Discovery of RO7185876, a Highly Potent γ-Secretase Modulator (GSM) as a Potential Treatment for Alzheimer’s Disease

Abstract

γ-Secretase (GS) is a key target for the potential treatment of Alzheimer’s disease. While inhibiting GS led to serious side effects, its modulation holds a lot of potential to deliver a safe treatment. Herein, we report the discovery of a potent and selective gamma secretase modulator (GSM) (S)-3 (RO7185876), belonging to a novel chemical class, the triazolo-azepines. This compound demonstrates an excellent in vitro and in vivo DMPK profile. Furthermore, based on its in vivo efficacy in a pharmacodynamic mouse model and the outcome of the dose range finding (DRF) toxicological studies in two species, this compound was selected to undergo entry in human enabling studies (e.g., GLP toxicology and scale up activities).

Alzheimer’s disease (AD) is an age-related chronic neurodegenerative disease that manifests in progressive cognitive decline followed by gradual changes in functioning ultimately resulting in death.¹ In histopathology, the disease is characterized by the presence of senile, amyloid plaques comprising a core of amyloid-β (Aβ) peptide fibrils^2,3 and neurofibrillary tangles of primarily hyperphosphorylated tau protein,^3,4 along with neuronal and synaptic loss in regions involved in learning and memory such as the hippocampus and cortical areas. Senile plaques are composed of beta amyloid (Aβ), which is a proteolytic fragment of the amyloid precursor protein (APP) produced through sequential proteolytic cleavages.^5,6 APP is first cleaved by BACE1 to generate a 99 amino acid C-terminal fragment (C99), which is then cleaved by γ-secretase in multiple consecutive steps, releasing Aβ peptides of different lengths.⁷ The most abundant product is the Aβ40 peptide, but the relatively minor species, Aβ42 peptide, is very prone to aggregate and may then lead to neurotoxicity.⁸

γ-Secretase is an enzymatic protein complex composed of four different subunits: presenilin (PS), nicastrin (Nct), anterior pharynx-defective 1 (Aph-1), and presenilin enhancer 2 (Pen-2) in a 1:1:1:1 stoichiometry, with PS forming the catalytic subunit of γ-secretase.⁹ It cleaves type I transmembrane proteins and has more than 90 reported substrates, of which APP and Notch are characterized best to date. Mutations in the substrate APP and in the γ-secretase component PS cause autosomal dominant inherited familial AD (FAD), in the majority of mutations leading to an increase in the ratio of Aβ42:Aβ40. This genetic evidence strongly supports the amyloid hypothesis of AD, which has since been refined to suggest that oligomeric Aβ, rather than the larger Aβ aggregates found in amyloid plaques, may be responsible for neurotoxicity.

γ-Secretase modulators (GSMs) are small molecules that selectively decrease Aβ42 and to a lesser extent also Aβ40 production.⁸ GSMs bind to presenilin, the catalytic subunit of γ-secretase, and stabilize the enzyme/substrate complex in such a way that further exoproteolytic cleavage events result in the generation of a greater proportion of shorter Aβ peptides, such as Aβ38 and Aβ37, relative to the longer peptides Aβ42 and Aβ40. These shorter Aβ peptides do not form toxic aggregates and may even function as inhibitors of the aggregation of Aβ42 in vitro.¹⁰ Moreover, recent data seem to confirm these beneficial effects of Aβ38 and Aβ37 in vivo in drosophila models.¹¹

An attractive feature of GSMs is that the effect on Aβ peptides is essentially the opposite of the PS mutations that cause familial AD (FAD). In individuals with PS FAD the proportion of Aβ42 relative to the levels of other smaller Aβ peptides is increased and associated with earlier age-of-onset AD. GSMs shift Aβ peptide production in a direction opposite to that of FAD, implying that they may have the potential to delay or prevent the onset of dementia.

From a safety perspective, GSMs are distinct from γ-secretase inhibitors (GSIs) which have been abandoned as potential AD therapies due to their serious safety liabilities (e.g., skin tumors, impaired B-cells, GI-toxicity, albumin depletion, worsening of memory) which are believed to be driven mainly by inhibition of Notch processing and/or other substrates of γ-secretase¹² and the accumulation of APP β-CTF fragment. In contrast, GSMs do not inhibit processing of Notch and other protein substrates of γ-secretase (such as CD44, E-cadherin, neurexin, and ERB4) and in addition do not cause accumulation of the potentially toxic APP C-terminal fragments (CTFs) in the brain.

All those aspects make GSM a promising approach for the potential treatment of AD and explain why a number of GSMs (including our own derivatives), reaching various development stages, were reported in the past decades.^8,13,14 We have recently discovered compound 2,^15,16 starting from 1,¹⁷ addressing multiple issues (Figure 1). One of the key structural modifications was the reduction of the aromatic rings number, replacing the trifluorophenyl moiety by an alkoxy group, hence preventing phototoxicity and improving the aqueous solubility. As the initial profile of 2 was looking promising, it was submitted to further characterization with an in vivo rat toxicological study. Unfortunately, 2 only displayed a low tolerability, at exposures below 10-fold above efficacious concentration, in our minitox study in rat (7 days treatment upon oral administration) preventing further development. Herein, we report our lead optimization work leading to the discovery of a novel chemical class, the triazolo-azepine, and the identification of (S)-3 (Figure 1), which was successful in both a rodent minitox study and the follow up dose range finding (DRF) toxicology studies (two species, 14 days treatment) supporting its entry into GLP toxicology studies (entry into human enabling study).

Figure 1.

Structures of compound 1, alkoxy 2, and the triazolo-azepine (S)-3 derivatives.

Lead Optimization

The low tolerability of compound 2 was attributed to its relatively high dose required to conduct the toxicological study, 200 mg/kg/day to achieve a calculated 30-fold window versus the efficacious plasma exposure. Furthermore, upon in vivo dosing at 200 mg/kg/day a lack of dose-exposure linearity limited the exposure to less than 10-fold above the efficacy. To circumvent that, the objective was the identification of an alternative compound with a similar physicochemical profile but with an improvment in vitro potency (2, IC₅₀: 31 nM), leading to a maximum tox-dose below 100 mg/kg upon oral administration to achieve a minimum of a 30-fold window versus the efficacious exposure. Initial efforts focused then on identification of novel analogs around compound 2 fulfilling these criteria. For this purpose, several libraries were prepared to explore the scope of three exit vectors (Figure 2). However, while all novel compounds were found to be rather potent, none was significantly superior than 2.

Figure 2.

Further expansion in the alkoxy series.

This initial strategy to reduce the aromaticity by replacing one aromatic ring by an alkoxy moiety did not yield derivatives with both a high level of potency and excellent physicochemical properties. Therefore, a new approach was considered consisting in the saturation of the second phenyl ring fused with the triazole moiety providing potentially two new subseries with either a carbon (C-saturated series) or a nitrogen atom (N-saturated series) in the benzylic position (Figure 3).

Figure 3.

Approaches to reduce the aromaticity.

We started our investigation with the C-saturated (series A, Figure 3). Gratifyingly, the replacement of the triazolopyridine by the corresponding saturated 6-membered ring tetrahydro-triazolopyridine moiety (with n = 1) led to derivatives with promising in vitro potency (e.g rac-5, rac-7, or rac-9, Table 1). We then evaluated the impact of the ring size by preparing the corresponding contracted five (n = 0) and homologated seven (n = 2) ring analogs. We systematically found that the five membered ring analogs showed a lower potency and that the highest potency was actually achieved with the seven member ring (tetrahydro-triazoloazepine derivatives).

Table 1. Ring Size Effect on Potency with C-Saturated Analogs (Subseries A).

^aRacemic compounds.

^bIC₅₀ Aβ42 in nM.

In parallel, we also investigated the subseries B, which contains a nitrogen atom in the benzylic position, potentially advantageous as it suppresses a stereogenic center. We have started by replacing the triazolopyridine by the corresponding saturated tetrahydro-triazolopyrimidine moiety (with n = 1), leading to derivatives with promising in vitro potency (e.g., 14 and 17, Table 2). As previously done with the corresponding C-saturated series A, we evaluated the impact of the ring size by preparing the corresponding contracted five (n = 0) and homologated seven (n = 2) ring analogs. Interestingly, although the difference of potency between the different ring size was more subtle than in the case of the series A, here in subseries B we systematically found that the six membered ring analogs provided the highest potency while both the correspond five and seven membered ring analogs led to lower potency (Table 2).

Table 2. Ring Size Effect on Potency with N-Saturated Analogs (Subseries B).

^aIC₅₀ Aβ42 in nM.

The ring size effect on the potency can be rationalized with our modeling activities using quantum mechanics¹⁸ (QM) calculation. For the subseries A (C-saturated), the ring size affects the phenyl moiety position very significantly (Figure 4). When n = 2 (7-membered ring, in red), the phenyl moiety is closely positioned to its initial position as compared when attached to the original triazolo-pyridine core (in gray). However, going down to a 6-membered ring (in green) or further contraction to a 5-membered ring (in blue) shifted the phenyl moiety even more, particularly for the 5-membered ring. This is in agreement with the measured potency where [n = 2] > [n = 1] ≫ [n = 0] (Table 1).

Figure 4.

Conformational effect of the ring size within subseries A (C-saturated).

For the subseries B (N-saturated), the ring size affects also the phenyl moiety position but with a less pronounced effect (Figure 5). In this case, it is when n = 1 (6-membered ring, in green) that the phenyl moiety is positioned closest to its initial position as comparedto when attached to the initial triazolo-pyridine core (in gray). Enlarging the ring size to a 7-membered ring (in red) or contracting it to a 5-membrered ring (in blue) led to a similar shift of the phenyl moiety. This is in agreement with the measured potency where [n = 1] > [n = 2] ≥ [n= 0] (Table 2).

Figure 5.

Conformational effect of the ring size within subseries B (N-saturated).

We therefore decided to focus on both subseries, with the tetrahydro-triazolo-azepine (in red) and the dihydro-triazolo-pyrimidine (in yellow) being the most potent, which could also be rationalized based on the alignment with the initial aromatic triazolo-pyridine in gray (Figure 6). For both, we planned to evaluate the phenyl substitution pattern (R1 moiety), in combination with various head groups (HG).

Figure 6.

Tetrahydro-triazolo-azepine (red) and dihydro-triazolo-pyrimidine (yellow) subseries A and B overlaid with the initial triazolo-pyridine core in gray.

A set of novel derivatives in the series A was prepared leading to promising analogs (Table 3).

Table 3. Aryl Substitution Patterns and Head Groups Evaluation in Series A.

^aRacemic compounds.

^bIC₅₀ Aβ42 expressed in nM.

^cSolubility expressed in μg/mL and measured in Lysa assay.

^dClearance in human hepatocytes expressed in μL/min/Mcells.

^eHuman P-gp transport as efflux ratio.

The use of a trifluorophenyl as an aryl group led to derivatives with a high potency and in combination with various head groups. However, with a chloropyrimidine headgroup (e.g rac-19) the aqueous solubility was below the lower limit of measurement. With the use of pyridazines as headgroup (e.g., rac-21 and rac-22) the highest solubility was achieved together with good clearance and potency, but at the expense of being strong P-gp transporter substrates. The methoxypyridine headgroup (e.g., rac-20) appears to offer a good compromise; however, upon further characterization it was found to be rather potent on hERG inhibition (IC₅₀/IC₂₀: 0.55/0.20 μM). Five membered ring oxadiazoles and isoxazole head groups (e.g., rac-24, rac-25, rac-26) led to a reduced potency. At this stage, the methylpyrimidine (e.g., rac-3) appears to be the best headgroup. Further efforts to improve the overall profile were attempted by modifying the aryl substitution pattern, for instance preparing the analogous difluoro derivatives (e.g., rac-27), but this led to a slight potency reduction and an increase of clearance. The meta-CF₃ substituted derivative (rac-28) retains potency and improves the solubility, but the clearance became high. We therefore focus on the compound rac-3 and proceed to a separation of both enantiomers by chiral HPLC. It was found that the S enantiomer from 3 was the most potent. The absolute configuration was established based on a crystal structure (Figure 7)¹⁹ and in agreement based on our conformational hypothesis.

Figure 7.

Molecular structure of (S)-3 determined by X-ray crystal structure analysis with thermal ellipsoids (drawn at the 50% probability level) for non-hydrogen atoms (heteroatoms F, N are colored in green, blue, respectively).

Similarly, in subseries B, we prepared a range of additional analogs to identify the best combination of headgroup and phenyl substituents. Table 4 summarizes the most promising examples.

Table 4. Aryl Substitution Patterns and Head Groups Evaluation in Series B.

^aIC₅₀ Aβ42 in nM.

^bSolubility expressed in μg/mL, measured in Lysa assay.

^cClearance in human hepatocytes expressed in μL/min/Mcells.

^dHuman P-gp transport as efflux ratio.

The optimal headgroup was also the methylpyrimidine (e.g., 32). Five-membered ring head groups were only modestly tolerated, with the most potent derivative being the oxadiazole 35. The methoxypyridine-pyridine derivative (e.g., 33) appears interesting, but upon further evaluation, it was also found to be quite potent on hERG (IC₅₀/IC₂₀: 0.7/1.4). Although in this subseries, interestingly, the use of pyridazine (e.g., 34) did not raise the P-gp efflux ratio unlike in subseries A, where the potency could not be improved further with the use of alternative phenyl substituents.

The best derivatives from each subseries (A and B) were fully characterized in vitro and displayed an excellent overall profile (Table 5).

Table 5. In-Vitro Profile of 32 and (S)-3.

	32	(S)-3
h IC50 Aβ42 H4: total/free [nM]	9/6	9/4
h IC50 Aβ42 Hek292: total/free [nM]	4/3	4/2
Mouse IC50 N2A: total/free [nM]	5/3	4/2
h Notch IC50: [nM]	>10000	>10000
LogD:	3.5	3.8
Cl heps (h/m/r): [μL/min/Mcells]	2.4/10/6.4	4.5/6.7/43
P-gp ER (h/m):	2.7/4.6	2.4/3.3
F.u. (h/m/r): [%]	1.0/0.6/1.4	5.4/1.4/1.4
GSH	clean	clean
Ames/MNT	clean/clean	clean/clean
hERG IC20/IC50 [μM]	>1/>2	0.3/1.4

It is noteworthy that the mode of action of this compounds as a gamma secretase modulator was further confirmed as both compounds reduce the amount of the large peptides Aβ42 and Aβ40 prone to aggregate and increase the fractions of smaller peptides Aβ37 and Aβ38 while leaving the total amount of Aβ unchanged. And in sharp contrast to the gamma secretase inhibitors (GSI), those gamma secretase modulator (GSM) compounds do not inhibit the Notch pathway (Figure 8).

Figure 8.

In vitro potency and Notch selectivity of 32 (panel A and B) and (S)-3 (panel C and D).

In Vivo Pharmacokinetics, Efficacy, and Toxicological Studies

Both compounds were evaluated in a single dose pharmakokinetic study (SDPK) in rodents upon oral and intravenous administration to assess their potential for upcoming toxicological study and preclinical efficacy studies. While for 32 we performed this SDPK study in rat, for (S)-3 we used the mice as a rodent species, due to a specific higher in vitro mice clearance (Table 6). A good oral bioavailability was observed, leading to high exposures. The half-life was in a good range for a once a day drug. Moreover, we also observed a good correlation between the observed in vivo clearance and the predicted clearance based on the in vitro data.

Table 6. In-Vivo SDPK Profile of 32 in Rat^a and (S)-3 in Mice^b.

	32	(S)-3
AUCinf_D (h·kg·ng/mL/mg)	4110	1795
Cl (iv) (mL/min/kg)	3.5	5.0
Terminal t1/2 (iv) (h)	4.6	3.4
Vss (iv) (l/kg)	1.1	1.1
F (%)	82	53
IVIVC clearancec	0.65	1.3

^aWistar rat (male), n = 3/group, iv: 1 mg/kg; po: 3 mg/kg.

^bC57BL/6J mice (male), n = 3/group, iv: 1 mg/kg, po: 3 mg/kg.

^cIVIVC clearance, ratio obs/predicted from Clint.

We next evaluated the compounds 32 and (S)-3 in a pharmacodynamic (PD) mice model to determine their efficacy. For this purpose we used the double transgenic mice model APP-Sweddish, the compounds were administered orally, and the different peptides levels (Aβ37, Aβ38, Aβ 40, Aβ42, and total Aβ) were determined overtime in brain (Figure 9). A free in vivo IC₅₀ Aβ42 was also measured for 32 and (S)-3 and found to be 3.0 nM and 2.5 nM in perfect agreement with the free in vitro potency in mice of 3.0 nM and 2.0 nM.

Figure 9.

In-vivo pharmacodynamic evaluation of 32 (panel A) and (S)-3 (panel B).

Additionally, an excellent PKPD correlation was observed (Figure 10). The predicted in vivo Aβ42 reduction (depicted in black Figure 10), based on the in vitro potency of the compounds and their observed plasma concentration (in blue), match perfectly the measured Aβ42 reduction (in gray).

Figure 10.

Aβ42 reduction: Prediction versus measurement for compounds 32 and (S)-3.

Both compounds 32 and (S)-3 were then evaluated in a minitox study upon daily oral administration in mice and rat, respectively, for 4 days. The mouse was selected as the rodent species for the compound (S)-3 as it displayed a specific high clearance in rat both in vitro and in vivo. To our delight, both exhibit a very good tolerability and no histopathological findings preventing further development. Hence we proceeded in the 2 weeks dose range finding studies (DRF) using in addition a nonrodent species. In that study, (S)-3 was evaluated as superior to 32 and was therefore selected as the compound of choice to enter GLP toxicological studies (entry into human enabling studies).

Synthesis of 32 and (S)-3

The synthesis of 32 is depicted Scheme 1.²⁰ A nucleophilic addition of the difluoro-aniline 33 onto diphenyl cyanocarbonimidate led to 34, which underwent a N-alkylation with 2-(3-bromopropoxy)tetrahydropyran to form 35. Upon reaction with hydrazine hydrate, the corresponding amino-triazole 36 was obtained. Following a deprotection of the alcohol residue, an intramolecular Mitsunobu reaction was performed to yield the fused derivative 37. A Sandmeyer reaction was then done to provide the versatile bromo-triazole intermediate 38. Finally, a cross-coupling Buchwald reaction with the intermediate 39 led to the formation of 32.

Scheme 1. Synthesis of 32.

Conditions: (a) Diphenyl cyanocarbonimidate (1.0 equiv), iPrOH, RT, 16 h, 54%; (b) 2-(3-bromopropoxy)tetrahydro-2H-pyran (1.5 equiv), K₂CO₃ (2.0 equiv), DMF, 85 °C, overnight, 51%; (c) NH₂–NH₂·H₂O 25% in H₂O (1.0 equiv), MeOH, RT, overnight, 80%; (d) (i) HCl 2 N (3.0 equiv), MeOH, RT, 1.5 h; (ii) Cyanomethylenetrimethyl phosphorane (3.0 equiv), THF, RT, overnight, 72%; (e) CuBr₂ (1.5 equiv), t-BuNO (1.5 equiv), CH₃CN, 75 °C, 0.5 h, 86%; (f) Compound 39 (1.1 equiv), Pd₂(dba)₃ (0.015 equiv), t-BuXPhos (0.03 equiv), t-BuONa (4.0 equiv), MeTHF, 75 °C, 1.0 h, 62%.

The synthesis of (S)-3 is depicted in Scheme 2.²¹ Trifluorophenylacetic acid 40 was alkylated with 1-chloro-4-iodobutane forming the racemic derivative 41. The carboxylic acid moiety was then activated by conversion to the corresponding acid chloride, which then was coupled with the amino-guanidine, and an intramolecular cyclization in the presence of sodium hydroxide afforded the amino-triazolo-azepine 42. A Sandmeyer reaction was used to form the racemic bromo-triazole derivative 43. Finally, a cross-coupling Buchwald reaction with the intermediate 39 led to the formation rac-3, which upon separation of the enantiomers by chiral SFC gave the optically pure (S)-3.

Scheme 2. Synthesis of (S)-3.

Conditions: (a) 1-chloro-4-iodobutane (1.1 equiv), NaHMDS (2.0 equiv), THF, −10 °C, 3 h, 77%; (b) (i) SOCl₂ (3.0 equiv), DMF (1.0 equiv), CH₂Cl₂, RT, 4 h, 100%; (ii) Amino-guanidine (1.1 equiv), NMP, 120 °C, 4 h, 100%; (iii) 20% NaOH/H₂O, pH = 10, 100 °C, 10 h, 68%; (c) CuBr₂ (1.2 equiv), t-BuNO (1.2 equiv), CH₃CN, 65 °C, 1 h, 59%; (d) Compound 39 (1.1 equiv), Pd₂(dba)₃·CHCl₃ (0.01 equiv), tBuylXphos (0.02 equiv), NaOtBu (4.0 equiv), 3-Me-THF, 70 °C, 1 h, 93%; (e) Chiral SFC (Column: ChiralPak AY, 150 × 4.6 mm I.D., 3 μ, Mobile phase: A for CO₂ and B for Isopropanol (0.05% DEA)) 43%.

Conclusion

In summary, our lead optimization program has led to the discovery of a novel chemical class, the triazolo-azepine series. Fine tuning activities to balance multiple physicochemical parameters (e.g., clearance, P-gp, and solubility for instance) yielded the identification of (S)-3. This compound clearly demonstrated an in vitro and in vivo exquisite modulation of the gamma secretase enzyme by increasing the amount of the smaller fraction peptides (Aβ37 and Aβ38), decreasing the potentially pathogenic larger fractions peptides (Aβ40 and Aβ40) while leaving unchanged the total Aβ peptide level without affecting the Notch pathway. The outcome of the 2-week dose range finding toxicological studies in two species was supportive of further development of this compound.

Glossary

Abbreviations

Aβ

amyloid beta

AD

Alzheimer’s disease

APP

amyloid precursor protein

DMF

N,N-dimethylformamide

DMPK

drug metabolism and pharmacokinetics

DRF

dose range finding

GLP

good laboratory practice

GS

γ-secretase

GSM

γ-secretase modulator

P-gp

P-glycoprotein

QM

quantum mechanics

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.0c00109.

• Detailed synthetic procedure for compounds 32 and (S)-3 and their complete analytical data (¹H NMR, ¹³C NMR, ¹⁹F NMR, [α]D²⁵, and HRMS), ¹H NMR spectra for 32 and (S)-3, detailled crystallographic information for (S)-3, in vitro cellular Aβ secretion assay, Notch cellular reporter assay, acute treatment of APP-Swe transgenic mice (PDF)

The authors declare no competing financial interest.