Selective Nonnuclear Estrogen Receptor Activation Decreases Stroke Severity and Promotes Functional Recovery in Female Mice

Abstract

Estrogens provide neuroprotection in animal models of stroke, but uterotrophic effects and cancer risk limit translation. Classic estrogen receptors (ERs) serve as transcription factors, whereas nonnuclear ERs govern numerous cell processes and exert beneficial cardiometabolic effects without uterine or breast cancer growth in mice. Here, we determined how nonnuclear ER stimulation with pathway-preferential estrogen (PaPE)-1 affects stroke outcome in mice. Ovariectomized female mice received vehicle, estradiol (E₂), or PaPE-1 before and after transient middle cerebral artery occlusion (tMCAo). Lesion severity was assessed with MRI, and poststroke motor function was evaluated through 2 weeks after tMCAo. Circulating, spleen, and brain leukocyte subpopulations were quantified 3 days after tMCAo by flow cytometry, and neurogenesis and angiogenesis were evaluated histologically 2 weeks after tMCAo. Compared with vehicle, E₂ and PaPE-1 reduced infarct volumes at 3 days after tMCAo, though only PaPE-1 reduced leukocyte infiltration into the ischemic brain. Unlike E₂, PaPE-1 had no uterotrophic effect. Both interventions had negligible effect on long-term poststroke neuronal or vascular plasticity. All mice displayed a decline in motor performance at 2 days after tMCAo, and vehicle-treated mice did not improve thereafter. In contrast, E₂ and PaPE-1 treatment afforded functional recovery at 6 days after tMCAo and beyond. Thus, the selective activation of nonnuclear ER by PaPE-1 decreased stroke severity and improved functional recovery in mice without undesirable uterotrophic effects. The beneficial effects of PaPE-1 are also associated with attenuated neuroinflammation in the brain. PaPE-1 and similar molecules may warrant consideration as efficacious ER modulators providing neuroprotection without detrimental effects on the uterus or cancer risk.

Premenopausal women have a lower risk of severe cerebrovascular events than similar age groups of men (1, 2), and women who have undergone surgical menopause have a higher risk of stroke than women who experience natural menopause (3). In a similar manner, young adult ovariectomized female rodents exhibit worse stroke outcomes than intact controls (4). Numerous preclinical studies demonstrated that 17β-estradiol (E₂) affords neuroprotection in the setting of stroke (5–7), neurodegenerative diseases (8, 9), and spinal cord injury (10). However, E₂-mediated neuroprotection has not been observed in clinical trials. Results from the Women’s Estrogen for Stroke Trial showed that stroke recurrence was similar between E₂-treated and placebo-treated women, and stroke severity and mortality were actually worse in the E₂ group (11–13). Moreover, in the Women’s Health Initiative Estrogen-Alone trial, which randomly assigned almost 11,000 women aged 50 to 79 years to daily conjugated equine estrogen vs placebo, the hazard ratio (95% CI) for ischemic stroke for conjugated equine estrogen vs placebo was 1.55 (1.19 to 2.01) over an average follow-up period of 7.1 years, and there was no difference in hemorrhagic stroke (14). Given the inconsistencies in preclinical and clinical observations, it is vital to understand in greater detail which mechanisms of action of estrogens potentially provide neuroprotection (15).

The biological effects of estrogens are mediated largely by estrogen receptors (ERs), and there is a complex and diverse distribution of ERs in the central nervous system (CNS) (16). The CNS cell types that express ERs include oligodendrocytes, endothelial cells, microglia, neurons, astrocytes, and vascular smooth muscle cells (16–19). The two original ER subtypes, ERα and ERβ, are members of the nuclear receptor superfamily (20), and they classically function as transcription factors. In addition, plasma membrane‒associated subpopulations of ERα and ERβ are coupled to kinase cascades and initiate nonnuclear intracellular signaling (21–24). Membrane-associated ERα has been identified in the hypothalamus, with coupling to metabotropic glutamate receptors via their colocalization in caveolae, with rapid nonnuclear signaling by ERβ also observed in the CNS (25). The plasma membrane‒spanning receptor G protein‒coupled estrogen receptor-1 (GPER-1), also known as G protein‒coupled receptor 30, is additionally expressed in the CNS, and rapid estrogen-mediated actions associated with GPER-1 have been measured in hypothalamic and hippocampal neurons (19). The processes regulated by estrogens and ERs in the CNS, through both nuclear and nonnuclear actions, include locomotor coordination (26, 27), cognitive function (28), synaptic plasticity, memory and learning, and modulation of social behaviors (29–31).

In early studies of nonnuclear ER function, E₂ was conjugated to BSA (E₂-BSA) in an attempt to provide an extracellular ER ligand (32). Since then, two types of molecules have been created to selectively activate nonnuclear ER actions. The first, the estrogen dendrimer conjugate (EDC), is composed of a large poly(amido)amine (PAMAM) dendrimer onto which ∼20 ethinyl E₂ molecules are stably attached. EDC binds to ERs with an affinity similar to that of free E₂, but it is excluded from the nucleus (33, 34). EDC promotes endothelial monolayer repair, attenuates neointima formation, and reduces cardiac ischemia‒reperfusion injury in mice in the absence of uterotrophic responses or breast cancer promotion (15, 35–37). The second type of nonnuclear ER agonists are pathway-preferential estrogens (PaPEs), which are minimally modified estrogenic ligands that have greatly reduced binding affinity for ERs. PaPEs do not affect reproductive or mammary tissues or breast cancer cells, but they elicit favorable effects in the vasculature, liver, and adipose tissue (38). Regarding the clinical utility of EDC or PaPEs, because delivery of the large PAMAM dendrimer to the CNS is minimal (39), EDC is most likely excluded from CNS entry. By contrast, because the PaPEs resemble steroidal estrogens, they are likely effectively delivered to the brain upon systemic administration (40).

To better understand how estrogens influence the magnitude of stroke injury, the present project investigated the effect of selective nonnuclear ER activation on stroke severity and functional recovery in mice after transient middle cerebral artery occlusion (tMCAo). To ensure that the pharmacologic intervention could be administered systemically (i.e., so clinical utility could be considered), selective nonnuclear ER activation was accomplished with PaPE-1 (38). Because plasma membrane‒associated ERs in neurons mediate rapid neuroprotective and neurotrophic signaling cascades in response to E₂ (41) and nonnuclear ER activation with EDC reduces cardiac ischemia‒reperfusion injury (35), we tested the proof-of-concept hypothesis that nonnuclear ER activation with PaPE-1 provides protection from stroke injury and its sequelae. Recognizing that the infiltration of certain leukocyte subsets into the brain has deleterious effects on ischemic injury and recovery (42–44), we conducted additional studies of leukocyte recruitment to the CNS to determine the basis by which PaPE-1 modifies the acute and long-term outcomes of tMCAo.

Methods

Female C57BL/6 mice were obtained from the University of Texas (UT) Southwestern Mouse Breeding Core and maintained in standard animal housing with 12/12 light cycles (lights off at 6:00 pm), cob bedding, and food and water as needed. All procedures were approved by the UT Southwestern Institutional Animal Care and Use Committee. Mice were ovariectomized at 8 weeks of age and randomized to receive subcutaneous pellets containing vehicle, E₂ (0.075 mg), or PaPE-1 (8 mg) implanted at 10 weeks of age, as previously established for efficacy (38). Lesion severity was assessed by MRI, and poststroke motor function was evaluated through 2 weeks after tMCAo. One cohort of mice received rotarod training daily for 9 days to establish baseline motor coordination before tMCAo. These mice were euthanized at 14 days after stroke. A second cohort of mice randomly assigned to the same three study groups was euthanized 3 days after tMCAo to evaluate neuroinflammation. At euthanasia, the uterus was quickly isolated, a wet weight was obtained, and uterine weight was expressed relative to body weight. Neurogenesis and angiogenesis were quantified at 2 weeks after tMCAo by histology. All treatment groups were randomized, and experimenters were blinded to the data/conditions.

tMCAo

Surgeries were performed under anesthesia (2% isoflurane/70% NO₂/30% O₂), and body temperature was maintained at 37°C. Before and after the tMCAo procedure (45), cerebral blood flow (CBF) was quantified using transcranial laser Doppler flowmetry (TSI, Inc., Shoreview, MN) on the exposed left middle cerebral artery following previously published methods (45–48). The middle cerebral artery was occluded with blunted suture (6.0-gauge nylon, 12 mm) by surgeons blinded to the experimental groups, and a CBF reduction >80% was considered indicative of successful occlusion. The surgeries were performed between 0800 and 1400, and each surgeon operated on a maximum of 10 mice per day. The mice were placed in a 34°C incubator and then reanesthetized for removal of the occluding sutures 45 minutes after their insertion. After occluding suture removal, CBF values >50% of baseline preocclusion CBF were considered indicative of successful reperfusion. Buprenorphine and lidocaine were used as analgesics. In total, 72 mice initially underwent tMCAo surgeries for all cohorts. Five mice did not meet CBF occlusion criteria, and 14 mice did not meet CBF reperfusion criteria. Three mice died before the end of experimentation. These mice were excluded from further study. This includes n = 8 from the vehicle group, n = 7 from the E₂ group, and n = 7 from the PaPE-1 group.

Infarct volume quantification

MRI was used for quantification of infarct volume in vivo at 3 and/or 7 days after tMCAo (i.e., the 2-week cohort of mice had serial imaging on days 3 and 7). Mice that received serial imaging on days 3 and 7 after stroke included n = 9 from the vehicle group, n = 10 from the E₂ group, and n = 9 from the PaPE-1 group. Mice that received imaging only on day 3 after stroke included n = 7 from the vehicle group, n = 6 from the E₂ group, and n = 6 from the PaPE-1 group. A 7-Tesla small animal MRI system (Agilent Inc., Palo Alto, CA) with a 400 mT/m gradient coil set and a 40-mm millipede coil was used. Animals were placed supine and headfirst into the middle of the device according to the center of the radiofrequency coil, and they were attached to a respiratory sensor (SA Instruments, Stony Brook, NY). Respiratory rates were monitored throughout the imaging session. Parameters used for the imaging were slice thickness = 1 mm, matrix size = 256 × 256, repetition time/echo time = 2500/60 ms, eight averages, field of view = 25.6 × 25.6 mm, total scan time of 10 minutes and 45 seconds, and 100 µm in-plane resolution. T2-weighted high-resolution images were obtained axially on the entire brain (fast spin-echo). Fiji software (Image J; National Institutes of Health, Bethesda, MD) was used to determine hyperintense areas (with higher water content) in the images, such as expanded or dilated ventricles and edema, and they were outlined through manual segmentation. Using a slice thickness of 1 mm and 15 slices total, the quantified areas were used to calculate the infarct volume (45).

Behavioral assessment

Rotarod behavioral tests were used to quantify motor coordination. The apparatus (IITC Life Science, Woodland Hills, CA) used was composed of a 10-inch elevated metal rod 1.25 inches in diameter with five semiclosed lanes, and grip was enhanced by a fine textured finish. Mice were allowed to stabilize their posture on the stationary rod before initiating rotation. The rod rotation direction was toward the investigator so that the mice preferentially faced away from the investigator while walking. Parameters for rotarod testing were as follows: start speed of 4 rpm, acceleration rate of 0.2 rpm/s, top speed of 44 rpm, and maximum test duration of 300 seconds. Average times from four trials were taken per day for each mouse. The trial was considered complete when the mouse fell from the rod onto the padded surface and triggered an automatic fall-detection sensor. In cases in which the mouse gripped the rod and held onto the rod without moving, the fall-detection sensor was manually triggered after a complete rotation with the mouse stationary. Also, if the mouse fell within 2 seconds after the initial rotation of the rod, it was considered startled, and the data were excluded.

Neurogenesis and angiogenesis measurements

Animals were euthanized with isoflurane overdose and transcardially perfused with 4% paraformaldehyde as described previously (49, 50). Brains were cryoprotected in 30% sucrose with 0.1% sodium azide before processing for immunohistochemistry. Coronal sections 30 μm thick were collected at dry ice temperature on a microtome. Sections spanned the entire cortex and hippocampus in a 1:6 series for maintaining consistency. Sections were stained for alkaline phosphatase (ALP; SK-5400 Kit; Vector Laboratories, Burlingame, CA) to identify existing capillaries as previously described (51). All histology was imaged using whole slide imaging (Nanozoomer 2.0HT; Hamamatsu Photonics, Hamamatsu-shi, Japan). Total number of ALP-positive blood vessels were quantified around the infarct area in five distinct microscopic fields of size 300 by 400 μm in the ipsilateral hemisphere and corresponding areas in the contralateral hemispheres separately. Capillary density was reported at 14 days after tMCAo as a ratio between the blood vessel counts in the ipsilateral hemisphere and contralateral hemisphere.

Neurogenesis was quantified in the dentate gyrus from the granule cell layer, modified from Whoolery et al. (52). Hippocampal sections, which were mounted onto superfrost-plus slides (Fisher Scientific, Waltham, MA), were allowed to dry for 2 hours and were incubated in 0.01 M citric acid (pH, 6.0) at 100°C for 15 minutes for antigen retrieval. Sections were incubated in 0.3% H₂O₂ (to quench endogenous peroxidases), followed by incubation with 0.3% Triton X-100 in 1X PBS and 3% normal donkey serum for 1 hour (to block nonspecific binding). Sections were incubated with primary antibody, goat polyclonal anti-doublecortin (anti-DCX; 1:500; Santa Cruz Biotechnology, Dallas, TX) overnight, followed by incubation with biotin-conjugated secondary antibody for 2 hours and ABC [1:50; Vector Laboratories; RRID: AB_10610966 (53)] solution for 90 minutes. 3,3′-Diaminobenzidine/metal concentrate (10×; Thermo Scientific, Waltham, MA) was used to visualize DCX⁺ cells, and Nuclear Fast Red (Vector Laboratories) staining was used to visualize nuclei. Tissues from all mice were processed simultaneously through immunohistochemistry staining. Total DCX⁺ positive cells were counted in both ipsilateral and contralateral hemispheres separately, and data were represented as the number of DCX⁺ positive cells per section (total cells for sections per animal/number of sections).

Flow cytometry

Spleens and brains were processed into single-cell suspensions, and 10⁶ cells per sample were subjected to fluorescence-activated cell sorting (FACS) using the following fluorescent antibodies (45): leukocytes [CD45-APC-Cy7; RRID: AB_312981 (54)], B cells [CD19-PE; RRID: AB_2621752 (55)], monocytes/macrophages [CD11b-APC; RRID: AB_2621556 (56)], neutrophils [Ly6C Ly6G-FITC; RRID: AB_394643 (57)], natural killer (NK) cells [NK1.1-Percp Cy5.5; RRID: AB_2621910 (58)], CD4 T cells [CD4-V450; RRID: AB_2621927 (59)], and general T cells [TCRβ-BV510; RRID: AB_2562349 (60)]. CD8 T cells are CD45⁺TCRβ⁺CD4⁻CD19⁻ cells. The antibody mixture was added to the cells and incubated for 30 minutes at 4°C, and the cells were washed with FACS buffer and fixed in 1% paraformaldehyde; FACS was then performed with a BD FACS Canto instrument (BD Biosciences, San Jose, CA). FACS Diva 6.0 software was used for processing the cells and data acquisition, and the collected data were analyzed using FlowJo 9.0 software (Tree Star, Ashland, OR) and gated into different leukocyte subsets. The gates were set on the basis of fluorescence minus one control. Viable cells were gated according to their forward scatter and side scatter (45, 47).

Statistical analysis

The data sets were analyzed for normality, and parametric or nonparametric post hoc statistical analysis was used accordingly. Specifically, MRI (five of six groups), poststroke edema, and uterine weight data sets exhibited normal distribution. The findings for E₂ and PaPE-1 effects on MRI infarct size and edema were similar at 3 days after tMCAo in the first and second study designs. Therefore, we combined results from both cohorts at 3 days after tMCAo and used an ordinary one-way ANOVA with Fisher least significant difference for analysis. Experimental N was too small for normality verification of neurogenesis, angiogenesis, blood, spleen, and brain flow cytometry data sets. Thus, the nonparametric Kruskal-Wallis test with Dunn multiple comparisons was used for analysis. Rotarod data sets were analyzed using repeated-measures two-way ANOVA with Tukey multiple comparisons. Wilcoxon matched-pairs signed rank test was used for comparing the corresponding ipsilateral and contralateral data sets. Graph Pad Prism 7.0 software (La Jolla, CA) was used for analysis, and values with P < 0.05 were termed significant. The data are reported as mean ± SD. Because we had not previously performed a focal stroke experiment of this type with ovariectomized female mice, a post hoc analysis was used to confirm proper group powering at 80% power and α = 0.05.

Results

PaPE-1 decreased stroke lesion severity

The treatment protocol is outlined in Fig. 1A. Compared with vehicle treatment, E₂ predictably caused a 392% increase in uterine wet weight compared with both vehicle-treated and PaPE-1‒treated cohorts (P < 0.0001 vs both) (Fig. 1B). During the tMCAo procedure, there was no effect of treatment on baseline red blood cell flux or magnitude of reperfusion after occlusion. Compared with vehicle, both E₂ and PaPE-1 treatments reduced infarct volumes at 3 days after tMCAo (E₂: 36%, P < 0.05; PaPE-1: 66%, P < 0.001) and at 7 days after tMCAo (E₂: 50%; PaPE-1: 55%; both P < 0.05) (Fig. 2A and 2B). Correspondingly, both E₂-treated and PaPE-1‒treated mice exhibited reduced brain edema 3 days after stroke (E₂: 60%; PaPE-1: 78%; both P < 0.05) (Fig. 2C). Poststroke swelling resolved by 7 days in all treatment groups. Thus, PaPE-1 reduced infarct volumes and edema to degrees that were at least as great as those observed with E₂. Of note, the favorable effect on stroke lesion severity and edema with PaPE-1 occurred without the undesired effect on the reproductive tract that occurs with E₂ treatment.

Figure 1.

E₂ and not PaPE-1 increased uterine weight. (A) Schematic for treatment procedure: female mice were ovariectomized (Ovex) at 8 weeks; received vehicle, E₂, or PaPE-1 delivered by subcutaneous pellet both before and after tMCAo; and underwent rotarod training and serial MRI. (B) E₂-treated mice (green squares) had higher uterine weight than vehicle-treated (black circles) and PaPE-1‒treated mice (red triangles; n = 9 to 10 per group). F_(2,26) = 60.65. Values are shown as mean  ±  SD. ****P < 0.0001.

Figure 2.

E₂ and PaPE-1 decreased stroke lesion severity. (A) Representative T2-weighted images of vehicle-treated, E₂-treated, and PaPE-1‒treated mice at 3 and 7 days after stroke (n = 9 to 16 per group; individual animals are designated by m number). (B and C) Mice treated with E₂ (green squares) and PaPE-1 (red triangles) exhibited (B) smaller infarct volumes (day 3: F_(3,47) = 7.016; day 7: F_(3,28) = 3.78) and (C) reduced ipsilateral edema (day 3: F_(3,47) = 6.02; day 7: F_(3,28) = 0.01) compared with mice treated with vehicle (black circles). Scale bar, 5000 μm. Values are shown as mean ± SD. *P < 0.05; ***P < 0.001.

PaPE-1 exerted minimal effect on the peripheral immune system

To evaluate potential early changes in immune cell populations in the spleen, circulation, and brain related to treatment, we used a second study design in which MRI and FACS analyses (gating strategy shown in Fig. 3) were performed 3 days after tMCAo (Fig. 4A). In the spleen, total leukocytes (Fig. 4B) and both adaptive cell subpopulations, including B cells, CD4⁺ T cells, CD8 T cells, and NK T cells (Fig. 4C‒4F), and innate cell subpopulations, namely NK cells, macrophages, and neutrophils (Fig. 4F‒4I), did not differ between vehicle-treated, E₂-treated, and PaPE-1-treated groups. Regarding circulating leukocyte subtypes (Fig. 5), no differences were observed 3 days after stroke with either E₂ or PaPE-1 treatment, with the exception of an 86% increase in circulating neutrophils in PaPE-1‒treated mice (P = 0.0535 vs E₂) (Fig. 5F). Thus, splenic and circulating immune cell populations were generally unaltered by E₂ and PaPE-1.

Figure 3.

Gating strategies for the spleen, brain, and peripheral blood samples after stroke. Specific leukocyte populations were identified in the gating first from a forward scatter and side scatter (not shown) to exclude doublets, (A) then gated for CD45⁺ cells (shown in boxed region) with number shown as percentage of parent gate. (B) This group was then gated on TCRβ (y-axis) and CD19 (x-axis) to identify T and B cells. (C) T cells were further gated for CD4 (y-axis) and NK1.1 (x-axis). (D and E) Non‒TCRβ⁺ and CD19⁺ cells were gated for (D) NK1.1 and (E) CDllb (y-axis) and Ly6CLy6G (x-axis). Subpopulations identified in boxes and gating designated by arrows.

Figure 4.

Splenocyte immune populations were not altered 3 days after stroke. (A) Schematic for treatment procedure: female mice were ovariectomized (Ovex) at 8 weeks and received vehicle, E₂, or PaPE-1 delivered by subcutaneous pellet both before and after tMCAo 3 days before euthanasia (Euth). (B‒I) Splenocytes were isolated for flow cytometry (n = 5 to 7 per group) from E₂-treated (green squares) and PaPE-1‒treated (red triangles) cohorts for comparison with vehicle controls (black circles). There were no poststroke differences in (B) general leukocyte counts; (C) B cells; T cells including (D) CD4⁺ T cells, (E) CD8 T cells, and (F) NK T cells; and innate populations, including (G) NK cells, (H) macrophages, and (I) neutrophils. Values are shown as mean  ±  SD.

Figure 5.

PaPE-1 increased circulating neutrophils 3 days after stroke. Percentage of circulating lymphocyte profiles for (A) B cells, (B) CD4 T cells, and (C) CD8 T cells, as well as innate populations for (D) NK cells, (E) monocytes, and (F) neutrophils are shown for vehicle-treated (black circles), E₂-treated (green squares), and PaPE-1‒treated (red triangles) cohorts (n = 5 to 7 per group). Individual leukocyte populations are shown on the y-axis. PaPE-1‒treated mice exhibited higher percentages of neutrophils only relative to E₂-treated mice (F_(3,21) = 5.75). Values are shown as mean  ± SD.

PaPE-1 reduced leukocyte infiltration into the postischemic brain

To determine whether PaPE-1‒induced neuroprotection is related to reductions in leukocyte infiltration into the brain parenchyma after stroke, we quantified immune cell subsets in the CNS 3 days after tMCAo (Figs. 6 and 7). In vehicle-treated mice, stroke induced an eightfold increase in the number of CD45⁺ leukocytes in the ischemic hemisphere (P < 0.05 vs the contralateral hemisphere) (Fig. 6A), similar to previously published results (61). PaPE-1 (P < 0.01) treatment drastically reduced poststroke leukocyte egress into the ischemic hemisphere by 93% (Fig. 6A). Although no significant differences were observed between NK cells in the ipsilateral and contralateral hemispheres and between cohorts (Fig. 6B), PaPE-1 induced a 91% reduction in magnitude for monocyte and macrophage egress (P < 0.05) (Fig. 6C). Neutrophil recruitment into the injured hemisphere was also decreased by 97% with PaPE-1 treatment (P < 0.01) (Fig. 6D) but not with E₂ treatment.

Figure 6.

PaPE-1 reduced poststroke innate cell infiltration into the brain. (A) Total leukocyte infiltration was reduced in the ischemic (i.e., ipsilesional; F_(3,20) = 10.09) cortex (gray bars) of PaPE-1‒treated mice (red triangles) in comparison with that of controls (black circles). (B) Neither PaPE-1 treatment nor E₂ treatment (green squares) affected NK cells in the brain. (C and D) PaPE-1 did reduce the number of (C) macrophages and monocytes (F_(3,20) = 8.08) and (D) neutrophils (F_(3,20) = 9.71) (n = 6 to 7 per group). Contralesional values were generally reduced compared with the ipsilesional hemispheres, but with greatest reductions in PaPE-1‒treated mice. Values are shown as mean ± SD. *P < 0.05; **P < 0.01. Asterisks without brackets indicate comparisons with the ipsilesional hemisphere.

Figure 7.

PaPE-1 and not E₂ reduced poststroke cytotoxic T cell and B cell infiltration into the brain. (A) PaPE-1 (red triangles) reduced CD19 B cell infiltration (F_(3,20) = 8.21) in the ischemic (i.e., ipsilesional) cortex (gray bars) compared with vehicle control (black circles). (B) Neither PaPE-1 nor E₂ (green squares) significantly reduced CD4 T cell recruitment. (C and D) PaPe-1 reduced cytotoxic CD8 T cells (F_(3,20) = 7.87) but not NK T cells (n = 6 to 7 per group). Not indicated by symbols, contralesional hemispheres exhibited PaPE-1‒induced reductions in (A) CD4 T cells (F_(3,20) = 6.90) and (C) CD19 B cells (F_(3,20) = 7.19). Values are shown as mean ± SD. *P < 0.05. Asterisk without a bracket indicates comparison with the ipsilesional hemisphere.

Regarding adaptive immune cell infiltration into the postischemic CNS, B cell recruitment as well was attenuated by 73% with PaPE-1 (P < 0.05) (Fig. 7A). E₂ and PaPE-1 treatments had no effect on CD4⁺ T cell egress into the ischemic hemisphere (Fig. 7B). However, in parallel to the innate and B cell populations, only PaPE-1 decreased cytotoxic T cell (CD8) recruitment into the injured hemisphere (P < 0.05) (Fig. 7C), with no effect on NK T cell abundance (Fig. 7D). As such, immune cell infiltration in the early poststroke period was uniquely attenuated by selective nonnuclear ER activation. These cumulative findings indicate that the favorable effects of nonnuclear ER stimulation on stroke severity are associated with decreases in leukocyte trafficking into the injured hemisphere.

PaPE-1 did not affect neurogenesis or angiogenesis in the postischemic brain

To assess long-term mechanisms of recovery (i.e., 2 weeks after tMCAo) potentially altered by PaPE-1, we first quantified hippocampal neurogenesis in the inner granular cell layer of the dentate gyrus by evaluating doublecortin (DCX) expression (62) (Fig. 8A and 8B). Both E₂ and PaPE-1 treatments had no effect on DCX⁺ cell numbers in the ipsilateral or contralateral hemispheres. It should be noted that several mice in each group had MRI-identified lesions extending caudally into the hippocampus (Fig. 8C), suggesting that the severity of the 45-minute tMCAo in ovariectomized female mice may be responsible for the overall low number of DCX⁺ cells in the hippocampus in all groups. We additionally quantified cortical angiogenesis within the tMCAo territory by ALP staining (63) (Fig. 8D and 8E). Capillary density in the ischemic brain sections did not differ between vehicle-treated, E₂-treated, and PaPE-1‒treated mice when normalized to the respective contralateral hemisphere. These data suggest that the observed earlier anti-inflammatory actions of E₂ and PaPE-1 had no effect on long-term poststroke neuronal or vascular plasticity.

Figure 8.

PaPE-1 improved motor recovery after stroke without long-term effects on neurogenesis or angiogenesis. (A) PaPE-1 (red triangles) and E₂ (green squares) did not affect neurogenesis, (B) as assessed by DCX⁺ cells in the dentate gyrus. (C) However, representative T2-weighted images from vehicle-treated, E₂-treated, and PaPE-1‒treated mice at 7 days after stroke exhibit tissue damage in the posterior hippocampus (white areas in the left hemisphere). (D) Treatment also did not affect cortical angiogenesis, (E) as assessed by ALP stain 14 days after stroke compared with vehicle controls (black circles) (n = 4 to 7 per group). (F) Rotarod training was performed before tMCAo, and testing was done after tMCAo to evaluate poststroke function (n = 9 to 10 per group). Only PaPE-1‒treated mice (red line, asterisks) showed significant recovery over vehicle controls (black line) at 6 days after tMCAo, though both PaPE-1 and E₂ (green line, asterisk) groups recovered at 13 days. Day (F_(11,286) = 19.63; P < 0.0001) and group (F_(2,26) = 5.46; P < 0.05) significance did not result in an overall significant interaction (F_(22,286) = 1.43; P = 0.10). (G‒I) The degree of motor coordination recovery did not correlate with MRI lesion volume in mice treated with (G) vehicle, (H) E₂, or (I) PaPE-1. Scale bar, 5000 μm. Values are shown as mean ± SD. *P < 0.05.

PaPE-1 promoted long-term functional recovery after stroke

During initial rotarod training, PaPE-1‒treated mice showed greater motor coordination at day 2 than vehicle-treated mice (Fig. 8F), but overall all groups of mice reached the plateau by day 5 of training. When tested 2 days after tMCAo, all groups displayed comparable motor deficits of 40% to 60% of baseline. Only PaPE-1 improved functional recovery to 85% of baseline at 6 days after stroke (P < 0.05 vs vehicle), but both E₂ and PaPE-1 improved motor coordination at 13 days after stroke, to 76% and 82% of baseline, respectively (both P < 0.05 vs vehicle). The degree of functional recovery did not correlate with infarct volume in any study group (Fig. 8G‒8I), though it should be noted that there was a large difference in infarct volumes between the groups. Thus, nonnuclear ER activation with PaPE-1 yielded a recovery of motor coordination that was at least as robust as that observed with E₂ treatment.

Discussion

It is now apparent that cellular, tissue, and physiologic responses to estrogens in the CNS entail the culmination of both nuclear and nonnuclear ER actions (21, 64, 65), including in the contexts of neuronal plasticity and behavior (29, 66). To better understand how estrogens affect the severity of stroke injury, we evaluated the contribution of nonnuclear ER-mediated processes in estrogen-related neuroprotection in mice after tMCAo. We chose PaPE-1, an agent likely to have CNS effect when given systemically, to invoke nonnuclear ER activation and compared its effects with those of E₂. We discovered that PaPE-1 had beneficial effects on stroke lesion severity and related edema to a degree that is at least comparable to that observed with E₂. In doing so, we demonstrated that selective nonnuclear ER activation afforded neuroprotection in the setting of focal cerebral ischemia‒reperfusion injury.

In early studies of nonnuclear ERs, E₂ was conjugated to BSA to prevent its cellular entry and provide selective activation of plasma membrane‒associated ERs. However, freshly prepared solutions of E₂-BSA contain free immunoassayable E₂ and are of very high molecular weight, suggesting extreme protein cross-linking; in addition, E₂-BSA binds to ERs poorly because the E₂ is linked to BSA through groups that are important for ER binding (32, 67). As important, E₂-BSA is biodegradable, making long-term experiments problematic and in vivo studies unapproachable. In recognition of these limitations of E₂-BSA, the EDC was developed entailing a hydrolytically stable linkage of E₂ to a large, positively charged, nondegradable PAMAM dendrimer. The short tether linking E₂ to PAMAM through a 17α-phenylethynyl unit provides optimal ligand access to ERs and yields a binding affinity comparable to that of E₂ alone (33, 34). Studies initially in breast cancer cells and then in endothelial cells demonstrated that EDC was excluded from the nucleus and readily activated nonnuclear signaling but was ineffective at stimulating ER target gene expression. EDC was then used in vivo first in studies of vascular injury in mice (15, 33, 34).

PaPE-1 was designed to preferentially activate extranuclear-initiated ER signaling not by restricting subcellular distribution like EDC, but rather by preserving essential chemical and physical features of estrogens while greatly reducing binding affinity for ERs. Compared with E₂, PaPE-1 has 50,000-fold lower binding affinity for ERα and ERβ and a greatly increased dissociation rate (38). The modulation of gene expression by nuclear ERs requires sustained action of the ER-hormone complex sufficient to displace heat shock proteins, recruit coregulators, stimulate ER binding to chromatin, modify chromatin structure and histones, and activate RNA polymerase II to initiate gene transcription. Such sustained action is provided by the kinetically stable ER-E₂ complex but not by the kinetically more labile ER‒PaPE-1 complex. Mirroring these unique features of extranuclear vs nuclear ER action, the selectivity of PaPE-1 for extranuclear ER activation was rigorously demonstrated in MCF-7 cells. Although both E₂ and PaPE-1 activate mTOR and MAPK-dependent processes, PaPE-1 does not cause high affinity binding of the cofactor steroid receptor 3 or ERα recruitment to chromatin. Furthermore, PaPE-1 does not stimulate reproductive or mammary tissues, but it triggers beneficial metabolic responses in adipose tissue and liver and accelerated vascular repair after endothelial injury (38). The present work now reveals that PaPE-1 additionally affords potential CNS benefit.

The effect of nonnuclear ER activation on global cerebral ischemia (GCI) was previously evaluated in rats in a study of EDC injected directly into the cerebral ventricles 60 minutes before GCI (68). Hippocampal neuroprotection and improved spatial learning and memory after GCI were observed and occurred concomitantly with enhanced activation of the prosurvival kinases ERK and Akt and decreased activation of the proapoptotic kinase JNK. However, in contrast to EDC, PaPE-1 has the major therapeutic advantage that it can be administered systemically. Also in a global ischemia model, diphenylacrylamide estrogenic compound or G-1 delayed the loss of hippocampal neurons (6), suggesting that GPER-1 may serve a neuroprotective effect. The present findings of neuroprotection with PaPE-1 are not likely mediated by GPER-1 because studies in MCF-7 cells, in which both ERα and GPER-1 were expressed and were silenced and in which the ERα antagonist and GPER-1 agonist ICI 182,780 was used, indicate that PaPE-1 activated ERα and not GPER-1 (38).

In addition to evaluating infarct size and regional swelling, the present work determined the long-term effect of stroke on motor function by rotarod testing. During the initial training, PaPE-1 caused an early increase in motor performance at least equal to that attained with E₂. These observations may relate to ER actions in brain regions associated with memory, learning, and motor functions, including the hippocampus, cerebral cortex, amygdala, and basal forebrain (16–18, 26, 27), to estrogen regulation of neurotransmission in the cerebellum (69), or to nonnuclear actions of estrogens participating in synaptic plasticity (29, 30). More importantly, PaPE-1 administration resulted in long-term improvement in motor coordination after stroke, with a favorable effect on early motor recovery that may be secondary to overall smaller infarct volumes. However, because the degree of motor recovery was independent of lesion size in all three study groups, the long-term functional benefits may be related to the promotion of poststroke functional plasticity in areas of the brain that support recovery of motor function after tMCAo in mice (70).

Because leukocyte infiltration into the brain has deleterious effects on expansion of ischemic injury and subsequent repair (42–44), we determined whether PaPE-1‒induced neuroprotection is also related to an anti-inflammatory effect on leukocyte recruitment into the ischemic brain. PaPE-1 consistently decreased recruitment of immune cells into the ischemic brain more than E₂. Although this is potentially secondary to the overall smaller initial infarct volumes in PaPE-1‒treated mice, regulation could also be occurring at the blood-brain barrier. We previously found that neuroprotective treatments could induce differential recruitment of leukocytes into the poststroke brain through altered chemokine, selectin, and integrin expression by endothelial cells (45, 47). These changes can occur in the absence of alterations in splenic or circulating immune cell populations, as was observed with PaPE-1 treatment, and the basis for altered immune cell recruitment is an appropriate target for future mechanistic studies. Interestingly, innate immune cell incorporation was blunted by both E₂ and PaPE-1, albeit nonsignificantly for E₂, whereas adaptive immune cell infiltration was uniquely decreased by PaPE-1. It is important to note that nonnuclear ER activation by PaPE-1 leads not only to changes in extranuclear processes but also to alterations in gene expression (38). In addition, observations unique to PaPE-1 or greater for PaPE-1 vs E₂ may reflect processes in which the actions of estrogens via nuclear receptors oppose the actions that occur through nonnuclear receptor activation. The effects of E₂ on immune responses in the CNS are recognized as being diverse and characterized by nonnuclear and nuclear process cross-talk (71–75). As such, numerous cellular processes may underlie the beneficial effects of PaPE-1 on stroke severity and long-term outcome, and the relevant cellular targets may be endogenous or exogenous to the CNS, particularly with respect to the poststroke adaptive immune response. Apart from the effects on leukocyte infiltration in the CNS, PaPE-1 may potentially regulate other functional aspects of peripheral and splenic immune cell populations in the long term, including subsets of T and B cells not identified in our current flow gating, which can be addressed in future studies.

Having observed long-term benefit with PaPE-1 treatment and noting that prior studies indicate that E₂ increases neurogenesis (76), cerebral perfusion, and angiogenesis after ischemic injury (77), we evaluated neuronal and vascular plasticity. We did not observe an effect of either E₂ or PaPE-1 on angiogenesis or neurogenesis. This disparity with prior findings for E₂ is potentially related to the use of the tMCAo stroke model, as some mice suffered lesions that extended into the hippocampal region wherein neurogenesis was evaluated. Alternatively, in the present studies the observed neuroprotective effects of E₂ or PaPE-1 on lesion volume, edema, and functional recovery may have all been related to their effects on the recruitment of leukocytes, particularly T cells, to the CNS after injury (78).

Summary/Conclusions

The preclinical promise of estrogens in stroke and other neurologic conditions has been limited in translation to human trials (79), highlighting the need for detailed understanding of the specific actions of estrogens that may provide neuroprotection. In the present work, we revealed that the selective activation of nonnuclear ERs with PaPE-1 is entirely sufficient to decrease stroke severity and improve functional recovery in mice. Of note, PaPE-1 was administered systemically, indicating that it has favorable CNS action without requiring central delivery. Furthermore, as observed in the present work and previously (38), PaPE-1 did not invoke a uterotrophic response or aberrant growth in mammary tissue or breast cancer cells. As such, PaPE-1 and similar molecules may warrant consideration as ER modulators to provide neurologic protection without detrimental effects on the reproductive system or cancer risk.

Glossary

Abbreviations:

ALP

alkaline phosphatase

CBF

cerebral blood flow

CNS

central nervous system

DCX

doublecortin

E₂

estradiol

E₂-BSA

estradiol conjugated to BSA

EDC

estrogen dendrimer conjugate

ER

estrogen receptor

FACS

fluorescence-activated cell sorting

GCI

global cerebral ischemia

GPER-1

G protein‒coupled estrogen receptor-1

NK

natural killer

PAMAM

poly(amido)amine

PaPE

pathway-preferential estrogen

tMCAo

transient middle cerebral artery occlusion

UT

University of Texas