Background
Accumulation of aggregated amyloid β peptides (Aβ) in the brain is proposed to be a key trigger in a complex neuropathological cascade that leads to Alzheimer’s disease (AD). Aβ is produced from the amyloid precursor protein (APP) through sequential proteolytic cleavages [
1]. APP is first cleaved by β-secretase to produce a soluble APPβ and a membrane anchored APP carboxyl terminal fragment (CTFβ). The CTFβ is then cleaved by γ-secretase to produce extracellular Aβ peptides and APP-intracellular domain (AICD) fragments. Notably, a number of Aβ peptides are normally produced, with Aβ40 being the most abundant species with minor species including, but not limited to, Aβ37, 38, 39 and 42 routinely observed in most studies. These various species are not produced by simple classic endoproteolysis at multiple sites, but appear to arise from both variation in the initial substrate cleavage site which produces longer Aβs (i.e., Aβ48, Aβ49, and Aβ51) and the cognate AICD, which is then followed by multiple cycles of step-wise, carboxyl-peptidase like cleavages, all of which are mediated by γ-secretase. Although all Aβ peptides normally produced appear to accumulate in the human AD brain, the minor Aβ42 species is typically the most prevalent form that accumulates in the brain parenchyma [
2,
3]. Additional lines of evidence further support the concept Aβ42 is the most pathogenic isoform [
4], whereas Aβ40 may, under some circumstances, be a protective isoform [
5,
6]. Many early onset familial AD (FAD) mutations linked with APP and Presenilin (PSEN, the catalytic subunit of γ-secretase) increase the relative levels of Aβ42 [
7‐
10].
In vitro studies show that Aβ1-42 has a much stronger tendency to aggregate than Aβ1-40 [
11]. In AD mouse model, Aβ42 plays a role as a seeding molecule for amyloid deposition but Aβ40 [
6] does not. In fact, Aβ40 appears to prevent mice from amyloid deposition [
5,
12]. Moreover, Aβx-42 is the earliest detectable Aβ isoform in the brain parenchyma [
13‐
16]. The role of other shorter carboxyl-terminal truncated species is at this point unclear, though it is hypothesized that they may behave like Aβ40 [
5,
17]. Altogether, there is ample rationale that decreasing the levels of Aβ42 could be a prophylactic approach to prevent accumulation of Aβ and, thereby, delay or prevent the development of AD.
There have been studies demonstrating that production and processing of Aβ can be influenced by membrane lipid composition [
18‐
21]. In particular, membrane cholesterol appears to play an important role [
18]. APP-CTFβ and γ-secretase are found in lipid rafts, composed primarily of cholesterol [
18]. Further, it has been shown that cholesterol directly binds to the APP-CTFβ substrate [
22,
23]. The interdependent interactions among the three components (APP-CTFβ, γ-secretase, and cholesterol) are postulated to create the optimal microenvironment for Aβ production. Indeed, it has been reported that γ-secretase activity is largely dependent on the amount of cholesterol, which affects Aβ production as a result [
18,
24] though others have not reproduced this finding [
25]. These observations suggest the potential for modulating γ-secretase activity and thus altering the overall Aβ levels or the ratios of Aβ isoforms produced by steroid derivatives as cholesterol surrogates.
Previously, we have reported steroid carboxylic acid γ-secretase modulators (GSMs) [
26]. Numerous acidic steroids decrease Aβ42 levels and increase Aβ38 levels without changing total Aβ or Aβ40 levels [
26]. Acidic steroid GSMs have gross structural similarity to the established-NSAID based GSMs in that a carboxylic acid group, that is key for GSM activity, is attached by a carbon tether chain to a highly lipophilic core structure [
26,
27]. 5β-Cholanic acid (ursocholanic acid) was the most potent steroid GSM identified in our previous study with an EC
50 of 5.7 μM, but the endogenous bile acids, lithocholic acid and ursocholic acid, were also found to be GSMs [
26]. Mechanistically, GSMs decrease production of Aβ42 selectively by promoting step-wise γ-secretase cleavage and, thus, inherently increase shorter Aβ peptides [
28,
29]. Because γ-secretase cleavage activity participates in a broad spectrum of cellular signaling mechanisms (i.e., Notch-1) [
30], indiscriminate inhibition of γ-secretase activity has been essentially abandoned as a therapeutic approach for AD due to debilitating side effects associated with target-based toxicity. In contrast, GSMs do not alter overall γ-secretase activity, appear to be relatively selective for APP, and are, therefore, thought to be an intrinsically safe mechanistic approach to AD therapy; however, it has been challenging to identify GSMs that are potent, have sufficient brain penetrance, and lack off-target toxicity.
Considering that GSMs derived from synthetic compounds have toxicity issues that are not associated with target-based toxicity, we have explored whether other naturally occurring acidic steroids might have sufficient potency to be therapeutically useful. An extended screening identified 3β-hydroxy-5-cholestenoic acid (CA) as a highly potent GSM with an EC
50 for Aβ42 lowering of 250 nM. As CA is produced endogenously during the course of cholesterol elimination in many extrahepatic organs including the brain [
31,
32] and is present in human plasma at concentrations near its EC
50 for GSM activity, we explored whether CA might function endogenously as a GSM. Our results showed that Cyp27a1−/− [
33,
34] and Cyp7b1−/− [
35] mice that reduce or increase brain CA, respectively, resulted in the predicted brain Aβ42 changes consistent with the hypothesis that CA is an endogenous GSM. Peripheral dosing of CA in wild type mice dramatically increased plasma CA levels, but not brain Aβ levels, suggesting limited brain exposure of peripheral CA. Structure-activity relationship (SAR) with multiple known and novel CA analogs studies failed to reveal CA analogs with increased potency. These studies show that though CA is a potent GSM that may act within the brain to regulate Aβ42 levels, exogenous administration of CA is not likely to be therapeutically useful for lowering Aβ42.
Discussion
In this study, we identified CA as a potent acidic GSM with an EC50 for lowering Aβ42 of ~250 nM, a concentration well within the normal range of CA levels in human plasma (~100-300 nM). This data raised the possibility that CA was an endogenous GSM and that increasing brain CA levels might be a safe approach to lower brain Aβ42 levels. Peripheral dosing of CA, however, did not lower brain Aβ42 despite extremely high CA levels in the plasma (~10 μM), indicating that either CA does appear to readily cross the blood brain barrier or, if it does, is rapidly exported from the brain. Unfortunately, using our methodology, we were not able to accurately measure CA levels with sufficient sensitivity to accurately measure CA levels in the brains of these mice.
Given the potency of CA as a GSM, we explored whether mice with genetic deletions of Cyp27a1 [
33,
34] and Cyp7b1 [
35], the two enzymes regulating CA levels in the brain [
31], showed alterations in mAβ42 levels. We found that the mAβ42/mAβ40 ratio was increased in the Cyp27a1−/− mice and mAβ42/mAβ40 ratio was decreased in Cyp7b1−/− mice, where CA levels were shown to decrease or increase CA levels, respectively [
36]. Given that these shifts in ratio in these knockout mice are precisely what would be predicted if CA demonstrated GSM activity, we concluded that CA is likely to be a bona fide endogenous GSM synthesized in a cholesterol elimination pathway in brain [
31]. Given the challenges of measuring levels of endogenous Aβ outside of the brain in wild type mice, we attempted to generate Cyp7b1−/−, APP+/− (CRND8) mice. Despite extensive efforts, we were unable to generate mice with this genotype that lived past 3 months. We did not attempt to cross the CRND8 mice with Cyp27a1−/− mice, because Cyp27a1−/− mice were even less fecund than the Cyp7b1−/− mice. Future studies in humans with genetic loss of function of
CYP27A1 that causes cerebrotendinous xanthomatosis (CTX) [
45,
36] or with genetic loss of function of
CYP7B1 deficiency that causes liver failure in children or spastic paraplegia 5 (SPG5) in adults [
46,
39,
36], might help to further establish the likelihood that CA is an endogenous GSM, as these patients show altered CA levels and would be predicted to have altered Aβ42/Aβ40 ratios [
39,
36]; however, due to the small number of patients with these rare disorders, and the severe disease induced by loss of these CYP enzymes, such studies may be challenging to sufficiently power and control.
Building off our previous studies to examine a large number of steroids for GSM and inverse GSM (iGSM) activity [
26], we synthesized a number of analogs to see if we can further increase potency. From these studies, we can conclude that CA seems to represent a relatively optimized steroid GSM, especially the C5 alkene tether linking the carboxylate group to the steroid backbone which appears to be optimal in length for maximizing steroid GSM potency. Indeed, there was a significant increase in GSM potency upon increasing the tether length from C3 to C5, but there was only a moderate loss of potency for increased C6 and C7 tether. Based on the observations from extended CA tether analogs, we explored the effects of C5 alkene tether carboxylates on other acidic GSM “scaffolds”. In all cases examined, this “grafting” approach decreased potency, indicating that the positon of the carboxylate group for optimal GSM potency is dependent on the overall structure of the molecule. Further modifications along the steroid backbone all decreased GSM activity relative to CA. For example, both endogenous CA metabolites 7α-OH-CA and 7α-OH-3-CA maintained GSM activity, but were less potent than CA.
Our findings that CA and other steroids can modify Aβ production expand the growing number of studies that demonstrate how cholesterol and other steroids can modulate Aβ profiles [
47‐
54]. Of particular interest are studies showing that cholesterol binds to APP CTFβ [
22,
23], albeit with low affinity, as this might suggest that CA, a cholesterol metabolite, could also interact with CTFβ. Our studies also show that CA behaves much like classic acidic GSMs and like all GSMs, exhibit a fairly flat SAR. Moreover, membrane lipids have been shown to alter the profile of Aβ produced [
55], and therefore it is theoretically possible that CA could alter γ-secretase in a similar manner. However, given the nanomolar potency of CA and the aforementioned flat SAR, we think that this mechanism of action is unlikely. As our data show that it is challenging to generate CA analogs that retain potency, we have not attempted to generate CA analogs that could be used for affinity studies to identify primary binding sites. Given the nanomolar potency of CA, we speculate that it almost certainly interacts with PSEN/γ-secretase. However, as we have previously hypothesized, we would propose that most GSMs alter γ-secretase through a complex interaction involving both substrate and γ-secretase and possibly even other lipid membrane components [
56,
57,
26,
58,
59]. Such a model is consistent with data showing that GSM effects are extremely sensitive to mutations within the substrate [
59,
58,
60] and could explain why different GSM affinity probes have been shown to bind PSEN, PEN2 or C99 [
56,
61‐
63]. It is important to consider that demonstrating binding with such a probe to a certain component does not rule out interaction with the other components, due to limitations where the reactive groups can be placed on the GSM affinity probes and the requirement for photoaffinity probes to have its photoaffinity label be in very close proximity to the bound protein.
In summary, although the endogenous metabolite CA is a potent γ-secretase modulator, i) its lack of ability to lower brain Aβ42 following peripheral dosing and ii) the inability to identify additional endogenous CA analogs with increased potency, suggests that pursuing CA or CA analogs for further preclinical development is not likely to be fruitful. Recent data show that CA can be toxic to primary mouse motor neuron in cultures [
36] and raises concerns for pursuing CA or CA derivate as possible new small molecule therapeutics for AD. As the immediate precursor of CA, 27-OHC cholesterol, readily crosses the blood brain barrier, a pro-drug approach using a modified 27-OHC might be considered as an alternative strategy; however, emerging data that elevated 27-OHC may be a risk factor for osteoporosis and breast cancer, raises concerns about a 27-OHC cholesterol prodrug strategy to increase CA levels as well [
64,
65].
Methods
Cell culture and drug treatment
Chinese hamster ovary (CHO) cells stably overexpressing APP695 (CHO-2B7 cells) [
66] were grown in Ham’s F-12 medium (Life Technologies) supplemented with 10 % fetal bovine serum and 100 units/ml of penicillin and 100 μg/ml streptomycin. Cells were grown at 37 °C in a humidified atmosphere containing 5 % CO
2 in tissue culture plates (Costar). The cells were harvested at confluence and then utilized for biochemical analyses. Compounds were dissolved in dimethyl sulfoxide (DMSO) and screened in CHO-2B7 cells. The cells were incubated for 16 h in the presence of the compound diluted into OptiMEM-reduced serum medium (Life Technologies, Carlsbad, CA, USA) containing 1 % fetal bovine serum. Compounds used for our study were either purchased from Avanti Polar Lipids, Inc. or synthesized by SAI Life Sciences Ltd. The synthesis schemes of the newly synthesized compounds are demonstrated in Additional file
1.
In vitro γ-secretase assay
Broken cell assays were performed with slight modifications from the previous studies [
67,
18]. The membrane derived from the H4 neuroglioma cells overexpressing APP695wt were prepared by carbonate extraction and incubated at 37 °C for 2 h with CA at various concentrations. Aβ levels were quantified by sandwich ELISAs. For Aβ and AICD spectra, the recombinant C100Flag proteins were overexpressed and purified from
Escherichia coli BL21 using a HiTrap Q-column (GE Life Science, Little Chalfont, U.K.) [
68,
69,
58]. The membrane containing γ-secretase was isolated from the CHO S-1 cell line using sodium carbonate (100 mM, pH 11.0) [
70]. For the
in vitro γ-secretase assay, C100Flag recombinant protein at 25 μM was incubated with the membrane (100 μg/mL) in the presence of CA (20 μM) and DMSO in 150 mM sodium citrate buffer (pH 6.8) containing complete protease inhibitor (Roche, Indianapolis, IN) for 2 h at 37 °C.
Mice
All procedures were performed according to the National Institute of Health Guide for the Care and Use of Experimental Animals and were approved by the University of Florida Institutional Animal Care and Use Committee. The Cyp27a1−/− (B6.129-Cyp27a1tm1Elt/J) and Cyp7b1−/− (B6;129S-Cyp7b1tmRus/J) strains were obtained from Jackson Laboratory (Bar Harbor, ME). Cyp27a1−/− mice were bred with C57BL/6 in order to produce the heterozygous littermates of Cyp27a1, and Cyp7b1−/− mice were bred with C57BL/6 mice to produce the heterozygous Cyp7b1 littermates. The wild type, heterozygous, and knockout littermates of Cyp27a1 and Cyp7b1 mice were generated from Cyp27a1+/− X Cyp27a1+/− and Cyp7b1+/− X Cyp7b1+/−, respectively.
Primary mixed neuron-glia culture
Primary mixed neuron-glia cultures were prepared from postnatal day 0 (P0) C3HBL/6 mouse brains (Harlan Labs). Cerebral cortices were dissected from P0 mouse brains and were dissociated in 2 mg/ml papain (Worthington) and 50 μg/mL DNAse I (Sigma) at 37 °C for 20 min. They were then washed three times in sterile Hank’s balanced salt solution (HBSS) to inactivate the papain and switched to 5 % fetal bovine serume (HyClone) in Neurobasal-A growth media (Gibco), which includes 0.5 mM L-glutamine (Gibco), 0.5 mM GlutaMax (Life Technologies), 0.01 % antibiotic-antimycotic (Gibco), and 0.02 % SM1 supplement (Stemcell). The tissue mixture was then triturated three times using a 5 mL pipette followed by a Pasteur pipette, and strained through a 70 μm cell strainer. The cell mixture was then centrifuged at 200xg for 3 min, and re-suspended in fresh Neurobasal-A media. They were then plated onto poly-D-lysine coated 96well plates at 100,000 cells/well. Cells were maintained in the Neurobasal-A growth media mentioned above without fetal bovine serum (FBS) at 37 °C in a humidified 5 % CO2 chamber.
CA IP injections
25(R)-CA powder was initially dissolved in DMSO (<4.5 % in the final mixture) and then combined with polyethylene glycol (15)-hydroxystearate (Solutol), ethanol, and water at a ratio of (15:10:75). One molar equivalent of sodium hydroxide was added to the mixture [
71,
72]. We performed CA intraperitoneal (IP) injections to wild-type mice (C57BL/6 or CF-1). The mice were injected with 25(R)-CA on the right side of the abdomen. The injections have been performed with various time points (30 min, 1 h, 2 h, and 3 h) and with multiple doses (30 mg per kg (mg/kg), 60 mg/kg, 75 mg/kg and 100 mg/kg). The number of each cohort is 6–8. We used 30 mg/kg of CA for the time-course experiments, and for the dose–response experiments the end-point was set at 30 min. The brains and serum are harvested and frozen for brain Aβ extraction.
The mouse brains were harvested at the age of 3 months. The brains were weighed and recorded. The Diethylamine/Sodium Chloride (DEA/NaCl) extraction buffer (0.4 % DEA) was added to each sample and homogenized using a sonicator. The samples were transferred to a poly-carbonate centrifuge tube and spun down at 50,000xg for 30 min at 4 °C. The supernatant was loaded on the vacuum manifold with the appropriate number of HLB Oasis columns. The samples were loaded on the conditioned column, filtered, and eluted using prepared elution buffer (90 % Methanol, 2 % NH4OH). The eluates are concentrated using the Thermo-Savant SpeedVac concentrator for a minimum of 2 h at 55 °C with radiant heat. The concentrated samples are reconstituted in a blocking buffer (0.67 % Bovine serum albumin (BSA)) at the appropriate volume.
Plasma CA analysis
The plasma samples were extracted using published solid phase extraction method (72) and analyzed by HPLC-MS-MS. Briefly, 0.1 ml mouse plasma samples after adding 20 μl of D3-CA as internal standard were preconditioned with 1.4 ml of ethanol (99.9 %), and 0.5 ml of water, centrifuged at 4 °C, 4000 rpm for 10 min. This solution was then loaded onto a Sep-Pak tC18 (SPE1) solid phase extraction cartridge which were preconditioned with 70 % ethanol. The sample was washed with one column volume of 70 % ethanol then eluted from the column by 2 + 1 ml of 99.9 % ethanol; it was dried in centrifuge evaporator. The residue was reconstituted in 100 μl of isopropanol. It was oxidized by adding 1 ml of 50 mM phosphate buffer (pH = 7) containing 3 μl of cholesterol oxidase and incubated at 37 °C for 1 h, quenched with 1.9 ml of methanol. The mixture was further processed by adding 150 μl glacial acetic acid and 1 smidgen (about 80 mg) GP reagent {1-(carboxymethyl) pyridinium chloride hydrazide} and incubated at room temperature overnight in the dark. On the next day, a second solid phase extraction [
73] was employed to separate the derivatized CA from the excess derivatization reagent using the following: Sep-pak C18 (SPE 2, different from SPE1) cartridge with 1 column volume of 99.9 % methanol and 1 column volume of 10 % methanol, after application of the sample wash with 10 % methanol, then elute with 2*1 ml of 100 % methanol. Mix 200 μl of the elution solution with 50 μl of water to obtain 250 μl of 80/20 (methanol/water, v/v) samples. 20 μl was injected onto HPLC-MS-MS for analysis.
HPLC-MS-MS conditions: HPLC contains a Perkin Elmer series 200 autosampler and a Perkin Elmer series 200 pump, MS-MS was Waters Quattro LC-Z, ES positive mode, Cone voltage 45 volts, collision energy 30volts, Desolvation temperature 350 °C. Source block temperature 120 °C. MS/MS transitions: CA 549.0/470.0; D3-CA 552.0/473.0. HPLC mobile phase was 80/20 Methanol/water(v/v) containing 0.1 %Formic Acid, HPLC column was ThermoFisher Hypersil Gold, 50*2.1 mm, 1.9 μ, flow rate 0.2 ml/min. Injection volume 20 μl, run time 4 min, CA retention time 1.4 min.
Antibodies and ELISAs
Monoclonal antibodies to Aβ were generated by the Mayo Clinic Immunology Core facilities (Jacksonville, FL, USA). Ab5 recognizes an epitope in the amino terminus of Aβ (Aβ1-16), recognizes both monomeric and aggregated Aβ, and is human specific. Ab13.1.1. was raised against Aβ35-40 and is specific for Aβx-40, and exhibits minimal cross-reactivity with other Aβ peptides. Ab 2.1.3 was raised against Aβ35-42 and is specific for Aβx-42. The Aβ38 antibody (Ab38), supplied by P. Mehta (Institute of Basic Research, Staten Island, NY, USA), specifically recognizes Aβx-38 and shows no cross-reactivity with other Aβ peptides [
74]. For cell-based screens, Aβ was captured from conditioned medium with either Ab5, Ab38, Ab13.1.1, or Ab2.1.3 (coated at 10-50 μg/ml in EC buffer: 5 mM NaH2PO4-H2O, 20 mM Na2HPO4, 400 mM NaCl, 2.5 mM EDTA-full name, 151.5 μM BSA, 813 μM CHAPS, and 7.7 mM NaN3) on Immulon 4HBX Flat-Bottom Microfilter 96-well plates (Thermo Scientific, Waltham, MA, USA). Total Aβ level was determined by capture with Ab5 and detected with horseradish peroxidase (HRP)-conjugated 4G8 (a monoclonal antibody against Aβ17-24; Covance, Waltham, MA, USA) with the other Aβ peptides detected with HRP-conjugated Ab5. For the cell-free assay and measuring mouse endogenous Aβ, HRP-conjugated 4G8 was used as the secondary detection antibody. Aβ standards (Bachem, King of Prussia, PA, USA) were prepared by dissolving in hexafluoroisopropanol (HFIP) at 1 mg/ml with sonication, dried under nitrogen, resuspended at 2 mg/ml HFIP, sonicated again and dried under nitrogen. The resulting Aβ was resuspended in 0.01 % ammonium hydroxide, portioned into aliquots in EC buffer, and frozen at −80 °C. Following these steps, the Aβ is monomeric, as determined by size-exclusion chromatography.
Immunoprecipitation-Mass spectrometry
Conditioned media from the CHO-2B7 cells and the samples prepared from
in vitro γ-secretase studies were used to analyze Aβ and AICD profiles using matrix-assisted laser desorption/ionization time of flight (MALDI-TOF) mass spectrometry analysis. The secreted Aβ peptides were analyzed as previously described with the following modifications [
2,
75,
76]. Briefly, the Aβ peptides were immunoprecipitated using Ab5 recognizing the Aβ1-16 epitope [
77] and sheep anti-mouse IgG magnetic Dynabeads (Life Technologies, catalog no. 11201D) and the AICD fragments were captured using anti-Flag M2 magnetic beads (Sigma). The samples were washed and eluted with 10 μM solution of 0.1 % trifluoroacetic acid (TFA) in water. Eluted samples were mixed 2:1 with saturated α-cyano-4-hydroxycinnamic acid (CHCA) matrix (Sigma) in acetonitrile: methanol (60:40 %) and loaded onto a CHCA pretreated MSP 96 target plate-polished steel (Bruker, Billerica, MA, USA - Part No.224989). Samples were analyzed using a Bruker Microflex LRF-MALDI-TOF mass spectrometer.
Statistics
In vitro data were expressed and graphed as the mean ± SEM using GraphPad Prism 5 software. Analysis was by one-way analysis of variance (ANOVA) followed by Dunnett’s multiple comparisons, and was by two-way analysis of variance (ANOVA) followed by bonferroni post-hoc testing for group differences. The level of significance was set at p < 0.05 in all tests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
JIJ participated in study design, performed in vitro and cell-based immunoassays, analyzed data, and drafted manuscript. APR and LAS participated in animal study design and performed ELISA. TBL and YR performed in vitro and cell-based assays. HJP and CCD performed primary neuronal culture and ELISA. GH and YT participated in pharmacokinetics study and performed LC-MS/MS. RA and SB synthesized CA analogs, which were designed by GS. EHK was involved in experimental interpretation and manuscript editing. GS, KMF, and TEG participated in study design and coordination and in manuscript preparation and editing. All authors read and approved the final manuscript.