Portal blood leukocytes play an important, but still poorly defined, role in the immune processes of the liver. In our previous studies, we showed that certain leukocyte subsets are selectively halted in the liver. These cells marginate in sinusoids and, together with resident Kupffer and endothelial sinusoidal cells, participate in antiviral and antitumor processes. The molecular mechanisms of margination and cooperation with resident sinusoidal cells require clarification. However, in vivo harvesting of portal blood leukocytes is associated with cumbersome cannulation of portal and hepatic veins and manipulation of the liver, causing major disturbances in splanchnic blood flow and liver blood supply, totally distorting sinusoidal blood perfusion and leukocyte margination. To overcome these difficulties, we have developed an in situ normothermic rat liver perfusion model permitting quantitative observations of blood leukocyte extraction in sinusoids. First, liver was flushed through the portal vein and the effluent leukocytes, named liver-associated leukocytes (LAL1), were collected from hepatic veins. Then, the liver was perfused for 60 min with 50 ml of blood using a semiclosed perfusion system. Upon completion of perfusion, the liver portal vasculature was flushed again to retrieve the leukocytes extracted from the perfusing blood (LAL2). These cells were characterized with respect to their phenotype and cytotoxicity. The mean leukocyte count of the washout before perfusion was 1.04±0.2×10 6/g of liver tissue and 0.9±0.1×10 6/g after 60 min of perfusion, indicating retention by the perfused liver, the live leukocyte extracting capacity. To further evaluate the efficiency of perfusion, FITC-labelled leukocytes were added to the perfusing leukocyte-free blood. Around 95% of the postperfusion washout LALs were FITC +. Heat-killed leukocytes did not marginate in sinusoids. Preincubation of leukocytes with substances able to lower adhesion capacity, such as lidocaine, trypsin and AAGM1, significantly decreased the postperfusion LAL2 washout population. The numbers of extracted postperfusion LAL2 CD5 +, CD4 +, CD8 +, CD56 + and class II + subsets did not differ statistically from those of preperfusion LAL1. Moreover, the cytotoxicity of LAL2 and LAL1 against CC531 and K562 remained at a similar level. Thus, perfused liver also retained its selective leukocyte extraction capacity. This model shows that the process of selective margination of portal blood cell subsets in the liver can be studied in an artificially perfused liver subject to physiological blood flow parameters, temperature, oxygenation and minimal ischemic time before connection to the perfusion device. Furthermore, it is suitable for studies of the selective recruitment of blood cells in sinusoids in a wide large of situations including liver tumors, infections, rejection after transplantation, graft vs. host disease, as well as in the investigation of the effect of drugs on these processes.