The objective of this study was to evaluate the utility of brain tissue slices to determine the effect of plasma and brain tissue nonspecific binding on the brain-to-plasma ratio (K(p)). Mouse or rat brain slices (400 microm) were prepared using a McIlwain tissue chopper (Surrey, UK) and incubated with 1 microg/ml of compound at 37 degrees C either in a physiological buffer to determine the buffer-to-slice concentration ratio, i.e., unbound fraction in brain tissue (f(u,slice)), or in plasma to determine the slice-to-plasma concentration ratio (C(slice)/C(plasma)). The unbound fraction in plasma, f(u,plasma), was determined using equilibrium dialysis. In vitro-in vivo correlation of the brain-to-plasma ratio was examined for 13 and eight model compounds in mice and rats, respectively. C(slice)/C(plasma) and f(u,plasma)/f(u,slice) predicted the K(p) in rats, and C(slice)/C(plasma) predicted the K(p) in FVB mice for non-P-glycoprotein substrates within 3-fold but overpredicted K(p) for P-glycoprotein substrates by more than 3-fold. However, C(slice)/C(plasma) predicted the K(p) in mdr1a/1b knockout mice for both non-P-glycoprotein and P-glycoprotein substrates. Our present study demonstrates that a brain slice method can be used to differentiate whether a compound having a low K(p) is due to the effect of low nonspecific binding to brain tissue relative to plasma proteins or because of efflux transport at the blood-brain barrier.