Gene Transfer between Salmonella enterica Serovar Typhimurium inside Epithelial Cells

Virulence and antibiotic resistance genes transfer between bacteriaby bacterial conjugation. Conjugation also mediates gene transferfrom bacteria to eukaryotic organisms, including yeast and humancells. Predicting when and where genes transfer by conjugationcould enhance our understanding of the risks involved in therelease of genetically modified organisms, including those beingdeveloped for use as vaccines. We report here that Salmonellaenterica serovar Typhimurium conjugated inside cultured humancells. The DNA transfer from donor to recipient bacteria wasproportional to the probability that the two types of bacteriaoccupied the same cell, which was dependent on viable and invasivebacteria and on plasmid tra genes. Based on the high frequenciesof gene transfer between bacteria inside human cells, we suggestthat such gene transfers occur in situ. The implications ofgene transfer between bacteria inside human cells, particularlyin the context of antibiotic resistance, are discussed.

INTRODUCTION

Top

Introduction

Genomic databases have confirmed the enormous contribution thathorizontal gene transfer (HGT) has made to the content and eventhe organization of bacterial genomes (for reviews see references23, 37, 42, and 48). One theory even attributes creation ofthe most famous of the bacterial chromosomal structures, theoperon (41), to HGT. Genes transfer between bacteria via manydifferent kinds of vectors, ranging from conjugative plasmidsand conjugative transposons to integrons and cis-acting signalsequences, which mediate transformation of organisms such asHaemophilus and Neisseria, to viruses. These vectors are collectivelyreferred to as horizontally mobile elements (HMEs) (9, 26, 45,52). If HGT is powerful enough to affect the organization ofchromosomes, then what effect does it have on the organizationand function of the elements that mediate gene transfer?

A central question in bacterial evolution, and a question thathas immediate medical interest, is why antibiotic resistanceand new virulence-determining genes are so often found on HMEsrather than on chromosomes of dominant clones of bacteria (10,27, 29, 59). Further study of HGT in both the environment andthe laboratory is necessary to answer the questions about howchromosomes and HMEs attract different kinds of genes (14, 15,44).

We sought to address such questions concerning the evolutionof antibiotic resistance by studying HGT between invasive bacteriathat cause disease in humans. Many bacteria invade animal cellsduring pathogenesis (17). This ability to invade animal cellshelps the bacteria evade both the host immune system and antibiotics(51). Eukaryotic cells protect internalized bacteria from otherwell-known stresses in a variety of ways. For example, somebacteria consumed by protozoans are protected from environmentaland chemical stresses (4), and many antibiotics penetrate animalcells poorly or accumulate insufficiently in the intracellularcompartments to kill invasive bacteria (51, 60). Thus, we askedwhether the cytoplasm of animal cells could be a niche for genetransfer between bacteria protected from antibiotics that mightotherwise inhibit gene transmission.

We postulated that if bacteria conjugate in the intracellularenvironment, such conjugation could partially explain both theevolution of virulence and the spread of antibiotic resistancein invasive pathogens. Antibiotic-susceptible bacteria thatescape the effects of antibiotics by entering human cells couldacquire resistance genes from pathogens residing in the samecells, from benign bacteria taken up by the cells (20), or throughmixing with the normal flora upon reemergence from the cells.Conversely, normally benign bacteria could acquire virulencegenes intracellularly from resident pathogens, since many newvirulence traits are plasmid borne (3, 54).

These ideas are plausible because of the elevated frequenciesof gene transfer between bacteria that are associated with protozoancells (55). Transmission of the conjugative plasmid RP4 betweenEscherichia coli strains increased 3 orders of magnitude inthe presence of the protozoan Tetrahymena pyriformis, suggestingthat plasmid transfer occurred within digestive vacuoles.

Using antibiotic resistance genes as markers and the invasivepathogen Salmonella enterica serovar Typhimurium as a model,we sought to determine whether conjugative plasmids could betransferred between bacteria inside cultured human cells.

How S. enterica serovar Typhimurium invades epithelial cellshas been well described previously (17, 34, 47) and so is summarizedonly briefly here. Epithelial cells take up Salmonella whenthe bacteria translocate signal transduction-altering proteinsto the cell. Cytoskeletal rearrangement, membrane ruffling,and bacterial uptake by macropinocytosis follow the translocationof bacterial proteins. Engulfed bacteria reside inside membrane-boundvacuoles within the cell cytoplasm. Whereas some bacteria lysevacuoles and are released into the cytoplasm, S. enterica serovarTyphimurium remains in the vacuoles.

Bacteria inside cultured mammalian cells are often distinguishedfrom extracellular bacteria by the fact that the intracellularbacteria are protected from gentamicin in the medium (36). Weexploited the fact that human cells exclude gentamicin to ensurethat viable recipient and donor bacteria, introduced into amonolayer of cells at different times, could not form transconjugantsin the extracellular medium. To our knowledge, this is the firstdemonstration of intracellular bacterial conjugation, and ourresults reveal both the functionality of the bacterial conjugativemachinery in the intracellular environment and the potentialfor transfer of genes between intracellular bacteria, particularlyin the presence of antibiotics.

MATERIALS AND METHODS

Top

Materials and Methods

Bacteria. ${Delta}$ invA strains were created by P22 transduction (13) of the invA61::TnphoAmutation from SB111 (21) to S. enterica serovar Typhimuriumwild-type strains SL1344^R (32) and 14028^R-P (5). The strainsand plasmids used in this study are described in Table 1.

View this table:
[in this window]
[in a new window]
TABLE 1. S. enterica serovar Typhimurium strains and plasmids

Cell culture. INT-407 cells (ATCC CCL 6) were cultured in minimal essentialmedium with Earle's salts (Gibco) supplemented with 2 mM nonessentialamino acids (Gibco), 2 mM L-glutamine (Gibco), and 10% fetalbovine serum (Gibco) at 37°C in a 10% CO₂ atmosphere.

Conjugation experiments. For intracellular conjugation, 24-well (1.9-cm²) trays (Nunc)were inoculated with 4 x 10⁵ tissue culture cells/ml (0.5 mlper well), and cells were grown to confluence (24 to 48 h).In some experiments, cells were grown on 10-mm membrane filterinserts with 2-µm-pore-size membranes (Anopore; Nunc).Prior to infection, monolayers were washed once with phosphate-bufferedsaline (PBS) (8 g of NaCl per liter, 0.2 g of KCl per liter,1.44 g of Na₂HPO₄ per liter, 1.43 g of KH₂PO₄ per liter; pH7.4), and they were maintained in minimal essential medium withEarle's salts supplemented with L-glutamine, nonessential aminoacids, and 1% fetal bovine serum throughout the experiment.

Exponential-phase cultures of recipient (plasmid-free) bacteriawere washed and diluted in PBS to a concentration of 1 x 10⁸bacteria/ml. Wells were inoculated with $~$ 10⁷ bacteria to givea multiplicity of infection (MOI) of approximately 10 and centrifugedat 317 x g for 5 min. Following 2 h of incubation at 37°Cin the presence of 10% CO₂, monolayers were washed three timeswith PBS and then incubated in medium containing 100 µgof gentamicin per ml for 1 h. The procedure described abovewas repeated with donor (plasmid-bearing) bacteria. One hourafter the second incubation with gentamicin, the gentamicinconcentration was decreased to 20 µg/ml. After an additional3 h the monolayers were washed with PBS and lysed with 0.5%deoxycholate to release the intracellular bacteria (39).

Recombinant selection and transmission frequency calculations. Intracellular bacteria were enumerated by serial dilution inPBS and plating on Luria-Bertani (LB) agar (Gibco) plates supplementedwith antibiotics appropriate for the selection of the recipientbacteria (60 µg of nalidixic acid per ml), the donor bacteria(100 µg of rifampin per ml), and the recombinant bacteria(60 µg of nalidixic acid per ml, 100 µg of kanamycinper ml). Control crosses on agar plates were performed as describedpreviously (28).

The transmission frequency was calculated by determining thenumber of recombinants per limiting intracellular parent (28,31). Because the number of bacteria that internalize in tissueculture cells can and did vary from experiment to experiment,frequencies had to be calculated for a common limiting parentfor meaningful comparisons between experiments. Generally, fewerdonors than recipients were recovered from tissue culture cells,but on occasion the recipient was the limiting parent.

In any conjugation assay, some recombinants can form after thedonor and recipient conjugants are transferred to the selectionplates (31). The proportion of recombinants that formed afterplating was determined by mixing lysates of donor-infected cellswith lysates of recipient-infected cells on LB agar plates supplementedwith nalidixic acid and kanamycin. Extracellular bacteria inthe culture medium were concentrated by centrifugation (15,000x g for 3 min), rinsed once in PBS, and then enumerated by platingon LB agar. The resulting colonies were transferred by replicaplating to LB agar supplemented with nalidixic acid and kanamycinto test for recombinant markers (43).

All recombinant colonies were transferred by replica platingto LB agar plates supplemented with rifampin to test whetherthey were nalidixic acid-resistant donors that arose spontaneously.

RESULTS

Top

Results

Plasmids were transmitted between invasive S. enterica serovar Typhimurium cells during infection of cultured human cells. Two conjugative plasmids were tested for transmission betweenbacteria internalized in human cells. The RP4 (IncP ${alpha}$ ) and F (IncF1)plasmids used express conjugation genes constitutively (25,35). Moreover, there are interesting differences in their frequenciesof transmission in liquid and solid environments (8). WhereasF plasmids are transmitted at high frequencies in both environments,IncP plasmids are transmitted between bacteria at much higherfrequencies on solid surfaces.

The transmission frequencies of RP4 and F were measured by matingexperiments performed on LB agar plates as previously described(28) (Table 2). The frequency of transmission of F between theSL1344 strains was only 1% of the frequency of transmissionof RP4 (Table 2). This was expected because the Salmonella virulenceplasmid, pSLT, inhibits transmission of F (53). The inhibitionof fertility caused by pSLT was relieved by using the pSLT-curedstrain 14028-P (Table 2).

View this table:
[in this window]
[in a new window]
TABLE 2. Agar plate transmission frequencies^a

Monolayers of the human intestinal cell line INT-407 were infectedwith recipient (plasmid-less) bacteria and donor bacteria carryingeither RP4 or F. In order to preclude the possibility that geneexchanges could occur outside human cells, cell monolayers wereinfected in series with intervening incubation in the presenceof gentamicin to kill extracellular bacteria (see Materialsand Methods). Donor invasion was halted by the addition of gentamicin(36). While some cells detached and were washed away, the majorityremained attached and appeared to be viable throughout the experiment.More sensitive assessments of viability were not made.

Recombinant bacteria were recovered after cell lysis by selectionof plasmid-borne antibiotic resistance markers in the recipientbacteria. Recipient bacteria with RP4 markers were recoveredat a frequency of 2 x 10^-4 recombinant colony per limiting parent(Table 3). F was transmitted between intracellular Salmonellaat a frequency of 7 x 10^-4 transconjugant colony per limitingparent (Table 3). All recombinants were positive for a secondplasmid marker, tetracycline resistance (RP4), or the abilityto use lactose as a sole carbon source (F42::mini-Tn10Kn).

View this table:
[in this window]
[in a new window]
TABLE 3. Intracellular transmission by Tra⁺ and Tra^- plasmids^a

Trivial explanations for the recovery of recombinants were eliminatedby control experiments. First, the possibility that the recombinantswere spontaneous nalidixic acid-resistant mutants of donors(nalidixic acid resistance was the marker used to select recombinantrecipients) was eliminated by confirming that putative recombinantswere still sensitive to rifampin (the donor-specific marker).Second, the abilities of equivalent numbers of bacteria thatwere released from independently infected cells onto selectivemedia to mate on the enumeration plates were determined (seeMaterials and Methods), and the results did not account fora significant proportion of the recombinants. The total numberof recombinant colonies observed after plating in eight independentexperiments performed in triplicate was 138. In the same experiments,9,038 bacterial recombinants were recovered from mixed infectionsof tissue culture cells. Thus, plasmid transmission after platingaccounted for an average of 2.3% of the recombinants and neveraccounted for more than 10% of the recombinants observed. Norecombinant colonies were observed after plating in three independentexperiments performed in triplicate when F plasmids were used,whereas 1,704 recombinants were observed in mixed infectionsof tissue culture cells. Therefore, less than 0.06% of the recombinantscould have resulted from gene transfer that occurred after bacteriawere plated on selection media.

Intracellular recombinants accumulated steadily over time. Intracellular Salmonella resides inside vacuoles (17), and ithas been shown that they reside inside the vacuoles of culturedcells for at least the length of time relevant to our experiments(16). We therefore predicted that for conjugation to occur,either intracellular donor- and recipient-containing vacuolesmust fuse or a small number of bacteria (less than the numberwhich could be detected by microscopy [e.g., reference 16])must be released into the cytoplasm.

Salmonella-containing vacuoles reportedly coalesce 12 h afterinfection, and bacteria have been observed inside one largeperinuclear vacuole (16, 18). If vacuole fusion were requiredfor intracellular conjugation, intracellular recombinants mightnot be detected before 12 h after donor internalization. However,we observed intracellular recombinants by 3 h after donor invasion(Fig. 1).

View larger version (20K):
[in this window]
[in a new window]
FIG. 1. Accumulation of intracellular transconjugants over time. Intracellular recipients (SL1344^N) ( ${blacktriangleup}$ ), donors [SL1344^R(RP4)] (•), and transconjugants ( ${diamondsuit}$ ) were enumerated over time beginning 6 h after recipient invasion began (or 2.75 h after donor invasion began). The number of transconjugants formed after plating ( ${diamond}$ ) was not significantly different at the first time point. Each value is an average based on three independent experiments performed in triplicate. Error bars indicate standard errors.

The number of intracellular recombinants per limiting parentincreased 2 orders of magnitude over 6 h, from our limit ofdetection of $~$ 10^-6 recombinant per limiting parent to an averageof 7 x 10^-4 ± 2 x 10^-4 recombinant per limiting parent(Fig. 1). This time-dependent increase in recombinant frequencywas not due to faster growth of newly formed recombinants thanof the parent bacteria. In reconstruction experiments, cellsinfected by newly recombinant bacteria did not accumulate intracellularrecombinant bacteria any faster than they accumulate parentalbacteria (data not shown). Thus, the increase in recombinantfrequency was likely due to accumulation of new recombinants.

Gene exchange did not occur extracellularly. Some proportion of extracellular recipient bacteria may haveescaped killing by gentamicin or may have emerged from cellsand mated with extracellular donors during the time that donorbacteria were infecting cells. If this were the source of therecombinants, then the recombinants should have been createdeven if the donor bacteria could not invade. Therefore, theintracellular gene transfer experiment with invasive donors(see above) was repeated using the noninvasive donors ${Delta}$ invA SL1344^Rand ${Delta}$ invA 14028^R-P, which are compromised for invasion by TnphoAinsertions in the invA gene (21). InvA is an inner membraneprotein that may form part of the type III secretion system(22). Although invA mutants can bind to the cell surface, theyinvade at least 100 times less frequently than the wild type(21).

Recombinants were rarely and inconsistently detected in theintracellular mating assay when donors were noninvasive (Table4), even though invA and wild-type donors transmit RP4 and Fat frequencies comparable to those of recipient bacteria mixedon LB agar plates (Table 2).

View this table:
[in this window]
[in a new window]
TABLE 4. Intracellular plasmid transmission requires donors to be invasive^a

No recombinants containing RP4 were recovered when invA mutantdonors were used, whereas a small number of recombinants containingF were recovered. The difference reflects the fact that theinvA mutation has different effects on the invasion efficienciesof SL1344 and 14028-P. On average, the number of intracellular ${Delta}$ invA 14028^R-P(F42::mini-Tn10Kn) bacteria recovered was 3% ofthe number of inv⁺ bacteria recovered, whereas the number ofintracellular ${Delta}$ invA SL1344^R(RP4) bacteria recovered was 0.1%of the number of inv⁺ bacteria recovered. It appears that athreshold of 10⁴ intracellular donors is required for detectionof the formation of recombinants by this assay.

Mechanism of plasmid transmission was conjugation. It was still possible that plasmids that were released fromdead internalized donors transformed competent internalizedrecipients. Transformation was an unlikely mechanism of transmissionbecause Salmonella is not naturally competent for DNA uptake.Nevertheless, vacuoles might induce competence in unknown ways.

If the mechanism of transmission is conjugation, it should besensitive to mutations in the tra (transfer) genes of the plasmids.To distinguish decisively between conjugation and transformationas the mechanism of RP4 transmission, a ${Delta}$ traJ derivative of RP4(57) and a traB derivative of F (31) were introduced into thedonor strains. TraJ forms part of the relaxosome complex requiredfor nicking and unwinding of single-stranded plasmid DNA (24).traB is necessary for conjugative pilus formation (35). ThetraJ deletion significantly reduced plasmid transmission onagar plates (Table 2). Likewise, traJ was necessary for theformation of intracellular recombinants (Table 3). Similarly,F plasmids with mutations in traB produced no intracellularrecombinants (Table 3).

In addition, if transformation contributes to the formationof recombinants, the effect should be enhanced with smallerplasmids. However, no recombinants were detected when donorbacteria carried either the 4.4-kb plasmid pBR322 or the 10.5-kbplasmid pSUP104 (49) (data not shown). Conjugation, not transformation,was therefore probably the mechanism of intracellular plasmidtransmission measured in these experiments.

Transconjugants did not form from conjugation on the well surfaces beneath the monolayers. Since Salmonella has been shown to traverse cell monolayers(18, 19), it was possible that conjugation occurred on the wellsurfaces beneath the cells in pockets not penetrated by gentamicin-containingmedium. To control for this possibility, we performed two experiments.In the first experiment, tissue culture cells were seeded ontopermeable membrane supports (see Materials and Methods) insertedinside wells. This allowed both sides of the monolayer to bebathed in gentamicin, thus exposing any potential bacterialescapees. Transconjugants were recovered at a frequency indistinguishablefrom that observed when cells were grown on a plastic surface(Table 5).

View this table:
[in this window]
[in a new window]
TABLE 5. Plasmid transmission frequencies in adherent and nonadherent cells^a

Many fewer transconjugants were recovered in experiments inwhich membranes were used. Since the surface area of an insertmembrane was approximately one-half that of a well, the bacteriareleased from cells grown on two inserts were pooled to makeone replicate. Despite this, the recovery of intracellular bacteriawas poor, a result of both poor adhesion of cells to the membraneand reduced bacterial invasion efficiency. In a second experiment,cells were released from the monolayer by treatment with trypsin(0.25% trypsin, 1 mM EDTA; Gibco) after the second gentamicin-bathingperiod and reseeded into wells as a cell suspension in gentamicin-containingmedium. Cells did not readhere during this experiment. Again,transconjugants were detected at a frequency indistinguishablefrom that observed with undisturbed monolayers (Table 5).

As a further control for extracellular mating, cells were infectedwith either recipient or donor bacteria and then were releasedfrom the monolayer with trypsin. Infected cells were then mixedand seeded into a clean well. Transconjugants were detectedonly at the frequency at which they formed after plating ofthe lysate (Table 5). Together, these experiments confirmedthat transconjugants were not formed by extracellular conjugationbetween emergent bacteria and suggested that donors and recipientsmust concurrently infect the same cell in order for transconjugantsto form.

Frequency of transconjugants decreased with decreasing likelihood of coinfection. For intracellular conjugation to occur, donor and recipientbacteria must make contact inside cells. Consistent with thisfact, the frequency of transconjugants was found to decreaseas the ratio of invading bacteria to cells (the MOI) decreased,which decreased the probability of a donor and a recipient occupyingthe same cell at the same time (Fig. 2A).

View larger version (17K):
[in this window]
[in a new window]
FIG. 2. Intracellular conjugation frequency increases with the probability of coinfection. (A) Transconjugant frequency measured as a function of MOI. (B) Invasion of tissue culture cells by recipients (•) and donors ( ${circ}$ ) measured as a function of MOI. Each value is an average based on three independent experiments performed in triplicate.

Plasmid transmission was measured in experiments conducted atMOIs ranging from 0.1 to 100. In each experiment, the same numberof donors as recipients was introduced into a cell culture.Thus, the ratio of the total number of bacteria to tissue culturecells was always twice the MOI reported. The number of recipientsinternalized after 3 h was determined by lysis and plating ofbacteria from replicate wells immediately after the gentamicin-bathingperiod. The initial number of intracellular recipients increasedas the MOI increased. MOIs greater than 10 had no additionaleffect on the number of internalized bacteria (Fig. 2B).

The plateau observed in the number of internalized bacteriais consistent with previously reported observations that Salmonellainvasion kinetics reach saturation at MOIs greater than 40.Saturation is attributed to the presence of a limited numberof binding sites on the cell surface (33, 40). The initial numberof donors that invaded cells that had been preinfected withrecipients at each MOI was therefore also determined. As expected,fewer donors invaded cells that had been preinfected with ahigh number of recipients (Fig. 2B). Again, this was likelydue to internalization of binding sites during the first infection(33).

The frequency of transconjugants (Fig. 2A) was proportionalto the frequency of coinfection (Fig. 2B). The transconjugantfrequency increased from 3 x 10^-5 ± 2 x 10^-5 transconjugantper limiting parent (where detectable) at an MOI of 0.1 to 6x 10^-4 ± 1 x 10^-4 transconjugant per limiting parentat an MOI of 10. The peak frequency at an MOI of 10 correspondedto the highest number of both donors and recipients internalized.Therefore, the intracellular conjugation frequency was dependenton the probability of coinfection. The frequency of transconjugantswas lower at MOIs greater than 10 for unknown reasons, but thismay have been because cells infected by large numbers of bacteriabecame more permeable to gentamicin or were otherwise lost dueto the cytotoxicity associated with a heavy infection.

DISCUSSION

Top

Discussion

Salmonella transferred plasmids within cultured INT-407 humancells. Mating was detected at frequencies greater than 10^-4per limiting parent for RP4 and F. This frequency is only 3orders of magnitude less than that detected on agar plates containingmuch higher concentrations of bacteria. The INT-407 cell linethat we used was not a special case because transconjugantsalso formed in dog kidney (MDCK) and human colon cancer (CaCo-2)cells (data not shown). These observations support suggestionsthat antibiotic resistance genes can and do transfer withinhumans and animals (50, 58).

Gene transmission between bacteria occurs inside animal cellsby conjugation. In this study, the requirement that donors beinvasive, the requirement that both parental types occupy thesame cell at the same time, and the requirement that transconjugantsremain inside the cell at all times prior to harvest showedthat gene transmission was intracellular.

The dependence on tra genes for the formation of recombinantsdemonstrated that genes were transmitted by conjugation andnot by transformation. The detection limits of these experimentsranged from $~$ 10^-5 to $~$ 10^-7 recombinant per limiting parent (Table3). This range can be attributed to the variation in internalizationof bacteria from experiment to experiment. In some replicates,fewer than the average number of intracellular bacteria wererecovered, reducing the sensitivity of the experiment. However,in three replicate experiments performed in triplicate, no recombinantcolonies were recovered when donors carried traJ or traB plasmids,compared with the 1,893 and 1,704 recombinant colonies obtainedwith Tra⁺ RP4 and F, respectively.

What is the frequency of intracellular gene transmission betweeninvasive pathogenic bacteria? The overall frequency of transmission, $~$ 10^-4 transconjugant per limiting parent, is the result of (i)how often donors and recipients infect the same cell, (ii) howoften the bacteria could form contacts within the cells, and(iii) the frequency of plasmid transmission between bacteriain contact with each other. Perhaps dual infection of singlehuman cells by donor and recipient bacteria is common and thefrequency is a true reflection of gene transmission efficiency.Alternatively, conjugation within human cells may be as efficientas conjugation on agar plates, but donors and recipients areless often in direct contact because coinfection of human cellsis infrequent or contact between bacteria inside human cellsis rare. Experiments are under way to distinguish between thesetwo explanations. Preliminary microscopy results suggest thatthe proportion of cells that are infected after the first roundof invasion is high (approximately 80% on average) and thatthe cells tend to be infected equally with a few bacteria ( $~$ 1to 20 bacteria) as opposed to a small subpopulation of cellsthat are infected with many bacteria. We are in the processof determining the frequency of coinfection.

Our experiments required donor and recipient bacteria that wereinitially in separate vacuoles to somehow meet despite thatfact that Salmonella is known to remain inside vacuoles forextended periods. The fact that the mechanism of gene transferwas found to be conjugation rather than transformation suggeststhat the bacteria did meet, because conjugation requires bacterium-bacteriumcontact. We propose two models to explain how the bacteria findone another. In the first model, we imagine that a small proportionof intracellular bacteria of each parental type escape the vacuole.Transconjugants are subsequently formed by conjugation in thecytoplasm. In the second model, we imagine that transconjugantsare formed after fusion of donor- and recipient-containing vacuoles,which is known to occur within 12 to 24 h (16, 18).

To our knowledge, there is no evidence that Salmonella escapesfrom vacuoles. If conjugation occurs outside vacuoles, thenit has to involve a proportion of intracellular bacteria smallerthan the proportion which could be detected by the experimentsof other workers designed to determine the intracellular locationof invasive Salmonella. The vacuole fusion model is supportedonly if vacuole coalescence occurs, at least at some level,before 12 h. To our knowledge, the time course of vacuole coalescencehas not been investigated. We are currently monitoring the intracellularlocation of transconjugants using donor and recipient bacteriamarked with fluorescent proteins of different colors.

Of what relevance are these observations to plasmid and bacterial evolution? Bacterial conjugation is probably the primary mechanism of antibioticresistance gene transmission (27, 46). The ubiquity of plasmid-borneresistance may be explained in part by the stamina of conjugationitself (27). Conjugation can occur even in environments thatotherwise kill a bacterium, allowing plasmids to replicate byhorizontal gene transfer in the presence of antibiotics andother environmental toxins that prevent bacterial reproduction(11, 27, 28).

HMEs carry a range of genes in addition to resistance determinants,including genes that accentuate virulence and symbiosis potentialin many different microbes (29), from soil microbes (2) to membersof the human flora (3, 9, 34, 38). What is the mechanism thatcreates linkage among various catabolic pathways, virulence,and resistance genes (29)? Here we considered the possibilitythat in at least one environment, gene transfer concentratesvirulence and resistance determinants, providing an opportunityfor the evolution of genetic linkage. When the invasive formof Salmonella is inside human cells, it can exchange plasmidswith other strains. Since virulence is partially dependent uponinvasion, the most virulent bacteria are probably inside cellsmore often than less virulent bacteria are. The most virulentpathogens are also more likely to attract the attention of antibiotic-dispensingclinicians who effectively remove other microbes from the extracellularniche and concentrate resistant microbes in the patient (61).Mammalian cells infected by Salmonella can subsequently internalizeother species of bacteria (20), which must be antibiotic resistantin patients treated with antibiotics. These other species maythen transfer antibiotic resistance genes to Salmonella.

Furthermore, it is known that antibiotic treatment causes overgrowthof resistant microbes, increasing translocation of these organismsfrom the gut to the mesenteric lymph nodes (7). Translocationmost likely occurs by intracellular passage of bacteria throughthe intestinal epithelium (6). Thus, antibiotic usage may increasethe potential for intracellular gene exchange, giving the pathogenand less virulent strains the opportunity to acquire the genesthat confer antibiotic resistance and virulence characteristics.The phenomenon of in situ mating provides an evolutionary mechanismfor the coevolution of resistance and novel virulence traits.

Tissue culture experiments have established the plausibilityof gene transmission in humans and animals, particularly insidetheir cells, but have not established that the process is relevantto evolution. The frequency of coinfection, intracellular contactbetween donors and recipients, and expression of conjugativefunctions may be different in the gut environment. The resultsof tissue culture experiments do, however, directly supportthe possibility that gene transmission occurs during antibiotictherapy.

HGT is a surprisingly robust and common phenomenon, defyingthe most concerted efforts to control microbial diseases (27,29, 30). Understanding the evolution of gene transfer shouldbe as relevant to developing a lasting strategy for controllinginfectious diseases as mining the genome for new drug targetsor restricting the application of antibiotics.

ACKNOWLEDGMENTS

This work was supported by the Marsden Fund (grant M1042 toJ.A.H.) and by a UC Roper Ph.D. Scholarship (to G.C.F.). M.A.K.is an HRC Senior Research Fellow.

We thank A. Miller for her assistance with cell culture, B.B. Finlay, A. J. Bäumler, and B. M. M. Ahmer for Salmonellastrains, and D. Thaler, H. Shuman, H. Hilbi, and I. Molineuxfor critical reviews of the manuscript.

FOOTNOTES

* Corresponding author. Mailing address: Department of Plant and Microbial Sciences, University of Canterbury, Private Bag 4800, Christchurch, New Zealand. Phone: 643 364 2926. Fax: 643 364 2083. E-mail: j.heinemann{at}pams.canterbury.ac.nz.

REFERENCES

Top