Cyclooxygenase (COX)-2 and COX-1 Potentiate β-Amyloid Peptide Generation through Mechanisms That Involve γ-Secretase Activity*

In previous studies we found that overexpression of the inducible form of cyclooxygenase, COX-2, in the brain exacerbated β-amyloid (Aβ) neuropathology in a transgenic mouse model of Alzheimer's disease. To explore the mechanism through which COX may influence Aβ amyloidosis, we used an adenoviral gene transfer system to study the effects of human (h)COX-1 and hCOX-2 isoform expression on Aβ peptide generation. We found that expression of hCOXs in human amyloid precursor protein (APP)-overexpressing (Chinese hamster ovary (CHO)-APPswe) cells or human neuroglioma (H4-APP751) cells resulting in 10-25 nm prostaglandin (PG)-E2 concentration in the conditioned medium coincided with an ∼1.8-fold elevation of Aβ-(1-40) and Aβ-(1-42) peptide generation and an ∼1.8-fold induction of the C-terminal fragment (CTF)-γ cleavage product of the APP, an index of γ-secretase activity. Treatment of APP-overexpressing cells with the non-selective COX inhibitor ibuprofen (1 μm, 48 h) or with the specific γ-secretase inhibitor L-685,458 significantly attenuated hCOX-1- and hCOX-2-mediated induction of Aβ peptide generation and CTF-γ cleavage product formation. Based on this evidence, we next tested the hypothesis that COX expression might promote Aβ peptide generation via a PG-E2-mediated mechanism. We found that exposure of CHO-APPswe or human embryonic kidney (HEK-APPswe) cells to PG-E2 (11-deoxy-PG-E2) at a concentration (10 nm) within the range of PG-E2 found in hCOX-expressing cells similarly promoted (∼1.8-fold) the generation of the CTF-γ cleavage product of APP and commensurate Aβ-(1-40) and Aβ-(1-42) peptide elevation. The study suggests that expression of COXs may influence Aβ peptide generation through mechanisms that involve PG-E2-mediated potentiation of γ-secretase activity, further supporting a role for COX-2 and COX-1 in Alzheimer's disease neuropathology.

A large number of epidemiological studies have indicated that the use of non-steroidal anti-inflammatory drugs (NSAIDs) 1 may prevent or delay the clinical features of Alzhei-mer's disease (AD) (1)(2)(3)(4)(5)(6). However, recent therapeutic studies with both NSAIDs (7)(8)(9) and steroids (10) have been unable to confirm this epidemiological evidence. The pharmacological activity of NSAIDs is generally attributed to the inhibition of COXs, which are rate-limiting enzymes necessary for the production of prostaglandins (PGs). Both COX-1 and COX-2, the constitutive and inducible forms of COX, respectively (11)(12)(13), are known to be involved in inflammatory responses and normal neuronal functions (14 -16).
We (17)(18)(19) and others (15, 20 -24) have shown that COX-2 expression in the brain and PG-E 2 content in the cerebrospinal fluid (24) are elevated in AD and further that COX-2 protein content in the brain correlates with the severity of amyloidosis and clinical dementia (19). Moreover there is evidence that COX-1 expression is also elevated in the AD brain, raising the possibility that both COX-1 and COX-2 may contribute to AD neuropathology (21,24,25). Thus, the characterization of COX activities and subsequent PG generation in the brain as well as their potential roles in amyloidosis is receiving a great deal of attention.
Further studies implicating COX in neuronal dysfunction in vivo include work by Andreasson et al. (26), demonstrating that COX-2-overexpressing transgenic mice developed memory dysfunction, neuronal apoptosis, and astrocytic activation in an age-dependent manner. Moreover a recent study has shown that overexpression of human (h)COX-2 in neurons of PSAPP transgenic mice (a transgenic mouse model of AD expressing both mutant amyloid precursor protein (APP swe ) and mutant presenilin-1 (A246E-PS1)) significantly potentiated amyloidogenic A␤ peptide generation and amyloid plaque deposition in the brain (27), indicating that conditions of elevated COX expression can promote neuronal dysfunction and amyloidosis in vivo.
In this study, we continued to explore the role of COXs in AD amyloidogenesis by testing the hypothesis that COXs may directly influence A␤ peptide generation in vitro. We found that a mechanism through which COX may promote amyloidogenic generation of A␤ peptides might involve PG-E 2 -mediated promotion of ␥-secretase activity. Understanding the apparent mechanistic relationship of COXs and A␤ generation is highly relevant to the successful development of COX inhibitors and other NSAID-based therapeutic strategies for AD.

EXPERIMENTAL PROCEDURES
Generation of hCOX-1 or hCOX-2 Adenoviruses-hCOX-1 and hCOX-2 cDNA constructs (described previously by our laboratories (28)) were introduced to the Adeno-X TM genome for generation of recombinant Adeno-X virus according to the Adeno-X expression system manual (Clontech). In brief, the full-length hCOX-1 cDNA was subcloned into the pShuttle vector cassette via MluI and ApaI, and the full-length hCOX-2 cDNA used for generation of hCOX-2 transgenic mice (28) was cloned into the pShuttle vector cassette via ApaI and XbaI. Both pShuttle/hCOX-1 and pShuttle/hCOX-2 were then transferred into the Adeno-X viral DNA via I-CeuI and PI-CseI sites; the identity of the hCOX-1 or hCOX-2 Adeno-X viral DNA was confirmed by nucleotide sequencing (not shown). The recombinant viruses were then packaged by transfecting PacI-linearized recombinant viral DNA into human embryonic kidney (HEK) 293 cells with the aid of LipofectAMINE (Invitrogen). HCOX-1 or hCOX-2 Adeno-X viral titer was determined by the tissue culture infectious dose 50 (TCID 50 ) method (29). This identical strategy was used to generate recombinant LacZ adenovirus (Clontech) expressing the bacterial ␤-galactosidase gene (Clontech), which served as a negative control.
Ibuprofen and 11-deoxyprostaglandin E 2 (Cayman Chemical, Ann Harbor, MI) and L-685,458 (a gift from Merck Sharp and Dohme Research Laboratories, Terlings Park, Harlow, UK) were stored at Ϫ20°C (in Me 2 SO). Disposable aliquots of Me 2 SO (final concentration, 0.01%) were also stored at Ϫ20°C to mimic freeze-thaw conditions in vehicletreated cultures. All cultures and reagents were demonstrated to be free of endotoxin (Ͻ10 pg/ml) by Limulus lysate assay (Sigma) (not shown).
hCOX-1, hCOX-2, and APP Immunodetection-Following adenoviral infection and/or incubation with drugs for 48 h, conditioned media were collected, and tissue cultures were lysed in RIPA buffer (1ϫ phosphatebuffered saline, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, and 0.1 mM EDTA) in the presence of a protease inhibitor mixture (Sigma) on ice and stored at Ϫ20°C. For immunoblot analysis, protein content was determined by the Bradford method (Bio-Rad), samples were boiled and centrifuged, and proteins were resolved electrophoretically by SDS-PAGE (10%). Proteins were transferred to nylon Transblot membranes (Bio-Rad) and immunoreacted with appropriate antibody. In these studies immunoreactivities were visualized by fluorescence autoradiography using enhanced chemiluminescence detection (Super-Signal chemiluminescent detection kit, Pierce).
Polyclonal C8 antibody (raised against amino acids 676 -695 of human APP, a gift from Dr. Selkoe) was used for detection of total human holo-APP. Monoclonal 22C11 antibody (Chemicon International, Temecula, CA) recognizing amino acids 60 -100 of an N-terminal epitope of human APP was used to quantify total soluble (s)APP released into the conditioned medium of CHO-APP swe cells. Monoclonal 6E10 antibody (Senetek, St. Louis, MO) recognizing amino acids 1-17 of the A␤ domain of APP (a site that constitutes the C terminus of sAPP peptide) was used to quantify the level of sAPP␣ released into the conditioned medium.
For detection of hCOX-2 (or hCOX-1) expression in transfected CHO-APP swe cells, specific antibodies raised against a synthetic peptide derived from the carboxyl region of hCOX-2 or hCOX-1 sequence were used (Cayman Chemical). Specificities of hCOX-2 and hCOX-1 antibodies were reported previously by our laboratories (28), and ␤-actin immunoreactivity (anti-␤-actin, Sigma) was used to control for variations in gel loading.
PG-E 2 Assay-PG-E 2 content was assessed in the same conditioned medium used for A␤ peptide determinations using a commercially available ELISA (Cayman Chemical) as described previously (34). In brief, conditioned medium was applied to 96-well plates precoated with goat anti-mouse IgG and incubated (18 h at 4°C) with PG-E 2 monoclonal antibody and a recovery tracer. After incubation with PG-E 2 monoclonal antibody, plates were rinsed fives times with washing buffer and developed (1 h at room temperature) using Ellman's reagent. Specific PG-E 2 concentration was determined spectrophotometrically and calculated by plotting (percentage of sample or standard bound/ maximum bound) the protein standard versus PG-E 2 concentration in pg/ml.
Statistical Analysis-All values are expressed as means Ϯ S.E. Differences between means were analyzed using a two-tailed Student's t test. In all analyses, the null hypothesis was rejected at the 0.05 level. All statistical analyses were performed using the Prism Stat program (GraphPad Software, Inc., San Diego, CA).

RESULTS
Adenovirus-mediated hCOX-1 or hCOX-2 Delivery System-In control studies the dose-dependent adenoviral mediated expression of hCOX-1 or hCOX-2 at 10, 20, and 40 m.o.i. 48 h postinfection was confirmed in CHO-APP swe and H4-APP 751 cells by Western blot analysis ( Fig. 1, a and b and d and e, respectively). No apparent endogenous COX-1 or COX-2 immunoreactivity was found in LacZ adenovirusinfected CHO-APP swe or H4-APP 751 cells ( Fig. 1, a and b); although endogenous expression of COX(s) in LacZ adenovirus-infected cells could also be detected at longer exposure time (not shown).
The functional expression of hCOX-1 and hCOX-2 in CHO-APP swe (Fig. 1c) and H4-APP 751 (Fig. 1f) cells was monitored by PG-E 2 generation by ELISA. We found that hCOX-1 or hCOX-2 infection at 10 m.o.i. resulted in 10 -25 nM PG-E 2 in the conditioned medium, which is within the physiological concentration of PG-E 2 described previously in the human brain (35). Consequently a virus titer of 10 m.o.i. was selectively utilized throughout this study to assess the role of hCOXs in A␤ peptide generation. In control adenoviral LacZ infection studies (using ␤-galactosidase staining as an indicator of adenoviral infection) we found that 10 m.o.i. resulted in 90% efficiency of infection with no apparent cytotoxicity as assessed by MTT assay 48 h postinfection (data not shown). hCOX-1 or hCOX-2 Promotes CTF-␥ Generation in CHO-APP swe Cells That Is Prevented by COX Inhibition-Based on the evidence that both hCOX-1 and hCOX-2 expression promotes A␤-(1-40) and A␤-(1-42) generation, we next examined the influence of expression of hCOXs on APP processing by assessing the generation of CTF-␥ APP cleavage product (known index of ␥-secretase activity). For this study, fresh membranes from CHO-APP swe cells were isolated and then incubated at 37°C for 2 h to allow generation of CTFs. Unlike other conventional assays, this in vitro ␥-secretase assay detects enzymatic cleavage under physiologic conditions wherein CTF-␥ is generated by cleavage of membrane-bound APP.

hCOX-1 or hCOX-2 Infection of CHO-APP swe Cells Promotes
In this experiment, we found that, compared with the LacZ control group, both hCOX-1 and hCOX-2 adenovirus infection in CHO-APP swe cells leads to a functional Ͼ3-fold induction of PG-E 2 content in the conditioned medium (Fig. 3g). Elevation of PG-E 2 content in the conditioned medium coincided with potentiation of the ϳ6-kDa CTF-␥ generation (Fig. 3, a and c, lanes 3 and 4, respectively, and b and d, relative to steady-state content of holo-APP) from fresh membrane preparations 48 h postinfection, strongly suggesting hCOX-1 and hCOX-2 overexpression leads to an induction of ␥-secretase activity. Increased CTF-␥ signal further coincided with significant elevation of A␤-(1-40) and A␤-(1-42) content in the conditioned medium of these same cultures (Fig. 3, e and f, respectively). As expected, in control studies, no detectable CTF-␥ cleavage product was detected in membrane preparations from either hCOX-1- (Fig. 3a, lanes 1 and 2) or hCOX-2 (Fig. 3c, lanes 1 and  2)-transfected CHO-APP swe cells kept on ice during the entire period of the reaction time.
To further address the physiological role of hCOX-1-and hCOX-2-mediated promotion of A␤ peptide generation, we next examined the role of the non-selective COX inhibitor ibuprofen on generation of CTF-␥ and A␤-(1-40) and A␤-(1-42) peptides. We found that hCOX-1-and hCOX-2-mediated promotion of the ϳ6-kDa CTF-␥ generation was prevented by co-treatment of CHO-APP swe cells with the non-selective COX inhibitor ibuprofen (Fig. 3, a and c, lane 4 versus lane 6, and b and d) at 1 M, which coincided with the reduction of A␤-(1-40) and A␤-(1-42) peptides (Fig. 3, e and f, respectively) and hCOX-1-and hCOX-2-mediated promotion of PG-E 2 content (Fig. 3g) to LacZ levels as assessed by ELISA of the conditioned medium 48 h postinfection.
In control studies, we found that blockade of hCOX-1-and hCOX-2-mediated induction of A␤ generation by the selective ␥-secretase inhibitor L-685,458 at 20 -100 nM was highly specific and did not influence the activities of hCOXs as suggested by a lack of change in PG-E 2 content (Fig. 4c) in the conditioned medium 48 h after treatment. Moreover no detectable cell toxicity in response to L-685,458 at any concentration tested relative to vehicle (Me 2 SO, 0.01%) was found in LacZ, hCOX-1, or hCOX-2 adenovirus-infected CHO-APP swe cells as assessed by MTT assay (not shown).
hCOX-1 and hCOX-2 Expression Do Not Influence the Nonamyloidogenic sAPP␣ Pathway in CHO-APP swe Cells-To further confirm that hCOX-1 and hCOX-2 selectively promoted A␤ generation through mechanisms involving ␥-secretase activity, we next assessed the potential role of hCOX-1 and hCOX-2 in non-amyloidogenic pathways by measuring sAPP secreted in the conditioned media.
We found that sAPP␣ content (defined as 6E10-immunoreactive sAPP in the conditioned medium, see "Experimental Procedures" for more information), which is assumed to be an ␣-secretase-cleaved form of APP, was not affected by hCOX-1 or hCOX-2 expression in CHO-APP swe cells (Fig. 5a, inset) relative to the LacZ-infected control group. Similarly no detectable change in total sAPP content (defined as 22C11-immunoreactive sAPP) was found in the conditioned medium of hCOX-1and hCOX-2-infected CHO-APP swe cells (Fig. 5a, inset) relative to LacZ-infected cells. Finally the sAPP␣/total sAPP ratio, indicative of ␣-secretase activity (Fig. 5a), was not altered in response to hCOX-1 or hCOX-2 infection (identical cultures were used for detection of sAPP␣/total sAPP ratio, A␤ peptides, and CTF-␥ cleavage product generation discussed above). Collectively this evidence suggests that the non-amyloidogenic ␣-secretase activity is not altered by expression of hCOXs in vitro. In further control studies we also confirmed that total steady-state holo-APP content (defined as C8-immunoreactive APP) in the cell lysate of CHO-APP swe cells was not altered by hCOX-1 or hCOX-2 adenovirus infection ( Fig. 5b) relative to LacZ adenovirus-infected cells.

11-Deoxy-PG-E 2 Treatment Promotes A␤ Peptide Generation in CHO-APP swe Cells and HEK-APP swe
Cells-Based on the evidence that hCOX-1 and hCOX-2 can promote CTF-␥ and A␤ peptide generation in vitro, we tested whether PG-E 2 , the major product of the COX enzymatic pathway, could promote similar responses at a concentration comparable to that elicited by hCOX-1 or hCOX-2 viral expression.
Similar to responses observed following infection with hCOXs, we observed that exposure to the PG-E 2 analog 11deoxy-PG-E 2 (11-dPG-E 2 , 10 nM) significantly potentiated the generation of the ϳ6-kDa CTF-␥ (relative to holo-APP content) in membrane preparations from CHO-APP swe cells (Fig. 6, a  and b) 48 h posttreatment. Further 11-dPG-E 2 -mediated induction of CTF-␥ cleavage product in CHO-APP swe (Fig. 6c) cells coincided with an approximate ϳ1.8-fold elevation of A␤-  and A␤-(1-42) content in the conditioned medium of the same cultures. No detectable change in holo-APP was observed in both CHO-APP swe (Fig. 6a, quantification not shown) cells following 11-dPG-E 2 treatment.
Moreover we found that 11-dPG-E 2 -mediated induction of CTF-␥ generation and subsequent elevation of A␤-(1-40) and A␤-(1-42) peptides in CHO-APP swe cells was highly selective and occurred in the absence of detectable alterations to "nonamyloidogenic" APP pathways as indicated by a lack of change in the sAPP␣/total sAPP ratio in the conditioned medium or total cellular holo-APP contents relative to vehicle-treated CHO-APP swe cells (Fig. 6, d and e), respectively. Finally the 11-dPG-E 2 (10 nM)-mediated induction of CTF-␥ generation and subsequent elevation of A␤-(1-40) and A␤-(1-42) peptides were confirmed in HEK-APP swe cells in the absence of detectable changes in holo-APP and sAPP␣ levels (data not shown). DISCUSSION The accumulation and aggregation of A␤ peptides in the brain is believed to be an early event in the pathogenesis of AD (1). A␤ peptides are generated by the sequential proteolytic cleavage of APP by ␤and ␥-secretase, and consequently inhibitors of these secretases are under investigation as potential A␤-lowering strategies for AD. Recent studies have suggested that ibuprofen among other NSAIDs may selectively reduce A␤-(1-42) production in vitro via direct modulation of ␥-secretase activity (36 -38) and not via the COX-inhibiting features characteristic of this class of drugs. However, the evidence that the expression of both the inducible and constitutive forms of COX (15,(17)(18)(19)(20)(21)(22)(23)(24) as well as PG-E 2 (39) are elevated in AD raises the possibility that COXs may contribute to AD pathology and that COX inhibition may be therapeutically relevant. This evidence is further supported by our previous work showing that hCOX-2 expression in PSAPP mice induced potentiation of brain parenchymal amyloid plaque formation coincidental with a 2-fold increase in PG-E 2 production (40). The goal of this study was to further test the hypothesis that COX may directly influence amyloidogenesis and explore the mechanisms through which COX-1 and COX-2 may promote A␤ peptide generation in vitro.
We found that hCOX-1 or hCOX-2 infection in CHO-APP swe and H4-APP 751 cells resulting in 10 -25 nM PG-E 2 , which is within the range of PG-E 2 concentrations observed in the hu- No detectable changes in total holo-APP (105 kDa) or CTF-␣ and -␤ cleavage products (13-and 10-kDa cleavage products) were observed in either hCOX-1-or hCOX-2-transfected CHO-APP swe cells relative to LacZ controls. b and d, quantification of the CTF-␥ fragment of APP (an index of ␥-secretase activity) following hCOX-1 or hCOX-2 infection. In a and c, the induction of CTF-␥ (lane 4 versus lane 3) following hCOX-1 and hCOX-2 infection was prevented by co-treatment of CHO-APP swe cultures with the non-selective COX inhibitor ibuprofen at 1 M. e and f, ELISA-based quantification of A␤-(1-40) and A␤-(1-42), respectively. g, PG-E 2 content following hCOX-1 or hCOX-2 adenoviral infection. In a and c, lack of incubation of purified CHO-APP swe membrane preparations (kept on ice versus 37°C, 2 h) resulted in negative CTF-␥ product formation as expected (lanes 1 and 2). In b and d-f, results are expressed as a percentage of LacZ infection in the control group. In g, changes in ibuprofen-treated groups are expressed as percentage of each respective condition in the vehicle-treated group. In b-g, values represent means Ϯ S.E. of determinations made in three separate culture preparations; n ϭ 3 per culture. In b and d, *, p Ͻ 0.01 versus LacZ in the same group; in f-g, **, p Ͻ 0.01 versus each respective vehicle (ethanol, 0.01%) group. man brain (35), coincided with a significant elevation of A␤-(1-42) and A␤-(1-40) peptide generation and increased generation of the ␥-CTF product of APP. This evidence strongly supports the hypothesis that COX-1 and COX-2 may have promoted A␤ generation via a novel mechanism resulting in potentiation of ␥-secretase activity. Consistent with this hypothesis, we found that the selective ␥-secretase inhibitor L-685,458 prevented hCOX-1-and hCOX-2-mediated promotion of A␤ peptide generation without altering PG-E 2 production. This finding coupled with the evidence that hCOX-1 and hCOX-2 overexpression did not influence sAPP␣ secretion into the conditioned medium supported the hypothesis that COX may have promoted A␤ peptide generation via downstream activation of ␥-secretase activities rather than interfering (inhibiting) with non-amyloidogenic secretory APP pathways.
In light of the observed relationship between PG-E 2 production (a major product of the COX enzymes) and A␤ production following hCOX infection, we hypothesized that increased PG-E 2 synthesis (as found in cerebrospinal fluid of AD (39)) may have exacerbated A␤ peptide generation by specifically influencing ␥-secretase activity. Consistent with our observa-tions that hCOX-1 and hCOX-2 overexpression promoted A␤ generation, we found that treatment of CHO-APP swe cells and HEK-APP swe cells with PG-E 2 (11-dPG-E 2 ) at a concentration comparable to that elicited by hCOX-1 or hCOX-2 viral expression achieved similar potentiation of both A␤ peptide generation and ␥-secretase activity. This novel finding tentatively suggested that expression of COXs in the brain might influence amyloidogenesis through activation of signal transduction pathways downstream of the PG receptor. In view of the evidence that endogenous EP4 PG-E 2 receptors are positively coupled to adenylyl cyclase (41) in CHO cells, it may the case that the EP4 receptor represents a novel target for blocking A␤ generation in the brain. However, given that characterization of PGs and their receptor families in the brain remains in developmental stages (42), the therapeutic utility of PG receptor antagonists, at present, may be limited.
Previous evidence indicated that ibuprofen (a non-selective COX inhibitor) can decrease AD type amyloid burden in a mouse model (43,44). Based upon this evidence and the fact that ibuprofen can selectively lower A␤-(1-42) levels in the brain and in APP-overexpressing cells (36,37), we further evaluated the role of ibuprofen in blocking A␤ peptide generation and ␥-secretase activity mediated by COXs in vitro. We found that ibuprofen (at a low 1 M concentration) significantly prevented hCOX-2-and hCOX-1-mediated promotion of the generation of A␤ peptides and ␥-secretase activity in CHO-APP swe cells as reflected by inhibition of the ␥-CTF product of APP generation (and decreased PG-E 2 content). Given that we did not observe any changes in "basal-endogenous" A␤-(1-40) and A␤-(1-42) peptide generation in control CHO-APP swe cells treated with ibuprofen at this concentration (consistent with previous evidence (36,37)), we suggest that at a low 1 M dose ibuprofen might modulate generation of A␤ peptides by inhibiting responses mediated by COXs/PG-E 2 . However, it remains indisputable that ibuprofen (among other NSAIDs) at doses higher than that used in our study can directly influence ␥-secretase activity and eventually selectively lessen A␤-(1-42) generation (36,37). However, we also note that recent in vivo evidence indicates that ibuprofen may significantly decrease both A␤-(1-40) (Ϫ19%) and A␤-(1-42) (Ϫ63%) content in the brain of a mouse model of AD (45), consistent with the possi-bility that ibuprofen may nonspecifically inhibit generation of both A␤ peptides under certain conditions in vivo. Finally we note that other studies also suggest an additional potential antiamyloidogenic role for ibuprofen via mechanisms that impact APP processing and favor sAPP-␣ formation in neuroblastoma SH-SY5Y cells (46).
Despite the implications of the present findings and growing evidence from our (17)(18)(19)40) and other (15, 20 -25, 47) labo-   peptides in the conditioned medium of these same cultures 48 h following 11-deoxy-PG-E 2 treatment. No detectable change in total holo-APP (105 kDa) was observed following 11-deoxy-PG-E 2 treatment when compared with vehicle-treated cells. d, bar graph representing the ratio of sAPP␣ (detected by anti-APP 6E10 antibody) to total sAPP (detected by anti-APP 22C11 antibody) content in the conditioned medium. Inset in d, no detectable changes in total sAPP or sAPP␣ immunoreactivity (content) in the conditioned media was found 48 h following 11-deoxy-PG-E 2 treatment versus the vehicle-treated group. e, quantification of total holo-APP protein content in the cell lysate from 11-deoxy-PG-E 2 -treated CHO-APP swe cells (ratio anti-APP C8-immunoreactive APP/␤-actin). In b-e, results are expressed as a percentage of vehicle-treated group; values represent means Ϯ S.E. of determinations made in three separate cultures; n ϭ 3 per culture. *, p Ͻ 0.01 versus each respective vehicle (Me 2 SO, 0.01%)-treated group. ratories implicating COX-1 and COX-2 in the pathophysiology of AD and models of AD type neuropathology, the role of COX in the clinical progression of AD is little understood (48,49). Moreover it remains that clinical trials with NSAIDs in AD have had marginal success (7-9) when applied therapeutically in moderate-severe AD. However, given that multiple in vivo studies now show that COX-inhibiting NSAIDs can mitigate cognitive impairment and amyloidosis in mouse models of AD type neuropathology (prophylactically) (36,43,44) and that epidemiological studies continue to support a beneficial role for NSAIDs in AD (1,2), it may be the case that efficient treatment with NSAIDs in AD will yield success when applied to early or preclinical AD dementia cases.
It remains unclear which COX-inhibiting NSAID is the most appropriate candidate for the treatment of AD. Thus further understanding of the role of COX activity (specifically COXderived PG) in mechanisms leading to A␤ generation is critical to the future development of NSAID therapy for AD. As shown in Scheme 1, our finding showing that COXs may promote A␤ generation via a PG-E 2 -mediated pathway and the current evidence suggesting that certain NSAIDs may also directly influence ␥-secretase activities support the hypothesis that NSAIDs may bear therapeutic relevance to antiamyloidogenic strategies.