Background
There are no drugs that prevent the death of nerve cells in Alzheimer’s disease (AD). Because age is the greatest risk factor for AD, we developed a drug discovery paradigm that is based upon phenotypic screens against old age-associated brain pathologies without requiring preselected molecular targets [
1,
2]. Six cell culture assays were designed to mimic multiple old age-associated toxicities, and drug candidates were selected that show efficacy in all of these assays [
3]. On the basis of these toxicity assays, we identified an exceptionally potent, orally active, neurotrophic molecule called
J147. J147 facilitates memory in normal rodents, and it prevents the loss of synaptic proteins and cognitive decline when administered to 3-month-old APPswe/PS1∆E9 mice for 7 months [
1,
2,
4], as well as in rapidly aging senescence-accelerated mouse-prone 8 (SAMP8) mice, a model of sporadic AD [
5]. It also reverses memory deficits and some AD pathology when fed to very old transgenic AD mice [
2].
We recently synthesized a derivative of J147 called
CAD-31 that has enhanced neurogenic activity over J147 in human neural precursor cells (NPCs). CAD-31 also stimulates the division of NPCs in the subventricular zone of old APPswe/PS1ΔE9 mice when fed starting from an early age, a preventive strategy [
4]. However, the neuroprotective properties of CAD-31 have not been well characterized, and CAD-31 needed to be tested for disease modification in a more relevant model of AD.
To more closely mimic the clinical setting, we examined the effect of CAD-31 in transgenic mice at a stage when pathology is significantly advanced and asked if the drug could rescue AD-associated deficits. APPswe/PS1∆E9 mice exhibit a subset of behavioral and pathological features of AD, including age-dependent accumulation of β-amyloid (Aβ) as well as learning and memory deficits at 10 months of age [
1,
6,
7]. These mice were previously used to demonstrate the neurogenic and neuroprotective and memory-enhancing effects of CAD-31 in a preventive paradigm in which CAD-31 was administered before pathology was present [
4]. In contrast, the APPswe/PS1∆E9 AD mice in this study were allowed to age to 10 months before being fed CAD-31 for 3 months. Here we show that under these conditions, CAD-31 normalized cognitive skills to those of age-matched wild-type (WT) mice, reduced markers of inflammation and synaptic loss, and shifted the metabolic profile of fatty acids toward the production of ketone bodies, a potent source of energy in the brain when glucose levels are low.
Methods
Materials
High-glucose DMEM and fetal calf serum were obtained from Invitrogen (Carlsbad, CA, USA). C57BL/6J mice were ordered from The Jackson Laboratory (stock number 000664; The Jackson Laboratory, Bar Harbor, ME, USA). The transgenic mouse APPswe/PS1∆E9, line 85, was a generous gift of Dr. J. L. Jankowsky (Department of Neuroscience, Baylor College of Medicine, Houston, TX, USA). The primary rabbit antibodies were used at a dilution of 1:1000 unless otherwise stated, and their sources were as follows: β-actin, mouse monoclonal HRP conjugate; voltage-dependent anion channel; Arc-1; clusterin; phospho-S51-eukaryotic initiation factor 2α (eIF2α) and total eIF2α; ubiquitin; adenosine monophosphate-activated protein kinase (AMPK); phosphor-S72-AMPK; vascular cell adhesion molecule (VCAM); receptor for advanced glycation endproducts (RAGE); oligomycin sensitivity-conferring protein (OSCP); doublecortin (DCX); drebrin; and phospho-S79-acetyl-coenzyme A carboxylase 1 (ACC-1) (all from Cell Signaling Technology, Danvers, MA, USA). All other materials were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless otherwise stated.
Phenotypic screening assays
The various phenotypic screening assays were conducted as previously described [
3]. Briefly, HT-22, primary cortical neurons, or MC65 was plated and exposed to the different environmental stresses. Cells were treated with varying concentrations of CAD-31, and half-maximal effective concentration (EC
50) was determined on the basis of cell viability.
Animal studies
All animal studies were carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Committee on the Ethics of Animal Experiments of the Salk Institute for Biological Studies. The number of mice per group was determined by a power analysis based upon published data from our laboratories and others using this strain of mice.
APPswe/PS1ΔE9-transgenic mice
The APPswe/PS1∆E9-transgenic mice (line 85) were characterized previously [
7]. Line 85 mice carry two transgenes, the mouse/human chimeric APPswe, linked to Swedish familial AD and human PS1∆E9. At 10 months of age, female transgenic mice were fed a defined diet (Harlan Teklad; Envigo, Indianapolis, IN, USA) with and without CAD-31 (200 ppm, approximately 10 mg/kg/day). Treatment continued for 3 months, followed by behavior testing and tissue harvesting. Mouse body weights and food consumption were measured weekly, and there were no significant differences between the groups (data not shown).
Tissue preparation and immunoblotting
Hippocampal tissue samples were homogenized in 10 volumes of radioimmunoprecipitation assay (RIPA) lysis buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.1% SDS, 0.5% deoxycholate, and 1% Nonidet P-40) containing a cocktail of protease and phosphatase inhibitors [20 mg/ml each of pepstatin A, aprotinin, phosphoramidon, and leupeptin; 0.5 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride; 1 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid; 5 mM fenvalerate; and 5 mM cantharidin]. Samples were sonicated (2 × 10 seconds) and centrifuged first at 10,000 × g for 10 minutes and then at 100,000 × g for 60 minutes at 4 °C. The 100,000 × g pellet was taken up either in 6 M guanidine for Aβ analysis or in SDS sample buffer for Western blot analysis. Protein concentrations in the cell extracts were determined using a bicinchoninic acid protein assay (Pierce Biotechnology, Rockford, IL, USA). Equal amounts of protein were solubilized in 2.5× SDS sample buffer, separated on 12% SDS-polyacrylamide gels, transferred to Immobilon-P (EMD Millipore, Billerica, MA, USA), and immunoblotted with the antibodies indicated in the Materials subsection above. For Western blot experiments, protein levels were normalized to actin levels. An unpaired t test was performed to compare two groups at a single time point. When comparing multiple groups, one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was used. All statistical analysis was conducted using GraphPad InStat software.
Aβ enzyme-linked immunosorbent assay
Aβ1–42 levels in hippocampal lysate were analyzed using Aβ1–42 enzyme-linked immunosorbent assay kits from Invitrogen (catalogue number KHB3442). All kit reagents were brought to room temperature before use. Standards were prepared according to the manufacturer’s guidelines, and samples were diluted as follows; RIPA fractions were diluted 1:10 for Aβ1–42, and RIPA insoluble fractions were diluted 1:5000 for Aβ1–42. A quantity of 50 μl of Aβ peptide standards and samples was added in duplicate to 96-well plates precoated with antibody to the NH2-terminal region of Aβ. Plates were incubated at 4 °C overnight, and then 50 μl of human Aβ42 detection antibody was added to each well except the chromogen blanks. Plates were incubated at room temperature with gentle shaking for 3 h and then washed four times with the provided wash buffer. At that time, 100 μl of antirabbit immunoglobulin G HRP working solution was added to each well except the chromogen blanks for 30 minutes at room temperature. Wells were then washed as before four times and incubated with 100 μl of stabilized chromogen for 25 minutes at room temperature in the dark. Stop solution was then added at 100 μl to each well, followed by reading the absorbance of each well at 450 nm. Curve-fitting software was used to generate the standard curve where a four-parameter algorithm provided the best standard curve fit. The concentrations of the samples were calculated from the standard curve and multiplied by the dilution factor.
Pharmacokinetics and free feeding CAD-31 assay protocols
Sprague-Dawley rats had free access to food and water. CAD-31 was given by gavage to rats at 20 mg/kg in corn oil and intravenously in 15% HS 15/PBS. Whole-blood samples were collected from the jugular vein at every time point and brain collected after 20-ml saline perfusion. CAD-31 was extracted with acetonitrile and compared with standards on a TSQ Quantiva™ Triple Quadrupole Mass Spectrometer (Thermo Scientific, Waltham, MA, USA). For the mouse feeding studies, mice had free access to food containing 200 ppm CAD-31. Blood was collected by heart puncture, and the brain concentration was determined as with rats.
Whole-transcriptome RNA-sequencing analysis
RNA was isolated from the hippocampus using the RNeasy Plus Universal Mini Kit (QIAGEN, Valencia, CA, USA). RNA sequencing (RNA-seq) libraries were prepared using the TruSeq Stranded mRNA Sample Prep Kit (Illumina, San Diego, CA, USA) according to the manufacturer’s instructions. Briefly, polyadenylation RNA was selected using oligo(dT) beads. Messenger RNA was then fragmented and reverse-transcribed. Complementary DNA (cDNA) was end-repaired, adenylated, and ligated with Illumina adapters with indexes. Adapter-ligated cDNA was then amplified. Libraries were pooled and sequenced single-end 50 bp on the HiSeq 2500 platform (Illumina). Sequencing reads were quality-tested using FastQC and aligned to the mm10 mouse genome using STAR Aligner version 2.4k. Mapping was carried out using default parameters (up to ten mismatches per read and up to nine multimapping locations per read). Raw gene expression was quantified across all gene exons by RNA-seq using the top-expressed isoform as a proxy for gene expression, and differential gene expression was carried out using the edge R package version 3.6.8, using replicates to compute within-group dispersion.
p Values are adjusted to control the false discovery rate (FDR) by the method of Benjamini and Hochberg. Differentially expressed genes were defined as having an FDR <0.05 and a log2 fold change >1. Further statistical analysis was carried out using AltAnalyze software (
http://www.altanalyze.org). Pathway analysis was conducted using the DAVID Bioinformatics Resource 6.8 [
16].
Metabolomic analyses were conducted at Metabolon (Durham, NC, USA) as described previously [
5]. For statistical analyses and data display, any missing values were assumed to be below the limits of detection and imputed with the compound minimum (minimum value imputation). An estimate of the FDR (
Q value) was calculated to take into account the multiple comparisons that normally occur in metabolomics-based studies, with
Q < 0.05 used as an indication of high confidence in a result. The MetaboAnalyst tool was used to generate heat maps.
Discussion
Because of the innate complexity of the CNS, essentially all of its associated diseases are multifactorial in the sense that there are a large number of toxicities that contribute to nerve cell death. Many, if not most, of these can be reproduced in cell culture assays, and compounds can be identified that inhibit these toxicities [
4]. Using this approach, we have established six screening assays, identified natural products that are neuroprotective, and synthesized a large number of derivatives that have excellent medicinal chemical and pharmacological properties and are active with high potency in all of our assays [
3,
40]. CAD-31 was selected from our chemical library on the basis of its neurogenic properties in human ES-derived NPCs [
4]. Because of the modest change in structure of CAD-31 from its parent molecule J147, we predicted that the neuroprotective properties of J147 would be retained. Table
1 shows that when J147 was compared directly with CAD-31 in six assays, the two compounds had similar but not identical neuroprotective activities.
Several hundred compounds alter Aβ metabolism or improve behavioral deficits in AD transgenic mice using a preventive strategy in which the compound is given before disease onset [
41], but none has translated to a viable therapeutic [
42]. The reason for this may be that many of these compounds are tested only when administered before definitive pathology arises [
43]. However, in humans, pathology is usually present at the time of diagnosis. To test the efficacy of CAD-31 in a more rigorous preclinical AD model, we treated mice using a therapeutic strategy that more accurately reflects the human symptomatic stage.
Tests that assess distinct aspects of human memory can be performed in rodents. Spatial memory is assessed using the Morris water maze [
44], and hippocampus-dependent contextual memory can be analyzed by using a fear-conditioning assay [
12]. In addition, the anxiety response of rodents can be measured using the elevated plus maze, an assay in which AD mice show a disinhibition phenotype [
45,
46]. In the AD reversal treatment strategy described here, CAD-31 reduced the cognitive defect to a level found in WT mice of the same age.
CAD-31 treatment did not result in a significant effect on Aβ metabolism, because the levels of Aβ
1–42 did not change in either the soluble or the insoluble fractions. This agrees with our early study which showed that CAD-31 had no effect on Aβ plaque density or size [
4]. These Aβ data with the therapeutic model are supported by the observation that the protein levels of neither β-secretase nor amyloid precursor protein changed with CAD-31 treatment, whereas they did change in the preventive model, where there was a reduction in Aβ level [
4].
RNA-seq analysis was conducted to determine the possible in vivo mechanisms of action of CAD-31. Significant changes in gene expression clearly clustered into three specific groups: a drug effect independent of the model, a drug effect only in the AD model, and an AD model group that is independent of drug treatment. The AD model grouping contained genes involved mainly in the inflammatory response, whereas the two drug-related groups displayed neuroactive ligand receptors. Further analysis identified genes differentially expressed between the groups. Whereas only 29 genes were significantly changed by CAD-31 in the WT mouse, pathway analysis suggests that these are involved in the modulation of synaptic function. CAD-31 displayed a greater effect in the AD model, where its effects on multiple pathways were identified, including AD, oxidative phosphorylation, long-term potentiation, and AMPK signaling. These data suggest that CAD-31 may be acting through anti-inflammatory, synapse-protective, and metabolic regulatory pathways. To verify these data, we measured the levels of key proteins in these pathways.
There is growing evidence that vascular inflammation may be directly involved in AD because inflammation and microvascular problems are ubiquitous features of the AD brain [
47]. VCAM is a marker for vascular inflammation. VCAM expression is induced by reactive oxygen species and other pro-oxidants, and it is elevated in patients with AD [
39,
48,
49]. VCAM is also elevated in 13-month-old APPswe/PS1∆E9 mice, and it is significantly reduced by CAD-31 along with another inflammation-related protein called
RAGE. RAGE is elevated in patients with AD [
50], and we show in the present study that CAD-31 reduces the level of RAGE well below the level in control mice.
Clusterin is a stress-induced chaperone molecule that is increased in AD and may be a biomarker for inflammation in the disease [
51,
52]. But clusterin has a complex biology, and its functional association with AD is not clearly understood. In the present study, we show that CAD-31 significantly lowers the expression of clusterin, but unlike RAGE and VCAM, clusterin is not lowered to control levels.
The accumulation of intracellular Aβ and other aggregated insoluble proteins are likely triggers for inflammation and cell death in aging and in AD [
33]. Moreover, the intraneuronal aggregated proteins very likely are the cause of ER stress in AD [
53]. Like most animals, APPswe/PS1∆E9 transgenic mice accumulate aggregated, ubiquitinated proteins as they age [
35], and here we show that CAD-31 reduces the amount of these proteins in old mice.
One way for cells to escape this form of stress is by the activation of the unfolded protein response (UPR). We have previously shown that the activation eIF2α, a protein that mediates this response, is neuroprotective both in vitro [
24] and in vivo [
1]. CAD-31 strongly stimulates the phosphorylation of eIF2α as well as AMPK, a central player in cellular metabolism that also mediates ER stress and autophagy [
54].
There is a loss of synapses and dendritic structure in the 13-month-old APPswe/PS1∆E9 mice used in these experiments [
2]. Two markers for synapses are drebrin and Arc-1. Drebrin and Arc are actin-binding synaptic proteins that have a role in synaptic plasticity [
55‐
57]. CAD-31 increases the expression of both Arc and drebrin in AD mice. OSCP is a protein that is a subunit of mitochondrial F1F0-ATP synthase that is reduced in expression in AD [
58]. Although it is not reduced in old AD mice, its expression is significantly elevated by CAD-31. These data are in line with the gene expression information showing that genes associated with ATP synthesis are upregulated by CAD-31. Together, these hippocampal protein expression data show that, when fed to APPswe/PS1∆E9 mice, CAD-31 reduces inflammation, enhances neuroprotective aspects of the UPR, and promotes synaptic structure.
Metabolic profiling was undertaken to identify potential biomarkers for CAD-31 therapy and to further understanding of the effect that CAD-31 has on small-molecule metabolism. In both the AD and WT groups, CAD-31 had the largest impact on lipid metabolites among the over 600 different molecules examined in the plasma and the cortex. In WT mice, CAD-31 increased levels of the ketone body 3-hydroxybutyrate and acyl carnitines, as well as acetyl-CoA levels in the brain. There was a decrease in fatty acids in both the brain and plasma. Ketone bodies, acyl carnitines, and acetyl-CoA are all associated with mitochondrial fatty acid metabolism, and they are all neuroprotective in various experimental paradigms [
59‐
61]. The decrease in fatty acids, coupled with the increase in their byproducts, suggests that CAD-31 is increasing the rate of their oxidation, thereby enhancing energy metabolism. This mechanism may be able to compensate for the reduced rate of glucose metabolism associated with the AD brain [
62]. CAD-31 also increases the levels of sphingolipids. Sphingolipid and sphingosine signaling are dysregulated in AD, and sphingosine-1-phosphate is a potential therapeutic for treating neurodegenerative diseases [
63,
64].
The metabolic shift toward the breakdown of fatty acids induced by CAD-31 could be a possible mechanism of neuroprotection. It has been demonstrated that AMPK activation inhibits ACC-1, leading to a decrease in fatty acid synthesis and an increase in fatty acid β-oxidation [
31]. The first step in fatty acid β-oxidation is the conjugation of a carnitine group for transport across the mitochondrial membrane. Once inside the mitochondria, the fatty acid is converted into acetyl-CoA, which can then be converted into ketone bodies for transport into the bloodstream. Ketone bodies are an important energy source for the CNS because long-chain fatty acids are unable to pass the blood-brain barrier. It has been reported that acetyl-CoA, acyl carnitines, and ketone bodies are dysregulated in AD, and all have shown promise as potential therapeutics. Metabolic regulation, such as through a ketogenic diet, has been successful in treating neurological disorders in the past, and clinical trials with this diet for AD are ongoing [
65]. Experiments using dietary modification suggest that metabolic pathways are a legitimate target for treating AD [
66]. CAD-31’s ability to upregulate three potentially therapeutic neuroprotective pathways is a possible mechanism for the reversal of AD pathology.