Viable and non-viable pathological bacterial translocation promote a self-perpetuating circle of dysfunctional immune activation and systemic inflammation facilitating infections and organ failure in advanced cirrhosis. Bacterial infections and sepsis are now recognized as a distinct stage in the natural progression of chronic liver disease as they accelerate organ failure and contribute to the high mortality observed in decompensated cirrhosis. The increasing knowledge of structural, immunological and hemodynamic pathophysiology in advanced cirrhosis has not yet translated into significantly improved outcomes of bacterial infections over the last decades. Therefore, early identification of patients at the highest risk for developing infections and infection-related complications is required to tailor the currently available measures of surveillance, prophylaxis and therapy to the patients in need in order to improve the detrimental outcome of bacterial infections in cirrhosis.

Bacterial infections are diagnosed in 25% to 47% of hospitalized patients with cirrhosis[1-4] and represent the most important precipitating event for acute decompensation[5]. Infections are increasingly recognized as a major trigger of systemic inflammation and organ failure in advanced cirrhosis leading to a four-fold increased mortality[6]. Despite recent advances in understanding the underlying pathogenic mechanisms of bacterial infections in cirrhosis[7-9], the progression of infections to multiple organ failure and septic shock is still associated with a short-term mortality of patients exceeding 75%[5,10]. In this review we will discuss susceptibility and impact of specific bacterial infections in cirrhosis, their natural course and the identification of risk factors for organ failure and death in order to help clinicians identifying patients at the highest risk that may benefit from intensified surveillance, prophylaxis and therapy.

Decompensated liver cirrhosis predisposes to delayed intestinal transit time, increased intestinal permeability and disturbed expression of intestinal antimicrobial peptides thereby facilitating the translocation of bacteria and bacterial products from the gastro-intestinal lumen through the lamina propria into mesenteric lymph nodes (MLN), ascitic fluid and the systemic circulation[11-13]. Whereas bacterial translocation (BT) can be controlled by the immune system of healthy individuals, pathological BT in cirrhotic patients is accompanied by a cirrhosis-associated immune dysfunction (CAID) supporting systemic inflammation as a consequence of non-viable BT as well as bacterial infections due to viable BT[8,9,14].

Among the intestinal microbiota, Gram-negative enteric bacilli more easily translocate to MLN than Gram-positive bacteria and obligate anaerobes[15]. Although the capability to translocate across the gastrointestinal barrier varies between different Escherichia coli (E. coli) strains[15], no single virulence factor responsible for BT in cirrhosis could be identified yet[16]. Furthermore, catabolic and inflammatory stress to the epithelial barrier can increase BT of various coliform bacterial strains[17-19]. With classical bacterial culture techniques, enteric organisms can be detected in MLN from up to 30% of patients with Child-Pugh C cirrhosis compared to 9%-15% of non-cirrhotic patients undergoing laparotomy[20,21].

This increased translocation of intestinal bacteria and bacterial products in cirrhosis has been attributed to both clinically significant portal hypertension (PH) and advanced liver failure. The role of PH is supported by the observations that the development of ascites is a prerequisite for significant translocation of viable bacterial to the MLN in the CCl4 model[22], and that the use of non-selective beta blockers (NSBB) reduces the translocation of viable BT to MLN in cirrhotic rats[23] and prevents infectious complications in humans with decompensated cirrhosis[24]. Furthermore, it has been shown that non-viable BT and immune activation also occur in the pre-ascitic stage[25] and correlate with the increase in portal pressure[26]. On the other hand, pathological BT to MLN also occurs in models of experimental acute liver failure in the absence of PH[27] but not in chronic PH in the absence of cirrhosis[28]. Moreover, NSBB are also effective in preventing spontaneous bacterial peritonitis (SBP) in patients showing no hemodynamic response to treatment[24] underlining that PH may not be the principal driver of viable pathological BT in advanced cirrhosis.

Therefore, alternative mechanisms must be involved in triggering pathological BT in liver failure and advanced fibrosis. There is evidence that quantitative and qualitative changes in the intestinal microflora might contribute to this phenomenon. Small intestinal bacterial overgrowth (SIBO) frequently occurs in decompensated cirrhosis and correlates with systemic endotoxemia[29] and the presence of bacterial DNA fragments (bactDNA) in the peripheral blood[30]. Small intestinal dysmotility[31] and altered bile composition in cirrhosis[32] contribute to SIBO and drive endotoxemia and infectious complications[33]. Once a systemic or local immune response has been established, proinflammatory cytokines like tumor necrosis factor (TNF) and interleukin (IL)-6 disrupt the intestinal integrity of the intestinal barrier by altering the structure of the apical junctional complex[34,35] thereby further increasing intestinal permeability (IP) generally observed in cirrhosis[12,26]. Although some cohort studies show an association of IP with septic complications[12,36,37] the only prospective study available did not find an association of IP with infectious complications[38] questioning whether increased IP is really the cause - or rather the consequence - of pathological BT and intestinal inflammation in cirrhosis.

Whether the composition of the intestinal microbiome, modulates BT and infectious complications in cirrhosis has not been thoroughly investigated yet. There is some evidence that the fecal microbiome in cirrhosis is less diverse and shows an abundance of Enterobacteriaceae, Streptococcus spp. and Enterococcus faecalis[39-41], which mirrors the microbial pattern observed in SBP[1,42,43]. The microbiome associated with the colonic mucosa might be of particular interest because the enrichment of Enterococcus, Proteus, Clostridium, and Burkholderia and the loss of non-pathogenic commensal bacteria in cirrhosis was associated with more severe liver disease, increased inflammation and endothelial activation[44,45]. The recent findings that the gut microbiome is modulated by the enterocyte inflammasome[46,47] which leads to increased nonviable BT and augmented hepatic inflammation[48], suggest a critical interaction of gut mucosa, intestinal dysbiosis and immune activation in cirrhosis.

To control viable BT in cirrhosis, immune control takes place at the epithelial barrier including the mucus layer, at gut-associated lymphatic tissue and the MLN and at the systemic level after bacteria or bacterial products have passed MLN via the portal vein or the thoracic duct[49]. In the context of pathological BT, organ-resident macrophages recognize pathogen-associated molecular patterns, such as lipopolysaccharide (LPS), muramyl dipeptide, bacterial lipoproteins and bactDNA, via extra- and intracellular Toll-like receptors (TLR) (e.g., TLR2, TLR4, TLR9) or intracellular NOD-like receptors (e.g., NOD2). Ligation of these conserved pattern recognition receptors (PRR) leads to classical macrophage activation, the secretion of pro-inflammatory cytokines (TNF, IL-1β, IL-12) and polarization to pro-inflammatory macrophages supported by interferon (IFN)-γ released from activated T cells[50]. In healthy individuals, binding of LPS to TLR4 increases expression of scavenger receptors, MHC class II and co-stimulatory molecules to accelerate phagocytosis and subsequent presentation of bacteria and bacterial products[51]. Several abnormalities of the innate immune system in cirrhosis contribute to non-control of BT, subsequent inflammation and the increased incidence of infections. These abnormalities have been described as CAID syndrome (CAIDS) stressing the role of insufficient processing of bacteria and bacterial products by phagocytes[8]. Importantly, CAIDS is often not associated with inflammatory anergy but with marked immune activation and a high degree of systemic inflammation, which correlates with the severity of liver disease and predicts survival in these patients[52].

The liver itself plays a major role in phagocytizing pathogens and scavenging macromolecules because organ-resident Kupffer cells and liver sinusoidal endothelial cells account for 80% of the human reticuloendothelial system (RES)[53]. The overwhelming sinusoidal influx of bacterial components in liver cirrhosis owing to mechanisms described above, switches the physiological immune-modulatory state of the local hepatic immune system elicited by Kupffer cells to a state in which the production of pro-inflammatory cytokines predominates accumulating in further demise of hepatic parenchyma cells and deterioration of liver architecture[54]. Due to portosystemic shunting, decreased opsonization and dysfunctional phagocytic activity, the hepatic clearance function is markedly reduced in cirrhosis and correlates with the severity of liver disease, bacteremia and survival[55-58]. In parallel, non-classical pro-inflammatory subsets of monocytes expand and contribute to an inflammatory state in cirrhosis[52,59-61]. This state of dysfunctional activation has been observed in neutrophils as well. Neutrophils from patients with cirrhosis are potent producers of pro-inflammatory cytokines and reactive oxygen species but also display decreased chemotaxis and inefficient phagocytosis[62-67]. Dysfunctional phagocytosis and increased activation can be transmitted to neutrophils from healthy individuals with plasma from cirrhotic patients suggesting repetitive priming mediated by soluble factors[66,67], which may result in the cellular depletion of antioxidants and increased oxidative damage[65,68].

Whereas phenotypic and functional abnormalities have been well described for the aspects of the innate immunity in cirrhosis, the state of the adaptive immunity is less well defined. Phenotypically, patients with cirrhosis display decreased numbers of total and naïve T cells due to defective thymic generation and splenic sequestration in parallel with increased activation, proliferation and turnover of memory T cells presumably due to repetitive antigen stimulation[69,70]. Data on the functional consequences of these findings are sparse. Although clinical studies show attenuated immune responses after vaccination in cirrhotic patients[71-73], in vitro analyses could not identify specific T cell defects[74,75] but suggest soluble factors such as IL-10 as modulators of inconsistent T cell responses in cirrhosis[75-77].

Whether an observed depletion and functional alteration of innate natural killer (NK) cells[78,79] and memory B cells[79,80] in cirrhosis contribute to pathological BT, inflammation and bacterial infections in cirrhosis has not been demonstrated.

In response to bacterial infection, patients with cirrhosis have a pronounced inflammatory response with elevated systemic concentrations of the pyrogenic cytokines IL-6 and TNF superimposing on high basal levels resulting in fever, leukocytosis and acute phase reaction[81]. As a consequence, systemic inflammatory response syndrome (SIRS) can be found in up to 67% of cirrhotic patients with bacterial infections compared to 37% of patients without, making sepsis a common complication of advanced cirrhosis[82-84].

Even in the absence of overt bacterial infection, the occurrence of systemic inflammation as indicated by SIRS is of prognostic relevance in patients with cirrhosis[83,85]. There is some evidence that beyond its role as an indicator of occult bacterial infection and advanced organ failure, systemic inflammation might even aggravate portal hypertension, renal failure and hepatic encephalopathy, thereby contributing to organ failure and death in cirrhotic patients by immunological, metabolic and hemodynamic mechanisms[83]. Due to hyperdynamic circulation with tachycardia or tense ascites leading to hyperventilation in patients with decompensated cirrhosis, SIRS criteria are most certainly less specific for inflammatory and infectious complications than in the general population and need to be interpreted with caution.

The magnitude of the acute phase response as indicated by C-reactive protein (CRP) or IL-6 levels might be more reliable tools than clinical SIRS criteria to estimate the risk for adverse outcome in cirrhotic patients. In patients with cirrhosis and SIRS concentrations of serum IL-6 correlate with organ failure, monocyte and neutrophil activation[86] but are also closely associated with the degree of portal hypertension[26] and hyperdynamic circulation[87]. The same observations have been made for CRP, where persistent elevation in the absence of infection indicates increased short-term mortality[88]. Discriminating infections from sterile systemic inflammatory response is only possible using CRP cut-offs are as high as 56 mg/L in advanced cirrhosis[89-91]. The large multicenter CANONIC study[5] identified elevated CRP levels and increasing WBC count as hallmark features to distinguish acute decompensation of cirrhosis from acute-on-chronic liver failure (ACLF) even when patients with bacterial infections were excluded. Patients who presented with ACLF had mean CRP levels of 33 mg/L (compared to 21 mg/L in patients without ACLF) and mean WBC counts of 9.4 Gpt/L (compared to 6.6 Gpt/L) at inclusion. Importantly, the probability of death in ACLF increased with the rise in WBC count[5] making WBC count an attractive, easily available linear variable to assess mortality risk associated with systemic inflammation in cirrhosis.

Since none of these established indicators of inflammation allow a precise distinction between sterile SIRS and bacterial infection, treating physicians must be aware of risk factors of bacterial infections to assess the likelihood of bacterial infections in patients with advanced cirrhosis presenting with signs of systemic inflammation. We have recently evaluated novel biomarkers that more precisely indicate specific infections and/or immune activation in advanced cirrhosis, such as mid-regional pro-adrenomedullin and soluble urokinase plasminogen activator receptor[52,90], which will expand the diagnostic armamentarium for predicting outcome independent of the presence of confounding systemic inflammation.

SBP and spontaneous bacteremia are the most thoroughly investigated complications in patients with cirrhosis because they occur frequently with a prevalence of 15%-20% in hospitalized patients[65,92] and cause high mortality after one month (33%-42%) and after one year (49%-66%)[6]. Supporting the concept of pathological BT of Enterobacteriaceae in cirrhosis, Gram-negative bacteria have been isolated from more than 70% of culture-positive bacterial infections in the past[1,42]. More recently however, bacterial infections caused by Gram-positive cocci dramatically increased in tertiary centers and now represent 60% of nosocomial culture-positive infections including the archetypal infectious complications spontaneous bacteremia and SBP[3,37,43]. This shift has been attributed to the use of antibiotics leading to intestinal dysbiosis favoring Gram-positive BT, but also to increased invasive procedures and associated episodes of Gram-positive bacteremia with secondary organ spread[3,43,93].

As a preventive measure, primary antibiotic prophylaxis is currently recommended in cirrhotic patients with gastrointestinal (GI) bleeding because bleeding facilitates pathological BT and infections on the one hand, and infections are associated with a higher rate of recurrent bleeding on the other hand[94]. Antibiotic prophylaxis over seven days in cirrhotic patients in with gastrointestinal bleeding is the current standard of care[94,95]. In nine controlled trials including 987 patients the pooled incidence of bacteremia following GI bleeding was 15% without antibiotic prophylaxis and 3% with antibiotic prophylaxis, indicating a risk reduction of 75% with antibiotic prophylaxis[96]. Results favoring antibiotic prophylaxis were also obtained when the outcomes SBP, pneumonia, urinary tract infection, overall bacterial infections and mortality were analyzed separately[96,97]. Although translocation of gut microbial flora into the bloodstream is likely to occur during endoscopy because of mucosal trauma related to the endoscopy procedure itself[98], the risk of bacteremia associated with endoscopic procedures is poorly investigated in non-bleeding cirrhotic patients. One study reported an incidence of 10% (6/58) bacteremia by possible contaminants (Gram-positive skin flora) after colonoscopy, but all cirrhotic patients with bacteremia remained asymptomatic[99]. Thus, international guidelines do not recommend routine antibiotic prophylaxis for cirrhotic patients undergoing colonoscopy[100] because of the lack of data in this setting, and it is recommended that clinicians should decide “on an individual case basis[101].

The meta-analysis by Arvaniti et al[6] comprising data from almost 12000 patients concluded a four-fold increased mortality owing to bacterial infections in cirrhotic patients - with respiratory tract infections, SBP and bacteremia as major contributors. We and others[84,123] could recently demonstrate that urinary tract infections in decompensated cirrhosis are less detrimental as these three major infections, but are still associated with a significantly increased risk of mortality, especially in the presence of concomitant SIRS. Fueled by excessive inflammation in cirrhosis, organ failure frequently occurs in the absence of septic shock, making bacterial infections a distinct stage in the natural history of cirrhosis progression[6] and the major precipitating event for the development of decompensation and ACLF[5]. There are three major factors that determine mortality of bacterial infection in cirrhosis irrespective of etiology: severity of underlying liver disease, concomitant renal failure and non-resolution of infection due to antimicrobial resistance.

Patients with decompensated cirrhosis represent a highly vulnerable population with structural, immunological and hemodynamic abnormalities, which render them susceptible to bacterial infections, pronounced systemic inflammation, organ failure, and death (Figure 1). Despite increasing understanding of underlying pathophysiology and mechanisms of organ failure, the outcome of bacterial infections in cirrhosis has remained poor over last decades. Antibacterial prophylaxis is effective in preventing infectious complications but increasing antimicrobial resistance demands its restriction to patients at the highest risk. Non-antibiotic approaches to decrease the risk of bacterial infections by reducing pathological BT (probiotics, prokinetics, bile acids) or restoring dysfunctional immune responses (anti-inflammatory strategies, immune therapy) are desperately needed as we enter the post-antibiotic era. Until then, basic measures of infection prevention (vaccinations, hygiene, nutrition), antimicrobial de-escalation strategies, and surveillance for early signs of organ dysfunction are required to reduce the burden associated with bacterial infections in cirrhosis.

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Self-perpetuating mechanisms contributing to bacterial infections, systemic inflammation and organ failure in patients with decompensated cirrhosis. BT: Bacterial translocation.