Expanding Biosensing Abilities through Computer-Aided Design of Metabolic Pathways

Abstract

Detection of chemical signals is critical for cells in nature as well as in synthetic biology, where they serve as inputs for designer circuits. Important progress has been made in the design of signal processing circuits triggering complex biological behaviors, but the range of small molecules recognized by sensors as inputs is limited. The ability to detect new molecules will increase the number of synthetic biology applications, but direct engineering of tailor-made sensors takes time. Here we describe a way to immediately expand the range of biologically detectable molecules by systematically designing metabolic pathways that transform nondetectable molecules into molecules for which sensors already exist. We leveraged computer-aided design to predict such sensing-enabling metabolic pathways, and we built several new whole-cell biosensors for molecules such as cocaine, parathion, hippuric acid, and nitroglycerin.

Engineering tunable biosensors for monitoring putrescine in Escherichia coli

“…For example, the L‐arabinose‐responsive transcriptional regulator AraC from Escherichia coli has been engineered to specifically respond to the levels of D‐arabinose (Tang, Fazelinia, & Cirino, ), triacetic acid lactone (Tang et al, ), and mevalonate (Tang & Cirino, ). Another strategy to expand the range of biologically detectable molecules is to transform nondetectable molecules via multistep biochemical reactions into molecules for which sensors already exist (Libis, Delépine, & Faulon, ). Therefore, the range of molecules that can be monitored has been expanding rapidly and now includes certain members of sugars, amino acids (Mahr & Frunzke, ; Mustafi, Grünberger, Kohlheyer, Bott, & Frunzke, ), lactams (Zhang et al, ), organic acids (Li & Yu, ), redox molecules (Zhang et al, ), heavy metals (Bereza‐Malcolm, Mann, & Franks, ; Cerminati, Soncini, & Checa, ), organophosphates (Chong & Ching, ), and flavonoids (Siedler, Stahlhut, Malla, Maury, & Neves, ).…”

“…For example, the L-arabinose-responsive transcriptional regulator AraC from Escherichia coli has been engineered to specifically respond to the levels of D-arabinose (Tang, Fazelinia, & Cirino, 2008), triacetic acid lactone (Tang et al, 2013), and mevalonate (Tang & Cirino, 2011). Another strategy to expand the range of biologically detectable molecules is to transform nondetectable molecules via multistep biochemical reactions into molecules for which sensors already exist (Libis, Delépine, & Faulon, 2016).…”

Reinforcement Learning for Bio-Retrosynthesis

“…We showcase RetroPath 3.0's modularity by biasing the search towards less toxic intermediates, and also provide expert users 2 strategies to speed-up the retrosynthetic search, either by storing results in a database (Supplementary Note 3) or extending a search from previously run search trees (Supplementary Note 4). Another feature of interest is the ability to use RetroPath 3.0 for biosensor design, as we also demonstrated with our previously developed tools Libis, Delépine, and Faulon 2016). This can be used in conjunction with a dataset of detectable compounds (Koch et al 2018) to allow for design of Sensing Enabling Metabolic Pathways.…”

Plug-and-Play Metabolic Transducers Expand the Chemical Detection Space of Cell-Free Biosensors

“…Compared to its in vivo counterpart 17 , the cell-free benzoic acid biosensor is faster (maximum signal reached in four hours, Supplementary Fig. 1 ) and far more sensitive ( Fig.…”

“…1b ). We used the SensiPath webserver that we previously designed and validated in vivo to determine the required metabolic cascade 17,18 .…”