Letter to the editor

The commentary by Wassenaar and Gunzer [1] on a recently published characterization study of a Crohn’s disease (CD) Escherichia coli combining virulence properties of multiple pathotypes [2] hits the complex issue on how to identify a pathogenic strain within this species, because of its high genetic and phenotypic heterogeneity and varying potential to cause disease. E. coli is a resident member of gut microbiota and a multi-disease organism alike, including a predominantly commensal and a smaller group of pathogenic strains [3]. The observation of a high numbers of E. coli in the gut of inflammatory bowel diseases (IBD) patients by many independent researchers in the last decades [4–6] has made these bacteria suspicious of involvement with these diseases as well. The main point expressed in the commentary is that the presence of virulence genes in an E. coli strain is not sufficient to grant it a role in disease because it has been found that pathogenic and probiotic strains of long recorded history of safe use share virulence genes and the administration of the later in a high dose to volunteers had no detrimental effects [7]. These findings are really important since they can instigate a discussion towards a better discernment of what among the many known E. coli traits can be considered as true virulence factors. Yet, in my view the data are still modest to sustain any conclusion and care must be taken to avoid generalizations. The number of virulence genes detected in the probiotics was limited and in general restricted to factors involved in colonization and fitness (toxins genes, for example, were not found). The comparison of EDL933 strain’s proteome with that of G3/10, one of the probiotic strain used in the study, showed that the protein genes typical of the former were mostly absent in the probiotic E. coli [7]. Analysis of genetic relationship put probiotics and commensal or non-pathogenic strains in a single cluster apart from E. coli reference pathogenic strains.

The commentary on da Silva Santos et al.’s paper seems not to consider that the focus of the original work was not restricted to investigate genetic traits, but also virulence associated phenotypes, notably the capacity to adhere and to invade epithelial cells. E. coli D92/09 strain was able to replicate inside macrophages and its efficiency of entry into these cells was higher than that shown by adherent and invasive E. coli (AIEC) strain LF82 [2]. Innumerous studies published to date [5, 8, 9] have demonstrated a high prevalence of adherent strains among E. coli from CD patients. The characterization of the invasive phenotype of a strain from one of these works [10] and subsequent studies [11] led to the description of the AIEC pathotype. Evidence gathered from these studies settled the basis for the recognition of AIEC as the most probable bacterial agent suspected of playing an active role in CD [12]. Concerning the strain described by da Silva Santos et al. [2], in addition to the adherent and invasive phenotype, the bacteria proved to be lethal to the host cell. This and other features reveal a virulence profile distinct from those expressed by AIEC (Table 1) and whose significance should be investigated in experiments with animal models. In addition, search for strains bearing markers similar to those observed in E. coli D92/09 in other CD patients should be done. In a recent PCR screening [13, 14] performed in my lab to detect some E. coli O serogroups in a bacterial collection from nine CD patients and eight control subjects, who routinely attended the São Paulo State University Hospital (HC-UNESP), a high prevalence of O25, O83 and O126 could be observed in both groups. One of these O25 strains, detected in a CD patient, belonged to the O25:H4 serotype and ST131, an ESBL producing virulent clone of global distribution [15, 16]. Strains of this clone bearing AIEC phenotype have been identified among both intestinal and extraintestinal pathogenic E. coli (ExPEC) [17]. O25 and O83, but not O126, include multidrug resistant serotypes frequently associated with nosocomial infections [18]. D92/09 strain belongs to O126:H27 (Table 1), an O:H type traditionally involved with diarrheal diseases [19], but which in this case marked a CD clinical isolate. Manual search in this strain’s genome revealed genes conferring resistance to aminoglycoside (strB), beta-lactam (bla TEM-1B), fluoroquinolone (QnrB19), tetracycline (tet(A)) and trimethoprim (dfrA8). This finding allied to the high prevalence of O126 strains among the patients, as mentioned above, could indicate that possibly strains within this serogroup can be involved with nosocomial infections in HC-UNESP and that D92/09 strain may have infected the patient in the hospital environment. Since many CD E. coli strains belong to serotypes of ExPEC [17], some of which associated with nosocomial infections, it is likely that by attending the hospital in a routine basis, it is reasonable to suppose that at least some of the CD patients may acquire infections by these bacteria in the hospital settings.

Table 1 Some features of E. coli D92/09 strain in comparison with AIEC LF82
Full size table

A significant consideration in the commentary refers to the absence of symptoms in the infected patient, a condition which was explained by the existence of an immune system control, preventing bacterial action [7]. Indeed, this might have been the case but not necessarily is to mean that the strain is devoid of any virulence potential. It has been shown that the abundance of E. coli in the ileal mucosa of post-operative CD patients may not be linked to disease activity [4, 5]. Nonetheless, the D92/09 host presented discrete clinical activity at the site of bacterial isolation insufficient to produce detectable symptoms [2].

Concerning the involvement in CD, doubts remain not only in regard to D92/09 strain but also to E. coli as a whole. Many of the disease provoked by this species are attributable to opportunists sharing colonization factors of pathogenic strains. Abnormal proliferation and involvement with disease could depend on additional and particular bacterial properties which could act in specific predisposing condition of the host. That seems to be the case of CD patients whose ileal enterocytes overexpress carcinoembryonic antigen-related cell adhesion molecule 6 (CEACAM6) which was shown to be a receptor for AIEC type 1 pili [20]. In theory, this predisposing condition would also favor strains of the probiotics such as Symbioflor2, possessing type 1 pili. A number of key factors described in LF82 strain, not present in the probiotics, such as the ability to replicate inside macrophages and potential to form granulomas constitute evidence strong enough to suspect of the participation of AIEC in a disease with the clinical and histopathological features typical of CD. In other words, the manifestation of virulence depends on the concerted action of multiple factors. Although not fully characterized, the available information on D92/09 strain, configures a profile with potential do cause damage in a condition of opportunism such as that in CD [2].

In conclusion, both the Wassenaar and Gunzer’s commentary on D92/09 strain and the results of their recently published work on virulence genes in probiotics are very interesting and should represent a starting point on a discussion which may broaden our knowledge on the different ways that E. coli interacts with humans, as probiotics, commensals or pathogens.