|Home | About | Journals | Submit | Contact Us | Français|
The naturally occurring sex difference in dendritic spine number on hypothalamic neurons offers a unique opportunity to investigate mechanisms establishing synaptic patterning during perinatal sensitive periods. A major advantage of the model is the ability to treat neonatal females with estradiol to permanently induce the male phenotype. During the development of other systems, exuberant innervation is followed by activity-dependent pruning necessary for elimination of spurious synapses. In contrast, we demonstrate that estradiol-induced organization in the hypothalamus involves the induction of new synapses on dendritic spines. Activation of estrogen receptors by estradiol triggers a non-genomic activation of PI3 kinase that results in enhanced glutamate release from presynaptic neurons. Subsequent activation of ionotropic glutamate receptors activates MAP kinases inducing dendritic spine formation. These results reveal a trans-neuronal mechanism by which estradiol acts during a sensitive period to establish a profound and lasting sex difference in hypothalamic synaptic patterning.
Determining the mechanisms underlying the formation of functional neuronal circuits is central to understanding brain development. Variation in neuronal activity, hormonal milieu, or environmental cues can permanently change the formation of specific circuits and the phenotype of individual neurons during a developmental sensitive period. This concept is well established in the visual cortex (Desai et al., 2002; Wong, 1999; Zhang and Poo, 2001), motor neuron circuits (Hanson and Landmesser, 2004; Haverkamp and Oppenheim, 1986), and the song control circuit in birds (Iyengar and Bottjer, 2002), in which, respectively, light, movement or sound affect spontaneous activity in the neurons and in so doing establish the proper synaptic connections for optimal function. Appropriate synaptic connections must also be achieved in male and female brains where fundamental neuronal differences mediate the well-known sex differences in physiology and behavior. In rodents, these sex differences are determined during a critical period of development by estradiol, which is aromatized from testicular androgens in the male (Naftolin et al., 1971). Early estradiol exposure permanently differentiates the male from the female rodent brain by establishing sex differences in synaptic connections and neuronal morphology to affect behavior (De Vries and Simerly, 2002; McCarthy, 2008).
Dendritic spines are anatomical loci of excitatory synapses on neurons. Spine formation and maintenance are sensitive to both electrical and chemical stimulation. Estradiol is an important regulator of dendritic spines in many brain regions during development (Amateau and McCarthy, 2002b; Matsumoto and Arai, 1980, 1986b; Pozzo Miller and Aoki, 1991; Todd et al., 2005) and adulthood (Calizo and Flanagan-Cato, 2000; Carrer and Aoki, 1982; Murphy and Segal, 1996; Woolley and McEwen, 1992), thereby influencing behavior and reproductive function. In newborn rats, males have twice the number of dendritic spines on hypothalamic neurons as females and treatment of females at birth with testosterone, which is centrally aromatized to estradiol in the brain, increases the number of dendritic spines to that of males (Todd et al., 2007). However, the mechanism(s) by which estradiol permanently organizes a sex difference in dendritic spines in the MBH remain unknown.
The classic model of estradiol action involves ligand binding to a nuclear receptor, dimerization and recruitment of co-activators, followed by initiation of gene transcription at estrogen response elements (ERE) found in the promoter region of estrogen-responsive genes. Estrogen receptors (ER) are members of a super-family of nuclear transcription factors that regulate gene expression (Beato, 1989; Uht et al., 1997). Dense concentrations of ER are found in neonatal hypothalamic neurons of males and females (Shughrue et al., 1997; Simerly et al., 1990) and are directly linked to sexual differentiation of the brain (McCarthy et al., 1993; McEwen et al., 1977; Quadros et al., 2002). Activated estrogen receptors can also indirectly promote gene transcription independent of EREs via activation of kinases that in turn activate specific promoter elements. In the hypothalamus, ligand-bound ER activates protein kinase A, CREB (Abraham and Herbison, 2005; Aronica et al., 1994; Lee et al., 2004; Zhou et al., 1996), the mitogen-activated protein kinase (MAP kinase) MEK 1/2 (Singh et al., 1999), and phosphatidylinositol 3-kinase (PI3 kinase) (Znamensky et al., 2003). Thus estrogen receptor activation can be localized to the membrane, yet induce a host of effects which induce or facilitate transcriptional activity dependent or independent of the ERE (Vasudevan and Pfaff, 2007). To date, membrane-initiated effects of estradiol have not been reported in the developing brain, and in fact would seem unlikely given that rapid membrane effects are presumed not conducive to permanent organizational changes. We now report that estradiol promotes spine formation in the developing hypothalamus via rapid, non-genomic activation of PI3 kinase in presynaptic neurons, enhancing glutamate release and glutamate receptor activation in postsynaptic neurons. AMPA and NMDA receptor-dependent activation of postsynaptic MAP kinases then triggers a protein synthesis-dependent increase in dendritic spine number.
On the day of birth, males had significantly more dendritic spines on Golgi-Cox impregnated hypothalamic neurons than females (main effect F3,19= 8.80, p = 0.0007, post-hoc p = 0.001; Fig 1a). Treatment of females with estradiol or NMDA for 6 hr significantly increased the number of spines compared to vehicle treated females (p = 0.001) and equivalent to levels seen on male neurons. Further analysis of the Golgi impregnated neurons revealed the increase in spine number was accompanied by an elongation of neuronal dendrites of 34% (F3,19=3.47, p = 0.036; Fig 1a). There was no difference in the density of dendritic spines (data not shown).
Newborn males also had significantly more spinophilin protein in the MBH than females, as determined by western blot analysis, confirming the previously reported correlation between increased dendritic spine number and increased spinophilin levels (Amateau and McCarthy, 2002a, 2004; Todd et al., 2005; Todd et al., 2007). Treatment of females with estradiol (250 μg) increased spinophilin to that of males within 6 hr (F2,15= 15.8; p = 0.002; Fig. 1b). To determine whether the effect of estradiol on dendritic spines at 6 hr was maintained at later time points, brains were also collected from animals 48 hr after a single injection of estradiol. Spinophilin levels remained significantly elevated 48 hr after estradiol treatment compared to control (F3,20= 4.09; p = 0.031, Fig 1c) and were not different from 6 hr of estradiol exposure (p = 0.901), confirming estradiol-induced dendritic spines are maintained for an extended period of time. The MBH is thus sexually dimorphic at birth with regard to spinophilin level and spine number, and the levels seen in males can be induced in females by treatment with estradiol or activation of the NMDA receptor.
We hypothesized that activation of glutamate receptors is necessary for the estradiol-induced increase in dendritic spines in the developing hypothalamus. We first treated animals with the AMPA-type glutamate receptor antagonist, NBQX (2.5 μg subcutaneously), prior to estradiol treatment. NBQX prevented the estradiol-induced increase in spinophilin (F3,17= 7.04; p = 0.002, Fig 2a). Treatment with NBQX alone had no effect. To determine if NMDA receptors are also required for estradiol-induced spinophilin, animals were pre-treated with the NMDA receptor antagonist, MK801 (10 μg subcutaneously), prior to estradiol treatment. MK801 completely blocked the estradiol-induced increase in spinophilin (F3,13= 13.29; p = 0.0003, Fig 2a). Treatment with MK801 alone had no effect on spinophilin levels.
These results suggest that NMDA receptor activation is necessary for estradiol to induce dendritic spines. The ability of NMDA treatment in vivo to increase dendritic spines detected by Golgi impregnation suggested it is sufficient as well. We further confirmed this by treatment of female pups on the day of birth with estradiol, NMDA (10 μg s.c.), or vehicle. Direct activation of NMDA receptors mimicked the effect of estradiol by increasing spinophilin to the same degree over control females (F2,20=7.70; p = 0.002; Fig 2b). These data confirm the NMDA-mediated increase in dendritic spines seen using Golgi impregnation of VMN neurons (Fig 1a) and are consistent with previous reports indicating activation of AMPA receptors also increases dendritic spines in the developing female hypothalamus to the same degree as estradiol (Todd et al., 2007). We conclude that estradiol does not induce increases in spinophilin protein directly, but rather does so indirectly in a glutamate receptor-dependent manner. Furthermore, activation of glutamate receptors is not only necessary but also sufficient to induce sex differences in dendritic spine number in the developing hypothalamus.
The pivotal observations from the previous experiments suggested activation of AMPA and NMDA receptors is required for estradiol-induced increases in dendritic spines. There are two potential mechanisms by which this could occur: 1) estradiol might alter the number or function of postsynaptic AMPA and/or NMDA receptors or 2) estradiol might act presynaptically to promote glutamate release directly from hypothalamic neurons.
To test the first hypothesis, male and female hypothalami were assessed by quantitative western blot for levels of the AMPA receptor subunits GluR1 and GluR2 and the requisite NMDA subunit, NR1. There was no sex difference and no effect of estradiol treatment on the amount of GluR1 (F2,18= 0.378; p = 0.690), GluR2 (F2,16= 0.488; p = 0.623) or NR1 (F2,14= 0.788; p = 0.474) in the neonatal MBH (Supplementary data Figure 1a).
In another experiment females were treated with or without estradiol (100 μg/0.1 ml sesame oil, sc) on PN0, 1 and 2; then AMPA-mediated miniature excitatory postsynaptic currents (mEPSCs) recorded from ventrolateral VMN (vlVMN) neurons on PN4−5, at which point estradiol-induced dendritic spines are fully established (Matsumoto and Arai, 1986a) (Fig 3b). We found no significant effect of estradiol on the median amplitude (t16=−0.87; p = 0.238), frequency (t16=−1.22; p = 0.396), or decay time (t16=−0.57; p = 0.572) of mEPSCs in the developing hypothalamus (n = 8 − 10 cells from vehicle or estradiol treated; Fig 3a). We conclude estradiol has no effect on the number of synaptic AMPA-type glutamate receptors in these neurons. We also found no effect on the computed ratio of synaptic AMPA-to-NMDA receptors, as seen by measuring the peak amplitude of EPSCs mediated by AMPA receptors and the amplitude of the NMDA receptor-mediated component of the EPSC (p = 0.8, n = 7 cells from each; (Fig 4a). Estradiol also had no effect on glutamate-induced calcium responses in cultured hypothalamic neurons as measured with the fluorescent ratiometric calcium-sensitive dye, fura-2. We found no difference in either the number of neurons that responded to glutamate (χ2= 1.65; p = 0.200) or the peak amplitude of the calcium response to glutamate (K-S test: χ2= 2.12; p = 0.697, Supplementary data Figure 1b).
Recorded neurons were filled with biocytin from the patch pipette and their morphology analyzed. Consistent with the morphological analysis of Golgi impregnated neurons (Fig 1a), neurons treated with estradiol had more dendritic spines along the primary dendrite compared to control (t7=2.67; p = 0.031) and the overall length of the dendrite was greater (t7=2.41; p = 0.060; Fig 3b). There was no effect of estradiol on the number of branches per dendrite or the dendritic spine density (data not shown).
Concluding that estradiol does not affect glutamatergic transmission postsynaptically, we hypothesized that estradiol promotes presynaptic glutamate release to increase dendritic spines. To test this hypothesis, we measured whole-cell voltage-clamp recordings to compare the paired-pulse ratio (PPR) in brain slices from female pups treated with or without estradiol treatment for up to 6 hr on PN2. The PPR is inversely correlated with release probability (Zucker and Regehr, 2002). The amplitude of the AMPA receptor-mediated EPSC elicited in response to the second of two stimuli delivered at an interstimulus interval of 50 ms was considerably facilitated compared to the first EPSC in cells in vehicle-treated slices (PPR = 1.85 ± 0.21, n = 11). In cells from estradiol treated slices, the amplitude of the second EPSC was also facilitated, but to a significantly smaller extent (PPR = 1.22 ± 0.09, n = 12) (t21=−2.72; p = 0.0127, Fig 4a). The decrease in the PPR in the estradiol treated slices is consistent with an estradiol-induced increase in the probability of release at glutamatergic synapses.
We next used the fluorescent styryl dye FM4−64 to assay the effects of estradiol on release probability more directly. Presynaptic boutons were labeled with FM4−64 and the probability of transmitter release was assayed by the rate of FM4−64 destaining in response to either KCl-induced depolarization or electrical field stimulation (Brager et al., 2003; Pyle et al., 1999; Ryan et al., 1996) (Supplementary data Fig 2). Post-hoc analysis of hypothalamic neurons revealed 100% co-localization of FM4−64 staining and VGLUT2 immunostaining, indicating the terminals analyzed for transmitter release were glutamatergic (vehicle: n=32 boutons; estradiol n=36 boutons from 2 separate cultures; Fig. 5a).
Treatment of hypothalamic neurons with estradiol for only 3 hr significantly increased the total fluorescence lost after 30 sec of depolarization with either 50 mM KCl or stimulation at 1 Hz via an electrode positioned near the imaged region, compared to untreated cultures (F3,16= 3.94, p = 0.027; Fig. 5b). After 120 seconds of depolarization, the total fluorescence lost for all groups was not significantly different (p = 0.223), indicating that the size of the recycling vesicle pool was not affected by estradiol.
Treatment with the estrogen receptor antagonist ICI 182,780 alone had no effect on the rate of destaining (p = 0.239; Fig 5c) but completely prevented estradiol-induced enhancement of FM4−64 destaining (F3,7=6.29, p = 0.032), indicating that estradiol enhanced glutamate release requires activation of the estrogen receptor.
There was no effect of estradiol treatment on the density of FM4−64 labeled boutons (F3,25= 0.86; p = 0.470) or the area of these boutons (F3,25= 0.26; p = 0.850) compared to controls (Supplementary data Figure 3).
When analyzed separately, neurons treated with estradiol for less than 2 hr showed an increase in neurotransmitter release when compared to vehicle treatment for the same amount of time (p = 0.05; Fig 5c inset), demonstrating that estradiol rapidly enhances the release of glutamate from developing hypothalamic neurons via activation of classical estrogen receptors. Notably, the increase in release probability within 2 hr precedes the increase in the number of dendritic spines, first detectable at 6 hr, and the activation of MAP kinase (see below) that are not detectable until 3 hr post-estradiol.
The rapid effect of estradiol on glutamate release from hypothalamic neurons suggested a non-genomic mechanism of action. Using 30μM cycloheximide to block RNA translation, we tested whether estradiol-enhanced neurotransmission required synthesis of new proteins. Blocking protein synthesis had no effect on estradiol-enhanced FM4−64 destaining in cultured hypothalamic neurons after 3 hr (F3,24= 3.66; p = 0.67; vehicle n = 8, E2 n = 8, CHX + E2 n = 8, CHX n = 4 coverslips), however, cycloheximide treatment did block the estradiol-induced increase in spinophilin protein (F3,18= 3.17; p = 0.0498; vehicle n = 6, E2 n = 6, CHX + E2 n = 6, CHX n = 4 coverslips Fig. 6a), confirming effective inhibition of protein synthesis and a requirement for protein synthesis in the promotion of spine formation. We conclude that estradiol can enhance neurotransmitter release by activating ER, however, the enhancement does not require the synthesis of new proteins.
A time course analysis of phosphorylation of the PI3 kinase substrate Akt at ser473, using western blot, indicated that estradiol rapidly activated PI3 kinase within only 1 hr (F5,21=3.89,p = 0.048; Fig. 6b) and this activation was maintained at 3 hr and 6 hr after estradiol treatment (p = 0.038 and p = 0.0247 respectively; n = 4−5 coverslips for each group). The activation of PI3K by estradiol therefore just precedes the enhancement of glutamate release, which occurs at 2−3 hr after treatment, suggesting that it may underlie the increase in release probability. We tested this hypothesis by blocking PI3 kinase activation with the specific inhibitor LY294002 (10 μM) and found it completely prevented estradiol-enhanced neurotransmitter release (F3.22= 3.91; p = 0.023, vehicle, n = 7, E2 n = 7, LY + E2 n = 8, LY n = 4 coverslips Fig. 6b). If enhanced glutamate release via estradiol-induced activation of PI3K is necessary for increasing dendritic spine number, then blocking estradiol-enhanced glutamate release with LY294002 would also block the ability of estradiol to increase dendritic spines in these neurons. As predicted, application of LY294002 also blocked estradiol-induced formation of dendritic spines as determined by western blot for spinophilin (F3,15= 4.39; p = 0.003; n = 4−5 coverslips per group, Fig. 6b).
These pharmacological experiments, taken together with the time course of events, are consistent with a model in which estradiol acts via estrogen receptors to activate PI3K via a non-transcriptional mechanism, PI3K activation enhances presynaptic glutamate release, thereby increasing postsynaptic AMPA and NMDA receptor activation to induce spine formation.
We also determined that estradiol activates the MAP kinase, MEK 1/2, as seen by phosphorylation of its substrate ERK 1/2, just prior to the estradiol-induced increase in spinophilin (Fig 7a). By 3 hr post-estradiol treatment, there was significant activation of MEK 1/2 (F2,12=6.60, p = 0.042), that was not seen at 1 hr but maintained for up to 6 hr after estradiol (p = 0.002; Fig. 7c). This time course suggests activation of MAP kinase by estradiol occurs at a point after the enhanced release of glutamate, and that estradiol-induced activation of MAP kinase occurs downstream of glutamate receptor activation to induce dendritic spine formation. This led to the prediction that blocking MAP kinase would have no effect on estradiol-enhanced neurotransmitter release in hypothalamic neurons, and indeed blocking MAP kinase with PD98059 had no effect on estradiol-enhanced FM4−64 destaining (F3,22= 6.67; p = 0.169, vehicle n = 7, E2 n = 7, PD + E2 n= 8, PD n = 4 coverslips Fig. 7a).
Earlier experiments revealed that treatment of females with the glutamate receptor antagonists NBQX or MK801 completely blocked estradiol-induced increases in spinophilin (see Fig. 2). In the same animals, NBQX treatment also completely blocked the estradiol-induced activation of MAP kinase (F3,17= 2.96; p = 0.041, Fig 7b). We determined that treatment with MK801 also prevented activation of MAP kinase by estradiol in vivo (F3,17= 4.79; p = 0.013, Fig 7b). Neither NBQX nor MK801 treatment alone effected basal levels of MAP kinase. These data confirm that activation of MAP kinase occurs downstream of glutamate receptor activation in the developing hypothalamus.
To determine if MAP kinase activity was required for the effect of estradiol on dendritic spines, animals were pre-treated with PD98059 (200 ng ICV). Indeed, PD98059 blocked both the estradiol-induced activation of MAP kinase (F3,16= 7.91, p = 0.002, Fig 7c) and the increase in spinophilin protein within 6 hours (F3,16= 12.8, p = 0.002; Fig. 7c). PD98059 treatment alone had no effect on spinophilin, indicating MAP kinase activity is not required to maintain basal spinophilin levels in the MBH.
We also determined that estradiol promoted an increase in spinophilin protein in cultured hypothalamic neurons generated from E18 embryos and assayed on day in vitro 5 (Supplementary data Figure 4a). We used this in vitro model to test the role of other kinases in the estradiol-induced increase in spinophilin. We found that blocking MEK 1/2 or the c Jun kinase (JNK) significantly blocked the estradiol-induced phosphorylation of ERK 1/2 (F4,19=3.77 p = 0.013; Supplementary data Figure 4b) and increase in dendritic spines (F4,19=8.38; p = 0.005; Supplementary data Figure 4c). Blocking activation of p38 MAP kinase, also located in the hypothalamus, had no significant effect on estradiol-induced ERK 1/2 activation (p = 0.936) or levels of spinophilin (p = 0.074). In addition, blocking protein kinase A (p = 0.130), protein kinase C (p = 0.067) and Cam Kinase II (p = 0.430) had no significant effect on estradiol-induced spinophilin in the hypothalamic neurons (Supplementary data Figure 5). We conclude that activation of glutamate receptors at a time after which estradiol enhances glutamate release, is necessary for estradiol-induced activation of at least two MAP kinases, MEK 1/2 and JNK, and possibly more. Consequently, activation of multiple MAP kinases is a necessary component of estradiol-induced dendritic spines in the developing hypothalamus.
If estradiol promotes dendritic spine formation by enhancing glutamate release from developing hypothalamic neurons, then direct enhancement of neurotransmitter release should mimic the effects of estradiol and increase spinophilin levels in MBH neurons. To test this prediction, we treated cultured hypothalamic neurons with estradiol (10 nM), as a positive control; the estrogen antagonist ICI 182,780 (1μm) with or without estradiol; or α-latrotoxin (3 pM). Latrotoxin is a secretagogue that at low concentrations stimulates neurotransmitter release by facilitating the presynaptic exocytotic machinery (Capogna et al., 1996). As seen previously, there was an overall effect of treatment on spinophilin levels in hypothalamic neurons (F4,20= 9.86; p = 0.001; Fig. 8a). Specifically, treatment of neurons with estradiol significantly increased spinophilin compared to controls (p = 0.001, n=5−6 coverslips). Pre-treatment with ICI 182,780 blocked the estradiol-induced increase in spinophilin (p = 0.007 compared to estradiol treated), but had no effect on basal spinophilin (p = 0.439). Stimulation of glutamate release with α-latrotoxin induced an increase in spinophilin similar to that observed in estradiol-treated cultures (p = 0.68), under conditions in which we demonstrated that estrogen receptors were fully blocked by ICI. With estrogen receptors fully blocked, α-latrotoxin increased spinophilin levels to those seen with estradiol treatment. This experiment provides positive evidence that enhanced glutamate release alone is sufficient to promote the formation of new dendritic spines in the developing hypothalamus, and rules out any possible permissive postsynaptic role for estrogen receptor activation in the induction of dendritic spines, suggesting a completely presynaptic mechanism of estrogen action. In a separate experiment, we also observed that combined α-latrotoxin and estradiol treatment induced an increase in spinophilin expression (F3,26=3.65 main effect) that was not different from the increase induced by either alone (p = 0.221 and p = 0.220; Fig 8b), suggesting that α-latrotoxin occludes the effect of estradiol on hypothalamic neurons. Therefore, enhancing neurotransmitter release is sufficient to increase dendritic spines on developing hypothalamic neurons.
A major role of steroids in the adult brain is modulation of neuronal plasticity, with an increasing emphasis on rapid but transient effects (Balthazart and Ball, 2006). In marked contrast, estradiol action in the developing brain permanently organizes the neural substrate, making a rapid membrane-initiated effect an unlikely mechanism by which this end would be achieved. To our surprise, we have found that the initial event in estradiol-mediated organization of dendritic spine synapses of the developing hypothalamus begins with a transcription-independent activation of PI3 kinase within 1 hr of steroid exposure, as demonstrated by an increase in the phosphorylation of its substrate, Akt. The activation of PI3 kinase was maintained for at least 6 hr under continuous estradiol exposure, as would be found in the developing brain.
Consequent to PI3 kinase activation we observed estradiol-enhanced glutamate release from hypothalamic neurons both in acutely prepared hypothalamic brain slices, as seen by a reduced paired-pulse ratio, and in primary cell cultures, as seen by examining the release of FM4−64 from presynaptic terminals. Application of a depolarizing stimulus, such as KCl or 1-Hz electrical stimulation, evoked calcium-dependent neurotransmitter release and destaining of FM4−64 from boutons that was significantly faster in neurons incubated with estradiol for only 2−3 hr prior to application of the depolarizing stimulus, compared to vehicle treated cultures. This action of estradiol on synaptic excitation was entirely presynaptic, as we detected no change in the number or function of postsynaptic glutamate receptors.
Estradiol-enhanced glutamate release required activation of PI3 kinase. Phosphoinositides are a key component of the membrane and their phosphorylation, by PI3 kinase, produces a signaling molecule important for the recruitment and phosphorylation of other proteins at the membrane (Cremona and De Camilli, 2001; Di Paolo and De Camilli, 2006; Toker and Cantley, 1997). As a result, they play an important role in the processes of exocytosis, endocytosis and trafficking of vesicles at the presynaptic membrane (Cousin et al., 2003; Lin et al., 2001; Opazo et al., 2003; Yokomaku et al., 2003). Though the exact mechanism of action remains unknown, phosphoinositides appear to act as a critical regulatory component of neurotransmitter release. We found that PI3 kinase inhibition did not block all neurotransmitter release, suggesting its role in estradiol-enhanced glutamate release may be to facilitate the fusion of synaptic vesicles to the target membrane (Rizzoli and Betz, 2002), either via changes in the dynamics of the membrane itself or phosphorylation of the proteins necessary for vesicle fusion (Cousin et al., 2003). PI3 kinase might also enhance glutamate release by facilitating the mobilization of calcium into the terminal (Gong et al., 2005; Holt et al., 2003; Mori et al., 2004). We conclude that the role of PI3 kinase in glutamate release appears specific, because inhibition of MAP kinase had no effect on estradiol-enhanced glutamate release; and blocking PKA, PKC and CamKII had no effect on estradiol-induced spinophilin. Moreover, there was no effect of the protein synthesis inhibitor, cycloheximide, on estradiol enhanced glutamate release, suggesting the activation of PI3 kinase by estradiol was due to a direct interaction between the two proteins, presumably in the presynaptic membrane.
The estradiol-enhanced glutamate release and glutamate receptor activation led to the activation of postsynaptic MAP kinases, as demonstrated by increased phosphorylation of ERK1/2. This effect was delayed compared to the activation of PI3 kinase, requiring at least 3 hr of estradiol exposure to become apparent, further confirming the temporal sequence of events. Blocking ionotropic glutamate receptors prevented activation of MAP kinase in response to estradiol and inhibition of the MAP kinases, MEK 1/2 and JNK, but not p38 Kinase, prevented the estradiol-mediated increase in spine number, leading to the conclusion that activation of multiple MAP kinases in the postsynaptic neuron is a consequence of enhanced glutamate receptor activation and not the result of a direct activation by the estrogen receptor. Inhibiting MAP kinase prevented the estradiol-induced increase in spinophilin, establishing a causal role of these enzymes in new spine construction in this system. Phosphorylation of ERK is implicated in the reorganization of actin filaments and thereby altering numerous cellular endpoints, including among many, the targeting of mRNA to dendritic spines (Huang et al., 2007), BK channel activation (O'Malley and Harvey, 2007), and AMPA receptor reorganization at the spine (Kim et al., 2005; Rumbaugh et al., 2006).
The observation that estradiol-induced signaling in the postsynaptic neuron does not involve direct actions of the estrogen receptor highlights an emerging principle in the organizational action of gonadal steroids on the developing brain, the role of cell-to-cell communication. Advances in the mechanism of sexual differentiation of the Drosophila brain highlight the importance of a single gene, fruitless (fru) (Manoli et al., 2005) and further demonstrate that fru acts in a cell autonomous manner to selectively change the morphology of individual neurons that express the protein (Kimura et al., 2005). One might predict, given the restricted distribution and expression of steroid receptors, a similar potential exists in the mammalian brain. Both the current results and our previous findings establish this is not the case. In the preoptic area, another brain region subject to estradiol-mediated organization of synaptic patterning, the initiating event is increased production of prostaglandin E2 (PGE2) following up regulation of the enzyme COX-2 (Amateau and McCarthy, 2004). Drawing on observations that PGE2 induces glutamate release from astrocytes (Bezzi et al., 1998), a working model proposes that in the POA the prostaglandin is released from the neuron and acts upon neighboring astrocytes to induce glutamate release, which in turn acts back on neurons to activate AMPA receptors and induce the formation of dendritic spines. Therefore, in this brain region, neuronal-astrocytic-neuronal communication is essential. Similarly, neurons in the arcuate nucleus of the hypothalamus undergo estradiol-mediated sexual differentiation that begins in neurons with an increase in GABA synthesis via up regulation of glutamic acid decarboxylase (GAD) (Mong et al., 2002). GABA is released from the neurons, presumably non-synaptically, and activates GABAA receptors on neighboring astrocytes and induces growth and branching of glial processes, i.e. stellation. Through mechanisms that remain poorly understood, the differentiated astrocyte communicates back to the neurons and suppresses the formation of new dendritic spines. Thus, in three circumstances in the mammalian brain in which the cellular mechanisms establishing sex differences in synaptic patterning have been well characterized, there is a defined role for cell-to-cell communication (McCarthy, 2008). The consequence of this mode of differentiation is the potential for steroid-mediated differentiation of neuronal morphology to extend beyond those cells that contain steroid receptors and to coordinate the organization of a more extensive region of the neuropil. As a result, the optimal neural circuitry for integrating sensory and physiological stimuli to maximize reproductive fitness can be achieved.
Female Sprague-Dawley rats (Charles River Laboratories, Wilmington, MA) maintained on a reverse 12 hr light/dark cycle and provided ad libitum food and water were mated in our animal facility and pregnancy confirmed by presence of sperm in a vaginal smear. Pregnant females were isolated and allowed to deliver normally. Cages were checked regularly for presence of pups to determine the time and day of birth [postnatal day 0 (PN0)].
All drugs were purchased from Sigma unless otherwise noted. The steroid, 17β-estradiol (250 μg), was dissolved in ethyl oleate. NBQX disodium salt, was dissolved (2.5 μg/0.05 ml) in 0.9% saline. MK801, was dissolved (10 μg/0.05 ml) in 0.9% saline. N-methyl-D-aspartate was dissolved (10 μg/0.05 ml) in 0.9% saline. Biocytin was obtained from Molecular Probes. PD98059, (Calbiochem) was dissolved (100 ng/μl) in 2% DMSO for ICV injections.
Estradiol and vehicle were injected subcutaneously (s.c.). PD98059, was infused intracerebroventricularly (ICV) 30 minutes prior to treatment with estradiol or its vehicle. ICV infusions were performed by penetrating the skin and skull with a beveled 23 gauge 1 μl Hamilton syringe stereotaxically lowered to a predetermined depth of 2 mm before infusion of 1 μl over a 60 second period. NBQX, MK801, NMDA, and saline injections were given subcutaneously immediately before estradiol and again four hours after estradiol injection. Exactly 6 hr after treatment, animals were euthanized and the MBH isolated (Fig 1b).
PN0 rats were separated and treated according to the following groups: 1) males + vehicle, 2) females + vehicle, 3) females + 250 μg 17β-estradiol and 4) females + 10 μg NMDA. Six hours after treatment, brains were collected and processed for Golgi impregnation as previously described (Mong et al., 1999; Todd et al., 2007). After processing of both Golgi and biocytin-filled neurons, 5 neurons from each were analyzed for the number of dendritic spines (> 5μm) on each primary dendrite and the length of the primary dendrite (including all branch lengths extending from that dendrite). Data were analyzed with one-way ANOVA, α < 0.05 and Tukey's post-hoc when appropriate.
Brain tissue from treated animals and hypothalamic cultures were homogenized in lysis buffer and prepared as described previously (Amateau and McCarthy, 2002a; Todd et al., 2007) and details provided in Supplementary Methods
For evoked excitatory postsynaptic currents (eEPSC), female pups were treated on PN2 with estradiol (250 μg/0.1 ml ethyl oleate) or vehicle and collected 6 hr later. For miniature excitatory postsynaptic currents (mEPSC), female rat pups were treated with estradiol (100 μg/0.1 ml) or vehicle once a day each day from PN0−2, and collected on PN4−5. Coronal hypothalamic slices (400 μm) containing the VMN region (Fig 3b) were prepared using a Vibratome (Vibratome series 1000, St. Louis, MO) in ice-cold sucrose solution, where NaCl in the artificial cerebral spinal fluid (ACSF) was replaced by an isosmotic concentration of sucrose. Slices were then incubated at 34°C for 20 min, and then allowed to recover for at least 1 hr at RT before experimentation. For evoked EPSCs, vehicle (DMSO 0.01%) or estradiol (10 nM) were included to the ACSF solution during slices preparation and recordings in vehicle or estradiol pretreated animals, respectively. The osmolarity of all solutions used was 305−315 mOsm. Details regarding recording of eEPSC's and mEPSC's can be found in Supplemental Methods.
Cumulative probability distributions of mEPSC amplitude were created either from control or treated cells. Evoked EPSC amplitude was computed from an average of 10 consecutive responses. The amplitude of NMDA receptor-mediated currents, recorded with the cell voltage-clamped at +40 mV, were collected at 60 ms following the stimulation, a time at which AMPA receptor-mediated currents had completely decayed. The ratios of AMPA and NMDA amplitudes were calculated in consecutive time periods with unchanged stimulation parameters. The paired-pulse ratio (PPR) was calculated as the average amplitude of the second of two EPSCs elicited with extracellular stimuli separated by 50 ms, divided by the average amplitude of the first EPSC in the pair. For these data, 10 individual trials were analyzed and the PPR was computed from the averaged trace. Data were expressed as mean ± SEM, and p values were derived from Student's unpaired t-test, with significance levels assessed at p < 0.05, unless otherwise noted.
Timed pregnant Sprague-Dawley female rats purchased from Charles River Laboratories (Wilmington, MA, USA) were killed on gestational day 18. The hypothalami were dissected from the brains of approximately 8−9 fetal rats and cultured as described in (Perrot-Sinal et al., 2001). Cells were plated onto poly-L-lysine coated glass coverslips at a density of 250,000 cells/plate with plating medium for western blot experiments and 150,000 cells/plate for imaging experiments (FM and calcium) in 2 ml of SCM. Plates were maintained in an incubator at 37° C and 5 % CO2 for 5 days in vitro before use. (DIV1= day of plating) Culture medium was changed one time on DIV3.
For experiments involving FM4−64 imaging, cultures were treated for 3 or 6 hours with 10 nM β-estradiol or veh, prior to imaging on DIV 5. Pre-treatment with antagonists such as ICI 182,780 (Tocris 1 μM in DMSO), PD98059 (10 μM in DMSO), LY294002 (10 μM in DMSO), or cycloheximide (30 μM in ethanol), occurred 30 minutes prior to estradiol or DMSO treatment. α-Latrotoxin (Calbiochem) was administered in saline at 3 pM in combination or separately from estradiol.
Hypothalamic neurons used for FM4−64 imaging were plated on gridded coverslips (Bellco Biotechnology),rinsed in an equilibrating buffer for 15 min before being loaded with 2 μM FM4−64 (SynaptoRed) using a loading buffer for 10 min. Immediately thereafter, neurons were washed in a rinsing buffer (see supplemental methods) for 10 min. Images were captured (red 600−700 nm emission filter) every 3 seconds to minimize photobleaching. After approximately 30 seconds of establishing the baseline fluorescence, the neurons were rinsed for 60 seconds with an unloading buffer (see supplemental methods) containing 50 mM KCl or stimulated at 1 Hz with an extracellular electrode placed in the bath. After ~2 min of image capture during destaining, unloading of the terminals continued for another 5 min to obtain a picture of the completely destained puncta. This image was considered to represent non-specific staining and was subtracted from all images prior to analysis. Images were analyzed using SimplePCI imaging software (Hamamatsu Inc.). The destaining rate was determined from the change in fluorescence emission over time as described in (Brager et al., 2003; Pyle et al., 1999; Ryan et al., 1996). The mean total fluorescence lost for all boutons on one coverslip was averaged, and this value was then averaged with other coverslips treated identically (n = one coverslip, 3−4 separate culture runs per experiment). The mean total fluorescence lost was compared across treatment groups using a one-way ANOVA. Criterion for significance levels was α<0.05. Neurons analyzed for the rate of neurotransmitter release were also analyzed for the number of loaded boutons and the area of each bouton.
Immediately after imaging, coverslips with adhered cells were fixed and prepared for immunocytochemical detection with anti-vesicular glutamate transporter 2 (vGLUT2; Sigma, 1:1,000) then developed as described previously (Speert et al., 2007).