Although belonging to the normal human intestinal flora, E. coli becomes highly pathogenic following the acquisition of genes coding for virulence factors, one of which being CNF1 34. CNF1-producing E. coli strains are occasionally detected in isolates from feces of children with diarrhea 353637, but, more frequently, are responsible of extraintestinal infections, particularly in the urinary tract (UTIs) 383940. Also, these strains can be detected in cases of bacteraemia 41 and of meningitis in neonates 42. CNF1, first described in 1983 by Caprioli and coworkers as a toxin capable of causing multinucleation ("cytotoxic") in cultured cells (Fig. 1a, b) and necrosis in rabbit skin ("necrotizing") 4344, is a single-chain multidomain protein toxin, containing a binding domain to a cell receptor (N-terminal domain), a translocation domain (middle domain) and an enzymatic one (C-terminal domain) 4546, which modifies a specific cellular target in the host cell cytosol. CNF1 binds to the surface of cultured epithelial cells with high affinity 47, and the 67-kDa laminin receptor has been suggested (67LR) as a putative receptor for CNF1 48. After binding to its receptor, CNF1 is endocytosed and routed to an endosomal compartment 47, from where the toxin injects its catalytic activity into the cytosol 47, by using two hydrophobic structures present in the middle part of the CNF1 molecule 49.

The cytoplasmic target of CNF1 is represented by the Rho GTPases, important molecular switches belonging to the Ras superfamily, that cycle between an inactive GDP-bound state and an active GTP-bound one, under the strict control of activators (guanine nucleotide exchange factors, GEFs) and inactivators (GTPase-activating proteins, GAPs) 50. The conformational changes, induced by the binding of GTP or GDP, occur inside two molecular domains of the Rho proteins called switches that are responsible of the coupling of the G-proteins with their downstream effectors (switch 1) and allow the interaction of the activated GTPase with GAPs (switch 2). The enzymatic activity of CNF1 consists in the deamidation of a specific glutamine residue located in the switch 2 domain of the G proteins (glutamine 63 of Rho 5152 or glutamine 61 of Rac and Cdc42 53). This glutamine residue is essential for the GTPase activity of Rho proteins, either intrinsic or GAP-mediated 54. By modifying glutamine into glutamic acid, CNF1 impairs the role of Rho GAP, allowing Rho proteins to be permanently locked in their activated GTP-bound state and, thus, enhancing the activity of these proteins on their effectors.

It is worth noting that the modification on Rho proteins induced by CNF1 corresponds to the modification found on Ras proteins in many tumors 55. In fact, Rho glutamine 63 corresponds to Ras glutamine 61, a well-known residue that, after mutagenization, leads to a permanent activation of the molecule and to tumor onset.

The Rho GTPases are pivotal in controlling the actin cytoskeleton architecture 50. In fact, one of the first described cell responses to CNF1 is a remarkable reorganization of the actin cytoskeleton in epithelial cells that consists in the assembly of F-actin in prominent stress fibers, membrane ruffles and filopodia 56 (Fig. 1c, d). The prominent ruffling activity promoted by CNF1 permits epithelial cells to behave as phagocytes, developing a macropinocytotic activity that allows the capture and engulfment of different types of particles, including bacteria 5758. This aspect is of particular relevance for the bacterial pathogenicity since the macropinocytic activity ensued by CNF1 in epithelial cells may possibly represent the route of entry of CNF1-producing E. coli, similarly to what occurs with other intestinal pathogens (Fig. 1e). Furthermore, upon activation by CNF1, the Rho GTPases undergo sensitization to ubiquitylation and subsequent proteosomal degradation 59, a process that would turn off the ruffling process, thus allowing an efficient internalization of bacteria inside the cells (Fig. 1f). It is worth noting that the activity of CNF1, with its ability to switch on the Rho GTPases and then coerce their degradation in the proteasome, is somehow similar to the activity of the intracellular bacterium Salmonella 60, for which a link with cancer has been evidenced 9. This bacterium first activates the Rho GTPases, by the GEF-like toxin SopE, to promote macropinocytosis that allows its entry into cells and soon after, once inside, deactivates the GTPases via a GAP-mimicking protein (SptP), thus allowing a moderate threshold of Rho protein activation for a high invasion efficiency 61.

By regulating the actin cytoskeleton, the Rho GTPases also play a crucial role in certain aspects of the malignant phenotype, such as tumor cell motility, invasiveness and metastasis 62. Indeed, RhoC has been implicated as a marker for highly angiogenic and aggressive breast cancer with a high metastatic ability 63. An additional clue in favor of our hypothesis comes from the observation that CNF1 provokes cell junctions disruption and strongly enhances cellular motility in uroepithelial 804G cells 59. The augmented motility of epithelial cells highlights another characteristic that somehow links this toxin to cancer.

It is also known that Rho proteins are crucially involved in the development of inflammatory processes and that a key player in the link existing between Rho, chronic inflammation and cancer is the Nuclear Factor-kB (NF-kB) 64. As mentioned in the Introduction, NF-kB is represented by a group of structurally related and evolutionarily conserved transcription factors involved in regulating the expression of genes that control different aspects of the tumor cell biology, including inflammation, cell growth, and suppression of apoptosis (reviewed in 65). Tumor cells can show elevated NF-κB-regulated transcription, which can inhibit TNF-α-induced apoptosis through the up-regulation of the anti-apoptotic proteins of the Bcl-2 family.

In this context, we previously reported that CNF1 can activate the nuclear factor-kB (NF-kB) 66 in epithelial cells. Such an activation, is responsible for the ability of the toxin to stimulate the expression of pro-inflammatory factors and to protect host cell from apoptotic stimuli. As concerns the pro-survival activity, we have shown that CNF1 can increase the expression of proteins related to cell adhesion (integrins, Focal Adhesion Kinase, cadherins, catenins), thus improving cell spreading and the ability of cells to adhere to each other and to the extracellular matrix 67. In fact, prolonged cell survival, together with increased adhesion to matrix components might have significant biological consequences and affect the tumorigenic potential of epithelial cells. Moreover, CNF1 protects epithelial cells against the drop of the mitochondrial membrane potential provoked by UVB radiation and increases the expression of the anti-apoptotic members of the Bcl-2 family, Bcl-2 and Bcl-X_L 68. Although the causal relationship between Bcl-2 and mitochondrial membrane potential has not yet been clarified, we have very recently demonstrated that the up-regulation of the anti-apoptotic protein Bcl-2 somehow controls the mitochondrial morphology, and that this depends on the activation of the pro-inflammatory Rac1/PI3K/Akt/IKK/NF-kB pathway 69. In fact, besides blocking the activity of the pro-apoptotic members of Bcl-2 family, Bcl-2 is involved in the regulation of mitochondrial morphology, since Bcl-2-over-expressing mitochondria present both increased volume and structural complexity 70. In living cells, mitochondria continuously divide (fission) and fuse (fusion) with one another 71 and Bcl-2 family members are involved in these processes, the pro-apoptotic members regulating fission whereas the anti-apoptotic regulating fusion of mitochondria 72. In this context, we very recently demonstrated that CNF1 can induce, in epithelial cells, the formation of a complex network of elongated and interconnected mitochondria with an increased average length [69; Fig. 1i, l]. Importantly, Bcl-2 silencing reduces the ability of CNF1 to protect cells against apoptosis and also prevents the CNF1-induced mitochondrial changes. Therefore, since the mitochondrial remodeling is of direct relevance for the role of these organelles in cell physiology and the mitochondrial dysfunction can contribute to a number of human disorders, including cancer 73, the role of CNF1 as a factor favoring transformation can be further supported by this novel finding.

As concerns inflammation, CNF1 ensues the transcription and release of pro-inflammatory cytokines, such as IL-6, IL-8 and TNF-α in uroepithelial 74 and endothelial 75 cells, probably contributing to the establishing of the inflammatory process due to CNF1-producing E. coli. The pro-inflammatory role of CNF1 is also supported by the demonstration that the toxin strongly up-regulate the transcription of cyclooxygenase- 2 (COX-2) 76, an immediate-early gene induced in response to pro-inflammatory cytokines, tumor promoters, and growth factors and over-expressed in cancers of the lung, colon, stomach, and breast 777879.

On the whole, it appears that CNF1 touches some of the signaling pathways that are engaged by carcinogens and tumor promoters, as schematized in Fig. 2. Particularly relevant are the activation of the transcription factor NF-κB, the pro-survival activity, the increased expression of RNA messengers for COX-2 and pro-inflammatory cytokines as well as the augmented cell motility. In addition, CNF1 impairs the cytokinesis, thus leading to multinucleation 4456, induces nuclear segmentation, amitotic cell division, multipolar mitosis (Fig. 1g, h) 80, and modulates autophagy 81, cellular phenomena that are frequently observed in different types of cancer cells, and blocks the cell cycle G2/M transition in epithelial cells 82. The ability of CNF1 to block the cell cycle progression suggests a strategy that permits to contain the host damage, and to induce specific cellular responses rather than rapid cell death. This could in turn facilitate the bacterial invasion of underlying tissues. The activity of CNF1, with its ability to switch on the Rho GTPases and then coerce their degradation in the proteasome, probably renders the CNF1-producing E. coli intracellular parasites, hence permitting their potentially harmful and presumably transforming activity inside the cell, where they can escape the host immune system attack. Finally, it is worth noting that CNF1 activity on cells shares several properties with CagA from the carcinogenic bacterium H. pylori, which can be indicated as the best example of a potentially carcinogenic toxin 29. In Table 1, the cellular effects of E. coli CNF1 and those of the major virulence factors (VacA and CagA) of H. pylori are compared.

Hence, taking altogether, we can hypothesize that once secreted by epithelium-associated E. coli, CNF1 switches on the Rho GTPases, initiating a pathway that leads to the actin cytoskeleton reorganization, but also to the activation of NF-κB, which in turn enter the nucleus and causes the transcription of anti apoptotic factors as well as pro-inflammatory cytokines. Production and secretion of pro-inflammatory factors recruits cells of the immune system creating an inflammatory environment. Epithelial cells with the Rho GTPases modified by CNF1 in a way that corresponds to the modification found on Ras proteins in many tumors 55, continue to grow without undergoing apoptosis in an environment rich of inflammatory factors. This can account, finally, for promotion or increase of the neoplastic risk.

ST, AF and CF participated in drafting the manuscript and in its approval.