Introduction
Deposition of extra-cellular amyloid-β peptide (Aβ) plaques in the brain is a feature of Alzheimer’s disease (AD) pathology as Aβ monomers can aggregate, which form fibrils and senile plaques [
1‐
3]. The ratio of Aβ42 to Aβ40 in plasma is a promising biomarker for selecting patients with brain amyloid accumulation. Low plasma Aβ42/40 ratio has been associated with increased risk of dementia, more pronounced decline in cognitive function, and increased fibrillary Aβ deposition in the brain [
4‐
6]. Aβ plaque brain deposition has been linked to cholesterol and lipid metabolism, both in the brain and in the periphery. In the brain, neuronal production of Aβ is controlled by membrane cholesterol content. Cholesterol content of neurons is kept low, inhibiting Aβ accumulation [
7]. Aβ is cleared from the brain into the peripheral circulatory system. Sequestration of Aβ by sLRP [
8] or by Aβ antibodies [
9] in the periphery can promote Aβ efflux from the brain. In plasma, Aβ also interacts with apolipoprotein E (apoE), apolipoprotein A-I (apoA-I), or apolipoprotein C-III (apoC-III) [
10]. Moreover, Aβ can interact with high-density lipoprotein (HDL) and very-low-density lipoprotein (VLDL) particles in the plasma and CSF [
11].
ATP binding cassette 1 (ABCA1) participates in the formation of nascent HDL particles [
12] and in the clearance of Aβ from the brain [
13]. Recently, studies using CS-6253, an alpha-helical peptide designed from the C-terminus of apoE, to induce ABCA1 activity have shown promising results in reducing AD-related pathology in animal models [
14]. With a greater binding affinity to ABCA1 than apoE, CS-6253 prevents ABCA1 degradation by stimulating ABCA1 recycling to the cell membrane which is associated with augmented cholesterol efflux to primarily apoE acceptor particles [
15,
16]. Consistent with ABCA1 regulating lipidation of apoE, treatment of apoE4-targeted replacement (ApoE4-TR) mice with the ABCA1 agonist, CS-6253, increased apoE4 lipidation. This was accompanied by a reversal of apoE4-related cognitive and brain pathologies, including intraneuronal Aβ42 accumulation [
14] and was associated with an increase in plasma apoE concentrations [
17]. Furthermore, in a similar model, CS-6253 decreased apoE4 and ABCA1 aggregation in hippocampal homogenates of ApoE4-TR mice [
16], supporting the importance of apoE lipidation in preventing its aggregation.
The effect of CS-6253 on plasma and CSF lipoproteins, together with measures of Aβ in primates, have not yet been studied. We hypothesized that treatment with CS-6253 by virtue of inducing ABCA1 activity would influence lipoprotein dynamics, including that of apoE particles, to promote Aβ clearance. We tested this hypothesis in monkeys, as part of the CS-6253 IND-enabling toxicology studies, in three cynomolgus monkey studies: the preliminary pharmacokinetics (PK) assessment study, the 10-day non-Good Laboratory Practice (GLP) dose-range finding (DRF) study, and the 30-day GLP study.
Methods
Study designs
The preliminary PK study included 2 male cynomolgus monkeys, each injected intravenously with a single dose of 25 mg/kg CS-6253. Blood samples were taken pre-injection, baseline, and at 5 min, 30 min, 1 h, 2 h, 4 h, 6 h, 12 h, 24 h, 48 h, and 72 h post-injection (p.i.). One aliquot of CSF was collected at baseline and 6 h after injection. The DRF study included three active CS-6253 groups: 75 mg/kg (low-dose), 150 mg/kg (mid-dose), and 225 mg/kg (high-dose) and a placebo group. Each of the 4 groups contained 2 male cynomolgus monkeys and 2 female cynomolgus monkeys. Monkeys were dosed every other day for a total of 5 times and blood samples were taken at baseline (2 weeks pre-injection), and at 10 min, 2 h, 4 h, 12 h, 24 h, and 48 h post-injection on days 1 and 9. Baseline measurements were used for normalization of the measurements. CSF was collected at 6 h after the last dose, at day 9. In the 30-day GLP dosing study CS-6253 10 mg/kg (low-dose), 25 mg/kg (mid-dose), and 75 mg/kg (high-dose) was injected every other day (QAD) for 28. The CS-6253 75 mg/kg high-dose and placebo groups consisted of 10 (5 male and 5 female) cynomolgus monkeys each. The CS-6253 10 mg/kg and 25 mg/kg groups consisted of 6 (3 male and 3 female) monkeys each. Animals were terminated 2 days after the last injection. Blood samples were collected at 5 min, 2 h, 4 h, 12 h, 24 h, and 48 h p.i. on days 1, 9, and 25. The measurements were normalized to the first reading that was taken at 5 min after the first injection. See also Table S
1 for details of each study. All experimental procedures were conducted according to the approved protocols from the relevant institutions: PK study, BTS research, OWAL Assurance ID: D16-00,768 (A4519-01)—IACUC: 19–023; DRF study, BASI / Inotiv, OWAL Assurance ID: D16-00,571 (A4058-01)—IACUC: 03-MK-2019; GLP study, Altasciences, OWAL Assurance ID: D16-00,639 (A4261-01)—IACUC: 147,820–01. The experiments at USC were approved by IACUC, protocol #21,225.
CSF collection
CSF samples were collected from the monkeys in the PK and DRF studies. In short, following standard procedures animals were anesthetized with intramuscular injection of ketamine and dexmedetomidine for the procedure. CSF was collected aseptically by cisterna magna puncture from all animals. Intramuscular atipamezole was administered as a reversal agent for dexmedetomidine after the procedure. CSF samples were divided into aliquots and frozen at 70 °C until shipped on dry ice by overnight delivery to the biomarker laboratory for analysis.
CS-6253 concentration analysis
CSF and plasma CS-6253 levels were assayed using an ultra-high performance liquid chromatography (UHPLC) with tandem mass spectrometry (MS/MS) bioanalytical method at Climax (San Jose, CA, USA).
Plasma and CSF were collected and Aβ42 and Aβ40 concentrations were measured by sensitive Single molecule array (Simoa) technique (Quanterix Corp., Billerica, MA, USA). Concentrations of APP and AP2B1 were measured using a targeted mass spectrometry method, as previously described [
18].
ApoE measurements in plasma
DRF study plasma samples were diluted 1:5000 and GLP study plasma samples were diluted 1:15,000. ApoE levels were measured using Sandwich ELISA. The readings were analyzed using Myassays Four Parametric Logistic Curve. Note that plasma apoE levels for the 75 mg/kg dose in the GLP study were not measured.
Plasma triglyceride, cholesterol, and pre-β-HDL measurements
Plasma triglyceride levels in the DRF study were measured using the L-Type Triglyceride M test (Fujifilm) according to the manufacturer’s instructions. Samples were diluted three times before the measurement. Total cholesterol levels in plasma were measured using Cholesterol E kit (Fujifilm). HDL cholesterol levels were measured using the HDL-Cholesterol E kit (Fujifilm). Plasma triglyceride, HDL, LDL, and total cholesterol levels in the GLP study samples were measured by IDEXX Laboratories. Data was analyzed using linear quantification. The plasma from the monkeys in the PK study were diluted 1:50 and pre-Beta HDL levels were measured using pre-β1 HDL ELISA kits (Daiichi Pure Chemicals, Inc.) according to the manufacturer’s instructions. The data was analyzed using Myassays Four Parametric Logistic Curve.
In a 96-well round-bottom plate compatible with an accompanying magnetic separator (EpiGentek, Cat. # Q10002-1), an aliquot (30 μL) of plasma from each timepoint was mixed with 70 μL of a primary precipitating solution, then incubated on ice for 15 min. The samples were then centrifuged (2000 RCF, 10 min, 4 °C). The resultant supernatants were mixed in equal proportion with a secondary precipitating solution containing dextran sulfate and incubated at room temperature for 3 min. A volume of 20 μL of magnetic beads (Sigma, Cat. # GE24152105050250) solution \(\left[3.5\frac{\mathrm{mg}}{\mathrm{ml}}\right]\) was added to each suspension, then incubated at room temperature for 3 min. The beads were washed with MQ water (60 μL × 2) following a magnetic pulled down. The beads were then washed with 30 μL of releasing buffer (× 2), subjected to a magnetic pulled down, and the supernatants were pooled for analysis.
Ion-mobility analysis
Monkey plasma samples treated with dextran sulfate were introduced into a charge-reducing electrospray (TSI Inc., model 3482) every 13 min by automated loop injections via an integrated autosampler (Teledyne CETAC Technologies, model MVX-7100). Electrospray settings were as follows: voltage 2.0 kV, CO2 flow 0.15 slmp, and airflow 1.5 slmp. The differential mobility analyzer (TSI Inc., model 3085), coupled to a condensation particle counter (TSI Inc, model 3788), scanned particles 4.45 to 63.8 nm for 180 s. The generated data of interest was analyzed on Fityk (version 1.3.1), as previously described, and graphed using OriginPro software (version 2021). Voigt probability distribution curves were generated from particle count (#/mL) vs diameter range for lipoprotein subclasses and normalized by dividing sub-classes with the sum of peak areas from all lipoproteins present within the spectrum.
Isolation and examination of CSF and plasma lipoprotein fractions using AF4, MRM, and DLS
CSF and plasma samples from the DRF study were sent to the CDC Division of Laboratory Sciences. 50 μL of each plasma sample, at three time points (5 min, 4 h, and 12 h) from two monkeys each, was injected into the asymmetric-flow field-flow fractionation (AF4) system, collecting a set of 40 fractions from each sample. The fractions and all unfractionated CSF and plasma were analyzed by three LC–MS/MS methods using multiple reaction monitoring (MRM) as described elsewhere [
19‐
21], quantifying proteins typically detected in HDL subclasses, main non-polar lipids (free cholesterol and cholesteryl ester), and phospholipid classes (PC, SM, LPC, PE, and PI). Particle sizes in the fractions were determined using dynamic light scattering (DLS) as previously described [
22]. The moles of analytes in the sized fractions were divided by the volume of plasma injected into the AF4 channel, giving equivalent analyte concentrations in plasma.
Statistical analysis
We used multiple linear regression models and linear mixed-effects models to analyze the effects of CS-6253 over time on levels of various biomarkers (baseline-normalized on a percentage scale with baseline values of 100%). Multiple linear regression models were used on those datasets from the DRF study with only two CSF measurement time points per monkey subject. These models were fitted using ordinary least squares estimation. In each model, the main CSF biomarker outcome of interest was modeled as a function of dose (with placebo dose indicated as 0) and timepoint (endpoint compared with baseline measurement), with an interaction term of dose and timepoint; a significant interaction term indicates a mean difference in the biomarker change compared with placebo. Linear mixed-effects models were used on datasets from both the DRF and GLP studies, where each monkey subject had repeated plasma measurements throughout the study. These mixed-effects models were fitted using restricted maximum likelihood estimation. In some of the mixed-effects models, the main plasma biomarker outcome of interest (cholesterol, triglycerides, apoE) was modeled as a function of fixed effects including treatment (active compared with placebo), and indicator variables for hours since injection (i.e., time of injection) and injection number, and total time under study; a random intercept of the subject was specified to model correlated outcomes arising from repeated measurements. Since Aβ40 and Aβ42 measurements were not obtained in placebo-treated monkeys at all time points at which treated monkeys were assessed, the mixed-effects models for these measurements (and the Aβ42/40 ratio) used only actively treated animals; fixed effects included indicator variables for treatment dose, injection number, and time of assessment (4 h and 48 h, each compared with the 5-min baseline timepoint). All data were standardized as needed to obtain standardized parameters. Wald approximation was used to obtain p-values and confidence intervals. A 2-sided p value of less than 0.05 was considered statistically significant. All models were evaluated for assumptions of normality and homoscedasticity using residual plots. Statistical analyses were conducted using the lme4 package in R version 4.0.5.
For some of the variables, which showed a treatment-related trend without statistical significance, we performed a sample size calculation analysis using the pwr package in R to determine the sample sizes needed to detect a certain effect size in each of those variables (Table S
2). All calculations were done with a two-sample independent
t-test, with power set at 0.8 and significance level alpha set at 0.05 and using a one-sided form of the alternative hypothesis. Cohen’s effect size, d was calculated as
$$d=\frac{|{X}_{1}-{X}_{2}|}{SD}$$
where X is the mean value of each sample groups and SD is the pooled standard deviation of the two groups.
Discussion
In this study, cynomolgus monkeys were treated with CS-6253 as part of IND-enabling studies and its effects on lipid metabolism and AD biomarkers were assessed in plasma and CSF.
Since aggregation of Aβ in the brain contributes to the pathogenesis of AD, Aβ-related biomarkers are used for selecting the prodromal stages of this neurodegenerative disease [
39,
40]. Particularly, recent studies have identified lower plasma Aβ42/40 ratio as a predictor of brain amyloidosis [
4‐
6]. Accordingly, plasma Aβ42/40 ratio were used to test the effectiveness of CS-6253. Indeed, treatment with CS-6253 increased the plasma Aβ42/40 ratio, suggesting CS-6253 was able to facilitate Aβ brain to plasma flux in cynomolgus monkeys. One mechanism for this observation is the shift of apoE to larger triglyceride-containing particles, absorbing Aβ from the CSF and transporting it to the liver for clearance [
11]. These results are consistent with previous studies in apoE4-targeted replacement mice, which showed that CS-6253 can counteract Aβ42 accumulation in hippocampal neurons and improve behavioral deficits [
14].
CS-6253 treatment did not have a significant effect on CSF lipoproteins or lipids, possibly due to its poor transport into the brain. CSF to plasma concentrations at 6 h post-dosing was < 1%. Even though these low CS-6253 levels may have direct ABCA1 effects in the CNS, the pharmacological effects of CS-6253 are interpreted to be indirect in nature. A possible explanation for these findings follows the Peripheral-Sink Hypothesis [
41], which postulates that Aβ-binding ligands in the periphery sequester Aβ, promoting efflux of Aβ from the CSF to the periphery. This aligns with the present study’s finding that CS-6253 was able to simultaneously increase Aβ42 concentrations in plasma and decrease them in CSF. Studies have shown support for this hypothesis, showing that increasing peripheral Aβ antibodies and Aβ-binding lipoproteins increase Aβ efflux [
8,
42] through LRP1 [
43,
44]. As Aβ is highly lipophilic, the majority of Aβ40 and Aβ42 in the circulation are bound to lipoproteins, particularly triglyceride-rich lipoproteins (TRLs) [
45,
46]. Since apoE plays an important role in lipoprotein association with Aβ, with apoE-containing human plasma lipoproteins able to absorb excess Aβ [
11]. It is likely then that Aβ may cross into the periphery with an increase in plasma apoE in TRLs. Indeed, the present report found that CS-6253 consistently caused a transient increase in plasma apoE concentrations in TRL particles. The transient nature of the plasma apoE and Aβ42/40 ratio increase may be explained by liver uptake of Aβ42 containing apoE particles by apoE receptors such as LRP1 [
47], thus forming a vector from the brain, then to plasma, and finally to the liver for degradation or excretion. Low plasma apoE levels are associated with increased risk of AD [
25‐
27]. However, the association between apoE levels and dementia risk does not appear to be linear. Examination of a prospective cohort with 105,949 white individuals revealed that an increase in plasma apoE level from 2.5 to 5 mg/dL was associated with lower dementia risk but when apoE level exceeded 5 mg/dL, the association with dementia risk was inversed [
48]. Very high levels of plasma apoE are associated with an increased vascular risk [
49].
In the periphery, apoE plays an important role in reverse cholesterol transport [
50,
51]. In plasma, both apoE and apoA-I receive cholesterol and phospholipids from the plasma membrane of peripheral cells, via ABCA1, a process most pronounced in monocyte-macrophage cells. This reverse cholesterol transport results in the formation of HDL particles, which transport excess cholesterol to the liver for secretion [
52]. In addition to demonstrating vasoprotective functions, plasma HDL particles have been implicated in protection from AD [
53,
54]. Higher levels of apoE-HDL have been shown to increase triglyceride levels by inhibiting displacement of hepatic lipase, an enzyme, which must be liberated to hydrolyze triglycerides [
55,
56]. The increase of apoE and triglycerides found in the present study suggests there may have been a substantial rise in apoE-HDL levels, but further investigation is necessary. The cooperation between apoE and HDL also has an important role in Aβ clearance. It has been shown that injecting apoE into the periphery in the presence of reconstituted HDL promoted the transport of Aβ across bioengineered cerebral blood vessels [
57]. This suggests that interactions between apoE and HDL have synergistic effects on the clearance of Aβ across vasculature.
Low plasma HDL cholesterol levels have been linked to greater cerebral Aβ deposition [
58]. Intravenous administration of HDL has been shown to reduce soluble levels of Aβ in the brain [
59]. Levels of plasma apoA-I and apoE, which are components of plasma HDL, are lower in AD patients [
60‐
63]. Isolated apoA-I binds to Aβ peptide and can prevent Aβ-induced toxicity and Aβ aggregation [
64,
65]. Furthermore, plasma lipoproteins have been linked to the transport and clearance of Aβ from the brain [
66]. Interestingly, adding CS-6253 to plasma has been shown to displace apoA-I from alpha-HDL particles, and stimulate the formation of pre-β HDL [
15]. CS-6253 mimics apoA-I’s ability to interact with ABCA1 to form functional, so-called nascent HDL particles that are actively remodeled in plasma [
15]. The capacity of CS-6253 to compete with other apolipoproteins such as apoE remains to be delineated.
The results of the present study validate previous findings in vitro which demonstrate the ability of CS-6253 to induce formation of pre-β HDL in plasma [
15]. Treatment with CS-6253 increased plasma pre-β levels as soon as 5 min following injection. Plasma pre-β plays an important role in reverse cholesterol transport, as it efficiently stimulates ABCA1-dependent cholesterol efflux. This study also found that CS-6253 decreased HDL-C levels, which may account for the decrease in total-C levels. While low levels of HDL-C have been associated with negative AD outcomes, recent studies suggest that HDL particle functionality is important, for example, through cholesterol efflux from macrophages by ABCA1 [
58,
67]. It has been shown that plasma HDL-C concentrations divided by apoA-I concentrations may be a better alternative to HDL-C levels alone in predicting mortality outcomes [
68]. This may provide more information about the quality of HDL, as HDL is a dynamic and heterogeneous particle. Particularly, this ratio is thought to represent the amount of cholesterol per HDL particle. Accordingly, lower HDL-C/apoA-I ratios would represent a higher number of lipid-poor HDL particles, which are better able to pick up cholesterol from peripheral tissues than cholesterol-rich HDL particles. Indeed, it has been shown that individuals with lower HDL-C/apoA-I ratios had a decreased likelihood of subclinical atherosclerosis and mortality [
68]. The present study found that when accounting for apoA-I, CS-6253 was able to decrease the HDL-C/apoA-I ratio, suggesting CS-6253 increases lipid-poor HDL particles. Interestingly, the time of the HDL-C/apoA-I decrease (4 h after treatment) correlates with the time of the plasma Aβ42/40 ratio increase. The significance of this is unclear and more work needs to be done to understand the plasma HDL-C/apoA-I in relation to AD. It is possible that exchangeable apolipoproteins such as apoE and apoA-I which are present on lipid-poor s-HDL may enter the brain and become lipidated via ABCA1 [
8,
69,
70]. This may allow for the transport of brain lipids and peripheral lipoproteins, which are important for Aβ clearance from the brain. However, we did not detect any changes in CSF lipids or apolipoproteins in this study.
Limitations
This study has some limitations. The findings presented do not show direct activation of ABCA1 in cynomolgus monkeys and therefore might not be directly related to reverse cholesterol transport. For example, CS-6253 treatment effects may not be on macrophages, but on liver or other cells. The study design precludes such an examination. Second, Aβ peptides were not measured on any lipoprotein fractions given the low abundance of Aβ peptides and a limited amount of samples available in this study. In addition, it remains to be demonstrated whether pre-beta HDL formed from CS-6253 or apoE containing TG-rich particles, or both are driving the clearance Aβ peptides with vasoprotective or AD protective effects. Finally, the small sample sizes, particularly for CSF studies, explain why the changes in CSF Aβ levels after treatment did not reach statistical significance.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.