In efforts to interfere with Aβ peptide production, most of the attention has been lavished on the β- and γ-secretase enzymes. Could, however, their neglected cousin α-secretase prove to be the better target? In an article published January 10, 2005, in the journal Public Library of Science Medicine, Sam Gandy’s group of Thomas Jefferson University, Philadelphia, with colleagues elsewhere, report evidence suggesting that statin drugs can boost α-secretase cleavage of AβPP via the Rho/ROCK1 protein phosphorylation pathway. If confirmed, this data could offer new insight into how one might tip the scales away from β cleavage of AβPP. It could also help explain the mechanism underlying the apparent beneficial effect of statin drugs on Alzheimer disease.
Recent evidence has shown that upregulating α-secretase reduces brain levels of Aβ in APP transgenic mice (see ARF related news story). Moreover, "shedding" of the α-secretase-cleaved APP ectodomain appears to be stimulated by statin drugs (see Parvathy et al., 2004). In the current paper, first author Steve Pedrini and colleagues take aim at one of the candidate pathways for mediating this effect (for review of pathways to boosting α-secretase cleavage, see Allinson et al., 2003).
Statins reduce cholesterol levels, at least in part, by inhibiting the ability of the enzyme hydroxymethylglutaryl-coenzyme A (HMGCoA) reductase to promote the conversion of HMGCoA into the cholesterol-synthesis intermediate mevalonic acid. But in doing this, statins also reduce levels of other intermediates on the way to cholesterol, including isoprenoids such as farnesyl pyrophosphate and geranylgeranyl pyrophosphate. Pedrini and colleagues point out that these lipids play other regulatory roles. For example, they are critical in activating the Rho family of GTPases, which in turn activate Rho-associated coiled-coil containing kinases (ROCKs). By inhibiting HMGCoA, then, statins might ultimately reduce ROCK-mediated protein phosphorylation. One member of this family, ROCK1, has been implicated in the regulation of γ-secretase cleavage (Zhou et al., 2003; but see also γ-secretase news story from the 2004 Society for Neuroscience meeting). Pedrini and colleagues suggest that the isoprenoid-Rho-ROCK1 pathway might also be involved in the statin-induced shedding of soluble α-secretase-cleaved AβPP (sAβPPα).
Working in mouse neuroblastoma cells transfected with the gene for APP carrying the Swedish AD-causing mutation, the researchers find evidence consistent with this hypothesis. A dominant-negative form of ROCK1 increased sAβPPα, as did an inhibitor of the enzyme required for isoprenylation of Rho. Conversely, a constitutively active ROCK1 was able to abolish statin-induced sAβPPα shedding. Further implicating the isoprenoid portion of this pathway in statins' actions on α cleavage, the researchers abolished statin-induced shedding of sAβPPα by adding mevalonic acid, effectively bypassing the statin inhibition of HMGCoA reductase. On the other hand, inhibiting the isoprenylation of Rho mimicked the effects of statins on sAβPPα.
"Taken together, these results suggest the existence of a reciprocal relationship between isoprenoid-mediated Rho/ROCK signaling and sAPPa shedding, i.e., activation of ROCK1 blocks basal and stimulated shedding while ROCK1 inhibition apparently relieves a tonic negative influence exerted on shedding by ROCK1 activity," write the authors. Whether statins will therefore block ROCK1 activity in neurons is something the authors are now investigating.—Hakon Heimer.
Synopsis: How Statins May Protect Against Alzheimer Disease PLoS Med 2(1):e22.
Q&A with author Sam Gandy
Q: There has been mention of the Rho-ROCK pathway possibly boosting Aβ42 production by effects on γ-secretase. How does this relate to your results?
A:We haven't yet addressed it directly.
Many papers (e.g., Zhou et al., 2003) used Y27632 (nominally a ROCK inhibitor) to imply ROCK actions. Surprisingly, we found opposing results on statin-activated α-secretion when we used dominant-negative ROCK vs. Y27632. The predicted result was that they would show identical effects. Since our conditionally active ROCK had effects that were the opposite of those of dominant-negative ROCK, we chose to pursue the results from those molecules that were in "agreement" and, for now, set aside the results with the drug (which we discuss in the paper).
The Rho/ROCK pathway is also very state-dependent. In fact, there are examples of ROCK activators and ROCK inhibitors doing the exact same thing even in the same system, rather than having opposing effects. Presumably there is some moment-to-moment balance of which pathways prevail. Obviously, there is a lot we don't know about ROCK regulation.
There may be real conflicts in predicted results because different cells or different cellular states were not controlled for. There may also be conflicting data depending on whether ROCK's role is implicated by drug or by molecular biology.
So, the short answer is that we must look directly at Aβ, and we must look at neurons. We are doing this now. If it seems like a mess, it is. The results are too unpredictable to guess. We just have to do the experiments.
Q: What about the effects of isoprenoid pathway (pathways?) on γ- vs. α-secretase? Could it be affecting both? Different isoprenoids? Is all that still to be investigated?
A:Again, the short answer is yes, multiple pathways could be differentially regulated, and we just have to do the experiments.
α- vs. β-secretase "competition" controls levels of Aβ only, while γ-secretase can control either levels or 40/42 ratio. It's really probably just the absolute level of Aβ42 that is the most important. Standardizing to Aβ40 was devised in the early 1990s by Steve Younkin and Dennis Selkoe as a convenient way of comparing one dish to another. Aβ42 is so vanishingly low in amount, especially before highly sensitive ELISAs were developed, that the standardization was crucial for getting interpretable data. The Aβ42 signal and variations could easily get lost in the background noise were it not for this innovative standardization technique. Even so, Aβ42 is the real culprit (this is also the title of a highly cited review by Steve Younkin in Annals of Neurology).
Theoretically, if you hyperactivate α-secretase enough, you might get so little Aβ made that the ratio wouldn't matter. Hyperactivating α-secretase would be tantamount to BACE inhibition (which, using the small-molecule BACE1 approach, has been very difficult so far due to the large catalytic pocket in BACE). The trouble in developing BACE inhibitors has provided some impetus for revisiting the strategy of indirect BACE inhibition by hyperactivating α-secretase.
There is some disagreement over what ROCK does to γ-secretase: Some say it inhibits generation of 42 specifically (Zhou et al., 2003). But at least one group at the Alzheimer Congress in Philadelphia and/or at the SFN Meeting in San Diego reported that ROCK controlled both Aβ40 and Aβ42. I haven't heard an update on how that has sorted out.
Among known isoprenylated substrates, farnesylation is most common, but it may be that we just don't know all the geranylgeranyl substrates yet. Certainly different Rhos can differentially regulate the same reaction: I can't immediately think of an example where farnesylation causes "reaction A" to proceed in one direction and geranylgeranylation causes the reverse. These pathways are receiving a lot of attention in oncology research, where a farnesyl transferase inhibitor is in human clinical trials.
Q: What are the principle dietary sources of isoprenoids?
A: Squalene is the most readily identifiable dietary isoprenoid (on MedLine), and olive oil is a rich source. Phytosteroids and oxidized sterols seem to be included in this class, as well. Dietary isoprenoids exert complex actions on cholesterol metabolism at the level of both HMGCoA reductase and LDL, and have been proposed as adjuncts to statins in ASCVD and as anti-neoplastics, but their anti-neoplastic properties still fall in the realm of alternative medicine.
Given the vicissitudes of this pathway, I would not hazard a guess as to which way they would modulate α- and γ-secretases. If I really wanted to know, I would do the experiment that Steve Paul suggests: feed some olive oil to a plaque-forming mouse, or a triple transgenic with both plaque and tangle etiology, and see what happens to pathology and behavior. Maybe the olive oil industry would support such a study? (Only half kidding!)
- Parvathy S, Ehrlich M, Pedrini S, Diaz N, Refolo L, Buxbaum JD, Bogush A, Petanceska S, Gandy S. Atorvastatin-induced activation of Alzheimer's alpha secretase is resistant to standard inhibitors of protein phosphorylation-regulated ectodomain shedding. J Neurochem. 2004 Aug;90(4):1005-10. PubMed.
- Allinson TM, Parkin ET, Turner AJ, Hooper NM. ADAMs family members as amyloid precursor protein alpha-secretases. J Neurosci Res. 2003 Nov 1;74(3):342-52. PubMed.
- Zhou Y, Su Y, Li B, Liu F, Ryder JW, Wu X, Gonzalez-DeWhitt PA, Gelfanova V, Hale JE, May PC, Paul SM, Ni B. Nonsteroidal anti-inflammatory drugs can lower amyloidogenic Abeta42 by inhibiting Rho. Science. 2003 Nov 14;302(5648):1215-7. PubMed.
No Available Further Reading
- Pedrini S, Carter TL, Prendergast G, Petanceska S, Ehrlich ME, Gandy S. Modulation of statin-activated shedding of Alzheimer APP ectodomain by ROCK. PLoS Med. 2005 Jan;2(1):e18. PubMed.