Discovery of novel PDE9 inhibitors capable of inhibiting Aβ aggregation as potential candidates for the treatment of Alzheimer’s disease

Recently, phosphodiesterase-9 (PDE9) inhibitors and biometal-chelators have received much attention as potential therapeutics for the treatment of Alzheimer’s disease (AD). Here, we designed, synthesized, and evaluated a novel series of PDE9 inhibitors with the ability to chelate metal ions. The bioassay results showed that most of these molecules strongly inhibited PDE9 activity. Compound 16 showed an IC50 of 34 nM against PDE9 and more than 55-fold selectivity against other PDEs. In addition, this compound displayed remarkable metal-chelating capacity and a considerable ability to halt copper redox cycling. Notably, in comparison to the reference compound clioquinol, it inhibited metal-induced Aβ1-42 aggregation more effectively and promoted greater disassembly of the highly structured Aβ fibrils generated through Cu2+-induced Aβ aggregation. These activities of 16, together with its favorable blood-brain barrier permeability, suggest that 16 may be a promising compound for treatment of AD.

Inhibitory properties of PDE9 inhibitors. The PDE9 inhibitory activities of the compounds 12-29 are summarized in Table 1. Bay73-6691, which showed an IC 50 of 48 nM, an inhibitory activity consistent with that reported in the literature (55 nM), was used as the positive control 20 . As indicated, most compounds were found to be potent PDE9 inhibitors with IC 50 values less than 100 nM. Qualitative structure-activity relationship analyses showed that the inhibitory activity was influenced by the structures of R 2 . From a comparison of the potency of 12 (IC 50 = 211 nM) with that of 15 (58 nM), and 28 (1160 nM) with that of 29 (38 nM), it appeared that the amine structures were more favorable than the corresponding imines for the inhibition of PDE9. However, this difference was negligible when methoxyl groups were present in the benzene ring of C (13, imine, 35 nM; 16, amine, 34 nM). However, in the series of compounds containing an imine structure, the tetrahydropyran-4-yl (R 1 ) appeared to be beneficial for the inhibitory activity. For example, compound 14, in which the c-pentyl (R 1 ) of 13 was replaced by tetrahydropyran-4-yl in the pyrazole moiety, showed the highest inhibitory effect, with an IC 50 value of 12 nM. However, compound 16 (34 nM) exhibited superior inhibitory activity than did compound 17 (51 nM), indicating that the c-pentyl group was superior to the tetrohydropyran-4-yl in the amine series. Considering the relative stabilities of the imines and the amine in vivo, we focused on studying the effect of different R 2 in the amine structures. Most compounds with different substituents on the moiety of C ring, such as 19-24, 26-27, and 29, showed excellent PDE9 inhibitory activity, with IC 50 values ranging from 32 nM to 59 nM. Compounds 16,19, and 20, with methoxyl groups in the C ring, and compounds 21-24, 26-27, and 29, with different substituents in the C ring, exhibited similar IC 50 values against PDE9.
Enantiopure drugs are very important in pharmaceuticals because different enantiomers of a chiral drug can bind different target receptors or enzymes. The results show that among the two pairs of the enantiomers, 15 vs 18 or 25 vs 26, the (R)-configurations resulted in higher activities than the (S)-enantiomers. This result was consistent with earlier reports 20,26,27 . CoMFA statistical studies for PDE9 inhibitors. The comparative molecular field analysis (CoMFA) 43,44 method was performed to determine the quantitative relationship between the structures and the IC 50 values toward PDE9 . The statistical parameters for the CoMFA models are shown in SI 1 (Supplementary information). Based on the IC 50 values, the CoMFA results generated a reasonable/acceptable model (q 2 = 0.554 and r 2 = 0.996) at optimal component six, which implied that the steric and electrostatic fields in this CoMFA model  were sufficient to explain the inhibitory effects of the target compounds described in Table 1. The contour maps (SI 1), graphically converted from the resulting CoMFA model, can offer valuable insights into the intermolecular interactions between these compounds and their receptor, which may be helpful in the rational design of PDE9 inhibitors.
Considering the balance between the PDE9 inhibitory activity and the antioxidant capacity, we chose compounds 16 and 22, which showed excellent performance in both assays, for subsequent evaluation of their specific affinity toward PDE families. The results, shown in Binding pattern of inhibitor 16 with PDE9. As assessed from the results of molecular docking experiments, inhibitor 16 showed a PDE9 binding pattern similar to that of our previously reported inhibitors 3r (SI 2) and 28s. Its pyrazolopyrimidinone ring formed two hydrogen bonds, 2.9 Å and 3.3 Å (relatively weak, not shown in SI 2), with the invariant Gln453 of PDE9 and was involved in aromatic π -stacking interactions with Phe456. These are two characteristic interactions of PDE9 inhibitors (3r and 28s) 20,46 with PDE9. Interestingly, the newly introduced amine N2 atom of 16 made a hydrogen bond, 3.0 Å with the side chain of the unique Tyr424 in PDE9, which may explain its 500-fold better PDE9 selectivity over PDE1 (Table 2). Similarly, in the crystal structure of PDE9 complexed with 28s, Tyr424 formed a hydrogen bond with the amide oxygen of L-Ala of 28s 20 , and changes to the nitrogen adjacent to D-Ala of 3r have also been observed 46 . As expected, compound 3r showed a more negative docking score of −51.2 kcal/mol (CDOCKER-INTERACTION-ENERGY) than 16, which showed a score of − 47.7 kcal/mol. This result is in accordance with the inhibitory effects of the two compounds (0.6 nM and 34 nM).
Lipid-water distribution coefficient and blood-brain barrier permeability in vitro. The 1-noctanol/water system was used to estimate the lipid-water distribution coefficient (logP) values 47 . These values were 1.10 and 1.50, respectively for compounds 16  Blood-brain barrier (BBB) permeability is another important feature of the drugs used in the treatment of CNS diseases. We measured the BBB permeability of 16 by using the parallel artificial membrane permeation assay 10,47,48 of the blood-brain barrier (PAMPA-BBB). The permeability values of 13 selected commercial drugs were compared with reported values to validate the assay (SI 4 in Supplementary information). The experimental data versus the reported values exhibited an excellent linear correlation:    Metal-Chelating properties of 16. The ability of 16 to chelate bio-metals was studied by UV-vis spectrophotometry 31,49,50 . The spectral pattern of 16 with or without metal ions such as Cu 2+ , Fe 2+ , Fe 3+ , and Zn 2+ are shown in Fig. 3. The pink line is the UV-vis spectrum of 16 between 200-600 nm; this spectrum showed two absorption peaks at 233 and 259 nm, respectively. After the incubation of 16 with Cu 2+ , the second peak shifted from 259 to 273 nm and a new peak appeared at 410 nm. Similar results were obtained upon incubation of 16 with Fe 2+ or Zn 2+ . For example, the absorption peak shifted from 259 nm to 271 nm, and the optical intensity increased markedly after Fe 2+ was added to the solution of 16. These changes in absorbance indicated the formation of 16-Metal ion (II) complex. The results of the UV-vis spectrophotometry assay showed that 16 failed to chelate Fe 3+ effectively. However, this result requires further confirmation.
To evaluate the stoichiometry of the 16-Cu 2+ complex, a series of UV-vis spectrophotometry assays of 16 titrated against Cu 2+ were performed. The final concentration of 16 was maintained at 40 μ M, and the absorption spectra were recorded after different concentrations of Cu 2+ were added. The stoichiometry of the 16-Cu2 + complex was evaluated by determining the changes in absorbance at 410 nm, where a new band had appeared (Fig. 3C). As shown in Fig. 3B, the absorption increased with an increase in Cu 2+ concentration and reached a plateau at approximately 40 μ M, which indicated that the stoichiometry of 16-Cu 2+ complex was 1:1.
The ability of 16 to halt copper redox cycling via metal complexation. Several bio-metals, especially the redox-active Cu 2+ , are involved in oxidative stress, which triggers neuronal cell death as seen in AD 31,51 .
To evaluate the ability of the target compounds to halt the copper redox cycling via metal complexation under aerobic conditions, the Cu-ascorbate redox system was used as a model (Fig. 4) 31,52 . Fluorescent 7-hydroxyl-CCA, which was produced from coumarin-3-carboxylic acid (CCA) in the presence of hydroxyl radicals (OH·), was used to measure the reduction of hydroxyl radicals during the copper redox-cycling in the presence of ascorbate. As shown in Fig. 4B, the fluorescence intensity increased linearly for the first 12 min and then reached a plateau at 15 min. This process was fully inhibited when 16 was co-incubated with the Cu-ascorbate system, indicating that 16 had the capacity to halt the copper redox cycling by chelating the metal ions.
The result of the TEM analysis of the Aβ species was also consistent with that of ThT fluorescence assay. The TEM assay showed that the Cu 2+ -treated sample of fresh Aβ produced more fibrils than did the non-treated sample (Fig. 5a,b). When compound CQ or 16 and Cu 2+ were incubated with Aβ, fewer Aβ fibrils were detected (Fig. 5c,d). When compound 16 was added to the samples, fewer fibrils were observed than in the presence of CQ.
Disaggregation of Cu 2+ -induced Aβ 1-42 aggregation fibrils. The ability of 16 to disaggregate the preformed Aβ 1-42 fibrils was also studied (Fig. 6) by using reported methods 31,37 . First, fresh Aβ samples were incubated with Cu 2+ at 37 °C for 24 h to obtain the Aβ fibrils. Then, 16 and CQ were added separately and incubated for an additional 24 h. The ThT binding assay showed that 16 and CQ markedly lowered the fluorescence intensity (16: 64.6% disaggregation; CQ: 50.9% disaggregation).
These results were also confirmed by the TEM assay. The incubation of Aβ 1-42 in the presence of Cu 2+ at 37 °C for 24 h produced well-defined Aβ fibrils (Fig. 6a). Notably, as assessed by TEM, incubation of the preformed fibrils with 16 or CQ for 24 h drastically reduced the amount of Aβ fibrils (Fig. 6Bc, 16; Fig. 6Bb, CQ).

Cell viability and intracellular antioxidant activity of 16. The antioxidant activity of 16 in SH-SY5Y
cells was evaluated by using the cell-permeable dichlorofluorescein diacetate (DCFH-DA) as an indicator of ROS 29,37,45 . Trolox, a Vitamin E analog, was used as the positive control. First, the cytotoxicity of 16 toward the SH-SY5Y cells were determined by the colorimetric MTT assay. The results (SI 5) showed that 16 had nearly no toxicity below the 10 μ M concentration. As shown in SI 6, the intracellular oxidative stress increased significantly after the treatment of SH-SY5Y cells with tert-butyl hydroperoxide, which resulted in the appearance of fluorescence (versus control). When the SH-SY5Y cells were incubated with tert-butyl hydroperoxide and the antioxidants (Trolox or 16), the fluorescence intensities decreased by varying degrees, confirming their antioxidant activities. As shown in SI 5, 16 showed superior antioxidant activity to that of Trolox, even at relatively low concentrations (for example, 5 μ M of 16 vs 10 μ M of Trolox).

Conclusion
In summary, a new series of multifunctional agents were designed and synthesized for the treatment of AD. These compounds combined the pharmacophores of PDE9 inhibitors and the bio-metal chelators. Among these compounds, 16 exhibited multivalent activities, such as an excellent inhibitory affinity of 34 nM towards PDE9, an antioxidant activity of 4.47 ORAC-FL units, significant inhibition of Cu 2+ -induced Aβ aggregation, and disaggregation of Aβ fibrils formed upon the treatment of Aβ with Cu 2+ . Moreover, our results showed that 16 is likely to cross the blood-brain barrier. All these properties suggest its potential as a compound for treatment of AD. Further investigations on candidate compounds are in progress.

Methods
General. All reagents used in reactions were obtained commercially and were used without further purification unless otherwise specified. Flash column chromatography was performed with silica gel (200-300 mesh) purchased from Qingdao Haiyang Chemical Co. Ltd. The mass spectra were recorded on an Agilent LC-MS 6120 instrument equipped with an ESI mass selective detector in positive ion mode. Melting points were determined on an SRS-Opti Melt automated melting point instrument. The NMR spectra were acquired on a Bruker Avance III spectrometer with TMS (Tetramethylsilane) as the internal standard. The purity (> 95%) of the samples was determined by high-performance liquid chromatography (HPLC) with a TC-C 18 column (4.6 × 250 mm, 5 μ m) and acetonitrile/water as mobile phase at a flow rate of 1.0 mL/min.

Docking methods.
To identify the binding pattern of 16 with PDE9, the CDOCKER docking method of the Accelrys Discovery Studio 2.5.5 software was used. The crystal structure of the catalytic domain of human PDE9 complexed with 3r (PDB code: 4QGE 46 ) was used for the docking studies. The water molecules in the crystal structure were removed, except those coordinated with the two metal ions Zn 2+ and Mg 2+ . Hydrogen atoms and charges were added to the receptor/ligand systems by using the CHARMm force field and the Momany-rone partial charge method, respectively. All ionizable residues in the systems were set to their protonation states at a neutral pH. The bound 3r was used as a reference chemical to define the active site of PDE9. The radius of the input site sphere was set as 9 Å from the center of the binding site, and 50 random conformations were generated for each ligand. Other docking parameters were set to default values unless otherwise specified. Before the docking procedures, the bound ligand 3r in 4QGE was redocked back to the same PDE9 enzyme, using different docking conditions and scoring parameters to assess the reliability of the CDOCKER method. In general, the docking may be considered successful if the RMSD (root mean square deviation) value of the optimum position is not more than a given threshold of 1.0 Å from the crystal structure after cluster analysis. As a result, the 25 positions of 3r with the top docking scores had a mean of 0.92 Å for their RMSDs, which suggested that CDOCKER was suitable for use with the PDE9 system. Therefore, the same docking procedures were applied to 16 to generate its binding pattern with PDE9.
Study of metal-chelating capacity. The experiments were performed according to previously reported methods 37 . The metal chelation was monitored spectrophotometrically using a UV-vis spectrophotometer.
Typically, a solution of compound 16 (40 μ M, final concentration) alone or 16 in the presence of CuSO 4 , FeSO 4 , Fe 2 (SO4) 3 , or ZnCl 2 (40 μ M, final concentration) in 30% (v/v) ethanol/buffer (20 mM HEPES, 150 mM NaCl, pH 7.4) was allowed to stand at room temperature for 30 min, and then the absorption spectrum was recorded at room temperature. The stoichiometry of the compound-Cu 2+ complex was determined from molar ratio method as follows: compound 16 (40 μ M, final concentration) was incubated with different concentration of CuSO 4 (range from 0 to 57 μ M), and the absorption spectra of the solutions were recorded after 30 min. The blank contained 30% (v/v) ethanol/buffer instead of Cu 2+ . The normalized absorbance of the newly formed absorption peak at 410 nm was plotted against the molar concentration of Cu 2+ . The breakpoint revealed the stoichiometry of the compound-Cu 2+ complex.
The determination of the lipid-water distribution coefficient. The distribution coefficients were determined by using the shake-flask method in 1-octanol/water system 47 . After shaking the tested compounds in 1-octanol/water (1:1) solution for 30 min, the distribution of the compounds in 1-octanol phase and water phase was determined by HPLC analysis. The lipid-water distribution coefficient was calculated according the following equation (eq. 3): where C o is the concentration of the test compound in water and C w is its concentration in 1-octanol. The logP are mean values from at least three independent tests.
In Vitro Blood-Brain Barrier Permeation Assay. The blood-brain barrier penetration capacity of the compounds was evaluated by using the parallel artificial membrane permeation assay (PAMPA). The drugs were purchased from Sigma and Alfa Aesar 10,47,48 . The porcine brain lipid (PBL) was obtained from Avanti Polar Lipids. The donor microplate (PVDF membrane, pore size of 0.45 nm) and the acceptor microplate were from Millipore. The acceptor 96-well microplate (COSTAR) was filled with 300 μ L of a PBS/EtOH mixture (7:3) and the filter membrane was impregnated with 4 μ L of PBL in dodecane (20 mg/mL). The compounds were dissolved in DMSO at a concentration of 5 mg/mL and diluted with the PBS/EtOH mixture (7:3) to a concentration of 100 μ g /mL. Then, 200 μ L of this solution was added to the donor wells and the wells were carefully placed on the acceptor plate, which was then incubated for 10 h at 25 °C in a vibrationless environment. After the incubation, the donor plate was removed and the concentration of compounds in the acceptor wells was determined with a UV plate reader (Flexstation 3). Each sample was analyzed at five wavelengths in four wells, and at least three independent runs were performed. The results are expressed as the mean ± standard deviation. In these experiments, 13 quality control standards of known BBB permeability were included to validate the analysis set. The P e was calculated from the following equation (eq. 4) as reported by Faller et al. and Sugano et al. For the TEM assay of copper (II)-induced Aβ 1-42 aggregation and the disaggregation of the fibrils, the samples were pretreated as described for the ThT assay. Samples (10 μ L) were placed on a carbon-coated copper/rhodium grid for 2 min. Then, each grid was stained with uranyl acetate (1%, 5 μ L) for 2 min. After the excess staining solution was drained off, the specimen was transferred for imaging by a transmission electron microscope (JEOL JEM-1400).
Cell Culture. Cell culture was performed as reported previously 37 . The human neuron-like cell line SH-SY5Y was obtained from the Institute of Biochemistry and Cell Biology, Shanghai Institute for Biological Sciences (Shanghai, China). The cells were cultured at 37 °C in a humidified atmosphere of 5% CO 2 in Dulbecco's modified Eagle's medium (DMEM, GIBCO) supplemented with 10% fetal calf serum (FCS, GIBCO), 1 mM glutamine, 100 IU/mL penicillin, and 100 μ g/mL streptomycin.

Determination of Cytotoxicity.
Cytotoxicity was determined as reported previously with the colorimetric MTT [3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2Htetrazolium bromide] assay 37 . The SH-SY5Y cells were seeded at a density of 5 × 10 3 cells/well in 96-well plates. After 24 h, the culture medium was replaced with medium containing the tested compound at different concentrations at 37 °C. After culturing for 48 h, 100 μ L of medium containing 0.5 mg/mL MTT was added to each well. The cell were then incubated for 4 h at 37 °C in the dark. The solution was then gently aspirated from each well and the formazan crystals formed were dissolved with 100 μ L of DMSO. The optical density of this solution was measured at 570 nm, and the cell viabilities were expressed as a percentage relative to the vehicle-treated control (0.5% DMSO was added to untreated cells).
Antioxidant Activity in SH-SY5Y Cells. The antioxidant activity was determined as reported previously 37 .
The SH-SY5Y cells were seeded at a density of 1 × 10 4 cells/well in a 96-well plate. After 24 h, the culture medium was replaced with medium containing tested compounds, and the cells were cultured for an additional 24 h. Then, the cells were washed with PBS and incubated with 5 μ M DCFH-DA (diluted by PBS) at 37 °C for 30 min. After discarding the solution and washing with PBS, the cells were treated with 0.1 mM t-BuOOH (a compound induce oxidative stress, diluted by PBS) for 30 min in the dark. Then, the fluorescence of the cells in each well was measured (λ excitation = 485 nm, λ emission = 535 nm) with a multifunctional microplate reader (Flex Station 3). The antioxidant activity was expressed as a percentage relative to that of the control cells and calculated using the formula (F t -F nt )/(F t ' -F nt ) × 100, where F t is the fluorescence value of the cells treated with the tested compound, F t ' is the fluorescence value of the cells not treated with the tested compound, and F nt is the fluorescence value of the cells treated with t-BuOOH. Statistical Analysis. The experimental results are expressed as the mean ± standard deviation of at least three independent measurements. The data were subjected to Student's t-test or one-way analysis of variance (ANOVA), followed by Dunnett's test. p values ≤ 0.05 were considered statistically significant.