Large Scale Engineering of Chimeric Histidine Kinases

Abstract

Sensor histidine kinases (SHKs) represent one of the most abundant and versatile protein families in nature, mediating cellular responses to an extensive range of environmental stimuli. Despite their ubiquity, the majority of SHKs remain functionally uncharacterized due to the limitations of traditional, low-throughput methods. This dissertation addresses the challenge of SHK deorphanization by developing a high-throughput platform for rationally engineering, synthesizing, and profiling the basal activation of chimeric SHKs.Central to this approach is Degenerate DropSynth, a multiplex gene synthesis technique adapted to assemble multiple variants per gene per droplet in a single emulsion reaction. Chimeric fusion-phase SHK variants were designed via residue insertions or deletions on either side of the fusion junction below the HAMP domain, with variants sampled at one to eight degeneracy levels per parent gene, leading to 21,724 total gene constructs, and this method synthesized these four libraries with a coverage of 75.8% at the amino acid level. Each fusion-phase variant introduces specific angular changes between helical domains to explore effects of register alignment on SHK activity. These libraries were cloned into barcoded plasmids and transformed into E. coli BW29655 (ΔompR ΔenvZ) carrying the response regulator plasmid pSR40.29. This plasmid couples SHK signaling to expression of the superfolderGFP fluorescent reporter, enabling quantification of signaling output. Following growth in supplemented minimal media, cells were sorted by fluorescence-activated cell sorting into six brightness bins. Barcodes from sorted cells were sequenced via NovaSeq, and per-variant brightness levels were inferred by calculating bin-weighted fluorescence, reported in MEFL units. Across the successfully profiled proteins, fusion phase exerted a dominant effect on basal brightness, with some phase shifts resulting in constitutively active “locked-on” phenotypes and others producing inactive or functional signaling behavior. Analysis revealed that the impact of fusion phase varied across sensor classes, highlighting the structural sensitivity of SHK domain interfaces. A random forest model trained on 13,170 variants was able to predict a fifth of the variance for MEFL brightness based on sequence and phase, offering a framework for computational pre-screening. This work establishes a scalable, structure-aware design-build-test-learn cycle for SHK engineering, enabling functional mapping at a scale previously inaccessible. This dissertation includes unpublished co-authored material.