Secretase Modulators in the Treatment of Alzheimer Disease

CLINICAL IMPLICATIONS OF BASIC NEUROSCIENCE RESEARCH

SECTION EDITOR: HASSAN M. FATHALLAH-SHAYKH, MD, PhD Potential Use of ␥-Secretase Modulators in the Treatment of Alzheimer Disease

Steven L. Wagner, PhD; Rudolph E. Tanzi, PhD; William C. Mobley, MD, PhD; Douglas Galasko, MD

lthough significant progress has occurred in the past 20 years regarding our under- standing of Alzheimer disease pathogenesis, we have yet to identify disease- modifying therapeutics capable of substantially altering the clinical course of this preva- lent neurodegenerative disease. In this short review, we discuss 2 approaches that are Acurrently being tested clinically (␥-secretase inhibition and ␥-secretase modulation) and empha- size the significant differences between these 2 therapeutic approaches. We also discuss certain genetic- and biomarker-based translational and clinical trial paradigms that may assist in devel- oping a useful therapeutic agent. Arch Neurol. 2012;69(10):1255-1258. Published online July 16, 2012. doi:10.1001/archneurol.2012.540

Genetic models have long been used to de- tin), and according to the American Heart vise and evaluate rational approaches to the Association (http://www.heart.org), stat- treatment of high-prevalence diseases such ins can safely elicit a selective lowering of as coronary artery disease and stroke. One low-density lipoprotein cholesterol levels of the most notable examples was in the and long-term administration of statins can pioneering studies of Mabuchi et al1 re- reduce the incidence of coronary artery dis- porting the highly beneficial effects of a ease and stroke by more than 25% to 30%. 3-hydroxy-3-methylglutaryl–coenzyme A Alzheimer disease (AD) affects approxi- reductase inhibitor, compactin, in pa- mately 1 in 8 Americans aged 65 years or tients with familial hypercholesterolemia. older. Unfortunately, unlike coronary ar- 3-Hydroxy-3-methylglutaryl–coenzyme A tery disease and stroke, there are currently reductase is the enzyme that catalyzes the no effective disease-modifying or preven- conversion of 3-hydroxy-3-methylglutaryl– tive therapies available. However, more than coenzyme A to mevalonate, the rate- 200 different mutations have been identi- limiting step in cholesterol synthesis. This fied in 3 genes associated with an autoso- historic study involving 7 patients with a mal dominant form of AD, referred to as mutation in the low-density lipoprotein re- early-onset familial Alzheimer disease ceptor gene (LDLR) reported the success- (EOFAD). It is an aggressive form of AD that ful lowering of low-density lipoprotein cho- has served as a template for many of the cel- lesterol levels without changing or even lular and transgenic animal models being slightly increasing high-density lipopro- used to study the pathogenesis of AD and tein cholesterol levels. That small proof- is relevant to the more common late-onset of-concept clinical pharmacodynamic study 2 eventually paved the way to the subse- AD. The 3 genes causally implicated in quent development of a related class of mol- EOFAD, which typically affects carriers be- ecules known as the synthetic statins (eg, fore age 65 years and sometimes as early as fluvastatin, atorvastatin, and rosuvasta- the third or fourth decades, are APP (amyloid precursor protein), PSEN1 (presenilin Author Affiliations: Department of Neurosciences, University of California, San 1), and PSEN2 (presenilin 2). The preseni- Diego, School of Medicine, La Jolla (Drs Wagner, Mobley, and Galasko); and lin genes encode 2 highly homologous pro- Genetics and Aging Research Unit, Department of Neurology, Massachusetts teins that can independently assemble along General Hospital, Charlestown (Dr Tanzi). with 3 other membrane-associated pro-

ARCH NEUROL / VOL 69 (NO. 10), OCT 2012 WWW.ARCHNEUROL.COM 1255

©2012 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/25/2021 substrate of ␥-secretase, and the Membrane Notch-1 receptor intracellular domain peptide plays an important role in the differentiation of multiple cell types.4 The GSIs also cause accumu- lation of the ␥-secretase substrate C99 (the C-terminal fragment of APP gen- Pen-2 erated by the A␤-cleaving enzyme γ Aβ BACE 1) as well as C-terminal frag- 40 PS-NTF 38 42 ments from a number of other type 1 37 39 1 membrane proteins believed to be Aβ 5 AICD degraded by ␥-secretase. Eli Lilly and

48 49 Co recently reported (http://www.lilly PS-CTF ε .com) that in 2 phase 3 clinical trials involving more than 2600 mildly to Nicastrin moderately affected patients with Aph-1 probable AD, the GSI semagacestat failed to slow disease progression and led to worsened cognition and per- formance of activities of daily living. In addition, semagacestat was associated with an increased risk of skin cancer. The explanation for these dis- appointing clinical outcomes is currently unknown but may be mecha- Figure 1. ␥-Secretase processing of amyloid precursor protein C-terminal fragments (CTFs). A cartoon illustration shows the various subunits of the ␥-secretase enzyme complex performing a number of different nistically related to inhibiting the putative ␥-secretase–mediated proteolytic activities. Assembly and maturation of the ␥-secretase enzyme ␥-secretase enzyme complex and at- begins with the formation of a binary complex between nicastrin and anterior pharynx defective 1 (Aph-1). tenuating the processing of 1 or more Presenilin (PS) binds to the nicastrin/Aph-1 complex to form an inactive trimeric complex. The PS N-terminal fragment (NTF) and the PS-CTF collectively harbor the catalytic center of ␥-secretase, which of its substrates including the C- requires binding of presenilin enhancer 2 (Pen-2) for the endoproteolysis of PS into the PS-NTF and the terminal fragments of APP, Notch-1, PS-CTF, rendering activation of ␥-secretase. The active ␥-secretase enzyme can cleave the amyloid and possibly other substrates.5 precursor protein CTF at numerous sites, yielding a host of peptide products (amyloid precursor protein intracellular domain [AICD], ␤-amyloid (A␤) 1-42, A␤1-40, A␤1-39, A␤1-38, and A␤1-37). The relative size Interestingly, the vast majority of of the curved arrows at the ␥ sites indicates the relative use of each processing site. The ε cleavages (larger EOFAD-linked mutations do not cor- curved arrows at amino acids 48 and 49 of the 99–amino acid–long amyloid precursor protein CTF) relate with an overall increase in generate the AICDs, which remain inside the cell (as indicated by the direction of the right arrow) and may ␥-secretase activity, nor do they ap- actually travel to the nucleus, while the various A␤ peptides are released extracellularly (as indicated by the ␤ direction of the left arrow) immediately on cleavage by ␥-secretase within the membrane. The different pear to increase total A peptide lev- colors correspond to the different proteolytic processing products. els. Instead, most appear to cause a 2-fold increase in the ratio of A␤42/ teins, namely, nicastrin, anterior phar- EOFAD, these and perhaps others A␤40 (Figure 2). This increased ynx defective 1, and presenilin that develop further downstream (eg, A␤42/A␤40 ratio is a highly repro- enhancer 2 (Pen-2), into a large mem- neurofibrillary tangles, another ma- ducible biochemical phenotype (in brane-bound enzymatic complex jor neuropathological hallmark of vitro and in vivo) for the predomi- (Figure 1) that plays a key role in AD) arise owing to a change in the ra- nant effect of the more than 200 metabolizing a large number of type tio of cerebral levels of A␤42 relative EOFAD-linked genetic mutations, in- 1 membrane proteins, including APP.2 to the next longest A␤ peptide, A␤40 cluding a number of mutations in APP The rationale for targeting this enzy- (which is by far the most abundant (those proximal to the ␥-secretase matic complex, referred to as ␥-secre- of all secreted A␤ peptides), increas- cleavage site in the APP protein), all tase, as a therapeutic approach for AD ing the A␤42/A␤40 ratio approxi- of those in PSEN2, and most of the stems from the fact that ␥-secretase mately 2-fold.2 Initial therapeutic in- large number of mutations in PSEN1.2 performs the terminal series of pro- terventions targeting ␥-secretase Therefore, a therapeutic rationale teolytic cleavages that generate and re- focused on curbing total A␤ peptide based on this EOFAD genetics model lease extracellularly a number of pep- production by directly inhibiting would entail modulating ␥-secretase tides known collectively as ␤-amyloid ␥-secretase activity through a class of activity (decreasing the A␤42/A␤40 (A␤).2 Most secreted A␤ peptides compounds referred to as ␥-secre- ratio), not inhibiting this important range in size from approximately 34 tase inhibitors (GSIs).3 The GSIs at- enzyme complex. The feasibility of to 42 amino acids (Figure 1). One of tenuate the production of all major A␤ this more selective EOFAD genetics- these, A␤42 (42 amino acids long), peptide variants (eg, A␤42, A␤40, based approach was first identified is the major peptide found in neu- A␤39, A␤38, and A␤37) equiva- using a series of nonsteroidal anti- ritic plaques and diffuse amyloid de- lently (Figure 2) and inhibit the gen- inflammatory drug (NSAID)–like posits, neuropathological lesions eration of an APP intracellular do- compounds, eg, ibuprofen and sulin- found in abundance in the brains of main and the Notch-1 receptor dac sulfide.6 However, the NSAID-like patients with AD at autopsy. In intracellular domain. Notch is a key compounds lack potency (Ն50-

ARCH NEUROL / VOL 69 (NO. 10), OCT 2012 WWW.ARCHNEUROL.COM 1256

©2012 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/25/2021 100µM in vitro half-maximal inhibitory concentrations for selective low- A ering of the A␤42 level in cell-based assays) and later proved to have very 99 6,7 poor brain penetrance. Despite hav- 42 ing less than stellar pharmacody- 40 namic and pharmacokinetic proper- Healthy ties, a large phase 3 clinical trial 39 involving more than 1600 mildly af- 38

fected patients with AD was con- Relative Peptide Level, AU ducted using the NSAID-like com- 37 pound R-flurbiprofen (Tarenflurbil; sponsored by Myriad Genetics, Inc). B

Unfortunately, this study failed to 99 show any clinical benefit.7 More recently, a potent series of 42 compounds referred to as the non– 40 NSAID-like or non–carboxylic acid– EOFAD ↑42/40 containing ␥-secretase modulators 39 (GSMs), harboring a completely dif- 38 ferent chemical scaffold from that of Relative Peptide Level, AU

NSAIDs, have received considerable 37 attention from pharmaceutical com- panies, biotechnology concerns, and C academic laboratories. Within the last 99 18 months, more than 60 new pat- ent applications have been pub- 42

lished on slight variations of GSM GSMs 40 chemical scaffolds highly similar to ↓42/40 those originally reported by mem- ↑38 and 37 39 8 bers of our team. This may seem sur- 38 prising given the recent failures of the Relative Peptide Level, AU GSI semagacestat and Tarenflurbil. 37 However, those 2 compounds had the limitations mentioned earlier. Al- D though GSIs and GSMs target the 99 ␥-secretase enzyme complex, they appear to have distinct mechanisms of 42 action as measured by their effects on ␤ GSIs 40 the levels of the various A peptide ↑CTFs variants and C-terminal fragments ↓Aβs 39 such as C99.3,8 Unlike the GSIs, some these nonacidic GSMs (a series of dia- 38 Relative Peptide Level, AU rylaminothiazoles) were found to be capable of potently inhibiting A␤42 37 production in vitro (cell-based half- Peptide Length, Amino Acids maximal inhibitory concentrations in the low nanomolar range) and in Figure 2. Semiquantitative estimation of the outcome in healthy individuals (A) and in individuals 8 vivo. At slightly higher concentra- harboring early-onset familial Alzheimer disease (EOFAD)–linked mutations before (B) and after receiving tions, they reduce A␤40 levels only either a ␥-secretase modulator (GSM) (C) or a ␥-secretase inhibitor (GSI) (D). The graphs represent moderately while concomitantly in- shifts in the different forms of ␤-amyloid (A␤) that result from processing by normal (wild-type) ␥-secretase, mutations in the presenilin component of ␥-secretase that result in EOFAD, and the effects creasing the levels of the shorter of treatment with GSIs or GSMs. The heights (ordinate) and lengths (abscissa) of each A␤ peptide variant A␤38 and A␤37 peptide variants correspond to the relative abundance and the number of amino acids, respectively. The GSMs elicit a (Figure 2).8 Moreover, one of these unique bidirectional biomarker profile with respect to their effects on A␤ peptide variants (decreasing A␤42 and A␤40 levels, while increasing A␤38 and A␤37 levels). AU indicates arbitrary units; nonacidic diarylaminothiazole- CTFs, C-terminal fragments. containing GSMs was shown to mark- edly reduce A␤ deposition in a trans- tion of GSIs in rodents such as gob- ure of R-flurbiprofen (Tarenflurbil) in genic mouse model of AD following let cell hyperplasia.8 the phase 3 trial highlights a critical long-term daily administration for 29 Potent GSMs that can selectively first step: the need to demonstrate that consecutive weeks and did not show decrease the production of A␤42 will a GSM engages its target and selec- adverse effects that are commonly as- need to be rigorously tested during tively lowers the A␤42 level in the sociated with long-term administra- translation to clinical trials. The fail- brain in humans. This can be exam-

ARCH NEUROL / VOL 69 (NO. 10), OCT 2012 WWW.ARCHNEUROL.COM 1257

©2012 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/25/2021 ined by pharmacokinetic and phar- addressing other potential mecha- Galasko. Analysis and interpreta- macodynamic studies that analyze ce- nisms of disease. Second, extensive tion of data: Mobley. Drafting of the rebrospinal fluid and plasma. Serial structural neuronal pathology is al- manuscript: Wagner, Tanzi, and sampling of cerebrospinal fluid via an ready present in the hippocampus Galasko. Critical revision of the manu- indwelling lumbar catheter may help and other brain regions in people script for important intellectual con- to define the precise time course and with mild to moderate AD, which tent: Tanzi and Mobley. Obtained selectivity of inhibition of A␤42 pro- limits the degree of recovery or im- funding: Mobley. Administrative, tech- duction and facilitate dose selection. provement that may be achieved, nical, and material support: Mobley. These studies can also analyze other even if treatment with GSMs leads Study supervision: Mobley. forms of A␤ such as A␤38, which, to a dramatic reduction in amyloid FinancialDisclosure:DrsWagnerand based on transgenic mouse studies, burden. For these reasons, clinical Tanzi are shareholders and cofound- may increase in response to a potent trials of GSMs are likely to be more ers of a privately held company (Neu- GSM.8 The studies can address the promising if conducted in people rogenetic Pharmaceuticals, Inc) that question of whether there is a re- with very early symptomatic AD (eg, holds rights to a GSM currently in pre- ␤ bound increase in A production mild cognitive impairment). clinical development. when drug levels fall—a problem pre- With many potent GSMs currently Additional Contributions: Nora 9 viously cited with GSIs. in various stages of preclinical and Joelson, BA, created the figures. We In planning clinical trials to test clinical development, perhaps we thank the Cure Alzheimer’s Fund, a potent GSM in relation to AD, sev- should take a lesson from the history the Down Syndrome Research and eral key issues need to be consid- of compactin and prioritize limited Treatment Foundation, and the Hill- ered. It is unknown how much the proof-of-concept trial(s) with well- ␤ blom Foundation for their past and A 42 level needs to be lowered to studied EOFAD kindreds whose continued support of the discovery have significant clinical benefits. Al- disease course, extending from pre- and development of GSMs. though dosing can be bracketed from clinicalphasestoclinicaloutcomes,has phase 1 pharmacokinetic and phar- been well mapped. Proof-of-principle macodynamic studies, it is unclear studies have been proposed among whether reduction of A␤42 produc- presymptomaticormildlyaffectedcar- REFERENCES tion by 25% (a moderate effect) or riers of an extended Colombian fam- 1. Mabuchi H, Haba T, Tatami R, et al. Effect of an by 50% or higher (a large effect) is ily of more than 5000 members carry- inhibitor of 3-hydroxy-3-methyglutaryl necessary to influence amyloid dep- ing a single PSEN1 genetic mutation coenzyme A reductase on serum lipoproteins and osition and neuritic plaque dynam- who have been thoroughly studied for ubiquinone-10-levels in patients with familial ics in the brain. Monitoring plasma thepastfewdecades.Tomeetthisneed, hypercholesterolemia. N Engl J Med. 1981;305 and cerebrospinal fluid during long- GSMs must be capable of modulating (9):478-482. term dosing and the use of biomark- a␥-secretaseenzymecomplexharbor- 2. Tanzi RE, Bertram L. Twenty years of the Alzhei- mer’s disease amyloid hypothesis: a genetic ers such as amyloid imaging can ing EOFAD-linked mutations. Inter- perspective. Cell. 2005;120(4):545-555. provide tools to monitor these para- estingly, unpublished experiments in 3. Anderson JJ, Holtz G, Baskin PP, et al. Reduc- meters. Biomarkers such as struc- 2 of our laboratories (R.E.T. and tions in ␤-amyloid concentrations in vivo by the tural magnetic resonance imaging S.L.W.) suggest that these types of ␥-secretase inhibitors BMS-289948 and can supplement clinical assess- aminothiazole-containingGSMswork BMS-299897. Biochem Pharmacol. 2005;69(4): 689-698. ment to provide insights into at least as well if not better against all 4. De Strooper B, Annaert W, Cupers P, et al. A pre- whether sustained lowering of the 5ofthemutantpresenilin1–expressing senilin-1-dependent ␥-secretase-like protease me- A␤42 level (and a concomitant cell lines tested so far compared with diates release of Notch intracellular domain. Nature. increase in the A␤38 level) has ben- the cell lines expressing wild-type pre- 1999;398(6727):518-522. efits on brain structure and func- senilin 1. Studies with carriers of fa- 5. Wakabayashi T, De Strooper B. Presenilins: members of the ␥-secretase quartets, but part-time so- tion. Toxic effects will need to be milial AD–linked mutations could loists too. Physiology (Bethesda). 2008;23:194- monitored carefully in view of the serve as important tests of therapeu- 204. experience with semagacestat. tic regimens targeting ␥-secretase and 6. Weggen S, Eriksen JL, Das P, et al. A subset of Although clinical trials have typi- the amyloid cascade hypothesis. NSAIDs lower amyloidogenic Abeta42 indepen- cally targeted people with sympto- dently of cyclooxygenase activity. Nature. 2001; 414(6860):212-216. matic AD at a mild to moderate stage Accepted for Publication: March 2, 7. Green RC, Schneider LS, Amato DA, et al; Taren- of dementia, this may not be the best 2012. flurbil Phase 3 Study Group. Effect of tarenflurbil approach for GSMs. First, the ex- Published Online: July 16, 2012. on cognitive decline and activities of daily living in tent of brain amyloid burden does doi:10.1001/archneurol.2012.540 patients with mild Alzheimer disease: a random- not correlate with the degree of cog- Correspondence: Steven L. Wag- ized controlled trial. JAMA. 2009;302(23):2557- 2564. nitive impairment in AD. It is pos- ner, PhD, Department of Neurosci- 8. Kounnas MZ, Danks AM, Cheng S, et al. Modula- sible that a cascade of neurodegen- ences, University of California, San tion of ␥-secretase reduces ␤-amyloid deposition erative processes eventually becomes Diego, School of Medicine, MTF in a transgenic mouse model of Alzheimer’s disease. independent of A␤ or amyloid- 150, 9500 Gilman Dr, Ste 0624, La Neuron. 2010;67(5):769-780. related toxic effects and leads to pro- Jolla, CA 92093 (slwagner@ucsd 9. Martone RL, Zhou H, Atchison K, et al. Begacestat (GSI-953): a novel, selective thiophene sulfon- gressive dementia in AD; decreas- .edu). amide inhibitor of amyloid precursor protein ing A␤42 production may not be Author Contributions: Study con- ␥-secretase for the treatment of Alzheimer’s disease. successful without simultaneously cept and design: Wagner, Tanzi, and J Pharmacol Exp Ther. 2009;331(2):598-608.

ARCH NEUROL / VOL 69 (NO. 10), OCT 2012 WWW.ARCHNEUROL.COM 1258