Establishing the parameters of host-pathogen interactions in pyogenic liver abscess caused by hypervirulent Klebsiella pneumoniae

Abstract

Klebsiella pneumoniae (Kp) is a Gram-negative bacterium, classically known to cause opportunistic infections in patients with comorbidities. Since the 1980s, strains of Kp causing pyogenic liver abscess, with dissemination to other tissues (brain, lungs, kidneys, etc.) in young and immunocompetent people have been described as "hypervirulent" (HvKp).

Beyond its clinical definition, the main microbiological characteristics that give HvKp its hypervirulent nature are a thick capsule, hypermucoviscosity (HMV) and the production of siderophores, molecules that help capturing iron necessary for infection. Encoded by various genes (subsequently associated with hypervirulence), these phenotypes have mainly been shown to enable HvKp to survive in adverse environments (such as the human body) and to avoid the immune system.

However, their precise role in HvKp’s infection process is currently unclear.

It has been shown that HvKp first colonises the intestine, before crossing the epithelial barrier to reach the portal system and form an abscess in the liver, but the mechanism of this switch from "commensal in the intestine" to "pathogenic in the liver" is unknown.

The hypothesis of this PhD project was therefore that HvKp presented specific interactions along the gut-liver axis, both to cross the gut barrier and to damage liver cells. The aim of this work was to verify this hypothesis by developing and validating an in vitro infection model to (1) explore the nature, interconnections and mechanisms of the interactions between gut/liver cells and HvKp and (2) establish the role of key microbe-associated virulence factors in the pathophysiology of these liver abscesses.

This model was based on ECIS (Electrical Cell Impedance System), a platform measuring the electrical resistance generated by a layer of eukaryotic cells, and its evolution in response to external stimuli (bacteria in our case) over time. Remarkably, this resistance is recorded at different current frequencies, representing various parameters: low frequencies (4kHz) mainly reflect tight junction (TJ) integrity and higher frequencies (64kHz) cell membrane integrity.

Cell lines representative of tissues of the gut-liver axis involved in the pathophysiological process of HvKp (Caco-2 for intestinal cells, HepaRG for liver cells and HUVECS for endothelial cells) were infected with different strains of HvKp, varying in their hypervirulent characteristics. These strains all to were SGH10, a classical HvKp strain belonging to the clonal group 23, SGH10 without its virulence plasmid (SGH10-p), and Ecl8, a strain belonging to a clonal group associated with hypervirulence but with attenuated virulence.

The first stage of our work involved characterising these strains, in terms of their bacterial phenotype, their interactions with eukaryotic cells and their lethality in a zebrafish larvae infection model. We were able to confirm that SGH10 was hypercapsulated, HMV and secreted siderophores; SGH10-p had lost this HMV characteristic and the secretion of siderophores; while Ecl8 was neither HMV nor hypercapsulated, but secreted siderophores. These strains’ specific characteristics were reflected in their interactions with gut and liver cells, with SGH10 adhering very poorly to epithelial cells, avoiding immune cells and significantly disseminating and killing zebrafish larvae. These results also demonstrate that HMV and hypercapsulation were key factors responsible for these interactions.

These differences between our strains allowed us to use them with confidence in our ECIS model. By first infecting intestinal Caco-2 cells, we demonstrated that SGH10 specifically disrupted the intestinal tight junctions more rapidly (up to 4 hours earlier) relative to SGH10-p and Ecl8, indicating a major role for HMV in this process. In the liver, the capsule seemed to be the main pathophysiological factor as SGH10 and SGH10-p both had a greater detrimental effect on the cell membrane than Ecl8.

All these findings were substantiated using alternative methods: the impact of SGH10 on intestinal TJs was confirmed using a TEER (Trans-epithelial electrical resistance) system, while the viability of liver cells was verified using flow cytometry. By studying the expression of TJ’s proteins (claudins, occludin, ZO-1), we were able to show that HvKp specifically downregulated claudin-1 and occludin expression, two proteins critical for epithelial integrity. Next, by adapting and developing our model, we were also able to show that these HvKp effects were specific to the gut-liver axis, that they required the presence of living bacteria, and that they were not linked to secreted factors or to the epithelial cells’ immune response.

We were finally able to show that, while there was no clear communication between the intestine and the liver, the HMV and hypercapsulated nature of HvKp meant that it passed through the intestinal epithelium using a paracellular route, taking advantage of the disruption of the tight junctions. Crucially, this preserved its subsequent toxicity on liver cells. Conversely, the strains that were neither HMV nor hypercapsulated were internalised in the intestinal cells, thus passing by a transcellular route, which was faster but resulted in a significantly reduced toxicity in the liver.

Overall, our model enabled us to describe for the first time that HvKp uses both its HMV and hypercapsule phenotype to specifically avoid contact with eukaryotic cells, allowing it to disrupt specific TJ proteins by downregulating claudin-1 and occludin, to pass through the intestinal epithelium via a paracellular route and to retain its significant toxicity when reaching the liver.