Parallel Mapping of Antibiotic Resistance Alleles in Escherichia coli

RESEARCH ARTICLE Parallel Mapping of Antibiotic Resistance Alleles in Escherichia coli Sophie J. Weiss1☯, Thomas J. Mansell1☯, Pooneh Mortazavi2, Rob Knight3,4, Ryan T. Gill1* 1 Department of Chemical and Biological Engineering, University of Colorado Boulder, 3415 Colorado Avenue, Boulder, Colorado, 80303, United States of America, 2 Department of Computer Science, University of Colorado Boulder, 1111 Engineering Drive ECOT 717, Boulder, CO 80303, United States of America, 3 Department of Pediatrics, University of California San Diego School of Medicine, 9500 Gilman Drive, MC 0602, La Jolla, CA 92093, United States of America, 4 Department of Computer Science & Engineering, University of California San Diego, 9500 Gilman Drive, MC 0404, La Jolla, CA 92093, United States of America ☯ These authors contributed equally to this work. * Abstract OPEN ACCESS Citation: Weiss SJ, Mansell TJ, Mortazavi P, Knight R, Gill RT (2016) Parallel Mapping of Antibiotic Resistance Alleles in Escherichia coli. PLoS ONE 11 (1): e0146916. doi:10.1371/journal.pone.0146916 Editor: Brian Frederick Pfleger, University of Wisconsin, UNITED STATES Chemical genomics expands our understanding of microbial tolerance to inhibitory chemicals, but its scope is often limited by the throughput of genome-scale library construction and genotype-phenotype mapping. Here we report a method for rapid, parallel, and deep characterization of the response to antibiotics in Escherichia coli using a barcoded genomescale library, next-generation sequencing, and streamlined bioinformatics software. The method provides quantitative growth data (over 200,000 measurements) and identifies contributing antimicrobial resistance and susceptibility alleles. Using multivariate analysis, we also find that subtle differences in the population responses resonate across multiple levels of functional hierarchy. Finally, we use machine learning to identify a unique allelic and proteomic fingerprint for each antibiotic. The method can be broadly applied to tolerance for any chemical from toxic metabolites to next-generation biofuels and antibiotics. Received: May 4, 2015 Accepted: December 23, 2015 Published: January 15, 2016 Copyright: © 2016 Weiss et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: Raw FASTQ files have been uploaded to the NCBI BioProject accession number PRJNA260093. A table of sequence counts is also available in the Supporting Information. Funding: SJW was funded by the NIH/CU Molecular Biophysics Training Scholarship (T32 GM-065103). Competing Interests: The authors have declared that no competing interests exist. Introduction Chemical genomics, or the study of the genome-scale response to small molecules, has rapidly advanced thanks to synthetic biology approaches. For example, studies of phenotype mapping of small molecule landscapes have led to elucidation of novel genetic functions and drug mechanisms [1–3]. These pioneering studies took large genomic libraries, usually painstakingly created [4, 5], and characterized them under a range of chemical and physical conditions using DNA microarrays. Studies of chemical tolerance have also used adaptive evolution methods to identify mutations that contribute to fitness [6, 7]. While these studies closely mimic responses to stresses in nature, the extent of genotyping is limited by the throughput of whole-genome sequencing. The increasing throughput and decreasing cost of multiplex oligonucleotide synthesis [8] and high-throughput sequencing [9] has enabled unprecedented advances in throughput of PLOS ONE | DOI:10.1371/journal.pone.0146916 January 15, 2016 1 / 17 Genome-Scale Analysis of Antibiotic Resistance genome engineering and analysis technologies [10–13]. For example, recent studies have leveraged high-throughput sequencing to expand the characterization of yeast deletion libraries [14]. Along these lines, we recently reported the trackable multiplex recombineering (TRMR) approach [15]: a one-pot construction of a barcoded, genome-scale library simulating overexpression and knockdown of over 4,000 genes in the Gram-negative bacterium E. coli. Initial experiments with the library focused on the genomic response to various carbon sources and biofuel-related inhibitory conditions using DNA microarrays and exploratory by-hand analyses [15, 16]. At sub-lethal antibiotic concentrations such as those found in wastewater and agricultural runoff, the contribution to microbial fitness of cellular factors is not nearly as well-studied [17] as horizontal gene transfer of specific resistance effectors [18]. Thus, understanding the response and resistance of microbes to antimicrobial compounds is of critical importance. To isolate gene products contributing to antibiotic resistance, several genomic and proteomic studies have been performed [19–23]. However, previous attempts to characterize genomescale responses to antibiotic challenges [1, 3, 6, 7, 24–26] relied on either (1) the low-throughput construction of large libraries or (2) many generations of adaptive evolution, where characterization was limited by sequencing surviving colonies. Here we report a method for the rapid and deep characterization of laboratory population dynamics in response to eight antibiotics by multiplex selection, next-generation sequencing, and multivariate analysis of E. coli TRMR libraries. Our findings support the development of multi-drug resistance and susceptibility genes as an important step in the evolution of antibiotic resistance in microbial populations at sub-lethal concentrations. Finally, to expand the throughput and extent of our bioinformatic analysis, we integrate the data gathered into the QIIME multivariate analysis pipeline, with which we examine the response at a pathway level and identify a unique genomic signature for each antibiotic. Methods Strains and plasmids The TRMR library was previously constructed [15]. Briefly, E. coli MG1655 (ATCC #700926) cells were subjected to multiplex recombineering using synthetic DNA cassettes containing either an “up” (strong promoter and RBS) or “down” (no promoter or RBS) phenotype with homology regions corresponding to over 4,000 genes in the E. coli genome. The synthetic cassettes also contained unique barcodes for rapid characterization and gene-trait mapping. In this study, a modified version of strain JWKAN, which is MG1655 with the kanamycin resistance gene neoR (from pKD13 [27]) inserted in a safe region and barcoded as in the rest of the library, was used as the wild-type control. Expression of FLP recombinase (pCP20 [28]) excised neoR from the genome using flanking FRT sites to create a barcoded MG1655 without kanamycin resistance, which we refer to as MG1655-BC. This phenotype was confirmed (...truncated)