The ramifications of endocrine and neural senescence converge in the hippocampus, particularly with respect to glutamatergic synapses. In this review, we will focus on current literature suggesting that potential synaptic alterations induced by estrogen in the hippocampus are mediated through interactions between ER-α and NMDA receptors. In addition, we will examine the data suggesting that these interactions may be uncoupled with aging. These studies demonstrate that while estrogen helps retain a youthful synaptic phenotype by some measures, the aged synapse may differ from the young synapse in several key respects that impact plasticity in general, and endocrine influences on the synapse, in particular.
Recent evidence indicates that changes in sex steroid hormones affect not only the hypothalamic circuits directly involved in reproductive behavior and physiology, but also brain regions and circuits that mediate cognitive functions, such as the hippocampal formation. Estrogens regulate multiple aspects of synaptic plasticity in the hippocampus, and these effects may influence memory. Rats given estrogen replacement following an ovariectomy show an increase in the number of spines and synaptic boutons in the CA1 region of the hippocampus relative to ovariectomized animals that lack estrogen replacement (Gould et al., 1990; Woolley and McEwen, 1992, 1993; Woolley et al., 1996). In addition, fluctuations in the dendritic spine density appear during the estrous cycle of the rat: dendritic spine density on CA1 pyramidal neurons decreases on days when levels of estrogen are low, i.e. proestrus, as compared to days when estrogen levels are higher, i.e. estrus (Woolley et al., 1990; Woolley and McEwen, 1992).
Current neurobiological evidence suggests that one mechanism by which estrogen alters synaptic plasticity is through N-methyl-D-aspartate (NMDA)-type glutamate receptors. Glutamate receptors are the primary mediators of excitatory transmission in the central nervous system (Hollmann and Heinemann, 1994). They play an important role in learning and memory, and have the capacity to mediate cell death under certain conditions, implicating them in several devastating neurodegenerative disorders and age-related functional decline (Olney, 1983; Rothman, 1984; Morris et al., 1986; Choi, 1988; Bliss and Collingridge, 1993; Lipton and Rosenbeerg, 1995). Pharmacologically, glutamate receptors can be divided into several classes, one metabotropic (G-protein linked) and three classes of ionotropic: NMDA, alpha-amino-3-hydroxy-5-methyl-4-isoxazole proprionic acid (AMPA), and kainate (KA) receptors (Hollmann and Heinemann, 1994). They are each likely to function differently in mediating excitation, learning and memory, and excitotoxicity. Each of the defined classes of glutamate receptors are composed of multimeric assemblies of protein subunits (Hollmann and Heinemann, 1994), with NR1 an obligatory subunit (NMDAR1), and functional attributes of the receptor complex partially dependent on which NR2 subunits are present.
Estradiol appears to have a direct effect on hippocampal NMDA receptors. For example, it increases NMDA agonist binding in the CA1 region of the hippocampus, but has no effect on AMPA receptors (Weiland, 1992). In addition, NMDA receptor antagonists block the estrogen-induced increase in dendritic spine density, and NMDA receptor dependent long-term potentiation is enhanced on days of proestrus, i.e. when estrogen levels area high (Woolley and McEwen, 1994; Warren et al., 1995; Murphy and Segal, 1996). A study in our laboratory using confocal laser scanning microscopy, demonstrated a 30% increase in NMDAR1 fluorescence intensity in the dendrites of CA1 pyramidal neurons with estrogen replacement (Gazzaley et al., 1996).
In summary, these data demonstrate conclusively that estrogens regulate synaptic plasticity in the young hippocampus, reinforcing the fact that sex steroids such as estrogen can directly impact regions and circuits underlying cognitive function, such as the hippocampal formation. The data summarized above are all from young rats, however, what role does estrogen play in age-related cognitive impairment? While many studies have addressed the effects of estrogen deprivation and replacement in hypothalamus in both young and aged animals (Wise and Ratner, 1980; Steger et al., 1983; Gee et al., 1984; Mobbs et al., 1984; Rubin et al., 1985; Caraty and Locatelli, 1988; Belisle et al., 1990; Hwang et al., 1990; Joshi et al., 1995), few studies have addressed the differential effects of estrogen and age on the hippocampus (Miranda et al., 1999). Thus, we proposed the following questions: (i) Does estrogen increase axospinous synapse density on CA1 pyramidal cells in aged rats in a manner similar to young animals? (ii) Is the dendritic increase in NMDAR1 manifested at the synapse? (iii) If so, is the synaptic effect equivalent in young and aged rats?
To answer these questions, we used quantitative ultrastructural approaches to reveal age-related differences in estrogen-induced synaptic plasticity in the CA1 region of the hippocampus [see Fig. 3 legend for animal ages and numbers (Adams et al., 2001)]. An analysis of axospinous synapse density revealed that while estrogen increases synapse density in the young hippocampus in a manner similar to that described by other investigators (Gould et al., 1990; Woolley et al., 1990; Woolley and McEwen, 1992, 1993; Woolley et al., 1996), it fails to increase it in the aged hippocampus (Adams et al., 2001). Moreover, there appears to be an overall loss of axospinous synapses that is occurring in the aged CA1. In order to determine whether NMDAR1 levels are shifting at the synapse we used post-embedding immunogold (Fig. 1A,B), and quantified the distribution of each gold particle using software developed in our laboratory (i.e. SYNBIN; Fig. 2), on the basis of principles regarding proximity to membranes articulated by Ottersen and colleagues (Blackstad et al., 1990; Ruud and Blackstad, 1999). Accordingly, the position of each gold particle (Fig. 2A) is determined as it relates to the post- and presynaptic membrane structures (Fig. 2B). The program analyzes the resulting data map (Fig. 2C) and objectively assigns each gold particle to a given bin (Fig. 2D), with bin sizes and targeted synaptic domains established prospectively. Through this process, a precise gold particle/bin density emerges that is an accurate reflection of gold particle distribution and density in different compartments of the synaptic complex (Adams et al., 2001). Only synapses that contained two or more gold particles associated with the postsynaptic density were considered in the current analyses. Moreover, minimal mitochondrial labeling was observed and controls omitting the primary antibody were also performed to assess the specificity of the secondary antibody.
We found that estrogen does not alter the expression of NMDAR1 per synapse in the young CA1 synapse, however, it does affect the aged CA1 synapse, increasing the amount of NMDAR1 per synapse in the aged animal to that of a young animal (Adams et al., 2001). Therefore, we concluded that the increase in dendritic NMDAR1 revealed with immunofluorescence (Gazzaley et al., 1996) serves to provide adequate NMDA receptors to the additional axospinous synapses induced by estrogen replacement therapy (ERT) in young female rats, but does not alter the NMDA receptor profile at the synapse in young animals. With age there is a loss of CA1 synapses that is not reversible by ERT, yet these synapses still respond to estrogen through increasing NMDA receptors at the synapse. Moreover, these changes in NMDA receptor levels are occurring in the absence of alterations in postsynaptic density lengths (Adams et al., 2001). Thus, while estrogen impacts the aged CA1 synapse in a manner that might help preserve hippocampal function, it does so in the context of a synaptic density compromised by age (Adams et al., 2001). In addition, an estrogen-induced increase in synaptic NMDA receptors may have implications for excitotoxicity that need to be explored further.
What is it about the aged CA1 spine that makes it less responsive to ERT? To this end we focused on the role of the estrogen receptor (ER). Recent evidence demonstrated that both estrogen receptor-alpha (ER-α) and estrogen receptor-beta (ER-β) are expressed in the hippocampus (Li et al., 1997; Shughrue et al., 1997; Weiland et al., 1997; Pau et al., 1998; Petersen et al., 1998; Register et al., 1998; Shughrue et al., 1998; Shughrue and Merchenthaler, 2000; Brinton, 2001). Additionally, Milner et al. (2001) found ultrastructural evidence for the localization of ER-α within dendritic spines in CA1. These data suggest that ER-α may act locally, as well as by regulating nuclear transcription. The potential local, non-genomic effects related to plasticity include regulating neurotransmitter release, signaling cascades, and mRNA translation (McEwen, 2002).
We performed a quantitative ultrastructural analysis of synaptic ER-α in the same animals that were analyzed for synapse density and NMDAR1 localization (Adams et al., 2002). Using postembedding immunogold we observed that the percentage of ER-α labeled synapses decreases with age, independent of estrogen status, by ∼50% (Adams et al., 2002). In addition, ERT led to a decrease in ER-α in both the terminal and the spine in young CA1, but did not change ER-α distribution in the aged spine. Thus, in young animals, peri-synaptic ER-α in CA1 is dynamic and responsive to circulating estrogen levels, but with aging, responsiveness to circulating estrogen decreases (Adams et al., 2002). These highly localized shifts in ER-α regulation and distribution may contribute to age-related decreases in synaptic plasticity. In addition, recent data suggests that the combined effects of aging and ovarian hormones can impact basal forebrain cholinergic function, and this may augment the functional impact of our observation in the aged animals (Gibbs, 2003).
While the study from our laboratory (Adams et al., 2002) confirms a previous study (Milner et al., 2001) demonstrating that ER-α-IR is present within the presynaptic terminal and spines of CA1 pyramidal neurons, and also that it is responsive to estrogen in young animals and decreases with aging, it is important to note that the distribution and nature of estrogen receptors in the hippocampal formation is far from conclusive. Interestingly, most of the ER-α-IR is within extra-synaptic membranes, the spine cytoplasm or associated with presynaptic vesicles, rather than within the post-synaptic density. It will be of great importance to determine the identity of proteins that are co-localized and potentially interact with the spinous ER-α-IR with greater precision. For example, it will be important to determine the degree to which the ER-α-IR is sequestered in membranous caveolae-like structures as has been hypothesized (Toran-Allerand, 2000), since such a localization would position ER-α to interact with the multiple signal transduction pathways that have been hypothesized to function within such an organelle in non-nervous system tissues (Schlegel et al., 1999). A critical role for ER-α in such caveolae was recently demonstrated in endothelial cells, in that caveolar ER-α was functionally and biochemically linked to nitric oxide synthase (NOS) in a critical signaling module capable of regulating the local calcium levels. Such a link between ER-α and NOS could have profound local effects on synaptic plasticity in CA1. However, the substrate for such a complex is hypothetical at present, since calveolin and calveolin-containing structures do not appear to be present in brain, though there may be a structural and functional homologue involving flotillin-anchored calveolae that may be a site for ER-α or a putative homologous ER (Kelly and Levin, 2001), that may be recognized by ER-α antibodies referred to by Toran-Allerand (2000) as ER-X. While such a scenario is compelling, given the rich environment that it would offer regarding local effects of ER-α or its homologue, its characterization in the nervous system in general and CA1 in particular will require extensive additional ultrastructural and biochemical analyses.
Our data suggests that there are potential synaptic alterations induced by estrogen in CA1 mediated through interactions between ER-α and NMDA receptors, and this interaction may be uncoupled with aging (Fig. 3A–D). In young animals, NMDAR1 levels per synapse are consistent across estrogen treatment groups (Fig. 3A,B). Thus, we interpret the increased levels of NMDAR1 in dendritic shafts (Gazzaley et al., 1996) as required to maintain normal levels of NMDAR1 in the new synapses. Furthermore, in the young animal, we hypothesize that only synapses that contain ER-α and NMDA receptors respond to estrogen through formation of multiple synaptic contacts. In contrast, in the aged animals, there is a dramatic decline in the number of spines/synapses that contain ER-α and thus a blunted capacity to form new synapses (Fig. 3C,D). In addition the aged synapse has less NMDAR1 per synapse in the absence of estrogen, yet retains a youthful NMDA receptor profile in the presence of estrogen (Fig. 3C,D). However, given the decrease in ER-α in the aged animal, we must consider that this NMDA receptor response to estrogen is mediated through some mechanism other than ER-α such as, perhaps, ER-β.
While these cellular and synaptic data are compelling with respect to estrogen’s effect on hippocampus and may be relevant to age-related cognitive decline, their relevance to hormone replacement in women is unclear at this time. First, endocrine senescence in rats is quite different from humans, with rodents entering into what is referred to as estropause with the cyclicity becoming irregular and then halted. Unlike humans, this estropause is not accompanied by a dramatic loss of estrogen (Huang HH et al., 1978; Lu et al., 1979; Wise and Ratner, 1980). Thus, the primate model may be a critical intermediate for relating the cellular and synaptic rodent data to human, though little data from non-human primates are presently available. First, like the human, the rhesus monkey has a 28 day menstrual cycle and experiences menopause with reproductive senescence, though it is later in life relative to humans. Recently our laboratory demonstrated that estrogen increased the spine density in the CA1 region of both young and aged female rhesus monkeys (Hao et al., 2002). Thus, unlike the rat, the CA1 region of the aged monkey hippocampus is responsive to estrogen replacement at the level of increased spine and synapse number.
In summary, estrogen impacts the aging synapse at the cellular and molecular level within cells and circuits that are involved in learning and memory. Additionally, estrogen affects performance on learning and memory tasks (Berry et al., 1997; Packard and Teather, 1997; Roberts et al., 1997; Sherwin, 1997; Stackman et al., 1997; Warren and Jurska, 1997; Rapp et al., 2003). Preliminary evidence suggests that primates and rodents may respond differently to ERT, but in both cases there is evidence to suggest that estrogen impacts the aging synapse. Thus, future studies need to be directed at integrating cellular/molecular data and behavior, and determining whether estrogen can help retain a youthful phenotype of plasticity in the aging synapse in both rodents and primates, and the impact of such synaptic effects on behavior. Such data will have important implications for ERT in humans particularly with respect to cognitive function.
Supported by NIH grant AG 16765.