Comparative genomics profiling of clinical isolates of Aeromonas salmonicida using DNA microarrays

Aeromonas salmonicida, the causative agent of furunculosis in salmonid fish, is a non-motile, Gram-negative bacterium, and one of the most studied bacterial pathogens of fish. Furunculosis is an important disease in wild and cultured stocks of fish with the potential for severe negative economic impact. The Canadian Aquaculture Industry Alliance estimated the total direct and indirect costs of infectious diseases within the Canadian aquaculture industry at over $400 M annually, with furunculosis contributing approximately $4 M annually to these losses [1]. A. salmonicida is not limited to salmonids and many species of fish are affected. Several excellent reviews of A. salmonicida and furunculosis are available [2–4].

Although Bergey's Manual of Systematic Bacteriology [5] recognizes five subspecies of A. salmonicida: salmonicida, achromogenes, masoucida, smithia, and pectinolytica, many laboratories currently classify A. salmonicida subsp. salmonicida as "typical" and any isolate deviating phenotypically as "atypical". Morphological and biochemical differences are used to distinguish typical and atypical isolates. These are pigment production, colony size and growth rate, haemolysis, and sucrose fermentation [4, 6, 7]. A. salmonicida subsp. salmonicida (i.e. typical) isolates grow well on blood agar with large colonies, produce brown, diffusible pigment, are haemolytic and do not ferment sucrose. Historically, typical strains are thought to be extremely homogenous [8, 9], and therefore any deviation in any of these characteristics has been considered enough evidence to classify a strain as "atypical" [8, 10]. However, some of these parameters are subjective, and are largely based on the experience of the operator/microbiologist, possibly leading to misclassification. Molecular techniques are not yet used on a regular basis in many laboratories that type A. salmonicida. Although typical isolates are generally phenotypically homogenous some variation does exist [10], particularly with regard to the production of hemolysin and degradation of casein. Antibiotic resistance patterns have shown potential as epizootiological markers in specific geographical origins. Typing schemes based on biochemical and other phenotypic methods are dependent on many factors, including plastic bacterial phenotypes and the often low consistency of inter-laboratory testing [11].

Genetic analysis provides a more stable basis for microbiological investigation than phenotypic methods. Several molecular techniques, including finger-printing by Randomly Amplified DNA Polymorphism (RAPD-PCR) [12, 13], Amplified Fragment Length Polymorphism (AFLP) [14], Pulsed-Field Gel Electrophoresis (PFGE) [9, 15], Restriction Enzyme Fragmentation Patterns (REFP) [16], plasmid profiling [17, 18], and ribotyping [19, 20], have been used to study different strains of A. salmonicida. While the results generally support the phenotypical evidence that typical strains are genetically homogenous and may well be clonal, and that atypical isolates are heterogeneous, there is little congruence between the techniques in establishing relationships between subspecies, strains and isolates. Austin et al. [1998] showed that among 52 isolates of atypical A. salmonicida taken from a wide host and geographical range, there was no agreement between PCR, RAPD, ribotyping, or phenotypic typing methods [8].

The use of genome arrays containing whole genomes or large sets of genes, either as a result of high-coverage genome sequencing or selected after suppressive subtractive hybridization have been used to study genome variability among strains of bacterial pathogens, including Campylobacter jejuni [21], Flavobacterium psychrophilum [22], Listeria monocytogenes [23], and Burkholderia species [24]. To examine the genomic diversity of virulence factors among clinical isolates of A. salmonicida and other aeromonads, we constructed a DNA microarray of 2024 selected virulence-related genes from A. salmonicida subsp. salmonicida strain A449. A449 is a wild-type typical virulent clinical isolate which has been sequenced and it is currently undergoing annotation.

We are undertaking a research program to examine the molecular mechanisms of interaction between A. salmonicida and Atlantic salmon (Salmo salar L.). The applied goal of this program is to develop vaccines, vaccine delivery systems and other health management tools. In order to develop vaccines we must first identify virulence-associated genes as potential vaccine candidates, the process known as "reverse vaccinology" [25]. Accurate typing is important in searching for vaccine candidates if there are known differences in pathogenicity of the sub-species, or if there is variation between the degree of conservation of virulence genes within each subspecies. Although all subspecies of A. salmonicida have been implicated in clinical cases in fishes, Aeromonas salmonicida subsp. salmonicida is more prevalent within salmonids and "typical" isolates are more often associated with disease outbreaks in salmon [3]. Moreover, there are indications that there are differences in the susceptibility and immune response of salmonids and other fishes to "atypical" isolates [14, 26]. Therefore it is essential that any characterization or typing schemes be accurate, if health management strategies are to be based, even partly, on those schemes.

We used M-CGH to study the genetic relationships between Aeromonas species, subspecies and strains based on gene conservation profiles and to examine genomic diversity. We compared strains and isolates from selected geographic areas and host species to explore correlations between geographic or host origin and conservation or diversity of genes. These strains were selected to provide a diverse sample of clinical and laboratory isolates from the differing geographic regions and host species infected. Our goal was to identify a set of virulence-associated genes conserved across these strains to consider for future consideration as vaccine candidates. We hypothesized greater variability of these virulence genes between subspecies and no correlation between the "typical" and "atypical" classifications and gene variability. We also hypothesized positive correlation between the degree of variability and the functional category of the variable genes involved.

In this study, patterns of variability of the 2024 selected virulence-associated genes of A. salmonicida subsp. salmonicida strain A449 within the sixteen various test isolates of A. salmonicida and other Aeromonas species were examined. It is noteworthy that genes divergent and absent in some tester strains seem to be distributed across the entire A449 genome sequence and across all the predicted functional categories whereas genes shown to be at higher copy number than in the reference strain were clustered in selected loci associated with plasmids. The number of divergent and absent genes correlated strongly with species and subspecies boundaries. For A. bestiarum and A. caviae ~16% of the genes were divergent or missing, whereas for the achromogenes and masoucida subspecies ~8% of the genes were divergent or missing. The range across the various A. salmonicida subsp. salmonicida strains was between 0.3% and 2%, again reinforcing the single clone theory for this species. Overall, for the 14 A. salmonicida isolates studied, approximately 19% of the genes were divergent in at least one sample. For comparison, our previous meta-analysis of M-CGH studies of 97 strains of C. jejuni showed ~33% of the genes in the genome were divergent in at least one strain [21]. It is likely that as we examine more strains of A. salmonicida the number of divergent genes will increase. One interesting finding is that almost 70% of the virulence-associated genes on the DNA microarray are conserved across all the Aeromonas species. Because more than 80% of these highly conserved genes appear to be chromosomal, this suggests that the divergence of A. bestiarum, A. caviae, and A. salmonicida may have occurred fairly recently.

The majority of A. salmonicida subsp. salmonicida isolates studied were isolated from Atlantic salmon and gave homogenous results, i.e., there is no correlation between host or geographic origin and M-CGH patterns in subsp. salmonicida isolates. Highest variability amongst the subsp. salmonicida isolates was observed in an isolate (SS70.1) from coho salmon and in an isolate (N2461) from turbot. However, more isolates from Pacific salmon should be analyzed before drawing definite conclusions on the variation between isolates from Atlantic and Pacific salmon. The relatively low number of divergent genes in most of the A. salmonicida subsp. salmonicida strains supports the findings of Garcia et al. who identified a single clone of A. salmonicida subsp. salmonicida as responsible for most outbreaks of disease worldwide [9].

Differentiation of typical and atypical strains of A. salmonicida is important economically if it can be shown that there are significant differences in the degree of variation in the genes associated with virulence and pathogenicity, and which affect the immune response of the host. If this is the case, then accurate typing methods and differentiation between typical and atypical strains and subspecies are required for diagnostic methods and for vaccine development. Lund et al. have shown variation in the protective ability of some commercial furunculosis vaccines (whole bacterins made with typical strains) against atypical strains infecting farmed spotted wolffish [14] and halibut (cited in [14]). In this context, M-CGH provides an objective means to type clinical isolates based upon the presence, absence or divergence of a large number of genes.

Phenotypical variation amongst subsp. salmonicida isolates is rare but has been described [31]. The increased resolution of M-CGH compared to phenotypic typing can be illustrated on consideration of strain N4705 which was considered an atypical strain on phenotypical grounds. However, both M-CGH and subsequent pulsed-field gel electrophoresis [Colquhoun, unpublished results] suggest that this strain is very closely related to, and should perhaps more accurately be considered a typical strain. Similarly, atypical strain N2517, may be more accurately placed within the subspecies achromogenes based upon M-CGH results.

Further comparative studies will be required to establish the level of strain resolution capable by M-CGH for aeromonads. However, in the absence of complete genome sequence data for each Aeromonas strain used, M-CGH, which in this study uses over 2,000 separate genetic markers, represents a comprehensive, high- resolution methodology for comparing genome information. Recent work by our group used M-CGH to examine the quantitative relationship between the Log Ratio and probe/target identity, and these analytical processes were used in the study [27].

It is preferable that vaccine candidates be strongly conserved among strains that cause disease. The M-CGH data from these experiments will provide a list of conserved genes, sorted by function category, which can be further analyzed to generate these candidates. For example, there are bioinformatics programs available to predict sub-cellular location [32] to select candidates which may be exposed to the host immune system, or to predict adhesins [33]. These downstream analyses, in combination with M-CGH analysis will reduce the number of genes required to be cloned in order to test the immune response of their encoded proteins. These experiments are currently underway.

It must be noted that M-CGH does not reveal information about gene expression or the specific role of targeted genes in pathogenicity or host immune responses. However, it is a powerful tool in reverse vaccinology and M-CGH is a sensitive and comprehensive technique that can determine genomic variation between pathogen strains, and it also can be used as a screening tool for target identification, and/or a typing method.

M-CGH is a powerful first screen for vaccine target identification, and is the basis of reverse vaccinology [25], as it facilitates identification of conserved and duplicated genes associated with virulent strains. This technique may also prove useful in strain typing and epidemiological studies.

Comparative analysis of genomic data of A. salmonicida and related organisms reveals candidates common to all clinical isolates or those strains shown to be associated with virulence and disease. Our work on knock-out mutants of A449 virulence genes, combined with proteomic analysis of selected isolates confirm the results of our M-CGH analyses and will enable us to select specific antigens for vaccine development [work in progress].

The findings reported here support our initial hypotheses that variability exists between subspecies, that there is no correlation between 'typical' and 'atypical' classifications and gene variability. We also found that there is a positive correlation between the degree of variability and the functional category of the genes involved.

Future work will include the use of microarray transcript profiling experiments to further focus the choice of target genes. This will allow us to develop an ongoing, dynamic list of vaccine candidates to be used in live challenges for virulence and protection assays and then in further vaccine development.