This study was performed in agreement with the Declaration of Helsinki, and with local and national laws. The Human Ethics Committee of the Hospital Universitario Santa Cristina (HUSC, Madrid, Spain) approved all procedures (PI-688A). This cross-sectional study included 20 patients with gallstones to whom a programmed laparoscopic cholecystectomy was performed in the HUSC. All participants gave a written consent for a perioperative liver biopsy and a postoperative respiratory polygraphy as part of an experimental protocol designed to evaluate the relationship between sleep disturbances and liver disease. All subjects included drank <20 g/day of alcohol, had no previous respiratory disorders, were not having potentially hepatotoxic drugs, had no analytical evidence of iron overload, and were seronegative for autoantibodies, for hepatitis B virus, hepatitis C virus, and human immunodeficiency virus.

The polygraphic studies were performed at night in the Sleep Laboratory of the HUSC (Madrid, Spain). For interpretation, the recommendations of the American Academy of Sleep Medicine (AASM) for the diagnosis of OSA were followed. The apnea/hypopnea index (AHI) was used as diagnostic criteria for severity of OSA: AHI <5, no OSA; AHI 5–14, mild OSA; AHI 15–29, moderate OSA; AHI ≥30, severe OSA. In addition, nocturnal hypoxemia parameters including oxygen desaturation index (ODI), cumulative sleep time percentage with oxyhemoglobin saturation (SpO₂) <90% (Tc90) and minimum SpO₂ were analyzed.

Twelve-weeks-old C57BL/6J mice were purchased from Charles River Laboratories (Saint Germain sur L'Arbresle, France) and divided into two groups of 10 mice. The control mice (C mice) were placed in conditions of normoxia while the experimental group (IH mice) was subjected to IH conditions. Every minute, IH mice received air containing an oxygen fraction of 5% for 20 s, followed by 40 s of room air, during 6 h per day, 5 days a week for a total of 8 weeks. Control mice were only exposed to room air (23). At the end of the experiment, mice were anesthetized, sacrificed and livers were harvested. All experimental procedures were approved by the Ethical Committee of the University of Barcelona (174/18−10268).

Liver sections (5 μm) were embedded in paraffin and cut using a Microm microtome (Midland, ON, Canada). After cutting, sections were stained with hematoxylin (1.09235.0500, PanReac AppliChem, Barcelona, Spain) and eosin (71211, Thermo Fisher Scientific, Inc., Madrid, Spain) and with Masson's Trichrome Solution (Masson Trichome Kit with Aniline Blue 04-010802, Milan, Italy). Once stained, the severity of steatosis was quantified by a single-blind hepatopathologist. Specifically, Kleiner's histological scoring system was employed to evaluate the degree of steatosis, lobular inflammation, hepatocellular ballooning, and the stage of fibrosis (24). The following percentages of steatotic hepatocytes were used in the histological assessment: 0–5% hepatocytes, grade 0; 5–33%, grade 1; 33–66%, grade 2; and >66%, grade 3. Histologic diagnosis of liver biopsies was classified into two groups: simple steatosis without hepatocellular ballooning nor lobular inflammation, also termed NAFL, and NASH. Minimal criteria for NASH included the combined presence of grade 1 steatosis, lobular inflammation and hepatocellular ballooning with or without fibrosis. NAFLD activity score was also calculated for each liver biopsy (24). To this end, three different lobular areas were analyzed in each sample. Representative images were taken using a Nikon Eclipse E400 optical microscope (Nikon, Tokyo, Japan) and the NIS Elements Imaging Software (Melville, NY, USA).

Liver tissue was embedded in Tissue-Tek® O.C.T.™ Compound (Sakura Finetek Europe, Netherlands). Sections (10 μm) were then cut using a Leica CM1510S cryostat (Leica Microsistemas S.L.U, Barcelona, Spain), stained using an Oil Red O biological stain (Sigma-Aldrich, St. Louis, MO, USA) working solution (60% ORO/isopropanol w:v), and counterstained with hematoxylin (1.09235.0500, PanReac AppliChem). Three different lobular areas were analyzed in each sample and photographed using a Nikon Eclipse E400 optical microscope (Nikon) and the NIS Elements Imaging Software (Melville). Intensity of red stain was quantified using ImageJ Biological Image Analysis (NIH) and reported as the average value in arbitrary units (a.u.).

Triglycerides (TGs) were extracted as described previously (25). Briefly, liver biopsy samples (15–20 μg) were homogenized in distilled water. Chloroform (−20°C) and methanol (−20°C) were added to each sample. Samples were centrifuged and the triglyceride-containing layer was collected. Once purified, TGs were suspended in isopropanol and analyzed using a colorimetric kit (SpinReact, Girona, Spain). Absorbance values were obtained using a Dynex Spectra MR Microplate spectrophotometer/computer software (Chantilly, VA, USA) and graphically expressed as mg/dl.

Liver biopsy samples (15–20 μg) were homogenized in an extraction buffer containing the following: 10 mM ethylene-diamine-tetraacetic acid (EDTA), 50 mM Hepes, 50 mM sodium pyrophosphate, 0.1 mM NaF, 10 mM Na₃VO₄, 1% Triton X-100 and protease inhibitors. Protein extracts were stored at −80°C after centrifugation. A small aliquot of sample was used for protein quantification (Bradford method). The samples were then prepared to be loaded into 8% SDS-PAGE gels. After running, proteins were further transferred to Inmunoblot nitrocellulose membranes (BioRad Inc., Madrid, Spain), blocked with 5% non-fat dry milk and incubated overnight with primary antibodies: CD36/SR-B3 (1:1000, NB400-144, Novus Biotechne, Abingdon, United Kingdom) and anti-βactin (1:5000, A-5441, Sigma Aldrich). Then, the corresponding secondary antibodies were added (Santa Cruz Biotechnology Inc., Heidelberg, Germany). Using the Bio-Rad Clarity™ Western ECL Substrate (BioRad Inc.), the immunoreactive bands were visualized by the ImageQuant LASD 4000 digital imaging system (GE Healthcare Europe, Barcelona, Spain). Densitometric analysis of the band was performed using ImageJ Biological Image Analysis (NIH), normalized against the loading control (βactin), and graphically expressed as fold change relative to control condition (1).

RNA was extracted from liver samples using the TRIzol® reagent (Vitro, Sevilla, Spain). Samples were then reverse transcribed using the Reverse Transcription System kit (Promega Inc., Madison, WI, USA). A BioRad T100™ Thermal Cycler was used to carry out the reverse transcription. Quantitative real-time polymerase chain reaction (RT-qPCR) was performed to assess gene expression using a StepOnePlus™ Real Time PCR System Sequence Detector (Thermo Fisher Scientific Inc.). Samples were prepared using a SYBER Green qPCR Kit (Promega Inc.) and d(N)6 random primers were purchased from Metabion (Planegg, Germany). Primer sequences are detailed in Supplementary Table 1. Each sample was run in duplicated, normalized in comparison to 36B4 gene expression, and graphically expressed as fold change relative to control condition (1).

Paraffin-embedded liver biopsy sections (5 μm) were deparaffinized and rehydrated. Sections were then placed in antigen retrieval buffer (10 mM sodium citrate, pH 6–7), boiled for 20 min at 95°C and incubated with a blocking solution for 1 h before being immunostained with the CD36 antibody (1:200, NB400-144, Novus) for 16 h in a moisture chamber. The EnVision™ FLEX Mini Kit, High pH (Link) (Agilent, Santa Clara, CA, USA) was used for visualization according to the manufacturer's instructions. Three different lobular areas were analyzed in each sample and images were captured using a Nikon Eclipse E400 optical microscope (Nikon) and the NIS Elements Imaging Software (Melville). Intensity of CD36 stain was quantified using the FIJI software (NIH) and reported as the average value in arbitrary units (a.u.).

Categorical variables were presented as percentage and were compared by the Pearson χ² test. Continuous data were shown as standard deviation (SD) or standard error of mean (SEM), and were compared using the unpaired t-test or Mann-Whitney U-test, as indicated. The Spearman's r-test was used to evaluate correlations. All statistical analyses were performed using the GraphPad Prism 6 software (GraphPad Software Inc., San Diego, CA, USA) and SPSS statistical software version 24.0 (IBM SPSS Statistics, Armonk, NY), with a p < 0.05 considered statistically significant.

Twenty patients undergoing programmed cholecystectomy had both a liver biopsy and a sleep study. Overall, the mean age of the study population was 46.5 years, 14 (70%) were female and 9 (45%) had a diagnosis of OSA by polygraphy (AHI > 5). Patient characteristics from the cohort included in this study are presented in Table 1.

OSA patients were significantly older than those without OSA (p = 0.004) and women predominated in both groups. In order to evaluate the presence of obesity in the entire study population, body mass index (BMI) was calculated and waist circumference was measured in each participant. Regarding BMI, the study population showed an overweight status and no significant differences were found among the different patient groups studied (p = 0.201), whereas waist perimeter was significantly higher in patients with OSA than in those without (p = 0.034).

As expected, OSA patients had a significantly higher rate of oxygen desaturation per hour of sleep (ODI) and percentage of sleep time with oxygen saturations lower than 90% (Tc90) with respect to patients without OSA (Table 1).

Regarding metabolic parameters, basal glucose levels did not significantly differ between groups, while insulin levels and the degree of insulin resistance assessed by HOMA-IR index were higher in OSA patients than in those without OSA, but these differences were not statistically significant (Table 1).

Although no significant differences were observed in liver enzymes between the two groups, there was a higher prevalence of NAFLD and evidence of more severe disease among patients with OSA (Table 1 and Figure 1A). Surprisingly, 36.4% of patients without OSA exhibited NAFLD, all of them featuring simple steatosis. In the OSA group, however, 55.6% of them had simple steatosis and 22.2% showed histological features of steatohepatitis (NASH) (Table 1 and Figure 1A). Hepatic fibrosis was not observed. All other variables did not significantly differ between groups.

Next, we wanted to investigate whether OSA might alter CD36 expression in the liver. Interestingly, hepatic mRNA levels of CD36 were significantly higher in patients with OSA when compared with control patients (Figure 1B), and its expression significant positively correlated with both AHI and ODI values in the entire study population (Figure 1C), but not with Tc90 (Supplementary Figure 1A). In parallel, an increase of CD36 protein expression was also observed in the livers from OSA patients compared with those from the control group detected by immunostaining (Figure 1D).

In order to investigate whether intermittent hypoxia (IH), one of the main features of OSA, contributes to liver steatosis and to the increase of CD36 expression observed in OSA patients, a mouse experimental model of OSA was used. After 2 months of IH exposure, histological examinations of liver tissue revealed that 60% of the mice submitted to IH exhibited signs of mild hepatic steatosis while control mice displayed normal liver features (Figures 2A,B). Liver fibrosis was not detected in any of the groups (Supplementary Figure 2).

Next, we investigated the amount of lipids performing an Oil Red O staining on sections of liver biopsy samples from all mice. The results indicated that there was a significant increase in average red-stain intensity (directly proportional to lipid content) among the IH mice when compared to control mice (Figure 2C). To further evaluate hepatic lipid content, triglycerides were extracted and quantified from liver biopsies of both IH and C mice. Triglyceride levels of the IH mice were greater than those observed in the C mice (Figure 2D).

tHEN, we analyzed the hepatic expression of genes involved in the regulation of lipid metabolism. We observed a significant increase in the expression of genes implicated in lipid synthesis, such as Fasn (fatty acid synthase) and Scd1 (stearoyl-CoA desaturase 1), among livers from mice exposed to IH (Figure 3A); however, no differences were observed in the expression of genes implicated in β-oxidation, such as Cpt1a (carnitine palmitoyltransferase I) and Ppara (peroxisome proliferator activated receptor alpha) (Figure 3A).

With respect to CD36, its hepatic mRNA expression was significantly increased in mice submitted to IH compared to those maintained in normoxic conditions (Figure 3B), which was also found in OSA patients. In parallel, its protein expression determined by both Western blot (Figure 3C) and immunohistochemistry (Figure 3D) was elevated in the livers of IH mice.