Increased intestinal permeability and tight junction alterations in nonalcoholic fatty liver disease


  • Potential conflict of interest: Nothing to report.


The role played by the gut in nonalcoholic fatty liver disease (NAFLD) is still a matter of debate, although animal and human studies suggest that gut-derived endotoxin may be important. We investigated intestinal permeability in patients with NAFLD and evaluated the correlations between this phenomenon and the stage of the disease, the integrity of tight junctions within the small intestine, and prevalence of small intestinal bacterial overgrowth (SIBO). We examined 35 consecutive patients with biopsy-proven NAFLD, 27 with untreated celiac disease (as a model of intestinal hyperpermeability) and 24 healthy volunteers. We assessed the presence of SIBO by glucose breath testing (GBT), intestinal permeability by means of urinary excretion of 51Cr-ethylene diamine tetraacetate (51Cr-EDTA) test, and the integrity of tight junctions within the gut by immunohistochemical analysis of zona occludens-1 (ZO-1) expression in duodenal biopsy specimens. Patients with NAFLD had significantly increased gut permeability (compared with healthy subjects; P < 0.001) and a higher prevalence of SIBO, although both were lower than in the untreated celiac patients. In patients with NAFLD, both gut permeability and the prevalence of SIBO correlated with the severity of steatosis but not with presence of NASH. Conclusions: Our results provide the first evidence that NAFLD in humans is associated with increased gut permeability and that this abnormality is related to the increased prevalence of SIBO in these patients. The increased permeability appears to be caused by disruption of intercellular tight junctions in the intestine, and it may play an important role in the pathogenesis of hepatic fat deposition. (HEPATOLOGY 2009.)

Nonalcoholic fatty liver (NAFLD) is one of the most common forms of chronic liver disease throughout the world.1 It includes a spectrum of pathological changes ranging from the simple accumulation of fat in the liver through nonalcoholic steatohepatitis (NASH) to fibrosis, cirrhosis, and even hepatocellular carcinoma. Considerable progress has been made toward understanding the pathogenesis of steatosis and the roles played in this process by insulin resistance, dyslipidemia, and obesity, but much less is known about the mechanisms underlying the transition from steatosis to steatohepatitis. There is a growing body of evidence suggesting that the progression of NAFLD depends on interactions between genetic2 and environmental factors.3 The latter include bacterial translocation through the intestinal wall and small intestinal bacterial overgrowth, which are believed by some authors to contribute to alcoholic steatohepatitis. Studies conducted in mice4, 5 and humans6 indicate that both these factors also may be involved in the pathogenesis of NASH. In light of this preliminary, and as yet unconfirmed, evidence, the current study was designed to investigate the prevalence and underlying mechanisms of intestinal hyperpermeability in patients with biopsy-confirmed NAFLD. For comparison, we also examined healthy volunteers and patients with untreated celiac disease, a condition well known to be associated with increased gut permeability.


51Cr-EDTA, 51Cr-ethylene diamine tetraacetate; BMI, body mass index; CD, celiac disease; GBT, glucose breath testing; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; SIBO, small intestinal bacterial overgrowth; TJ, tight junctions; ZO-1, zonula occludens-1.

Patients and Methods

The protocol for this observational, noninterventional study was preapproved by the institutional review board at our Institution. All participants provided written informed consent to all study procedures and to publication of the results.

Patients and Controls

The study population consisted of 35 patients with biopsy-proven NAFLD who were consecutively seen in the outpatient Liver Unit of our medical center for chronically (at least 6 months) elevated aminotransferase levels of unknown origin and ultrasound evidence of hepatic steatosis. Exclusion criteria were as follows: refusal to undergo upper gastrointestinal endoscopy, smoking (current or past), significant alcohol consumption (>20 g/day),7 a history of viral hepatitis, diarrhea, diverticulosis, irritable bowel syndrome, inflammatory bowel diseases, or bariatric or other abdominal surgery, seropositivity for anti-endomysial immunoglobulin A antibodies, impaired renal function (which can interfere with intestinal permeability testing),8 the use of nonsteroidal anti-inflammatory drugs, antibiotics, probiotics, or antisecretory drugs capable of causing achlorhydria within the 2 months preceding enrollment, or evidence of immunoglobulin A or immunoglobulin deficiency (both of which produce confounding effects during assessments of intestinal permeability and small intestinal bacterial overgrowth [SIBO]).

For comparison purposes, we enrolled two other groups of subjects, matched for sex, age, and ethnic origin (white) with those of the study group. The first consisted of patients with newly diagnosed celiac disease (CD) who had not commenced on a gluten-free diet. The second was composed of healthy volunteers from hospital and university staff, who had normal serum aminotransferase levels, no evidence of a “bright liver” on hepatic ultrasound examination, and no family history of diabetes, metabolic disease, CD, or dyslipidemia.

Complete medical histories and the results of the physical examination were recorded for all participants. Metabolic syndrome was diagnosed when at least three of the five criteria recommended by the National Cholesterol Education Program's Adult Treatment Panel III9 were present. The degree of insulin resistance was estimated with the homeostatic model assessment equation, as follows: (insulin × glucose)/22.5.10 Body mass index (BMI) was calculated as the ratio of body weight to height [kg × (m2)−1]. Patients with BMIs between 25 and 29.9 were considered overweight; those with BMIs ≥ 30, obese.

Liver Biopsy

A single pathologist (F.M.V.) reviewed the liver biopsies (at least 1.6 cm length and 5 μm thick) from all patients with NAFLD. The tissues were stained with hematoxylin and eosin, Masson's trichrome stain, and periodic acid-Schiff stain and examined under blinded conditions by a single experienced pathologist. The Clinical Research Network (CRN)-NAFL Activity Score scoring system was applied for staging NAFLD.11 The diagnosis of NASH was made when NAFL Activity Score was greater than 5.

Upper Gastrointestinal Endoscopy and Duodenal Biopsy

Upper gastrointestinal endoscopy was performed on all participants (including healthy controls) with a high-resolution, high-magnification (2×) videoendoscope (Fuji, model EG-485ZH, Omiya, Japan). The duodenal villous profile was evaluated with the immersion technique, and biopsy specimens were collected from the descending duodenum (two from the first and two from the second portion) according to current guidelines.12 The diagnosis of CD disease was confirmed with evidence from duodenal biopsy specimens of typical features of villous pattern according the protocol previously reported.13

Chromium-51 Ethylene Diamine Tetraacetate Excretion Testing

The urinary excretion of chromium-51 ethylene diamine tetraacetate (51Cr-EDTA) was measured as an index of intestinal permeability 1 week before the liver biopsy. After an overnight fast, patients drank 1.85 MBq 51Cr-EDTA (Amersham Health, England) in 10 mL water, and urine was collected for the next 24 hours. Two 3-mL samples of the collected urine were assessed with a gamma counter (LKB-Wallac 1282 Compugamma, Turku, Finland), as previously described.14

51Cr-EDTA clearance was calculated using the following formula: [(Mean Urinary Counts × Urinary Volume) × (Standard counts × 50)]−1. Results were expressed as percentages of the ingested dose.

Glucose Breath Testing

One week before the endoscopic examination, the glucose breath test (GBT) was administered to all participants to detect the presence of SIBO.15 Subjects ate a carbohydrate-restricted meal the evening before the test and fasted for at least 12 hours to minimize basal hydrogen excretion. Smoking and physical exercise were not allowed for 30 minutes before or during the test. Patients rinsed their mouths with 20 mL chlorhexidine 0.05%, and basal end-alveolar breath samples were collected. Immediately thereafter, the patient drank 250 mL of an iso-osmotic solution containing 50 g glucose, and end-alveolar breath samples were collected every 10 minutes for the next 2 hours with a two-bag system composed of a mouthpiece, a T valve, and two collapsible bags for collection of dead-space and alveolar air. Samples were aspirated from the alveolar-air bag into a 20-mL plastic syringe and analyzed immediately for hydrogen (H2) using gas chromatography (Quintron Gas Chromatographer, model DP; Quintron Instrument Co., Milwaukee, WI). Results were expressed as parts per million. An increase in H2 excretion of 12 parts per million or more over basal levels within the 2-hour sampling period was considered indicative of SIBO.

Immunohistochemical Studies of Duodenal Zonula Occludens-1 Expression

Duodenal tissue specimens were fixed in 4% buffered formalin, processed in the usual manner, and embedded in paraffin. Four sections of each block were stained with hematoxylin-eosin for histological evaluation. For immunohistochemical studies of zonula occludens-1 (ZO-1) expression, two sections were incubated with pre-diluted rabbit polyclonal ZO-1 antibody (Zymed Laboratories, San Francisco, CA) or rabbit pre-immune serum (negative controls) and then with avidin-biotin-peroxidase (Vector, Burlingame, CA). Reactions were detected with 3.3′-diaminobenzidine (Vector, Burlingame, CA).

The results were evaluated independently by two expert pathologists (R.R. and F.M.V.), who were blinded to the group origin of the specimen. Results were expressed as previously reported16; briefly, mucosal labeling intensity was scored as follows: 0, no labeling; +, weak labeling; + +, strong labeling. Nuclear labeling at the villous and crypt levels was expressed as the percentage of analyzed cells displaying stained nuclei. At least 3000 cells were examined for each count, and nuclear labeling in villous cells was based on the assessment of six or more villi.

Statistical Analysis

Sample size calculations indicated that, at a confidence level of 95% (alpha = 0.05), at least 24 individuals per group had to be tested to detect a differences in gut permeability or SIBO with a power of 80% (beta = 0.20).

The data were analyzed using a descriptive statistical method (Statistical Package for Social Sciences 13.0) and the results expressed as median and quartile for continuous variables unless otherwise specified. Kruskal-Wallis and Mann-Whitney tests were used to determine statistically significant differences (P < 0.05) between three and two groups, respectively. The chi-squared test was used to test for differences in the prevalence of categorical variables. Spearman's correlation coefficient was used to evaluate the relation between 51Cr-EDTA excretion values and the other continuous variables.


Clinical Features

Table 1 summarizes the clinical features of the three age-matched and sex-matched study groups. Significant intergroup differences emerged for three of the five variables included in the National Cholesterol Education Program's Adult Treatment Panel III guidelines for diagnosis of metabolic syndrome (waist circumference, systolic and diastolic blood pressure, and serum triglycerides). As expected, the prevalence of this syndrome was markedly higher patients with NAFLD (34.3% versus 14.8% in those with CD and 0% in healthy volunteers; P < 0.001). Compared with the CD patients and healthy controls, patients with NAFLD also had significantly higher levels of insulin, total cholesterol, and liver enzymes. NASH was diagnosed in 17 of 35 (48.5%) of NAFLD subjects.

Table 1. Characteristics of Patients with NAFLD, CD, and Healthy Controls
VariableHealthy Volunteers (HV)NAFLDP* (NAFLD Versus HV)CDP (CD Versus NAFLD or HV)
  • Results for continuous variables are expressed as median and interquartile range; for categorical variables as number and percentage.

  • Histological score applied: Steatosis score: 0 = (<5% of hepatocytes involved); 1 = mild (5%-33%); 2 = moderate (33%-66%); 3 = severe (>66%). Lobular inflammation scores at 200×: 0 = no foci, 1 = 2 foci, 2 = 2-4 foci, 3 = >4 foci. Ballooning: 0 = none; 1 = few; 2 = many. Fibrosis scores: 0 = no fibrosis; 1 = perisinusoidal or periportal (including the subgroups A, B, and C); 2 = perisinusoidal and portal/periportal; 3 = bridging; 4 =cirrhosis.

  • *

    NAFLD versus healthy volunteers.

  • CD versus

  • NAFLD or

  • §

    healthy volunteers.

  • SBP, systolic blood pressure; DBP, diastolic blood pressure; EMA, endomysial antibody; IgA, immunoglobulin A anti-endomysium; MS, metabolic syndrome.

N2435 27 
Age33 (28-45)42 (32-54)NS47 (38-55) 
BMI (kg/m2)24.69 (23.85-25.22)26.19 (24.39-27.97)NS25.18 (23.96-26.58)0.032
Waist (cm)87 (80-90)96 (91-100)<0.0593 (87-95)<0.05
Creatinine (mg/dL)0.8 (0.6-0.9)0.8 (0.6-0.9)NS0.8 (0.7-1.0)NS
SBP (mmHg)135 (130-140)140 (130-145)<0.05140 (130-140)NS
DBP (mmHg)80 (75-80)90 (80-90)<0.0585 (80-90)<0.05§
Glucose (mg/dL)86 (79-91)93 (85-100)<0.0590 (84-97)NS
Insulin (μIU/mL)6.75 (5.25-8.00)14.00 (8.80-18.00)<0.0016.06 (4.68-8.15)< 0.001
HOMA-R1.43 (1.03-1.74)3.12 (1.94-4.00)<0.0011.31 (1.01-1.82)< 0.001
Tot. Chol. (mg/dL)182 (167-199)201 (175-220)<0,05179 (158-198)<0.05
HDL-Cholesterol (mg/dL)44 (42-48)45 (42-53)NS50 (43-71)NS
Triglycerides (mg/dL)101 (93-118)139 (103-187)<0.05104 (92-139)<0.05
ALT (IU/L)21 (18-27)58 (42-84)<0.0543 (24-61)<0.05§
AST (IU/L)23 (16-25)34 (27-45)<0.0534 (23-45)NS
gGT (IU/L)27 (20-34)45 (27-87)<0.0534 (21-58)<0.05
EMA IgA positive, n (%)00NS27/27 (100)<0.001
MS (Yes), n (%)012/35 (34,3)<0.054/27 (14,8)0.014
Liver Histology     
111 (31.4) 
214 (40) 
310 (28.6) 
Lobular Inflammation     
120 (57.1) 
215 (42.9) 
02 (5.7)   
133 (94,3)   
NASH (NAS score >5)17 (48.5)   
02 (5.7) 
110 (28.6) 
218 (51.4) 
35 (14.3) 

Intestinal Permeability as Measured by Urinary Excretion of 51Cr-EDTA

The 51Cr-EDTA excretion values for the three groups are shown in Table 2 and Fig. 1A. Compared with the values observed in the healthy volunteers, 51Cr-EDTA excretion was significantly increased in both the NAFLD and CD groups. Within the NAFLD group, intestinal permeability was significantly increased in patients moderate/severe steatosis (>33% of fatty infiltration) versus mild steatosis (P < 0.05) (Fig. 1B); however, it was unrelated to the severity of lobular inflammation (P = 0.74), ballooning (P = 0.15), fibrosis (P = 0.99), or the presence of NASH (P = 0.47).

Table 2. Small Intestinal Bacterial Overgrowth (SIBO), Intestinal Permeability (%51Cr-EDTA), and Immunohistochemical Staining for ZO-1 in the Three Groups Studied
  Healthy versusNAFLDP*CDP
  • Results of ZO-1 immunohistochemistry are reported quantitatively for the nuclear staining and semiquantitatively for the cytoplasmic staining according the following score: 0, absence of labeling; +, mild labeling; + +, robust labeling; –, undetectable as absence of villi.

  • *

    NAFLD versus healthy volunteers.

  • CD versus

  • NAFLD or

  • §

    healthy volunteers.

  • Results are expressed as mean ± SD for numerical continuous variables and as number and percentage of total for categorical variables.

  • e

    P value assessed by chi-squared test for categorical variables and by Kruskall-Wallis for continuous variables.

n 2435 27 
%51Cr-EDTA 2.04 (2.00-2.37)4.88 (3.14-6.65)<0.0019.18 (7.91-11.56)<0.001§
SIBO positive, n (%) 5/24 (20.8)21/35 (60)<0.00115/27 (55.5)<0.001§
CryptsZO-1 nuclei (%)58 (42-69)10 (6-19)<0.0012 (1-2)<0.001§
 ZO-1 cytoplasm n. (%)  0.056 <0.05§
 000 13/27 (48,1) 
 0/+00 9/27 (33,3) 
 +12/24 (50)26/35 (77.1) 5/27 (18,5) 
 ++12/24 (50)9/35 (22.9) 0 
VilliZO-1 nuclei (%)7 (4-12)1 (0-3)<0.001-<0.001
 ZO-1 cytoplasm n (%)  <0.001 <0.001
 000 - 
 +3/24 (12.5)19/35 (54.3) - 
 ++21/24 (87.5)16/35 (45.7) - 
Figure 1.

Intestinal permeability. (A) Intestinal permeability in the three groups studied. Bars represents the mean of 51Cr-EDTA recovery dose as expression of intestinal permeability. (Bars show mean ± 2 standard error of the mean). *P < 0.001 NAFLD versus HV, CD versus HV, CD versus NAFLD. (B) Intestinal permeability increases in subjects with histological evidence of moderate or severe steatosis (>33%). (Bars show mean ± 2 standard error of the mean). *P < 0.001.

To determine the features of subjects with increased intestinal permeability, we used the median value of 51Cr-EDTA excretion in the NAFLD cohort to identify subjects with increased permeability (51Cr-EDTA > 4.88%). Comparing the two groups of NAFLD subjects with respect to permeability (Table 3), we found that subjects with increased permeability had a significant (P < 0.05) increased prevalence of metabolic syndrome, whereas no significant differences were found with regard to other metabolic parameters (Table 3). With respect to liver histology, subjects with increased permeability had an increase prevalence of moderate or severe steatosis according to Clinical Reseach Network–NAFL Activity Score score (P = 0.025).

Table 3. Intestinal Permeability (%51Cr-EDTA) and Small Intestinal Bacterial Overgrowth (SIBO) in NAFLD Patients (n = 35)
Normal permeabilityIncreased permeability*PNegativePositiveP
  • *

    Increased intestinal permeability was defined with a 51Cr-EDTA value above 4.88%.

n1718 1421 
Sex (M)3/142/16NS3/112/19NS
Age44 (34-53)41 (32-54)NS46(32-54)40 (32-53NS
BMI24.97 (24.39-28.01)26.19 (25.18-27.45)NS25.74 (23.88-28.0126.15 (24.58-27.46)NS
Waist (cm)95 (92-98)96 (91-100)NS97 (92-100)96(91-98)NS
Creatinine (mg/dL)0.80 (0.60-0.90)0.80 (0.70-0.90)NS0.78 (0.60-1.00)0.80 (0.70-0.80)NS
Glucose (mg/dL)96 (87-100)93 (85-100)NS92 (87-96)95(85-101)NS
Insulin (μUI/mL)14.00 (10.70-18.00)13.40 (7.25-18)NS13.50 (8.80-15.50)14.80(10.10-19.00)NS
HOMA3.15 (2.67-4.27)2.69 (1.91-3.78)NS2.99(1.91-3.24)3.12 (2.37-4.00)NS
MS positive1/17 (0.5%)11/18 (61.1)<0.0011 (7.2)11 (52.3)<0.001
Triglycerides127 (110-148)153 (100-206)NS133 (110-148)146 (85-206) 
ALT (<45 UI/L)58 (40-77)59 (47-87)NS55(40-70)66(47-93) 
AST (<45 UI/L)35 (27-46)34 (29-44)NS35(27-46)34(25-44) 
gGT (<45 UI/L)48 (23-114)34 (27-56)NS47(27-114)33(27-86) 
Steatosis  0.025  0.003
000 00 
19 (53.0)2 (11.1) 9 (64.3)2 (9.6) 
24 (23.5)10 (55.5) 3 (21.4)11 (52.3) 
34 (23.5)6 (33.4) 2 (14.3)8 (38.1) 
Lobular Inflammation  NS  NS
000 00 
110 (58.8)10 (55.5) 7(50.0)13 (61.9) 
27 (41.2)8 (44.5) 7(50.0)8 (38.1) 
300 00 
Ballooning  NS  NS
02 (11.7)0 2 (14.3)0 
115 (88.2)18 (100) 12 (85.7)21 (100) 
200 00 
NASH (NAS5)8/17 (47%)9/18 (50%)NS7(50%)10 (47,6)NS
Fibrosis  NS  NS
01 (5.8)1 (5.5) 1(7.2)1 (4.8) 
16 (35.3)4 (22.2) 6 (42.8)4 (19.1) 
27 (41.1)11 (61.2) 5 (35.7)13 (61.9) 
33 (17.6)2 (11.1) 2 (14.3)3 (14.2) 
400 00 
SIBO positive5 (29.4)16 (88.8)<0.001021 (100)<0.001
%51Cr-EDTA3.14 (2.81-4.22)6.60 (5.46-8.00)<0.053.05(2.08-4.48)6.12 (4.88-7.32)<0.001
Crypts ZO-1 nuclear stain (%)9 (1-11)15 (10-27)<0.058 (1-10)15 (10-28)<0.05
Villi ZO-1 nuclear stain (%)1 (0-2)2 (0-3)NS1 (0-2)1 (0-3)NS

SIBO Detected with GBT

The GBT results were indicative of SIBO in 21 of the 35 (60%) NAFLD patients, 15 of the 27 (55.5%) with CD, and five of the 24 (20.8%) healthy controls (P < 0.001) (Table 2).

The prevalence of SIBO in the NAFLD group was more than twice that observed among controls (P < 0.001). In patients with NAFLD, the prevalence of SIBO increased significantly with the severity of steatosis (P < 0.001) (Fig. 2), and the Spearman's test revealed a positive correlation between SIBO and steatosis (r = 0.578, P < 0.001). We did not found a significant correlation between SIBO and presence of NASH (P = 0.84), lobular inflammation (P = 0.60), or fibrosis score (P = 0.97). In the NAFLD cohort, compared with subjects without SIBO, those with SIBO had an increased prevalence of moderate and severe steatosis independent of metabolic parameters, such as BMI, waist, homeostatic model assessment, and triglycerides. Interestingly, NAFLD patients with increased permeability had a prevalence of SIBO more than twice that of patients with normal permeability (88.8% versus 29.4%, P < 0.001). Moreover, subjects with SIBO had increased intestinal permeability (P < 0.05) and significant increase of metabolic syndrome (P < 0.001). (Table 3).

Figure 2.

Small intestinal overgrowth and severity of steatosis. The prevalence of SIBO increases in subjects with moderate or severe steatosis. Shadow bars represent the prevalence of GBT positivity as a surrogate marker of SIBO. *P < 0.001.

Integrity of Intestinal Tight Junctions Reflected by ZO-1 Immunohistochemistry

Interobserver agreement in reading the results of ZO-1 immunohistochemistry was good (kappa coefficient, 0.78). The results observed in the three groups are summarized in Table 2, and representative ZO-1 labeling patterns are shown in Fig. 3.

Figure 3.

Immunohistochemical analysis of ZO-1 in duodenal mucosa. Healthy subjects: At crypt level (A), a high percentage of nuclei labeled is present with low cytoplasm staining (original magnification 400×); at villous level (B), strong labeling of cytoplasm of enterocytes covering the tip (arrows) with a progressive decrease toward the basis of villi, whereas nuclear staining is low (original magnification 400×). NAFLD: At crypt level (C), a moderate percentage of nuclei labeled is present with low cytoplasm staining (original magnification 250×); at villous level (D), moderate labeling of cytoplasm of enterocytes covering the tip (arrows) with a progressive decrease toward the basis of villi, whereas nuclear staining is low (original magnification 400×). Untreated celiac disease (E): very low labeling of both nuclei and cytoplasm, at crypt level and in enterocyte covering atrophic mucosa (original magnification 250×).


In healthy subjects, well over half (57.18% ± 16.41%) of the nuclei of intestinal crypt cells were strongly labeled for ZO-1, whereas the percentages of labeled nuclei in crypts from CD patients were very low. In patients with NAFLD, crypt cell nuclear expression of ZO-1 was higher than that of CD patients but still far lower than that of healthy controls (P < 0.001). Cytoplasmic staining intensity of ZO-1 in the epithelium of hyperplastic crypts from CD patients was significantly lower than that observed in the other two groups (P < 0.05), and cytoplasmic labeling in the NAFLD group was reduced compared with control values (P = 0.056) (Table 2).


At the villous level, nuclear staining for ZO-1 was observed in a median value of seven of all cells in the normal mucosa of healthy subjects but only one of those from patients with NAFLD (P < 0.0001). In healthy subjects, immunohistochemical analysis showed a strong labeling at the villous tip that reduced progressively along the villous axis; NAFLD patients had the same progressive reduction along the villous axis (from tip to basis of villous) but with less strong cytoplasmic labeling (P < 0.001). Because of the absence of villi in the mucosa of patients with active CD, villous expression of ZO-1 was not evaluated in this group. In the NAFLD group, no differences in ZO-1 staining of crypt or villous were observed according to histological severity.

Interestingly, subjects with NAFLD and increased intestinal permeability (Table 3) had a significant increase of SIBO, and the increase of intensity of crypt cells nuclear ZO-1 staining was significantly correlated with increasing intestinal permeability (Spearman's rho: 0.425; P = 0.011) (Fig. 4).

Figure 4.

Correlation between intensity of crypt cell nuclear ZO-1 staining and intestinal permeability.


The main findings of this study are that both intestinal permeability and the prevalence of SIBO are increased in patients with NAFLD and correlate with the severity of steatosis and that the intensity of duodenal ZO-1 staining in both crypts and villi is significantly lower in patients with NAFLD, suggesting that disruption of tight junction integrity may explain the increased permeability in these patients.

The idea that increased intestinal permeability and gut flora might contribute to the development of several diseases was first suggested in 1890 (Llewellyn Jones: “Theory of auto-intoxication from gut bacteria”).17 The occurrence of cross-talk between the gut and the liver is an intriguing hypothesis that could explain the hepatobiliary changes associated with several inflammatory and infectious intestinal diseases (that is, inflammatory bowel disease, celiac disease, infections caused by Salmonella, Yersinia, and so forth)18 Evidence supporting a role for the liver–gut axis in the pathogenesis of NAFLD has been slowly accumulating over the past 7 years.5, 19, 20

The primary aim of this study was to determine whether intestinal permeability is indeed increased in patients with NAFLD. We chose the 51Cr-EDTA excretion test over the lactulose/rhamnose intestinal permeability tests used by Wigg et al.5 because the latter test is influenced by the presence of SIBO. In fact, in the article by Wigg et al., although there was a significant difference for the SIBO prevalence, the authors found no difference between NAFLD and controls with respect to intestinal permeability. In normal subjects, only 1%-3% of an orally administered dose of 51Cr-EDTA is absorbed from the gastrointestinal tract. Under conditions of enhanced intestinal permeability, a much larger fraction of the dose enters the bloodstream and is rapidly cleared through glomerular filtration. The 51Cr-EDTA crosses the intestinal epithelium along the paracellular pathway, and its increased presence in the urine is thus a reflection of disruption of the tight junctions (TJs).21 Arslan et al.22 found that 51Cr-EDTA excretion is a good indicator of the severity of gut-mucosal inflammation and loss of TJ integrity. Increased urinary excretion of this marker has been well documented in other small bowel disorders,23, 24 including CD.25 Celiac disease represents a good model of disruption of the gut barrier.21

To investigate the possible presence of SIBO, we use the GBT, because it is a simple, noninvasive, low-cost method, which has exhibited good sensitivity when compared with culture of intestinal aspirates, the current gold standard.26, 27 It can detect bacterial concentrations proximal to the distal ileum of greater than 105 per milliliter.28 Using GBT, we demonstrated that the prevalence of SIBO in patients with NAFLD of our cohort was more than twice as high as that found in healthy subjects (60% versus 20.8%, P < 0.001).

In the current study, recovery of 51Cr-EDTA in the urine of CD patients was markedly increased over that of healthy controls, and that observed in NAFLD patients was also clearly higher than control values. This finding indicates that subjects with NAFLD had a significant increase in gut permeability, and it was positively correlated with the extent of hepatic fat deposits but unrelated to scores of hepatic inflammation or fibrosis. Interestingly, the presence of SIBO in NAFLD patients was also positively associated with the degree of steatosis, and the prevalence of SIBO was twice as high in NAFLD patients with increased gut permeability compared with those with normal permeability.

According to these results, we can hypothesize that SIBO or the associated increase in gut permeability may be causally linked to pathogenesis of steatosis. Data from studies in mice2 support the hypothesis that gut bacteria contribute to the pathogenesis of NAFLD by: (1) increasing gut luminal ethanol production29; (2) metabolizing dietary choline (required for very-low-density lipoprotein synthesis and hepatic lipid export)4; or (3) releasing lipopolysaccharide,30 which is likely to activate steatogenic, proinflammatory cytokines in the luminal epithelial cells, liver macrophages, or both. The recently reported beneficial effect of probiotics on high-fat diet–induced steatosis provides further experimental evidence supporting a role of gut-derived endotoxemia in the onset of the steatosis.31 Our results are consistent with experimental data suggesting an effect of gut flora on the development of steatosis in humans. Recent studies in humans have revealed that obesity is associated with a decrease in the ratio of Bacteroidetes to Firmicutes, the dominant groups of beneficial microbes in the human gut, and this obese microbiome has an increased capacity for extracting energy from the diet, which results in increased caloric intake and ultimately fat deposition.32, 33 The characteristics of the gut flora were not assessed in our patients, but our data support the hypothesis that changes in the gut microbiota can contribute to hepatic steatosis in patients with NAFLD. Importantly, the lack of any association between either SIBO or intestinal permeability and steatohepatitis or fibrosis argues against a primary role for gut bacteria in the development of progressive NAFLD.

Evidence of TJ disruption has recently been reported in an animal model of obesity and metabolic syndrome.34 To test the hypothesis that subjects with NAFLD had increased permeability as result of damage to the intestinal TJ, we evaluated the expression of ZO-1, one of several specific proteins involved in the TJ structure, in the duodenal specimens of our cohort. ZO-1 was the first TJ-associated protein to be identified,35 and it is widely considered an excellent marker for detecting intact cell-to-cell contacts and assessing TJ integrity.36 Accumulation of ZO-1 in the cell nucleus appears to correlate with biologically significant phases of cellular maturation, suggesting that TJ components contribute to the regulation of nuclear processes and of cellular growth and differentiation, and the nuclear expression may be used as marker of TJ integrity.37 On the whole, ZO-1 expression (nuclear and cytoplasmic) in the duodenal mucosa of patients with NAFLD was lower than that observed in healthy subjects but higher than that associated with CD. In our NAFLD patients, the increased nuclear staining was accompanied by increased staining of the cytoplasm as well, in both the villi and the crypts. This pattern seems likely to suggest an attempt by the cells to upregulate ZO-1 expression in response to incompetent TJs. The fact that increased excretion of 51Cr-EDTA was associated with stronger (++) labeling of the cytoplasm in crypt cells is consistent with the putative role of TJ disruption in the increased gut permeability documented in NAFLD.

The design of our study did not allow us to answer the question of which come first, SIBO or increased intestinal permeability, although NAFLD is clearly associated with both. The association between intestinal hyperpermeability and SIBO may reflect qualitative or quantitative changes in the bacterial flora of the gut (or in toxic products produced by these organisms), which lead to disruption of the intestinal barrier documented in our NAFLD patients and subsequently bacterial translocation across the gut mucosa and the development of portal endotoxemia. With respect to the factors responsible for the increase of SIBO in NAFLD, we can only speculate because this study was not designed to address this issue. Considering data from literature, diabetes could enhance SIBO by provoking intestinal dysmotility, which would favor bowel stasis,38 and a delay in oro-cecal transit has also been observed in NAFLD,39 suggesting a common explanation for SIBO. However, in our patients with NAFLD, the presence of SIBO showed no correlation with either the fasting glucose level or the presence of diabetes. Most intriguingly, recent studies in both rodent models of NAFLD and patients with NAFLD have shown that dietary habits including fat and fructose content can influence gut flora and endotoxin levels.40–42 Given the association between NAFLD and both a high-fat43 and high fructose-containing diet,44 this observation offers a potential explanation for the increased prevalence of SIBO in patients with NAFLD. Our study collected data on only total consumption of protein, carbohydrate, and fat and the number of portions of fruit per day and found no significant difference between the three groups (data not shown). The association between SIBO, intestinal permeability, and accurately determined dietary intake in patients with NAFLD is clearly an area for further study.

It has recently been demonstrated that lipopolysaccharide contributes to the development of the subclinical inflammatory state and insulin resistance associated with type 2 diabetes and obesity by stimulating the innate immune system and triggering the release of proinflammatory cytokines from adipose tissue.45 This insulin resistance is associated with steatosis.40 A further potential indirect link between SIBO/increased permeability and steatosis severity is therefore that the resulting endotoxemia leads to inflammation-mediated extrahepatic insulin resistance with a subsequent increased supply of fatty acids to the liver and steatosis.31, 32, 33, 34, 40, 46 In our cohort, the correlation we observed between intestinal permeability, the presence of the metabolic syndrome, triglyceride levels, and the marginally higher insulin levels in NAFLD patients with increased intestinal permeability offers some support for this hypothesis.

In conclusion, we have demonstrated that NAFLD is associated with increased intestinal permeability and SIBO and that these factors are associated with the severity of hepatic steatosis. We report evidences of “leaky” intestine with TJ disruption in human NAFLD with evidence of increased SIBO. These data may suggest that the increased intestinal permeability may be the condition sine qua non for the hypothesis of the contribution of gut–liver axis to the development of NAFLD. The evidence of correlation between SIBO and fatty infiltration requires further studies to understand the causative role of intestinal bacteria in hepatic fat metabolism. These studies may lead to investigation of the microbiome as new target therapy of NAFLD patients.