The Diversity of Coliphages and Coliforms in Horse Feces Reveals a Complex Pattern of Ecological Interactions

Abstract

The diversity of coliphages and indigenous coliform strains (ICSs) simultaneously present in horse feces was investigated by culture-based and molecular methods. The richness of coliforms (as estimated by the Chao1 method) is about 1,000 individual ICSs distinguishable by genomic fingerprinting present in a single sample of feces. This unexpectedly high value indicates that some factor limits the competition of coliform bacteria in the horse gut microbial system. In contrast, the diversity of phages active against any selected ICS is generally limited to one to three viral genotypes present in the sample. The sensitivities of different ICSs to simultaneously present coliphages overlap only slightly; the phages isolated from the same sample on different ICSs are usually unrelated. As a result, the titers of phages in fecal extract as determined for different Escherichia coli strains and ICSs may differ by several orders of magnitude. Summarizing all the data, we propose that coliphage infection may provide a selection pressure that maintains the high level of coliform diversity, restricting the possibility of a few best competitors outgrowing other ICSs. We also observed high-magnitude temporal variations of coliphage titers as determined using an E. coli C600 test culture in the same animal during a 16-day period of monitoring. No correlation with total coliform count was observed. These results are in good agreement with our hypothesis.

Bacteriophages exert a significant influence on natural microbial communities (2, 24, 33). They are responsible for 20 to 80% of bacterial mortality in freshwater and marine ecosystems (25, 33) and increase bacterial biodiversity (references 14, 15, 31, and 33 and references therein) due to preferential attack on the dominant species or strains and redistribution of the organic matter. The role of bacteriophages could be even more important in microbial systems where high densities of active bacteria are achieved. Among these systems are intestinal microbial populations of animals and humans where bacteria (17, 28) and bacteriophages (5, 6, 9, 10, 13, 16, 17) are present at high densities. The gut is the natural habitat for Escherichia coli and for coliphages, which are highly suitable for work in culture, making this system attractive as a model for phage ecology.

For our study, we selected the horse as the macro host. The cellulolytic microbial community localized in the horse large intestine is very complex and includes bacteria, archaea, fungi, and protozoa (18). In contrast to rumen communities, the microbial biomass in the horse intestine is not subjected to digestion and is excreted with the feces. The conditions in the horse gut seem more stable than those in the intestines of many other species, as the time taken to digest grass is about 72 h (18), and the intervals between food intake and defecation are normally much shorter. A spatial complexity is present in the gut (9). The mucosal surface and the lumen contents are different ecological niches for bacteria. It has been shown in the mouse model that in the lumen, E. coli cells are less sensitive to externally administrated phages and may even starve (9, 23). However, a study of horse intestinal microflora (11) by rRNA gene sequencing revealed no differences in microbial composition on the mucosal surfaces and in the lumen or along the different parts of the large intestine.

Bacteriophage-like particles were first reported for the horse large intestine in 1970 (1). In our recent study (20), up to 69 morphological phage types were registered in a single specimen of horse feces. However, we repeatedly observed some particles that had identical dimensions and morphologies. The most abundant phage type had an unusual morphology, with an isometric head 100 nm in diameter and a very long (about 700-nm) flexible noncontractile tail. The fraction of these particles was about 10%. Recently, a metagenomic study of a viral community from equine feces was published (6). As is typical for this type of study, only about 20% of the sequences were known virus-related sequences, and among them, siphoviruses and myoviruses (bacteriophages with long noncontractile and contractile tails, respectively) predominated. The authors suggest that the total community may represent several hundred viral genomes. Similar results were reported earlier for a viral community from human feces, where the estimated complexity was approximately 1,200 viral genotypes (4). More data on total abundance of bacteriophages in equine fecal material are still needed, but judging by the yield of phage DNA from horse feces reported by Cann et al. (6), we can expect about 10¹⁰ to 10¹¹ phage particles per g of feces.

Studies involving the culturing of equine intestinal bacteriophages were pioneered by Felix d'Herelle. As early as 1921 (12), he reported that the vast majority of 62 samples of horse feces examined were positive for E. coli and Shigella phages. In later literature reports on the coliphages in equine feces (5, 13, 17), a considerable range was observed in the titers found in different individuals. In some animals, no phages were detected on conventional E. coli strains (e.g., C600), while in others, titers as high as 10⁷ PFU/g were found in the same host. No coherent explanations for such variations were offered. The main objective of this study is to investigate the relationship between coliphages and their bacterial hosts in the microbial community of the horse gut.

MATERIALS AND METHODS

Sampling.

Fecal samples were collected from four horses (one mare and three geldings 7 to 11 years old, referred to hereafter as animals A, B, C, and D) every 2 days for an 18-day period. At each date, a single large specimen (about 200 g) of freshly voided feces was collected from each animal. This sample was homogenized, and two averaged subsamples were taken: for bacteriophage enumeration, 20 g of feces was suspended in 80 ml of phage solution (0.2 M NaCl, 0.1 g sodium azide, 1 g Tween 20, and water [total volume, 1 liter]). The suspension was shaken in an orbital shaker for 1 h at 120 rpm at room temperature, aliquoted and centrifuged for 2 min at 12,000 rpm in a microcentrifuge, and used for plating either directly or with the appropriate dilution. For coliform CFU, a 0.9% NaCl solution was used instead of buffer, with a 20-min shaking time. No centrifugation step was performed before plating. Five milliliters of phage-containing extracts was supplemented with glycerol up to 15% (vol/vol) and stored at −70°C. An additional 50 g of feces was stored at −70°C for subsequent analysis and for isolation of additional sets of indigenous coliform strains (ICSs) when they were required.

Bacterial strains and cultivation of bacteria and phages.

As the reference test cultures, the following E. coli strains (Fermentas, Lithuania) were used: K-12 C600; NM522 (F′), BL21, and Be/1 (kindly provided by V. V. Mesyanzhinov; IBCh RAS, Moscow, Russia).

A standard Luria-Bertani (LB) medium was used for cultivation of all E. coli strains and ICSs. Phage enumerations in fecal extracts were performed by the double-layer plate titration method (26). Conventional solid LB was used for plates, and the same medium with 0.6% agar was used as a top agar.

To propagate the phages in liquid culture, 6 ml of LB broth was inoculated by a single colony from a fresh plate or, alternatively, by 1% (by volume) of an overnight culture of the appropriate strain and grown up to mid-log phase (optical density at 600 nm = 0.5 to 0.6) at 37°C with vigorous agitation. The culture was inoculated by adding a single plaque and grown until signs of the lysis were visible (the incubation time varied for different phage isolates). Then, a 0.5% volume of chloroform was added, and 1 to 5 h later, the lysates were centrifuged at 10,000 × g for 10 min to remove bacterial debris. For larger-scale preparation, 50 ml of liquid culture was infected with 200 μl of primary lysate and processed by the same protocol.

Coliform enumeration and isolation of ICSs.

To obtain ICSs from horse feces, 100-fold dilutions of the fecal extracts were plated on lauryl-tryptose agar (LTA) plates (tryptose [20.0 g; Difco], lactose [5.0 g], sodium chloride [5.0 g], dipotassium hydrogen phosphate [2.75 g], potassium dihydrogen phosphate [2.75 g], sodium lauryl sulfate [0.1 g], agar [15 g], and addition of water to achieve a total volume of 1 liter, pH 6.8 ± 0.2) selective for coliforms. From each sample, 50 individual bacterial colonies were transferred onto Endo agar plates (Difco). Eighty to 100% of them appeared as typical lactose-positive coliform colonies (dark red with a metallic sheen). A subset of 120 strains was tested for indole production with Kovac's reagent (Sigma); 96 of them were positive. We did not exclude from our study “atypical” isolates (lactose negative or producing no metallic sheen), because in the course of our experiments (data not shown), we observed that these isolates may share phages with “typical” isolates as well as with standard strains of E. coli. The mutations that extend the coliphage host range over new species of enterobacteria were demonstrated earlier (30; see also reference 12). So, for the purpose of this study, we name the whole set of bacteria able to grow on LTA plates ICSs; no attempts at further identification were made except for the strains that were used for phage isolation. These strains were analyzed by 16S rRNA gene sequencing (see below and Table 2). For technical reasons, we excluded from all experiments the rare isolates (about 5%) giving no PCR products in our genomic fingerprint test (see below). It also has to be stated here that in the boundaries of this work, we use the term coliphage in the sense of “coliform bacterium phage”.

TABLE 2.

Host ranges of bacteriophage isolates tested on selected sets of ICSs and on laboratory E. coli strains^a

Test strain	Result for bacteriophage isolated on indicated strain
	ICS							E. coli C600
	Sc5		Sc7	Sc8	Sc6	Sc4
	1c	2c	4c	5c	6c	7c	8c	A1	A2	B1	B2	C1	C3	C2	D1	D2	D3
Sc4	−	±	−	+	+	+	+	+	−	−	+	−	+	+	−	+	+
Sc5	+	+	−	−	−	+	−	−	−	−		−	−	−	−	−	−
Sc6	−	−	−	−	+	−	−	−	−	−	−	−	−	−	−	−	−
Sc7	−	−	+	−	−	−	−	−	−	−	−	−	−	−	−	−	−
Sc8	+	−	−	+	−	−	−	−	+	+	+	+	−	+	+	+	+
S1/7	−	−	−	−	−	+	+	−	−	−	−	+	+	+	−	+	+
S1/2	−	−	−	−	−	−	−	−	−	−	−	−	−	+	−	−	+
S4/2	−	−	−	±	−	+	+	−	−	−	−	−	+	+	−	−	+
S3/104	+	−	−	+	−	−	−	+	+	+	+	+	+	+	+	+	+
S3/117	−	−	−	−	−	−	−	+	+	−	+	−	−	−	+	−	−
S2/33	−	−	−	±	−	−	−	−	−	−	−	+	−	−	+	+	+
S2/34	+	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−
C600	+	−	+	±	−	−	−	+	+	+	+	+	+	+	+	+	+
Be/1	±	−	+	+	−	+	+	−	−	−	−	+	+	+	+	+	+
BL21	−	−	−	−	−	+	+	−	−	−	−	−	+	−	−	−	−
Z85	−	−	−	±	−	+	+	+	+	+	+	+	+	+	+	+	+

^aAll 12 ICSs mentioned here were identified by 16S rRNA gene sequencing as E. coli strains except for Sc5, which appears to be a Klebsilella sp. strain. +, good growth similar to that observed on the strain originally used to phage isolation; ±, weak growth observed only around the trace of the toothpick, with the plaque much smaller than that on the original strain; −, no plaque observed. Letter A, B, C, or D in the phage code indicates the animal source. Phages isolated on E. coli C600 were obtained from the samples from four animals and from of the same date.

Whole-cell PCR-based fingerprinting of ICSs.

Isolated ICS colonies grown on Endo plates were picked with sterile toothpicks and suspended in 100 μl of MilliQ sterile water. The cell suspension was heated for 5 min at 95°C and mixed well on a Vortex shaker, and the debris was pelleted in a microcentrifuge for 3 min at maximum speed; 1 μl of supernatant was used as the PCR template. The fingerprints were obtained with IS1tr (5′-ATCAGTAAGTTGGA [G/A] [T/G] CATTACC-3′), a single primer hybridizing to the terminal repeats of IS1 elements, with the 3′ end directed outward. Thus, the amplification occurs between closely located copies of IS1 and/or a number of potential hybridization sites outside IS1 elements that are detected in the E. coli genome by a BLAST search for short, nearly exact matches, most likely traces of lost insertion elements. The PCR master mix contained 67 mM Tris-HCl (pH 8.3), 17 mM (NH₄)₂SO₄, 0.001% Tween 20, 2.5 mM MgCl₂, 25 pmol of the IS1tr primer, and 1.25 units of Taq polymerase (Sigma). PCR was performed on an MJ mini thermal cycler (Bio-Rad). The program parameters were as follows: denaturation at 94°C for 30 s; 30 cycles consisting of 94°C for 15 s, 56°C for 30 s, and 72°C for 45 s; and a final extension step at 72°C for 2 min. The PCR products were analyzed by electrophoresis in a 1% agarose gel (26), and strain-specific band patterns (IS1 profiles) were observed.

Sequencing of 16S rRNA genes.

The strains used for phage isolations and phage host range analysis (Table 2) were examined. Genes of 16S rRNA were amplified using universal eubacterial primers 27F (5′-AGA GTT GAT CMT GGC TCA G) and 1492R (5′-GGY TAC CTT GTT ACG ACT T) as described in reference 21. PCR fragments were sequenced from the 27F primer.

Estimation of the strain richness of the coliform population.

To estimate the richness of horse fecal coliform communities, a nonparametric estimator, Chao1, was employed (8). The number of strains distinguishable by their IS1 profiles was estimated using EstimatesS software (http://viceroy.eeb.uconn.edu/EstimateS) with the bias-corrected formula for Chao1: S_Chao = S_obs + [F₁ × (F₁ − 1)]/ [2(F₂ + 1)], where S_obs is the number of different strains observed, F₁ the number of strains observed once, and F₂ is the number of strains observed two times.

DNA extraction and restriction fragment length polymorphism (RFLP) analysis.

The individual phage lysates, each propagated from a single plaque in LB broth, were DNase treated (0.01 mg/ml for 1 h at ambient temperature) and then sedimented by ultracentrifugation (90,000 × g for 1 h) in a bucket rotor. Phage pellets were carefully resuspended in SM buffer (50 mM Tris HCl [pH 7.6], 50 mM NaCl, 5 mM MgCl₂) and used for phenol extraction of phage DNA as described in reference 26. The isolated phage DNA was digested with restriction enzymes (Fermentas, Lithuania) EcoRV, EcoRV-HindIII, and DraI according to the manufacturer's recommendations and separated on a 1% or 1.5% Tris-acetate-EDTA agarose gel containing ethidium bromide (26).

Characterization of bacteriophages by sequencing of genomic fragments.

EcoRV-SspI or DraI digestions of DNA from one phage of each genotype listed in Table 1 were ethanol precipitated (26), and random clones were obtained using a T-system PCR cloning kit (Promega). The cloning procedure was accomplished as recommended by the manufacturer for blunt-ended PCR fragments. Several clones for each phage were sequenced from the standard M13F primer (Fermentas, Lithuania). In total, 1 to 2 kb of sequence was obtained per phage genotype. The sequences were analyzed by n-BLAST searches at www.ncbi.gov. For amplification of major head protein genes of bacteriophages related to coliphages JS98 and T5 (see Results), the following primers were used: Mzia1 (5′-GAT ATT TGI GGI GTT CAG CCI ATG A) and Mzia2 (5′-CGC GGT TGA TTT CCA GCA TGA TTT C) for JS98 (29) and T5-hdD (5′-TC TCT GGA AGG TCT GAC CG) and T5-hdR (5′-GC TGA GTA ACG TAG TAC GC) for T5. PCR conditions were the same as those described above for genomic profiling. Fragments were sequenced from the same primers.

TABLE 1.

Genotypes of the bacteriophages isolated on E. coli C600 and morphologies of the particles

Genotype	Subtype	Phage isolate code(s)	Head diam (nm)	Tail length (nm)	Family	Closest relative
I		C3	100 × 75b	100	Myoviridae	Coliphage JS98a
II	1	C1, D2	70	190	Siphoviridae	Coliphage T5
	2	C2, D3
	3	D1
III	1	A1	65	110	Myoviridae	Coliphage Felix 01
	2	A2
IV		B1, B2	65	110	Myoviridae	Coliphage Felix 01

^aRelated to T-even phage.

^bThis phage has an elongated head with a length of 100 nm and a width of 75 nm.

DNA sequencing.

All nucleotide sequences were determined using an automated sequencing device (Avant 3150; Applied Biosystems) with standard chemistry.

Electron microscopy.

A drop of phage suspension purified as for DNA extraction (see above) was applied to a Formvar-carbon-coated copper grid for 5 min and then removed by filter paper. The grids were stained with 1% uranyl acetate and examined using a Jeol 100S (Japan) microscope operated at ×25,000 magnification.

Nucleotide sequence accession numbers.

The sequences of genomic fragments of bacteriophages were deposited in GenBank under the following accession numbers: for type I, EF618556, EF618562, and EF618653; for type II, EF618557 and EF618560; for type III, EF618558, EF618564, and EF618566; and for type IV, EF618559 and EF618565. The accession numbers for the ICS 16S rRNA gene sequences are EF622097 to EF622110.

RESULTS

The dynamics of coliphages and their hosts.

We performed the initial phage monitoring experiment in duplicate with four horses, using a single test strain, E. coli C600, as a host. The observed range of phage titer variation was 2 to 3 orders of magnitude over the experimental period (Fig. 1). These data are in agreement with the results of Havelaar et al. (17), who observed a large variability of coliphage titers among feces of individual animals. For the same samples, we determined the titers of coliforms by duplicate plating on LTA plates. The curves are shown in Fig. 1. No apparent correlation between coliform number and coliphage titer in the same animal was observed. The most probable explanation for that is the high heterogeneity of the coliform population at the strain level.

FIG. 1.

Dynamics of coliforms (white circles) and coliphages (black dots) plated on E. coli C600 in animals A, B, C, and D.

We then tested 40 randomly chosen phage plaques obtained from strain C600 (10 from each animal from the same date) on 20 randomly chosen ICSs (5 per animal isolated from the same sample). Only 5 of our 20 ICSs were able to support the growth of any of the tested phages, and only 15 of 40 phages were able to grow on at least 1 of the tested ICSs. This indicates that the total population of coliforms in the horse gut is not accurately reflected by the subpopulation that supports the growth of phages counted and isolated from C600 lawns.

Genetic fingerprinting of coliform strains.

We developed a simple system for genomic differentiation of coliform strains by PCR with IS1tr, a single oligonucleotide primer that anneals to the inverted terminal repeats of the IS1 element, the most abundant insertion sequence in the E. coli genome (3, 27). The IS1 profile obtained for each strain is highly reproducible and specific; the reaction is robust and nonsensitive to minor variations of PCR conditions (data not shown). Coliform strains giving no amplification with this protocol are rare (about 5%) and were excluded from further experiments; most yielded 3 to 10 distinct bands (Fig. 2).

FIG. 2.

Examples of IS1 patterns of 10 ICSs isolated from animal B on day 4.

The majority of IS1 profiles were not related to each other, so any grouping of them is difficult. At the same time, the IS1 profiles are easy to distinguish by eye, in contrast to many other reported fingerprinting systems. The fingerprinting data are fully compatible with phage sensitivity examinations (see below), so this test is perfectly suitable for differentiating the coliform isolates obtained from horse feces.

A series of 30 randomly chosen ICSs from one sample from each animal (from day 4) was analyzed. Judging by IS1 PCR product patterns, all the strains of sample A were different, samples C and D each contained only one pair of identical strains, and sample B had two pairs of identical isolates. These data suggest an unexpected level of polymorphism in the coliform strains inhabiting the horse gut. To investigate this further, we analyzed an additional subset of 63 isolates from sample B. On the basis of data from 93 tested isolates, the Chao1 estimator predicts the total coliform community to contain more than 1,000 ICSs (range, 684 to 1,500; P < 0.05). We analyzed three additional samples from animal B on days 2, 6, and 60. For each sample, 30 ICSs were analyzed. No consistent dominant genotypes were observed.

Overlap of sensitivity of ICSs to simultaneously occurring phages.

We tested the phage plaque counts in the extract of a randomly chosen sample (day 4, animal C) on 10 genetically distinct ICSs isolated from the same sample. The phage titers ranged from less than 1 × 10² to 6 × 10⁵ PFU/g (mean, 1.7 × 10⁵ PFU/g); the titer in E. coli C600 was 10⁴ PFU/g.

For each ICS (except for the one that yielded no phages), we selected 10 phage plaques at random and repicked all these phages onto the lawns of all 10 strains. Only 31 phages from the 90 tested were able to form plaques on any strain besides the one used for its isolation. These phage isolates were divided into 15 host range types: 7 phage types that grow on their original strains only, 6 types that are able to lyse one additional strain, and 2 that propagate on two additional strains (data not shown). If we consider the plating of each phage type as an independent test, the frequency of a negative test result (F_neg, where no suitable host is found among nine examined strains) in our case can be calculated as F_neg = 7/(7 + 6 + 2) = 7/15. If we define P as the probability that a random ICS is the host for the phage under examination, the basic theory of probability allows us to calculate the probability of such a negative test result from the equation F_neg = (1 − P)ⁿ, where n is the number of independent trials (in our case, n = 9). The P value found from this equation is 0.08. It means that on average, a randomly chosen coliphage from the sample would lyse about 8% of the ICSs present in the community. To test this estimation directly, we carried out an additional experiment. Two genetically distinct phages isolated from animal B (on days 0 and 6, using ICS isolated from the same animal at day 6) were used to prepare phage agar (about 10⁹ PFU were added to the top layer, and no test culture was added). Two hundred thirty individual colonies of ICSs from days 0 and 6 from the same animal were transferred by toothpicks onto both phage agar plates and the Endo plate. In all cases, four to nine sensitive colonies, corresponding to 2 to 4% of the strains shown to be sensitive to each phage, were recovered.

We also examined five pairs of ICSs that appear identical in our genomic profiling assay, obtained from three different samples. The titers of the phages in fecal extracts plated on lawns of these strains were almost identical, and all individual plaques obtained on such indicator cultures could be successfully replated to the strains that have the same genomic profile (data not shown).

Diversity of coliphages.

As we have shown, the coliphages able to grow on a given strain can differ significantly in population size and may include several distinct host range types. We have studied by RFLP analysis the DNA from 10 phage isolates obtained from E. coli C600 from four animals on the same day (Fig. 3). Based on DraI digestion (data not shown), genotypes II and III may be subdivided into subtypes (Table 1). The virion morphology of each RFLP type is shown in Fig. 3. For each genotype, the relations with known coliphages were determined by sequencing genome fragments (see Materials and Methods). In all cases, close relatives (>92% nucleotide identity) were identified. The results are shown in Table 1. Types I and II were shown to be related to coliphages JS98 (pseudo-T-even phage) and T5, respectively. Types III and IV both are closely related to the coliphage Felix 01; however, since their RFLP profiles differ significantly, two distinct lineages of Felix 01-related bacteriophages are present in the phage pool of tested animals. The identifications were confirmed by amplification and sequencing of major capsid protein genes from all six isolates attributed to types I and II. It is interesting that all 10 phage isolates identified here belong to virulent phages. A rather frequent occurrence of identical or closely related phages in small sample groups indicates a limited level of diversity among C600-specific phages in the sample. In contrast, phages isolated from the same fecal specimen but obtained on different host strains are usually not related (Fig. 4). This conclusion is confirmed by an RFLP study of five series of six phages isolated from three animals on five ICSs (obtained from the same samples and distinct by IS1 profiles). In all cases, the phages within each series represented a single genotype (sometimes with few subtypes), unrelated to the genotypes of other series (data not shown).

FIG. 3.

RFLP analysis of DNA of phages isolated from E. coli C600. (A) EcoRV digestion. The letter in the isolate code indicates the animal host. M, DNA size marker. (B) Morphology of C600 phages of all identified RFLP types (Table 1). C3, type I; D1, type II; A2, type III; and B1, type IV. The morphologies of phages of different subtypes within one genotype are identical (not shown). Bar, 100 nm.

FIG. 4.

RFLP analysis of DNA of phages isolated from five different ICSs from a fecal sample from animal C. EcoRV enzyme digestion is shown. Isolates on lines 1 and 2 were obtained from the same strain, isolates on lines 6 and 7 were obtained from another, and isolates on lines 3, 4, and 5 were each obtained from a particular ICS.

It is interesting to note that all the bacteriophages studied (except for the serial isolates mentioned above that were not tested) had different host specificities, as determined with a panel of laboratory E. coli strains and indigenous strains (Table 2). The ICSs used for host range tests were analyzed by 16S rRNA gene sequencing. All of them were identified by their closest GenBank homologs as E. coli strains except for Sc5, which appears to be a Klebsiella sp. strain. It is interesting that the 1c and 7c phages infecting strain Sc5 are also able to grow on some laboratory E. coli strains and field isolates (Table 2). These results indicate a considerable genetic plasticity in the bacteriophage host range determinants in this microbial community.

DISCUSSION

A community of bacteria and their phages requires some control mechanisms to maintain the high densities and balance among its constituents. Recent studies on phage infection in benthic environments (2, 14, 15) indicate that despite the presence of a large number of viral particles, the impact of viral infection on bacterial mortality is quite modest (1 to 6%) compared to the mortality in the overlying water column (more than 60%). One explanation of this “infection paradox” (14) is the significantly higher level of diversity of bacteria in the benthos than in the surface waters. Substantial genetic diversity in E. coli populations has been reported for environmental niches such as feedlots (35) as well as for the feces of humans and animals (7, 19).

Based on the data present here, we make three main conclusions. (i) The system that we have developed for high-resolution genomic profiling of intestinal coliform strains allows the differentiation of genetically close strains, which may, however, differ in their phage sensitivities. This has allowed us to analyze the diversity of equine fecal coliforms and to investigate their relationship with respect to phage sensitivity. The diversity of the coliform population in horse feces can exceed 1,000 strains. This result suggests that there is some factor that prevents the displacement of the majority of ICSs by a few competitors most fit for the ecological niches of the gut. (ii) The divergence of the phages that can form plaques on the lawn of any ICS is quite limited, which probably reflects the competition for available host cells. On the other hand, the different ICSs support growth of genetically unrelated phages in the same sample. (iii) The host range overlap of coliphages isolated from individual host strains present in the same sample is quite limited. This result is in agreement with observations made for freshwater coliphages and coliforms (22).

Based on the second and third conclusions, we suggest that the individual coliform strains (and their phages) composing the population are at least partially independent of the larger population. Summarizing all the data, we speculate that under the conditions of the horse gut, the coliphages are able to efficiently control the populations of their hosts, to select for phage-resistant variants, and thus to maintain a very high level of diversity of coliforms at the strain level. The transient dominance of some bacterial strains (presumably good competitors) may occur, but this is limited by the accumulation of specific phages. The observed temporal fluctuations of the phage titers determined for various strains (20; also this work) could reflect such predator-prey waves. To prove this suggestion rigorously, the simultaneous monitoring of the temporal variations in the abundance of an individual indigenous strain and its specific phages is required.

Obviously, various ecological factors in the gut besides host cell availability will influence phage infection. For example, environmental sequestering could protect the bacteria from phage attack (by adsorption to the mucosal surface or to nondigested food particles, biofilm formation, or inhibition of phage adsorption by bile salts). Alternatively, a fraction of the host population could be in a physiological state that is unsuitable for phage multiplication (e.g., cells in stationary phase). These factors may have a substantial influence on the determination of the threshold concentrations of any particular virus and host that can coexist stably in this ecological system.

The isolation of phages that are almost identical by their morphologies and restriction profiles from different individuals held in one stable suggests that animals from our experimental group may contribute to a common phage pool. The fact that the phage isolates closely related by RFLP differ in their host ranges may reflect an intensive selection for new host range variants. The abilities of phages to adapt to a new host presumably involve point mutations as well as recombination events of a larger scale. Also, the effect of formation of small and unstable fractions of virus particles of some phages that have extended ranges of host adsorption, known as “nascent phage quality” (32, 34), may contribute to phage host range variability.

Our observations suggest that the mechanisms maintaining the diversity of the E. coli population in the gut may be responsible for the inefficiency of the introduction of probiotic strains when they are used in humans or animals.