Identification of Cathepsin B as a Mediator of Neuronal Death Induced by Aβ-activated Microglial Cells Using a Functional Genomics Approach*

Alzheimer's disease is a progressive neurodegenerative disease characterized by senile plaques, neurofibrillary tangles, dystrophic neurites, and reactive glial cells. Activated microglia are found to be intimately associated with senile plaques and may play a central role in mediating chronic inflammatory conditions in Alzheimer's disease. Activation of cultured murine microglial BV2 cells by freshly sonicated Aβ42 results in the secretion of neurotoxic factors that kill primary cultured neurons. To understand molecular pathways underlying Aβ-induced microglial activation, we analyzed the expression levels of transcripts isolated from Aβ42-activated BV2 cells using high density filter arrays. The analysis of these arrays identified 554 genes that are transcriptionally up-regulated by Aβ42 in a statistically significant manner. Quantitative reverse transcription-PCR was used to confirm the regulation of a subset of genes, including cysteine proteases cathepsin B and cathepsin L, tissue inhibitor of matrix metalloproteinase 2, cytochrome c oxidase, and allograft inflammatory factor 1. Small interfering RNA-mediated silencing of the cathepsin B gene in Aβ-activated BV2 cells diminished the microglial activation-mediated neurotoxicity. Moreover, CA-074, a specific cathepsin B inhibitor, also abolished the neurotoxic effects caused by Aβ42-activated BV2 cells. Our results suggest an essential role for secreted cathepsin B in neuronal death mediated by Aβ-activated inflammatory response.

Alzheimer's disease is a progressive neurodegenerative disease characterized by senile plaques, neurofibrillary tangles, dystrophic neurites, and reactive glial cells. Activated microglia are found to be intimately associated with senile plaques and may play a central role in mediating chronic inflammatory conditions in Alzheimer's disease. Activation of cultured murine microglial BV2 cells by freshly sonicated A␤42 results in the secretion of neurotoxic factors that kill primary cultured neurons. To understand molecular pathways underlying A␤-induced microglial activation, we analyzed the expression levels of transcripts isolated from A␤42-activated BV2 cells using high density filter arrays. The analysis of these arrays identified 554 genes that are transcriptionally up-regulated by A␤42 in a statistically significant manner. Quantitative reverse transcription-PCR was used to confirm the regulation of a subset of genes, including cysteine proteases cathepsin B and cathepsin L, tissue inhibitor of matrix metalloproteinase 2, cytochrome c oxidase, and allograft inflammatory factor 1. Small interfering RNA-mediated silencing of the cathepsin B gene in A␤-activated BV2 cells diminished the microglial activation-mediated neurotoxicity. Moreover, CA-074, a specific cathepsin B inhibitor, also abolished the neurotoxic effects caused by A␤42-activated BV2 cells. Our results suggest an essential role for secreted cathepsin B in neuronal death mediated by A␤activated inflammatory response.
Alzheimer's disease (AD) 1 is a progressive neurodegenerative disorder with a loss of higher cognitive function (1). The main pathological hallmarks include senile amyloid plaques, neurofibrillary tangles, dystrophic neurites, significant neuronal loss in selected regions of the brain, and a chronic inflammatory response characterized by activated microglia, reactive astrocytes, complement factors, and increased expression of inflammatory cytokines (2)(3)(4). Chronic inflammation plays an important role in AD pathogenesis because numerous epidemiological studies have revealed significant protective effects of nonsteroidal anti-inflammatory drugs (5)(6)(7)(8)(9)(10).
Microglia, the resident macrophages in the brain, play a central role in mediating chronic inflammatory conditions in AD. They are closely associated with amyloid plaques and exhibit a reactive phenotype with elevated expression of cell surface markers, including CD45, Mac1, and major histocompatibility complex class II antigens (11). They are also capable of releasing numerous acute phase proteins, such as ␣-antichymotrypsin, ␣-antitrypsin, complement proteins, and proinflammatory cytokines, such as interleukins 1␣, 1␤, and 6, tumor necrosis factor (TNF)-␣, and others (12)(13)(14)(15)(16).
There is compelling evidence that A␤ peptides serve as inflammatory stimuli to provoke a microglial-mediated inflammatory response that contributes significantly to neuronal loss and cognitive decline (3). In vitro studies have shown that stimulation of microglia by high concentration of fibrillar A␤ results in TNF-␣-dependent expression of inducible nitric oxide synthase and neuronal apoptosis (16,17). Other studies have shown that A␤ can amplify microglial activation by other coexisting inflammatory stimuli, such as lipopolysaccharide (LPS), interferon ␥ (IFN␥), and advanced glycation end products (18). Besides fibrillar A␤42, soluble A␤42 also appeared to be a potent activator of microglia. Purified dimeric and trimeric components of A␤42 peptides from neuritic and vascular amyloid deposits elicit a profound neurotoxicity in hippocampal neurons in the presence of microglia (19). In addition, although the identity of the neurotoxic agent(s) has not been determined, studies indicate that activation of cultured microglia by senile/ neuritic plaque fragments leads to the release of toxic factors that kill cultured hippocampal neurons (20,21). Therefore, identification of molecular targets involved in the initiation and maintenance of microglial activation caused by A␤42 peptides may lead to a better understanding of inflammatory processes leading to AD.
In this study, we investigated molecular mechanisms underlying the A␤42-mediated inflammatory response in microglial cells using large scale profiling of transcriptional induction by A␤42 peptides in murine microglial BV2 cells. Cathepsin B was identified to be one of the 554 genes transcriptionally induced by A␤42. Specific inhibition of cathepsin B in BV2 cells using either small interference RNA (siRNA)-mediated gene silencing or a specific cathepsin B inhibitor leads to diminished toxic effects on primary neurons. Our studies indicate that cathepsin B is a key player in microglial-mediated neuronal death. In addition, by combining large scale array analyses with siRNAmediated gene silencing, our approach provides a useful plat-form to systematically probe molecular pathways involved in inflammation-mediated neuronal degeneration in AD.

A␤ Preparation and Treatment of BV2 Cells with A␤ Peptides-
Freshly sonicated 〈␤42 was prepared by adding 1 ml of DMEM/F12/N2 medium to 1 mg of 〈␤42 lyophilized powder (California Peptide, Napa, CA) and sonicating for 5 min. The resulting 〈␤42 solution was then diluted with DMEM/F12/N2 to a final concentration of 5 or 11 M and added to the BV2 cells for 15-16 h.
Aged A␤ was prepared using a modified protocol from Dr. LaDu's laboratory (24). Briefly, 〈␤42 powder lyophilized with hydroxyfluroisopropanol (California Peptide, Napa, CA) was dissolved in Me 2 SO and resuspended in DMEM/F12 at 44 M before each use. The resulting solution was aged for 48 h at 4°C. On the day of the treatment, the A␤42 solution was diluted with DMEM/F12/N2 to a final concentration of 22 M and added to the BV2 cells for 15-16 h.
RNA Preparation and Library Construction-Total RNA from cells was isolated using Tri Star reagent according to the manufacturer's instructions (Invitrogen). The resulting RNA was treated with RNasefree DNase (Ambion, Austin, TX) for 40 min at 37°C. Isolation of mRNA was carried out using Oligotex (Qiagen). To ensure the quality of mRNA preparations, denaturing agarose gel electrophoresis was used to assess the integrity and purity of each preparation. To decrease the redundancy and increase the relative abundance of rare transcripts, subtractive and normalized libraries were made. Reverse transcription and second strand synthesis were performed using the Superscript II reverse transcription kit (Invitrogen) with an anchored oligo(dT) (30) primer. Subtractive and normalized cDNA libraries were generated using suppression PCR as described previously (26).
Array Production, Probe Generation, and Hybridization-Subtracted and normalized cDNA libraries were cloned into the pCR2.1 vector (Invitrogen) and plated. Single colonies were picked into 384-well plates using an automated colony picker (Q-Pix, Genetix, Cambridge, UK) to grow overnight to serve as templates for PCR. PCR products were spotted on Nylon membranes using a 384-pin head arrayer (Q-bot, Genetix, Cambridge, UK). More than 50,000 individual clones along with standards and empty wells were spotted onto three filters in each array. The probes were generated from three independent BV2 cultures treated with freshly sonicated 〈␤42 peptide (California Peptide) along with controls. Briefly, mRNA was isolated from those cells as described previously. Double-stranded cDNA was generated by reverse transcription and second strand synthesis and digested with RsaI (New England Biolabs, Beverly, MA). The probes were generated through labeling with [ 33 P]dCTP (PerkinElmer Life Sciences) using a Decamer labeling kit (Ambion). Hybridization was performed at 42°C for 24 h in a hybridization solution containing 50% formamide.
Array Analysis and Gene Annotation-A Cyclone phosphorimaging system (Packard Biosciences, Meriden, CT) and GenePix software analysis (Axon Instruments, Foster City, CA) were used to obtain densitometry of the 33 P signal. After normalization with the global background calculated from 1100 empty array spots, the mean and standard deviation from triplicate measurements for each clone was determined to calculate the confidence level (Student's t test) and ratio of expression between control and A␤-activated BV2 cells. Arrays experiments were performed twice with the same probes, and the clones confirmed to be up-regulated in either experiments were submitted for sequencing. Sequence analysis was performed using Phrap (CodonCode, Dedham, MA), and a comprehensive homology analysis was conducted using a Blast algorithm against the RefSeq (NCBI) data base on accelerated hardware (TimeLogic, Reno, NV). Gene classifications were generated by mapping nonredundant genes to gene ontology (GO) "molecular function" terms that are assigned and approved by a GO curator or inference from electronic annotation (27).
Cell Culture and Transfection with siRNA-The BV2 immortalized murine microglial cell line was kindly provided by Dr. Virginia Bocchini and has been described previously (22). Briefly, BV2 cells were maintained in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 100 IU/ml penicillin G and 100 g/ml streptomycin. The cells were kept at 37°C in a humidified atmosphere containing 5% CO 2 .
To inhibit expression of cathepsin B, cathepsin L, TIMP2, and AIF1 in BV2 cells, the cells were transiently transfected with siRNAs (Dharmacon Research, Lafayette, CO) corresponding to part of their coding sequences. The corresponding sequences are aagaagctgtgtggcactgtc for cathepsin B, aagtccacgcacagaagactg for cathepsin L, aaggagtatctaattgcagga for TIMP2, and aacaagcaattcctcgatgat for AIF. The transient transfections with siRNAs were performed with LipofectAMINE Plus Reagent (Invitrogen) as described previously (23). Three hours after transfection, the cultures were washed once and incubated with DMEM/F12/N2 overnight. The next day, the cells were washed with DMEM/F12/N2 before being treated with 100 ng/ml of LPS and 100 ng/ml of IFN␥ or 5-11 M of sonicated A␤42 for 15-16 h. The supernatants were collected to treat primary cortical neurons, and the cells were harvested for total RNA extraction and quantitative RT-PCR.
Primary Cortical Culture and Neurotoxicity Assay-Primary cortical neurons were cultured in serum-free media. Briefly, embryonic day 16 -18 rat cortices were dissected and digested using trypsin and DNase I. After dissociation with repetitive trituration, the cells were plated in Neurobasal medium supplemented with SATO (25) and B27 (Invitrogen). Three days after plating, half of the medium was removed, and fresh neurobasal medium with N2 was added with 5-fluoro-2Ј-deoxyuridine (Sigma) to inhibit rapidly dividing glial cells.
To test the neurotoxic effects of factors secreted by activated BV2 cells, seven days after plating cortical neurons were treated with supernatants from BV2 cells, which were stimulated by LPS/IFN␥, freshly sonicated A␤42, or aged A␤42 for 15-16 h (see Fig. 2A). For the siRNAmediated knock-down studies, BV2 cells were stimulated with freshly sonicated A␤42 for 15-16 h 2 days after the transfection of siRNAs corresponding to sequences of cathepsin B, cathepsin L, TIMP2, and AIF1 genes. For the inhibitor study, BV2 cells were pretreated with 0.5 or 1 M of CA-074 for half an hour before A␤42 stimulation in the presence of 0.5 or 1 M of CA-074. The resulting supernatants were added to primary neurons for 72 h to induce neuronal death. The amount of neuronal death was quantified with the CellTiter-Glo cell viability assay (Promega, WI) per the manufacturer's instructions in a microplate luminometer reader (LB 96V; EG&G Berthold, Germany).
Quantitative RT-PCR-Total RNA isolation and cDNA generation was performed as described under "RNA Preparation and Library Construction." SYBR Green real time PCR amplifications were performed in an Icycler Real-Time Detection System (Bio-Rad). The reactions were performed in duplicate in 25 l of reaction volume with the following PCR conditions: 50°C for 2 min and 95°C for 10 min, followed by 45 cycles of 95°C for 15 s and 60°C for 30 s, followed by 72°C for 40 s. The primers were designed using Primer3 developed by the Whitehead Institute for Biomedical Research. The PCR primers used were TGAG-GTAGACAGGGGAGTCTTT and GTGTTTGGAGGGTAGCTCTGAT for cathepsin B; TTGGGAACAGCCTTAGCTACTC and ATATCGCTCTC-CTCCACTCTTC for cathepsin L; GGAGACGTTCAGCTACTCTGACT and ACCAGTTGGCCTCTTGTGTT for AIF1; GCACATTACCCTCTGT-GACTTC and AGCGAGTGATCTTGCACTCA for TIMP2; and GACCA-TCAGCACCACTTCAC and GCTCTGTAGAGGAGGGCATTAG for cytochrome c. A standard curve was plotted for each primer set with the critical threshold cycle (C t ) values obtained from amplification of 10-fold dilutions of cDNA from whole brain to calculate the PCR efficiency of the primer set. Glyceraldehyde-3-phosphate dehydrogenase was used an internal reference to normalize the target transcripts, and the relative differences were calculated using the PCR efficiencies according to Pfaffl (28).

Activation of BV2 Cells by Freshly Sonicated A␤, Not Aged A␤, Induced Neuronal
Death-Numerous studies have described how microglial lineage cells secrete neurotoxic factors upon stimulation with synthetic A␤ peptides or A␤ isolated from human plaques. Most of the studies focused on stimulation of microglial lineage cells with fibrillar A␤ at very high concentrations or in combination with IFN␥ (12, 16 -18, 29). More recent studies demonstrate that soluble A␤ peptides, including oligomers and amyloid-derived diffusible ligands, are also neurotoxic. We reasoned that soluble A␤42 might play an important role in microglial-mediated neurodegeneration. In our in vitro system that mimics microglial-dependent neuronal death in AD, we compared neuronal damage in primary rat cortical neurons upon treatment with the conditioned medium from murine microglial BV2 cells stimulated with either freshly sonicated soluble A␤42 or aged A␤42 peptides.
Consistent with prior studies, aged 〈␤42 peptide alone caused significant direct neuronal death in primary cortical neurons (Refs. 30 -32 and Fig. 1). However, when aged 〈␤42 peptide was incubated with BV2 cells, the conditioned medium caused no significant neurotoxicity. The loss of neurotoxicity is likely due to the removal of A␤42 peptides from the conditioned medium after incubation with BV2 cells, probably through phagocytosis as well as proteolysis by secreted degrading enzymes (33)(34)(35). In contrast, freshly sonicated A␤42 peptide alone did not induce direct toxicity in primary cultured neurons (Fig. 1). However, when sonicated 〈␤42 peptide was incubated with BV2 cells, the conditioned medium induced a profound neurotoxicity, suggesting the presence of neuron-damaging factors in the BV2 conditioned medium (Fig. 1). Not surprisingly, LPS and IFN␥, which are well known to activate microglial cells to secrete neurotoxic factors, such as TNF-␣ and nitric oxide, behave similarly as freshly sonicated A␤42. Although they induce no direct neurotoxicity when added directly, LPS/ IFN␥ activated BV2 cells to cause more than 60% cell death in primary cortical neurons (Fig. 1).
Our data suggest that diverging molecular pathways may be involved in the activation of BV2 microglial cells induced by aged or soluble A␤. Indeed, activation of BV2 cells by aged A␤42 results in loss of A␤-mediated neurotoxicity. This effect is likely due to secretion of A␤-degrading enzymes and/or phagocytosis of aggregated A␤. In contrast, activation by freshly sonicated A␤ results in secretion of neurotoxic factors. Therefore, identification of genes specifically involved in freshly sonicated 〈␤42-induced activation of BV2 cells may lead to a better understanding of neurodegeneration caused by inflammatory conditions in AD.
Identification of Genes That Are Up-regulated by Freshly Sonicated A␤42-To identify genes involved in BV2 activation by freshly sonicated 〈␤42, a large scale cDNA array expression profiling experiment was performed using BV2-specific arrays containing ϳ50,000 clones ( Fig. 2A). The cDNA clones were generated by amplifying the recombinant clone inserts derived from cDNA libraries enriched for transcripts from A␤-treated BV2 cells through normalization and subtraction. This procedure decreased redundancy and increased the relative abun-dance of rare transcripts. To account for biological variability, the arrays were hybridized with three individual probes from BV2 cells treated with 5 M freshly sonicated 〈␤42 for 24 h. In the array screen of 50,000 clones, 9248 clones up-regulated by 〈␤42 by a ratio of Ն1.2 in at least one of the A␤-treated BV2 cell samples were identified and sequenced. Of the 9248 clones sequenced, 6929 clones yielded annotable sequence, based on the base call quality. These 6929 clones were contiged to 2176 unique transcripts to estimate the redundancy of the up-regulated clone selection. From the 6929 annotable clones, we selected 2851 clones that were up-regulated by 〈␤42 by a ratio Ն 1.2 in at least two BV2 cell samples ( Fig. 2A). Following gene annotation of these 2851 clones, we identified 554 nonredundant genes whose expression was induced by more than 1.2fold. Clones lacking annotation to a gene in NCBI RefSeq were not analyzed further. The 554 up-regulated genes were mapped using GO molecular function terms (27). We found a high percentage of functional mappings to enzymes (40%), especially hydrolases (14%), such as cathepsin B and cathepsin L, and oxidoreductases (7%) among other enzymes. In addition, of all functional mappings for the 554 genes, 29% of them map to the binding factors, including protein-binding (5%), nucleotidebinding (9%), nucleic acid-binding (10%), metal ion-binding (2%), and other binding factors (3%) (Fig. 2B). Finally, 32 of the 554 genes mapped to the GO molecular function unknown category, representing 3% of all functional mappings and 5% of all genes. The NCBI RefSeq accession numbers for all 554 up-regulated genes, along with corresponding array data, has been deposited in the Gene Expression Omnibus hosted by NCBI (www.ncbi.nlm.nih.gov/geo/, series submission number GSE772 and platform accession numbers GPL561 and GPL562).
Expression Analysis of Cathepsins B and L, TIMP2, AIF1, and Cytochrome c-From the 554 genes, we selected a subgroup of genes from the two major functional mapping groups, namely enzymes and binding factors for further analysis. From the enzyme group, cathepsin B, cathepsin L, TIMP2, and cytochrome c oxidase were further analyzed for their roles in mediating inflammatory responses in microglia. AIF1, an IFN␥inducible Ca 2ϩ -binding EF-hand protein, was selected from the binding factor group. Cathepsin B and L are lysosomal proteinases that are found extracellularly at high levels in the senile plaques of AD brain, which are known to contain activated microglia (36). Moreover, the activation and differentiation of mononuclear phagocytes is known to be accompanied by increased cathepsin B/L enzymatic activities (37). TIMP2, a tissue inhibitor of matrix metalloproteinases, is one of the major endogenous counter-regulators of matrix metalloproteinases. Both matrix metalloproteinases and TIMPs are implicated in the pathogenesis of inflammatory disorders of the central nervous system (38). Mitochondrial cytochrome c oxidase is known to be involved in apoptosis induced by various stressful stimuli, including inflammation (39). AIF1 has been associated with microglial activation in experimental models and in human cerebral infarctions (40).
We first used quantitative RT-PCR to confirm the gene regulation in sonicated A␤-activated BV2 cells and compare gene regulation induced by sonicated and aged 〈␤42. In addition, regulation by LPS/IFN␥, strong inflammatory stimuli, was used as a positive control for inflammatory responses. Of the five selected genes, we confirmed that the mRNA levels of all of them, except that of TIMP2, were significantly up-regulated by freshly sonicated A␤42 in three or four independent experiments using quantitative RT-PCR (Table I, p Ͻ 0.05). Interestingly, three of the five selected genes, AIF1 and cathepsins B and L, were up-regulated by both LPS/IFN␥ and freshly soni- cated 〈␤42, suggesting overlapping pathways shared by LPS/ IFN␥ and freshly sonicated A␤42 stimulation. However, none of the three genes were up-regulated by aged 〈␤42, supporting our previous finding that stimulation of BV2 cells by freshly sonicated 〈␤42 and aged 〈␤42 involves divergent pathways (Fig. 1). Although in the array analyses, mRNA of TIMP2 was shown to be up-regulated by freshly sonicated 〈␤42 in a statistically significant manner (Average ratio ϭ 1.35 Ϯ 0.09, p Ͻ 0.01), we were not able to confirm the finding in three independent quantitative RT-PCR analyses. This discrepancy is likely due to nonspecific hybridization with other isoforms of TIMPs in the large scale array analysis because large cDNA fragments were used as probes. It is therefore very important to confirm the results from large scale profiling studies in more detailed studies, such as quantitative RT-PCR using isoformspecific primers.
Knock-down Analysis of TIMP2, and AIF1-To identify whether any of the five selected genes play a key role in mediating the neurotoxic effects of A␤-activated microglia, we used a strategy outlined in Fig. 3A. Expression of specific genes in BV2 cells was inhibited using siRNA (41) by transient transfection, followed by activation with freshly sonicated 〈␤42. The rat primary cortical neurons were then exposed to the conditioned medium from activated BV2 cells transfected with siR-NAs. We reasoned that specific inhibition of a key mediator in BV2 cells would abolish the neuron-killing effects of the toxic factors released by BV2 cells.
Gene-specific siRNAs induced highly efficient knock-down of mRNA levels for both TIMP2 and AIF1 (Fig. 3B). The efficiency of knock-down was similar in both control cells and cells treated 〈␤42 (data not shown). However, the inhibition of TIMP2 and AIF1 expression did not affect the toxicity exerted

FIG. 2. Schematic view of gene discovery and expression profile analysis and categorization of genes up-regulated by A␤42 in microglial BV2 cells.
A, subtraction and normalization of RNA derived from A␤42-activated and nontreated BV2 cells was conducted to enrich for the most relevant transcripts and to generate BV2 specific cDNA libraries. 50,000 clones were hybridized with probes from three samples of A␤-activated BV2 cells and three control probes from nontreated BV2 cells. Confirmed candidate clones were sequenced and annotated. B, representation of GO mappings for 554 up-regulated genes. The GO analysis was conducted based on molecular function terms that were assigned and approved either by a GO curator or from electronic annotation. The "unknown" category represents genes without any molecular function terms currently mapped in GO. The "others" category represents GO molecular function mappings representative of less than 1% of the total mappings. The RefSeq accession numbers for all 554 up-regulated genes, along with corresponding array data, are deposited in two parts in the Gene Expression Omnibus hosted by NCBI (www.ncbi.nlm.nih.gov/geo/, series submission number GSE772 and platform accession numbers GPL 561 and GPL 562). by the conditioned medium of A␤42-activated BV2 cells, suggesting that these proteins do not play a crucial role in this pathological process (Fig. 3C).
Cathepsin B, Not Cathepsin L, Mediates the Neurotoxic Effects of Activated Microglia-We next specifically inhibited the expression of cathepsin B and L by siRNAs (Fig. 4A) and compared their functional roles in A␤-activated BV2 cells (Fig.  4B). Inhibition of cathepsin B expression in BV2 cells completely abolished the neurotoxic effects exerted by conditioned medium from A␤42-activated BV2 cells. However, inhibition of cathepsin L expression in BV2 cells did not affect the neurotoxic effects mediated by A␤42-activated BV2 cells, indicating that cathepsin B, not cathepsin L, plays a crucial role in this process (Fig. 4B).
We then addressed the question of whether intracellular or extracellular cathepsin B is essential in mediating neurotoxicity in primary cortical neurons. CA-074, a nonpermeable cathepsin B inhibitor, was used to selectively inhibit extracellular cathepsin B activity (42). Indeed, inhibition of the enzymatic activity of cathepsin B in the conditioned medium of A␤42-activated BV2 cells abolished its neurotoxic effects on primary neurons (Fig. 4C). Significant rescue effects were observed at both the 0.5 and 1 M concentrations (Fig. 4C). Our data suggest that cathepsin B, especially extracellular cathepsin B, is probably a crucial factor released by A␤-activated microglial cells to mediate the neurotoxicity in cortical neurons or BV2 cells. Interestingly, CA-074 did not block the neurotoxicity mediated by LPS/IFN␥-activated BV2 cells (data not shown), suggesting that cathepsin B may not be a crucial factor in LPS/IFN␥-mediated microglial activation. Our studies using siRNA-mediated gene silencing and a highly specific cathepsin B inhibitor strongly indicate that cathepsin B plays an essential role in neurtoxicity inflicted by A␤-activated microglial cells. Additionally, these results highlight the importance of cathepsin B in inflammation-mediated neurodegeneration in AD brain. DISCUSSION In the present study, we used a functional genomics approach to identify crucial molecular mediators of the neurotoxic effects caused by A␤42-activated microglial BV2 cells. Our approach integrated large scale expression profiling with siRNA-mediated gene-specific silencing to identify cathepsin B as one of the key players in this process. Its crucial role was further confirmed using a highly specific inhibitor, CA-074, which completely abolished the neurotoxic effects mediated by A␤42-activated microglial BV2 cells.
One important aspect of the pathophysiology of AD is its chronic inflammation conditions associated with microglial activation, which is known to play multi-functional roles in the neurodegeneration process (43). Proinflammatory factors and acute phase proteins released by chronically activated microglia are known to play important roles in promoting neurodegeneration in AD brain (3,4). However, there is also mounting evidence demonstrating beneficial roles of microglial activation and inflammatory responses in plaque clearance (44 -46) and neurodegeneration (47).
Interestingly, we found that activation of BV2 cells by different forms of 〈␤42 resulted in completely different neurotoxic effects in primary cortical neurons (Fig. 1). Freshly sonicated 〈␤42 did not cause neuronal death when added directly to neurons, but activated BV2 cells to release toxic factors that caused significant neuronal death (Fig. 1). On the contrary, aged A␤42 elicited significant direct neuronal toxicity, but its toxic effect on neurons was abolished after incubating with BV2 cells. Our current observation that aged A␤42 peptide does not induce microglial-mediated neurotoxicity is different from some previous studies in which aged fibrillar A␤42 was found to be effective. This difference could be explained by the fact that those studies used much higher concentrations (e.g. 60 M) of aged fibrillar A␤ or a combination of aged A␤42 and IFN␥ to stimulate microglial cells to release toxic factors (16 -18, 29). More importantly, we found that a relatively low concentration (11 M) of freshly sonicated A␤42 alone can induce microglia-mediated neurotoxicity. Although TNF-␣ and nitric oxide appear to be intimately involved in microglial-mediated neurotoxicity with a high dose of fibrillar A␤40 (16,17), no TNF-␣ or nitric oxide was detected in the conditioned medium from BV2 cells activated by freshly sonicated A␤42, 2 indicating different pathways involved. More detailed studies will be needed to understand the roles of different species of 〈␤42 peptides in microglial activation as well as their pathophysiological relevance in the AD brain.
To understand the molecular pathways activated by freshly sonicated A␤42, we conducted a large scale expression profiling analysis using filter-based cDNA arrays made from BV2 cDNA libraries enriched for 〈␤42-activated microglial genes. The distinct advantage of this approach over presynthesized oligonucleotide arrays is that the clone collections are cell type-and disease-specific and are enriched with transcripts that are induced in specific disease conditions. Using this approach, we identified a total of 554 genes that are up-regulated by freshly sonicated A␤42, of which 40% are enzymes. ⌻he unusual enrichment of enzymes, especially proteases, suggests that proteases activated in the chronic inflammatory conditions might play important roles in neuronal degeneration in AD (48).
To identify crucial players in A␤-mediated inflammatory response for therapeutic intervention, we used siRNA-mediated gene knock-out to establish a causative functional link between candidate genes with microglial-mediated neurotoxicity. Indeed, transfection of siRNAs corresponding to the candidate genes in BV2 cells resulted in 80 -90% inhibition of expression of all four selected genes, indicating that siRNA-mediated gene silencing is a very powerful tool in analyzing gene function and identifying therapeutic targets in microglial BV2 cells.
Our study found that inhibition of cathepsin B expression using siRNA completely abolished the neurotoxicity mediated by A␤-activated BV2 microglial cells, suggesting that cathepsin 2 S. Yi and L. Gan, unpublished observations. shown to protect against ischemia-induced neuronal damage (55), which is at least in part due to inflammatory and immunological reactions (56,57).
In the present study, CA-074, an irreversible cathepsin Bspecific inhibitor with very low membrane permeability, completely abolished the toxicity in the conditioned medium released by activated BV2 cells. Our study indicates that cathepsin B, most likely extracellular cathepsin B released from A␤42-activated BV2 cells, is a key mediator in causing neuronal death in primary neurons. To the best of our knowledge, our study is the first to establish the functional link of cathepsin B with A␤-mediated microglia activation and highlight the potentially crucial role of cathepsin B in AD pathogenesis.
In summary, we identified cathepsin B as a key player in microglial-mediated neuronal death using a strategy that inte- FIG. 3. Lack of TIMP2 and AIF1 function in mediating the toxic effects caused by the conditioned medium of A␤-activated BV2 cells. A, schematic view of the functional assay to identify whether a target microglial gene plays a causative role in mediating neurotoxicity. Specific inhibition of gene functions in BV2 cells was achieved by transient transfection of gene-specific siRNAs followed by activation with A␤42. For cathepsin B, a specific inhibitor, CA-074, was used to pretreat the BV2 culture to inhibit the cathepsin B activity. The supernatants were applied to the primary cortical neurons for 72 h to induce cytotoxicity, which was quantified using a CellTiter-Glo Luminescent cell Viability Assay. Quantitative RT-PCR was used in parallel to quantify siRNA-induced gene silencing. B, expression of TIMP2 (n ϭ 8) and AIF1 (n ϭ 8) was strongly inhibited by siRNAs with corresponding sequences, but not by siRNA with a scrambled sequence (si-Control). The graph represents the means Ϯ S.E. from duplicate wells in four independent experiments. C, inhibition of AIF1 and TIMP2 expression in BV2 cells did not abolish the neurotoxicity caused by the supernatant from activated BV2 cells. Significant neuronal toxicity was induced by conditioned medium from A␤-activated BV2 cells transfected with control siRNA (si-Control, p Ͻ 0.001), siRNA corresponding to AIF1 (si-AIF1, p Ͻ 0.001), and TIMP2 (si-TIMP2, p Ͻ 0.001). Neuronal viability was quantified as described for A and expressed as relative luminescence units (RLU). The graph represents the means Ϯ S.E. from quadruplicate wells in two independent experiments (n ϭ 8). grated large scale expression profiling with custom cDNA arrays and siRNA-based functional analysis of selected gene products. Our approach offers an efficient strategy for rapid identification of underlying molecular pathways leading to neurodegeneration associated with chronic inflammation in the AD brain. Subsequent pharmaceutical inhibition of the genes crucial in this process may delay or prevent neurodegeneration in this devastating disease. FIG. 4. Cathepsin B plays a key function in mediating the toxic effects caused by conditioned medium from A␤42-activated BV2 cells. A, expression of cathepsin B (n ϭ 8) and cathepsin L (n ϭ 8) was strongly inhibited by siRNAs with corresponding sequences but not by siRNA with a scrambled sequence (si-Control). The graph represents the means Ϯ S.E. from duplicate wells in four independent experiments. B, supernatants from BV2 cells transfected siRNAs were applied to primary cortical neurons for 72 h. Neuronal viability was quantified using CellTiter-Glo Luminescent cell viability assay and expressed as relative luminescent signal units (RLU). Significant neuronal toxicity was induced by conditioned medium from A␤-activated BV2 cells transfected with control siRNA (p Ͻ 0.05) and siRNA corresponding to cathepsin L (si-CatL, p Ͻ 0.05). Transfection of siRNA corresponding to cathepsin B (si-CatB) abolished the BV2-mediated neurotoxicity (p ϭ 0.29). The graph represents the means Ϯ S.E. in triplicate wells. Similar results were obtained in three independent experiments with different A␤ preparations. C, inhibition of cathepsin B activity leads to increased neuronal viability after treatment with A␤-activated BV2 supernatants. BV2 cells were pretreated with 0.5 or 1 M CA-074 for 30 min followed by treatment with 11 M freshly sonicated A␤42 for 15-17 h in the presence of CA-074. Neuronal viability was quantified 72 h as described. The graph represents the means Ϯ S.E. from quadruplicate wells. Similar results were obtained in three independent experiments. Significant microglial-dependent neuronal toxicity was observed in control ([em], p Ͻ 0.05) but not in cells treated with 0.5 M (p ϭ 0.09) or 1 M (p ϭ 0.55) of CA-074.