Differential Regulation of Interleukin-1 Receptor-associated Kinase-1 (IRAK-1) and IRAK-2 by MicroRNA-146a and NF-κB in Stressed Human Astroglial Cells and in Alzheimer Disease*

Specific microRNAs (miRNAs), small non-coding RNAs that support homeostatic gene expression, are significantly altered in abundance in human neurological disorders. In monocytes, increased expression of an NF-κB-regulated miRNA-146a down-regulates expression of the interleukin-1 receptor-associated kinase-1 (IRAK-1), an essential component of Toll-like/IL-1 receptor signaling. Here we extend those observations to the hippocampus and neocortex of Alzheimer disease (AD) brain and to stressed human astroglial (HAG) cells in primary culture. In 66 control and AD samples we note a significant up-regulation of miRNA-146a coupled to down-regulation of IRAK-1 and a compensatory up-regulation of IRAK-2. Using miRNA-146a-, IRAK-1-, or IRAK-2 promoter-luciferase reporter constructs, we observe decreases in IRAK-1 and increases in miRNA-146a and IRAK-2 expression in interleukin-1β (IL-1β) and amyloid-β-42 (Aβ42) peptide-stressed HAG cells. NF-κB-mediated transcriptional control of human IRAK-2 was localized to between −119 and +12 bp of the immediate IRAK-2 promoter. The NF-κB inhibitors curcumin, pyrrolidine dithiocarbamate or CAY10512 abrogated both IRAK-2 and miRNA-146a expression, whereas IRAK-1 was up-regulated. Incubation of a protected antisense miRNA-146a was found to inhibit miRNA-146a and restore IRAK-1, whereas IRAK-2 remained unaffected. These data suggest a significantly independent regulation of IRAK-1 and IRAK-2 in AD and in IL-1β+Aβ42 peptide-stressed HAG cells and that an inducible, NF-κB-sensitive, miRNA-146a-mediated down-regulation of IRAK-1 coupled to an NF-κB-induced up-regulation of IRAK-2 expression drives an extensively sustained inflammatory response. The interactive signaling of NF-κB and miRNA-146a further illustrate interplay between inducible transcription factors and pro-inflammatory miRNAs that regulate brain IRAK expression. The combinatorial use of NF-κB inhibitors with miRNA-146a or antisense miRNA-146a may have potential as a bi-pronged therapeutic strategy directed against IRAK-2-driven pathogenic signaling.

The innate immune response and inflammatory signaling play determinant roles in brain homeostasis, neuroprotection, and repair; however, altered or excessive signaling in these injury defense systems contributes to the irreversible degeneration of brain cells, as typified in the common, age-related neurodegenerative disorder Alzheimer disease (AD). 2 In innate immune signaling members of the Toll-like receptor (TLR) or IL-1 receptor (IL-1R) superfamily, via their common Toll and IL-1R (TIR) domains, act as extracellular sensors to detect pathogens and cytotoxic molecules. This enables cells to respond to toxins and inflammatory cytokines by mounting effective neuroprotective immune responses (1)(2)(3)(4)(5). A family of interleukin-1 receptor-associated kinases (IRAKs) in the human genome, including IRAK-1, IRAK-2, IRAK-4, and IRAK-M, are key mediators in the immune pathways utilized by TLR/IL-1R (TIR) signaling (3)(4)(5)(6)(7). By means of their integral kinase and multiple adaptor functions, IRAKs initiate diverse downstream signaling processes and a cascade of events that can eventually lead to the induction of pro-inflammatory transcription factors such as NF-B. Further recruitment of NF-B essential modulator (NEMO/IKK␣/␤) and adaptor proteins either enhance or mis-regulate both the innate immune response and inflammatory gene expression (5)(6)(7)(8).
Abundant DNA array, Northern, RT-PCR, and Western gene expression analysis of AD brains have repeatedly shown a significant disruption in the homeostatic expression of essential brain genes and a progressive up-regulation of inflammatory gene expression, driven in part by overactivation of transcription factor NF-B. This supports both the development and progression of neurodegenerative disease processes (9 -16). Indeed the TLR/IL-1R-IRAK-NF-B signaling axis is substantially over-stimulated in AD brain (6, 9 -12). Components of this innate immunity and inflammatory pathway are known to play a central role in driving neuropathology, in part via overexpression of interleukin-1 ␤ (IL-1␤) and upregulating the generation of the 42-amino acid amyloid ␤ 42 (A␤42) peptide. These in turn induce transcription from the * This work was supported, in whole or in part, by National Institutes of Health Grant AG18031 (NIA; to W. J. L.). This work was supported in part by a Translational Research Initiative grant from Louisiana State University (to W. J. L.) and by Alzheimer Association Investigator Initiated Research Grant IIRG-09-131729 (to W. J. L.). 1  proinflammatory prostaglandin synthase cyclooxygenase-2 (COX-2) gene and stimulate apoptotic brain cell death and neural tissue degeneration (5, 9 -12, 17-19). MicroRNAs (miRNAs) have emerged as important epigenetic, post-transcriptional, regulators of brain gene expression and the immune response and have been recently implicated in a surprisingly wide variety of human brain disorders including AD (20 -27). A mouse and human brain abundant miRNA-146a has been specifically associated with up-regulated inflammatory signaling in temporal lobe epilepsy (21), in prion-induced neurodegeneration (22), in down-regulating IRAK-1 in endotoxin-and cytokine-challenged human monocytes (28), and in the down-regulation of complement factor H, an important repressor of inflammatory signaling in AD brain (29). The miRNA-146a has been found to be further up-regulated in cytokine-, A␤42-, oxidation-, or neurotoxic metal-stressed primary human neuronal glial (HNG) co-culture cell models of AD (11,29,30,32,33).
In this report we have significantly expanded, refined, and advanced our earlier studies on miRNA-146a-mediated signaling in AD brain and in stressed brain cells. No miRNAs have as yet been functionally linked to specific IRAK signaling in the human brain. These studies were undertaken to further understand the involvement of the brain-enriched, stressinduced miRNA-146a in the molecular-genetic mechanism that drives TLR/IL-1R-IRAK-NF-B-mediated inflammatory signaling and AD-type change in primary cultures of human astroglial (HAG) cells, a highly specialized cell type known to contribute to the brain innate immune and inflammatory response.
Human Brain Tissues-With the emergence of sophisticated techniques for gene expression analysis, the quality of tissues being studied becomes increasingly critical. Brain tissues used in these studies were carefully selected from several hundred potential specimens obtained from archived tissues or extracts at the LSU Neuroscience Center, New Orleans LA, the University of California at Irvine, California, and the Oregon Health Sciences Center, Portland, OR. Human brain tis-sues were used in accordance with the institutional review board at the LSU Health Sciences Center and donor institutions (6,14,15). Table 1 summarizes the selection of brain tissues used in this study. All AD brain tissues were sporadic; as post-mortem interval (PMI; death to brain-freezing interval) is a factor that can affect RNA quality (2, 7, 9 -14, 22), all RNAs were derived from control or AD tissues having a mean PMI of 3.1 h or less. Consortium to Establish a Registry for Alzheimer's Disease/NIH criteria were used to categorize AD tissues in accordance with established guidelines; AD tissues used in these studies had a clinical dementia rating (an index of cognitive decline) ranging from a clinical dementia rating of 0.5 to 3.0, indicating mild to a severe stage of this neurological disorder (29,36).
miRNA Isolation from Human Brain Tissues and Cells-In brain tissue studies 10-mg wet weight samples were isolated from the superior temporal lobe neocortex (Brodmann area A22), hippocampus, brain stem, thalamus, or cerebellum of AD brain or age-matched controls. Tissues were gently homogenized using a mini-pestle homogenizer in tissue isolation buffer (Kontes; Fisher; Refs. 10,11,29). In HAG cell studies treated or untreated cells from 3 to 5, 70% confluent 3.5-cm diameter 6-well CoStar plates were scraped and pooled, taken up into a 20-ml syringe, RNase-and DNase-free, diethyl pyrocarbonate plasticware-treated, and gently packed using centrifugation (10,30). For both tissues and cells a guanidine isothiocyanate/ silica gel-based membrane total RNA purification system was utilized to isolate total miRNA RNA from each sample (PureLink TM , Invitrogen). Total RNA concentrations were quantified using RNA 6000 Nano LabChips and a 2100 Bioanalyzer (Caliper Technologies, Mountainview, CA; Agilent Technologies, Palo Alto, CA). Total RNA yield was typically about 1.3 g of total RNA per mg wet weight of tissue. No significant differences between the spectral purity or molecular size of small RNA between AD and control tissue samples were noted (Table 1). Total small RNA samples were typically run out on 15% Tris borate-EDTAurea polyacrylamide denaturing gels (TBE-urea; Invitrogen) and after ethidium bromide staining total miRNA species (Ͻ25 nucleotides) were excised and end-labeled using [␥-32 P]␦ATP (6000 Ci/mmol) according to the manufacturer's protocols (Invitrogen) and as previously described (11,29,30,32).
DNA Arrays and Brain-enriched miRNAs-As a preliminary screen and to obtain general trends for miRNA abundance, total miRNA was pooled and analyzed as an AD group (n ϭ 36) and an age-matched control group (n ϭ 30) (Fig. 1) using commercially available miRNA arrays (LC Sciences, Houston TX; Ref. 29 and data not shown). Specific controls and miRNAs showing strong hybridization signals in disease or controls were studied further using robotically generated miRNA panels (11,29,30,32). Briefly, DNA targets for human 5 S ribosomal RNA (5 S RNA), miRNA-132, and miRNA-146a (Table 2) were spotted onto GeneScreen Plus nylon membranes either by hand pipetting or by using a Biomek 2000 laboratory automation work station (Beckmann, Fullerton, CA). These mini-miRNA array panels were then miRNA-146a and IRAK-2 Signaling in AD Brain cross-linked, baked, hybridized, and probed according to the manufacturer's protocol (NEN Research Products, Boston MA) (11,29,30,32). Every second mini-miRNA array panel generated was normalized by probing with purified single radiolabeled miRNAs (5 S RNA, miRNA-132 and/or miRNA-146a) to ascertain equivalent 5 S RNA and individual miRNA loadings (29). Mini-miRNA panels were subsequently probed with total labeled miRNAs isolated from various AD brain regions or stressed HNG or HAG cells and controls; AD, HAG cell, or control extracts (20 g) containing miRNA or 5 S RNA (5 g) were run out on 15% TBE-urea denaturing gels, transferred to GeneScreen membranes, cross-linked, baked, hybridized, and probed with specific DNA oligomers corresponding to specific miRNAs ( Table 2), radiolabeled using [␥-32 P]␦ATP (6000 Ci/mmol) and a T4 polynucleotide kinase labeling system (Invitrogen) (11,29,30,32).
HNG and HAG Cells in Primary Culture-HNG cells were cultured as previously described (11,15,29) (Fig. 2). Human astroglial (HAG) cells (CC-2565; Lonza) were grown using an astrocyte growth medium consisting of astrocyte basal medium (Lonza CC-3187) supplemented with astrocyte growth medium SingleQuots (Lonza). Astrocyte basal medium was changed at 2-day intervals; after 1 week of culture HAG cells received at each astrocyte growth medium change IL-1␤ (10 ng/ml; I4019, Sigma) plus A␤42 peptide (5 M; Sigma); control HAG cells received cell culture grade human serum albumin (Sigma; containing no biological activity) at the same concentrations as a control. After the additions, HAG cells were cultured for 0.5 additional week after which total RNA and protein fractions were prepared (14 -16).
Signal Quantitation, Data Analysis, and Interpretation-5 S ribosomal RNA (5 S RNA), an abundant 107 nucleotide structural RNA, and the 22 nucleotide miRNA-132 are two human brain-abundant small RNAs used as internal controls for miRNA-146a signal determinations in each brain tissue or cell sample. Relative miRNA-132, miRNA-146a, IRAK-1, and IRAK-2 mRNA signal strengths were quantified against 5 S RNA in each sample using data-acquisition software provided with a GS250 molecular imager (Bio-Rad). IRAK-1 and IRAK-2 promoters were searched for transcription factor binding sites using the PROMO 3 (University of Catalonia, Barcelona, Spain) or the TRANSFAC (BioBase Biological Databases, Wolfenbuettel, Germany) DNA sequence search algorithms. Graphic presentations of resultant data were analyzed using Excel algorithms (Microsoft, Seattle, WA), and figures were generated using Adobe Photoshop 6.0 and Adobe Illustrator CS3 (Adobe Systems, San Jose CA). Statistical significance was analyzed using a two-way factorial analysis of variance (p, ANOVA; SAS Institute, Cary, NC). A p Ͻ 0.05 was deemed as statistically significant; experimental values are expressed as the means Ϯ 1 S.D. Table 1 shows the number, age, age range, PMI, RNA optical quality (A 260/280 ), RNA 28 S/18 S ratios, and RNA yields of adult control (n ϭ 30) and Alzheimer (n ϭ 36) brain tissues, many of which were previously analyzed for global miRNA and mRNA gene expression patterns (9 -12, 29, 30, 32, 39) (Fig. 1). PMIs for age-matched control or AD human brain tissues were all Յ3.1 h; in focused studies involving IRAK-1, IRAK-2, and IRAK-4 abundance, shorter PMIs were selected from this larger group (Fig. 2). The entire study group tissues exhibited no significant differences in age (71. 5 45, p Ͻ 0.91) age-matched control and AD, respectively. We further noted no differences in total RNA yields between the control and AD groups, although there was a trend in some miRNA-146a and IRAK-2 Signaling in AD Brain DECEMBER 10, 2010 • VOLUME 285 • NUMBER 50 younger adult brains and the cerebellum for a slightly higher total RNA yield (29,30).

Selection of Human Brain Tissues and Messenger RNA Quality-
miRNA-146a Up-regulation in AD Brain-miRNA-146a was found to be consistently up-regulated in AD brain (Fig. 1,  A and B) as a function of disease progression (Fig. 1C), and this change was focused in specific brain regions that exhibit AD neuropathological change (Fig. 1D).

IRAK Signaling in AD Neocortex and in Stressed HAG
Cells-In relation to control ␤-actin levels, we observe in control brain tissues higher relative levels of IRAK-1 compared with IRAK-2 within the same tissue sample (Fig. 2). However, in relation to a control ␤-actin levels within the same tissue sample, we observe in AD-affected brain tissues higher relative levels of IRAK-2 and lower relative levels of IRAK-1 (Fig.

TABLE 1 Summary of tissues from control and Alzheimer groups used in this study
n ϭ number of individual brain samples. Age is time of death in years. Age range indicates ranges of the individual means. PMI (death to brain freezing interval) range is the range in mean in hours. RNA A260/280 and RNA 18 S/28 S mean ratios are indicative of high brain tissue RNA spectral quality (10 -12, 29, 39). There was no significant difference between the mean yield of total RNA between the control or AD tissues. Characterization of control and AD total RNA message: a , mean death to brain freezing interval in hours at Ϫ81°C; b , average yield in total g of RNA/mg wet weight brain tissue.

miRNA-146a and IRAK-2 Signaling in AD Brain
2, A and B). Notably we found no change in the relative levels of IRAK-4 or IRAK-M between either control or AD-affected brain. Because levels of miRNA-146a were significantly higher in HAG cells versus HNG cells when either stressed with IL-1␤ and A␤42 peptides and because HAG cells are an integral part of the brain neurovascular unit and inflammatory response (33)(34)(35)(36), stressed and control HAG cells were used in all subsequent experiments (Fig. 2C).
Because the proinflammatory cytokine IL-1␤ and A␤42 peptide production are significantly up-regulated in AD brain and play a central role in the regulation of inflammatory responses and increase in abundance as the disease progresses (12)(13)(14)(15)(16)(17)(18)(19), we chose next to examine the effects of these inducible pathogenic peptides on HNG and HAG cells in primary culture. Compared with HNG cells, HAG cells exhibit a significantly higher miRNA-146a up-regulation (Fig. 2C). Either IL-1␤, A␤42 peptide, or IL-1␤ϩA␤42 peptide together induced IRAK-2 and miRNA-146a but not IRAK-1 in this particular cell type (Fig. 2D). Because IL-1␤ϩA␤42 peptide together had a significantly stronger induction of IRAK-2 and miRNA-146a, this combination was used in all subsequent experiments.
DNA Sequence Characteristics of Oligonucleotides Used- Table 2 and Fig. 3 show the sequences of synthetic oligonucleotides used in these studies. DNA sequence characteristics and stabilities for 5 S RNA, miRNA-132, miRNA-146a, AM-146a, and AM-146ac have been previously described in detail (29,39) . Figs. 3, A and B, further describe the human IRAK-1 immediate promoter located at human chromosome Xq28 containing 7 Sp1 GC-rich DNA binding sites and the human IRAK-2 immediate promoter located at human chromosome 3p25.3. Fig. 3C shows the miRNA-mRNA complementarity sequence alignment between miRNA-146a and the IRAK-1 mRNA 3Ј-UTR containing 15/22 matches, or 68% complimentarity, and a calculated free energy of association of Ϫ29.1 kcal/mol (MIRBASE algorithm; Cambridge UK). Notably, this feature of the miRNA-146a-IRAK-1 mRNA 3Ј-UTR recognition is conspicuously absent from the IRAK-2 mRNA 3Ј-UTR, although it is present in the 3Ј-UTRs of other inflammation-related mRNAs such as those encoding complement factor H and the tumor necrosis factor receptor-associated factor 6 (TRAF6 (11, 28 -30, 32)).  Table 1. B, quantified levels from A in the bar graph format are shown; note the inverse relationship between IRAK-1 and IRAK-2 in control and AD brains, which could be in part due to containment of an NF-B binding site in the human IRAK-2 but not the human IRAK-1 immediate promoter (Fig. 3). No significant changes were observed in the levels of IRAK-4 or IRAK-M between control and AD brains (data not shown); a horizontal dashed line at 1.0 is included for ease of comparison. C, in comparison to stressed HNG primary cells, stressed HAG primary cells exhibit a greater upregulation of miRNA-146a, perhaps due to their nature as an immune-responsive cell type (33,34). Neuronal cells are stained with neuron-specific ␤-tubulin (red; ϭ 690 nm), glial cells are stained with glial-specific glial fibrillary acidic protein (GFAP; green; ϭ 525 nm), and nuclei are stained with Hoechst 33258 (blue; ϭ 470 nm). Magnification, 20ϫ. A horizontal dashed line at 1.0 corresponds to control levels of miRNA-132 in HNG cells; *, p Ͻ 0.05; **, p Ͻ 0.01. D, relative IRAK-1, IRAK-2, and miRNA-146a abundance in control and IL-1␤ϩA␤42-stressed HAG cells is shown. A horizontal dashed line at 1.0 corresponds to control levels of IRAK-1, IRAK-2, and miRNA-146a; n ϭ 3-6; significance over control: *, p Ͻ 0.05; p Ͻ 0.01 (ANOVA).

TABLE 2 DNA sequences of 5 S RNA, miRNA-132, miRNA-146a, and anti-miRNA-146A (AM-146A) and AM-146a control probes used in this study
The 5 S RNA probe was derived from the first 22 nucleotides of the 107 nucleotide human 5 S ribosomal RNA (5 S RNA) (29). AM-146a control (AM-146ac), containing the same nucleotide composition as anti-miRNA-146a, is a scrambled anti-miRNA-146a oligonucleotide used as a gene expression control (Fig. 5).

JOURNAL OF BIOLOGICAL CHEMISTRY 38955
Relative Abundance of Sp1 and NF-B in IL-1␤ϩA␤42stressed HAG Cells-Using a gel shift assay, stressed-HAG cells exhibited a significant 3.6-fold up-regulation of relative signal strength for NF-B with no change in relative signal strength for Sp1. This NF-B activation was quenched using three different NF-B inhibitors (Fig. 4, A-C).
miRNA-146a Is Transcribed from a NF-B-regulated Gene-miRNA-146a is transcribed from a pre-miRNA-146a gene, and the 5Ј regulatory region of that gene has been sequenced, revealing 3 upstream NF-B-DNA binding sites in the immediate promoter (NT_023133 (28,29)). The current data suggest that a significant part of the inflammatory signaling pathway in IL-1␤and A␤42-triggered HAG cells involves an up-regulation of NF-B, an important transcription factor responsible for driving transcription of not only miRNA-146a but also of related pro-inflammatory genes and their interrelated pathogenic signaling cascades (Fig. 4B) (10, 14).
Effects of NF-B Inhibitors on miRNA-146a, IRAK-1 and IRAK-2 Promoter-luciferase Reporters-Up-regulation of NF-B was associated with a direct increase in the IRAK-2 but not the IRAK-1 promoter reporter (Fig. 4C). The inclusion of NF-B inhibitors was associated with a significant down-regulation in IRAK-1 promoter activity and a stimu-lated expression of the IRAK-1 luciferase reporter (Fig. 4C). A coordinated interplay of miRNA-146a and NF-B signaling thereby appears to regulate IRAK-1 and IRAK-2 expression in IL-1␤ϩA␤42-stressed HAG cells; miRNA-146a was induced about 3.5-fold over control, and miRNA-146a activation was quenched by three different NF-B inhibitors. As monitored by luciferase reporter, IRAK-1 was induced 1.2-fold over control, and IRAK-2 was induced 3.5-fold over controls. In the FIGURE 3. DNA sequence structure of the human IRAK-1, human IRAK-2 gene immediate promoters, and a highly stable miRNA-146a-IRAK-1 mRNA-3-UTR interaction. A and B, the human IRAK-2 promoter contains a single NF-B binding consensus sequence from Ϫ111 to Ϫ102 bp of the IRAK-2 promoter. This feature is missing from the immediate IRAK-1 promoter. Further studies showed that the IRAK-2, but not the IRAK-1 gene, is under NF-B transcriptional control (Fig. 4). XhoI and BglII restriction sites (underlined) were used for ligation into pGL3 vectors (see text). The bent arrow at ϩ1 indicates the start of transcription. AP2␣, NF-B, and Sp1 binding sites are bold and highlighted in yellow. C, the sequence of the 22 nucleotide miRNA-146a (highlighted in red) shows highly specific complementarity to the human IRAK-1 mRNA 3Ј-UTR (the target sequence is highlighted in yellow). 14 of 22 base pairs of the 5Ј end of miRNA-146a align. The structural stability of the 22 nucleotide oligomer is Ϫ29.1 kcal/mol. This feature is absent from the IRAK-2 mRNA 3Ј-UTR (data not shown). Expression data suggest that the IRAK-1 mRNA 3Ј-UTR, but not the IRAK-2 mRNA, is regulated by miRNA-146a (28).

miRNA-146a and IRAK-2 Signaling in AD Brain
presence of NF-B inhibitors, miRNA-146a is down-regulated, resulting in the up-regulation of IRAK-1 expression; similarly, in the presence of NF-B inhibitors, the expression of IRAK-2 is quenched to near control levels. In these studies PDTC was found to be the most efficient NFB, miRNA-146a and IRAK-2 inhibitor (Fig. 4) and was used in all subsequent studies.
Specific Inhibition of miRNA-146a and NF-B-We next studied the effects of anti-miRNA-146a (AM146a) using AM-146ac as a control and the highly efficient NF-B inhibitor PDTC on IRAK-1, IRAK-2 mRNA, and miRNA-146a expression in control and IL1␤ϩA␤42-stressed HAG cells (Fig.  5A). AlthoughAM146a de-repressed IRAK-1 and up-regulated IRAK-1 to Ͼ3-fold over controls, no significant effect was noted on IRAK-2 levels. The control AM-146ac had no significant effect on either IRAK-1 or IRAK-2 expression.
Similarly, when the NF-B inhibitor PDTC was included in the assay, which down-regulated miRNA-146a (Fig. 4A), IRAK-1 increased in expression, and IRAK-2 decreased in expression. The inclusion of both AM-146a and PDTC significantly up-regulated IRAK-1 expression and down-regulated both IRAK-2 and miRNA-146a. These effects are reiterated in the expression profiles of IRAK-1 and IRAK-2 protein shown in Fig. 5B.

DISCUSSION
As techniques in quantitative analytical gene expression in human neurobiology advance, the quality of the brain tissue samples being studied becomes increasingly critical (38 -41). This is especially important in the study of regional gene expression patterns in human brain where the PMI is a key determinant of total RNA quality (38,39,41). As a measure of human superior temporal lobe messenger RNA integrity, previous studies utilizing total RNA isolation or run-on gene transcription and Northern dot-blot hybridization of newly synthesized RNA indicated that human brain extracts of up to about 4 h PMI were efficient in incorporating [␣-32 P]UTP radiolabel into new DNA transcription products highly representative of the physiological status of the cell, after which there was a precipitous decline in de novo mRNA biosynthetic capacity and RNA quality (38). Other more recent studies have reiterated these observations and concerns (39 -41). In the current studies, control and AD-affected human brain tissues (n ϭ 66) were carefully selected from hundreds of potential specimens in domestic brain tissue banks, and tissues were carefully selected with a mean PMI averaging 3.1 h or less (Table 1). In these same tissues we show a consistent, significant increase in the AD brain of miRNA-146a compared with the internal controls 5S RNA and an unchanging brain abundant miRNA-132 (Fig. 1, A and B), a significant increase in miRNA-146a as the severity of AD progresses (Fig. 1C), and a significant compartmentalization of these effects to the hippocampus and temporal lobe neocortex of the brain, but not in the brain stem, cerebellum, or thalamus of anatomical regions unaffected by the AD process and within the same brain (Fig. 1D). Importantly, the temporal lobe neocortex, a brain area of higher-order cognitive function, is specifically targeted by AD neuropathology and is a primary region of interest that is used to indentify early events in Alzheimer-type neurodegenerative change (37,38,42).
Interestingly, we observe IRAK-1 to be absent in young human brains but relatively abundant in aged human neocortex ( Fig. 2A and Ref. (6)). In aged human neocortex we further observe a highly significant increase in the expression of IRAK-1 in control over AD and, conversely, an up-regulation of IRAK-2 in AD over control. We further note no significant change in either IRAK-4 or IRAK-M expression in comparison to ␤-actin controls within the same brain tissue samples (Fig. 2, A and B; data not shown).
Because the proinflammatory cytokine IL-1␤ and A␤42 peptide production are strongly up-regulated and accumulate in AD brain and also play determinant roles in driving immune and inflammatory responses, we next chose to examine the effects of these inducible pathogenic peptides in primary A, we provide evidence that a complex interplay of miRNA-146a and NF-B signaling regulates IRAK-1 and IRAK-2 signaling in IL-1␤ϩA␤42stressed HAG primary cells. In HAG cells, IRAK-1 mRNA, IRAK-2 mRNA, and miRNA-146a were found to be induced about 1.1-, 3.1-, and 4.0-fold over controls. The effects of AM-146a on up-regulating IRAK-1 and PDTC on down-regulating IRAK-2 suggest that these are, respectively, miRNA-146aand NF-B-regulated genes. Further evidence for this is reflected in the structure of their immediate gene promoters (Fig. 3) and from the pattern of IRAK-1 and IRAK-2 protein abundance compared with the levels of an internal ␤-actin control in the same cell sample (B); n ϭ 4 -5; significance over control: *, p Ͻ 0.05, p Ͻ 0.01 (ANOVA). DECEMBER 10, 2010 • VOLUME 285 • NUMBER 50 cultures of HNG and HAG primary cells. HAG cells were found to exhibit a significantly up-regulated expression of the inflammation-associated miRNA-146a in comparison to HNG cells (Fig. 2C). HAG cells are of further interest as they are an important cell type of the human neurovascular unit and a major modulator of the brain innate immune and inflammatory response (33)(34)(35)(36). Either IL-1␤ alone or A␤42 peptide alone and especially IL-1␤ϩA␤42 peptide together induced IRAK-2 and miRNA-146a but not IRAK-1 in this particular cell type (Fig. 2D). Because IL-1␤ϩA␤42 peptide together had a significantly stronger induction of IRAK-2 and miRNA-146a, this combination was used in all subsequent experimentation. We further examined the mechanism of this up-regulation by examining the promoter structure of the human IRAK-1 and IRAK-2 genes, which are located on two different human chromosomes (chr Xq28 and chr 3p25.3 respectively; Fig. 3, A and B). The human IRAK-1 immediate promoter is extremely GϩC-rich (83%) and contains 11, sometimes overlapping, binding sites for the GC box element 5Ј-(C/G)CC(C/G)N(C/G)N(C/G)-3Ј recognized by the zinc finger transcription factor Sp1 (Fig. 3A). We also note an AP-2␣ binding site (5Ј-AGAGGC-3Ј) centered at Ϫ97 bp from the start of transcription of the IRAK-1 gene; however, as analyzed by gel shift assay, the levels of AP-2␣ were found not to significantly change after IL-1␤ϩA␤42 peptide-induced stress in HAG cells (data not shown). On the other hand the human IRAK-2 immediate promoter is far less GϩC-rich (41%) and contains putative 2 Sp1 binding sites and a single NF-B binding site 82% homologous to the canonical NF-B binding site of 5Ј-GGGGRNNYY(C/G)(C/ G)-3Ј (19, 28 -30) located from Ϫ111 to Ϫ101 bp upstream of the IRAK-2 start of transcription. We note that the NF-B binding site partially overlaps with a relatively nonspecific zinc finger-rich CTCF binding site (consensus 5Ј-CNNNNNNNCCCTC-3Ј), which has been previously shown to be important in transcriptional control of the IRAK-2 promoter in non-neural cell types (45). We further found that DNA sequences upstream of Ϫ130 bp (Ϫ270 to Ϫ130 bp) in the human IRAK-2 promoter had insignificant effects on luciferase reporter activities in stressed HAG cell studies ( Fig. 3; data not shown).

miRNA-146a and IRAK-2 Signaling in AD Brain
To further assess the importance of these transcription factors in an IL-1␤ϩA␤42-induced HAG cell response, we next examined Sp1-and NF-B-DNA binding using a gel shift assay (5,19,28). We observed a significant up-regulation of NF-B-DNA p50/p65 binding in stressed HAG cells. Although Sp1-DNA was not significantly induced during this treatment, NF-B-DNA binding was increased 3.6-fold over controls. The NF-B inhibitors CAY10512, curcumin, and PDTC sharply quenched this induction (Fig. 4A). Previous studies have shown NF-B-DNA binding to be significantly up-regulated in AD brain and in IL-1␤, A␤42 peptide, or oxidationstressed human brain cells (19,29).
We next studied relative luciferase signal yield in promoter constructs of the human IRAK-1 and IRAK-2 gene promoter transfected into control and stressed HAG primary cultures. Because miRNA-146a is an NF-B-regulated gene, we also examined the activity of the A547 construct containing the miRNA-146a promoter with luciferase reporter (29). miRNA-146a was found to be up-regulated about 3.5-fold by IL-1␤ϩA␤42 peptide, but this induction was significantly attenuated in the presence of three different classes of NF-B inhibitors (Fig. 4B). In related studies the IRAK-2 promoter construct was found to be induced 3.6-fold over controls in the presence of IL-1␤ϩA␤42 peptide; however, this induction was quenched to non-significant levels in the presence of 3 different NF-B inhibitors (Fig. 2C). This suggests that IRAK-2 is under NF-B regulatory control, as would be suggested by an NF-B binding site in the IRAK-2, but not the IRAK-1 immediate promoter. In these experiments IRAK-1 was found not to be up-regulated in the presence of IL-1␤ϩA␤42 peptide and an up-regulated NF-B but instead was moderately up-regulated to about 2.3-fold over controls when 3 NF-B inhibitors were present (Fig. 2C). Given the known decrease of miRNA-146a in the presence of NF-B inhibitors, we reasoned that increases in IRAK-1 were the result of decreases in miRNA-146a, a negative regulator of IRAK-1, due to its strong interaction with the IRAK-1 mRNA 3Ј-UTR (Fig. 3C) (28,29).
To further sort out this apparent NF-Band miRNA-146a-mediated control of IRAK-1 and IRAK-2 expression, we next used differential combinations of a specific miRNA-146a antisense sequence (anti-miRNA-146a; AM-146a) and the three different NF-B inhibitors to dissect the activation of IRAK-1, IRAK-2, and miRNA-146a in control and stressed HAG cells. Although AM-146a had significant inducing effects on IRAK-1 mRNA, no significant effects were observed on the abundance of IRAK-2 mRNA (Fig. 5A). In the presence of the highly efficient NF-B inhibitor PDTC, IRAK-2 and miRNA-146a expression was inhibited to less than control levels, suggesting that the regulation of these two elements by NF-B and the de-repression of IRAK-1 via miRNA-146a expression-quenching by PDTC (Figs. 4A and 5A). Interestingly, the inclusion of both AM-146a and PDTC very significantly quenched both IRAK-2 and miRNA-146a expression (Fig.  5A). In this instance IRAK-1 levels increased significantly again due to the inhibition of miRNA-146a in the system (Fig.  5A). Expression of IRAK-1 and IRAK-2 at the protein level was found to be similarly affected (Fig. 5B). Although IRAK expression has shown cell-type specificity (Fig. 2C) (45), these results suggest that a significant part of IRAK-1 or IRAK-2 transcriptional control lies within the immediate promoter region in stressed HAG cells. We cannot exclude that other transcription factors or DNA sequences further upstream from the human IRAK-1 or IRAK-2 promoter sequences studied here have ancillary regulatory controls on IRAK-1 or IRAK-2 gene expression in different brain cell types.
A normally functioning TIR and innate immune response is required to maintain healthy immune defense; both acute and chronic inflammatory pathological conditions arise if these systems are induced either too strongly or for too long. IRAK-1 was originally thought to be central to the TLR/IL-1R-NF-B signaling axis; however, recent data show that it is dispensable for NF-B activation via some TLRs and that IRAK-2 is the more critical element for a more sustained TLR-IL-1R-mediated NF-B activation (4 -8, 44). In AD, miRNA-146a and IRAK-2 Signaling in AD Brain emerging evidence supports the hypothesis that the TLR/IL-1R-IRAK-NF-B innate immunity pathway plays a regulatory role in mediating the neuropathological effects of the A␤42 peptide (3,10). In AD this signaling pathway appears to provide a critical link between immune stimulants, such as A␤42 peptides, and the initiation of host defense, as A␤42-mediated TLR-IL-1R activation further modulates the release of inflammatory cytokines and COX-2 via downstream effects (2,3,43). Interestingly, A␤42 peptide stimulation of TLR3, TLR5, TLR8, TLR9, and TLR10 transcription has been found to be severely depressed in AD mononuclear cells, and down-regulation of these TLRs may impair microglia-mediated clearance of A␤42-deposits in the AD brain (43,44). Interestingly, IL-1␤ϩA␤42 peptides are both increased in AD brain, and IL-1␤ϩA␤42 peptide-stressed brain cells emulate many of the neurochemical, pathological, and gene expression changes characteristic of AD (10, 11, 14 -16 , 29, 30, 32). These studies, therefore, provide evidence in an immune-responsive brain cell type that a stress-induced NF-B-activated, miRNA-146a-mediated down-regulation IRAK-1 coupled to an NF-B-driven up-regulation of IRAK-2 provides an important basis for a self-perpetuating inflammatory signaling loop. Notably, in primary macrophages, IRAK-2 activation has been shown to be essential to mount an optimal response to TLR/ IL-1R (8). The presence of an alternately regulated IRAK-1/ IRAK-2 signaling system adds a further degree of complexity to the variety of mechanisms known to regulate the brain innate immune and inflammatory response both in health and disease and in primary cell models used to study AD-type neurodegenerative mechanisms (11,26,27,29,30). Importantly, in these studies no significant changes were observed in IRAK-4, IRAK-M, MyD88, or TRAF6 abundance either in AD or in stressed HAG cells ( Fig. 2 and data not shown). It will be interesting to study other adaptor components of the TLR/IL-1R-NF-B cascade and in particular NEMO/IKK␣/␤ signaling to learn how they further modulate pathogenic activities.
In summary, miRNA-146a is emerging as a key small RNA regulator of the innate immune response and pro-inflammatory signaling in several human neurodegenerative diseases associated with a strong inflammatory component (21,22,29,39,46,47). The current work describes an NF-Band miRNA-146a-mediated IRAK-1 and IRAK-2 expression network in IL-1␤ϩA␤42-treated HAG cells stressed for up to one-third of their in vitro lifespan. Analogous features are observed in specifically affected anatomical regions of AD brain. Notably, this TLR/IL-1R-IRAK-2-NF-B self-reinforcing pathogenic loop occupies a key central position within the TLR/IL-1R-NF-B signaling axis (Fig. 6). AD and stressed brain cells also associate with a significant down-regulation in another miRNA-146a target encoding complement factor H, a key repressor of inflammatory signaling in the complement cascade (11,29,30). NF-B and miRNA-146a are also sharply up-regulated in human brain cells in response to an HSV-1 challenge and to other neurotrophic viral infections as well as in other human neuroinflammatory diseases (11,21,22,46,47). These findings further illustrate an important instance where a specific pro-inflammatory transcription factor and an immune system-related miRNA cooperate to orchestrate pathogenic responses in brain disease and further suggest that up-regulation of miRNA-146a is an important contributor to inflammatory neuropathology (11, 21, 22, 26 -30). The use of directed anti-miRNA strategies to repress the effects of upregulated miRNAs may represent an effective therapeutic approach; antisense-miRNA strategies may be preferred over antisense-mRNA strategies because of the potential of a single antisense-miRNA to affect the regulation of multiple diseaserelated genes. Indeed the combinatorial use of AM-146a with selective NF-B inhibitors may have potential as an effective bi-pronged therapeutic attack against chronically mis-regulated, pathogenic IRAK-2-mediated signaling pathways that drive neurodegenerative disease processes (Fig. 6). FIGURE 6. Schematic illustration of the TLR/IL-1R-IRAK-NFB connectivity and signaling in AD brain and in IL-1␤؉A␤42 peptide-stressed HAG cells. IRAK-1 and IRAK-2 occupy important early control points in TLR/IL-1R-NF-B signaling pathway (6 -8, 43, 44). A membrane-spanning TIR transduces signals triggered by extracellular toxins, peptides, and inflammatory cytokines via MyD88, IRAK-1, and IRAK-2. These ultimately trigger NF-Bmediated inflammatory gene expression, neuropathology, and an altered innate immune response. The TLR/IL-1R shown here is generic and may represent multiple types of the IL-1 receptor sensor whose identities in the human brain are not well understood (6,8,31,44). The normally high abundance of IRAK-1 in control aging human brain and HAG cells (Fig. 2) and the IRAK-1-NF-B circuit (dotted lines) appears to be virtually shut down by an NF-B-mediated up-regulation of miRNA-146a (a negative regulator of IRAK-1) and is associated with a shift to strong up-regulation of an NF-Bactivated IRAK-2. IRAK-2 transcription appears to be highly cell-specific as 10-fold differences in the same IRAK-2 immediate promoter construct between A549 and THP-1 cells have been noted (45). The current data support previous observations that the IRAK-2-NF-B signaling loop is strongly selfreinforcing. This may contribute to the sustained activation of NF-B and its pathogenic consequences in the neurodegenerative process (7,8,44). Although MyD88, NEMO, IKK␣/␤, IRAK-4, TRAF6, and other adaptor proteins are known to contribute to homeostatic TIR-IRAK-NF-B signaling, their interactive role in this pathogenic scheme is currently not known and warrants further study. We have found no change in MyD88, IRAK-4, IRAK-M, or TRAF6 abundance in either AD or stressed HAG cells (Fig. 2, A and B; data not shown), suggesting that IRAK-2 up-regulation alone may be of primary importance in pathogenic signaling. The combinatorial use of miRNA-146a/ AM146a (Fig. 5) along with selective NF-B inhibitors (heavy horizontal black bars) may have significant potential as an effective multi-target therapeutic strategy against pathogenic IRAK-2-mediated inflammatory signaling that drives neurodegenerative processes. DECEMBER