M344 promotes nonamyloidogenic amyloid precursor protein processing while normalizing Alzheimer's disease genes and improving memory.

Alzheimer's disease (AD) comprises multifactorial ailments for which current therapeutic strategies remain insufficient to broadly address the underlying pathophysiology. Epigenetic gene regulation relies upon multifactorial processes that regulate multiple gene and protein pathways, including those involved in AD. We therefore took an epigenetic approach where a single drug would simultaneously affect the expression of a number of defined AD-related targets. We show that the small-molecule histone deacetylase inhibitor M344 reduces beta-amyloid (Aβ), reduces tau Ser396 phosphorylation, and decreases both β-secretase (BACE) and APOEε4 gene expression. M344 increases the expression of AD-relevant genes: BDNF, α-secretase (ADAM10), MINT2, FE65, REST, SIRT1, BIN1, and ABCA7, among others. M344 increases sAPPα and CTFα APP metabolite production, both cleavage products of ADAM10, concordant with increased ADAM10 gene expression. M344 also increases levels of immature APP, supporting an effect on APP trafficking, concurrent with the observed increase in MINT2 and FE65, both shown to increase immature APP in the early secretory pathway. Chronic i.p. treatment of the triple transgenic (APPsw/PS1M146V/TauP301L) mice with M344, at doses as low as 3 mg/kg, significantly prevented cognitive decline evaluated by Y-maze spontaneous alternation, novel object recognition, and Barnes maze spatial memory tests. M344 displays short brain exposure, indicating that brief pulses of daily drug treatment may be sufficient for long-term efficacy. Together, these data show that M344 normalizes several disparate pathogenic pathways related to AD. M344 therefore serves as an example of how a multitargeting compound could be used to address the polygenic nature of multifactorial diseases.

Alzheimer’s disease (AD) is the sixth leading cause of death in the United States and is presently the only top-10 cause of death that has no prevention or effective treatment. With a cost greater than $220 billion for the year 2015 in the United States, AD is a significant burden to the health care system (1). It is expected to reach a prevalence of ∼16 million people in America by the year 2050 (1). Currently approved treatments for AD lack efficacy and are palliative at best. These drugs include cholinesterase inhibitors (donepezil, rivastigmine, and galantamine) or NMDA receptor antagonists (memantine). None of these treatments addresses the molecular pathology present in the brains of AD patients.

AD is confirmed by the diagnosis of dementia associated with the presence of extracellular beta-amyloid (Aβ) plaques in the brain parenchyma and the accumulation of intracellular neurofibrillary tangles—the latter consisting mostly of aggregated hyperphosphorylated tau protein. The accumulation of Aβ results from sequential cleavage of mature (N- and O-glycosylated) amyloid precursor protein (APP) by proteases β- and γ-secretase in the late protein secretory pathway, constituting the amyloidogenic pathway (2, 3). The “amyloid cascade hypothesis” places Aβ at the origin of AD, triggering downstream AD-related events such as tau hyperphosphorylation, neuroplasticity deficits, learning and memory impairments, and, eventually, death (4–9). The accumulation of Aβ can be prevented via the nonamyloidogenic processing of APP by α-secretase cleavage within the Aβ sequence, releasing the neuroprotective metabolite sAPPα and the C-terminal fragment-α (CTF-α, C83). An increase in α-secretase cleavage has been hypothesized as a possible therapeutic target for AD, but currently, due to the difficulties of increasing the activity of an enzyme, most Alzheimer’s drug discovery efforts have aimed at three main strategies to reduce Aβ peptide: immunotherapy, inhibition of β-secretase activity, or inhibition of γ-secretase activity. While there are still some single-target drugs in clinical trials, until this date these approaches have been disappointing at treating AD patients (10–14). It is important to note that several other hypotheses have been proposed to explain AD etiology and pathogenesis. Such hypotheses include—but are not limited to—the mitochondrial cascade, the tau, the vascular, and the neuroinflammation hypotheses that, respectively, place decreased mitochondrial activity, hyperphosphorylated-tau pathology, cerebral hypoperfusion, and/or increased inflammatory events (microgliosis, astrogliosis, and proinflammatory cytokines) as root causes of AD (15–19).

Several AD susceptibility genes, identified in patients through linkage and genome-wide association studies, suggest that AD is a complex polygenic disease (5, 20–23). Due to the polygenicity of AD and the vast number of failures with the single-target approach, many have hypothesized that it will be necessary to utilize combination therapies, and/or treatment at preclinical or prodromal stages for this disease. Here, we tested this hypothesis with an epigenetic approach where we hypothesized a single small molecule could simultaneously affect the expression of many AD-related drug targets, thus bypassing the need for drug combinations. Moreover, since gene expression changes through the remodeling of chromatin play an important role in memory formation (24–27), and epigenetic changes are widely reported in AD brain (28–31), such an epigenetic-directed compound could also prevent memory decline in an AD mouse model.

We provide data describing that the histone deacetylase inhibitor (HDACi) M344 {4-(diethylamino)-N-[7-(hydroxyamino)-7-oxoheptyl]benzamide} modifies several of the AD-related pathways and thus holds some therapeutic potential. M344 was first synthesized in 1999 by Jung et al. (32) and, while little studied compared with many other HDACis, it was reported to significantly increase survival motor neuron 2 (SMN2) gene expression—a gene associated with the severity of proximal spinal muscular atrophy, an orphan disease (33, 34). In the experiments described below we show that M344 favorably addresses a number of key genes reported to be involved in early- and late-onset AD pathogenesis and attenuates cognitive decline in a chronically treated AD mouse model.

Results

Compound Selectivity Profile.

Since little is known about the HDAC selectivity profile of M344 (32), we tested its potency at inhibiting all 11 known zinc-dependent HDACs. The half-maximal inhibition (IC50) concentration of M344 was calculated for each HDAC with a ten-point concentration response curve in duplicates, using titration of 1:3 dilutions (BPS Bioscience). Each HDAC was also inhibited by an appropriate positive control, such as vorinostat (suberoylanilide hydroxamic acid, SAHA) or trichostatic acid (TSA). The HDAC activity profile revealed that M344 showed potent activity for class I (HDACs 1, 2, 3, and 8) and IIB (HDACs 6 and 10) in the submicromolar to the micromolar range (Table1), suggesting selectivity for these classes. Detailed concentration curves are provided in Fig. S1.

Table 1.

M344 HDAC selectivity profile

HDACs	IC50, µM
	M344	Reference
HDAC1	0.048	0.083 (SAHA)
HDAC2	0.12	0.19 (SAHA)
HDAC3	0.032	0.046 (SAHA)
HDAC4	26.80	2.21 (TSA)
HDAC5	15.82	1.18 (TSA)
HDAC6	0.0095	0.027 (SAHA)
HDAC7	17.09	1.34 (TSA)
HDAC8	1.34	0.61 (TSA)
HDAC9	48.80	5.20 (TSA)
HDAC10	0.061	0.089 (SAHA)
HDAC11	>100 μM 18% at 100 μM	27 (TSA)

Summary of half-maximal inhibitory concentration in biochemical activity assay for each zinc-dependent HDAC. Each sample was tested in duplicate, with a 10-point dose–response of one to three dilutions, starting at 100 μM. M344 shows greater potency at inhibiting classes I and IIB HDACs.

Fig. S1.

Concentration response curves demonstrating selectivity profile of M344. Each compound is tested in duplicates in biochemical HDAC activity assays (BPS Biosciences). Each data point represents the mean percent activity ± SEM. M344 has submicromolar to low micromolar IC50 values for HDACs 1, 2, 3, 6, 8, and 10.

Effects of M344 on AD-Related Genes.

Using NanoString nCounter technology (35, 36), we investigated the effects of M344 on 71 AD-related genes after 48 h treatment of HEK cells overexpressing the familial APP Swedish double mutation (KM670/671NL) (5)—HEK/APPsw—a well-characterized AD cell model (6, 37, 38). The heat map generated from this experiment illustrates the differential expression of genes after M344 treatment. With a false discovery rate (FDR) less than 5%, several AD-related and neuroplasticity genes are significantly up- and down-regulated by M344 (Fig. 1 and Table S1). Interestingly, several genes reported to be neuroprotective when up-regulated in AD are increased by M344 treatment. Among these genes with increased expression are brain-derived neurotrophic factor (BDNF) (3.4-fold, P < 0.0001), neuregulin (NRG1) (4.8-fold, P < 0.0001), NAD-dependent deacetylase sirtuin-1 (SIRT1) (1.6-fold, P < 0.0001), a disintegrin and metalloprotease 10 (ADAM10) (1.40-fold, P < 0.0001), ADAM19 (1.5-fold, P < 0.01), and repression element-1 silencing transcription factor (REST) (1.2-fold, P < 0.0001). Of particular interest are the ADAM family members and SIRT1, which promote nonamyloidogenic APP processing thought to be beneficial in both early- and late-onset AD.

Fig. 1.

Heat map summarizing up- and down-regulation of genes after M344 treatment of HEK/APPsw cells. Green indicates down-regulation of gene expression. Red indicates up-regulation. Changes are considered significant if FDR < 0.05, P < 0.05, and fold change > 1.2. n = 6. NanoString Data were analyzed using nSolver software 3.0.

Table S1.

Summary of gene expression changes in HEK/APPsw cells after M344 treatment

Gene name	Accession no.	Class name	Fold change	P value	FDR
IL8	NM_000584.2	Endogenous	34.9	<0.0001	<0.0001
PAI-1	NM_001165413.1	Endogenous	26.3	<0.0001	<0.0001
CD40	NM_001250.4	Endogenous	6.4	<0.0001	<0.0001
NRG1	NM_004495.2	Endogenous	4.8	<0.0001	<0.0001
CLU(APOJ)	NM_203339.2	Endogenous	4.3	<0.0001	<0.0001
BDNF	NM_170732.4	Endogenous	3.4	<0.0001	<0.0001
SIRT4	NM_012240.1	Endogenous	3.1	<0.0001	<0.0001
GM-CSF	NM_016584.2	Endogenous	3.0	<0.0001	<0.0001
APP	NM_000484.3	Endogenous	2.9	<0.0001	<0.0001
GM-CSF-R	NM_006140.3	Endogenous	2.4	<0.0001	<0.0001
Tau(MAPT)	NM_016834.3	Endogenous	2.2	<0.0001	<0.0001
BIN1	NM_004305.2	Endogenous	2.2	<0.0001	<0.0001
ABCA7	NM_033308.1	Endogenous	2.1	<0.0001	<0.0001
SIRT7	XR_430032.1	Endogenous	2.0	<0.0001	<0.0001
IL1	NM_000575.3	Endogenous	1.8	<0.0001	<0.01
GSK3-alpha	NM_002093.2	Endogenous	1.7	<0.0001	<0.0001
SIRT6	NM_001193285.1	Endogenous	1.6	<0.0001	<0.0001
SIRT1	NM_012238.4	Endogenous	1.6	<0.0001	<0.0001
ADAM19	NM_023038.3	Endogenous	1.5	<0.01	0.06
SOD1	NM_000454.4	Endogenous	1.5	<0.0001	<0.0001
HDAC11	NM_024827.3	Endogenous	1.4	<0.0001	<0.05
ADAM10	NM_001110.2	Endogenous	1.4	<0.0001	<0.05
HDAC4	NM_006037.3	Endogenous	1.4	<0.0001	<0.05
PSEN1	NM_000021.2	Endogenous	1.3	<0.0001	<0.01
BRPF1	NM_001003694.1	Endogenous	1.3	<0.01	<0.05
TNF-alpha	NM_000594.2	Endogenous	1.3	<0.05	0.26
BRD2	NM_005104.2	Endogenous	1.3	<0.0001	<0.0001
PSEN2	NM_000447.2	Endogenous	1.2	<0.0001	<0.0001
HDAC9	NM_014707.1	Endogenous	1.2	<0.05	0.27
PICALM	NM_007166.2	Endogenous	1.2	<0.05	0.21
REST	NM_001193508.1	Endogenous	1.2	<0.0001	<0.05
PEN2	NM_172341.1	Endogenous	1.2	<0.0001	<0.0001
CD147	NM_198590.1	Endogenous	1.2	<0.01	0.06
BRD4	NM_014299.2	Endogenous	1.1	<0.0001	<0.01
HDAC3	NM_003883.3	Endogenous	1.1	<0.01	<0.05
PPP1CA	NM_002708.3	Endogenous	−1.2	<0.01	<0.05
JARID1C/SMCX	NM_004187.2	Endogenous	−1.2	<0.0001	<0.0001
HDAC6	NM_006044.2	Endogenous	−1.3	<0.0001	<0.0001
SIRT2	NM_012237.3	Endogenous	−1.3	<0.0001	<0.01
SETDB1	NM_001145415.1	Endogenous	−1.3	<0.0001	<0.0001
ADAM9	NM_001005845.1	Endogenous	−1.3	<0.05	0.22
GSK3-beta	NM_019884.2	Endogenous	−1.4	<0.0001	<0.0001
SIRT5	NM_012241.3	Endogenous	−1.5	<0.0001	<0.0001
FXN	NM_001161706.1	Endogenous	−1.5	<0.0001	<0.0001
CD40L	NM_000074.2	Endogenous	−1.5	<0.01	0.09
HTT	NM_002111.6	Endogenous	−1.6	<0.0001	<0.05
BRD3	NM_007371.3	Endogenous	−1.6	<0.0001	<0.0001
BACE1	NM_012104.3	Endogenous	−1.7	<0.0001	<0.0001
APOE(e4)	NM_000041.3	Endogenous	−1.8	<0.0001	<0.0001
APH-1	NM_001077628.1	Endogenous	−1.8	<0.0001	<0.0001
APOE(e3)	NM_000041.3	Endogenous	−1.8	<0.0001	<0.0001
EHMT1	NM_024757.3	Endogenous	−2.0	< 0.0001	<0.0001
CXCR2	NM_001557.2	Endogenous	−2.0	<0.0001	<0.05
HDAC10	NM_032019.5	Endogenous	−2.0	<0.0001	<0.0001
BRD1	NM_014577.1	Endogenous	−2.1	<0.0001	<0.0001
HDAC7	NM_001098416.2	Endogenous	−2.2	<0.0001	<0.0001
NCSTN	NM_015331.2	Endogenous	−3.2	<0.0001	<0.0001
BACE2	NM_012105.3	Endogenous	−3.2	<0.0001	<0.0001

A similar trend toward nonamyloidogenic processing and anti-AD protection is also observed in significantly down-regulated genes depicted in the NanoString nCounter heat map. Among them are glycogen synthase kinase 3-β (GSK3β) (−1.4-fold, P < 0.0001), Nicastrin (NCSTN) (−3.2-fold, P < 0.0001), anterior pharynx-defective 1 (APH1) (−1.8-fold, P < 0.0001), β-site APP-Cleavage Enzyme 1 (BACE1) (−1.7-fold, P < 0.0001), BACE2 (−3.2-fold, P < 0.0001), cluster of differentiation 40 ligand (CD40L) (−1.5-fold, P < 0.01), and C-X-C Motif Chemokine Receptor 2 (CXCR2) (−2.0-fold, P < 0.0001), which are all genes hypothesized to counter AD phenotype and pathogenesis (37, 39–41). In the case of late-onset AD (LOAD) genes, apolipoprotein-E-ε4 (APOEε4) is reduced (−1.8-fold, P < 0.0001), which may be therapeutically beneficial (21, 42). There is also a significant increase observed with the bridging integrator 1 (BIN1) (2.2-fold, P < 0.0001)—reported to increase tau pathology and BACE1-dependent processing of APP (43, 44). Adenosine triphosphate-binding cassette subfamily A member 7 (ABCA7) is also up-regulated (2.1-fold, P < 0.0001), which is thought to be protective. ABCA7 loss of function is a risk factor for LOAD, and deficiency in ABCA7 increases production of Aβ (45, 46). Several Alzheimer’s-related genes tested such as complement receptor 1 (CR1), interleukin 10 (IL10), cluster of differentiation 33 (CD33) and APOE-ε2 showed no change in gene expression by M344, showing that this molecule does not randomly affect all genes.

M344 Effects on α- and β-Secretases and APP Processing.

With the observation of significant increases in several α-secretases and decreases in β-secretases in the NanoString experiments we confirmed the effect of M344 on ADAM10 and BACE1 (the two predominant α- and β- secretases involved in brain APP processing) using real-time (RT) qPCR and Western blotting (Fig. 2). Treatment of HEK/APPsw cells with 10 μM of M344—a concentration that will inhibit target HDACs (Table 1), and which we show displays no toxicity (Fig. 3)—resulted in significant increase in ADAM10 gene expression (1.80-fold, P < 0.0001) and protein levels (121.0%, P < 0.001), similar to results obtained with the NanoString. BACE1 gene expression (−3.6-fold, P < 0.0001) and protein level (−58.1%, P < 0.0001) also were confirmed to decrease after treatment of HEK/APPsw cells with M344, replicating the NanoString results (Fig. 2).

Fig. 2.

Effects of M344 on ADAM10, BACE1, and APP processing in HEK/APPsw cells. (A) RT-qPCR data showing significant increase of ADAM10 and (B) significant decrease of BACE1 after M344 treatment. (C) Representative Western blots of APP metabolites ADAM10 and BACE1 after M344 and garcinol treatments. Densitometry of bands from Western blots show significant increases in (D) sAPPα and (E) CTFα after M344 treatment. (F) There is a significant increase in the ADAM10 98kDa precursor compared with DMSO controls and (G) a significant decrease in BACE1 expression with M344. (H) M344 significantly increases immature APP and (I) decreases mature APP. All cells were treated with either 0.2% DMSO buffer or 10 μM of compounds in 0.2% DMSO. *P < 0.05, **P < 0.01, ****P < 0.0001; n = 3; mean ± SEM.

Fig. 3.

Effects of different HDAC inhibitors on Aβ42/Aβ40 ratio and cell viability. (A) Several HDAC inhibitors significantly reduce Aβ42/Aβ40 ratio at 10 μM concentration. (B) M344 presents no effect on cell viability, whereas SAHA, oxamflatin, and trichostatin significantly reduce cell viability. All drugs were tested in duplicates. Mean ± SEM; ***P < 0.001, ****P < 0.0001.

Because we observed significant regulation of several APP-cleaving secretases after treatment of HEK/APPsw cells with M344, we hypothesized that there will be an increase in full-length APP (holo-APP) in the presence of M344. Unexpectedly, we observed a significant increase (361.9%, P < 0.0001) of immature APP after treatment with M344 (Fig. 2). We also investigated the levels of sAPPα and CTF-α, two APP metabolites that result from α-secretase cleavage of APP, and observed significant increases (118.0%, P < 0.0001 for sAPPα and 35.9% for CTF-α, P < 0.05), functionally supporting the increase of α-secretases and decrease in β-secretase observed in the NanoString, RT-qPCR, and with Western blots. As an additional control, we used garcinol, a histone acetyl transferase (HAT) inhibitor of p300 and PCAF (47), hypothesizing that a HAT inhibitor would cause opposite effects from those observed with M344. Garcinol caused significant increases in both mature APP (37.9%, P < 0.01) and BACE1 (54.3%, P < 0.0001), whereas M344 treatment resulted in significant decreases in these APP processing parameters, as described above. Moreover, garcinol treatment caused sAPPα to significantly decrease (−33.6%, P < 0.01) compared with a significant increase of 118% observed with M344, further supporting a histone acetylation-dependent mechanism. We also show, in these cells, that M344 significantly increases acetylation of H3K27 (245.3%, P < 0.01) and H4K12 (95.5%, P < 0.05) after 48 h of treatment (Fig. S2). We also show a time-dependent increase of both pan-lysine and H4K12 acetylation (Fig. S3). We further support an HDAC-dependent effect by shRNA silencing of class I and IIb HDACs—targets of M344—and show significant increases in CTFα, ADAM10, and holo-APP protein levels with silenced HDACs 1, 2, 3, and 6 (Fig. S4).

Fig. S2.

Representative Western blot of histone acetylation in HEK/APPsw cells. Purified histone extracts from cells treated with M344 for 48 h present significant increases of acetylation at (A) H3K27 and (B) H4K12 residues. n = 3; mean ± SEM; *P < 0.05, **P < 0.01.

Fig. S3.

Effects of M344 treatment of HEK/APPsw cells on histone acetylation over time. Purified histones from M344-treated HEK/APPsw cells show that M344 increases histone acetylation in a time-dependent manner as revealed by both increased H4K12 acetylation and increased pan-acetylation of lysine residues. n = 3; mean ± SEM; *P < 0.05, **P < 0.01. ns, not significant.

Fig. S4.

Effects of silencing of classes I and IIb HDACs in HEK/APPsw cells. (A) RT-qPCR results showing that HDACs 1, 2, 3, 6, 8, and 10 are silenced in the HEK/APPsw cells. (B) Representative Western blot and quantification showing that silencing HDACs 1, 2, 3, and 6 significantly increases ADAM10 protein levels. (C) Representative Western blot and quantification showing significant increases in holoAPP and in CTFα when HDACs 1, 2, 3, and 6 are silenced. (D) ELISA results showing that Aβ1–42 significantly decreases with silencing of HDACs 1, 2, 3, 6, and 10 but increases with silencing of HDAC 8. (E) ELISA results showing that Aβ1–40 significantly decreases with HDACs 1, 2, 3, 6, and 10 but increases with HDAC 8. (F) Calculation of Aβ1–42/Aβ1–40 ratio shows that only HDAC 3 significantly decreases the ratio whereas HDACs 1, 2, 8, and 10 significantly increase the pro-AD Aβ1–42/Aβ1–40 ratio. n = 3; mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Effects of M344 on Aβ Accumulation.

Since there was a shift toward nonamyloidogenic processing, we hypothesized that Aβ level would decrease in the presence of M344. We indeed observed a significant decrease of Aβ in HEK/APPsw cells treated with 10 μM of M344. To further validate an HDAC mechanism we tested several other HDAC inhibitors that also significantly reduced Aβ1−42/Aβ1–40 accumulation in these cells (Fig. 3). We then verified that these effects were caused by an effect on Aβ accumulation and not due to cytotoxicity by performing a cell viability assay (CellTiter-Glo; Promega) on treated cells versus controls (Fig. 3). The cell viability results demonstrate that cells treated with M344 showed no cell death and that M344 appears to be less toxic than the other HDAC inhibitors tested.

Effects of M344 on APP Trafficking Genes.

Considering immature APP (N-glycosylated) is significantly increased with M344 treatment (Fig. 2 and Fig. S5), and with the knowledge that immature APP localizes mostly to the early protein secretory pathway, then matures upon trafficking to the late secretory pathway where it is cleaved by β-secretase, we hypothesized that M344 affects gene expression of proteins involved in APP trafficking. MINT2 (APBA2, X11L) and FE65 (APBB1) expressions are two important regulators of APP endocytosis shown to increase immature APP in the early secretory pathway, subsequently preventing APP interaction with BACE in the late endosome (2, 48). Interestingly, we observed significant increases of both MINT2 (2.7-fold, P < 0.01) and FE65 (1.7-fold, P < 0.05) gene expression in HEK/APPsw cells treated with M344 (Fig. 4 ). These data further support the anti-AD profile of M344 because increased MINT2 or FE65 has been linked to decreased Aβ production and less amyloid deposition in APP transgenic mice brain (49, 50).

Fig. S5.

Representative Western blot of cell surface APP. The N termini of cell surface proteins were biotinylated on live HEK/APPsw cells with sulfosuccinimidyl-6-[biotin-amido]hexanoate (Sulfo-NHS-LC-Biotin; Thermo Fisher Scientific). Cell surface proteins were then captured with streptavidin beads, boiled in Laemmli buffer, separated by polyacrylamide gel electrophoresis, and transferred onto PVDF membranes. Hybridization of these membranes with anti-APP-CTF antibody (Calbiochem) shows that there is an accumulation of both mature and immature APP at the cell surface, further supporting an effect of M344 on APP trafficking. n = 4, mean ± SEM; **P < 0.01.

Fig. 4.

Gene expression levels of APP trafficking and neuroprotective genes after M344 treatment of HEK/APPsw cells. Treatment with 10 μM of M344 causes a significant increase in (A) FE65 and (B) MINT2. (C) RT-qPCR results show BDNF gene expression is significantly increased in the presence of 10 µM M344. (D) RT-qPCR data show that the overexpression of APPsw decreases REST levels below baseline, comparing HEK + DMSO versus HEK/APPsw + DMSO. Compound M344 significantly increases REST gene expression in HEK-293 cells and in HEK/APPsw cells. (E) Western blot showing that M344 significantly increases BDNF protein level. n = 3–6; mean ± SEM; *P < 0.05, **P < 0.01, ****P < 0.0001.

M344 Effects on Neuroprotective Genes BDNF and REST.

Treatment of HEK/APPsw cells for 48 h revealed significant increases of both BDNF (7.1-fold, P < 0.0001) and REST (4.2-fold, P < 0.001) gene expression (Fig. 4 ) and of BDNF protein expression (42.3%, P < 0.01) (Fig. 4). Due to lack of a reliable REST antibody we were unable to determine REST protein levels. M344 also increased REST gene expression in control HEK-293 cells, although to a lesser extent compared with HEK/APPsw (1.2-fold, P < 0.05) (Fig. 4). Fig. 4 also shows that the presence of APPsw in the cells significantly reduced REST expression (−2.0-fold, P < 0.05), reminiscent of the human condition reported in AD patients (51).

M344 Is Brain-Penetrant and Increases Histone Acetylation in Vivo.

We conducted pharmacokinetic studies with 10 mg/kg of M344 injected i.p. Fig. 5 show that M344 concentrations peak rapidly at 15 min in both plasma and the brain. Of note, M344 reaches brain concentrations of 47 ng/mL (P < 0.05), a value equivalent to 0.13 μM that is sufficient to inhibit HDACs 1, 2, 3, 6, and 10 as shown in Table 1 and Fig. S1. Fig. 5 show that M344 significantly increases acetylation of histone H4K12 in the frontal cortex, but not in the cerebellum. Fig. 5 shows that after 15 min of 10 mg/kg i.p. injection there is ∼1.4% brain/plasma ratio with 3.1 ± 0.71 μM free concentration and 0.39 ± 0.04 free fraction of M344 in plasma.

Fig. 5.

Pharmacokinetics and histone acetylation after M344 treatment of wild-type mice. (A) I.p. injection of 10 mg/kg of M344 results in significant increase of M344 brain concentration after 15 min of treatment. (B) The same treatment also causes significant increase of M344 plasma concentrations after 15 min. (C) Representative Western blots from purified histone extracts show increased acetylation of H4K12 in the frontal cortex but not in the cerebellum after i.p. injection with 10 mg/kg of M344 for 30 min. (D) Quantification of frontal cortex Western blots from purified histone extracts shows significant increase in H4K12 acetylation after M344 treatment. (E) Quantification of cerebellum Western blots shows there is no significant difference in acetylation at the H4K12 residue in purified histone extracts from the cerebellum after M344 treatment. (F) Summary of M344 free fraction, free concentration, and brain/plasma ratio levels after 10 mg/kg i.p. treatment. n = 3; mean ± SEM; *P < 0.05, ****P < 0.0001. ns, not significant.

Effects of M344 on Y-Maze Spontaneous Alternation in 3xTg APPsw/PS1M146V/TauP301L Mice.

We then tested M344 in the 3xTg AD mice overexpressing APPsw, TauP301L, and Presenilin 1 (PS1, PSEN1) (52) using a battery of behavioral tests. In 3xTg AD mice that were repeatedly i.p. treated with M344 (for ∼4 mo, as described in ) we observed a dose-dependent increase in Y-maze spontaneous alternation (3 mg/kg, 67.0%, P < 0.05; 10 mg/kg, 71.2%, P < 0.01) compared with vehicle controls (Fig. 6). No difference in total number of arm entries was observed (Fig. 6), demonstrating no deficits in motor function and supporting that the increased spontaneous alternation observed in treated mice is due to increased spatial memory and willingness to explore new environments.

Fig. 6.

Effects of M344 treatment on behavior of the 3xTg AD mice. (A) I.p. injection of M344 increases Y-maze spontaneous alternation in mice at both 3 mg/kg and 10 mg/kg, with (B) showing no significant differences in total arm entries. (C) Open field test shows that animals treated with M344 have no locomotion deficits and (D) travel at similar velocity compared with controls. (E) Injection of 3 mg/kg and 10 mg/kg of M344 increases novel object recognition performance of mice as determined by duration of exploration and (F) frequency of novel object exploration. (G) Barnes maze acquisition trials for these mice show significantly fewer errors in trial 3 for mice treated with 10 mg/kg of M344 and in trial 5 for mice treated with 3 mg/kg or 10 mg/kg. (H) Barnes maze probe trial shows that M344 significantly increases spatial memory as determined by decreased errors in treated mice. Vehicle: n = 10; 3 mg/kg: n = 9; 10 mg/kg: n = 8; mean ± SEM; *P < 0.05, **P < 0.01. ns, not significant.

Effect of M344 on Open Field Behavior and Novel Object Recognition.

We further tested the effects of M344 on locomotor behavior using the open field test. No significant difference was observed between M344-treated animals and controls for distance traveled or velocity (Fig. 6 ). Having observed no difference in locomotor behavior in the treated and control mice, we proceeded to test for novel object recognition in the same open field arena. In this test, at both 3 mg/kg and 10 mg/kg doses, treated mice significantly outperformed control mice in novel object exploration duration (3 mg/kg: 66.5%, P < 0.05; 10 mg/kg: 57.2%, P < 0.05) (Fig. 6) and novel object exploration frequency (3 mg/kg: 47.8%, P < 0.05; 10 mg/kg: 47.3%, P < 0.05) (Fig. 6).

Effects of M344 on Barnes Maze Performance.

We further evaluated spatial memory in the 3xTg AD mice treated with 3 mg/kg and 10 mg/kg of M344 using a Barnes maze. We observed significantly fewer errors in acquisition trials 3 and 5 for mice that received 10 mg/kg of M344 (P < 0.05) and fewer errors in trial 5 for those treated with 3 mg/kg (P < 0.05) compared with vehicle-treated controls (Fig. 6). After 24 h of rest, in the probe trial both mice treated with M344 committed fewer errors than controls (3 mg/kg: −44.6%, P < 0.05; 10 mg/kg: −53.8%, P < 0.01) (Fig. 6), indicating increased spatial memory in these mice.

Effects of M344 on AD-Like Pathology in the Hippocampus of 3xTg AD Mice.

M344 significantly decreased Aβ1–42 in the hippocampus of mice treated with doses of 3 mg/kg (−42.7%, P < 0.05) and 10 mg/kg (−35.6%, P < 0.05) (Fig. 7). M344 significantly increased ADAM10 gene expression only in the hippocampus of mice treated with 10 mg/kg (2.1-fold, P < 0.05) (Fig. 7). Only treatment with 3 mg/kg of M344 resulted in a significant decrease of BACE1 gene expression (−1.8-fold, P < 0.05) (Fig. 7). We also observed significant decrease in phosphorylation of tau at Ser396—a residue found in paired helical filaments in brain neurofibrillary tangles of AD patients—at both 3 mg/kg (−58.2%, P < 0.01) and 10 mg/kg (−57.7%, P < 0.01) (Fig. 7).

Fig. 7.

Analysis of Aβ1–42, BACE1, ADAM10, and phospho-tau Ser396 in the hippocampus of 3xTg AD mice. (A) M344 significantly reduces levels of Aβ1–42 at 3 mg/kg, as determined by ELISA. (B) RT-qPCR results show that ADAM10 gene expression is significantly increased in the hippocampus of 3xTg mice treated with 10 mg/kg. (C) BACE1 mRNA level is significantly reduced in the hippocampus of mice treated at 3 mg/kg. (D) Both 3 mg/kg and 10 mg/kg of M344 significantly decrease tau phosphorylation at serine residue 396, as determined by ELISA. Mean ± SEM; *P < 0.05, **P < 0.01. ns, not significant.

Discussion

We show that the HDAC inhibitor M344 is a potent inhibitor of class I and class IIB HDACs that simultaneously regulates the expression of several high-priority genes related to EOAD, LOAD, synaptic plasticity, and neuroprotection in the HEK/APPsw cell model (Table 1, Fig. 1, and Fig. S1). In support of the gene expression data, we show that M344 significantly reduces Aβ1–42/Aβ1–40 ratio with no negative effects on cell viability, while also appearing to have a better in vitro toxicity profile than other HDAC inhibitors tested (Figs. 1–3). A mechanism that can explain this decrease in Aβ1–42/Aβ1–40 ratio is the M344-induced down-regulation of γ-secretase complex components NCSTN and APH1 (Fig. 1), which would reduce APP cleavage at the relevant sites. However, down-regulating the γ-secretase complex—comprising PSEN1 or PSEN2, PEN2, NCSTN, and APH1—is troublesome since γ-secretase also cleaves NOTCH, a transmembrane protein whose cleavage products are reported to promote neurogenesis. Inhibition of NOTCH processing has been cited as a possible cause of the recent γ-secretase inhibitor clinical trial failures (12–14). Although both NCSTN and APH1 are significantly down-regulated with M344, we would not expect a decrease in the processing of NOTCH since M344 also increases other components of the complex in PEN2, PSEN1, and PSEN2 (Fig. 1), which have been demonstrated to be sufficient for γ-secretase–dependent NOTCH processing (53). The decrease in Aβ1–42/Aβ1–40 ratio could also be the result of a combined effect of decreasing the expression of CXCR2, NCSTN, and APH1. Depletion of CXCR2 has been reported to reduce γ-secretase cleavage of APP (40). The effect of M344 or other HDACs on CXCR2-mediated γ-secretase APP processing is not known and deserves further investigation. Because we only observed significant reduction of Aβ1–42/Aβ1–40 ratio in cells silenced for HDAC3 (Fig. S4), it may be worth investigating the effects of HDAC3 function on CXCR2 and γ-secretase components.

In cases of EOAD involving APP mutations such as APPsw, the overproduction of Aβ is often due to excess cleavage by β-secretases (6). Here, we show that compound M344 significantly reduces β-secretases BACE1 and BACE2, concurrent with observed decreases in the accumulation of Aβ1–42 (Figs. 2, 3, and 7), indicating that HDAC inhibition is an alternative approach to BACE1 inhibition. Of note, a BACE1 inhibitor, verubecestat (MK-8931), recently failed in phase III trials (10). Thus, an epigenetic compound that is able to reduce BACE1-mediated metabolites as one of its targets in the AD network represents a novel way to regulate BACE activity, which has been challenging (54).

The up-regulation of α-secretase has been proposed as a highly desirable therapeutic target for AD. Here, we report that the M344 compound also significantly increases the gene expression of α-secretases ADAM10 and ADAM19, concurrent with the observed increases in the metabolites sAPPα and CTFα after M344 treatment (Fig. 2 and Fig. S6). Similar effects have been reported with the HDAC inhibitor apicidin up-regulating the expression of ADAM10 via an HDAC2/3 mechanism involving the transcription factor USF-1 (55). Here we show in HEK/APPsw cells that it is possible that the effect on ADAM10 is mediated by HDACs 1, 2, 3, and 6 because M344 inhibits these HDACs, and silencing experiments caused increased protein levels of ADAM10 and CTFα (Fig. S4 ). We also observe a significant increase in SIRT1 expression, which has been shown to promote ADAM10 cleavage of APP (56). Thus, it is also possible that the increased nonamyloidogenic processing induced by M344 is partially a SIRT-1-mediated effect, making M344 an HDAC inhibitor affecting both zinc-dependent and class III NAD-dependent HDACs.

Fig. S6.

Representative Western blot of M344 treatment of CHO cells overexpressing wild-type APP. These data show that treatment of CHO/APP cells with M344 results in significant increase of sAPPα and CTFα, supporting that the effect of M344 is not cell-dependent, and that its effect on APP also occurs in the wild-type version of the protein. n = 3; mean ± SEM; *P < 0.05, ***P < 0.001, ****P < 0.0001.

We also observed, in HEK/APPsw cells, significant increases in the APP trafficking genes MINT2 and FE65, and of the neuroprotective genes BDNF, NRG1, and REST, all genes reported to be beneficial against AD if up-regulated (Fig. 4). Increased REST expression correlates with healthy aging, cognitive preservation, and longevity (51). Our findings support studies by others that have shown that the HDAC inhibitor SAHA and inhibition of HDACs 2 and 3 increase BDNF gene expression (57, 58). Further, large concentrations—0.8–5 mM—of β-hydroxybutyrate increase BDNF expression via inhibition of HDACs 2 and 3 (59). Since M344 inhibits these two HDACs, M344-mediated induction of BDNF expression (Fig. 4 and Fig. S7) is likely due to activity on HDACs 2 and 3. To our knowledge, inhibition of class I and IIb HDACs has not previously been shown to increase the expression of APP trafficking genes MINT2 and FE65 involved in decreased APP cleavage.

Fig. S7.

Gene expression of BDNF gene expression in different cells. Treatment of HEK-293, SH-SY5Y, and N2A cells with 10 μM of M344 significantly increases BDNF gene expression, demonstrating that effects are not cell-type-specific. n = 6; mean ± SEM; **P < 0.01, ****P < 0.0001.

Among LOAD-related genes, M344 decreases the expression of APOEε4—for which the presence of just one ε4 allele represents the greatest risk factor of developing AD (21, 42). M344 also increases the expression of BIN1—the second-greatest reported LOAD risk factor (43, 44). Effects of M344 at both genes would be expected to be protective. Increased APOEε4 also elevates Aβ accumulation (60). Decreased BIN1 has been reported to promote tau pathology (43). Similarly, M344 also up-regulates other LOAD risk-factor genes (i.e., ABCA7 and PICALM) whose deficiencies have been shown to promote AD pathology (43–46, 61). M344 also shows significant decrease of CD40L, the cognate ligand of CD40, which has been proposed as a diagnostic biomarker in LOAD (62), and whose signaling has been reported to increase Aβ-induced microglial activation and plaque-associated tau phosphorylation in AD mice (39, 63). The M344-mediated reduction in CD40L likely results in interruption of CD40-CD40L interaction. Such a disruption of CD40–CD40L signaling has been shown to be beneficial in reducing AD-like pathogenesis and increase cognition in AD mice (37, 39, 64).

We show that i.p. treatment of mice with 10 mg/kg of M344 causes a maximum plasma concentration of ∼8.8 μM (Cmax) and gets into the brain with a peak of about 0.13 μM (Cmax) after 15 min of treatment. That concentration is high enough to reach the IC50 values of HDAC1 (0.048 μM), HDAC2 (0.12 μM), HDAC3 (0.032 μM), HDAC6 (0.0095 μM), and HDAC 10 (0.061 μM) but not HDAC8 (1.34 μM). With a low molecular weight (307.4), a LogP of ∼1.06, as well as high free fraction and free concentration levels in plasma (Fig. 5), M344 has the properties of a brain-penetrant compound. The fact that brain plasma ratio ranges from 1.4% at 15 min to 1.7% at 30 min suggests quick removal by brain Pgp and Bcrp efflux transporters, similar to what is observed with SAHA, a related compound (65). Despite its high rate of removal in the brain, M344 causes significant increases of H4K12 acetylation in the cortex of mice, but not in the cerebellum (Fig. 5). Deregulation of H4K12 acetylation is linked to cognitive impairment associated with aging, and increased acetylation at that mark may rescue memory (66, 67). Our data with the 3xTg mice indicate that one dose per day of M344 at 3 mg/kg or 10 mg/kg for 4 mo is enough to trigger an anti-AD profile without observable adverse effects. It is plausible that over the course of 4 mo the relatively low Cmax is the reason no toxicity is observed with M344 treatment.

We show that treatment of the well-established 3xTg AD mouse model with doses as low as 3 mg/kg of M344 results in improvement of learning and memory in different behavioral tests, with no effects on locomotor activity (Fig. 6). Indeed, we observed significant increases in Y-maze spontaneous alternation, a measure of hippocampus-dependent spatial memory and the willingness of mice to explore new environments (68, 69). We also observed superior performance of the 3xTg AD mice treated with M344 in both the novel object recognition test and the probe test of the Barnes maze—a spatial memory test similar to the Morris water maze. Interestingly, Tg2576 AD mice treated with 25 mg/kg and 50 mg/kg of SAHA (also known as vorinostat), a compound closely related to M344, has shown positive effects on synaptic plasticity at the long-term potentiation level, but not behaviorally in the fear conditioning paradigm (65). Such a discrepancy with our study could be due to the different tests used, the animal model, age of animals, and length of treatment. In a different animal model (aged APP/PS1 mice), Kilgore et al. (70) saw improvement of cognitive behavior with i.p. injections of 50 mg/kg of SAHA, supporting an HDAC class I inhibition approach in their paper. Another study using SAHA administered 2 mg/d orally to aged (10-mo-old) APP-PS1-21 AD mice observed partial improvement of spatial memory, reduction of transcriptional inflammatory response, and increased H4K12 acetylation, with no significant differences observed in Aβ plaques (66). Although they used different animals and paradigms, focusing more on aging and the transcriptome, the Kilgore et al. (70) and the Benito et al. (66) studies both support our findings that targeting several HDACs with a small-molecule inhibitor provides a multifactorial approach to normalize AD-related genes. The lack of difference in Aβ levels observed in Benito et al. (66) is likely due to three main differences between the studies: (i) the length of treatment (4 wk versus 4 mo in our case), (ii) the stage of AD-like symptoms of the animals at the start of treatment (postdisease state versus presymptomatic), and (iii) the Aβ measurement technique (immunohistochemistry versus ELISA). Overall, the two studies are concordant. Other studies with HDACis such as 200 mg/kg sodium 4-phenylbutyrate in Tg2576 AD mice (71) or 50 mg/kg of the class II mercaptoacetamide compound W2 in 3xTg AD mice (72) also show positive effects on memory, further suggesting feasibility of the approach. However, somewhat in contrast, treatment of 3xTg AD mice with even low dose of M344 significantly decreases levels of the molecular targets BACE1, Aβ1–42, and phospho-tau Ser396 in the hippocampus. Separate studies have reported other HDACis to be beneficial for AD, either due to induced increase in BDNF gene expression, or decrease in GSK3β expression, or decrease in tau phosphorylation, or decrease in Aβ accumulation to increase cognition in AD mouse models (31, 70, 73–77). Here, we propose that such effects are due to the multitarget nature of these HDAC inhibitors, similarly to M344, and not to a single target.

It is important to emphasize that most of the work performed with HDACis on AD models in the literature has been conducted on old animals, after AD-like disease onset. This approach has yielded poor results in AD patients. Many clinical trial failures have been on old patients with mild to moderate disease, and reversing the pathology may not readily result in alleviation of symptoms. As the field is moving toward trials on preclinical/prodromal AD populations (10), we opted to start treatment of mice before the development of disease. Since 3xTg mice have been reported to present overt molecular and behavioral AD-like pathology at the age of 6 mo, we started treatment of 3-mo-old presymptomatic 3xTg mice 5 d a week for about 4 mo to evaluate the possibility of low doses of this drug as a preventive measure for AD. This method was successful at preventing AD-like pathogenesis at molecular and behavioral levels. This study does not, however, show whether M344 would continue to be beneficial for longer-term studies (i.e., beyond 7 mo of age when AD-like symptoms are more severe). Other limitations of the work presented here include the possibility of HDACis increasing the acetylation of proteins other than histones, such as tau acetylation reported to be a promoter of tau pathology (78, 79). We have, for instance, shown that treatment of HEK/APPsw cells with M344 results in an approximately sixfold increase in acetylated α-tubulin (Fig. S8), a target of HDAC6 (80). However, other possible mechanisms are beyond the scope of the present study showing that a multitarget approach is a plausible alternative to the one-target–one-disease paradigm in the context of AD.

Fig. S8.

Effects of M344 on α-tubulin acetylation in HEK/APPsw cells. Treatment of HEK/APPsw cells with 10 μM of M344 results in a significant increase of the α-tubulin acetylation, an HDAC6 target. n = 3; mean ± SEM; *P < 0.05.

Finally, pharmacokinetic studies demonstrated that M344 is brain-penetrant, and that our in vivo dosing regimens resulted in sufficiently high CNS concentrations (comparable to the concentrations required to affect gene expression in vitro), but that the drug clears rapidly from plasma and brain. We therefore suggest that drug efficacy relates to Cmax (discussed above) and that prolonged daily exposure is likely not required. The observations that (i) in vivo efficacy was observed with much lower doses of this HDACi than typically used in mouse models of cancer and other CNS disorders (81, 82) and that (ii) short periods of high brain exposure seem to be sufficient for efficacy indicate that it may be possible to avoid adverse effects in possible future attempts to use M344 (or related compounds) to treat humans with AD or related disorders.

Conclusion

Using a multifactorial approach to fight a multifactorial disease is necessary. Since the single-target approach has been essentially unsuccessful to date in the treatment of AD, we aimed to use a broader-acting molecule to address the polygenic nature of this disease. In this paper, using an epigenetic approach, we show that it is possible to use one drug compound that simultaneously addresses several aspects of AD, including down- and up-regulation of key AD and neuroprotective genes. We demonstrated that M344, displaying sufficient but transient brain exposure, can prevent memory impairment in the 3xTg (APPsw/PSEN1M146L/TauP301L) AD mouse model. Efficacious in vivo doses of this HDAC inhibitor appear to be much lower than those typically used in animal models for cancer, and brief daily brain exposure seems sufficient. More work is needed with other small-molecule epigenetic compounds to identify the ideal anti-AD profile.

Materials and Methods

Detailed materials and methods are provided in . Below are brief descriptions.

Cell Culture.

HEK (HEK-293) cells were purchased from ATCC. They were cultured under standard conditions (37 °C, 5% CO2, 95% air) in Advanced DMEM supplemented with 10% (vol/vol) FBS, penicillin (100 U/mL), streptomycin (100 μg/mL), and Primocin (100 μg/mL). HEK cells overexpressing APP with the Swedish mutation (HEK/APPsw) were a gift from Dennis Selkoe, Brigham and Women’s Hospital and Harvard Medical School, Boston, and were cultured in the same media as the HEK-293 cells supplemented with 250 μg/mL of G418 as a selection agent.

NanoString Gene Expression Analysis.

For NanoString experiments, cells were treated with either 10 μM of M344 in 0.2% DMSO buffer or 0.2% DMSO buffer alone, in T-75 flasks (n = 6). Total RNA was extracted and then used to perform NanoString experiments described in detail by our group (83). The nCounter analysis system (NanoString Technologies) was used to quantify target RNA molecules using these color-coded molecular barcodes. Genes whose fold-change expression was statistically significant and FDR was less than 5% were used for further analysis. A P value threshold was set at 0.05.

HDAC Activity Assay.

The selectivity profile of M344 was determined biochemically by performing activity assays in duplicate with each of the 11 zinc-dependent HDACs at 10-point 1:3 dilutions, starting at 100 μM (BPS Biosciences). SAHA was used as a positive control for HDACs 1, 2, 3, 6, and 10 since it is known to inhibit those enzymes. TSA was used as a positive control for HDACs 4, 5, 7, 8, 9, and 11 as it has been reported to inhibit these HDACs. All HDAC substrates, buffers, and developers were from BPS BioSciences. Fluorescence signal was measured at 360-nm excitation and 460-nm emission using a Tecan Infinite M1000 microplate reader. Curves were generated with GraphPad Prism 6.0, using a four-parameter nonlinear curve fit to determine the concentration causing IC50 values.

ELISAs and Western Blots.

Aβ1–40 and Aβ1–42 were measured from the media and from brain tissue by ELISA with the Novex kit (Life Technologies). An AlphaLISA kit from PerkinElmer was also used to measure Aβ1–42 levels in cells. The phosphorylation level of tau at Ser396 was measured using an ELISA kit from Thermo Fisher Scientific. All of the kits were used per the manufacturer’s instructions. For Western blots, electrophoresed proteins were transferred onto PVDF membranes. All of the primary antibodies were used at 1:1,000 dilution. Membranes were developed using the Clarity ECL detection reagents (Bio-Rad), visualized and then quantified by densitometry, using the Image J software from the NIH.

RT-qPCR.

After total RNA extraction, cDNA was synthesized with random hexamers and Moloney Murine Leukemia Virus (M-MLV) reverse transcriptase. Extracted cDNA was used for RT PCR with primers and Taqman Master Mix from Life technologies/Thermo Fisher Scientific. Samples were then amplified for 40 cycles using the Applied Biosystems FAST Real-Time PCR Detection System 7900HT or the Applied Biosystem Quantstudio Flex Real-Time PCR System and analyzed with the SDS Real-Time PCR analysis software (Applied Biosystems). The results presented are based on fold change using the 2^−ΔΔCt method.

Animals and Treatment.

We used the triple transgenic (3xTg-AD) mice that overexpress three human transgenes: the APP Swedish double mutation KM670/671NL (APPsw), the presenilin-1M146V mutation (PS1M146V), and the TauP301L mutation (52). Mice were purchased through The Jackson Laboratory, from the NIH-supported Mutant Mouse Regional Resource Center (mmrrc). A cohort of 30 mice was used (50% males and females). Three groups of 10 (5 males and 5 females) were treated intraperitoneally with either vehicle, 3 mg/kg, or 10 mg/kg of M344 diluted in vehicle. An AD prevention paradigm was used where animals were treated from the age of 3 mo, before onset of AD-like pathology. The treatment regimen consisted of 5 d of injection per week for about 4 mo, until animals were killed after behavioral experiments. Behavioral tests were conducted on the mice in the following order: Y-maze spontaneous alternation, open field, novel object recognition, and Barnes maze. For wild-type animal studies, groups of three mice were i.p. treated with 10 mg/kg of M344 at the following time points: 15 min, 30 min, 60 min, 120 min, and 24 h. Brain and plasma were collected as described above and in .

All experiments were approved by the University of Miami Miller School of Medicine Institutional Animal Care and Use Committee and conducted according to specifications of the NIH as outlined in the Guide for the Care and Use of Laboratory Animals (84).

Brain Distribution and in Vivo Pharmacokinetics.

Methods for brain distribution and in vivo pharmacokinetics studies are described in ref. 85.

Statistical Analyses.

Unpaired Student’s t test was used whenever only two means were being compared. One-way ANOVA with either Bonferroni, Dunnett’s, or Tukey post hoc analysis was used for multiple comparisons when several means were being compared. Repeated measures two-way ANOVA with Tukey post hoc analysis was used to analyze daily performance.

SI Materials and Methods

HEK (HEK-293) cells were purchased from ATCC. They were cultured under standard conditions (37 °C, 5% CO2, 95% air) in Advanced DMEM supplemented with 10% (vol/vol) FBS, penicillin (100 U/mL), streptomycin (100 μg/mL) and Primocin (100 μg/mL). HEK cells overexpressing APP with the Swedish mutation (HEK/APPsw) cells were a gift from Dennis Selkoe and were cultured in the same media as the HEK-293 cells supplemented with 250 μg/mL of G418 as a selection agent. CHO/APP cells were also a gift from Dennis Selkoe. They were cultured under standard conditions in Advanced DMEM/F12 supplemented with 10% (vol/vol) FBS, penicillin (100 U/mL), streptomycin (100 μg/mL), Primocin (100 μg/mL), and 250 μg/mL of G418. With the exception of Primocin—purchased from Invivogen—all of the cell culture reagents were purchased from Thermo Fisher Scientific.

The selectivity profile of M344 was determined biochemically by performing activity assays in duplicate with each of the 11 zinc-dependent HDACs at 10-point 1:3 dilutions, starting at 100 μM (BPS Biosciences). SAHA was used as a positive control for HDACs 1, 2, 3, 6, and 10 since it is known to inhibit those enzymes. TSA was used as a positive control for HDACs 4, 5, 7, 8, 9, and 11 as it has been reported to inhibit these HDACs. All HDAC substrates, buffers, and developers were from BPS BioSciences. Dilutions of the compounds were prepared with 10% DMSO in HDAC assay buffer and 5 μL of the dilution was added to a 50-μL reaction, resulting in a 1% final DMSO concentration in all of the reactions. The enzymatic reactions consisted of a 50-μL mixture containing HDAC assay buffer, 5 μg BSA, an HDAC substrate, an HDAC enzyme, and a test compound, all incubated at 37 °C for 30 min. To end the reactions, 50 μL of HDAC developer was added to each well and the plate was incubated for 20 min at room temperature. Fluorescence signal was measured at 360-nm excitation and 460-nm emission using a Tecan Infinite M1000 microplate reader. For analyses, the fluorescent intensity (FH) measured in the absence of compound was defined as 100% activity. The fluorescent intensity (FL) measured in the absence of HDAC was defined as 0% activity. The percent HDAC activity was calculated according to the following equation: % activity = (F − FL)/(FH − FL), where F is the fluorescent intensity in the presence of the compound. Curves were generated with GraphPad Prism 6.0, using a four-parameter nonlinear curve fit to determine the concentration causing half-maximal percent activity (IC50) values.

Compounds.

Small-molecule compounds were purchased from Enzo Life Sciences, Tocris Biosciences, or Selleckchem. All compounds were dissolved in 100% DMSO stock, and cells were treated at 0.2% DMSO final concentration and appropriate drug concentrations to measure drug effects on protein and gene expression.

CellTiter-Glo.

Cytoxicity of different HDAC inhibitors was evaluated using the CellTiter-Glo luminescent cell viability assay from Promega, as per the manufacturer’s instructions. Briefly, cells were seeded overnight, treated for 48 h with each compound or DMSO control, and ATP was measured after treatment using the CellTiter-Glo reagent. Since only live metabolically active cells produce ATP, this assay allows measurement of viable cells in each well. Data are presented as percent of DMSO control.

For NanoString experiments, cells were treated with either 10 μM of M344 in 0.2% DMSO buffer or 0.2% DMSO buffer alone, in T-75 flasks (n = 6). Total RNA was extracted by lysing cells and performing phase separation using the TRIzol (Thermo Fisher Scientific) chloroform method as per the manufacturer’s instructions, followed by column extraction using the Qiagen RNeasy kit. The total RNA was then given to the Sylvester Oncogenomics Core to perform NanoString experiments, described in detail in Geiss et al. (35) and by our group in Pastori et al. (36). Briefly, two 100-bp probes consisted each of one capture and one reporter probe specific to 80 human mRNA targets—70 AD-related or epigenetic enzyme genes and 10 housekeeping genes. In the multiplexed reaction, all probes were simultaneously hybridized to 100 ng of total mRNA per sample. The affinity-purified capture–reporter–target complexes were immobilized to a streptavidin-coated cartridge via a biotin tag on the capture probe. Systemic variability and nonspecific background signal were normalized by including positive and negative controls in each sample lane. The reporter probe for each target RNA has a unique color-coded sequence that is detected with an automated imager after elongation, alignment, and immobilization of the molecule using electrophoresis.

The nCounter analysis system (NanoString Technologies) was used to quantify target RNA molecules using these color-coded molecular barcodes. First, raw count NanoString data produced by the nCounter analysis system was corrected to positive controls, allowing for correction of sample-to-sample variations that arise from assay-specific factors (e.g., differences in input reagents). Second, a negative corrections method was applied as follows. The mean of the negative controls for a given sample was subtracted from the positively corrected data. Next, sample content normalization was determined based on the type of statistical analysis selected for further analysis. Our data were analyzed using t test, and therefore normalization of the mRNA to housekeeping genes was applied. The normalization was calculated by , where c is the data count, m is the average of the sum of all housekeeping genes across all samples, and s is the sum of the housekeeping genes for a given sample. Finally, differential expression of all genes was calculated. Our data are assumed to be normally distributed, and hence t test was used to calculate differentially expressed genes between DMSO- and M344-treated cells. Genes whose fold-change expression was statistically significant and FDR was less than 5% were used for further analysis. A P value threshold was set at 0.01%.

Aβ1–40 and Aβ1–42 were measured from the media and from brain tissue by ELISA with the Novex kit (Life Technologies). AlphaLISA kit from PerkinElmer was also used to measure Aβ1–42 levels in cells. The phosphorylation level of tau at Ser396 was measured using an ELISA kit from Thermo Fisher Scientific. All of the kits were used as per the manufacturer’s instructions. ELISA results were corrected for the amount of protein in each treated well. Results are thus expressed in picogram per milligram of protein. For Western blots, clarified cell extracts containing equivalent amounts of proteins were mixed with equal volumes of Laemmli sample buffer (125 mM Tris⋅HCl, pH 6.8, 4% SDS, 20% glycerol, 2% DTT, and 5% β-mercaptoethanol) and denatured at 99 °C for 10 min.

Western blots were performed under denaturing conditions with Bio-Rad anykD gels for SDS/PAGE using the Bio-Rad Minigel apparatus. Separated proteins were transferred electrophoretically overnight at 4 °C or over 1 h on ice, onto PVDF membranes. Membranes were then blocked in blocking milk solution [5% (wt/vol) nonfat blocking milk (Bio-Rad) and 0.1% Tween 20 (Bio-Rad) in TBS (TBST)] for 1 h at room temperature. Membranes were then hybridized with appropriate primary antibodies (diluted in blocking milk solution) for 1 h at room temperature or overnight at 4 °C, washed with TBST for 30 min (three 10-min washes), and hybridized with appropriate HRP-conjugated secondary antibodies (Santa Cruz Biotechnology) for 1 h. Membranes were developed using the Clarity ECL detection reagents (Bio-Rad). The following antibodies were used: anti-APP-C-terminal fragment (CTF) (Calbiochem/EMD Millipore) 1:1000 dilution; 6E10 for sAPPα detection (Biolegend) 1:1,000 dilution; anti-BACE1 (Abcam) 1:1,000; anti-β-actin (AC15) (Santa Cruz) 1:1,000 dilution, anti-BDNF (Milipore) 1:1,000 dilution; anti-ADAM10 (Abcam) 1:1,000 dilution; and anti-GAPDH (Santa Cruz) 1:1,000 dilution. The developed membranes were visualized using the LI-COR C-DiGiT Chemiluminescence Western Blot Scanner or the Protein Simple FluorChem Imaging System. Protein bands were quantified by densitometry, using the ImageJ software from the NIH.

For cell culture experiments, after 48 h of treatment with compounds cells were lysed with TRIzol reagent for extraction of total RNA as per the manufacturer’s instructions (Thermo Fisher Scientific). The RNA was converted into cDNA with random hexamers and Moloney Murine Leukemia Virus (M-MLV) reverse transcriptase (Life Technologies). The extracted cDNA was prepared for RT PCR with primers and Taqman Master Mix from Life technologies/Thermo Fisher Scientific Inc. Samples were then amplified for 40 cycles using the Applied Biosystems FAST Real-Time PCR Detection System 7900HT or the Applied Biosystem Quantstudio Flex Real-Time PCR System, and analyzed with the SDS Real-Time PCR analysis software (Applied Biosystems). The results presented are based on cycle threshold (Ct) values. We calculated the differences between the Ct values for target and reference genes (β-actin or GAPDH) as ΔCt and the difference between the resulting ΔCt and that of the vehicle control (calibrator sample) to obtain the ΔΔCt. Results are presented as fold change (RQ = 2−ΔΔCt) for mRNA expression relative the vehicle. All genes tested by qPCR in these studies were amplified with Taqman primers from Life Technologies/Thermo Fisher Scientific.

Histone Extractions.

Core histones were purified from cells and brain tissue using a Histone Purification Mini Kit (40026; Active Motif) according to the manufacturer’s protocol. Briefly, after cells were grown to 90% confluency, media was discarded and cells were washed twice in low-serum media and collected into ice-cold extraction buffer. Brain tissue was homogenized in ice-cold extraction buffer, using a dounce homogenizer. After homogenization on a rotating platform, extracts were centrifuged and neutralized with neutralization buffer until a pH of 8 was achieved. Column input control (crude histone extract) was collected for further analysis. After spin columns were equilibrated with equilibration buffer, the remaining crude histone extract was passed through the column, washed three times with wash buffer, and eluted with elution buffer. Following overnight histone precipitation with 4% perchloric acid, samples were centrifuged and washed with perchloric acid and acetone. Purified histones were resuspended in sterile water, quantified, and used for analyses.

For the in vivo work we chose to investigate the effects of M344 on behavior and AD-like pathology in the triple transgenic (3xTg-AD) mice developed by Frank Laferla’s laboratory. These mice overexpress three human transgenes: the APP Swedish double mutation KM670/671NL (APPsw), the presenilin-1M146V mutation (PS1M146V), and the tau P301L mutation (APPsw/PS1M146V/TauP301L) (52). The 3xTg-AD mouse is one of the few comprehensive AD models to present both Aβ deposits and neurofibrillary tangles, indicative of the human condition. Furthermore, in addition to showing AD-like cognitive and motor impairment, these mice also present synaptic dysfunction as early as 6 mo of age. They were purchased through The Jackson Laboratory, from the NIH-supported mmrrc. A cohort of 30 mice was used (50% males and 50% females). Mice were housed five animals per cage under a regular 12-h/12-h light/dark cycle and had ad libitum access to food and water. Mice were housed in a humidity- and temperature-controlled, AAALAC-accredited animal facility at the University of Miami Miller School of Medicine. All experiments were approved by the University of Miami Miller School of Medicine Institutional Animal Care and Use Committee (IACUC) and conducted according to specifications of the NIH as outlined in the Guide for the Care and Use of Laboratory Animals (84).

Three groups of 10 (5 males per group and 5 females per group) were treated i.p. with either vehicle (saline, 5% Tween, and 2.5% DMSO) or 3 mg/kg or 10 mg/kg of M344, both prepared in vehicle. Since this is an AD prevention paradigm, animals were treated from the age of 3 mo, before onset of AD-like pathology. The treatment regimen consisted of 5 d of injection per week for ∼4 mo, until animals were killed after behavioral experiments. Three animals died during the course of treatment (one in the 3 mg/kg group and two in the 10 mg/kg group).

Behavior.

Y-maze test.

The Y-maze spontaneous alternation test measures exploratory behavior based on the willingness of the mice to visit a new arm of the maze rather than a familiar arm. It is a test of hippocampal function but also includes use of other parts of the brain such as the septum, basal forebrain, and prefrontal cortex. The apparatus we used consisted of three enclosed arms (30-cm length, 5-cm width, and 10-cm height) in the shape of a Y. We first placed a mouse in a randomly selected start arm of the Y maze. Upon leaving the start arm, the mouse chooses between entering either the left or the right goal arm. With repeated trials, a mouse with no cognitive impairment typically shows less of a tendency to enter a previously visited arm. The percentage of alternation is calculated as per Arendash et al. (86), with the formula % alternation = 100 × number of alternation/(total arm entries −2).

Open field test.

To examine the effects of M344 on locomotor activity, mice were individually placed in the center of an open field arena (27 × 27 × 23 cm) in a quiet, well-lit room for 10 min. Horizontal activity was detected using a ceiling-mounted camera and Ethovision automated tracking software, and the total distance as well as the distance that each mouse traveled over that time period were recorded. Arenas were cleaned with 70% ethanol between mice.

Novel object recognition test.

For the object recognition test, we adapted the protocol from Roach et al. (64). The test was conducted in the same arena as the open field test described above, thus reducing the need for prolonged habituation trials. We used a 5-min exposure trial with two identical objects followed by a 30-min intertrial interval (ITI) in the home cage. Mice were then tested in a 5-min retention trial with a single one of the previous object and a novel object. Objects were made out of plastics and shapes included were balls or cubes. Total time spent investigating and facing each object within 1.5 cm was recorded. This test is based on the spontaneous tendency of rodents to spend more time exploring a novel object than a familiar one. Exploration of the novel object reflects the use of learning and recognition memory. A memory index defined as [(novel object investigation time)/(total investigation time of both objects)*100] was calculated to compare memory retention of treated versus control animals. A second memory index was also calculated, defined as [(frequency of novel object investigation)/(total frequency of investigation of both objects)*100]. All trials were recorded and analyzed by the automated Ethovision tracking software (Noldus).

Barnes maze test.

For spatial learning and memory assessment, we used a ∼1-m-diameter Barnes maze elevated ∼90 cm above the floor and that contains 20 holes, each 5 cm in diameter, equally spaced around the perimeter of the apparatus. One of the holes led to a plastic escape box whose base was covered with bedding material. The testing room was rich in spatial cues located in constant, fixed positions around the maze so that they would be visible to the mouse. A buzzer sound (∼85 db) and bright light served as mildly aversive environmental stimuli, from which mice tried to escape.

We used a modification of the shortened Barnes maze protocol from Attar et al. (87) to test the effects of M344 on spatial learning and memory. First, the mice were each given one 5-min habituation session on day 1 to get used to the maze and the existence of the escape box. That same day, after 30 min ITI, they were given the first training/acquisition session. Then, for two additional days, mice were given training sessions with two trials per day, also with 30-min ITI. On the fourth day, the mice were allowed to rest for 24 h. Then, on the final day, we administered the probe test, where the escape box was removed and the time spent in the escape box zone was recorded as a measure of spatial memory retention. For all of the training tests, the mouse was placed in the middle of the maze under a start cylinder. After 10 s, the cylinder was lifted, the buzzer initiated, and the mouse allowed to escape or was gently guided to the escape box. Once the mouse was inside the box the buzzer was turned off, and the animal was left inside for 2 min. It was then returned to its home cage during the ITI.

The amount of time that each mouse took to enter the escape box (escape latency) was recorded for each trial. For the probe trials the setup was similar, but with the escape box removed and exploration of the goal hole (former location of the escape box) recorded. All trials were recorded by a ceiling-mounted digital camera, and performance was automatically scored using Ethovision analysis software. We hypothesized that M344 treated mice would show decreased cognitive decline, demonstrated by greater decrease in escape latency across training sessions compared with controls, and spend more time investigating the goal hole.

Tissue Collection.

After the completion of all of the behavior tests, at the age of ∼7 mo, all of the animals were anesthetized with isoflurane and cervically dislocated and the brains were extracted. Brains were then quickly dissected on ice to collect the hippocampus, which was immediately frozen in liquid nitrogen. The samples were then stored at −80 °C until processed for RNA or protein extraction. For each mouse, half the hippocampus was used for total RNA preparation and qPCR and the other half for protein work such as ELISAs and Western blots as described above. Protein lysates were obtained by sonicating the hippocampus in mammalian protein extraction reagents supplemented with protease inhibitor (cOmplete protease inhibitor mixture; Roche) and phosphatase inhibitor mixture (G-Biosciences).