Cardiac myofibroblasts (MFs) have recently been regarded as important in arrhythmogenesis. There is growing evidence that coupling of cardiomyocytes (CMs) with MFs in vitro alters impulse propagation via electrotonic effect by MFs. However, exact initiating mechanisms of the MF-mediated arrhythmia remain elusive because studies on the arrhythmogenesis have been conducted mostly on mixed co-cultured monolayers of CMs with MFs. On the hypothesis that spatially non-uniform coupling of CMs with MFs is requisite for reentrant arrhythmogenesis, we analysed spatiotemporal changes in impulse propagation of the CM monolayer cultured on a filter membrane (25-mm diameter; pore diameter, 8 μm; pore density, 6×105 /mm2) with culture of either MFs or CMs on the reverse side. CMs and MFs were obtained from neonatal rat hearts excised under generalised anaesthesia by i.p. injection of sodium pentobarbitone (0.2mg/g body wt). CMs (2×105 cells/cm2) were plated on the upper membrane layer, and either CMs (2×105 cells/cm2, CM/CM model) or MFs (2×104 cells/cm2, CM/MF model) were plated on the reverse side. After 5-day culture, spatiotemporal changes in fluo-4-fluorescence (11x7mm, 192×128 pixels, 200 frames/s) were imaged on the confluent CM monolayers on the upper side at 33°C. Data (mean±SD), statistically analysed by ANOVA, were considered as significant at p<0.05. We found that during 2-Hz pacing at the margin of the upper layer, CMs exhibited uniform wavefronts from the pacing site in the CM/CM model with CV of 76.6±10.9mm/s (n=20), whereas those in the CM/MF model exhibited irregular wavefronts with slower CV (51.8±12.2mm/s, n=23, p<0.01). Gap-junction coupling between the upper and lower layers was demonstrated by calcein-dye transfer and its inhibition by heptanol. During 3-Hz pacing CM monolayers exhibited reentrant arrhythmias with greater induction rate in CM/MF models (67.4%, n=43) than that in the CM/CM models (22.2%, n=27, p<0.01). In contrast, high-K+ (8mM) perfusion in the CM/CM models, which slowed CV (41.3±7.6mm/s) equivalent to that in the CM/MF models under control K+ (4mM) perfusion, barely induced arrhythmias (10.0%, n=10). In confocal immunofluorescence analysis for connexin 43 (Cx43) plaques at the membrane pores, the CM/MF model revealed denser distribution (35.5±3.5%, n=6) in the reentrant cores than in the surroundings (24.6±5.0%, n=6, p<0.01), suggesting greater electrotonus of MFs to CMs on the core than the surroundings. However, the CM/CM models that failed to induce arrhythmia showed no regional difference in Cx43 density (n=6). In conclusion, heterocellular coupling of CM monolayer with MFs depresses CV, but spatial non-uniformity in the coupling is requisite for reentrant arrhythmogenesis.