Bacterial genetics and molecular pathogenesis in the age of high throughput DNA sequencing

Abstract

When Stanley Falkow introduced Molecular Koch’s Postulates (Falkow, 1988) as a conceptual framework to identify microbial factors that contributed to disease, he reaffirmed the prominent role that the basic principles of genetic analysis should play in defining genotype-phenotype associations in microbial pathogens. In classical bacterial genetics the nature of mutations is inferred through cis-trans complementation and by indirectly mapping their relative position and physical distance through recombination frequencies — all of which were made possible by the genetic tools of the day: natural transformations, conjugation and transduction. Unfortunately, many of these genetic tools are not always available to study pathogenic bacteria. The recombinant DNA revolution in the 1980s launched the field of molecular pathogenesis as genes could be treated as physical units that could be cut, spliced and transplanted from one microbe to another and thus not only ‘prove’ that an individual gene complemented a virulence defect in a mutant strain but also could impart pathogenic properties to otherwise benign microbes. The recombinant DNA revolution also enabled the generation of newer versions of genetic tools to generate mutations and engineer microbial genomes.

The last decade has ushered in next generation sequencing technologies as a new powerful tool for bacterial genetics. The routine and inexpensive sequencing of microbial genomes has increased the number and phylogenetic scope of microbes that are amenable to functional characterization and experimentation. In this review, we highlight some salient advances in this rapidly evolving area.

Whole genome sequencing

The options for microbial whole genome sequencing (WGS) include platforms for shorts reads, primarily Illumina, as well as ‘third generation’ DNA sequencing platforms, such as PacBio’s single molecule real-time sequencing (SMRT) [1] and Oxford Nanopore Technology (ONT) [2^••]. Illumina sequencing reads can be used for de novo genome assemblies; however, its short reads can fail to assemble repetitive or transposed regions of a genome. In contrast, SMRT and ONT produce long reads that span repetitive regions and enable assembly over challenging sequences [3]. All three methods can be used to generate complete bacterial genomes de novo. WGS is revealing remarkable variation even within single strains, including populations with single nucleotide polymorphisms (SNPs), variance in gene copy number, and transpositions and inversions [2^••]. WGS also has important public health implications, such as global tracking of foodborne pathogens [3], and it can be used for prediction, diagnosis, and surveillance of antimicrobial resistance (reviewed in Refs. [4,5]). The accuracy and depth of modern DNA sequencing are continuously increasing, and the option to multiplex several samples in a single run has reduced the time and cost associated with WGS. Increasingly, microbiologists will be expected to routinely sequence and re-sequence their strains to confirm the accuracy and reproducibility of their observations (Figure 1).

Figure 1.

Applications of high throughput sequencing in bacterial genetics. (a) Whole genome sequencing is often the first step in the analysis of gene function. DNA sequencing can be accomplished using Illumina, SMRT, or ONT sequencing technologies. Depending on the size of a genome, multiple isolates can be multiplexed and sequenced in parallel while still obtaining high coverage. Once a genome is assembled, it can be used for comparative genomics or in other downstream applications. (b) Genome sequencing with SMRT also captures epigenetic modifications such as methylation. The high-resolution detection of modified bases can be used to determine a strains methylome. (c) Transposon insertion sequencing combines transposon mutagenesis with high throughput sequencing. Mutants present in the input and output pools are identified by mapping the insertion sites back to a reference genome. Insertion in genes that disappear from the input pool are considered conditionally essential. (d) RNA-seq uses Illumina sequencing to sequence cDNA. Sequenced reads are mapped back to a reference genome, capturing information about gene expression, transcriptional start sites, operon structure, and small RNAs. (e) Chemical mutagenesis enables rapid genetic analysis in genetically intractable bacteria. This is a particularly useful approach for novel bacterial isolates without established genetic tools. Mutants are generated with chemical mutagens and subjected to phenotypic screening. A pool of mutants enriched for the phenotype of interest is then sequenced and the mutations are mapped to a reference genome.

WGS and comparative genomics

WGS has been used to identify genes that are associated with virulence and host adaptation by comparing large collections of bacterial isolates from around the world [6,7^•], and from isolates within a single host [8,9,10]. Genome analysis can give insights into strain variation, and have revealed new virulence factors in pathogens such as Mycobacterium spp. [6,11], Listeria monocytogenes [12], and Staphylococcus aureus [13]. A recent comparison of Neisseria meningitidis strains exemplifies how WGS can be used to trace the origins of an emerging pathogen. While multi locus sequence typing (MLST) indicated that a meningococcal outbreak strain was identical to a known benign strain, WGS showed that the strain had acquired a capsule locus and a phage called Meningococcal Disease-Associated island (MDAФ), which were sufficient to transform it into an invasive pathogen [14^•]. Similarly, WGS of an Escherichia coli strain linked to hemolytic uremic syndrome showed that the strain had a duplicated copy of the Shiga toxin stx2 gene, thus more toxin production and enhanced virulence [15]. Although acquisition of virulence factors is a clear route to pathogenicity, genetic differences contributing to virulence can also be more subtle. For instance, a comprehensive analysis of over 2000 Streptococcus pyogenes isolates identified a single nucleotide insertion within an intergenic region that resulted in upregulation of a secreted virulence factor that enhanced virulence [7^•].

Beneficial microbes play a key role in health, and efforts to identify the genetic basis for these activities is of great medical and commercial interest. For example, a comparative analysis of the gut commensal Bifidobacterium adolescentis identified genes for glycan degradation, indicating adaptation to the host diet [16]. Analysis of intestinal metagenomic data aimed at identifying groups of genes differentially distributed among individuals, called structural variants, identified a butyrate production pathway in Anaerostipes hadrus linked to metabolic health [17]. These types of analyses that link specific genes to microbial influence on host health, could potentially drive the selection and engineering of probiotic strains in the future.

WGS for detection of epigenetic modifications

Next generation sequencing platforms such as SMRT can give single nucleotide resolution of epigenetic modifications (reviewed in Refs. [18–20]). Methylation is a common form of epigenetic modification in bacteria that is catalyzed by methyltransferases, typically resulting in N6 methyladenine, N4 methylcytosine, or C5 methylcytosine. One function of these modifications is to shield DNA from the restriction enzymes in restriction-modification (RM) systems, which protect against invading foreign DNA by cutting at specific unmethylated motifs. Methyltransferases also regulate gene expression and replication [18]. A second type of epigenetic modification, phosphorothioation on the sugar-phosphate backbone, appears to have similar functions and has been detected with both SMRT [21] and Illumina sequencing [22]. In some instances, phosphorothioation co-occurs with methylation [21].

DNA methylation regulates the expression of virulence factors such as pap pili and the surface protein antigen 43 in E. coli [23,24], and controls cell division in many alphaproteobacteria [25–27]. However, before modern sequencing techniques, bacterial epigenetic modifications were difficult to detect and often required prior knowledge of the methylated DNA sequence, thus the broader biological roles of these modifications are only just beginning to be understood [20]. With methylome analysis using SMRT sequencing, the DNA sequence targeted by a particular methyltransferase can be distinguished and used to determine the degree of methylation.

Such analyses determined that differentially methylated motifs are enriched in promoter regions, consistent with a role in gene regulation [28,29]. For instance, in S. pyogenes a combination of SMRT sequencing and RNA-seq identified a set of 20 virulence genes that had altered expression in absence of methylation [30^••]. Similar mechanisms may be widespread, and there is evidence that links methylation to gene regulation in diverse bacterial species, including Helicobacter pylori, Streptococcus pneumoniae, Campylobacter jejuni, Borrelia burgdorferi, and Photorhabdus luminescens [28,29,31–33].

In addition, bacterial methyltransferases are frequently subject to phase variation. As a result, a methyltransferase can be inactivated in a subpopulation of cells [34], or have its specificity altered to recognize a different DNA motif [31]. This generates global changes in gene expression called phase variable regulons (phasevarions) [20]. The H. pylori methyltransferase, modH, has 19 known versions, and pathogens such as N. meningitidis [35] and Streptococcus suis [36] encode multiple phase variable methyltransferases, potentially generating extensive genetic variability within a population. Phasevarions have primarily been analyzed in pathogens where they play an important role in virulence. For instance, a phase variable methyltransferase in H. pylori controls flagella production and expression of virulence genes [37]. Similarly, phase variation from DNA excisions and inversions in the S. pneumoniae methylation specificity gene, hsdS, result in altered colony morphologies, each with different abilities to colonize and cause systemic infection [31]. Consistent with a role in pathogenicity, methylome analysis of Haemophilus influenzae isolates from otitis media infections showed selection for specific phasevarion types, which were confirmed to show enhanced virulence in animal models [34].

With the ability to map entire methylomes, the intricate networks of phasevarions will continue to be dissected. Moreover, there is growing evidence hinting that epigenetic regulation is also active at the mRNA level, and that it can be detected by direct sequencing of mRNA using ONT [38,39]. As sequencing technologies and the ability to detect epigenetic modifications improves, the full extent of these modifications and their role in infection will begin to be appreciated.

Transposon insertion sequencing

Genomic sequences provide insight into gene content, but gene function often relies on homologies to genes previously characterized in model organisms. A powerful tool for functional analysis of genes is transposon mutagenesis. Typically, transposon (Tn) mutants are selected or screened for a particular altered function and the location of the Tn insertion in the genome is then determined using various PCR-based methods. Tn-insertion sequencing (abbreviated TIS herein) can be used to identify the insertion sites for thousands of mutants in parallel. Several variations of TIS have been described, all of which follow a similar negative selection approach: first, Tn mutagenesis is used to generate a pool of mutants, next the mutant pool is subjected to a selective pressure, and finally the Tn insertion junctures are sequenced to assess the abundance of each mutant in the input versus the output pool. Tn mutants that are depleted in output pool are assumed to contribute to survival in the selective pressure applied. Methods such as Tn-seq [40] and IN-seq [41] both use a modified mariner transposon with a point mutation that generates a recognition site for the type II restriction enzyme MmeI. When DNA is digested with MmeI, the enzyme generates a short flanking sequencing of genomic DNA adjacent to the transposon that is used to map the transposon insert location. Alternatively, approaches such as transposon-directed insertion sequencing (TraDIS) [42] and high-throughput insertion tracking by deep sequencing (HITS) [43] sequence randomly sheared DNA adjacent to transposon inserts and can therefore be used with any transposon. A further modification to standard TIS methods called random barcode Tn-seq (RB Tn-seq) incorporates a barcode into the transposon, which facilitates rapid quantification of TIS outputs and makes it feasible to test hundreds of different conditions [44]. Collectively, these approaches have helped identify core essential genes, genes required for microbial colonization, and to associate phenotypes with genes of unknown function.

TIS to identify essential genes

TIS data from multiple strains and growth conditions have been used to identify genes that are essential for growth under most conditions [45^••,46]. Recently, TIS was used to re-examine the core essential genes for some well characterized bacteria. For example, densely saturated Tn libraries of the laboratory strain E. coli K12 and manual curation identified a core set of 248 essential genes, which corrected multiple essential genes that had been misidentified in the single gene Keio collection due to errors and gene duplications [47]. In Pseudomonas aeruginosa, an analysis of 90 TIS datasets with nine different strains revealed dramatic variation in essential genes among strains, but with cross comparisons, a shared set of 321 core essential genes was defined [45^••]. These genes represent attractive targets for antimicrobial development that would potentially be effective across most Pseudomonas species. As sequencing bacterial genomes de novo becomes more prominent and economical, it is now possible to carry out TIS on virtually any bacteria that are amenable to genetic manipulation. TIS has been used to identify essential gene sets in a growing list of pathogens including Mycobacterium tuberculosis, Enterococcus faecalis, Streptococcus aglactiae, Proteus mirabilis, and C. jejuni, for both laboratory strains and patient isolates [46,48–52]. Deciphering essential genes has the potential to predict antibiotic resistance and to identify novel species specific, and possibly even strain specific, antimicrobial targets [45^••,46].

TIS to map host-microbe interactions

One of the first applications of TIS was to identify genes that contribute to host colonization, and host-microbe interactions remain a key application for TIS [41]. Genes required for colonization during infection have been determined for Vibrio cholerae, P. miribalis, Klebsiella pneumoniae, Yersinia pseudotuberculosis, and S. pyogenes [50,53–58]. These studies identified gene sets and pathways involved in infection of specific tissues [50,54,57,59], and the structure of microbial populations over the course of infection [56,57]. Interestingly, these types of experiments often identify a subset of mutants with enhanced fitness as well [56,60].Although cross complementation from neighboring cells can mask certain growth defects [61], analyses of polymicrobial interactions consistently identify distinct gene sets that are required for colonization with other microbes [59,61]. An exciting example of the translational implications of TIS is a study that used TIS to define a set of genes required for transmission of S. pneumoniae in a ferret model of infection, and the products of these genes were effective vaccine antigens [62^•].

TIS to define gene function

A common challenge in bacterial genomics is that genes frequently lack informative annotations. By combining very large numbers of mutants and growth conditions, however, TIS can give insight into protein function. For example, a TIS analysis of Pseudomonas simiae tested the fitness of genes required for root colonization in 90 in vitro conditions, which allowed even hypothetical proteins to be associated with specific phenotypes [60]. In a subsequent study, Price et al. used almost 5000 fitness tests in 32 different bacteria to identify phenotypes for 11 779 hypothetical proteins, in addition to correcting an additional set of incorrect or underannotated genes [63^••].A third study used TIS to identify novel genes involved in amino acid biosynthesis pathways across six different genera [64]. Because of the remarkable scale of these TIS experiments, it was possible for the authors to identify novel conserved genes and pathways among their data sets, suggesting that their findings are likely applicable across many bacterial species.

Chemical mutagenesis and whole genome sequencing

For many microbes the necessary genetic tools or DNA delivery mechanisms for Tn mutagenesis do not exist. One method to circumvent the need for molecular genetic tools is to induce mutagenesis through DNA damaging agents and mapping the resulting mutations by WGS [65]. Mutagens such as UV radiation and the DNA alkylating compounds ethyl methanesulfonate (EMS) and N-ethyl-N-nitrosourea (ENU) introduce point mutations randomly throughout the genome, which can inactivate or truncate a gene. As in the case of Tn-seq, WGS has transformed chemical mutagenesis into a viable high throughput method for microbial genetics.

Chemical mutagenesis and WGS was applied to generate mutants in the obligate intracellular bacterium Chlamydia [66,67]. The genomes of arrayed libraries of mutagenized C. trachomatis were sequenced in pools to map all resulting mutations. The library has been used for both reverse genetics, to analyze specific genes of interest, and in a forward genetic screen that identified an effector involved in host actin assembly [65] and in the protection from host defense mechanisms [68]. A similar approach was used to define the components of magnetosomes in Desulfovibrio magneticus RS-1, where mutagenesis and sequencing led to the discovery of novel, species specific magnetosome genes [69]. In addition to phenotypic screens, chemical metagenesis has also been used to investigate how specific amino acid residues contribute protein function in a method called Mut-Seq. This approach uses deep sequencing to compare genomes with and without mutagenesis to identify residues important for protein function based solely on their mutability, which can be used to determine features such as enzyme active sites and other functionally important motifs [70].

To further demonstrate how chemical mutagenesis and whole genome sequencing can be used in bacteria without established tools, a method termed mutational enrichment analysis after phenotypic selection (MEAPS) was developed and validated in the gut firmicute Exiguobacterium acetylicum [71]. A collection of chemically mutagenized E. acetylicum was screened for non-motile mutants, which were subsequently analyzed by WGS. The genes identified included conserved motility genes, such as flagella components, as well as novel genes that lacked motility associated annotations. Thus, with this approach, any culturable organism can be subjected to genetic analysis in the absence of classical molecular genetic tools.

Large scale transcriptional profiling (RNA-seq)

High throughput sequencing has had a particularly transformative impact on the analysis of gene expression. RNA-seq has long replaced microarray as the standard method to assess global patterns of gene expression. While alternative methods such as direct sequencing of RNA molecules with ONT [38], and cDNA sequencing with SMRT [72] have recently been described, Illumina sequencing of cDNAs remains the most approach for RNA-seq. With high levels of sequencing depth, RNA-seq can to deliver information on transcriptional activity and other aspects of RNA biology. For example, an SMRT sequencing-based method that was used to analyze operon structures in E. coli found an unexpected read-through at transcriptional stops sites that suggested a potential new mechanism of gene regulation [72]. In another approach, Illumina sequencing of RNA transcript ends created a detailed map of transcriptional start sites and terminators in E. coli [73]. These techniques, along with a rapidly growing list of specialized RNA-seq applications, can been used to determine operon structure, transcriptional start sites, small non-coding RNAs, and to identify RNA-binding proteins (recently reviewed in Ref. [74]).

RNA-seq is increasingly used to explore host-microbe interactions, using a modified version of RNA-seq called dual RNA-seq. Dual RNA-seq has been applied to probe gene expression in intracellular pathogens including C. trachomatis, uropathogenic E. coli, and H. influenzae [75–77]. This type of analysis can be applied to mRNA and small RNA expression during infection. For example, RNA-seq in Salmonella led to the discovery of the small non-coding RNA pinT that is expressed during infection and regulates genes required for intracellular replication. Simultaneous changes in host gene expression in response to pinT mutants revealed Salmonella’s ability to control inflammation via the JAK-STAT pathway [78]. A similar approach was used to investigate the extracellular pathogen Y. pseudotuberculosis during infection in mice. Analysis of Y. pseudotuberculosis collected from Peyer’s patches showed that the type three secretion system and its effectors were upregulated in a temperature dependent manner that involved control from small RNAs. In response, the host upregulated genes involved in inflammation and sequestering metal ions [79]. RNA-seq has also been applied to plant-microbe interactions, where it was used to test gene expression in the plant pathogen Pseudomonas syringae during infection with Arabidopsis mutants to test the roles of various plant immunity genes [80]. RNA-seq has also been used to directly test gene expression in human infections with pathogens including C. jejuni [81^•], Haemophilus ducreyi [82], and Mycobacterium leprae [83^•]. Thus, with sufficient sequencing depth, it has become possible to dissect a range of host-microbe interactions and to better understand the dynamics of infection.

Future directions

Next generation DNA sequencing has transformed bacterial genetics and opened previously intractable organisms to genetic analysis. As technologies develop, costs decrease and more robust bioinformatic analysis tools emerge, we expect that increased applications for WGS in single bacterial cell analysis, to reveal heterogeneity in populations, and to facilitate analysis of unculturable organisms. For example, single bacteria WGS is possible [84] and technical advances in cell isolation and genome amplification will make this approach more accessible [85]. Single cell analysis also has potential to be applied in TIS, with a method called droplet TN-seq where single cells are separated into droplets by microfluidics and cultured in isolation [86]. Single cell RNA-seq for bacterial cells is another area where methods are being developed that will likely shed new light on cell to cell diversity [87]. At the other end of the spectrum, analysis of complex microbial communities, such as the intestinal microbiota, with metatranscriptomics is an exciting area for RNA-seq that can not only identify the members of a community, but also reveal their transcriptional activities [88].

With the high speed and low cost of modern sequencing platforms, genetic analysis is no longer restricted to well characterized type strains with published genomes. By combining WGS, RNA-seq, and mutagenesis, virtually any microbe can be analyzed to better understand genetic factors that contribute to colonization or pathogenesis.