Aeromonas species are ubiquitous aquatic bacteria that have received increasing attention as opportunistic and primary pathogens in humans. The genus Aeromonas can be classified into two main groups. The first is the psychrophilic non-motile strains, primarily Aeromonas salmonicida, that infect fish. The second and larger group is motile, mesophilic aeromonads associated with human diseases such as gastrointestinal diseases, wound infections, and septicemia (Janda and Abbott, 2010; Parker and Shaw, 2011). Of the approximately 30 recognized species of Aeromonas (Martínez-Murcia et al., 2016), a subset of four species are more frequently implicated in human infections (Aeromonas hydrophila, A. caviae, A. veronii biovar sobria, and A. dhakensis) (Janda and Abbott, 2010; Teunis and Figueras, 2016).

In a previous study, we detected Aeromonas spp. in 71% of fresh retail ready-to-eat (RTE) sushi boxes purchased in Norwegian supermarkets (Hoel et al., 2015). These products are offered as complete meals with a selection of different sushi and are distributed cold (≤4°C) with a shelf life of 2–3 days after production. In addition to acidified rice (pH < 4.6), the sushi meals typically consist of various species of both raw fish and vegetables, which are excellent substrates for bacterial growth. The widespread occurrence of Aeromonas ssp. in fish and seafood, vegetables, meat, poultry, raw milk, and different water sources has been confirmed by others (Villari et al., 2000; Kingombe et al., 2004; Heaton and Jones, 2008; Xanthopoulos et al., 2010; Ottaviani et al., 2011; Carvalho et al., 2012; Latif-Eugenín et al., 2016).

Most aeromonads grow optimally around 28°C but can grow at a wide range of temperatures (4–42°C) (Palumbo et al., 1985a). Environmental isolates (such as food isolates) seem to be more adapted to low temperatures and can proliferate at refrigeration temperatures (Mateos et al., 1993). Growth at temperatures as low as -0.1°C has been reported for some strains (Daskalov, 2006). Species belonging to the genus Aeromonas are generally recognized as sensitive to low pH (min. pH∼6), but food isolates have demonstrated increased tolerability to low pH (Hwang, 2010).

The pathogenesis of Aeromonas-mediated infections is multi-factorial, and the role of the virulence determinants in human infections have not been conclusively established (Pablos et al., 2010). An assortment of virulence factors enables these bacteria to colonize, invade, and infect different hosts. The most reviewed virulence factors are the pore-forming hemolytic toxins, hemolysin, and aerolysin, (Howard et al., 1987; Parker and Shaw, 2011), and three different enterotoxins, including cytotoxic (Act), heat-labile cytotonic (Alt), and heat-stable cytotonic (Ast) enterotoxins. All these toxins have been linked to cases of diarrhea (Galindo et al., 2006). Furthermore, the expression of lateral or peritrichous flagella is associated with enhanced adherence and invasiveness in addition to the ability to biofilm formation (Gavín et al., 2003).

The taxonomy of the genus Aeromonas is complex and has undergone numerous changes with the implementation of new phylogenetic markers. Conventional phenotypic tests do not necessarily correspond to results achieved by genetic methods, and this is especially evident in environmental isolates of Aeromonas species (Ørmen et al., 2005; Puthucheary et al., 2012). Sequencing of the 16S rDNA has proven unsuccessful for identification of Aeromonas species due to low taxonomic resolution (Morandi et al., 2005; Nagar et al., 2013). Therefore, sequencing of the housekeeping genes gyrB (encoding the B-subunit of DNA gyrase, a type II DNA topoisomerase) (Yanez et al., 2003) and rpoD (encoding the σ⁷⁰ factor conferring promoter-specific transcription initiation on RNA polymerase) (Soler et al., 2004) is now preferred for phylogenetic analysis and species identification of Aeromonas spp.

The aim of this work was to assess the species identity and phylogenetic relationship of Aeromonas strains isolated from fresh retail sushi from three producers and to characterize the isolates with respect to genetic and phenotypic virulence factors. Furthermore, we evaluate the potential hazard associated with their presence in minimally processed RTE seafood such as sushi. To the best of our knowledge, no studies have been published which concurrently assess prevalence, species distribution, and genetic characterization of several Aeromonas species in seafood intended for raw consumption. In light of the taxonomical controversies, including recent species reclassifications and the lack of definite virulence markers, studies providing updated characterization of current environmental strains are required in order to provide insight into the impact of these bacteria as foodborne pathogens.

In this study, Aeromonas strains isolated from retail sushi products from three producers were identified at species level and characterized genetically and phenotypically with respect to virulence traits, hemolysis and motility. We identified seven different Aeromonas species, equally distributed between the three producers, with the exception of A. hydrophila and A. dhakensis, which were isolated from one producer a year after the initial sampling. The variability of species distribution over time for producer A demonstrates the microbiological complexity of these products based on an assortment of raw ingredients. The most frequently isolated species, A. salmonicida, is by definition known as a psychrophilic, non-motile species, associated with fish disease (Austin et al., 1998). However, the A. salmonicida isolates in the present study fall into a group of so-called non-typical A. salmonicida characterized as mesophilic, sometimes motile, bacteria with no or reduced pigment production (Dallaire-Dufresne et al., 2014). Infections attributed to non-typical A. salmonicida are an increasing problem in aquaculture, mainly affecting the cleaner fish (wrasse) (Hjeltnes et al., 2015; Biering et al., 2016). A high prevalence of mesophilic A. salmonicida was also reported in studies of Italian food samples (Ottaviani et al., 2011) and in Mexican frozen fish samples (Castro-Escarpulli et al., 2003).

Phylogenetic analysis of Aeromonas spp. based on 16S rDNA sequencing has revealed that the genus is composed of a very tight group of species, some of them differing by only a few nucleotides (Martinez-Murcia et al., 1992; Kupfer et al., 2006). As an example, the inter-species similarity for the 16S rDNA sequences of A. salmonicida, A. bestiarum, and A. piscicola were reported as high as 99.8–100%, and the species were thus impossible to separate by 16S analysis (Figueras et al., 2011). Moreover, the overall mean 16S rDNA sequence similarity for Aeromonas spp. was 97.3%, highlighting the poor discriminatory power of the 16S rRNA gene (Nagar et al., 2013). The discriminatory power of some housekeeping genes was reported to be considerably higher than 16S rDNA, as demonstrated by an 89 and 92% mean sequence similarity for rpoD and gyrB, respectively (Soler et al., 2004; Nagar et al., 2013). The mean gyrB sequence similarity in the present study was 93%, and thus is in accordance with previous studies. The range of nucleotide substitutions within the tighter group of species A. salmonicida, A. bestiarum, and A. piscicola (2.7–3.4%) was on the border between the values of intra- and inter-species sequence divergence observed here, as reported previously (Yanez et al., 2003). Still, the sequence divergence between the three species were higher than the overall intra-species value (1.8%), demonstrating the power of gyrB to separate these phylogenetically close species.

The phylogenetic analysis based on partial gyrB sequences resulted in neighbor-joining and maximum-likelihood trees with robust and well-separated clusters of species. Of particular interest, all A. salmonicida and A. bestiarum isolates were separated into two robust clusters with a 100% bootstrap value (Figure 2). Isolate SU58.3 formed a separate sub-branch between the A. salmonicida and A. bestiarum clusters. Comparison with its closest relatives in the GenBank database produced significant alignment with one sequence of A. pisciciola (type strain sequence) and several sequences of A. bestiarum. However, isolate SU58.3 clustered with the type strain sequence of A. piscicola and their sequence similarity was 98.9%.

The taxonomical controversy of the genus Aeromonas was further demonstrated by the recent reclassification of A. hydrophila subsp. dhakensis (Huys et al., 2002) and A. aquariorum (Martinez-Murcia et al., 2008) as A. dhakensis (Beaz-Hidalgo et al., 2013). A nucleotide database search with the gyrB sequences from the six A. dhakensis isolates included in our study resulted mainly in alignments with sequences annotated as A. hydrophila. In fact, it is estimated that 30% of the genomes deposited in the GenBank database under the name A. hydrophila do not belong to this species (Teunis and Figueras, 2016). However, the phylogenetic analysis of the obtained gyrB sequences demonstrated that the six isolates clustered strongly with known sequences of A. aquariorum/A. dhakensis (Figure 2), and they were clearly separated from the A. hydrophila cluster. For decades, A. dhakensis has been mistaken for A. hydrophila, which might have contributed to an overestimation of the clinical significance of A. hydrophila. Increasing evidence indicates that A. dhakensis is widely distributed in the environment and must be recognized as a potent human pathogen (Chen et al., 2016).

The Aeromonas isolates in the present study were considered potentially pathogenic due to the high frequency of virulence-related genes. Aeromonas pathogenicity is associated with numerous virulence factors, but there is no definite link between the presence of specific toxin genes and clinical presentation. The lack of classification as a true pathogen is also related to the low number of registered outbreaks of foodborne disease linked to Aeromonas spp. Nevertheless, several studies have established an epidemiological link between the sources of infection and clinical isolates of Aeromonas spp. (Khajanchi et al., 2010; Pablos et al., 2011; Teunis and Figueras, 2016). As for other opportunistic pathogens, infections by potentially pathogenic Aeromonas may not always lead to disease due to host responses and the infectious bacterial dose. However, studies of clinical isolates from patients with various infections have brought more attention to some virulence factors, including the enterotoxins, hemolysins, and their secretion systems, and the clinical Aeromonas isolates have been reported to harbor a variety of toxin gene profiles (Rasmussen-Ivey et al., 2016). Furthermore, several studies have effaced the differences between clinical and environmental strains with respect to pathogenicity. The prevalence of toxin genes in isolates from food, environmental, and clinical sources was compared by Ottaviani et al. (2011), and no differences between these groups were detected. In agreement with previous studies of clinical and environmental isolates (Sen and Rodgers, 2004; Khajanchi et al., 2010), we detected multiple virulence-related genes in all sushi isolates. Moreover, we found heterogeneity in the distribution of toxin genes among the isolates, also within species. In our study, the combination of act/alt/hlyA/aerA was the most common, observed in 63% of the isolates. Interestingly, we observed variability in the combination of enterotoxin genes between species, but some of the species were represented by only one or a few isolates. In accordance with previous studies, we demonstrated lack of or reduced prevalence of the classical virulence genes in A. caviae (Kingombe et al., 1999; Carvalho et al., 2012). A significant proportion of clinical A. caviae strains that lacked all the toxin genes or harbored only the aerA gene alone were reported (Ottaviani et al., 2011). However, A. caviae is considered as one of the most potent human pathogen belonging to the genus Aeromonas and is frequently associated with infections (Pablos et al., 2010; Puthucheary et al., 2012). This suggests that for this species, there is a possible involvement of other virulence factors, and it highlights the relevant role of the aerolysin toxin in pathogenesis (Aguilera-Arreola et al., 2007; Ottaviani et al., 2011).

The act gene encoding a cytotoxic enterotoxin has been shown to be highly prevalent in strains from drinking water (70%) and seafood (75%) (Sen and Rodgers, 2004; Yano et al., 2015). For this gene, we observed a species-dependent prevalence consisting of a 100% prevalence in A. salmonicida, A. bestiarum, A. hydrophila, and A. piscicola (represented by only one isolate). In contrast, the act gene was not detected in any A. dhakensis, A. caviae, or A. media isolates. Act and aerolysin are pore-forming toxins, and the hemolytic, cytotoxic, and enterotoxic activities of these proteins have been previously demonstrated (Xu et al., 1998; Buckley and Howard, 1999; Sha et al., 2002). The vast majority of Aeromonas isolated from sushi in our study harbored both the aerA and hlyA genes, but not all of these isolates displayed β-hemolysis on blood agar. Shared characteristics of the non-hemolytic isolates were the absence of act, the faint aerA band observed on the agarose gel, and the presence of hlyA. For these isolates, a link between the absence of the act gene and the observed lack of hemolysis is suggested. In contrast, all A. dhakensis (act negative) isolates in the present study displayed strong hemolysis, suggesting that hemolysis cannot be attributed to a single factor. Wong et al. (1998) proposed a two-toxin model for aerA and hlyA, in which both genes must be knocked out to reduce virulence. The vast majority of Aeromonas isolated from sushi in our study harbored both the aerA and hlyA genes, but not all of these isolates displayed β-hemolysis on blood agar. Thus, the lack of observed hemolysis might be a result of altered gene expression and/or posttranslational processes not analyzed here.

Regardless of geographical origin, the alt and ast genes seem to be less prevalent, and their impact on pathogenesis has not been completely recognized. In the present study, the proportion of isolates harboring these genes were higher than reported previously (Di Pinto et al., 2012; Senderovich et al., 2012; Yano et al., 2015). It has been hypothesized that the species A. caviae and A. veronii lack the ast gene (Puthucheary et al., 2012). However, three out of six A. caviae isolates from the present study harbored this gene. Moreover, we detected ast in the A. veronii biovar veronii reference strain. Keeping in mind the heterogeneous distribution of various virulence genes in strains from various sources and geographical origin and the possibility of horizontal gene transfer, it is unlikely that the lack of specific genes can be attributed to certain species.

Human Aeromonas-mediated infections have occasionally been associated with hemolytic uremic syndrome (HUS), which can lead to acute renal failure and death. HUS is mainly associated with shiga-like toxins (Stx1 and/or Stx2) produced from enterohemorrhagic E. coli strains (EHEC) (Figueras et al., 2007). These toxins are encoded by stx genes in the genome of a lysogenic bacteriophage (Stx phage) of EHEC and represent a horizontal transfer mechanism (Palma-Martínez et al., 2016). The stx-1 and stx-2 genes have been detected in clinical and environmental isolates of A. hydrophila, A. caviae, and A. veronii biovar sobria (Snowden et al., 2006; Alperi and Figueras, 2010; Palma-Martínez et al., 2016). None of our isolates was positive for the stx-1/-2 genes, and to our knowledge, these genes have not been detected in foodborne isolates.

Bacterial motility, enabled by a lateral or polar flagella, have a number of biological functions in pathogens such as chemotaxis, adhesion, and invasion as seen in E. coli, P. aeruginosa, and Clostridium difficile (Haiko and Westerlund-Wikström, 2013; Qin et al., 2016). A recent study demonstrated that flagellar motility is necessary for A. hydrophila to adhere to the host mucus (Qin et al., 2016). We were able to demonstrate motility on soft agar plates in a significant proportion of A. salmonicida (71%) and A. hydrophila (100%) isolates, and less in the other species. However, it must be noted that perturbations in the motility assays such as inoculum state, incubation temperature, and brand of agar can give false negative results (Grim et al., 2013).

Most modern approaches for controlling levels of food contamination by microorganisms are effective against Aeromonas spp. (Daskalov, 2006). However, the increased consumption of raw or mildly processed RTE food poses some new safety concerns because the food is intended for consumption without further preparation such as washing or heat treatment. We have previously reported relatively high numbers of Aeromonas spp. (>log 4 CFU/g) in retail sushi before its stated expiration date. The risk of acute gastrointestinal disease associated with concentrations of Aeromonas spp. found in different water and food matrices was recently simulated by Teunis and Figueras (2016). Of particular concern, the concentration of Aeromonas spp. in the salad identified as the source of a substantial Chinese outbreak was estimated to be in the range of log 2.3–3.0 CFU/g. Several studies have reported the proliferation of Aeromonas spp. at low temperatures (Palumbo et al., 1985b; Palumbo and Buchanan, 1988; Mano et al., 2000; Papageorgiou et al., 2006). However, most studies were done in laboratory cultures or in products not comparable to sushi. Little information is available on the growth rates of Aeromonas spp. in sushi during refrigerated storage. Moreover, fluctuations in temperature are likely to occur during production, distribution, and display in stores. We have seen that the growth rates of other bacterial groups increased in sushi during storage at 8°C compared with storage at 4°C (submitted manuscript). Furthermore, analysis of sushi ingredients revealed that Aeromonas spp. can be introduced into the product through both raw vegetables and fish (Hoel et al., 2015). In the present study, we isolated Aeromonas spp. from a homogenized product containing two species of fish and vegetables (spring onion). Considering the number of species detected, and the fact that all isolates were mesophilic, these bacteria probably originated from more than one ingredient.

Our results demonstrate that potentially pathogenic Aeromonas species are widespread in retail sushi products. The isolates were assigned to seven different species according to gyrB sequencing. All isolates harbored several virulence genes, and a significant proportion were hemolytic and motile, thus showing virulence properties comparable to those in clinical strains (Wu et al., 2007; Puthucheary et al., 2012; Senderovich et al., 2012). The results highlight the potential microbiological hazard linked to the presence of Aeromonas spp. in sushi and in other perishable RTE food not subjected to heat treatment. Because of the multi-factorial virulence, a rapid discrimination between pathogenic and non-pathogenic strains remains a challenge for the genus Aeromonas.

SH performed the experiments, phylogenetic analyses, and wrote the manuscript. OV did the statistical analysis, contributed to experimental planning and critically commented and revised the manuscript. AJ contributed to experimental planning and critically commented and revised the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.