Accurate assignment of protein function from sequence remains a fascinating and difficult challenge.   The periplasmic binding protein (PBP) superfamily present an interesting case of function prediction, because they are both ubiquitous in prokaryotes, and they tend to diversify through gene duplication “explosions” that can lead to large numbers of paralogs in a genome.  An engineered version of the moderately thermostable glucose-binding PBP from Escherichia coli has been used successfully as a reagentless fluorescent biosensor both in vitro and in vivo.  To develop more robust sensors that meet the challenges of real-world applications, we report the discovery of thermostable homologs that retain a glucose-mediated conformationally coupled fluorescence response.  Accurately identifying a glucose-binding PBP homolog among closely related paralogs is challenging.  We demonstrate that a structure-based method that filters sequences by residues that bind glucose in an archetype structure is highly effective.  Using fully sequenced bacterial genomes we found that this filter reduced high paralog numbers to single hits in a genome, consistent with accurate separation of glucose binding from other functions.  We expressed engineered proteins for eight homologs, chosen to represent different degrees of sequence identity and tested their glucose-mediated fluorescence responses.  We accurately predicted the presence of glucose binding down to 31% sequence identity.  We also have successfully identified suitable candidates for next-generation robust, fluorescent glucose sensors.

Bioinformatics: the NCBI database of fully sequenced bacterial genomes was searched for sequence of homologs of the E. coli periplasmic glucose-binding protein (ecGBP).  True glucose-binding homologs were selected by applying sequence filters derived from analysis of the 3D structure of ecGBP.  The deposited information includes the results from these searches, the software package that was developed to carry out this type of search, and the scripts used for this project.

Protein Chemistry: synthetic genes for glucose-binding proteins homologs were built, and proteins produced by over-expression.  The glucose-binding properties and glucose-dependent thermostabilities of their fluorescent conjugates were determined experimentally.  The deposited information comprises the raw data sets, and the software used to analyze these.