β-Site amyloid precursor protein-cleaving enzyme 1 (BACE1) is considered to be a promising target for treating Alzheimer's disease. However, all clinical BACE1 inhibitors have failed due to lack of efficacy, and some have even led to cognitive worsening. Recent evidence points to the importance of avoiding BACE2 inhibition along with careful dose titration. In this study, we focused on the fact that the 10s loop lining the S3 pocket in BACE1 can form both "open (up)" and "closed (down)" conformations, whereas in BACE2, it prefers to adopt a "closed" form; thus, more space is available in BACE1. By leveraging the difference, we designed fused pyridine analogues that could reach the 10s loop, leading to 6 with high selectivity and significant Aβ reduction. The cocrystal structures confirmed that 6 significantly increased B-factors of the 10s loop in BACE2 relative to those in BACE1. Thus, the destabilization of BACE2 seems to offer structural insights into the reduced BACE2 potency of 6, explaining the significant improvement in BACE1 selectivity.