Metagenomic mining of regulatory elements enables programmable species-selective gene expression

Abstract

Robust and predictably performing synthetic circuits rely on the use of well-characterized regulatory parts across different genetic backgrounds and environmental contexts. Here, we report the large-scale metagenomic mining of thousands of natural 5′-regulatory sequences from diverse bacteria and their multiplexed gene expression characterization in industrially-relevant microbes. We identified regulatory sequences with broad and host-specific expression properties that are robust in various growth conditions. We further observed significant differences between species’ capacity to utilize exogenous regulatory sequences. Finally, we demonstrated programmable species-selective gene expression that produces distinct and diverse output patterns in different microbes by leveraging regulatory sequences with pre-defined host-specificities. Together, these findings provide a rich resource of characterized and annotated natural regulatory sequences and a framework to engineer synthetic gene circuits with unique and tunable cross-species functionality and properties.

Multiplex transcriptional characterizations across diverse bacterial species using cell‐free systems

“…Across different regulatory libraries, transcription levels spanned > 5 orders of magnitude and were highly reproducible using the same and separately prepared cell lysates on different days (Fig B, Appendix Fig S2A). Similar to previous reports (Johns et al , ), transcriptional values were consistent across libraries using alternate N‐terminal barcodes and downstream genes (sfGFP or mCherry), which supports that gene‐specific reverse transcription or barcode identity does not significantly affect our activity measurements (Appendix Fig S2B and C). We observed that relative transcription levels remained constant across different TXTL reaction times (0.5, 1, 2, or 4 h).…”

“…We did not find distinct patterns of expression across different classes of antibiotic resistance genes from which the regulatory sequences were mined (Appendix Fig S9). However, we did observe striking phylogenetic trends, with regulatory sequences derived from the Firmicutes tending to be more highly active overall compared with those from Proteobacteria, which is likely the result of differences in genomic GC contents between these groups (Fig E; Johns et al , ). These findings provide a more informed basis for regulatory element use and design in diverse bacterial species and provide a path for further examination of the sequence features underlying these observed differences.…”

“…A major roadblock in identifying these species‐specific regulatory properties is the low efficiency of DNA transformation in many bacteria, which limits multiplex characterization in them. To address this challenge, we first tested a small 234‐member regulatory library (RS234; Johns et al , ) using DRAFTS in seven species with available genetic tools and efficient transformation methods for in vivo comparison to obtain multiplex transcriptional data from cell‐free systems. These in vitro results were compared with in vivo transcription levels measured from library‐harboring populations of each species harvested at mid‐exponential growth phase in rich media (Appendix Tables S1 and S4).…”

“…We observed increased dissimilarity of transcriptional profiles between phylogenetically more distant species, suggesting a functional evolutionary divergence of transcriptional regulation across the natural microbial biome (Iyer et al , ). Differences between microbial species could be exploited to build more refined regulatory circuits that function in specific organisms or multiple targeted species (Johns et al , ). Using these high‐throughput datasets, simple linear regression models of transcriptional activation could be trained to predict promoter strengths in a wide array of organisms.…”

“…Custom code was used to identify the barcode within each read and map the region corresponding to the regulatory sequence to designed library constructs. Transcriptional activities were calculated as done in previous studies (Kosuri et al , ; Johns et al , ) by normalizing each construct's individual RNA abundance by its DNA abundance using merged counts from two replicates. The log 10 ‐transformed expression levels were converted to Z ‐scores for further comparative analysis across samples.…”