In the natural aquatic ecosystems, mortality of prokaryotes is caused, to a great extent, by protozoan grazing (Menon et al., 2003). Ingestion rates vary widely across the different groups of protozoa (amoebae, heterotrophic nanoflagellates, and ciliates), depending on their feeding behavior, prey size, and prey abundance (Pernthaler, 2005). The daily reduction of the bacterial standing stock by ciliate grazing may range between 1 and 8% (Kemp, 1988; Wieltschnig et al., 2003; Tuorto and Taghon, 2014). Clearance rates of heterotrophic nanoflagellates have been estimated to account for up to 50%, although grazing efficiencies vary strongly according to the study (Weisse and Müller, 1990; Wieltschnig et al., 2003; Bettarel et al., 2004). While ciliates and heterotrophic nanoflagellates are effective grazers in the open water column, amoeba graze primarily on biofilms (Zhang et al., 2014). Bacterial and viral losses through grazing are most often influenced by the trophic status of the water source and by season (Jacquet et al., 2005). Losses of the bacterial standing stock in the eutrophic lakes have been measured at values up to 28%, whereas researchers have reported losses in the oligotrophic lakes of up to 70% (Šimek et al., 1997; Domaizon et al., 2003).

Although other small protozoa and bacteria are the favored food sources, studies have reported that Tetrahymena and other ciliates, heterotrophic nanoflagellates, and amoeba take up non-attached viral particles as well (Groupé and Pugh, 1952; Knorr, 1957, 1960; Suttle and Chen, 1992; Gonzalez et al., 1993; Bettarel et al., 2005; Evans and Wilson, 2008; Bouvy et al., 2011). Deng et al. (2014) compared the reduction in the population of the model bacteriophage MS2 in the presence of three heterotrophic flagellates: the filter-feeding flagellate Salpingoeca sp.; the benthivorous grazer Thaumatomonas coloniensis; and the active raptorial feeder Goniomonas truncate. The experiment was performed in the presence of a natural bacterial community in groundwater. Grazing by Salpingoeca sp. or T. coloniensis, decreased the MS2 titer by six orders of magnitude within a 90-day period (Figure 2). In the absence of protozoa, a reduction of only 2 log units was observed, and that was attributed to the antagonistic activities of the bacterial community. Although ingested, viruses only marginally contribute to the protozoan diet in terms of carbon (Deng et al., 2014). Hennemuth et al. (2008) demonstrated that protozoa may directly reutilize viral amino acids for their own protein biosynthesis.

As grazing efficiency is strongly affected by prey size, motility, nutritional quality, and cell-surface characteristics, there is debate regarding whether the pathogenic bacteria and viruses are consumed by protozoa selectively, or merely by chance. Early evidence was provided by Bahr (1954) that there is some selective discrimination. A mixture of different bacterial strains was offered to various ciliate cultures. After some days, the individual ciliate cells were picked and transferred into a saline solution to induce cell lysis. Bacteria released from the bursting cells were cultured on selective agar (endoagar). Although E. coli could be isolated in the early stages of the experiment, at later time points only Bacillus subtilis, and to a lesser extent Staphylococcus aureus, were found present inside the ciliates. Moreover, besides the evidence for selective avoidance, E. coli was even found harmful for the ciliate grazers, as uptake of these species led to protozoan decay. It is important to consider that not all particles ingested by the protozoa are inactivated and/or digested. Occasionally, some bacteria, like V. cholerae or Legionella pneumophila, have revealed resistance to phagocytosis (e.g., by preventing the fusion of the lysosome with the phagosome, thus avoiding digestion) and may even replicate within amoebas or cause the death of their grazers (Barker and Brown, 1994; Kirby et al., 1998; Hägele et al., 2000; Greub and Raoult, 2004; Abd et al., 2007). Co-evolutionary selective forces continuously drive the development of bacterial-evading mechanisms by altering the cell surface molecules, cell morphology, speed of motility, biofilm formation, or toxin release (Matz and Kjelleberg, 2005; Matz et al., 2005; Pernthaler, 2005; Siddiqui and Khan, 2012). Similar anti-grazing mechanisms are not known for viruses, to date, but may indeed exist.

Protozoa play a key role in balancing the bacterial populations not only in natural aquatic systems but also in the wastewater treatment plants or other kinds of bioreactors. In particular, during the biological phase of the wastewater treatment process, protozoa are important for the flocculation and reduction of bacterial biomass (Lee and Welander, 1996; Pauli et al., 2001). The effectiveness of protozoa as biocontrol agents against human pathogenic bacteria and viruses in both manmade and natural aquatic systems depends upon many factors, including protozoan abundance, growth and grazing rates, predation (in) specificity, pathogen abundances and growth rates, as well as rates of predation on protozoa by higher organisms (e.g., copepods) (Brabrand et al., 1983; Sigee et al., 1999). Improved removal of enteric bacteria due to protozoan grazing has been observed in the biological filters (Stevik, 1998). Although concentrations of undesired bacteria and/or viruses, as well as of grazers, are less in the natural aquatic environments, similar effects may be expected; however, no conclusive data are available at this point.

Viruses that prey exclusively on prokaryotes are called bacteriophages or phages. In the aquatic environment, phages outnumber bacteria and archaea by 10-fold or more (Fuhrman, 1999) and phage-induced lysis of prokaryotes accounts for 5–50% of the day-to-day bacterial mortality (Fuhrman and Noble, 1995; Wommack and Colwell, 2000; Weinbauer, 2004; Suttle, 2007; Brussaard et al., 2010). Phages generally display certain specificity for a host; however, that range can be very narrow, for a particular species only, or relatively broad, including various species within a common grouping. This includes human pathogenic bacteria. It is assumed that every organism has its own subset of viruses to which it is susceptible. In a study undertaken by Khan et al. (2002), it was reported that viral predation can even encompass both Gram-negative and Gram-positive bacteria. Indeed, lytic phages influence the microbial diversity and population structures, thus adding a significant selective pressure on the microbial communities (Letarov and Kulikov, 2009; De Paepe et al., 2014). In the oligotrophic environments, phage-induced lysis may stabilize the co-existence of bacteria by avoiding the overgrowth of a single species (a scenario known as the “killing-the-winner” theory) (Thingstad and Lignell, 1997; Shapiro et al., 2010; Winter et al., 2010). Additionally, fitness costs for carrying phage-resistance genes in nutrient-poor environments, such as groundwater, are comparably high. This indicates that, in this type of environment, phages cause minor, but continuous long-term diminishing effects on the bacterial biomass (Lopez-Pascua and Buckling, 2008). Likewise, the evolution of phage resistance in these environments reveals a strong association with the presence of co-occurring phages (Gómez and Buckling, 2011). As a type of protective function, certain bacteria may organize themselves into biofilms, which are more difficult for phages to access; however, as a remedy to the bacterial solution, some phages have evolved polysaccharide depolymerases attached to their tail fibers that digest the bacterial extracellular polymeric substances (EPSs), gaining access to bacterial cell surfaces (Adams and Park, 1956; Hughes et al., 1998a,b). Temperate phages (or prophages) can contribute to phenotypic changes via horizontal gene transfer, driving bacterial evolution and adaptation to new habitats, and this is often accompanied by an increase in bacterial virulence (Jiang and Paul, 1998). Both human-pathogenic serotypes of V. cholerae (O1 and O139) can acquire two pivotal virulence factors (toxin-co-regulated pilus and cholera toxin) that were found being mediated by phages (Waldor and Mekalanos, 1996; Karaolis et al., 1999). Indeed, most toxin-coding genes are linked to a lysogenic lifestyle, as with the diphtheria toxin or the cholera toxin, leading to a great risk of emerging new pathogenic bacteria (Freeman, 1951; Waldor and Mekalanos, 1996; Brüssow et al., 2004; Tinsley et al., 2006).

Of all the participants acting in antagonistic microbial interactions, viruses (and here mainly bacteriophages) probably represent the most powerful ones. Isolated bacteriophages were used in the early 1920s to treat the pathogenic bacteria in humans, a therapy that has recently regained attention due to the growing number of multidrug-resistant pathogenic bacteria (Thiel, 2004; Viertel et al., 2014). To date, lytic activities of phages against a broad variety of pathogenic bacteria have been reported (e.g., against S. aureus, P. aeruginosa, Salmonella enterica, V. cholerae, or E. coli) (Slopek et al., 1987; Capparelli et al., 2010; Ceyssens and Lavigne, 2010).

A 1-year surveillance study by Mookerjee et al. (2014) impressively presented data of a phage controlling a human pathogen that was indigenous to an aquatic environment. The team monitored the dynamics of the toxigenic V. cholerae strain O1 and its lytic phage (vibriophage) at two different sites of the Hooghly River in West Bengal over the three seasons, summer, monsoon, and winter. With an increasing abundance of V. cholerae, the corresponding phage titer increased, then leading to a responsive decline in the V. cholerae load (Figure 3). These repeating patterns strikingly underline the potential control that phages may render on pathogenic bacteria in a natural setting.

Active “phage therapy” in the environment has been suggested for several years. At present, phages are used routinely for the biocontrol of herbal or food borne pathogens, or for decontamination in aquacultures and food industries (Nakai and Park, 2002; Balogh et al., 2010; Goodridge and Bisha, 2011). The rapid generation time of cyanophages, for example, makes them attractive agents for controlling the toxic and bloom-forming cyanobacteria (Sigee et al., 1999 and references therein). When isolated from lake water and treated with a natural viral cocktail, Microcystis aeruginosa decreased in abundance by 95% within only a few days (Tucker and Pollard, 2005). In the aforementioned study, two phages displaying a T7-like morphology and belonging to the Podoviridiae group (short tails) were assumed to be responsible for killing the cyanobacterial strain. Yoshida et al. (2006) isolated a cyanophage (Ma-LMM01) that specifically infected and killed Microcystis aeruginosa. Baudoux and Brussaard (2005) isolated several lytic viruses from freshwater infecting the eukaryotic algae Phaeocystis globosa, an abundant and harmful, bloom-forming phytoplankton. Phages infecting Vibrio coralliilyticus and Thalosomonas loyaeana, both aggressive coral pathogens, were isolated and applied to curing infected corals (Efrony et al., 2007; Cohen et al., 2013). Another example of beneficial phage use is the dewatering process of sludge in wastewater treatment plants, which is an important process for condensing the sludge volume. High levels of EPS (up to 99% water content) are problematic in this step of the treatment, generally interfering with effective volume reduction (Costerton, 1999). The presence of some extensive EPS producers, like Zoogloea and Thauera, may be controlled by the application of selective bacteriophages (Kang et al., 1989; Thomas et al., 1993; Sanin and Vesilind, 1994). Another interesting feature for phage application has been observed recently by Chan et al. (2016). This group isolated a naturally occurring phage that forces a desired genetic trade-off between phage and antibiotic resistance, thus favoring a development toward increased antibiotic sensitivity for P. aeruginosa in the presence of this particular phage. Besides the successful application of phages in patients and other hot spots of pathogens in the laboratory and the environment, the effectiveness of bacteriophages selectively inactivating and killing target hosts in natural settings remains unclear and co-evolutionary mechanisms of hosts’ phage resistance needs further attention as it may limit a long-term application (see below).

In complex and diverse communities, competition for nutrients and space is high. Interspecific competition between prokaryotes often is mediated by the use of a variety of antimicrobials, such as secondary metabolites (e.g., lactic acids from lactobacilli), extracellular enzymes (e.g., lysozymes, exotoxins, bacteriocins), or antibiotics (e.g., streptomycin, tetracycline, or vancomycin) (Jack et al., 1995; De Boer et al., 2005; Riley and Chavan, 2007; Hibbing et al., 2010; de Lima Procópio et al., 2012). While we all got accustomed to the use of antimicrobial substances, such as antibiotics, against pathogens, we often forget that these substances are naturally produced by microbes, giving them a competitive advantage. Indeed, natural microbial communities have a yet unrecognized arsenal of substances that they apply daily in their “microbial war” (Hibbing et al., 2010). Bacteriocin production, for example, is found in a vast majority of bacteria (e.g., within the genera Myxococcus, Lysobacter, and Bacillus). Bacteriocins, such as colicin, are primarily active against closely related species, and work by degrading the antagonist’s inner membrane or nucleic acids. Gram-negative bacteria, in particular, lack a specific secreting system for bacteriocins; therefore, the release of these substances occurs via their own cell lysis. This indicates that only a small fraction of the bacterial population produces bacteriocins, thus providing a competitive edge to their population (Cascales et al., 2007). Nevertheless, for Gram-positive bacteria, bacteriocin production is not necessarily lethal, as some express a bacteriocin-specific transport system that is used for shuffling the antimicrobial back out of the cell (Riley and Wertz, 2002). The expression of antimicrobial substances generally occurs in the stationary phase of the bacterial growth cycle, when they are running short of nutrients. These compounds may enable or disable the invasion of a strain into an established community; they may provide the release of nutrients by cell lysis, and they may even affect the interbacterial communication (e.g., quorum sensing) (Miller and Bassler, 2001; Riley and Wertz, 2002).

Early studies on “lytic” bacteria emphasized their potential in controlling specific groups of microbes; however, those studies were primarily descriptive. Hirsch and Rades-Rohkohl (1983) isolated aerobic bacteria from groundwater and monitored their ability to reduce E. coli K12 numbers using a simple agar overlay method. From the 214 different bacterial isolates tested, 24% revealed inhibitory/lytic effects against E. coli. In total, 39% of the isolates displayed negative interactions against the fecal pathogen. As another example of potentially controlling cyanobacteria in aquatic environments, scientists found a direct statistical correlation between the chlorophyll-a concentration, the cyanobacterial biomass, and the abundance of lytic bacteria (Daft et al., 1975; Fallon and Brock, 1979). Similarly, the filtrate of different actinomycetes (e.g., Streptomyces sp.) revealed antimicrobial properties against 50% of more than 400 prokaryotic strains tested (Safferman and Morris, 1962). Similar studies on antimicrobial substances released by algae and actinomycetes reported not only antimicrobial effects but also antiviral effects (e.g., Coxsackievirus or Poliovirus) (Husmann, 1966; Cliver and Herrmann, 1972; Daubner, 1972; Herrmann and Cliver, 1973; Rehse, 1977). Nasser et al. (2002) examined the potential for proteases and elastases produced by Pseudomonads to reduce different viral titers, and found that the effects were dependent on the virus type (Figure 4). Cox-A9 virus and Hepatitis A virus were significantly affected by the presence of extracellular bacterial enzymes, whereas Polio-1 virus remained unaffected. Although similar in size, the viruses were characterized by pronounced differences in the compositions of their capsid proteins, as evidenced by differences in the isoelectric points. Accordingly, different viruses may react in a distinct manner in the presence of extracellular enzymes (Nasser et al., 1991, 2002).

Some reports are available on the bacteriocin responses against certain medically important human pathogenic Gram-negative bacteria (e.g., Campylobacter, Heliobacter, and Neisseria) (Mota-Meira et al., 2000). In the food industry, antimicrobials are used routinely as preservative agents and for the reduction of specific unwanted germs (Burnham et al., 1981; Riley and Gordon, 1992; Lin and McBride, 1996; Gautam and Sharma, 2009). For example, Nisin is a bacteriocin produced by Lactococcus lactis spp., and is used worldwide against a wide variety of Gram-positive bacteria (e.g., lactic acid bacteria) or heat-resistant bacterial spores (e.g., Clostridium botulinum) (Brewer et al., 2002; López-Pedemonte et al., 2003; Sobrino-López and Martín-Belloso, 2006; Lucera et al., 2012). Lysozymes are another type of antimicrobial compounds used specifically against Gram-positive bacteria, because they act by hydrolyzing the murein layer (Cunningham et al., 1991). The examples we have described here only provide insight into the actions of single antimicrobial substances that have been tested in the laboratory settings or applied under controlled conditions. Environmental-based studies, including those for human bacterial pathogens, are scarce. The effectiveness of intrinsically produced antimicrobial compounds at ambient concentrations in a heterogeneous and complex aquatic environment, as well as the co-evolving development of resistance, remains unclear and a better understanding is urgently needed (see below).

An antagonistic interaction that is rarely considered is the predation of bacteria by other bacteria. Bacteria that share this feeding mode are commonly described as “Bdellovibrio and like organisms” (BALO), with Bdellovibrio being the most studied and best characterized organism in this group. BALOs can be found in several environments including soils, waters of various qualities, and in wastewater treatment plants (Martin, 2002). Bdellovibrio is a Gram-negative Deltaproteobacterium, known for invading the periplasm of other bacteria. Suitable prey comprise mostly Gram-negative, planktonic, or attached bacteria (Dashiff et al., 2010). Bdellovibrio attacks by entering the periplasm of its prey, where it forms a bdelloplast, septates, and finally lyses its prey, releasing progeny cells (Stolp and Starr, 1963; Rendulic et al., 2004; Lambert et al., 2006; Davidov and Jurkevitch, 2009; Sockett, 2009). The bdelloplast serves as a protective shield against phototoxic and chemical damage or phage attack (Friedberg, 1977; Markelova, 2002). Other, less extensively studied BALOs, like Micavibrio, Ensifer, Vampirococcus, or Daptobacter have evolved different feeding behaviors (Guerrero et al., 1986; Yair et al., 2003; Davidov et al., 2006; Dashiff et al., 2010). Micavibrio, for example, is an Alphaproteobacterium that attaches to the surface of various planktonic and sessile bacteria (Burkholderia, Enterobacter, Klebsiella, Pseudomonas, etc.) without entering the prey, but instead, works to exhaust it from the outside (Davidov et al., 2006; Dashiff et al., 2010; Koval et al., 2013). While the exact mechanisms are not yet clear, a transcriptome analysis has provided the first evidence for the involvement of porins that may facilitate the uptake of metabolites derived from the degrading prey cells (Wang et al., 2011). Another extracellular predation strategy is applied by Vampirococcus, where the predator attaches via cytoplasmic bridge structures to the cell membrane of Chromatium, a phototrophic purple sulfuric bacterium living in freshwater. Subsequently, the introduction of hydrolytic enzymes leads to the degradation of the prey’s cytoplasm and the ingestion of its contents (Guerrero et al., 1986; Martin, 2002). Another example is Daptobacter, a Gram-negative, facultative anaerobic freshwater bacterium, that is also endobiotic, meaning it resides and replicates within the cytoplasm of its prey (e.g., the phototrophic Chromaticeae) (Guerrero et al., 1986).

Given the aforementioned highlighted details, our understanding of bacterivorous bacteria is still far from complete. There is, for example, still no proof for how BALOs are attracted to suitable prey. Chemotaxis toward certain amino acids and attraction to high bacterial concentrations, prey or not, seem to play important roles (LaMarre et al., 1977; Straley and Conti, 1977; Rendulic et al., 2004); however, it is not clear at this point (i) how they identify and distinguish their Gram-negative prey from Gram-positive bacteria or particles, (ii) how they come into contact with their prey, or (iii) how they manage to survive changes in osmolarity or pH that would be prevalent when encountering the periplasm of their victims (Rendulic et al., 2004; Sockett, 2009). In particular, BALOs’ preferred temperature range is 18–30°C, which questions its activity in cold aquatic habitats, such as the deep sea or groundwater in temperate regions (Filip et al., 1991; Dashiff et al., 2010).

The ability for bacterivorous bacteria to significantly reduce the pathogenic bacteria in vitro and in vivo has raised high expectations. Dashiff et al. (2010) confirmed the activity of Bdellovibrio bacteriovorus and M. aeruginosavorus strains against several pathogenic bacterial genera, like Aeromonas, Burkholderia, Enterobacter, Salmonella, Shigella, Vibrio, and Yersinia. Moreover, a strong reduction potential for the predatory bacteria against multidrug-resistant Acinetobacter baumannii, E. coli, Klebsiella pneumoniae, P. putida, and P. aeruginosa was indicated by Kadouri et al. (2013). Two experimental treatments testing the application of bacterivorous bacteria have proven successful: (1) oral applications of B. bacteriovorus in chickens infected with S. enterica, and (2) their topical applications in cows suffering from Moraxella bovis infections (Atterbury et al., 2011; Boileau and Clinkenbeard, 2011). It is important to note that individual BALOs are tolerant, or even immune, to some toxic, antibiotic, or antiseptic agents due to the presence and activities of specific efflux pumps (Markelova, 2002). Even in human saliva, which acts as an antibacterial agent through protective and antimicrobial proteins (e.g., peroxidases, mucins, or lysozymes), some BALOs are able to retain their activities (Slowey et al., 1968; Van Nieuw Amerongen et al., 2004).

Studies addressing the application of bacterivorous bacteria in specific ecosystems are scarce, and little is known about the quantitative effects of bacterial predation on pathogens. Like other antagonistic processes, previous research was mainly restricted to the well-defined laboratory experiments rather than the field studies. In a microcosm experiment, the BALO Bacteriovorax was inoculated simultaneously with two pathogenic Vibrio species (V. vulnificus and V. parahaemolyticus) and the change in optical density (OD) was monitored over a period of 120 h (Figure 5). During this time, a constant decrease in OD was observed, indicating a reduction in the Vibrio strains. As depicted in Figure 5, Bacteriovorax revealed a higher preference toward V. parahaemolyticus (Chen et al., 2011). A successful application of BALOs in an aquaculture system was documented by Chu and Zhu (2010), where induced Aeromonas hydrophila infections in fishes were cured by the administration of B. bacteriovorus. Considering the very few environmental applications, a better and fundamental understanding of the role of BALOs in natural aquatic ecosystems is greatly desired.