Imaging with GECIs has become a widely used method in physiology and neuroscience [1][2][3] . According to readout mode, the design of the sensors has followed two different pathways, leading to single-wavelength sensors and FRET-based ratiometric sensors [4][5][6][7][8] . Among the most popular single-wavelength sensors are the G-CaMPs 9-13 , R-CaMPs 14 and GECOs 15 . FRET sensors include yellow cameleon 3.60 (refs. 16,17), D3cpv 18 , yellow cameleon Nano 19 and TN-XXL 20 .Quantification by ratiometric FRET imaging is more accurate than single-channel measurements and may be more suitable for long-term functional imaging studies, as it is less influenced by changes in optical path length, excitation light intensity and indicator expression level and by tissue movement and growth during development. In addition, FRET indicators are substantially brighter than single-wavelength sensors at rest, allowing better identification of expressing cells and their subcellular structures. Another practical feature of FRET-based indicators is their ability to measure basal Ca 2+ levels within cells, for example, to distinguish between resting and continuously spiking neuronssomething that cannot easily be achieved with single-wavelength indicators 21 . Increased basal Ca 2+ levels within the brain are also observed at the onset of neurodegenerative processes, and ratiometric FRET calcium imaging has been used in these conditions to monitor disease progression 22,23 . Moreover, ratiometric indicators are advantageous for monitoring calcium in motile cells.Both calmodulin and troponin C (TnC), the calcium binding proteins within the various GECIs, consist of two globular domains connected by a central linker 24,25 . Each domain possesses two calcium-binding EF hand motifs. Thus, currently available GECIs are highly nonlinear sensors binding up to four calcium ions per sensor. Identification of a smaller calciumbinding domain with fewer binding sites could help to reduce buffering during long-term chronic GECI expression 26 , make the sensor smaller and further minimize the risk of cytotoxicity. It may also help to simplify response properties and facilitate the biophysical modeling of sensor behavior.Here we report several improvements of FRET-based calcium sensors for in vivo imaging. First, we identified a minimal calcium binding motif based on the C-terminal domain of TnC with only two or one remaining calcium binding sites per sensor molecule, thus reducing the overall calcium buffering of the sensors. Second, by sampling TnCs from various species we identified a new TnC variant from the toadfish Opsanus tau, which offered the possibility of generating minimal domains with high-affinity calcium binding. Third, we used a large-scale, two-step functional screen to optimize the FRET changes in the sensor by linker diversification. This approach allowed us to identify Twitch sensors with a superior FRET change and may become useful for optimizing other types of FRET sensors. Finally, we improved brightness and photostability o...