Development of multiple strain competitive index assays for Listeria monocytogenes using pIMC; a new site-specific integrative vector

Background The foodborne, gram-positive pathogen, Listeria monocytogenes, is capable of causing lethal infections in compromised individuals. In the post genomic era of L. monocytogenes research, techniques are required to identify and validate genes involved in the pathogenicity and environmental biology of the organism. The aim here was to develop a widely applicable method to tag L. monocytogenes strains, with a particular emphasis on the development of multiple strain competitive index assays. Results We have constructed a new site-specific integrative vector, pIMC, based on pPL2, for the selection of L. monocytogenes from complex samples. The pIMC vector was further modified through the incorporation of IPTG inducible markers (antibiotic and phenotypic) to produce a suite of four vectors which allowed the discrimination of multiple strains from a single sample. We were able to perform murine infection studies with up to four EGDe isolates within a single mouse and showed that the tags did not impact upon growth rate or virulence. The system also allowed the identification of subtle differences in virulence between strains of L. monocytogenes commonly used in laboratory studies. Conclusion This study has developed a competitive index assay that can be broadly applied to all L. monocytogenes strains. Improved statistical robustness of the data was observed, resulting in fewer mice being required for virulence assays. The competitive index assays provide a powerful method to analyse the virulence or fitness of L. monocytogenes in complex biological samples.


Background
Historically, the measurement of bacterial virulence has involved the use of infection models to access the ability of a pathogen to cause disease. One widely used method is the LD 50 assay, which is defined as the number of bacteria required to kill 50% of the infected hosts. This method yields valuable data pertaining to the cumulative, absolute virulence of the bacterium, which can theoretically be compared between laboratories. But as detailed by Beuzón and Holden [1], it is a crude method as it does not relay information on the infection kinetics. Also, if the deletion of a gene does not increase the LD 50 , it does not necessarily mean that the gene product does not play a role in the virulence of the bacterium. If two or more strains (e.g. wild type to mutant or independent isolates) could be compared within a single host, the kinetics of infection could be monitored, which could expose subtle differences. Intra-animal experiments would help to minimise inherent inter-animal biological variation and also improve the identification of mutations or isolates with reduced competitive fitness within the host [2]. This form of assay has been termed a "competitive index" and is becoming an increasing popular method to examine bacterial virulence [3][4][5]. One crucial factor is to establish the ability to discriminate strains, without adversely impacting on the natural fitness of the organism.
Listeria monocytogenes is a foodborne pathogen, which can cause fatal infections in a susceptible host [6]. In humans, the initial step of gastrointestinal invasion is mediated via the interaction of a bacterial cell surface exposed protein, Internalin A (InlA), with the host cell surface ligand, Ecadherin [7]. Subsequent intracellular replication and spread can lead to systemic disease. While the mouse is a poor model of listerial gastrointestinal disease due to the limited affinity of InlA for murine E-cadherin, the tools required for systemic proliferation are functional once the gastrointestinal tract is bypassed via intravenous or intraperitoneal injection [8]. Identification and assessment of genes required for the virulence of L. monocytogenes has been greatly helped by the sequencing of multiple genomes [9,10]; however, functional post genomic analysis requires the development of improved techniques for the discrimination of virulence potential. LD 50 measurements are still widely used, but for the reasons described above, competitive indices could provide a more elegant alternative. Even though competitive index assays have been previously applied to L. monocytogenes [11][12][13], these are not broadly applicable and have inherent limitations (see Discussion section).
Here, we have established a method to stably tag L. monocytogenes strains and assess their competitive fitness in complex samples. Through the development of a new sitespecific integrative vector, pIMC, we have produced four derivatives with different IPTG inducible antibiotic and phenotypic markers for their subsequent discrimination from complex samples.

Construction of pIMC
We have modified the Listeria site-specific integrative vector, pPL2 [11], by a series of SOE-PCR reactions to generate pIMC (Fig 1a). The backbone of pPL2 was conserved, including the pBluescriptII KS multiple cloning site, low copy E. coli origin of replication, conjugative transfer functions, and the listerial phage integrase genes. To improve antibiotic selection and decrease vector size, the gramnegative cat gene was deleted and the inducible gram-positive cat promoter [12] replaced with the highly expressed listerial promoter (Phelp) [13]. As shown in Fig 1b, L. monocytogenes strain EGDe, when transformed with pPL2, exhibited variable colony size and took 48 h to yield col-Plasmid maps for pIMC and IPTG inducible marker expres-sion in pIMC3 Figure 1 Plasmid maps for pIMC and IPTG inducible marker expression in pIMC3. The plasmid (A) pIMC was created by SOE PCR (see Material and Methods) from oligonucleotides described in Table 1. The backbone of pIMC, derived from pPL2 [11] encodes the p15A low copy E. coli origin of replication and RP4 conjugative origin of transfer. No grampositive origin of replication is present on the plasmid, therefore upon transformation, chromosomal integration into L. monocytogenes tRNA ARG is directed by the Listeriophage PSA integrase. Antibiotic selection is supplied by the chloramphenicol acetytransferase (cat) fused to the highly expressed listerial promoter (Phelp) [13]. Restriction sites labelled on pIMC are unique and for sequencing purposes, T3 and T7 primer binding sites are present before the KpnI and after the SacI restriction sites, respectively. The plasmid sequence is accessible under the EMBL nucleotide accession number AM940001. A comparison of chloramphenicol selection (7.5 μg/ml) is shown in (B), with pIMC exhibiting uniform colony size compared to pPL2 transformed EGDe. (C) Antibiotic markers (kanamycin (aphA3), erythromycin (ermAM) and tetracycline (tetM)) and the beta-glucuronidase marker (gusA) were subcloned from pIMK3 into pIMCa as a SacI/PstI fragment (see Materials and Methods).
onies. Even upon passage in the presence of chloramphenicol, EGDe::pPL2 yielded a variable colony size and could not grow at chloramphenicol concentrations above 7.5 μg/ml (data not shown). These results suggested that the inducible gram-positive cat gene, as a single copy on the chromosome, was a poor selection marker [12]. In contrast, EGDe transformed with pIMC yielded uniformly sized colonies, which were visible after 20 h (Fig 1b). EGDe::pIMC exhibited similar growth kinetics to EGDe and was able to grow in the presence of up to 75 μg/ml of chloramphenicol. The integration of pIMC in EGDe and two additional L. monocytogenes strains (10403S and F2365) was highly stable. After 10 serial passages at 37°C in BHI broth (1:1000 dilution) without antibiotic selection, none of the 100 colonies screened from each strain had lost the plasmid. This vector should prove useful for genetic complementation and as a marker for in vitro and in vivo strain selection.

Development of in vivo multiple competitive index assay
Comparative analysis of virulence potential is a cornerstone in understanding the ability of a pathogen to cause disease and in determining the relative contribution of putative virulence factors. We applied pIMC to establish a widely applicable method to evaluate virulence potential during mixed infections, in the form of a competitive index (CI) [1]. To exploit the high level of chloramphenicol resistance of pIMC, which allows selection from complex samples, we subcloned three IPTG-inducible antibiotic resistance markers ermAM (Ery R ), aphA3 (Kan R ), tetM (Tet R ) and an IPTG-inducible phenotypic marker, gusA (Gus) into pIMCa (see Material and Methods). These markers enable strains to be distinguished directly from each other by antibiotic selection or through X-gluc hydrolysis (leading to a blue colony colour). These plasmids were electroporated into EGDe and the resultant strains were evaluated in vitro for growth rate and in vivo for virulence. No differences in growth rate were observed between the strains when grown in BHI (data not shown).
In the initial in vivo experiments, the four EGDe tagged strains were mixed at a 1:1:1:1 ratio and then administered by intravenous injection to 15 BALB/c mice. On each subsequent day, 5 mice were sacrificed, with the bacterial loads in the spleen and liver enumerated. Total counts recovered from the spleen and liver throughout a three day time course were consistent with a typical EGDe infection (Fig 2a and data not shown), suggesting the tagging and marker expression do not impact on the virulence of EGDe. We were able to distinguish the four tagged strains with a double selection, using both chloramphenicol and an additional marker for selection or discrimination. We found that chloramphenicol at a concentration of 7.5 μg/ml was sufficient to remove background host bacterial flora on BHI agar from mouse faecal and intestinal samples. Counts from both organs indicated that each strain maintained a competitive index close to 1.0 (Fig 2b), thereby maintaining the original Growth kinetics within Balb/c mice after quadruple intrave-nous infection with EGDe::pIMC3 derivatives Figure 2 Growth kinetics within Balb/c mice after quadruple intravenous infection with EGDe::pIMC3 derivatives. (A) A tail vein injection with L. monocytogenes isolate EGDe was administered into 15 Balb/c mice with a total inoculum of 2 × 10 4 CFU per 100 μl. The inoculum was composed of an equal ratio of EGDe wt (5 × 10 3 CFU) tagged with the four different pIMC3 plasmids. On each subsequent day, 5 mice were sacrificed, with the spleens and livers enumerated on different BHI agars as described in the materials and methods. Bars correspond to EGDe; green (::pIMC3kan), red (::pIMC3ery), grey (::pIMC3tet) and blue (::pIMC3gus). Data are presented as mean CFU with standard deviation from five organs. The relative virulence ratio (RVR) (B) was calculated from the data presented in (A). Per mouse, the proportion of each strain comprised within the organ (liver or spleen) (output %) was divided by the proportion of the strain in the initial inoculum (input %) and presented as an individual data point (using the formula in additional file 2). The mean RVR was calculated from the average of 5 organs from each strain. A value of 1 indicates no change in the relative ratio. Symbols correspond to EGDe; green circle (::pIMC3kan), red diamond (::pIMC3ery), grey triangle (::pIMC3tet) and blue square (::pIMC3gus). (C) A tail vein infection with L. monocytogenes isolate EGDe was administered into 15 Balb/c mice with a total inoculum of 2 × 10 4 CFU per 100 μl. The inoculum was composed of EGDe wt tagged with pIMCkan (0.75%), pIMCery (2.75%), pIMCtet (13.25%) and pIMCgus (83.25%) in a skewed ratio. The number on each bar gives the mean percentage for each strain recovered from the spleen or liver. Bars correspond to EGDe; green (::pIMC3kan), red (::pIMC3ery), grey (::pIMC3tet) and blue (::pIMC3gus).
inoculum ratio within each mouse. This demonstrates that none of the four markers had an adverse impact on virulence. The kinetics of infection through the IV route revealed that the spleen was more permissive to initial infection than the liver, consistent with the results of other researchers [14]. In a second experiment, we tested whether the ratios would be maintained throughout an infection if the composition of the initial inoculum was skewed (83.3(gus):13.2(tet):2.75(ery):0.75(kan)). As before, ratios were maintained throughout the course of the infection (Fig 2c). The Gus tagged strain is discriminated by a colourimetric change, which is not directly selective and thus may be overlooked at very low recovery rates (e.g. if virulence has been dramatically affected). The other three markers have the advantage of clean selection; however if four strains are required, we recommend that the dominant strain (e.g. the wild type) should be tagged with this marker.

Virulence comparison of different L. monocytogenes strains
Finally, we tested three genome-sequenced L. monocytogenes strains in a 1:1:1 infection; namely EGDe (Kan R ) 10403S (Ery R ) and F2365 (Tet R ), to examine the kinetics of the individual strain infection and obtain information on the relative virulence of these commonly used laboratory strains. We observed differences in the multiplication of each strain within the organs analysed, with counts of EGDe predominating in both the spleen and liver over the other two strains (Fig 3a). By day 3 in the liver, EGDe outcompeted 10403S, while 10403S was recovered in significantly higher levels than F2365. In almost all cases, the relative virulence ratio method (RVR) resulted in greatly improved statistical validity when compared to the use of raw CFU data (Fig 3b). This was especially true for organ counts after 72 hours, where a large degree of variation in the infection of livers and spleens was observed. However, when relative ratios were examined, the increased statistical robustness of the data resulted in easier identification of significant differences. The RVR can only be calculated when strains are mixed within a single mouse and relative ratios are calculated. An absolute value for the ability to compete is presented in Fig 3c, as the competitive index of each strain relative to EGDe. Here we show that in the spleen after 48 h, both 10403S and F2365 exhibit significantly reduced loads when compared to EGDe, with this margin narrowed at 72 h. In the livers, F2365 was significantly impaired throughout the three day time course compared to both EGDe and 10403S. Taken together, these results show that the number of bacteria present in the organs at 72 h can be ranked as follows; spleen -EGDe > 10403S = F2365 and liver -EGDe > 10403S > F2365.
Growth kinetics within Balb/c mice after IV infection with three L. monocytogenes strains Figure 3 Growth kinetics within Balb/c mice after IV infection with three L. monocytogenes strains. (A) A tail vein injection containing an equal ratio of three L. monocytogenes strains (EGDe, 10403S and F2365) was administered into 15 Balb/c mice with a total inoculum of 2 × 10 4 CFU per 100 μl. Each of the strains was tagged with a different antibiotic marker. On each subsequent day, 5 mice were sacrificed, with the spleens and livers enumerated on different BHI agars as described in the materials and methods. Bars correspond to; green (EGDe::pIMC3kan), red (10403S::pIMC3ery) and grey (F2365::pIMC3tet). Data are presented as mean CFU per organ with standard deviation. Statistical analyses were conducted using the one sample T-test (see additional file 1) measuring the raw CFU differences between two strains to calculate the mean difference per organ over 5 mice. The relative virulence ratio (RVR) (B) was calculated from the data presented in (A) as described in figure legend 2. Symbols correspond to; green circle (EGDe::pIMC3kan), red diamond (10403S::pIMC3ery) and grey triangle (F2365::pIMC3tet). Statistical analyses were conducted using the one sample T-test measuring the RVR differences between two strains to calculate the mean difference of 5 organs (see additional file 2). The calculated P-values are presented, with values below 0.05 considered significant (P < 0.05 (*), P < 0.005 (**) and P < 0.0005(***)). (C) The competitive index score relative to EGDe (set as 1) was calculated by dividing the percentage of the strain (10403S or F2365) within an organ by the percentage value obtained for EGDe. The CI score per organ were plotted (with the same symbols as in (B)) and the mean CI written below. A CI score of 1 denotes no difference in the virulence compared to EGDe (for CI calaulation see additional figure 3). Underlined scores denote were statistical significance from (B) was observed.

Discussion
Our prior experience with the pPL2 plasmid [15,16] indicated that the chloramphenicol resistance expressed was relatively low, therefore we were unable to directly select L. monocytogenes out of complex samples. This could be due to the inducible nature of gram-positive cat genes as described by Lovett [12]. To circumvent the requirement for induction of cat expression, we remodelled pPL2 to decrease plasmid size and exchange the inducible cat promoter for a highly expressed constitutive version. This increased the MIC of the new vector, pIMC, ten-fold compared to pPL2. The improved selection allows the isolation of tagged L. monocytogenes from complex environments such as intestinal contents or food samples. Further development of pIMC involved the creation of a suite of four vectors with diverse antibiotic and phenotypic markers to allow the discrimination of individual L. monocytogenes isolates. These were used to establish a competitive index assay to compare the relative fitness of L. monocytogenes strains. The tags did not compromise the fitness of the transformed strains as determined by growth in complex media and by the maintenance of an initial infection ratio through out a three day time course in Balb/c mice. The method established allows the direct discrimination of strains without post-enumeration manual scoring, such as PCR or replica plating [17,18]. The introduction of post enumeration scoring can introduce bias in the discrimination of bacteria, as only a small sample size is processed per organ. When samples can be directly enumerated, larger numbers can be analysed and the introduction of human error reduced. Also, the IPTG inducible promoter should help to decrease any potential consequences of marker expression during competitive growth.
We also observed improved statistical robustness when analysing data from competitively infected mice. We found that ratios were maintained throughout the course of infection (Fig 2b), suggesting that, as observed with Salmonella typhimurium [1], true replication of the inoculum occurs rather than clonal expansion. The statistical significance was especially evident when data from the third day post infection was analysed (Fig 3a). Large differences in the CFU counts isolated from individual livers and spleens were observed due to inherent variation in the infection process and animal-animal variability. However, analysis of the ratio of strains demonstrated improved statistical validity (Fig 3b). For the experiment reported in Fig 3, 15 mice were used, whereas traditionally 45 mice would have been required if each strain were analysed individually. Therefore, using mixed infections, fewer mice are required to produce better results, a powerful demonstration of the 'reduce and refine' principles.
Competitive index assays for L. monocytogenes have previously been described by two research groups. Auerbuch et al [18] established a strain of L. monocytogenes (10403S) tagged with Tn917 (Ery R ), which mimics wild type 10403S infections. This has been widely applied by researchers using the 10403S strain [19][20][21][22]. However, the assay developed is only applicable to 10403S, requires further analysis post enumeration (i.e. patching of colonies) and only allows the comparison of wild type to a second unmarked strain. In 2006, a fluorescence based method for the discrimination of L. monocytogenes strains was established in a replicative plasmid system [23]. A total of five fluorescent proteins were shown to be functional, with three suitable for simultaneous use to distinguish strains (due to emission spectra). While the tagging method is non-invasive, prolonged incubation is required for chromophore maturation and in the absence of selective pressure, significant and variable plasmid loss was observed, depending on the fluorescent marker used [23]. The authors also describe difficulty in discriminating large virulence attenuations due to the non-selective nature of the marker.
From the comparison of three commonly used laboratory strains of L. monocytogenes, we observed distinct differences in the kinetics of infection. Upon re-examination of the sequence of the inlB gene from F2365, it was shown to encode a nonsense mutation leading to premature stop codon in the gene [24]. This may help to explain the impaired ability to establish infection within the liver compared to the 10403S and EGDe, as InlB plays an important role in liver colonization [8].
We have applied the competitive index experiments to in vivo experiments, but it equally could be used in vitro to examine growth or survival within both sterile and complex systems. Additionally, we envisage the use of internal controls in experiments, such as the comparison of mutant, wild type and complemented mutant, or the dual tagging of a wild type and two separate mutants in murine or tissue culture assays (e.g. adhesion, invasion and/or intracellular multiplication). We acknowledge that problems may be encountered in certain situations when comparing multiple strains in a single host, e.g. if a molecule involved in quorum sensing complemented a virulence defect or when the route of infection can be hijacked by a defective competing strain [25]. Nonetheless, we believe our protocol displays significant advantages over previously described methods and provides an additional tool for use in infection biology.

Conclusion
We have established a method for the comparison of the relative fitness of L. monocytogenes strains utilising a novel, stable integrative plasmid, pIMC, based on the pPL2 phage integrase plasmid. This method allows for the selection and discrimination of up to four isolates from com-plex samples, without the requirement for postenumeration processing. There were no detrimental effects on virulence or growth rate of the L. monocytogenes strains due to tag expression, as demonstrated by systemic spread in a BALB/c mouse model similar to that seen with a wild type strain. The utility of the method was demonstrated through the direct comparison of virulence of three commonly used L. monocytogenes strains (EGDe, 10403S and F2365). The data yielded improved statistical robustness when the ratios were examined rather than the cumulative CFU counts from spleens and livers. EGDe was shown to predominate in the liver and spleen by day three of infection compared to the other two strains. F2365 exhibited similar kinetics to 10403S in spleen replication but was significantly impaired in the ability to replicate within the liver when compared to both strains.
The comparison of the tagged strains' ability to compete within a mouse was determined as follows: the CFU per organ from each strain within a single mouse was divided by the cumulative total of the four strains (addition of each single count) to obtain a ratio of each strain per organ. This ratio was divided by the initial ratio of the inoculum to obtain a relative virulence ratio (RVR). A RVR score of 1 denotes no change the ability to compete, e.g. for the comparison of EGDe, 10403S and F2365 the RVR scores were represented relative to EGDe (set as 1), thus obtaining a competitive index (CI) value. All procedures involving animals complied with relevant legislation and were approved by the animal ethics committee at University College Cork.
Statistical analysis (One sample T-test, Microsoft Excel) was applied to the raw CFU and RVR counts through the calculation of CFU or ratio difference (e.g. strain 1 to strain 2, strain 1 to strain 3 and strain 2 to strain 3) per organ (n = 5). A Microsoft Excel spreadsheet is attached for the calculation of the RVR and subsequent one sample T-test of CFU and RVR (additional files 1, 2 and 3). Competitive index was calculated relative to EGDe with the RVR of 10403S and F2365 per organ divided by the RVR of EGDe (additional file 3). The average CI from 5 organs is presented in Fig 3c.