Tularemia as a waterborne disease: a review

ABSTRACT Francisella tularensis is a Gram-negative, intracellular bacterium causing the zoonosis tularemia. This highly infectious microorganism is considered a potential biological threat agent. Humans are usually infected through direct contact with the animal reservoir and tick bites. However, tularemia cases also occur after contact with a contaminated hydro-telluric environment. Water-borne tularemia outbreaks and sporadic cases have occurred worldwide in the last decades, with specific clinical and epidemiological traits. These infections represent a major public health and military challenge. Human contaminations have occurred through consumption or use of F. tularensis-contaminated water, and various aquatic activities such as swimming, canyoning and fishing. In addition, in Sweden and Finland, mosquitoes are primary vectors of tularemia due to infection of mosquito larvae in contaminated aquatic environments. The mechanisms of F. tularensis survival in water may include the formation of biofilms, interactions with free-living amoebae, and the transition to a ‘viable but nonculturable' state, but the relative contribution of these possible mechanisms remains unknown. Many new aquatic species of Francisella have been characterized in recent years. F. tularensis likely shares with these species an ability of long-term survival in the aquatic environment, which has to be considered in terms of tularemia surveillance and control.


Introduction
Francisella tularensis is a small, Gram-negative coccobacillus and the causative agent of the zoonosis tularemia [1]. Because it is highly infectious and can be spread by aerosols, this microorganism is classified in the category A of potential agents of biological threat by the US Centers for Disease Control and Prevention (CDC) [2]. The species F. tularensis is classically divided into four subspecies: F. tularensis subsp. tularensis (Type A strains), F. tularensis subsp. holarctica (Type B strains), F. tularensis subsp. mediasiatica, and F. tularensis subsp. novicida, which may be considered a different species (F. novicida) because it is an aquatic bacterium of low virulence in humans [3,4]. Type A and type B strains of F. tularensis are the two etiological agents of tularemia. The former subspecies is mainly localized in North America, although it has been occasionally detected in arthropods in Slovakia and Austria [5]. The latter one is found throughout the Northern Hemisphere [1], but has recently been detected in Australia [6].
F. tularensis has multiple reservoirs. Firstly, this bacterium can infect a multitude of animal species, including lagomorphs and small rodents, which are the primary sources of human infections [1,7,8]. Secondly, Ixodidae ticks are vectors but also a probable reservoir of F. tularensis owing to the transstadial transmission of this pathogen in these arthropods [1,8]. Mosquitoes and deer flies can also transmit F. tularensis to humans and animals in specific areas, but are not considered long-term reservoirs of this pathogen [1,8]. Finally, a hydro-telluric reservoir of F. tularensis is suggested by numerous studies but has not been extensively characterized [1,8]. Water-borne tularemia was first described in the 1930s in the USSR [9]. Human infection with F. tularensis may occur from these different reservoirs, and the clinical presentation of tularemia is primarily related to the mode of infection. Six major clinical forms of tularemia are classically recognized. The most frequent route of contamination is through the skin, through contact with an infected animal (especially hares in hunters) or arthropod bites. The ulceroglandular form of tularemia combines a skin ulcer at the site of F. tularensis inoculation and regional lymphadenopathy. The glandular form corresponds to regional lymphadenopathy without any visible skin lesion. Infection with F. tularensis through the conjunctiva (e.g. hand to eye contamination) or the oral route (ingestion of contaminated water or food) correspond to the oculoglandular and oropharyngeal forms, respectively. The involvement of lungs through inhalation of infected aerosols or hematogenous spread of bacteria corresponds to the pneumonic form. Finally, whatever the portal of entry of bacteria, severe sepsis often associated with confusion and F. tularensis bacteremia corresponds to the typhoidal form. The present review summarizes the literature data on human tularemia cases infected from aquatic sources, detection of F. tularensis in aquatic environments, and potential mechanisms of F. tularensis survival in water environments.

Search strategy and selection criteria
Data on tularemia cases related to aquatic sources, on the presence of Francisella species in water environments, and the mechanisms of survival of these bacteria in water environments were collected from the English literature in the PubMed database. Only articles published in the last two decades (1998-2018) were included. They were extracted using the keywords 'tularemia' or 'Francisella' and 'case report' or 'water' or 'mosquito' or 'biofilms' or 'amoeba.' In addition, data on other aquatic Francisella species were added, including F. novivida, F. philomiragia, F. halioticida, F. hispaniensis, F. noatunensis, F. salina, F. frigiditurris, Allofrancisella (formerly Francisella) gangzhouensis, F. marina, F. ulginis and F. endociliophora.

Tularemia and drinking water
In the last 20 years, tularemia cases linked to drinking water have been reported in Turkey, Kosovo, Bulgaria, Georgia, Macedonia, Norway, Sweden, Italy, and Germany (Table 1).
In Turkey, tularemia reemerged in 1988 in the form of water-borne outbreaks of oropharyngeal tularemia cases [10]. Between 1988 and 2018, 28 tularemia outbreaks and non-outbreak tularemia cases linked to consumption of contaminated water were reported in this country . The Turkish outbreaks were frequently of large scale, involving more than one hundred people for some of them [11][12][13][14][15]. Patients were almost exclusively suffering from the oropharyngeal form of tularemia, in accordance with the oral route of contamination with F. tularensis. In addition, cases were observed in the whole population, with a male/ female sex ratio close to 1 or even lower, and both children and adults involved [11][12][13][14][15][16][17][18][19]31,[33][34][35]. This is in sharp contrast to countries where infections usually occur through contact with animals or tick bites, and therefore predominate in middle-aged men due to more frequent occupational or leisure exposures [8]. Less frequently, exposure to contaminated water resulted in glandular [12,15], oculoglandular [14,15,19,25,26], and pneumonic [24] forms of tularemia. In some reports, Francisella species or more specifically F. tularensis were detected by PCR or culture in the suspected water sources [12,[15][16][17][18][19][20]24,32]. Moreover, the same F. tularensis genotypes were detected concomitantly from water sources and related tularemia patients in some studies [24,32,37], confirming the epidemiological link between drinking water and human infection. In the majority of cases, the source of contamination was spring water, water from the community water supplies or water from wells. These water sources were not or inadequately disinfected. Contamination of the water by infected animal carcasses or excreta was the most likely hypothesis [10]. Although F. tularensis subsp. holarctica was identified only for some of these outbreaks, only type B strains have been associated with human infections in Europe and Asia.
In Kosovo, two huge oropharyngeal tularemia outbreaks occurred in 1999-2000 and 2001-2002, each one involving more than two hundred patients [38,39]. It was not possible to establish if the primary source of infection was contaminated water or food, but it can be assumed that both were involved in the human transmission of F. tularensis [38,39]. These outbreaks were likely a collateral damage of the Kosovo war, which ended in 1999 and left the country with bad sanitary conditions. Indeed, people reported that mice and rats invaded their villages during this period, which may correlate with contamination of water wells and cellars with F. tularensis. This bacterium was detected in a field mouse recovered from a water well in an affected village [39].
A long-lasting outbreak occurred between 1997 and 2005 in Bulgaria, involving 285 people [40]. F. tularensis was isolated from water samples collected from private wells. Thus, the route of human contamination was likely water consumption. The majority of patients suffered from the oropharyngeal form of tularemia [40]. Similarly, in Georgia, a tularemia outbreak involving 26 patients occurred in 2006 [41], with a predominance of oropharyngeal cases. The water source was the community water supply from which F. tularensis was isolated [41]. Finally, a small outbreak involving 13 patients was described in 2015 in Macedonia [42].
Water-borne tularemia cases are also reported in Scandinavia. In Norway, three small outbreaks, involving eight to 39 people, linked to the consumption of contaminated water were reported over 20 years [43][44][45]. Most of the patients suffered from oropharyngeal tularemia [43][44][45], although ulceroglandular, glandular, pneumonic and typhoidal forms were also observed [43][44][45]. The source of these infections was water collected from private water wells, as confirmed by PCR-detection of F. tularensis DNA [43][44][45]. F. tularensis-infected rodent carcasses probably contaminated these water wells. In one report, a lemming carcass retrieved from the contaminated water well was PCR-positive for F. tularensis [43]. The involved F. tularensis subspecies was not identified, but only type B strains are found in this country. In Sweden, a small outbreak and sporadic cases of oropharyngeal tularemia were also associated with the consumption of contaminated water from private water wells [46,47].
Tularemia cases related to F. tularensis-contaminated water consumption have been occasionally observed in other countries. In Italy, in 1988, a tularemia outbreak linked to the use of water collected from  an aqueduct involved 24 people living in the same village, most of which developed oropharyngeal tularemia [48]. More recently, a German oropharyngeal case of tularemia was suspected to be associated with drinking of non-purified surface water during a holiday in Bavaria [49].

Tularemia and mosquitoes
In Sweden and Finland, tularemia is mainly transmitted through mosquito bites. Large tularemia outbreaks, frequently involving more than one hundred people, have occurred almost annually in Sweden during the last two decades [46,[50][51][52][53][54]. These outbreaks mainly occurred in central Sweden (counties of Örebro, Stockholm, Södermanland, Västmanland, Värmland, and Dalarna), in wet and woody areas of the Boreal Forest, during the late summer period, because of the mosquito vector habitat and seasonality [46,[50][51][52][53][54]. A large scale epidemiological study in Sweden over 29 years identified that tularemia incidence was positively correlated with the presence of lakes and rivers [55]. Similarly, outbreak modeling demonstrated that tularemia in Sweden is concentrated in a few high-risk regions, with a high incidence in summer likely related to the population dynamics of the mosquito vector [56]. Due to this specific mode of transmission, the ulceroglandular form of tularemia is predominant in Sweden. In addition, cases occur in the exposed population regardless of age, but more frequently in men than women probably because of more outdoor activities in men [46,50,51,53]. Swedish mosquitoes collected in Örebro were PCR-positive for F. tularensis subsp. holarctica, still arguing their role in tularemia transmission in this country [57]. Tularemia is also predominantly a mosquito-borne disease in Finland, where several tularemia outbreaks [58] or sporadic cases [59,60] have been linked with this mode of transmission. Here again, tularemia is a seasonal disease occurring during late summer, and ulceroglandular tularemia is the primary clinical presentation [58,61].
The potential role of mosquitoes in the transmission of F. tularensis to humans has been occasionally reported in other countries, such as Germany [62].
All these reports involved F. tularensis subsp. holarctica as they occurred in Europe. Table 2 summarizes the literature reports of mosquito-borne tularemia cases. Interestingly, F. tularensis DNA was detected in Alaskan mosquitoes, suggesting that tularemia could be a mosquito-borne disease in this U.S. state [63].

Tularemia of other aquatic sources
All around the world, human contamination with F. tularensis has occasionally occurred through other types of water exposure (Table 3).
Tularemia cases were reported after near-drowning accidents in France [64] and Finland [65]. For these two cases, patients developed a pneumonic form of the disease after inhalation of contaminated water. For the French case, contamination occurred after inhalation of freshwater and was caused by F. tularensis subsp. holarctica. The involved subspecies and water salinity were not specified for the Finnish case. Four tularemia cases reported in Turkey were related to swimming activities [66,67]. Three of these cases were oropharyngeal forms that developed after swimming in the same lake [67]. The fourth patient suffered from a glandular form after swimming in a natural aquatic environment in a valley [66]. Otomastoiditis caused by F. tularensis subsp. holarctica were reported in three patients (two from France, one from Germany) after canyoning in the same river in France, between 2009 and 2014 [68]. Otitis likely occurred after exposure to F. tularensis contaminated water, although the source was not specifically identified [68].
Tularemia cases also occurred after handling F. tularensis-contaminated aquatic animals [69][70][71]. A small tularemia outbreak involving 19 patients occurred in Spain after crayfish fishing [69]. The patients developed a glandular or ulceroglandular form of the disease through skin injuries while catching or cleaning red swamp crayfishes sinned in the same river. Cases were grouped over three weeks after which fishing was prohibited [69]. The source of contamination was confirmed by PCR-detection of F. tularensis DNA in one crayfish and two water samples from the river [69]. In Canada, a patient developed ulceroglandular tularemia after a finger injury while cleaning pike and pickerel, which are freshwater fishes [70]. Similarly, in the USA, a patient contracted an ulceroglandular tularemia due to F. tularensis subsp. holarctica after a finger injury while fishing on a freshwater lake [71]. The particular ecology of Martha's Vineyard (Massachusetts), an island in the east coast of the USA should be highlighted. Two outbreaks of pneumonic tularemia occurred on this island, the first in 1978 [72] and the second from 2000 to 2006 [73,74]. F. tularensis subsp. tularensis was isolated from one dead patient and rabbits [73,75]. According to a case-control study, landscaping activities, especially lawn mowing and brush-cutting, were the most likely sources of exposure to F. tularensis through aerosols from environmental material [73,76]. However, the environmental reservoir of F. tularensis could not be characterized. Skunks and raccoons were found to be seropositive for F. tularensis, but other animal species tested had not or very rarely been infected with this pathogen [75]. A low prevalence of infection and a high degree of Francisella genetic diversity in ticks argued against the role of these arthropods as a source of human contamination [77]. In contrast, the role of the aquatic environment as a long-term reservoir of F. tularensis was strongly suspected. PCR detection of Francisella sp. DNA was frequently reported from brackish-water samples collected on the island [78].
The role of brackish water as a potential reservoir of F. tularensis is supported by a more recent study in Nunavik territory (Canada) demonstrating that tularemia seroprevalence in the human population was positively correlated with residence near the coast [79].

Other Francisella species and aquatic reservoir
The genus Francisella includes many other species with a proven or probable aquatic habitat. Genetic analyses have suggested that the ancestral Francisella species originated in a marine habitat [80]. Here, we review water-borne human infections caused by Francisella species other than F. tularensis ( Table 4). Some of these species are opportunistic pathogens mainly Table 3. Tularemia outbreaks and sporadic cases related to other aquatic sources.

Country
Year of occurrence   causing diseases in immunocompromised patients exposed to an aquatic environment. Among them, infections caused by F. philomiragia or F. novicida have been most frequently reported. Less than 20 human infections with F. philomiragia have been published in the English literature [81][82][83][84][85][86][87][88]. These infections occurred in healthy individuals that have survived a near-drowning accident [86] or in immunocompromised patients (especially people suffering from chronic granulomatous disease) after exposure to an aquatic environment [81][82][83][84][85][86]. Interestingly, human contamination usually occurred after exposure to salt-or brackish-water [81][82][83][84][85][86]. Infections with F. philomiragia most often occurred during recreational activities in sea, ocean or bay connected to the ocean [81][82][83]85,86]. A young patient was contaminated after a skin lesion caused by a saltwater crab [84]. In a review of 14 cases, Wenger et al. showed that most F. philomiragia infections occurred in patients living within 50 miles of a salt-water coastline [86], again suggesting a clear association between salt-water exposure and F. philomiragia infections.
The geographical distribution of F. philomiragia is probably wide as human infections with this species have been described in the USA [82,85,86], Canada [84], Europe [83] and Turkey [81]. Frequent detection of F. philomiragia in water samples by culture or PCR indicates that the aquatic environment is likely the primary reservoir of this bacterium [78,[89][90][91][92][93]. In two studies, in Norway [92] and the USA [78], F. philomiragia was only found in salt-or brackishwater but not in fresh-water, suggesting that water salinity is a major element in the natural life cycle of F. philomiragia. However, this species was also isolated from spring water near a salt lake in the USA [89] and cooling towers in China [91], reflecting its distribution in different aquatic reservoirs. F. philomiragia DNA was also detected in ballast water from cargo traveling all around the world [90], which might be a mode of diffusion of this species at the global scale. Although the aquatic reservoir appears to be predominant, F. philomiragia was isolated in a sick muskrat and in brackish water surrounding it [94] suggesting that a mammal reservoir may exist for this bacterium. F. philomiragia was also detected in Dermacentor ticks [95].
F. novicida is also a rare human pathogen. A dozen cases have been published in the English literature so far [96]. For all cases with an identified mode of contamination, an aquatic source was involved [89,[97][98][99][100]. In the USA, a case of F. novicida bacteremia occurred after a near-drowning accident in the Atlantic Ocean [97], another case after exposure to hot spring water near a salt lake [89], and the last one after exposure to water from a private water well [100]. A striking outbreak of F. novicida bacteremia occurred among inmates in the USA after consumption of ice from ice machines from which F. novicida DNA was detected by PCR [98]. Finally, in Australia, a patient developed an ulceroglandular form of infection with F. novicida after cutting himself in brackish water [99]. Most of the patients suffering from F. novicida infection were immunocompromised or had underlying health conditions [98,100]. The identification of F. novicida has never been reported in animals or arthropods [96]. The only known reservoir of this bacterium is the aquatic environment as attested by its repeated isolation from water samples [78,89,101,102]. As for F. philomiragia, salinity seems to impact F. novicida survival in water, as this bacterium was detected by culture or PCR only in seawater [101,102], brackish-water [78] and spring water near a salt lake [89].
Other Francisella species have been rarely associated with human infections originating from aquatic sources. F. hispaniensis was first isolated in 2003 from a Spanish patient suffering from bacteremia [103,104]. The source of contamination was not identified [103]. However, a few years later, F. hispaniensis was isolated in the blood of an Australian immunocompromised patient after he cut himself with a fishhook while fishing [105].
The genus Francisella also includes species that are pathogenic for marine animals. F. noatunensis subsp. orientalis and Francisella noatunensis subsp. noatunensis are widely described as warm-and cold-water fish pathogens, responsible for 'piscine francisellosis.' This disease causes high morbidity and mortality in many fish species worldwide and is responsible for economic losses in aquaculture [106][107][108][109]. F. noatunensis does not seem to be pathogenic in humans [106]. F. halioticida can infect Haliotis mollusks (abalones) [110,111] and F. marina sp. nov. was recently identified as causing disease in Spotted Rose Snapper fishes [112]. Interestingly, a novel Francisella species very close to F. halioticida was isolated in the USA, in a diabetic patient, from an infected skin wound developed after contact with brackish water [113]. Finally, other Francisella species are endosymbionts of marine ciliates such as F. endociliophora [114].
Over the past ten years, new Francisella species have been isolated from the aquatic environment. F. salina and F. uliginis were isolated from sea-water in the USA [101,102]. Water from cooling towers also seems to be a reservoir of Francisella species as attested by recent isolation from these air conditioning systems of F. frigiditurris in the USA [102] and F. guangzhouensis in China and Germany [91,[115][116][117]. These two latter species have been transferred to the new genus Allofrancisella [115]. Table 5 summaries studies dealing with the detection of Francisella species in water samples, either using culture, PCR or both methods. It shows the broad spectrum of Francisella species found in aquatic reservoirs.
Regarding PCR methods, it is important to notice that Francisella species that are not yet characterized may not be PCR-amplified from water samples due to the use of inadequate primers [78]. Even if there is PCR amplification with a new species, it may not be accurately identified as a novel species due to a lack of resolution within the utilized amplicon. Full identification of a novel species requires whole genome sequencing.

Mechanisms of F. tularensis survival in water environments
The high frequency of water-borne tularemia cases implies the persistence of F. tularensis in the aquatic environment. However, the mechanisms of F. tularensis survival in this environment have not been elucidated so far.

Long-term survival of F. tularensis in water
Experimental studies suggest long-term survival of F. tularensis in various water environments. Several authors described the survival of F. tularensis in water microcosm from 1 to 70 days [118][119][120]. F. tularensis survival seems to be influenced by both water temperature [118] and salinity [119]. Gilbert and Rose observed that F. tularensis subsp. holarctica remains cultivable after a stay in water for one day at 5°C or 25°C, but up to 28 days at 8°C [118]. Berrada and Telford showed that both type A and type B strains of F. tularensis remain cultivable after a stay of 8-10 days in fresh-water, but 30-42 days in brackishwater, both at 21°C [119]. In another study, F. tularensis subsp. holarctica remained cultivable after a stay in fresh water at 8°C up to 70 days [120].
Interestingly, when F. tularensis became uncultivable on agar plates, bacteria could still be detected and were metabolically active in water [118,120]. This phenomenon was observed by Gilbert and Rose [118] and Forsman et al. [120] for F. tularensis subsp. holarctica. The latter authors described the persistence of metabolic activity for F. tularensis in water up to 140 days [120]. This state is defined as 'viable but non-culturable' (VBNC) and could be responsible for longterm survival of bacteria in the water environment. The VBNC state has been defined as a state from which bacterial cells cannot be cultured but maintain a metabolic activity and cellular integrity [118]. In addition, the VBNC state may be reversible, as bacteria may become cultivable under certain conditions. This reversion in the ability to grow on acellular media is called 'resuscitation' of VBNC bacteria. The VBNC state has been described for a wide range of bacteria such as Vibrio sp. [121,122], Campylobacter sp. [123], Escherichia coli [122] and Legionella pneumophila [124]. Depending on the bacterial species, the infectious nature and pathogenic potential of VBNC cells are variable, as well as the ways of their resuscitation. Forsman et al. reported that F. tularensis VBNC cells were no longer virulent in mice, and could not be resuscitated [120]. Finally, a recent study demonstrated that F. tularensis subsp. holarctica possesses a mechanosensitive channel that protects this bacterium from hypo-osmotic shock when it is released from an infected animal to water [125].
Long-term survival in water has also been reported for other Francisella species. Berrada and Telford showed that F. novicida (like F. tularensis) remains cultivable after a stay of up to 30-42 days in brackishwater, at 21°C [119]. The fish-pathogen Francisella species are able to survive in water in the absence of a suitable fish host. Indeed, F. noatunensis subsp. orientalis remains cultivable after a stay up to 2 days in freshwater and up to 3 days in seawater [126]. F. noatunensis subsp. noatunensis remains cultivable after up to 12 days in freshwater and up to 50 days in seawater [127]. Interestingly, F. noatunensis was also able to enter in a VBNC state after a period of stay in the water, and these VBNC were not pathogenic to cods [127].
Long-term survival of F. tularensis in natural aquatic environments is suggested by a number of studies detecting this bacterium by culture or PCR in environmental water samples [12,16,17,19,20,32,40,41,43,44,47,78,[128][129][130][131][132]. Not surprisingly, long-term detection of F. tularensis in water environments was reported in countries were water-borne tularemia cases are frequent and predominant. In Turkey, two environmental studies identified F. tularensis subsp. holarctica in water samples collected from the aquatic environment or from village water supply systems highlighting the role of different water sources as common and persistent reservoirs of F. tularensis in this country [32,130]. In Sweden, F. tularensis subsp. holarctica was PCR-detected in water samples collected during outbreak and non-outbreak periods, in tularemia endemic areas [131]. However, F. tularensis was also detected from water samples in countries where tularemia is not or rarely a water-borne disease [128,129,132]. Hightower et al. isolated F. tularensis strains from water in Ukraine and considered that the aquatic environment was the third main F. tularensis reservoir after arthropods and mammals in this country [129]. In Germany, following the reemergence of tularemia in 2004, an ecological study in outbreak areas found a river water sample PCR-positive for F. tularensis suggesting a natural aquatic reservoir for this bacterium [132]. In the Netherland, tularemia also re-emerged in 2011 leading to the surveillance of F. tularensis prevalence in the environment [128]. Surface water samples were PCR-positive for F. tularensis subsp. holarctica in 10% of the randomly collected samples and 88% of the samples collected in areas where tularemia cases among hares or humans had been reported [128]. These studies show the almost constant presence of F. tularensis in randomly collected water samples at different time periods. This observation can only be explained by repeated contaminations of the aquatic environment, particularly from the animal reservoir during epizootics, but also by the persistence of these bacteria in the water environment during non-epizootic periods. Table 5 summarizes studies dealing with F. tularensis detection in water environments. PCR was more effective than culture for detection of F. tularensis in water samples, possibly because of a low bacterial inoculum, the encroachment of F. tularensis by other bacterial species, or a VBNC state of bacteria. However, PCR results should be interpreted with caution since it has been shown that F. tularensis specific primers (such as those targeting fopA or tul4 genes) may also amplify DNA from Francisella species other than the tularemia agents [78].

Survival in biofilms
An essential mechanism for survival and persistence of bacteria in the water environment is biofilm formation. Biofilms are defined as naturally formed adherent communities of bacteria within an extracellular polymeric matrix [133]. A number of bacterial species, such as Vibrio cholerae [134], Legionella pneumophila [135], Helicobacter pylori [136] and Pseudomonas aeruginosa [137] form biofilms to promote their survival under environmental water conditions. In vitro studies have demonstrated that both F. tularensis subsp. holarctica [138,139] and F. tularensis subsp. tularensis [139] can form biofilms. The aquatic Francisella species, F. novicida [139][140][141] and F. philomiragia [142] have also been capable of biofilm formation experimentally. F. novicida was also demonstrated to be able to form biofilm in chitin surface, the second most abundant biopolymer in nature, providing the structure of arthropods, insects, and fungi [139]. F. philomiragia was shown to form more biofilm at 25°C than at 37°C , which is compatible with its natural aquatic reservoir [142]. The fish pathogen F. noatunensis was also demonstrated to form biofilm in vivo. [143]. Until now, biofilm formation has not been associated with virulence in Francisella species [133]. Thus, biofilm formation is most likely a way of environmental survival and persistence in these species [133]. To our knowledge, Francisella biofilm has never been described in natural environmental water microcosm.

Survival in amoebae
Free-living amoebae are ubiquitous organisms in soil and water environments. Several human pathogens, such as L. pneumophila and some Mycobacterium species resist phagocytosis and digestion by the freeliving amoebae and may survive in water environment inside amoebae [144]. Moreover, these bacteria may survive for long periods in amoeba cysts [144]. This mode of survival may apply to Francisella species, especially F. tularensis. Multiple experimental studies have focused on the interaction between Francisella species and several amoeba species [145][146][147][148][149][150][151][152][153][154]. Berdal et al. demonstrated that F. tularensis could penetrate in the amoeba Acanthamoeba castellanii and be released from it [152]. Then, other authors described the multiplication of F. tularensis subsp. tularensis in A. castellannii 24 h post-infection [150]. F. novicida and F. philomiragia were also able to multiply within Hartmanella vermiformis and A. castellanii [142,146,148,150,151]. F. noatunensis was also able to infect and replicate within the amoeba Dictostelium discoideum [149]. Francisella cells were localized within vacuoles in amoeba trophozoites [146] but were also able to survive in amoebal cysts for several weeks [150]. This latter finding suggested that amoeba could be a long-term reservoir of Francisella spp. in water environments. Other authors described an enhanced survival of F. tularensis subsp. tularensis, F. tularensis subsp. holarctica and F. novicida in co-culture with amoebae such as A. castellanii, A. polyphaga, Vermamoeba vermiformis or Ochromonas danica [145,147,153]. These authors described the presence of Francisella sp. inside and outside the amoeba in the co-culture model, suggesting that bacterial survival could be related to intra-amoebal replication, to favorable interaction between extracellular bacteria and amoeba, or both [145,147,153]. Interestingly, Gustafsson et al. demonstrated that growth supernatant of A. palestinensis (without the presence of amoebae) increased multiplication of F. tularensis [154]. After five days in co-culture with amoeba, F. tularensis subsp. holarctica was also shown to enter in a VBNC state [147]. In contrast to L. pneumophila, passage through amoebae did not increase Francisella novicida virulence [148]. Despite some disagreements about interaction mechanisms, all these reports argue that Francisella spp. are resistant to free-living amoeba and that protists may contribute to the survival of Francisella sp. in the water environment. However, to our knowledge, Francisella species have never been detected within amoebae in environmental water samples, in contrast to L. pneumophila and Mycobacterium spp. [155].

Survival in mosquito larvae
Mosquito larvae may also represent a long-term F. tularensis reservoir in the aquatic environment. It has been shown that these larvae can ingest F. tularensis subsp. holarctica that are present in water and ingested bacteria survive throughout the different maturation stages of these arthropods up to adult mosquitoes [57,63,138,156,157]. The fact that mosquito-borne tularemia cases have occurred over years in Sweden and Finland should be considered a further evidence of the existence of a long-term aquatic reservoir of F. tularensis. However, it should be notified that mosquitoes could also be infected at the adult stage after a blood meal on an infected host [57]; in this case tularaemia is not water-borne.

Discussion
Although water-borne tularemia was first described in the 1930s [9], this route of human contamination has been largely underestimated. Tularemia cases linked to the aquatic reservoir are common and can occur as large epidemics. Thus, tularemia is a major public health problem in countries where water-borne tularemia cases predominate. These cases may occur through consumption of F. tularensis-contaminated drinking water, such as in Turkey  and its neighboring countries [38][39][40][41][42], and in Norway [43][44][45] where large-scale tularemia outbreaks caused by F. tularensis subsp. holarctica are regularly reported. Drinking water as a source of human infections with type B strains of F. tularensis have also been occasionally encountered in Sweden [46,47] and central Europe [48,49]. Human infections have occurred after consumption of contaminated water from the community water supplies, especially in countries with bad sanitary conditions, old water networks and inappropriate water treatment. Consumption of unsanitized surface water or well water was also a source of contamination. Consequently, it is of primary interest for practitioners to keep in mind that tularemia may correspond to sub-acute or chronic pharyngitis associated with cervical lymphadenopathy, especially in patients living or traveling in countries where water-borne tularemia cases are frequent. In addition, medical questioning regarding tularemia exposure should include a statement about unsanitized water consumption. Until now, this mode of human contamination has been observed in restricted areas and only linked to F. tularensis subsp. holarctica. However, similar cases could occur throughout the Northern hemisphere where type B strains are encountered. F. tularensis subsp. tularensis has also been associated with the water environment [72][73][74], and could also cause tularemia cases related to drinking water.
Mosquito-borne tularemia is also related to contamination of the aquatic environment by F. tularensis. Tularemia outbreaks caused by F. tularensis subsp. holarctica in Sweden [46,[50][51][52][53][54] and Finland [58][59][60] are primarily related to mosquito bites. Because of skin inoculation of bacteria, the ulceroglandular form of tularemia predominates. Infections usually occur during the warm season at the time of maximum activity of mosquitoes.
Scandinavian practitioners are now particularly aware of this situation and usually diagnose tularemia early in the course of the disease. In contrast, this mode of transmission is most often unknown by physicians in other parts of the word, and mosquito-borne tularemia cases could be missed in patients returning from Scandinavian countries. Also, potential transmission of tularemia through mosquito bites outside Scandinavia has not been thoroughly evaluated. A probable autochtonous mosquito-borne tularemia case was reported in Germany [62]. In Alaska, mosquitoes tested positive for F. tularensis DNA [63] arguing that these arthropods could also transmit tularemia to humans in this US state. In this time of global warming, mosquito vectors could spread to new geographic areas, leading to a rise in mosquito-borne tularemia cases [158].
Finally, tularemia cases may occur through other types of aquatic exposure such as near drowning accident [64,65], swimming [66,67], canyoning [68] and fishing activities [69][70][71], due to penetration of F. tularensis through the skin, conjunctiva, or digestive and respiratory tracts. These cases have been reported in Europe [64,65,68,69], Turkey [66,67] and North America [70,71], suggesting a wide aquatic distribution of F. tularensis. Type B strains of F. tularensis were likely involved in all these cases, except one case occurring in Canada for which type A and type B strains could be involved [70]. These data suggest that F. tularensis subsp. holarctica could be more frequently associated with aquatic reservoirs than F. tularensis subsp. tularensis There is currently no explanation regarding the preferential association of type B strains with water. Williamson et al. demonstrated that type B strains can resist to hypoosmotic shock when released into water [125]. It would be interesting to compare the osmotic shock resistance of type A versus type B strains. Noticeably, almost all human contaminations occurred after contact with fresh-water, suggesting that F. tularensis may better survive in such aquatic environment. F. tularensis subsp. tularensis may also infect humans through water exposure, as attested by the particular ecology of the Martha's Vineyard island in the USA [72][73][74]. In this case, brackishwater was considered as a potential reservoir of F. tularensis [78]. Altogether, available data indicate that human contamination with F. tularensis may occur from a wide diversity of aquatic sources and activities.
Several mechanisms could be involved in the survival of F. tularensis in aquatic environments. Experimental and epidemiological studies have demonstrated that both F. tularensis subsp. tularensis and F. tularensis subsp. holarctica can survive for long periods in water microcosms [32,78,[118][119][120]125,[128][129][130][131][132]. Interestingly, after a long stay in water, F. tularensis has been shown experimentally to evolve to a VBNC state [118,120], which could account for long-term survival of this bacterium in water. Both type A and type B strains of F. tularensis can form biofilms in vitro [138,139], another potential survival mechanism of these bacteria in aquatic environments. Experimentally, both subspecies are able to multiply in amoebae or at least interact with these protozoa to enhance their survival [145,147,150,152,153]. Finally, F. tularensis subsp. holarctica is also able to infect mosquito larvae in vitro and survive during larvae maturation up to the adult stage [57,63,138,156,157]. It should be mentioned, however, that all these mechanisms have been evaluated experimentally, but not yet confirmed in natural water environments. It is very likely that all these mechanisms exist and are entangled. Infected animals and animal carcasses may contaminate water environments, in which protozoa, mosquito larvae and biofilm communities may become contaminated with F. tularensis and serve as reservoirs for this bacterium.
In conclusion, our goal was to demonstrate that F. tularensis is likely able to survive for prolonged periods in various aquatic environments, which likely constitute a significant reservoir for this bacterium. Figure 1 summarizes the probable tularemia aquatic cycle as it can be pictured from current literature data. In terms of public health, it is important to remember that providing people with access to safe drinking water via treatment of municipal and private sources remains a priority. Better characterization and control of the aquatic reservoir of F. tularensis would also be of tremendous importance following a bioterrorist attack. Following a bioterrorist attack, F. tularensis could survive for months in the environment, leading to a high number of secondary tularemia cases. Water and mosquito reservoirs should be monitored in the overall tularemia surveillance, in addition to the wildlife reservoir.

Disclosure statement
No potential conflict of interest was reported by the authors.