Using CYP3A4-expressing Caco-2 cell monolayers, we assessed the roles of CYP3A4-mediated metabolism, P-glycoprotein (P-gp)-mediated efflux, and serum protein binding in determining the extent of the intestinal first-pass extraction (E(i)) of saquinavir. Saquinavir (5-40 microM) was added to the apical compartment of culture inserts. After 3 h, apical and basolateral media and cell scrapings were analyzed for saquinavir and a major CYP3A4-mediated metabolite (M7). The intracellular concentration of saquinavir was estimated from the degree of inhibition of CYP3A4 catalytic activity (midazolam 1'-hydroxylation). Compared with vehicle, the P-gp inhibitor LY335979 (zosuquidar trihydrochloride) (0.5 microM, apical) increased saquinavir cell content and M7 formation rate, but decreased the E(i) by approximately 50% due to a >90% increase in the amount of saquinavir recovered in the basolateral compartment. Compared with LY335779, physiological concentrations of basolateral serum proteins [human serum albumin and alpha1-acid glycoprotein (AAG)] increased saquinavir permeability by a similar degree but decreased the E(i) by approximately 50% due to a marked reduction in M7 formation. Increasing AAG concentration (1.0-2.5 g/l) had no additional effect on permeability or E(i). An estimate of the range of the E(i) of saquinavir (7-60%) was less than has been predicted based on in vitro data (>99%) but was consistent with a clinical study involving grapefruit juice. The incidental finding of greater M7 formation after basolateral compared with apical dosing could not be explained by differences in saquinavir cell content. We conclude that variable intestinal first-pass extraction of saquinavir in human immunodeficiency virus-infected patients could reflect variation in P-gp-mediated efflux and/or CYP3A4-catalyzed metabolism, but not in blood AAG levels.