Inhibition of delta-secretase improves cognitive functions in mouse models of Alzheimer's disease

δ-secretase, also known as asparagine endopeptidase (AEP) or legumain, is a lysosomal cysteine protease that cleaves both amyloid precursor protein (APP) and tau, mediating the amyloid-β and tau pathology in Alzheimer's disease (AD). Here we report the therapeutic effect of an orally bioactive and brain permeable δ-secretase inhibitor in mouse models of AD. We performed a high-throughput screen and identified a non-toxic and selective δ-secretase inhibitor, termed compound 11, that specifically blocks δ-secretase but not other related cysteine proteases. Co-crystal structure analysis revealed a dual active site-directed and allosteric inhibition mode of this compound class. Chronic treatment of tau P301S and 5XFAD transgenic mice with this inhibitor reduces tau and APP cleavage, ameliorates synapse loss and augments long-term potentiation, resulting in protection of memory. Therefore, these findings demonstrate that this δ-secretase inhibitor may be an effective clinical therapeutic agent towards AD.

(a) Color-coded electrostatic surface potential of δ-secretase (red: negative charge, blue: positive charge) calculated at pH 7.0. The covalent YVAD-cmk inhibitor labeling the active site is shown in orange sticks, compound 11 in blue sticks.
(b-c) 2d interaction diagrams of compound 11 (b) and compound 11 b (c) with δ-secretase. The diagrams were prepared using the LigPlot + software. Residues on δ-secretase interacting with compound 11(b) are labelled in green; a sulfate ion is indicated as yellow sphere.
(d) (Small molecule) binding sites on caspases and legumain cluster around the β6 strand. The β6 strand serves as the dimerization interface in caspases (c). Consistently, selected mutagenesis in the dimer interface leads to allosteric inhibition of caspases (M). Several activity modulating molecules have been identified which bind close to β6 (3 -6). Likewise, we find compound 11 binding close to the β6 strand on legumain (11). Additionally, a sulfate ion (S) and two glycosylation sites (Asn167 and Asn272; g) are located near β6 in legumain.
(e) Sequence alignment showing the conservation of the allosteric compound 11(b) binding site in human (Q99538) and mouse (NP_776526.1) δ-secretase. Interacting residues are labelled. The sequence numbering is based on human legumain. Interacting residues are not conserved in human caspase-1 (P29466). The alignment was created with ClustalW and modified with Aline. Figure 7. Binding of compound 11b to δ-secretase.

Supplementary
(a) Similar to compound 11, compound 11b is binding to active site liganded and free δ-secretase. Compound 11b was immobilized on a sam®5BLUE biosensor chip and binding of C189S-prolegumain (red curves), in trans activated C189S-legumain only (dark blue curves) and complexed to cystatin E (light blue curves), δ-secretase (dark green curves) and δ-secretase covalently inhibited with YVAD-chloromethyleketone (δ-secretase + YVAD-cmk, light green curves), was tested at pH 5.0. Cystatin E (black curves) served as a control to test for unspecific binding.
(b) Binding of compound 11 to the δ-secretase active site is substrate-like. Stereo-view on the 1 active site of YVAD-cmk and compound 11 bound δ-secretase. Compound 11 is mimicking a substrate in P1 and P2 position. Furthermore, a water molecule is occupying the oxyanion pocket. Thereby, compound 11 together with the oxyanion generates a transition state analog.
The active site is labelled with the covalent YVAD-cmk inhibitor and shown in orange sticks, the catalytic Cys189 and His148 in green sticks and compound 11b in blue sticks.
(d) Zoom-in view on the compound 11b binding site.
(e) Compound 11b binds to the δ-secretase active site, similar to compound 11.
(f) Zoom-in view on the active site. The morpholino group binds into δ-secretase's S1-pocket.
Furthermore, the oxyanion-pocket, formed by Cys189, Gly149 and His148, is also occupied in the structure. The electron density (2F obs -F calc ) defining compound 11b is contoured at 1σ over the mean. Interacting residues on δ-secretase are shown as green sticks.
(g) SDS-PAGE gel of human pro-δ-secretase and pH-activated δ-secretase used for crystallization experiments confirming its purity of > 95%.
(h) Stereo-view on the active site of δ-secretase in complex with compound 11b. The electron density (2F obs -F calc ) defining S1-specificity residues (purple sticks), catalytic residues (green sticks) and compound 11b (blue sticks) is contoured at 1σ over the mean. (a) Reversed phase HPLC chromatograms (C18 column) of compound 11b and 11 were detected at 220 nm absorption, resulting in unimodal elution peaks at 19.7 and 26.9 min, respectively.
(b-d) The 1 H, 13 C and two-Dimensional NMR of compound 11.

Supplementary Figure 11. In vivo pharmacokinetic and pharmacodynamic study of compound 11 in mice.
(a) Pharmacokinetic study of compound 11. Two months old CD-1 male mice were orally administrated 10 mg kg -1 of compound 11 or i.v. injected 2 mg kg -1 of compound 11. The serum was collected at indicated time points with each group of 3 mice (mean ± s.d.). The lowest limit of quantification of compound 11 in plasma was 2 ng ml -1 .
(b-c) Compound 11 time-dependently decreases Aβ concentration in brain and serum. 5XFAD mice were orally administrated with 10 mg kg -1 compound 11. 1 h, 2 h, and 4 h after administration, the concentrations of Aβ in the brain and serum were determined using ELISA.
(d-e) Compound 11 dose-dependently decreases Aβ concentration in brain and serum. 5XFAD mice were were orally administrated vehicle, 5 mg kg -1 , or 10 mg kg -1 compound 11. 4 h later, the concentrations of Aβ in the brain and serum were determined using ELISA (mean ± s.e.m.; n = 10 mice per group).
(f) Pharmacokinetics of compound 11 in 5XFAD mice. 5XFAD mice were orally administrated with vehicle, 5 mg kg -1 , or 10 mg kg -1 compound 11. 4 h after administration, the concentrations of compound 11 in the brain were determined.
(g) Pharmacokinetics of compound 11 in 5XFAD mice. 5XFAD mice were orally administrated with 10 mg kg -1 compound 11. 1 h, 2 h, and 4 h after administration, the concentrations of compound 11 in the brain were determined. (d) The percentage of time spent in the target quadrant in the probe trail was decreased in mice overexpressing tau 1-368 than the GFP-expressing mice, and was not affected by compound 11 treatment. *p < 0.01 compared with GFP group.

Supplementary
(e) Number of crossings over the previous location of the target platform was decreased in tau 1-368 mice than GFP-expressing mice, and was not affected by compound 11. *p < 0.01 compared with GFP group.
(f) The ratio of paired pulses in tau 1-368 mice was not affect by compound 11 treatment (mean ± s.e.m.; n = 3 per group).
(g) LTP of fEPSPs was similar between vehicle-and compound 11-treated mice (mean ± s.e.m.; n = 6 per group). (c) Hematoxylin and eosin staining of the bone marrow, brain, heart, kidney and liver. Bar=20 µm. (a) Time-dependent effect of compound 11 on the deposition of Aβ in the brain. 5XFAD mice were treated with compound 11 at a dose of 10 mg kg -1 d -1 for 1.5 month or 3 month, respectively. The deposition of Aβ was determined using thioflavin-S staining. Scale bar, 100 µm.
(e) Dose-dependent effect of compound 11 on the deposition of Aβ in the brain. 5XFAD mice were treated with compound 11 at a dose of 2, 5, or 10 mg kg -1 d -1 for 1.5 month. The deposition of Aβ was determined using thioflavin-S staining. Scale bar, 100 µm.
(g-h) Concentrations of Aβ 1-40 and Aβ 1-42 in the brain determined using ELISA (mean ± s.e.m.; *P <0.05, one-way ANOVA).     The structures were determined from single crystals. *Highest resolution shells are shown in parentheses.