This is one of the most interesting papers on γ-secretase modulation I have seen in a while. Applying sophisticated NMR experiments and subtle detergent/C99 preparations, it describes effects on NSAID aggregation at high concentrations. The paper contradicts findings in Kukar et al., 2008.

The authors are careful in the interpretation of their data by saying that their results apply to mono-disperse C99 preparations. This, in turn, poses more questions than answers, like every good experiment does. For example, do mono-disperse C99 preparations of up to one micromolar concentration provide a reliable model for γ-secretase modulation? The mono-dispersity of the preparation challenges the findings on substrate dimerization by Gerd Multhaup and colleagues (Kaden et al., 2008, who write, “The presence of SDS was sufficient to convert native APP dimers entirely into monomers”). It also challenges recent experiments by Eddie Koo et al. (Eggert et...  Read more

This is one of the most interesting papers on γ-secretase modulation I have seen in a while. Applying sophisticated NMR experiments and subtle detergent/C99 preparations, it describes effects on NSAID aggregation at high concentrations. The paper contradicts findings in Kukar et al., 2008.

The authors are careful in the interpretation of their data by saying that their results apply to mono-disperse C99 preparations. This, in turn, poses more questions than answers, like every good experiment does. For example, do mono-disperse C99 preparations of up to one micromolar concentration provide a reliable model for γ-secretase modulation? The mono-dispersity of the preparation challenges the findings on substrate dimerization by Gerd Multhaup and colleagues (Kaden et al., 2008, who write, “The presence of SDS was sufficient to convert native APP dimers entirely into monomers”). It also challenges recent experiments by Eddie Koo et al. (Eggert et al., 2009). Does any GXXXG-derived substrate dimerization occur in the NMR tube? Is the detergent LMPG a good model for lipid raft-like environments?

However, these same arguments hold true for the findings in Kukar et al.: did the CHAPSO concentration of 0.25 percent provide a reliable environment for this complex process? Did the preparation include sufficient amounts of dimeric or even monomeric substrate? The supplementary Figure 4b of Kukar et al. does indicate the presence of such dimers, but not the high-molecular-weight aggregates that Charles Sanders (co-author of this paper) postulates to be dominant at 0.25 CHAPSO concentrations. Furthermore, Kukar et al. labeled the substrate at one micromolar concentration, whereas Sanders and John Jordan investigated the substrate at concentration of five to 250 micromolar and 25 to 2,000 micromolar of flurbiprofen. This naturally requires higher amounts of solubilizing agent and may thus lead to different findings.

NSAID concentrations exceeding 200 micromolar may interfere with membrane and protein integrity. For instance, flurbiprofen accumulates within the palisade layer of Tween 20 micelles (Saveyn et al., 2009). This effect is not limited to NSAIDs or GSMs; it is frequently observed for low-potency and poorly soluble kinase inhibitors. Flurbiprofen has an intrinsic solubility in water of just 40 micromolar (Li and Zhao, 2003). Therefore, more potent and at least equally soluble GSMs (GSM-1 IC/EC50 <200 nM, C8-Carprofen IC/EC50 <10 uM) should be investigated in this setting.

In short, this paper presents very nice data and raises many questions to spark debate.

View all comments by Boris Schmidt

Strong evidence has been provided that γ-secretase modulators (GSMs) directly affect enzyme activity by the demonstration that GSMs selectively lower Aβ42 production and increase shorter Aβ species in cell-free γ-secretase assays (Takahashi et al., 2003; Weggen et al., 2003). However, surprisingly, and in contrast to the findings with γ-secretase inhibitors, the primary binding site of GSMs has been reported to reside in the substrate APP and not in presenilin (PSEN) or one of the three accessory subunits that form the γ-secretase enzyme complex (Kukar et al., 2008). This new NMR study by Beel et al. now challenges the specificity of the observed interaction between GSMs and APP.

In the earlier study by Kukar et al., photo-activatable derivatives of GSMs based on the GSM flurbiprofen and the inverse GSM fenofibrate incorporating benzophenone as a photo-active moiety were synthesized and employed for biochemical labeling studies. Given the...  Read more

Strong evidence has been provided that γ-secretase modulators (GSMs) directly affect enzyme activity by the demonstration that GSMs selectively lower Aβ42 production and increase shorter Aβ species in cell-free γ-secretase assays (Takahashi et al., 2003; Weggen et al., 2003). However, surprisingly, and in contrast to the findings with γ-secretase inhibitors, the primary binding site of GSMs has been reported to reside in the substrate APP and not in presenilin (PSEN) or one of the three accessory subunits that form the γ-secretase enzyme complex (Kukar et al., 2008). This new NMR study by Beel et al. now challenges the specificity of the observed interaction between GSMs and APP.

In the earlier study by Kukar et al., photo-activatable derivatives of GSMs based on the GSM flurbiprofen and the inverse GSM fenofibrate incorporating benzophenone as a photo-active moiety were synthesized and employed for biochemical labeling studies. Given the low potency of the parent compounds, the photo-probes were applied in high concentrations (10-100 μM). Photolysis in the presence of partially purified γ-secretase demonstrated that none of the four γ-secretase subunits was labeled by the photo-probes. In contrast, a purified recombinant APP substrate consisting of the last 100 C-terminal amino acids of APP (APP-CTF) was readily labeled by the photo-probes, and binding was competed by the parent compounds and other NSAID-type GSMs such as sulindac sulfide and indomethacin. In addition, labeling of APP-CTFs and full-length APP was achieved in the presence of crude membrane preparations from cells overexpressing APP. Mapping of the binding region of the photo-activatable GSMs within APP demonstrated binding to residues 29-36 of the Aβ domain.

However, several earlier observations do not easily conform to the concept of substrate targeting GSMs.

1. Kinetic studies have demonstrated a mode of non-competitive inhibition of Aβ42 production for GSMs in γ-secretase in vitro assays (Takahashi et al., 2003; Beher et al., 2004). This indicates that the inhibitory effect of GSMs on Aβ42 production cannot be overcome by increasing substrate concentrations, which would be expected for compounds with a primary binding site within the substrate.

2. While GSMs do not impair γ-secretase cleavage close to the cytosolic membrane border and the release of intracellular signaling domains from several γ-secretase substrates in cell-based and in vitro assays, it has been convincingly demonstrated that GSMs do modulate γ-secretase cleavage events in the middle of the Notch-1 TMD (S4-cleavage), which are analogous to the cleavage events in the APP TMD that generate the Aβ peptides (Okochi et al., 2006).

3. Signal peptide peptidase (SPP), an intramembrane protease homologous to PSEN, has been demonstrated to harbor a binding site for GSMs, and high concentrations (500 μM) of the GSMs sulindac sulfide and indomethacin were able to shift the major cleavage site of a recombinant SPP substrate under cell-free assay conditions (Sato et al., 2006; Iben et al., 2007). However, in contrast to γ-secretase, SPP functions without accessory proteins, and γ-secretase and SPP substrates do not show any overlap.

In the new study, the authors investigated the NMR complexation-induced chemical shift (CIS) of several GSMs on APP-CTFs. CIS analysis is an extensively used method in structure-based drug discovery that relies on the measurement of the NMR chemical shift perturbations of the target upon interaction with a ligand, the largest perturbations occurring at the binding site due to changes in the chemical environment. The presented data suggest that the examined GSMs are non-specific binders of the APP-CTF peptide, supported by negligible 15N chemical shifts with fenofibrate and sulindac sulfide, and only very small shifts with flurbiprofen and indomethacin. Although CIS is usually applied qualitatively, the observed CIS magnitudes for fenofibrate and sulindac sulfide indeed reflect non-specific binding, whereas flurbiprofen and indomethacin showed a minor unsaturable effect on three specific APP residues (689V, 691F and 739D), which are distinct from the binding site proposed by Kukar et al. Although the tested GSMs have only moderate potencies in vitro with IC50s in the high micromolar range, a more pronounced CIS effect would have been expected over the broad range of molar equivalencies used in the study. Moreover, the absence of 19F chemical shift for flurbiprofen even in the presence of large excess of the APP-CTF peptide is in line with the 15N-CIS observations.

Based on these results, the authors suggest that the target of the compounds might be an aggregated form of the peptide due to the experimental procedure. The precise nature of those aggregates has not been identified; however, the presence of typical amyloid structures might indeed be relevant in the assay, given that membranes/micelles have been reported elsewhere to have the ability to catalyze the formation of amyloid aggregates (Evers et al., 2009). Moreover, many GSMs share common structural features with known aggregation inhibitors, such as extended conjugated aromatics and anionic moieties that are characteristic of binders to hydrophobic patches. Indeed, some GSMs have been reported as weak Aβ aggregation inhibitors, and vice versa; the study by Kukar et al. identified potent aggregate binders such as X-34 and chrysamine G as moderate GSMs. Of note, the benzophenone cross-linking moiety used in the study by Kukar et al. also shows strong structural similarity with fenofibrate and, in addition, with potent amyloid binders in the anthraquinone family. Hence, the cross-activities in γ-secretase modulation and/or APP-binding might only reflect the high hydrophobicity of the putative targets and are not mutually exclusive in view of the available data.

The literature offers very little precedent for small-molecule inhibitors targeting enzyme substrates. This is most likely due to the difficulties in developing high-affinity compounds to small and linear epitopes with minimal 3D structure on the surface of proteins. Furthermore, the cellular concentration of a substrate is generally far in excess of its processing enzyme; this further complicates the development of potent compounds, as higher drug concentrations are required to inhibit an enzymatic reaction by targeting the substrate. In this respect, the concept of GSMs as substrate targeting protease modulators does have far-reaching implications for drug development beyond the potential use of GSMs in AD therapeutics. Accordingly, conformation of the findings by Kukar et al. with photo-affinity ligands based on GSMs with substantially improved potency or with other experimental approaches is highly anticipated.

View all comments by Bruno Bulic
View all comments by Sascha Weggen