Horizontal gene transfer in the phytosphere


Author for correspondence: Jan Dirk van Elsas Tel: +31 317 476210 Fax: +31 317 423110 Email: j.d.vanelsas@plant.wag-ur.nl


Here, the ecological aspects of gene transfer processes between bacteria in the phytosphere are examined in the context of emerging evidence for the dominant role that horizontal gene transfer (HGT) has played in the evolutionary shaping of bacterial communities. Moreover, the impact of the putative capture of genetic material by bacteria from plants is discussed. Examples are provided that illustrate how mobile genetic elements (MGEs) influence the behaviour of bacteria in their natural habitat, especially in structured communities such as biofilms on plant surfaces. This community behaviour is used as a framework to pose questions on the evolutionary role and significance of gene transfer processes in plant-associated habitats. Selection within the highly structured phytosphere is likely to represent a dominant force shaping the genetic make-up of plant-associated bacterial communities. Current understanding of the triggering and impact of horizontal gene transfer, however, remains limited by our lack of understanding of the nature of the selective forces that act on bacteria in situ. The individual, colony, population and community level selection benefits imposed by the ability to use specific carbon sources or survive selective compounds are clear, but it is not always possible to assess what drives gene transfer and persistence. The role of HGT in the adaptation of host bacteria to their environmental niche is still not fully understood.


Horizontal gene transfer (HGT) between bacteria is the exchange of genetic material from donor to recipient cells, and stable persistence (either by integration or autonomous replication) in the latter. Whole genome sequence comparisons have revealed that detectable HGTs among bacteria are more numerous than previously thought and have greatly influenced bacterial evolution (for review see Koski et al., 2000; Ochman et al., 2000; Koonin et al., 2001). More than 50 yr ago, the three essential mechanisms for horizontal gene transfer – transformation, transduction and conjugation – were described for the first time. The first mechanism, transformation, allows the uptake and stable integration of naked DNA by bacterial cells. The second derives from errors in lysogenic phage integration into and/or excision from the chromosome of their host, or incorrect packaging of nonphage DNA into phage particles in the lytic cycle. The third involves an active process of bacterial mating pair formation through specified structures to enable DNA transfer from the donor to the recipient cell. The involvement of mobile genetic elements (MGEs) – bacteriophages, plasmids, and transposons – in the latter two processes has since gained wide acceptance (Thomas, 2000) to the extent that it is now apparent that MGEs represent key vehicles for gene transfer between bacteria (Wilkins, 1995). As such, MGEs contribute to bacterial evolution, and there is evidence suggesting that they have affected speciation processes (Levin & Bergström, 2000; Ochman et al., 2000; Bailey et al., 2001; Majewski, 2001). Mobile genetic elements have also been linked to the spread of intermittently adaptive and symbiotic traits involved in host survival (Lilley et al., 2000; Sullivan & Ronson, 1998). While there is a wealth of information, collected in vitro, on the microbiological and molecular mechanisms of gene transfer, there are only limited data on gene transfer processes in natural habitats. Furthermore, the ecological and evolutionary impact of these in situ processes remains largely enigmatic (van Elsas et al., 2000; van Elsas & Bailey, 2002).

It should also be remembered that evolution is inherently a stochastic process. Random mutational drift may alter protein function or regulation, but whether the mutation is successful (i.e. persists into subsequent generations) is dependent upon the prevailing conditions. Different conditions impose different selection pressures and determine whether or not the mutation is beneficial or detrimental in a given niche at a given time. The same is true for the establishment of horizontally transferred DNA into a new host; beneficial or adaptive combinations are likely to compete better than their neighbours and so proliferate. Real gene transfer processes in the environment are difficult to measure, not only because of the problem of differentiating transfer and proliferation, but also because, to date, only a minority of the total bacterial diversity in the environment can be cultivated on artificial growth media in the laboratory. Thus, it is likely that measures of HGT are underestimates, since transfer of DNA into nonculturable bacteria is unlikely to be detected or confirmed using standard methods. Possibly, one of the more important stochastic factors that influences HGT in the environment will be the proximity of donor and recipient cells (i.e. population structures at the microscale); moreover, it is important which genotypes are present, whether nutrients are available and whether selection pressures occur. The following discussion accepts implicitly all of these chance events and the impact that they have on our perception of HGT in the phytosphere.

The first exploratory studies on HGT between bacteria in (semi)natural soil habitats were performed in the 1970s (Weinberg & Stotzky, 1972; Graham & Istock, 1978). Since that time, there has been an increasing number of studies addressing environmental HGT, using microcosm and field studies (van Elsas et al., 2000). Until recently, most of these studies focused on the effects of key environmental factors such as temperature, moisture, nutrient availability, and the presence of grazing, competing or antagonistic organisms on the rates at which gene transfer processes take place (van Elsas et al., 2000; Timms-Wilson et al., 2001). The basic conjecture was that these factors are determinants of natural gene transfer rates and, if understood, could provide tools with which the frequencies of HGT in nature could be predicted (Timms-Wilson et al., 2001). A major impetus to the field was provided by societal questions about the impact of genetically modified organisms on nature and the spread of antibiotic resistance genes into natural habitats. However, the intricacies of HGT processes in natural settings, involving highly mixed communities, rather than simple mixtures studied under laboratory conditions, are far from being fully understood. Complications include in situ phenomena of intra/intercellular signalling and other molecular triggers of DNA transfer or uptake, as well as physical phenomena such as cell-to-cell and cell–MGE contacts. Furthermore, the ecological consequences, the ways in which these processes impact the microbial communities in their adaptation to natural habitats and the dependence of all of these processes on the environment, have received much less attention.

Advanced molecular techniques have been developed over the last decade that are applicable to studies of microorganisms in natural settings (Akkermans et al., 1995). These have enabled researchers to investigate environmental gene transfer processes at much more refined levels of resolution than before. For example, the information contained in full sequences of plasmids (Tauch et al., 2002) or bacterial genomes (Salanoubat et al., 2002), when employed in DNA microarrays, allows the study of specific plasmid and/or host gene expression in relation to HGT in the natural habitat. In addition, the resolving power offered by reporter gene technology, especially that using green fluorescent protein (GFP), has allowed detailed analyses of the patterns of gene transfer in microcolonies or biofilms (Christensen et al., 1998). The impetus of these developments in fostering our understanding of environmental HGT and its impact on natural microbial communities is only beginning to emerge.

This review examines the current understanding of bacterial HGT processes in the phytosphere and soil. We briefly review the impact of key environmental factors on HGT in these settings. We then focus on the potential of plant-associated bacteria to act as samplers of genetic material, the role of MGEs, the importance of structured microbial consortia, selective forces and how these affect HGT processes, and the adaptive responses of bacterial communities.

Factors affecting HGT in soil and the phytosphere

Most soils are restricted in resources for microbial growth, thus severely limiting population densities, growth and activity. Consequently, this restricts microbial processes that are dependent on density and metabolic activity, such as all HGT mechanisms (van Elsas et al., 2000; Timms-Wilson et al., 2001). However, some sites in soil, mostly related to soil or plant surfaces, have been shown to provide conditions for bacterial growth, colonization and mixing, resulting in the occurrence of locally enhanced densities of active cells. These sites are often conducive to HGT processes (van Elsas et al., 2000), and have been termed ‘hot’ spots for bacterial gene transfer activity. Figure 1 depicts the key hot spots for HGT processes in soil- and plant-related habitats. The conditions in these hot spots that influence HGT can be divided into abiotic (e.g. temperature, pH, moisture content, micro- and macro-nutrient availability, presence of surfaces and O2) and biotic (presence of plants or grazing, antagonistic, competing or syntrophic organisms) factors. As natural environments are heterogeneous and dynamic, conditions vary in time and space (for review see van Elsas et al., 2000). It is difficult to predict the extent to which these factors affect natural conjugation, transduction and transformation, as they have differential effects under different conditions. A range of hot spots for HGT processes in soil environments (van Elsas, 1992) is discussed in detail later.

Figure 1.

Hot spots for horizontal gene transfer activity among plant- and soil-associated bacteria. These hot spots can be characterized by their capacity to enhance bacterial metabolic activity, resulting in enhanced rates of genetic exchange (and mutational change) processes, which leads to an enhanced adaptability of hot-spot-associated bacteria. I. The rhizosphere: enhancement of metabolic rates and mobility of cells and mobile genetic elements (MGEs). II. The phyllosphere: enhanced metabolism and possibility of cells to cluster, forming microcolonies or biofilms. III. Decomposing organic (plant- or animal-derived) material stimulating bacterial activities resulting in enhanced horizontal gene transfer (HGT) rates; IV. Manured soil: enhanced nutrient and water availability and mixing. V. In planta: plants can maintain high levels of endophytic bacteria in conditions of enhanced mixing and activity, particularly when pathogens invade; this may even lead to plant-to-bacterium gene transfer (Kay et al., 2002b); VI. Guts of soil animals: enhanced microbial mixing and activity.

Soil is heterogeneous with regard to the distribution of gaseous, liquid or solid compounds (Smiles, 1988). Clay–organic matter complexes are important sites for soil microorganisms because of their negatively charged surfaces and enhanced nutrient availability (Smiles, 1988). Water availability in soil is a second important factor driving microbial activity. In bulk soil, bacterial cells are mainly adsorbed to surfaces, often in the form of microcolonies, are refractory to movement and so do not contact cells at different locations. Hence, most bacterial cells in soil only interact with other cells in their immediate vicinity and are influenced by local conditions. Possibly, one of the most significant parameters in determining the physiological status of soil-dwelling bacterial cells is the presence of large interfaces between mineral and organic phases in soil. At these interfaces, which primarily comprise plant phytospheres, surfaces of decaying organic matter and guts of soil animals such as earthworms (Thimm et al., 2001) and Collembola (Hoffmann et al., 1998), nutrients can become concentrated (Table 1). The rhizosphere of many plants represents a region in soil with a (transient) high availability of organic carbon as well as nitrogen, phosphate and sulphur. Moreover, water flow in soil induced by plant roots may enhance bacterial movement. Similarly, aboveground plant parts (the phyllosphere) provide nutrient-rich surfaces. The guts of a range of soil animals also provide nutritional hot spots in which cells are physiologically activated and where gut movements stimulate cell-to-cell contacts.

Table 1.  Conceptual approaches for studying horizontal gene transfer between bacteria in natural habitats
Type level−1*Method1SystemMechanism and references
  • 1

    Adapted from van Elsas & Bailey (2002). Approaches are divided in direct experimental and retrospective (indirect) approaches. The direct approaches can be based on a system-disruptive (extractive) or a nondisruptive (gfp based) methodology. MGE, mobile genetic element.

I: Direct, disruptiveExtractive, cultivation-based coupled to molecular analysisPlant/soil microcosmsTransfer of MGE ( van Elsas et al., 1988, 1998; Lilley et al., 1994, 2002; Björklöf et al., 1995; Hoffmann et al., 1998; Kroer et al., 1998) Transformation (Nielsen & van Elsas, 2001; Kay et al., 2002b)
II: Direct, disruptiveExtractive, cultivation-based coupled to molecular analysisPlants in field soilMGE transfer in the phytosphere (Lilley et al., 1994; Lilley & Bailey, 1997; van Elsas et al., 1998, 2000)
III: RetrospectiveIsolation of MGE, molecular (sequence) analysisField: soil (aquatic, clinical)Detection of (sequences of) MGE, providing evidence of gene transfer (Hill et al., 1994; Götz et al., 1996; Levin & Bergström, 2000; Ochman et al., 2000; Smalla et al., 2000)
IV: Direct, nondisruptiveDonor-repressed gfp-labelled MGEBiofilms in laboratory (and nature)Detection of MGE transfer in biofilms (Christensen et al., 1996, 1998); In situ transfer (Dahlberg et al., 1998a, 1998b)

Evidence for the involvement of hot spots in HGT

A major hot spot for bacterial activity in soil is the phytosphere of plants. There is plentiful evidence that pinpoints, in particular, the rhizosphere as a major hot spot. First, the enhanced nutrient input and water fluxes in the rhizosphere have been implicated in the stimulation of conjugative plasmid transfers between inhabitants of this system, such as pseudomonads (van Elsas et al., 1988, 1998; Lilley et al., 1994; Pukall et al., 1996; Kroer et al., 1998). Second, the phyllosphere of plants has been shown to be equally conducive to conjugative plasmid transfer (Björklöf et al., 1995). Recent studies have further indicated that the horizontal gene pool in the phytosphere is highly mobile and directly linked to host fitness (i.e. the successful colonization by successive populations on developing plants) (Lilley & Bailey, 1997; Bailey et al., 2001). The host range of related plasmids in a mesocosm environment was also assessed. This was achieved using a genetically tagged plasmid in the Stellaria media (chickweed) phytosphere. The Pseudomonas spp. hosts that had acquired the plasmid 74 d after sowing varied widely (Lilley et al., 2003). The host range crossed populations of pseudomonads, niches (roots and shoots) and sampling times. These findings agreed with studies that investigated the associations of plasmids and hosts from crop, pasture and wild plants in the field, in which spatial and temporal (plant growth stage) variation was observed in the plasmid populations. One plasmid type was found on a wide range of plants and was carried by diverse pseudomonad hosts. Together, these studies indicated that the plasmids’ host range was not only diverse but also dynamic. A temporal component was found in the transfer behaviour and frequency of occurrence of these plasmids (Bailey & Lilley, 2002). While there are clear correlations between plasmid distribution and host plant species (niche) or growth stage (temporal variation), the distribution among pseudomonad hosts is less apparent. This seasonal or plant-development-based behaviour may be better understood when the traits conferred to the hosts have been identified.

An in vitro model system using a gfp–marked pseudomonad plasmid, pQBR11, was used to follow plasmid transfer in situ on agar-supported membrane biofilms (Lilley et al., 2003). These studies revealed peak transfer rates of 1 : 1000–1 : 1500 cells h−1; further, key factors for plasmid establishment were the size and activity (growth phase) of the donor populations and the effects of plasmid carriage on host fitness (assayed as population size). Secondary transfers, from transconjugant cells to other cells, played a minor role in plasmid establishment throughout the growth of the biofilm up to the stationary phase. These in vitro studies highlighted the conditions necessary for these phytosphere plasmids to transfer between, and persist in, populations in biofilms. These experiments point to the importance of cellular activity, which is likely to affect population size and therefore cell–cell contacts, on the transfer process. This, in turn, explains why the presence of decomposing material in soil is another hot spot conducive to HGT. For example, the addition of manure was clearly indicated as a stimulator of the persistence of Pseudomonas putida hosts and their mobilization of MGEs (Götz & Smalla, 1997).

The phytosphere is also a habitat in which HGT by transduction can be stimulated. As with all pathogens, bacteriophages exist in a trade-off between virulence (infectious spread among a population) and reproductive capacity (long-term persistence within a population involving vertical transmission) (Messenger et al., 2001). Highly infectious phages that induce high levels of mortality/morbidity risk local extinction of their host species, and so risk their own local extinction. Many phages have the capacity to escape their lytic lifestyle by adopting benign forms, either as lysogens integrated into the host genome or as noninfectious, autonomously replicating forms in the cell cytoplasm. In both instances, the host cell acquires resistance to infection by closely related phage, aiding host survival and vertical transmission of the phage. Stephens et al. (1987) showed that a naturally occurring lytic phage was responsible for the decline of introduced pseudomonads in the sugar beet rhizosphere, indicating an enhanced infection rate of the strain studied, and possibly suggesting that the indigenous populations were resistant to the phage. Mendum et al. (2001) recently focused on the potential for transduction between rhizobia in the rhizosphere and soil. The results suggested that adopting a benign form is a common phage survival strategy in these populations. The results also indicated that, where phage and susceptible bacteria coincide, especially in regions of enhanced bacterial growth such as the rhizosphere, infection would probably occur, increasing the chance of transduction events. Because phage DNA is often packaged in relatively resilient phage coats, it is possible that transducing phages provide a reservoir of bacterial genes under localized conditions where the host might not survive.

Burroughs et al. (2000) addressed the dynamics of phage–host interactions in soil using Streptomyces lividans, coelicolor and actinophage ΦC31. Strikingly, under soil growth conditions, ΦC31 showed a burst size (average number of phage released per bacterial cell) of up to 300, compared to 130 found in vitro. Moreover, only recently germinated spores were susceptible to phage infection in soil. These factors have a profound influence on the frequency of gene transfer and lysogeny, as they indicate that predictions made on the basis of in vitro experiments may actually underestimate some HGT frequencies in nature.

In another study, phages obtained from phytosphere enterobacteria and pseudomonads were shown to infect related bacterial species (Ashelford et al., 2000). This overlap in phage susceptibility, which provides the potential for interspecies HGT, was interpreted as a demonstration of an extended gene pool. Interestingly, not only did individual bacterial types have a specific range of phages that infected them, but also their susceptibility varied through the growing season. The apparent successions in abundance of host bacteria and infecting phage genotypes were recorded for both enterobacteria and pseudomonads. The interaction between these bacteria and their phages was found to be the most important factor driving bacterial persistence, succession and population density.

Transformation can also be enhanced by a range of different compounds exuded by plant roots in soil (Nielsen & van Elsas, 2001). Specific organic acids and amino acids had a significant stimulatory effect on the transformation of Acinetobacter sp. BD413. Other in planta plant parts are also seem conducive to transformation, even allowing the successful acquisition of plant-derived genes by bacteria (Kay et al., 2002b). The experiments showed that Acinetobacter sp. BD413, coinfecting tobacco plants with the plant pathogen Ralstonia solanacearum, was able to capture a bacterial marker present in the plant-derived DNA. The presence of homology between the captured DNA and the recipient genome was a prerequisite for successful acquisition, by recombination into the chromosome, of the plant-derived sequences.

The guts of soil animals, including Collembola and earthworms (Lumbricus rubellus), have been shown to stimulate HGT (Hoffmann et al., 1998; Thimm et al., 2001). Thimm et al. (2001), following pioneering work of Daane et al. (1996), confirmed this hot spot for HGT by providing evidence that earthworms can facilitate the spread of conjugative plasmids from nonindigenous bacteria to soil bacteria. Lumbricus rubellus stimulated the spread of plasmids from Escherichia coli donor strains harbouring marker-tagged plasmids with different properties (i.e. narrow and broad host range replication, conjugative, mobilizable and nonmobilizable). In microcosm studies with soil at 12°C, transconjugants were only detected in the casts of L. rubellus, indicating that gut passage was a precondition for plasmid transfer. Plasmid RP4 marked with a luciferase (luc) gene was transferred to indigenous bacteria at higher frequencies than detected in filter matings. Transconjugants were identified as a range of different gamma Proteobacteria.

Together, the above data confirm the earlier postulate (van Elsas et al., 2000) that HGT by conjugation, transduction and transformation is promoted in soil under conditions in which bacterial activity is stimulated. Another important observation is that possible outcomes of laboratory studies cannot be extrapolated to the field, as key determinative conditions may be different.

Gene mobilizing capacity of soil systems

As indicated above, MGEs are key mediators of gene transfer in specific environments within soil (van Elsas et al., 2000). However, there still is a paucity of information on the diversity and distribution of naked DNA and MGEs, especially phage, in soil and phytosphere habitats. Most is known about plasmids since, using a range of isolation procedures, plasmids with diverse characteristics with respect to incompatibility (Inc) group, host range, transfer proficiency and the type of accessory genes present have been obtained. One of the most effective methods to obtain plasmids that are transfer-proficient is (bi- or tri-parental) exogenous isolation. These methods capture such plasmids directly from environmental samples into recipient strains that can be grown in the laboratory (Bale et al., 1988), and have been successfully applied to soil and phytosphere habitats (Lilley et al., 1994; van Elsas et al., 1998). One such plasmid, denoted pIPO2, was shown to self-transfer and to mobilize IncQ plasmids to a range of diverse gram-negative bacteria in the wheat rhizosphere in the field (van Elsas et al., 1998). The fact that plasmids can mobilize other, often unrelated, plasmids that themselves are unable to transfer to other hosts is an important observation, as it suggests that any piece of DNA can potentially be transferred between cells. On the basis of specific primers and probes applied to environmental DNA, it was suggested that pIPO2 occurs primarily in the rhizospheres of a variety of plants such as wheat, grass, potato and tomato (Tauch et al., 2002). While the plasmid clearly can be a mediator of in situ transfer processes in rhizospheres (van Elsas et al., 1998), its role and fitness benefits to the host are still unclear.

Plasmids carrying mercury resistance determinants able to mobilize IncQ plasmids like RSF1010 have frequently been isolated (van Elsas et al., 2000). This in part reflects the usefulness of mercury as a selective marker for environmental plasmids; however, an adaptive role for mercury resistance is inferred by enhanced prevalence of these plasmids under conditions of mercury stress. Recently, plasmids obtained by exogenous isolation from soil were tagged with gfp and their transfer into soil bacteria observed in vitro (Drönen, 1998). Three plasmids showed high transfer frequencies and very broad host ranges, whereas five others with narrower host ranges transferred at lower rates. Together, these results point to the natural occurrence of a broad diversity of plasmids that can serve as mediators of HGT in soils and related environments. However, we are still far from understanding the full diversity of plasmids and the extent of their roles in bacterial communities and as transfer mediators.

Effects of selection on HGT between bacteria in soil

Selective pressure is the key environmental factor that can exacerbate the effects of gene transfer processes. The effects of selective pressure are most easily seen in cases in which the HGT event confers a selective (growth) advantage to bacterial hosts. Top et al. (2002) recently reviewed the issue of selection acting on gene transfer in soils and presented diverse lines of evidence indicating strong effects of selection. For example, a catabolic plasmid encoding a 2,4-dichloropropionate (DCPA) degradative gene was transferred from an Alcaligenes xylosoxidans donor to members of the indigenous community in soil, but transfer was detected only in soil treated with DCPA. In addition, Enterobacter agglomerans carrying the self-transmissible biphenyl-degadative plasmid RP4::Tn4371 was introduced as a (nonexpressing) donor into soil with or without added biphenyl. The introduced donor strain was unable to persist and declined to undetectable levels very quickly. By contrast, indigenous transconjugants belonging to the genera Pseudomonas and Comamonas able to degrade biphenyl and carrying RP4::Tn4371 were shown to thrive but, again, only in the soil that had been pretreated with biphenyl (Top et al., 2002). Later, the in situ acquisition of biphenyl-degradative genes by introduced bacteria was again seen only in biphenyl-treated soil. These observations have two explanations. First, it is possible that the enhanced metabolic advantage gained by access to an abundant carbon source means that donors are more abundant and have greater metabolic activity than recipients, and that the transfer rates are inherently higher in the treated soils. Conversely, transfer rates might be similar in each site but fixation of HGT processes might be higher owing to the selective advantage in the contaminated sites and therefore detection rates are higher.

The transfer of the 2,4-dichlorophenoxy acetic acid (2,4-D) degradative plasmid pJP4 from Ralstonia eutropha JMP134 to Variovorax paradoxus was only detectable in the presence of high levels of 2,4-D in soil (Top et al., 2002). Furthermore, such high concentrations of 2,4-D in soil enhanced the apparent transfer rate of the 2,4-D degradative plasmid pJP4 from introduced strain JMP134 (which died out rapidly under these conditions) to indigenous Pseudomonas and Burkholderia spp. (Top et al., 2002). Finally, Top and coworkers attempted to transfer the 2,4-D degradative plasmids pJP4 and pEMT1 to the indigenous bacteria of the soil A- and B-horizons (Top et al., 2002). No 2,4-D degradation occurred in the uninoculated control soil during at least 89 d, while inoculation and subsequent plasmid transfer resulted in complete degradation of 2,4-D within 19 d. Transconjugants related to Burkholderia graminis, Burkholderia caribensis and R. eutropha formed and proliferated in this system. Overall, this work clearly demonstrated that gene transfer and establishment among soil bacteria is directly affected by the presence of driving forces such as a utilizable carbon source, resulting in bioaugmentation in these examples.

In addition to the role of simple whole plasmid or gene/operon transfers, there was strong indirect evidence for a role of genetic recombination in the adaptation of bacterial aquifer communities to chlorobenzene (CB) (van der Meer et al., 1998). Bacterial adaptation resulting in a novel CB-degradative pathway presumably occurred following gene transfer and recombination.

Studies of the ecology of MGE performed by Wellington and coworkers (Herron et al., 1998) have focused on the impact of selection on plasmid transfer in streptomycete hosts in soil. The spread of selectable traits (e.g. drug resistance) was studied. Transfer was detected in soil, and frequencies at which transconjugants were recovered were affected by the presence of selection for plasmids with antibiotic resistance. These investigations clearly demonstrated the overriding forces exerted by selection for either the utilization of a specific carbon source or a resistance to antimicrobial compounds, in shaping bacterial populations in nature as a result of HGT.

Bacteria as ‘samplers’ of environmental DNA – relevance to plant-associated communities

To date, 90 prokaryotic genomes have been sequenced and 134 are in progress (Spiers et al., 2001). For current numbers, see http://www.ncbi.nlm.nih.gov/PMGifs/Genomes/micr.html. These provide a unique resource to investigate the impact that HGTs have had during bacterial evolution. Direct comparisons between genomes of single species are the easiest to interpret and understand. This was first possible with the publication of two E. coli genomes, the first of strain K12 (Blattner et al., 1997), the laboratory workhorse, and the other one of the enterohaemorrhagic strain O157:H7 (Perna et al., 2002). Strain O157:H7 has 1.34 Mbp of sequence, or 1387 (26%) genes that are absent from K12. The majority of these genes were identified as phage (or phage-like) sequences or virulence functions described in other enterobacteria and were presumed to have been acquired by HGT processes (Perna et al., 2002). A similar comparison of the genomes of two phytopathogenic Xanthomonas species (Da Silva et al., 2002) showed that these shared more than 80% of genes. The species-specific differences could often be correlated with virulence and host-range functions and were predicted to have arisen due to HGT. The preponderance of genomic differences in niche-adaptive (host colonization/pathogenicity) traits is indicative of the importance of HGT in plant-associated bacterial processes.

In the absence of multiple sequences from the same species, the next most informative comparisons involve closely related species from the same genus or family. In the context of soil- and plant-associated bacterial genome sequences, rhizobia are numerically the best represented in this context. To date, the genomes of three plant-associated bacteria (Mesorhizobium loti MAFF303099, NC_002678, Sinorhizobium meliloti 1021, NC_003047 and Agrobacterium tumefaciens C58, NC_003305) and two animal-pathogenic bacteria (Brucella melitensis NC_003317 and Brucella suis NC_004310) have been sequenced in this context. Perhaps informatively, the genomes of the facultative intracellular animal pathogens, B. melitensis and B. suis, are the smallest (Table 2) by 2.3 Mbp. These organisms show very limited evidence for acquisition of external DNA (Tsolis, 2002). This might indicate that they spend most of their time in the intracellular niche and exist only transiently in the environment. Reduced genome size, as is apparent for the brucellas, is a common feature of intracellular pathogens and is thought to reflect gene losses due to alleviation of purifying selection in such nutritionally rich and homogenous environments (Itoh et al., 2002). It is therefore tempting to speculate that much of the additional DNA in the soil/plant-associated species represents an adaptation to a highly variable environment. This hypothesis is supported by the observations that substantial amounts of the additional DNA contained in the plant-associated species is associated with plasmids or other mobile genetic elements (including the 0.5 Mbp symbiotic island present on the chromosome of M. loti). Indeed, in all of three of the plant-associated rhizobial genomes sequenced to date, the genes that make them interesting to humans, either nitrogen-fixing symbioses with legumes (M. loti and S. meliloti) or tumour induction (A. tumefaciens), are carried on MGEs. Numerous population studies have clearly demonstrated that these traits are dynamic, often moving between strains and species in a host-plant responsive manner (Sullivan et al., 1996).

Table 2.  Summary of fully sequenced genomes of plant-associated and other bacteria
SpeciesGenomic componentSize (Mbp)%GCAccession number
  • 1

    Figure includes 0.5 Mbp symbiotic island.

  • 2

    2 These plasmids can not be cured from the host and may represent secondary chromosomes.

Mesorhizobium loti MAFF 303099Chromosome7.062.71NC_002678
Sinorhizobium meliloti 1021Chromosome3.762.7NC_003047
Agrobacterium tumefaciensC58Chromosome2.859.4NC_003062
Linear chromosome2.159.3NC_003063
Brucella melitensisChromosome I2.157NC_003317
Chromosome II1.257NC_003318
Brucella suisChromosome I2.157.2NC_004310
Chromosome II1.257.3NC_004311
Ralstonia solanearumChromosome3.767.0NC_003295
Xanthomonas axonopodisChromosome5.264.8NC_003919
Xylella fastidiosaChromosome2.752.7NC_002488

Both the Xanthomonas and rhizobial genome and population ecology data clearly suggest that HGT among plant-associated bacteria is an important mechanism enabling these bacteria to sample the available gene combinations present in the community and so exploit the host plants that are available at a given moment. The genome of another plant pathogen, R. solanacearum, shows similar high levels of exogenously acquired DNA, much of which contains pathogenicity determinants (Salanoubat et al., 2002). Thus, there is mounting evidence that HGT among plant-associated bacteria is an important factor in genome adaptation and evolution. At least some bacteria may sample DNA from other sources, as indicated by the finding that both R. solanacearum and Acinetobacter sp. develop competence in planta. (Bertolla et al., 1999; Kay et al., 2002a), thus increasing their chances of acquiring genetic material from either other bacteria or plant cells (Kay et al., 2002b).

Lawrence & Ochman (1997) proposed that anomalous G + C content (especially at third codon positions) of genes was indicative of recent horizontal acquisition from an organism with a different overall G + C profile. From their analyses, they inferred that 17% of the coding sequence of E. coli had been acquired by HGT since the divergence of E. coli and Salmonella typhimurium c. 100 million years ago. However, Koski et al. (2000) demonstrated that this is not exclusively true. They showed that some genes with anomalous G + C content are unlikely to be the products of HGT, whereas, as Ochman et al. (2000) also pointed out, ancient HGT events would not be detectable using this criterion because of the process of amelioration. Amelioration describes how horizontally acquired DNA from a distantly related organism with a distinct genomic signature (such as G + C content) will, because of mutational drift and selection, gradually acquire the genomic signature of its new host. The time window in which anomalous G + C signatures are detectable is dependent upon the level of discrepancy between the host genomic signatures and those of the horizontally transferred sequences; more similar signatures are more likely to be ameliorated more rapidly. Putative HGT events must therefore be assessed on a case-by-case basis and take into consideration other known factors such as the likely source of the transferred DNA and the taxonomic distribution of orthologous genes (Ochman et al., 2000).

The genomic signatures of the chromosome and associated MGEs of each the three genomes of the soil-dwelling rhizobia are different (Table 2). Downie & Young (2001) suggested that this could reflect differential mutational and selection pressures acting on the different genomic components (chromosome vs MGEs). To reiterate and extend their argument in light of the additional A. tumefaciens genome: the symbiotic genes and tumour-inducing genes characteristic of plant-associated rhizobia are all associated with species that have relatively high genomic G + C content. Similarly, the plasmid replication sequences (repABC) identified on many rhizobial plasmids (Table 2) have so far only been identified in rhizobia and closely related α-proteobacterial genera, again all with relatively high G + C genomes. If these traits and plasmids, which have anomalous G + C signatures, have recently been acquired from organisms with a substantially different G + C content, where are these other organisms in the environment? Why should such cryptic bacteria maintain these ecologically useful genes but apparently not use them, since they have not been found associated with plants? The occurrence of lower G + C content signatures in plasmids compared with their associated genome is also apparent in the full genome sequences of other plant-associated species (Table 2). This general pattern, identified in unrelated plasmids and species, provides independent and more general support to the hypothesis of Downie & Young (2001) that genes in the accessory gene pool may be under different mutational and/or selection pressures than chromosomal genes. Whether this is true, and how these differences might be maintained is yet another challenge on the road to fully understand the evolution and role of HGT in bacteria.

Horizontal gene transfer as a community phenomenon – advantages to bacterial populations or to MGE?

To date, a considerable body of knowledge on HGT processes has accumulated. However, there has been too little attention in this context for the role of microbial population ecology. Events in natural environments such as microcolony distributions, biofilm formation, cell-to-cell signalling and habitat sensing are likely to affect HGT processes. In addition, impact/penetration of selective agents on highly structured microbial communities, such as those found in a biofilm, is poorly understood. Consideration of gene regulation and expression processes that influence the behaviour and population dynamics of bacterial communities is key to understanding the significance and impact of gene transfer. Current evolutionary concepts dictate that every function in (population) biology has a cost in addition to a potential benefit, and that biological systems evolve towards minimizing costs while maximising (fitness) benefits. Microbiological systems in natural settings most often face energy limitations, making any additional expressed trait potentially ‘costly’. In other words, the cost of any added functions lies in the energy the cell has to spend to make the function operational. Benefits of the added traits can be of varying kinds, which includes the very obvious benefits of, for example, the degradation of specific carbon sources and the resistance to specific antimicrobial compounds. These (simplified) concepts predict three situations in the ecology of the interaction between bacterial host and horizontally acquired DNA (HADNA):

  • 1The host gains a net selective disadvantage from the interaction with the HADNA owing to expression of a nonbeneficial trait;
  • 2The host gains a clear net selective advantage from carriage of the HADNA;
  • 3The host–new DNA interaction is ecologically neutral (i.e. there is no obvious advantage or disadvantage from carriage of the HADNA).

The net outcome of host–HADNA interactions, though, is not always as clear-cut as described by these three simple interactions. This results from the complexity and heterogeneity, in both time and space, of any ecosystem, which impose different selective pressures that act on the isolated host–HADNA combinations and drive them to proliferate, to go extinct or to form new associations. As hosts and MGEs have often effectively coevolved, they may share or complement strategies for interaction and mutual survival. In addition, it is probable that no two host cells are physiologically identical: populations, even of the same bacterial type, exist as a collection of interacting and structured communities, or as dispersed single cells. Such variance in the life cycle of bacterial cells probably results in the largely divergent outcomes observed when cells and HADNA interact. These range from the very rapid (infectious) spread of traits associated with plasmids like RP4 through a bacterial community, which can eventually occupy up to 100% of potential recipient cells in just a few hours under the appropriate conditions, to the apparent reluctance of structured microbial communities to accept exogenous DNA or active MGEs. Different populations and strains respond differently to different MGE (frequently owing to host-range constraints) and the outcome is therefore difficult to predict. However, the interactions that bacteria exhibit with, and in response to, their environment and upon contact with potential donors of DNA are both intimate and complex.

Horizontal gene transfer from the perspective of an MGE

Whereas the fitness gain from HGT for populations of bacterial cells may often be apparent, there is less clarity about the strategies that have evolved in MGEs to assure their self-perpetuation. Mobile genetic elements (plasmids, bacteriophages) will be affected by the varying strategies, including concerted differentiation events, developed by their potential hosts to colonize and explore their environmental niches. It seems obvious that MGEs, in order to persist within microbial consortia, have developed strategies that enable them to either be replicated and maintained within growing cells in an ecologically relevant way (vertical transmission), or to be transferred at ecologically optimal rates (horizontal transmission), or both (Thomas, 2000). Vertical transmission may be dependent on plasmid-encoded replication systems, or on integration into the host chromosome. Horizontal transmission is equally dependent on plasmid-encoded functions, and sometimes requires functions from other sources such as the host or other MGEs. Individual plasmids have adopted different strategies that facilitate their self-perpetuation, and apply them in accordance with the ecophysiological status of host cells (Wilkins, 1995; Thomas, 2000). Hence, it is likely that, at the level of the community, different types of interactions, including vertical transmissions, horizontal transfers and even integration or excision events occur simultaneously.

A mechanistic description of the factors that influence the behaviour and function of MGEs and their response to the strategies adopted by their hosts remains unexplored. This is likely to persist until more sensitive methods are available that allow the study of gene expression and the phenotype of single cells. Hence, novel research strategies are needed to gain an understanding of the genetic solutions evolved by MGEs to combat or complement those of their host to ensure their survival. Solutions will certainly include tools that come from the emerging fields of ‘-omics’. First, the explosive development of bacterial genomics, in which whole genomes, including their associated MGEs, can be sequenced and compared, provides new opportunities. These in turn facilitate transcriptomics and proteomics to be applied in studies on the variation in gene and protein expression under different conditions. Complementing these technologies is the potential offered by promoter trapping and related approaches to evaluate gene expression under ecologically representative conditions and to determine their impact on bacterial fitness (survival in vivo) and phenotype.

Horizontal gene transfer in biofilms

There is emerging evidence that microbial life in natural settings mainly takes place in microbial consortia. These frequently occur in the form of highly complex structures such as microcolonies or biofilms rather than as isolated cells (Tolker-Nielsen & Molin, 2000). In these structured communities, cell-to-cell signalling (quorum sensing), cellular differentiation, perception of the local environment via two-component regulatory systems, and responses to environmental stresses play key roles in determining the behaviour at the level of the individual and the community as a whole. This interdependency of the role and activity of free cells, cells in biofilms or those in microcolonies produces populations that are morphologically, physiologically and even genetically very different (Rainey & Travisano, 1998). For example, a biofilm consisting of unstable gfp-marked Pseudomonas putida cells (Andersen et al., 1998) established on a glass slide differentiated into several sections, each one of which could support different morphotypes (J. Haagensen, unpublished). Subsequent challenge of the biofilm with sublethal concentrations of the antibiotics kanamycin and tobramycin, followed by vital staining revealed that throughout its development the biofilm comprised a heterogeneous collection of quiescent to metabolically active cells.

The community lifestyle of most microorganisms in biofilms has special implications for HGT processes. An as yet conceptual model dictates that in young biofilms, HGT by conjugation is related to the growth of donor and recipient cells (Prosser et al., 2000). To this end, it is not surprising that the initial nutrient concentration was found to be a key determinant of HGT rates. However, the model did not address spatial heterogeneity within biofilms or events in mature biofilms, and it is likely that more complex models are needed for these. Christensen et al. (1996, 1998), studying the spread of a gfp-marked TOL plasmid in a mature biofilm, noted that transfers of the TOL plasmid were confined to the outer (upper) biofilm regions with presumably the highest metabolic activity. The lack of detectable fluorescence by cells deeper in the biofilm might suggest that they are unable to serve as plasmid recipients. Alternatively, this might reflect a failure of the plasmid to establish itself within the new host due to an excessive genetic or metabolic load. A third option is that this reflects an inability of the new host to achieve sufficient GFP expression to enable detection of the transfer event. The latter scenario might be informative, since, if the new recipient is metabolically unable to meet the challenge of GFP production, it is also unlikely to be able to meet the metabolic load required to build the complex conjugation machinery necessary for plasmid transfer to its neighbours. The results of Sternberg et al. (1999) using an unstable GFP, suggest that the metabolic activity of cells away from the surface of biofilms is lower compared with those at the surface. In mature biofilms, HGT probably depends on the physiological state of donor and recipient cells and/or on the ability of cells to move within the biofilm (Lilley & Bailey, 2002). In addition, the spatial and temporal heterogeneity of biofilms will play a role. Finally, HGT in biofilms may be affected by the action of other microorganisms on that biofilm.

Despite this emerging information, key questions regarding the functioning and impact of HGT mechanisms in biofilms still need to be answered. For example, are HGT processes specifically triggered, and how are these responses coordinated between the complex of recipient, donor and MGE? If they are, do control mechanisms act at the population level rather than at the individual cell level? Some of the answers to these questions are already available. For example, it has long been known that HGT between streptococci is dependent on a quorum sensing mechanism in which a small peptide serves as the signalling compound (Clewell, 1993). In other words, plasmid transfer in this case is a highly orchestrated population event, involving multiple partner cells at least in the signalling phase. To further illustrate this point, in the mechanisms stimulating Ti plasmid transfer between Agrobacterium cells, apparently the traR product, which is regulated by opines, plays a central role. The autoinducer AAI was shown to coact with traR product in the induction of transfer, and, thus, cell density was crucial. This complexity of the regulatory circuits of Ti plasmid transfer gene expression demonstrated that further understanding of gene transfer requires the study of the signals that regulate the expression of conjugative systems.

From the foregoing, we suggest that not only stress responses, quorum sensing and diffusible factors, but also the physiological state of donor and recipient cells need to be investigated by direct experimentation. The necessary tools for such in situ work have been or are being developed in several laboratories. Furthermore, it is tempting to speculate that the Agrobacterium transfer system is widespread and developed primarily for transfer of mobile genetic material, but how certain can we be of this? The transfer apparatus encoded by the Ti plasmid involves a type IV secretion system, and similar systems have been found on other environmentally isolated conjugative plasmids such as pIPO2 (Tauch et al., 2002), but also in bacteria with highly divergent excretion functions (e.g. pathogens that excrete toxic proteins).

Concluding remarks

To elucidate the outcome of gene transfer in the phytosphere, it is required that consideration of the ecological principles that affect transfer, such as temperature, host nutrient status, cell-to-cell contact rates and selective pressure be supplemented by other measures at a more refined level. These include assessments of how bacterial host cells sense, and respond to, environmental changes, how they interact with other – similar or dissimilar – cells (by mechanisms such as signalling, competition, antagonism, predation or parasitism), and how these community-level processes affect HGT. In addition, and perhaps most importantly, there is a need to identify and define the fitness-affecting traits that are acquired or carried by the horizontal gene pool. This potential is largely cryptic and may pertain to functions that are difficult or impossible to mimic in the laboratory, yet are of extreme importance for the survival of the host and HADNA alike. However, the combination of molecular biology, genomics and ecology is bound to reveal many of the best-kept secrets of the HADNA. For example, novel gene clusters containing open reading frames, which probably encode small proteins of unknown function, have been described on many plasmids and chromosomes (Tauch et al., 2002). It may well be that these proteins play a role in the cell-to-cell contact mechanism leading to HGT in the natural habitat. Alternatively, they may function in the interaction between the HADNA host, with the rhizosphere as its preferred niche, and the plant. HADNA carriage may thus be advantageous to the host dwelling in this natural setting. This is supported by the observation that the carriage or transfer of large plasmids by Pseudomonas spp. appeared to be advantageous only at a very particular period during growth of the host plant (Lilley & Bailey, 1997). This may suggest that carriage of these plasmids and expression of plasmid functions promote fitness of the host only under certain conditions. Plasmid pSym carriage was shown to be a disadvantage in rhizobia surviving in bulk soil in the field, yet to enhance fitness in the rhizosphere (Sullivan & Ronson, 1998). This pointed to a function other than just nodulating capacity as the mechanism behind the fitness-enhancing trait. These observations provide interesting leads for future fundamental research on the ecology of HGT between bacteria in plant-related habitats. In particular, there seems to be a much more intimate relationship, in ecological terms, between transfer and host fitness than previously thought. In other words, the mechanism and triggering of MGE transfer may have very intricate links to host fitness and/or sensing of conditions in the natural habitat.

Finally, it is axiomatic that the traits carried over in successful HGT events, that (positively) affect host fitness, are of ecological and evolutionary importance. However, we need to better understand the ecological aspects of HGT, particularly in respect to questions on what types of conditions, compounds or signals stimulate HGT, resulting in stable incorporation into the new host. Furthermore, resolving those environmental conditions that affect cellular metabolism and cell-to-cell contacts remains the key focal point for many researchers. The application of new molecular tools, such as DNA arrays, will allow the study of in situ expression of transfer-related genes. It is hoped that this will shed more light on the factors in natural habitats that trigger the events leading to cell-to-cell contact (mating pair formation) and HGT. Alternatively, the use of reporter genes downstream of promoters involved in triggering the expression of these genes can be used.


This work was supported by grants from the European Union (Projects RESERVOIR and MECBAD), the Dutch Ministry of Agriculture (DWK352) and the Natural Environment Research Council, UK. We thank Rob Pastoor for providing help in the preparation of Fig. 1.