MicroRNA-339-5p Down-regulates Protein Expression of β-Site Amyloid Precursor Protein-Cleaving Enzyme 1 (BACE1) in Human Primary Brain Cultures and Is Reduced in Brain Tissue Specimens of Alzheimer Disease Subjects*

Background: BACE1 is the rate-limiting enzyme in the synthesis of Aβ from amyloid precursor protein. Results: Human miR-339-5p negatively regulates BACE1 and Aβ in human brain cultures and is reduced in AD specimens. Conclusion: Human miR-339-5p physiologically regulates human BACE1 protein expression and Aβ and is dysregulated in the AD brain. Significance: miR-339-5p represents a novel drug target in AD. Alzheimer disease (AD) results, in part, from the excess accumulation of the amyloid-β (Aβ) peptide as neuritic plaques in the brain. The short Aβ peptide is derived from the large transmembrane Aβ precursor protein (APP). The rate-limiting step in the production of Aβ from APP is mediated by the β-site APP-cleaving enzyme 1 (BACE1). Dysregulation of BACE1 levels leading to excess Aβ deposition is implicated in sporadic AD. Thus, elucidating the full complement of regulatory pathways that control BACE1 expression is key to identifying novel drug targets central to the Aβ-generating process. MicroRNAs (miRNAs) are expected to participate in this molecular network. Here, we identified a known miRNA, miR-339-5p, as a key contributor to this regulatory network. Two distinct miR-339-5p target sites were predicted in the BACE1 3′-UTR by in silico analyses. Co-transfection of miR-339-5p with a BACE1 3′-UTR reporter construct resulted in significant reduction in reporter expression. Mutation of both target sites eliminated this effect. Delivery of the miR-339-5p mimic also significantly inhibited expression of BACE1 protein in human glioblastoma cells and human primary brain cultures. Delivery of target protectors designed against the miR-339-5p BACE1 3′-UTR target sites in primary human brain cultures significantly elevated BACE1 expression. Finally, miR-339-5p levels were found to be significantly reduced in brain specimens isolated from AD patients as compared with age-matched controls. Therefore, miR-339-5p regulates BACE1 expression in human brain cells and is most likely dysregulated in at least a subset of AD patients making this miRNA a novel drug target.

Alzheimer disease (AD), 2 the most common form of dementia (1), is a complex disease, with genetics, environment, diet, and lifestyle thought to play a role in disease development (2,3). From a molecular viewpoint, the disease is thought to arise in part from excess deposition of the amyloid-␤ peptide (A␤) (4,5). A␤ is derived from its parental molecule, A␤ precursor protein (APP), during a series of processing steps that result in the liberation of various soluble fragments (6). The ␤-site APPcleaving enzyme 1 (BACE1) is an aspartyl protease (7-10) that initiates A␤ production by internally cleaving APP at the A␤ N-terminus, releasing truncated secreted ␤ form of APP. Cleavage by the ␥-secretase complex at the A␤ C-terminus results in the release of the soluble A␤ peptide and APP C-terminal fragments. Promiscuous cleavage by the ␥-secretase complex results in the release of two C-terminal length variants of the A␤ peptide (A␤  and A␤ ) that are the main constituents of one of the hallmark pathologies of the disease, extracellular neuritic plaques (11). BACE1 processing of APP serves as the rate-limiting step in this biosynthetic pathway. It is worth noting that APP is not the exclusive substrate for BACE1. Indeed, neuregulin 1, neuregulin 3, APP-like protein 1 (APLP1), APLP2, and ␣-2,6-sialyltransferase are also known substrates (12)(13)(14)(15). BACE1 protein levels are known to be increased in the brains of some patients with sporadic AD (16 -18). BACE1 levels also appear elevated in the cerebrospinal fluid (CSF) of AD patients, and these levels correlate with elevated levels of phospho-Tau (19). Therefore, therapeutic strategies that aim to reduce BACE1 expression may be useful as a means to reduce A␤ production and normalize BACE1 expression in sporadic AD.
Given the rate-limiting role that BACE1 plays in the production of A␤ from APP, elucidating the native regulatory pathways that underlie control of BACE1 expression is vital to uncovering novel strategies for modifying BACE1 expression. The spectrum of regulatory schemes utilized by the cell to control BACE1 expression is complex and still not fully clear. BACE1 mRNA is synthesized from a 30.6-kb region of chromosome 11q23.2-11q23.3 containing nine exons and eight introns (20). The BACE1 promoter has been fairly well characterized (21)(22)(23). In neurons, the BACE1 promoter activity is significantly weaker as compared with the APP promoter (24). The promoter lacks canonical TATA or CAAT boxes but contains core or minimal promoter elements that span the transcription start site with some aspects of the minimal promoter located in the genomic 5Ј-UTR (21,22). Several transcription factor sites have been validated in terms of physical interactions and functional effects. Sp1 (22), Yin Yang 1 (YY1) (25), NF-B (26), and possibly even A␤ itself (27,28) bind to the BACE1 promoter and regulate expression. Transcriptional regulation of BACE1 by p25/cdk5 leads to enhanced amyloidogenic processing (29).
Several modes of BACE1 post-transcriptional regulation have also been discovered. The BACE1 5Ј-UTR contains multiple predicted upstream AUGs (uAUGs) and open reading frames (uORF) (30), a feature characteristic of gene products under strict translational control. The presence of multiple uAUGs and uORF generally inhibits mRNA translation because ribosomal scanning initiated from the cap will result in binding and translation of the uORF instead of the authentic ORF. Indeed, multiple studies have identified the second uORF in the BACE1 5Ј-UTR as a potent inhibitor of BACE1 translation (30 -33).
Another post-transcriptional mechanism employed by human cells to control BACE1 levels is the expression of a BACE1 antisense noncoding RNA (34). This RNA binds to ϳ106 complementary nucleotides (nt) from exon 6 in the BACE1 mRNA and stabilizes the transcript. The mechanism involves protecting a microRNA recognition element against targeting by miR-485-5p (35). Despite containing a longer 3Ј-UTR than APP, no novel regulatory mechanisms targeting the 3Ј-UTR have been described for BACE1.
It is clear that, as with APP, transcriptional and post-transcriptional mechanisms for regulating BACE1 expression in human cells are complex and varied. Our understanding of the full regulatory network is still incomplete. Therefore, continued study of the mechanisms that regulate BACE1 expression in human cells is warranted.
MicroRNAs (miRNAs) are small (18 -24 nucleotides) noncoding RNAs that interact with target mRNAs and mediate inhibitory controls on protein production (36). They generally base pair to sites in the 3Ј-UTR of target mRNAs with imperfect complementarity, with the exception of a region at the 5Ј end of a miRNA termed the seed sequence. Studies have shown that near perfect complementarity between the seed sequence and target mRNA is required for a functional interaction (37,38). Notably, miRNAs exist in complex with protein mediators as part of the RNA-induced silencing complex (39), with AGO proteins serving as a primary effector protein. Interactions between miRNAs and their target mRNAs bring the mRNA in close association with effector proteins that generally inhibit protein production either by transcript destabilization or translational inhibition (40), although recent studies suggest that transcript destabilization is the primary mechanism (41).
We and others have begun to describe the contributions that miRNAs bring to the post-transcriptional control of gene products implicated in AD, including APP (42)(43)(44)(45)(46)(47)(48)(49)(50). Others have previously identified and partially characterized miRNAs that also appear to negatively regulate BACE1 expression (18,(51)(52)(53). However, many additional miRNA target sites are predicted in the BACE1 3Ј-UTR. These miRNAs may mediate potent inhibitory effects and participate in the network of molecular regulators that control APP expression.
Here, we demonstrate that hsa-miR-339-5p, or simply miR-339-5p, inhibits expression of BACE1 in a human glioblastoma cell line and in human primary brain cultures via two specific target sites in the BACE1 3Ј-UTR and is a participant in the endogenous molecular network that controls physiological BACE1 expression. We further show that miR-339-5p is dysregulated in a subset of AD patients.

Culture and Transfection of Continuous Cell Lines-HeLa
(human cervical carcinoma) and U373 MG (human glioblastoma) cells were obtained originally from American Type Culture Collection (ATCC). Standard cell culture procedures were employed in the culture and maintenance of all cell lines. HeLa and U373 MG (U373 used throughout) cells were cultured in Minimum Essential Media (Mediatech) supplemented with 10% FBS (Atlanta Biologicals) and penicillin/streptomycin/amphotericin solution (Mediatech) at 37°C in a 5% CO 2 humidified incubator. Antibiotics and antimycotics were omitted from the media during all transfections. For co-transfections of DNA constructs and miRNA mimics (Dharmacon, Thermo Scientific), HeLa cells were cultured on 96-well plates (5 ϫ 10 4 cells per well) and transfected with 150 ng of DNA and 40 nM miRNA using 0.2 l of TransFectin TM (Bio-Rad). For single transfections of siRNA (Applied Biosystems) or miRNA mimics, HeLa cells were cultured in 24-well plates (1.35 ϫ 10 5 cells per well) and reverse-transfected with 20 nM siRNA or 50 nM miRNA using 0.5 l of Lipofectamine RNAiMAX (Invitrogen).
The remaining experiments were single transfections of siRNA and miRNA mimics into U373 cells using RNAiMAX reagent. U373 cells (7.5 ϫ 10 4 cells per well) were cultured in 24-well plates and reverse-transfected by adding transfection complexes to suspension cultures at the same time cells were plated. Transfection complexes were prepared in 50 l of Opti-MEM serum-free media (Invitrogen) with 10 -15-min incubation periods prior to mixing with cell suspensions. U373 cells were transfected with 75 nM miRNA mimics using 3.5 l of RNAiMAX per well.
Culture and Transfection of Primary Human Brain Cultures-Primary cultures of mixed human fetal brain (HFB) cells were prepared from the brain parenchyma of aborted fetuses (80 -100 days gestational age), as described previously (45). Briefly, tissues were obtained from the Laboratory of Developmental Biology, University of Washington, Seattle, after shipping overnight in chilled Hibernate-E medium (Invitrogen) supplemented with B27 (Invitrogen), GlutaMAX (Invitrogen), and antibiotic/antimycotic solution (Cellgro). Tissues were digested in 0.05% trypsin, 0.53 mM EDTA solution and incubated in a shaking water bath (150 RPM) at 37°C for 15 min, transferred to Hibernate-E medium, and triturated several times using a siliconized fire-polished pipette followed by centrifugation at 800 ϫ g for 10 min. After an additional round of trituration and centrifugation, the pellet was resuspended in culture medium and plated on poly-D-lysine-coated 24-well plates in Neurobasal medium (Invitrogen), supplemented with 1ϫ B27, 0.5 mM GlutaMAX, 5 ng/ml basic FGF (Invitrogen), and antibiotic/antimycotic mixture. Half-media changes were performed every 4th day of culture. The protocol was approved by the Indiana University School of Medicine Institutional Review Board (IRB) and was in compliance with state and federal regulations.
HFB cultures were transfected on day in vitro (DIV) 17-18 in 24-well plates. Cultures were transfected with 20 nM siRNA, 150 nM miRNA mimics, or 1 M custom-designed target protectors (Qiagen) using 1.25 l of RNAiMAX. Target protectors were synthetic, modified oligonucleotides designed to block the interaction of a miRNA with a putative target site on a target transcript. Sequences of target protectors were not provided. Antibiotics and basic FGF were omitted from media during transfections.
Generation of BACE1 3Ј-UTR Reporter Construct-A BACE1 3Ј-UTR reporter construct was prepared. The parental construct used to prepare this construct was psiCHECK-2 (Promega, Madison, WI). This plasmid is 6.2 kb in length and contains a Renilla luciferase coding sequence (CDS) driven by the SV40 promoter, a multiple cloning site (MCS) located in the Renilla 3Ј-UTR, and a synthetic polyadenylation signal. A firefly luciferase CDS is located downstream and is driven independently by the HSV-TK promoter. To prepare the construct, the full-length BACE1 3Ј-UTR (3.9 kb) was PCR-amplified from pooled human genomic DNA (Roche Applied Science). Forward and reverse PCR primers were designed with 5Ј extensions compatible with the Infusion cloning system (Clontech). The forward primer sequence was as follows (extension underlined): 5Ј-taggcgatcgctcgagagatagagattcccctggac-3Ј; the reverse primer sequence was as follows (extension underlined): 5Ј-ggccgctctaggtttaaacgcctcagtattgttttagcc-3Ј. The amplicon was then inserted into XhoI and PmeI double-digested psiCHECK-2 using the Infusion cloning system.
Site-directed Mutagenesis of Predicted miR-339-5p Target Sites-Two predicted miR-339-5p target sites in the BACE1 3Ј-UTR reporter construct were mutated using the QuikChange Lightning site-directed mutagenesis kit (Agilent). The following primers were utilized to introduce seed sequence mutations at target site 1: sense 5Ј-ggagaggatgcacagtttgctatttgctttagacggatccactgtataaacaagcctaacattg-3Ј and antisense 5Ј-caatgttaggcttgtttatacagtggatccgtctaaagcaaatagcaaactgtgcatcctctcc-3Ј. For site 2 the following primers were used: sense 5Ј-aagaggagaaggagagggagtacaaacggatccaatagtgggatcaaagctaggaaagg-3Ј and antisense 5Ј-cctttcctagctttgatcccactattggatccgtttgtactccctctccttctcctctt-3Ј. For double mutation, site-directed mutagenesis was applied sequentially to first generate a single mutant followed by the generation of a double mutant.
Luciferase Reporter Assays-HeLa cells were transfected with the WT and mutant BACE1 3Ј-UTR reporter constructs either alone or in combination with miRNA mimics, as described above. Forty eight h post-transfection, the Renilla and firefly luciferase activity was assayed independently using the Dual-Luciferase reporter system (Promega) on a Turner Biosystems Veritas luminometer. Ratios of Renilla/firefly luminescence values were calculated and scaled relative to the value for the BACE1 3Ј-UTR reporter alone transfection.
Western Blotting Analysis-For protein analysis by Western blotting, cell lysates were prepared at various time points in culture as indicated in the figure legends and as described previously (45). Briefly, cells were first washed with PBS and then lysed on-plate with vigorous shaking using the mammalian protein extraction reagent (Pierce) supplemented with 0.1% SDS and protease inhibitor mixture set III (Calbiochem). Lysates were centrifuged at 30,000 ϫ g for 10 min at 4°C and cleared lysates collected. A small volume of lysate (15 l) was utilized for assessment of cell viability at the time of harvest using the CellTiter-Glo (CTG) kit (Promega). Lysate protein concentrations were assayed by BCA (Pierce), and equal amounts of lysate protein (1-5 g) were loaded onto BisTris XT denaturing 10% polyacrylamide gels containing SDS (Bio-Rad). Proteins were resolved by SDS-PAGE and transferred onto PVDF membranes. Protein bands on each blot were stained with 0.1% Ponceau-S (Sigma) solution prepared in 5% acetic acid to confirm complete and even transfer across different lanes. Membranes were blocked for 1 h in 5% nonfat milk and then incubated overnight separately with primary antibodies against BACE1 (clone 3D5 (54), kindly provided by Dr. Robert Vassar, Northwestern University), ␣-tubulin (B-5-1-2, Sigma), and ␤-actin (AC15, Sigma). Membranes were then incubated with HRPconjugated goat anti-mouse secondary antibody (Rockland Immunochemical) for 1 h. Bands were visualized using ECL reagent (Pierce), detected on autoradiographic film, and scanned.
ELISA Analysis of A␤  and A␤ -Both U373 and HFB cultures secrete human-specific A␤ species at detectable levels in the culture medium. Therefore, levels of A␤(1-40) and A␤  were measured independently in the conditioned media (CM) samples of U373 and HFB cultures, as described previously (45). Sensitive and specific commercially available ELISAs (IBL America) were used to assay these analytes. For A␤  assay, an equal volume of CM (25 l) was loaded onto a plate pre-coated with anti-human A␤ (35)(36)(37)(38)(39)(40) antibody (clone 1A10) and incubated overnight. This kit uses anti-human A␤ (11-28) as detection antibody. For A␤ , 50 l of CM was loaded overnight onto an antibody pre-coated plate (anti-human A␤ 38 -42). The overall assays were performed according to the manufacturer's instructions.
Absolute A␤ values (in pg/ml of CM) were measured and corrected for well-to-well variations in cell number by either normalizing to total protein yield of crude cell lysates (U373 cultures) or to a measure of cell viability (CTG) for HFB cultures. These values were then scaled relative to mock transfection values.
Quantification of BACE1 mRNA in Cell Culture-Total RNA was extracted from U373 cultures using the miRVana miRNA miR-339-5p Negatively Regulates BACE1 Expression isolation kit (Ambion). RNA quantity and purity were assessed using a Nanodrop instrument (Thermo Scientific). RNA integrity was assessed on a Bioanalyzer (Agilent). All samples had acceptable A 260 /A 280 ratios and RIN values greater than 8.5. BACE1 mRNA levels were quantified by reverse transcription quantitative PCR (RT-qPCR). Briefly, RNA was first converted to cDNA using the High Capacity RNA-to-cDNA kit (Applied Biosystems). cDNA was then subjected to qPCR analysis using TaqMan hydrolysis probe assays (Applied Biosystems) on a 7300 Real Time PCR instrument (Applied Biosystems). Relative quantification was performed using the delta-C q method and normalized to the geometric mean of at least three reference genes (55). To determine gene-specific amplification efficiencies for each TaqMan assay, aliquots of every RNA sample in the analysis were pooled and used to create a relative standard curve by serial dilution. This standard curve was then converted to cDNA and analyzed by qPCR in parallel with unknown samples. The slope of the plot of C t versus standard curve dilutions was used to calculate amplification efficiency. The reference genes used in this study were GAPDH, B2M, and TBP. Assay names and IDs are listed as follows: human BACE1 (Hs00201573_m1); human GAPDH (4333764T); human B2M (4333766T), and human TBP (4333769T).
Human Brain Specimens and Processing-Frozen brain specimens isolated from BA9 of the frontal cortex were obtained from the University of Kentucky Alzheimer's Disease Center Brain Bank. Both age-matched controls (n ϭ 5) and AD patients (n ϭ 20) were kindly provided by Dr. Peter Nelson. Sample demographics, post-mortem interval, and neuropathological characterization were all provided and are discussed under "Results." Specimens were initially pulverized using a stainless steel pulverizing chamber pre-chilled with liquid nitrogen and were quickly aliquoted, avoiding sample thawing, and stored at Ϫ80°C.
Brain specimens were prepared as described previously (45). Briefly, 1 aliquot of each sample was processed for protein analysis. This frozen aliquot was immersed in mammalian protein extraction reagent supplemented with 0.1% SDS and protease inhibitor mixture and immediately sonicated using a Sonifier Cell Disruptor 350 (Branson) until visible clumps disappeared. Lysates were then incubated with 50 units/ml Benzonase enzyme (Calbiochem) for 10 min at 37°C to reduce nucleic acid content and associated viscosity. Lysates were then centrifuged down at 30,000 ϫ g for 2 h to clear debris. Protein concentrations of the cleared lysates were determined by BCA. Western blot analyses were performed as described above.
A second aliquot was processed for RNA analysis. The frozen aliquot was immersed in Cell Disruption Buffer from the miRVana miRNA isolation kit and immediately homogenized using Polytron (Kinematica). These homogenates were then processed per the manufacturer's instructions for tissue samples. RNA quality control was performed as described above. All samples were of high quality with RIN Ͼ6.5.
Quantification of Small RNA Expression Levels in Human Brain Specimens-RT-qPCR was performed on these RNA extracts as described above but with additional modifications. miRNA levels were detected using absolute quantification methods.
To perform absolute quantification in miRNA analyses, HPLC-purified synthetic oligoribonucleotide standards identical in sequence to hsa-miR-339-5p and hsa-miR-124 were obtained commercially (Sigma). The oligoribonucleotide was resuspended, and exact concentrations were measured by A 260 readings on a Nanodrop instrument. Based on measured concentrations, standard curves with absolute copy counts were prepared by serial dilution. These serially diluted standards were converted to cDNA and analyzed by qPCR in parallel with unknown samples. Copy counts per reaction were then determined based upon standard curve analysis. Given that each unknown reaction was loaded with a known amount of total RNA (generally 10 ng), copy counts were then presented as copy counts/15 pg of total RNA. This serves as a rough estimate of copy counts per average human cell. Assay names and IDs are listed as follows: hsa-miR-339-5p (002257) and mmu-miR-124a (001182).
Data and Statistical Analysis-Densitometric analysis of Western blots was performed using ImageJ software (56). qPCR data analysis and normalization was performed using qbase PLUS software (57). Western blot image processing was performed using Adobe Photoshop. Western blot images were adjusted for contrast and brightness and some extraneous sections of blots between boxed regions were removed for the sake of clarity. No manipulations have been made to images that alter data quantification or interpretation. Statistical analyses were performed and charts prepared using Prism GraphPad. In all cases, data are expressed as the means Ϯ S.E. For comparison between two categories, two-tailed Student's t test was performed. For comparison across multiple categories, analysis of variance was performed followed by post hoc Dunnett's t test. The ␣ threshold for statistical significance was set at 0.05.

Global Knockdown of miRNA Function Elevates BACE1
Expression-HFB cultures were utilized to investigate the effects of miRNA regulation on BACE1 expression. An example of primary HFB culture morphology and cell type distribution has been recently published and is reproduced in Fig. 1A (45). We also examined the temporal profile of BACE1 expression in primary HFB cultures. We assayed levels of BACE1 protein at DIV 7, 10, 14, 18, 22, and 26 to better characterize the nature of the culture with respect to this key analyte. Western blot analysis revealed very high levels of BACE1 protein at DIV 7 that plateaued over time with the lowest expression levels observed at DIV 18 ( Fig. 1, B and C). This finding is quite similar to the profile of APP expression observed in this culture system, as published recently by us (45). In this mixed cell type culture system, we consistently observe multiple isoforms of BACE1 that are closely spaced on Western blot and difficult to discern as individual bands. This is in contrast to other cell types, such as U373, in which a major BACE1 isoform is more easily distinguished (see below).
To begin to dissect the role that miRNAs play in the basal regulation of BACE1 protein expression, global miRNA function was inhibited in primary HFB cultures. To accomplish this goal, AGO2 expression was knocked down using siRNA transfection. Given that AGO family proteins are the core constitu-ent of RNA-induced silencing complex (58), this approach would be expected to modulate expression of gene targets that are under direct basal regulation by miRNAs. Primary HFB cultures were either mock-transfected or transfected with a negative control or AGO2-specific siRNA at DIV 18. Cultures were harvested 48 h post-transfection (DIV 20), and BACE1 protein expression was assayed by Western blot. BACE1 protein levels were significantly elevated following knockdown of AGO2 as compared with negative control siRNA transfection (Fig. 2), suggesting that the global complement of miRNA acts to basally inhibit BACE1 protein expression in cultured human brain cells.
Computational Prediction of miR-339-5p Target Sites in BACE1 3Ј-UTR-We used various web-based bioinformatic algorithms to predict favorable miRNA interactions in the BACE1 3Ј-UTR, including TargetScan 6.2 (38, 59 -61), PicTar (62), DIANA-MicroT version 4.0 (63, 64), miRanda-mirSVR (65)(66)(67), PITA (68), and rna22 (69). Two putative miR-339-5p target sites in the BACE1 3Ј-UTR were identified (Fig. 3A). When all transcripts with putative miR-339-5p target sites predicted by TargetScan were ranked by context ϩ score, BACE1 was ranked third out of 3117 transcripts, indicating the BACE1 target sites are highly ideal. The first site with seed sequence located at 484 -491 nucleotides relative to start of the BACE1 3Ј-UTR was predicted by TargetScan, miRanda-mirSVR, PITA, and rna22. The second site with seed sequence at 611-618 nucleotides relative to start of 3Ј-UTR was predicted by Target-Scan, DIANA-MicroT, miRanda-mirSVR, and rna22. Previously validated target sites in the BACE1 3Ј-UTR for miR-107 and miR-29a/b (18,51) were also predicted. Scores generated by the algorithms were compared between the validated miR-107 and miR-29a/b target sites and the predicted miR-339-5p target sites in Table 1. Sequence comparison between the predicted miR-339-5p target sites in the human BACE1 3Ј-UTR and orthologous sequences from multiple mammalian species revealed several sequence differences (Fig. 3B). Both the sites are considered poorly conserved according to the conserved     Validation of Predicted miR-339-5p Target Sites-To validate the functionality of the putative miR-339-5p/BACE1 3Ј-UTR interaction, a reporter construct was prepared containing the full-length BACE1 3Ј-UTR. This reporter was generated by PCR-amplifying the BACE1 3Ј-UTR from human genomic DNA and inserting the amplicon downstream of a Renilla luciferase CDS. A separate firefly luciferase CDS under independent transcriptional control was also present in this construct to serve as an internal control (Fig. 3C).
Co-transfection of the reporter construct along with miR-339-5p mimic in HeLa cells resulted in significantly reduced Renilla activity relative to co-transfection with negative control mimic or transfection of reporter construct alone (53% of negative control mimic) suggesting an inhibitory regulatory interaction between miR-339-5p and the BACE1 3Ј-UTR (Fig. 3D). To confirm that the inhibitory effect of miR-339-5p on BACE1 3Ј-UTR reporter expression was mediated specifically via one of two predicted miR-339-5p target sites located in the BACE1 3Ј-UTR, mutations were introduced in the seed sequences of both target sites in the reporter construct (Fig. 3C). A double mutant reporter was constructed containing both seed sequence mutations. Perfect complementarity at the seed sequence is critical for functional miRNA interactions (36), and mutation at this position should eliminate effective interaction between miRNA and target site. These mutant reporter constructs were then cotransfected along with miR-339-5p mimic into HeLa cells and reporter expression compared with wild-type reporter (Fig.  3D). Mutation of site 2 partially eliminated the inhibitory effect of miR-339-5p mimic on reporter expression, whereas mutation of both sites completely reversed the miR-339-5p response. To confirm the significance of this change following site 2 mutation or double site mutation, a comparison of difference analysis was completed. The difference in reporter expression following miR-339-5p transfection versus negative control mimic transfection was significantly attenuated in both the site 2 mutant (0.57 Ϯ 0.09-fold) and the double site mutant (Ϫ0.22 Ϯ 0.17-fold) as compared with the wild-type construct (p Ͻ 0.01 by analysis of variance followed by post hoc Dunnett's t test). Due to technical difficulties, a construct containing mutation of site 1 alone was not generated. Therefore, miR-339-5p mediates its inhibitory effect on BACE1 3Ј-UTR reporter expression by interacting with at least one of two predicted target sites in the BACE1 3Ј-UTR. However, we cannot definitively quantify the relative contribution of either site in isolation due to the inability to generate a mutant construct for site 1 alone.
miR-339-5p Inhibits BACE1 Expression and A␤ Production in Human Glioblastoma Cell Line U373-We next examined whether miR-339-5p directly reduces endogenous BACE1 levels. Human U373 cells were selected for initial analysis because they are both readily transfectable and express moderate levels of BACE1 protein. U373 cells were transfected with either negative control, miR-29b (positive control), or miR-339-5p mimic (Fig. 4, A and B). miR-29b has been previously shown to downregulate endogenous BACE1 protein expression (18). Transfection with miR-29b and miR-339-5p mimics did not affect culture morphology as visualized by phase contrast microscopy or cell viability as assessed by CTG assay (0.93 Ϯ 0.04-and 1.13 Ϯ 0.03-fold change relative to negative control mimic transfection, respectively; p Ͼ 0.05 by post hoc Dunnett's multiple comparison test). BACE1 protein levels were then directly assayed by Western blot. Notably, BACE1 protein expression in U373 cells was significantly reduced following transfection with both miR-29b (87.5% reduction) and miR-339-5p (62% reduction) as compared with transfection with the negative control miRNA mimic. Therefore, endogenous BACE1 levels are inhibited significantly by miR-339-5p delivery in human U373.
In a separate experiment, RNA was extracted from U373 cells 48 h after transfection with negative control mimic or miR-339-5p and BACE1 mRNA levels analyzed by RT-qPCR. BACE1 mRNA levels were decreased significantly (74% reduction) following transfection of miR-339-5p mimic as compared with negative control (Fig. 4C). This suggests that exogenous miR-339-5p likely inhibits BACE1 expression via mRNA destabilization.
We next examined if down-regulation of BACE1 following delivery of miR-339-5p results in reduced secretion of A␤ peptides. Levels of A␤(1-40) (Fig. 4D) and A␤(1-42) (Fig. 4E) in the CM of U373 cultures were analyzed by specific sandwich ELISA 48 h after transfection with negative control or miR-339-5p mimic. Both A␤ species were significantly reduced following transfection of miR-339-5p as compared with negative control (61% reduction for A␤(1-40) and 33% reduction for A␤ ). Therefore, miR-339-5p delivery in U373 cells negatively regulates A␤ secretion (both short and long forms), presumably via its direct effect on BACE1 expression. miR-339-5p Inhibits Expression of BACE1 and A␤ Production in Primary Human Brain Cultures-To determine the role of miR-339-5p in human primary mixed brain cultures, we tested the effect of miR-339-5p delivery on BACE1 protein expression in primary HFB cultures.
The primary HFB cultures were transfected at DIV 17 with negative control or miR-339-5p mimics and harvested 48 h post-transfection. Western blot analysis revealed significantly reduced BACE1 levels (14% reduction) following delivery of miR-339-5p as compared with negative control (Fig. 5, A and  B). Because cultured HFB cells secrete A␤ species into the culture medium, levels of A␤(1-40) (Fig. 5C) and A␤(1-42) (Fig.  5D) were also assayed by ELISA and found to be significantly decreased after delivery of miR-339-5p as compared with negative control (21% reduction for both A␤ species). The smaller reduction in BACE1 and A␤ levels as compared with U373 cells is most likely secondary to the reduced transfection efficiency in HFB as compared with U373 cells. Therefore, miR-339-5p delivery negatively regulates BACE1 expression and A␤ secretion in primary HFB cultures.
Endogenous miR-339-5p Inhibits BACE1 Expression via Predicted 3Ј-UTR Target Sites-To establish whether endogenous miR-339-5p regulates BACE1 expression by interacting with the two target sites in primary HFB cultures, two separate custom target protectors were utilized to block the interaction of endogenous miR-339-5p with each validated BACE1 3Ј-UTR target site. Primary HFB cultures were transfected with miR-339-5p target protectors at DIV 17, and BACE1 protein expression was assayed by Western blot 48 h post-transfection (Fig. 6). Transfection of miR-339-5p site 1 target protector elevated miR-339-5p Negatively Regulates BACE1 Expression BACE1 expression (35% elevation), but the increase was not statistically significant compared with transfection with negative control target protector. Transfection of miR-339-5p site 2 target protector significantly elevated BACE1 expression compared with transfection with negative control target protector (49% elevation). Therefore, endogenous miR-339-5p actively inhibits BACE1 expression in primary HFB cultures under physiological culture conditions and does so by potentially interacting with site 1 and definitely by interacting with site 2 in the BACE1 3Ј-UTR. Of note is that the site 2, although itself poorly conserved by TargetScan standards, exhibits a higher level of cross-species conservation than does site 1. Therefore, the reduced efficacy of the site 1 target protector in modulating BACE1 expression and the lower level of site 1 cross-species conservation may suggest that this site is less physiologically relevant than site 2.
We observed that the negative control target protector transfection resulted in reduced BACE1 levels as compared with mock transfection. However, it is important to point out that significantly higher quantities of nucleic acid were transfected in the target protector (1000 nM) as compared with the siRNA (20 nM) transfection condition. Therefore, it is likely that the reduction in BACE1 expression following negative control tar-get protector transfection is likely secondary to a nonspecific effect from a high RNA load. This highlights the importance of comparing the site-specific target protector transfection conditions to the negative control.
miR-339-5p Is Dysregulated in the AD Brain-Several miRNA species have been previously reported to be dysregulated in the post-mortem AD brain (18, 51, 70 -72). We next investigated whether miR-339-5p might also be dysregulated in AD brain specimens. Specimens were obtained from the University of Kentucky AD Brain Bank. These specimens were isolated from BA9 of the frontal cortex and consisted of well characterized control (n ϭ 5) and AD (n ϭ 20) specimens. Demographic and neuropathological details for these specimens are listed in Table 2. These specimens were age-matched, with a mean age for control specimens of 84.0 Ϯ 2.2 and mean age for AD specimens of 80.0 Ϯ 2.1 years. All AD specimens had advanced AD neuropathology (Braak stage V-VI (73) and CERAD score C (74)). Importantly, all specimens were collected following a short post-mortem interval (range 1.75-8 h).
We first analyzed BACE1 protein levels by Western blot in these brain specimens to determine whether these specimens had dysregulated protein levels, as observed in several previous studies (16 -18). Indeed, we found that BACE1 levels were sig-   FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8

miR-339-5p Negatively Regulates BACE1 Expression
nificantly elevated in AD specimens as compared with control specimens (41% increase) (Fig. 7). We next examined miR-339-5p levels by RT-qPCR. Notably, miR-339-5p levels were significantly decreased compared with controls (45% reduction) (Fig. 8B). Expression levels of an additional miRNA not specifically tested for BACE1-regulating effects in these studies but previously reported to regulate BACE1 expression was also assessed, miR-124. There were no significant differences in miR-124 (Fig. 8A) levels in AD specimens as compared with negative control specimens. However, levels did trend toward decreased expression despite not reaching statistical significance. Therefore, miR-339-5p levels are negatively dysregulated in AD specimens from this cohort.

DISCUSSION
This report is the first, to our knowledge, to describe a novel inhibitory interaction between the known miRNA miR-339-5p and BACE1 3Ј-UTR that participates in the physiological regulatory scheme employed in human cultured primary brain cells to regulate BACE1 expression levels and is dysregulated in AD brain specimens. Specifically, we found that miR-339-5p potently inhibited BACE1 3Ј-UTR reporter expression via interaction with at least one of two poorly conserved target sites in the BACE1 3Ј-UTR. BACE1 protein levels were dramatically decreased following miR-339-5p transfection in human U373 cells and less potently (but significantly) so in primary human brain cultures. We also found that endogenous miR-339-5p regulates basal BACE1 expression in primary human brain cul-tures as evidenced by elevated BACE1 expression following target protector transfection. We also observed elevated BACE1 protein levels and decreased miR-339-5p levels in AD brain specimens relative to control specimens. Direct miR-339-5p delivery or targeting pathways that elevate expression of endogenous miR-339-5p in the CNS may represent viable therapeutic strategies. Therefore, miR-339-5p is a putative drug target for AD.
Other studies have profiled miRNA expression in the AD brain (18, 46, 48, 51, 70 -72, 75), in peripheral blood mononuclear cells (76), and in peripheral blood (77) and identified miR-NAs that are dysregulated (e.g. miR-107, miR-29a/b, miR-106b, miR-124, etc.) (see Ref. 78). To our knowledge, ours is the first to report differentially expressed miR-339-5p in the AD brain. One potential explanation is that most previous studies have employed global miRNA analyses that must be statistically corrected for a high number of multiple comparisons. Therefore, the statistical power of these analyses may have been too low to detect the decrease observed here. Alternatively, CNS expression levels of miR-339-5p may be too low to readily detect using probe-based microarray protocols.
The quality of the brain specimens utilized in this study is very high as evidenced by very short post-mortem intervals. Subsequent RNA extract quality was also very high. No samples with RIN values below 6 were used in the analysis. Therefore, we are very confident that our analysis is not biased by significant post-mortem degradation. However, certain shortfalls are

miR-339-5p Negatively Regulates BACE1 Expression
also present in our brain expression analysis. Primarily, the sample size here is small (n ϭ 5-20 per group), and the reliability of our finding would be greatly aided if it was observed in an independent cohort of a larger sample size. However, it should also be noted that procuring a large sample of high quality, low post-mortem interval AD brain tissue is a significant challenge. We also realize that dysregulation of miR-339-5p might be an epiphenomenal manifestation of cell type distribution changes that occurs during the progression of AD. Clarification could be provided by the analysis of miR-339-5p by in situ hybridization in sections from AD and control specimens, allowing for cellular level resolution. In our analysis, miR-339-5p levels in human brain specimens were significantly lower than the highly abundant neuronal miR-124. However, the actual copy counts per cell for miR-339-5p are still likely high enough to mediate regulatory effects. This statement is based upon the fact that expression levels here were measured in post-mortem brain specimens, where miRNA decay is known to be rapid (79). Although the postmortem interval in this study was quite short by typical autopsy standards, it is likely that miRNA copy counts in living CNS cells are much higher than detected here.
The precise level of miRNA expression necessary for functional regulation is likely to vary for each combination of miRNA and mRNA target and their specific environmental context. Titration experiments have demonstrated that miRNA efficacy is influenced both by the levels of target expression as well as miRNA levels, such that target thresholds are generated around which target levels are highly sensitive to miRNA effects (80,81). However, in some conditions, as few as 100 miRNA copies per cell have been reported to be sufficient for repression (81). By this standard, we anticipate that expression levels of miR-339-5p in the human brain would be sufficient to mediate its regulatory effects on BACE1 expression.   (74) and as follows: 0 ϭ no disease; A ϭ possible AD; C ϭ definite AD. c Assessment of AD pathology is based on National Institutes of Health NIA-Reagan Institute consensus recommendations as described previously (96). FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 5193
Consistent with what has been reported in previous studies (16 -18, 54), we observed increased BACE1 protein levels in AD specimens as compared with control specimens. This suggests a tantalizing possibility, i.e. decreased miR-339-5p levels may drive enhanced BACE1 expression in the AD brain. However, cells likely have built-in redundancies that homeostatically control gene expression, reducing the impact of dysregulation of a single gene regulator on gene product output. Therefore, it is more likely that dysregulation of multiple miRNA (including miR-339-5p) along with other gene regulators synergize to produce dysregulated BACE1 levels in the AD brain.
miR-339-5p has been implicated in other physiological and pathological roles aside from its regulatory effects on BACE1 expression, including hematopoiesis, various cancers, and neural tube defects (NTD). More specifically, miR-339-5p expression was shown to vary during the course of in vitro erythroid differentiation and is predicted to target gene products (EpoR and STAT-5) previously implicated in erythroid differentiation (82).
The role of miR-339-5p in cancer is still unsettled and may vary by tissue origin. miR-339-5p expression is significantly upregulated in B-cell precursor acute lymphoblastic leukemia (83). In contrast, its expression is significantly decreased in highly invasive breast cancer cell lines (MDA-MB-231 and MDA-MB-468) as compared with poorly invasive cell lines (MCF7) (84). Reduced expression was also observed in breast cancer tissue with lymph node metastasis as compared with breast cancer tissue without node metastases. This study also showed that miR-339-5p had no effect on cell proliferation but that overexpression reduced breast cancer cell migration and inhibition enhanced migration. This effect may be mediated by the inhibitory effect of miR-339-5p against BCL6. Finally, endogenously expressed miR-339-5p in colorectal cancer cell lines and glioblastoma multiforme cell lines apparently reduces susceptibility of malignant cells to cytolysis by cytotoxic T-lymphocytes by inhibiting expression of ICAM1 (85,86). In toto, these studies suggest that miR-339-5p suppresses the malignant potential of transformed cells in some forms of cancer but not necessarily others.
miR-339-5p may contribute to the formation of NTD. An NTD-prone animal model, Splotch (Pax3 Ϫ/Ϫ ), that can be res- FIGURE 7. Analysis of BACE1 protein levels in control and AD human brain specimens. A, Western blot analysis of BACE1 in brain specimens (BA9 of frontal cortex) from AD and control patients. Ponceau stain of blotted region is also presented to demonstrate equal protein loading across lanes. B, blot from A was densitometrically analyzed and BACE1 levels normalized to total protein in the blotted region based on Ponceau stain and scaled relative to control BACE1 levels. BACE1 levels were significantly higher in AD specimens as compared with control specimens (*, p ϭ 0.0046 by two-tailed Student's t test). Ponceau normalization was employed due to variable degradation of ␣-tubulin and ␤-actin across lanes despite equal protein loading. C, control group (n ϭ 5); AD, AD group (n ϭ 15). AD*, specimens were excluded from analysis due to uncertain medication exposure history. These specimens were excluded a priori for analyses in Fig. 7. FIGURE 8. miR-339-5p levels are dysregulated in AD brain specimens. RT-qPCR analysis of expression levels for miR-124 (A) and miR-339-5p (B) in brain specimens (BA9 of frontal cortex) from AD and control patients. Expression levels were quantified in absolute terms as miRNA copy counts per 15 pg of total RNA. Copy counts were calculated from standard curves prepared by serial dilutions of miRNA oligonucleotide standards with known concentrations. miR-124 levels were somewhat lower in AD specimens as compared with control specimens but not statistically significant (p ϭ 0.1928 by twotailed Student's t test). miR-339-5p levels were significantly lower in AD brain specimens as compared with control brain specimens (0.55 Ϯ 0.04-fold change; *, p ϭ 0.0339 by two-tailed Student's t test). n ϭ 5 for control, n ϭ 14 for AD in A and n ϭ 15 for AD in B. A contains one fewer AD specimens due to technical failure in the qPCR run for this specimen.

miR-339-5p Negatively Regulates BACE1 Expression
cued from NTD development by folate supplementation was shown to have reduced expression of a lysine-specific demethylase, KDM6B (87). Repressed KDM6B expression was shown to inhibit gene expression of several genes essential for normal neural tube formation (Hes1 and Neurog2) by promoting histone methylation at the promoters of these genes. Reduced expression of KDM6B in Splotch animals was likely mediated by increased expression of several miRNAs, including miR-339-5p that repressed KDM6B levels via interactions with its 3Ј-UTR. Folate supplementation reversed dysregulated miR-339-5p expression, KDM6B expression, promoter methylation, and Hes1 and Neurog2 expression in these animals (87).
As has been described previously with APP (42,44,(47)(48)(49)45), there is likely an extensive network of miRNAs that act in collaboration with miR-339-5p to modulate BACE1 expression. Although the full repertoire has not been elucidated, several BACE1-targeting miRNAs have been described in the literature. For example, miR-107 regulates BACE1 expression via interaction with a 3Ј-UTR target site and is down-regulated early in the temporal cortex of the AD brain (51,88). miR-107 also associates with progranulin (PRGN) mRNA in the RNAinduced silencing complex and down-regulates expression via interaction with a target site in the PRGN CDS (89). This interaction is relevant because mutations in PRGN have been previously linked to frontotemporal dementia (90,91). miR-29a/b both down-regulate BACE1 expression via 3Ј-UTR target sites (18), and both have been reported to be down-regulated in the AD brain (18,92). Surprisingly, miR-29b has also been shown to down-regulate PRGN expression but via interactions in the 3Ј-UTR, as compared with the CDS for miR-107 (93). Furthermore, miR-29c down-regulates BACE1 expression in both cell culture and in vivo miR-29c transgenic animal models via a 3Ј-UTR target site (53). miR-124 overexpression down-regulates BACE1 expression, whereas inhibition elevates BACE1 protein levels (52). This finding has significant implications because miR-124 is one of the most abundant neuronal miR-NAs and has been previously shown to be decreased in the AD brain (48,70). While this finding was not replicated in this study, we did observe a trend toward decreased expression. Additionally, miR-124 functions to regulate neuronal APP mRNA splicing (48). Finally, mouse (mmu) miR-298 and miR-328 also down-regulate BACE1 via mouse BACE1 3Ј-UTR target sites (94).
In summary, the full complement of miRNAs that participate in the regulation of BACE1 expression has yet to be fully elucidated. Here, we identified a previously known miRNA, miR-339-5p, as a member of this network in primary human brain cells. We also discovered that miR-339-5p appears to be dysregulated in at least a subset of AD patients. These results are consistent with the premise that down-regulation of miR-339-5p would in turn disinhibit expression of BACE1, the ratelimiting enzyme for A␤ generation from the proteolytic cleavage of APP, and would trigger the amyloidogenic pathway believed implicated in AD pathogenesis. Therefore, we conclude that miR-339-5p is a novel drug target in AD. We would anticipate that therapeutic manipulations that enhance miR-339-5p levels would reduce BACE1 expression and subsequent A␤ production, resulting in salutary effects in AD.
This finding is especially significant given the news of recent failures of several clinical drug trials for AD, including a BACE1 drug trial. In this context, a miRNA-based therapeutic approach may serve as an alternative strategy (95). Thus, our discovery is novel and significant with great potential translational impact. The next steps will require assessing whether miR-339-5p manipulations in vivo ameliorate AD pathology and memory deficits in appropriate AD animal models.