The key genes, phosphoproteins, processes, and pathways affected by efavirenz-activated CYP46A1 in the amyloid-decreasing paradigm of efavirenz treatment

Abstract

Efavirenz (EFV) is an anti-HIV drug, and cytochrome P450 46A1 (CYP46A1) is the major brain cholesterol hydroxylase. Previously, we discovered that EFV activates CYP46A1 and improves behavioral performance in 5XFAD mice, an Alzheimer’s disease model. Herein, the unbiased omics and other approaches were used to study 5XFAD mice in the amyloid-decreasing paradigm of CYP46A1 activation by EFV. These approaches revealed increases in the brain levels of postsynaptic density protein 95, gephyrin, synaptophysin, synapsin, glial fibrillary acidic protein, and CYP46A1 and documented altered expression and phosphorylation of 66 genes and 77 proteins, respectively. The data obtained pointed to EFV effects at the synaptic level, plasmin-depended amyloid clearance, inflammation and microglia phenotype, oxidative stress and cellular hypoxia, autophagy and ubiquitin-proteasome systems as well as apoptosis. These effects could be realized in part via changes in the Ca²⁺-, small GTPase, and catenin signaling. A model is proposed, in which CYP46A1-dependent lipid raft rearrangement and subsequent decrease of protein phosphorylation are central in EFV effects and explain behavioral improvements in EFV-treated 5XFAD mice.—Petrov, A. M., Mast, N., Li, Y., Pikuleva, I. A. The key genes, phosphoproteins, processes, and pathways affected by efavirenz-activated CYP46A1 in the amyloid-decreasing paradigm of efavirenz treatment.

Cytochrome P450 46A1 (CYP46A1) is a key enzyme for cholesterol homeostasis in the brain. CYP46A1 mainly resides in neurons and converts cholesterol to 24S-hydroxycholesterol (24HC), a transport form of cholesterol from the brain to the systemic circulation (1, 2). About 75–85% and 40–50% of cholesterol excess is eliminated via cholesterol 24-hydroxylation from human and mouse brain, respectively (3, 4). Cholesterol 24-hydroxylation is tightly linked to local cholesterol biosynthesis (3), the major source of brain cholesterol (5), and thereby controls brain cholesterol turnover (3, 6). 24HC is not only a cholesterol elimination product but also a biologically active molecule. 24HC is a potent activator of liver X receptors and a positive allosteric modulator of NMDA receptors (7–9). Liver X receptors are important transcription factors (10–12), whereas NMDA receptors mediate excitatory transmission throughout the CNS and are involved in memory and learning (13). In addition, CYP46A1 activity or expression levels could affect protein prenylation and protein phosphorylation (14–16). Protein prenylation is linked to CYP46A1 via cholesterol biosynthesis and availability of the nonsterol intermediates used for prenylation (14, 16). Protein phosphorylation could be altered by CYP46A1 (15) via the enzyme-dependent flux of cholesterol and 24HC through plasma membranes and lipid rafts, which serve as signaling platforms for activation of different protein kinases (17–20).

Clinical significance of CYP46A1 is highlighted by the association between plasma 24HC levels and Alzheimer’s disease (AD) (21–23) as well as changes in CYP46A1 expression pattern in the brain under different pathologic conditions (24–28). In addition, some but not all genetic studies linked CYP46A1 intronic polymorphisms to AD (29). Genetic modulation of Cyp46a1 expression in mice supports the CYP46A1-AD link and suggests that CYP46A1 should not only be considered as a target for AD but also Huntington disease and conditions of neuronal sclerosis or epileptic activity (30–35).

Previously, we identified compounds that can enhance CYP46A1 activity pharmacologically and tested one of them, the anti-HIV drug efavirenz (EFV), on C57BL/6J mice (6). We found that EFV exerted a dual effect on CYP46A1; it activated this enzyme and enhanced brain cholesterol turnover at a low dose of 0.1 mg/kg of body weight yet inhibited the P450 at the higher doses (6). Subsequent experiments with purified CYP46A1 suggested that at a low dose EFV likely binds to the CYP46A1 allosteric site, which is away from the enzyme active site, and makes cholesterol 24-hydroxylation in the active site more efficient. However, at high doses, EFV probably binds to both the CYP46A1 allosteric and active sites and inhibits CYP46A1 because of the competition with cholesterol for binding to the active site (6). We then tested the activating EFV dose on 5XFAD mice (36), a model of AD, in which amyloid deposition begins at 2 mo of age and behavioral deficits start to develop at 4 mo of age (37).

Two treatment paradigms (Fig. 1A), both using the same drug dose of 0.1 mg/kg of body weight, have been evaluated so far. In the first treatment paradigm (1TP), 5XFAD mice received EFV from 1 to 9 mo of age and were assessed after 4 and 8 mo of EFV treatment (i.e., at 5 and 9 mo of age, respectively) (36). There was a decrease in the amyloid β (Aβ) burden at both times of evaluation and impaired behavioral performance after 4 mo of EFV treatment, but there was improved behavioral performance after 8 mo of drug treatment. Different effects on behavioral performance were attributed to the brain cholesterol depletion during the first 3 mo of drug dosing when EFV was administered to young mice whose brains and a pool of brain cholesterol were still expanding (36). In the second treatment paradigm (2TP), EFV was given to mice from 3 to 9 mo of age, and animals were only evaluated at the end of treatment (i.e., at the age of 9 mo) (38). The overall Aβ burden in EFV-treated 5XFAD mice remained unchanged; nevertheless, behavioral performance was improved (38). Herein, we conducted additional characterizations of 9-mo-old 5XFAD mice from the 1TP so that we can compare the following: 1) the 1TP and 2TP for the levels of synaptic and other proteins in mice of the same age of 9 mo (Fig. 1B); 2) EFV effects on the brain gene expression after 4 and 8 mo of treatment (Fig. 1C); and 3) EFV effects on the brain phosphoproteome of EFV-treated mice with that of the age-matched Cyp46a1^−/− mice (Fig. 1D). Collectively, these comparisons identified the biologic processes and proteins involved in these processes that were affected in 5XFAD mice by EFV treatment. This study also pointed to the processes and pathways not yet considered in the context of CYP46A1 activation.

Figure 1.

Study design. A) Previous EFV treatments of 5XFAD mice. Magnifying glasses indicate evaluation points. A brief summary of the results is in colored boxes. ↑, increase, ↓, decrease, and ↔, no change. B) Schematic location in the excitatory (glutamatergic) and inhibitory (GABAergic) axodendritic synapses of the proteins (calbindin, gephyrin, Munc13-1, PSD-95, synapsin-1, and synaptophysin, in bold) whose levels were measured in the present work in 9-mo-old 5XFAD mice from the 1TP. These quantifications were then compared with those conducted previously in 9-mo-old 5XFAD mice from the 2TP. C) The characterization of the brain transcriptome of 9-mo-old 5XFAD mice from the 1TP. The data obtained were then compared with EFV effects on the brain transcriptome of 5-mo-old 5XFAD mice from the same treatment paradigm. Gene expression was assessed by RNA-Seq. D) Studies of the brain phosphoproteome of 9-mo-old male 5XFAD mice from the 1TP and a subsequent comparison with the brain phosphoproteome of 9-mo-old Cyp46a1^−/− (KO) and wild-type (WT) female mice. Cntr, control 5XFAD mice; P, phosphorylation; Tx, EFV-treated 5XFAD mice.

MATERIALS AND METHODS

Animals

5XFAD mice were hemizygotes for the mutant human amyloid precursor protein 695 and mutant human presenilin 1 protein. Amyloid precursor protein contained the Swedish (K670N, M671L), Florida (I716V), and London (V717I) familial AD mutations. Presenilin 1 harbored the M146L and L286V familial AD mutations. 5XFAD hemizygous mice were obtained by crossing wild-type B6SJL female mice (100012; The Jackson Laboratory, Bar Harbor, ME, USA) and hemizygous 5XFAD male mice (34840-JAX; The Jackson Laboratory). Only F1 generation of hemizygous animals was used. These mice were free of the Pde6b^rd1 mutation, which leads to retinal degeneration and blindness and is present in the B6SJL strain. The Pde6b^rd1 mutation was bred out of our colony. Animals were kept on a 12-h light/dark cycle and provided food and water ad libitum. All animal experiments were approved by the Case Western Reserve University’s Institutional Animal Care and Use Committee and conformed to recommendations of the American Veterinary Association Panel on Euthanasia. Cyp46a1^−/− mice were obtained from Dr. D. Russell (University of Texas Southwestern, Dallas, TX, US) and were on the mixed C57BL/6J;129S6/SvEv background.

EFV treatment

The S-isomer of EFV (Toronto Research Chemicals, Toronto, ON, Canada) was used and administered to 5XFAD mice as previously described in Mast et al. (6). Briefly, EVF stock (200 mg/L) was prepared by dissolving the drug in water containing 0.2% (v/v) Tween 80 (MilliporeSigma, Burlington, MA, USA). This stock was then diluted into drinking water at a final concentration of 0.42 mg/L. EFV-containing water was placed in light-protected bottles and given to mice. Mice consumed ∼6.0–6.5 ml of water per day, which equaled to about 0.1 mg of EFV/day/kg body weight. EFV treatment started when animals were 1 mo old and continued for 8 mo. Control mice received the vehicle (0.0004% aqueous Tween 80). At the end of treatment, mice were not fed overnight and were euthanized the next morning.

Western blots

Brains were isolated and dissected along the midline. One hemisphere from each brain was used for homogenization (10%, w/v) in 20 mM Tris-HCl, pH 7.4, containing 250 mM sucrose, 0.5 mM EDTA, 0.5 mM EGTA, and a cocktail of protease inhibitors (Complete; Roche, Basel, Switzerland). Brain homogenates were then subjected to centrifugation at 1500 g for 15 min, and the supernatants obtained were used for protein separation (10–20 µg protein per lane) by SDS-PAGE on 10 or 4–20% Tris/Glycine gels (Bio-Rad, Hercules, CA, USA). After separation, proteins were transferred on the nitrocellulose membranes (P/N 926–31092; Li-Cor Biosciences, Lincoln, NE, USA) followed by membrane blocking in the Odyssey blocking buffer (927–40000; Li-Cor Biosciences) for 1.5–3 h at room temperature. Incubations with primary antibody (overnight at 4°C) and secondary antibody (1 h at room temperature) were in the Odyssey blocking buffer supplemented with 0.1% Tween-20. The source, clonality, and dilutions of primary and secondary antibodies are indicated in Supplemental Table S1. The intensity of fluorescent bands on the nitrocellulose membranes was quantified by an Odyssey infrared imaging system (Li-Cor Biosciences).

Whole transcriptome sequencing

This was carried out as previously described in Mast et al. (36). Briefly, each sample (a total of 3 biologic replicates per group) was prepared from a brain hemisphere, which was homogenized (10%, w/v) in Trizol reagent (Thermo Fisher Scientific, Waltham, MA, USA). Total RNA was then isolated, and individual libraries were prepared from each RNA sample; these libraries were tagged with a unique adapter index. The libraries were next pooled together and run on a HiSeq 2500 System (Illumina, San Diego, CA, USA). Reads that passed the quality filters were aligned to the mouse mm10 reference genome (Genome Reference Consortium mouse build 38; https://www.ncbi.nlm.nih.gov/grc) using the TopHat software (39); these alignments were analyzed by the Cufflinks RNASeq analysis package (40). Differentially expressed genes between the control and treatment groups were identified using a value of q (adjusted P value for multiple testing) ≤ 0.05.

Real-time quantitative PCR

Total RNA (1 µg) was converted to cDNA by SuperScript III Reverse Transcriptase (Thermo Fisher Scientific) according to the manufacturer’s protocol. PCR reactions (performed in triplicate) were carried out using 2 μl of cDNA, a pair of gene-specific primers (Supplemental Table S2), and a FastStart Universal SYBR Green Master (Rox) (Roche Life Science, Indianapolis, IN, USA). The reaction volume was 25 μl, and the final concentration of the primers was 1 μM. Changes in the relative mRNA levels were calculated by the 2^−ΔΔCt method (41) after the normalization of gene expression to the expression of β-actin.

Brain phosphoproteomics

This was carried out as previously described in Mast et al. (15). Briefly, each sample (a total of 3 biologic replicates per group) was prepared from a brain hemisphere, which was lysed in 20 mM HEPES, pH 8.0, containing 9 M urea and 1X Halt phosphatase inhibitor cocktail (Thermo Fisher Scientific). Brain lysates were digested with trypsin following protein reduction with DTT and alkylation with iodoacetamide. Trypsin digests were spiked with internal standards and enriched with phosphoserine- and phosphothreonine-containing peptides. Enriched digests were cleaned-up on C18 cartridges (360 mg, WAT05190; Waters, Milford, MA, USA) and applied to a Dionex Acclaim Pepmap C18 reversed-phase capillary chromatography column connected to a Finnigan LTQ-Obitrap Elite hybrid mass spectrometer (both are from Thermo Fisher Scientific). The digest was analyzed using the data-dependent multitask capability of the instrument acquiring both full scan mass spectra and product ion spectra. The data were searched by Sequest in Proteome Discoverer 1.4 (Thermo Fisher Scientific) against the reference mouse databases. The quantification of the identified phosphorylated peptides was performed by a label-free method. Data were normalized to the total ion chromatogram. All differentially expressed phosphopeptides were manually validated using a second normalization to the synthetic phosphopeptides added as internal standards.

Prediction of protein kinases

Protein kinases that could phosphorylate serine, threonine, or tyrosine amino acid residues in the identified phosphorylation sites were predicted by Netphos 3.1 (42). These predictions were made for the following 17 kinases: ATM, CKI, CKII, CAMK2, DNAPK, EGFR, GSK3, INSR, PKA, PKB, PKC, PKG, RSK, SRC, CDC2, CDK5, and p38 MAPK. Only the kinases with the prediction score of ≥0.50 indicative of positive predictions were then considered.

Statistical analysis

All data represent means ± sd. Statistical analysis (except the omics approaches) was by unpaired Student’s t test assuming a 2-tailed distribution. Results were considered statistically significant when values were P ≤ 0.05. All Western blots (repeated at least 2 times) are representative of observations in 4–5 mice/group.

RESULTS

Brain levels of synaptic and other proteins

The levels of 6 synaptic proteins were assessed by Westerns blots (Fig. 2A–F), the same 6 proteins that were measured in 9-mo-old mice from the 2TP (38). EFV treatment did not affect protein expression of protein Unc-13 homolog A (Munc13-1; a presynaptic component) (43) and calbindin (a neuron-specific Ca²⁺ buffering protein) (44) but increased the expression of PSD-95 (a postsynaptic scaffold protein in excitatory/glutamatergic synapses) (45), gephyrin (a postsynaptic scaffold protein in inhibitory synapses) (46), synaptophysin (the most abundant protein in synaptic vesicles) (47), and synapsin-1 (one of the main anchor protein connecting synaptic vesicles in vesicular pools) (48). Thus, similar to the 2TP, the 1TP also changed the levels of some of the synaptic proteins. However, these changes were not identical in the 2 treatment paradigms with only PSD-95 being increased in both (Fig. 2G). The 2 treatments were also different in EFV effects on the levels of glial fibrillary acidic protein (GFAP) (a marker of astrocyte activation), Iba1 (a marker of microglia activation), and CYP46A1 (Fig. 3A–C). The levels of GFAP and CYP46A1 were increased in 9-mo-old mice from the 1TP, whereas those of Iba1 were unchanged. Like with synaptic proteins, these were different changes as compared with the 2TP with only GFAP levels being increased in both treatments (Fig. 3D). A decreased (1TP) or unchanged (2TP) amyloid load in the 2 treatment paradigms could possibly underlie differential effects of EFV on protein abundance in the 5XFAD brain.

Figure 2.

EFV effects on the levels of synaptic proteins in 9-mo-old 5XFAD mice from the 1TP. A–F) Representative Western blots of brain homogenates from control (Cntr, n = 4) and EFV-treated (Tx, n = 5) male mice. Each lane, except those with MW markers, represents a sample from an individual animal. G) The quantification of the relative protein expression in A–F after the normalization to the β-actin signal. Light and dark green bars indicate relative protein levels in control and EFV-treated 5XFAD mice, respectively, from the 1TP. Light and dark magenta bars indicate relative protein levels in Cntr and Tx mice from the 2TP, which is shown for comparison. Data for the 2TP from (38). The results are presented as means ± sd of measurements in individual mice. Asterisks are significant changes relative to the corresponding control mice as assessed by a 2-tailed, unpaired Student’s t test. *P ≤ 0.05, **P ≤ 0.01.

Figure 3.

EFV effect on the levels of GFAP, Iba1, and CYP46A1 in 9-mo-old 5XFAD mice from the 1TP. A–C) Representative Western blots of brain homogenates from control (Cntr, n = 4) and EFV-treated (Tx, n = 5) male mice from the 1TP. Each lane, except those with MW markers, represents a sample from an individual animal. D) The quantification of the relative protein expression in A–C after the normalization to the β-actin signal. Light and dark green bars indicate relative protein levels in control and EFV-treated 5XFAD mice, respectively, from the 1TP. Light and dark magenta bars indicate relative protein levels in Cntr and Tx mice from the 2TP, which is shown for comparison. Data for the 2TP from (38). The results are presented as means ± sd of measurements in individual mice. Asterisks are significant changes relative to the corresponding control mice as assessed by a 2-tailed, unpaired Student’s t test. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 by a 2-tailed, unpaired Student’s t test.

Brain gene expression

Brain gene expression was assessed by whole transcriptome sequencing (RNA-Seq), an approach used previously for the characterization of 5-mo-old 5XFAD mice from the same treatment paradigm (32). More than 27,670 genes were detected, of which 66 genes had at least a 2-fold difference (an arbitrary cutoff limit, q ≤ 0.05) in the expression between EFV-treated and control groups (Fig. 4A, B). Of these 66 differentially expressed genes, 40 genes were down-regulated up to 8.4-fold; 21 genes were up-regulated up to 8.5-fold, and the expression of the remaining 5 genes was found only in 1 group: either in EFV-treated mice (Trac, 2010001K21Rik, and Gm12976) or control mice (Gm2800 and Gm38360). Differentially expressed genes were next grouped by process, and changes in their expression were analyzed for the positive effects on AD. Thirty-six such changes were found, out of which 21 were next assessed by real-time quantitative PCR (Fig. 4C). A change in expression was confirmed for 13 genes (Angptl4, Ccr5, Egr2, Gbp5, Hif3a, Iigp1, Itgax, Lilrb4, Maff, Plau, Plin4, Serpine1, Xdh), whereas the remaining genes showed either unaltered expression (Arc, Erbb3, Hspa8, Npas4, and Rsad2) or the opposite direction of change (C3, Gpnmd, and Clec7) as compared with the RNA-Seq data. A summary of the RNA-Seq and real-time quantitative PCR data was then generated (Table 1).

Figure 4.

EFV effects on the brain gene expression in 5XFAD mice. A, B) gene expression in 9-mo-old 5XFAD mice from the 1TP as assessed by RNA-Seq. Biologic replicates represent individual animals (n = 3 per group, all male mice and littermates). The volcano plot (A) of changes in the gene expression in EFV-treated (Tx) vs. control (Cntr) 5XFAD mice. The vertical dashed lines are the boundaries of a 2.0-fold change (Log₂ of 1, an arbitrary cutoff) in the gene expression; the horizontal dashed line is the boundary of the value of P = 0.05. Genes (small circles) with ≥2.0-fold change in expression and q ≤ 0.05 are colored in green (the down-regulated genes) and magenta (the up-regulated genes). The heat map (B) of the up-regulated (↑, boxed region above the horizonal black line) and down-regulated (↓, boxed region below the horizonal black line) genes in Tx mice with the values of P and q ≤ 0.05. Each row and column correspond to the same gene and biologic replicate, respectively. Rows are scaled and represent Z scores (the deviation from the mean by the sd units): negative Z scores are shown in green for both Cntr and Tx mice and positive Z scores are shown in blue for Cntr and orange for Tx mice, see the color code at the panel bottom. The heat map was generated by the Heatmapper software. C) The quantifications by real-time quantitative PCR of some of the differentially expressed genes in 9-mo-old 5XFAD mice from the 1TP. The results are presented as means ± sd of the measurements in individual animals (n = 4 per group, all male mice and littermates). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 by a 2-tailed, unpaired Student’s t test. D) A Venn diagram comparing the number (in ovals) of differentially expressed genes in Tx vs. Cntr 5XFAD mice from the 1TP at 5 and 9 mo of age. These are the up- (↑) and down- (↓) regulated genes with at least a 2-fold difference in expression and the values of P and q being ≤0.05 if characterized by RNA-Seq only, and statistically significant changes (P ≤ 0.05) in expression, if characterized by both RNA-Seq and real-time quantitative PCR. C3 was the only common gene between 5- and 9-mo-old 5XFAD mice, whose expression was decreased as a result of EFV treatment. For each age group, a brief summary of EFV effects on the levels of Aβ and Iba1 as well as behavioral performance is shown at the panel bottom; ↑, ↓, and ↔ indicate increase, decrease, and no change, respectively.

TABLE 1.

EFV effects on the brain gene expression in 9-mo-old 5XFAD mice from the 1TP

Processes and conditions	↓Tx/Cntr, positive effect of gene down-regulation	↑Tx/Cntr, positive effect of gene up-regulation
Aß/APP degradation	Serpine1 (0.5)	Plau (2.1)
Inflammation and microglia phenotype	C3 (0.3), Fkbp5 (0.4), Gbp5 (0.5), Iigp1 (0.2), Zbtb16 (0.5)	Ccr5 (1.6), Itgax (2.2), Lilrb4 (1.7)
Lipid metabolism	Angptl4 (0.7), Plin4 (0.3)
Myelinization		Egr2 (1.4), Nkx6-2 (2.1)
Hypoxia	Apold1 (0.3), Cdkn1a (0.3), Hba-a1 (0.2), Hba-a2 (0.1), Hbb-bs (0.2), Hbb-bt (0.1), Hif3a (0.2)
Extracellular matrix formation		Col1a1 (2.4), Col3a1 (2.5)
Oxidative stress	Maff (0.2), Xdh (0.5),
Apoptosis	Map3k6 (0.2), Nkx3-1 (0.3), Pmaip1 (0.3)
Glucocorticoid-dependent stress	Nfkbia (0.4), Slc4a1 (0.1), Sgk1 (0.4), Tph2 (0.2), Tsc22d3 (0.5)
Autophagy and UP systems	Ctla2a (0.3), Dennd3 (0.4), Ube2l6 (0.4)
Genes with unknown/unclear or context-specific effects	Arrdc2 (0.3), Atp5g1 (0.3), B230303O12Rik (0.4), Fmo2 (0.4), Gm2800 (Cntr only), Gm38360 (Cntr only), Gm37509 (0.5), Gm42644 (0.3), Gm7292 (0.5), Gm9855 (0.2), Nostrin (0.4), Rpl26 (0.4), Slc18a2 (0.5), Tmem252 (0.1)	2010001K21Rik (TX only), Cep78 (2.4), Gm6969 (2.2), Gm12976 (TX only), Gm15772 (2.9), Gm6394 (3.6), Hes5 (2.3), Myo1f (2.2), RP23-359P19.1 (8.5), Trac (TX only)

Down- (↓) and up-regulated (↑) genes are grouped either by processes or sensitivity of the gene expression to specific conditions. The number in parenthesis indicates a fold change in EFV-treated (Tx, n = 3 mice) vs. control (Cntr, n = 3 mice) animals as determined by RNA-Seq (q and P are ≤0.05). Genes in bold are those whose expression was confirmed by real-time quantitative PCR (P ≤ 0.05); their fold change represents that determined by real-time quantitative PCR. For clarity, each gene is shown only in 1 group despite involvement in multiple processes or conditions. APP, amyloid precursor protein.

We seemed to identify the genes (Serpine1 and Plau) whose altered expression can explain a decrease in extracellular Aβ degradation in the 1TP (Supplemental Fig. S1) as well as the genes (Ccr5, Itgax, and Lilrb4, Table 1), which are associated with the protective microglia phenotype (49–53). In addition, the genes with a decreased expression suggested a decrease in neuroinflammation (C3, Fkbp5, Gbp5, Iigp1, and Zbtb16) (54–59), hypoxia (Apold1, Cdkn1a, Hba-a1, Hba-a2, Hbb-bs, Hbb-bt, and Hif3a), oxidative stress (Maff and Xdh), apoptosis (Map3k6, Nkx3-1, and Pmaip1), glucocorticoid-dependent stress (Nfkbia, Sgk1, Slc4a, Tph2, and Tsc22d3) as well as possible restoration of lipid metabolism (Angptl4 and Plin4), autophagy, and ubiquitin-proteasome (UP) systems (Ctla2a, Dennd3, and Ube2l6) (60–76). The genes with increased expression pointed to increases in neuronal myelinization (Egr2 and Nkx6-2) and extracellular matrix formation (Col1a1 and Col3a1) (77–79). The functions of the remaining differentially expressed genes are currently either unknown, unclear, or context specific. All of the processes, except Aβ degradation, inflammation, and microglia phenotype, identified by RNA-Seq were not indicated by our previous studies.

A comparison of the gene expression in 9-mo-old vs. 5-mo-old 5XFAD mice from the 1TP (Fig. 4D) revealed a totally different pattern of changes with no overlap in either up- or down-regulated genes. The only exception was C3, which was down-regulated 2.7- and 3.3-fold in 5- and 9-mo-old EFV-treated 5XFAD mice, respectively.

Brain phosphoproteome

Previously, this approach was used for studies of Cyp46a1^−/− and wild-type mice (15). Herein, we characterized the brain phosphoproteome of 9-mo-old 5XFAD mice from the 1TP. A total of 4838 phosphopeptides from 1613 proteins were identified with 93 phosphopeptides from 77 proteins showing statistically significant differences in abundance between EFV-treated and control mice (Fig. 5A, B and Supplemental Table S3). All of these phosphopeptides (except the 3 from CTNND2) had changes in the abundance within 1 protein in the same direction: 76 phosphopeptides (from 63 proteins) were less abundant in EFV-treated 5XFAD mice and 14 phosphopeptides (from 13 proteins) were more abundant in EFV-treated mice. In CTNND2, 2 phosphopeptides were less abundant in EFV-treated mice, and 1 peptide was only detected in EFV-treated mice.

Figure 5.

Brain phosphoproteome in 9-mo-old 5XFAD mice from the 1TP. A) The volcano plot of changes in the phosphopeptide abundance in EFV-treated (Tx) vs. control (Cntr) 5XFAD mice. Biologic replicates represent individual animals (n = 3 per group, all male mice and littermates). Phosphopeptides (small circles) with decreases and increases in abundance are colored in green and magenta, respectively. The horizontal dashed line is the boundary of the P value of 0.05. B) The heat map representation of the phosphopeptides with increased (↑, boxed region above the horizonal black line) and decreased (↓, boxed region below the horizonal black line) phosphorylation (P ≤ 0.05). Each row and column correspond to the same phosphopeptide and biologic replicate, respectively. Rows are scaled and represent Z scores (the deviation from the mean by the sd units): negative Z scores are shown in green for both Cntr and Tx mice and positive Z scores are shown in blue for Cntr and orange for Tx mice, see the color code at the panel bottom. The heat map was generated by the Heatmapper software. C, D) Schematic locations of the differentially phosphorylated proteins. Differentially phosphorylated proteins that are associated and not associated with lipids rafts are in bold green and bold black fonts, respectively; all other proteins are in regular font. C) Only 1 protein location, either in a pre- or postsynaptic compartment, is shown based on high protein abundance in this compartment. Gray ovals outline the 3 complexes of the differentially phosphorylated proteins important for neurotransmitter release and postsynaptic signaling.

Differentially phosphorylated proteins were next analyzed for common location, biologic process, and signaling pathway. Analysis for common location (Fig. 5C) revealed that of the 77 differentially phosphorylated proteins, 21 were associated with synapses, mainly or in part (AKAP5, CACNA1B, CTNND2, DGKQ, DPYSL2, EPB41, EPB41L1, GAP43, GPRC5B, LPPR4, LRRC7, MARCKS, PALM, PCLO, PRKCA, PRKCE, RIMS1, SLC12A5, SNPH, SPTBN1, and SRCIN1), including 13 proteins (AKAP5, CACNA1B, CTNND2, DGKO, GAP43, GPRC5B, LPPR4, MARCKS, PCLO, PRKCA, PRKCE, PALM, and SLC12A5), which could reside in synaptic lipid rafts (80–112). Proteins in synapses form complexes, which were found to contain a number of differentially phosphorylated proteins (Fig. 5C). These were the complexes of the: 1) anchoring protein LRRC7 with the protein kinase CAMK2α; 2) scaffold protein AKAP5 with the protein kinases PRKAP1α, PRKCA, and PRKCE and the PPP3CA phosphatase; and 3) Ca²⁺-channel CACNA1B with DPYSL2, PCLO, and RIMS1 (80–82, 84, 87, 88, 90, 98, 113). In addition to synapses (Fig. 5D), 12 differentially phosphorylated proteins could be present in neurofibrillary tangles (DPYSL2, EPB41, MAP1B, MAP4, MAP6, and SPTBN1) or amyloid plaques (ARHGEF2, CAMK2A, DPYSL2, GAP43, MARCKS, PALM, PCLO, and SPTBN1) (113–120).

Analysis of proteins with altered phosphorylation in EFV-treated mice for common biologic process (Table 2) revealed an overlap with the processes, which were identified by the changes in the brain gene expression (Table 1). These were inflammation (3 and 1 proteins with decreased and increased phosphorylation, respectively, in EFV-treated vs. control mice), apoptosis (1 and 2 proteins, respectively) along with autophagy and UP systems (all 7 proteins had a decreased phosphorylation in EFV-treated vs. control mice). In addition, a big group of differentially phosphorylated proteins pertained to synaptic plasticity and neurite growth (44 and 9 proteins with the decreased and increased phosphorylation, respectively, in EFV-treated vs. control mice).

TABLE 2.

EFV effects on protein phosphorylation in the brain of 9-mo-old 5XFAD mice from the 1TP

Processes and signaling pathways	↓Tx/Cntr (decreased phosphorylation)	↑Tx/Cntr (increased phosphorylation)
Grouping by process
Synaptic plasticity and neurite growth	AKAP5, ARHGEF12, ARHGEF2, ATXN2, CACNA1B, CAMKV, CANX, CDC42EP4, CSDC2, CTNND2, DGKQ, DPYSL2, EPB41, EPB41L1, GAP43, GPRC5B, KCNC3, KCNH7, LPPR3, LPPR4, LRRC7, MAP1B, MAP4, MAP6, MAP7D1, MARCKS, MINK1, PALM, PCLO, PKP4, PCM1, PLCH2, PRKAR1A, PRKCE, RAP1GAP, RIMS1, SH3KBP1, SLC12A5, SNPH, SOLO/TRIO8, SPTBN1, SRCIN1, STIM2, ULK1	ARHGAP17, CAMK2A, CBX5, CTNND2, FXYD7, MECP2, PPP3CA, PRKCA, TIAM1
Inflammation	LSM14A, RLTPR, TRIM9	ZC3H18
Apoptosis	CAST	NEK1, SRSF2
Autophagy and UP systems	DENND3, FOXK1, HECW1, HSPA12A, HSPA4, TRAPPC10, ULK1
Grouping by signaling pathway
Calcium signaling and phosphorylation	AKAP5, CACNA1B, CAST, CCNK, DGKQ, EPB41L1, GAP43, GPRC5B, HSPA12A, KCNH7, LPPR3, LPPR4, LRRC7, MINK1, PHLDB1, PLCH2, PRKAR1A, PRKCE, SH3KBP1, SPTBN1, STIM2, ULK1	CAMK2A, NEK1, PPP3CA, PRKCA, WBP11
Small GTPase-dependent signaling	ARHGEF12, ARHGEF2, CAMKV, CDC42EP4, DENND3, RAP1GAP, RLTPR, SOLO/TRIO8, TRAPPC10	ARHGAP17, TIAM1
Catenin signaling	ALDOA, CTNND2, FOXK1, PKP4, PTPRU, SRCIN1	CTNND2

Only proteins with statistically significant changes (P ≤ 0.05) in the phosphopeptide abundance in EFV-treated (Tx, n = 3 mice) vs. control (Cntr, n = 3 mice) animals are shown; if more than 1 phosphopeptide was detected, each protein, except CTNND2, had changes in the phosphopeptide abundance in the same direction. Differentially phosphorylated proteins are grouped by process and signaling pathway. For clarity, each phosphoprotein is shown in only 1 group despite the involvement in multiple processes or pathways.

Finally, analysis of the differentially phosphorylated proteins for common signaling pathway, identified 3 such pathways (Supplemental Fig. S2 and Table 2). These were Ca²⁺-signaling and control of phosphorylation (27 proteins), small GTPase-dependent signaling (11 proteins), and catenin signaling (7 proteins).

Changes in the protein phosphorylation in 9-mo-old EFV-treated vs. control 5XFAD mice from the 1TP were next compared with those in age-matched Cyp46a1^−/− vs. wild-type mice (15). Both EFV treatment (i.e., CYP46A1 activation) and Cyp46a1 ablation mostly decreased protein phosphorylation but affected different amino acid residues, except of S687 in MAP6 and S992 in MAP1B (Fig. 6A, B). S687 in MAP6 had a decrease in phosphorylation in both conditions, whereas S992 in MAP1B had opposite changes in phosphorylation between EFV-treated and Cyp46a1^−/− mice. Proteins affected by CYP46A1 activation and Cyp46a1 ablation were also different, except CTNND2, DPYSL2, GAP43, MAP1B, MAP6, MARCKS, PKP4, and SAMD14 (Fig. 6B), which all pertain to neurite growth excluding SAMD14 (122–127).

Figure 6.

Effects of CYP46A1 activation and Cyp46a1 ablation on the brain phosphoproteome in 9-mo-old mice. Changes in protein phosphorylation in EFV-treated (Tx) vs. control (Cntr) 5XFAD mice from the 1TP were compared with those in Cyp46a1^−/− (KO) vs. wild-type (WT) mice. A) A Venn diagram showing the number of phosphopeptides with statistically significant (P ≤ 0.05) increases (↑) and decreases (↓) in abundance as a result of CYP46A1 activation or Cyp46a1 ablation. Only a decrease in the phosphorylation of S687 in MAP6 was common between the 2 conditions. B) The overlapping phosphoproteins between the conditions of CYP46A1 activation and Cyp46a1 ablation and the phosphorylated sites in these proteins with differential abundance; fold changes in the phosphopeptide abundance are in parenthesis. The identical sites between the 2 conditions are in bold; the sites with the known significance of phosphorylation are underlined. C, D) Prediction of protein kinases, which can phosphorylate the peptides with differential abundance in the conditions of CYP46A1 activation and ablation. The NetPhos 3.1 software was used to make these predictions; only protein kinases with the prediction score of ≥0.5 were considered. Protein kinases with a known association with lipid rafts are above the bars; all other protein kinases are under the bars. Unsp, unspecified protein kinases.

The identified phosphorylation sites were then analyzed for protein kinases, which could phosphorylate amino acid residues within these sites (Fig. 6C, D). In both control 5XFAD and wild-type mice, 4 protein kinases (CDK5, p38 MAPK, GSK3, and CK2) were predicted to phosphorylate >50% of the identified phosphosites. Yet, in both EFV-treated 5XFAD and Cyp46a1^−/− mice, there were no major protein kinases that were predicted to perform phosphorylations. Thus, it could be a decrease in the activity of CDK5, p38 MAPK, GSK3, and CK2 that was responsible for an overall decrease in protein phosphorylation under the conditions of CYP46A1 activation or Cyp46a1^−/− ablation. A decrease in phosphorylation of DPYSL2 at S522 (known to be catalyzed by CDK5) and T509, T514, and S518 (known to be catalyzed by GSK3) (128, 129) supports the validity of this interpretation (Fig. 6B). Similarly, there was a decrease in GAP43 phosphorylation at S193 known to be catalyzed by CK2 (129). Thus, there should be common factors in EFV-treated 5XFAD mice and Cyp46a1^−/− mice that have similar effects on the activity of protein kinases, despite opposite changes in the activity of CYP46A1.

DISCUSSION

Herein, the omics approaches and measurements of some of the brain protein levels by Western blots were used to conduct additional characterization of 5XFAD mice treated with EFV from 1 to 9 mo of age (the 1TP). Important insights were obtained. RNA-Seq pointed to increases in degradation of Aβ by plasmin and microglia as well as a decreased inflammation via the suppression of the NF-κB signaling. Brain phosphoproteomics revealed a remarkable general decrease in phosphorylation and that the majority of differentially phosphorylated proteins pertain to synaptic processes and the 3 signaling pathways: Ca²⁺-, GTPase, and catenin signaling. Finally, the measurements of the synaptic proteins confirmed EFV effects at the synaptic level.

The expression of Plau (a plasminogen activator) and Serpine1 (a plasminogen activator inhibitor-1) were increased and decreased respectively in EFV-treated mice from the 1TP at 9 mo of age (Supplemental Fig. S1 and Table 1). Both genes encode the proteins that control the levels of plasmin, an Aβ-degrading enzyme in the CNS (49). Accordingly, if translated into changes in the protein expression, the altered expression of Plau and Serpine1 could synergistically increase the plasmin content in the brain of 9-mo-old EFV-treated 5XFAD mice and thereby Aβ cleavage and a decrease in the Aβ load. Moreover, changes in the Serpine1 and Plau expression would explain why there were remarkably similar Aβ decreases (by ∼30%) in EFV-treated mice of 5 and 9 mo of age, the time points when control 5XFAD mice had >10-fold difference in the Aβ burden (36). Furthermore, 9-mo-old EFV-treated 5XFAD mice from the 1TP had increases in the levels of markers (Ccr5, Itgax, and Lilrb4, Supplemental Fig. S1) indicative of the protective, Iba1-positive microglia phenotype (50–53). These microglia could have enhanced amyloid plaque phagocytosis, a decreased expression of the proinflammatory (Gbp5 and Iigp1) and pro-oxidant (Maff and Xdh) genes (Table 1) along with the suppressed NF-kB signaling (54–59, 61, 62, 131). A decrease in the NF-κB signaling is supported by increased expression of Lilbr4 encoding an inhibitor of NF-κB signaling (132), decreased expression of Gbp5 encoding an activator of NF-κB signaling (133) as well as altered phosphorylation of the negative (TRIM9) and positive (GRPC5B, LSM14A, RLTPR, and ZC3H18) NF-κB regulators (134–138).

The investigation of the brain phosphoproteome of 9-mo-old EFV-treated 5XFAD mice from the 1TP showed that 63 of the 77 identified phosphoproteins had a decreased phosphorylation, and of them, 41 are known to be important for synaptic plasticity and neurite growth (Table 2). Moreover, 7 of these proteins (CACNA1B, GAP43, DPYSL2, DGKQ, PCLO, SNPH, and RIMS1) are highly specific for presynaptic compartments and regulate neurotransmitter release (80–86), whereas 9 proteins (AKAP5, CTNND2, EPB41L1, LPPR4, LRRC7, MARCKS, SLC12A5, SPTBN1 and SRCIN1) are abundant in postsynaptic compartments (Fig. 5C) and contribute to postsynaptic signaling (87–97). In addition, EFV effects on Ca²⁺, GTPase, and catenin signaling were identified, the novel and important findings because aberrant signaling through all of these pathways is tightly linked to progression of AD (139–142). Effects on Ca²⁺ signaling are supported by differential phosphorylation of 20 proteins of pertinence to this pathway (Supplemental Fig. S2A and Table 2) including CAMK2A and PRKCA, one of the main Ca²⁺-dependent protein kinases, as well as PPP3CA, the Ca²⁺-dependent phosphatase (also known as calcineurin). Changes in the GTPase signaling are indicated by differential phosphorylation of the important regulators of GTPase activity: DENND3 for Rab12, RAP1GAP for Rap1 along with ARHGEF2, ARHGEF12, ARHGP17, CAMKV, RAP1GAP, SOLO/TRIO8, and TIAM1 for the Rho family (Supplemental Fig. S2C and Table 2). Finally, the modulation of catenin signaling is suggested by the altered phosphorylation of the neuron-specific CTNND2 and ubiquitous PKP4, the 2 catenins from the p120ctn family (124), as well as ALDOA1, FOXK1, PTPRU, and SRCIN1 (Supplemental Fig. S2C and Table 2), the proteins, which modulate β-catenin signaling (143–146). Specific effects of the altered phosphorylation in 5XFAD mice are mostly unknown, yet 9-mo-old EFV-treated 5XFAD mice from the 1TP had behavioral improvements (36) and changes in the levels of synaptic proteins (Fig. 2) reflective of a possible increase in synaptic stability (Supplemental Fig. S3A, B).

A comparison of the data obtained with the results of our previous studies (15, 36) provided additional important insights into the mechanisms of EFV effects. Indeed, a pattern of changes in the expression of the 6 studied proteins (calbindin, gephyrin, Munc13-1, PSD-95, synaptophysin, and synapsin-1) was mainly different in 9-mo-old EFV-treated 5XFAD mice from the 1TP and 2TP (Fig. 2). This suggests differences at the synaptic level, which we link to a decreased and unchanged Aβ levels in the 1TP and 2TP, respectively (Supplemental Fig. S3A–C). Likewise, a comparison of the 5- and 9-mo evaluation points for EFV effects on gene expression within the 1TP (Fig. 1C) revealed very few similarities, with only a common decrease in the C3 expression (Fig. 4D). C3 encodes complement component 3, a protein of the immune system, which mediates microglia-dependent synaptic pruning in the CNS (147). C3 could be produced by microglia and astrocytes (147); accordingly, a decrease in the C3 expression in 5-mo-old EFV-treated mice could be secondary to a decrease in the levels of Iba1-positive microglia. However, in 9-mo-old EFV-treated mice, the Iba1 levels were unchanged. Yet there was increased expression of Ccr5, Itgax, and Lilrb4 (Supplemental Fig. S1 and Table 1), the genes that suggest an increase in the population of the protective, Iba1-positive microglia (50–53). Thus, at both evaluation times, a decrease in the C3 expression could reflect EFV effects on Iba1-positive microglial cells but these effects are treatment time-specific.

A comparison of changes in the brain phosphoproteome of 9-mo-old EFV-treated mice from the 1TP with those in 9-mo-old Cyp46a1^−/− mice uncovered several commonalities (Fig. 6). These were: 1) an overall decrease in protein phosphorylation; 2) a prediction of reduction in the activity of the same 4 proteins kinases (CDK5, p38 MAPK, GSK3, and CK2); and 3) a small set of overlapping phosphoproteins. All these commonalities could be explained with the consideration of the effect of CYP46A1 activation (EFV treatment) and Cyp46a1 ablation on lipid rafts. In both conditions, there is an altered cholesterol turnover in the brain (3, 36) and hence altered cholesterol flux through lipid rafts. Accordingly, irrespective of whether this flux is increased (in EFV-treated mice) or decreased (in Cyp46a1^−/− mice), there would be a CYP46A1-dependent effect on lipid raft integrity. Cholesterol serves as the glue for lipid raft assembly (148, 149), and the modulation of Cyp46a1 expression was shown to alter the cholesterol content in lipid rafts (30, 31, 150). Hence, there could be a lipid raft rearrangement as a result of changes in CYP46A1 activity and subsequent impairment in targeting of the protein kinases to their protein substrates and vice versa protein substrates to their kinases within lipid rafts (17–20). If so, there would be a common decrease in phosphorylation in EFV-treated and Cyp46a1^−/− mice as well as a reduction in the activity of the same protein kinases. However, the set of the affected kinase substrates will mostly be different (as shown herein) because different proteins will likely be associated with lipid rafts under the conditions of CYP46A1 activation and Cyp46a1 ablation. Inhibition of GSK3, CDK5, and p38 MAPK is considered as a potential treatment for AD (151–153); all 3 kinases can phosphorylate tau, and their overactivity leads to synaptic dysfunction, neurodegeneration, and cognitive deficits in different Alzheimer models (151–153). Also, inhibition of GSK3 and CDK5 or neuronal-specific deletion of p38α MAPK was shown to ameliorate cognitive deficits in 5XFAD mice (154–156). Accordingly, for 5XFAD mice, the putative decrease in the activity of GSK3, CDK5, and p38 MAPK is likely a positive effect of EFV treatment.

On the basis of data obtained and considering our comparative analyses, we propose a model of EFV effects in 5XFAD mice after 8 mo of drug administration (Fig. 7). In this model, EFV treatment activates CYP46A1 and leads to increases in the 24HC levels and the rate of brain cholesterol turnover. Then, EFV effects could be split into 3 branches: signaling, synaptic, and AD. We hypothesize that the first 2 branches will be common between the 1TP and 2TP, whereas the third branch will be paradigm-specific. Also, we reason that central in EFV effects is the synaptic branch because in both treatment paradigms there were behavioral improvements in 5XFAD mice, which are not possible without changes at the synaptic level. These specific synaptic changes, namely EFV effects on synaptic plasticity and neurite growth as well as synaptic Ca²⁺-signaling (Fig. 2, Supplemental Fig. S2A, B, and Table 2), were identified in the present work by the brain phosphoproteome analysis and the measurements of the synaptic proteins. Changes in the brain phosphoproteome could in turn be secondary to the putative lipid raft rearrangement followed by a decrease in protein phosphorylation and the activity of the protein kinases. The phosphorylation of DPYSL2 at S522 and S518 (Fig. 6B) is increased in AD (129) and was shown to lead to axon degeneration and impairment of synaptic plasticity in mouse models (157–159). Similarly, S140 of MARCKS (Fig. 6B), is among the hyperphosphorylated sites in AD (114). Hence, combined with behavioral improvements in 5XFAD, a decreased phosphorylation at these sites of DPYSL2 and MARCKS, one of the key players in AD pathogenesis (159, 160), is probably a beneficial effect of EFV treatment. Likewise, differential phosphorylation of MECP2 and CBX5, 2 of the 13 proteins with increased phosphorylation, could be a positive effect as well. Indeed, the complex formation between MECP2 and CBX5, which regulate the transcription of the genes related to synaptic plasticity including Egr2 (Supplemental Fig. S2B and Table 1), depends on phosphorylation of MECP2 at S80, and a lack of this phosphorylation was shown to impair spatial memory in mice (161, 162). Thus, the present work revealed the link between CYP46A1-initiated enhancement of the brain cholesterol turnover, altered (mainly decreased) protein phosphorylation, synaptic processes, and behavioral improvements in 5XFAD mice. Specific synaptic-related proteins essential for this link were identified as well.

Figure 7.

Proposed model of EFV effects in 9-mo-old 5XFAD mice from the 1TP. Three branches are shown from left to right: signaling 1), synaptic 2), and AD 3) branch. These branches are highlighted in sage, lemon, and pink, respectively. The affected biologic processed identified in our previous studies are shown in regular black font and those found in the present work are emphasized with bold color fonts; gray regular font indicates the processes suggested by literature data. ↑, ↓, and ↔ arrows indicate increase, decrease, and no change, respectively. The central branch links CYP46A1 activation with lipid raft rearrangement and subsequent decrease in protein phosphorylation followed by improvements in synaptic processes (synaptic plasticity and neurite growth) and behavior. Altered Ca²⁺ signaling is a part of this branch and affects protein phosphorylation as well as synaptic processes. Lipid raft rearrangement and changes in protein phosphorylation are the major events, which link the central branch to the side branches. In addition, increased synaptic plasticity and neurite outgrowth could increase neuronal survival and vice versa increased neuronal survival may promote synaptic processes. The signaling (left) branch starts from an increase in protein prenylation affecting in turn small GTPase signaling and synaptic processes. The catenin signaling is a part of this branch and regulates both GTPase signaling and synaptic plasticity along with neurite growth. Finally, the AD (right) branch ties amyloidogenesis and glial cells (microglia and astrocytes, whose activation can reciprocally affect each other). The putative shift in the microglia phenotype to a protective one could counteract amyloid load and inflammation and thereby increase cell survival, which in turn could lead to behavioral improvements and increase in the levels of CYP46A1, normally a neuronal enzyme. Other processes, known to contribute to the pathogenesis of AD (hypoxia, oxidative stress, autophagy and UP system as well as apoptosis), are included in this branch and placed in the box. Each of the boxed processes contributes to cell survival and synaptic plasticity along with neurite outgrowth. Thus, all 3 branches can act on synaptic processes and thereby improve behavioral performance in EFV-treated 5XFAD mice.

Signaling through Rho GTPases and catenins represents a separate branch of EFV treatment in our model (Fig. 7), with both signaling systems being involved in the cytoskeleton organization as well as the formation and remodeling of synaptic contacts (163, 164). Importantly, catenin signaling is known to be attenuated in AD models and contribute to synaptic and cognitive defects (140). This branch starts in our model with protein prenylation, one of the first processes revealed to be affected by CYP46A1 (14, 16, 165). Then CYP46A1-induced increase in protein prenylation was demonstrated to alter the function and membrane localization of Rho GTPases and increase the expression of PSD-95, synaptophysin, and Shank 3, the only 3 synaptic proteins studied in this work (16, 166). Herein, we found differential phosphorylation of the main regulators of Rho GTPases (Supplemental Fig. S2C), increases in the PSD-95 and synaptophysin levels (Fig. 2), and changes in the phosphorylation of proteins related to synaptic plasticity and neurite growth (Table 2). Therefore, we attribute changes in the phosphorylation of the main small GTPase regulators and the components of catenin signaling to improvements of synaptic plasticity and cognition observed in EFV-treated 5XFAD mice (36). Notably, CTNND2 and PKP4 from catenin signaling (Supplemental Fig. S2C) can modulate Rho GTPases as well (163). The present works is the first indication that CYP46A1 activation may affect catenin pathways, and this effect is realized via protein phosphorylation.

Finally, the third branch in our model focuses on the Aβ levels (Fig. 7). The Aβ production from amyloid precursor protein (APP) occurs in lipid rafts (167, 168). Therefore, the likely lipid raft rearrangement in the 1TP could affect the Aβ levels. This would be in addition to the altered Serpine1 and Plau expression (Supplemental Fig. S1), putative shift in the microglia phenotype, and the predicted inhibition of GSK3, CDK5, and p38α MAPK (Fig. 6C), which were shown to reduce the Aβ deposits in 5XFAD mice (154–156). As a part of the third branch, we included hypoxia and oxidative stress, which can both cause synaptic and neuronal loss and promote amyloidogenesis (169–172). The expression of many genes involved in hypoxia and oxidative stress was decreased in EFV-treated mice including Hif3a, Maff, and Xdh (Table 1). HIF3A is a key transcription factor that up-regulates gene expression in response to hypoxia (60). MAFF is a transcriptional repressor of the antioxidant genes (61). XDH is an enzyme producing reactive oxygen species (62). Autophagy and UP systems are the other 2 processes that could be affected by EFV treatment. These processes directly control the quality and protein levels in synapses and hence are implicated in synaptic plasticity and regulation of the Aβ levels (173–176). EFV effects on autophagy and UP systems are supported by changes in both gene expression and protein phosphorylation (Tables 1 and 2): ULK1 and DENND3 participate in the autophagy initiation (63), whereas UBE2L6 and HECW1 are important ubiquitin ligases that target abnormal proteins for degradation (64, 177). Finally, apoptosis, whose attenuation can increase neuronal survival and hence the levels of neuronal (e.g., CYP46A1) and synaptic proteins could also be affected by EFV treatment. EFV effect on apoptosis is indicated by a decreased Pmaip1 expression and altered phosphorylation of NEK1 (Tables 1 and 2). PMAIP1 is a positive regulator of apoptosis (67), and NEK1 is important for keeping cells alive after DNA damage (178). Thus, the third branch of EFV effects encompasses different biologic processes, but all of them could lead to behavioral improvements following CYP46A1 activation.

It is always possible that some of the observed changes in 5XFAD mice are due to off-target EFV effects that are not related to CYP46A1 activation. We minimized this possibility by using a very small EFV dose in our treatments, and we are also testing EFV on Cyp46a1^−/−5XFAD mice that we generated.

In summary, the present work provided new insights into the mechanisms of EFV effects on the Aβ levels and behavioral improvements in 5XFAD mice. Multiple processes along with the key genes and proteins that could be affected by EFV treatment were mapped. CYP46A1 activation was linked to changes in protein phosphorylation and signaling through Ca²⁺, small GTPases, and catenins. A model of EFV effects on 5XFAD mice was generated.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

1TP

first treatment paradigm

2TP

second treatment paradigm

24HC

24S-hydroxycholesterol

Aβ

amyloid β

AD

Alzheimer’s disease

CYP46A1

cytochrome P450 46A1

EFV

efavirenz

GFAP

glial fibrillary acidic protein

Munc13-1

protein Unc-13 homolog A

PSD-95

postsynaptic density protein 95

RNA-Seq

whole transcriptome sequencing

UP

ubiquitin-proteasome

AUTHOR CONTRIBUTIONS

I. A. Pikuleva conceived and designed the study; A. M. Petrov, N. Mast, and Y. Li performed the experiments; A. M. Petrov, N. Mast, and I. A. Pikuleva analyzed the data; A. M. Petrov and I. A. Pikuleva wrote the manuscript; I. A. Pikuleva acquired funding.