High-throughput screening is essential for current pharmaceutical drugs development where the number of target compounds to be screened is in the range of 105 – 106 trials and constantly increasing (Macarron et al., 2011). Specifically, cardiotoxicity is one of the main causes of drug rejection and withdrawal from the market. Additionally, current in vitro research based on animal models does not truly predict the response of the human heart (Mathur et al., 2015). Multi-electrodes arrays (MEAs) represent a powerful tool for in vitro electrophysiological studies of both animal or human cardiac models. They allow non-invasive and long-term monitoring of cardiac action potentials, including both extra- and intra-cellular recordings. Therefore, they represent an effective strategy for high-throughput and high-resolution cardiotoxicity studies. To further increase the throughput of MEAs, microfluidic devices can separate the surface of chips into multiple, independent cell culture chambers. Microfluidic devices enable parallelization and automation as well as the reduction of reagents and cell consumption. This leads to a decrease in the final cost of the assay. Nevertheless, the integration of microfluidics on complementary metal-oxide semiconductors (CMOS) integrated circuits presents numerous technological challenges that need to be thoroughly assessed and faced (Huang and Mason, 2013). To overcome these challenges, we developed a novel strategy to integrate a polystyrene (PS)-based microfluidic system on top of a CMOS MEA (Lopez et al., 2018). A thin perforated glass interposer was bonded onto the CMOS MEA, while the PS-based microfluidic system was glued on top. The PS-device determined the height of the microfluidic chamber and their locations. The microfluidic chip separated the chip surface into 16 independent cell culture chambers. Therefore, up to 16 different experiment conditions are possible per device. The microfluidic cell culture chambers (area 0.62 mm^2, height 1.25 mm) are addressed by an individual inlet and an individual outlet. Each chamber contains 1,024 recording electrodes that are grounded by on-chip integrated reference electrodes. Therefore, no external reference electrodes were required. The electrodes in each chamber are independently addressable for individual measurements. The chip architecture allows for 6 different types of measurements on the same chips: extracellular recordings, intracellular recording, constant voltage stimulation, constant current stimulation, impedance monitoring at a fixed frequency and impedance spectroscopy between 10 Hz and 1 MHz (Lopez et al., 2018). At first, we optimized the protocol for the microfluidic culture of rat ventricular cardiomyocytes in prototype devices. In the first-generation prototypes, the CMOS MEA chip was replaced with a transparent glass slide for easier optical access. External syringe pumps controlled the perfusion of fluids through the microfluidic chambers. Flow rates were optimized for each of the steps of the cell culture, including sterilization, coating, seeding, and nutrients delivery. Cardiomyocytes were successfully cultured for up to 9 days in vitro (DIV). Cardiac functionality was confirmed both by visualizing cardiac contractions using phase-contrast microscopy and by lens-free imaging for in situ monitoring. Cell viability was assessed through calcein and Hoechst stainings. Specifically, from the fluorescent staining a cell density of 1640 ± 280 nuclei/mm^2 was measured inside the microfluidic chambers at 2 DIV. In the second-generation prototypes, the glass slide was replaced with a passive high-density MEA. The optimized protocol for culturing microfluidic cardiac cell cultures was successfully transferred and clear extracellular recordings were detected in the electrodes of size 11x11 µm^2 from DIV 3 – 5. An average peak-to-peak voltage amplitude of 601 ± 16 µV was measured at 4 DIV with a beating rate of 51.5 ± 1.1 bpm. Therefore, the presented approach is compatible with Si-based MEAs. Currently, the optimized protocols developed in the prototype device are being applied to a microfluidic system integrated onto the CMOS MEA substrate. As syringe pumps are bulky and not scalable, they will be replaced by embedded peristaltic pumps to internally control the fluids flows. Human induced pluripotent stem-cells will be then used to assess drug toxicity assays exploiting the high-throughput 16 wells configuration. Our approach is compatible with the 96 well plate standard. Moreover, both the number of CMOS MEAs and of microfluidic chambers per system can be extended so to further increase the throughput and reach a 96 wells configuration.