Phage host range
The host specificity of bacteriophages is generally quite narrow, restricting any particular phage to a defined set of bacterial strains; yet in some cases bacteriophages can infect several host species. D’Herelle (1921) first suggested that propagation of Yersinia pestis phages isolated by him from rat faeces 3 months after the end of a plague outbreak could be explained by multiplication of these viruses in other bacterial species. Comeau et al. (2005) observed the persistence of vibriophages in oysters in the absence of detectable hosts (in winter) and suggested the possibility of vibriophage multiplication on different host species. So, multi-species coverage by some phages may have certain ecological significances in animal-associated microflora. Nevertheless, even closely related host strains may differ significantly by their phage sensitivity. This means that the impact of phage infection on populations of different species or even different strains of bacteria populating the same site of the animal body may vary dramatically. Culture-based approaches are today perhaps the only methodology that allows the analysis of ecological interactions in microbial community at bacterial strain-level resolution.
Numerous reports of phage isolation on cultivable micro-organisms have been published, and an extensive review of this literature is out of the scope of this paper. Briefly, the phages of Streptococcus bovis, S. durans, Prevotella bryantii and Bifidobacterium ruminale were isolated from rumenal contents of sheep and cattle (reviewed in Tarakanov 2006); the faeces of humans and different animals were shown to contain phages of E. coli, Salmonella, Bacteroides, Klebsiella and other bacteria (Dhillon et al. 1976; Furuse et al. 1983; Havelaar et al. 1986; Cornax et al. 1994; Grabow et al. 1995; Calci et al. 1998; Gantzer et al. 2002; Schaper et al. 2002; Cole et al. 2003; Lusiak-Szelachowska et al. 2006; see also references in these articles). In the vast majority of these studies, the isolation of bacteriophages from animals was performed on laboratory bacterial strains, or on previously existing/characterized environmental isolates. This approach is logical in numerous studies, where phages are considered the indicators of faecal pollution of water.
The occurrence and titres of DNA-containing coliphages can vary significantly between animal species and individuals of the same species, which is in agreement with data indicating that total viral communities are highly individual, at least in the rumen of sheep and cattle (Klieve and Swain 1993; Swain et al. 1996b). At the same time, there is no indication that the particular types of DNA-containing phages are specifically associated with any animal species. By contrast, the occurrence of RNA-containing F-specific (F-RNA) coliphages (members of the Leviviridae family) in animal waste shows certain species specificity. These phages can be subdivided into four genetic groups that can be also distinguished serologically. The incidence of these serotypes varies significantly between species: horse faeces, for example, rarely contain them, while the faeces of more than 70% of chicken contain high titres of these phages (105–107 PFU g−1). Only about 10–20% of human faeces contain F-RNA coliphages, but the occurrence of group II and III F-RNA phages in these samples is much higher (>80% of all isolates) than in animals, where groups I and IV are prevalent at the same level (Furuse et al. 1978; Havelaar et al. 1986; Schaper et al. 2002; Cole et al. 2003). No coherent explanation for this group specificity has been suggested yet.
The possible association of some genetic subgroups within established genera of the tailed phages may be obscured by the fact that no simple technique for selective isolation of related phages from environmental samples has yet been developed. Recently, however, Chibani-Chenoufi et al. (2004a) reported an almost selective isolation of T4-related phages from the stools of paediatric diarrhoea patients in Bangladesh, when E. coli K803 was used as the host. The same samples plated on an E. coli O127:K63 enteropathogenic strain yielded a totally different set of phages, all of them belonging to the Siphoviridae family.
Culturing indigenous GI tract phages
Furuse et al. (1983) found that faecal coliphage titres in healthy humans are low, and the pools of free virions in faeces are represented mainly by temperate phages. These results suggest that in healthy people most of the released phage particles are produced by the induced lysogenic cells and therefore that phage multiplication may have a limited impact on intestinal coliform microflora. In contrast to healthy people, the phage populations in some patients with internal and leukaemic diseases contained a substantial fraction of virulent phages as well as an increased faecal coliphage background (Furuse et al. 1983). In several patients, phage titres did increase when the severity of the clinical symptoms increased.
In good agreement with the data of Furuse et al. (1986), attempts to isolate bacteriophages from dog faeces using indigenous coliform strains (ICSs) made by Ricca and Cooney (2000) were largely unsuccessful. Over 500 ICSs isolated from six specimens of dogs kept in private homes did not detect phages in the same samples, and only one of these samples yielded phages on the laboratory E. coli C strain. However, in 16 dogs from a kennel, somatic coliphages were detected at variable titres from 0 to 107 PFU g−1. The authors suggested that a low abundance of coliphages in home-kept dogs may be due to the isolation from other dogs and ‘too-clean’ living conditions. The recontamination by faecal microbes is also limited in humans that could partially explain the results (Furuse et al. 1986) mentioned above.
Our recent results obtained on horse model (Golomidova et al. 2007; Kulikov et al. 2007) are nearly opposite to the findings of Furuse et al. (1986). In horses, the cellulolytic microbial community localized in the large intestine is very complex and includes bacteria, archaea, fungi and protozoa. In contrast to rumen communities, the microbial biomass in the horse intestine is not subjected to digestion and is excreted with the faeces. The conditions in the horse gut seem to be more stable than in many other species, as the time required to digest grass is about 72 h – much longer, than the intervals between feeding and defecations (Hintz and Cymbaluk 1994). To investigate the relationships between coliphages and their hosts in this system, we followed the dynamics of coliphage titres on an E. coli C600 test strain. Furuse et al. (1986) did not observe in nine serial samples from 19 healthy humans at 2-week intervals any large temporal variations in either coliphage titres or diversity. In horses we observed high-magnitude fluctuations (2–4 log units in different animals over 16 days). At the same time, the titres of total coliform bacteria were much more stable and oscillate around 5 × 105 CFU g−1. This difference may reflect higher impact of lytic phage infection on coliform bacteria ecology in equine gut compared to human GI tract (see also above).
We noticed no direct correlation between coliform and coliphage titres. This result may be explained by the high strain diversity of intestinal coliform phage hosts (Yoshida et al. 2007; see also Muniesa et al. 2003). To differentiate the closely related coliform bacterial strains potentially possessing different phage sensitivity, a simple whole-cell PCR-based system for enterobacteria genomic fingerprinting was created (IS1 profiling; Golomidova et al. 2007). Using this system, we found that the coliform bacterial pool in the equine gut includes hundreds of individual strains (up to 1500 by Chao1 estimation). We also estimated that only 2–8% of ICSs present at any moment may be infected by a single purified phage isolated from the same sample. The diversity of bacteriophages in faecal samples active against any given ICS (obtained from the same sample) is very low, in most cases only one to two genotypes distinguishable by RFLP. The diversity of phages detected on E. coli C600 lawns was limited to one to three phage RFLP types in any studied sample. This presumably indicates competition between viruses for available host cells.
Phage impact on GI tract bacteria
Microbial biomass may contribute up to 54% of the total weight of human faeces (Stephen and Cummings 1980). Coliform bacteria (>80% of them generally are E. coli) are normally present at 105–108 CFU g−1 in human and animal faeces (Havelaar et al. 1986). Streptococcus bovis is present at 106–107 CFU ml−1 of sheep rumenal fluid (Iverson and Mills 1977; Tarakanov 2006). These cell densities are well above the reported threshold level of about 104 cells ml−1 for exponential phage growth (Wiggins and Alexander 1985). Thus, some mechanisms stabilizing the coexistence of phages and their hosts have to be present.
We propose that coliphage infection may provide a selection pressure that maintains high levels of coliform diversity in the horse gut, restricting the possibility for a few ‘best-fit’ competitors to outgrow other ICSs. On the other hand, this diversity limits the availability of the host for any particular coliphage, thus stabilizing the system. The high complexity and network-like organization of the community makes it difficult to follow experimentally the dynamics of specific phage strains and their hosts, complicating a direct test of this hypothesis. However, we recently studied a case of antibiotic-induced dysbiosis in a horse given enrofloxacine orally for an extended time during the treatment of a serious wound. This treatment caused a dramatic decrease of total coliform and coliphage diversity while the total counts of coliform bacteria were rapidly re-established to normal levels. We are in the process of using these antibiotic-treated animals as a model system. The investigation of microflora establishment in the newborns may be another good approach to this problem.
The observed fine structure of the equine intestinal coliform–coliphage community is consistent with the results of mathematical and experimental modelling of phage–host communities in which the co-evolution of both components is allowed (Weitz et al. 2005; Poullain et al. 2008). The diversification of the host population into multiple lineages with different susceptibilities to the phages was observed. The selection of the phage strains that had a tendency of extending their host range at the price of reduced adsorption efficiency was also demonstrated. Thus, the complexity of the two-component community dramatically increased (Poullain et al. 2008). A similar ecological situation was recently described by Holmfeldt et al. (2007) in a marine ecosystem. Using an approach based on the combination of PCR fingerprinting of bacteria, phage RFLP and host range analysis, the authors demonstrated a large variability in abundance and phage host range in flavobacteria and their phages in coastal sea water.
Bacteriophage activity may also be involved in the microfloral succession that can take place in the equine gut upon carbohydrate overload. Such an experimental overload by inuline is commonly used as a model of nutrition-induced laminitis. It was proposed that some toxins of intestinal microflora are liberated into the bloodstream and cause direct or indirect damage to the basal membrane of hoof-producing epithelium. Recently it was shown (Milinovich et al. 2006) that the marked numeric increase in streptococci of S. bovis/equinus complex occurs 8–16 h after oligofructan challenge, however the viability of these bacteria is then rapidly diminished. A possible explanation is phage-induced bacterial killing as equivalent to reports of control of algal blooms by viruses (reviewed in Weinbauer 2004).
The high impact of bacteriophages naturally occurring in the large intestine on populations of Campylobacter jejuni was demonstrated recently in chickens. The incidence of C. jejuni phages in the chicken caecum was shown to correlate negatively with the colonization level of the host (mean 105 CFU g−1 in the samples containing phages vs 107 CFU g−1 in phage-free samples; Atterbury et al. 2005). Later it was shown that the presence of bacteriophages selects for C. jejuni variants with large-scale genomic rearrangements that occur via lateral gene transfer (Scott et al. 2007a) or as a result of intragenomic inversions between Mu-like prophage elements (Scott et al. 2007b). In both cases, bacteriophage-sensitive variants had a significant competitive advantage over the resistant ones in colonization of broiler chickens. In the absence of phages, the population rapidly reverted to a sensitive phenotype. It was suggested that genomic instability of C. jejuni in the avian gut serves as an adaptive mechanism to temporarily survive bacteriophage predation and subsequent competition for resources.
Impact of externally added phages on GI tract bacteria
The effective elimination of bacterial pathogen populations from the gut by bacteriophages has been convincingly demonstrated many times in phage therapy experiments (extensively reviewed in Sulakvelidze and Barrow 2005; Sulakvelidze and Kutter 2005). This may involve, however, simply reducing the pathogen to low-enough levels that other host mechanisms can keep bacterial growth under control. It was shown, in addition, that phage doses as low as 102 PFU may prevent development of infection in animals upon experimental inoculation by pathogenic E. coli (reviewed in Brussow 2005), though this low dosage of phage is much less effective if administrated after inoculation by bacteria. This suggests that the pathogen population is more susceptible to phage attack before it colonizes a specific ecological niche in the gut.
The impact of externally added phage on the resident and introduced E. coli populations in mice was studied by Chibani-Chenoufi et al. (2004b). The mice used in these experiments continuously excreted E. coli at a titre of about 106 CFU g−1 of faeces. Some temporal variation in E. coli counts was observed in individual animals, but no naturally occurring phages were detected on E. coli K803 indicator strain, sensitive to all of the phages used in subsequent experiments. The mice were exposed to up to 107 PFU ml−1 of phage cocktail added to drinking water. These mixed phages were shown to be active in vitro (in spot tests) against nearly 100% of the resident E. coli isolates obtained from the same animals before treatment. Nevertheless, the influence of external phage administration on the faecal coliform counts was quite mild, and phages could not propagate continuously on the resident host population. Escherichia coli isolates obtained during the treatment showed the same pattern of sensitivity to the components of the phage cocktail under cultivation conditions as before treatment, ruling out the replacement of the resident E. coli population by phage-resistant strains or other enterobacteria. At the same time, their selected phages were highly effective in reducing the populations of sensitive E. coli strains introduced to germ-free mice a week before phage application, and efficient phage replication in the gut was demonstrated. However, the residual E. coli population remained sensitive to phages, suggesting the presence of some shelters in the gut of mice where E. coli may be physically or physiologically protected from the phages. As almost the whole population in conventional mice was found to be protected, these ‘shelters’ appear to cover the majority of ecological niches suitable for E. coli replication in mice. Interestingly, the mice purchased by this group from the same breeder 1 year later contained almost no E. coli and were colonized by another species of enterobacteria (Brüssow, personal communication). Consistent with these observations, Kasman (2005) found that less than 20% of 48 surveyed mice carried E. coli in their GI tracts, with coliphages present only in minute amounts. The author concluded that the main barrier in mice for widespread coliphage infection of resident E. coli is due to lack of sufficient host populations.
Environmental factors may contribute to the protection of bacteria in the gut. Bile salts and carbohydrates were shown, for example, to inhibit adsorption of a variety of coliphages (Gabig et al. 2002). This effect is suppressed if Ag43 protein, mediating cell aggregation and attachment, is present on the surface of the bacteria. The expression of Ag43 is regulated in a phase-variation manner. Thus, phages may in some circumstances select against increased biofilm formation. It has to be mentioned, however, that for Bacteroides phages the addition of the bile to the medium had an opposite effect. That is, its addition improved phage plating efficiency (Araujo et al. 2001).
Growth in biofilms on the surfaces of mucosa and food particles may also contribute to bacterial anti-phage protection. There is also some evidence that the E. coli population in the mouse gut lumen is starving and thus represents poor hosts for bacteriophages (Poulsen et al. 1995). The actively-replicating population in fact may be limited to microcolonies found on the mucosal surface (Krogfelt et al. 1993). Both studies, however, employed streptomycin-treated mice artificially inoculated by a single E. coli strain, BJ4, of known phenotype. In addition, the physiology of the digestive tract in mice is not identical to that of larger animals. In such a small animal, the rates of metabolism are much higher, and the time of food passage through the gut is short. The mucosal surface to volume ratio in the intestine is also much higher than in big animals, which facilitates rapid nutrient uptake. An important question therefore is to what degree results from mouse models may be extrapolated to larger animals.
Body regions of low phage impact
Data suggest that phages in the gut, at least in some species and/or conditions, manage to overcome physical and chemical barriers and represent a limiting factor for the host bacterial population. Considering the open nature of the gut for matter exchange with the external environment, a high influence of phages is expected. The situation seems to be somewhat different however in other animal body cavities heavily populated by bacteria. Hitch et al. (2004), for example, failed to isolate bacteriophages from the oral cavity for indigenous bacteria, making a conclusion that this community is not significantly impacted by phages. No coherent explanations of these observations were suggested.
In the rumenal microbial system, some bacterial species persist at high population levels. Streptococcus bovis may be present at 106–107 CFU ml−1 (Iverson and Mills 1977; Tarakanov 2006). As was demonstrated by Iverson and Mills (1977), the whole population of S. bovis in sheep may belong to a single phage-sensitivity type. The authors reported rapid changes of one dominant strain for another, but they failed to detect any S. bovis-specific phages in the rumenal fluid. The presence of S. bovis phages was, however, reported in the rumen (Stryiak et al. 1989) at concentrations 101–5 × 104 PFU ml−1, varying between individuals.
The ecology of rumenal streptophages was extensively studied by Tarakanov et al. in the 1970–80s. Unfortunately, their results are published in rare and poorly available journals, or even as national database depositions written only in Russian. Here we refer to the book by Tarakanov (2006) where these studies are reviewed in great detail. They report concentrations of S. bovis phages at 101–104 PFU ml−1 and demonstrated in a highly controlled field study that administration of an S. bovis phage preparation (streptophagin) in cattle led to an increase of the free phage concentration by four to five orders of magnitude, coupled with a significant decrease of amilolytic bacteria counts, decrease of the total amilolytic activity in the rumen and increase of the cellulolytic activity. These changes exerted some favourable influence on animal productivity that was rigorously controlled and documented. Phage titres subsequently dropped rapidly to the level detected before phage administration and the counts of amilolytic bacteria rose. These results indicate that single phage preparations impacted a significant fraction of rumen S. bovis populations in experimental animals which is in good agreement with the results of Iverson and Mills (1977), suggesting a low level of intraspecific diversity of this bacterium in the rumen.
It has to be suggested that there is some factor that prevents effective multiplication of the Streptococcus phages on the resident host population in the rumen. Swain et al. (1996a) demonstrated that tannic acid at physiological concentrations may inhibit bacteriophage replication, precipitating the free-phage particles. Tarakanov (2006) has shown that the rumenal fluid of the cow and acetic acid at physiological concentrations may inactivate S. bovis phages. The sensitivity of different phages also varied significantly and depended on the concentration used. The inhibition of phage activity by various chemical compounds in the rumen thus may play a key role in limiting phage lysis of resident bacteria. This model is consistent with the conclusions of Wells and Russel (1996) that mass lysis of the microbial biomass in the rumen, which has a negative impact on the productivity of the animals, happens mostly due to bacterial autolysis rather than to phage-induced lysis.
The human vagina is also a site extensively colonized by bacteria. The most significant part of the vaginal microflora is comprised by lactobacilli. The species and strains do vary in the human population (Antonio et al. 1999), but normally in any particular individual one to two strains of lactobacilli heavily dominate (Antonio and Hillier 2003). The density of colonization is quite high (106–107 CFU per vaginal swab). Thus, one can assume that episodes of mass killing of lactobacilli by phages may occur, but all attempts to detect free bacteriophages in vaginal swabs were unsuccessful (Kiliçet al. 2001). At the same time, lysogenic strains of lactobacilli were shown to be prevalent in this environment (Kiliçet al. 2001; J. Suarez, personal communication).
Though the vagina is less subjected to exchange of bacteria and viruses with the external environment, except as mediated by sexual activity, a sudden breakdown of Lactobacillus populations is frequently observed in examinations of vaginal swabs from clinically healthy women. Some individuals also develop anaerobic bacterial vaginosis syndrome, when lactobacilli became replaced by anaerobic bacteria such as Gardnerella vaginalis, Prevotella, Porphyromonas and Mobiluncus species. This condition has an epidemiology of a sexually transmitted disease (reviewed in Blackwell 1999). It was proposed by Blackwell (1999) that the causative agent triggering the breakdown of the normal vaginal microbiota may be a bacteriophage attack. We speculate that such event may take place if the lysogenic strain able to produce a phage infectious for the major resident strain(s) is acquired.
The frequency of lysogenization of host cells during infection by temperate phages is normally very low (10−7–10−2; Kihara et al. 2001; Brown et al. 2006; Riipinen et al. 2007), so the multiplication of such a virus in a niche, densely populated by one or few susceptible strains, would lead to killing off the majority of the host cells that could be subsequently replaced by the lysogenic strain which initially liberated the phage. This scenario was modelled both mathematically and experimentally (Brown et al. 2006) on E. coli populations. Taking into consideration that the rate of spontaneous phage production is less than 10−8 PFU per cell (Kiliçet al. 2001) in lysogenic vaginal lactobacilli cultures, any factors contributing to lysogen induction may increase the probability of triggering an ‘ecological catastrophe’ in the vaginal microflora. This may also explain the epidemiological link between anaerobic bacterial vaginosis and tobacco smoking (reviewed in Blackwell 1999) as components of smoke were shown to increase phage production in lysogenic lactobacilli cultures (Pavlova and Tao 2000). Detailed investigations of temporal dynamics of Lactobacillus strains in the vagina may allow testing of this hypothesis.