Escherichia coli interactions with Acanthamoeba: a symbiosis with environmental and clinical implications

Acanthamoeba is a free-living protozoan pathogen that is widely distributed in the environment, including air, soil, tap water and swimming pools (Khan, 2003; Marciano-Cabral & Cabral, 2003; Schuster & Visvesvara, 2004). Acanthamoeba can cause Acanthamoeba keratitis (AK), a painful sight-threatening infection, primarily associated with contact-lens usage, and a fatal granulomatous amoebic encephalitis (GAE), mostly limited to immunocompromised patients. During the last two decades there has been an upsurge in Acanthamoeba infections, due to an increase in the number of people wearing contact lenses, and a rise in the number of immunocompromised patients. In addition to its direct role in causing human infections, it is now well established that Acanthamoeba acts as a host for various bacterial pathogens, including Legionella pneumophila (the causative agent of Legionnaires' disease) (Rowbotham, 1980), Coxiella burnetii (Q fever) (La Scola & Raoult, 2001), Pseudomonas aeruginosa (keratitis) (Michel et al., 1995), Vibrio cholerae (cholera) (Thom et al., 1992), Helicobacter pylori (gastric ulcers) (Winiecka-Krusnell et al., 2002), Listeria monocytogenes (listeriosis) (Ly & Muller, 1990) and Mycobacterium avium (respiratory infections) (Krishna-Prasad & Gupta, 1978; Steinert et al., 1998), and may act as a vector to transmit these pathogens to susceptible hosts.

Among Gram-negative bacteria, Escherichia coli is the leading cause of neonatal meningitis. At least 80 % of E. coli strains that cause meningitis possess the capsular polysaccharide antigen K1 capsule (Xie et al., 2004). The pathogenesis of E. coli K1 meningitis involves bacterial entry into the bloodstream and the development of bacteraemia, followed by the crossing of the bloodbrain barrier to produce disease (Wang et al., 2004). A high level of bacteraemia (>10³ c.f.u. bacteria per millilitre of blood) is a prerequisite of E. coli K1 meningitis (Xie et al., 2004). However, we are only beginning to understand the mechanisms associated with E. coli K1 survival and multiplication in the bloodstream, E. coli survival of the macrophage onslaught, and crossing of the bloodbrain barrier. Here, we studied Acanthamoeba interactions with the invasive E. coli K1 and the non-invasive E. coli K-12 strains, and identified the bacterial determinants responsible for K1 interactions with amoebae.

Culture of Acanthamoeba. All chemicals were purchased from Sigma, unless otherwise stated. A clinical isolate of A. castellanii belonging to the T4 genotype, isolated from a keratitis patient, was used in the study. The amoebae were grown without shaking in 15 ml PYG medium (0.75 %, w/v, proteose peptone; 0.75 %, w/v, yeast extract; 1.5 %, w/v, glucose) in T-75 tissue-culture flasks at 30 °C, as previously described (Khan, 2001), the medium being refreshed 1720 h prior to all experimentation. This resulted in more than 95 % of the amoebae in the trophozoite form.

E. coli strains and growth conditions. E. coli K1, used in the present study, is a rifampicin-resistant mutant of strain RS218 (serotype O18 : K1 : H7). This strain is a clinical isolate from the cerebrospinal fluid of a neonate with meningitis. A non-invasive E. coli K-12 laboratory strain, HB101, was used as a non-pathogen. In addition, we used several isogenic mutants of K1, including ΔfimH, constructed by deleting the entire fimH gene and replacing it with an antibiotic-resistance cassette using standard molecular methods (Teng et al., 2005). FimH is a 29 kDa protein and is expressed on the tip of bacterial fimbriae. Other mutants included an outer-membrane protein A (OmpA) mutant (ΔompA) (OmpA is a 35 kDa protein expressed in the outer membrane of E. coli; Prasadarao et al., 1999), and a cytotoxic necrotizing factor-1 (CNF1) mutant (Δcnf1) (CNF1 is a 110 kDa AB type bacterial toxin; Khan et al., 2002). In addition, a rough LPS mutant, constructed using chemical mutagenesis, was used (Kim et al., 1992). For simplicity, the rough LPS mutant is referred to as ΔLPS, even though there were no genetic manipulations. All bacteria were grown in LuriaBertani (LB) broth overnight with appropriate antibiotics: kanamycin (40 µg ml¹) or chloramphenicol (25 µg ml¹).

E. coli association assays. To study E. coli interactions with A. castellanii, association assays were performed (Fig. 1). Briefly, A. castellanii was grown in 24-well plates in PYG medium (5x10⁵ amoebae ml¹ per well) until confluent. The cells were washed once with PBS. Next, E. coli strains [2x10⁶ c.f.u. per well (per 0.5 millilitre PBS)] were added, and the plates incubated for 1 h at room temperature. Following this incubation, amoebae were washed with PBS three times to remove non-adherent bacteria, and counted using a haemocytometer. Finally, amoebae were lysed by adding SDS, 0.5 % final concentration, to each well for 20 min, and the number of bacteria was enumerated by plating on nutrient agar plates. The bacteria associated with A. castellanii were calculated as follows: number of bacteria/number of amoebaex100=percentage bacteria associated with A. castellanii.

(29K): Fig. 1. Stages in E. coli association assays, E. coli invasion assays and E. coli intracellular assays.

E. coli invasion assays. To determine the ability of bacteria to invade or be taken up by A. castellanii, invasion assays were performed (Fig. 1). Briefly, amoebae were grown until confluent in 24-well plates followed by the addition of 2x10⁶ E. coli cells, as described above. After 1 h incubation, the wells were washed three times with PBS, followed by the addition of gentamicin (100 µg ml¹ final concentration, for 45 min) to kill extracellular bacteria. Finally, amoebae were counted and the intracellular bacteria enumerated as described above. The intracellular bacteria were calculated as follows: number of bacteria/number of amoebaex100=percentage intracellular bacteria.

Intracellular survival assays. To determine the long-term effects of A. castellanii and E. coli interactions, intracellular survival assays were performed (Fig. 1). Briefly, amoebae were incubated with E. coli, followed by the addition of gentamicin for 45 min. After incubation, wells were washed three times with PBS and subsequently incubated in 0.5 ml PBS for 24 h at 30 °C. Finally, amoebae and E. coli were enumerated as described above, and intracellular bacteria after 24 h incubations were calculated as follows: number of bacteria/number of amoebaex100=percentage bacteria after 24 h. To determine the effects of environmental conditions on A. castellanii and the intracellular bacteria, survival assays were performed in the presence of PYG, instead of PBS, for 24 h.

Invasive E. coli K1 exhibits higher association and increased invasion/uptake by A. castellanii than non-invasive E. coli K-12
To determine the ability of E. coli to interact with A. castellanii, association assays were performed. Our findings revealed that the non-invasive K-12 strain exhibited significantly less association with A. castellanii than the invasive K1 strain (47±14 and 122±2.2 %, respectively) (P <0.05). Here, the term association is used to describe both E. coli that were inside the amoebae and those that were attached to the amoebae. Next, to determine the numbers of intracellular E. coli, invasion assays were performed. We observed a higher recovery (0.5±0.01 %) of intracellular E. coli K1 than E. coli K-12 (0 %).

Invasive E. coli K1 survive intracellularly in A. castellanii, while non-invasive E. coli K-12 are killed
To determine the fate of E. coli in long-term interactions with A. castellanii, intracellular survival assays were performed by incubating E. coli with A. castellanii in PBS for 24 h. Our findings revealed that once inside the cell, E. coli K1 remained viable and multiplied, while K-12 were killed (2.8±0.46 and 0 %, respectively). It is important to emphasize that A. castellanii remained intact for this period of incubation. To determine the effects of favourable environmental conditions, E. coli were incubated with A. castellanii in PYG medium for 24 h, instead of PBS. Under these conditions, E. coli K1 lysed the amoebae and grew exponentially, whereas E. coli K-12 exhibited minimal growth (16 785±988 and 12±1.4 %, respectively). In addition, E. coli K-12 had no effect on Acanthamoeba viability as determined by the trypan blue exclusion test (data not shown).

OmpA and LPS are important determinants required for E. coli K1 association with amoebae
To identify the bacterial determinants responsible for E. coli K1 association with A. castellanii, several mutants lacking known virulence determinants were used in tandem with the invasive E. coli K1. Firstly, we studied the role of FimH, which is expressed at the tip of fimbriae, owing to its importance as an initial point of contact with the host cells. An isogenic fimH deletion mutant derived from E. coli K1 was used. Both the wild-type and the ΔfimH mutant exhibited similar levels of association with A. castellanii (122 and 152.6 %, respectively) (Fig. 2).

(6K): Fig. 2. OmpA and LPS are crucial determinants for E. coli K1 binding/association with A. castellanii. To determine the bacterial determinants responsible for E. coli K1 interactions with A. castellanii, several mutants lacking important virulence determinants were used. Briefly, amoebae were grown in 24-well plates in PYG medium until confluent. The cells were washed and incubated with the invasive E. coli in PBS for 1 h, and amoebae and E. coli enumerated as described in Methods. Note that both the rough LPS mutant and the ompA deletion mutant exhibited significantly decreased binding/association with A. castellanii compared to wild-type E. coli K1. Results are the mean of three independent experiments performed in duplicate. Error bars represent standard error.

Next, we determined the role of OmpA, which is one of the proteins embedded in the external membrane of E. coli. OmpA resembles one of the Neisseria proteins, Opa, that is implicated in the invasion of epithelial cells. Using an isogenic ompA deletion mutant (ΔompA), the results revealed that OmpA is a crucial determinant for E. coli K1 association with A. castellanii (Fig. 2). The ΔompA mutant exhibited significantly decreased levels of association compared with the non-invasive K-12. Similarly, the E. coli K1 strain expressing rough LPS exhibited significantly less association with A. castellanii than its parent strain, E. coli K1 (P <0.05) (Fig. 2). Interestingly, CNF1 (a dermanecrotic protein toxin) exhibited no effects on E. coli K1 association with A. castellanii. This was shown by the demonstration that an isogenic cnf1 deletion mutant, Δcnf1, exhibited amoebae association similar to its parent strain, E. coli K1 (Fig. 2).

OmpA and LPS, but not FimH and CNF1, are important determinants required for E. coli K1 invasion of and/or uptake by A. castellanii
To study the effect of the aforementioned factors on bacterial invasion of and/or uptake by A. castellanii, invasion assays were performed. The results demonstrated that the ΔfimH and Δcnf1 mutants exhibited similar levels of invasion/uptake to those of the wild-type E. coli K1 (0.46, 0.51 and 0.49 %, respectively) (Fig. 3). In contrast, the ΔompA and rough LPS mutants exhibited significantly reduced levels of invasion/uptake by A. castellanii (P <0.05) (Fig. 3).

(5K): Fig. 3. OmpA and LPS, but not FimH and CNF1, are important determinants required for E. coli K1 invasion of and/or uptake by A. castellanii. To determine the bacterial determinant responsible for E. coli K1 invasion/uptake in A. castellanii, several mutants lacking important virulence determinants were used. Briefly, amoebae were incubated with E. coli for 1 h, as described above. After this incubation, gentamicin was added to kill the extracellular E. coli, followed by amoebae and E. coli counting as described in Methods. Note that both the rough LPS mutant and the ompA deletion mutant exhibited significantly decreased invasion/uptake in A. castellanii compared to wild-type E. coli K1. Results are the mean of three independent experiments performed in duplicate. Error bars represent standard error.

OmpA is a crucial determinant required for E. coli K1 survival inside A. castellanii
To identify the bacterial determinants required for intracellular survival of A. castellanii, intracellular survival assays were performed as described in Methods. Both E. coli and A. castellanii were incubated for 24 h in a non-nutrient environment, PBS. The results showed that all mutants tested exhibited a decreased ability to survive intracellularly compared with the parent K1 strain (Fig. 4a). Both the rough LPS mutant and the cnf1 deletion mutant showed some survival ability, but significant decreases in their ability to survive within A. castellanii were observed (Fig. 4a). The ompA deletion resulted in the loss of the ability of E. coli K1 to survive inside A. castellanii. To determine the effects of favourable conditions on E. coli K1 interactions with A. castellanii, assays were performed in the presence of PYG medium, instead of PBS. Again, the ΔompA mutant exhibited a limited ability to survive intracellularly, even under favourable conditions, i.e. PYG medium (Fig. 4b). Overall, these findings indicate that OmpA and LPS are critical determinants for bacterial binding, invasion, and/or uptake and survival inside A. castellanii.

(12K): Fig. 4. OmpA is a crucial determinant required for E. coli K1 survival inside A. castellanii. (A) To identify the bacterial determinants required for intracellular survival of A. castellanii in the absence of nutrients, assays were performed using several isogenic deletion mutants derived from E. coli K1. Briefly, amoebae were incubated with E. coli for 1 h, followed by the addition of gentamicin to kill the extracellular bacteria. Finally, A. castellanii plus intracellular E. coli were incubated in PBS (A) or PYG (B) for 24 h. Note that in the presence of PBS, the ompA deletion resulted in the loss of the ability of E. coli K1 to survive inside A. castellanii. Results are the mean of three independent experiments performed in duplicate. Error bars represent standard error.

It is well known that Acanthamoeba acts as a host for several bacteria, and its ability to host bacterial pathogens, such as L. pneumophila, C. burnetii, P. aeruginosa, V. cholerae, H. pylori, List. monocytogenes and M. avium, has gained particular attention. This is a consequence of the finding that Acanthamoeba may host bacterial pathogens under harsh environmental conditions, and may help transmit bacterial pathogens to susceptible hosts, thus offering a potential route of entry into the human body. However, the precise mechanisms associated with bacteriaamoebae interactions remain incompletely understood. Further complexity is added by the fact that amoebae feed on bacteria. So why bacteria interact with amoebae and how pathogenic bacteria maintain survival inside amoebae, while non-pathogens are killed, is unclear. Here, we studied E. coli interactions with A. castellanii. Using clinical and non-clinical isolates of E. coli, we demonstrated that the outcome of E. coli interactions with A. castellanii varies, depending on the virulence of E. coli. The invasive E. coli K1 has the ability to gain entry to amoebae and remain viable, while the non-invasive E. coli K-12 is killed.

It is interesting to observe that both K1 and K-12 exhibited binding-association with amoebae, 122 and 47 %, respectively (although K-12 showed reduced levels of association). This is in contrast to previous findings, which have shown that E. coli K1 and K-12 exhibit more than 10-fold differences in binding to other cell types, such as human brain microvascular endothelial cells, and have identified FimH as an important adhesion determinant (Prasadarao et al., 1999; Teng et al., 2005). However, our findings are not surprising. The fact that amoebae feed on bacteria suggests that the initial interactions between E. coli and Acanthamoeba may not be totally dependent on the virulence properties of E. coli. In support, we observed that a fimH deletion mutant exhibited similar levels of association with amoebae to those of the parent strain. However, it is the post-invasion events that vary greatly between invasive and non-invasive E. coli strains in their interactions with A. castellanii. In this regard, we observed that under harsh conditions, the invasive K1 isolate remains intracellular and viable but does not kill the host, i.e. A. castellanii. However, under favourable conditions (presence of nutrients), the invasive K1 isolate grows exponentially and lyses the host cells. In addition, we identified OmpA as a critical bacterial determinant required for E. coli K1 association with, invasion/uptake by, and intracellular survival within A. castellanii.

One intriguing aspect is that these findings show remarkable similarities to E. coli K1 interactions with human macrophages. For example, recent studies have shown that E. coli K1, but not K-12, is able to enter murine and human macrophages, and survive and replicate intracellularly (Sukumaran et al., 2003), and this property may be crucial for E. coli survival in the bloodstream, a primary step in the development of meningitis. Moreover, these properties of E. coli K1 have been directly attributed to the expression of OmpA protein. The fact that Acanthamoeba resembles human macrophages in many ways, particularly in its phagocytic activity and cell surface receptors (Yan et al., 2004), and that macrophages and Acanthamoeba exhibit parallel mechanisms in their interactions with E. coli K1, suggests that Acanthamoeba may provide an alternative model to study E. coli pathogenesis and to understand its immune evasion mechanisms. The implication of this observation remains unclear, but will be the subject of further studies.

Although we observed clear differences between invasive and non-invasive E. coli strains in their ability to survive intracellularly in A. castellanii, the precise mechanisms of E. coli K1 intracellular survival remain unknown. Interestingly, previous studies of the interactions of L. pneumophila with A. castellanii have demonstrated the ability of L. pneumophila to inhibit the fusion of lysosomes with phagosomes as a critical step in the intracellular survival of this bacterium (Bozue & Johnson, 1996). E. coli K1 may use similar mechanisms to evade the host-cell defences; however, this remains to be determined. Overall, our findings suggest that the interactions of E. coli and A. castellanii are highly complex and depend on the virulence of E. coli. Acanthamoeba may act as a bacterial predator, or as a reservoir or Trojan horse for bacteria, with environmental and clinical implications.

This work was supported by the Korean Research Foundation Grant funded by the Korean Government Ministry Of Education and Human Resources Development (MOEHRD) (KRF-2005-214-E00040), and partially supported by grants from the Faculty Research Grant, University of London, and the Royal Society.