The SKAR Strategy.

This strategy consists of a chemical kinetics platform applied to drug discovery against misfolding diseases to (i) correlate the chemical structure of a candidate therapeutic molecule with its ability to inhibit oligomer production during the aggregation of a target protein and (ii) optimize this ability by rationally changing the chemical properties of the molecule to generate more-effective inhibitors that reduce oligomer production even further (Fig. 1). In this approach, from a specific scaffold of a parent molecule, individual chemical modifications are introduced while keeping the remainder of the molecule unchanged, thus generating a pool of derivatives with similar chemical characteristics, except for the specific groups that have been altered in this manner (Fig. 1A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Schematic illustration of the SKAR strategy. An in vitro kinetic analysis is used to determine the effects of a pool of derivatives from a particular molecular scaffold on the aggregation of the target aggregating protein, here Aβ42. (A) A pool of molecule derivatives from a particular scaffold is first collected for the study. (B) These molecules are then tested in vitro to determine quantitatively their effects on the different microscopic steps of the Aβ42 aggregation. (C) Kinetic rates related to the production of oligomers in the presence of small molecules are then extracted from the kinetic profiles, and three parameters are estimated: (i) OIC₅₀^PT (green), which is associated with a delay in the flux toward oligomers; (ii) OIC₂₅^T (pink), which is associated with a reduction in the total number of oligomers generated; and (iii) the macroscopic parameter KIC₅₀^M (purple), which is associated with the macroscopic kinetic inhibition. (D) Using these three parameters, the KIA fingerprint is constructed (SI Appendix, Table S1) to compare the different molecules and select the most potent ones. SKAR can be applied in an iterative manner to identify more potent derivatives or explore different chemical scaffolds.

The next step in the procedure is the measurement of the macroscopic aggregation kinetics of the protein in the presence of the derivatives through a ThT-based assay (23) (Fig. 1B). In this way, any change in the half-time of aggregation (t_1/2) induced by a given small molecule can be obtained (Fig. 1C). By carrying out such experiments in the presence of a range of concentrations of the small molecule, it is also possible to estimate the changes in the reactive flux toward oligomers from global fitting to the experimental curves (24) (Materials and Methods, Fig. 1C, and SI Appendix, Eq. S2). To carry out a quantitative analysis, we have adopted three specific parameters associated with the aggregation kinetics, which, taken together, correspond to the IC₅₀ parameter commonly used in SAR. These parameters are (i) the concentration of a given molecule that results in a 50% increase in the t_1/2 value of the overall aggregation reaction, denoted here as KIC₅₀^M, where KIC refers to “kinetic inhibitory concentration” and M refers to “macroscopic”; (ii) the concentration of the small molecule that results in a 50% increase in the time to reach the peak value of the generation of oligomers, denoted here as OIC₅₀^PT, where OIC refers to “oligomer inhibitory concentration” and PT refers to “peak time”; and (iii) the concentration of the small molecule that results in a 25% decrease in the total number of oligomers formed during the reaction, denoted here as OIC₂₅^T, where T refers to “total concentration of oligomers” (Materials and Methods and Fig. 1C).

More generally, we note that SKAR is an approach that enables the quantitative determination of the extent to which a small molecule affects the reactive flux toward any target species of interest in an aggregation reaction, and then to enhance its efficacy. The use of appropriate parameters corresponding to the SAR IC₅₀ enables the determination of the potency of any compound for a specific process, for example, to reduce the peak levels of oligomer formation, or to change the total quantity of fibrils generated over time.

While KIC₅₀^M reflects a macroscopic effect obtained directly from the overall aggregation process, OIC₅₀^PT and OIC₂₅^T reflect the changes induced by a small molecule on specific microscopic steps in the aggregation process. Taking these three parameters together, a unique fingerprint of a molecule is generated, which is denoted here as the “kinetic inhibitory attributes” (KIA) fingerprint. The KIA fingerprint reflects the potency of a compound in inhibiting the specific steps in the aggregation process responsible for the production of the majority of oligomeric species (Fig. 1D and SI Appendix, Table S1). Thus, the KIA fingerprint allows a comparison to be made between the potencies of different compounds in reducing the number of potentially toxic species generated during aggregation. For instance, a KIA fingerprint such as the one shown in green in Fig. 1 reflects a molecule whose major effect is on primary events in aggregation, hence generating a delay in the formation of oligomeric species, i.e., leading to an increase in the time taken to reach the peak of the reactive oligomer flux (21). On the other hand, a KIA fingerprint such as that shown in red indicates a molecule that inhibits the surface-catalyzed secondary nucleation, thus resulting in both a delay in the reactive flux toward oligomers and a reduction in the total number of oligomers formed (20). In addition, it is evident from their KIA fingerprints that both molecules show a greater potency in all three attributes compared with that of the molecule with the blue KIA fingerprint.

The detailed understanding offered by the KIA fingerprint is a key element guiding the optimization and search for drugs with the potential to combat AD, because it allows a relationship to be established between the chemical modifications introduced into a parent molecule and the resulting changes in the rate and quantity of the production of Aβ42 oligomers. With such a relationship, this process can then be used iteratively to systematically and rationally design and optimize small molecules with increasing potency. Here we show two examples where SKAR against Aβ42 has been applied to two different parent scaffolds leading to (i) an understanding of the chemical properties that are responsible for an inhibitory effect on Aβ42 aggregation and (ii) an exploitation of this understanding to increase the potency of a different parent molecule.

Lead Compound Optimization Against Aβ42 Oligomer Production.

We illustrate the SKAR strategy by its application to bexarotene, a small molecule found to inhibit strongly the nucleation of Aβ42 aggregation (21). Ten molecules bearing different substituents of R₁ (the polar moiety), R₂ (the linker group), and R₃ (the apolar moiety) were considered for this study, some of which (E, H, and G) had two of these features modified concomitantly. We generated the KIA fingerprints for all 10 molecules and compared their effects on the different microscopic steps in the process of Aβ42 aggregation (13, 24) (Fig. 2 and SI Appendix, Fig. S1).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Derivation of the SKAR rules from the analysis of the bexarotene scaffold. The 10 derivatives (A through J) of bexarotene used in this study are shown with components R₁, R₂, or R₃ in different colors. The component similar to that of the parent molecule is shown in black. (A) Six derivatives (B, D, E, F, H, and J) inhibited the aggregation of Aβ42. For each of these compounds, the time dependence of the reactive flux toward oligomers (SI Appendix, Eq. S2) for the aggregation of a 2 μM Aβ42 solution in the presence of either 1% DMSO (black) alone or with 2 μM (green), 6 μM (orange), or 10 μM (red) of the compound is shown; these reactive fluxes were estimated from the data shown in SI Appendix, Fig. S1 and literature (13). The OIC₅₀^PT, OIC₂₅^T, and KIC₅₀^M values determined from these reactive fluxes are shown for the positive compounds, i.e., with a significant effect on the aggregation kinetics. (B) Four other compounds (A, C, G, and I) did not affect the aggregation of Aβ42 significantly.

The aggregation kinetics of a 2 μM Aβ42 sample were monitored in the absence and presence of each compound at concentrations ranging from 2 μM to 10 μM [datasets for bexarotene, molecules D through H, and J, were obtained from a previous study (13)]. For four compounds (A, C, G, and I), no significant effects could be observed on the kinetics (SI Appendix, Fig. S1). For the remaining six compounds (B, D, E, F, H, and J), however, a progressive delay in the aggregation reaction could be observed in Aβ42 aggregation with increasing concentrations of the compounds (SI Appendix, Fig. S1). These six derivatives exhibited different effects on the t_1/2 of Aβ42 aggregation, each of which was estimated by determining its KIC₅₀^M value (SI Appendix, Fig. S2A).

We next carried out a semiempirical quantitative analysis of the effects of all seven positive compounds that inhibit Aβ42 aggregation by fitting the experimental aggregation profiles to a single rate law (19), thus obtaining the combined microscopic kinetic rate constants k₊k_n, which represent the overall rate constants of the primary pathways of aggregation, and k₊k₂, which represent the overall rate constants of the secondary pathways. These aggregation profiles in the presence of the inhibitors were then compared with the corresponding values evaluated in the absence of the compounds, to define the systematic perturbations of the microscopic combined rate constants in the rate law (see Materials and Methods). The results showed that the rates of both k₊k_n (primary pathways), and k₊k₂ (secondary pathways) were both decreased as a function of the concentration of small molecules (SI Appendix, Fig. S2 B and C).

To decouple the effects of the small molecules on the combined rate constants k₊k_n and k₊k₂, we carried out an additional set of experiments where Aβ42 aggregation in the presence of each molecule was monitored following the addition of 30% of preformed fibril seeds (SI Appendix, Fig. S3); under these conditions, the formation of amyloid fibrils is driven mainly by elongation processes (13). We observed that none of the molecules detectably affected the kinetics under these conditions at concentrations as high as 10 μM, indicating strongly that none affects significantly the rate of elongation of Aβ42 fibrils. Thus, the decrease in the combined rate constants obtained from the unseeded aggregation in the presence of the small molecules, k_nk₊ and k₂k₊, can be attributed solely to the decrease in k_n and k₂ (SI Appendix, Fig. S2).

A crucial aspect of the SKAR strategy is the analysis of the rate constants of the primary and secondary nucleation steps, which allows calculation of the reactive flux toward oligomers over time in the absence and presence of increasing concentrations of molecules (Materials and Methods, Fig. 2A, and SI Appendix, Fig. S4). Strikingly, the formation of Aβ42 oligomers over time was found to be both delayed and reduced by one or more of the set of all seven molecules discussed in the present study. The shift in the maximum reactive flux toward oligomers (OIC₅₀^PT, SI Appendix, Fig. S4A) and the total number of oligomers (OIC₂₅^T, SI Appendix, Fig. S4B) generated could be determined as a function of the concentration of each small molecule.

The values of KIC₅₀^M, OIC₅₀^PT, and OIC₂₅^T were then used to construct the KIA fingerprint for each compound to compare their potency in inhibiting macroscopic aggregation, delaying oligomer formation, and reducing the total number of oligomers formed (Fig. 3). The KIC₅₀^M values indicated that four molecules (J, F, D, and H) inhibited the overall rate of fibril formation in Aβ42 aggregation to a greater extent than the parent compound. In particular, molecule H, the most potent inhibitor, is three times more effective than its parent compound, bexarotene (Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

KIA fingerprint of the bexarotene scaffold. KIA fingerprints are shown for bexarotene and its positive derivatives (represented by different colors) to compare their potencies.

Furthermore, a particularly important observation is that the same four molecules (J, F, D, and H) also showed an increased potency with respect to that of bexarotene in delaying the formation of oligomers (OIC₅₀^PT), and reducing the total number of oligomers over time (OIC₂₅^T) (Fig. 3). More specifically, molecule J exhibited the same potency as bexarotene in delaying the flux toward oligomers, but exhibited an increased potency in reducing the total number of oligomers formed. On the other hand, molecule E, which is less potent than bexarotene in inhibiting the aggregation kinetics and delaying the flux toward oligomers, was slightly more effective than bexarotene in reducing the total number of oligomers generated during the reaction.

Establishment of the SKAR Rules: Identification of the Physicochemical Parameters That Lead to an Increased Potency in Reducing Oligomer Production.

Taken together, the effects described in this study provide an opportunity to identify chemical features that could account for the potency of the molecules in inhibiting the aggregation of Aβ42. They indicate, in particular, that the combination of polar and apolar (R₁ and R₃) components is essential for the activity of the parent molecule. Thus, neither molecule A nor molecule I, both of which lack the carboxylic acid group present in bexarotene, affects the aggregation of Aβ42 (Fig. 2B and SI Appendix, Fig. S1), suggesting that a polar moiety containing an acidic ionizable group is essential for inhibitory activity. In addition, molecule C, which lacks the cyclohexane group of bexarotene, is inactive, suggesting that the apolar nature of the component R₃ is also crucial for an inhibition of the aggregation process. Furthermore, although compound F, where the linker group is modified, is still able to inhibit the aggregation of Aβ42, compound G, which contains an additional polar carbonyl group in the apolar component (R₃) compared with molecule F, no longer has activity. These findings suggest that essential apolar interactions are made between the R₃ moiety of the molecule and Aβ42, such that modifications to include a polar group would eliminate this interaction.

On the basis of these results, we generated a specific pharmacophore based on the chemical structures of all of the bexarotene derivatives found to possess inhibitory activity (SI Appendix, Fig. S5) (25⇓⇓–28). The features of this pharmacophore are fully consistent with the findings discussed above, including the conclusion that the detailed nature of the linker is not essential for inhibiting Aβ42 aggregation, while the apolar and polar groups play key roles in this process.

In particular, all compounds with an altered linker component (B, D, E, F, and H) but that preserve the nature of the polar R₁ and apolar R₃ moieties of bexarotene were effective in inhibiting the aggregation of Aβ42. Minor modifications to the linker group are thus possible without losing activity. In addition, analysis of the physicochemical properties of the linker components (SI Appendix, Table S2) did not reveal any correlation between the potency of the compound and the overall polarity, length, or flexibility of the linker region (29). These results suggest that the R₂ component is not responsible for forging any critical interactions required for the inhibition of Aβ42 aggregation. We therefore infer that the linker is primarily responsible for orientating the polar and apolar components in the optimal positions. This conclusion is also consistent with our pharmacophore model (SI Appendix, Fig. S5). We do note, however, that molecules F and J have a longer linker length compared with bexarotene. This greater distance, which could allow a higher degree of flexibility of binding for components R₁ and R₃, could account for the higher inhibitory potency observed in molecules F and J compared with bexarotene (SI Appendix, Table S2).

As we compared the apolarity of the R₃ component between bexarotene and molecules E and H, we found that the apolarity of the R₃ component is correlated with the potency of the associated scaffold (SI Appendix, Table S3) (30). In particular, when comparing molecules H and F, we found that an extended apolar structure results in an overall more pronounced KIA fingerprint.

Implementation of the SKAR Rules: Generation of Active Derivatives from an Inactive Parent Compound.

To illustrate how the SKAR rules derived above could be applied in a rational drug discovery program directed toward AD, we applied them to a scaffold different from that of bexarotene, which contains a central rhodanine functional group. Rhodanine-based compounds are frequently used in medicinal chemistry, since they are well tolerated and exhibit a variety of pharmacological activities (31). In particular, rhodanine-based compounds have been found to inhibit the aggregation of tau, and thus have been suggested, in a different context, as potential drugs against AD (31). The central molecule of the present SKAR procedure is molecule K, which has three chemical components: polar R₁ and R₃ components and an R₂ linker component. When we monitored the aggregation of a 2-μM sample of Aβ42 in the presence of the parent molecule K, no effect of the compound could be observed (Fig. 4A and SI Appendix, Fig. S6).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Application of the SKAR rules to the development of a rhodanine-based active compound. The eight derivatives of the rhodanine-based compound K (L through S) used in this study are shown with components R₁, R₂, or R₃ in different colors. The component that is similar to that of the parent molecule is shown in black. (A) Four compounds (M, N, O, and R) inhibited the aggregation of Aβ42. For each of these compounds, the time dependence of the reactive flux toward oligomers (SI Appendix, Eq. S2) for the aggregation of a 2 μM Aβ42 solution in the presence of either 1% DMSO alone (black) or with 2 μM (green), 6 μM (orange), 10 μM (red), or 20 μM (blue) of the compound is shown; these reactive fluxes were estimated from the data shown in SI Appendix, Fig. S6. The OIC₅₀^PT, OIC₂₅^T, and KIC₅₀^M values determined from these reactive fluxes are shown for the positive compounds. (B) Four other compounds (L, P, Q, and S) did not affect the aggregation of Aβ42 significantly.

Molecule K, however, lacks the dual polar−apolar nature of the R₁ and R₃ moieties in the pharmacophore, which, in the study of the bexarotene derivatives discussed above, was judged as essential for the activity of the compound. We therefore studied the effects of a set of derivatives of molecule K (M, N, O, and R), which we modified to include such polar and apolar components, on the aggregation kinetics of Aβ42 with a concentration range between 2 μM and 20 μM. As expected from the conclusions above, all four compounds were found to inhibit the aggregation of Aβ42; further studies showed that this inhibition showed a clear concentration dependence (Fig. 4A and SI Appendix, Fig. S6).

As negative controls, we considered molecules P and Q, in which, like the parent molecule K, both R₁ and R₃ moieties are polar, and neither was found to have a detectable effect on the aggregation of Aβ42. As a further negative control, molecule L, which possesses an apolar R₃ moiety but an R₁ component of lower polarity than in compound K, again showed no significant effect on Aβ42 aggregation, thus further supporting the importance of the dual apolar−polar character of the pharmacophore. Along similar lines, molecules M, N, O, and R showed no effect on Aβ42 aggregation under seeding conditions (SI Appendix, Fig. S7), indicating that the elongation process of Aβ42 was not affected by these molecules.

Since extending the apolar nature of the aggregation inhibitors discussed above was found to increase their potency, we sought to assess whether or not extending the polar nature could have a similar effect on the compounds based on molecule K. We therefore studied Aβ42 aggregation in the presence of molecule S, in which a primary amine extension is added. We found that the activity of the molecule was essentially abolished (Fig. 4B and SI Appendix, Fig. S6 and Table S4) (30), although we note that this lack of activity might be also be due to the fact that the R₁ component was not apolar enough.

As observed for bexarotene, these results indicate that an extended apolar nature leads to a higher potency of the molecule in inhibiting Aβ42 aggregation. Indeed, molecules N and R, which both have extended apolar components compared with the positive molecule O (with R having the most apolar component), are able to delay oligomer formation, as well as inhibit more effectively the overall oligomer production (Fig. 4 and SI Appendix, Fig. S8 and Table S5) (30). The KIA fingerprints of the positive molecules were generated as previously (SI Appendix, Fig. S8) and showed that the molecules inhibit the aggregation kinetics of Aβ42 with overall potencies in the order R > N > M > O (Fig. 5).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

KIA fingerprint of the rhodanine-based compounds. KIA fingerprints are shown for the positive rhodanine-based compounds (represented by different colors) to compare their potencies.