The genus Aeromonas, includes Gram-negative, non-spore-forming rods that are autochthonous of the aquatic environments worldwide (Austin et al., 1996; Janda and Abbott, 2010; Figueras and Beaz-Hidalgo, 2015; Graf, 2015). In humans diarrhea, and wound infections are the most common presentation followed by bacteremia (Figueras, 2005; von Graevenitz, 2007; Janda and Abbott, 2010; Figueras and Beaz-Hidalgo, 2015). Isolation of Aeromonas in patients with diarrhea varies from ca 2 to 10% but a higher incidence (13%) has been found in children from Nigeria (Figueras and Beaz-Hidalgo, 2015). Also a significantly higher incidence of Aeromonas is found in patients with diarrhea than in those considered asymptomatic carriers (Figueras, 2005; Janda and Abbott, 2010; Figueras and Beaz-Hidalgo, 2015). Typical susceptible patients in the different studies include babies, young children and the elderly and especially those with a pre-existing illness and/or immunocompromised (Janda and Abbott, 2010; Figueras and Beaz-Hidalgo, 2015). According to Janda and Abbott (2010) it is important to report the presence of Aeromonas in stools of immunocompromised patients even if they are only asymptomatic because the risk of invasion and dissemination is considered inherently high for these patients. Nevertheless the role of Aeromonas in gastroenteritis has been questioned (Janda and Abbott, 1998, 2010; Chu et al., 2006; von Graevenitz, 2007), mainly because 55 out 57 challenged volunteers with a high dose of Aeromonas did not developed any symptoms of enteric illness, and due to the few reported outbreaks (Morgan et al., 1985; Figueras et al., 2007b; Janda and Abbott, 2010; Figueras and Beaz-Hidalgo, 2015). However, there are also many publications providing support for a causal relation between Aeromonas and enteric disease (Figueras et al., 2007a,b; Janda and Abbott, 2010; Figueras and Beaz-Hidalgo, 2015). Although 10 of the ca. 30 species that comprise the genus have been isolated in association with gastroenteritis, only 4, i.e., Aeromonas caviae, A. veronii, A. dhakensis, and A. hydrophila are the dominating species accounting for ca. 96% of the recovered isolates from this origin in different studies (Figueras and Beaz-Hidalgo, 2015).

Consumption of contaminated water and food are considered the main routes of transmission and there are few reports of well documented outbreaks that include information about the ingested doses of Aeromonas (Krovacek et al., 1995; Granum et al., 1998; Zhang et al., 2012). An association between diarrheal cases and consumption of untreated or contaminated drinking water has been established in several occasions (Holmberg et al., 1986; Martin Delgado et al., 2001; Khajanchi et al., 2010; Pablos et al., 2011; Ventura et al., 2015). Aeromonas was isolated from drinking water and stools in some patients that were diagnosed with traveler’s diarrhea among the tourists that visited specific hotels in the coastal area of Tenerife and Canary Islands in Spain (Martin Delgado et al., 2001). Failures in the drinking water distribution system caused fecal contamination of the water, as revealed by the presence of bacterial indicators of fecal pollution, high concentrations of organic matter and the detection of Aeromonas (Martin Delgado et al., 2001). Once the water distribution system was repaired the incidence of diarrhea among the tourist population returned to normal. Even a recurrent case of Aeromonas bacteremia has been attributed to the consumption of contaminated well water (Katz et al., 2015). Use of contaminated water can cause secondary contamination of food products, and this can be the source of food-borne outbreaks (Altwegg et al., 1991; Krovacek et al., 1995; Granum et al., 1998; Figueras and Borrego, 2010; Wadhwa et al., 2012; Zhang et al., 2012; Figueras and Beaz-Hidalgo, 2014).

It has become generally accepted that only a subset of Aeromonas strains can cause gastroenteritis in humans (von Graevenitz, 2007; Janda and Abbott, 2010; Grim et al., 2014). However, nowadays it is clear that infection is a complex process in which not only the virulence of the colonizing strain is important, but also its interaction with other microbes that are present in the gut, as co-infecting pathogens or in the natural microbial ecosystem, together with the specific physiological status of the host (Lund and O’Brien, 2011; Leggett et al., 2012; Beaz-Hidalgo and Figueras, 2013; Mosser et al., 2015; Ribet and Cossart, 2015; Denny et al., 2016). In fact, the overall crosstalk and interactions between commensal bacteria, enteric pathogens, and host physiology is what is considered crucial to the establishment and progression of intestinal disease (Leggett et al., 2012; Ribet and Cossart, 2015).

Challenge studies have been essential, to establish a causal relation between exposure and health effects and for quantifying the dose response relation (Teunis et al., 1996; Teunis and Havelaar, 2000). However, some outbreak reports include data on the numbers of person exposed, and even information allowing an estimate of the magnitude of the dose involved. Such outbreaks may be treated as ’natural experiments or natural exposure’, comparable to a challenge experiment with a single dose group (Teunis et al., 2004, 2008, 2012b). Both data sources have been successfully used in several studies that investigated the potential of human pathogens different from Aeromonas for producing colonization (infectivity) and acute illness (pathogenicity) (Teunis et al., 2005, 2010).

In the present study we use dose response models for microbial infection and acute illness (Teunis et al., 1999; Teunis and Havelaar, 2000) to re-assess the Morgan et al. (1985) challenge study and combine these data with outbreak investigations to determine the dose response relation for producing infection and acute enteric disease by Aeromonas spp. Furthermore, to illustrate the outcomes, we estimate the risk associated with waterborne exposure to these bacteria using reference concentrations found in different water or food matrices. In addition we provide evidence in support of Aeromonas acting as a true enteropathogen.

The major clinical manifestation produced by Aeromonas is diarrhea affecting principally young children and immunocompromised patients (Janda and Abbott, 2010; Figueras and Beaz-Hidalgo, 2015). The single available human challenge study for Aeromonas shows low susceptibility, in particular to developing acute enteric symptoms with only 2/57 (3.5%) challenged being positive (Table 1; Morgan et al., 1985). In contrast, the “natural exposure events” indicate that a high risk of illness may be associated with exposure to Aeromonas even at moderate to low doses. Thus, host susceptibility may be a strong determinant for the illness risk of Aeromonas sp. The dose response assessment presented in this study corroborates this conclusion (median 1% illness dose was ~1.4 × 10⁴ times higher in “the natural exposure events” compared to the clinical challenge), and shows how these two subsets of illness data lead to different estimates of susceptibility.

Although invaluable for understanding infection and pathogenesis, clinical challenge may not quantitatively represent natural infection that occur during an outbreak. Clinical challenge may underestimate the risk because pathogen inocula may decrease in virulence because of safety testing, often requiring repeated culture in the laboratory. Likewise, volunteers are selected for being healthy and immunocompetent to reduce the risk of serious complications. Conversely, outbreaks select for virulent pathogen strains and susceptible hosts, as this increases the probability that a cluster of cases is detected (Teunis et al., 2004). Consequently, challenge studies and outbreaks may both be biased, in opposite directions: each may represent an extreme in the continuum of dose response relations as discussed earlier (Teunis et al., 2005, 2010; Thebault et al., 2013). Outbreaks or natural exposure events may select for highest susceptibility in the affected host population producing a high illness risks, even when the dose is moderate or low. For instance in the food outbreak that produced acute diarrhea in over 200 college students’ described by Zhang et al. (2012) the estimated concentration of Aeromonas in the cucumber salad determined in the present study ranged from 200 to 1000 cfu/g (Figure 1). This estimated concentration assumed a homogeneous distribution of bacteria (Teunis et al., 2005) in the cucumber salad. As the origin of the contamination in this outbreak was attributed to rinse water of the vegetables, contaminated with sewage, occurrence of bacteria may not have been severely clustered. In case there is severe clustering, some subjects would have been exposed to high doses, while others would not have been exposed at all. Only when there is extreme clustering, the slope of the dose response relation is affected, not its location and the mean dose remains proportional to the amount of contaminated food that was ingested (Teunis et al., 2008, 2012a). As no information regarding clustering in the contaminated cucumber salad was provided by Zhang et al. (2012), any assumptions regarding clustering would be speculation.

Little is known about who is susceptible to any specific strain or pathogen in the population at large, or how many of the already considered susceptible like infants or immunocompromised would develop acute illness. It has been estimated that in developed countries the part of the population that could be susceptible to acquiring a foodborne disease would be near 20% (Lund and O’Brien, 2011; Lund, 2015). Challenge studies, on the other hand, tend to select for low susceptibility, and may represent the opposite extreme of the susceptibility spectrum. The results presented here for Aeromonas appear to support the same conclusion, where the outbreaks represent a worst case situation, while most challenged exposures result in low risk of infection and illness.

As indicated before the results of the Morgan et al. (1985) challenge study have been used to indicate that Aeromonas is not a true gastrointestinal pathogen (von Graevenitz, 2007; Janda and Abbott, 2010). However, it must be noted that 15 out of 57 (26%) were infected. The occurrence of few illness outcomes is not rare in human challenge studies. For instance in a challenge study for Giardia lamblia 40 subjects were challenged, with doses up to 10⁶ cysts resulting in 21 infected subjects, yet not a single of these infected subjects showed any symptoms of enteric illness (Rendtorff, 1954). Nevertheless, the significance of Giardia as a human enteric pathogen is not doubted. The same may be found for Plesiomonas shigelloides with 22 subjects challenged, resulting in eight infected and none of these becoming ill (Herrington et al., 1987). The 3.5% acute diarrhea incidence of the Morgan et al. (1985) study is similar to the 2% of adult diarrhea incidence found in several studies (Svenungsson et al., 2000; Vila et al., 2003; Figueras, 2005; Senderovich et al., 2012; Figueras and Beaz-Hidalgo, 2015).

The above analysis suggests that the natural exposures involve much lower doses than the challenge study. Notwithstanding the low risk predicted from the challenge study data, the same study suggests that the risk of infection is not extremely low, in particular for two of the strains 6Y and 3647 (Morgan et al., 1985). This means that the risk cannot be ignored, even though few cases of acute diarrhea were observed. Janda and Abbott (2010) underlined that the strain SSU (CDC diarrheal isolate), which is probably the most well-characterized Aeromonas strain with respect to virulence factors (i.e., carriage and expression of enterotoxin genes, and potential colonization factors see studies directed by Dr. Chopra, i.e., Grim et al., 2014 and Ponnusamy et al., 2016, and references therein) did not produce infection (colonization) nor diarrhea in the challenge study, so this cannot support the idea that the strains were wrongly selected. However, it is still possible that a critical virulence or colonization factor could have been lost when subculturing the original strains (Morgan et al., 1985; Janda and Abbott, 2010). Experimental studies in a murine animal model showed that the injection of strain SSU generated the death of all mice in 48 h (Grim et al., 2014; Ponnusamy et al., 2016). Strain SSU has been recently identified as Aeromonas dhakensis, a highly virulent species (Beaz-Hidalgo et al., 2013; Morinaga et al., 2013; Figueras and Beaz-Hidalgo, 2015), instead of A. hydrophila as originally thought (Morgan et al., 1985). It is now known that the importance attributed to A. hydrophila is due to misidentification of the majority of strains as belonging to this species, using phenotypic identification systems (Soler et al., 2003; Figueras, 2005; Beaz-Hidalgo et al., 2010; Morinaga et al., 2013). In fact even 30% of the genomes deposited at the GenBank database with the name of A. hydrophila do not belong to this species (Figueras et al., 2014; Beaz-Hidalgo et al., 2015). Old literature on clinical aeromonads limited the identification only to three species, i.e., A. hydrophila, A. sobria (the correct terminology for the clinical strains is A. veronii biovar sobria) and A. caviae (Janda and Abbott, 1998; Figueras, 2005; von Graevenitz, 2007). However, strains under these names may belong to other species (Janda and Abbott, 1998; Soler et al., 2003; Figueras, 2005). For instance in the Morgan et al. (1985) study among the five strains named A. hydrophila three (6Y, B158, and 3284) were suspected to belong to A. sobria (A. veronii biovar sobria), and strain SSU, as commented above, corresponds to A. dhakensis. A recent review showed that when using molecular identification methods ca. 96% of the recovered isolates (313/327) from human feces in different studies belong to 4 species: A. caviae (37.6%), A. veronii (27.2%), A. dhakensis (16.5%), and A. hydrophila (14.5%) (Figueras and Beaz-Hidalgo, 2015). A higher prevalence of A. caviae in diarrhea cases over the other species is reported also in previous reviews (von Graevenitz, 2007). However, studies on traveler’s diarrhea where an adult population is involved are dominated by A. veronii (Vila et al., 2003; von Graevenitz, 2007 and references therein). Also in a recent study performed by Chen et al. (2015) in and adult population from Taiwan A. veronii was the prevailing species (54.6%) followed by A. caviae (27.6%), and with one strain of each of the species A. dhakensis (9.1%) and A. sanarellii (9.1%). Interestingly one patient in the latter study had a history of eating lettuce with salad prior to the illness. In a recent study tomatoes and parsley irrigated with the same water showed the same genotype of a strain of A. sanarellii and the same occurred for an A. caviae strain recovered from the water and the irrigated lettuces (Latif-Eugenín, 2015).

As we commented in the introduction it is generally assumed that the virulence of Aeromonas is multifactorial and that only a subset of Aeromonas strains is capable of causing gastroenteritis in humans (von Graevenitz, 2007; Janda and Abbott, 2010; Beaz-Hidalgo and Figueras, 2013; Figueras and Beaz-Hidalgo, 2014; Grim et al., 2014; Ponnusamy et al., 2016). In fact some Aeromonas strains possess several virulence toxins and secretion systems, among which the Type III secretion system and the Shiga toxin genes are similar to those present in other important pathogenic bacteria (Figueras et al., 2007a; Alperi and Figueras, 2010; Figueras and Beaz-Hidalgo, 2014). The large variation observed among Aeromonas strains maybe linked to the site where the virulence is expressed (Leggett et al., 2012). For instance carriage of virulence genes that have a local action within the host cells like, i.e., the injection of toxins through the Type III or Type IV secretion systems may generate a more virulent response than other strains that exert virulence at more distant sites (Leggett et al., 2012). The latter occurs when pathogens cause secretion of proteins binding to host cells (i.e., immune modulators delivered by the general secretary pathway, or by Type I, II, and V secretory systems). This different pathway or the synergetic effect of such pathways may lead to differences in infective doses (Leggett et al., 2012). In addition the host susceptibility is very relevant, now we know that the genetic polymorphism in the host population determines the variability observed in the type or intensity of responses against the encountered specific pathogens ranging from asymptomatic infections to fatal disease (Ribet and Cossart, 2015).

The concentrations of Aeromonas that were selected in the exposure scenarios to assess risk of infection, were those that can normally be found in oysters, i.e., MPN 9.3/100 g or 10 cfu/100 ml have been reported in several drinking water studies (Abeyta et al., 1986; Austin et al., 1996; Figueras et al., 2005; Egorov et al., 2011; Robertson et al., 2014) or for instance in Milk (1-2 × 10³) or meat products (10²-10³) as found by Borrell et al. (1998). In fact a MPN 9.3/100g was the concentration of Aeromonas found in oysters considered to be the source of an outbreak that affected 472 persons suffering from gastroenteritis in Louisiana (Abeyta et al., 1986). Concentrations as high as 10⁵ or 10⁶ cfu/100 ml can be found in treated wastewater and in reclaimed waters used for irrigation of vegetables (Austin et al., 1996; McMahon and Wilson, 2001; Janda and Abbott, 2010; Fernandez-Cassi et al., 2016). Such sources could contaminate ready to eat vegetables and pose a risk for human health as was recently demonstrated finding the same genotype of Aeromonas in the irrigated water and in the irrigated vegetables as commented above (Latif-Eugenín, 2015).

The dose response assessment reported here shows that when Aeromonas is present, the probability of infection may not be negligible, and in a susceptible host, there may be a high risk of acute enteric illness. It is therefore important to determine how rare it is for a host to be highly susceptible. Outbreaks or other natural experiments are hard to find. This may be because they are so rare, but there could be underreporting because Aeromonas is often not considered among the microbes to be analyzed during investigation of outbreaks of infectious gastroenteritis (Janda and Abbott, 2010; Figueras and Beaz-Hidalgo, 2015). In fact the recognition of Aeromonas in the clinical setting occurs accidentally on routine enteric isolation media designed for other enteropathogens like XLD (MacConkey, Xylose Lysine Dextrose Agar), SS (Salmonella–Shigella Agar), or CIN (Cefsulodin–Irgasa–Novobiocin Agar) as described earlier (Figueras, 2005; Janda and Abbott, 1998, 2010; Figueras and Beaz-Hidalgo, 2015).

The carriage of an infectious microorganism in the general population cannot be easily extrapolated from (observable) symptomatic cases, because of underreporting, but also because in the population at large, a great majority of the infections may remain asymptomatic. Biomarker studies using serum antibodies for estimating infection (seroconversion) rates in the general population have shown that for Campylobacter there may be more than 1,000 asymptomatic infections for every notified case of campylobacteriosis (Teunis et al., 2012b, 2013). Similar ratios have been found for Salmonella (Simonsen et al., 2011). Serology could be helpful in determining whether there could be substantial carriage of Aeromonas in the general population. Serological evidence of infection has been provided in some cases, supporting the true enteropathogenicity of Aeromonas (see review by Janda and Abbott, 1998; Crivelli et al., 2001). In fact a specific secretory immunoglobulin A (sIgA) response at the intestinal mucosa against the extracellular products that appeared in the feces of patients with Aeromonas diarrhea was demonstrated by Crivelli et al. (2001).

Prevalence or incidence of HIV, hepatitis C have been estimate using the “evidence synthesis”, that is a well-accepted methodology for integrating various sources of data to estimate a quantity of interest for which there are no or limited direct data (McDonald et al., 2015 and references therein). This approach was applied recently to determine the incidence of symptomatic pertussis infection in the Netherlands (McDonald et al., 2015) and may help in determining whether the low numbers of symptomatic cases are consistent with frequent isolation of Aeromonas.

The predicted risks of infection and illness show considerable uncertainty. This is caused by the variation observed between outcomes produced by different strains, but it also results from the small sizes of the exposed groups, in particular in the challenge study (Table 1). Unfortunately it is not likely that there will be more or better data available, to improve the precision of the estimates. Despite the small sample sizes and uncertain doses, what is relevant is that the low and high susceptibility estimates are clearly different.

Combined evidence collated from clinical studies in humans and outbreaks shows that Aeromonas should be treated as a human enteropathogen. Exposure to low doses of Aeromonas sp. may lead to infection, but most infections may remain asymptomatic. Given the omnipresence of Aeromonas in the environment, seroprevalence studies in the general population are needed to reveal frequent carriage. Further investigations are needed to determine the specific combination of host, environment and pathogen factors that lead to the occurrence of acute enteric symptoms (illness) associated with Aeromonas infections. The risk of illness may be considerable, even when exposed to moderate doses as shown for the studied outbreaks. Therefore, as suggested earlier, patients with underlying malignancies or immunosuppressing conditions should be closely supervised, considering the inherently high risk of invasion and dissemination associated with this population group.

PT and MF contributed equally to the conception and design of the study, PT performed the mathematical modeling and PT and MF jointly interpreted the data and wrote the manuscript. Both authors approved the final version of the paper.

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.