The endogenous human gut microbiome has several important functions including nourishment, the training of innate immunity and the regulation of epithelial development [1]. Although the Escherichia coli population represents a rather small portion of the intestinal bacterial microflora, E. coli nonetheless occupy an important niche with regard to their close proximity to intestinal epithelium, wherein they utilize available oxygen and facilitate anaerobic growth [2]. Intestinal microflora also prevent the growth of pathogenic bacteria, either by competing for nutrient sources, or through direct bacterial antagonism mediated by bacteriocins and bacteriophages [3].

E. coli is a highly diverse species with respect to its gene content, phenotype and virulence [4]. Based on different virulence factors, E. coli strains can be classified into three main groups: commensal, intestinal pathogenic and extraintestinal pathogenic E. coli (ExPEC) [5]. Commensal strains are commonly considered to be non-pathogenic. It has been shown that intestinal and extraintestinal pathogenic E. coli strains can develop from commensal strains by acquisition of virulence factors [6, 7].

Intestinal pathogenic (diarrhea-associated) E. coli is a typical mucosal pathogen which uses different pathogenic strategies including invasion of host cells (enteroinvasive E. coli, EIEC), production of enterotoxins (enterotoxigenic E. coli, ETEC) and production of Shiga-like toxins (enterohemorrhagic E. coli, EHEC) [8]. Enteropathogenic E. coli (EPEC) strains cause attaching-and-effacing (A/E) lesions and harbor the EAF plasmid [8]. Diffuse-adherent strains of E. coli (DAEC) are characterized by continuous adherence to eukaryotic cells mediated by afimbrial adhesins [9], while entero-aggregative (EAggEC) strains produce an aggregative adherence (AA) pattern [10] when adhering to HEp-2 cells.

ExPEC strains carry different combinations of virulence factors. Johnson et al. (2003) defined ExPEC strains as those possessing 2 or more of the following virulence factors: P fimbriae, S/F1C fimbriae subunits, Dr-antigen binding adhesins, aerobactin receptor and group 2 capsule synthesis [11].

Another important characteristic of human E. coli strains is production of bacteriocins. Colicins and microcins are antimicrobial agents with a relatively narrow spectrum of activity [12–14]. In general, microcins are known to have a wider spectra of antibacterial activity compared to colicins [14, 15]. Colicin Js [16, 17] is unique in that it shares features of both colicins and microcins. The ecological role and molecular evolution of bacteriocinogeny are less clear but synthesis of bacteriocins may have both invasive and defensive functions in microbial communities [18]. Bacteriocin production is a common feature of many probiotic bacteria. For example, microcins H47 and M are active substances produced by the non-pathogenic, probiotic E. coli strain Nissle 1917 [19].

At the same time, several studies have shown an association between the production of some types of bacteriocins and pathogenic E. coli strains [20–23]. Previous studies have found that genes encoding colicin E1, as well as microcins H47, M, I47, E492 and V were associated with extraintestinal pathogenic E. coli strains [20–23]. Colicin E1 is also known to have toxic effects on eukaryotic cells and is considered to be a virulence factor in UPEC strains [21, 24, 25]. Microcins H47 and M are induced when iron is a limiting factor and are associated with competition for iron [22, 26]. Synthesis of microcin H47 and M could therefore be advantageous in bacteremia of urinary tract origin [22, 26].

Previously published studies have only provided partial insight into the association between bacteriocin production and bacterial virulence, as they were primarily focused upon UPEC strains and differed in the number of detected bacteriocin and virulence genes. Azpiroz et al. (2009) screened 5 microcin types in 125 UPEC strains and 9 virulence factors [20], while Budič et al. (2011) and Petkovšek et al. (2012) analyzed 14 virulence factors (primarily those associated with urinary tract infections) and 19 bacteriocin types in 105 UPEC strains [22, 23]. Similarly, Abraham et al. (2012) analyzed 16 bacteriocin types and 18 virulence factors in a collection of 159 UPEC strains [27]. Together, these studies identified an association between microcins (M, H47, V, B17 and L) and several virulence genes [20, 22, 23, 27]. Studies by Gordon et al. (2005) and Gordon and O’Brien (2006) focused on 266 fecal E. coli strains and identified an association between strains encoding at least one microcin type, and four genes involved in iron acquisition (from a total of 29 tested virulence determinants) [26, 28].

The main aim of our study was to test the association between bacteriocin production and bacterial virulence within a large collection of E. coli strains (n = 1181) isolated from human gut microflora. In this study, new associations between bacteriocin-encoding genes and virulence determinants were found in human fecal E. coli strains.

In this study, the average prevalence of bacteriocinogenic E. coli strains isolated from fecal microflora was 54.4%. This value is close to the upper range limit seen in previous studies, where the prevalence of bacteriocinogenic E. coli strains varied from 25 to 55% [15, 21, 26, 27, 29–31]. However, these studies differed in a number of important ways including cultivation conditions and indicator bacteria used for detection of bacteriocin production and/or in the number of detected bacteriocin genes. Older studies on the prevalence of bacteriocinogeny in fecal E. coli strains only focused on the identification of colicin production [30, 32]. While Šmarda and Obdržálek (2001) used five different indicator strains to detect colicin production in the fecal E. coli strain 1043 [32], Achtman et al. (1983) used 2 indicator strains for the detection of colicin producers in a sample of 234 fecal E. coli strains [30]. More recently, Gordon and O’Brien (2006) used PCR with 19 bacteriocin genes to screen 266 fecal E. coli strains (38% of which were bacteriocinogenic) [26], and Šmajs et al. (2010) detected 29 bacteriocin types in 411 fecal E. coli strains (55% of which were bacteriocin-encoding strains) [21].

Our results have revealed that the frequency of bacteriocinogeny in E. coli strains positively correlates with the detected number of virulence determinants. Bacteriocinogeny increased by as much as 75–80% depending on the number of encoded virulence factors. To our knowledge, this is the first time that the correlation between bacteriocinogeny frequency and the number of encoded virulence factors has been shown. This finding suggests that at least some bacteriocin-encoding genes should be considered as factors which increase the virulence of E. coli strains.

E. coli strains encoding only fimbriae type I did not show differences in the frequency of bacteriocinogeny compared to strains without genes for virulence factors. The role of fimbriae type I in development of human infections is not clear. Although deletion of the fim gene cluster from virulent E. coli strain O1:K1:H7 attenuated virulence in the urinary tract infection (UTI) model [33]; possession of fimbriae type 1 in E. coli strains from different sources was not found to correlate with the ability to cause UTIs [34–39]. Several virulence factors, typical for diarrhea-associated E. coli strains, including pCVD432 (EAggEC), ial/ipaH (EIEC), eaeA/bfpA (EPEC) and afaI (DAEC) were not found to be associated with bacteriocin genes. Bacteriocin producers therefore appear to be mainly associated with ExPEC virulence factors (E. coli strains containing combinations of sfa, pap, aer, iucC, cnf1, α-hly determinants). The occurrence of these virulence factors were associated with both chromosomally (microcins H47 and M) and plasmid encoded colicin (E1, Ia and S4) and microcin types (B17, V).

Presently, several bacteriocins including colicin E1, and microcins H47, I47, E492, M, and V are considered virulence factors in extraintestinal pathogenic E. coli strains [20–23]. Azpiroz et al.[20] and Budič et al.[22] found an association between production of microcins H47, I47, E492, M, and V and the distribution of virulence factors (i.e. hlyA, cnf1, usp, iroN, iroCD, fyuA, papC, papG and tcpC) in uropathogenic strains of E. coli; from these results they assumed that production of these bacteriocin types could contribute to development of bacteraemia. Although different sets of virulence determinants and bacteriocin genes were used in these studies, our findings match with these observations.

We also found these associations between bacteriocin production and ExPEC virulence determinants among human fecal E. coli isolates. Moreover, we have found new associations between 3 bacteriocin types (microcin B17, colicins Ia and S4) and the ExPEC virulence characteristics of human fecal E. coli strains. Given that colicin Ia and microcin B17 are known to be encoded on relatively large plasmids, it is possible that the association with more virulent strains is due to other genes being harbored on these plasmids, and not by colicin synthesis itself. However, colicin S4 was found to be encoded on a small plasmid (7.4 kb) [40] that was similar to the colicin E1-encoding plasmid (6 kb) [21]. Since these small plasmids do not encode virulence factors, colicin S4 appears to be a potentially important virulence factor and/or an important factor of resident E. coli strains.

The presence of virulence determinants (e.g. genes encoding P-fimbriae, siderophore aerobactin, hemolysin and expression of O antigens, which are typical for ExPEC strains; and capsular types K1 and K5) are associated with resident E. coli strains [41–44]. At the same time, ExPEC strains causing extraintestinal infections like urinary tract infections and sepsis/meningitidis are believed to originate from the gut microflora. Their virulence determinants including adhesins (P-fimbriae, S-fimbriae), toxins (e.g. hemolysin, cytotoxic necrotizing factor) and siderophores (e.g. aerobactin) appear to be important for E. coli strains to survive in the extraintestinal environment [45–47].

On the other hand, we found that diarrhea-associated strains from our set of 1181 fecal E. coli had a lower prevalence of bacteriocinogeny and a lower frequency of several bacteriocin producers. In addition, no specific bacteriocin types appear to be associated with virulence determinants that are typical for these strains. Unlike fecal strains which have the characteristics of ExPEC strains, diarrhea-associated strains are not considered to be resident human E. coli strains, which may explain the lower prevalence of bacteriocin genes.

In summary, bacteriocin synthesis is linked to strains with ExPEC characteristics and appears to increase the ability of E. coli to reside in the human gut. Moreover, at least several bacteriocin-encoding genes should be also considered as factors which increase the virulence of ExPEC strains.

The frequency of bacteriocin-encoding genes was found to be positively correlated with the frequency of E. coli virulence determinants. E. coli with virulence characteristics of ExPEC strains, i.e. strains encoding virulence factors S-fimbriae (sfa), P-fimbriae (pap), cytotoxic necrosis factor (cnf1), α-hemolysin (α-hly) and aerobactin biosynthesis (aer, iucC) were more often found to harbor genes encoding synthesis of microcins (H47, M, V and B17) and colicins (E1, Ia and S4) than other tested E. coli strains.

The data set of 294 colicin gene sequences supporting the results of the article has been deposited in the GenBank/EMBL/DDBJ under accession numbers AB923519 - AB923812.