Thinking outside the box about COX-1 in Alzheimer's disease

Abstract

This article from Coma et al. shows that a salicylic acid derivative Triflusal, a platelet aggregation inhibitor and irreversible inhibitor of COX-1, can correct defects in axonal curvature and cognition in an AD transgenic mouse model (Tg2576) (Coma et al., 2010). Here we discuss the controversy over the role of COX-1 in AD, which has not been considered carefully in part due to the presumed adverse gastrointestinal effects of COX-1 antagonism. However, recent clinical data from this group as well as other groups challenges this assumption that COX-1 antagonism will be associated with side effects. Most importantly this article raises critical questions about the role of COX-1, versus COX-2 versus both in Aβ pathogenesis. The animal model data in this article as well as the recently published trial data suggest that COX-1 may play an important role in early pathogenesis and should not be ignored as a potential target for early intervention.

The salicylic acid derivative Triflusal, a platelet aggregation inhibitor and irreversible inhibitor of COX-1, was effective in reducing cognitive deficits and indices of neurodegeneration. As with other NSAID effects observed in animal models, longer duration of Triflusal treatment appeared more effective, exerting a greater impact on reducing dense cored-plaque load and associated, glial cell proliferation, pro-inflammatory cytokine levels and abnormal axonal curvature, and the better the correction of cognitive deficits. The small Triflusal effects on amyloid burden (reducing dense cored plaques but not total burden) was somewhat surprising, considering that mixed COX inhibitors, including ibuprofen, have been reported to have a much more pronounced effect on reducing amyloid burden in AD models, provided intervention was pre-pathology and long duration (Choi et al., 2009; Heneka et al., 2005; Jantzen et al., 2002; Lim et al., 2000, 2001; McKee et al., 2008; Morihara et al., 2005; Weggen et al., 2001; Yan et al., 2003). Indomethacin, which like Triflusal, has more specificity for COX-1, exerted a similarly weak (but significant) anti-amyloid effect reducing amyloid burden by 20%. In contrast, NCX-2216, nitro-S-flurbiprofen (also preferential COX-1) had an intermediate effect on reducing amyloid burden (Jantzen et al., 2002). This may indicate that some of the anti-Aβ effects of mixed COX inhibitors like ibuprofen, indomethacin and flurbiprofen also benefit from pleiotropic actions including γ-secretase modulation (Weggen et al., 2001), anti-aggregation (Agdeppa et al., 2002) and share significant COX-related mechanisms involving cytokine modulation (e.g. phospholipase A2, inducible nitric oxide synthase, interleukin-1, and α-1 anti-chymotrypsin) (Morihara et al., 2005).

Trials in MCI or AD using a drug selective for COX-2 such as rofecoxib clearly failed (Aisen et al., 2008). Ultimately, COX-2 inhibitors have not worked in clinical trials after onset of cognitive deficits, for example, the COX-2 inhibitor rofecoxib was even associated with increased AD in a trial with MCI patients (Aisen et al., 2008). Similarly, a COX-2 inhibitor did not work for primary prevention in the ADAPT which was enriched with subjects with first degree relatives of AD patients (Laino, 2009). The anti-inflammatory steroid agent prednisolone failed (Aisen et al., 2000),which may relate to a known effect of glucocorticoids in inhibiting COX-2 in astrocytes (O'Banion et al., 1996).

Non selective COX inhibitors such as indomethacin (Rogers et al., 1993) and diclofenac (Scharf et al., 1999) showed preferential inhibition of COX-1 and possible positive clinical findings in AD trials. These “some what positive” trials were underpowered and full of safety issues including low tolerability. But with the positive results and good safety trial with Triflusal (Gomez-Isla et al., 2008), does that mean we should dismiss COX-2 as a target for prevention? Well the answer to that question requires a better understanding of stage-dependent pathogenesis. For example, COX-2 is elevated early in AD brain (Hoozemans et al., 2008). In AD models, COX-2 inhibitors can correct early deficits in LTP and cognition (Kotilinek et al., 2008) and early intervention trials (preMCI) with COX-2 specific inhibitor celecoxib showed a cognitive benefit with mild memory complaints and glucose utilization measured by FDG PET (Small et al., 2008). COX-1 and COX-2 are hypothesized to play different roles at different stages in AD (Hoozemans et al., 2008), which may explain why COX-2 antagonists can help if intervention is preMCI, while COX-1 antagonism may be somewhat effective up to MCI, but still probably not a viable target for AD. COX-2 has been associated with excitotoxicity but now it is known that COX-1 can also play a role in excitotoxicity (Pepicelli et al., 2005) that may have been inhibited by Triflusal in this Coma et al. report where it was associated with restoration of BDNF. Because neuronal COX-2 plays a critical role in memory retention (Rall et al., 2003), intervention with COX-2 antagonists at later stages could be detrimental after onset of neuron loss. In addition, some COX-2 inhibitors even increase amyloid burden (Jantzen et al., 2002; Xiang et al., 2002) or production (Kukar et al., 2007). Another reason not to dismiss COX-2 entirely as a target for prevention is that mixed COX inhibitors like ibuprofen, which has a better safety profile than naproxen, generally have a more robust impact on Aβ pathogenesis in AD models (Choi et al., 2009; Heneka et al., 2005; Jantzen et al., 2002; Lim et al., 2000, 2001; McKee et al., 2008; Morihara et al., 2005; Weggen et al., 2001; Yan et al., 2003) than Triflusal had in this AD model study.

The new data presented in this manuscript showing efficacy of Triflusal in an AD model (Coma et al., 2010) along with some positive results and good safety profile from the recent MCI trial with Triflusal (Gomez-Isla et al., 2008) renews discussions about COX-1 and GI side effects. Although COX-1 inhibitors are generally considered to be associated with adverse side effects, it should be mentioned that this is not necessarily the case for Triflusal (Gomez-Isla et al., 2008). Surprisingly, there were no serious adverse gastrointestinal effects with Triflusal, in contrast to serious intestinal bleeding with subjects on aspirin, diclofenac or indomethacin (Gomez-Isla et al., 2008). The dogma that COX-1 inhibition inevitably causes gastrointestinal toxicity has recently been questioned by reports that newer selective COX-1 inhibitors do not necessarily have significant GI side effects (Tanaka et al., 2009). COX-2 specific inhibitors also have issues, most notably increasing cardiovascular risk (Rao and Knaus, 2008). COX-2 inhibitors cause selective suppression of vasodilatory and antithrombotic prostracyclin (PGI2) in vascular endothelial cells, so suppression of COX-2 can increase thromboxane (TxA2) levels and thus platelet aggregation (Jespersen et al., 2009).

Interestingly, like indomethacin (Rogers et al., 1993) and diclofenac (Scharf et al., 1999) naproxen is a mixed (COX-1 and COX-2) inhibitor that is more preferential for COX-1 than COX-2. In the large ADAPT trial, naproxen appeared to prevent conversion to AD, improved CSF tau and Aβ42 biomarkers and reduced conversion rates, but only in subjects with no cognitive defects prior to onset of study (Laino, 2009). Whether the early halting of the ADAPT trial was necessary for naproxen's prevention of conversion to MCI (median 1.5 years duration on naproxen, 2 years off) remains a major question. Naproxen appeared to have more of an adverse cognitive effect on subjects with baseline deficits (Laino, 2009) possibly because of a stage-dependent alteration in the cytokines or microglial function (Hoozemans et al., 2008; Streit et al., 2009; Walker et al., 2009), raising serious questions about the potential detrimental effects of late stage intervention. Adverse effects observed at late intervention (post-MCI) with mixed COX inhibitors may also be independent of inflammation, for example by interfering with neuronal COX-2 required for normal memory retention (Rall et al., 2003). Alternatively, COX-1 and/or COX-2 may not be the ideal inflammatory target, since highly specific COX-1 or COX-2 inhibitors may upregulate arachidonic acid, a substrate for lipoxygenase pathway, also implicated in AD pathogenesis (Yao et al., 2005; Ikonomovic et al., 2008).

Although lagging effects of drug efficacy (such as seen in the ADAPT) would be missed unless trials could be followed up to 4 years or later, collectively results from AD trials support the hypothesis that targeting early steps of pathogenesis are unlikely to be helpful after onset of cognitive decline. For example, prior to cognitive decline there is a pro-dromal period of several years duration where plaque and tangle buildup occur and are followed by an increase in CSF tau/ Aβ42 ratios. Immediately preceding memory decline, there is pronounced brain atrophy and dramatic post-synaptic protein loss (Jack et al., 2010).

Unfortunately the role of neuroinflammation in Alzheimer's disease and mild cognitive impairment is still “a field in its infancy”, also a title of McGeer's recently published article (McGeer and McGeer, 2010). This is exemplified by the fact that ten years ago several experts in neuroinflammation convened to write a special issue on relevance of neuroinflammation to Alzheimer's, including prospects of the amyloid vaccine and other inflammatory mechanisms like complement activation (Akiyama et al., 2000). The conclusions of this article were that inflammation in AD is complex and has beneficial as well as detrimental effects (Akiyama et al., 2000). Ten years later the field remains full of controversy in part because of the frustration with trials having negative or unspectacular findings (both vaccine and NSAIDs). This paper (Coma et al., 2010) is likely to reignite a more focused discussion to the role of COX-1 at or before MCI and for more research on stage specific effects of inflammatory pathways and drug targets. This study also reinforces the need for animal models in developing and planning trials. Nevertheless epidemiology with aspirin, which may have similar mechanisms to Triflusal, showed weak or negative effects in prevention, and the fact that mixed COX inhibitors may have more robust effects on amyloid than Triflusal raises concerns about whether mixed COX inhibitors may work better to prevent AD.

In summary, although it is conceivable that a target other than COX-1 could play a role in Triflusal mechanism, this paper along with previous literature supports the hypothesis that COX-1 may be playing a crucial role in cognitive loss and neuroinflammation in the AD model. These strong animal model data suggest that COX-1 might be a viable target in Alzheimer's for prevention or even at MCI, but not play a major role in Aβ pathogenesis. In addition the lack of adverse side effects with Triflusal in the previous positive clinical trial results as well as in the animal models is encouraging.

Abbreviations

AD

Alzheimer's disease

ADAPT

Alzheimer's Disease Anti-inflammatory Prevention Trial

ApoE

Apolipoprotein E

Aβ

β-amyloid peptide

COX

Cyclooxygenase

MCI

Mild cognitive impairment

NSAIDs

Nonsteroidal anti-inflammatory drugs