The development and use of an extracorporeal liver support device depends upon the isolation of a large number of viable, functioning hepatocytes from whole or partial livers. Current practice, however, produces nonoptimal yields, given that a large percentage of hepatocytes initially presented are not successfully isolated. The normal hepatocyte isolation protocol consists of sequential perfusion with calcium chelating and collagenase buffers, and then separation of viable hepatocytes from non-viable and nonparenchymal cells, usually on the basis of cell density. In order to improve understanding regarding the metabolic and perfusion state of the liver during this perfusion protocol, ATP, pH, and tissue perfusion were evaluated using nuclear magnetic resonance (NMR). Perfusion with calcium chelating buffer was found to have minimal effect on the metabolic and perfusion parameters, whereas subsequent perfusion with collagenase buffer produced large declines in ATP, pH, and homogeneity of perfusion within 3 min. Perfusion with calcium-chelating buffer alone, or perfusion with calcium chelating buffer followed by a short period of ischemia to mimic the perfusion disruption of collagenase, did not produce the same decline in metabolic parameters. This NMR data suggested that enhancing the early perfusion and penetration of collagenase or prolonging the nontoxic calcium-chelation step may improve the yield and/or functionality of isolated cells. Therefore, several altered perfusion protocols were evaluated in terms of yield of viable parenchymal hepatocytes and hepatocyte albumin production. Although increasing the perfusion flow rate and initial perfusion with inactive (cold) collagenase did not produce significant improvements when compared with the control protocol (control cell yield 226 +/- 42 x 10(6) viable hepatocytes for 10- to 14-week-old female Lewis rat), prolonging and enhancing the calcium-chelating perfusion step or increasing the collagenase concentration did yield a significantly great number of viable parenchymal hepatocytes (393 - 44 and 328 +/- 39 x 106 viable hepatocytes, respectively) with no change in albumin production per seeded viable cell.