consequence of adding PREGS (Fig. 4C), although this manipulation caused the expected enhancement of simultaneously recorded R2 fEPSP population spikes (Fig. 4D). These results indicate that the presynaptic enhancement of facilitation caused by PREGS is downstream from [Ca²⁺]_pre, and we therefore sought to support this observation further.

View larger version (29K): [in this window] [in a new window]   Figure 4.  Effect of PREGS on fEPSPpopulation spike PPF as a function of [Ca2+]res Simultaneously recorded F/F0 (top) and fEPSP (bottom) in ACSF (Aa) and after addition of 1 µM PREGS (Ab). In the representative experiment and in 5 additional experiments, the PREGS fEPSP population spike facilitation was consistently enhanced by 78.5 ± 26.6% (n = 6, P < 0.05) over ACSF, while F/F0 facilitation was not significantly different (P = 0.193, n = 6). Field potentials were positioned so that stimulus artifacts were aligned with the beginning of the F/F0 signal. B, for each recording site (n = 6), fEPSP PPF was normalized to the maximum value in ACSF and plotted against simultaneously recorded [Ca2+]res at that interpulse interval. Recordings were made using the interpulse interval protocol. ACSF data (grey) and 1 µM PREGS data (black) are fitted with least squares linear regressions. Note that if PREGS enhanced PPF solely by enhancing [Ca2+]res, then PREGS-enhanced PPF (black) would be represented by points that fell further to the right along an extension of the same continuous grey line that was obtained for the ACSF data. C, relationship of R2 F/F0 to R1 F/F0 at 6 recording sites (ACSF: • and grey line, least squares regression slope = 1.31, c.d. = 0.9924; PREGS at the same recording sites: and black line, least squares regression slope = 1.31, c.d. = 0.9886). D, fEPSP data from 5 of the 6 recording sites shown in C. See Discussion and Appendix for an explanation of the grey (c.d. = 0.9594) and black (c.d. = 0.9246) curves. Points from individual recording sites are connected with thin grey lines. Data from the 6th recording site yielded a linear relationship with slope = 1.20, c.d. = 0.9957 for ACSF and slope = 1.36, c.d. = 0.9564 for PREGS and was omitted for clarity. Dotted grey lines in C and D indicate unity slopes.

View larger version (31K): [in this window] [in a new window]   Figure 5.  Effect of PREGS on [Ca2+]res-subtracted R2 F/F0 A, increase of [Ca2+]res-subtracted R2 F/F0 normalized to R1 F/F0 using the interpulse interval protocol (black, 1 µM PREGS and grey, ACSF). One-way ANOVA with Bonferroni post hoc test against control: ACSF, 50 ms interpulse interval, P < 0.05; ACSF, 100 ms interpulse interval, P < 0.001; ACSF, 150 ms interpulse interval, P < 0.05; no significant difference at any interpulse interval in PREGS (n = 11). B, there was not a significant difference in the F/F0 for R1 between PREGS and ACSF (Student's paired t test: P = 0.924, n = 11). C, there was a significant difference in f and s from double exponential fits between R1 and R2 in both ACSF and PREGS at a 50 ms interpulse interval (Student's paired t test: f, ACSF, *P < 0.05, PREGS, *P < 0.05; s, ACSF, ***P < 0.005, PREGS, *P < 0.05; n = 11) D, there was a significant increase in the initial amplitude of the fit to the F/F0 signal between R1 and R2 at 50 ms interpulse interval in ACSF, but not in PREGS (Student's paired t tests: ACSF, *P < 0.05; PREGS, P = 0.36, n = 11).

Pregnenolone sulphate has no significant effect on presynaptic Ca²⁺ buffering

One interpretation of the data in Fig. 5 is that PREGS alters the contribution of buffer occupancy to facilitation. In order to independently assess the roles of cytoplasmic Ca²⁺ buffers and [Ca²⁺]_res as the site of action of PREGS, we carried out experiments in which we varied [Ca²⁺]_o while holding constant the divalent ion concentration (see Methods). In the Schaffer collateral presynaptic terminals in CA1, changes of [Ca²⁺]_o produce significant changes in the PPR (Fig. 6). With a buffer-independent mechanism for facilitation downstream from [Ca²⁺]_res, the PPR would be expected to decrease as the [Ca²⁺]_o increases (Blatow et al. 2003). We observed such a [Ca²⁺]_o-dependent change in the PPR both in ACSF (ACSF PPR) and after addition of 1 µM PREGS (PREGS PPR; Fig. 6). Pregnenolone sulphate was found to enhance the PPR at all values of [Ca²⁺]_o-tested (Fig. 6B). Furthermore, the percentage enhancement of the PPR caused by PREGS was similar at each [Ca²⁺]_o (Fig. 6Aa, ) indicating that PREGS does not alter the inverse relationship between PPR and [Ca²⁺]_o. This is better demonstrated by equally normalizing the PREGS PPR at each [Ca²⁺]_o, which produced a plot that accurately overlaps the ACSF PPR at all values of [Ca²⁺]_o (Fig. 6Aa, and grey closed circles). Likewise, when the PREGS PPR was normalized to the control PPR at the same [Ca²⁺]_o, the resulting ratios were similar among different values of [Ca²⁺]_o (Fig. 6Ab). Thus PREGS does not alter the inverse relationship between PPR and [Ca²⁺]_o as would be expected if its action were predominately through a change in the contribution of presynaptic Ca²⁺ buffering.

Pregnenolone sulphate leads to significant changes in the R2 input–output relationship

Since 1 µM PREGS does not affect basal glutamate release following a single stimulus (Schiess & Partridge, 2005), those components of short-term facilitation that depend solely on the amplitude of the first response should be relatively unchanged between ACSF and PREGS. Any remaining differences in R2 between ACSF and PREGS should reflect the process of enhancement caused by PREGS. We thus compared input–output relationships for fEPSP population spikes versus [Ca²⁺]_res-subtracted R2 F/F₀ in ACSF and in PREGS in order to selectively assess the effects of PREGS on the enhancement of PPF. Hill function fits to the R2 input–output relationships in PREGS showed a significant increase in y_max and decrease in EC₅₀ over ACSF (Fig. 7B and C) without a significant change in n_H, and these differences were not seen in vehicle control experiments. The increase in y_max and decrease in EC₅₀ caused by PREGS predict an increased R2 fEPSP population spike for a given [Ca²⁺]_pre (Fig. 7) and suggest that neurosteroid enhancement occurs at a facilitatory site where [Ca²⁺]_pre exerts an effect.

View larger version (21K): [in this window] [in a new window]   Figure 7.  Hill function fit for [Ca2+]res-subtracted R2 input–output data using the stimulus strength protocol A, typical R2 input–output curves for ACSF (•, grey line) and 1 µM PREGS (, black line). B, ymax from least squares fit to Hill function (Student's paired t test: P < 0.05, n = 6). C, EC50 from least squares fit to Hill function (Student's paired t test: P < 0.05, n = 6). There was not a significant difference in the Hill coefficient, nH (Student's paired t test: P = 0.32, n = 6).

	Discussion

Top Abstract Introduction Methods Results Discussion Appendix References

Paired-pulse facilitation is accompanied by an increase in [Ca²⁺]_pre

Other studies have found either no change (Kamiya & Ozawa, 1998) or a decrease (Wu & Saggau, 1994) in the amplitude of the [Ca²⁺]_res-subtracted R2 F/F₀ at a 50 ms interpulse interval. This could be because of differences in the affinity of the indicators used, or owing to differences between animal models. When we integrated the [Ca²⁺]_res-subtracted R2 F/F₀, though, we found a significant increase over R1 F/F₀ that resulted largely from a slowing of the decay time course. The double exponential fit to the F/F₀ decay suggests a variation from a single compartment model (Neher & Augustine, 1992; Koester & Sakmann, 2000), and the significant increase of _f and _s for F/F₀ decay between a single stimulus and R2 without [Ca²⁺]_res subtraction suggests that some underlying component of calcium decay is changed during PPF.

The increase in the integral of the [Ca²⁺]_res-subtracted R2 F/F₀ (Fig. 2B) reflects an effect that largely outlasts the time during which facilitated transmitter release occurs in R2; however, it is likely to reflect presynaptic processes that are involved in PPF. We found no change in (F/F₀)/t during PPF, while a decrease in this parameter has been reported in fura-2 AM measurements at these same synapses (Wu & Saggau, 1994). In either case, it appears that the increased [Ca²⁺]_pre during R2 does not result from an increased Ca²⁺ influx at this synapse, although facilitation of presynaptic I_Ca can underlie PPF at other synapses (Borst & Sakmann, 1998; Cuttle et al. 1998). Furthermore, our data (Fig. 6) are consistent with previous reports that buffer occupancy is not a major contributor to PPF in Schaffer collateral terminals (Blatow et al. 2003). One additional possibility is that the increased [Ca²⁺]_pre during R2 results from a component of Ca²⁺ release from internal stores (Emptage et al. 2001; Cabezas & Buno, 2006), although there is other evidence that release from stores does not contribute to PPF (Carter et al. 2002). Another intriguing possibility is that Ca²⁺ extrusion processes are saturated or depressed during R2.

Presynaptic Ca²⁺ buffering is not a major factor in short-term facilitation

Since the presynaptic Ca²⁺ influx is a major factor in the probability of vesicle release, there should be an inverse relationship between [Ca²⁺]_o and depletion of the readily releasable pool for a given stimulus. Furthermore, since facilitation depends on both the number of available vesicles during R2 and some facilitatory process (Bark et al. 2004), a similar inverse relationship might be expected to exist between [Ca²⁺]_o and facilitation. However, if there is significant buffering of [Ca²⁺]_pre, the initial depletion of the readily releasable pool will be diminished and the time course of [Ca²⁺]_res will be prolonged. Likewise, the closer [Ca²⁺]_pre is to saturating the buffer during R1, the less [Ca²⁺]_pre will be buffered during R2, and the greater the probability of release will be during R2. Thus at increased [Ca²⁺]_o, presynaptic Ca²⁺ buffering will increase the amount of PPF. As shown in Fig. 6, we observed the inverse relationship between [Ca²⁺]_o and PPR that demonstrates a buffer-independent facilitation mechanism (Blatow et al. 2003). Likewise, while a buffer saturation mechanism would lead to a more rapid extrusion of facilitated [Ca²⁺]_res such that the time constants of F/F₀ decay would be shorter for R2 than for R1, we observed decay time constants that were, in fact, longer for R2 than for R1 (Fig. 2C). Importantly, these relationships were not altered in the presence of PREGS, thereby minimizing the possibility that a consequential component of the action of PREGS is to alter [Ca²⁺]_pre buffering.

Paired-pulse facilitation and its enhancement by PREGS are accompanied by consistent changes in the synaptic input–output relationship

The most unambiguous interpretation of synaptic input–output data results from measurements of failures versus successes for minimal stimulation (Dobrunz & Stevens, 1997) or from direct measurements in specialized large synapses (Felmy et al. 2003). In our experiments, increasing stimulus strength leads to an increase in the number of presynaptic fibres recruited, which would also be reflected in the F/F₀ signal. We found no evidence for a change in excitability of the presynaptic fibres between R1 and R2 in ACSF or during PPF enhancement by PREGS. It is thus reasonable to make comparisons among these different conditions at a particular stimulus strength and stimulating electrode placement. In addition, comparisons of R2 fEPSP population spike amplitude versus F/F₀ before and after PREGS should eliminate effects of differences in the numbers of fibres recruited and should thereby provide a reasonable assessment of the effect of PREGS in enhancing facilitation independent of the underlying presynaptic mechanisms of PPF. In general, however, input–output relationships generated from population measurements as were used here need to be interpreted with caution. While the observed changes in EC₅₀ and y_max reflect, in a general way, changes in the effectiveness of [Ca²⁺]_pre in the process of neurotransmitter release, it would be inappropriate to draw conclusions from the Hill function fits to the data about receptor affinity or co-operativity at a single site.

Errors in the input–output relationships could be introduced by inaccuracies in the measurement of F/F₀. In particular, bleaching of the intracellular Magnesium Green AM over the course of multiple measurements could produce an underestimation of the [Ca²⁺]_pre. This is unlikely to have affected the apparent shift in EC₅₀ between R1 and R2 (Fig. 3A), since F/F₀ measurements from R1 and R2 were by necessity alternated and no consistent change was observed in R1 F/F₀ (e.g. see Fig. 1Aa). A more significant opportunity for bleaching was present during experiments that compared R2 in ACSF and in PREGS (Fig. 5A), since all of the PREGS measurements were made after completion of the ACSF measurements. However, we found no significant difference in either the amplitude of the R1 F/F₀ between ACSF and PREGS (Fig. 5D) or in the F/F₀ between these two conditions (Fig. 5B).

Paired-pulse facilitation and its enhancement by PREGS are well characterized by a model for transmitter release

Interestingly, the percentage facilitation of R2 F/F₀ with the stimulus intensity protocol is quite linear (Fig. 1B), while there is considerable variability in the facilitation of simultaneously recorded fEPSP population spikes (Fig. 1C). Presynaptic [Ca²⁺]_res is one factor in determining the amount of PPF; however, downstream events must contribute significant amounts of variability to the postsynaptic response. One very apparent site for such variability is the readily releasable pool of vesicles, which, if sufficiently depleted following R1, could lead to PPD even after significant accumulation of [Ca²⁺]_res. Thus, interactions between PPF and PPD frequently dictate an inverse relationship between PPF and R1 (Dobrunz & Stevens, 1997); however, this relationship is not obligatory and depends, for instance, on the actions of [Ca²⁺]_res (Cabezas & Buno, 2006). Other potential sources of variability are the access of Ca²⁺ to a facilitatory site and the effectiveness of this site in facilitating subsequent release.

In an effort to understand the relationship between R1 fEPSP and R2 fEPSP (Fig. 1C) better, we have attempted to fit these data based on our previously described model (Bark et al. 2004) in which [Ca²⁺]_res changes the probability of release of a subpopulation of the readily releasable pool of vesicles, possibly by acting on a second Ca²⁺-mediated facilitatory site. (Other models with similar assumptions have also been proposed for PPF (e.g. Jiang & Abrams, 1998; Dittman et al. 2000; Rozov et al. 2001), and our model is discussed in more detail in the Appendix.) Interestingly, we consistently found that although all recording sites showed PPF, at most recording sites this model provided a good fit to the data (Fig. 1C, • and continuous black line), while at a few recording sites, the data could be better fitted with a straight line with a slope > 1 (Fig. 1C, and dotted black line). Since the relationship between R1 F/F₀ and R2 F/F₀ was indistinguishable for these two groups (Fig. 1B), we hypothesize that in some circumstances [Ca²⁺]_res leads to facilitation by directly interacting at the release site (Rozov et al. 2001), while in other circumstances [Ca²⁺]_res has a significant effect at a second facilitatory site, which increases the release probability of a fraction of the vesicles (Atluri & Regehr, 1996). Although the recording sites shown in Fig. 4 had less PPF than those in Fig. 1C, they showed facilitation that was consistently enhanced following the addition of PREGS. Again, the relationship between R1 F/F₀ and R2 F/F₀ was linear with a slope of 1.3 both in PREGS and ACSF (Fig. 4C). At five of these six recording sites, the relationship between R1 fEPSP and R2 fEPSP in both ACSF and PREGS (Fig. 4D) was accurately fitted with our model. A comparison of the parameters of the fit of these data by the model predicts that the most significant factor in the enhancement of PPF by PREGS is an increase in the fraction of vesicles influenced by [Ca²⁺]_res.

Pregnenolone sulphate enhances facilitation at a site downstream from [Ca²⁺]_pre

Pregnenolone sulphate enhances facilitation of the postsynaptic fEPSP response without increasing [Ca²⁺]_pre (Figs 4B and 5). Thus PREGS must have a significant action at a site downstream from [Ca²⁺]_pre. We have presented here four lines of experimental evidence that support a PREGS-induced increase in presynaptic neurotransmitter release for a given [Ca²⁺]_pre. First, there is no change as a result of PREGS in the 30% increase in R2 F/F₀ at each R1 F/F₀, although simultaneously recorded fEPSP PPF is enhanced (Fig. 4C and D). Second, in the presence of PREGS, there is a smaller increase in R2 F/F₀ than in ACSF (Fig. 5) as a result of a reduced increment of initial F/F₀ amplitude and a smaller increase in _f and _s of F/F₀ decay. It has been reported previously that PREGS inhibits voltage-gated calcium channels (ffrench-Mullen et al. 1994) in cultured CA1 neurons, while in Schaffer collateral terminals we saw no change of (F/F₀)/t. Thus the fEPSP enhancement by PREGS occurs without an increase of [Ca²⁺]_pre. Third, if PREGS altered facilitation by a significantly different proportion at different values of [Ca²⁺]_o and thereby produced a noticeable shift in the [Ca²⁺]_o dependence of the PPR (Fig. 6A), it could indicate that PREGS has an effect on presynaptic Ca²⁺ buffering (Blatow et al. 2003). Since we found no change in this relationship, we believe that the mechanism by which PREGS enhances short-term facilitation is independent of any effect that PREGS may have on Ca²⁺ buffering. Fourth, there is a leftward shift, in the presence of PREGS, of the input–output relationship of the R2 fEPSP population spike amplitude to [Ca²⁺]_res-subtracted F/F₀ (Fig. 7). Since PREGS has a minimal effect on the number of presynaptic fibres recruited at a given stimulus intensity, and since this input–output curve compares ACSF and PREGS R2, both of which have undergone short-term facilitation, this leftward shift indicates an increased amount of transmitter release for a given [Ca²⁺]_pre in the presence of PREGS. These data are thus consistent with a dominant role of the action of PREGS downstream from [Ca²⁺]_pre such as at a synergistic high-affinity Ca²⁺ binding site associated with the vesicle release protein complex (Atluri & Regehr, 1996).

Neurosteroids are important modulators of synaptic efficacy, and the presynaptic action of PREGS to enhance short-term facilitation may underlie crucial aspects of hippocampal function (e.g. Partridge & Valenzuela, 2002; Thomas et al. 2005). The evidence that we present here for a site of action of PREGS downstream from [Ca²⁺]_pre in Schaffer Collateral terminals in CA1 raises interesting possibilities for the involvement of specific proteins both in paired pulse facilitation and in the enhancement of this facilitation by neurosteroids.

	Appendix

Top Abstract Introduction Methods Results Discussion Appendix References

 
In an effort to understand the non-linear relationship between R1 fEPSP and R2 fEPSP at some recording sites (Figs 1C and 4D) better, we have attempted to fit these data based on our previously described model (Bark et al. 2004) in which [Ca²⁺]_res changes the probability of release of a subpopulation of the readily releasable pool of vesicles. According to this model:

(2)

(3)

where R1/N₁ and R2/N₁ are directly proportional to first and second fEPSP population spikes, respectively, and N₁ is proportional to the size of the readily releasable pool; P₁ is the basal release probability and P₂ is the larger, facilitated release probability; and a is the fraction of vesicles influenced by [Ca²⁺]_res. We assumed that a was sigmoidally related to P₁ by a Hill function:

(4)

where a_max is related to the maximum effectiveness of [Ca²⁺]_res in facilitating release; K(a) is related to the sensitivity of R1 in determining a; and n_H1, the Hill coefficient, determines the steepness of this relationship.

Table A1 gives the values of these parameters from least squares fits of the model equations to the data in Figs 1C and 4D.

	References

Top Abstract Introduction Methods Results Discussion Appendix References

 
Atluri PP & Regehr WG (1996). Determinants of the time course of facilitation at the granule cell to Purkinje cell synapse. J Neurosci 16, 5661–5671.[Abstract/Free Full Text]

Bark C, Bellinger FP, Kaushal A, Mathews JR, Partridge LD & Wilson MC (2004). Developmentally regulated switch in alternatively spliced SNAP-25 isoforms alters facilitation of synaptic transmission. J Neurosci 24, 8796–8805.[Abstract/Free Full Text]

Barrot M, Vallee M, Gingras MA, Le Moal M, Mayo W & Piazza PV (1999). The neurosteroid pregnenolone sulphate increases dopamine release and the dopaminergic response to morphine in the rat nucleus accumbens. Eur J Neurosci 11, 3757–3760.[CrossRef][Medline]

Baulieu EE (1998). Neurosteroids: a novel function of the brain. Psychoneuroendocrinology 23, 963–987.