a, Structures of the parent GSMs (fenofibrate and tarenflurbil) and photoprobe derivates (Fen-B and Flurbi-BpB) are shown. b, The absence of PSEN1, NCSTN, APH1 and PEN2 labelling by the GSM Fen-B in a purified γ-secretase preparation (from CHO γ-30 cells12) and immunoprecipitation with streptavidin. The ratios of sample relative to the starting material are shown. Start and unbound lanes contain 5% of the immunoprecipitated material (lane 3), therefore the ratios are 1, 1 and 20. Asterisk denotes nonspecific reactivity with streptavidin. c, GSM photoprobes (Flurbi-BpB, closed circles, and Fen-B, open triangles) label a recombinant APP γ-secretase substrate (APP(C100)–Flag) with similar potency. A, absorbance; data are mean ± s.e.m., n = 2. d, Labelling of APP(C100)–Flag by Fen-B (10 μM) is competed by Aβ42-lowering and -raising GSMs (100 μM) but not by sulindac sulphone, a non-GSM NSAID. Data are presented as percentage control ±s.e.m., n = 2. Asterisk, P <0.05; two asterisks, P <0.01; ANOVA with Dunnett’s post-hoc analysis. e, GSM photoprobes label APP CTF from cells. CHAPSO solubilized membrane fractions from H4 APP-CTF–alkaline phosphatase cells were crosslinked with Fen-B and Flurbi-BpB (50 μM) and analysed by immunoprecipitation with streptavidin and immunblotting for APP (antibody CT20). Both GSMs label a fragment of APP that co-migrates with APP(C83). UV, ultraviolet. f, A GSM photoprobe preferentially labels a recombinant APP substrate (APP(C100)–Flag; left panel) relative to Notch (Notch(C100)–Flag; right panel). Samples were analysed by western blotting for incorporation of Fen-B. Green, biotin; red, Flag; yellow, dual reactivity; LiCor Odyssey.

a, A cell-based screen of Aβ-binders identified molecules that increase Aβ42 (DAPH) or decrease Aβ42 (Bis-ANS, X-34 or chrysamine G (CG)). Data are mean ± s.e.m., n = 3. b, X-34 is an Aβ42-lowering GSM. Changes in amyloid-β peptide amounts after X-34 treatment are shown. Data are presented as percentage control ±s.e.m., n = 3. EC₅₀ values were calculated as described in Methods. Total amyloid-β did not decrease. NA, not applicable. c, X-34 binds to APP(C100)–Flag and decreases labelling by GSM photoprobes. Biotin incorporation into APP by Fen-B and Flurbi-BpB is presented as percentage of control (peptide without X-34) ±s.e.m., n = 2. d, X-34 lowers Aβ42 in Tg2576 mice after 4 h. X-34 (n = 7) and tarenflurbil (n = 5) reduce Aβ42 selectively; control (n = 7). Data are presented as amyloid-β percentage of control relative to vehicle ±s.e.m. Animals per group (n). Asterisk, P <0.05; two asterisks, P <0.01; ANOVA with Dunnett’s post-hoc analysis.

a, Fen-B labels Aβ1–40 and Aβ1–36 but not Aβ1–28, suggesting that the binding site for Fen-B is located between residues 28 to 36 of amyloid-β, which are highlighted in italics. b, Flurbi-BpB and Fen-B label Aβ1–36 (biotin incorporation), whereas the photoaffinity tag alone (BpB) shows minor labelling. c, Flurbi-BpB and Fen-B preferentially label Flag-tagged Aβ25–36. Data are presented as biotin incorporation (absorbance, A) ±s.e.m., n = 3. d, The peptide fragment I1 (NH₂-FEGKF-CONH₂) increases Aβ42 in H4 cells expressing APP similar to the GSM fenofibrate.

a, b, The sequences of wild-type (WT; a) APP and the mutated substrate (b) containing the homologous region (italic, underlined) of the NOTCH transmembrane domain (TMD) in APP where GSMs are hypothesized to bind. Top row, X-34 lowers Aβ42 (EC₅₀ 5.9 μM) from APP wild-type cells (a) but did not change either Aβ40 or Aβ42 concentrations in the APP-NOTCH TMD line (b). Bottom row, FT-1 (12.5 μM) raised Aβ42 200% in APP wild-type cells (a); however, in APP-NOTCH TMD cells (b), FT-1 caused minimal changes in the Aβ42 (95%) or Aβ40 (90%) signal. Data are presented as amyloid-β percentage of control ±s.e.m., n= 3.