AIMTo determine the influence of the construction design over the biological component’s performance in an experimental bio-artificial liver (BAL) device.METHODSTwo BAL models for liver microorgans (LMOs) were constructed. First, we constructed a cylindrical BAL and tested it without the biological component to establish its correct functioning. Samples of blood and biological compartment (BC) fluid were taken after 0, 60, and 120 min of perfusion. Osmolality, hematocrit, ammonia and glucose concentrations, lactate dehydrogenase (LDH) release (as a LMO viability parameter), and oxygen consumption and ammonia metabolizing capacity (as LMO functionality parameters) were determined. CPSI and OTC gene expression and function were measured. The second BAL, a “flat bottom” model, was constructed using a 25 cm2 culture flask while maintaining all other components between the models. The BC of both BALs had the same capacity (approximately 50 cm3) and both were manipulated with the same perfusion system. The performances of the two BALs were compared to show the influence of architecture.RESULTSThe cylindrical BAL showed a good exchange of fluids and metabolites between blood and the BC, reflected by the matching of osmolalities, and glucose and ammonia concentration ratios after 120 min of perfusion. No hemoconcentration was detected, the hematocrit levels remained stable during the whole study, and the minimal percentage of hemolysis (0.65% ± 0.10%) observed was due to the action of the peristaltic pump. When LMOs were used as biological component of this BAL they showed similar values to the ones obtained in a Normothermic Reoxygenation System (NRS) for almost all the parameters assayed. After 120 min, the results obtained were: LDH release (%): 14.7 ± 3.1 in the BAL and 15.5 ± 3.2 in the NRS (n = 6); oxygen consumption (μmol/min·g wet tissue): 1.16 ± 0.21 in the BAL and 0.84 ± 0.15 in the NRS (n = 6); relative expression of Cps1 and Otc: 0.63 ± 0.12 and 0.67 ± 0.20, respectively, in the BAL, and 0.86 ± 0.10 and 0.82 ± 0.07, respectively, in the NRS (n = 3); enzymatic activity of CPSI and OTC (U/g wet tissue): 3.03 ± 0.86 and 222.0 ± 23.5, respectively, in the BAL, and 3.12 ± 0.73 and 228.8 ± 32.8, respectively, in the NRS (n = 3). In spite of these similarities, LMOs as a biological component of the cylindrical BAL were not able to detoxify ammonia at a significant level (not detected vs 35.1% ± 7.0% of the initial 1 mM NH4+ dose in NRS, n = 6). Therefore, we built a second BAL with an entirely different design that offers a flat base BC. When LMOs were placed in this “flat bottom” device they were able to detoxify 49.3% ± 8.8% of the initial ammonia overload after 120 min of perfusion (n = 6), with a detoxification capacity of 13.2 ± 2.2 μmol/g wet tissue.CONCLUSIONIn this work, we demonstrate the importance of adapting the BAL architecture to the biological component characteristics to obtain an adequate BAL performance.