Pure rabbit antithrombin III was isotope labeled (with 125I or 3H) by two different methods; neither procedure caused a loss of antithrombin activity although both methods affected the affinity of the protein for Sepharose-heparin. From segments from freshly excised rabbit aorta, the uptake of isotope-labeled antithrombin III by the endothelium was rapid and saturable, although relatively small compared to the uptake of thrombin; binding of 3H-antithrombin III to the endothelium resembled that of 125I-antithrombin III. Transendothelial passage of antithrombin III into the subendothelial layers (intima-media) was slow and progressive. Endothelium binding was not affected by pretreating the vessel with either heparin, thrombin, or glycosaminoglycan-specific enzymes. Endothelium-bound antithrombin III was not selectively displaced by either heparin or thrombin. In contrast, endothelium-bound thrombin was rapidly dislodged by antithrombin III as a thrombin- antithrombin III complex. The surface of the deendothelialized aorta (ie, subjected to a balloon catheter) bound antithrombin III avidly. Pretreatment of the deendothelialized vessel with glycosaminoglycan- specific enzymes, particularly heparitinase, decreased intima-media binding by up to 80%. 125I-antithrombin III, when bound to the deendothelialized vessel surface, was actively displaced by either heparin, thrombin, or by unlabeled antithrombin III. The relatively poor binding of antithrombin III compared with that of thrombin by the endothelium in vitro supports an earlier proposal (Lollar P, Owen WG: J Clin Invest 66:1222–1230, 1980) that thrombin bound to high-affinity sites, possibly pericellular proteoglycan, of the endothelium is inactivated by plasma antithrombin III in vivo. Such a situation probably holds for large arteries at least.