In an attempt to increase selectivity vs Cathepsin D (CatD) in our BACE1 program, a series of 1,3,4,4a,10,10a-hexahydropyrano[4,3-b]chromene analogues was developed. Three different Asp-binding moieties were examined: spirocyclic acyl guanidines, aminooxazolines, and aminothiazolines in order to modulate potency, selectivity, efflux, and permeability. Using structure-based design, substitutions to improve binding to both the S3 and S2' sites of BACE1 were explored. An acyl guanidine moiety provided the most potent analogues. These compounds demonstrated 10-420 fold selectivity for BACE1 vs CatD, and were highly potent in a cell assay measuring Aβ1-40 production (5-99 nM). They also suffered from high efflux. Despite this undesirable property, two of the acyl guanidines achieved free brain concentrations (Cfree,brain) in a guinea pig PD model sufficient to cover their cell IC50s. Moreover, a significant reduction of Aβ1-40 in guinea pig, rat, and cyno CSF (58%, 53%, and 63%, respectively) was observed for compound 62.