Oxidative stress is associated with suspected non‐alcoholic fatty liver disease and all‐cause mortality in the general population

Abstract Background & Aims Non‐alcoholic fatty liver disease (NAFLD) is characterized by excessive lipid accumulation, inflammation and an imbalanced redox homeostasis. We hypothesized that systemic free thiol levels, as a proxy of systemic oxidative stress, are associated with NAFLD. Methods Protein‐adjusted serum free thiol concentrations were determined in participants from the Prevention of Renal and Vascular End‐Stage Disease (PREVEND) cohort study (n = 5562). Suspected NAFLD was defined by the Fatty Liver Index (FLI ≥ 60) and Hepatic Steatosis Index (HSI > 36). Results Protein‐adjusted serum free thiols were significantly reduced in subjects with FLI ≥ 60 (n = 1651). In multivariable logistic regression analyses, protein‐adjusted serum free thiols were associated with NAFLD (FLI ≥ 60) (OR per doubling of concentration: 0.78 [95% CI 0.64‐0.96], P = .016) even when adjusted for potential confounding factors, including systolic blood pressure, diabetes, current smoking, use of alcohol and total cholesterol (OR 0.80 [95% CI 0.65‐0.99], P = .04). This association lost its significance (OR 0.94 [95% CI 0.73‐1.21], P = .65) after additional adjustment for high‐sensitive C‐reactive protein. Stratified analyses showed significantly differential associations of protein‐adjusted serum free thiol concentrations with suspected NAFLD for gender (P < .02), hypertension (P < .001) and hypercholesterolemia (P < .003). Longitudinally, protein‐adjusted serum free thiols were significantly associated with the risk of all‐cause mortality in subjects with NAFLD (FLI ≥ 60) (HR 0.27 [95% CI 0.17‐0.45], P < .001). Conclusion Protein‐adjusted serum free thiol levels are reduced and significantly associated with all‐cause mortality in subjects with suspected NAFLD. Quantification of free thiols may be a promising, minimally invasive strategy to improve detection of NAFLD and associated risk of all‐cause mortality in the general population.


| INTRODUC TI ON
Non-alcoholic fatty liver disease (NAFLD) is defined as an abnormal accumulation of triglycerides (TG) in hepatocytes in the absence of excessive alcohol consumption. NAFLD is emerging as the most prevalent chronic liver disease in Western countries.
NAFLD encompasses a spectrum of diseases that ranges from simple steatosis to non-alcoholic steatohepatitis (NASH), in combination with fibrosis. NASH can subsequently lead to cirrhosis with its known complications, such as hepatocellular carcinoma (HCC). 1 Many co-morbidities coincide with the development of NAFLD, such as obesity, insulin resistance and metabolic syndrome, including type 2 diabetes (T2D). 2,3 In the general population, suspected NAFLD can be estimated by calculating proxies of the disease, including the Fatty Liver Index (FLI) or the Hepatic Steatosis Index (HSI). Both of these scoring systems are considered to be potential predictors for NAFLD and are based on prominent risk factors, including obesity indices, plasma triglycerides, gamma-glutamyl-transferase (GGT), body mass index (BMI) and liver transaminases. 4,5 A number of previous studies demonstrated that inflammation significantly contributes to the progression of NAFLD. During NAFLD, hepatocytes no longer tolerate the toxicity of accumulated fatty acids, resulting in dysfunction of cellular homeostasis, including mitochondrial β-oxidation and endoplasmic reticulum stress.
Following this, an overproduction of endogenous reactive species (consisting of reactive oxygen species [ROS], reactive nitrogen species [RNS] and reactive sulphur species [RSS]) as well as an inflammatory signalling cascade in the liver is being generated. 6,7 An increased production of reactive species subsequently leads to hepatocellular injury, which in turn results in secretion of inflammatory cytokines (TNF-α, IL-6, IL-10) and cellular death. The pro-inflammatory signalling pathways, increased β-oxidation in mitochondria and peroxisomes involved in this process lead to dysregulation of antioxidant homeostasis. 8 Thiols (R-SH) comprise a group of organosulphur compounds that can be found mainly in proteins (e.g. albumin) that contain sulphur-based amino acids (SAAs) as well as in low-molecular-weight (LMW) molecules like cysteine, homocysteine and glutathione.
Thiols are known to be involved in various biological processes, such as enzymatic catalysis, cell signalling and metal complexing in the body. 9 Most importantly, plasma or serum thiols are considered as a global marker of the systemic load of reactive species and as potent anti-oxidants because of their high reducing activity. 10 According to recently proposed terminology, reactive species can be identified as ROS, as well as RNS and RSS, which are collectively referred to as the 'Reactive Species Interactome' (RSI). 9 Depending on their redox state, thiols are classified as reduced or "free" thiols (R-SH) and oxidized or "bound" thiols, in which case a thiol is bound to another thiol via a disulfide bridge (R-SS-R'). In the circulation, the largest share of free thiols is embedded within the single cysteine residue (Cys 34 ) of albumin (HSA-SH) which exerts its antioxidant capacity. Remaining free thiols are classified as LMW free thiols, and the sum of protein free thiols and LMW free thiols is defined as total free thiols. Free thiols are able to scavenge reactive species and form disulphide bonds. Generally, total free thiol levels in serum could be interpreted as a direct and reliable reflection of the systemic redox system since they are readily oxidized by reactive species. 11,12 Typically, high concentrations of serum free thiols are representative of a more beneficial or 'healthy' redox status. Changes in serum free thiol levels have been reported for many risk factors in which reactive species are known to play a prominent role, such as ageing, smoking, alcohol consumption, as well as for several diseases including inflammatory bowel disease (IBD), cardiovascular disease (CVD), obesity and ischaemia-reperfusion injury. [13][14][15] Only one study reported that total serum thiol concentration is reduced while thiol-disulphide level is increased in NASH patients, compared to healthy controls. 16 In this study, we determined systemic levels of serum free thiols in 5562 participants included in the Prevention of Renal and Vascular End-stage Disease (PREVEND) cohort, a large population-based cohort study from the Northern part of the Netherlands. Firstly, we compared protein-adjusted serum free thiol levels between subjects with FLI < 60 and FLI ≥ 60 values and established associations between free thiol levels and multiple clinical, biochemical and NAFLD-specific parameters. Secondly, we investigated the association between baseline protein-adjusted serum free thiol concentrations and the risk of all-cause mortality during a follow-up of 10 years.

| Study population
This study used data from the Prevention of REnal and Vascular ENd-stage Disease (PREVEND) cohort study. 17 This is a large, prospective population-based cohort study with participants from the Northern part of the Netherlands. The PREVEND study was set up to investigate cardiovascular and renal disease outcomes.
From 1997 to 1998, 85 421 inhabitants aged 28-75 years from the Northern part of the Netherlands, received a questionnaire asking information about demographics, medication use, cardiovascular disease and pregnancy, including a request to supply an early morning urine sample. Participants who had a previous diagnosis of type 1 diabetes mellitus, insulin-treated type 2 diabetes mellitus and pregnant women were excluded from the study. In total, 40 856 subjects responded to the questionnaire and were analyzed for urinary albumin concentrations. Subjects with a urinary albumin concentrations ≥10 mg/L (n = 6000) were invited to visit the outpatient research clinic, as well as a random selection of participants with urinary albumin concentrations <10 mg/L (n = 2592). The PREVEND study consisted of a total of 8592 participants who completed the full study program. 18 However, for the current study we excluded subjects (n = 3030) of which data on serum levels of free thiols and clinical and biochemical variables to calculate the Fatty Liver Index (FLI), as a proxy of NAFLD, were not available. This study was approved by the Institutional Review Board (IRB) of the University Medical Center Groningen (UMCG). The study was conducted in accordance with the principles of the Declaration of Helsinki (2013).
All study participants provided written informed consent.

| Data collection
All study participants visited the outpatient research clinic of the UMCG, Groningen, the Netherlands. During the first visit, participants were requested to complete a questionnaire that contained information about demographics, health status, history of cardiovascular diseases (CVD), use of medications and lifestyle (e.g. self-reported smoking and alcohol consumption). Smoking was categorized as either current smoking or never or previous smoking. Alcohol consumption was documented with the assumption of one alcoholic drink to contain 10 grams of alcohol. History of cardiovascular disease included the following: hospitalization for myocardial ischaemia, obstructive coronary artery disease or revascularization procedures. Subsequently, anthropometric measurements were performed, including height (meters), weight (kilograms), body-mass index (BMI, weight divided by squared height), waist circumference (cm, defined as the smallest girth between rib cage and iliac crest), and waist/hip ratio (waist circumference divided by the largest girth between waist and thigh). 19,20 During the second visit, systolic and diastolic blood pressure was measured automatically every minute until 8 minutes in supine position (Dinamap XL Model 9300 series device, Johnson & Johnson Medical). Blood pressure was defined as the average of the last two measurements in this procedure. Next, venous serum samples were withdrawn after an overnight fast while the participants had rested for 15 minutes. In addition, patients were asked to collect 24-hours urine specimens after they were provided with both oral and written instructions. In the current study, data were used of participants who completed the second screening evaluation in the PREVEND study.  Ultimately, serum free thiol concentrations were adjusted to total serum protein levels (measured according to standard procedures) by calculating the free thiol/total protein ratio (μmol/g of protein).

| Measurement of serum free thiols
This adjustment was performed since serum proteins harbour the largest amount of free thiols and therefore largely determine the amount of potentially detectable free thiols. 24

| Study outcomes and definitions
The estimated glomerular filtration rate (eGFR) was calculated using as determined by ultrasonography. 5 Therefore, FLI ≥ 60 was used as a definition of suspected NAFLD, which is used nowadays as one of the best-validated steatosis scores for large scale screening studies. 26 Alternatively, we used the Hepatic Steatosis Index (HSI), which has been used previously in predominantly Asian populations and is defined as follows 4 : HSI = 8 × ALT/AST ratio + BMI (+2, if diabetes; +2, if female). The optimal cut-off value of the HSI for detecting NAFLD is a score of 36. In the above equations, BMI was expressed as kg/m 2 , triglycerides as mmol/L and gamma-glutamyltransferase confidence intervals (CI). Stratified analyses were performed to examine the association between serum free thiols and NAFLD across various subgroups. Survival distributions for subjects with and without NAFLD were assessed according to tertiles of protein-adjusted serum free thiol concentrations using Kaplan-Meier curves and compared to each other using log-rank tests. Survival time was defined from baseline (time of serum sample withdrawal) until the date of the last examination participants attended, either death or January 1, 2010 (end of follow-up period). Subsequently, Cox proportional hazards regression analyses were performed to assess associations between protein-adjusted serum free thiol concentrations and the risk of all-cause mortality, expressed as hazard ratios (HRs) (per doubling) with corresponding 95% CIs. Univariable associations were followed by multivariable models to adjust for potential confounding factors. Data analysis was performed using SPSS Statistics 25.0 software package (SPSS Inc) and data visualization using GraphPad Prism 5.0 (GraphPad software). Two-tailed P-values ≤.05 were considered statistically significant.

| Baseline characteristics of the study population
Baseline characteristics of the study population are presented in Table 1. The study population consisted of 5562 participants, of whom 1651 (29.7%) subjects were classified with a FLI ≥ 60.
Participants classified with a FLI ≥ 60 were significantly older, as compared to subjects with a FLI < 60 (56.0 years vs 49.8 years, P < .001). In addition, subjects with a FLI ≥ 60 more frequently had a history of cardiovascular disease (P < .001), MetS (P < .001) and more often used antihypertensive medication (P < .001) and lipidlowering drugs (P < .001). Moreover, anthropometric tests (ie BMI, waist circumference, waist/hip ratio), cholesterol levels and liver transaminase levels were higher in subjects with a FLI ≥ 60 (P < .001 for all). Conversely, LDL-cholesterol levels were not found to be significantly different between groups. With regard to serum levels of protein-adjusted free thiols, we observed significantly reduced concentrations in subjects with a FLI ≥ 60, as compared to subjects with a FLI < 60 (4.91 μmol/L/g vs 5.05 μmol/L/g, P < .001).

| Associations between protein-adjusted serum free thiol levels and FLI and HSI scores
Multivariable logistic regression analyses were subsequently performed in order to establish the extent to which serum levels of free thiols were associated with a FLI ≥ 60 ( Table 2). In the age-and gender-adjusted analysis, we found a significant association between protein-adjusted free thiols ( 2 log-transformed, per doubling of con- Similar results were observed in the analysis for HSI (Table S1). For instance, the association between HSI and serum levels of proteinadjusted free thiols ( 2 log-transformed, per doubling of concentration) only lost its significance after additional adjustment for hs-CRP (OR 0.87 [95% CI 0.68-1.10], P = .24). Stratified analyses for the association between protein-adjusted serum free thiols (per doubling) and FLI scores are presented in Table 3. Stratification by gender, the presence of hypertension and the presence of hypercholesterolemia showed significant differences between groups. Corresponding HRs were lower for female subjects (P interaction = 0.02), subjects without hypertension (P interaction = 0.001) and subjects without hypercholesterolemia (P interaction = 0.003). Comparable results were obtained in stratified analyses when using the HSI instead of the FLI (Table S2).

| Protein-adjusted serum free thiols and risk of all-cause mortality
During follow-up, 291 (5.2%) subjects died (FLI < 60, n = 162 (4.1%), FLI ≥ 60, n = 129 (7.8%)). Kaplan-Meier survival analysis showed a significantly differential survival distribution between tertiles of protein-adjusted serum free thiols among subjects with a FLI < 60 and FLI ≥ 60 (Figure 1, P < .001, log-rank test). Cox proportional hazards regression analyses showed a significant inverse predictive association between 2 log-transformed protein-adjusted serum free thiol concentrations and the risk of all-cause mortality for subjects with a FLI < 60 (Table 4A, (Table 4B, (Table 4B, (Table 4B, (Table S3). However, statistical significance vanished after adjustment for potential confounders in subjects of both subgroups, with the exception of the highest tertile of protein-adjusted serum free thiol concentrations in the group with HSI ≥ 36 (Table S3B,

| D ISCUSS I ON
In this study, we reported that protein-adjusted serum free thiol concentrations, as a marker of the systemic redox status, were lowered in subjects with suspected NAFLD (FLI ≥ 60). In addition, proteinadjusted serum free thiols were significantly associated with an increased risk of all-cause mortality in subjects with suspected NAFLD in this population-based cohort. Multivariable regression analyses showed maintenance of this significant association after adjustment for potential confounding factors, including the adjustment for systolic blood pressure, diabetes, current smoking, use of alcohol and total cholesterol in subjects with FLI ≥ 60. As expected, this association lost its significance after additional adjustment for TA B L E 1 Clinical and laboratory characteristics including protein-adjusted serum free thiols in 3911 subjects with a fatty liver index (FLI) < 60 and 1651 subjects with a FLI ≥ 60 Data are presented as mean ± standard deviation (SD) for normally distributed data or median with interquartile ranges (IQR) for non-normally distributed data. These results were consistent in subjects having an HSI > 36. Altered serum thiol balance in NAFLD has been reported in only one study before. Asil et al reported that serum total thiols were reduced in patients with NASH and simple steatosis as compared to healthy controls (n = 90). 16 In comparison to our data, that study focused on total/native thiol ratios, included relatively low numbers of patients and applied liver biopsy to define NAFLD. Several other studies reported that there were no significant differences with regard to total serum thiol concentrations in subjects with insulin resistance (IR), type 2 diabetes (T2D). [30][31][32] Additionally, in paediatric subjects, increased serum thiols such as cysteine and homocysteine were observed in patients with NAFLD, while they were reduced in patients with NASH or liver fibrosis. 33 However, these studies focused on thiol/disulphide homeostasis using different measurement protocols i.e. distinct thiol-reactive reagents which compromises comparability of results between studies as measurements of either free thiols or total thiols lead to different classifications and terminology. 34 In addition, all these studies were based on datasets with relatively low numbers of study participants or they focused on different types of populations, e.g. solely on paediatric or female subjects.
Oxidative stress is referred to as an imbalance between oxidant and anti-oxidant substances. In NAFLD, the antioxidant system is TA B L E 2 Multivariable logistic regression analysis to test the relationship between FLI and serum levels of protein-adjusted serum free thiols ( 2 log-transformed) disrupted because of excessive fat accumulation-mediated endoplasmic reticulum (ER) stress and mitochondrial β-oxidation dysfunction, leading to oxidative stress-induced complications caused by endogenous production of reactive species. 6,7 It should be noted that serum free thiols have been considered a prominent antioxidant marker in serum because of their potent capacity to scavenge reactive species. 9,14,31 High-sensitive C-reactive protein (hs-CRP) has been reported to be a prominent ROS-induced inflammatory marker in NAFLD. 35 A diminished antioxidant capacity is significantly associated with hs-CRP during disrupted redox homeostasis in multiple oxidative stress-related diseases. [36][37][38][39] Similarly, in our study, serum hs-CRP levels were significantly increased in subjects with FLI ≥ 60. In addition, in multivariable regression analyses, the persistent statistically significant associations of 2 log-transformed protein-adjusted serum free thiols and systolic blood pressure, diabetes, current smoking, use of alcohol and total cholesterol with FLI ≥ 60 lost their significances after adjustment for hs-CRP. The same results were obtained in the analysis of the HSI > 36 group.
This similar association of hs-CRP and thiols has been observed in several studies related to antioxidant homeostasis. For instance, one study found a negative correlation between hs-CRP levels and thiol/disulphide ratio and a positive correlation with total thiols during acute appendicitis in children (n = 80). 40 In addition, in patients with inflammatory bowel disease (IBD), hs-CRP was also significantly inversely associated with free thiols. 14 A recent meta-analysis study reported a significant positive association between NAFLD and all-cause mortality. 46 Thus, there is importance for an early and non-invasive screening method to enable prediction for all-cause mortality in NAFLD. 47 Of note, measuring free thiols in serum is relatively minimally invasive. In this study, using Cox proportional hazard regression analysis, we showed a significant predictive association between protein-adjusted serum free thiols and the risk of all-cause mortality for subjects with FLI ≥ 60. This association lost its significance after adjustment for potential confounders in subjects with a FLI < 60 and remained significant in subjects with a FLI ≥ 60 (Table 4). Since serum free thiols could be a potential therapeutic target in NAFLD, interventions targeted to increase the free thiol pool could also potentially predict the risk of all-cause mortality. Taken together, protein-adjusted serum free thiols could be a prominent predictor of all-cause mortality in NAFLD. However, it is important to further investigate the association between serum free thiol levels and different stages of NAFLD.
Our study has several strengths and limitations that need to be acknowledged. For example, to the best of our knowledge, this is the first large study to report a significant association between serum free thiols -as a minimally invasive method to quantify systemic oxidative stress -and NAFLD. Most importantly, the protein-adjusted serum free thiol level was significantly associated with the risk of all-cause mortality in patients with identified NAFLD. We were able to establish this association in a population-based cohort study with a large sample size (n = 5562) that enabled us to properly adjust for potential confounding variables with sufficient study power. Furthermore, the association of serum free thiols with suspected NAFLD individuals in the general population were determined using two different, but accurate proxies of NAFLD: the FLI and HSI indices. However, FLI cannot identify absolute clinical NAFLD because of the lack of discrimination between severe steatosis levels and liver fat, but it is considered to be an acceptable method to indicate NAFLD in large-population based studies. 48 Although the HSI has only been validated in Asian populations, results were comparable in our cohort. Indeed, both methods are widely accepted and recommended to characterize NAFLD in large population-based cohort studies. 4 However, it was not possible to correlate free thiols with the different stages of NAFLD, e.g. NASH, fibrosis or cirrhosis because of the lack of necessary data to enable this characterization. Similarly, it was not possible to exclude other potential causes of liver disease as these data were not available in the present cohort.
In conclusion, protein-adjusted serum free thiol concentrations were significantly reduced in subjects with suspected NAFLD, even after adjustment for known risk factors for NAFLD.
Furthermore, protein-adjusted serum free thiols were significantly associated with the risk of all-cause mortality in subjects with suspected NAFLD. Future studies are warranted that focus on the clinical utility of systemic free thiols in patients with NAFLD and the detailed discovery of potential associations with therapeutic outcome, disease course and overall prognosis. As free thiols are known to be receptive for therapeutic manipulation, future thiol-targeted therapy should be investigated as well to ameliorate disease outcome in NAFLD.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.