Ceramide as a key regulatory factor in the impact of dietary inflammatory index on plaque stability

Abstract

Diet influences plaque stability through its potential to modulate inflammation, a process that involves complex interactions among various lipid metabolites. This study aims to utilize metabolomics to identify key metabolites involved in this pathway and to elucidate the mechanisms by which dietary factors affect plaque stability. The Dietary Inflammatory Index (DII), derived from dietary data, was used to assess the inflammatory potential of individual diets. Propensity score matching categorized serum samples from coronary heart disease (CHD) patients into an anti-inflammatory group (n = 108) and a pro-inflammatory group (n = 108). A comprehensive analysis of lipid profiles was performed using an UPLC-MS/MS detection platform combined with the broad-targeted lipid metabolomics technique, and lipid metabolites with significant differences were screened out. Concurrently, we measured serum levels of inflammatory factors and plaque stability. A Bayesian network model was then applied to elucidate the causal relationships among DII, lipid metabolites, inflammatory factors, and plaque stability. A lipidomics analysis identified 22 differentially expressed lipid metabolites, which were associated with sphingolipid metabolism pathways in the KEGG (Kyoto Encyclopedia of Genes and Genomes) database, particularly involving nine ceramide species. The Bayesian network model exploring the impact of DII on plaque stability comprises 16 nodes and 23 directed arcs. It revealed multiple causal relationships among DII, ceramide species, inflammatory factors, and plaque stability. Specifically, six ceramide species [Cer(d16:0/20:1), Cer(d24:3/15:1), Cer(t14:1/21:0), Cer(t20:0/18:2), Cer(t22:1/16:1), Cer(t26:1/12:1)] and five inflammatory factors (IFN-γ, IL-1β, IL-8, IL-12, IL-13) were found to be involved in these associations. Ceramide species emerged as differential lipid metabolites that distinguish between the anti-inflammatory and pro-inflammatory groups, simultaneously serving as key lipid metabolic products through which diet exerts its influence on plaque stability.

Introduction

Coronary heart disease (CHD), a severe condition marked by high rates of morbidity, disability, and mortality. It primarily results from atherosclerosis in the coronary arteries, leading to stenosis or occlusion, which in turn causes myocardial ischemia, hypoxia, or necrosis¹. The prevailing hypothesis suggests that CHD is a chronic inflammatory disease, driven by lipid metabolic disorders. When lipid accumulation exceeds the metabolic capacity of intravascular macrophages, it triggers an inflammatory cascade that undermines plaque stability². Diet, as the primary source of lipid components, plays a central role in modulating inflammation and carries significant inflammatory potential³. The lipid metabolites derived from the diet can either elevate or reduce inflammatory factor levels within the body⁴. Numerous studies have emphasized the critical role of the Dietary Inflammatory Index(DII) in the onset and progression of CHD^5–7. Our preliminary research has also identified the involvement of the DII in the formation of unstable plaques⁸. However, the specific lipid metabolites and inflammatory factors that drive this complex process remain largely unidentified. Therefore, there is an urgent need for further investigation into these metabolic and inflammatory mediators, as they may reveal novel therapeutic targets and intervention strategies for CHD.

Under the conditions of inflammation and oxidative stress, alterations in lipid metabolism and dyslipidemia become key internal drivers that facilitate the transition of atherosclerotic plaques from a stable to an unstable state. Research highlights the pivotal roles that disruptions in glucose-lipid metabolism, aromatic amino acid metabolism, oxidative stress, and inflammatory pathways play in the progression of atherosclerosis⁹. Biswapriya and others have noted that metabolites generated under pro-inflammatory, high-cholesterol, and high-fat diet conditions are linked to the plaque formation¹⁰. This conclusion is further supported by animal studies demonstrating that feeding ApoE-/- mice with a high-fat and high-cholesterol diet triggers macrophage polarization, and inflammatory factor secretion, which in turn promotes the development of atherosclerotic plaques¹¹. Additionally, research has pointed out that metabolic biomarkers can differentiate varying degrees of inflammatory states, with glycerophospholipid metabolism and arginine-proline metabolism potentially influencing systemic immunity and low-grade inflammatory changes¹². Based on these findings, we hypothesize that lipid metabolites and inflammatory factors may serve as key mediators in the process by which dietary inflammatory potential impacts plaque stability.

Metabolomics, an evolving technology for investigating small molecular substrates, intermediates, and products inherent to cellular metabolism, is experiencing rapid advancements. It employs sophisticated analytical methodologies to conduct qualitative and quantitative multi-parametric assessments of metabolites in biological fluids like blood, urine, and saliva, within a specified time frame¹³. Recent innovations in detection techniques, notably high-resolution mass spectrometry, matrix-assisted laser desorption/ionization mass spectrometry, and mass spectrometry imaging, have expanded metabolomics to lipidomics. Lipid metabolism represents the primary category of metabolism, accounting for approximately 70% of bioactive molecules in plasma, which are composed of lipids. Lipidomics research significantly enhanced our understanding of the biological roles of lipids and underscored their importance in refining cardiovascular disease (CVD) risk prediction^14,15. The basis of CHD is atherosclerosis, which is mainly caused by inflammation, oxidative stress, lipid metabolism disorders triggered by glycerol phospholipid, sphingomyelin, cholesterol, ceramide, and accumulation of vascular lumen stenosis or obstruction, resulting in myocardial ischemia, hypoxia, or necrosis^16,17. However, the relationship between inflammatory markers, lipid metabolites, and plaque stability from a dietary perspective has received limited attention. Clarifying this complex interplay could provide novel insights into the mechanisms of plaque destabilization through the lens of dietary lipid metabolism.

This study employs a broad-target lipidomics approach based on the UPLC-MS/MS detection platform to analyze lipid components. Building on our previous research, we aim to investigate the potential lipid metabolites and their associated metabolic pathways in the context of inflammation among patients with CHD. Furthermore, by employing a Bayesian network model, we seek to elucidate the underlying pathways linking dietary inflammatory potential to plaque stability, thereby revealing the intrinsic mechanisms by which dietary inflammation influences plaque stability.

Materials and methods

Participants

Subjects were recruited from the cardiology department of the Second Affiliated Hospital of Harbin Medical University between March 2022 and March 2023. The exclusion criteria included: patients with other severe CVD and systemic diseases(n = 5); daily energy intake less than 600 kcal or more than 4000 kcal, as well as those with digestive system diseases(n = 3); pregnant or lactating women(n = 1); poor clarity of optical coherence tomography (OCT) images(n = 10); incomplete clinical data, and unqualified blood samples (n = 17). A total of 261 eligible patients were identified. To more clearly compare the effects of different dietary patterns, propensity score matching was performed in a 1:1 ratio based on disease diagnosis (acute myocardial infarction or chronic myocardial infarction), gender and age. Ultimately, 216 patients were included in the study: 108 in the pro‑inflammatory group and 108 in the anti‑inflammatory group. All participants provided informed consent and underwent questionnaire surveys, including for age, gender, BMI, family history of CHD, hypertension, hyperlipidemia, diabetes mellitus, and smoking habits. Optical coherence tomography (OCT) was employed to assess the plaque stability of the patients. Approximately 5 milliliters of whole blood samples were collected from the patients, and a comprehensive analysis of lipid components in the samples was performed using an ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) detection platform combined with the broad-targeted lipid metabolomics technique. Meanwhile, a multi-factor MSD kit was used to detect the inflammatory factors.

This research protocol was approved by the Ethics Committee of the Second Affiliated Hospital of Harbin Medical University. All experiments were conducted in accordance with relevant guidelines and regulations.

Laboratory examination

The laboratory indicators encompassed blood glucose levels, total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), triglycerides (TG), and apolipoprotein A (apoA).

DII assessment

This study utilized a semi-quantitative food frequency questionnaire (SQFFQ) to collect data on the consumption frequency and typical serving sizes of various foods over the past year, covering a total of 45 food categories. Based on the SQFFQ, the intake levels of 21 dietary components were derived, consisting of 8 pro‑inflammatory components (carbohydrates, cholesterol, energy, total fat, saturated fatty acids, protein, iron, and vitamin B12) and 13 anti-inflammatory components (vitamin A, thiamine, riboflavin, vitamin B6, vitamin C, vitamin D, vitamin E, dietary fiber, folic acid, niacin, magnesium, zinc, and selenium). The Dietary Inflammatory Index (DII) was calculated as follows: (1) Standardization of dietary intake: the daily intake of each nutrient/food in the study population was compared with a globally representative mean daily intake to obtain a Z-score for each individual nutrient/food. (2) Z-score centering: the Z-scores were converted into percentiles to achieve a symmetric distribution centered at 0 and ranging from − 1 (maximally anti-inflammatory) to + 1 (maximally pro-inflammatory). (3) DII score for each nutrient: the centered Z-score was multiplied by its corresponding literature-derived inflammatory effect score to obtain the nutrient-specific DII score. (4) Total DII score: the DII scores of all nutrients were summed to yield the overall DII score¹⁸. A lower DII score signifies an anti-inflammatory diet, whereas a higher score indicates a pro-inflammatory diet.

Broad-target metabolomics by high performance liquid chromatography-mass spectrometry

Approximately 5 mL of whole blood samples were collected from the patients, processed following standardized procedures, and stored in a −80 °C refrigerator. Subsequently, the sample extracts were analyzed using a liquid chromatography-electrospray ionization tandem mass spectrometry (LC-ESI-MS/MS) system, which was integrated with the UPLC ExionLC AD and QTRAP^® System. For chromatography, a Thermo Accucore™ C30 column (2.6 μm, 2.1 mm × 100 mm i.d.) was used, employing a solvent system consisting of solvent A (acetonitrile/water, 60/40 V/V, containing 0.1% formic acid and 10 mmol/L ammonium formate) and solvent B (acetonitrile/isopropanol, 10/90 V/V, with the same additives). A pre-programmed gradient elution was initiated with an A/B ratio of 80:20 V/V, gradually adjusting to various proportions, and concluding with a return to A/B 80:20 V/V at 20 min. The flow rate was maintained at 0.35 mL/min, with a column temperature set at 45 °C, and an injection volume of 2 µL.

Subsequently, the effluent was introduced into the QTRAP^® LC-MS/MS system via an ESI Turbo Ion-Spray interface, operating in both positive and negative ion modes, and controlled by Analyst 1.6.3 software. The ESI source parameters were configured with a Turbo Spray ion source, a source temperature of 500 °C, and ion spray voltages of 5500 V (positive ion mode) or −4500 V (negative ion mode). The ion source gases, GS1, GS2, and CUR, were set to 45 psi, 55 psi, and 35 psi, respectively, while the collision gas was set to medium intensity. Instrument tuning and mass calibration were performed using polypropylene glycol solutions at concentrations of 10 and 100 µmol/L in QQQ (triple quadrupole) and LIT modes, respectively. In QQQ mode, MRM experiments were conducted with nitrogen as the collision gas at 5 psi, monitoring specific sets of MRM transitions for each elution period based on the eluted metabolites, with further optimization of declustering potential (DP) and collision energy (CE). Additionally, LIT scans were acquired to gather further fragment ion information.

OCT measurement

OCT is currently the gold standard for evaluating the composition and stability characteristics of coronary atherosclerotic plaques. The OCT measurement indicators include lipid plaques, which are characterized by blurred edges, high back-reflection, and strongly attenuated areas, with a high-signal fibrous cap on the surface of the low-signal area¹⁹; thin-cap fibroatheroma (TCFA): with the thinnest part of the fibrous cap < 65 μm and the lipid core angle ≥ 90°²⁰, serving as a specific indicator in this study; cholesterol crystals manifest as thin linear areas with high signal intensity and low attenuation, usually located in the fibrous cap or necrotic core of lipid plaques along with lipids²⁰; plaque rupture: manifested as discontinuity of the fibrous cap with cavity formation in the image²¹.

Inflammatory factor

The Multi-Factor MSD Kit was utilized to quantify inflammatory factors (IFN-γ, IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12, IL-13, and TNF-α). The experimental protocol comprised the following steps: Blood samples from patients with CHD, stored at −80 °C, were centrifuged at 4 °C and 3000 g for 20 min, and the resulting supernatant was collected; According to the manufacturer’s instructions, 25–50 µl aliquots of standards, quality control samples, and test samples were sequentially added to the assay plates and incubated at 700 rpm for 2 h at room temperature; Following three washes with 150 µl of wash buffer, 25 µl of detection antibody was added and incubated at 700 rpm for an additional 2 h at room temperature; After another three washes with 150 µl of wash buffer, 150 µl of reading solution was added, and the plates were read using a dedicated instrument.

Statistical analysis

Normally distributed continuous data were analyzed with the independent two-sample t-test and are presented as mean±standard deviation, whereas non-normally distributed continuous data were analyzed using the Mann-Whitney U test and are expressed as median and interquartile range. To control for potential confounding bias due to imbalanced baseline characteristics, propensity score matching was employed to obtain more comparable groups. Propensity scores were matched using 1:1 nearest-neighbor matching with a caliper width of 0.02. After excluding 45 unmatched cases, a total of 108 patients were retained in each of the anti-inflammatory and pro-inflammatory groups. Categorical variables are reported as a percentage, and chi-square tests were used to compare groups. First, Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA), a supervised pattern recognition method, is adopted to effectively eliminate the influences related to the research, so as to screen differential metabolites. In the OPLS-DA model, Variable Importance in the Project (VIP) values and Fold Change (FC) values were computed for all peaks. Potential metabolic biomarkers were selected based on the criteria of VIP > 1, absolute logFC > 1, and a p-value < 0.05 derived from the nonparametric Kruskal-Wallis test. Subsequently, Pathway enrichment analysis was performed utilizing the Kyoto Encyclopedia of Genes and Genomes (KEGG) database to gain insights into the underlying biological pathways associated with these biomarkers.

The Pearson correlation coefficient was employed to assess the correlation among DII, inflammatory factors, differential lipid metabolites, and plaque stability. Using the bnlearn package in R 4.3.2 for Bayesian network learning, the Directed Acyclic Graph (DAG) is constructed through the Max-Min Hill Climbing algorithm within a hybrid learning framework. Prior knowledge is incorporated by setting a blacklist of variables. A Bootstrap resampling method is applied to perform 1000 iterations of the learning process. Following this, a significance test is conducted for each arc in the resulting DAGs, and only arcs with a significance level below 0.05 are included in the final network. For Bayesian network inference, the Maximum Likelihood Estimation (MLE) method is employed. To visualize the DAG, the strength.plot function is utilized to illustrate the network, highlighting the strength of each arc.

Results

Characteristics of patients

The demographic and clinical characteristics of the subjects involved are shown in Table 1, with a total of 216 patients included (121 in stable plaque group and 95 in the unstable plaque group). The median age of the unstable plaque group was 60 years old, with 72.6% being male. while the median age of the stable plaque group was 58 years old, with 67.8% being male. As shown in Table 1, significant differences were observed in the levels of 9 inflammatory factors between stable plaque and unstable plaque group, with statistical significance (p < 0.05). Among them, the levels of IL-12, IL-1β, IL-4, IL-6, and TNF-α were significantly higher than those in unstable plaque group. Given that sex is a key modifying factor in the influence of inflammation on plaque stability, we conducted a subgroup analysis (Table 2). No significant difference was observed in the association pattern between the Dietary Inflammatory Index (DII) and plaque stability among male and female patients (p > 0.05).

Table 1.

Baseline characteristics of patients in unstable plaque and stable plaque.

Variable	Total (n = 216)	Unstable plaque (n = 95)	Stable plaque (n = 121)	p-value
Age M (P25, P75), years	59(51,66)	60(50,66)	58(52,66)	0.770
Gender (male, %)	151(69.9)	69(72.6)	82(67.8)	0.439
Hypertension (%)	99(45.8)	39(39.4)	60(60.6)	0.211
Hyperlipidemia (%)	95(44.0)	38(40.0)	57(60.0)	0.296
Diabetes (%)	56(25.9)	26(46.4)	30(53.6)	0.668
Family history of CHD (%)	73(33.8)	29(30.5)	44(60.3)	0.368
BMI, M (P25, P75), kg/m2	25.4(22.93,27.58)	25.3(22.6,27.5)	25.4(23.4,27.7)	0.237
TC, M (P25, P75), mmol/L	4.62(3.59,5.51)	4.56(3.53,5.39)	4.69(3.71,5.58)	0.331
TG, M (P25, P75), mmol/L	1.76 ± 1.30	1.38(1.06,2.16)	1.44(0.95,2.0)	0.717
LDL-C, M (P25, P75), mmol/L	2.83(2.00,3.60)	2.77(1.96,3.51)	2.91(2.04,3.65)	0.673
HDL-C, M (P25, P75), mmol/L	1.01(0.87,1.23)	0.99(0.87,1.20)	1.03(0.87,1.27)	0.405
BG, M (P25, P75), mmol/L	6.86(5.57,9.00)	6.76(5.55,8.93)	7.17(5.60,9.22)	0.328
apoA, M (P25, P75), g/L	1.19(1.06,1.37)	1.17(1.07,1.36)	1.21(1.04,1.37)	0.753
Ln INF-γ [pg/ml, M(SD)]	0.65 ± 0.43	0.78 ± 0.46	0.55 ± 0.37	< 0.001
Ln IL-10 [pg/ml, M(SD)]	0.38 ± 0.69	0.28 ± 0.66	0.47 ± 0.70	0.044
Ln IL-2 [pg/ml, M(SD)]	−0.51 ± 0.50	−0.38 ± 0.39	−0.62 ± 0.54	< 0.001
Ln IL-4 [pg/ml, M(SD)]	−1.55 ± 0.79	−1.84 ± 0.63	−1.32 ± 0.83	< 0.001
Ln IL-6 [pg/ml, M(SD)]	−0.27 ± 1.44	0.10 ± 0.68	−0.56 ± 1.17	< 0.001
Ln IL-8 [pg/ml, M(SD)]	0.54 ± 0.36	0.61 ± 0.36	0.48 ± 0.35	0.010
Ln IL-12 [pg/ml, M(SD)]	−1.55 ± 0.79	−0.37 ± 0.68	−0.62 ± 0.56	0.004
Ln IL-13 [pg/ml, M(SD)]	−0.36 ± 0.46	−0.46 ± 0.46	−0.28 ± 0.44	0.004
Ln IL-1β [pg/ml, M(SD)]	−0.75 ± 0.49	−0.59 ± 0.40	−0.87 ± 0.52	< 0.001
Ln TNF-α [pg/ml, M(SD)]	0.38 ± 0.26	0.44 ± 0.30	0.33 ± 0.21	0.001

Continuous variables are expressed as mean and standard deviation (SD) or quartile (P25,P75), and categorical variables are expressed as counts and percentages.

CHD coronary heart disease; BMI body mass index; TC total cholesterol; LDL-C low-density lipoprotein; HDL-C high-density lipoprotein; TG triglycerides; BG blood glucose; IFN-γ interferon-gamma; IL-1 βinterleukin-1 beta; IL-2 interleukin-2; IL-4 interleukin-4; IL-6 interleukin-6; IL-8 interleukin-8; IL-10 interleukin-10; IL-12 interleukin-12; IL-13 interleukin-13; TNF-α tumor necrosis factor alpha.

Table 2.

Subgroup analysis of the effect of dietary inflammatory index on plaque stability in patients with coronary heart Disease.

Variables	n(%)	OR (95%CI)	P	P for interaction
All patients	216(100.00)	13.918 (7.140–27.129.140.129)	< 0.001
Gender				0.055
Female	151(69.90%)	9.25 (4.37–19.53)	< 0.001
Male	65(30.10%)	54.857 (10.449–287.991.449.991)	< 0.001

OR Odds ratio, 95%CI: Confidence Interval.

Discovery and identification of potential metabolic biomarkers in anti-inflammatory and pro-inflammatory groups

Based on previous studies, patients were divided into two groups using propensity matching scores: a pro-inflammatory group with 108 patients and an anti-inflammatory group with 108 patients. After instrumental analysis, peak detection and alignment, and metabolite recognition, a total of 1213 annotated metabolites were detected in all collected samples. A t-test was performed on the relative content of the differential lipid metabolites between the two groups, and a threshold of VIP > 1, P < 0.05 and |LogFC|>1 was used for screening. A total of 22 significantly different lipid metabolites were selected in Fig. 1, with their concentrations in the anti-inflammatory group being significantly higher than those in the pro-inflammatory group. The detailed information of these lipid differential metabolites is shown in Table 3.

Fig. 1.

Difference multiples bar chart of lipid metabolite content differences.

Table 3.

List of information on 22 significantly different lipid metabolites between anti-inflammatory group (n = 108) and pro-inflammatory group (n = 108).

Identity	VIP	p-Value	FC	Log2FC	AUC
Cer (d16:0/20:1)	3.616	< 0.001	2.223	1.152	0.84(0.783–0.897)
Cer (d21:3/16:0)	3.354	< 0.001	2.330	1.220	0.786(0.723–0.848)
Cer (d24:3/15:1(2OH))	3.207	< 0.001	2.228	1.156	0.757(0.692–0.823)
Cer (t14:1/21:0)	4.078	< 0.001	2.209	1.144	0.881(0.835–0.927)
Cer (t18:1/20:1(2OH))	4.302	< 0.001	2.058	1.041	0.889(0.840–0.937)
Cer (t20:0/18:2(2OH))	4.570	< 0.001	2.662	1.413	0.914(0.876–0.953)
Cer (t20:2/18:0(2OH))	4.536	< 0.001	2.196	1.135	0.894(0.844–0.944)
Cer (t22:1/16:1(2OH))	4.010	< 0.001	2.092	1.065	0.905(0.860–0.950)
Cer (t26:1/12:1(2OH))	4.681	< 0.001	2.430	1.281	0.905(0.861–0.949)
DG (12:0_16:1)	2.578	< 0.001	2.650	1.406	0.726(0.658–0.794)
DG (O-16:1_18:1)	3.092	< 0.001	2.251	1.171	0.763(0.701–0.826)
DG (O-18:1_18:1)	2.676	< 0.001	2.239	1.163	0.726(0.695–0.793)
HexCer (t14:0/24:1(2OH))	3.990	< 0.001	2.125	1.088	0.871(0.821–0.921)
HexCer (t14:0/24:2(2OH))	4.724	< 0.001	2.326	1.218	0.904(0.858–0.950)
HexCer (t14:1/24:0(2OH))	4.471	< 0.001	2.198	1.136	0.891(0.843–0.940)
HexCer (t16:0/22:2(2OH))	4.029	< 0.001	2.103	1.072	0.871(0.820–0.922)
HexCer (t18:1/20:0(2OH))	4.471	< 0.001	2.154	1.107	0.891(0.842–0.940)
HexCer (t18:2/20:0(2OH))	4.639	< 0.001	2.485	1.313	0.898(0.852–0.944)
HexCer (t22:1/16:1(2OH))	4.716	< 0.001	2.516	1.331	0.905(0.860–0.950)
HexCer (t24:1/14:1(2OH))	4.199	< 0.001	2.356	1.236	0.887(0.838–0.936)
PE (P-22:0_22:6)	5.262	< 0.001	11.563	3.531	0.982(0.976–0.998)
TG (8:0_16:0_18:1)	4.530	< 0.001	2.096	1.068	0.932(0.877–0.969)
Cer (d21:3/16:0)	3.616	< 0.001	2.223	1.152	0.786(0.723–0.848)
Cer (d24:3/15:1(2OH))	3.354	< 0.001	2.330	1.220	0.757(0.692–0.823)

VIP variable importance in the project; FC fold change; AUC area under the curve; Cer ceramide; DG diglyceride; HexCer hexosylceramide; PE phosphatidylethanolamine; TG triglyceride.

Pathway enrichment analysis was conducted by KEGG database²². Finally, we obtained 22 significantly different lipid metabolites that were eventually enriched in sphingolipid metabolic pathways involving nine ceramide species, including Cer(d16:0/20:1), Cer(d21:3/16:0), Cer(d24:3/15:1), Cer(t14:1/21:0), Cer(t18:1/20:1), Cer(t20:0/18:2), Cer(t20:2/18:0), Cer(t22:1/16:1), and Cer(t26:1/12:1) (Fig. 2). Meanwhile, these nine metabolites can serve as stable biomarkers to diagnose the anti-inflammatory and pro-inflammatory groups.

Fig. 2.

KEGG enrichment plot of differential lipid metabolite. KEGG Kyoto Encyclopedia of Genes and Genomes.

Lipid metabolites and inflammatory factors associated with four unstable plaque traits

Following logarithmic transformation, an independent sample t-test was conducted to identify lipid metabolites and inflammatory factors that are correlated with 4 plaque stability characteristics, (TCFA, lipid plaque, cholesterol crystals, and plaque rupture). Table 4 illustrates the differential expression of nine lipid metabolites and inflammatory factors among four comparative groups. IL-10 shows significant differences specifically between the TCFA and non-TCFA groups. IL-8 exhibits significant variations in the three groups other than the lipid-rich plaque group(p < 0.05), and TNF-α demonstrates significant differences within the TCFA and non-TCFA groups, lipid-rich plaque and non-lipid-rich plaque groups (p < 0.05). Additionally, IL-2, IL-6, IL-12, IL-1β, IL-4, IL-13, and IFN-γ all display statistically significant differences across the four groups (p < 0.05).

Table 4.

Differences in lipid metabolites and inflammatory factors between stable plaque and unstable plaque groups (n = 216).

Variable	TCFA (n = 95)	Non-TCFA (n = 121)	p-Value	Lipid plaque (n = 123)	Non-lipid plaque (n = 93)	p-Value	Cholesterol crystal plaque (n = 87)	Non-cholesterol crystal (n = 129)	p-Value	Rupture plaque (n = 48)	Non-rupture plaque (n = 168)	p-Value
Cer(t18:1/20:1)	4.94 ± 0.21	4.74 ± 0.28	< 0.001	4.90 ± 0.25	4.73 ± 0.27	< 0.001	4.93 ± 0.21	4.76 ± 0.28	< 0.001	4.93 ± 0.21	4.76 ± 0.28	< 0.001
Cer(t22:1/16:1)	4.57 ± 0.31	4.39 ± 0.32	< 0.001	4.55 ± 0.31	4.36 ± 0.32	< 0.001	4.57 ± 0.30	4.40 ± 0.33	< 0.001	4.57 ± 0.30	4.40 ± 0.33	< 0.001
Cer(t20:2/18:0)	5.27 ± 0.22	5.04 ± 0.29	< 0.001	5.22 ± 0.26	5.04 ± 0.29	< 0.001	5.25 ± 0.22	5.07 ± 0.30	< 0.001	5.25 ± 0.22	5.07 ± 0.30	< 0.001
Cer(t14:1/21:0)	4.24 ± 0.25	3.98 ± 0.32	< 0.001	4.17 ± 0.30	3.99 ± 0.31	< 0.001	4.22 ± 0.25	4.01 ± 0.33	< 0.001	4.22 ± 0.25	4.01 ± 0.33	< 0.001
Cer(d16:0/20:1)	4.57 ± 0.36	4.34 ± 0.42	< 0.001	4.53 ± 0.40	4.33 ± 0.40	< 0.001	4.58 ± 0.36	4.35 ± 0.42	< 0.001	4.58 ± 0.36	4.35 ± 0.42	< 0.001
Cer(d24:3/15:1)	5.25 ± 0.46	4.96 ± 0.53	< 0.001	5.19 ± 0.50	4.95 ± 0.52	0.001	5.26 ± 0.49	4.98 ± 0.52	< 0.001	5.26 ± 0.49	4.98 ± 0.52	< 0.001
Cer(d21:3/16:0)	5.07 ± 0.44	4.78 ± 0.50	< 0.001	5.01 ± 0.47	4.77 ± 0.50	< 0.001	5.07 ± 0.47	4.79 ± 0.48	< 0.001	5.07 ± 0.47	4.79 ± 0.48	< 0.001
Cer(t26:1/12:1)	4.98 ± 0.24	4.73 ± 0.31	< 0.001	4.93 ± 0.27	4.72 ± 0.31	< 0.001	4.96 ± 0.25	4.76 ± 0.32	< 0.001	4.96 ± 0.25	4.76 ± 0.32	< 0.001
Cer(t20:0/18:2)	4.46 ± 0.30	4.20 ± 0.34	< 0.001	4.40 ± 0.33	4.19 ± 0.34	< 0.001	4.46 ± 0.27	4.21 ± 0.36	< 0.001	4.46 ± 0.27	4.21 ± 0.36	< 0.001
IFN-γ	0.78 ± 0.46	0.55 ± 0.37	< 0.001	0.73 ± 0.41	0.56 ± 0.42	0.003	0.75 ± 0.41	0.58 ± 0.43	0.004	0.84 ± 0.45	0.60 ± 0.40	0.001
IL-10	0.28 ± 0.66	0.47 ± 0.70	0.044	0.34 ± 0.70	0.44 ± 0.67	0.267	0.32 ± 0.71	0.42 ± 0.67	0.299	0.30 ± 0.71	0.41 ± 0.68	0.355
IL-2	−0.38 ±0.39	−0.62 ± 0.54	< 0.001	−0.45 ± 0.40	−0.60 ± 0.60	0.031	−0.42 ± 0.37	−0.57 ± 0.56	0.024	−0.27 ± 0.36	−0.58 ± 0.51	< 0.001
IL-4	−1.84 ± 0.63	−1.32 ± 0.83	< 0.001	−1.68 ± 0.74	−1.38 ± 0.83	0.005	−1.72 ± 0.67	−1.43 ± 0.84	0.008	−1.86 ± 0.45	−1.46 ± 0.84	0.002
IL-6	0.10 ± 0.68	−0.56 ± 1.17	< 0.001	−0.08 ± 0.85	−0.53 ± 1.20	0.001	0.01 ± 0.81	−0.46 ± 1.13	0.001	0.29 ± 0.51	−0.43 ± 1.09	< 0.001
IL-8	0.61 ± 0.36	0.48 ± 0.35	0.010	0.57 ± 0.37	0.49 ± 0.34	0.095	0.60 ± 0.36	0.50 ± 0.35	0.033	0.73 ± 0.41	0.49 ± 0.32	< 0.001
IL-12	−0.37 ± 0.68	−0.62 ± 0.56	0.004	−0.37 ± 0.67	−0.69 ± 0.52	< 0.001	−0.38 ± 0.70	−0.59 ± 0.56	0.018	−0.42 ± 0.74	−0.53 ± 0.59	0.278
IL-13	−0.46 ± 0.46	−0.28 ± 0.44	0.004	−0.42 ± 0.48	−0.29 ± 0.43	0.040	−0.52 ± 0.46	−0.26 ± 0.43	< 0.001	−0.49 ± 0.46	−0.32 ± 0.45	0.026
IL-1β	−0.59 ± 0.40	−0.87 ± 0.52	< 0.001	−0.66 ± 0.40	−0.87 ± 0.57	0.001	−0.62 ± 0.40	−0.84 ± 0.53	0.001	−0.55 ± 0.39	−0.81 ± 0.50	0.001
TNF-α	0.44 ± 0.30	0.33 ± 0.21	0.001	0.43 ± 0.27	0.31 ± 0.22	0.001	0.40 ± 0.27	0.37 ± 0.24	0.425	0.42 ± 0.30	0.37 ± 0.24	0.267

TCFA thin-cap fibroatheroma; Cer ceramide; IFN-γ interferon-gamma; IL-10 interleukin-10; IL-2 interleukin-2; IL-4 interleukin-4; IL-6 interleukin-6; IL-8 interleukin-8; IL-12 interleukin-12; IL-13 interleukin-13; IL-1β interleukin-1 beta; TNF-α tumor necrosis factor alpha.

Bayesian network model

To further investigate the impact of DII on various plaque stability traits, a Bayesian network analysis was conducted, incorporating a total of 22 variables: DII, 9 lipid metabolites, 8 significant inflammatory cytokines based on intergroup comparisons, and 4 plaque stability traits. The initial associations between these variables are provided in the Fig. 3. To obtain a stable Directed Acyclic Graph (DAG) for the Bayesian network, the Bootstrap method was employed to sample from the data 1000 times. The 1000 DAGs were then averaged, and conditional independence significance tests were performed on the directed arcs within the DAG. The findings revealed directed arcs between DII and 6 lipid metabolites, 2 lipid metabolites and IL-1β, 3 lipid metabolites and IFN-γ, 2 lipid metabolites and IL-8, 2 lipid metabolites and IL-12, and 1 lipid metabolite and IL-13. Additionally, directed arcs were observed between IL-1β and IFN-γ with TCFA, IL-1β, IFN-γ, and IL-8 with rapture plaque, IL-12 with lipid plaque, and IL-13 with cholesterol crystals. The thickness of the lines represents the strength of the influence among factors as illustrated in Fig. 4.

Fig. 3.

Initial relationship plot between DII and plaque stability. DII dietary inflammatory index; TCFA thin-cap fibroatheroma; Cer ceramide; IFN-γ interferon-gamma; IL-4 interleukin-4; IL-6 interleukin-6; IL-8 interleukin-8; IL-12 in-terleukin-12; IL-13 interleukin-13; IL-1β interleukin-1 beta; TNF-α tumor necrosis factor alpha.

Fig. 4.

Causal pathways between DII and 4 plaque stability characteristics. DII dietary inflammatory index; TCFA thin-cap fibroatheroma; Cer ceramide; IFN-γ interferon-gamma; IL-8 interleukin-8; IL-12 in-terleukin-12; IL-13 interleukin-13; IL-1β interleukin-1 beta.

Discussion

In this study, we employed a broad-targeted lipid metabolomics approach to analyze the overall lipid metabolic profiles in the serum of two patient groups, aiming to elucidate the metabolic characteristics of lipids under varying dietary inflammatory potential conditions. Using Bayesian network analysis, we explored the complex relationships among dietary inflammatory potential, lipid metabolites, inflammatory factors, and plaque stability. The key findings of this research are as follows: (1) significant differences in metabolites were observed between the anti-inflammatory and pro-inflammatory groups; (2) nine ceramide species were notably higher in the pro-inflammatory group compared to the anti-inflammatory group, with these species primarily enriched in sphingolipid metabolic pathways; and (3) ceramides may serve as crucial lipid metabolites through which dietary factors influence inflammatory potential.

The formation of atherosclerotic plaques involves an inflammatory and oxidative stress process, beginning with endothelial injury, followed by the invasion of local immune cells, lipid accumulation, and vascular wall remodeling. Diet, as an intervenable exposure variable, has been shown to influence inflammation, and the DII provides a comprehensive assessment of the overall inflammatory potential of an individual’s diet. Multiple studies have confirmed the association between DII and the incidence of CVD and its subtypes, with pro-inflammatory diets associated with a 38% higher risk of CVD compared to anti-inflammatory diets⁵. Our previous research has also demonstrated that a higher DII is linked to the formation of unstable plaques⁸. However, the calculation of DII is complex, requiring detailed dietary data and intricate computations. Therefore, identifying biomarkers that differentiate between anti-inflammatory and pro-inflammatory states is of significant clinical importance. In this study, we identified nine ceramide species that effectively differentiate between the anti-inflammatory and pro-inflammatory groups. These nine distinct metabolites serve as non-invasive biomarkers for the diagnosis of anti-inflammatory and pro-inflammatory states, primarily enriched in sphingolipid metabolic pathways.

Serum lipid profiles provide an estimation of lipid content in the bloodstream. However, this rough estimate is often inadequate for precisely identifying or comprehensively reflecting the status of atherosclerotic plaque formation or the development of unstable plaques in individuals. In contrast, metabolomics offers enhanced sensitivity for detecting variations in lipid metabolites within the body. Metabolomic studies have highlighted the crucial role of sphingolipid metabolism in driving lipid metabolic disturbances among patients with CHD²³. Notably, despite the modest dietary intake of sphingolipids, minute quantities of radiolabeled sphingolipid bases persistently manifest in blood, liver, and lymphatic fluids²⁴. This observation underscores the importance of not underestimating the role of sphingolipids when evaluating diet. A prospective case-cohort study within the PREDIMED trial demonstrated that the mediterranean diet can positively impact the prevention of CVD by mitigating the harmful effects of elevated plasma ceramide concentrations²⁵. Our research has refined this understanding by elucidating the precise mechanisms through which anti-inflammatory diets, by modulating sphingolipid metabolism, exerts a stabilizing influence on coronary plaques. This nuanced insight highlights the potential of anti-inflammatory dietary interventions as a promising therapeutic approach.

Ceramides are the most abundant sphingolipids in cells and are the basic building blocks of complex sphingolipids²⁶. Studies have shown that different ceramide concentrations are associated with varying degrees of predictive power for cardiovascular mortality risk^27,28, particularly Cer(16:0) and Cer(24:1), which contribute to plaque instability^28,29. Notably, plasma levels of Cer(d18:1/16:0) and Cer(d18:1/24:0) have been significantly associated with vulnerable plaques, but not with the severity of coronary artery stenosis. This suggests a stronger correlation between specific ceramides and the risk of plaque rupture, rather than with the progression of atherosclerosis^30,31. Bayesian networks analysis indicates that ceramide species at different levels play a crucial role in the interaction between dietary inflammatory potential and various plaque stability traits. Specifically, Cer(t22:1/16:1), Cer(d16:0/20:1), Cer(d24:3/15:1), and Cer(t14:1/21:0) exert varying degrees of influence on the TCFA and plaque rupture through inflammatory factors, while Cer(t20:0/18:2) and Cer(t26:1/12:1) play differential roles in the formation of lipid plaques and cholesterol crystal plaques. The dominance of distinct ceramide species and cumulative effects becomes evident as the risk of plaque instability increases, with TCFA being a pivotal point. This may be related to the increasing variety and quantity of unbalanced dietary intake. Therefore, if the functional ceramide species could be traced back to their dietary sources and the dietary origins specific to different plaque stability characteristics could be identified, it would provide a theoretical basis for enhancing plaque stability. Such insights would facilitate the development of targeted dietary interventions aimed at improving plaque stability and potentially reducing the risk of cardiovascular events.

There is growing evidence that inflammatory signals play a crucial role in regulating lipid metabolism, with ceramide serving as a key mediator in this process^32,33. Ceramide is synthesized through three main pathways, each of which is associated with inflammation. One of the key enzymes involved in this process, ceramide synthase, is upregulated during inflammatory stimuli, leading to an increase in ceramide synthesis. The sphingomyelinase (SMase) pathway generates ceramides directly from sphingomyelin, and its activity is activated by various inflammatory mediators, including TNF-α, IFN-γ, and IL-1β. Additionally, the salvage pathway provides an alternative route for ceramide synthesis. Once ceramide is generated by any of these pathways, it can activate proinflammatory transcription factors^34,35. Animal studies have shown that IL-10 can regulate ceramide metabolism in inflammatory conditions, both in vivo and in vitro³⁶. Furthermore, a positive feedback loop exists between inflammation and ceramide. For example, treatment of endothelial cells with TNF-αrapidly augments mitochondrial ROS production, primarily through ceramide-dependent signaling pathways³⁷. Research has demonstrated that the beneficial effects of the aSMase inhibitor amitriptyline on endothelial cells are mediated by a reduction in ROS formation, accompanied by an anti-inflammatory effect via inhibiting TNF-α-induced adhesion molecule expression³⁸.

Bayesian networks was used to clarify the causal pathways between dietary inflammatory potential and plaque stability. Under the influence of dietary inflammatory potential, a causal relationship was observed among ceramide species, inflammatory factors, and various plaque stability traits, forming an upstream-downstream linkage. Six ceramide species and five inflammatory were identified as key mediators in the impact of dietary inflammatory potential on plaque stability. Ceramides are recognized for inducing the expression and release of multiple inflammatory cytokines. Previous studies have confirmed that Cer induces macrophage polarization, promoting the transition from the anti-inflammatory M2 phenotype to the pro-inflammatory M1 phenotype. This transition activates the Txnip/NLRP3 inflammasome, leading to the secretion of various pro-inflammatory cytokines such as IL-1β, IL-6, TNF-α, and IFN-γ^27,39–41. Notably, IL-1β interferes with the browning of white adipose tissue, contributing to lipid metabolism disorders and, subsequently, atherosclerosis progression⁴². Some studies also point out that sphingolipid molecules like sphingosine-1-phosphate (S1P), by activating specific receptor molecules, increase the expression of inflammatory factors (TNF-α, IL-1, IL-6), leading to endothelial dysfunction, promoting inflammation, and accelerating the atherosclerosis process^43–45. The interdependence between lipid metabolism and inflammation provides a solid foundation for treating coronary artery disease based on these two ideologies.

This study systematically demonstrates that specific lipid metabolites play a key mediating role in the process through which dietary inflammatory potential influences plaque stability. We identified a panel of nine ceramide species that may serve as potential circulating biomarkers for distinguishing different dietary inflammatory statuses and assessing the risk of plaque instability. Compared with the complex DII calculated from questionnaire data, the measurement of serum ceramide levels offers a more straightforward and objective metabolic phenotyping tool for clinical use, which could aid in establishing a screening and early-warning model for high-risk patients with unstable plaques. Furthermore, this research elucidates the mechanism by which an anti-inflammatory diet enhances plaque stability through the modulation of sphingolipid metabolism, particularly by influencing ceramide synthesis pathways. These findings suggest that in the secondary prevention and rehabilitation of cardiovascular disease, nutritional interventions should extend beyond traditional macronutrient control and instead promote an overall dietary pattern with anti-inflammatory potential. This includes increasing the intake of foods rich in omega-3 fatty acids, dietary fiber, and natural antioxidants, thereby inhibiting the activity of key enzymes in ceramide synthesis, blocking its pathway toward promoting plaque instability, and ultimately improving cardiovascular outcomes in patients.

Our study has a few limitations. First, although the association between the DII and plaque stability is generally considered linear, and such dichotomization is commonly employed in nutritional epidemiology to facilitate comparisons between distinct dietary patterns, this binary classification remains an oversimplification that cannot fully capture the complexity of in vivo inflammatory states^46,47. Second, our study included only patients with confirmed coronary heart disease and did not incorporate a healthy control group. This limits the generalizability of our findings to the general population or to the stage of disease onset. Future studies should include healthy or at risk individuals for further validation. Thirdly, being a cross-sectional and observational study, it is inherently challenging to establish definitive causal conclusions and precise associations between metabolic markers and inflammation. Last, to validate causal effects, more interventional and prospective studies are warranted. our analysis, which explores the potential influence of metabolites on plaque stability from a lipid metabolomics perspective, assumes a universal lipid metabolic abnormality among CHD patients. However, this characteristic may not be universally applicable to all CHD individuals, necessitating further validation.

Conclusion

In summary, our study investigates the interplay between dietary inflammatory potential and plaque stability, elucidating the underlying mechanisms that connect these two factors. By utilizing LC-MS-based plasma metabolomics, we identified metabolic biomarkers enriched in sphingolipid metabolic pathways, which can accurately differentiate between anti-inflammatory and pro-inflammatory groups. This is the first study to explore the mechanisms by which diet contributes to decreased plaque stability in coronary heart disease patients, providing precise targets for future interventions aimed at enhancing plaque stability. Our findings lay the groundwork for developing dietary-based strategies to improve cardiovascular health.

Abbreviations

Cer

Ceramides

CHD

Coronary heart disease

CVD

Cardiovascular disease

DAG

Directed Acyclic Graph

DII

Dietary inflammatory index

FC

Fold change

KEGG

Kyoto Encyclopedia of Genes and Genomes

OCT

Optical coherence tomography

VIP

Variable Importance in the Project

Author contributions

Zhenjuan Zhao was responsible for conceptualization, methodology and writing-review&editing; Rui Wang was responsible for investigation, writing—original draft and writing-review&editing; Yini Wang was responsible for conceptualization and funding acquisition; Ping Wang, Jiaonan Ni and Ting Xiong were responsible for data collection and formal analysis; Xinrui Ma and Guojie Liu were responsible for visualization and methodology; Shaohong Fang, Huai Yu and Bo Yu were responsible for resources and data curation; Xueqin Gao and Ping Lin were responsible for project administration, supervision and funding acquisition.

Funding

This research was funded by Heilongjiang Provincial Key R&D Program (2024ZX12C29).

Data availability

The data that support the findings of this study are available on request from the corresponding author, (Ping Lin, linping_1962@163.com), upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Ethics Committee of the Second Affiliated Hospital of Harbin Medical University. (protocol code: YJSKY2022-129).