Approaches to querying bacterial genomes with transposon-insertion sequencing.

In this review, we discuss transposon-insertion sequencing, variously known in the literature as TraDIS, Tn-seq, INSeq, and HITS. By monitoring a large library of single transposon-insertion mutants with high-throughput sequencing, these methods can rapidly identify genomic regions that contribute to organismal fitness under any condition assayable in the laboratory with exquisite resolution. We discuss the various protocols that have been developed and methods for analysis. We provide an overview of studies that have examined the reproducibility and accuracy of these methods, as well as studies showing the advantages offered by the high resolution and dynamic range of high-throughput sequencing over previous methods. We review a number of applications in the literature, from predicting genes essential for in vitro growth to directly assaying requirements for survival under infective conditions in vivo. We also highlight recent progress in assaying non-coding regions of the genome in addition to known coding sequences, including the combining of RNA-seq with high-throughput transposon mutagenesis.

Introduction

A common approach to identifying genomic regions involved in survival under a particular set of conditions is to screen large pools of mutants simultaneously. This can be done with defined mutants;, however, the construction of defined mutant libraries is labor intensive and requires accurate genomic annotation, which can be particularly difficult to define for non-coding regions. An alternative to defined libraries is the construction and analysis of random transposon-insertion libraries. The original application of this method used DNA hybridization to track uniquely tagged transposon-insertions in Salmonella enterica serovar Typhimurium over the course of BALB/c mouse infection. DNA hybridization was eventually superseded by methods that used microarray detection of the genomic DNA flanking insertion sites, variously known as TraSH, MATT and DeADMAn (reviewed in ref. 4). However, these methods suffered from many of the problems microarrays generally suffer from: difficulty detecting low-abundance transcripts, mis-hybridization, probe saturation and difficulty identifying insertion sites precisely.

The application of high-throughput sequencing to the challenge of determining insertion location and prevalence solves many of these problems. Interestingly, the first application of transposon-insertion sequencing, developed by Hutchison et al., actually predates the development of microarray-based methods. However, this was applied to libraries of only approximately 1,000 transposon mutants in highly reduced Mycoplasma genomes, and the difficulty of sequencing at the time prevented wide spread adoption or high resolution. Modern high-throughput sequencing technology allows the methods discussed in this review to routinely monitor as many as one million mutants simultaneously in virtually any genetically tractable microorganism.

Protocols

Several methods were developed concurrently for high-throughput sequencing of transposon-insertion sites: TraDIS, INSeq, HITS and Tn-seq followed by Tn-seq Circle and refinements to the INSeq protocol. All of these protocols follow the same basic workflow with minor variations (see Fig. 1, Table 1): transposon mutagenesis and construction of pools of single insertion mutants; enrichment of transposon-insertion junctions and, finally, in some protocols a purification step either precedes or follows PCR enrichment before sequencing.

Figure 1. An illustration of the workflow typical of transposon-insertion sequencing protocols. Transposons are represented by pink lines, sequencing adaptors by blue, genomic DNA by black and PCR primers by green. Mutants are generated through either in vivo or in vitro transposition and subsequent selection for antibiotic resistance. These mutants are pooled, and optionally competed in test conditions, then genomic DNA is extracted and fragmented by restriction digest or physical shearing. Sequencing adaptors are ligated, some protocols then perform a step to purify fragments containing transposon insertions, and PCR with transposon- and adaptor-specific primers is used to specifically enrich for transposon-containing fragments. The fragments are then sequenced and mapped back to a reference genome to uniquely identify insertion sites with nucleotide-resolution. Dashed boxes indicate steps which differ between protocols.

Table 1. Applications of transposon-insertion sequencing

Study	Publication date	Organism	Total mutants	Unique insertion sites; Density	Application	Tn used	Name coined
Hutchison III et al., Science5	Oct 1999	M. genitalium/M. pneumoniae	1,291918	685; ~1/850 bp669; ~1/850 bp	Required gene sets	Tn4001	Global Transposon Mutagenesis
Goodman et al., Cell Host and Microbe7	Sep 2009	B. thetaiotaomicron	2 × 35,000	35,000;1/182 bp	Establishment in human gut as a natural habitat	Mariner	INSeq
Gawronski et al., PNAS8	Sep 2009	H. influenzae	75,000	55,935; 1/32 bp	Prolonged survival in lung in vivo	Mariner	HITS
van Opijnen et al., Nature Methods9	Oct 2009	S. pneumoniae	6 × 25,000	23,875; 1/91bp	Transcriptional regulation and carbohydrate transport	Mariner	Tn-seq
Langridge et al., Genome Research6	Dec 2009	S.Typhi 1	1.1 million	370,000;1/13 bp	Gene requirements, bile tolerance	Tn5	TraDIS
Gallagher et al., mBio10	Jan 2011	P. aeruginosa	~100,000	95,905; 1/65 bp	Tobramycin resistance	Mariner	Tn-seq (circle method)
Eckert et al., J Bact.24	Jan 2011	E. coli	19 × 95	1,645	Colonization of bovine intestinal tract; retrospective re-evaluation of a STM study	Tn5	-
Christen et al., Mol Syst Biol.16	July 2011	C. crescentus	800,000	428,735; 1/8 bp	Gene/ncRNAs/promoter requirements	Tn5	-
Griffin et al., PLOS Pathogens27	Sept 2011	M. tuberculosis	2 × 100,000	36,488;1/120 bp	Gene requirements and cholesterol utilization	Mariner	-
Khatiwara et al., AEM35	May 2012	S. Typhimurium	16,000	~8,000; ~1/610	Bile, low nutrient and heat tolerance	Tn5	-
Mann et al., PLOS Pathogens46	July 2012	S. pnuemoniae	~9,000–24,000	~8,000‒22,000	Determining roles of sRNAs in pathogenesis	Mariner	-
van Opijnen and Camilli, Genome Research22	July 2012	S. pnuemoniae	~4,000‒30,000	Varying	Stress response and metabolism in vitro and murine in vivo colonization	Mariner	-
Brutinel and Gralnick, Molecular Microbiology51	Aug 2012	S. oneidensis	50,000	26,793; ~1/191 bp	Gene requirements and Metabolism	Mariner	-
Zhang et al., PLOS Pathogens23	Sept 2012	M. tuberculosis	2 × 100,000	36,488; ~1/120 bp	Identifying genes, regulators and ncRNAs required for growth	Mariner	-
Klein et al., BMC Genomics52	Oct. 2012	P. gingivalis	N/A	54,000; 1/43 bp	Gene requirements	Mariner	-
Pickard et al., J Bact36	Jan 2013	S.Typhi 1	1.1 million	370,000; 1/13 bp	Bacteriophage infection	Tn5	-
Barquist et al., Nuc. Acids Res15	March 2013	S. Typhi1S. Typhimurium	1.1 million930,000	370,000; 1/13 bp550,000; 1/9 bp	Comparison of coding and non-coding gene requirements between two Salmonella serovars	Tn5	-

1 The same mutant library was used in these studies. A collection of studies to date utilizing transposon-insertion sequencing. Columns: (1) study reference, (2) date of publication, (3) organism mutagenized, (4) number of mutants generated, (5) number of unique insertion sites recovered from sequencing, (6) brief description of the application, (7) transposon used, (8) method name coined, if any.

Transposon mutagenesis

Most studies have used either Tn5 or mariner transposon derivatives. Tn5 originated as a bacterial transposon which has been adapted for laboratory use. Large-scale studies have shown that Tn5, while not showing any strong preference for regional GC-content, do have a weak preference for a particular insertion motif.- Transposon-insertion sequencing studies performed with Tn5 transposons in S. enterica serovars have reported a slight bias toward AT-rich sequence regions., However, this preference does not appear to be a major obstacle to analysis given the extremely high insertion densities obtained with this transposon,,(see Table 1). Additionally, Tn5 has been shown to be active in a wide range of bacterial species, though the number of transformants obtained can vary significantly depending on the transformation efficiency of the host.

Mariner/Himar1 transposons on the other hand originate from eukaryotic hosts and have an absolute requirement for TA bases at their integration site,, with no other known bias besides a possible preference for bent DNA. This can be a disadvantage in that it limits the number of potential insertion sites, particularly in GC-rich sequence. However, this specificity can also be used in the prediction of gene essentiality in near-saturated libraries: as every potential integration site is known and the probability of integration at any particular site can be assumed to be roughly equal, it is straight-forward to calculate the probability that any particular region lacks insertions by chance. Himar1 transposition can also be conducted in vitro in the absence of any host factors, and inserted transposons can then be transferred to the genomes of naturally transformable bacteria through homologous recombination. This can be advantageous when working with naturally transformable bacteria with poor electroporation efficiency., It is worth noting that Tn5 is also capable of transposition in vitro, and could potentially be used to increase insertion density and, hence, the resolution of the assay, particularly in GC-rich genomic regions.

Pool construction

Once mutants have been constructed, they are plated on an appropriate selective media for the transposon chosen, and colonies are counted, picked and pooled. A disadvantage of this is that the mutants must be recreated for follow up or validation studies. Goodman et al. introduced a clever way around this in the INSeq protocol: by individually archiving mutants, then sequencing combinatorial mutant pools, it is possible to uniquely characterize 2n insertion mutants by sequencing only n pools. Each mutant is labeled with a unique binary string that indicates which pools it has been added to. These binary strings can then be reconstructed for each insertion observed in these pools by recording their presence or absence in sequencing data, providing a unique pattern relating insertions to archived mutants. The authors control false identifications due to errors in sequencing by requiring that each binary label have a minimum edit distance to every other label, allowing for a robust association of labels with insertions despite sometimes noisy sequencing data. As a proof of concept, the authors were able to identify over 7,000 Bacteroides thetaiotaomicron mutants from only 24 sequenced pools. This effectively uses methods for the generation of random transposon pools to rapidly generate defined mutant arrays, though it is heavily dependent on liquid-handling robotics.

Enrichment of transposon-insertion junctions

Once pools have been constructed, they are grown in either selective or permissive conditions, depending on the experiment, and then genomic DNA is extracted. Fragmentation proceeds either through restriction digestion in the case of transposons modified to contain appropriate sites,,or via physical shearing,, then sequencing adapters are ligated to the resulting fragments. PCR is performed on these fragments using a transposon-specific primer and a sequencing adaptor-specific primer to enrich for fragments spanning the transposon-genomic DNA junction. Some protocols purify fragments containing transposon insertions using biotinylated primers,or PAGE before and/or after PCR enrichment. The purification step from the Tn-seq Circle protocol is particularly unusual in that restriction-digested fragments containing transposon sequence are circularized before being treated with an exonuclease that digests all fragments without transposon insertions, theoretically completely eliminating background. Given the success of protocols that do not include a purification step and the lack of systematic comparisons, it is currently unclear whether including one provides any major advantages.

Reproducibility, Accuracy and Concordance with Previous Methods

A number of studies have looked at the reproducibility of transposon-insertion sequencing. Multiple studies using different protocol variations have repeatedly shown extremely high reproducibility in the number of insertions per gene (correlations of ~90%) in replicates of the same library grown and sequenced independently,,, and good reproducibility (correlations between 70–90%) in independently constructed non-saturated libraries., Van Opijnen and Camilli compared traditional 1 × 1 competition experiments between wild-type and mutant Streptococcus pneumoniae to results obtained by transposon-insertion sequencing and showed that there was no significant difference in results over a range of tested conditions.

The accuracy of transposon-insertion sequencing in determining library composition has also been assessed. Zhang et al. constructed a library of identified transposon-insertion mutants in known relative quantities, and then were able to recover the relative mutant prevalence with transposon-insertion sequencing. Additionally, by estimating the number of PCR templates prior to enrichment, this study showed that there is a high correlation between enrichment input and sequencing output.

Two studies have evaluated concordance between results obtained with transposon-insertion sequencing and microarray monitoring of transposon insertions in order to demonstrate the enhanced accuracy and dynamic range of sequencing over previous methods. In the first, 19 libraries of 95 enterohemorrhagic Escherichia coli (EHEC) transposon mutants that had previously been screened in cattle using signature-tagged mutagenesis (STM), were pooled and re-evaluated using the TraDIS protocol. The original STM study had identified 13 insertions in 11 genes attenuating intestinal colonization in a type III secretion system located in the locus of enterocyte effacement (LEE). By applying sequencing to the same samples, an additional 41 mutations in the LEE were identified, spanning a total of 21 genes. Additional loci outside the LEE, which have been previously implicated in intestinal colonization but had not been detected by STM, were also reported by TraDIS.

The second study re-evaluated genes required for optimal growth determined by TraSH in Mycobacterium tuberculosis., The greater dynamic range of sequencing as compared with microarrays allowed easier discrimination between insertions that were nonviable and those that were only significantly underrepresented. The authors estimate that genes called as required by sequencing in their study are at least 100-fold underrepresented in the pool. In comparison, the threshold in the previous microarray experiment reported genes that had log probe ratios at least 5-fold lower than average between transposon-flanking DNA hybridization and whole genomic DNA hybridization. Additionally, the nucleotide-resolution of insertion sequencing allowed the authors to identify genes which had required regions, likely corresponding to required protein domains, but which tolerated insertions in other regions. Altogether, the authors increase the set of genes predicted to be required for growth in laboratory conditions in M. tuberculosis by more than 25% (from 614 to 774).

Gene Requirements

The earliest application of transposon-insertion sequencing was to determine the minimal set of genes necessary for the survival of Mycoplasma. This essential genome is of great interest to synthetic and systems biology where it is seen as a foundation for engineering cell metabolism, and in infection biology and medicine where it is seen as a promising target for therapies. However, it is important to remember that “essentiality” is always relative to growth conditions: a biosynthetic gene that is non-essential in a growth medium supplying a particular nutrient may become essential in a medium that lacks it. Traditionally, gene essentiality has been determined in clonal populations;,, since the high-throughput transposon sequencing protocols described here necessarily contain a short period of competitive growth before DNA extraction, many of these studies prefer to refer to the “required” genome for the particular conditions under evaluation.

Because of this short period of competitive growth, and because many otherwise required genes tolerate insertions in their terminus,,or outside essential domains, the determination of required genomic regions is not completely straight-forward and a number of approaches have been taken to counter this. These include only calling genes completely lacking insertions as required, determining a cut-off based on the empirical or theoretical distribution of gene-wise insertion densities.,,, Additionally, windowed methods have been developed which can be used to identify essential regions in the absence of gene annotation,, and have had success in identifying required protein domains, promoter regions and non-coding RNAs (ncRNAs). The organisms that have been evaluated for gene requirements under standard laboratory conditions are summarized in Table 1.

In agreement with previous studies,, many required genes identified by transposon-insertion sequencing are involved in fundamental biological processes such as cell division, DNA replication, transcription and translation,,,,and many of these requirements appear to be conserved between genera and classes., However, a recent study defining required gene sets in Salmonella serovars has found that phage repressors, necessary for maintaining the lysogenic state of the prophage, are also required, even though mobile genetic elements such as phage are usually considered part of the accessory genome. This study also highlights the need for temperance when interpreting the results of high-throughput assays of gene requirements. For example, many genes in Salmonella Pathogenicity Island 2 (SPI-2) did not exhibit transposon-insertions, despite clear evidence from directed knockouts showing that these genes are non-essential for viability or growth. Under laboratory conditions, SPI-2 is silenced by the nucleoid-forming protein H-NS,, which acts by oligermerizing along silenced regions of DNA blocking RNA polymerase access. A previous study has shown that transposon insertion “cold spots” can be caused by competition between high-density proteins and transposases for DNA. This suggests that H-NS may be restricting transposase access to DNA, though this has not previously been observed in transposon-insertion sequencing data, and will require additional work to confirm.

Defining Conditional Gene Requirements

One of the most valuable applications of the transposon-insertion sequencing method is the ability to identify genes important in a condition of interest, by comparing differences in the numbers of sequencing reads from input (control) mutant pools to output (test) pools that have been subject to passaging in a certain growth condition. Insertion counts are compared from cells in the input pool and those after passage, thereby identifying genes that either enhance or detract from survival and/or growth in the given condition, defined by decreased or increased insertion frequency, respectively. A further application of this method involves comparing insertions between biologically linked conditions, such as cellular stresses or different stages of a murine infection, to gain insight into complex systems.

So far, transposon-insertion sequencing has been used to investigate a number of interesting biological questions: bile tolerance in S. Typhi and S. Typhimurium, bacteriophage infection of S. Typhi, antibiotic resistance in Pseudomonas aeruginosa, cholesterol utilization in M. tuberculosis and a number of stress and nutrient conditions in S. pneumoniae. Transposon-insertion sequencing of populations passed through murine models have been used to assess genes required to establish the gut commensal B. thetaiotaomicron in its niche, for Hemophilus influenzae infection, as well as S. pneumoniae responses to two in vivo niches, the lung and nasopharynx. A further extension of the method examined double mutant libraries, that is transposon mutant libraries generated in a defined deletion background, to tease apart complex networks of regulatory genes.

Two studies in particular illustrate the power of using transposon-insertion sequencing to identify conditionally required genes. In the first, Goodman et al. set out to determine the genes necessary for the establishment of the commensal B. thetaiotaomicron in a murine model. First, the growth requirements of transposon mutant populations in the cecum of germ-free mice was assessed, and genes required for growth in monoassociation with the host were found to be enriched in functions such as energy production and amino acid metabolism. By further comparing monoassociated transposon mutant libraries with those grown in the presence of three defined communities of human gut-associated bacteria, the authors identified a locus upregulated by low levels of vitamin B12 that is only required in the absence of other bacteria capable of synthesizing B12. This showed that the gene requirements of any particular bacterium in the gut are at least partially dependent on the metabolic capabilities of the entire community and emphasizes the importance of testing in vivo conditions to complement in vitro study.

The second study, conducted by van Opijnen and Camilli, aimed to map the genetic networks involved in a range of cellular stress responses in S. pneumoniae. Seventeen in vitro conditions were tested, including: pH, nutrient limitation, temperature, antibiotic, heavy metal and hydrogen peroxide stress. Approximately 6% of disrupted genes resulted in increased fitness in some condition, suggesting that some genes are maintained despite being detrimental to the organism under particular conditions. These would be interesting candidates for further functional and evolutionary study, as the maintenance of these genes is presumably highly dependent on the conditions the bacteria faces, and may have implications for our understanding of e.g., gene loss in the process of bacterial host adaptation. Two additional in vivo experiments were performed in a murine model, where cells were recovered from the lung and nasopharynx. Combining this data, over 1,800 genotype-phenotype genetic interactions were identified. These interactions were mapped and pathways identified. Between the two in vivo niches, certain stress response pathways were markedly different. For example, temperature stress produced a distinct response in the lung, compared with the nasopharyanx, which is perhaps to be expected as temperature varies greatly between these two sites. By further examining sub-pathways required in the two different niches and comparing them to in vitro requirements, the authors were able to draw conclusions regarding the condition S. pneumoniae faces when establishing an infection. This comprehensive mapping of genotype-phenotype relationships will serve as an important atlas for further studies.

Monitoring ncRNA Contributions to Fitness

To date, four studies have used transposon-insertion sequencing to examine the contribution of non-coding RNAs (ncRNAs) and other non-coding regions to organismal fitness (see Table 1). Two of these examined requirements for non-coding regions in the relatively under-explored bacterial species Caulobacter crescentus and M. tuberculosis. Both utilized analytical techniques that allowed for the identification of putative required regions in the absence of genome annotation. Twenty-seven small RNAs (sRNAs) had previously been detected in C. crescentus; six were found to be depleted in transposon insertions indicating an important role in basic cellular processes. Additionally, the well-characterized ncRNAs tmRNA and RNaseP, as well as 29 non-redundant tRNAs, were found to be required. An additional 90 unannotated non-disruptable regions were identified throughout the genome, implying an abundance of unexplored functional non-coding sequence.

While the non-coding transcripts of M. tuberculosis have been explored more thoroughly than those of C. crescentus, most remain functionally uncharacterized, though there are hints that some of these may be involved in pathogenicity. Using a mariner transposon-based assay and a windowed statistical analysis that accounted for the distribution of potential TA integration sites, 35 intergenic regions were identified as putatively required in the M. tuberculosis genome. In common with the C. crescentus study, the RNA component of RNase P, required for the maturation of tRNAs and tmRNA, involved in the freeing of stalled ribosomes, were identified as required (Fig. 2A) together with 10 non-redundant tRNAs and potential promoter regions. However, due to the lower overall insertion density and lack of TA sites in some GC-rich regions, there were some regions that could not be assayed and the resolution was limited to 250 bases.

Figure 2. Applications of transposon-insertion sequencing to non-coding RNAs. (A) Plots of genomic regions in Mycobacterium tuberculosis containing the required non-coding RNAs RNase P (top) and tmRNA (bottom). Tracks, from top to bottom, 1. Histogram of insertion counts, 2. Comprehensive heat-map of requirement of 500-bp windows, 3. Position of annotated genes, 4. Position of TA dinucleotide sites, 5. Position of non-coding RNA. Reproduced from reference 23. (B) 1 × 1 competition assays validate attenuating Streptococcus pneumoniae sRNA mutants identified by transposon-insertion sequencing. Mice were infected with defined deletions of sRNAs identified as attenuating by Tn-seq and wild-type S. pneumoniae TIGR4 at the body site indicated and bacterial densities were compared 24 h post-infection. These plots show the derived competitive index in blood (top) and the nasopharnyx (bottom). Each point represents the result of a competition experiment between an sRNA deletion mutant and wild-type TIGR4. A competitive index of 1 indicates equivalent numbers of mutants and wild-type were recovered. Modified from reference 46.

A recent study has examined ncRNA requirements in the S. enterica serovars Typhi and Typhimurium. Using the tRNAs as a model set of ncRNAs, this study showed that the high transposon insertion density achieved by the TraDIS protocol is capable of assaying the requirement for genomic regions as small as 70‒80 bases. S. enterica, together with the closely related E. coli, has served as a model organism for the discovery and elucidation of ncRNA function, and extensive annotations of non-coding transcripts are available.- As a result, this study was able to assay approximately 300 non-coding regions with evidence for function or transcription. Among the ncRNAs identified as required were RNase P; the RNA component of the signal recognition particle, involved in targeting proteins to the plasma membrane; and a number of known autoregulatory ribosomal protein leader sequences, as well as providing evidence for a novel leader sequence, StyR-8, which appears to be involved in the autoregulation of the rpmB gene. In total, this study identified 15 confirmed and putative ncRNAs required for robust competitive growth on rich media in both serovars, including a number of known sRNAs involved in stress response.

A particularly exciting study has been conducted in S. pneumoniae TIGR4 combining RNA-seq with transposon-insertion sequencing. To identify sRNA loci, the authors first sequenced size-selected RNA from the wild-type and three two-component system knockouts, identifying 89 putative sRNAs, 56 of which were novel. Fifteen of these candidates, selected on the basis of high expression and low predicted folding free energy, were assayed for their ability to establish invasive disease in a murine model. Of these, eight sRNA deletions showed a significant attenuation of disease. To more broadly establish the roles of sRNAs in infecting particular organs, transposon insertion libraries were administered directly to the nasopharnyx, lungs or blood of mice, and bacteria were harvested following disease progression. Twenty-six, 28 and 18 sRNAs were found to attenuate infection in the nasopharnyx, lung and blood, respectively. These results were then validated with targeted deletions of 11 sRNAs (Fig. 2B). In addition to establishing the role of sRNAs in S. pneumoniae virulence, this study illustrated the power of combining RNA-seq and transposon-insertion sequencing to rapidly assign phenotypes to non-coding sequences.

Limitations

In this review, we have largely focused on the potential of transposon-insertion sequencing. However, this technology does have a number of important limitations, which we collect here and summarize in Table 2. As discussed previously, requirements for particular nucleotides at insertion sites, such as the TA required by the Mariner transposon, or preference for certain sequence composition, such as the AT bias exhibited by Tn5, can limit the density of observed insertions in certain genomic regions. This may impact any downstream analysis, and can potentially bias results, particularly the determination of gene requirements. Even if this bias has been accounted for, transposon-insertion screens will always over-predict gene requirements in comparison to targeted deletion libraries as discussed previously. However, this over-prediction can be controlled either through careful consideration of known insertion biases as in many Mariner-based studies, or by high-insertion densities, such as those achieved in several Tn5-based studies (Table 1). Once the library has been created, only regions that have accumulated insertions in the conditions of library creation will be able to be assayed for fitness effects in further conditions. This means that regions that lead to slow growth phenotypes when disrupted in standard laboratory conditions may be difficult to assay in other conditions. Additionally, the dynamic range of fitness effects detected will depend on the complexity of the input library(s). The absence of insertions may be a particular problem for assaying small genomic elements, such as sRNAs or short ORFs. Finally, the validation of hypotheses derived from transposon-insertion sequencing will require the construction of targeted deletions, as individual mutants cannot be recovered from pools unless specialized protocols have been followed during library construction (as in Goodman, 2009).

Table 2. Advantages and limitations of transposon-insertion sequencing

Advantages	Limitations
Library construction is extremely rapid in comparison to targeted deletion libraries.	Requirements for particular nucleotides at transposon-insertion sites or insertion biases can limit resolution.
Gene requirements and fitness effects can be quickly assayed in a wide range of conditions.	Determination of gene essentiality is dependent on insertion density, and is less conclusive than targeted gene deletion in clonal populations.
The precise location of transposon insertions can be determined due to the nucleotide resolution of high-throughput sequencing.	Only genomic regions that tolerate insertions under the conditions of library creation may be assayed for fitness effects in further conditions.
Wide dynamic range compared with older microarray-based technologies.	The dynamic range for fitness effects is dependent on mutant abundance in the initial library and may be limiting for some genes.
Requirements and fitness effects of genomic regions can be determined independently of annotation.	Mutants must be reconstructed for follow-up experiments in the absence of specialized protocols and robotics (see e.g., Goodman, 20097).

The Future of Transposon-Insertion Sequencing

Transposon-insertion sequencing is a robust and powerful technique for the rapid connection of genotype to phenotype in a wide range of bacterial species. Already, a number of studies have demonstrated the effectiveness of this method and the results have been far-reaching: enhancing our understanding of basic gene functions, establishing requirements for colonization and infection, mapping complex metabolic pathways and exploring non-coding genomic “dark matter.” Due to the range of potential applications of transposon-insertion sequencing, along with the decreasing cost and growing accessibility of next-generation sequencing, we believe that this method will become increasingly common in the near future.

A number of bacterial species have already been subjected to transposon-insertion sequencing (Table 1). Microarray-based approaches to monitoring transposon mutant libraries have even been applied to eukaryotic systems, and similarly transposon-insertion sequencing can potentially be applied to any system where the creation of large-scale transposon mutant libraries is technologically feasible. Recently, the Genomic Encyclopedia of Bacteria and Archea (GEBA) has been expanding our knowledge of bacterial diversity through targeted genomic sequencing of underexplored branches of the tree of life. Applying transposon-insertion sequencing in a comparative manner across the bacterial phylogeny will provide an unprecedented view of the determinants for survival in diverse environments. While most transposon-insertion sequencing studies to date have focused on pathogenic bacteria, these techniques could also have applications in energy production, bioremediation and synthetic biology.

The combination of transposon-insertion sequencing with other high-throughput and computational methods is already proving to be fertile ground for enhancing our understanding of bacterial systems. For instance, by using transposon-insertion sequencing in a collection of relatively simple conditions combined with a computational pathway analysis, van Opijnen and Camilli were able to provide a holistic understanding of the genetic subsystems involved in a complex process such as S. pneumoniae pathogenesis. In the future, methods to assay phenotype in a high-throughput manner,may be combined with transposon-insertion sequencing to provide exhaustive simple genotype-phenotype associations with which to understand complex processes in a systems biology framework. We look forward to the adoption of these data sets by the community as an important tool for rapid hypothesis generation.