Introduction

Alzheimer’s disease (AD) is the most widespread dementia and afflicts aging populations worldwide. The cause of AD is the result of many intertwining pathologies. However, progressive neuronal loss found in AD is theorized to result from the accumulation of extracellular amyloid plaques comprised of amyloid-ß (Aß) which is generated through proteolytic cleavage of amyloid precursor protein (APP) by ß- and γ-secretases [1]. More and more investigations suggest that AD is strongly correlated with metabolic syndrome (MetS) involving oxidative stress and post-translational modification (PTM) [2,3,4]. MetS is a group of five conditions, namely obesity, hypertension, hyperglycemia, hypertriglyceridemia, and reduced high-density lipoprotein-cholesterol (HDL-C) in the blood. Imbalanced blood levels of glucose and lipids enhance lipid peroxidation, producing acrolein, 4-hydroxynonenal (HNE), and other aldehydic adducts in AD brain, cerebrospinal fluid (CSF), and blood. Nationwide surveys from Taiwan and Korea also supported that metabolic aberration could be a good indicator to diagnose the development of dementia specifically AD [5, 6]. However, the detailed mechanism of pathogenesis from MetS to AD remains elusive.

Studies suggest that increased lipid peroxidation in membranes causes increased levels of acrolein and HNE in the amygdala, hippocampus/parahippocampal gyrus, and inferior parietal lobule [1A, right panel). The bands looked blurry possibly due to long exposure in image development. It suggests that the AD cells may undergo metabolic disturbance with increased acrolein adducts probably formed on Aß peptides.

Fig. 1
figure 1

Acrolein adducts in the transfected cells and model mouse tissues. A Immunoblotting of the cell lysates. The transfected neuro-2a cells were analyzed with SDS-PAGE in triplicate. One gel was stained with CBB as a loading control. The other two gels were immunoblotted with anti-C-terminal APP and acrolein antibodies, respectively. The full-length APP (APP-FL) and APP C-terminal fragments (APP-CTFa and APP-CTFß) were indicated. B Morris Water Maze analyses. The time of the mice (n = 9) spent in the target quadrant at 2 and 9 months old was recorded and compared. C Congo red staining of Aß plaques in the mouse brain. Aß plaques in the cortex and hippocampus of the mice (n = 6) were stained. The plaque burden was quantified and compared. D APP immunoblotting of mouse serum and brain lysates. The IgG-removed serum and brain tissues of the mice at 9 months old were immunoblotted with anti-C-terminal APP antibody, and the gel stained with CBB was used as a loading control. The full-length APP (APP-FL), APP C-terminal fragments (APP-CTFs), and Aß oligomers were indicated. E APP immunoprecipitation and acrolein immunoblotting of mouse serum and brain lysates. The IgG-removed serum and brain tissues of the mice at 9 months old were immunoprecipitated with anti-C-terminal APP antibody and then subjected to immunoblotting with anti-acrolein antibody. A duplicated gel was stained with CBB as a loading control. The prestained protein ladder showed some luminance. AD, Alzheimer’s disease; APP, amyloid precursor protein; CBB, Coomassie brilliant blue; CTF, C-terminal fragment; IP, immunoprecipitation; M, marker; V, vector plasmid p-CAX; WT, wild-type. *** p < 0.001

Full size image

To investigate the acrolein adduction in an animal model, we used 3xTg-AD mice and tested them using Morris water maze, histochemistry, immunoprecipitation, and immunoblotting. The water maze test revealed that 3xTg-AD mice spent a shorter time in the target quadrant than did WT with a statistical significance at 9 months old (p < 0.001) (Fig. 1B). Congo red-stained mouse brain sections showed an increased Aß plaque burden in AD mice compared to WT mice at 9 months old, especially in the cortical and hippocampal regions (Fig. 1C). Immunoblotting showed higher band intensity of APP isoforms and fragments including APP-CTFs which contain Aß peptides in the serum and brain lysates of the AD mice compared to WT mice (Fig. 1D). After the AD mouse model was verified, serum and brain lysates of the mice were used to analyze acrolein adduction. Pooled mouse serum and brain lysates were immunoprecipitated with a C-terminal APP antibody and then immunoblotted with acrolein. The level of acrolein adducts was increased obviously in the pooled AD mouse serum and brain lysates compared to those in WT mice, especially on APP-CTFs (Fig. 1E). These results further demonstrated the specific acrolein modification on Aß-containing peptides.

Correlation of acrolein adducts with MetS in AD patients

The levels of Aß and acrolein adducts in human serum were measured. The concentrations of biochemicals in routine clinical tests were provided by the Biobank as described in Materials and Methods. As shown in Fig. 2, the average levels of Aß and total protein seemed unchanged in the serum of AD patients compared to those of the control subjects, whereas acrolein adducts were significantly increased in AD serum (p < 0.001) (Fig. 2A). Interestingly, Pearson correlation analysis revealed that the level of acrolein adducts correlated positively with Aß and negatively with total protein (Fig. 2B). It suggests that acrolein adducts may be largely formed on Aß, which supports the findings in the AD models (Fig. 1A and E). Furthermore, the negative correlation of acrolein adducts with total protein in human serum may indicate a reduction of certain types of protein (i.e., antibodies) when acrolein adducts were increased in AD blood. Notably, significant correlations were also found between acrolein adducts and MetS parameters, including fasting glucose, triglycerides, and HDL-C (Fig. 2C). Therefore, both control and AD groups were further divided according to MetS-related laboratory test results as described in the “Materials and methods” section. However, the average age of the AD subjects was unexpectedly older than the control subjects with statistical significance. Therefore, Pearson correlation analysis was also performed to clarify the effects of age on these parameters, especially acrolein adducts, in the AD and non-AD control groups. It was plausible that age was not correlated with acrolein adducts or most of the tested parameters, except the negative correlation of age with total cholesterol and HDL-C in the control group (Table S1).

Fig. 2
figure 2

Correlation of acrolein adducts with specific metabolites in human serum. A Levels of Aß, acrolein adducts, and total protein (TP) in human serum samples. All the values were normalized with the mean of control. In the box plots, the dots represent the 5th and 95th percentiles; the error bars cover the 10th to 90th percentiles and the box covers the 25th to 75th percentiles; the solid and dash lines within the box represent the median and mean values, respectively. B Correlation of acrolein adducts with Aß and TP. Serum Aß and TP values were plotted against the level of acrolein adducts in control and AD subjects, respectively. The solid and broken lines represent the linear regression of the values in control and AD subjects, respectively. Pearson product moment correlation coefficients (r) of acrolein adducts with Aß and TP were shown with p values. C. Correlation of acrolein adducts with the metabolites related to metabolic syndrome. Serum fasting glucose (Glc), triglycerides (TG), and HDL-C values of control or AD subjects were plotted against the level of acrolein adducts. The solid and broken lines represent the regression of the values in control and AD subjects, respectively. Pearson product moment correlation coefficients (r) of acrolein adducts with Glc, TG, and HDL-C were shown with p values. Aß, amyloid-beta; AD, Alzheimer’s disease; Ctrl, control; HDL-C, high-density lipoprotein-cholesterol; TG, triglycerides; TP, total protein.*** p < 0.001

Full size image

Clinical profile and laboratory test results of the four groups, namely HC, MetS, AD-N, and AD-M, were summarized with mean ± SD in Table 1 and p values in Table S2. As expected, significant differences in fasting glucose, triglycerides, and HDL-C were observed between the groups with and without metabolic disturbance due to the sorting criteria. Likewise, HbA1c, LDL-C, and creatinine were also significantly different between the groups owing to MetS-related conditions. Although the average concentrations of both HDL-C and LDL-C were significantly lower in AD-M than in the MetS group, they were not considered different because of the age effect as mentioned (Table S1). No statistical differences were seen in gender, ALT, total cholesterol, and total protein between the groups.

Table 1 Demographic and clinical profile of the groups (mean ± SD)
Full size table

Enhanced acrolein-Aß adducts in the serum of AD patients with metabolic disturbance

Among the four groups, no significant difference in Aß level was observed (Fig. 3A). There were no significant differences in acrolein adducts observed among HC, MetS, and AD-N groups, either (Fig. 3B). Only in the AD-M group, the level of acrolein adducts was significantly increased compared to all other groups. The best AUC value of ROC analysis on acrolein adducts was observed in distinguishing AD-M from MetS (AUC = 0.729). These results suggest that acrolein adducts may be a better biomarker than Aß for not only AD pathology but also the pathogenesis from metabolic disturbance to neurodegeneration in AD.

Fig. 3
figure 3

Acrolein adducts on APP fragments and Aß peptides in human serum. A The relative levels of Aß in different groups of human serum sample. B The relative levels of acrolein adducts in different groups of human serum sample. All the values were normalized with the mean of healthy control (HC). The dots represent the 5th and 95th percentiles; the error bars cover the 10th to 90th percentiles and the box covers the 25th to 75th percentiles; the solid and dash lines within the box represent the median and mean values, respectively. The receiver operating characteristic curves were plotted with the area under curve values shown in brackets when the p value is < 0.001 between groups. C APP immunoprecipitation and acrolein immunoblotting of human serum. Pooled human serum samples from each group were IgG-removed and immunoprecipitated with anti-C-terminal APP antibody and then subjected to SDS-PAGE in triplicate. Two gels were immunoblotted with anti-C-terminal APP and acrolein antibodies, respectively. Another gel was stained with CBB as a loading control. The prestained protein ladder showed some luminance. Aß, amyloid beta; AD-M, Alzheimer’s disease with metabolic disturbance; AD-N, Alzheimer’s disease with normal metabolism; APP, amyloid precursor protein; CBB, Coomassie brilliant blue; HC, healthy control; IP, immunoprecipitation; M, marker; MetS, metabolic syndrome.***p < 0.001

Full size image

To detect acrolein adducts in more detail, human serum samples were pooled and subjected to immunoprecipitation and immunoblotting. The major APP species appeared near 48 kDa, and a minor species was seen near 28 kDa in the serum samples (Fig. 3C). They have been suggested as truncated APP-CTFs [28]. The low molecular-weight fragments were highly reactive to anti-acrolein antibody, especially in AD-M samples. These results were consistent with the findings in AD models and the ELISA data from human subjects, indicating that the formation of acrolein-Aß adducts may be highly correlated with metabolic disturbance in AD pathogenesis.

In vitro acrolein-modified Aß analyses

To further analyze acrolein-Aß adducts and their responding autoantibodies, we used synthetic Aß1-16 and Aß17-28 peptides and modified them with acrolein in vitro. Acrolein modification of Aß was verified by LC–MS/MS. Native and modified peptides were identified as b-ion (N-terminal fragments) and y-ion (C-terminal fragments) series. Native Aß1-16 fragment ion showed the [M + H]+ at 1953.87 and Aß17-28 at 1324.67 (Figs. 4A and 5A and Table 2). The molecular weight of acrolein is 56.06 Da. When acrolein binds proteins, four different types of adduct including schiff-base (+ 38 Da), aza-michael (+ 56 Da), Ne-lysine (MP-lysine) (+ 76 Da), and FDP-lysine (+ 94 Da) can be formed [29]. Acrolein modification on synthetic Aß peptides were observed on H6, H13, H14, and K16 on Aß1-16, and K28 on Aß17-28 (Figs. 4B and 5B and Table 2). Mainly aza-michael adducts were found on histidine residues, while all four types of adduct were found on lysine residues with schiff-base in the majority. The highest intensity of acrolein adducts was seen on K28 (Fig. 5B), while the most spectral count of the adducts was on K16 (Table 2). These results were consistent with the finding of acrolein modification on Aß-containing peptides as seen in mouse serum and brain lysates (Fig. 1) and human serum (Fig. 3).

Fig. 4
figure 4

Mass spectra of the native and acrolein-modified Aß1-16. The ion density (y-axis) was plotted against m/z ratio (x-axis). The b-ions (blue peaks) represent N-terminal fragments, whereas y-ions (red peaks) represents C-terminal fragments. Acrolein-modified histidine and lysine were pointed with arrows. A Native Aß1-16 peptide. B Acrolein-modified Aß1-16 at H6, H13, H14 or K16. Only the selected spectra were shown

Full size image
Fig. 5
figure 5

Mass spectra of the native and acrolein-modified Aß17-28. The ion density (y-axis) was plotted against m/z ratio (x-axis). The b-ions (blue peaks) represents N-terminal fragments, whereas y-ions (red peaks) represents C-terminal fragments. Acrolein-modified histidine and lysine were pointed with arrows. A Native Aß17-28 peptide. B Acrolein-modified Aß17-28 at K28. Only the selected spectra were shown

Full size image
Table 2 Aß peptides with acrolein-modified residues
Full size table

Alterations in anti-acrolein autoantibodies in AD patients with metabolic disturbance

Pearson correlation analysis revealed a negative correlation of acrolein adducts with total protein, especially in AD blood (Fig. 2B). We hypothesize that the increase of acrolein adducts as seen in AD-M condition (Fig. 3B) may be due to the depletion of acrolein-responding autoantibodies which are supposed to be an abundant protein type to neutralize toxic acrolein adducts. To test this hypothesis, we measured anti-acrolein antibody isotypes IgG and IgM in the human serum samples using acrolein-modified synthetic Aß1-16 and Aß17-28 peptides as antigens in ELISA. Native Aß peptides were used as control items.

No differences were seen in the relative levels of IgG against native Aß peptides among all groups (Fig. S2A). However, the ratio of anti-Aß1-16 IgG to the peptide was reduced slightly in the AD-M group compared to HC although the ratio of anti-Aß17-28 IgG to the peptide remained no difference among all groups (Fig. S2B). Unlikely, the relative level of IgG against acrolein-modified Aß1-16 was increased in the AD-M group moderately when compared to MetS and largely when compared to HC (p < 0.001) with an AUC value of 0.711 in ROC analysis (Fig. 6A). Nevertheless, the level of IgG against acrolein-modified Aß17-28 showed the same among all groups. Notably, the ratio of anti-acrolein-Aß1-16 IgG to acrolein adducts became no difference among all groups, while the ratio of anti-acrolein-Aß17-28 IgG to acrolein adducts was reduced moderately in AD-M group compared to MetS (Fig. 6B). Since the immune responses shall depend on the antigen levels, the ratio of the antibody to the specific antigen may provide a more useful evaluation. Therefore, these results suggest that the responding IgG levels, despite anti-native Aß or anti-acrolein-modified Aß, were reduced with slight to moderate amounts when MetS patients developed AD.

Fig. 6
figure 6

Change of IgG autoantibodies against acrolein-Aß adducts in human serum. A The relative levels of IgG autoantibody recognizing acrolein-modified Aß1-16 and Aß17-28 peptides in each group. All the values were normalized with the mean of healthy control (HC). B The ratio of the responding IgG to acrolein adducts of each subject in each group. In the box plots, the dots represent the 5th and 95th percentiles. The error bars cover the 10th to 90th percentiles and the box covers the 25th to 75th percentiles. The solid and dashed lines within the box represent the median and mean values, respectively. The receiver operating characteristic curves were plotted with the area under curve values shown in brackets when the p value is < 0.001 between groups. Aß, amyloid beta; AD-M, Alzheimer’s disease with metabolic disturbance; AD-N, Alzheimer’s disease with normal metabolism; HC, healthy control; MetS, metabolic syndrome.**p < 0.01, ***p < 0.001

Full size image

Similarly, both the level of IgM against native Aß1-16 and the ratio of anti-Aß1-16 IgM to the peptide were reduced moderately in the AD-M group compared to both MetS and HC (Fig. S3). In contrast, the level of IgM against acrolein-Aß17-28 and the ratio of both anti-acrolein-Aß1-16 and anti-acrolein-Aß17-28 IgM antibodies to acrolein adducts were reduced largely in AD-M group compared to MetS (p < 0.001) with AUC values from 0.693 to 0.755 in ROC analyses (Fig. 7). It was supportive in Pearson correlation analyses that the level of acrolein adducts was negatively correlated with anti-acrolein-Aß17-28 IgM although it was positively correlated with the specific IgG against native Aß1-16 in all samples (Table S3). These results suggest that the specific IgM autoantibodies against acrolein adducts, such as acrolein-Aß, were reduced largely when MetS patients developed AD and are potential biomarkers in AD pathology. In addition, acrolein-Aß adducts may be considered for the immunotherapy of AD, especially when AD is complexed with MetS.

Fig. 7
figure 7

Change of IgM autoantibodies against acrolein-Aß adducts in human serum. A The relative levels of IgM autoantibody recognizing acrolein-modified Aß1-16 and Aß17-28 peptides in each group. All the values were normalized with the mean of healthy control (HC). B The ratio of the responding IgM to acrolein adducts of each subject in each group. In the box plots, the dots represent the 5th and 95th percentiles. The error bars cover the 10th to 90th percentiles and the box covers the 25th to 75th percentiles. The solid and dashed lines within the box represent the median and mean values, respectively. The receiver operating characteristic curves were plotted with the area under curve values shown in brackets when the p value is < 0.001 between groups. Aß, amyloid beta; AD-M, Alzheimer’s disease with metabolic disturbance; AD-N, Alzheimer’s disease with normal metabolism; HC, healthy control; MetS, metabolic syndrome.* p < 0.05, **p < 0.01, ***p < 0.001

Full size image

Discussion

A recent Korean study demonstrates the association between MetS to developed AD in eight years [6]. It suggests that MetS patients are 11.48 times more prone to have AD than those without MetS. Statistical analysis of Taiwanese primary data from the National Health Insurance Research Database with encryption also reports the link between AD progressions from MetS [5]. This study explained the statistically significant relationship between dementia and worsen MetS compared to non-persistent MetS, improved MetS, or non-MetS in the follow-up period of 10 years. MetS is a prolonged state of oxidative stress which is resulted from abnormal metabolism of carbohydrates and lipids, inducing lipid peroxidation to generate aldehydic byproducts, especially acrolein adducts. Along with the accumulation of Aß in the diseased brain, the immune system may be stimulated to worsen AD progression.

In this study, acrolein adducts were found to correlate positively with triglycerides and negatively with HDL-C (Fig. 2C and Table S3). Research suggests that precautionary measurement in early (35–50 years) to middle (51–60 years) adulthood to maintain HDL-C and triglycerides levels may lower the risks of AD [30]. High circulating HDL levels have been linked to a lower risk of Alzheimer's disease in numerous epidemiological studies [24]. HDL appears to protect against memory loss, neuroinflammation, and cerebral amyloid angiopathy by lowering vascular Aß accumulation and attenuating Aß-induced endothelial inflammation. In contrast, high level of triglycerides can cause cognitive decline by blood brain barrier (BBB) dysfunction or amyloid metabolism imbalance [31]. However, cognitive function can be improved by coconut oil or medium-chain triglycerides. Omega-3 polyunsaturated fatty acids (ω-3-PUFAs) are key components of the neuronal membrane which acts as neuroprotective agent [32]. Impaired neurotransmission was observed in ω-3-PUFA-deficient animal [33]. Reduction in Aß production by altering APP pathological processing pathway is due to docosahexaenoic acid (22:6ω-3). A negative correlation was found in ω-3/ω-6 ratio with AD incidence and cognitive decline. Reduced docosahexaenoic acid and eicosapentaenoic acid (20:5ω-3) levels have been observed in AD brain autopsy, serum, and plasma.

Byproducts of PUFA oxidation are acrolein and HNE specifically from ω-6 PUFAs, such as arachidonic acid which is abundant in gray matter [2, 34]. Arachidonic acid oxidation in the presence of Aß in vitro produces HNE earlier than acrolein. It causes a high level of HNE at first and then leads to increased HNE-protein adducts but little extractable free HNE. In contrast, fewer acrolein-protein adducts and more extractable acrolein is produced. It has been shown that acrolein is much more toxic than HNE as seen in the primary culture of rat hippocampal neurons [37]. Polyamine like spermine helps for the suppression of Aß aggregation. A high level of acrolein forms conjugation with substrate and catalyst involved in tricarboxylic acid cycle, glycolysis, and carbon metabolism. These conjugations inhibit the energy metabolism under oxidative stress in AD conditions. Among acrolein-related risk factors for AD, smoking is also important due to increased oxidative stress and specially acrolein production. Impaired nitric oxide synthesis due to smoking alters glucose metabolism and cerebral blood flow in cerebral vascular endothelial cells of the brain. It stimulates Aß production, cerebral hypoperfusion, cytokine activation, and immune response, which are also helpful for AD diagnosis [38, 39].

Conclusion

Based on the present data, we conclude that metabolic disturbance may induce acrolein production and form adducts largely on Aß through oxidative stress. These adducts may lead to alterations in the autoantibodies, in which an increased level helps to slow disease progression by neutralizing and degrading acrolein adducts, whereas a decreased level of autoantibodies may lead to worse neurodegeneration and cause severe AD. This study explains the close relationship between AD and MetS by suggesting the mechanism that acrolein adduction in MetS may deplete responding autoantibodies and cause neurodegeneration in AD (Fig. 8).

Fig. 8
figure 8

Proposed reactions of acrolein-Aß adduct formation and degradation in correlation with metabolic disturbance and neurodegeneration. Under metabolic disturbance, oxidation of lipids generates aldehydic byproducts such as acrolein, which further reacts with peptides such as Aß and forms adducts. These adducts are responsible for neurodegeneration. However, the responding autoantibodies may inhibit the disease progression by neutralizing and degrading these adducts. Depletion of the responding autoantibodies by acrolein adducts may lead to worse neurodegeneration

Full size image