MDA epitopes are present in livers of patients with NASH
To determine whether OSE are generated during the spectrum of NAFLD, immunohistochemical
staining for MDA epitopes with the monoclonal antibody MDA2 was performed on liver
biopsies of patients with biopsy-proven NASH. As illustrated in Fig. 1a–c, small clusters of MDA2-positive staining were detected between swollen hepatocytes
of liver biopsies from patients with NASH, indicating the presence of MDA epitopes.
Fig. 1. Malondialdehyde (MDA) epitopes in livers of patients with non-alcoholic steatohepatitis (NASH) and relationship between plasma immunoglobulin M (IgM) titers and non-alcoholic fatty liver disease (NAFLD). Immunohistochemical detection of MDA epitopes in liver sections (magnification
20×) of a a negative control and b two patients with NASH – arrows indicate localization of MDA2-positive staining in
the liver tissue. Plasma IgM antibody titers against c MDA-low-density lipoprotein (LDL), d malondialdehyde-acetaldehyde (MAA)-LDL, e P1 mimotope, f CuOx-LDL, g PC-BSA, and h total IgM levels in patients with NAFLD and healthy controls. i Ratio of anti-P1 IgM titers to total IgM levels in patients with NAFLD and controls.
Data are expressed in log10 relative light units (RLU)/100 ms. *p??0.05, **p??0.01, ***p??0.001
IgM antibody titers towards different OSE are lower in patients with NAFLD
To investigate the potential relationship between antibodies specifically targeting
epitopes of OxLDL and NAFLD, IgM and IgG antibody titers against MDA, MAA (an immunodominant
advanced MDA-lysine adduct), P1 (a highly specific peptide mimotope of MAA 10]), Cu-OxLDL, and PC-BSA were measured in the plasma of patients with NAFLD (?=?71) and control participants with confirmed absence of steatosis (?=?68). The values of the different measurements were tested for normality with Shapiro–Wilks
test and the results indicated that data were not normally distributed for most of
the measurements in either of the groups (Additional file 4: Table S1). Therefore, all data were logarithmically transformed (log10) and represented
using a log10 scale in Fig. 1. The observed relative lights units can be found in Additional file 4: Table S2.
In line with our expectations, patients with NAFLD had significantly lower IgM titers
than controls against MDA-LDL (4.755 versus 4.819 log10 RLU/100 ms, p?=?0.0019; Fig. 1c, Additional file 4: Table S2), MAA-LDL (4.801 versus 4.868 log10 RLU/100 ms, p?=?0.0026; Fig. 1d, Additional file 4: Table S2), P1 (3.877 versus 4.08 log10 RLU/100 ms, p??0.0001; Fig. 1e, Additional file 4: Table S2), CuOx-LDL (4.46 versus 4.612 log10 RLU/100 ms, p?=?0.0004; Fig. 1f, Additional file 4: Table S2), and PC-BSA (4.255 versus 4.383 log10 RLU/100 ms, p?=?0.0096; Fig. 1g, Additional file 4: Table S2). No differences were found in specific IgG antibody titers towards the
different epitopes of OxLDL in patients with NAFLD compared to controls (Additional
file 4: Table S3).
Interestingly, total IgM antibodies were also significantly lower in patients with
NAFLD compared to healthy participants (3.773 versus 3.712 log10 RLU/100 ms, p?=?0.0044; Fig. 1h, Additional file 4: Table S2). Notably, IgM titers towards the specific P1 mimotope, which showed the
strongest reduction in patients with NAFLD compared to controls, remained significant
after adjusting for total IgM levels (0.3074 versus 0.1658 (ratio), p?=?0.0002, Fig. 1i). These data point towards a negative association between IgM antibodies and oxidized
lipids and fatty liver disease, and indicate an important role for lipid oxidation
in NAFLD.
Plasma anti-P1 IgM levels are reduced in the early phase of NAFLD and are not influenced
by inflammation or fibrosis
Because anti-P1 IgM levels showed the strongest reduction in patients with NAFLD and
due to its high clinical reproducibility, further analyses were conducted using the
anti-P1 IgM measurements. To determine in which phase in the disease spectrum of NAFLD
that plasma IgM titers towards P1 are reduced, participants were divided according
to their ALT levels (25, 25–50, 50 IU/L) because a normal ALT level is expected
to be 50 IU/L. Interestingly, P1-IgM levels were reduced in participants with moderately
elevated ALT levels (25–50 IU/L) or high ALT levels (50 IU/L) when compared with
the P1-IgM levels of participants with ALT levels 25 IU/L (Fig. 2a). Moreover, more than 90 % of control participants were classified as having low
ALT levels. These data suggest that IgM titers are already lower in early phases of
NAFLD, before liver damage is detected based on ALT levels.
Fig. 2. Relationship between plasma P1-immunoglobulin (IgM) levels and non-alcoholic fatty liver disease (NAFLD) upon further characterization of liver biopsies. a IgM titers towards P1 in controls and patients with NAFLD grouped based on their
ALT levels. b Plasma IgM antibody titers against the P1 mimotope in patients with NAFLD in which
the liver biopsy was evaluated and scored as simple steatosis versus non-alcoholic
steatohepatitis (NASH). Anti-P1 IgM titers after scoring liver biopsies for c steatosis, d inflammation, e hepatocyte ballooning, f applying the NAFLD activity score (NAS), and g scoring fibrosis. Data are expressed in relative light units (RLU)/100 ms and presented as means. **p??0.01
Next, we evaluated the relationship between IgM to P1 and different classifications
based on histological scoring of liver biopsies taken from patients with NAFLD. First,
the potential of anti-P1 IgM to distinguish simple steatosis from NASH was tested.
Whereas IgM titers were lower in NAFLD compared with controls, no difference was observed
in specific IgM levels between patients with steatosis and NASH (Fig. 2b). Moreover, no differences were observed in plasma P1-IgM titers after classifying
the patients according to a score for steatosis (Fig. 2c), inflammation (Fig. 2d), hepatocyte ballooning (Fig. 2e), the NAFLD activity score (NAS) (Fig. 2f), or for fibrosis (Fig. 2g). These data further suggest that antibodies recognizing oxidized lipids are involved
during early processes in the development of NAFLD.
Low IgM titers against the P1 mimotope have a high predictive value for the presence
of NAFLD
To determine the power of low plasma IgM titers against the P1 mimotope to predict
the presence of fatty liver disease, age, gender, BMI, total IgM levels, and anti-P1
IgM levels were combined in a single logistic regression model. To account for skewness
in their distribution, base-2 logarithms of P1-specific IgM levels were used in the
model. Thus, adjusted odds ratios (OR) for this variable reflect the change in odds
for an increase of one log2 (the equivalent of a doubling of the value) in the measure.
Without adjustment for other risk factors, age (OR?=?1.206; confidence interval?=?1.039–1.4),
BMI (OR?=?1.353; confidence interval?=?1.228–1.491), total IgM (OR?=?0.619; confidence
interval?=?0.431–0.888), and P1-IgM (OR?=?0.428; confidence interval?=?0.282–0.65)
were all significant predictors for the presence of NAFLD, and female gender was not
(Additional file 4: Table S4). Importantly, P1-IgM levels (OR?=?0.419; confidence interval?=?0.216–0.813)
were still a significant predictor for the presence of NAFLD after adjustment for
age, gender, BMI, and total IgM (Fig. 3a, Additional file 4: Table S5). Moreover, adjusting for waist circumference instead of BMI in our model
did not change the significance for anti-P1 IgM levels (OR?=?0.432; confidence interval?=?0.261–0.813)
in predicting the presence of NAFLD (Additional file 4: Table S6 and Additional file 5: Figure S1). In line with our expectations, BMI (OR?=?1.343; confidence interval?=?1.211–1.49)
and waist circumference (OR?=?1.666; confidence interval?=?1.354–2.049) were still
significant predictors for NAFLD after adjustment, whereas other variables lost their
significance (Additional file 4: Tables S5 and S6).
Fig. 3. Single logistic regression model, odds ratio determination, and receiver operating
characteristic (ROC) curve analysis for P1-specific IgM levels in non-alcoholic fatty liver disease (NAFLD). a Odds ratios for the predictive power of anti-P1 immunoglobulin (IgM) titers (log2 scale) for fatty liver disease after adjustment for age, gender, body
mass index (BMI), and total IgM. b ROC curve analysis for the diagnosis of NAFLD with P1-IgM. c IgM titers towards P1 were divided into quartiles (black: controls; red: NAFLD) and d the odds ratios P1-IgM (log2 scale) for NAFLD were computed across these quartiles
after adjusting for age, gender, BMI, and total IgM levels. CI confidence interval
A receiver operating characteristic (ROC) curve analysis was performed to assess the
diagnostic value of anti-P1 IgM titers for NAFLD. An area under the curve (AUC) of
0.703 (confidence interval: 0.617–0.790) for the diagnosis of NAFLD using IgM titers
towards P1 was found (Fig. 3b). In addition, the measurements of IgM against P1 were divided into quartiles and
the OR for NAFLD were computed across these quartiles after adjusting for age, gender,
BMI, and total IgM levels (Fig. 3c, Table 3). Our results indicate that participants in quartile 1 (lowest P1-IgM levels) had
an odds ratio of 5.8 (p?=?0.01, confidence interval: 1.209–27.818), indicating an increased likelihood of NAFLD
with decreasing P1-IgM levels, independent of age, gender, BMI, and total IgM levels
(Fig. 3d, Table 3). Taken together, these data indicate that P1-specific IgM titers have a high predictive
value for the presence of NAFLD, independent of total IgM levels.
Table 3. Odds ratios of P1-specific immunoglobulin M levels for non-alcoholic fatty liver disease
Anti-P1 IgM titers are not correlated with lipid levels and negatively correlate with
markers for liver damage and systemic markers of inflammation
First, the correlation between P1-IgM and plasma lipid levels in all participants
was tested. No correlation was found between IgM levels towards P1 and plasma triglycerides
(Pearson R: ?0.11, Fig. 4a), cholesterol (Pearson R: ?0.09, Fig. 4b), or LDL (Pearson R: ?0.03, Fig. 4c), well-known indicators for steatosis and risk factors for NAFLD. Next, we assessed
the potential correlation with markers for obesity: BMI, waist circumference, and
leptin concentration. A strong inverse correlation was observed between P1-IgM and
BMI (Pearson R: ?0.28, p?=?0.0009, Fig. 4d), waist circumference (Pearson R: ?0.22, p?=?0.02, Fig. 4e), and leptin concentration (Pearson R: ?0.32, p?=?0.0026, Fig. 4f). These data are in line with a previously reported study showing elevated serum
leptin levels during NAFLD 16]. Moreover, obesity, which is a known risk factor for NAFLD, has been associated with
increased leptin levels 17]. Additionally, we tested the correlation between P1-IgM and circulatory markers for
liver damage, inflammation, and adipokines. Interestingly, a strong inverse correlation
was found between P1-IgM levels and monocyte chemoattractant protein-1 (MCP-1; Pearson
R: ?0.39, p?=?0.0006, Fig. 4g), IL-6 (Pearson R: ?0.24, p?=?0.041, Fig. 4h), and cathepsin D (Pearson R ?0.25, p?=?0.0033, Fig. 4I), which we recently validated as a marker for pediatric NASH. Moreover, anti-P1-IgM
titers were inversely correlated with alanine aminotransferase (ALT; Pearson R: ?0.17,
p?=?0.047, Additional file 6: Figure S2A) and aspartate aminotransferase (AST; Pearson R: ?0.18, p?=?0.037, Additional file 6: Figure S2B) and a trend was observed with gamma-glutamyltransferase (?GT; Pearson
R: ?0.11, Additional file 6: Figure S2C). These data indicate that increased liver damage, assessed by AST and
ALT levels, is associated with reduced levels of IgM antibody titers towards OSE.
No correlation was found with IL-1? (Pearson R: 0.16, Additional file 6: Figure S2D) or IL-8 (Pearson R: ?0.10, Additional file 6: Figure S2E). Taken together, these data suggest that immune recognition of oxidized
lipids is affected during obesity and plays an important role during related diseases,
such as the early stages of hepatic inflammation in NAFLD. These results also suggest
an indicative role of low P1-IgM levels in regard to systemic markers of inflammation
and liver damage.
Fig. 4. Correlation between plasma P1-immunoglobulin M (IgM) levels and plasma lipids and systemic markers of inflammation. Shown are the Pearson
R correlation between plasma IgM titers towards P1 and a plasma triglycerides (TG), b cholesterol, low-density lipoprotein (LDL), d body mass index (BMI), e waist circumference, f leptin, and the markers for inflammation g monocyte chemoattractant protein 1 (MCP1), h interleukin 6 (IL-6), and i cathepsin D. RLU relative light unit
Low OSE-specific IgM levels are not associated with hepatitis C or IBD-dependent inflammation
To determine the association between plasma P1-IgM levels and other chronic inflammatory
diseases, a separate analysis was conducted on two cohorts consisting of patients
with hepatitis C or autoimmune IBD. Total and OSE-specific IgM and IgG antibody titers
were assessed in both cohorts. Total IgM titers were unchanged between patients with
hepatitis C and controls (167 versus 117 mg/dL, Fig. 5a, Additional file 4: Table S7) whereas total IgG levels were higher in patients with hepatitis C than
in controls (1506 versus 1091 mg/dL, Additional file 4: Table S7). Importantly, in contrast to our findings in NAFLD, IgM titers towards
P1 were higher in patients with hepatitis C than in controls (99,785 versus 62,828
RLU/100 ms, p??0.0001, Fig. 5b, Additional file 4: Table S7). Adjusting for total IgM antibody levels resulted in no significant differences
between the two groups (686 versus 644 (ratio), Fig. 5c). OSE-specific IgG levels were higher in patients with hepatitis C than in controls
(Additional file 4: Table S8). These data further suggest that low levels of IgM antibodies targeting
oxidized lipids are specific for obesity-related diseases such as NAFLD.
Fig. 5. Plasma immunoglobulin M (IgM) titers in cohorts consisting of patients with hepatitis C or patients with inflammatory
bowel disease (IBD). a Total, b anti-P1 IgM titers, and c P1/total IgM ratio in healthy controls and patients with hepatitis C. Relationship
between the activity score during IBD and d total IgM, e P1-specific IgM, and f P1/total IgM ratio. Inactive disease was given score 0, mild to moderate disease
activity was given a score of 1–2, and severe activity was given a score of 3. Data
are expressed in mg/dl or relative light units (RLU)/100 ms and presented as means and standard deviations. *p??0.05, ***p??0.001
In the IBD cohort (?=?62; mean male age 49.4 years and female age 54.8 years), 46 had Crohn’s disease (8
ileal, 17 colonic, and 21 ileocolonic), 15 ulcerative colitis (1 procolitis, 10 distal,
and 4 pancolitis), and one patient had indeterminate colitis. IBD is characterized
by fluctuations in the inflammatory state of the gut 18]. The patients were classified during endoscopy according to the activity state of
inflammation at the time of blood sampling with a score ranging from 0 to 3, in which
0 means inactive (?=?23), 1–2 refers to mild to moderate activity (?=?10), and 3 indicates severely active disease (?=?29). In contrast to our measurements in patients with NAFLD, total IgM levels were
not influenced by the inflammatory activity state in patients with IBD (36,025 versus
37,013 versus 40,280 RLU/100 ms, Fig. 5d). Next, IgM antibody titers towards the P1 mimotope were determined in these patients,
reflecting the IgM titers against OSE. In contrast to the decreased levels found in
the plasma of patients with NAFLD, slightly higher P1-specific IgM titers were found
in the plasma of patients classified with high activity IBD (score 3) compared to
patients with inactive disease (score 0) (31,052 versus 21,121 RLU/100 ms, p?=?0.0383, Fig. 5e). We could not detect a difference in P1-specific IgM in patients scored as 0 and
1–2 (21,121 versus 23,060 RLU/100 ms, Fig. 5e), nor between patients scored as 1–2 and 3 (23,060 versus 31,052 RLU/100 ms, Fig. 5e), respectively. Additionally, no difference was observed between the groups after
normalization for total IgM levels (0.628 versus 0.608 versus 0.755 (ratio), Fig. 5f). These data suggest that decreased IgM antibody titers towards MDA-type epitopes
are more specifically associated with lipid-induced inflammation, as present in obesity-related
diseases like atherosclerosis and in the spectrum of NAFLD.