Advanced glycation end products (AGEs) accumulate in fatty livers and may contribute to low-grade inflammation (LGI), potentially via their receptor, RAGE. It is unknown if the AGE accumulation in fatty livers results in elevated circulating AGEs. In a cohort study, we investigated the association of liver fat and hepatocellular damage with circulating AGEs and soluble RAGE (sRAGE) and subsequently the association of circulating AGEs and sRAGE with LGI. Cross-sectional associations of liver fat percentage (eLF%; ln-transformed) and liver enzymes (LE score; standardized) with circulating AGEs (free CML, CEL, and MG-H1 in nM and protein-bound CML, CEL, and pentosidine in nmol/mmol lysine; ln-transformed) and sRAGE (pg/ml, ln-transformed) and additionally of AGEs and sRAGE with LGI (standardized) were determined by multiple linear regression. eLF% was positively associated with circulating free CEL (
Nonalcoholic fatty liver disease (NAFLD) is a spectrum of liver abnormalities ranging from steatosis (fatty liver) to nonalcoholic steatohepatitis (NASH), fibrosis, and potentially even cirrhosis. NASH is characterized by both steatosis and inflammation, of which the latter causes hepatocellular injury and over time irreversible liver injury [
Fat accumulation in the liver, i.e., steatosis, can cause oxidative stress, increased lipid peroxidation, and release of inflammatory cytokines [
We previously showed that CML accumulates in human fatty livers. This accumulation was more pronounced in livers with more severe steatosis and inflammation [
Cohort on Diabetes and Atherosclerosis Maastricht (CODAM) is a cohort study of 574 subjects, selected based on a moderately increased risk of cardiometabolic disease from a large cohort in the general population as previously described [
Liver fat percentage (eLF%) was estimated using a magnetic resonance spectroscopy- (MRS-) validated equation developed by Kotronen et al. [
ASAT, ALAT, and GGT were measured in EDTA plasma as previously described [
Free CML, CEL, and MG-H1 and PB-CML, PB-CEL, and protein-bound pentosidine (PB-pentosidine) were analysed in EDTA plasma by ultraperformance liquid chromatography tandem mass spectrometry (UPLC MS/MS). Details of the measurement of plasma AGEs have been previously described [
sRAGE (esRAGE and shed forms of RAGE) levels were determined in EDTA plasma using a human RAGE quantikine ELISA (R&D Systems) according to the manufacturer’s instructions. Intra- and interassay coefficients of variation were 7.6 and 2.6%, respectively.
Body mass index (BMI), smoking status, mean alcohol consumption, medication use, prior CVD, and glomerular filtration rate (eGFR; estimated using the short Modification of Diet in Renal Disease equation) were determined as previously reported [
Study population characteristics were compared across tertiles of liver fat percentage using standard one-way ANOVA with a Bonferroni post hoc test for normally distributed variables, Kruskal-Wallis one-way ANOVA with a Bonferroni post hoc test for non-normally distributed variables, and a Chi-square test with a Bonferroni correction for categorical variables. A combined liver enzyme (LE) or low-grade inflammation (LGI) score was calculated as previously described [
The cross-sectional associations of liver fat and liver enzymes (main independent variables) with AGEs and sRAGE (outcomes) and additionally of AGEs and sRAGE (main independent variables) with LGI (outcome) were examined with the use of multiple linear regression analyses with adjustments for the following potential confounders: age (years), sex (male/female), alcohol consumption (g/day), current smoker (Y/N), prevalent CVD (Y/N), use of medication (glucose-lowering (Y/N), lipid-modifying (Y/N), and antihypertensive (Y/N)), eGFR (ml/min/1.73 m2), and BMI (kg/m2). All linear regression analyses were performed with four different models as follows: model 1: crude model; model 2: adjustment for age and sex; model 3: adjustment for age, sex, alcohol, smoking, CVD, medication, and eGFR; model 4: adjustment for age, sex, alcohol, smoking, CVD, medication, eGFR, and BMI. All statistical analyses were performed using the SPSS for Windows, version 25.0 (IBM Corp.), and all data were considered statistically significant at
In Table
General characteristics of the study population (
All participants | Liver fat according to tertiles of eLF% | ||||
---|---|---|---|---|---|
( |
Lowest ( |
Middle ( |
Highest ( | ||
Age (years) | 0.057 | ||||
Male sex (%) | 62.4 | 54.2 | 60.9 | 72 |
0.003 |
NGM/IGM/T2DM (%) | 53.5/22.6/24.0 | 83.3/15.5/1.2 | 56.8/30.2/13.0 |
20.2/22.0/57.7 |
<0.001 |
CVD (%) | 26.9 | 19.6 | 23.7 | 37.5 |
0.001 |
Current smoker (%) | 20.8 | 22.6 | 21.9 | 17.9 | 0.853 |
Alcohol (g/day) | 8.6 (1.3–22.8) | 8.5 (2.4–22.7) | 9.9 (1.4–24.2) | 7.1 (0.5–21.2) | 0.311 |
Antihypertensive medication (%) | 37.4 | 23.2 | 35.5 |
53.6 |
<0.001 |
Lipid-modifying medication (%) | 18.2 | 12.5 | 16.6 | 25.6 |
0.006 |
Glucose-lowering medication (%) | 12.1 | 0.6 | 5.3 |
30.4 |
<0.001 |
Body mass index (kg/m²) | <0.001 | ||||
Waist circumference (cm) | <0.001 | ||||
Fasting glucose (mM) | 5.58 (5.20–6.34) | 5.20 (4.94–5.44) | 5.57 (5.22–6.00) |
6.57 (5.84–7.62) |
<0.001 |
Triglycerides (mM) | 1.4 (1.0–1.9) | 1.0 (0.8–1.4) | 1.5 (1.1–2.0) |
1.8 (1.4–2.2) |
<0.001 |
Cholesterol (mM) | 0.197 | ||||
LDL cholesterol (mM) | 0.120 | ||||
HDL cholesterol (mM) | <0.001 | ||||
HbA1c (%) | 5.80 (5.50–6.20) | 5.60 (5.30–5.80) | 5.80 (5.50–6.10) |
6.20 (5.80–6.90) |
<0.001 |
HOMA2-IR | 1.58 (1.09–2.51) | 1.02 (0.80–1.26) | 1.56 (1.21–2.00) |
3.21 (2.37–4.55) |
<0.001 |
IL-6 (pg/ml) | 1.56 (1.12–2.28) | 1.25 (0.96–2.18) | 1.60 (1.17–2.20) |
1.80 (1.33–2.57) |
<0.001 |
IL-8 (pg/ml) | 4.34 (3.58–5.53) | 4.18 (3.44–5.23) | 4.30 (3.50–5.13) | 4.78 (3.83–6.30) |
<0.001 |
TNF- |
6.25 (5.23–7.61) | 5.94 (5.02–7.01) | 6.34 (5.30–7.51) | 6.60 (5.43–7.96) |
0.002 |
CRP (mg/l) | 2.04 (0.92–3.97) | 1.07 (0.59–2.72) | 2.20 (1.20–4.33) |
2.71 (1.46–5.00) |
<0.001 |
SAA (mg/l) | 1.42 (0.98–2.27) | 1.20 (0.87–2.16) | 1.52 (1.02–2.41) |
1.52 (1.05–2.33) |
0.004 |
sICAM-1 (ng/ml) | 212.5 (186.8–242.7) | 195.4 (177.8–221.5) | 211.0 (189.1–236.5) |
231.8 (204.9–257.8) |
<0.001 |
Low-grade inflammation score | <0.001 | ||||
PB-pentosidine (nmol/mmol lysine) | 0.43 (0.36–0.53) | 0.46 (0.39–0.55) | 0.44 (0.36–0.52) | 0.41 (0.35–0.50) |
0.015 |
PB-CML (nmol/mmol lysine) | 34.6 (29.6–41.0) | 37.1 (32.3–44.9) | 35.2 (31.1–40.6) | 31.1 (26.4–38.0) |
<0.001 |
PB-CEL (nmol/mmol lysine) | 23.3 (19.0–29.2) | 22.8 (19.5–26.9) | 24.2 (19.0–30.3) | 22.9 (18.6–29.5) | 0.250 |
Free CML (nM) | 79.5 (61.2–98.6) | 76.0 (60.3–92.9) | 80.4 (61.0–100.1) | 82.2 (64.1–102.5) | 0.144 |
Free CEL (nM) | 45.5 (37.0–58.0) | 42.7 (34.8–52.4) | 45.9 (38.7–56.7) | 51.0 (38.2–63.0) |
<0.001 |
Free MG-H1 (nM) | 123.8 (87.5–176.6) | 127.2 (90.8–168.1) | 121.0 (85.6–172.5) | 119.6 (85.0–198.3) | 0.810 |
sRAGE (pg/ml) | 1250 (893–1604) | 1402 (1112–1756) | 1229 (838–1567) |
1155 (850–1440) |
<0.001 |
ALAT (U/l) | 22.2 (17.2–27.9) | 17.1 (14.3–21.2) | 22.4 (18.0–26.7) |
28.6 (23.2–36.3) |
<0.001 |
ASAT (U/l) | 19.8 (16.5–24.2) | 18.2 (14.7–21.4) | 19.3 (16.6–23.3) |
22.7 (19.0–27.6) |
<0.001 |
GGT (U/l) | 24.0 (17.0–37.0) | 18.0 (13.0–23.8) | 26.0 (18.0–37.5) |
34.0 (24.0–48.8) |
<0.001 |
Liver enzyme score | <0.001 | ||||
eLF% (%) | 4.79 (2.35–8.62) | 2.11 (1.79–2.35) | 4.79 (3.80–5.89) |
10.64 (8.60–14.53) |
<0.001 |
FLI | <0.001 | ||||
eGFR (ml/min/1.73 m2) | 0.028 |
Data are expressed as mean ± SD, median (interquartile range) or percentages. The minimum and maximum of eLF% tertiles were (0.85–2.91), (2.92–6.97), and (6.98–36.65) %, respectively. eLF%: estimated liver fat %; NGM: normal glucose metabolism; IGM: impaired glucose metabolism; T2DM: type 2 diabetes mellitus; CVD: cardiovascular disease; LDL: low-density lipoprotein; HDL: high-density lipoprotein; HbA1c: hemoglobin A1c; HOMA2-IR: homeostasis model assessment insulin resistance; IL; interleukin; TNF-
In linear regression analyses, eLF% was positively associated with free CML (
Cross-sectional associations of liver fat and liver enzymes with free and protein-bound AGEs and sRAGE.
Outcome | Model | eLF% (%) | Liver enzyme score | ||||||
---|---|---|---|---|---|---|---|---|---|
95% CI | 95% CI | ||||||||
Free CML (nM) | 1 | 0.042 | 0.002 | 0.082 | 0.030 | 0.000 | 0.060 | 0.052 | |
2 | 0.031 | -0.008 | 0.070 | 0.122 | 0.030 | -0.001 | 0.061 | 0.055 | |
3 | 0.041 | -0.002 | 0.084 | 0.061 | 0.046 | 0.015 | 0.077 | ||
4 | 0.037 | -0.014 | 0.088 | 0.151 | 0.044 | 0.012 | 0.076 | ||
Free CEL (nM) | 1 | 0.096 | 0.058 | 0.135 | 0.047 | 0.018 | 0.077 | ||
2 | 0.083 | 0.045 | 0.121 | 0.037 | 0.007 | 0.067 | |||
3 | 0.095 | 0.054 | 0.136 | 0.049 | 0.019 | 0.079 | |||
4 | 0.090 | 0.041 | 0.139 | 0.040 | 0.009 | 0.072 | |||
Free MG-H1 (nM) | 1 | 0.043 | -0.019 | 0.105 | 0.170 | -0.009 | -0.056 | 0.038 | 0.719 |
2 | 0.023 | -0.038 | 0.084 | 0.464 | -0.019 | -0.067 | 0.028 | 0.423 | |
3 | 0.024 | -0.042 | 0.090 | 0.481 | 0.006 | -0.042 | 0.053 | 0.820 | |
4 | 0.036 | -0.042 | 0.115 | 0.367 | 0.007 | -0.043 | 0.057 | 0.789 | |
PB-pentosidine (nmol/mmol lysine) | 1 | -0.059 | -0.100 | -0.018 | -0.015 | -0.047 | 0.016 | 0.342 | |
2 | -0.076 | -0.117 | -0.035 | -0.028 | -0.060 | 0.004 | 0.088 | ||
3 | -0.102 | -0.146 | -0.058 | -0.032 | -0.064 | 0.001 | 0.055 | ||
4 | -0.051 | -0.102 | 0.001 | 0.055 | -0.007 | -0.040 | 0.026 | 0.676 | |
PB-CML (nmol/mmol lysine) | 1 | -0.115 | -0.143 | -0.087 | -0.061 | -0.083 | -0.039 | ||
2 | -0.122 | -0.150 | -0.093 | -0.067 | -0.090 | -0.044 | |||
3 | -0.105 | -0.136 | -0.074 | -0.054 | -0.077 | -0.031 | |||
4 | -0.071 | -0.108 | -0.034 | -0.037 | -0.060 | -0.013 | |||
PB-CEL (nmol/mmol lysine) | 1 | 0.000 | -0.036 | 0.036 | 0.992 | 0.003 | -0.025 | 0.031 | 0.828 |
2 | -0.002 | -0.039 | 0.035 | 0.911 | -0.002 | -0.030 | 0.027 | 0.919 | |
3 | 0.013 | -0.028 | 0.055 | 0.527 | 0.007 | -0.023 | 0.037 | 0.652 | |
4 | -0.003 | -0.052 | 0.047 | 0.920 | 0.001 | -0.030 | 0.033 | 0.939 | |
sRAGE (pg/ml) | 1 | -0.097 | -0.146 | -0.048 | -0.072 | -0.109 | -0.034 | ||
2 | -0.076 | -0.125 | -0.028 | -0.040 | -0.078 | -0.002 | |||
3 | -0.076 | -0.130 | -0.023 | -0.027 | -0.067 | 0.012 | 0.169 | ||
4 | -0.039 | -0.102 | 0.025 | 0.232 | -0.010 | -0.050 | 0.031 | 0.642 |
Next, we investigated the association of the LE score (as a measure of hepatocellular injury) with AGEs and sRAGE (Table
In our cohort, we next investigated the associations between liver fat or liver enzymes and the LGI score (Table
Associations of liver fat and liver enzymes with low-grade inflammation.
Outcome | Model | eLF% (%) | Liver enzyme score | ||||||
---|---|---|---|---|---|---|---|---|---|
95% CI | 95% CI | ||||||||
LGI | 1 | 0.494 | 0.387 | 0.601 | 0.310 | 0.226 | 0.393 | ||
2 | 0.493 | 0.386 | 0.599 | 0.347 | 0.262 | 0.431 | |||
3 | 0.533 | 0.415 | 0.650 | 0.356 | 0.270 | 0.442 | |||
4 | 0.429 | 0.290 | 0.568 | 0.292 | 0.205 | 0.380 |
AGEs can trigger inflammation via RAGE, potentially contributing to LGI and could thereby explain, at least in part, the association of liver fat and liver enzymes with LGI. Therefore, we investigated the association of AGEs and sRAGE with the LGI score. Free CML (
Cross-sectional associations of free and protein-bound AGEs and sRAGE with low-grade inflammation.
Outcome | Model | Free CML (nM) | Free CEL (nM) | Free MG-H1 (nM) | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
95% CI | 95% CI | 95% CI | |||||||||||
LGI | 1 | 0.443 | 0.194 | 0.692 | 0.359 | 0.103 | 0.616 | 0.204 | 0.014 | 0.041 | 0.366 | ||
2 | 0.322 | 0.067 | 0.577 | 0.282 | 0.023 | 0.541 | 0.128 | -0.037 | 0.293 | 0.127 | |||
3 | 0.344 | 0.086 | 0.603 | 0.259 | -0.006 | 0.525 | 0.056 | 0.128 | -0.041 | 0.297 | 0.137 | ||
4 | 0.297 | 0.049 | 0.545 | 0.152 | -0.104 | 0.409 | 0.243 | 0.131 | -0.031 | 0.293 | 0.112 |
Outcome | Model | PB-CML (nmol/mmol lysine) | PB-CEL (nmol/mmol lysine) | PB-pentosidine (nmol/mmol lysine) | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
95% CI | 95% CI | 95% CI | |||||||||||
LGI | 1 | -0.631 | -0.965 | -0.298 | -0.047 | -0.325 | 0.231 | 0.741 | -0.059 | -0.303 | 0.185 | 0.637 | |
2 | -0.704 | -1.032 | -0.377 | -0.028 | -0.302 | 0.246 | 0.840 | -0.154 | -0.398 | 0.091 | 0.219 | ||
3 | -0.547 | -0.888 | -0.207 | -0.024 | -0.293 | 0.245 | 0.863 | -0.089 | -0.338 | 0.160 | 0.483 | ||
4 | -0.251 | -0.592 | 0.090 | 0.149 | -0.078 | -0.335 | 0.180 | 0.553 | 0.120 | -0.125 | 0.365 | 0.337 |
Outcome | Model | sRAGE (pg/ml) | |||
---|---|---|---|---|---|
95% CI | |||||
LGI | 1 | 0.048 | -0.154 | 0.251 | 0.641 |
2 | 0.019 | -0.189 | 0.226 | 0.858 | |
3 | 0.037 | -0.170 | 0.244 | 0.725 | |
4 | 0.143 | -0.057 | 0.342 | 0.161 |
The present study revealed that liver fat and liver enzymes were positively associated with circulating free AGEs, while inverse associations were observed with circulating protein-bound AGEs. In addition, free AGEs were positively associated with LGI, but an inverse association was determined between protein-bound AGEs and LGI, albeit this latter association was explained by BMI. Lastly, there was no association observed of liver fat and liver enzymes with sRAGE or between sRAGE and LGI after adjustment for BMI.
We demonstrated that a higher hepatic fat content, a hallmark of NAFLD, is associated with AGEs and sRAGE, but a key role is played by obesity and adipose tissue as adjustment for BMI attenuated these associations. However, even after adjustment for this potential confounder, a positive association between liver fat and free AGEs (only CEL and not CML) and an inverse association between liver fat and protein-bound AGEs (pentosidine and PB-CML) was still observed. Lipid peroxidation contributes to CEL formation making its generation likely in metabolically active and lipid-rich environments such as the fatty liver [
The inverse relationship between hepatic steatosis and circulating protein-bound AGEs might suggest either trapping of AGEs in the liver or an elevated breakdown of (modified) proteins in fatty livers. We previously reported trapping of PB-CML in adipose tissue of obese mice and reduced levels of circulating PB-CML in obese subjects [
We found a strong inverse association between liver fat and sRAGE, but this association was largely dependent on BMI. Several studies have shown that obesity is a strong determinant of sRAGE levels explaining why the association is no longer present after adjustment for this confounder [
Our study implies that hepatic injury, represented by liver enzyme levels, is associated with higher levels of circulating free AGEs (CML and CEL), but lower levels of protein-bound AGEs (PB-CML). These findings are in agreement with our previous study revealing elevated levels of CML in livers with more severe steatosis and inflammation, which causes liver injury [
The association of hepatocellular injury with elevated circulating free CML, but lower circulating PB-CML, suggests enhanced breakdown of proteins by liver injury and the accompanying hepatic inflammation. As previously described, protein degradation is enhanced in NAFLD patients [
Our results also show an association of liver fat and liver injury with LGI corroborating that liver disease might contribute to elevated inflammatory cytokines and thereby maintenance of low-grade inflammation [
The observed positive association of free CML and CEL with LGI suggests that these plasma-free AGEs might influence LGI. However, the association of CEL with LGI disappeared after adjustment for BMI implying that this association depends more on adiposity and the adipose tissue inflammation that accompanies it. Free CML remained associated with LGI even after adjustment for BMI. However, the very modest effect of additional adjustment for free CML suggested that the strong association of hepatic steatosis and injury with LGI was not mediated by circulating free CML levels. Indeed, it has been described that free CML is unable to stimulate RAGE and cause inflammation [
In line with this notion, we observed that sRAGE was not associated with LGI despite its potential role as a decoy receptor for its ligands or as indicator of AGE-RAGE axis activity [
Of note, AGEs can form exogenously and are present in many food products [
Our study has some limitations. Given our cross-sectional study design, we cannot draw causal conclusions on the investigated relationships. Prospective studies are required to better understand the contribution of liver disease to AGE formation and subsequently circulating AGEs and low-grade inflammation. Another important consideration is the estimation of liver fat employed in the study. We used two different equations to estimate liver fat content but did not measure liver fat content using imaging techniques or quantify hepatic steatosis or inflammation in liver biopsies. These procedures were not feasible in our large human cohort due to cost or ethical concerns. However, the two measures of liver fat (eLF% and FLI) used in this study have been validated in previous studies and revealed similar associations in the present study [
Our findings suggest an elevated formation of AGEs in fatty and injured livers accompanied with an enhanced breakdown of the formed protein-bound AGEs resulting in altered levels of circulating AGEs. Potentially, circulating AGEs can contribute to LGI.
Advanced glycation end product
N
N
Cardiovascular disease
Low-grade inflammation
N
Nonalcoholic fatty liver disease
Nonalcoholic steatohepatitis
Receptor for AGEs
Tumour necrosis factor
Ultraperformance liquid chromatography tandem mass spectrometry.
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.
This work is based on a chapter in the dissertation of Mitchell Bijnen.
The authors declare that there is no conflict of interest regarding the publication of this paper.
MB analysed the data and wrote the manuscript. MMJvG collected and analysed data and provided statistical advice. CJHvdK collected and analysed data. JLS designed methods and performed measurements. MPHvdW performed measurements. CDAS supervised analyses and revised the manuscript. KW designed the study, supervised data analyses, and revised the manuscript. CGS designed the study, supervised data analyses, and revised the manuscript.
The authors thank Ying Xin and Nynke Simons for providing statistical advice. This study was financed by the Netherlands Heart Foundation (2013T143) grant to Kristiaan Wouters. Initiation of CODAM was supported by grants of the Netherlands Organisation for Scientific Research (940-35-034) and the Dutch Diabetes Research Foundation (98.901).
Suppl. Table 1: associations of FLI with free and protein-bound AGEs and sRAGE.