The proximal tubule absorbs water and small molecules back into the body after they have been filtered by the glomerulus. Diseases can interrupt this process and lead to kidney dysfunction. Nephrotoxicity can occur in any part of the kidney but most drug induced injury impacts the proximal tubules in some way and this can generate systemic effects. Drug development includes multiple stages before a compound is ready for market. A vast array of information about the drug, its usefulness as a therapy, and possible side effects must be assembled. A key cause of drug failure is kidney dysfunction or nephrotoxicity, which accounts for 19 % of all drug failures during testing in phase III clinical trials. It is generally perceived that modern toxicity tests are inadequate for determining cell responses during early drug development and this can lead to costly failures later in the testing process. Therefore, a more accurate toxicity assay that utilises organ-on-a-chip technology and can mimic the in vivo microenvironment of the kidney is needed. An ideal assay would model the transport properties of the proximal tubule and reproduce physiological responses to known nephrotoxic agents. It was shown that current assays could mimic some in vivo drug responses. However, all these bioreactors used some form of synthetic or manmade membrane or hydrogel. Cells are sensitive to their surroundings so most current assays fail to appropriately model native extracellular matrix as synthetic scaffolding can be considerably different and could directly impact any toxicity model prediction. What is needed is a system where the cells forming a membrane are self-supporting thus removing the substrate interactions entirely. This thesis describes a model of the kidney proximal tubule in which a sheet of proximal tubular epithelial cells can be fixed as a membrane separating two fluid chambers without the presence of a supporting structure. To accomplish this a bioreactor was constructed and its development protocols optimised. Computational fluid dynamics modelling and albumin perfusion testing further characterised the bioreactor as a viable assay. Renal proximal tubule epithelial cells (RPTECs) survived culture within the bioreactor and showed partial alignment parallel to flow; indicating the bioreactor could generate a cell response. In the process of forming and suspending cell sheets 3T3 cells, human kidney 2 (HK-2) proximal tubule cells, and RPTECs were used. Cell sheets were harvested using poly(N-isopropylacrylamide) temperature responsive substrates. Saturated oil and alginate based cell sheet forming techniques were also investigated but were unable to produce cell sheets. Growth on the temperature responsive substrates was optimised, with 3T3 cells needing no optimisation, HK-2 cells unable to form sheets, and RPTECs needing high density seeding combined with coating of the substrate with collagen IV. The cell sheets were then suspended over the window of an insert that would sit within the constructed bioreactor. A protocol involving insert coating with fibronectin and direct cell seeding was developed to functionalise the insert surface for cell sheet bonding. 3T3 cell and RPTEC sheets were able to be suspended and were viable. However, complete insert window coverage by a cell sheet was elusive. It was concluded that the bioreactor was a viable model for suspending cells sheets and as a toxicity assay, suspension of the cell sheets was possible, and that further optimisation of the suspension techniques could yield a completed assay for disease modelling and drug development. This model could then be applied to other transport tissues of the body; facilitating the increased accuracy of in vitro cell responses to stimuli, improving our understanding of how cells function, and aiding in the development of personalised medicine.