Retinoid-X-Receptor as a Mediator of Poststroke Recovery by Reversing Age-Associated Phenotypes of Microglia/Hematogenous Macrophages

Abstract

After stroke, microglia and hematogenous macrophages, together referred to as MΦ, clear dead cells and cellular debris in the infarcted brain through phagocytosis as an essential part of the recovery process. However, the phagocytic capability of MΦ declines with age. Furthermore, aged MΦ become overactivated in response to stroke, enhancing secondary brain injury. In this study, we demonstrated that by reversing the age-related dysfunctions in MΦ through activating the retinoid X receptor (RXR), the recovery after stroke in the aged brain could be improved. Using RNA-sequencing, we compared the transcriptomes between MΦ isolated from the brains of young and aged male mice. We observed higher levels of proinflammatory genes and lower levels of phagocytosis-facilitating genes (Cd206 and Cd36) expressed by aged MΦ. Meanwhile, the treatment with RXR agonist bexarotene (BEX) reversed the signature genes of microglia aging in the aged MΦ. With the in vivo phagocytosis model, we showed that BEX enhanced the phagocytic ability of aged MΦ. Using the MCAo stroke model and male and female mice, we established that BEX improved sensorimotor and cognitive recovery after MCAo in a myeloid-RXRα–specific and myeloid-RXRα–dependent manner. In conclusion, we showed that activating RXRα partially restores age-related MΦ dysfunctions and that RXRα deficiency in MΦ limits the therapeutic effect of RXR in improving poststroke recovery in the aged brain.

Significance Statement

Aging is the most robust unmodifiable risk factor for ischemic stroke. The majority of stroke patients are individuals older than 65 years (Virani et al., 2020). Aging is also a predictor for worse outcomes (increased neurological deficit and mortality) after stroke. Despite this, only a limited number of studies attempted to elucidate the differences in poststroke recovery between young and elderly subjects. One likely reason for worse repair/recovery in aged subjects after stroke is the decline in phagocytic and reparative properties of aged MΦ (Hefendehl et al., 2014; Koellhoffer et al., 2017). Aged MΦ may fail to clear cell debris and proinflammatory molecules from the ischemic brain, thereby worsening secondary injury and recovery after stroke (Shen et al., 2019).

Introduction

Ischemic stroke triggers a series of pathological responses known as the ischemic cascade, leading to cell death and brain damage through various mechanisms (Lo et al., 2005; Wang and Qin, 2010; Chen et al., 2011; Xing et al., 2012). The dead/injured cells release/spill various harmful toxic molecules, including damage-associated molecular patterns (DAMPs) and free radicals that harm the ischemia-affected brain. Locally, DAMP-affected microglia produce/secrete proinflammatory cytokines/chemokines, proteases that orchestrate the recruitment and activation of other immune cells, including neutrophils, monocytes, and T/B cells, to the site of injury. This may cause exacerbation of proinflammatory responses and augmentation of the damage (also known as secondary injury), creating an environment that impedes poststroke repair and recovery (Helming, 2011; Hu et al., 2015; Xiong et al., 2016).

Aging is the strongest unmodifiable risk factor for ischemic stroke. The majority (∼75%) of stroke patients are older than 65 years (Tsao et al., 2023). Aging also predicts worse outcomes (increased neurological deficit and mortality) and worse poststroke recovery. Despite the need for more research on this topic, only a small proportion of studies attempted to elucidate the differences in the pathobiology of the recovery process between young and elderly subjects. One potential reason for worse poststroke recovery in aged subjects is age-related decline in reparative phagocytic function of microglia/hematogenous macrophages (together refer to as MΦ; Hefendehl et al., 2014; Koellhoffer et al., 2017). Clearance of the infarcted/dead brain tissue (source of toxicity and inflammation) by MΦ is an essential step in facilitating brain repair. Aged MΦ may be less effective in clearing cell debris and other toxic and proinflammatory molecules from the ischemic brain. The prolonged presence of the stroke-damaged brain tissue (source of inflammation) may extend the duration of inflammation, worsen secondary injury, and impair repair/recovery after stroke (Shen et al., 2019).

Prior studies, including from our team, demonstrated that the pleiotropic transcription factor retinoid X receptor (RXR) in MΦ is important in coordinating the effective phagocytic activities (Roszer et al., 2011; Cramer et al., 2012; Natrajan et al., 2015; Ting et al., 2020). Furthermore, the expression of the RXRα (a primary isoform of RXR in MΦ) signaling pathway is compromised in the MΦ of aged animals (Natrajan et al., 2015). Therefore, we hypothesize that in response to RXR activation, the MΦ are driven toward their phagocytic phenotypes, which rescue the aged-associated dysfunctions of MΦ to achieve stronger reparative activities. Specifically, we anticipate enhancing phagocytosis-mediated clearance of cytotoxic brain cell debris by MΦ that would reduce secondary brain injury and improve repair after stroke. We also anticipated that as a therapeutic target, modulation of MΦ could render an extended therapeutic window since the recruitment of MΦ into the ischemic brain to produce effective cleanup and repair takes hours and then continues for days (Planas, 2018).

Materials and Methods

All experiments were performed using randomization and blinded approaches. Before the experiment started, the mice were assigned to each group randomly (one- or two-coin toss). The group assignment was coded throughout the experiment. All analyses were performed by investigators blinded to the group and treatment assignment. The sample sizes were calculated by a priori power analysis and stated in the figure legend.

Materials and reagents

All reagents and antibodies used in this study are listed in Table 1.

Table 1.

List of reagents and antibodies used in this study

Reagents	Source	Catalog #
BD Cytofix/Cytoperm Kit	BD Biosciences	554714
NP-40	Calbiochem	492016
DMSO	Corning	25-950-CQC
Percoll Plus	Cytiva	17544502
Absolute ethanol	Thermo Fisher Scientific	BP2818-500
PBS	GenDEPOT	P2101-050
RPMI 1640	Invitrogen	11875-135
DMEM	Invitrogen	11995-065
FBS	Invitrogen	10082-147
HBSS	Invitrogen	14185-052
pHrodo Deep Red TFP Ester	Invitrogen	P35358
Fluoromount-G, with DAPI	Invitrogen	00-4959-52
Zombie Red Fixable Viability Kit	BioLegend	423110
Live/Dead Fixable Near-IR Stain Kit	Invitrogen	L34975
Anti-Ly6G MicroBeads	Miltenyi Biotec	130-120-337
MS Columns	Miltenyi Biotec	130-042-201
CD11b (Microglia) MicroBeads	Miltenyi Biotec	130-093-636
M-CSF	Prospec	CYT-439
Rneasy Micro Kit	QIAGEN	74004
Dnase I	Roche	10104159001
BEX	Sigma-Aldrich	SML0282
Poly-D-lysine	Sigma-Aldrich	P7886
PFA	Thermo Fisher Scientific	J19943-K2
Papain	Worthington	LK003176
Antibodies	Source	RRID
Anti-mouse CD16/CD32 Fc shield	Tonbo Biosciences	AB_2621487
Anti-mouse CD36 AF488	BioLegend	AB_528792
Anti-mouse CD11b APC	BioLegend	AB_312795
Anti-mouse CD45 Bv421	BD Biosciences	AB_2651151
Anti-mouse CD206 PE	BioLegend	AB_10895754
Anti-mouse TNF FITC	BD Biosciences	AB_395379
Anti-mouse IL-1α PE	BD Biosciences	AB_397339
Anti-mouse CD11b BV510	BD Biosciences	AB_2737913
Anti-mouse Ly6G FITC	BD Biosciences	AB_10562567
Anti-mouse CD45 PE	BD Biosciences	AB_2870024
Anti-mouse CD11b FITC	Tonbo Biosciences	AB_2621676
Anti-mouse NK-1.1 PE/Cy7	BioLegend	AB_2888852
Anti-mouse Ly6G PE/Cy7	BioLegend	AB_1877261
Anti-mouse CD3e PE/Cy7	BioLegend	AB_312684
Anti-mouse CD19 PE/Cy7	BioLegend	AB_313655
Anti-mouse Iba1 rabbit IgG	Fujifilm	AB_839504
Anti-mouse CD11b rat IgG	Bio-Rad Laboratories	AB_321292
Anti-mouse NeuN AF488	Millipore	AB_2149209
Anti-mouse CD68 rat IgG	Bio-Rad Laboratories	AB_324217
Anti-mouse MMR goat IgG	R&D Systems	AB_2063012
Anti-mouse CD163 rabbit IgG	Abcam	AB_2753196
Anti-mouse CD36 rabbit IgG	Novus Biologicals	AB_10003498
Goat anti-rat IgG AF546	Invitrogen	AB_2534125
Goat anti-rat IgG AF488	Invitrogen	AB_2534074
Goat anti-rabbit IgG AF488	Invitrogen	AB_2576217
Goat anti-rabbit IgG AF546	Invitrogen	AB_2534093
Donkey anti-goat IgG AF647	Invitrogen	AB_2535864
Fab fragment donkey anti-mouse IgG	Jackson ImmunoReasearch Laboratories	AB_2307338

Animals

All animal studies were approved by the Animal Welfare Committee of UTHealth. Mice were fed a standard rodent diet and housed in standard cages on a 12 h light/dark cycle. Young (2–4 months old) and aged (18–20 months old) C57BL/6 male mice used in the RNA-sequencing and the in vivo phagocytosis experiments were purchased from Charles River Laboratories. Aged (18–20 months old; both male and female) RXRα^fl/fl and Mac-RXRα^−/− mice (Ricote et al., 2006) used in the MCAo stroke experiment were bred in-house. These mice were the progeny of LysM-cre⁺/RXRα^−/− mice crossed with RXRα^fl/fl. The genotypes of mice were age- and gender-matched littermates, either RXRα^fl/fl control (LysM-Cre⁻/RXRα^fl/fl) or Mac-RXRα^−/− knock-out (LysM-Cre⁺/RXRα^fl/fl).

Administration of RXR agonist

Bexarotene (BEX, Sigma-Aldrich) dissolved in 3% DMSO was administered intraperitoneally at 5 mg/kg, first 24 h after surgery, and then once a day for 7 d. DMSO (3%) in saline was used for vehicle control.

Ischemia model in mice

Focal ischemia in mice was induced by the tandem left middle cerebral artery (MCA) and left common carotid artery (CCA) occlusion, as we described (Aronowski et al., 1997; Aronowski and Labiche, 2003; Jeon et al., 2020). Briefly, animals were anesthetized, and a burr hole was drilled rostral to the fusion of the zygomatic arch with the squamosal bone to expose the left MCA rostral to the rhinal fissure. A 0.005 in. diameter stainless steel wire (Small Parts) was placed underneath the left MCA rostral to the rhinal fissure, proximal to the major bifurcation of the MCA, and distal to the lenticulostriate arteries. The left CCA was occluded using atraumatic Heifetz aneurysm clips. Reperfusion was established 60 min after occlusion. We experienced no procedural or poststroke mortality when performing experiments in this study.

Single-cell suspension preparation from mouse brain for MΦ isolation and flow cytometry analysis for in vivo phagocytosis and MΦ phenotyping, ex vivo

The mouse brains were physically and enzymatically dissociated using gentleMACS C tubes (Miltenyi Biotec) and enzyme mix papain (20 units/ml) and DNase I (100 μg/ml). The enzyme mix was prewarmed at 37°C for 30 min. The brains were cut into 2 × 2 mm pieces and transferred into the prewarmed enzyme mix in C tubes. The C tubes were then placed on gentleMACS Dissociator and run brain tissue program 1–3. After each program finished, the C tubes were incubated at 37°C for 5 min. After the final incubation, the dissociated brain cells were passed through a 100 μm strainer and washed with DMEM. The washed cells were then reconstituted in 10 ml of 30% Percoll and centrifuged at 1,000 × g for 10 min, with a full break. The cell pellet was reconstituted in staining buffer (PBS with 10% FBS) for further analysis.

MΦ isolation from the cell suspension for RNA-sequencing

The single-cell suspension prepared from the brains of aged or young C57BL/6 mice was added Ly6G microbeads (1:10) and incubated for 10 min at 4°C before being sorted with MS Column (Miltenyi Biotec). The flow-through was collected (Ly6G-negative selection) and washed with staining buffer. The cell pellet was reconstituted in staining buffer with CD11b microbeads (1:10) and incubated for 15 min at 4°C before being sorted with MS Column (CD11b-positive selection).

RNA isolation

RNA was extracted using RNeasy micro kits (QIAGEN) following the manufacturer's protocol. Briefly, the harvested MΦ were lysed and homogenized with a 28 G needle attached to a 1 ml syringe before passing through the RNA adsorption column. Genomic DNA was digested on-column with DNase I (30 units). The adsorbed RNA was washed in ethanol before being eluted and redissolved in RNase-free water.

RNA-sequencing and analysis for MΦ

RNA-sequencing, including library construction, was performed by the Cancer Prevention and Research Institute of Texas (CPRIT RP180734) at UTHealth. After quality assessment, the clean reads were mapped to the reference genome using the STAR software (Dobin et al., 2013), and the mapping results were visualized with Integrative Genomics Viewer. Differential expression analysis was performed using the DESeq2 R package (Anders and Huber, 2010); Gene Set Enrichment Analysis (GSEA), with Molecular Signatures Database, was performed using the GSEA software (Mootha et al., 2003; Subramanian et al., 2005).

Flow cytometry analysis for MΦ phenotyping and in vivo phagocytosis

The single-cell suspension prepared from the brain of mice subjected to vehicle or BEX treatment was divided into two portions for surface markers (CD36 and CD206) and intracellular cytokines (IL-1 and TNF) panels. For the surface marker panel, the cells were incubated with surface markers staining antibody cocktails (including CD45, CD11b, Ly6G, CD3, CD19, NK1.1, CD36, and CD206) for 20 min, on ice; avoid light. After incubation, the cells were washed with PBS before incubated with viability dyes (1:1,000 in PBS) for 10 min, on ice; avoid light. The cells were then washed with staining buffer and fixed with 4% PFA for 15 min. After fixation, the cells were washed with staining buffer and analyzed using Cytek Aurora spectral flow cytometry.

For the intracellular cytokines panel, the cells were incubated with BD GolgiPlug protein transport inhibitor (1:1,000) in a CO₂ incubator (5% CO₂) at 37.0°C for 4 h before staining for the surface markers (CD45, CD11b, Ly6G, CD3, CD19, NK1.1). The cells were then washed with PBS and stained for viability dyes for 10 min. After incubation, the cells were washed with staining buffer, followed by fixation and permeabilization using a BD fixation/permeabilization kit. The fix/perm cells were then incubated with intracellular staining antibody cocktails (IL-1 and TNF) for 30 min on ice; avoid light. After washing with staining buffer, the cells were analyzed using Cytek Aurora spectral flow cytometry. The fluorescence minus one controls were used for gating.

Brain debris preparation for in vivo phagocytosis

The whole brain of the tissue donor mouse was intracardially perfused with ice-cold PBS. The dissected brain was cut into 2 × 2 mm pieces before being transferred into a C tube (Miltenyi Biotec) containing 5 ml of papain (20 units/ml) and DNase I (100 μg/ml). The C tube was then placed on gentleMACS Dissociator and run spleen tissue program 1–3. After each program finished, the C tube was incubated at 37°C for 5 min. After the final incubation, the brain debris was passed through a 100 μm strainer and washed with PBS. The brain debris was then labeled with pHrodo Deep Red TFP amine reactive dye (Invitrogen) following the manufacturer's protocol.

In vivo phagocytosis

The pHrodo-labeled brain debris was used to measure MΦ phagocytic activities. The labeled brain debris (∼1 × 10⁹ debris in 10 μl PBS) were stereotactically injected into the left striatum (2 mm lateral to the bregma and 3.5 mm below dura mater) of the recipient mouse brain. After 3 d, the mice were killed, and the left striatum was processed for flow cytometry analysis, as described above. The debris-engulfing MΦ in the ischemic brains were defined as Live/Dead⁻ CD45^hi CD11b⁺ Ly6G⁻ APC-debris⁺ populations. The MΦ from the contralateral striatum were used as control.

Behavioral Tests to Assess Neurological and Cognitive Deficits

The corner turn test and foot fault test were conducted from 2:00 P.M. to 5:00 P.M. The cognitive tests, including the Barnes maze, Y maze, and novel object recognition test, were conducted after 5:00 P.M. The mice were calmed in the testing room for at least 1 h before each test. A pretest for the corner turn and foot fault tests was conducted 1–3 d before the MCAo surgery to exclude mice with preexisting asymmetry in behavioral tests.

Corner turn test

The mice were placed in a fresh cage (identical to the home cage) with two Plexiglass boards attached at a 30° angle (Fig. 2A,B). As the mice entered the corner, both sides of the vibrissae were stimulated symmetrically, and the mice then turned back toward either the right or left direction. Ten trials were recorded, and the asymmetry of the turns was calculated with the following formula: asymmetry index = | right turn − left turn |/10.

Figure 2.

BEX enhances lipid/cholesterol metabolisms in MΦ from young and aged mice and reverses age-related phenotypes of MΦ. MΦ were isolated from the brains of young (4–6 months old) and aged (18–20 months old) mice using MACS positively selecting CD11b+ cells followed by Ly6G-negative selection. BEX or vehicle was intraperitoneally injected daily for 3 d before MΦ isolation. The volcano plots show the DEGs between MΦ isolated from vehicle- and BEX-treated (A) young mice and (B) aged mice. C, Top five pathways enriched in MΦ isolated from BEX-treated young mice versus vehicle-treated young mice. Data were analyzed using the GSEA software with the GO gene set database. D, The heatmap shows the representative core enrichment (leading-edge genes) of apoptosis gene sets (p < 0.001) between microglia isolated from vehicle- and BEX-treated young mice. Each column represents microglia samples isolated from one mouse. The color reflects the z score of the gene expression. The enrichment plots showing the gene set of (E) sterol homeostasis (p < 0.001) and (F) fatty acid β-oxidation (p = 0.015) from the GO database was enriched in microglia isolated from aged BEX-treated mice versus aged vehicle-treated mice. Data were analyzed with the GSEA software. G, The heatmap shows the representative core enrichment of the gene sets in microglia isolated from BEX-treated aged mice that facilitate apoptotic cell clearance (p = 0.0017; GO) and antioxidative stress (p = 0.0016; Wikipathway). Each column represents microglia samples isolated from one mouse. The color of the heatmap reflects the z score of the gene expression. H, The enrichment plot showing the gene set of brain myeloid microglial cell aging was enriched in microglia isolated from aged vehicle-treated mice versus BEX-treated mice. Data were analyzed with the GSEA software. NES = −1.3; p = 0.007. The heatmap shows the core enrichment genes. Each column represents microglia samples isolated from one mouse. The color reflects the z score of the gene expression. To validate conclusion from the RNA-sequencing data, the phenotype of MΦ using the same experimental setup was further assessed by flow cytometry. The mean fluorescent intensity (MFI) of (I) CD36, (J) CD206, and (K) TNFα measured on/in CD11b⁺ Lin⁻ (CD19, CD3, NK1.1, and Ly6G) MΦ. The data are expressed as mean ± SEM (n = 3 mice in each group). *p < 0.05. Two-way ANOVA followed by Fisher's LSD test.

Foot fault test

The mice were placed on an elevated ladder bridge, and 10 uninterrupted steps were recorded by the camera as the mice traveled across the bridge. The percentage of missteps with the affected front limb (right limb) was calculated.

Y maze

The mouse was placed in one of the three arms of the Y-shape arena and was able to explore the arena ad libitum for 5 min. The order of the arm that the mouse entered was recorded, and the alternation rate was calculated using the Noldus EthoVision XT software. For every three consecutive entries, the mouse made without repetition of the Y maze arm is counted as a correct alternation. The alternation rate is calculated with the formula: the number of correct alternations / total possible alternations (the number of total entries, 2).

Open-field and novel object recognition test

During the open-field test, the mice were placed in a 15 × 15 in. empty arena and ad libitum explored the arena for 20 min. The video was recorded and analyzed using the Noldus EthoVision XT software. The total distance the mice moved was measured as an index for mobility; the ratio of time the mice spent in the center versus at the edge of the arena was also recorded as an index for anxiety (Seibenhener and Wooten, 2015).

Twenty-four hours after the open-field test, the mice were placed back in the same arena with two identical objects at each side of the arena (familiarization). The mice were given 20 min to explore the objects, and the time they spent investigating each object was recorded. Then, the mice were placed back in the home cage. One hour later, the mice were placed back in the same arena, and one of the two objects switched to a novel object with a different shape. The mice were able to explore ad libitum in the arena for 20 min. After each session, the arena was thoroughly cleaned with 70% ethanol to remove odors.

The mice's time spent investigating each object was measured, and the discrimination index was calculated using the formula DI = (time exploring the novel object − time exploring the familiar object) / total time exploring both objects. The animals whose total exploring time with both objects was shorter than 20 s were excluded from the analysis. To eliminate the bias due to preexisting preference, the mice with DI > 0.5 in the training session were also excluded from the analysis.

Barnes maze

The procedures of the Barnes maze to assess spatial learning and memory followed the protocols previously reported with minor modifications (Rodriguez et al., 2013; Pitts, 2018). The videos were recorded and analyzed using the Noldus EthoVision XT software. In the habituation phase, the mice were given 2 min to explore the arena ad libitum during the habituation phase. After 2 min of free exploration, the mice were gently guided to the goal box and spent an additional 2 min inside the box. There were no distal visual cues at this phase.

The acquisition phase began at least 24 h after habituation. It consisted of four training sessions per day for four consecutive days (16 training sessions in total). At the beginning of a session, the mouse was placed under the start box (opaque cylinder with cap) in the center of the arena for 10 s before the start box was removed. Each session was terminated when the mouse entered the goal box or 3 min had elapsed. There were four visual cues on the walls outside the arena with different shapes and colors for the mice to orient the goal box. The visual cue did not directly point to the goal box. During training sessions, the search strategies the mice used were recorded. The search strategies were classified into three categories: random, serial, and reference search. The random search defines the mice searched randomly. Serial search defines the mice searching the holes one by one for more than three holes. Reference search defines the mice used visual cues to orient in the arena and reached the goal box without making more than two erroneous attempts, indicating the mice acquired the spatial memory to locate the goal box. After each session, the Barnes maze was thoroughly cleaned with 70% ethanol to remove odors.

Seventy-two hours after the last training session, the mice were given a 90 s probe session to assess their long-term spatial memory. The latency to the goal box (primary latency) and the number of investigations of false holes before reaching the goal (primary errors) were measured. Since the probe session is aimed at assessing long-term memory, the mice that failed to locate the goal box within 30 s in the last training session, indicating these mice did not form the spatial memory required for the test, were excluded from the probe session.

Animal sacrifice

The mice were killed 31 d after ischemic stroke onset by overdose of anesthesia (0.5 mg/kg body weight chloral hydrate, i.p.). The blood was collected from the right cardiac ventricle into heparin-containing tubes. The animals were then perfused with PBS through the left ventricle, with the right atrium punctured. The spleen and the two femurs were collected before cardiac perfusion with 4% PFA. The descending aorta was clamped to prevent leakage during PFA perfusion.

Brain sample preparation

After the mice were killed and perfused with a 4% PFA solution, the brains were collected and postfixed in PFA for another 48 h at 4°C. The fixed brains were then submerged in 15% glucose in a PBS solution overnight or until sunk before they were transferred into 30% glucose in a PBS solution. Once sunk in 30% glucose in a PBS solution, the brains were snap-frozen and stored at −80°C until further analysis.

Frozen section preparation

The frozen brains were brought to a cryostat (−20°C) for at least 30 min to equilibrate with ambient temperature. They were then sectioned into and 40-μm-thick sections.

Free-floating immunofluorescent staining

The 40 μm frozen brain slices were fixed with PFA solution in a 24-well plate for 15 min. Each section was held in a grid basket. After fixation, sections were permeabilized with 0.5% NP40 in PBS for an additional 15 min, followed by applying blocking solution (10% goat serum + donkey anti-mouse IgG) for 30 min. The slices were then incubated with primary antibodies overnight at 4°C. After washing thoroughly in PBS, the sections were incubated in secondary antibodies for 1 h, followed by three times washing with PBS. The section was then mounted on a microscope glass slide with Fluoromount-G (Invitrogen).

Immunofluorescent imaging and data analysis

The immunofluorescent images were captured using a Zeiss LSM 800 confocal microscope. The images were analyzed using the Zeiss Zen Blue software. For the cell density analysis in hippocampal CA1 neurons, the coronal brain sections −2 mm from the bregma were used. The neurons in the CA1 area were identified by NeuN staining. For the MΦ phenotype analysis, the brain sections containing the peri-infarct area were used. The MΦ were identified by CD11b or Iba-1 staining, and the object-based colocalization analysis was performed using the Zen Blue software to assess the phenotype markers on MΦ.

Statistics

The statistical analysis for RNA-sequencing was performed using the DESeq2 package on the R software. The data have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE287142 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287142). All other statistical analyses were performed using the GraphPad Prism 10 software. Student’s t test was used to compare the data with two groups. Two-way analysis of variance (ANOVA) was used to assess data with two grouping variables (e.g., genotype × treatment). Two-way ANOVA repeated measures were used to analyze the behavioral tests with multiple time points. Three-way ANOVA was used to analyze three grouping variables (genotype × treatment × sex).

Results

Transcriptomic analysis of MΦ harvested from the aged brain of naive mice reveals the aged-related alterations in microglia, which could be reversed by activating RXR

To characterize the aged-associated changes in RXR signaling in MΦ, we performed the transcriptomic analysis on MΦ isolated from the brains of control young (4–6 months old) and aged (18–20 months old) C57BL/6 mice. The MΦ were isolated using magnetic cell selection (MACS) based on Ly6G-negative and CD11b-positive selection. This method yielded CD11b⁺ myeloid phagocytes (MΦ) with purity higher than 90% (Fig. S1). Among the isolated cells (CD11b⁺ MΦ), ∼90% of the cells are CD45^int microglia along with a small proportion (∼10%) of CD45^hi macrophages (Fig. S1). To probe if some of the age-related changes in MΦ can be rescued with RXR activation, RXR agonist BEX or vehicle was intraperitoneally injected once a day for 3 d before MΦ isolation.

Using principal component analysis (PCA), we observed that the datasets of these four groups (young vehicle, young BEX, aged vehicle, and aged BEX) clustered distinctly (Fig. 1A). Differentially expressed genes (DEGs; Fig. 1B) for MΦ isolated from aged and young mice showed that MΦ from aged mice had transcriptomic signatures similar to disease-associated microglia (DAM; Holtman et al., 2015; Keren-Shaul et al., 2017), including Cst7, Clec7a, Axl, Itgax, Cd22, and Lpl. Also, genes essential in phagocytic functions, such as scavenger receptors (Cd206, Cd163, and Cd36), decreased in aged MΦ (Fig. 1C). GSEA with hallmark gene set showed the genes that respond to inflammation were enriched in aged MΦ, likely reflecting the chronic inflammatory environment of the aging brain (Fig. 1D). Also, GSEA with cell type signature gene sets showed that microglial aging markers are enriched in aged MΦ (Tabula Muris Senis Brain Myeloid Microglial Cell Aging; NES = 2.84; p < 0.001; Fig. 1E).

Figure 1.

Transcriptomic analysis of MΦ revealed age-associated phenotypes of MΦ. MΦ were isolated from the brains of young (4–6 months old) and aged (18–20 months old) male mice using MACS positively selecting CD11b+ cells followed by Ly6G-negative selection. BEX (5 mg/kg) or vehicle was intraperitoneally injected daily for 3 d before MΦ isolation. A, PCA of the four groups of microglia used in the transcriptomic analysis. B, A volcano plot showing the DEGs between microglia isolated from aged and young vehicle-treated mice. C, The heatmap showing DEGs between microglia isolated from aged and young vehicle-treated mice that are identified as the transcriptomic signatures of DAM and phagocytosis-facilitating genes. Each column represents MΦ samples isolated from one mouse. The color reflects the z score of the gene expression. The enrichment plots showing the gene sets of (D) inflammatory responses (p < 0.001) from the hallmark database and (E) the gene set of brain myeloid microglial cell aging (NES = 2.84; p < 0.001) were enriched in MΦ isolated from aged vehicle-treated mice analyzed with the GSEA software.

In response to BEX, RXR signaling genes, including Abcg1, Abca1, and Tgm2, were upregulated in both young and aged mice (Fig. 2A,B, respectively). In young mice, the top five pathways (analyzed with GSEA with GO gene sets) promoted by BEX in MΦ are related to cholesterol and sterol metabolism and transport (Abcg1, Abca1, Pparg, and Lrp1; Fig. 2C). Also, BEX suppressed genes involved in inflammation and apoptosis pathways in young MΦ, including Tnf, Tlr4, Atf3, Vim, and Dynll1 (Fig. 2D).

In aged mice, as in young mice, GSEA with GO gene sets showed that pathways related to sterol transport and homeostasis were enriched in BEX-treated MΦ (Abcg1, Abca1, Eepd, and Lrp1; Fig. 2E). Also, the gene set that facilitates fatty acid β-oxidation was enriched by BEX treatment (Cp1a, Abcd2, and Acat3; Fig. 2F). Importantly, gene sets for apoptotic cell clearance and antioxidative stress were enriched in BEX-treated aged MΦ, suggesting that stimulating RXR in aged MΦ enhances their ability to clean up and digest apoptotic cells (Fig. 2G). Furthermore, GSEA with cell type signature gene sets revealed that BEX treatment reduced the gene set markers for microglial aging (Tabula Muris Senis Brain Myeloid Microglial Cell Aging; NES = −1.3; p = 0.007; Fig. 2H). The enriched genes include markers for highly activated microglia in aged brain (Lpl, Cd63, Capg; Jin et al., 2021), immune response (Cd48, Ddah2), and endoplasmic reticulum stress (Erp29). The gene expression profiles between MΦ harvested from young versus aged mice treated with BEX were also analyzed (Fig. S3).

To validate the results of the gene expression data, we also assessed the protein expressions of key receptors for phagocytosis, CD36 and CD206, on MΦ under the same experimental setup using flow cytometry (Fig. S2). Note that without injury, most of the CD11b⁺ MΦ in the brain are CD45^int microglia. In agreement with the transcriptomic analysis, CD36 decreased in aged MΦ compared with young MΦ and that BEX increased the expression of CD36 in both young and aged MΦ (F_(1,8) = 17.33; p = 0.003 in age effect; F_(1,8) = 10.3; p = 0.012 in treatment effect, two-way ANOVA; Fig. 2I). Also, CD206 is decreased in the aged MΦ (F_(1,8) = 10.87; p = 0.011 in age effect, two-way ANOVA; Fig. 2J) in accordance with the gene expression data. To probe the proinflammatory cytokine production, we also measured the intracellular TNFα in MΦ using flow cytometry. In line with the gene expression data, BEX reduced TNFα in both young and aged MΦ (F_(1,8) = 48.78; p = 0.020 in treatment effect, two-way ANOVA; Fig. 2K).

Taking together, the validated RNA-sequencing results show that activating RXR in aged mice can reverse many of the dysfunctions in aged MΦ, including processing for restoring apoptotic cell clearance and reducing proinflammatory responses.

Activating RXR after ischemic stroke in aged animals suppresses inflammatory responses of MΦ in the injured brain

To further assess whether MΦ’ RXR signaling induction could combat the toxic environment in the aged ischemic brain, we performed transcriptomic analysis on MΦ isolated from the brain of aged mice 3 d after ischemic stroke (MCAo). The cells were isolated using MACS, based on Ly6G-negative and CD11b-positive selections, as described above. Note that in the ischemia-injured brain, both CD45^int microglia and CD45^hi macrophages were collected for transcriptomic analysis (Fig. S4). To activate RXR, BEX or vehicle were intraperitoneally injected once a day for 3 d starting from 24 h after MCAo onset.

We identified 53 DEGs with 34 upregulated genes (0.11%) and 19 downregulated genes (0.062%) between MΦ isolated from vehicle- and BEX-treated ischemic brains (Fig. 3A). RXR gene sets that related to sterol and lipid homeostasis were enriched in BEX-treated MΦ in the ischemic brain (Lipc, Abcg1, and Abca1; Fig. 3B). GSEA with GO gene sets for leukocyte activation (Stap1, Tnf, and Batf2; Fig. 3C), proinflammatory cytokine production (Gbp5, Il1b, and Myd88; Fig. 3D), and lymphocyte recruitment (Ccl8, Cxcl12, and Ccl7; Fig. 3E) were downregulated by RXR activation.

Figure 3.

BEX suppresses MΦ inflammatory responses to ischemic stroke in the aged mouse brain. MΦ isolated from the brain of aged mice 3 d after ischemic stroke (MCAo). BEX (0.5 mg/kg) or vehicle were intraperitoneally injected once a day for 3 d starting from 24 h after ischemic stroke onset (n = 3 per group). A, A volcano plot showing the DEGs between MΦ isolated from vehicle- and BEX-treated aged mice 3 d after ischemic stroke. B, The enrichment plot showing the gene set of sterol homeostasis (p < 0.001) from the GO database was enriched in MΦ isolated from BEX-treated stroked mice. Data were analyzed with the GSEA software. The enrichment plots showing the gene sets of (C) myeloid leukocyte activation (p < 0.001), (D) interleukin 1 production (p < 0.001), and (E) lymphocyte chemotaxis (p < 0.001) from the GO database was enriched in MΦ isolated from vehicle-treated stroked mice. Data were analyzed with the GSEA software. The heatmap shows the core enrichment genes. Each column represents MΦ samples isolated from one mouse. The color reflects the z score of the gene expression. To validate conclusions from the RNA-sequencing data, the phenotype of MΦ in the aged ischemia-injured brain was assessed by flow cytometry. F, The mean fluorescent intensity (MFI) of TNFα and IL-1α measured on/in CD11b⁺ Lin⁻ (CD19, CD3, NK1.1, and Ly6G) MΦ. G, The population of MΦ is further divided into CD45^int microglia and CD45^hi macrophages, and their expressions of TNFα and IL-1α were also measured. The data are expressed as mean ± SEM (n = 4 mice in each group). *p < 0.05. Unpaired t test.

To validate the results of transcriptomic analysis, we used flow cytometry to assess protein for the proinflammatory cytokines TNFα and IL-1α in MΦ (Fig. S4) using the same experimental settings as for RNA-sequencing. We observed that, in line with the gene expression data, BEX reduced both TNFα (t₍₆₎ = 3.35; p = 0.015; unpaired t test) and IL-1α (t₍₆₎ = 2.47; p = 0.048; unpaired t test) in MΦ in the aged brain 3 d after stroke (Fig. 3F). Intriguingly, with further separation of CD45^int microglia and CD45^hi macrophages, we found that the reductions of proinflammatory cytokines are attributed to the microglia (Fig. 3G).

These data suggest that activating RXR suppresses the MΦ inflammatory responses to ischemic injury in the aging brain.

Activating RXR promotes MΦ phagocytic ability in the aged brain

Since we identified in transcriptomic analysis that the ability to clear apoptotic cells by MΦ in the aged brain could be improved by activating RXR in these cells, we performed in vivo phagocytosis assay to validate these results. In this experiment, as a proof-of-concept, we used brain debris as the phagocytic targets to simulate the cleanup of brain debris, similar to what happens after stroke. To detect the lysosomal degradation after phagocytosis (normally occurring during phagocytosis), the brain debris were labeled with pHrodo Deep Red dye (Invitrogen). This pH-sensitive dye emits fluorescence only in low pH environment of late endosomes and lysosomes. The pHrodo-labeled debris was stereotaxically injected into the brains (striatum) of aged mice to measure the MΦ-mediated phagocytosis. To activate RXR, BEX was intraperitoneally injected at 2, 24, and 48 h after the debris injection. At 3 d after postdebris injection, the affected brain (striatum) was harvested and processed for flow cytometry. The phagocytes (MΦ) were gated as Live/Dead⁻ CD45⁺ CD11b⁺ Ly6G⁻, which contains hematogenous macrophages (MDM) and microglia (Sedgwick et al., 1991; Rangaraju et al., 2018; Shukla et al., 2019). The percentage of phagocytes that engulfed brain debris (pHrodo-debris) was calculated (Fig. 4A). In agreement with the conclusions provided by gene expression profiling, BEX increased the percentage of pHrodo-labeled debris–engulfing phagocytes in the aged brains (t₍₄₎ = 3.54; p = 0.024; unpaired t test; Fig. 4B).

Figure 4.

In vivo phagocytosis assay of MΦ. The pHrodo-labeled brain debris were stereotactically injected into the brains of aged (18–20 months) mice to assess the MΦ-mediated phagocytosis in vivo. To activate RXRs, BEX (5 mg/kg) or vehicle were intraperitoneally injected at 2, 24, and 48 h after injection. The mice were killed 3 d later, and the brains were processed for measuring phagocytosis with flow cytometry. A, The gating strategy of debris engulfed phagocytes. The engulfing phagocytes are gated as Live/Dead⁻ CD45^hi CD11b⁺ Ly6G⁻ and APC-debris⁺. B, Dot graphs demonstrate the engulfing phagocytes in the percentage of total phagocytes. The data are expressed as mean ± SEM (n = 3 mice in each group). *p = 0.024, unpaired t test.

Activating RXR in MΦ improves neurological deficits after ischemic stroke

We have established that BEX may reverse the age-related MΦ phenotype by suppression of their proinflammatory responses to ischemic injury and improving their phagocytic activities. Both pathways are crucial for the cleanup and healing process in the ischemic brain. Next, we aimed to test if activation of RXR in MΦ could help to improve poststroke recovery in the aging brain. To test this notion, we used the MCAo ischemic stroke model in aged (18–20 months old) mice. BEX administration was used to activate RXR. To single out the effects of RXR activation for MΦ, we used aged myeloid-specific RXRα knock-out mice (Mac-RXRα^−/−) that received either BEX or vehicle. To avoid the effect of BEX on hyperacute pathogenesis of stroke-induced damage, in all the animals, BEX or vehicle were intraperitoneally injected 24 h after the stroke onset and then once a day for 7 d. The survival of the mice was recorded, and we found a worse survival rate only in female Mac-RXRα^−/− mice in the vehicle-treated group (Fig. 5A,B).

Figure 5.

Aged female Mac-RXRα^−/− vehicle-treated mice had a significantly lower survival rate than other groups. The survival curves for (A) all, (B) female, and (C) male mice throughout the experiment. *p = 0.024 log-rank test. N = 15 in male RXRα^fl/fl BEX; n = 13 in male RXRα^fl/fl vehicle and female RXRα^fl/fl BEX; n = 12 in female RXRα^fl/fl vehicle; n = 10 in male Mac-RXRα^−/− vehicle and female Mac-RXRα^−/− BEX; n = 9 in male Mac-RXRα^−/−.

The severity of neurological deficit caused by MCAo was measured with foot fault and corner turn tests on Days 3, 7, 14, and 28 after the stroke, with both tests assessing loss of the sensorimotor function. The foot fault test showed that BEX-treated RXRα^fl/fl (control genotype) mice (male and female together) showed a robustly improved recovery compared with all the other experimental groups (Fig. 6A). The most insightful regarding the role of phagocytes in the recovery was that BEX provided improved recovery (foot fault test) only in mice that express RXR in MΦ (RXRα^fl/fl) but not in the Mac-RXRα^−/− (Fig. 6A–C), mice that are RXRα deficient in MΦ. Together with the fact that the reversal of the dysfunction was obvious only during the later recovery phase, our data may imply that indeed the phagocytosis-mediated cleanup processes orchestrated by RXRα in MΦ could be essential to optimize poststroke recovery. It is also important to note that RXRα proficiency in MΦ without its activation (no BEX) was insufficient for effective recovery, as Mac-RXRα^−/− and control mice showed similar recovery patterns. Furthermore, a subanalysis comparing foot fault missteps of females versus males (Fig. 6B,C, respectively) found similar recovery trends for both males and females. However, only RXRα^fl/fl females and not males statistically benefitted from BEX treatment.

Figure 6.

BEX improves neurological deficits recovery after ischemic stroke in an RXR-dependent pattern. Aged Mac-RXRα^−/− and RXRα^fl/fl (control genotype) mice were subject to a 1 h MCAo stroke model. BEX or vehicle was intraperitoneally injected 24 h after stroke onsite and then once a day for 7 d. Foot fault test results of all (A), female (B), and male (C) aged stroke mice from d3 to d28. Corner turn test results of all (D), female (E), and male (F) aged stroke mice. The data are mean ± SEM (n = 10 in male RXRα^fl/fl vehicle, female RXRα^fl/fl vehicle, male RXRα^fl/fl BEX, and female Mac-RXRα^−/− BEX; n = 12 in female RXRα^fl/fl BEX; n = 7 in male and female Mac-RXRα^−/− vehicle; n = 6 in male Mac-RXRα^−/− BEX). *p < 0.05 between vehicle and BEX-treated RXRα^fl/fl mice. Two-way ANOVA repeated measure followed by Fisher's LSD test.

The corner turn analysis of males and females pooled together showed a trend toward improved recovery with BEX in RXRα^fl/fl (control) mice (Fig. 6D). Interestingly, despite the lack of beneficial effect from BEX treatment in aged male mice (Fig. 6F), the treatment with BEX improved the RXRα^fl/fl performance in aged female RXRα^fl/fl mice at Day 14 after the stroke (Fig. 6E). Similar to the foot fault test, the beneficial effects of activating RXR (BEX) were absent in Mac-RXRα^−/− mice, in which the MΦ does not express RXRα. These results suggest that the beneficial effects come from RXR signaling in MΦ.

BEX prevents cognitive declines in RXRα-proficient aged mice after stroke

Cognitive functions are often impaired in aging subjects, and stroke in these subjects may promote further cognitive decline. Thus, we assessed the cognitive functions using a comprehensive battery of tests performed 21 d after the stroke onset to avoid confounders introduced by the surgery itself.

The open-field test was first performed to assess the mice's anxiety level and locomotive function, which could represent the confounders in performing the cognitive and behavioral tests. The results showed no difference in mobility among all the groups (Fig. 7A). The working memory was assessed using Y maze spontaneous alternation. The results showed no difference among all four groups (Fig. 7B), suggesting that RXR signaling does not affect working memory in aged brains after stroke.

Figure 7.

Activating RXR prevents short-term memory loss after ischemic stroke in the aged brain without affecting mobility or working memory. A, The open-field test was performed to assess the mobility and anxiety levels of the aged mice after ischemic stroke. Total distance traveled in the open-field arena of all aged mice. B, Y maze spontaneous alternation assessing working memory on Day 29 after stroke. The mice were given 5 min to explore the Y-shaped arena ad libitum. The order of the mice entering each arm was recorded, and the alternation rate was calculated. C, Novel object recognition test assessing short-term (1 h) memory grouped. The data are mean ± SEM. *p < 0.05. Two-way ANOVA followed by Fisher's LSD multiple-comparison test.

The short-term (1 h) memory was assessed by the novel object recognition test on Day 30 after the ischemic stroke onset. The results suggest that BEX treatment may prevent short-term memory loss in the aged brain (F_(1,44) = 9.26; p = 0.004; two-way ANOVA; Fig. 7C). Fisher's LSD multiple-comparison tests revealed that BEX improved short-term memory performance after stroke in RXRα^fl/fl (control genotype) mice (p = 0.017). Despite the strong trend toward improved memory in the Mac-RXRα^−/− mice, the beneficial effects were not significant (p = 0.062).

The Barnes maze was used to assess the aged mice's spatial learning and long-term (72 h) memory after ischemic stroke (Fig. 8). The heatmap visualized the traces of all mice within the same treatment group for each day's training session (Fig. 8A), showing that BEX-treated RXRα^fl/fl mice performed better than all other groups in learning to locate the goal box with visual cues in that the traces converged around the goal box progressively. Further analyzing their search strategies, we found that random search (indicating no learned strategy) is the least used (p = 0.026; χ² test for trend) strategy in the BEX-treated RXRα^fl/fl group on the last day of training (23 d after the stroke onset; Fig. 8B), suggesting the superior spatial learning ability in BEX-treated RXRα^fl/fl mice. Also, BEX-treated RXRα^fl/fl mice traveled the shortest distance of all groups to locate the goal box on the fourth-day training session (Fig. 8C). All suggest that BEX improved spatial learning in RXRα^fl/fl, but not Mac-RXRα^−/− mice after ischemic stroke.

Figure 8.

Spatial learning and long-term memory test after ischemic stroke. Starting from Day 21 postischemic stroke, each mouse was given four training sessions per day for 4 consecutive days to learn to use visual cues in the arena to locate the goal box. A, The heatmap visualizes the cumulative traces of all animals per group each day. B, The proportion of each search strategy the mice used in each day's training to locate the goal box. The search strategies were classified into three categories: random, serial, and reference search. Random search defines the mice searched randomly. Serial search defines the mice searching the holes one by one. Reference search defines the mice used visual cues to orient in the arena. C, The total distance the mice traveled before reaching the goal box in the first training session of each day's training. D, Heatmap showing the cumulative traces of each group during the probe session. The probe session took place 72 h after the last training session. E, Total time it took the mice to reach the goal box during the probe session. The data are mean ± SEM (n = 14 RXRα^fl/fl vehicle; n = 16 RXRα^fl/fl BEX; n = 10 Mac-RXRα^−/− vehicle; n = 12 Mac-RXRα^−/− BEX). *p < 0.05. Two-way ANOVA followed by Fisher's LSD test. F, Immunofluorescence of NeuN⁺ neurons (green) in the hippocampal CA1. G, Correlation analysis of primary latency in novel object recognition test (x-axis) and neuronal density in hippocampal CA1 (y-axis). Pearson's correlation coefficient, r₍₁₄₎ = −0.526. p = 0.0366. N = 16 mice.

Seventy-two hours after the last training session, the mice were tested once in a probe session to assess their long-term memory. Since the long-term memory was assessed in the probe session, the mice that failed to locate the goal box within 30 s in the last training session, indicating these mice did not form the spatial memory required for the test, were excluded from the probe session. We found that 2–3 mice from each group failed to learn to locate the goal box after 16 training sessions. The failure rate is similar among all groups (p = 0.99; χ² test). For the probe session, BEX improved long-term memory after stroke in the aged RXRα^fl/fl mice. BEX-treated RXRα^fl/fl mice spent a shorter time locating the goal box than vehicle-treated RXRα^fl/fl mice. This beneficial effect was not observed in Mac-RXRα^−/− mice (Fig. 8D,E). To get more insight into relevance of better performance in memory testing, in a subgroup of animals, we used immunofluorescence-based analysis to establish the relationship between neuronal integrity (NeuN⁺ neurons count) in the hippocampal CA1 region, 31 d after stroke (Fig. 8F) and cognitive performance. The analysis revealed a correlation (r₍₁₄₎ = −0.526; p = 0.0366) between primary latency in probe session and neuron density in CA1 (Fig. 8G). This may suggest that the retention of long-term spatial memory, shorter time to reach the goal box mice, which was achieved with BEX treatment could be related to higher neuron density in CA1.

Taken together, we demonstrated that the RXR signaling in MΦ prevents cognitive decline in aged mice after ischemic stroke. Intriguingly, aged female mice benefited more from activating MΦ’ RXR signaling than aged male mice.

RXR signaling in myeloid cells is important in promoting prophagocytic and reparative MΦ in the brain after stroke

To better understand the processes associated with activating RXR in MΦ after stroke, we also analyzed the immunophenotype of MΦ in the perilesional brain 31 d after ischemic stroke, conducting histological analyses. In the peri-infarct area of the stroke-affected brain, the number of MΦ showing phagocytic/reparative markers, including CD163 (Fig. 9A,C), CD206 (Fig. 9B,E), and CD36 (Fig. 9D), increased only in BEX-treated RXRα^fl/fl (RXRα-proficient) mice. These results suggest activating RXR in MΦ promotes their phagocytic/reparative phenotype in the aged ischemic-stroked brain.

Figure 9.

BEX promotes MΦ with phagocytic/reparative markers in the aged brain after ischemic stroke. Immunohistochemistry of MΦ expressing phagocytic markers CD163 (A) and CD206 (B) in the peri-infarct area 31 d after ischemic stroke. Quantitative results of immunohistochemistry labeling (C) CD163, (D) CD36, and (E) CD206 colocalized with MΦ marker CD11b/Iba-1 in the peri-infarct area. The data are mean ± SEM (n = 3 in each group). *p < 0.05. Two-way ANOVA followed by Sidak's multiple-comparison test.

Discussion

Our study demonstrates the important role of RXRα in myeloid phagocytes in improving poststroke recovery in the aging brain. With myeloid-specific RXRα knock-out mice, we showed that BEX, via activating nuclear receptor RXRα selectively in myeloid phagocytes, improves poststroke recovery through reversing/offsetting many of MΦ age-associated dysfunctions. Our study did not intend to distinguish microglia from hematogenous macrophages (MDM) in the brain, as both microglia and MDM (together, we refer to as MΦ) are recruited into the ischemia-injured brain and contribute to the poststroke recovery through mechanisms involving a variety of reparative functions.

Based on an extensive body of work, one of the hallmarks of the MΦ phenotype in the aging brain is their proinflammatory status (age-priming) that often leads to overexaggerated immune responses to pathological stimulation (Sierra et al., 2007; Norden et al., 2015). MΦ overactivation (characterized by secretion of proinflammatory cytokines and free radicals and reduced production of trophic factors) and inability to effectively clear cellular debris in the aging brain are the important contributing factors to age-related degenerative diseases of the central nervous system, including Alzheimer's disease (Hickman et al., 2008; Hearps et al., 2012; Gabande-Rodriguez et al., 2020). Our transcriptomic analysis of aged MΦ corroborates the proinflammatory “age-primed” state of MΦ isolated from the aged brains in that the elevated expression in the gene set of the innate immune response, including Mpeg1, chemokines (Ccl3, Ccl4, and Ccl12), and interferon-induced proteins (Ifit3, Ifit2, and Ifit1). Furthermore, we observed the expression of signature genes for DAM/primed MΦ, such as Cst7, Clec7a, Axl, Itgax, and Lpl, elevated in aged MΦ. Many of these DAM signature genes may lead to facilitation of MΦ-mediated phagocytosis (e.g., Axl, Cst7, Clec7a) and lipid metabolism (Lpl) functions. Intriguingly, despite the phagocytic nature of age-primed MΦ phenotype (Keren-Shaul et al., 2017), we observed the decreased in expression of genes that encode scavenger receptors (Cd206, Cd163, and Cd36) that facilitate phagocytosis in the MΦ from the aged brain. Besides the reduction of expression of some genes involved in promoting phagocytosis, Cd22, a negative regulator of phagocytosis (Pluvinage et al., 2019), was found to be upregulated in the aged MΦ, suggesting that some of the phagocytic functions of aged MΦ may indeed be compromised. We also demonstrate that BEX promotes the expression of gene sets regulating apoptotic cell clearance in aged MΦ, including Tgm2, Timd4, Anxa1, and Cd36. These results suggest that activating RXR in aged MΦ may enhance their ability to clean up the environment in the aging brain.

A unique population of MΦ loaded with lipids in the aging brain has been identified (Shimabukuro et al., 2016; Marschallinger et al., 2020). These lipid-laden MΦ produce higher levels of oxidative stress/proinflammatory cytokines and show compromised ability in conducting phagocytosis. The accumulation of lipids in aged MΦ could result from their defective cholesterol efflux ability (Cantuti-Castelvetri et al., 2018). Indeed, over 20% of total cholesterol contents are in the central nervous system in the form of cell membranes and myelin (Hamilton et al., 2007). Myelin itself contains ∼50% of cholesterol (Poitelon et al., 2020). In the events of aging and brain injury (such as ischemic stroke) that leads to demyelination, the burden on MΦ to clear cholesterol and other lipid components derived from apoptotic/necrotic cells, and myelin is increased. Elevating cholesterol inside MΦ activates the nuclear receptor liver X receptors (LXRs), a dimer partner of RXR. It promotes the transcription of genes that facilitate cholesterol efflux, including ATP-binding cassette transporters (ABCs; Zhao and Dahlman-Wright, 2010; Daniel et al., 2014). Our transcriptomic analysis demonstrates that gene sets facilitating cholesterol efflux (Abca1, Abcg1, and Eepd1) are upregulated by RXR agonist BEX in both young and aged MΦ. These pathways have beneficial effects on reducing lipid accumulation in macrophages and suppressing their inflammatory responses (Claudel et al., 2001; Joseph et al., 2002; Zhou et al., 2019). Another dimer partner of RXR, the nuclear receptor peroxisome proliferator-activated receptors (PPARs), also plays a crucial role in reducing MΦ lipid accumulation (Ye et al., 2019). PPAR:RXR regulates genes involved in peroxisomal/mitochondrial β-oxidation that breaks down long-chain fatty acid in MΦ and releases energy (Plutzky and Kelly, 2011). Our data also confirm that BEX promotes gene sets of fatty acid oxidation (Cpt1a, Etfbkmt, and Acaa2), a pathway that could redirect MΦ toward alternative activation and reduce lipid accumulation (Namgaladze et al., 2014; Namgaladze and Brüne, 2016). Taken together, the results demonstrate that activating RXR promotes pleiotropic signaling pathways that reverse MΦ age-primed phenotypes in the aging brain.

An underlying reason for the poor prognosis of ischemic stroke in the aging brain may include the age-primed phenotypes of MΦ that exacerbate proinflammatory responses to ischemic injury, which in turn increases secondary brain injury and hinders poststroke recovery (Badan et al., 2003; Finger et al., 2022; Arbaizar-Rovirosa et al., 2023). Our transcriptomic analysis shows that activating RXR reduces the aged MΦ phenotype and their proinflammatory responses to ischemic stroke. In agreement with the results mentioned above, LXR-controlling genes that facilitate cholesterol efflux (Abcg1 and Abca1) were upregulated in response to BEX, which is given 24 h after an ischemic stroke in the aged brain. Moreover, BEX downregulates the gene sets of inflammatory responses in MΦ in the ischemia-affected brain, including leukocyte activation (Stap1, Tnf, and Batf2), proinflammatory cytokine production (Gbp5, Il1b, and Myd88), and lymphocyte recruitment (Ccl8, Cxcl12, and Ccl7). Together, we demonstrate that pharmacological activation of RXR with BEX can partially reverse age-associated MΦ phenotypes and suppress the reactivity of MΦ in the ischemic-affected brain.

Our results of gene expression profiles show that BEX promotes the expression of a gene set of apoptotic cell clearance in aged MΦ. These findings were further confirmed using a functional assay measuring the phagocytic ability of MΦ and its response to BEX in the aged brain. BEX treatment was capable of promoting MΦ phagocytic ability in both young and aged brains. The brain debris was used in this experiment as phagocytic targets to model the poststroke clearance of the brain by MΦ in the ischemia-injured brain. The used of pH-sensitive dye (pHrodo)-labeled brain debris enhanced the specificity to detect phagocytosis, as the pHrodo-labeled debris is only fluorescent in the acidic environment, such as these existing in lysosomes. One limitation of this approach was that this assay only allows for establishing the percentage of all MΦ that are involved in debris engulfing instead of testing the phagocytic capacity (amount of engulfment) of individual MΦ. Thus, our data could only demonstrate that more MΦ is recruited to perform cleanup in response to BEX.

The corner turn test and foot fault test are reliable behavioral tests to assess long-term sensorimotor recovery after ischemic stroke (Zhang et al., 2002; Encarnacion et al., 2011). Both tests showed beneficial effects of BEX only in RXRα-proficient mice. Intriguingly, we found that females benefit more from the treatment with BEX in neurological recovery. Sex dimorphisms in ischemic stroke have been well documented in both clinical and animal studies (Reeves et al., 2008; Liu et al., 2009; Haast et al., 2012). Multiple factors contribute to the sex dimorphisms in stroke outcomes and recovery, including inflammatory responses (Offner et al., 2009; Villa et al., 2018). Sex steroid hormones also play a pivotal role in the sex discrepancy after stroke. Literature showed that sex hormones such as estrogen protect young female mice from ischemic stroke, and this protection is no longer present in aged female mice (Manwani et al., 2013). As a nuclear receptor, RXR:PPAR could activate estrogen-responsive elements (Keller et al., 1995; Nuñez et al., 1997). This could be a potential reason that activating RXR is more beneficial in aged female mice than in aged male mice.

Cognitive decline is a hallmark of the aging process (Brito et al., 2023; Healy et al., 2024). The pathology of stroke, including brain atrophy and inflammation, further impairs cognition in aged subjects (Mattson and Arumugam, 2018; Aamodt et al., 2023). Therefore, assessing poststroke cognitive functions is particularly important in aged mice. We utilized three different cognitive tests to assess an extended scope of learning and memory performance. These tests are also moderate in physical demands and do not introduce distress, such as electrical shock or swimming, to aged mice. We showed that activating RXR with BEX prevents the stroke-induced declines in cognitive functions, specifically short-term and long-term memory, in aged RXRα^fl/fl (control genotype expressed normal RXRα in MΦ) mice. The beneficial effects of BEX diminished in myeloid-specific RXRα knock-out mice, indicating the importance of MΦ’ RXRα in improving poststroke recovery.

Literature suggests that over-reactivity of the aged-associated MΦ to immune challenges of the aging brain leads to excess proinflammatory cytokines (particularly IL-1) production in response to ischemic stroke, which impairs cognitive functions (Norden and Godbout, 2013; Patterson, 2015). As shown in our transcriptomic analysis, RXR activation suppresses MΦ inflammatory response to ischemic stroke. Furthermore, BEX treatment promoted the phagocytic/reparative phenotype of MΦ in an RXRα-dependent manner. These could be the underlying mechanisms that explain why BEX treatment improves cognitive performance in the aging brain after stroke.