|Home | About | Journals | Submit | Contact Us | Français|
While recent evidence suggests that the hippocampus is a source of 17β-estradiol (E2), the physiological role of this neurosteroid E2, as distinct from ovarian E2, is unknown. One likely function of neurosteroid E2 is to acutely potentiate excitatory synaptic transmission, but the mechanism of this effect is not well understood. Using whole-cell voltage-clamp recording of synaptically evoked EPSCs in adult rat hippocampal slices, we show that, in contrast to the conclusions of earlier studies, E2 potentiates excitatory transmission through a presynaptic mechanism. We find that E2 acutely potentiates EPSCs by increasing the probability of glutamate release specifically at inputs with low initial release probability. This effect is mediated by estrogen receptor (ER)β acting as a monomer, whereas ERα is not required. We further show that the E2-induced increase in glutamate release is due primarily to increased individual vesicle release probability and is associated with higher average cleft glutamate concentration. These two findings together argue strongly that E2 promotes multivesicular release, which has not been shown before in the adult hippocampus. The rapid time course of acute EPSC potentiation and its concentration dependence suggest that locally-synthesized neurosteroid E2 may activate this effect in vivo.
Recent studies support the idea that the hippocampus, a known target of the steroid hormone 17β-estradiol (E2), is also a site of E2 synthesis. The possibility that the brain produces its own steroids, neurosteroids, was first proposed almost 30 years ago (Corpechot et al., 1981). More recently, all enzymes needed to synthesize E2 from cholesterol, as well as significant levels of intermediate metabolites and E2 itself, have been demonstrated in the adult rat hippocampus (Kimoto et al., 2001; Hojo et al., 2004; Hojo et al., 2009). This raises the question: what is the physiological role of neurosteroid E2, as distinct from ovarian E2?
One likely answer lies in the ability of E2 to acutely potentiate excitatory synaptic transmission. Within minutes of application to hippocampal slices, E2 increases fEPSP slope (Teyler et al., 1980; Sharrow et al., 2002; Kim et al., 2006; Kramar et al., 2009), and potentiates intracellularly recorded EPSPs and EPSCs in CA1 (Wong and Moss, 1992; Foy et al., 1999; Rudick and Woolley, 2003). E2 also acutely increases CA1 neurons responses to glutamate receptor agonists applied to slices (Wong and Moss, 1992) or dissociated cells (Gu and Moss, 1996; 1998). Several observations suggest that brain-derived E2, not ovarian E2, is the endogenous steroid that activates these effects in vivo. First, although acute potentiation can occur with E2 concentrations matching peak circulating levels (~100 pM), it is more robust with higher concentrations, and potentiation of agonist-evoked responses requires 10 nM E2. Second, acute potentiation occurs in slices from male as well as female hippocampus (Teyler et al., 1980; Kramar et al., 2009). Finally, the ability of E2 to act within minutes is much more rapid than any fluctuations in circulating E2.
The mechanism(s) of acute potentiation of synaptic transmission by E2 are not well understood. Intracellular recording studies have shown that only some CA1 cells are E2-responsive, suggesting that E2 action is cell-specific. Further, the studies using exogenous agonist application led to the conclusion that E2 enhances postsynaptic sensitivity to glutamate. It is not clear, however, whether conclusions based on non-synaptic stimulation apply to E2 effects on synaptic transmission. Therefore, we studied acute potentiation of synaptic responses by E2 using whole-cell voltage-clamp recording of synaptically evoked EPSCs in adult rat hippocampal slices. Consistent with previous reports, we observed a response to E2 in only a subset of experiments. Interestingly, however, the E2-responsive subset shared the characteristic of relatively high initial paired-pulse ratio (PPR) and E2 decreased PPR in parallel with synaptic potentiation, suggestive of presynaptic mechanisms. Indeed, we found that E2 acutely potentiates excitatory synaptic transmission in an input-specific manner, through an increase in the probability of glutamate release specifically at inputs with low initial release probability. This potentiation depends on activation of estrogen receptor (ER)β and not ERα. Further investigation of mechanisms involved in the E2-induced increase in glutamate release revealed that E2 increases individual vesicle release probability as well as average cleft glutamate concentration, strongly suggesting that E2 promotes multivesicular release.
All animal procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Northwestern University Animal Care and Use Committee. Adult female Sprague Dawley rats (50–60 days old, Harlan) were ovariectomized under ketamine (85 mg/kg, i.p, Bioniche Pharma) and xylazine (13 mg/kg, i.p., Lloyd Laboratories) anesthesia using aseptic surgical procedures. Either 3 or 7 days after surgery, each rat was given 2 injections (s.c.) of 10 μg of 17β-estradiol benzoate in 100 μl of sesame oil 24 h apart and slices were prepared 2 days after the second injection. This estradiol pre-treatment produces serum levels of ~30–40 pg/ml (corresponding to peak proestrus levels, Smith et al., 1975) at the time of sacrifice (Woolley and McEwen, 1993), and has been shown to increase CA1 responsiveness to acute effects of estradiol on excitatory synaptic transmission (Wong and Moss, 1992).
Stock solutions of the following drugs were prepared in DMSO: (+)-bicuculline, cyclothiazide (CTZ), propylpyrazole triol (PPT), diarylpropionitrile (DPN), ICI 182,870, NBQX, G-1 (all from Tocris Bioscience), and 17β-and 17α-estradiol (from Sigma-Aldrich). Stock solutions of γ-D-glutamylglycine (γDGG), and (RS)-CPP (both from Tocris Bioscience) were prepared in ddH2O. Stock solutions were diluted in artificial CSF (aCSF) on the day of recording to the final concentrations indicated. Control aCSF contained equivalent concentration of DMSO (<0.1%). Other chemicals were either from Fisher Scientific (NaCl, NaHCO3, dextrose, KCl, NaH2PO4), or from Sigma-Aldrich (sesame oil, CaCl2, MgCl2, K-gluconate, HEPES, Na2-creatine phosphate, MgATP, NaGTP, biocytin, QX314).
Rats were deeply anesthetized with sodium pentobarbital (80 mg/kg, i.p., Virbac Animal Health) and perfused transcardially with ice-cold oxygenated aCSF containing, in mM: 125 NaCl, 25 NaHCO3, 25 dextrose, 2.5 KCl, 1.25 NaH2PO4, 1 MgCl2, 2 CaCl2 (osmolarity 315, pH 7.4). Transverse slices of dorsal hippocampus (300 μm) were cut into an ice-cold bath of oxygenated aCSF using a Leica VT1000S oscillating tissue slicer. Slices were allowed to recover submerged in oxygenated aCSF at 35°C for 30 min, then were kept at room temperature until recording. During recording, slices were perfused with warm (32–35°C) or room temperature (20–22°C, for high-frequency train experiments and γDGG/NBQX experiments) oxygenated aCSF at a rate of ~2 mL/min in a recording chamber mounted on a Zeiss Axioskop. Somatic whole-cell voltage-clamp recordings were obtained from visually identified CA1 pyramidal cells using patch electrodes (3–5 MΩ) filled with intracellular solution containing, in mM: 115 K-gluconate, 20 KCl, 10 HEPES, 10 Na2-creatine phosphate, 2 MgATP, 0.3 NaGTP, 0.1% biocytin, 1–5 QX314 (osmolarity 280, pH 7.3). One or two glass bipolar stimulating electrodes filled with aCSF were placed in the stratum radiatum, ~300 μm apart. At the beginning of every double-stimulation experiment, it was confirmed that no paired-pulse facilitation (IPI = 100 ms) occurred when the first pulse was delivered at one stimulation site and the second pulse was delivered at the other stimulation site, indicating no overlap of stimulated inputs. During an experiment, paired-pulse stimulation (IPI = 100 ms or 50 ms) was delivered every 15 s; in double-stimulation experiments, stimulation was delivered every 15 s, alternating between the two stimulation sites; in high-frequency train experiments, 2 min of paired-pulse stimulation (delivered every 15 s) alternated with 3 min of train stimulation (100-pulse 20-Hz trains delivered every 45 s). In experiments using both 50 ms and 100 ms IPI, there were no effects of IPI on either baseline paired-pulse ratio (PPR) or the E2 effect on PPR. Series resistance was monitored using 5 mV 10 ms voltage steps, and ranged between 5 and 25 MΩ between experiments. Only experiments with stable series resistance and holding current were included in the analysis, and series resistance did not differ between E2-responsive and E2-nonresponsive experiments. Synaptically evoked AMPA receptor-mediated EPSCs were recorded at a holding potential of −70 mV in the presence of the GABAA and NMDA receptor blockers, (+)-bicuculline (10 μM) and (RS)-CPP (10 μM), respectively. Data were acquired with Axopatch 200B amplifier and pClamp 9.0 software (Molecular Devices), filtered at 2 kHz, and digitized at 20 kHz using a Digidata 1322A data acquisition system (Molecular Devices).
After a stable 10–15-min baseline recording in control aCSF, different drugs were bath applied as follows: E2 (100 pM – 100 nM) for 10–15 min, in most cases followed by a 10–30-min wash-out period; CTZ (100 μM), PPT (100 nM or 200 nM), DPN (100 nM or 500 nM), ICI 182,780 (100 nM), or G-1 (100 nM) alone for 12 min, followed by co-application with E2 (100 nM) for 12 min, followed by E2 (100 nM) alone (see Fig. 4); γDGG (1 mM) for 10 min either in the absence or in the presence of E2 (100 nM, see Fig. 6); NBQX (200 nM) for 15 min either in the absence or in the presence of E2 (100 nM, see Fig. 6). Effects of different drugs were evaluated by comparing data from the last 5 min of recording in each condition, or from a 5-min period after the maximal effect of a drug was reached (in γDGG and NBQX experiments). Because effects of E2 were not readily reversible, E2 and ER agonists were applied to each slice only once.
In experiments using high-frequency stimulus trains, a 15–20-min baseline recording was followed by a 20-min recording in the presence of E2 (100 nM). Effects of E2 were evaluated by comparing data from the last 10 min of E2 application to data from the last 10 min of baseline, in order to include a sufficient number of train sweeps. The analysis of trains was based on Schneggenburger et al. (1999), and Wesseling and Lo (2002). Baseline EPSC trains showed early facilitation, followed by gradual depression, and reached a steady state after ~70 pulses, consistent with previous reports (Wesseling and Lo, 2002). Cumulative EPSC charge during the last second of the train (sum of pulses 81–100) was used to estimate the rate of steady-state release due to vesicle replenishment, assumed to be constant throughout the train. The relative size of the readily-releasable vesicle pool (RRP) was estimated from cumulative EPSC charge recorded before the steady-state was reached (sum of pulses 1–80) after subtracting EPSC charge due to vesicle replenishment during the train (4 × sum of pulses 81–100). The proportion of the total RRP released by the first stimulus was used to estimate relative individual vesicle release probability, Pves.
Data were analyzed off-line using Igor Pro (Wavemetrics) software, and statistical significance was evaluated with Student’s t-tests or ANOVA followed by Bonferroni post-hoc tests, as indicated. A p-value of <0.05 was considered significant.
We investigated the ability of 17β-estradiol (E2) to rapidly potentiate excitatory synaptic transmission in adult female rat hippocampus. We stimulated the Schaffer collateral pathway in acute slices and made whole-cell voltage-clamp recordings of synaptically evoked EPSCs from CA1 pyramidal cells. In a subset of experiments, bath application of E2 (100 pM – 100 nM) potentiated EPSC amplitude within minutes (Fig. 1A). This effect was not readily reversible, possibly due to slow wash out of E2 from the slice, or initiation of a persistent effect that outlasts the presence of E2. We quantified EPSC potentiation by normalizing average EPSC amplitude recorded during the last 5 min in E2 to baseline EPSCs during the last 5 min before E2 application. This showed that the magnitude of E2-induced potentiation ranged from 0.76 to 1.76, with a clearly bimodal distribution among experiments (Fig. 1B). Based on this distribution, we subsequently divided all experiments into two groups: those showing >20% potentiation were classified as E2-responsive and the rest as E2-nonresponsive.
Using the >20% potentiation criterion, we found that both the frequency of responders and the extent of potentiation in responders increased with increasing concentrations of E2. At 100 pM, E2 potentiated EPSCs in 6/14 (43%) experiments, by 26 ± 2%; 1 nM E2 potentiated EPSCs in 9/22 (41%) experiments, by 29 ± 3%; 10 nM E2 potentiated EPSCs in 8/17 (47%) experiments, by 33 ± 4%; 100 nM E2 potentiated EPSCs in 18/28 (64%) experiments, by 42 ± 3% (Fig. 1C). The effect of E2 was stereospecific, as no potentiation was observed with 100 nM 17α-estradiol (n = 5).
Previous studies indicated that E2 acutely increases postsynaptic responsiveness to AMPA receptor agonists. The range in magnitude of EPSC potentiation we observed, however, raised the possibility that variation in E2-responsiveness might be related to differences in glutamate release properties, which are known to vary widely among CA1 synapses (Turner et al., 1997; Dobrunz and Stevens, 1997). Thus, we investigated potential presynaptic effect(s) of E2 by analyzing paired-pulse ratio (PPR), a measure related to neurotransmitter release probability and commonly used to assess changes in presynaptic function.
We recorded EPSCs evoked by paired-pulse stimulation (IPI = 100 ms) of the Schaffer collateral pathway and examined the effect of E2 on PPR in E2-responders and nonresponders. This revealed, first, that responders and nonresponders differed in baseline PPR even before E2 was applied (Fig. 2A, B). Baseline PPR in responders was 1.48 ± 0.03, significantly higher than 1.37 ± 0.02 baseline PPR in nonresponders, (t-test: p<0.01). In addition, E2 decreased PPR within minutes of application specifically in responders (Fig. 2A; one-way ANOVA: p<0.01), while it had no effect on PPR in nonresponders (Fig. 2B; one-way ANOVA: p>0.7). Since PPR is inversely related to release probability, these results suggested that E2 acts preferentially on synapses with relatively low initial probability of release to potentiate EPSC amplitude, at least in part, by increasing glutamate release probability.
Although AMPA receptor desensitization does not seem to influence EPSC amplitude or PPR during basal synaptic transmission (Hjelmstad et al., 1999), it could be upregulated during a potentiated state. If so, glutamate released by the first pulse in a paired-pulse experiment may desensitize some postsynaptic receptors, limiting the postsynaptic current recorded on the second pulse and decreasing PPR. To address this possibility, we examined the effects of E2 (100 nM) on EPSC amplitude and PPR in the presence of cyclothiazide (CTZ, 100 μM), which blocks AMPA receptor desensitization (Trussell et al., 1993). In 9/14 experiments with CTZ, E2 rapidly increased EPSC amplitude by >20% (on average by 33 ± 4%, Fig. 2C). In these E2-responsive experiments, PPR was initially high and decreased in E2, from 1.53 ± 0.08 to 1.35 ± 0.05 (Fig. 2C; one-way ANOVA for amplitude: p<0.01; one-way ANOVA for PPR: p<0.01), similar to results without CTZ. In the remaining 5/14 experiments in CTZ that were E2-nonresponsive, PPR was low and remained constant in E2 (Fig. 2D). Thus, the E2-induced decrease in PPR is not dependent on AMPA receptor desensitization.
PPR also can be influenced by differences in the quality of voltage-clamp between experiments or over the course of an experiment. To investigate possible voltage-clamp artifact, we analyzed EPSC rise times in a subset of E2 experiments and found them to be comparable in responders and nonresponders and unchanged by E2 application (Fig. S1A). Additionally, analyzing the coefficient of variation of EPSC amplitude (CV = SD/mean), a measure related to presynaptic parameters (high CV reflects low release probability and vice versa; Malinow and Tsien, 1990), showed that baseline CV was higher in responders than nonresponders, and that E2 acutely decreased CV specifically in responders (Fig. S1B; two-way ANOVA: interaction p<0.01). Thus, the most plausible interpretation of our results is that E2 rapidly increases glutamate release probability preferentially at those synapses where baseline release probability is relatively low. To our knowledge, this is the first demonstration that E2 acutely regulates presynaptic physiology in the hippocampus.
The observation that E2-responsiveness is related to baseline PPR, a presynaptic characteristic, is difficult to reconcile with the postsynaptic mechanism of E2 action deduced from previous studies. Moreover, the rapid decrease in PPR observed during E2 application points to a presynaptic effect. These results suggest that E2 responsiveness could be a property of specific inputs to a postsynaptic cell, rather than of the postsynaptic cell itself.
To examine this possibility, we delivered paired-pulse stimulation to two non-overlapping sites in the stratum radiatum and recorded EPSCs from a common postsynaptic CA1 pyramidal cell. Interestingly, bath application of E2 often affected the two sets of inputs on a single postsynaptic cell differently. In the most extreme cases (n = 9/31, Fig. 3A–C), EPSCs were potentiated by >20% at one set of inputs (average normalized EPSC amplitude in E2 = 1.30 ± 0.02) whereas no EPSC potentiation occurred at the other set of inputs to the same cell (average normalized EPSC amplitude in E2 = 1.00 ± 0.04). Further, even when recorded from a common postsynaptic cell, E2-responsive inputs were characterized by a higher baseline PPR that decreased during E2 application, while E2-nonresponsive inputs had a lower baseline PPR that remained unaffected by E2 (Fig. 3B, C; two-way ANOVA: effect of group p<0.05, effect of E2 p<0.05, interaction p<0.05). There was no consistent relationship between E2-responsiveness and distance of the stimulating electrode from the cell body layer: in 3/9 experiments, the E2-responsive inputs were more distal than the E2-nonresponsive inputs, and in the other 6/9 experiments, E2-responsive inputs were more proximal. In addition, there was no difference in EPSC rise times between E2-responsive and E2-nonresponsive inputs from a common postsynaptic cell (not shown; paired t-test: p>0.1), arguing that the observed input-specific differences in response to E2 were not caused by voltage clamp error. Thus, the double-input experiments show that acute E2-induced EPSC potentiation (and the corresponding decrease in PPR) are input-specific, rather than postsynaptic cell-specific.
The observation that E2-induced EPSC potentiation depends on presynaptic inputs influences interpretation of the bimodal distribution of E2 responsiveness (Fig. 1B). EPSCs in our experiments reflect the composite response of multiple individual synapses and thus should contain both E2-responsive and E2-nonresponsive individual synapses in some proportion. This suggests that EPSCs resulting from activation of many inputs are likely to contain a mixture of E2-responsive and E2-nonresponsive individual synapses, and thus show EPSC potentiation of relatively small magnitude occurring in most experiments. In contrast, EPSCs resulting from activation of fewer inputs would be more likely to contain mostly E2-responsive or mostly E2-nonresponsive synapses, particularly if the distribution of E2-responsive vs. E2-nonresponsive synapses is somewhat patchy, and to show EPSC potentiation of a larger magnitude but less frequently.
To test these predictions, we compared the effects of E2 in a subset of experiments with relatively small (74 ± 4 pA, n = 25) vs. large (185 ± 25 pA, n = 31; Fig. S2A) EPSCs selected to have the same baseline PPR (each 1.45 ± 0.03; Fig. S2B). Thus, small and large EPSCs had similar proportions of low and high release probability synapses, on average, but small EPSCs likely contained fewer total synapses than large EPSCs. Consistent with our prediction, we found that large EPSCs tended to respond to E2 more frequently than small EPSCs (71% vs. 64%), but the magnitude of their response was significantly smaller than for small EPSCs (35 ± 3% vs. 47 ± 5%; t-test: p<0.05; Fig. S2C). PPR decreased in E2 similarly for small and large EPSCs (two-way ANOVA: effect of E2 p<0.01, effect of initial EPSC size p>0.2, interaction p>0.3; Fig. S2D). Thus, our results are consistent with the idea that the E2 response of a composite EPSC reflects heterogeneous E2-responsiveness among the individual synapses it contains. Indeed, E2-induced potentiation of some individual synapses may be much more robust than the observed potentiation of composite EPSCs. This analysis also demonstrates that E2-responsiveness does not depend on baseline release probability alone, since small and large EPSCs with the same baseline PPR responded differently to E2. One likely possibility is that a synapse responds to E2 only if its initial release probability is sufficiently low, and it possesses some additional factor, such as the appropriate estrogen receptor.
Which estrogen receptors (ERs) mediate acute EPSC potentiation? The hippocampus contains both classical ERs, ERα and ERβ, and a portion of ERα and ERβ in CA1 is found at excitatory synapses both pre- and postsynaptically (Milner et al., 2001; 2005). We used the ERα and ERβ selective agonists, PPT (Stauffer et al., 2000) and DPN (Meyers et al., 2001), respectively, to investigate the roles of ERα and ERβ in acute EPSC potentiation. Based on the binding affinities of PPT and DPN for their preferred ERs, we used PPT at concentrations 100 or 200 nM and DPN at concentrations 100 or 500 nM to approximate the dose of E2 that was most effective (100 nM).
DPN both mimicked and occluded the effect of E2 to potentiate EPSCs, whereas PPT had no effect (Fig. 4A, B). In 12/23 experiments, DPN rapidly potentiated EPSC amplitude by >20% (on average by 30 ± 2%; Fig. 4C). Additionally, in 15 of these experiments (8 DPN-responsive, 7 DPN-nonresponsive), DPN application was followed by E2 (100 nM), which induced no further potentiation (Fig. 4A, C). Importantly, DPN also decreased PPR in parallel with potentiating EPSC amplitude, specifically in responsive inputs (Fig. 4F, G; paired t-test: p<0.01 for responders, p>0.2 for nonresponders). In 5/11 DPN experiments done with two stimulating electrodes to activate non-overlapping inputs to a single cell, one input was potentiated by >20% whereas the other was not responsive, and PPR decreased specifically at the potentiated input (not shown; paired t-test: p<0.05 for responsive inputs, p>0.2 for nonresponsive inputs). Thus, like with E2, DPN-induced EPSC potentiation is input-specific, not postsynaptic cell-specific.
In contrast to results with DPN, PPT failed to potentiate EPSC amplitude in any of 15 experiments. In 8 experiments, PPT application was followed by E2 (100 nM), and in 5 cases, E2 after PPT induced robust EPSC amplitude potentiation (Fig. 4B, D), showing that inputs that failed to respond to PPT were capable of responding to E2. Paired-pulse ratio also was unchanged by PPT (Fig 4H; paired t-test: p>0.1). Together, these results indicate that acute actions of E2 to potentiate EPSCs and decrease PPR are mediated by ERβ and not by ERα. That DPN mimics both E2 effects is consistent with the idea that EPSC potentiation and decreased PPR are related, such that E2 acutely potentiates EPSCs by increasing glutamate release probability.
To further investigate the role of ERs in mediating acute EPSC potentiation, we used ICI 182,780, which binds ERα and ERβ with similar high affinity and behaves as an antagonist of classical nuclear ER activity in transcriptional assays (Sun et al., 2002) by blocking receptor dimerization (Fawell et al., 1990; Pike et al., 2001). Surprisingly, ICI 182,780 (100 nM) alone potentiated EPSC amplitudes in 8/26 experiments. In 14 of these experiments (4 ICI-responsive, 10 ICI-nonresponsive), ICI 182,780 application was followed by E2 (100 nM), which induced no significant additional potentiation (Fig. 4E). Similar to E2 and DPN, ICI 182,780 decreased PPR in parallel with potentiating EPSC amplitude (Fig. 4I, J; paired t-test: p<0.05 for responders, p>0.3 for nonresponders) and potentiated EPSCs in an input-specific manner (4/12 double-input experiments). The finding that ICI 182,780 alone can mimic and occlude acute effects of E2 on excitatory synaptic transmission suggests that ERβ mediates acute E2-induced EPSC potentiation without dimerizing.
Because ICI 182,780 has been reported to activate the G protein-coupled membrane ER, GPR30 (Filardo et al., 2002), we also tested whether a selective agonist for GPR30, G-1, mimicked E2-induced EPSC potentiation, as has been suggested previously (Lebesgue et al., 2009). E2 and G-1 have similar binding affinities for GPR30 (Ki of 5.7 and 11 nM, respectively) and 100 nM G1 is sufficient to induce maximal Ca++ mobilization in GPR30-transfected COS7 cells (Bologa et al., 2006). We found that G-1 (100 nM) potentiated EPSCs by >20% in only 2/15 experiments, by 25% and 33% (data not shown). G-1 decreased PPR in these experiments from 1.53 to 1.47 and 1.42 to 1.24 (data not shown). G-1 did not affect EPSC amplitude or PPR in the remaining 13/15 experiments. Thus, while we cannot exclude some role for GPR30 in EPSC potentiation, the small proportion of G-1 responsive experiments indicates that GPR30 is unlikely to fully account for EPSC potentiation by E2. The results with G-1 were in stark contrast to the effects DPN, which closely mimicked and occluded EPSC potentiation by E2, arguing that E2 acts primarily through ERβ to potentiate EPSCs.
To investigate mechanism(s) of E2-induced potentiation, we considered two major factors that influence glutamate release probability: the number of neurotransmitter vesicles in the readily-releasable pool (RRP), and the probability of release of an individual vesicle (Pves) from this pool. Long trains of high-frequency stimulation can be used to deplete the RRP (Schneggenburger et al., 1999). Steady-state release at the end of a train is related to the rate of vesicle replenishment. Assuming vesicle replenishment takes place at the same constant rate throughout the stimulus train, relative RRP size can be estimated by subtracting cumulative steady-state EPSC charge from cumulative EPSC charge recorded while the RRP empties, before the steady state is reached. Relative Pves can then be estimated as the fraction of the RRP released in response to the first pulse of the train (Wesseling and Lo, 2002).
Trains of EPSCs evoked by 100-pulse 20-Hz stimulation at the elevated temperature used for previous experiments (32–35°C) failed to reach a steady state (not shown), but trains evoked by identical stimulation at room temperature (20–22°C) reached a steady state after ~70 pulses (Fig. 5A; Wesseling and Lo, 2002; Garcia-Perez et al., 2008). The range of paired-pulse ratios at the two temperatures was comparable, consistent with others findings that baseline release probability is unaffected by temperature (Allen and Stevens, 1994). Similarly, the effects of E2 (100 nM) at room temperature were similar to results at elevated temperature (Fig. S3A–D). E2 increased EPSC amplitude at room temperature in 24/37 (65%) experiments (Fig. S3C, D; normalized EPSC amplitude in E2 in responders = 1.38 ± 0.04, normalized EPSC amplitude in E2 in nonresponders = 1.03 ± 0.02) and induced a robust decrease in PPR specifically in responders (Fig. S3C, D).
We then tested whether acute E2 effects on excitatory synaptic transmission are associated with changes in relative RRP size and/or relative Pves estimated using 100-pulse 20-Hz stimulus trains at room temperature. The biggest effect of E2 was to increase the first few EPSCs in the train (Fig. 5B, C). Comparison of EPSC trains evoked before and during E2 application revealed no difference in cumulative steady-state EPSC charge between the two conditions (Fig. 5D; p>0.2), suggesting that E2 does not modulate the steady-state rate of vesicle replenishment. While there was a small increase in cumulative RRP charge in E2 (Fig. 5E; p<0.01), E2 caused a more robust increase in the fraction of the RRP released in response to the first pulse, indicating an increase in Pves (Fig. 5F; p<0.01). Furthermore, examining individual experiments showed that E2-induced EPSC potentiation was correlated with the increase in Pves, but not in the RRP. There was no correlation between the relative change in RRP and E2-induced EPSC potentiation (Fig. S3E, r = 0.3597, p>0.1). In contrast, a large relative change in Pves was consistently associated with greater EPSC potentiation by E2 and vice versa (Fig. S3F, r = 0.7136, p<0.01). This observation, and the fact that the magnitude of the Pves increase in E2 (36 ± 9%) was similar to the degree of EPSC potentiation (38 ± 4%), suggests that E2 potentiates EPSC amplitude primarily by increasing individual vesicle release probability.
Increased Pves could result in a larger fraction of stimulated synapses releasing a single vesicle and/or cause some synapses to release more than one vesicle upon stimulation (i.e., multivesicular release, MVR). Compared to single vesicle release, MVR leads to higher glutamate concentration in the synaptic cleft following stimulation. Evidence for MVR has come from experiments using low-affinity glutamate receptor antagonists that block EPSCs in a manner dependent on cleft glutamate concentration. Manipulations known to increase Pves, such as elevated extracellular Ca2+ or 4-AP, increase average cleft glutamate concentration whereas manipulations that decrease Pves, such as adenosine, have the opposite effect, consistent with changes in Pves being related to changes in MVR (Tong and Jahr, 1994; Christie and Jahr, 2006). Based on these studies, we hypothesized that acute E2 application, which we found increases Pves, might enhance MVR, resulting in higher average cleft glutamate concentration.
We used γDGG (1 mM), a low-affinity AMPA receptor antagonist, to investigate whether E2-induced EPSC potentiation is associated with a higher average cleft glutamate concentration. If so, the degree of EPSC block by γDGG should be smaller in the presence than in the absence of E2. Because of the lifetime of a recording, it was not feasible to examine the degree of γDGG block both before and after E2 application for the same cells. Instead, we examined γDGG block in the absence of E2 (Fig. 6A) in one group of cells, and in the presence of E2 (Fig. 6B) in a separate group of cells. In experiments where γDGG was applied in the absence of E2, E2 was applied after γDGG washout to determine whether the stimulated inputs were E2-responsive or nonresponsive (Fig. 6A). In agreement with the MVR hypothesis, E2 decreased γDGG block (corresponding to a larger fraction of the EPSC remaining in γDGG), indicating higher average cleft glutamate concentration in E2 (Fig. 6E; t-test p<0.05). Importantly, this was true only in E2-responsive experiments, and not in E2-nonresponsive experiments (not shown, t-test: p>0.3), demonstrating that the E2-induced increase in cleft glutamate concentration is specifically associated with E2-induced EPSC potentiation. The magnitude of E2’s effect on γDGG block is comparable to the effect of increasing extracellular Ca2+ from 1.5 mM to 2.5 mM (Christie and Jahr, 2006), a manipulation known to significantly influence neurotransmitter release. To confirm that the difference between the degree of γDGG block in the absence vs. presence of E2 in responders was not caused by voltage-clamp error, we repeated this experiment with NBQX (200 nM), a high-affinity AMPA receptor antagonist for which block is independent of cleft glutamate concentration (Fig. 6D–F). As expected, and unlike γDGG block, the degree of NBQX block was the same in the presence and absence of E2 (Fig. 6F; t-test: p>0.4). Thus, these results show that E2 increases average cleft glutamate concentration, which, together with our result that E2 increases Pves (Fig. 5F), strongly suggests that E2 enhances MVR.
Finally, we asked whether E2 could occlude the effect of a manipulation known to increase MVR. Paired-pulse facilitation is related largely to higher Pves on the second pulse in a pair due to residual presynaptic Ca2+ during the second pulse. Others have shown that increased Pves on the second pulse increases MVR, as indicated by less γDGG block of the second EPSC in a pair and therefore a higher PPR in γDGG (Christie and Jahr, 2006). If E2 increases Pves and MVR by enhancing presynaptic Ca2+ influx, it might occlude MVR on the second pulse in a pair and thus the ability of γDGG to increase PPR. Alternatively, if E2 promotes glutamate release through a mechanism other than by increasing presynaptic Ca2+, the effect of γDGG on PPR should be similar in the absence and presence of E2.
We first evaluated the effect of γDGG on paired EPSCs and PPR (50 ms IPI) in the absence of E2 (Fig. 7A) and confirmed that γDGG blocks the second EPSC in a pair to a lesser extent than the first, resulting in an increase in PPR (Fig. 7C, D). Then we tested whether E2 could occlude the effect of γDGG on paired EPSCs and PPR (Fig. 7B). This showed that in the presence of E2, γDGG block of the second EPSC was still less than block of the first EPSC (Fig. 7C; two-way ANOVA, interaction p>0.4). As a result, γDGG increased PPR similarly in the absence and presence of E2, by 11 ± 7% vs. 11 ± 3%, respectively (Fig. 7D; two-way ANOVA: interaction p>0.7). This indicates that although E2 increases average cleft glutamate concentration during the first pulse, glutamate concentration increases even further on the second pulse. As with single-pulse γDGG experiments, we excluded a possible contribution of voltage-clamp error by showing that the two EPSCs in a pair were blocked equally by NBQX, both in the absence and presence of E2, resulting in no change in PPR in NBQX (Fig. 7E–H). Thus, these experiments demonstrated that the E2-induced increase in cleft glutamate concentration, assayed by γDGG block, is additive with the increase due to paired-pulse facilitation.
The failure of E2 to occlude γDGG’s effect on PPR is in contrast to the effect of increasing Pves by elevating extracellular Ca2+. Elevating Ca2+ from 1.5 mM to 2.5 mM partially occludes the ability of paired-pulse facilitation to increase cleft glutamate concentration further (Christie and Jahr, 2006), presumably because both manipulations increase Pves through the same mechanism, increased presynaptic Ca2+ concentration. Our observation that E2 does not occlude γDGG’s effect on PPR suggests that the increase in Pves and enhanced MVR due to E2 likely involves a factor(s) other than increased presynaptic Ca2+ influx. For example, E2 may increase Pves and MVR by mobilizing vesicles closer to sites of Ca2+ influx, and/or by increasing Ca2+ sensitivity of vesicle release machinery. These possibilities are especially intriguing in light of reports of synaptic vesicle-associated estrogen receptors (Milner et al., 2001; 2005; Hart et al., 2007), whose function is currently unknown.
We find that E2 acutely potentiates excitatory synaptic transmission in the hippocampus through a presynaptic mechanism, by increasing the probability of glutamate release at synapses with a relatively low initial release probability. This effect is mediated by ERβ, not ERα, and is mimicked by ICI 182,780 indicating that ERβ acts as a monomer. We further show that E2 increases glutamate release primarily by increasing individual vesicle release probability and that it increases average cleft glutamate concentration, which together argue strongly that E2 enhances multivesicular release. The concentration-dependence and time course of acute EPSC potentiation suggest that locally-synthesized neurosteroid E2, not ovarian E2, may activate this effect in vivo.
Earlier studies suggested that E2 exerts its acute effect(s) by enhancing postsynaptic sensitivity to glutamate in a subset of cells (Wong and Moss, 1992; Gu and Moss, 1996). We find instead that E2-sensitivity is input-specific, rather than cell-specific; inputs characterized by relatively high initial PPR (low probability of release) are those at which E2 subsequently potentiates EPSCs. Also in contrast with previous studies, but consistent with a presynaptic effect, E2 decreases PPR in parallel with increasing EPSC amplitude.
Our experiments differed from previous studies in several ways. Perhaps most importantly, we explicitly considered heterogeneity of E2-responsiveness. By dividing inputs into E2-responsive and E2-nonresponsive, we were able to detect a change in PPR specifically in E2-responsive inputs. Previous studies that evaluated PPR without separating E2-responsive and nonresponsive inputs used 100 pM (Kim et al., 2006) or 1 nM E2 (Foy et al., 1999; Kramar et al., 2009). Considering the subset of our experiments with these concentrations, the increase in EPSC amplitude for E2-responsive and E2-nonresponsive inputs together was statistically significant, whereas the decrease in PPR was not. Additionally, particularly for previous studies in dissociated cells (Gu and Moss, 1996), it is possible that agonist application activated primarily extrasynaptic receptors, which may not accurately reflect properties of synaptic transmission. Finally, while our results strongly support a presynaptic mechanism for synaptic potentiation by E2, they do not exclude the possibility that E2 also increases postsynaptic sensitivity to glutamate.
We used high-frequency stimulus trains to investigate mechanism(s) by which E2 enhances glutamate release. We found a small increase in cumulative release during a train, indicating increased RRP size, and a stronger effect on the fraction of RRP released by the first pulse, indicating increased Pves. While it is possible that E2 increases both the RRP and Pves in parallel, a more parsimonious explanation is that increased Pves accounts for both observations.
Although the RRP has been functionally defined as cumulative release during various depletion protocols, including stimulus trains (Stevens and Tsujimoto, 1995; Rosenmund and Stevens, 1996; Wesseling and Lo, 2002), release during a train is influenced by factors in addition to how many vesicles are available for release at the beginning of stimulation. For example, a common assumption in estimating RRP from trains is that the rate of vesicle replenishment during the train is constant as estimated from steady-state charge (Schneggenburger et al., 1999). However, there is evidence that the rate of replenishment increases in a Ca2+-dependent manner during the first 10–20 pulses, reaching an elevated level that persists for the rest of the train (Stevens and Wesseling, 1998; Wesseling and Lo, 2002). A change in the time course of such activity-dependent upregulation of vesicle replenishment could result in more total vesicles released during the train and a larger RRP estimate. Second, the identity/properties of vesicles constituting RRP may vary under different conditions. For example, under conditions of increased Pves, additional, “reluctant” vesicles may be released during a train, resulting in a larger RRP estimate (Moulder and Mennerick, 2005). These considerations suggest that E2 could increase both the RRP estimate and Pves by a single mechanism. For example, E2 could elevate intracellular Ca2+, mobilize vesicles closer to Ca2+ sources, or upregulate one of the biochemical steps in the release process, any of which could result in higher Pves, faster vesicle replenishment, and/or release of additional, “reluctant” vesicles.
The observation that E2 consistently and robustly increased Pves, suggested that E2 might promote MVR. The possibility and physiological significance of MVR at hippocampal synapses has been questioned based on minimal stimulation experiments (Stevens and Wang, 1995) and estimates of glutamate receptor saturation following release of a single vesicle (Clements et al., 1992). However, other studies suggest that MVR does occur, particularly under conditions of elevated Pves (Oertner et al., 2002; Tong and Jahr, 1994; Christie and Jahr, 2006), and argue that glutamate receptors are not saturated by a single vesicle (Liu et al., 1999; McAllister and Stevens, 2000). While evidence for MVR in the immature hippocampus is growing, no studies to date have reported MVR in the adult.
We used the low-affinity AMPAR antagonist γDGG to compare relative cleft glutamate concentration in the presence vs. absence of E2. We found that γDGG blocked EPSCs less effectively in E2, indicating that E2 increases average cleft glutamate concentration. Although there are several mechanisms by which E2 could increase cleft glutamate concentration, MVR is the most likely. One alternative is that E2 downregulates astrocytic glutamate uptake, resulting in increased spillover. This is unlikely, however, because glutamate uptake is extremely efficient (Diamond and Jahr, 2000), such that even when it is pharmacologically reduced, the degree of γDGG block of EPSCs remains unaffected (Christie and Jahr, 2006). E2 also could increase cleft glutamate by causing faster or more complete vesicle emptying. This too is unlikely, as it should result in faster EPSC rise times in E2 (Choi et al., 2000), which we did not observe. Thus, the best explanation for our findings that E2 increases both Pves and average cleft glutamate concentration is that E2 increases MVR. To our knowledge, this is the first demonstration of MVR in adult hippocampus. Further, our observation that the increase in cleft glutamate due to E2 fails to occlude the increase due to paired-pulse facilitation suggests that E2 may enhance MVR by a mechanism other than elevating presynaptic Ca2+. This finding argues for the possibility that E2 mobilizes vesicles toward sites of Ca2+ influx and/or regulates vesicle release machinery.
In neurons, E2 acutely increases Ca2+ influx through L-type Ca2+ channels and activates protein kinases including Src, Erk1/2, CaMKII, and PKA (Lee et al., 2004; Wu et al., 2005; Gu and Moss, 1996). Most of these effects are mediated by classical ERs acting outside the nucleus (Wade and Dorsa 2003; Wu et al., 2005; Zhao and Brinton 2007), but neither the initial steps of extranuclear ER activation, nor the consequences of rapid E2 signaling are well understood.
We found that E2-induced EPSC potentiation is mimicked by both an ERβ agonist and ICI 182,780. ICI compounds interfere with ER dimerization (Fawell et al., 1990; Pike et al., 2001), and consequently block classical nuclear ER activity (Sun et al., 2002). In contrast, rapid E2 signaling is less consistently blocked by ICI compounds (Singh et al., 1999) and may even be activated by them (Zhao et al., 2006). Our results are consistent with these latter reports, and suggest that ERβ acts as a monomer to activate acute EPSC potentiation.
Interestingly, targets of E2-activated kinases include synaptic vesicle proteins critical for vesicle mobilization and Ca2+-dependent release (Chi et al., 2003; Kushner et al., 2005; Menegon et al., 2006; Onofri et al., 2007; Hilfiker et al., 1999), suggesting that E2 could enhance release by regulating synaptic vesicle proteins. If E2 initiates a cascade of molecular events leading to enhanced vesicular release, such as vesicle protein phosphorylation, this may explain our observation that EPSC potentiation persisted after E2 washout. Since EM immunocytochemical studies have found ERβ in presynaptic boutons in CA1 (Milner et al., 2005), exploring the possibility that E2 acts via vesicle-associated ERs to regulate synaptic vesicle proteins will be an important area of future research.
While EPSC potentiation occurred with 100 pM E2, the effect was more robust at higher than circulating E2 concentrations. Additionally, the time course of E2-induced EPSC potentiation is faster than fluctuations in circulating E2 in vivo. These issues raise the question of whether acute E2-induced EPSC potentiation occurs physiologically. Importantly, recent studies show that the hippocampus can generate neurosteroid E2, resulting in high local E2 concentrations that can change rapidly (Hojo et al., 2004; Hojo et al., 2009). Additionally, EM immunocytochemistry shows that aromatase, the rate-limiting enzyme in E2 synthesis, is present both pre- and postsynaptically at a subset of synapses in hippocampal CA1 (Hojo et al., 2004). Together, these findings suggest that neurosteroid E2 could activate EPSC potentiation in vivo. Indeed, the concentration-dependence of EPSC potentiation likely protects hippocampal synaptic transmission from modulation by the relatively slow and low amplitude fluctuations in circulating E2 from the ovaries. Thus neurosteroid E2 may interact with presynaptic ERβ to activate a distinct suite of cellular/molecular events that increase glutamate release resulting in rapid, input-specific potentiation of hippocampal synaptic transmission.
This research was supported by NS037324, MH067564, and RR015497. The authors wish to thank Indira Raman for helpful discussions.