The endothelial function biomarker soluble E‐selectin is associated with nonalcoholic fatty liver disease

Abstract Background & Aims Plasma soluble E‐selectin (sE‐selectin) is a frequently used biomarker of systemic endothelial dysfunction. The present study explored the relationship between nonalcoholic fatty liver disease (NAFLD) and plasma sE‐selectin levels. Methods Expression of E‐selectin in liver, visceral adipose tissue (VAT) and muscle was studied in relation to plasma sE‐selectin in severely obese individuals (n = 74). The course of hepatic E‐selectin expression in relation to hepatic steatosis and inflammation was examined in C57BL/6J LDLR−/− mice on a Western‐type diet. The relationship between biomarkers of NAFLD, that is, plasma aminotransferase (ALT) and NAFLD susceptibility genes (rs738409 [PNPLA3] and rs1260326 [GCKR]), and plasma sE‐selectin was studied in the combined CODAM (n = 571) and Hoorn (n = 694) studies. Results E‐selectin expression in liver, not VAT or muscle, was associated with plasma sE‐selectin in severely obese individuals (β = 0.26; 95% CI: 0.05‐0.47). NAFLD severity was associated with hepatic E‐selectin expression (P = .02) and plasma sE‐selectin (P = .003). LDLR−/− mice on a Western‐type diet displayed increased hepatic E‐selectin expression that followed the same course as hepatic inflammation, but not steatosis. In the CODAM study, plasma ALT was associated with plasma sE‐selectin, independent of potential confounders (β = 0.25; 95% CI: 0.16‐0.34). Both rs738409 and rs1260326 were associated with higher plasma sE‐selectin in the combined CODAM and Hoorn studies (P = .01 and P = .004 respectively). Conclusions NAFLD and related markers are associated with higher expression of hepatic E‐selectin and higher levels of plasma sE‐selectin. Further studies are required to investigate the role of E‐selectin in the pathogenesis of NAFLD and the applicability of sE‐selectin as a plasma biomarker of NAFLD/NASH.


| INTRODUC TI ON
Nonalcoholic fatty liver disease (NAFLD) is emerging as the principal cause of end-stage liver disease requiring transplantation. 1 The pathogenesis is complex and involves genetic, inflammatory, (gut) microbiotic, metabolic, nutritional and lifestyle factors. 2 Recent studies have implicated a role for liver sinusoidal endothelial cells in the pathogenesis of NAFLD. Liver sinusoidal endothelial dysfunction favours the development of steatosis and recruitment of inflammatory cells. 3 This premise is largely based on animal studies, given the difficulties of acquiring human liver tissue to conduct endothelial function tests.
Plasma soluble E-selectin (sE-selectin) and vascular cell adhesion molecule-1 (sVCAM-1) have often been used as biomarkers of systemic endothelial activation. [4][5][6] Of interest, we and others have previously shown that NAFLD is a significant determinant of plasma endothelial biomarkers, in particular sE-selectin. 7,8 Furthermore, it has been shown that hepatic E-selectin is actively involved in hepatic neutrophil infiltration in mice exposed to chronic plus binge ethanol feeding. 9 The aim of the present study was, therefore, to study the relationship between NAFLD and sE-selectin in more detail. We aimed to assess (a) the relationship between NAFLD and he- Results: E-selectin expression in liver, not VAT or muscle, was associated with plasma sE-selectin in severely obese individuals (β = 0.26; 95% CI: 0.05-0.47). NAFLD severity was associated with hepatic E-selectin expression (P = .02) and plasma sE-selectin (P = .003). LDLR −/− mice on a Western-type diet displayed increased hepatic E-selectin expression that followed the same course as hepatic inflammation, but not steatosis. In the CODAM study, plasma ALT was associated with plasma sE-selectin, independent of potential confounders (β = 0.25; 95% CI: 0. 16-0.34). Both rs738409 and rs1260326 were associated with higher plasma sE-selectin in the combined CODAM and Hoorn studies (P = .01 and P = .004 respectively).
Conclusions: NAFLD and related markers are associated with higher expression of hepatic E-selectin and higher levels of plasma sE-selectin. Further studies are required to investigate the role of E-selectin in the pathogenesis of NAFLD and the applicability of sE-selectin as a plasma biomarker of NAFLD/NASH.

K E Y W O R D S
endothelium, E-selectin, genetic epidemiology, nonalcoholic fatty liver disease, translational research

Key points
• The pathogenesis of nonalcoholic fatty liver disease (NAFLD) is multifactorial. Liver sinusoidal endothelial cells have been suggested to play a role.
• The aim of the present study was to assess the relationship between NAFLD and markers of endothelial activation (ie expression and plasma levels of E-selectin) in mice, severely obese individuals undergoing bariatric surgery and the general population.
• NAFLD severity is associated with hepatic E-selectin expression in mice and severely obese individuals.
• E-selectin expression in liver, but not in other organs, is associated with plasma sE-selectin levels in severely obese individuals.
• NAFLD susceptibility genes, that is, PNPLA3 and GCKR, and plasma ALT levels are associated with plasma sE-selectin levels, independent of potential confounding factors.
• These findings favour further study to elucidate the role of E-selectin in the pathogenesis of NAFLD and the applicability of sE-selectin as a plasma biomarker of NAFLD/NASH.

| Liver biopsies in severely obese individuals undergoing bariatric surgery
Details of this study have been described elsewhere. 10,11 Briefly, liver, muscle and visceral adipose tissue (VAT) biopsies were collected from severely obese individuals (BMI > 40 kg/  The average score was used for statistical analysis. RNA was isolated from the liver using Trizol reagent (Ambion). Next, E-selectin and TNF gene expression were determined using IQ SensiMix SYBR master mix (Bioline) on a CFX96 Touch with CFX manager software (Biorad). The geometric mean of Cyclophillin and Beta2-microglobulin was used as reference (see Table S1 for primer sequences) and the ΔΔCT method was applied to calculate the gene expression levels. 17 Plasma sE-selectin levels were determined using a DuoSet ELISA kit (R&D Systems). All experiments were approved by the Animal Experiments Committee of Maastricht University (Maastricht, The Netherlands) and in compliance with the relevant guidelines from the Directive 2010/63/ EU of the European Parliament on the protection of animals used for scientific purposes.

| The combined Hoorn and CODAM studies
The Hoorn and Cohort on Diabetes and Atherosclerosis Maastricht cose tolerance test to determine glucose metabolism state, that is, normal glucose metabolism (NGM), impaired glucose metabolism (IGM) (comprising both impaired fasting glucose and impaired glucose tolerance) and T2DM. All participants provided informed consent. The Hoorn and CODAM studies were performed according to the Declaration of Helsinki 12 and approved by the local Medical Ethical Committees. As the protocols of both studies were similar, data were combined for statistical analyses, as has been done before. 21,22 In the combined cohorts, 1286 individuals were available for genotyping. Genotyping of GCKR (rs1260326) and PNPLA3 (rs738409) was performed with validated Invitrogen TaqMan assays (Thermo Fisher Scientific). In CODAM, GCKR (rs1260326) was genotyped as part of a genome-wide association study array, sE-selectin and sVCAM levels, as described before. 18,20 Plasma levels of sE-selectin and sVCAM were determined using a MSD multiplex assay (Meso Scale Diagnostics), which is comparable to single-biomarker methods, such as ELISA, but could result in a different absolute concentration. 23 In the Hoorn study, plasma ALT levels were only determined at the baseline visit (1989)(1990)(1991)(1992), not during the follow-up measurements (2000-2001) that were used for the present study. The relationships of plasma ALT with sE-selectin and sVCAM levels were therefore studied in participants of the CODAM study only. Since genotype is time-independent, the associations of rs1260326 and rs738409 with plasma ALT (and triglycerides, sE-selectin and sVCAM) were studied in the combined Hoorn and CODAM studies.

| Statistical analyses
Data are presented as percentage of total, mean ± SD or median (interquartile range). Non-normally distributed variables were log transformed before further analyses. Differences between groups or associations between variables were analysed using one-way ANOVA or linear/logistic regression, respectively, with adjustments for age, sex, and -if applicable -cohort (ie the Hoorn and CODAM studies). For readability purposes, figure data are presented after adjustment for age and sex (and cohort) in which the residuals are normalized to the average age and sex (and cohort) of that study population.
Multivariable linear regression was used to study the associations between plasma ALT and sE-selectin and sVCAM levels, independent of potential confounding factors, that is, BMI, smoking, alcohol intake, plasma lipids (ie total cholesterol and HDL-cholesterol), systolic blood pressure, HbA1c, use of glucose-lowering medication, use of lipid-modifying medication, inflammatory markers (ie C-reactive protein, amyloid A, interleukin-6, interleukin-8 and tumour necrosis factor α) and history of cardiovascular disease. Finally, the associations between the NAFLD susceptibility genes PNPLA3 (rs738409) or GCKR (rs1260326) and plasma ALT, triglycerides, sE-selectin and sVCAM levels were analysed under the assumption of an additive mode of inheritance, based on previous reports. [24][25][26] Results were considered statistically significant at P < .05. All statistical analyses were carried out by the IBM Statistical Package of Social Sciences (SPSS) version 23 for Windows (SPSS Inc).

| NAFLD histological stage is associated with hepatic E-selectin mRNA expression in severely obese individuals
Hepatic E-selectin and VCAM mRNA expression patterns were studied in severely obese individuals undergoing bariatric surgery (see Table 1 15 and therefore allows a disentanglement of hepatic steatosis from inflammation. Indeed, the onset of hepatic steatosis was observed after 72 hours and increased until 3 weeks on a Western-type diet (P = .01 and P = .001 for 1 and 3 weeks versus chow diet, respectively, Figure 1A), whereas hepatic inflammation reached its peak at 1 week, but appeared to decrease at week 3 (P = .001 and P = .08 for 1 and 3 weeks versus chow diet, respectively, Figure 1B). In line with this inflammation score, hepatic TNF mRNA expression was strongly upregulated only after 1 week Western-type diet (P < .001, Figure 1C). Hepatic E-selectin mRNA expression followed a similar pattern and was only significantly different from the chow diet on week 1 (P = .001, Figure 1D). Of note, this was not the case for plasma sE-selectin levels, which already increased after 72 hours and remained stable thereafter ( Figure S1).

| Hepatic E-selectin mRNA expression is associated with plasma sE-selectin levels in severely obese individuals
Analogous to the hepatic mRNA expression pattern, we observed a significantly positive relationship between NAS and plasma sEselectin levels in severely obese individuals (P for trend = .003, Figure 2A). In contrast to the negative trend between NAS and hepatic VCAM mRNA expression, a positive trend was now found for the relationship between NAS and plasma sVCAM levels (P = .07, Figure 2B). The discriminatory ability of sE-selectin levels to distinguish NASH (ie NAS ≥ 5) from non-NASH (ie NAS ≤ 2) was statisti-

| Plasma ALT levels are independently associated with plasma sE-selectin levels in the CODAM study
Linear regression models were constructed to determine the (independent) contribution of the liver to plasma sE-selectin levels at the population level, that is, in the CODAM study (n = 571). Plasma

| NAFLD susceptibility genes co-segregate with plasma sE-selectin levels in the CODAM and Hoorn studies
To further examine the relationship between NAFLD and plasma sEselectin levels, independent of potential confounders, we studied F I G U R E 2 Relationship between E-selectin mRNA expression and plasma sE-selectin levels. Association between nonalcoholic fatty liver disease (NAFLD) severity (according to NAFLD activity score) and plasma sE-selectin levels (A; n = 58) and sVCAM levels (B; n = 58) in severely obese individuals; E-selectin mRNA expression in liver (C; n = 62), visceral adipose tissue (VAT) (E; n = 62) and muscle (F; n = 43) in relation to plasma sE-selectin levels. Hepatic VCAM mRNA expression in relation to plasma sVCAM levels (D; n = 62). Data are expressed as mean (adjusted for age and sex) ± SD, analysed with one-way ANOVA (A and B), or as individual data points (adjusted for age and sex), analysed with linear regression (C-F) the association of two NAFLD susceptibility genes, that is, PNPLA3 (rs738409) and GCKR (rs1260326), with plasma sE-selectin levels in the combined CODAM and Hoorn studies (n = 1265, see Table 1 for characteristics). Previous studies have shown that both gene variants predispose to NAFLD (including NASH) and elevated liver enzymes, [26][27][28] but have opposing effects on factors that are associated with systemic endothelial activation, that is, plasma lipids, 24,29 T2DM 30 and coronary artery disease risk 31,32 (Table S3) Table 3). Additional adjustment for glucose metabolism state (ie NGM, IGM, and T2DM) did not affect the outcomes (data not shown).  Previous experiments have shown that E-selectin is actively involved in the development of hepatic inflammation in mice exposed to chronic plus binge ethanol feeding. 9 In the present study, we noted The association of NAFLD with sE-selectin, independent of factors that are associated with systemic endothelial activation, was also demonstrated by the use of common NAFLD susceptibility genes, that is, PNPLA3 and GCKR. 27 Since individuals are 'randomized' at conception to receive an allele that either predisposes to or protects from NAFLD, these gene variants can serve as instruments to make causal inferences about the relationship between NAFLD and sE-selectin levels. Although pleiotropic effects have been reported for both PNPLA3 and GCKR -which may violate this 'Mendelian randomization' assumption 40 ; it should be noted that these variants have opposing pleiotropic effects on plasma lipids, 24,29 and risk of type 2 diabetes and coronary artery disease, [30][31][32] as summarized in Table S3. As a combination, they therefore serve as a model to disentangle NAFLD from factors that affect systemic endothelial function, despite the presence of pleiotropy. Again, we observed that sE-selectin levels co-segregated with plasma ALT levels, not with plasma triglycerides. These associations were not found for sVCAM.

| D ISCUSS I ON
This study has several strengths and limitations. A major strength is the adoption of different methodologies, that is, animal experiments, liver biopsy studies and cohort studies (with both classical biomarkers and the use of distinctive genetic markers), that enabled us to demonstrate that the liver, more specifically NAFLD, is closely linked to circulating sE-selectin levels. The observational nature of our studies does, however, not allow to exactly determine if and to what extent sE-selectin is derived from the liver. Other organs besides the inflamed liver can contribute to circulating sE-selectin as well. In the present study, plasma ALT was used as a liver-specific biomarker of parenchymal damage. Previous studies have shown that plasma ALT is not a perfect biomarker of NAFLD severity, although simple steatosis and NASH have been associated with greater plasma ALT levels. 41,42 It is therefore likely that there is a residual 'NAFLD effect' on sE-selectin levels when plasma ALT is entered as an independent variable in the regression model. This phenomenon could explain the significant, independent association of BMI -a strong determinant of NAFLD 43 -with sE-selectin levels in our study.
In conclusion, the present study shows that higher expression of hepatic E-selectin and higher levels of plasma sE-selectin are associated with NAFLD and related markers. These findings favour further study to elucidate the role of E-selectin in the pathogenesis of NAFLD and the applicability of sE-selectin as a plasma biomarker of NAFLD/NASH.

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