Statistics from Altmetric.com
What is the key question?
Are alterations in gut–liver axis involved in the pathogenesis of obstructive sleep apnoea syndrome (OSAS)-associated liver injury in paediatric non-alcoholic fatty liver disease (NAFLD)?
What is the bottom line?
Recent data linked the presence and severity of OSAS to the presence and severity of NAFLD, but mechanism(s) connecting OSAS to liver injury are unclear.
Why read on?
OSAS may promote liver injury on one side by impairing intestinal barrier function and promoting endotoxemia and on the other side by sensitising the liver to endotoxin and proinflammatory stimuli: these pathways may mediate OSAS-associated liver injury in NAFLD.
Non-alcoholic fatty liver disease (NAFLD) affects 10% of the general paediatric population and 50–70% of obese children, and its prevalence is rising along with the obesity epidemic.1 NAFLD encompasses a histological spectrum ranging from simple steatosis to non-alcoholic steatohepatitis (NASH): while the former has a benign hepatological course, NASH can progress to cirrhosis and hepatocellular carcinoma (HCC).2 Although NAFLD is generally a slowly progressive disease, approximately 8% of children undergoing liver biopsy for suspected NAFLD have cirrhosis,3and onset of HCC on background NAFLD has been reported as early as age 7.4 These data indicate the need for early recognition and treatment of NASH to prevent liver-related complications in adulthood, and recent American Gastroenterological Association/American College of Gastroenterology/American Association for the Study of Liver Diseases guidelines state that onset of NAFLD in childhood may be at greater risk for severe liver-related complications later in life.5
Besides being an emerging cardiometabolic risk factor, OSAS has been recently connected to the presence and severity of liver disease in paediatric NAFLD, independently of whole-body/abdominal obesity, insulin resistance and metabolic syndrome (MS),8 ,9 but the mechanism(s) connecting OSAS to liver injury are unknown.
Current therapeutic approaches for paediatric OSAS have disappointing success rates, and elucidating mechanisms mediating hypoxia-associated liver injury would offer novel therapeutic opportunities for OSAS-associated NASH and fibrosis.10 Two pathways for liver disease progression have been recently identified in NAFLD.
The first is an altered gut–liver interaction, including an impaired intestinal barrier integrity with enhanced lipopolysaccharide (LPS) translocation and increased plasma endotoxemia, leading to chronic low-grade inflammation and liver injury through activation of the LPS–toll-like-receptor (TLR)-4 signalling pathway.11 ,12
The second pathway for liver disease progression is an expansion of hepatic progenitor stem cell (HPC) pool and an altered pattern of adipokine expression by HPCs and hepatocytes, which appear more tightly related to liver injury than circulating adipokines.13 The mechanisms regulating gut–liver interaction and HPCs pool and phenotype in NAFLD are unknown.
Chronic intermittent hypoxia (CIH) profoundly affects intestinal epithelial and entero-endocrine cell integrity and function experimentally, and hypoxia is a major regulator of progenitor cell proliferation and phenotype in diverse disease models.14–18
Currently, there are no in vitro or in vivo data on the role of gut–liver axis and of HPC compartment as a possible mediator of OSAS-associated liver injury in NAFLD.
We hypothesised that intermittent hypoxia affects liver injury by modulating gut–liver axis and/or HPC proliferation and adipokine phenotype in paediatric NAFLD.
The study was performed at the Bambino Gesù Children's Hospital during the period January 2012 to January 2014.
Eighty-one consecutive children with an ultrasonographic diagnosis of NAFLD, persistently (≥6 months) elevated serum aminotransferases and symptoms/signs suggestive of sleep apnoea,8 were prospectively seen and offered liver biopsy and polysomnography (PSG) as part of the experimental protocol, designed to evaluate the relationship of sleep disturbance with liver disease.1 ,8
Among these 81 patients, 80 children were assented and parents consented to participate. Hepatitis virus infection, other competing causes of hepatic steatosis and coeliac disease were excluded according to standard guidelines (see online supplementary file).
Body mass index (BMI), waist circumference and their SD score (z-score) were calculated.19 ,20 MS was defined as the presence of ≥3 of the following five criteria:21 abdominal obesity as defined by a waist circumference ≥90th percentile for age and sex,20 hypertriglyceridemia as defined by triglycerides (TG) >95th percentile for age and sex;22 low high-density lipoprotein cholesterol (HDL-C) as defined by <5th percentile for age and sex;23 elevated blood pressure (BP) as defined by systolic or diastolic BP >95th percentile for age and sex,24 and impaired fasting glucose, impaired glucose tolerance or type 2 diabetes mellitus.25
Patients and responsible guardians underwent a 1 h nutritional interview with experienced dieticians, and their dietary intake was recorded as previously described.11 ,13 The patients received no treatment for NAFLD or OSAS before liver biopsy and polysomnography were performed.
Aspartate aminotransferase, alanine transaminase, gamma-glutamyl transpeptidase, total TG and total/HDL-C were assessed using standard laboratory methods. All participants underwent a standard oral glucose tolerance test, performed with 1.75 g of glucose/kg of body weight (up to 75 g). Two indices of insulin sensitivity were calculated as previously described:8 the homeostasis model assessment of insulin resistance and the insulin sensitivity index.
Inflammatory markers and cytokines
Serum C-reactive protein was determined via a high-sensitivity latex agglutination method on HITACHI 911 Analyser (Sentinel Ch., Milan).
Serum adiponectin, tumour necrosis factor-α, interleukin-6, leptin, resistin (RayBiotech, Norcross, Georgia, USA) and retinol-binding protein-4 (Dade Behring, Newark, Delaware, USA) were measured by sandwich ELISA. Sensitivity and intra-assay and inter-assay CVs of each kit are detailed in online supplementary file).
Markers of hepatocyte apoptosis and of extracellular matrix deposition
Circulating cytokeratin (CK)18 fragments and hyaluronic acid, two validated markers of hepatocyte apoptosis and extracellular matrix deposition, respectively, in paediatric NAFLD6 were measured (see online supplementary file).
Intestinal permeability test and plasma LPS
Children underwent the previously described and validated intestinal permeability test with lactulose and mannitol,11 and the ratio of the fractional urine excretion of lactulose to the fractional urine excretion of mannitol (L/M ratio) was calculated. L/M ratio, as an index of intestinal permeability, was entered as both a continuous and a dichotomous variable: an L/M ratio of ≥0.03 was considered abnormal based upon previously established references.9
Plasma LPS concentration was measured by a commercially available kit (Amebocyte Lysate (LAL) LAL Chromogenic Endpoint Assay. Cambrex Limulus kit; Hycult Biotech, Uden, the Netherlands).
Liver histology: Liver biopsy was performed after an overnight fast, and specimens processed as detailed in online supplementary file. Biopsies were evaluated by a single blinded pathologist, with a long-time experience in the field. Steatosis, inflammation, hepatocyte ballooning and fibrosis were scored using the NAFLD Clinical Research Network criteria, as recently recommended (detailed in online supplementary file).5 Additionally, the presence of portal fibrosis, a distinctive feature of early fibrosis in paediatric NAFLD,1 was recorded.
Features of steatosis, lobular inflammation and hepatocyte ballooning were combined to obtain the NAFLD activity score (NAS). As recommended by current guidelines, biopsies were subdivided into not NASH and definite NASH subcategories, on the basis of overall injury pattern (steatosis, hepatocyte ballooning, inflammation), as well as the presence of additional lesions (eg, zonality of lesions, portal inflammation and fibrosis).8 ,13
Liver immunofluorescence and immunohistochemistry Liver immunofluorescence (IF) and immunohistochemistry (IHC) were performed on 2-µm-thick sections obtained from formalin-fixed tissue embedded in paraffin. Antigen retrieval was performed with EDTA (pH 8) (Dako, Glostrup, Denmark).
Each case was analysed by IF for α-smooth muscle actin (α-SMA), CK8/18, CD68 and TLR-4. For the identification of activated hepatic stellate cells (HSCs) and portal/septal myofibroblasts, the primary antibody used was anti-α-SMA (dilution 1:200 overnight; mouse monoclonal, clone 1A4, Novus Biological, Littleton, Colorado, USA); for the detection of Kupffer cells and macrophages, the primary antibody used was anti-CD68 (dilution 1:200 overnight; mouse monoclonal, clone KP1, Abcam, Burlingame, California, USA); and for the detection of hepatocytes, the primary antibody used was CK8/18 (dilution 1:100 incubated for 1 h, mouse monoclonal, Vector Laboratories, Burlingame, California, USA) (details provided in online supplementary file).
Double CD68/TLR-4 and α-SMA/TLR-4-positive cells were manually counted in at least 10 fields at ×20 magnification for each sample. Data were expressed both as a percentage of positive cells. The intensity average of TLR-4 fluorescence in CK8/18-positive hepatocytes was assessed in at least 200 regions of interest for sample using MetaMorph software.
Observations were processed and images were analysed by two independent pathologists blinded to the patient's data.
Adipokine expression by HPCs
The primary antibody against CK7 (dilution 1/200; mouse monoclonal, clone OV-TL 12/30, Dako) was used for the identification of HPCs as previously described by Roskams et al.26
For IHC, sections were incubated overnight at 4°C with primary antibodies against CK7, adiponectin, resistin, p21Waf1 and cleaved caspase-3 (primary antibodies are listed in online supplementary table S1).
The number of HPCs within the ductular reaction was counted within the entire section and expressed as the number of CK7-positive cells per high-power field (HPF; at ×20).
Intermediate hepatocytes (IHs) were defined as cells with sizes between those of hepatocytes and HPCs (<40 but >6 µm in diameter), with faint CK7 immunoreactivity in the cytoplasm and reinforcement at the plasma membrane. Based on our recent data, IHs were scored as present or absent.12
Adiponectin and resistin expression by CK7-positive HPCs was evaluated in serial sections and confirmed with IF. Data were expressed both as number of positive cells per HPF and as a percentage of positive cells. Adipokine expression by hepatocytes was semiquantitatively evaluated and expressed as a percentage of positive cells.
Hepatocyte apoptosis and cell-cycle arrest
Apoptosis and cell-cycle arrest were assessed by counting the number of hepatocytes that stained strongly positive for cleaved caspase-3 (at the cytoplasm level) and p21waf1 (at the nuclear level). The apoptotic and p21 indices were calculated by dividing the average number of positive cells by the average number of hepatocytes and expressing the quotient as a percentage for each section. At least 30 lobular fields at ×40 magnification were analysed (≈1000 hepatocytes) for each section.
All biopsy-proven patients with NAFLD underwent an overnight PSG using standard techniques.
The patients received no dietary or lifestyle counselling or any other treatment between liver biopsy and PSG, which was performed within 3 months of the histological diagnosis of NAFLD.
PSG was carried out in a quiet room in the sleep laboratory of our hospital; all recordings started at the patients’ usual bedtime and continued until spontaneous awakening.
No hypnotic drugs were allowed for at least two weeks before sleep recording. All patients were accompanied by one of their parents throughout the night. No oxygen was supplemented or respiratory stimulants were used.
The PSG montage included four EEG channels C3-A2, C4-A1, O1-A2 and O2-A1, left and right electrooculogram, chin electromyogram, ECG, nasal cannula, thoracic and abdominal respiratory effort, oxygen saturation (Siesta, Compumedics, Abbottsford, Australia), and end-tidal pCO2 (ETpCO2) was monitored simultaneously with other parameters (Capnostream, Oridion).
All recordings were manually and visually scored and interpreted according to current guidelines (2007 AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications V.2.0).27
Central, obstructive and mixed apnoea events were counted according to the criteria established by the American Thoracic Society (2007): an obstructive apnoea was defined as the absence of airflow, with continued chest wall and abdominal movement, for a duration of at least two breaths; a central apnoea was defined as the absence of airflow with the cessation of respiratory effort, lasting >20 s and associated with bradycardia and desaturation; central apnoea occurring after gross body movements or after sighs was not considered as a pathological finding; a mixed apnoea was defined as an apnoea that usually begins as central and ends in obstruction according to changes in the chest, abdominal and flow traces; hypopnoea was defined as a decrease in nasal flow of at least 50% with a corresponding decrease in SaO2 of at least 4% and/or an arousal; the apnoea/hypopnoea index (AHI) was defined as the number of apnoeas and hypopnoeas per hour of total sleep time (TST). Patients with an AHI ≥1 event/h were considered to have OSAS, while subjects with an AHI ≥5 events/h were considered to have severe OSAS, according to current guidelines.10 ,26 The following parameters were also recorded: oxygen desaturation index, that is, the number of haemoglobin desaturations (drop in SaO2 ≥4% of baseline value) per hour of sleep, mean SaO2, nadir SaO2, total duration of haemoglobin desaturation (SaO2 <90%), expressed as %TST and mean ETpCO2.
The presence of adenotonsillar hypertrophy was assessed and graded according to a standard four-stage scale (from 0= absent to 4=severe), and other OSAS-related symptoms/signs were recorded at the time of PSG, as indicated by current guidelines and previously described.6 ,10
Sample size calculation: Based on limited available data on OSAS,8 ,9 parameters related to gut–liver axis11 and HPC13 in paediatric NAFLD, at least 74 subjects were needed to detect a significant (p<0.05) difference in LPS, intestinal permeability, HPC number and hepatic adipokine expression between patients with and without OSAS with a power of 80%.
Data were expressed as mean±SD. Differences were considered statistically significant at p<0.05.
Differences across groups were analysed by analysis of variance and Bonferroni correction, when variables were normally distributed; otherwise, the Kruskal–Wallis test, followed by the post hoc Dunn test, was used to compare non-parametric variables. Normality was evaluated by Shapiro–Wilk test. χ2 test or Fisher's exact test was used to compare categorical variables, as appropriate. Spearman's rank correlation coefficient was used to estimate the relationship between different variables.
Multivariable standard logistic regression analysis was used to identify independent predictors of NASH, portal fibrosis and significant (stage ≥2) fibrosis, included as a dichotomous variable. For this analysis, quartiles of continuous variables were included. The variables significantly associated with selected outcomes on univariable analysis were entered in each model. Multiple linear regression analysis was applied to identify predictors of NAS and fibrosis score and of different continuous variables. Continuous variables with skewed distribution were log-transformed. STATISTICA 5.1 (StatSoft Italia, Padua) was used for all analyses.
Anthropometric, dietary and laboratory characteristics
In total, 80 out of 81 children assented to participate to the study. Clinical, laboratory and histological data of the included children are described in tables 1 and 2. In the whole NAFLD cohort, 73% subjects were obese and 16% had MS, 61% had NASH and 44% had significant (F ≥2) fibrosis. No patient had cirrhosis. Compared with patients without NASH, children with NASH had a higher BMI, higher intestinal permeability and LPS levels and lower plasma adiponectin levels and a higher prevalence of OSAS (table 1). Polysomnographic parameters showed a skewed distribution and were all log-transformed for correlative analyses.
Dietary intake did not differ between NASH and non-NASH subjects and between OSAS and non-OSAS subjects and included a 40–45 cal/kg/day, derived from carbohydrate (50–60%), fat (23–30%, constituted of two-thirds unsaturated fatty acids and one-third saturated fatty acids) and protein (15–20%).
NASH is associated with increased hepatic TLR-4 expression and expansion of adiponectin-defective HPC compartment
On liver IHC/IF, NASH was associated with an increased expression of TLR-4 by hepatocytes, Kupffer cells and HSCs, with increased hepatocyte apoptosis and cell-cycle arrest, and with an expansion of HPC compartment, coupled with a defective adiponectin expression by hepatocytes and HPCs (table 2 and online supplementary figure S1).
OSAS is associated with increased intestinal permeability and endotoxemia
Compared with non-OSAS patients, OSAS was associated with lower plasma adiponectin levels and with increased intestinal permeability and plasma LPS (table 3). Except that for plasma adiponectin statistically significant differences between the above-mentioned parameters were also observed between mild and severe OSAS (table 3).
OSAS and histological, IF and IHC parameters
Compared with non-OSAS patients, OSAS was associated with a greater prevalence of NASH, and of portal and significant fibrosis, an increased expression of TLR-4 by hepatocytes, Kupffer cells and HSCs, with increased hepatocyte apoptosis and cell-cycle arrest indices (table 4). OSAS was also associated with an expansion of HPC compartment and with a reduced adiponectin expression by hepatocytes and HPCs (table 4).
Statistically significant differences between the above-mentioned parameters were also observed between mild and severe OSAS (table 4).
Impact of OSAS on liver disease and IHC/IF parameters in non-obese children
OSAS and obesity often coexist, and obesity has been linked to impaired intestinal barrier function and increased endotoxemia.7 We therefore separately assessed the impact of OSAS on gut–liver axis and liver IHC/IF parameters in obese and non-obese children with NAFLD (see online supplementary tables S2 and S3): even when restricting the analysis to non-obese participants (n=20), OSAS was associated with increased intestinal permeability and endotoxemia, increased TLR-4 expression by mature liver cells and expansion of HPCs with reduced adiponectin expression.
To further evaluate the additive effect of BMI and OSAS on endotoxemia, we plotted quartiles of BMI z-score and presence/absence of OSAS: at each BMI quartile, the presence of OSAS was associated with significantly higher plasma LPS levels (figure 1).
Correlative analysis: predictors of liver histology
On multivariable logistic regression analysis, the presence of NASH was predicted by plasma LPS, by SaO2 <90% (%TST), by % hepatocytes expressing TLR-4 and by the number of Kupffer cells expressing TLR-4 (table 5). The presence of portal fibrosis was predicted by plasma LPS, by the number of HSCs expressing TLR-4 (TLR-4-positive HSCs) and (inversely) by the number of HPCs expressing adiponectin (adiponectin-positive HPCs). The presence of significant (stage F≥2) fibrosis was predicted by plasma LPS, by SaO2 <90% (%TST) and by the number of HSCs expressing TLR-4 (table 5).
On multivariable linear regression analysis, SaO2<90% (%TST), plasma LPS and TLR-4 expression by hepatocytes, Kupffer cells and HSCs variably predicted single histological features of NASH, NAS score and fibrosis stage (table 6).
Correlative analysis: predictors of parameters related to gut–liver axis and of hepatic IHC/IF parameters
On multiple linear regression analysis, SaO2 <90% (%TST) independently predicted intestinal permeability (expressed as L/M ratio), plasma LPS, hepatocyte cell-cycle arrest and apoptotic indices, TLR-4 expression by hepatocytes, Kupffer cells and HSCs, the number of HPCs and (inversely) the number of HPCs expressing adiponectin (tables 7 and 8 and see online supplementary figure S2A–E).
Apoptosis and cell-cycle arrest were assessed by counting the number of hepatocytes that stained strongly positive for cleaved caspase-3 (at the cytoplasm level) and p21waf1 (at the nuclear level). The apoptotic and p21 indices were calculated by dividing the average number of positive cells by the average number of hepatocytes and expressing the quotient as a percentage for each section.
Novel findings of this study are the following:
In paediatric NAFLD, OSAS is characterised by an altered gut–liver axis, represented by an increased intestinal permeability and endotoxemia and by upregulated endotoxin receptor TLR-4 expression by hepatocytes, Kupffer cells and HSCs.
OSAS is also characterised by an expanded adiponectin-defective HPC pool in the liver.
The duration of nocturnal hypoxaemia independently predicts the above-mentioned alterations in the gut–liver axis and in HPC pool.
OSAS has been recently connected to the presence and severity of paediatric NAFLD, independently of whole-body/abdominal obesity, MS and insulin resistance.8 ,9 Unravelling mechanisms mediating OSAS-associated liver injury in NAFLD would have major research and clinical implications, as current therapeutic approaches to paediatric OSAS yield disappointing results: severe OSAS persists in as many as 50% of children after adeno-tonsillectomy, especially in the presence of obesity and severe OSAS;23 CPAP therapy improves polysomnographic parameters and surrogate markers of NAFLD,28 ,29 but its feasibility and patient adherence in children remain an issue.10 Therefore, it is important to investigate the novel alternative or complementary strategies for the treatment of paediatric OSAS.
An increased LPS–TLR-4 axis activation has been recently implicated in liver injury in NAFLD,12 but the mechanisms modulating the LPS–TLR-4 axis activation are unclear. We found that OSAS-related features, and specifically the duration of nocturnal hypoxaemia, were independently associated with increased plasma LPS and TLR-4 expression by hepatocytes, Kupffer cells and HSCs, key features of NASH and fibrosis in our patients (tables 5 and 6).
An increased hypoxia-induced intestinal permeability could account, at least in part, for the increased endotoxemia observed in our population (table 6, see online supplementary figure S2A and B): besides facing the anoxic gut lumen, the intestinal mucosa physiologically experiences profound daily fluctuations of perfusion and oxygenation, with the nadir occurring during fasting, a condition that may synergize with nocturnal hypoxaemia to exacerbate gut injury and break intestinal barrier in OSAS;30 ,31 alternatively. hypoxia may shape gut microbiota composition towards LPS-producing bacterial strains.
In our children, OSAS was also associated with increased TLR-4 expression by hepatocytes, Kupffer cells and HSCs, thereby enhancing hepatic susceptibility to circulating LPS (table 5).
Consistent with our findings, growing experimental evidence supports a role for intermittent hypoxia in TLR-4-induced inflammation in different disease models, including myocardial remodelling32 and lung and kidney ischaemia-reperfusion injury.33 ,34 Intriguingly, TLR-4 antagonists reversed inflammation in these models 33 ,34 and could represent a novel therapeutic target for OSAS-associated liver injury, as well.
Another novel finding of our study was the association of nocturnal hypoxaemia with the expansion of an adiponectin-defective HPC pool. The expansion of HPC pool following chronic liver injury has been proposed to trigger portal fibrosis, an early step in the fibrogenic process in paediatric NASH,1 ,13 but the mechanisms triggering HPC proliferation and their switch to a pro-fibrogenic phenotype are unclear. Our data indicate nocturnal hypoxaemia as a potential regulator of HPC pool size and adiponectin expression, thereby linking OSAS to the pro-fibrogenic HPC activation.13
Notably, these associations were independent of BMI and occurred also in non-obese individuals, suggesting that CIH is per se associated with altered gut–liver axis integrity and hepatic adiponectin expression independently of obesity in NAFLD (table 6; figure 1, see online supplementary tables 1 and 2).
We found no relationship between the severity of liver histology and of OSAS and insulin resistance indices in our patients with NAFLD, which is at odd with part, but not all, of the literature in adult NAFLD. These results are consistent with previous findings in paediatric NAFLD population with OSAS8 ,9 ,11 ,13 and confirms common indices of insulin resistance are more tightly related to the presence of fatty liver rather than to the severity of liver histology or sleep apnoea.8 ,9 An alternative, not exclusive, explanation could be that adipose tissue insulin resistance may be more tightly related to the severity of liver injury and of OSAS than currently used indices, which predominantly reflect hepatic and muscle insulin resistance.35
In conclusion, we showed that the CIH of OSAS is accompanied by gut–liver axis dysregulation. Future studies need to assess the causal role of CIH in these abnormalities, which cannot be ascertained by the cross-sectional design of our study, and to explore potential therapeutic implications of our findings in OSAS-related liver injury, including gut microbiota manipulation and TLR-4 antagonists.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online supplement
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.