A subset of the family of fluorescent proteins are the non-fluorescent chromoproteins which are promising probe molecules for use in photoacoustic imaging and as acceptor chromophores in Förster resonance energy transfer (FRET)-based biosensors. Typical approaches for fluorescent protein optimization by screening of large libraries of variants cannot be effectively applied to chromoproteins due to their characteristic lack of fluorescence. To address this challenge， we have developed a directed evolution method to iteratively screen large libraries of protein variants on the basis of their photoacoustic signal levels. By applying this procedure to the promising Ultramarine and cjBlue chromoprotein templates， we were able to identify improved variants with a 02-04 fold increase in photoacoustic signal-to-noise ratio after only a few evolutionary steps. These improved variants enable more accurate spectral de-mixing and localization of protein-producing bacteria in vivo and serve as effective FRET acceptors for both fluorescence- and photoacoustic-based detection of protease activity.