Bacterial translocation in cirrhosis is not caused by an abnormal small bowel gut microbiota


Correspondence: Sandra Macfarlane, Gut Group, University of Dundee, Ninewells Hospital, Dundee DD1 9SY, UK. Tel.: +44 0 1382 632 535; fax: +44 0 1382 633 952; e-mail:


Sepsis is common in liver cirrhosis, and animal studies have shown the gut to be the principal source of infection, through bacterial overgrowth and translocation in the small bowel. A total of 33 patients were recruited into this study, 10 without cirrhosis and 23 with cirrhotic liver disease. Six distal duodenal biopsies were obtained and snap frozen for RNA and DNA extraction, or frozen for FISH. Peripheral venous bloods were obtained from 30 patients, including 17 chronic liver disease patients. Samples were analysed by real-time PCR, to assess total bacteria, bifidobacteria, bacteroides, enterobacteria, staphylococci, streptococci, lactobacilli, enterococci, Helicobacter pylori and moraxella, as well as TNF-α, IL-8 and IL-18. There was no evidence of bacterial overgrowth with respect to any of the individual bacterial groups, with the exception of enterococci, which were present in higher numbers in cirrhotic patients (P = 0.04). There were no significant differences in any of the cytokines compared to the controls. The small intestinal mucosal microbiota in cirrhotic patients was qualitatively and quantitatively normal, and this shifts the focus of disease aetiology to factors that reduce gut integrity, failure of mechanisms to remove translocating bacteria, or the large bowel as the source of sepsis.


Cirrhotic liver disease is an increasing problem in Scotland, in terms of social, economic and health care costs. Patients with cirrhosis are at particular risk of bacterial infections, especially spontaneous bacterial peritonitis (SBP) and hepatic encephalopathy. Overt infections, hepatic decompensation (liver failure) and some episodes of hepatic encephalopathy may in part be driven by endogenous sources of infections in these patients. Most interest focuses on the gut as the source of infections at present. The intestinal microbiota also has a major role to play in hepatic encephalopathy.

The incidence of bacterial infections in cirrhotic patients ranges from 15% to 47% (Yoshida et al., 1993). About one-third of deaths in cirrhotic patients are attributable to sepsis. Seventy to eighty per cent of these infections are attributable to Gram-negative bacteria, predominantly Escherichia coli, which can be isolated from ascitic fluid (Garcia-Tsao, 1992; Thulstrap et al., 2000). The gut has been shown in animals to be the source of these organisms, which invade through a mechanism called bacterial translocation (Pérez-Paramo et al., 2000). This supposes that viable microorganisms migrate from the intestinal lumen to mesenteric lymph nodes and other organs. Animal work suggests that the threshold for bacterial translocation is lower in the small bowel than in the large gut (Koh et al., 1996). More recently, TNF-α has been used as a surrogate measure for bacterial translocation, and patients with high TNF-α levels were found to have high Child-Pugh (CP) scores, a measure of the severity of liver disease, and were the only ones to develop bacterial infections during the first month post-transplant (Genescà et al., 2003).

Examination of jejunal aspirates in chronic alcoholics found a higher incidence of anaerobic bacteria (Bode et al., 1984). More recently, a study of severe chronic hepatitis patients found that patients with severe disease had reduced levels of faecal bifidobacteria and bacteroides, together with raised levels of yeasts and enterobacteria in their faeces, and that these changes were associated with raised levels of serum TNF-α (Li et al., 2001). However, faecal samples are not entirely representative of the intestinal microbiota, and if stored, even for a relatively short period of time, qualitative and quantitative alterations in the microbial communities can occur (Crowther, 1971; Poxton et al., 1997).

Hepatic encephalopathy can be associated with acute and chronic hepatocellular failure. Traditionally, this has been associated with bacterial products from the gut because of the beneficial effects of lactulose and neomycin therapies (Weber et al., 1987). Most studies measuring amino acids in humans, dogs, monkeys and rats with hepatic encephalopathy and chronic liver disease demonstrated grossly abnormal plasma concentrations; with raised levels of aromatic amino acids (AAA), reduced levels of branched-chain amino acids (BCAA) (leucine, isoleucine and valine) and a reduced BCAA : AAA ratio. Humans are unable to synthesize BCAA, which are crucial in protein and neurotransmitter synthesis, and depend on the gut microbiota (Macfarlane et al., 1992). Other well-known abnormalities of protein metabolism correlating with encephalopathy include elevated plasma ammonia, mercaptans and tyramine, as well as a reduced capacity to synthesize urea.

Ammonia is principally a product of amino acid fermenting bacteria in the large bowel and of the effects of glutaminase on glutamine in the small bowel (Smith & Macfarlane, 1997, 1998). The ammonia produced in the small bowel is thought to be more important because encephalopathy can still occur in germ-free animals (Schalm & Van Der May, 1979). There are other toxins produced by colonic bacteria that have been implicated in hepatic encephalopathy, such as gamma-amino-butyric acid and benzodiazepine-like substances (Schafer et al., 1981; Yuraydin et al., 1995). If changes to gut microbiotas are sufficient to cause malabsorption in patients with inflammatory bowel disease, then it is likely that a similar mechanism for altered BCAA : AAA ratios may exist in chronic liver disease.

The aim of this exploratory study was to facilitate identification, quantification and visualization of specific bacterial species colonizing the small bowel mucosa of patients with chronic liver disease, and to compare these with individuals who did not have liver disease. Secondary outcomes included assessing whether or not any particular differences were associated with evidence of bacterial translocation, by measuring TNF-α levels.

Materials and methods


The project was based in Ninewells Hospital and Medical School, which is situated in the east of Scotland, Tayside. There are approximately 10 000 people in Tayside who have been identified with liver disease (Steinke et al., 2002). Ethical approval for the study was granted by the Tayside Research Ethics Committee.


A total of 33 consecutive patients were recruited prior to attending for a pre-arranged upper gastrointestinal endoscopy in the hospital. Their ages were between 18 and 70 years, 10 were without liver disease, while 23 had known cirrhosis. Patients were excluded if they were on or had received probiotics or antibiotics in the last 3 months. They completed a questionnaire relating to medical history and lifestyle, including an Alcohol Use Identification Test (AUDIT) questionnaire. Baseline patient characteristics are shown in Table 1.

Table 1. Baseline characteristics of patients involved in the study
Average age5657
Proton pump inhibitor use412
AUDIT score2 (low risk  drinking)6 (low risk  drinking)
Childs-Pugh A18
Childs-Pugh B5
Childs-Pugh C0
Alcoholic liver  disease (ALD)14
Primary billiary  cirrhosis (PBC)2
Hepatitis B (HepB)1
Hepatitis C (HepC)1
Nonalcoholic  steatohepatitis (NASH)2
Cystic fibrosis1

Peripheral venous blood samples were obtained from 30 patients, 8 controls, 4 nonalcoholic fatty liver disease, 1 post-liver transplant and 17 chronic liver disease patients (10 Childs A, 5 Childs B, 2 Childs C). Childs grade A indicates well-compensated disease, grade B significant functional compromise and grade C decompensated disease. Of the patients with chronic liver disease, seven had primary sclerosing cholangitis (PSC), one had autoimmune hepatitis, two had nonalcoholic steatohepatitis and seven had alcoholic liver disease (ALD). One of the PSC patients was on long-term antibiotics and two of the ALD patients had received antibiotics within the last month.


In consenting patients, six extra duodenal (D2) biopsies were obtained during the endoscopic procedure and transported to the laboratory, where they were stored at −80 °C for subsequent microbiological and immunological analyses. Biopsies for FISH were covered in a freezing medium (Tissue-Tek; Sakura Finetek, the Netherlands) to protect the bacteria and to preserve mucosal integrity.

DNA extraction from the mucosa

Bacterial DNA was extracted from mucosal biopsies and peripheral venous blood according to manufacturer's instructions using a DNeasy blood and tissue kit (Qiagen, West Sussex, UK).

RNA and cDNA preparation

Samples were macerated using liquid nitrogen snap-freezing and mechanical grinding, before undergoing a clean-up stage using a Qiashredder column (Qiagen) and RNA purification using the RNeasy kit (Qiagen). Biopsies were reverse-transcribed using the AMV RT kit (Promega, Madison, WI) as per manufacturer's instructions, before being aliquoted for storage at −80 °C.

Preparation of standards for quantification of DNA

Bacterial DNA and cDNA from mononuclear blood cells were amplified using specific PCR primer pairs (see Table 2). Product of correct sequence and size was purified using the Qiaquick PCR purification kit (Qiagen) and ligated into a vector using the pGEM-T easy vector system I (Promega). JM109 competent E. coli (Promega) were transformed with each ligated vector of cDNA or DNA, and positive colonies were selected after overnight incubation. The plasmid was purified from each colony using the Wizard plus SV miniprep system (Promega). Agarose gel electrophoresis with known standards (New England Biolabs, Beverly, MA) was used to determine concentration of plasmid preparation, and samples were diluted to 1010 molecules μL−1, aliquoted and stored at −80 °C.

Table 2. Specific PCR primer pairs used for bacteriological and cytokine analyses
Primer setTarget groupSequenceAnnealing temperature (oC)Product size (bp)Reference
Uni 330F Uni 530RAll eubacteriaACTCCTACGGGAGGCAGCAGT GTATTACCGCGGCTGCTGGCAC58200Nadkarni et al. (2002)
Entero379F Entero568REnterobacteriacaeCATTGACGTTACCCGCAGAAGAAGC CTCTACGAGACTCAAGCTTGC63195Bartosch et al. (2004)
Coccus1135F Coccus1275REnterococcus genusCCCTTATTGTTAGTTGCCATCATT ACTCGTTGTACTTCCCATTGT61144Rintilla et al. (2004)
Staph961F Staph1167RStaphylococcus genusCTTACCAAATCTTGACATCCTTTGAC CCACCTTCCTCCGGTTTGTCACC62207A.R. Smith, S. Macfarlane, G.T. Macfarlane (unpublished data)
Bif 8F Bif 164RBifidobacterium genusAGGGTTCGATTCTGGCTCAG CATCCGGCATTACCACCC62156Kok et al. (1996)
Lact368F Lact693RLactobacillus genusAGCAGTAGGGAATCTTCCA CACCGCTACACATGGAG58341Walter et al. (2001), Heilig et al. (2002)
Bac 605F Bac 732RBacteroides genusGTCAGTTGTGAAAGTTTGC CAATCGGAGTTCTTCGTG55127Bernhard & Field (2000)
E. coli 395F E. coli 470REscherichia coliCATGCCGCGTGTATGAAGAA CGGGTAACGTCAATGAGCAAA6095Huijsdens et al. (2002)
Hpylori 402F Hpylori 520RHelicobacter pyloriGAA GAT AAT GAC GGT ATC TAA C ATT TCC ACC TGA CTG ACT AT58139Rintilla et al. (2004)
TNF-αF TNF-αRTumour necrosis factor alphaTCT CGA ACC CCG AGT GAC AA TAT CTC TCA GCT CCA CGC CA56123Furrie et al. (2005a)
IL-18F IL-18RInterleukin 18GAC GCA TGC CCT CAA TCC CTA GAG CGC AAT GGT GCA ATC58105Boeuf et al. (2005)

Real-time quantitative PCR

The appropriate plasmid preparation was diluted to give a standard curve of 106–101 molecules μL−1 for all assays, except glyceraldehydes-3-phosphate dehydrogenase (GAPDH) which had a standard curve of 108–101 molecules μL−1 (Furrie et al., 2005a,b). Real time PCR was carried out using an iCycler and the iQ SYBR Green Supermix (Biorad, Hercules, CA). Test samples were added in duplicate at 2 μL per well in a 20 μL total reaction volume.


Samples were stored in Tissue-Tek (Sakura Finetek) and frozen at −80 °C until FISH analysis, when they were sectioned into 10-μm sections with a cryomicrotome, and placed on Teflon-coated 10-well glass slides (VWR; Merck Eurolab). After air-drying, they were fixed in 4% paraformaldehyde in PBS (pH, 7.0) for 1 h, washed with PBS (10 min), and dehydrated in 50%, 80% and 96% ethanol (3 min each). After drying, the cy3 and fluorescein isothiocyanate (FITC) labelled oligonucleotide probes (Invitrogen) were used at concentrations of 30 and 50 ng μL−1, respectively (Table 3). Samples were allowed to hybridize overnight on glass slides at temperatures depending on the type of probe (Table 3). Hybridized slides were washed with hybridization buffer, rinsed with distilled H2O and air-dried. Citifluor (Citifluor Ltd, London, UK) was used as a mounting medium, and the slides were visualized with a Nikon Eclipse E800 upright microscope attached to a Nikon PCM 2000 confocal system. Bacteria labelled with FITC probes were detected with a 488-nm argon laser, and cells labelled with cy3 probes were detected with a 543-nm helium-neon laser. Images were captured and overlaid with c-imaging software (Compix Inc., Cranberry Township, PA). Unsectioned samples were also analysed under direct examination with the eubacterial probe using the method described above.

Table 3. FISH probes used in this study
ProbeTarget bacterial group/speciesSequenceFTLReference
  1. Hybridization conditions: F, formamide concentration in hybridization buffer (%); T, hybridization temperature (°C); L, lysozyme treatment (min).

Eub338Universal eubacterial probeGCTGCCTCCCGTAGGAGT0500Amann et al. (1990)
Bif164Bifidobacterium spp.CATCCGGCATTACCACCC0450Langendijk et al. (1995)
Lacb0722Lactic acid bacteriaYCACCGCTACACATGRAGTTCCACT205415Sghir et al. (1998)
EntEnterobacteriaceae except Proteus spp.CCCCCWCTTTGGTCTTGC30500Kempf et al. (2000)
Bac303Bacteroides/PrevotellaCCAATGTGGGGGACCTT0450Manz et al. (1996)
Enc131EnterococciCCCCTTCTGATGGGCAGG0500Meier et al. (1997)


Significant differences between the cirrhotic and noncirrhotic groups were assessed for bacterial DNA and cytokine results using the Mann–Whitney test for nonparametric analysis and a t-test for parametrically distributed data, with 95% confidence intervals, which were themselves statistically analysed using the F-test to allow comparisons of components of variance. Significance was given for P-values < 0.05.


Bacteriological analyses

All results were obtained and analysed in the form of mean gene copy numbers according to the housekeeping gene GAPDH, to normalize for biopsy size (Table 4). Total bacteria were present in the biopsy material in numbers ranging from 101 to 102. There was no statistical difference in mean total numbers of bacteria per 1000 GAPDH molecules between the cirrhotic and normal groups. There was no statistical difference in mean total numbers of enterobacteria between the cirrhotic and normal groups. Mean numbers of enterococci were significantly higher in the chronic liver disease group than in the control group (P = 0.0416). However, when these results were analysed in stratified for proton pump inhibitor (PPI) use there was no significant difference. There was no statistical difference in mean total numbers of lactobacilli between the cirrhotic and normal groups or in mean total numbers of bifidobacteria in the small bowel of either group. There was no statistical difference in mean total numbers of bacteroides between the two groups, or in mean total numbers of E. coli. Similarly, no statistical differences were seen with respect to staphylococci, moraxellas or Helicobacter pylori between the two groups.

Table 4. Results of gene copy number per bacteria and TNF-α mRNA molecules in the normal and cirrhotic group, with subgroup analysis of patients taking a proton pump inhibitor (PPI)
Group (patients)EubacteriaEnterobacEnterocoBifidLactBactE. coliH. pyloriStaphMoraxellaTNF-α
  1. Mean 16S rRNA gene copy number per 1000 GAPDH molecules for each bacterial group ± 95% CI.

  2. a

    Significant P < 0.05.

Normal (10)48.9 ± 13.93.6 ± 0.70.2 ± 0.21.1 ± 0.52.1 ± 0.60.4 ± 0.13.2 ± 0.70.2 ± 0.10.1 ± 0.01.0 ± 1.71383 ± 486
Cirrhotic (23)94.9 ± 16. 74.9 ± 1.70.5 ± 0.3a4.5 ± 1.52.5 ± 0.80.9 ± 0.24.3 ± 1.50.5 ± 0.10.2 ± 0.03.2 ± 2.11192 ± 177
PPI (15)62.9 ± 17.36.3 ± 2.20.6 ± 0.27.1 ± 2.6a3.6 ± 1.21.1 ± 0.2a5.6 ± 2.20.5 ± 0.20.3 ± 0.03.5 ± 1.3972 ± 158
Non-PPI (18)110 ± 16.93.2 ± 0.60.3 ± 0.11.5 ± 0.51.4 ± 0.30.4 ± 0.12.6 ± 0.60.2 ± 0.10.1 ± 0.01.7 ± 0.81481 ± 313
Normal: PPI (4)66.9 ± 34.02.8 ± 1.40.3 ± 0.22.0 ± 1.23.3 ± 1.50.6 ± 0.32.0 ± 0.70.3 ± 0.20.2 ± 0.02.3 ± 1.91203 ± 413
Normal: non-PPI (6)36.9 ± 6.54.1 ± 0.80.04 ± 0.010.5 ± 0.11.4 ± 0.20.2 ± 0.03.7 ± 1.00.1 ± 0.00.1 ± 0.00.2 ± 0.01506 ± 795
Cirrhotic: PPI (10)125 ± 18.17.6 ± 3.30.7 ± 0.27.8 ± 0.33.7 ± 1.61.2 ± 0.36.7 ± 2.90.5 ± 0.20.3 ± 0.14.0 ± 1.7888 ± 163
Cirrhotic: non-PPI (13)74.9 ± 24.62.7 ± 0.90.4 ± 0.21.9 ± 0.71.9 ± 0.70.5 ± 0.22.0 ± 0.70.3 ± 0.00.2 ± 0.02.4 ± 1.71470 ± 288

Cytokine analyses

Results are shown in Table 5. There were no significant differences observed in mean normalized gene copy numbers between the cirrhotic and normal groups in levels of TNF-α, IL-18 or IL-8. However, levels of IL-18 were significantly higher in the normal population on PPI than those in the cirrhotic population on PPI.

Table 5. Cytokine analysis of normal and cirrhotic patients and subgroup analysis for proton pump inhibitor use
Group (patient number)TNF-αIL-18IL-8
  1. a

    P = 0.007, significantly higher than the cirrhotic PPI group.

Normal (10)1383 ± 4861648 ± 435779 ± 252
Cirrhotic (23)1192 ± 1771442 ± 323912 ± 276
PPI (15)972 ± 1581281 ± 312799 ± 256
Non-PPI (18)1481 ± 3131690 ± 398921 ± 303
Normal: PPI (4)1203 ± 4132568 ± 881a887 ± 447
Normal: non-PPI (6)1506 ± 7951034 ± 264725 ± 333
Cirrhotic: PPI (10)888 ± 163813 ± 319770 ± 320
Cirrhotic: non-PPI (13)1470 ± 2882018 ± 5671028 ± 439

CP scores

There was no significant difference between current CP grade A and grade B cirrhosis in any of the bacterial or cytokine groups analysed above. This group was then divided into CP grade A, B and C according to their worst grade from their past medical history, which afforded 11 patients who had never been worse than a grade A, 7 grade B and 5 grade C cirrhosis. There was no statistical difference between the mean gene copy numbers per 1000 GAPDH molecules for eubacteria between any of the three groups. Means were 122 ± 57, 47 ± 18 and 67 ± 56, respectively.

Proton pump inhibitors

The cirrhotic and noncirrhotic groups were also analysed by separating them into those on PPI and those not on PPI for each bacteria and cytokine. Statistically, when comparing the total number of patients receiving a PPI (15) against those who were not (18), there was a significant increase in the mean number of bifidobacteria (P = 0.0133) and bacteroides (P = 0.0173) per 1000 GAPDH molecules only. Subanalysis of all groups, that is, cirrhotics on PPI against cirrhotics not on PPI, and normal volunteers on a PPI against those who were not, showed no statistical difference in any of the bacteria or cytokine analysed. PPI did not alter the total number of bacteria or lactobacilli, enterobacteria, enterococci, E. coli, H. pylori, staphylococci, moraxella or TNF-α (Table 6).

Table 6. PPI use as a confounding factor in bacteriological analysis
Group (patients)EubacteriaEnterobacEnterocoBifidLactBactE. coliH. pyloriStaphMoraxella
  1. a

    P = 0.013.

  2. b

    P = 0.017.

PPI (15)62.9 ± 17.36.3 ± 2.20.6 ± 0.27.1 ± 2.6a3.6 ± 1.21.1 ± 0.2b5.6 ± 2.20.5 ± 0.20.3 ± 0.03.5 ± 1.3
Non-PPI (18)110 ± 16.93.2 ± 0.60.3 ± 0.11.5 ± 0.51.4 ± 0.30.4 ± 0.12.6 ± 0.60.2 ± 0.10.1 ± 0.01.7 ± 0.8


The FISH analysis was done on all of the biopsies, and showed minimal bacterial colonization of the mucosal surfaces (not shown).

Venous blood analyses

Blood samples were analysed for E. coli and no statistical difference was found in levels of these bacteria in the liver disease group and control groups.


This study casts doubt on a major belief regarding the pathophysiology of the decompensation of chronic liver disease; namely that the gut, and in particular the small bowel, is the site of bacterial overgrowth that spills over to cause translocation of bacteria. This study clearly shows that the small bowel mucosal microbiota in patients with cirrhosis is qualitatively and quantitatively normal. Therefore, the bacteria are not likely to be the initiating factor in any translocation events. This does not remove translocation as a pathogenic mechanism, but puts the focus on factors that reduce bowel integrity or failure of mechanisms to remove translocating microorganisms.

These results contradict the animal data and the jejunal aspirate data that there is overgrowth of bacteria in the small bowel of patients with chronic liver disease. With the exception of enterococci, none of the pathogenic species found as the cause of infections in cirrhotic patients such as enterobacteria, staphylococci and moraxella were present in significantly higher numbers (Betzl et al., 1990), suggesting that there is no real bacterial overgrowth in the small bowel in stable cirrhotics. Therefore, bacterial overgrowth may not be the cause of translocation and subsequent infection as it was previously suspected to be. The results that are considered statistically significant are still dealing with relatively small numbers of bacteria, numbers so low that they cannot be converted to a logarithmic scale. The TNF-α was used as surrogate marker of translocation, and there was no evidence of different levels between the populations, from which we can infer that no translocation was taking place. Interleukin-8 is a chemokine, and an important mediator in the innate immune response. With IL-18, it works together with other cytokines to induce a cell-mediated immune response following administration of lipopolysaccharide. Again, the equivalent levels in both patients in this investigation group suggest that there was no evidence of bacterial translocation taking place.

There were no patients with Childs C cirrhosis in this study (it is difficult to find such patients not on, or recently exposed to antibiotics), however, many of the patients had recently been Childs C, and so a significant difference in microbiota composition between the two classes would be unlikely. Many of the patients with ALD in this group were abstaining or drinking at low risk levels, so no inference can be made on the effects of alcohol on total numbers of bacteria in the small bowel. Interestingly, PPI which are usually regarded as a confounding factor in bacterial population analysis, as a result of hypochlorhydia, did not alter the total number of mucosal bacteria, or the numbers of most individual species, despite widely regarded evidence that bacteria are sensitive to pH (O'May et al., 2005). Other in vitro work does support the principal that PPI can alter the growth and morphology of bacteria (Altman et al., 2008), and duodenal aspirate cultures have shown dramatic bacterial overgrowth in patients treated with these drugs (Lewis et al., 1996). Support for this theory tend to come from traditional microbiology culturing techniques, which have some limitations. There is little evidence using sensitive molecular techniques to support or refute this theory, and trials looking for an association between PPI and SBP in cirrhotic patients have failed to find a link (Campbell et al., 2008).

Lactic acid bacteria require amino acids for growth, specifically BCAA (Read et al., 1966). The BCAA account for 20% of the total protein amino acids in Lactobacillus lactis and E. coli (Loguercio et al., 1995). Most studies measuring amino acids in humans, dogs, monkeys and rats with hepatic encephalopathy and chronic liver disease demonstrated grossly abnormal concentrations; with raised levels of plasma AAA, reduced levels of BCAA and a reduced BCAA : AAA ratio (Marchesini et al., 2003). It might have been expected that lactobacilli would be present in increased numbers in the small bowel, and play a role in contributing to the altered circulating amino acid ratio that is related to encephalopathy, but the results show that there is no statistical difference in the numbers of lactobacilli, suggesting that any ratio alteration is related to substrate ingestion rather than bacterial metabolism of amino acids.

Bifidobacteria are considered to promote good health and are often numerically reduced in diseases of the gut (Macfarlane et al., 2004); however, they were present at the same levels in cirrhotic patients as in controls.

With no evidence of bacterial overgrowth in the small bowel, the case for probiotic use in treating patients with liver disease is significantly weakened. More positively, any concerns that probiotics would confound an existing problem, by adding to the burden of bacterial overgrowth, is alleviated and it would be reasonable to proceed with probiotic trials in liver disease with the purpose of modulating the immune response, promoting intestinal barrier integrity by regulating tight junctions or displacing more potentially pathogenic bacteria. There have been mixed results of trials with probiotics and synbiotics in liver patients, and substrate manipulation using BCAA oral supplements has led to improved nutritional status, reduced frequency of complications, reduced total number of days in hospital and improvements in the CP classification (Mills & Thomas, 1921; James et al., 1976; Garault et al., 2000). This warrants further investigation, so that future work could be aimed at developing a targeted probiotic or synbiotic to reduce the incidence of infectious complications and hepatic encephalopathy in patients with chronic liver disease, leading to improved nutritional status, as has been demonstrated in patients with ulcerative colitis (Furrie et al., 2005a).

This pilot human study has contradicted the animal evidence in this area and the one human trial, which was limited to looking at luminal bacteria rather than mucosal samples, and has not found any evidence of bacterial overgrowth in the small bowel. If bacterial translocation is the source of infections in patients with chronic liver disease, there is no evidence that it is occurring in the small bowel in grade A and B cirrhosis. The theory of bacterial translocation caused by intestinal bacterial overgrowth in this patient group has been thrown into doubt and when combined with the mixed results obtained from intestinal permeability studies in this group and the major mechanisms for the causes of such translocation are rapidly diminishing (Feld et al., 2006; Kalaitzakis et al., 2006). Equally it would appear unlikely that bacterial translocation from the gut is the initiating factor in hepatic decompensation, although once this has occurred, for instance a variceal bleed with hypotension, or massive ascites causing mechanical distortion, then gut integrity may be impaired and allow translocation of normal bowel microbial communities to contribute as a promoting factor to the syndrome of hepatic decompensation.

The gut remains the most likely source of bacteria causing infectious episodes in cirrhosis, and therefore it would be reasonable to examine the large bowel for quantification of bacterial species and immune function, but any difference would require a credible explanation as to why the small bowel has been spared. If the large intestine could be identified as the source of infection in cirrhosis, then this opens up the possibility of nutritional intervention in this patient group, since probiotics, prebiotics and synbiotics are known to have beneficial properties in the large bowel.

Research involving greater numbers of patients is of course required, and an expansion to include patients with chronic ascites and Childs C cirrhosis may indeed yield a difference in bacterial populations. Should that prove the case, then these results support the hypothesis that on improvement of the cirrhosis, the microbiota also reverts back to a precirrhotic state. Also of use would be a more detailed examination of the local immune function of the intestine in cirrhotic patients, which may further illuminate whether or not the gut is actually continuing to function at a level of relative normality in this high morbidity/mortality patient group.


The authors thank the Foundation for Liver Research, London and the Tayside Tissue Bank.