| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Review |
1 Department of Molecular, Cellular, and Developmental Biology, University of Colorado, Boulder, CO 80309, USA
2 Institute of Arctic and Alpine Research (INSTAAR) and Environmental Studies Program, University of Colorado, Boulder, CO 80309, USA
3 Department of Chemistry and Biochemistry, University of Colorado, Boulder, CO 80309, USA
Correspondence
Rob Knight
Rob.Knight{at}colorado.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT IntroductionInferring HGT from genomic...Detecting mechanisms of...Detecting mechanisms of...Prospects for sequence-based...ConclusionREFERENCES |
|---|
|
Introduction |
|---|
|
TOPABSTRACTIntroductionInferring HGT from genomic...Detecting mechanisms of...Detecting mechanisms of...Prospects for sequence-based...ConclusionREFERENCES |
|---|
; Frost et al., 2005). HGT can promote adaptation to new environments by allowing molecular innovations to be traded within ecological niches. For example, HGT has enhanced bacterial adaptation to life in environments containing heavy metals (Liebert et al., 1999) and toxic organic compounds (Top & Springael, 2003). Unfortunately, HGT has also allowed adaptation to the human host environment, both by spreading pathogenicity determinants (Canchaya et al., 2003) and by allowing pathogenic bacteria to acquire resistance to multiple classes of antibiotics (Walsh, 2006). Thus, HGT has diverse and important societal implications, both positive and negative, and deserves detailed study.
An understanding of how microbial genomes have been shaped by HGT events has been pursued in parallel using genetic, biochemical and computational approaches. Genetic and biochemical approaches to HGT have succeeded in elucidating many of the mechanisms by which HGT is accomplished (Curcio & Derbyshire, 2003; Averhoff, 2004; Chen et al., 2005). However, because the pattern of horizontal transfers that has affected each genome is unique, and because these horizontal transfers have occurred over very long timescales, the history of HGT in a particular genome cannot be uncovered in an experimental setting. Instead, past horizontal transfer events must be inferred from the molecular fingerprints that those events have left on extant microbial genomes. These inferences are made using computational approaches.
|
Inferring HGT from genomic data |
|---|
|
TOPABSTRACTIntroductionInferring HGT from genomic...Detecting mechanisms of...Detecting mechanisms of...Prospects for sequence-based...ConclusionREFERENCES |
|---|
; Page & Charleston, 1997). Additionally, even in a genome in which all genes share the same history, the sequence of some genes may not support the true phylogeny. This effect is especially important where the phylogenetic signal is weak, and thus susceptible to being overwhelmed by random noise.
Techniques related to these phylogenetic methods include distributions of BLAST hits (e.g. Nelson et al., 1999; Lander et al., 2001), inference of gene presence/absence within species phylogenies (e.g. Jordan et al., 2001; Mirkin et al., 2003), and ratios of evolutionary distances (e.g. Farahi et al., 2003; Kechris et al., 2006). These techniques, however, have been shown to detect only a subset of horizontally transferred genes, and therefore should be treated as approximations to, not replacements for, phylogenetic methods (Ragan, 2001; Kinsella & McInerney, 2003; Ragan et al., 2006).
Compositional methods, in contrast, examine sequence characteristics that vary in different taxa but are relatively conserved within a taxon. If a gene or genomic fragment possesses unusual sequence characteristics, it may have been transferred from a taxon in which those characteristics are typical. Unlike phylogenetic methods, compositional methods do not require alignment of homologous sequences, and are therefore better suited to examine poorly conserved or very rapidly evolving genes. Sequence characteristics that have been used to study HGT include GC content (in the whole gene or at the third position in each codon), the codon adaptation index, amino acid usage, relative synonymous codon usage and dinucleotide usage (Sharp & Li, 1987; Lawrence & Ochman, 1997, 1998; Karlin et al., 1998; Hooper & Berg, 2002). Markov models, including frame-dependent Markov models (Nakamura et al., 2004) and variable-order Markov models (Vernikos & Parkhill, 2006) have also been used. One disadvantage of compositional methods is that transfer between related or unrelated strains with similar compositional characteristics will not be detected. Conversely, if a sequence characteristic varies within a genome, then classes of genes, such as those encoding ribosomal proteins, that have unusual compositions may appear to have been horizontally transferred. Even when the composition of a transferred gene is initially distinct from the composition of the recipient genome, the composition of the gene will drift toward that of the recipient genome over time. Thus, the traces of ancient transfers may be obliterated in sequences that have had time to equilibrate fully (Lawrence & Ochman, 1997).
Benefits of differentiating between HGT mechanisms
Current phylogenetic and sequence composition-based methods for detecting horizontally transferred genes in genomic data treat all horizontally transferred genes as equivalent. We argue that these computational approaches to identifying HGT could be improved by incorporating recent insights from genetic and biochemical studies into the mechanisms that allow gene transfers to occur. Molecular biology has uncovered sequence features associated with many, though by no means all, HGT mechanisms, and these sequence features can be exploited to categorize many putatively transferred genes according to the mechanism of transfer.
Categorizing HGT events that can be identified by mechanism could provide three important benefits. First, such categorization could allow testing of whether the genes transferred by different mechanisms vary in properties such as composition or cellular function. Second, automatic identification of sequence features could allow for the construction of high-confidence datasets of horizontally transferred genes. Third, such categorization could allow the construction of mechanism-specific models that may detect HGT with greater sensitivity than the general techniques currently employed.
The first benefit of categorizing HGT genes by transfer mechanism is that such categorization could allow for quantification of differences in the properties of mechanisms that can be identified by sequence features. Different mobile elements or mechanisms of transfer may tend to transmit distinct classes of genes, predominate in specific environmental conditions, or transfer genes between hosts separated by variable phylogenetic distances. It is further possible, as suggested by one of our reviewers, that genes with different degrees of divergence in sequence or in compositional characteristics may be transferred by specific mechanisms (although the latter case may be difficult to disentangle from compositional differences induced by residency in a particular mobile element). It is also possible that several of these factors may interact in complex ways to determine which genes are transferred by a particular mechanism. Exploring these potential differences between mechanisms of transfer may help us to understand the different evolutionary implications of distinct classes of transfer events.
The second benefit of categorizing putative HGT genes by their mechanism of transfer is that such categorization may be used to generate a higher-confidence pool of HGT genes from a large set of putative HGT genes detected using phylogenetic or compositional methods. This can be accomplished by searching the region around each putatively transferred gene for the presence or absence of sequence features that indicate a mechanism of HGT. For example, if a gene is found to be integrated into a tRNA gene, and several phage genes are located nearby, then not only has very strong additional support been generated for the hypothesis of transfer, but a specific mechanism (phage transduction) can be proposed. However, the converse does not hold. Because not all mechanisms of transfer can be distinguished on the basis of sequence features alone, the absence of sequence features indicating a known mechanism does not by itself demonstrate that a gene was not transferred (Table 1). The pool of high-confidence HGT genes generated in this manner will necessarily be limited to genes transferred by mechanisms identifiable from their sequence features.
|
). IslandPath performs several kinds of compositional analyses, and annotates integrases and transposases in a genome. Information about the location of genes associated with mobility was later used to provide support for genomic islands predicted on the basis of unusual dinucleotide usage (Hsiao et al., 2005). This approach could also be used to estimate what proportion of putative gene transfers are accounted for by mechanisms identifiable from their sequence features. Genes with evidence of transfer but that lack sequence features associated with a known transfer mechanism could be transfers effected by known mechanisms that are not identifiable by sequence features, false positives of the compositional or phylogenetic method used to detect transfer, or genes transferred by a previously unknown mechanism.
The third benefit of categorizing HGT genes by transfer mechanism is that distinguishing between different types of transfer could allow for the construction of specialized, mechanism-specific compositional methods for detecting HGT events effected by different mechanisms. Existing compositional methods are general, and attempt to detect all HGT events. Categorizing HGT events by mechanism would allow for mechanism-specific compositional models to be built, which could improve sensitivity when searching for genes transferred by that mechanism. For example, the GC content of plasmids tends to be lower than that of the recipient genome; thus we might expect that building compositional models for this element would improve the detection of genes transferred by plasmid integration (van Passel et al., 2006).
Thus, we propose that automated annotation of the mechanism of gene transfer using sequence features present in the recipient genome will, where it is possible, provide a useful complement to existing compositional and phylogenetic methods for studying HGT.
Mobile DNA elements are mosaics of functional modules
One significant barrier to incorporating information about the mechanism of HGT into genomic analyses is the sheer diversity of mobile elements that are known to exist. Mobile elements as varied as conjugative transposons, transposable prophages, integrative plasmids and mobile integrons appear on first inspection to be extremely difficult to detect and categorize systematically. Indeed, early attempts to classify mobile elements as phage, plasmid or transposon were thwarted by the discovery of new elements, such as conjugative transposons, which possess unique combinations of previously known mobility determinants (Franke & Clewell, 1981; Osborn & Boltner, 2002). A key insight that has helped to organize and interpret this diversity of gene transfer mechanisms is that mobile genomic elements can be thought of as the sum of a set of relatively independent and exchangeable functional units (Osborn & Boltner, 2002; Toussaint & Merlin, 2002). These functional units can broadly be categorized as promoting intercellular or intracellular mobility, or replication, selection, stability and maintenance (Toussaint & Merlin, 2002).
Rather than creating an ever-expanding list of new categories of mobile elements, the mechanism-centric approach to mobile element diversity attempts to link particular mobility functions to particular sets of genes and sequence features. Thus, although the combinatorial nature of functional modules composing mobile elements makes classification difficult at the level of the element as a whole, this approach could allow for the functional capabilities of a particular mobile element to be predicted from its sequence, and provide the necessary framework to identify and describe mobile elements that do not fit into classical categories. This way of thinking about the diversity of mobile elements is now the basis of the ACLAME database, which seeks to annotate mobile genetic elements on the basis of the functional modules they contain (Leplae et al., 2004).
In order to facilitate comparisons of the types of genes transferred by different mechanisms, we review the known mechanisms of inter- and intragenomic gene transfer, emphasizing the distinct evolutionary pressures and mechanistic restraints each imposes, and the characteristics likely to prove most useful in distinguishing HGT mechanisms when examining genomic data.
|
Detecting mechanisms of intercellular DNA mobility using genomic data |
|---|
|
TOPABSTRACTIntroductionInferring HGT from genomic...Detecting mechanisms of...Detecting mechanisms of...Prospects for sequence-based...ConclusionREFERENCES |
|---|
). Mechanisms of intercellular DNA transfer include conjugation, phage transduction, and transformation.
|
; Franke & Clewell, 1981; Burrus et al., 2002). Conjugative DNA transfer is widespread, having been detected in both Gram-positive and -negative bacteria, and in Crenarchaeote but not Euryarchaeote archaea (Prangishvili et al., 1998 and references therein; Stedman et al., 2000).
DNA transfer during conjugation is effected by type IV secretory systems (TIVSSs). TIVSSs transfer only single-stranded DNA, so the DNA to be transferred must be separated from its complementary strand prior to transfer. In order to separate the two strands, the DNA is nicked. This nicking is achieved by a relaxase, which binds to a cis element called an origin of transfer (oriT). Relaxase binding promotes the formation of a relaxosome complex, and nicks the DNA at a conserved nic motif. Although relaxases can be supplied in trans (for example by a helper plasmid), conjugative DNA elements must contain an origin of transfer in cis in order to be moved through the conjugation apparatus (Grohmann et al., 2003 and references therein). Because oriT elements appear to be an absolute cis requirement for conjugal transfer, these sequences are good candidates for the identification of mobile elements or genomic regions that may be capable of movement by conjugative transfer (Table 2). A second property that makes these sequences good candidates for genomic markers of transfer by conjugation is their conservation (Fig. 2). The nic motifs of oriT sequences necessary for conjugative transfer fall into five conserved classes. Origins of transfer also fall into five conserved families, and alignments of these sequences are available (e.g. Lanka & Wilkins, 1995; Zechner et al., 2000). Transfer by conjugation could be detected by searching the region around each putatively transferred gene for origins of transfer using similarity searches against profiles of sequences from each oriT family (Altschul et al., 1997).
|
|
and references therein). These clt sequences lack sequence similarity to origin of transfer (oriT) sequences required for traditional conjugation, and appear to act by a novel mechanism. Unlike conventional conjugation, which requires a number of components, only a single tra gene is necessary in trans for conjugation to occur (Grohmann et al., 2003) (Table 2
). The tra genes from several Streptomyces plasmids appear to be related to the SpoIIIE/FtsK septal DNA transport proteins. SpoIIIE is involved in the transport of double-stranded DNA into the pre-spore of Bacillus subtilis (Errington, 2001), suggesting that TraA may also transfer double-stranded DNA (dsDNA) rather than single-stranded DNA as in traditional conjugation. Experiments using restriction enzymes support this notion. Conjugative transfer by TraA is sensitive to dsDNA-specific SalI restriction digestion, and is thus also likely to involve the transfer of dsDNA (Possoz et al., 2001). So far, the novel conjugation mechanism of the PIJ101 and pJV1 plasmids of Streptomyces has not been explored thoroughly enough to provide sequences that could be exploited in genomic studies. Indeed, although direct and/or inverted repeats appear to play a role in all known examples, the sequence and secondary structure of clt sequences appear to be highly variable (Servin-Gonzalez, 1996; Franco et al., 2003). Developing methods for detecting and annotating clt sites in silico remains a challenge for the future.
Phage transduction
Phage transduction as a mechanism of genetic exchange was discovered by Zinder & Lederberg (1952) in Salmonella typhimurium (Table 3). Phage gain entry to their microbial host cell by binding to extracellular receptors on the cell surface. Following entry into the host cell, lytic phage typically replicate numerous copies of their genome, and then exit the cell en masse, causing cell lysis. In addition to immediate replication and destruction of the host cell, temperate or lysogenic phage are also capable of integrating into the host genome, where they can lie dormant, replicating along with the host.
|
).
Phage transduction can be broadly separated into specialized and generalized transduction (Table 2).
Specialized transduction.
Specialized transduction occurs in phage that preferentially integrate at particular attachment sites in the microbial host genome (attB) through integrase-mediated recombination with a second attachment site in the phage genome (attP). If such a phage is incorrectly excised from its host genome, then small regions of host DNA on either side of the prophage may be packaged along with the prophage genome (Canchaya et al., 2003). Since the site of phage integration is highly specific as long as the host attachment site is present, the variety of genes that can be transferred from a particular host by specialized transduction is usually quite limited. Nonetheless, pathogenicity determinants have been shown to have been acquired in this manner (Canchaya et al., 2003 and references therein).
Generalized transduction.
In contrast to specialized transduction, generalized transduction can in principle facilitate the movement of any gene. The best-studied example of generalized transduction is Mu phage. Mu phage integrates randomly, and upon excision packages its DNA by recognition of a pac site within the Mu prophage sequence (Harel et al., 1990). Occasionally, however, Mu may package a segment of DNA from its bacterial host instead. In that case the Mu integrase is not packaged, and the transduced DNA must enter the recipient genome by homologous recombination in a RecA-dependent fashion (Bushman, 2002). Even when acting normally, Mu can transduce smaller regions of host DNA. The Mu head can hold more DNA than the Mu phage itself encodes; up to 100 bp of DNA upstream of the prophage, and 2000 bp downstream of the prophage may be included to fill the phage head (Bushman, 2002). Following infection of a new bacterial host, this extra genetic material can be integrated alongside the phage genome.
Genes transduced by phage may be detected by their proximity to genes with sequence similarity to known phage integrases and structural genes, the presence of direct repeats, and integration near tRNA genes (Williams, 2002) (Table 4). Two programs for detecting prophage in genomic data include Phage_Finder and ProphageFinder (Bose & Barber, 2006; Fouts, 2006). Phage_Finder functions by using BLAST and phage-specific sequences to search genomes for regions containing phage genes. It then removes putative integrons and bacteriocins, and is also capable of further subclassifying phage (for example, by identifying Mu-like phage) (Fouts, 2006). ProphageFinder uses BLASTX to search user-defined regions for phage genes (Bose & Barber, 2006).
|
; Bertani, 1999). These elements are termed gene transfer agents (GTAs), and were originally identified in the photosynthetic purple nonsulphur bacterium Rhodobacter capsulatus (Marrs, 1974). Subsequently, GTAs have been identified in the Euryarchaeote Methanococcus voltae PS, and in a variety of diverse bacteria (Humphrey et al., 1997; Bertani, 1999; Lang & Beatty, 2007 and references therein). Since no phage integrase is packaged, these small transduced fragments must enter the genome through recombination. At least one phylogenetic analysis suggests that GTAs arose only once in the 
-proteobacteria (Lang & Beatty, 2007), although the origin of the archaeal GTA in Methanococcus voltae PS almost certainly represents a separate event.
Because GTAs do not transfer phage structural genes or integrases, and because they are believed to rely on homologous recombination to enter the target genome, GTA-mediated transfers are likely to be much more difficult to detect, and to distinguish from other gene transfer mechanisms than traditional phage transduction. Two criteria may nonetheless help to do so. First, regions of DNA packaged into GTAs are believed to be short (
4.5 kb). Secondly, the organism donating DNA by this mechanism must contain GTA genes. Thus, in cases where the donor lineage can be discerned, the possibility of transfer by a GTA could be established by demonstrating that GTAs are common in the donor lineage. For example, the widespread distribution of GTAs in -proteobacteria suggests that GTA-mediated transfer should be considered a possibility when examining short transferred fragments of -proteobacterial origin. In the future, identification of the cell-surface receptors for known GTAs could help to identify lineages that have the capability of acquiring DNA in this manner.
Transformation
Transformation is the uptake of circular or linear DNA from the environment (Dubnau, 1999). Natural transformation was first documented in bacteria, a discovery that led to the finding that DNA is the genetic material (Griffith, 1928; Avery et al., 1944). More recently, natural transformation has also been reported within the archaea Methanococcus voltae PS (Bertani & Baresi, 1987), Methanobacterium thermoautotrophicum Marburg (Worrell et al., 1988) and Thermococcus kodakaraensis KOD1 (Sato et al., 2003), although the transformation mechanism used by the archaea is not yet known.
Transformation has several stages: induction of competence (in strains that are not constitutively competent), availability of DNA in the medium, DNA binding, DNA fragmentation, transport across the outer membrane (in Gram-negative bacteria) or the cell wall (in Gram-positives), conversion to single-stranded DNA, and transport across the cytoplasmic membrane (Dubnau, 1999; Averhoff, 2004; Chen et al., 2005). If the transformed DNA is to be stably maintained in the cell, it must then recombine with host, plasmid or phage DNA or, in the case that a plasmid was taken up from the environment through the transformation apparatus, recircularize (for several excellent reviews, see Lorenz & Wackernagel, 1994; Averhoff, 2004; Chen et al., 2005; Thomas & Nielsen, 2005).
In all naturally transformable bacteria studied to date, with the notable exceptions of Helicobacter pylori and Camplyobacter jejuni (see below), transformation depends on some protein components also utilized in type IV pilus formation, twitching mobility, and the type II secretory system (TIISS) (Dubnau, 1999; Averhoff, 2004). These proteins are exemplified by the Com proteins of Bacillus subtilis. Although components of type IV pili are necessary for natural competence, it appears that the pili themselves are dispensable (Averhoff, 2004). This could be explained if the transformation apparatus utilizes a pseudopilus, which competes for common components with type IV pili: Chen et al. (2005) have recently proposed a model for transformation in B. subtilis in which the uptake of DNA is driven by the force of repeated cycles of polymerization and depolymerization of the pseudopilus, where depolymerization of the pseudopilus is driven by the proton-motive force. Once single-stranded DNA has begun to enter the cytoplasm, the binding of single-strand binding (SSB) proteins is proposed to aid uptake through a Brownian ratchet mechanism (Chen et al., 2005 and references therein).
In Neisseria and the Pasteurellaceae, the transformation apparatus will only take up environmental DNA that contains a specific 9–11 nucleotide sequence. These sequences, called uptake signal sequences (USSs) or DNA uptake sequences (DUSs), are the most common sequences of their length in their genome (Davidsen et al., 2004 and references therein) (Fig. 2). No receptor that recognizes USSs has yet been found. Other naturally transformable bacteria such as Bacillus subtilis and Streptococcus pneumoniae do not appear to require such sequences (Dubnau, 1999). Likewise no requirement for signal sequences has yet been found in naturally transformable archaea.
The genomes of organisms that require signal sequences to achieve DNA uptake provide useful model systems for studying transformation-mediated HGT, because genes within them lacking a DUS are unlikely to have been acquired by transformation in the recent past (Table 2). Conversely, overrepresentation of DUSs within a class of genes may indicate that those genes are frequently horizontally transferred by transformation. As an example of this type of reasoning, Davidsen et al. (2004) demonstrated that DNA uptake signals are overrepresented in repair and recombination genes that promote genome maintenance, suggesting that transformation and recombination may help to protect DNA repair genes from genotoxic stress. This overrepresentation could not be accounted for by the dinucleotide composition of genome maintenance genes, and was found in all genomes known to contain DUSs. No overrepresented DNA sequences of similar length were found in E. coli, which is not believed to be naturally transformable (Lorenz & Wackernagel, 1994; Davidsen et al., 2004).
Transformation in H. pylori and C. jejuni.
An unusual system for transformation is found in Helicobacter pylori (Table 3). Transformation in Helicobacter is mediated by a TIVSS that contains components homologous to many elements of the Agrobacterium tumefaciens conjugal T-DNA transfer system (Hofreuter et al., 2001). A transmissible TIVSS involved in transformation has also been reported on a plasmid of Camplyobacter jejuni (Bacon et al., 2000). As yet, no special signal sequences have been found that are required for DNA uptake through the TIVSS of Helicobacter or Camplyobacter.
|
Detecting mechanisms of intracellular DNA and RNA mobility in genomic data |
|---|
|
TOPABSTRACTIntroductionInferring HGT from genomic...Detecting mechanisms of...Detecting mechanisms of...Prospects for sequence-based...ConclusionREFERENCES |
|---|
Integration and excision
Excision from the donor genome and integration into the recipient genome are crucial steps in the horizontal transfer of many types of mobile elements, including prophage, integron gene cassettes, and ICEs.
Site-specific integrases catalyse integration at a specific site in the host genome. For phage integrases, these are called attB sites. Frequently, attB integration sites are located within tRNA or tmRNAs, with different integrase subfamilies possessing specificity for different sublocations (such as the region coding for the anticodon loop) within the tRNA gene (Williams, 2002). Related integrase genes found on plasmids also allow site-specific integration, and share similar attachment site preferences (Williams, 2002).
A unique site-specific integrase is found in the archaea. The Sulfolobus spindle virus 1 (SSV1) integrase appears to recognize an attP site within the integrase gene itself (Table 3). This causes the integrase gene to be partitioned into an N-terminal and C-terminal fragment upon entry, apparently trapping it in the genome until it is excised by a related integrase residing on a non-integrated plasmid or virus (She et al., 2004). Integrase gene fragments on either side of a region of horizontally transferred genes may therefore suggest transfer by this mechanism. However, other archaeal integrases, including the putative integrase of the plasmid pNOB8, do not appear to disrupt the gene that encodes them, so it is unclear how widespread the SSV1 integrase mechanism may be.
Integrons.
Integrons are a unique class of mobile elements that utilize site-specific integration. Integrons were first identified in the late 1980s as a potentially mobile element found in association with a variety of antibiotic resistance genes (Stokes & Hall, 1989). The integron system consists of three essential elements: a tyrosine recombinase integrase gene (intI), a strong promoter (pC) and a recombination site (attI) (Fluit & Schmitz, 2004). Integrons do not encode the determinants for self-mobilization; rather, they facilitate the transfer of gene cassettes, which are promoterless open reading frames with attC (also called 59 bp) recombination sites. Specifically, the IntI enzyme catalyses a site-specific recombination event between the attI and attC sites, and can integrate, rearrange, or excise gene cassettes. Once incorporated into the integron, gene cassettes are expressed from the integron-based promoter pC; the number of integrated cassettes can range from 0 to well over 100 (Rowe-Magnus et al., 2001 and references therein). Because these mobile genes are integrated into a segment of DNA that is already capable of replication (bacterial chromosome or plasmid), and because they are integrated behind a host-compatible promoter, integrons may be important in overcoming limitations associated with the expression and replication of horizontally transferred DNA, and may therefore be an important mechanism for HGT between distantly related lineages. Finally, because integrons can be located on other mobile gene elements such as plasmids or transposons (e.g. Nesvera et al., 1998), they can themselves be passed between bacteria.
There are many different classes of integrons, each class possessing a unique integrase sequence. The first integrons discovered, which contain the intI-1 gene, were named ‘type 1’ integrons (Stokes & Hall, 1989). These integrons are common in pathogenic bacteria, and are one important factor contributing to the spread of antibiotic resistance (Fluit & Schmitz, 2004). Four other classes of integrons that are important in the transfer of resistance genes have been identified, and these are generally found on plasmids or transposons (Mazel, 2006). This association with mobile elements has facilitated their transfer between a diverse array of bacteria, including many species of 
-proteobacteria and Gram-positive bacteria (Fluit & Schmitz, 2004, and references therein). Integrons associated with mobile elements typically contain one to ten gene cassettes with highly variable attC sites.
Whole-genome sequencing has also revealed the presence of a wide variety of integrons on the chromosomes of a diverse suite of bacteria (Rowe-Magnus et al., 2001, 2003; Fluit & Schmitz, 2004 and references therein; Nemergut et al., 2004) (Table 3). These integrons are thought to be less likely to transfer resistance genes and to be horizontally transferred less frequently than integrons associated with other mobile elements, and they typically contain a suite of gene cassettes with homogeneous attC sites (Rowe-Magnus et al., 2001, 2003).
Notably, although integrons have been classified according to their location (e.g. chromosomal vs mobile) or the phenotype of gene cassettes that they contain (e.g. resistance), these qualities of integrons are not stable (Stokes et al., 2006), and therefore these designations should be avoided (Hall et al., 2007).
Unlike other integrases, the integron integrases contain unique additional domains (see Fig. 2), which have been implicated in recombination and DNA binding (Messier & Roy, 2001 and references therein). This unique set of features is useful for detecting integron integrases in genomic or environmental data, and in distinguishing them from other tyrosine recombinases. Following detection of an integron integrase, gene cassettes can be identified using a scheme for the automated identification of gene cassette attC sites described by the lab of Didier Mazel (Rowe-Magnus et al., 2003). Comparative analysis of the codon usages and other genomic signatures of gene cassettes, as well as sequence comparisons of gene cassette attC sites, may yield insight into diversity or homogeneity of donors for the gene cassettes in an integron.
Homologous recombination
Horizontally transferred DNA segments that do not possess any special features that might allow integration or transposition still have a chance to be incorporated into the recipient genome. Most sequences transferred between cells by transformation, GTAs, or conjugative transfer of a chromosomal region, and many sequences transferred by generalized transduction, will depend on homologous recombination to enter the genome (Thomas & Nielsen, 2005). Rather than introducing novel DNA, homologous recombination typically replaces one DNA sequence with another, related sequence. Homologous recombination functions most efficiently when sequences are related. Thus, we should expect that these processes will be most important as mechanisms of transfer between related lineages, and they may be particularly important for promoting intraspecies orthologue displacement (Lawrence & Hendrickson, 2003).
Transposition
Although transposition per se does not allow for the intercellular transport of DNA, and thus is not, in itself, a mechanism of HGT, transposable elements facilitate gene transfer in several ways. First, intracellular transposition of genes from a microbial genome to a plasmid or phage may provide otherwise sedentary regions of DNA with a vehicle to enter other cells. Second, duplicate copies of transposable elements within a genome may aid the entry of foreign DNA carrying the same transposable elements by providing regions of homology that promote homologous recombination.
Like other mobile elements, insertion sequences (ISs) and transposons exhibit extraordinary variation both in the genes that they transfer and in their mechanism of action. ISs, the simplest form of mobile gene elements that move by transposition, consist of a transposase gene (either a serine or DDE recombinase) flanked by two inverted repeats. The transposase enzyme recognizes the inverted repeats, and catalyses recombination events that transfer the IS from one site to another within the genome. Composite transposons arise when two ISs insert near one another on a genome (e.g. Galimand et al., 2005). In these cases, the outermost inverted repeats can be targets for the transposase enzyme, and the intervening sequence can be transferred along with the ISs (Fig. 2). A variety of genes have been ‘trapped’ on composite transposons in this fashion, including antibiotic resistance determinants (Galimand et al., 2005), genes involved in metal tolerance and detoxification (Liebert et al., 1999), and genes for the catabolism of xenobiotic compounds (Top & Springael, 2003). Non-composite transposons also contain a transposase and an intervening sequence, but they only contain flanking inverted repeat sequences, and not entire ISs (e.g. Schaefer & Kahn, 1998).
ISs and transposons vary in their transposition mechanism; some move via ‘replicative’ transposition, while others spread via ‘cut-and-paste’ transposition (Curcio & Derbyshire, 2003). The cut-and-paste transposons are characterized by a transposase enzyme that excises the IS or transposon from its original site and ligates it into a new location. In contrast, replicative transposons are first copied, and the new molecule is inserted into a distal site in the genome; replicative transposons generally require the activity of both a transposase and a resolvase. However, there are mechanisms to increase the number of cut-and-paste transposons within a genome as well. For example, they can be transposed immediately after the DNA replication fork passes and be transferred to a region of the genome that has not yet been replicated.
The names and DNA sequences of many known ISs are entered into the ISFinder database, which also categorizes ISs into families (Siguier et al., 2006). Further, the IslandPath program is capable of annotating transposases found near regions of putative HGT genes (‘genomic islands’), which may aid in the identification of novel transposable elements (Hsiao et al., 2003).
|
Prospects for sequence-based identification of HGT mechanisms |
|---|
|
TOPABSTRACTIntroductionInferring HGT from genomic...Detecting mechanisms of...Detecting mechanisms of...Prospects for sequence-based...ConclusionREFERENCES |
|---|
). Published programs are available that can be used to find specific mobile elements such as prophage and transposons (Hsiao et al., 2003; Bose & Barber, 2006; Fouts, 2006). An algorithm for annotating integron gene cassettes has also been developed (Rowe-Magnus et al., 2003). Although no programs designed specifically to detect origins of transfer are currently available, published alignments of the various oriT families make the automated identification of canonical origins of transfer possible. Incorporating information about these mechanisms of transfer into algorithms for HGT detection is possible now, and could provide novel insights into how horizontal transfers inferred from genomic data were produced.
Existing programs for prophage and gene cassette annotation have proven capable of identifying mechanisms in cases where independent evidence suggests horizontal transfer. In Salmonella typhimurium LT2, for example, Phage_Finder was able to successfully detect the presence of all prophage in a manually curated dataset (Casjens, 2003), including the Gifsy-1, Gifsy-2, Fels-1 and Fels-2 prophage previously implicated in horizontal transfer of pathogenicity determinants amongst strains of Salmonella enterica (Figueroa-Bossi et al., 2001). Similarly, the algorithm of Rowe-Magnus et al. (2003) for the detection of attC sites was used to compare the attC sites of gene cassettes that varied within the Vibrionaceae. These gene cassettes were found to possess a post-segregational killing (PSK) toxin–antitoxin system more commonly associated with plasmids, which was suggested to have been acquired by transfer because of its distinct GC content compared to the genome as a whole. Thus, mechanism-based approaches have been able to detect both previously described and novel cases of gene transfer.
Differentiating between some other mechanisms of transfer is likely to be more challenging. In lineages that do not require specific DUSs for natural transformation, there may not be sequence features that can be used to annotate sites of transformation and recombination. Natural transformation followed by recombination may therefore be difficult to distinguish from other HGT processes, such as introduction of DNA by a GTA or general transducing phage followed by recombination. Similarly, until more is known, it will be difficult to develop algorithms for automatic detection of conjugation sites, like the Streptomyces cis-acting locus of transfer, which promote dsDNA transfer. So far, known clt sequences do not share a high degree of sequence similarity, although this may simply be a symptom of the few examples that are known. Further study will be necessary to determine which clt characteristics are best used for identification. If features of secondary structure, but not sequence, are generally conserved, then it may be necessary to make use of stochastic context-free grammars to detect and annotate those structural features (Eddy & Durbin, 1994).
Although, as outlined above, there are several promising applications for the automated annotation of mechanisms and mobile elements that are currently identifiable, other potential applications are limited by the current inability to distinguish between or detect several known mechanisms of gene transfer from sequence features alone. It would be very interesting, for example, to know how many genes in a particular genome were transferred by each mechanism. Such an analysis is currently necessarily limited to identifying genes transferred by known and identifiable mechanisms, and placing other putatively transferred genes into an ‘other’ category. Nonetheless, even a limited analysis of this nature may explain a substantial proportion of gene transfers. For example, Hsiao et al. (2005) found that 75 % of the genomic islands analysed (from a curated set of islands previously described in the literature) contained an identifiable mobility gene such as a transposase or integrase.
It should be noted, however, that the presence or absence of mobile elements or other indications of transfer within an individual genome should not be taken as representative of the strain as a whole, because horizontal transfers, or intragenomic expansions of mobile elements, can occur at the strain (e.g. Wagner, 2006) or population level (e.g. van Essen-Zandbergen et al., 2007).
A further potential limitation of mechanism-based HGT studies is the presence of variant or entirely novel HGT mechanisms. We should be sceptical of the notion that all HGT mechanisms that occur in nature have already been described – even for mechanisms of gene transfer or mobility that have been very well studied in model organisms, many variations will undoubtedly have developed over the course of microbial evolution, and these variations may make detection of a gene transfer mechanism more difficult.
Variations could include mechanisms of gene transfer that (1) make use of divergent homologues of genes involved in well-studied mechanisms, (2) involve new mobile elements comprising novel combinations of known mobility genes, (3) are variants on existing mechanisms that use homologues of known mobility genes in combination with unknown genes, and (4) are entirely novel mechanisms. Divergent homologues of known mobility genes can be detected using standard sequence similarity comparison techniques, and new mobile elements composed of novel combinations of genes encoding known transfer mechanisms can be approached using the mechanism-centric approaches to mobile elements discussed above. The techniques outlined in this review cannot directly search for entirely novel mechanisms of gene transfer. However, attempts to identify those putative transfers that do contain sequence features consistent with transfer by a well-understood mechanism could nevertheless aid efforts to uncover wholly or partially novel mechanisms of gene transfer.
Indeed, one of the great benefits of rigorous attempts to assign plausible mechanisms to horizontal transfer events is that cases will almost certainly be found in which no known mechanism can be identified. These cases could arise because: (1) no transfer occurred, but the HGT detection method inferred one anyway (false positive), or (2) HGT occurred by a known mechanism, but that mechanism could not be identified, or (3) HGT occurred by a novel, undescribed mechanism (Table 1). Careful analysis of collections of genomic regions that are predicted to have been recently transferred, but to which no known mechanism can plausibly be ascribed, could help us to better understand situations that bamboozle existing HGT detection algorithms, and may even lead to the identification of previously unknown types of mobile elements.
Mechanism annotation and global estimates of HGT frequency
One of the major remaining challenges in the area of HGT research is to achieve accurate estimates of the overall rate of HGT, and to characterize the properties of genes that are transferred most frequently. Although there are many well-established individual cases of HGT, estimation of the overall frequency of HGT on the tree of life is a difficult problem and an active area of research. Several recent attempts to estimate the global extent of HGT produced strikingly different results, probably as a result of the different methodologies used (Nakamura et al., 2004; Ge et al., 2005; Choi & Kim, 2007).
The large differences between the values reported using different methodologies could be due to several factors. The methods may be detecting different subsets of transfers (e.g. transfers of different ages Ragan, 2001; Ragan et al., 2006), or some methods may simply be more robust to particular types of error than others. Comparison and benchmarking of the available methods against high-confidence sets of known HGT genes as well as against simulated datasets may aid understanding of the properties of these diverse methods.
Automated identification of sequence features indicative of transfer by a particular mechanism could be useful for generating the high-confidence pools of HGT genes necessary for such analyses. Such pools would also be useful because they could be used to test whether different compositional methods perform equally across identifiable transfer mechanisms. In particular, it seems possible that different compositional metrics may be particularly adept at identifying transfers facilitated by different types of mobile elements. New mechanism-specific compositional metrics for HGT detection could also be designed by training on pools of elements transferred by a similar mechanism. Finally, as discussed above, mechanism-based studies could also allow new questions to be asked about the differences, if any, between the distributions of genes transferred by various mechanisms.
|
Conclusion |
|---|
|
TOPABSTRACTIntroductionInferring HGT from genomic...Detecting mechanisms of...Detecting mechanisms of...Prospects for sequence-based...ConclusionREFERENCES |
|---|
|
ACKNOWLEDGEMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTIntroductionInferring HGT from genomic...Detecting mechanisms of...Detecting mechanisms of...Prospects for sequence-based...ConclusionREFERENCES |
|---|
Ambur, O. H., Frye, S. A. & Tonjum, T. (2007). New functional identity for the DNA uptake sequence in transformation and its presence in transcriptional terminators. J Bacteriol 189, 2077–2085.
Averhoff, B. (2004). DNA transport and natural transformation in mesophilic and thermophilic bacteria. J Bioenerg Biomembr 36, 25–33.[CrossRef][Medline]
Avery, O., MacLeod, C. & McCarty, M. (1944). Studies on the chemical nature of the substance inducing transformation of pneumococcal types. J Exp Med 79, 137–158.[Abstract]
Bacon, D. J., Alm, R. A., Burr, D. H., Hu, L., Kopecko, D. J., Ewing, C. P., Trust, T. J. & Guerry, P. (2000). Involvement of a plasmid in virulence of Campylobacter jejuni 81176. Infect Immun 68, 4384–4390.
Bertani, G. (1999). Transduction-like gene transfer in the methanogen Methanococcus voltae. J Bacteriol 181, 2992–3002.
Bertani, G. & Baresi, L. (1987). Genetic transformation in the methanogen Methanococcus voltae PS. J Bacteriol 169, 2730–2738.
Bose, M. & Barber, R. D. (2006). Prophage Finder: a prophage loci prediction tool for prokaryotic genome sequences. In Silico Biol 6, 223–227.[Medline]
Burrus, V., Pavlovic, G., Decaris, B. & Guedon, G. (2002). Conjugative transposons: the tip of the iceberg. Mol Microbiol 46, 601–610.[CrossRef][Medline]
Bushman, F. (2002). Lateral DNA Transfer: Mechanisms and Consequences. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press.
Canchaya, C., Fournous, G., Chibani-Chennoufi, S., Dillmann, M. L. & Brussow, H. (2003). Phage as agents of lateral gene transfer. Curr Opin Microbiol 6, 417–424.[CrossRef][Medline]
Casjens, S. (2003). Prophages and bacterial genomics: what have we learned so far? Mol Microbiol 49, 277–300.[CrossRef][Medline]
Chen, I., Christie, P. J. & Dubnau, D. (2005). The ins and outs of DNA transfer in bacteria. Science 310, 1456–1460.
Cho, E. H., Nam, C. E., Alcaraz, R., Jr & Gardner, J. F. (1999). Site-specific recombination of bacteriophage P22 does not require integration host factor. J Bacteriol 181, 4245–4249.
Choi, I. G. & Kim, S. H. (2007). Global extent of horizontal gene transfer. Proc Natl Acad Sci U S A 104, 4489–4494.
Court, D. L., Oppenheim, A. B. & Adhya, S. L. (2007). A new look at bacteriophage lambda genetic networks. J Bacteriol 189, 298–304.
Curcio, M. J. & Derbyshire, K. M. (2003). The outs and ins of transposition: from mu to kangaroo. Nat Rev Mol Cell Biol 4, 865–877.[CrossRef][Medline]
Davidsen, T., Rodland, E. A., Lagesen, K., Seeberg, E., Rognes, T. & Tonjum, T. (2004). Biased distribution of DNA uptake sequences towards genome maintenance genes. Nucleic Acids Res 32, 1050–1058.
Dubnau, D. (1999). DNA uptake in bacteria. Annu Rev Microbiol 53, 217–244.[CrossRef][Medline]
Eddy, S. R. & Durbin, R. (1994). RNA sequence analysis using covariance models. Nucleic Acids Res 22, 2079–2088.
Errington, J. (2001). Septation and chromosome segregation during sporulation in Bacillus subtilis. Curr Opin Microbiol 4, 660–666.[CrossRef][Medline]
Farahi, K., Whitman, W. B. & Kraemer, E. T. (2003). RED-T: utilizing the Ratios of Evolutionary Distances for determination of alternative phylogenetic events. Bioinformatics 19, 2152–2154.
Figueroa-Bossi, N., Uzzau, S., Maloriol, D. & Bossi, L. (2001). Variable assortment of prophages provides a transferable repertoire of pathogenic determinants in Salmonella. Mol Microbiol 39, 260–271.[CrossRef][Medline]
Fluit, A. C. & Schmitz, F. J. (2004). Resistance integrons and superintegrons. Clin Microbiol Infect 10, 272–288.
layout adapted from Court et al. (2007)