Enterobacteriaceae were isolated from fresh biopsies of CRC and UC patients as described before (2). Biopsies were subjected to three mild ultrasound-wash cycles to discard both transient and loosely attached bacteria. Each cycle consisted of 30 s at 50 Hz, followed by washing with 1 ml of 1× PBS (phosphate-buffered saline). A mild osmotic shock was applied by incubation for 5 min in distilled water to release any intracellular bacteria. During this time, eukaryotic cells are disrupted without compromising bacterial cells viability. Biopsies and supernatants were cultured in Tryptone Bile X-Glucuronide Medium (TBX, Oxoid, Cambridge, UK). Up to 96 colonies were collected per sample, either glucuronidase positive or negative, and confirmed using the indole assay. E. coli from CeD patients were isolated from the stools of children in Violet Red Bile Dextrose (VRBD) agar and identified as previously described (29).

Following the work of Darfeuille-Michaud et al. (1), isolates considered AIEC in this study were those that presented the ability to adhere to intestinal epithelial cells with an adhesion index (ADH_I) ≥1, to invade intestinal epithelial cells with an invasion index (INV_I) ≥1%, and to survive and/or replicate within macrophages with a replication index (REPL_I) ≥100%. In addition, this definition also pointed out that these E. coli lack virulence genes linked to diarrheagenic E. coli. Because some toxins such as cdt have been reported to be linked to acute diarrhea (30), in the context of this work positive strains for cdt and cnf have been referred to as “AIEC-like” throughout the manuscript. The methodology to identify putative AIEC isolates and to confirm the phenotype is depicted in the following sections.

Adherent-invasive isolates identified were confirmed to be E. coli by performing additional phenotypic tests, which were: lactose, citrate, and urease tests, swarming phenotype; and by quantitative Polymerase Chain Reaction (qPCR) using primers and conditions described in (31). Ct values >35 were considered no amplification.

To determine the clonality of 15 AIEC isolates recovered, we performed PFGE as described elsewhere (32). Agarose plugs containing genomic DNA were digested with XbaI for 4 h. Electrophoresis was carried out in a CHEF-DR III System for 19 h following the protocols of CDC PulseNet (32). TIFF images were analyzed using GelComparII software.

To characterize virulence features, E. coli strains were plated in LB agar media and incubated at 37°C in an anaerobic chamber Whitley DG250 Anaerobic Workstation (Don Whitley Scientific, Inc., Shipley, UK) for 12 h. A colony-PCR targeting multiple genes (chuA, yjaA, and tspE4C2) was used for the phylogenetic group characterization following the method of Clermont et al. (33). The multiplex PCR was performed as previously described (29). Isolates were assigned to phylogenetic groups as follows: ChuA+ YjaA+, group B2; ChuA+ YjaA−, group D; ChuA− TspE4C2+ group B1; ChuA− TspE4C2− group A.

The E. coli strains were also characterized for the presence of virulence genes using two sets of multiplex PCR as described by Nowrouzian et al. (34). The first multiplex PCR detected the presence of the virulence-associated genes fimA (type-1 fimbriae), papC (P fimbriae), sfaD/E (S fimbriae), and draA (Dr haemagglutinin). The second multiplex PCR detected the presence of hlyA (hemolysin), neuB (capsule K1), kfiC (capsule K5), and iutA (aerobactin).

Also, the virulence genes cdt and cnf were screened by PCR as previously described (35) with the following PCR program: denaturation 94°C - 5 min, 30 cycles of 1 min–94°C, 1 min–55°C, 1 min–72°C, with a final extension step of 10 min–72°C. Pks virulence gene was also detected following the same protocol as for the cdt and cnf genes but annealing temperature was at 60°C.

The afa/draBC virulence gene was detected using a modification of the protocol by Mora et al. (36) using the following PCR program: denaturation 94°C - 5 min, 30 cycles of 1 min–94°C, 1 min–58°C, 1 min–72°C, with a final extension step of 10 min - 72°C.

Virulence genes characteristic of diarrheagenic E. coli pathotypes (stx1, stx2, eae, ipaH, pCDV432, eltA, and est) were also investigated as previously described (37).

The primers used can be checked in Supplementary Table 2 . All PCRs reactions consisted of a 0.2 mM mix of each deoxynucleoside triphosphate, 1× Taq DNA polymerase buffer, 0.25 μM of each primer, and 2.5 U of Taq polymerase. PCR products were separated in a 2% agarose gel electrophoresis and visualized by using GelRed^® Nucleic Acid Gel Stain (Biotium, cat number #41003).

I-407 cells were seeded in 24-well plates at a density of 4 × 10⁵ cells/well. Monolayers with fresh medium (EMEM) not containing antibiotics, with inactivated FBS, were infected at a MOI of 10. Infected cells were incubated for 4 h at 37°C with 5% CO₂ and monolayers were examined by microscopy to search for cytopathic effects. Then, both cells in the supernatants and cells adhered to the well, were harvested from the plate, and subjected to Gibco™ Trypan Blue staining to quantify the percentage of dead cells.

AIEC prevalence data (frequency of patients colonized by AIEC) was compared between groups of patients with Fisher’s exact test. This test was also used to assess gender distribution among groups, distribution of phylogenetic origin, and virulence gene frequency in AIEC strains between groups of patients. Quantitative data such as the abundance of AIEC (proportion of AIEC with respect to total Enterobacteriaceae) or adhesion, invasion, and replication indices were compared by non-parametric tests (Mann–Whitney U-test for comparison of two groups, Kruskal–Wallis test for more than two groups).

Although E. coli abundance has been related to disease status in UC patients (38, 39) and adherent E. coli has been detected in the colon of UC patients (40–43), the AIEC population needs more investigation in larger UC cohorts. In the present study, higher AIEC prevalence in UC patients (35.7%) was reported in comparison with previous studies (0.0–10.0%) (1, 15, 16) but was similar to recent studies conducted in Asia (37.5%) (18) and Australia (40.0%) (19). These studies showed controversial results regarding AIEC prevalence in UC in comparison with CD, since some reported lower prevalence in UC than in CD (1, 16), but others reported similar (18, 44) or even higher prevalence in UC (15, 19). These differences might be due to the size of the cohort studied or the different methodologies applied. We used the same methodology to assess AIEC prevalence and abundance in UC as we did in the previous study for CD and H (2). Therefore, the results are comparable between cohorts. In UC patients, we found an AIEC prevalence lower in comparison with CD patients but higher than H, and no statistically significant differences were found between UC and these two groups of subjects. The prevalence in patients with proctitis (33.3%) and left-sided colitis (44.4%) was similar to the prevalence in CD patients with colonic disease (40.0%) but lower than CD patients with ileal involvement (60.0%). No conclusions can be extracted from patients with pancolitis since only two patients had this condition. Our results are in agreement with a recently published meta-analysis concluding that the AIEC prevalence is higher in UC patients compared to controls (despite not reaching statistical significance), being AIEC a pathobiont that could be involved in both CD and UC pathogenesis (45). Interestingly, 50% of UC patients with proctitis and left-sided colitis had AIEC in the ileum, whereas 0 and 44% of these patients had AIEC in the colon, respectively. AIEC probably colonize more easily the ileum without causing inflammation, and they can later colonize a compromised inflamed colon, perhaps contributing to the disease.

The massive characterization of E. coli isolates obtained from UC patients regarding their invasion ability allowed us to estimate the abundance of AIEC strains with respect to the total Enterobacteriaceae population. This approach revealed that the abundance of AIEC is quite similar between UC, CD, or H subjects ( Figure 2 ), which suggests that once an individual is colonized by AIEC, this bacterium occupies the niche similarly in all intestinal conditions. Generally, AIEC represents less than 10% of Enterobacteriaceae, except for some individuals (AIEC abundance reached 27 and 50% of Enterobacteriaceae in two I-CD patients, respectively, the 52% in one H subject, and the 19% in a UC patient with mild left-sided colitis).

Regarding the phenotypic characteristics of strains, studies that analyzed the invasion rates of E. coli isolates from UC determined that (I) these had lower Caco-2 invasion values (17), (II) these showed similar I-407 (20) and Caco-2 (46) invasion values or (III) these had higher invasion in Hep-2 cells than those E. coli isolated from CD patients (18). Although only AIEC strains were quantitatively assessed in the present study, similar UC-AIEC and CD-AIEC I-407 adhesion and invasion values were obtained. UC-AIEC strains had a lower ability to replicate inside J774 macrophages than CD-AIEC. This result is in contrast with previous observations (18, 21). Sepehri et al. (21) reported higher intramacrophage replication in RAW264.7 of UC-isolated E. coli in comparison with E. coli isolated from control subjects and a moderate higher replication in comparison to CD-E. coli. Lee et al. (18) reported similar intracellular survival of CD- and UC-E. coli in THP-1. Further studies and meta-analysis are needed to confirm which is the degree of virulence of UC-AIEC on different cellular models.

A meta-analysis that reviews the association of E. coli phylogenetic groups with IBD subtypes showed that the B2 phylogroup is frequently associated with IBD, particularly with UC (47). Of note, the AIEC isolated from UC in this study corresponded only to the A or B1 phylogenetic groups. This was an intriguing result considering that AIEC strains isolated from CD patients of the same geographical region are mostly represented by the B2 phylogroup (2).

Although no previous results have compared specifically the virulence gene carriage of AIEC from UC patients with that of AIEC from other disorders, previous reports have shown that UC-E. coli have particular virulence genes in comparison with CD- E. coli (21, 42, 46, 48, 49), while others showed similar profiles (18, 20, 50, 51). Virulence gene characterization of the strains included in this study indicated that most UC-AIEC strains harbored Afa/Dr adhesins. In comparison with other AIEC strains, it represents a remarkable difference since only 14% of CD-AIEC had genes encoding for the Afa/Dr family adhesins (afa/draBC gene tested) (52). Afa/Dr adhesins are commonly present in Diffusely Adhering E. coli (DAEC), a pathotype involved in urinary tract infections, pregnancy complications, and acute diarrhea (53). In particular, Dr fimbriae can adhere to several human carcinoembryonic antigen cell adhesion molecules (hCEACAMs), such as CEACAM5 which has been reported to be upregulated in UC patients (54). Therefore, in comparison with AIEC strains present in CD patients, which mostly adhere to human enterocytes via the interaction between type 1 fimbriae FimH adhesin and the CEACAM6 receptor (55), this study reveals that AIEC strains present in UC patients may use Dr fimbriae to adhere colonic enterocytes via the interaction with human receptors increased in UC such as CEACAM5. Further studies investigating the commonalities and differences between DAEC and AIEC are needed, since DAEC can also invade intestinal epithelial cells in vitro (56). However, the ability to survive and replicate in macrophages was used as the defining phenotypic feature to identify AIEC and differentiate them from DAEC strains (57). Another virulence gene that was frequent in strains from UC patients was the cdt gene which encodes for toxins with DNase activity (30). None of the CD-AIEC from our previous work harbored this gene (2), but it was present in E. coli strains from CRC patients (26). The presence of this gene and the cnf gene in the strains excludes them from being classified as AIEC, so we have referred to them as AIEC-like in the manuscript. CDT is a bacterial genotoxin present in gram-negative bacteria that is capable of modulating the eukaryotic cell cycle by pausing the G2/M transition. As a DNase, CDT damages the host DNA activating the DNA damage response. In most cases, the repair system fails to resolve the situation, which leads to cell death or senescence, but a small portion of cells outpace cell cycle arrest and continue to proliferate accumulating DNA lesions, being more likely to develop tumors (58). The high prevalence of this gene in UC-AIEC-like strains is of great relevance considering that UC patients have an increased risk for CRC (59). CDT-producing AIEC colonizing the UC intestine may be responsible of inflammatory response activation and epithelial barrier disruption, but also it could represent a risk for cancer development. Other genotoxins, the cytotoxic necrotizing factor (cnf) and the colibactin (pks), were also detected in the UC-AIEC strains, yet less frequently. It remains to be determined the prevalence of cdt, cnf and pks in non-AIEC-like E. coli, which would be interesting in future studies to assess if this is a commonality in E. coli inhabiting the gut of UC patients.

The high prevalence of genotoxins in UC-AIEC-like strains leads us to study the cytopathic effects of these bacteria on intestinal epithelial cells. Only the PL23F02 strain, which belongs to A phylogroup, induced cell death at 4 h post-infection. Of all genes tested, this strain only had the cdt gene. CDT-induced cytopathic effects can only be detected in cell–lysate-interaction tests, and are characterized by enlarged nuclei and cell distension (60). Therefore, we suspect that this strain harbors other virulence factors, different from CDT, and is responsible for such cytotoxic effects. Similarly, E. coli from the A phylogroup, devoid of known cyclomodulin-encoding genes, were found to induce DNA damage in vitro (60).

In summary, AIEC-UC strains identified here had a different genetic profile in comparison with AIEC-CD, being mostly B1 strains with Afa/Dr adhesins and cyclomodulins, characteristics not found in the AIEC-CD investigated previously (2). We hypothesize that AIEC can show a myriad of faces and their virulence particularities may determine the disease type. Nonetheless, we think that a standardized protocol for AIEC identification is needed, as well as further studies comparing the genetic and phenotypic features (e.g., adherence phenotype, invasion abilities, or cytopathic effects caused on host cells, etc.) between close E. coli pathotypes, such as AIEC and DAEC, and genotoxin-producing strains.

For a long time, intestinal E. coli populations have been suspected of playing a role in CRC. Mucosa-associated and intracellular E. coli have been more frequently identified in colonic tissue from patients with adenocarcinomas than in controls (22, 23), and within colon cancer more abundance has been found in tumors than in the mucosa (24). Moreover, high levels of mucosa-associated and internalized E. coli have been correlated with poor colorectal carcinoma prognostic factors and a higher proliferative index of epithelial cells (24). These strains are frequently cyclomodulin-positive (25, 60, 61) and have pro-carcinogenic activities in murine models (24, 62–65) and human organoids (66).

Few works have addressed the adhesion, invasion, and intracellular replication abilities of CRC-associated E. coli. Buc et al. found no differences in adhesion abilities of E. coli strains isolated from CRC or diverticulosis patients (60). On the other hand, Sobieszczańska et al. showed that E. coli isolated from a pediatric cohort with polyposis had the highest invasion efficiency on I-407 cells compared with E. coli strains isolated from IBD children (20). Raisch et al. (26) determined the adhesion and invasion abilities of B2 E. coli isolated from CRC and diverticulosis patients, and they found that both adhesion (below 4) and invasion (below 0.5%) levels were much lower for CRC-E. coli in comparison with the AIEC reference strain LF82 (adhesion: 53.23 ± 6.63, invasion: around 8%). However, it is unclear whether the strains could be considered AIEC or not following the criteria described in (1), since strains with adhesion levels higher than 1 and invasion levels higher than 0.1% were found, but macrophage intracellular replication was not studied. Darfeuille-Michaud et al. (1) found a low prevalence of AIEC in ileal specimens of right colonic cancer patients (6.2%). Only one out of 16 patients harbored AIEC (one strain with an adhesion level of 16 and an invasion level of 0.2% in I-407 cells and an intramacrophage replication level of 833 in J774 cells). The AIEC prevalence observed in our study was similar to these results since we found that AIEC was present only in 6.7% of CRC patients, which was much lower than AIEC prevalence in UC (35.7%) and CD (60.0%) patients, and similar to H controls. In comparison with AIEC strains isolated from CD and H subjects, CRC-AIEC-like strains of the present study showed lower adhesion values to intestinal epithelial cells but similar invasion abilities, and similar intramacrophage replication indices ( Figure 3 ). Thus, our results suggest no particular relationship between AIEC and CRC.

Prorok-Hamon et al. reported that E. coli from CRC patients frequently present afaC and pks genes, and the presence of these virulence genes correlated with higher adhesion and invasion capacities (25). In the present study, only AIEC-like strains were characterized in terms of virulence gene carriage, and we found that none of the two CRC-AIEC-like strains harbored these genes. Nonetheless, it does not rule out the possibility that other non-AIEC strains of the CRC gut frequently harbor these genes. The CRC-AIEC-like strains isolated here harbored draA and cdt genes. As explained in the previous section, CDT is a genotoxin that may be responsible for tumor induction. Nonetheless, although B2 cyclomodulin producing E. coli has been largely incriminated in cancer induction, further evidence is needed to prove this hypothesis or demonstrate whether the flourishment of these strains in the mucosa of CRC patients is an effect of the disease (67). To establish the prevalence of cdt, cnf, and pks in non-AIEC-like E. coli from CRC patients will be of interest to know if this is a widely distributed feature in the strains colonizing the gut affected by colorectal cancer.

Dysbiosis characterized by increased E. coli abundance has been observed in untreated CeD children (27, 28). Also, an increased virulence-gene carriage has been reported in E. coli isolates from treated and untreated pediatric patients compared to healthy controls (29). However, no previous studies assessing adhesion, invasion, and intracellular replication abilities of E. coli present in the gut of CeD patients have been reported. We have phenotypically characterized the E. coli strains isolated in a previous study from the feces of children with active CeD, symptom-free CeD patients, and H children (29).

AIEC prevalence in H children was very similar to that previously observed in H adults [20.0% vs. 21.4% (2)], which was noticeable considering the differences between the cohorts compared. For example, E. coli from H children were isolated from fecal samples, whereas those of CD patients were obtained from biopsies, and the number of strains screened was much higher for the latter. Nonetheless, this evidences that AIEC might also be present in the gut during childhood and possibly infancy. Surprisingly, none of the CeD patients had AIEC. Thus, we rule out the hypothesis that AIEC could be associated with CeD. However, we are aware that a reduced number of E. coli strains have been studied, and these strains have been isolated from feces instead of mucosal samples. Therefore, this hypothesis should be further confirmed. In turn, the fact that the 20% of strains from H children were AIEC in contrast with the 0% from CeD suggests that other types of bacteria are favored in the CeD intestine.

Although no AIEC strains were identified in CeD patients, interestingly, we found that the invasion abilities of the strains of E. coli isolated from CeD patients with active disease presented higher invasion ability than those isolated from symptom-free CeD and non-AIEC isolated from H patients ( Figure 4 ). On average, CeD active-E. coli showed invasion indices almost four times higher than strains of the other two groups. We suggest that the intestinal inflammatory state of untreated CeD patients may generate a more permissive environment that facilitates the colonization of more virulent E. coli strains, with a phenotype with increased invasion and/or intracellular survival capacity on intestinal epithelial cells. In turn, these virulent E. coli strains can contribute to the inflammatory state creating a vicious circle. In fact, the contribution of an E. coli strain isolated from CeD children triggering the gluten-induced immunopathology in mice has been proven (68). Nonetheless, further studies should be warranted to understand how the persistence of an inflammatory state could eventually displace non-virulent E. coli strains in favor of more virulent strains or exert a selective pressure driving the evolution of existing E. coli strains towards more invasive phenotypes. The study of AIEC prevalence in CeD adult patients with a long history of the disease would also be highly interesting.