OUP user menu

The distributed genome hypothesis as a rubric for understanding evolution in situ during chronic bacterial biofilm infectious processes

Garth D. Ehrlich, Azad Ahmed, Josh Earl, N. Luisa Hiller, J. William Costerton, Paul Stoodley, J. Christopher Post, Patrick DeMeo, Fen Ze Hu
DOI: http://dx.doi.org/10.1111/j.1574-695X.2010.00704.x 269-279 First published online: 1 August 2010


Most chronic infectious disease processes associated with bacteria are characterized by the formation of a biofilm that provides for bacterial attachment to the host tissue or the implanted medical device. The biofilm protects the bacteria from the host's adaptive immune response as well as predation by phagocytic cells. However, the most insidious aspect of biofilm biology from the host's point of view is that the biofilm provides an ideal setting for bacterial horizontal gene transfer (HGT). HGT provides for large-scale genome content changes in situ during the chronic infectious process. Obviously, for HGT processes to result in the reassortment of alleles and genes among bacterial strains, the infection must be polyclonal (polymicrobial) in nature. In this review, we marshal the evidence that all of the factors are present in biofilm infections to support HGT that results in the ongoing production of novel strains with unique combinations of genic characteristics and that the continual production of large numbers of novel, but related bacterial strains leads to persistence. This concept of an infecting population of bacteria undergoing mutagenesis to produce a ‘cloud’ of similar strains to confuse and overwhelm the host's immune system parallels genetic diversity strategies used by viral and parasitic pathogens.

  • chronic bacterial pathogenesis
  • bacteria
  • horizontal gene transfer
  • distributed genome hypothesis
  • supragenome
  • pangenome


Biofilms serve as population-level virulence factors as they confer the resident bacteria with virulence attributes that a single bacterium does not possess. Most of these biofilm-related population-level virulence traits are protective for the bacteria, allowing them to persist in the host in the face of both the innate and the adaptive immune systems. Thus, they are chiefly of a chronic nature as opposed to planktonic virulence factors, such as toxins, which make the host acutely ill. In addition to providing protection and enabling persistence, biofilms associated with the middle-ear mucosa also often induce the host to produce effusions and/or to promote hyperplastic growth of the surrounding host tissue by downregulating apoptosis (Post & Ehrlich, 2007, 2009). Thus, there is interkingdom signaling that serves to provide a constant nutrient source for the biofilm bacteria that helps to maintain the infectious process.

Biofilms also provide an ideal setting for elevated levels of gene transfer among the resident bacteria, both among strains of a species and among related species (Wang et al., 2002; Molin & Tolker-Nielsen, 2003; Sørensen et al., 2005). These gene transfers occur because nearly all of the chronic bacterial pathogens that form biofilms also contain inducible energy-requiring horizontal gene transfer (HGT) mechanisms that serve a non-nutritive purpose (as opposed to using the DNA simply as a food source). These microbial gene transfer capabilities have long been recognized by the infectious disease and clinical microbiological communities, but only in a very narrow sense. Capsular switching among the pneumococci was recognized as resulting from HGT among strains, as were differences in virulence associated with the capsule type during epidemics (Ramirez & Tomasz, 1999), but these observations were not generalized until 2001 when we advanced the distributed genome hypothesis (DGH) (Ehrlich, 2001; Ehrlich et al., 2005), which posits that bacterial biofilms associated with chronic infections are composed of multiple strains of a single species (as well as often being polymicrobial or polykingdom communities) and that real-time HGT among the component strains (and species) leads to the continuous generation of a cloud of new strains with a novel combinations of genes, thereby providing the bacterial community with a means to thwart the adaptive immune response of the host.

Evolutionary consequences of HGT

Bacterial HGT is defined as the movement of genes (almost always in a unidirectional manner) between two, often unrelated, bacterial cells. It is important to understand that the donor cell from which the horizontally transferred DNA arose does not have to be viable at the time of HGT, and in fact, is definitely not the case in two of the three major HGT mechanisms used by bacterial species. HGT mechanisms usually result in the transfer of one or more relatively small blocks of donor DNA into the recipient cell and thus provide for only the partial replacement of the receiving bacterium's chromosome. The mean sizes of horizontally acquired gene blocks for those species such as Haemophilus influenzae, Streptococcus pneumoniae, and Staphylococcus aureus that have been studied extensively are usually only between 1 and 2 kb (Hiller et al., 2007; Hogg et al., 2007; Hall et al., 2010), but larger horizontally acquired regions of 50–100 kb in size are not uncommon.

Detailed comparative whole chromosomal analyses among large numbers of strains of H. influenzae (Hogg et al., 2007) and S. pneumoniae (Table 1) have revealed that, on average, each strain contains between 200 and 400 insertions/deletions (indels) throughout their chromosome relative to other strains of the species. Thus, each chromosome is highly mosaic with respect to the origin of its own component genes, and further, each strain's chromosome is highly unique with respect to its gene possession complement. In fact, gene possession differences among the strains of a species account for the vast majority of the genetic heterogeneity within a species and dwarf the number of allelic differences observed within genes (Hall et al., 2009). Exhaustive pair-wise comparisons among all of the genomically sequenced strains for each of the species H. influenzae, S. pneumoniae, S. aureus, and Gardnerella vaginalis reveal that there are 385, 407, 246, and 608 gene possession differences, respectively, on average between every pair of strains that has been sequenced within these species (Hiller et al., 2007; Hogg et al., 2007). The 12-strain G. vaginalis supragenome (pangenome) contains 2248 genes, of which only 719 are core, with the remaining 1529 genes being distributed (noncore) among the 12 strains. Thus, more than two-thirds of the species' genes are found in only a subset of strains. Because the average individual strain contains only 1257 genes, on average, each genome contains 538 (42%) distributed genes. These extraordinary gene possession differences can only arise via HGT mechanisms.

View this table:
Table 1

Summary of indels between the reference strain R6 and 21 clinical pneumococcal isolates

StrainNo. of insertionsMedian insert length (bp)Mean insert length (bp)Total insert length (kb)Maximum insert length (kb)No. of deletionsMedian deletion length (bp)Mean deletion length (bp)Total deletion length (kb)Maximum deletion length (kb)Total no. of indels
BS 397106277165817616111499204022733217
PAT 6428720770320217131344120415828418
TIGR 4632431388871779789236318737142
TIGR 670673247160511717101690306030948174
  • Bold values show the number of genomic insertions, deletions, and total indels per strain compared with the Streptococcus pneumoniae reference strain, R6. The bottom row shows the 21-strain means for each measure.

HGT is defined in contrast to vertical gene transfer, which is the standard mechanism by which a mother cell replicates her entire complement of DNA and then passes along identical (or nearly so) copies of each chromosome and plasmid to each of her daughter cells during cell division. Genes and chromosomes that are acquired solely though vertical transmission can be used to construct phylogenetic relationships among bacterial strains, species, and higher taxa; however, genes that are acquired through HGT mechanisms produce mosaic chromosomes in which each part of the chromosome that was acquired horizontally has a different ancestry from every other part of the chromosome (unless there are two or more simultaneous transformative events arising from the uptake of DNA from a single donor/competence event), which therefore makes phylogenetics at the whole chromosome level very difficult. In other words, for any set of strains containing mosaic chromosomes, each individual gene that has been horizontally transferred and then used to build a phylogenetic tree will produce a different tree structure from the same set of strains (Fig. 1) (Shen 2005; Hall et al., 2010). Extensive HGT does not always completely obliterate the average chromosomal phylogenetic signal as has been demonstrated recently for S. pneumoniae (Donati et al., submitted); however, because of extensive HGT, strains that are phylogenetically related may have profoundly different genic compositions and thus produce very different disease phenotypes (Buchinsky et al., 2007).

Figure 1

Phylogenetic trees were built for each of seven Streptococcus pneumoniae core genes using the same set of eight clinical strains. Each gene produces a different tree structure (from the same set of strains), demonstrating the mosaic nature of the S. pneumoniae genome in general that results from extensive and continuous HGT. (a)–(g) Phylogenetic trees showing the relative distances of seven housekeeping loci among the eight clinical strains of S. pneumoniae including: (a) the aroE gene encoding shikimate dehydrogenase; (b) the gdh gene encoding glutamate dehydrogenase; (c) the gki gene encoding glucose kinase; (d) the recP gene encoding the RecP recombination protein; (e) the spi gene encoding a signal peptidase; (f) the xpt gene encoding xanthine phosphoribosyltransferase; and (g) the ddl gene encoding d-alanyl-d-alanine ligase. Reference sequences were downloaded from http://www.mlst.net. mlst-aroE, mlst-gdh, mlst-recP, mlst-spi, mlst-xpt, and mlst-ddl; (h) Phylogenic tree reconstructed using concatenated sequences from all seven housekeeping genes (3199 bp) of S. pneumoniae. Concatenated sequence mlst-7 gene was created using seven reference gene sequences from http://www.mlst.net.

Bacterial HGT mechanisms

HGT is accomplished largely through three fundamentally different mechanisms: competence and transformation, mating or conjugation, and viral transduction. Some species of bacteria use only one of these mechanisms, whereas others utilize two or even all three. Transformation and mating are active processes and require significant energetic expenditures by the recipient and the donor bacteria, respectively, as well as the maintenance of entire genetic regulons that encode the necessary machinery for the uptake and transfer of DNA, respectively (Mann et al., 2009). Thus, the bacteria that possess and maintain these systems must receive an evolutionary advantage in order for them to persist, particularly in the face of strong genomic deletatory mechanisms present in bacteria that are designed to minimize the genomic burden and eliminate unwanted foreign DNA — particularly that of bacteriophages (Brussow et al., 2004). Viral transduction, on the other hand, is a passive process engendered by temperate phage. The widespread possession of HGT mechanisms among pathogenic bacterial species, regardless of phylogeny and gram status, was one of the chief observational points on which the DGH was built (Ehrlich, 2001; Shen et al., 2003, 2005, 2006; Ehrlich et al., 2005, 2008; Hu & Ehrlich, 2008).

Historically, transformation was the first HGT mechanism identified. In 1928, Griffith reported the ‘transformation’ of rough, avirulent live pneumococci into smooth, virulent pneumococci by the addition of factors from dead, smooth, virulent pneumococci (Griffith, 1928). Thus, from its first recognition, transformation was demonstrated to be a population-level virulence factor (Hu & Ehrlich, 2008); however, this very important clinical aspect of Griffith's seminal work was overshadowed for generations by the even larger basic science implications that derived from this same work. Griffith's work also suggested the chemical nature of the gene and demonstrated conclusively that individual genes were not living entities in and of themselves. His observations also supported Mendel's concept of there being discrete genes associated with specific phenotypes (Mendel, 1866), but from a practical basis, this work provided the means, through purification, to identify the hereditary molecule. In 1944, Avery, McLeod, and McCarty, in a series of follow-up experiments to Griffith's work demonstrated, to the surprise of the world at that time, that DNA, not protein, was the pneumococcal transforming substance (Avery et al., 1944), and in so doing, ushered in the era of mechanistic molecular biology.

Competence and transformation are actually two separate molecular processes. Competence is the metabolic state of being able to take up foreign DNA into the cell, and transformation results if and when foreign DNA is integrated into the host chromosome, changing the genotype and ultimately the phenotype of the cell. In most bacterial species in which competence has been studied, it has been determined to be an inducible phenomenon associated with nutrient limitation or part of an SOS response (Herriott et al., 1970; Håvarstein et al., 2006; Kreth et al., 2006; Prudhomme et al., 2006; Claverys & Håvarstein, 2007; Claverys et al., 2007; Thomas et al., 2009). Therefore, these processes, which increase the probability of mutation considerably, are triggered when the bacteria are under stress and indicate that bacteria can control their mutational rate based on environmental conditions. This is in stark contrast to the widely held view of evolution that mutational rates are invariant and are not able to be controlled by the organism. Viewed teleologically, the bacteria ‘realize’ that they must ‘change their spots’ to survive and thus activate an energetic system to increase the likelihood of genetic recombination and genic reassortment. In so doing they are utilizing a strategy very similar to that of sexual species in which they are recombining genes (all of which have proven their fitness under various conditions) in different combinations, to produce new organisms with the expectation that some of the recombinants will have a selective advantage under the prevailing conditions in the host, as well as ensuring the survival of the bacterial population and the bacterial gene pool as a whole.

Bacterial mating or ‘conjugation’ as it was dubbed by its discoverer, Joshua Lederberg, who was looking for a sexual phase in the life cycle of bacteria, can result in the transfer of either episomal (plasmid) elements and/or parts of the bacterial chromosome from a donor cell to a recipient cell (Lederberg & Tatum, 1946) and unlike transformation requires cell : cell contact for transfer of the donated DNA (Davis, 1950). Bacterial conjugation, like transformation, is a bacterial equivalent of sex as both of these prokaryotic HGT mechanisms involve genetic exchange. However, neither of these processes includes the entire genomes of the parental pair, but rather in both cases, one bacterium serves as a donor that provides a section of DNA that, if chromosomal, replaces a section of the chromosomal DNA in the recipient strain, usually through homologous recombination. In the case of conjugation, as opposed to transformation where the donor cell must be dead, the conjugative donor must be viable as it contains either a conjugative plasmid, or mobilizable genetic element integrated into the chromosome, that encodes the molecular machinery to support the creation of a proteinaceous bridge, a pilus, through which the DNA is mobilized, as well as the enzymatic machinery to make a copy of the donor's DNA for transport through the pilus into the recipient. For these reasons, the bacteria initiating conjugation are referred to as male. This brings up a fundamental mechanistic dichotomy between these two energy-requiring bacterial HGT processes. In the case of transformation, the recipient cell is the one expending energy and has evolved to either scavenge extracellular DNA (eDNA) or kill its neighbors to ensure an eDNA supply (vide infra), whereas with conjugation, it is the donor cell that is expending most of the energy and thus its conjugative elements can be viewed as genetic parasites that evolved to spread themselves into new hosts. However, the conjugative elements often bring beneficial genes with them as well, including those encoding antibiotic and heavy metal resistances, the ability to utilize novel metabolites, or virulence determinants such as adhesins, iron acquisition systems, and serum tolerance.

Transduction, also first discovered in Lederberg's lab (Zinder & Lederberg, 1952), results when a temperate or a lysogenic bacteriophage that has been integrated into the host chromosome excises itself and an adjacent section of the host chromosome as part of the lytic phase and then transfers the previous host's chromosomal region to its next host upon chromosomal integration. Transduction, unlike competence/transformation and mating, is a passive process on the part of both the donor and the recipient bacteria as it does not require any energy expenditure or host mechanistic genes to accomplish.

HGT in biofilms

Four elements are necessary for HGT to occur in situ within bacterial biofilms and result in the generation of diversity within the infecting bacterial population. First, it must be demonstrated that chronic infections, in general, are indeed associated with bacteria adopting a biofilm mode of growth. Second, it must be demonstrated that there is a supply or a means to generate a supply of DNA for HGT within the biofilm community. Third, there need to be mechanisms (vide supra) for the transfer of DNA into live organisms. Fourth, and perhaps most importantly, the infecting bacterial population must be polyclonal in nature, i.e. be made up of multiple independent strains of the same bacterial species that are present simultaneously. The necessity for polyclonality derives from the need to generate diversity. If the infection–colonization is monoclonal, it means that each bacterium in the biofilm contains the same set of genes and the same set of allele forms of each gene; thus, exchanging DNA between any two cells in such an environment would not produce a new strain with new combinations of genes and alleles. In such a case, an extensive energy output would be rewarded with no possible gain in terms of creating a more competitive organism. Finally, it must be demonstrated that gene exchange indeed does occur, in real time, among strains within a polyclonal biofilm population and that some of the recombinant strains persist and expand their presence over time (i.e. prove to have a reproductive advantage under the prevailing conditions in the host) and in turn serve as recipients or donors of DNA in further HGT processes.

Chronic infectious conditions possess all the elements necessary for HGT

An examination of the conditions present during the bacterial colonization of eukaryotic hosts, and during the subsequent chronic infectious disease processes, demonstrates that all of the criteria exist for fruitful genic reassortments (Hu & Ehrlich, 2008).

Chronic bacterial infections are characterized by biofilms

Bacterial infections associated with chronic disease states are nearly universally found to have adopted a biofilm phenotype (Hu & Ehrlich, 2008). The bacterially elaborated extracellular matrix of the biofilm, associated with the final irreversible attachment of bacterial cells to a surface, is composed of multiple extracellular polymeric substances (EPS) including exopolysaccharides, eDNA, proteins, and lipids, and provides a protective physical barrier for the bacteria within. The cooperative creation of the matrix on host tissues or implantable devices by a community of bacteria is a population-level virulence trait as it provides for a community of bacteria that are collectively more difficult for the host to eradicate than individual free-swimming or individual attached bacteria would be. Once initiated, a biofilm acts like a single dynamic living organism that can grow, change its physical properties in response to its environment, evolve through mutation to be better adapted to its environment (Boles et al., 2004; Kraigsley & Finkel, 2009), and incorporate other pathogenic species into an integrated polymicrobial community. Importantly, not only can multiple species coexist with a biofilm, but within spatially structured environments, they will mutate and evolve in such a way as to improve the interaction with other resident species, producing a more stable and productive community than their ancestral counterparts (Hansen et al., 2007). The biofilm serves as a skeleton for large numbers of bacteria within a single structure and confers the population of interacting organisms with protection, one of the hallmarks of multicellular organisms. Extending the skeletal metaphor, the biofilm matrix also plays important roles in signaling control and nutrient availability. Rheological studies by Stoodley and colleagues have demonstrated that the hydrated EPS matrix is highly viscoelastic and can be rapidly remodeled in response to changes in shear and other environmental stressors (Dunsmore et al., 2002; Klapper et al., 2002; Stoodley et al., 2002; Towler et al., 2003; Shaw et al., 2004). Thus, in this regard, it displays qualities similar to endochondral bone in that the strength of the extracellular matrix is modifiable by the cellular component in accordance with the external load.

Detailed imaging and metabolic studies spurred by the development of the confocal microscope and PCR-based diagnostics have revealed that many disease conditions that were previously thought of as being chronic inflammatory in nature are actually indolent bacterial biofilm infections. These include osteomyelitis associated with S. aureus and Staphylococcus epidermidis (Marrie & Costerton 1985; Costerton, 2005); gall bladder disease (Sung et al., 1991; Stewart et al., 2007); various chronic middle-ear disease processes associated with H. influenzae, S. pneumoniae, Moraxella catarrhalis, and Pseudomonas aeruginosa including otitis media with effusion, recurrent otitis media, and otorrhea (Rayner et al., 1998; Ehrlich et al., 2002; Post et al., 2004; Dohar et al., 2005; Hall-Stoodley et al., 2006; Bakaletz, 2007; Kerschner et al., 2007; Post & Ehrlich, 2007, 2009; Apicella, 2009); chronic rhinosinusitis associated with H. influenzae, S. aureus, and other bacteria (Palmer, 2006; Sanderson et al., 2006; Psaltis et al., 2007; Prince et al., 2008); cholesteatoma associated with P. aeruginosa (Chole & Faddis, 2002); tonsillitis (Chole & Faddis, 2003); and adenoiditis associated with H. influenzae, S. pneumoniae, and M. catarrhalis (Zuliani et al., 2006; Nistico et al., 2009). In addition, there is substantial evidence to support a bacterial biofilm etiology for many chronic infections of the urogenital systems of both men and women including cystitis, prostatitis, vaginitis, and endometritis (Nickel et al., 1994; Hua et al., 2005; Swidsinski et al., 2008), and recently, both S. aureus and S. epidermidis have been demonstrated to form biofilms at surgical site infections (Kathju et al., 2009). Biofilms are also associated with dental infections including plaque, endodontitis (Carr et al., 2009), and periodontitis (Marquis, 1995; Paster et al., 2006).

Moreover, biofilms represent the overwhelming bacterial phenotype associated with chronic nonhealing wounds such as venous and diabetic ulcers, pressure sores, and burn wounds. These infections are often complex polymicrobial and polykingdom communities (Davis et al., 2006; Wolcott & Ehrlich, 2008). These chronic wound infections and foreign body infections associated with implantable medical devices and indwelling catheters (Ehrlich et al., 2004, 2005; Stoodley et al., 2005, 2008) are nearly impossible to eradicate without aggressive debridement and removal of the device, and have become the bane of many permanent and long-term interventional strategies, including artificial joints, central vascular lines, urinary catheterizations, cardiac pace makers and defibrillators, ventricular-peritoneal shunts, and dialysis ports (reviewed in Ehrlich et al., 2004).

Availability of bacterial DNA within biofilms

These observations of bacterial phenotype are important because both transformation and mating have been demonstrated to be up to 104-fold higher in biofilms than in planktonic forms (Molin & Tolker-Nielsen, 2003; Sorenson et al., 2005). High transformation rates in biofilms likely result from the fact that one of the major constituents of the biofilm matrix is eDNA (Fig. 2), thus providing a ready source of genetic raw material. In the case of mating, the close spatial juxtaposition of bacterial cells in the biofilm and the physical stability conferred by the biofilm matrix likely support pilus attachment and reduce the likelihood that the conjugal bridges through which the donor DNA is exported will be broken due to hydrodynamic shear stresses. The Bakaletz lab has further demonstrated that the biofilm matrix of H. influenzae, in addition to containing DNA, also contains very high concentrations of type IV pili (Jurcisek & Bakaletz, 2007). Subsequently, Juhas (2007a, b) demonstrated that some H. influenzae strains encode pilus genes that have been shown to support conjugal DNA transfer.

Figure 2

Confocal micrograph of a Pseudomonas aeruginosa PAO1 biofilm grown on a glass substratum demonstrating eDNA within the biofilm matrix surrounding the bacteria. (a) Plan view and (b) sagittal section with the glass substratum is shown in blue by reflected light. The biofilm was stained with the DNA stain Syto9. The cells (an example of a nonmatrix enclosed cell is shown with a white arrow) were slightly overexposed to show the more diffuse signal from the eDNA between the individual bacteria. This image demonstrates: (1) the close proximity of the cells within the biofilm; (2) that within biofilms, which are relatively quiescent (unlike laboratory planktonic cultures, which are continually mixed), the cells have more time to interact (i.e. for conjugation), and (3) there is a pool of eDNA surrounding the cells that confers structural stability and also serves as a source for transformation. Scale bars=25 µm.

The biofilm matrices of all bacterial species that have been characterized for molecular composition including P. aeruginosa, H. influenzae, S. pneumoniae, Streptococcus mutans, S. aureus, and Enterococcus faecalis contain large amounts of eDNA (Whitchurch et al., 2002; Jurcisek & Bakaletz, 2007; Hall-Stoodley et al., 2008; Mann et al., 2009; Perry et al., 2009; Thomas et al., 2009). Even more interestingly, the laboratories of Shi, Clavery, Havarstein, Cvitkovitch, and Hancock have convincingly demonstrated a temporal link between conspecific fratricide and the development of competence among the streptococci and the enterococci as a means to ensure a source of species-specific eDNA for those cells first becoming competent (able to take up foreign DNA). The streptococci, just before they become competent, produce and release bacteriocins that will kill their neighbors, thus ensuring a ready supply of DNA for transformation (Kreth et al., 2006; Prudhomme et al., 2006; Claverys & Håvarstein, 2007; Perry et al., 2009), whereas the enterococci utilize a toxin–antitoxin system that kills quorum nonresponders of their own species (Thomas et al., 2009). Haemophilus influenzae and the other naturally competent Pasteurellaceae utilize a different mechanism to ensure that they primarily take up DNA from their own and highly related species. Within their genomes, they have a highly repeated uptake signal sequence (USS), which is present at approximately one copy per gene and their competence apparatus has evolved to selectively take up only DNAs that contain their species-specific USS (Redfield et al., 2006; Maughan & Redfield, 2009).

Third, and most importantly, for HGT mechanisms, colonization is nearly always polyclonal, an observation that had long been missed due to the medical microbiology community's adherence to Koch's postulates, which teach that a single clonal isolate must be obtained from an infected individual and subsequently demonstrated to cause the same disease in a second host to establish etiology. The mantra of always purifying a single clone put blinders on the medical microbiology community because any diversity that was present was never observed. Over the past decade and a half, the laboratories of Smith-Vaughan, Murphy, and Gilsdorf have repeatedly demonstrated, by examining OM patients, COPD patients, and the normal nasopharynx, respectively, that nearly all persons who are infected or colonized with H. influenzae are polyclonally infected — sometimes with >20 strains simultaneously (Smith-Vaughan et al., 1995, 1996, 1997; Murphy et al., 1999; Ecevit, 2004, 2005; Farjo et al., 2004; Mukundan et al., 2007; Lacross et al., 2008). Similarly, the de Lencastre laboratory and independently Dowson's group have observed polyclonal infection with pneumococcus (Muller-Graf et al., 1999; Sá-Leão et al., 2002, 2006, 2008; Jefferies et al., 2004), and Hoiby's and Molin's groups in Denmark have seen polyclonal P. aeruginosa infections in the CF lung (Jelsbak et al., 2007).

Polyclonality is critical to the DGH as it posits that at the species and local population levels, there exists a supragenome (pangenome) that is much larger in terms of the total number of genes (not just alleles) than the genome of any single strain within that species or population. Thus, under this rubric, the majority of genes within a species are not possessed by all strains of that species, but rather each strain contains a unique distribution of noncore genes from the species-level supragenome, as well as the species core genome (those genes that are carried by all strains of a species). Thus, we predicted that the bacteria's possession of HGT mechanisms and the polyclonality of chronic infections would provide a setting in which new strains with unique combinations of distributed genes would be continually generated. Furthermore, some of these novel strains will have improved survival characteristics under the current prevailing conditions in the host.

Recently, we have obtained direct evidence of massive and repeated HGT among pneumococcal strains during a polyclonal pediatric chronic infection that supports the above hypotheses. In this study, we identified a dominant strain that, over a period of 7 months, underwent more than a dozen transformation events, leading to the replacement of approximately 7% of its genome. The fact that we were able to recover multiple recombinant strains when isolating only one strain per time point suggests that these recombinant strains did indeed have a selective advantage in the host environment.


Our laboratory, as well as those of our colleagues (Tettelin et al., 2005; Hall et al., 2010; Harris et al., 2010) have used whole-genome sequencing to characterize the sizes of the supragenomes and determine the average number of gene possession differences of multiple independent clinical or environmental strains for over two dozen bacterial species including Escherichia coli, H. influenzae, Pseudomonas fluorescens, S. pneumoniae, Streptococcus agalactiae, S. aureus, and G. vaginalis. These studies have validated the DGH for all species examined and demonstrated that the noncore genes in each strain comprise on average one-fifth to one-third of each strain's genome and that the species-level supragenomes are often three to four times the size of the core genomes (Tettelin et al., 2005; Hiller et al., 2007; Hogg et al., 2007; Hall et al., 2009; Ahmed et al., submitted; Donati et al., submitted). The predictions of the DGH and the observation that there are enormous gene possession differences among the strains of nearly all bacterial species combine to suggest that during chronic infections, the bacteria, through HGT mechanisms, create a ‘cloud’ of related strains, each with distinct antigenic and virulence profiles that serve to keep the bacterial population ‘one step ahead of the host's immune system’. Such a strategy would be analogous to what has been demonstrated for other classes of chronic pathogens such as HIV (Lee et al., 2009) and the trypanosomes that use error-prone nucleic acid polymerases and programmed gene cassette swapping to generate a cloud of diverse strains to avoid immune clearance. Thus, it would appear that diversity generation, regardless of its precise mechanism, is key to the maintenance of a chronic infectious disease state.

These observations on diversity generation by bacteria during chronic infectious processes suggest that interventional therapeutic strategies could be developed to target this aspect of microbial pathogenesis. One such strategy would be STAMP (specific targeted antimicrobial peptides) technology, wherein a bifunctional peptide is constructed that contains a generic bacteriolytic segment and a species-specific ligand for targeting. By targeting the DNA uptake system of S. mutans, the Shi laboratory has demonstrated a multilog kill of S. mutans in the presence of other streptococcal species that were relatively unharmed (Eckert et al., 2006), providing a ‘magic bullet’ for the pathogen of interest. This strategy is currently being adopted within our laboratory for S. pneumoniae and should be generally applicable to a broad array of pathogenic bacteria.


The authors thank Ms Mary O'Toole for help in the preparation of this manuscript. This work was supported by Allegheny General Hospital, Allegheny Singer Research Institute, Grants from the Health Resources and Services Administration (HRSA); a system usage grant from the Pittsburgh Supercomputing Center (G.D.E.); and NIH grants DC04173 (G.D.E.), DC02148 (G.D.E.), DC02148-16S1 (G.D.E), and AI080935 (G.D.E.).


  • Editor: Roger Bayston


View Abstract