Physical Linkage of Estrogen Receptor α and Aromatase in Rat: Oligocrine and Endocrine Actions of CNS-Produced Estrogens

Abstract

Rapid-signaling membrane estrogen receptors (mERs) and aromatase (Aro) are present throughout the central nervous system (CNS), enabling acute regulation of CNS estrogenic signaling. We previously reported that spinal membrane Aro (mAro) and mERα oligomerize (1). As their organizational relationship would likely influence functions of locally produced estrogens, we quantified the mAro and mERα that are physically associated and nonassociated in two functionally different regions of rat CNS: the spinal cord, which has predominantly neural functionalities, and the hypothalamus, which has both neural and endocrine capabilities. Quantitative immunoprecipitation (IP), coimmunoprecipitation, and Western blot analysis were used to quantify the associated and nonassociated subpopulations of mAro and mERα. Regardless of estrous-cycle stage, virtually all mAro was oligomerized with mERα in the spinal cord, whereas only ∼15% was oligomerized in the hypothalamus. The predominance of nonassociated mAro in the hypothalamus, in combination with findings that many hypothalamic Aro-immunoreactive neurons could be retrogradely labeled with peripherally injected Fluoro-Gold, suggests that a portion of hypothalamic estrogens is secreted, potentially regulating pituitary function. Moreover, circulating estrogens increased hypothalamic Aro activity (quantified by the tritiated water-release assay) in the absence of increased Aro protein, revealing nongenomic regulation of Aro activity in the mammalian CNS. The demonstrated presence of associated and nonassociated mAro and mERα subpopulations in the CNS suggests that their selective targeting could restore impaired estrogen-dependent CNS functionalities while minimizing unwanted effects. The full physiological ramifications of brain-secreted estrogens remain to be explored.

Spinal aromatase is mostly physically associated with mERα, but hypothalamic aromatase is mainly independent of mERα. Some hypothalamic estrogens are likely secreted to act outside of the hypothalamus.

Originally thought to act exclusively as hormonal transcription factors outside the central nervous system (CNS), estrogens are now known also to have a profound influence on the CNS (2, 3). Because estrogen receptors (ERs) that are expressed throughout the CNS are indistinguishable from those in the periphery (4), estrogenic modulation of CNS functions, extrapolated from peripheral estrogen actions, was originally thought to result solely from their modulation of protein synthesis. This perspective was challenged by the discovery of estrogenic effects on the CNS that were seconds to minutes in onset and offset [e.g., Filardo et al. (5), Levin (6), and Liu et al. (7)], much less than that required for transcriptional regulation of protein synthesis to be manifested (8).

It is now well established that the two classical nuclear ERs, ERα and ERβ, traffic to the plasma membrane (9), where their activation rapidly induces multiple membrane-initiated estrogenic signaling cascades (10–13), regulating neuronal excitability (14–20). Importantly, the onset and offset of membrane-initiated estrogenic signaling via membrane ERs [mERs; which also include G protein–coupled receptor 30 (21–24), Gq-mER (25–27), and ER-X (28)] are rapid, occurring within seconds to minutes. The harnessing of the activity of these rapid signaling CNS mERs requires a source of estrogens whose availability can be regulated within a similar time frame.

This prerequisite is fulfilled by the presence of the enzyme aromatase (Aro; which catalyzes the conversion of androgens to estrogens) throughout the spinal cord (29–32), as well as the brain (33–36). CNS Aro localizes to presynaptic boutons (37, 38) and the surface of synaptic vesicles (39–41), enabling accessibility of CNS mERs to a source of estrogens that is not only independent of ovarian production but also rapidly regulated (42–46). This realization fueled the synaptocrine hypothesis (47), which postulates that estrogens function as neurotransmitters. Recently, this laboratory demonstrated that Aro and mERs are not only coexpressed in some spinal neurons, but are also contained within the same multimeric signaling complex—an organization that allows estrogens to function as intracellular messengers whose synthesis and actions are confined to the same signaling oligomer (1). We termed this subcellular modality of estrogenic signaling “oligocrine” (1).

The relative abundance of neuronal Aro and mERs, which are either associated or exist independent of each other, is not known. Also unknown is whether the relative proportions of associated and nonassociated (free) populations of membrane Aro (mAro) and mERs vary among functionally distinct regions of the CNS. Such information would indicate the predominant modality of estrogenic signaling in those regions, which would suggest possible physiological functions of the locally synthesized estrogens.

Accordingly, we investigated the proportion of neuronal mAro and mERα that are either free or physically associated with each other, in spinal cord and hypothalamus. We chose these CNS regions because the spinal cord has predominantly neural functionality, whereas the hypothalamus has both neural and endocrine capabilities. We selected mERα for study because its state of activation is reflected by its Ser 118 phosphorylation (p-mERα) (48, 49), which can be quantified by Western analysis. Experiments were performed using tissue obtained from diestrous and proestrous rats to investigate the influence of the estrous cycle on the physical linkage of CNS mAro and mERα.

Results revealed that the relative abundances of mAro and mERα that are associated with or independent of each other are dynamic and vary in a CNS region-specific, as well as in an estrous cycle-dependent fashion. This suggests that the associated and free subpopulations of mAro and mERα perform distinct functions and are likely to be differentially activated. Findings provide an organizational context for the diversity of physiological functions subserved by rapid CNS estrogenic signaling.

Materials and Methods

Animals and tissue harvesting

All experimental procedures were reviewed and approved by the Animal Care and Use Committees of the State University of New York, Downstate Medical Center or University of Minnesota. Adult Sprague-Dawley rats (Charles River Laboratories; females 225 to 275 g; males 250 to 300 g) were kept on a 12-hour light/dark cycle, with food and water available ad libitum. Estrous-cycle stage was determined using vaginal smear histology. Rats were decapitated and the spinal cord and brain quickly removed and submerged in cold oxygenated artificial cerebrospinal fluid (124 mM NaCl, 26 mM NaHCO₃, 5 mM KCl, 10 mM d-glucose, 1.6 mM MgCl₂, and 2.0 mM CaCl₂, pH 7.4). The spinal cord was then frozen on dry ice and stored at −80°C. The brain was sectioned using Leica VT 1000S vibratome at 500 μm, between plates 17 and 23 of The Rat Brain in Stereotaxic Coordinates (50). Hypothalamus was excised, immediately frozen on dry ice, and stored at −80°C.

Aro activity assay

Aro activity was quantified using the tritiated water-release assay, as previously described (51–53), with minor modifications. In brief, hypothalamus or spinal cord was homogenized (300 μL/60 mg tissue) in 150 mM KCl, 10 mM Tris, 1 mM EDTA, pH 7.2 buffer. Sample (50 μL) was added to 150 μL reaction mixture [150 mM KCl, 10 mM Tris, 1 mM EDTA, pH 7.2 buffer; 10 pmol 1β-³H-androstenedione (PerkinElmer); and 0.5 μmol reduced form of nicotinamide adenine dinucleotide phosphate (MilliporeSigma)]. Specificity of reaction was verified by near-elimination of tritiated water formation following addition of Aro inhibitors letrozole and formestane (1 μM each; MilliporeSigma). Reaction took place overnight at 37°C with agitation. One aliquot of each sample (prepared immediately before termination of reaction) was used for background subtraction. Purification and quantification of the generated tritiated water were performed as previously described (54). An LS6500 scintillation counter (Beckman Coulter) was used to measure disintegrations per minute.

Membrane preparation, coimmunoprecipitation electrophoresis, and Western blotting

Spinal cord membranes were prepared and solubilized, as we reported (7, 55). The authenticity of our membrane preparation was validated by demonstrating (via Western analysis) that ∼81% and ∼98% of the total cell membrane protein marker cadherin (Abcam; catalog no. ab22744, RRID: AB_447300; 1:1000) was present in our membrane fraction from the spinal cord and hypothalamus, respectively (with the remainder in the 100,000 g microsomal fraction). Importantly, our membrane fraction contained only ∼25% and ∼20% of the microsomal marker cytochrome P450 reductase (Abcam; catalog no. ab180597; 1:10,000) in the spinal cord and hypothalamus, respectively, indicating that microsomal contamination is likely not a substantial confound.

Equivalent amounts of total protein (as determined by Bradford) from each sample were immunoprecipitated using either anti-Aro or anti-ERα antibodies and Protein A or G agarose beads (Roche) overnight at 4°C. Thereafter, the supernatant of this immunoprecipitate was immunoprecipitated using either anti-Aro [if the first immunoprecipitation (IP) used anti-ERα antibody] or anti-ERα antibodies (if the first IP used anti-Aro antibody; see Fig. 1). Either 10 or 60 μg total protein (for assessment of mAro or mERα/p-mERα protein levels, respectively) or immunoprecipitate eluate was separated on 4% to 12% Bis-Tris SDS Gels (Thermo Fisher Scientific) and Western blotted. Chemiluminescence was captured using the G:Box charged-coupled device camera (Syngene), and the signals were quantified for comparison using Genetools software (Syngene).

Figure 1.

Schematic representation of the method used to quantify mAro and mERα–associated and –free subpopulations.

Antibodies used for IP and Western blots (WBs) were raised in different hosts to avoid cross-recognition by secondary antibodies. Calculations of free and complexed percentages were possible because (1) all IPs were quantitative, i.e., no directly targeted protein remained in the supernatant following its IP; and (2) Western quantification of the coimmunoprecipitated protein was always normalized against an internal standard (10 and 60 μg total protein from a standard spinal membrane preparation for mAro and mERα/p-mERα, respectively). This permitted quantitative comparisons across experiments (compensating for variations in antibody dilutions, image capture exposure times, etc.).

Antibodies

The following antibodies were used: for IP, anti-ERα (sc-71064, RRID: AB_1122667; 3 μg/800 μg total protein) and anti-Aro (sc-14245, RRID: AB_2088684; 3 μg/800 μg total protein), both purchased from Santa Cruz Biotechnology; and for WB, anti-ERα (Santa Cruz Biotechnology; sc-542, RRID: AB_631470; 1:250), anti-Aro (Abcam; ab124776; RRID: AB_10972863; 1:40,000), and anti-phospho-Ser 118 ERα (p-ERα; Santa Cruz Biotechnology; sc-12915-R, RRID: AB_653188; 1:250). Specificity of these antibodies was recently affirmed (1). The following antibody was also used for immunohistochemistry (IHC), anti-Aro (Abcam; ab124776; RRID: AB_10972863; 1:50,000). Specificity was affirmed by the absence of the IHC signal when performed using preadsorbed antibody flow-through (Supplemental Fig. 1).

Ovariectomy and estrogen supplementation

Bilateral ovariectomy was performed under isoflurane anesthesia, as previously described (56) and performed in our laboratory (57). Estradiol (E2) was administered via subcutaneous implantation of Silastic tubing (10 mm/100 g body weight), filled with a solution of E2 in sesame oil (182.4 μg E2/1 mL) to reach circulating E2 levels comparable to proestrus (58). Ovariectomized (Ovx) rats were implanted 1 week following ovariectomy. Diestrous rats were implanted on day one of diestrus. Control rats were implanted with Silastic tubing containing vehicle (sesame oil). All rats were killed 48 hours after implantation.

Fluoro-Gold injections and IHC

Fluoro-Gold (FG) injections were performed using a variation of the method of Dickson et al. (59). Four milligrams of the retrograde tract-tracer FG (hydroxystilbamidine; FluoroChrome) in 1 mL sterile saline was injected intravenously through the tail vein of four female rats. Two to 4 days later, rats were perfused with 4% formaldehyde. Sections (50 μm) of the forebrain were stained with anti-Aro antibody, followed by cyanine 3–conjugated secondary antibody (Jackson ImmunoResearch). Sections were examined using a FluoView 1000 confocal microscope (Olympus; FG: 405 nm excitation; cyanine 3: 543 nm excitation). Brightness and contrast of comparable images (e.g., normal staining, preadsorption controls, etc.) were adjusted identically using Photoshop (Adobe).

Statistical analysis

SPSS Statistics version 20 (IBM) was used for data analysis. Student t test was used to compare normalized WB signals and Aro activity (femtomoles of converted androstenedione) between proestrous and diestrous groups. One-way ANOVA was used to compare Aro activity among groups. Specific comparisons that achieved significance were identified using Tukey post hoc test. P < 0.05 (two-tailed) was considered significant.

Results

Spinal protein levels of mAro and mERα do not vary between diestrus and proestrus

We compared levels of mAro and mERα in spinal membranes obtained during proestrus and diestrus (Fig. 2A and 2B). In WBs, mAro was visualized at the expected molecular mass (∼55 kDa). No significant difference was observed in the spinal content of mAro protein between proestrous and diestrous rats [t (12) = 1.017, P > 0.05, n = 7]. Likewise, no differences were found in levels of the previously reported (60–62) molecular mass species of mERα (∼65, ∼50, and ∼45 kDa) between proestrus and diestrus [t (12) = −0.250, 0.318, and 1.784 for 65, 50, and 45 kDa, respectively, P > 0.05 for all, n = 7; ∼40- and ∼32-kDa mERα species yielded low-intensity signals precluding their reliable quantification].

Figure 2.

Spinal protein levels of mAro and mERα do not vary between diestrus and proestrus. Equivalent amounts of total solubilized membrane protein (10 μg for mAro and 60 μg for mERα), obtained from spinal cords of proestrus and diestrus rats, were resolved on SDS-PAGE and Western blotted. There were no significant differences in the levels of either (A) mAro or (B) mERα in spinal cords obtained during proestrus and diestrus (n = 7 rats per group). Bar graphs represent mean chemiluminescence + SEM. In (B), we did not quantify the ∼40- and ∼32-kDa mERα species, given their weak Western signal. D, diestrus; P, proestrus.

Virtually all spinal mAro is physically associated with mERα

We previously reported that mAro and mERα coimmunoprecipitate (1). Here, we examined the relative amounts of mAro that associate with and exist independently of mERα, as well as the influence of estrous-cycle stage on these subpopulations (Fig. 3B). In these experiments, the supernatant from quantitative mERα IP was immunoprecipitated using an anti-Aro antibody. Westerns of the anti-ERα IP (first IP, representing complexed mAro) and anti-Aro IP (second IP, representing free mAro) were both blotted for Aro.

Figure 3.

Associated and free subpopulations of mAro and mERαin spinal cord. Solubilized spinal cord membrane fractions were immunoprecipitated using specific antibodies against Aro and ERα. (A) Quantitative IP of mERα (upper blot) and mAro (lower blot). Upper blot, lane 1: ERα WB of mERα IP; lane 2, ERα WB of mERα IP obtained from the supernatant of the initial mERα IP. The near absence of signal indicates quantitative removal of mERα by the initial mERα IP; one of three replications. Lower blot, lane 1: Aro WB of mAro IP; lane 2, Aro WB of Aro IP obtained from the supernatant of the initial mAro IP. The absence of signal indicates quantitative removal of mAro by the initial mAro IP; one of three replications. (B) Equivalent amounts of total solubilized membrane protein obtained from spinal cords of proestrous and diestrous rats were quantitatively immunoprecipitated using the anti-ERα antibody. The resulting supernatant (devoid of mERα) was immunoprecipitated using the anti-Aro antibody. Immunoprecipitates from each were resolved on SDS-PAGE and Western blotted for Aro. The graph shows the quantification of mAro coimmunoprecipitating with mERα or remaining in the supernatant following quantitative removal of mERα. Data represent mean percentage (+SEM) of total (calculated by equating the sum of free and associated mAro chemiluminescent intensity to 100%). Approximately 98% of spinal mAro was associated with mERα, without differences between proestrus and diestrus (n = 7 rats per group). ^#P < 0.05 for comparison of mERα-associated vs free mAro during proestrus, as well as diestrus. (C) Equivalent amounts of total solubilized membrane protein obtained from spinal cords of proestrous and diestrous rats were quantitatively immunoprecipitated using the anti-Aro antibody. The resulting supernatant (devoid of mAro) was immunoprecipitated using the anti-ERα antibody. Immunoprecipitates from both were resolved on SDS-PAGE and Western blotted for ERα. Quantification of mERα coimmunoprecipitating with mAro and that remaining in the supernatant following quantitative removal of mAro is shown in the bar graph (mean chemiluminescence + SEM). There was a small but significant increase in free mERα during proestrus (∼22% for ∼50 kDa; ∼21% for ∼32 kDa; n = 7 rats per group). *P < 0.05 for comparisons between proestrus and diestrus. D, diestrus; P, proestrus.

Approximately 98% of spinal mAro was found complexed with mERα. This preponderance of complexed vs free mAro did not differ between diestrous and proestrous rats [t (12) = 0.729, P > 0.05, n = 7]. The near absence of free spinal mAro suggests that spinally produced estrogens act predominantly, if not exclusively, within the spinal cord, signaling in an oligocrine fashion (1).

Free mERα in spinal cord expands during proestrus

To assess the influence of estrous-cycle stage on the physical association of spinal mERα with mAro, we quantified the mERα that coimmunoprecipitated with mAro, as well as that remaining in the supernatant following quantitative IP of mAro during proestrus and diestrus (Fig. 3C). The supernatant from the immunoprecipitate obtained using the anti-Aro antibody (first IP, complexed mERα) was immunoprecipitated using the anti-ERα antibody (second IP, free mERα), and both immunoprecipitates were Western blotted for ERα.

Two molecular mass species of mERα, ∼50 and ∼32 kDa, coimmunoprecipitated with mAro. This represented ∼62% of the ∼50-kDa species and ∼82% of the ∼32-kDa species. The levels of spinal mERα complexed with mAro did not vary between diestrus and proestrus [t (12) = −1.379, −1.134 for ∼50 and ∼32 kDa, respectively, P > 0.05, n = 7]. However, there was a significant increase in the pool of free mERα during proestrus [∼50 kDa: ∼22%; t (8.3) = 2.7, P = 0.03; ∼32 kDa: ∼21%; t (12) = 3.985, P = 0.002, n = 7]. As the pool of mAro-associated mERα did not change, the increase in free mERα (that could enhance responsiveness to synaptically derived or circulating estrogens) most likely did not result from a redistribution of mERα between its subpopulations.

Spinal mERα activity does not vary between diestrus and proestrus

ERα is phosphorylated at Ser 118 in response to estrogen binding, which can be used to reflect its activation status (49). Using p-ERα as a marker of mERα activity, we assessed if the overall mERα activity in the spinal membranes is altered with the change in circulating estrogen levels (Fig. 4A). Two major bands of p-mERα were visualized: ∼50 and ∼32 kDa. No significant differences were detected between p-mERα obtained from spinal membranes during proestrus and diestrus [t (12) = 1.139, −0.990 for 50 and 32 kDa, respectively, P > 0.05 for both, n = 7].

Figure 4.

Levels of p-mERα (activated mERα) in spinal cord during proestrus and diestrus. (A) Equivalent amounts of total solubilized membrane protein obtained from spinal cords of proestrous and diestrous rats were Western blotted using anti-p-ERα antibody. No significant differences were observed in p-mERα levels (shown as mean chemiluminescence + SEM; n = 7 rats per group). (B) Equivalent amounts of solubilized membrane protein obtained from spinal cords of proestrous and diestrous rats were quantitatively immunoprecipitated using the anti-Aro antibody. The resulting supernatant (devoid of mAro) was immunoprecipitated using a pan–anti-ERα antibody (which recognizes p-ERα as well as non-p-ERα). This ERα immunoprecipitate was resolved on SDS-PAGE and Western blotted for p-mERα. No significant differences in levels of mAro-free p-mERα (shown as mean chemiluminescence + SEM) were observed in spinal cords obtained from proestrous vs diestrous rats (n = 7 rats per group). Quantification of p-mERα, coimmunoprecipitating with mAro from the spinal cord of proestrous and diestrous rats, is not shown, as these data replicate earlier findings that spinal mAro-associated p-mERα significantly increases during diestrus compared with proestrus (1). D, diestrus; P, proestrus.

Differential activation of oligomerized and free spinal mERα

We quantified spinal mAro-oligomerized and free p-mERα to compare the state of activation of mERα in these subpopulations. In these experiments, the p-mERα oligomerized with mAro was quantified but is not shown, as these data replicated findings published previously (1). We found that the majority of spinal p-mERα (∼89% of ∼50-kDa species and ∼67% of 32-kDa species) was not associated with mAro, irrespective of stage of estrous cycle [∼50 kDa: t (12) = 1.800, P > 0.05; 32 kDa: t (12) = 1.265, P > 0.05, n = 7; Fig. 4B]. This suggests that in spinal cord, the majority of mERα activation results from the action of endocrine and/or synaptocrine estrogens. Because the increase in free mERα during proestrus (see above) was not accompanied by an increase in its activation, free mERα is likely not targeted by circulating estrogens (under normal physiological conditions) but instead, may function on demand.

In contrast to free p-mERα, the amount of mAro-associated p-mERα (which reflects the activity of the associated Aro) is significantly higher in diestrus than proestrus (1) (data not shown). The observed increase in mAro-associated p-mERα during diestrus underscores that circulating levels of estrogens can be inversely related to mERα signaling; i.e., ovarian and CNS synthesis of estrogens can be independently regulated.

Hypothalamic protein levels of mAro and mERα do not vary between diestrus and proestrus

We extended our findings in the spinal cord to the hypothalamus, which subserves endocrine as well as neural functionalities. There was no significant difference in mAro protein in the crude membrane fraction, obtained from hypothalami of proestrous and diestrous rats [t (5.12) = 1.137, P > 0.05, n = 5 to 7; Fig. 5A]. Likewise, there were no differences in any of the quantified mERα molecular mass species during diestrus and proestrus [t (6) = 2.106, −0.017, −1.221 for 65, 50, and 45 kDa, respectively, P > 0.05 for all, n = 4; Fig. 5B]. (As in spinal cord, the ∼40- and ∼32-kDa mERα species yielded a low-intensity signal, precluding their reliable quantification.)

Figure 5.

Hypothalamic protein levels of mAro and mERα do not vary between proestrus and diestrus. Equivalent amounts of total solubilized membrane protein obtained from hypothalami of proestrous and diestrous rats were resolved on SDS-PAGE and Western blotted for Aro and ERα. No significant differences were found in levels (shown as mean chemiluminescence + SEM) of either (A) mAro or (B) mERα (n = 4 rats per group). (The weak signal corresponding to the ∼52-kDa mAro seen in spinal cord was not visible in the hypothalamus.) (B) We did not quantify the ∼40- and ∼32-kDa mERα species, given their weak Western signals. D, diestrus; P, proestrus.

Predominance of free mAro in the hypothalamus

Although virtually all of the mAro coimmunoprecipitated with mERα in the spinal cord, the preponderance of mAro in the hypothalamus was not oligomerized with ERα; only ∼15% of the mAro coimmunoprecipitated with mERα (Fig. 6). This distribution of hypothalamic mAro between its mERα-associated and nonassociated subpopulations did not change with estrous cycle [t (6) = −0.156 and −0.085 for associated and free mAro, respectively, P > 0.05 for both, n = 4]. The vast preponderance of nonoligomerized mAro in the hypothalamus could suggest that hypothalamic-derived estrogens mostly act outside of the hypothalamus.

Figure 6.

Hypothalamic mERα-associated and free populations of mAro. Equivalent amounts of total solubilized membrane protein obtained from hypothalami of proestrous and diestrous rats were quantitatively immunoprecipitated using the anti-ERα antibody. The resulting supernatants of these immunoprecipitates (devoid of mERα) were immunoprecipitated using the anti-Aro antibody. Both immunoprecipitates were resolved on SDS-PAGE and Western blotted for Aro. In striking contrast to the spinal cord (see Fig. 3), only ∼15% of mAro was associated with mERα in the hypothalamus. As in the spinal cord, no significant differences were observed between proestrus and diestrus. Data represent mean percentage (+SEM) of total mAro (calculated by equating the sum of free and associated mAro chemiluminescent intensity to 100%; n = 4 rats per group). ^#P < 0.05 for comparison of mERα-associated vs free mAro in proestrus as well as diestrus. D, diestrus; P, proestrus.

Aro-containing hypothalamic neurons project outside the blood-brain barrier

The finding that the majority of hypothalamic mAro was independent of mERα suggested that hypothalamic estrogens were likely to be secreted. Accordingly, we used the retrograde tract-tracer FG (injected intravenously) to identify neurons that project their axons outside of the blood-brain barrier. We then used IHC to determine if these neurons also expressed Aro. Retrogradely labeled neurons were observed in the following nuclei within the hypothalamus: paraventricular (PVN), periventricular (PeVN), supraoptic (SON), and arcuate (ArcN) (Fig. 7). Labeling was not observed among neurons not projecting axons outside of the blood-brain barrier, e.g., cerebral cortex or basal ganglia (data not shown).

Figure 7.

Hypothalamic and forebrain neurons that project axons outside of the blood-brain barrier frequently express Aro. (A) PVN (lateral to the dorsalmost part of the third ventricle) and PeVN (immediately surrounding the third ventricle). (B) SON. (C) ArcN. (D) Median eminence (MedEm). (E) Subfornical organ (SFO). In each case, columns 1 and 2 show lower-magnification images of FG or Aro-ir, respectively. The boxed outlines in columns 1 and 2 are shown at higher magnification in columns 3 and 4. (A, B, C, and E) The arrows point out cells that are double-labeled for FG and Aro. The expression of Aro in PVN, SON, and ArcN suggests that estrogens synthesized by hypothalamic neurons are directed to the pituitary. Consistent with that hypothesis, Aro-immunoreactive fibers are found in the MedEm [arrows (D)]. Expression of Aro in retrogradely labeled cells in the SFO (E) suggests that SFO releases estrogen directly into the systemic circulation. 3v, third ventricle; OT, optic tract.

Most retrogradely labeled neurons in or near the diencephalon manifested Aro immunoreactivity (ir). These were found in regions projecting axons to the posterior pituitary and in regions projecting axons to the hypophyseal portal system. Among the latter, double-labeled neurons were found in the PeVN (outlining the dorsal half of the third ventricle), PVN (also projecting axons to the posterior pituitary), and ArcN (Fig. 7A and 7C). Among nuclei that innervate the posterior pituitary, SON and PVN were double labeled with FG and anti-Aro antibody (Fig. 7A and 7B). Interestingly, although many Aro-ir neurons in these regions were not labeled with FG, the vast majority of FG retrogradely labeled neurons expressed Aro. These findings suggest that at least a portion of hypothalamic estrogens can be released into the circulation and act in an endocrine fashion.

Because neurons double-labeled with FG and anti-Aro antibody were present in PVN and ArcN, which project to the median eminence (MedEm), it was not surprising that the MedEm contained varicose fibers immunoreactive for Aro (Fig. 7D). The MedEm is the entry point of hypothalamic hormones into the hypophyseal portal system. Therefore, hypothalamic estrogens are well positioned to regulate anterior pituitary function.

Interestingly, in addition to hypothalamic neurons, neurons in the subfornical organ (SFO) were also frequently double labeled with FG and anti-Aro antibody (Fig. 7E). The SFO is one of the circumventricular organs—brain regions that are highly vascularized and lack a blood-brain barrier. This observation, together with our observations in the hypothalamus, suggests that the brain is an endocrine organ that secretes estrogens.

Association of hypothalamic mERα with mAro is augmented in proestrus vs diestrus

As in spinal cord, two molecular mass species of mERα, ∼50 and ∼32 kDa, coimmunoprecipitated with hypothalamic mAro (Fig. 8). During both proestrus and diestrus, the ∼50-kDa mERα species was roughly equally distributed between mAro-free and mAro-complexed populations. In contrast, in diestrus, more of the ∼32-kDa species (∼61%) was complexed with mAro than was free. Furthermore, hypothalamic levels of the mAro-associated, ∼32-kDa mERα species were stage-of-cycle dependent; i.e., there was ∼73% more during proestrus vs diestrus [t (6) = 3.204, P = 0.019, n = 4; Fig. 8]. No significant changes were observed in the ∼50-kDa mERα species [t (6) = 1.347 and −0.552 for associated and free, respectively, P > 0.05 for both, n = 4]. The increase in mERα, physically associated with mAro in proestrus vs diestrus, suggests that during proestrus, estrogens synthesized via oligomerized mAro (although remaining a minority of hypothalamic Aro) are increasingly relevant to intrahypothalamic estrogenic signaling.

Figure 8.

Hypothalamic mAro-associated and mAro-free populations of mERα. Equivalent amounts of total solubilized membrane protein obtained from hypothalami of proestrous and diestrous rats were quantitatively immunoprecipitated using anti-Aro antibody. The resulting supernatants of these immunoprecipitates (devoid of mAro) were immunoprecipitated using anti-ERα antibody. Both immunoprecipitates were resolved on SDS-PAGE and Western blotted for ERα. Data represent mean chemiluminescence + SEM. Free mERα was unchanged between proestrus and diestrus. In contrast, the associated ∼32-kDa mERα species significantly increased (∼73%) during proestrus vs diestrus (n = 4 rats per group). *P < 0.05. D, diestrus; P, proestrus.

Interestingly, a complementary decrease in free ∼32-kDa mERα was not observed [t (4.232) = 1.908, P > 0.05, n = 4]. This absence of reciprocal changes in mERα subpopulations indicates that the increment in complexed ∼32-kDa mERα, observed during proestrus, did not result from a redistribution between mERα subpopulations. The source(s) of the increased mERα complexed with mAro remain to be determined.

Subpopulations of free and mAro-associated hypothalamic mERα are differentially activated during diestrus and proestrus

Whereas in the spinal cord, the majority of p-mERα was not physically associated with mAro, the majority of hypothalamic p-mERα was associated (Fig. 9; ∼69% for ∼50-kDa and ∼80% for ∼32-kDa species), irrespective of estrous-cycle stage [∼50 kDa: t (6) = 0.222, P > 0.05; 32 kDa: t (6) = −1.530, P > 0.05, n = 4]. In the aggregate, these results indicate that the majority of hypothalamic mERα signaling results from estrogens synthesized by mAro physically associated with mERα, i.e., oligocrine estrogens. Thus, whereas only a minority (∼15%) of hypothalamic mAro is oligomerized with mERα, oligocrine estrogens are likely to be the primary source of the estrogens activating hypothalamic mERα under normal physiological conditions.

Figure 9.

Distribution of activated hypothalamic mERα (p-mERα) between mAro-associated and mAro-free populations during proestrus and diestrus. Equivalent amounts of total solubilized membrane protein obtained from hypothalami of proestrous and diestrous rats were subjected to quantitative Aro IP. The resulting supernatant (devoid of mAro) was then immunoprecipitated using a pan anti-ERα antibody (which recognizes p-ERα, as well as non-p-ERα). Immunoprecipitates from both IPs were resolved on SDS-PAGE and Western blotted for p-ERα. The p-mERα WB signals from each immunoprecipitated fraction were quantified and compared. No significant differences were observed for either the ∼50- or ∼32-kDa p-mERα molecular mass species between proestrus and diestrus. Data represent mean percentage (+SEM) of total (calculated by equating the sum of free and associated p-mERα chemiluminescent intensity to 100%; n = 4 rats per group). ^#P < 0.05 for associated p-mERα vs free for both molecular mass species during proestrus as well as diestrus. D, diestrus; P, proestrus.

Catalytic activity of hypothalamic mAro is dependent on stage of estrous cycle

Because only a minor fraction (∼15%) of hypothalamic mAro is linked with mERα (in contrast to essentially all in the spinal cord), we were concerned that the p-mERα that coimmunoprecipitated with mAro, although reflecting the activation state of mERα, would not accurately reflect the aggregate activity of hypothalamic mAro. Accordingly, we quantified total hypothalamic Aro activity using the tritiated water release assay (Fig. 10). Hypothalami from proestrous rats showed ∼40% more Aro activity than hypothalami from diestrous rats [188.0 ± 10.7 fmol vs 104.5 ± 5.7 fmol per hypothalamus of proestrous vs diestrous rats; t (7) = 6.394, P < 0.001, n = 4 to 5].

Figure 10.

Influence of circulating estrogens on Aro activity in hypothalami of female rats. Aro activity was quantified using the classical tritiated water release assay in hypothalami obtained from proestrous (n = 5), diestrous (n = 4), and Ovx (n = 6) rats; Ovx + E2 rats (n = 4); diestrous rats with E2 supplementation (D + E2; n = 4); and diestrous rats treated with sesame oil vehicle (D + veh; n = 3). Aro activity was significantly less than that in proestrous rats, in all cases except diestrous rats treated with E2. *P < 0.05 vs P; ^#P < 0.05 vs D + veh. These findings demonstrate the ability of circulating estrogens to stimulate hypothalamic Aro activity. Increased hypothalamic Aro activity following E2 supplementation of intact diestrous rats, but not Ovx rats, underscores the requirement for an ovarian factor(s) in addition to estrogen (e.g., androgens, activin, inhibin) and/or an intact hypothalamic-pituitary-gonadal axis. Data are shown as mean femtomoles (+SEM) of estrone produced per whole hypothalamus (∼60 mg tissue). D, diestrus; P, proestrus.

Elevated levels of circulating estrogens stimulate hypothalamic mAro activity

Parallelism between the catalytic activity of hypothalamic Aro and circulating levels of estrogens suggests their causal association. To test this inference, we investigated the effects of manipulation of circulating estrogens on hypothalamic Aro activity (Fig. 10). Manipulations included the following: (1) ovariectomy alone and (2) ovariectomy in combination with estrogen replacement to achieve proestrous levels of circulating estrogen. To explore possible contributions of estrogens when acting in concert with other ovarian factors (e.g., androgens, activin, inhibin), we (3) supplemented circulating estrogen levels of intact diestrous rats to proestrous levels.

We found significant differences [F(5,20) = 9.86, P < 0.001; one-way ANOVA] in Aro activity among hypothalami obtained from (1) diestrous rats (104.5 ± 5.7 fmol, n = 4), (2) proestrous rats (188.0 ± 10.6 fmol, n = 5), (3) Ovx rats (121.6 ± 3.5 fmol, n = 6), (4) Ovx rats receiving E2 supplementation (Ovx + E2: 132.1 ± 15.2 fmol, n = 4); (5) diestrous rats receiving E2 supplementation (185.4 ± 20.1 fmol, n = 4), and (6) diestrous rats receiving vehicle (D + veh; 103.0 ± 18.3 fmol, n = 3; Fig. 10). Planned comparisons using Tukey post hoc test revealed the following: (1) hypothalamic mAro activity was significantly greater during proestrus than diestrus (P = 0.001), consistent with results of their initial comparison by t test; (2) Aro activity in hypothalami obtained from Ovx rats did not significantly differ from those obtained during diestrus (P > 0.05) but did significantly differ from hypothalami obtained during proestrus (P = 0.002); and (3) although proestrous levels of circulating estrogen in Ovx rats failed to stimulate hypothalamic Aro activity (P > 0.05 for Ovx + E2 vs Ovx), (4) the same estrogen supplementation in intact diestrous rats did stimulate hypothalamic Aro activity to levels characteristic of physiological proestrus (P > 0.05 for proestrus vs simulated proestrus in intact rats). The sesame oil vehicle had no effect (P > 0.05 for diestrus vs D + veh). These findings indicate that circulating estrogens are indeed critical, positive modulators of hypothalamic Aro activity, but their ability to do so requires a complementary ovarian factor(s) and/or an intact hypothalamic-pituitary-gonadal axis.

Disconnect between Aro activity and corresponding protein levels

The increment in hypothalamic Aro activity during proestrus vs diestrus was not paralleled by any increase in Aro protein (Fig. 11A and 11B). This disconnect was also evidenced by the barely detectable letrozole/formestane-inhibitable Aro activity in rat spinal cord using the same tritiated water assay, despite the fact that rat spinal cord contained ∼4× more mAro protein than did the hypothalamus (obtained from the same proestrous animal, processed, and Western blotted in parallel [t (2) = 4.667, P = 0.043, n = 3]; Fig. 11B). Moreover, hypothalami of males contained ∼60% more mAro protein but ∼300% more activity than did hypothalami of proestrous females [mAro protein: t (8) = −4.720, P = 0.002 for male vs proestrous rats, n = 3 and 7, respectively; Aro activity: 544.5 ± 54.1 fmol vs 188.0 ± 10.6 fmol per hypothalamus of male vs proestrous rats, respectively, t (5.4) = −6.469, P = 0.001, n = 5 to 6; Fig. 11A and 11B]. The striking disparity between mAro protein and Aro activity in hypothalamus, as well as spinal cord, underscores the importance of nongenomic regulatory mechanisms in mammalian CNS to modulating Aro activity.

Figure 11.

Disconnect between changes in Aro enzyme activity and Aro protein levels. (A) Aro enzyme activity was quantified as described in the legend for Fig. 10. Aro activity [represented as mean femtomoles (+SEM) of estrone produced] in hypothalami of male rats (M) was approximately threefold and fivefold higher than in hypothalami of rats in proestrus and diestrus, respectively (P and D reproduced from Fig. 10 to facilitate comparison with M); Aro activity in the spinal cord (SC) of proestrous rats was virtually undetectable. *P < 0.05 when comparing hypothalamic Aro activity of males (n = 6) with that of proestrous (n = 5) and diestrous (n = 4) females or the SC of proestrous females (n = 3). ^#P < 0.05 for comparison of Aro activity in SC vs hypothalami. (B) Equivalent amounts of total solubilized membrane protein obtained from SCs of proestrous rats or hypothalami of proestrous, diestrous, and male rats were resolved on SDS-PAGE and Western blotted for Aro. Hypothalamic and SC protein levels of mAro are shown relative to the mean (+SEM) mAro protein levels in hypothalami of proestrous rats (set to 100%; n = 7). Despite the approximately threefold and fivefold higher Aro activity in hypothalami of males vs proestrous or diestrous females, respectively, Aro protein was only ∼60% higher in males (n = 3). *P < 0.05 vs proestrous. Even more striking, Aro activity in SC (n = 3) was essentially undetectable, notwithstanding that spinal levels of mAro protein were more than fourfold higher than those found in hypothalamus (^#P < 0.05).

The Aro activity assay was performed on tissue homogenates, whereas Aro protein was quantified using solubilized crude plasma membrane fractions. We eliminated the possibility that the observed disparity between changes in Aro activity and Aro protein results from the presence of microsomal/cytoplasmic Aro in its activity assay but not its protein quantification. This is evidenced by demonstrating (1) the plasma membrane fraction contained ∼85% of the total Aro activity (n = 2; data not shown), and (2) no significant differences were observed in hypothalamic microsomal/cytoplasmic Aro protein levels between diestrus and proestrus [t (4) = 1.952, P > 0.05, n = 3; data not shown].

Discussion

This study tested the hypothesis that the extent of oligomerization of mAro and mERα differs between CNS regions with different functions. We found that the relationship between mAro and mERα in the spinal cord and hypothalamus had strikingly different quantitative aspects. Main findings included the following: (1) mAro and mERα physically associated in both spinal cord and hypothalamus; (2) the distribution between mERα-associated and free mAro was essentially opposite in these regions: in the spinal cord, essentially all of the mAro was oligomerized with mERα, whereas in the hypothalamus, ∼85% of the mAro was free; (3) in the hypothalamus, many Aro-expressing neurons were retrogradely labeled by intravenous FG, indicating that hypothalamically produced estrogens are likely released; (4) in the spinal cord, the preponderance of activated mERα (p-mERα) was not physically associated with mAro, whereas in the hypothalamus, the opposite was true; (5) hypothalamic Aro activity varied in synchrony with changes in circulating levels of estrogens over the estrous cycle; and (6) there was a striking disconnect between relative tissue levels of mAro protein and Aro activity.

Given their lipid solubility, ovarian estrogens might be expected to cross the blood-brain barrier and activate CNS mERs indiscriminately, trivializing the physiological significance of locally synthesized estrogens. However, peripheral estrogens should not be expected to have unrestricted access to all CNS ERs because the ability of peripheral estrogens to activate them would be influenced by the activity and distribution of estrogen-metabolizing enzymes (63–68), estrogen-binding proteins, etc. Moreover, (1) in certain brain areas, concentrations of estrogens exceed those measured in the circulation (54, 69, 70), and (2) some CNS estrogenic signaling, reflected by p-mERα, is out of phase with circulating concentrations of estrogens (1), neither of which would be expected if peripheral estrogens had unrestricted access to CNS ERs. The current observation that overall spinal p-mERα did not change over the estrous cycle also underscores that circulating estrogens may have limited relevance to CNS mERα signaling.

Estrogenic signaling strategies include the following: endocrine, synaptocrine (47), and oligocrine, in which estrogens function as intracellular messengers that are synthesized and act within the same macromolecular signaling complex (1) (restricting loci of action). The ability of spinal and hypothalamic mAro and mERα to pair physically allows for exceedingly rapid and spatially precise regulation of CNS estrogenic signaling on a subcellular level. Importantly, such local production of oligocrine estrogens would result in much higher concentrations of estrogens in the immediate vicinity of mERs than that achieved by diffusion from the periphery. This would fulfill the requirement for higher concentrations of estrogens to activate mERs compared with nuclear ERs (69, 71).

The organization of mERs and mAro is likely to determine the specific functions of locally synthesized estrogens and the mechanisms by which they act. In the spinal cord, virtually all mAro was physically associated with mERα. This organization could be compatible with the synaptocrine mode of action. However, it would require that oligocrine estrogens not diffuse beyond the multimeric complex in which they were generated, the occurrence of which remains to be determined. This consideration suggests that spinally synthesized estrogens most likely signal predominantly in an oligocrine fashion, which allows for the selective activation/deactivation of different populations of mAro-associated mERs, as well as their exquisite subcellular spatial segregation. Nevertheless, as the majority of spinal p-mERα (activated mERα) was not associated with mAro, the preponderance of spinal mERα signaling likely results from the actions of endocrine and/or synaptocrine estrogens [the latter likely originating from endoplasmic reticulum Aro (39), which was not investigated in this study].

Most of hypothalamic mAro was found not to be associated with mERα, suggesting that overall, estrogens synthesized in the hypothalamus have broader loci of action than those synthesized in the spinal cord. Indeed, a considerable portion of the hypothalamic Aro borders on the third ventricle (35), suggesting that hypothalamic-derived estrogens are secreted into the third ventricle, thus contributing to cerebrospinal fluid estrogens (72). Moreover, the ability of E2 implanted into the MedEm to enhance the release of pituitary luteinizing hormone from the anterior pituitary (73) supports the likely physiological relevance of hypothalamic-derived estrogens acting outside of the hypothalamus. Secretion of hypothalamic-derived estrogens into the portal circulation, their transport to the anterior pituitary, and activation of ERα, which is robustly expressed therein (74–76), would represent a dimension of hypothalamic neuroendocrine action that has substantial implications for reproductive physiology.

Our observations of Aro-ir in neurons that are retrogradely labeled with systemically injected FG suggest that the endocrine function of the brain is not limited to the synthesis and release of peptide hormones but includes release of sex steroids as well. The presence of Aro in FG-labeled neurons of the SFO underscores the brain’s ability to function as an endocrine organ, suggesting that it contributes to circulating estrogens. Although the magnitude of the CNS contribution to circulating estrogens remains to be defined, as do the physiologic conditions in which such contributions are critical, it is interesting to note that approximately one-half of circulating levels of estrogens persists following ovariectomy (77). Moreover, our observations of Aro-expressing neurons in the hypothalamic nuclei, retrogradely labeled with FG, suggest that brain-derived estrogens have the potential to alter hormonal release from both the anterior and posterior pituitary. Consistent with our observations, previous studies have reported modulation of anterior pituitary function by circulating estrogens (78, 79).

Populations of mERα that are mAro associated and free would be expected to have different functions: mERα physically associated with mAro would likely subserve intracellular, oligocrine estrogenic signaling, whereas free mERα would likely mediate signaling in response to synaptic and/or endocrine estrogens. Notably, during proestrus, the ∼32-kDa mERα, physically associated with mAro, expanded in the hypothalamus, suggesting increased physiological importance of oligocrine estrogens at this time (notwithstanding the preponderance of free hypothalamic mAro and the implied dominance of synaptocrine and/or endocrine estrogenic signaling). In contrast, in the spinal cord of proestrous rats, the mAro-free population of mERα was augmented, suggesting increased responsiveness to circulating estrogens (notwithstanding the vast preponderance of mAro that is associated with mERα and thus, the implied virtually exclusive use of oligocrine estrogenic signaling by spinally synthesized estrogens). In the aggregate, our data indicate that locally synthesized estrogens are used differently across CNS regions and that mERα subpopulations can be regulated independently in a tissue- and estrous-cycle stage-specific manner.

Several current observations indicate that the relationship between Aro protein and corresponding activity is strikingly dichotomous. Although the spinal cord contains substantially (∼4×) more mAro protein per milligram of total protein than does the hypothalamus, spinal Aro activity remains essentially undetectable. Moreover, although total Aro activity in hypothalami of males is almost threefold higher than that of proestrous females, the corresponding increment in protein is only ∼60%. An analogous disconnect occurs between changes in Aro activity and Aro protein in hypothalami of proestrous vs diestrous rats, where the former is augmented by ∼40% in the absence of any corresponding increase in Aro protein. This dichotomy likely indicates the importance of acute covalent modulation of CNS Aro activity. Aro phosphorylation has been demonstrated in the CNS of birds (quail) (43, 52, 53) and in human breast cancer cells (80) (having inverse effects on Aro activity). Notably, the discrepancy we observed between the magnitude of change in Aro protein and Aro activity is an indication that analogous acute regulation of Aro activity also occurs in mammalian CNS.

Our finding that supplementation of E2 in intact diestrous rats elevates hypothalamic Aro activity to that seen during proestrus indicates that the proestrus-associated increment in circulating estrogens positively modulates hypothalamic Aro activity. The inability of estrogen supplementation to elevate hypothalamic Aro activity in Ovx animals indicates that whereas elevated peripheral estrogens are necessary for increased hypothalamic Aro activity during proestrus, they are not sufficient—an additional ovarian factor(s) is required. This requirement could be satisfied by ovarian androgens, in which secretion is also elevated during proestrus (81) and which have been reported to increase Aro activity (82). Alternatively, an intact hypothalamic-pituitary-gonadal axis could be necessary.

Interestingly, although ovariectomy profoundly reduced hypothalamic Aro activity, it was not abolished. This suggests that a portion of hypothalamic Aro activity is independent of peripheral estrogens (and/or other ovarian factors). The likelihood of this is underscored by the approximately threefold higher Aro activity in hypothalami of males than proestrous females, a difference clearly independent of circulating estrogens and which may be a result of circulating androgens (see above). Identification of Aro regulatory factors, in addition to estrogens, remains to be elucidated.

The current study, together with its immediate predecessor (1), delineates a novel dimension of CNS estrogenic signaling. The propensity of mAro either to physically pair with or exist separately from mERα creates two populations of each. This, in combination with a synaptic relationship between mAro and mERα, enables locally produced estrogens to act with a high degree of spatial selectivity (ranging from the subcellular to the synaptic), facilitating highly nuanced regulation and plasticity. Factors that coordinate the ability of estrogens to act at subcellular, synaptic, and CNS regional levels are likely to be prime integrative agents. A future challenge will be to associate the relative contributions of the various estrogenic signaling modalities with specific CNS functions. The identification of factors that influence the interaction between mAro and mERs and the development of pharmacophores that selectively target specific pools of mAro and/or mERs could be of high clinical importance.

Glossary

Abbreviations:

ArcN

arcuate nucleus of the hypothalamus

Aro

aromatase

CNS

central nervous system

D + veh

diestrous rat supplemented with vehicle for estradiol

E2

estradiol

ER

estrogen receptor

FG

Fluoro-Gold

IHC

immunohistochemistry

IP

immunoprecipitation

ir

immunoreactivity

mAro

membrane aromatase

MedEm

median eminence

mER

membrane estrogen receptor

Ovx

ovariectomized

Ovx + E2

ovariectomized and supplemented with estradiol

p-ERα

estrogen receptor α phosphorylated at Ser 118

p-mERα

membrane estrogen receptor α phosphorylated at Ser 118

PeVN

periventricular nucleus of the hypothalamus

PVN

paraventricular nucleus of the hypothalamus

SFO

subfornical organ

SON

supraoptic nucleus of the hypothalamus

WB

Western blot