Chemometrics-guided plasma lipidomics insights underlying the decrease of plasma total cholesterol and LDLc in children with hypercholesterolemia following habitual intake of an EVOO-based phytosterols-enriched spreadable cream

Abstract

Background

Higher LDL cholesterol levels are associated with increased risk of cardiovascular disease, and early intervention in children and regular follow-up are keystones for its prevention. Hypercholesterolemia can be partially delayed adopting healthy lifestyles, including healthy diets such as the Mediterranean diet and physical activity. Extra-virgin olive oil (EVOO), the main edible oil in the Mediterranean diet, has demonstrated cardiovascular benefits.

Objective

The present randomized double-blind crossover clinical study aimed to evaluate the efficacy of the daily consumption of a spreadable cream based on extra virgin olive oil enriched in phytosterols (0.04–0.06 g plant sterols/kg of body weight/day) for 8 weeks separated by a wash-out period of 4 weeks on children (6–18 years old) with hypercholesterolemia.

Methods

Fifty children were included although only 23 reported at least 70% of compliance and were used for final analyses. We used a dual approach combining the determination of anthropometric and biochemical plasma parameters and the lipidomics profile of plasma samples.

Results

Spreadable cream enriched in phytosterols consumption led to a decrease in total cholesterol and low-density lipoprotein cholesterol plasma levels. Multivariate analysis of the lipidomic plasma profile indicated that phospholipids and sphingolipids were mostly involved in modulating the observed effects.

Conclusions

Overall, this study provides a multi-level integrated approach for the characterization of a novel functional food to counteract hypercholesterolemia from a nutritional perspective.

Clinical Trial Registry

Clinical Trial Registry number and website where it was obtained https://clinicaltrials.gov/ct2/show/NCT05460208

Supplementary Information

The online version contains supplementary material available at 10.1007/s11306-025-02368-3.

Introduction

Cardiovascular diseases (CVDs) constitute the major cause of death in developed countries, representing 32% of annual deaths worldwide and involve high economic costs for public health systems (Available from, 2021). High low-density lipoprotein cholesterol (LDLc) plasma levels are associated with CVDs (Cholesterol-lowering bioactive foods & nutraceuticals in pediatrics: clinical evidence of efficacy & safety., 1526). Childhood is the optimal period for early hypercholesterolemia screening, since although cardiovascular events are not common in children, they may have morphological and functional modifications in the vessel wall, since the atherosclerotic process can start very early in life (Kusters Dorothé et al., 2014). Hypercholesterolemia is usually countered by a healthy lifestyle (diet and physical activity) and cholesterol-lowering medications that prevent eventual clinical complications associated with these diseases (Ademi et al., 2023). Due to the long-term efficacy and adverse events of pharmacological interventions, public health programs claim the importance of adopting healthy diets and lifestyle habits, mainly in children and young adults, committed to the improvement of plasma lipid profile, intending to delay pharmacological interventions and the onset of clinical complications (Kaminsky et al., 2022).

Among different CVD-preventing interventions, a wide range of clinical trials highlight the effectiveness of the Mediterranean diet as a traditional balanced diet and lifestyle habits, in which extra-virgin olive oil (EVOO) plays a key role as the main culinary fat (Martínez-González et al., 2022). The health-enhancing properties of EVOO are attributed to the presence of several constituents, including unsaturated fatty acids, mostly represented by oleic acid, and other minor phytochemical constituents, such as polyphenols i.e., tyrosols, oleuropein, and other secoiridoids; tocopherols; triterpenoid acids, such as maslinic and oleanoic acids; and phytosterols such as campesterol, β-sitosterol, fucosterol, and stigmasterol (Paz et al., 2021). Phytosterols are considered efficient plant-derived hypocholesterolemic agents because of their structural similarity to cholesterol, which interferes with their absorption, biosynthesis, and metabolism in the human body. Therefore, they have recently been considered promising candidates for the treatment of hypercholesterolemia and associated complications in children (Wójcik, 2022).

Lipidomics has emerged as a powerful tool to comprehensively characterize lipid profiles and to elucidate their roles in health and disease. In this framework, phytosterols are of particular interest, as their structural similarity to cholesterol allows them to modulate lipid metabolism and influence membrane dynamics. Integrating phytosterol research into lipidomics not only provides a deeper understanding of their biological functions but also opens avenues for identifying novel biomarkers and therapeutic targets related to metabolic disorders, cardiovascular health, and nutrition.

The present interventional, crossover, clinical trial aimed to evaluate the effect of the daily intake of a spreadable cream based on EVOO enriched in phytosterols (PESC) for 8 weeks on children and adolescents with untreated hypercholesterolemia compared with the intake of a control EVOO-spreadable cream without phytosterols (CSC). Anthropometric and biochemical parameters related to cardiovascular risk and a lipidomics-based workflow, via liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS), were performed to decipher the effect of PESC administration at a lipid metabolome-wide level. Chemometric analyses and the application of multivariate statistics-based chemometric tests were used to better interpret the particular effects of either single or combined factors on the plasma lipid profile of children with hypercholesterolemia. Our data opens new perspectives for the management of hypercholesterolemia and paves the way for the design of novel functional food products.

Methods

Extra virgin olive oil-based phytosterols-enriched spreadable cream and control spreadable cream

The PESC experimental spreadable cream, based on EVOO and enriched in phytosterols, and the CSC control cream were prepared in 15 g single-dose containers using a base of 90% EVOO. Each single dose of PESC preparation was added with 1.2 g of free phytosterols (Vegapure 95E, Reference: 30,534,530, CAS: 83-48-7, BASF Nutrition & Health, Germany), whereas the CSC was the same EVOO-spreadable cream without phytosterols. The experimental dose of phytosterols was adjusted in compliance with the European regulation 2000/500/CE, which sets a maximum daily amount of 3 g/day of free plant sterols, which represents 8% of free phytosterols in the experimental spreadable cream. The EVOO-spreadable creams were prepared in ACER CAMPESTRES S.L. Table 1 shows the chemical compositions of the two spreadable creams. Phytosterols were analyzed using gas chromatography at the Instituto de la Grasa (CSIC, Sevilla), as previously described (García-González et al., 2018).

Table 1.

Chemical composition of the spreadable creams used in the study

	CSC	PESC
Energy and macronutrients
Energy, kcal	900	900
Energy, kJ	3762	3762
Fat, g	100	100
Of which saturated, g	24	24
Carbohydrates, g	0	0
Of which sugars, g	0	0
Proteins, g	0	0
Salt, g	0	0
Composition of sterols, %
Cholesterol		0.08
Brassicasterol		2.42
24-Me-Cholesterol		0.27
Campesterol		13.89
Campestanol		1.04
Stigmasterol		0.53
D7-Campesterol		0.19
D5,23-Stigmastadienol		0.00
Clerosterol		0.00
B-Sitosterol		71.49
Sitostanol		7.80
D5-Avenasterol		1.06
D5,24-Stigmastadienol		0.38
D7-Stigmastenol		0.55
D7-Avenasterol		0.30

CSC control extra virgin olive oil-spreadable cream; PESC phytosterols-enriched spreadable cream

Considering an average weight of approximately 20 kg for 6-year-old children, 40 kg for 12-year-old and 60 kg for 16-year-old, the doses of plant sterols were adjusted for weight ranges. Children weighing > 20 kg and < 40 kg received a single daily 15-g dose of spreadable cream (0.03–0.06 g plant sterols/kg of body weight/day), whereas children weighing ≥ 40 kg and up to 60 kg received two 15-g doses per day (0.04–0.06 g plant sterols/kg of body weight/day). Since both PESC and CSC were EVOO-based, participation in this study did not entail any potential risks.

Experimental design

A randomized, double-blind, crossover, controlled trial was conducted in the Reina Sofía University Hospital of Córdoba (Spain), comprising the daily consumption of PESC compared with the control CSC creams. The intervention was conducted for 5 months divided into two periods of 8 weeks separated by a wash-out period of 4 weeks, after which crossover was carried out. Allocation into the two sequences of spread cream administration was performed using a stratified block randomization method adjusted for age and sex. The randomization list was concealed in a light-proof sealed envelope. The sealed envelope was maintained by an independent statistician. Therefore, the children and researchers were blinded to treatment allocation throughout the study.

Blood pressure, body mass index, and blood sampling were performed before and after the two 8-week intervention periods (t1 and t2). The participants were instructed to return the individual spreadable containers that were not consumed during each intervention period for compliance determination. Intervention compliance was calculated as a percentage by dividing the dosages consumed by those indicated for the entire period of intervention.

The protocol was conducted following the Declaration of Helsinki and Good Clinical Practice Guidelines (Vijayananthan & Nawawi, 2008; World Medical Association, 2024). All investigators were qualified to conduct and supervise the trial. The parents or tutors of the participants provided written informed consent before study entry.

The sample size was calculated assuming a decrease in total cholesterol of 10%, an alpha error of 0.05, a type-2 Beta error of 0.2, and a loss of children of 15% throughout the study, indicating 25 children for each group were required. The clinical trial protocol has been registered previously in: https://clinicaltrials.gov/ct2/show/NCT05460208.

Participants and eligibility

Fifty-six children and adolescents (33 boys and 23 girls) were recruited from the pediatric Clinical Management Unit of the Reina Sofía University Hospital in Córdoba (Spain). The inclusion criteria were as follows: children between 6 and 18 years old who agreed to participate in the study and whose parents were willing to provide written informed consent for adherence to the protocol, diagnosed with hypercholesterolemia, and not receiving any treatment for hypercholesterolemia or any other pathology. The exclusion criteria were as follows: intake of hypocholesterolemic drugs or any other drug, suffering from other metabolic diseases, and unwillingness to sign the informed consent. Six children did not meet the inclusion criteria, and only 50 children were finally included (30 boys and 20 girls).

Children were randomly divided into two groups. Daily, children allocated to group A (n = 24; 13 boys and 11 girls) received the corresponding dose of the control CSC (depending on their body weight, as described above) for 8 weeks, whereas children in group B (n = 26; 17 boys and 9 girls) received the corresponding dose of the enriched PESC during the same period. Eight weeks after the administration of the spreadable cream, it was interrupted, and a wash-out period of 4 weeks was initiated. One boy and two girls from group A dropped out during the first period of intervention because they did not like the spread texture. Afterward, a crossover for both groups was implemented, and children in group A (n = 21; 12 boys and 9 girls) took the PESC cream, while those in group B (n = 26; 17 boys and 9 girls) took the CSC cream during the following 8 weeks. During the second intervention period, one boy from group B and three girls from group A dropped out because they did not want to be drawn again. Thus, only 18 children (12 boys and 6 girls) from Group A and 25 children (16 boys and 9 girls) from Group B remained in the study at the end of the intervention. Additionally, to find a reliable effect of phytosterol intake, we calculated adherence to the intervention compliance and for statistical data analysis we established a cut-off value to select those samples from subjects presenting at least 70% of treatment adherence. Thus, 10 and 13 children were included in groups A and B, respectively, for metabolomic data analyses. A flowchart of the study is shown in Fig. 1.

Fig. 1.

The flow chart of the study. CSC, control spreadable cream; PESC, phytosterols-enriched spreadable cream

Clinical and biochemical parameters

Body weight and height were measured using standardized procedures. Body mass index (BMI, kg/m²) and BMI z score were calculated for all participants using the standard growth percentile charts for the Spanish population (Begoña et al., 2004). Overweight and obesity were defined as BMI z scores ≥ 1 and ≥ 2, respectively, in participants under 18 years old and BMI > 25 and 30 kg/m² in those older than 18 years and children are classified as underweight if their BMI z score ≤ -2 according to the WHO guidelines (World Health Organization, 2023). Systolic and diastolic blood pressures (mmHg) and heart rate were measured with a mercury sphygmomanometer (Dinamap V-100) after a minimum of 10 min of rest in the seated position, and the average of two measurements was recorded before and after each intervention period.

Additionally, after a 12 h overnight fasting period, blood samples were taken at time points t1 and t2 (basal and after 8 weeks of intervention, respectively) of both periods to perform biochemical and metabolomics analyses. Serum and plasma were obtained by centrifugation in tubes containing a gel separator (for serum) and EDTA (for plasma) for 10 min at 3700 rpm and 4 °C. Aliquots of serum and plasma were stored at −80 °C for specific analysis.

The primary outcomes were total cholesterol (TC) and LDLc serum concentrations. TC, LDLc, high-density lipoprotein cholesterol (HDLc), triacylglycerols (TAG), glucose, urea, creatinine, uric acid, bilirubin, total proteins, ferritin, transferrin concentrations, and alanine aminotransferase (ALT), aspartate aminotransferase (AST), and gamma-glutamyl transferase (GGT) activity were analyzed in serum using standardized clinical methods with Architect c16000 and i2000SR autoanalyzers (Abbott Diagnostics®, Abbot Laboratories, Madrid, Spain). Insulin was analyzed using chemiluminescent microparticle immunoassays. All parameters were analyzed at the Clinical Analysis Laboratory of the Reina Sofia University Hospital. External and internal quality controls were performed according to the hospital protocols.

Sample preparation for lipidomic analysis

Plasma lipidomic analysis was performed as a secondary outcome. Plasma aliquots samples were thawed and stored at 4 °C until analysis, and 200 μL was mixed with 800 μL of a cold methanol/acetonitrile (1/1, v/v) solution, vortexed for 30 s, and then subjected to ultrasound-assisted extraction for 10 min. Next, 10 μL of 1% formic acid (v/v) was added and incubated at −18 ℃ overnight to allow for protein precipitation. The samples were then centrifuged for 15 min at 17,000 rpm and 4 °C. Finally, the supernatants were filtered into analytical vials using 0.22-micron cellulose syringe filters. Quality control (QC) was performed by mixing a small aliquot of each sample (10 μL) into the same vial.

Untargeted UHPLC-Orbitrap-MS lipidomics analysis of plasma samples.

The untargeted lipidomic profiling of each sample was performed using a Q Exactive™ Focus Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Scientific, Waltham, MA, USA) coupled to a Vanquish ultra-high-pressure liquid chromatography (UHPLC) pump and equipped with a HESI-II probe (Thermo Scientific, USA). The chromatographic separation was performed using a BEH C18 (2.1 × 100 mm, 1.7 µm) column maintained at 40 °C. The mobile phase consisted of 5 mM ammonium formate and 0.1% formic acid in water/methanol (95/5, v/v; solution A) and 5 mM ammonium formate and 0.1% formic acid in 2-propanol/methanol/water (65/30/5, v/v/v; solution B). The linear gradient and flow rate increased, considering time (min), %B, flow rate (µL/min): (0, 10, 200), (5, 50, 200), (15, 80, 250), (28, 100, 250), (30, 100, 250), (30.9, 10, 250), and (35, 10, 250). The data acquisition was performed in full-scan MS mode using both positive and negative ionization at m/z 200 with a mass resolution of 70,000. The automatic gain control target (AGC target) was 106, and the maximum injection time (IT) was 100 ms. Furthermore, randomized injections of QC samples were performed in a data-dependent (Top N = 3) MS/MS mode with a full-scan mass resolution of 17,500 at m/z 200, an AGC target value of 105, a maximum IT of 100 ms, and an isolation window of 1.0 m/z. The injection volume was 5 µL and the m/z range for the full scan analysis was 150–1500 m/z. The Top N ions were selected for further fragmentation (data-dependent MS/MS workflow) under stepped normalized collision energies (10, 20, and 40 eV). The parameters of the heated electrospray ionization (HESI) were spray voltage 3.5 kV for HESI + and 2.8 kV for HESI-, capillary temperature 320 °C, sheath gas flow 30 arbitrary units (arb) and auxiliary gas flow 10 arb. The mass spectrometer was calibrated using positive and negative ion calibration solutions (Pierce™; Thermo Fisher Scientific, San Jose, CA, USA). To avoid possible bias, the injection sequence was randomized.

The collected RAW files were further processed using two open-source software packages, MS-DIAL (version 4.70) (Hiroshi et al., 2015) and MS-Finder (Hiroshi et al., 2016). Therefore, automatic peak finding, LOWESS normalization, and spectral matching annotation against the LipidBLAST database were performed. For peaks with a minimum peak height of 10,000 cps for HESI + and HESI- polarities, a mass range of 100–1500 m/z was inspected. In addition, the MS tolerances for peak centroiding were 0.01 Da and 0.05 Da for MS/MS. The identification step (set > 50% for annotation) was based on mass accuracy, isotopic pattern, and spectral matching, considering the usual ion adducts for lipidomics. To fill in missing peaks, gap-filling based on the peak finder algorithm was used considering a 5-ppm tolerance for m/z values. To conclude, using an identification score > 5, in in-silico fragmentation of the not fully annotated MS/MS features was achieved using MS-Finder software, considering the available database (i.e., Lipid Maps).

Chemometrics analyses, data processing and statistics

The experimental design included three factors: group (A or B), treatment (PESC for the experimental group, and CSC for the control group), and sampling time (before and after the intervention, t1 and t2, respectively). Lipidomics results were thoroughly processed using different chemometric analyses. First, to determine the influence of each factor on the lipid profile of plasma samples, unsupervised hierarchical cluster analysis (HCA) was performed using Mass Profiler Professional software (Agilent Technologies®, Santa Clara, CA, USA), as previously described (El-Nakhel et al., 2023). For this purpose, the abundance of annotated features was subjected to log₂ transformation, normalization at the 75th percentile, and baselining against the median value of all samples. The resulting hierarchical cluster was obtained based on a fold-change-based heatmap and combined with a dendrogram built on Euclidean distances by applying the Ward’s linkage rule.

Additionally, a supervised multivariate analysis based on the analysis of variance (ANOVA) multi-block orthogonal partial least squares (AMOPLS) was performed using the “rAMOPLS” package (version 0.2) in the R software (version 4.2.3), as previously described (Boccard et al., 2019). For this purpose, log₂-transformed abundances of annotated features were subjected to AMOPLS modeling to determine the ANOVA-mediated discriminant power of each factor: “treatment”, “time”, and “group” as described above, as well as their interactions. Models were built by performing 100 permutations, and 10-step stratified sub-sampling was carried out to cope with the sample population differences between groups due to the unbalanced experimental design. The results for the modeling were obtained by optimizing the smallest number of orthogonal components showing statistical significance (α = 0.01). The model interpretation was based on the relative sum of squares (RSS), which indicates the reliability of observed effects by reflecting the proportion (in percentage) of the variability attributed to each factor; residual structure ratio (RSR), which represents the ANOVA consistency of each effect concerning residuals, assuming that the highest values correlate with the highest contribution of each effect; p-value, which represents the statistical significance of each effect; principal predictive components (in percentage), which represent the major block contributions associated with each effect; and orthogonal component (in percentage), which represents the contribution of the orthogonal predictive component. Additionally, AMOPLS models were analyzed based on the variable importance in the projection (VIP), represented by VIP² markers, corresponding to the metabolites showing the strongest discriminating influence for each factor. The top-20 VIP² markers exhibiting the highest score were selected and combined with their logarithm of fold change values (logFC) and probability values to achieve pairwise comparisons by volcano analysis.

Baseline anthropometric, clinical and biochemical parameters were statistically analyzed by T test and χ² test for quantitative and qualitative variables, respectively, after checking normality of variables and homogeneity of variance to compare the two groups of intervention. Non-normal variables were log transformed and used for analysis. Finally, the results after the intervention from anthropometric, clinical, and biochemical parameters were statistically analyzed by factorial ANOVA followed by Duncan’s post hoc test. P-value < 0.05 was considered significant, using SPSS software (version 25, IBM®, Armonk, NY; US).

Results

Basal description of volunteers included in the study

Table 2 describes the anthropometric, clinical, and serum biochemical parameters of 23 children who met ≥ 70% compliance included in group A (starting with CSC) and group B (starting with PESC) before the beginning of the intervention. No statistical differences were observed in these values, except for serum HDLc serum levels, which were higher in group A than in the children included in group B. At baseline, 74% of children were normal-weight, 12% were overweight and 14% were obese children.

Table 2.

Baseline values for anthropometric, clinical and biochemical variables of children with hypercholesterolemia included in the study

	Group A (Starting with CSC), n = 10			Group B (Starting with PESC), n = 13			p value
	Mean	±	SEM	Mean	±	SEM
Gender, male n (%)	6 (60)			10 (77)			0.382
Age, years	13	±	1	11	±	1	0.183
Weight, kg	51	±	8	52	±	5	0.908
Height, m	1.5	±	0.0	1.5	±	0.0	0.549
BMI, kg/m2	20.3	±	2.0	21.7	±	1.2	0.288
BMI z-score	0.09	±	0.48	1.36	±	0.42	0.061
SBP, mmHg	119	±	5	114	±	4	0.453
DBP, mmHg	67	±	2	67	±	3	0.971
Heart rate, beats	83	±	6	83	±	5	0.976
Glucose, mg/dL	83	±	3	84	±	2	0.786
Insulin, uU/mL	8	±	1	12	±	2	0.131
Total cholesterol, mg/dL	224	±	8	213	±	11	0.445
HDLc, mg/dL	68	±	4	55	±	3	0.019
LDLc, mg/dL	141	±	7	141	±	11	0.995
Triacylglycerides, mg/dL	79	±	11	87	±	14	0.666
Urea, mg/dL	33	±	1	35	±	2	0.526
Creatinine, mg/dL	0.6	±	0.0	0.5	±	0.0	0.478
Bilirubin, mg/dL	0.8	±	0.0	0.6	±	0.0	0.203
Uric acid, mg/dL	4.2	±	0.5	5.1	±	0.3	0.084
AST, U/L	26.30	±	2.18	28.84	±	2.25	0.436
ALT, U/L	18.80	±	1.54	22.77	±	2.81	0.268
GGT, U/L	15.00	±	1.83	14.46	±	1.11	0.794
Alkaline phosphatase, U/L	219	±	33	325	±	33	0.036
Proteins, g/dL	7.3	±	0.1	7.3	±	0.1	0.946

Values are expressed as the means ± SEMs. All values were within normal ranges except total cholesterol and LDLc. T test and χ² test for quantitative and qualitative variables, respectively, were used

ALT alanine aminotransferase; ANOVA analysis of the variance; AST aspartate aminotransferase; BMI body mass index; CSC control spreadable cream; GGT gamma-glutamyl transferase; PESC phytosterol enriched spreadable cream

Clinical and biochemical parameters of selected volunteers included in lipidomic analysis

After selecting volunteers who complied with more than 70% adherence to the treatment (n = 10 for group A and n = 13 for group B), we found significant differences in the baseline for HDLc and ALP (Table 1). Similarly, we found differences between groups for HDLc, ALP and uric acid, as well as SBP (“Group × Treatment”) and LDLc (“Group × Treatment” and “Group × Time”) (Supplementary Table S1). Serum TC and LDLc levels were affected differently by the treatments, with significantly lower concentrations in the PESC-treated samples: 11.3%, p < 0.01, and 17.2%, p < 0.001 for TC and LDLc, respectively, compared to the control (Fig. 2 and Supplementary Table S1). In addition, when evaluating the “Time × Treatment” effect, we found that both TC and LDLc serum levels significantly decreased (p = 0.050 and p = 0.022, respectively) in volunteers after the intake of either PESC or CSC (Fig. 2). No significant effects were found on the rest of the anthropometric, blood pressure, and biochemical parameters due to the intake of any spreadable creams (Supplementary Table S1).

Fig. 2.

Boxplots for the effect “Time × Treatment” of selected individuals involved in the study that complied > 70% intervention. A. Total cholesterol (TC). B. Low-density lipoprotein cholesterol (LDLc). CSC, control spreadable cream; PESC, phytosterols-enriched spreadable cream. Different letters indicate significant differences between groups, according to Duncan’s post hoc test

Lipidomics combined with chemometrics analysis reveals the effect of PESC on plasmatic lipid profile

The results from the plasma lipidomic profile of subjects involved in the present clinical trial showed a total of 1,948 features effectively annotated using the database LipidBlast for all samples, from which 455 metabolites were annotated at the MS² level (165 annotated in silico) and the rest were annotated at the MS-only level. For each annotated feature, the name, annotation level, average retention time (min), average mass (u), related adducts, molecular formula, ontology, International Chemical Identifier (InChiKey), simplified molecular-input line-entry system (SMILES), annotation score, MS isotopic spectrum, MS² spectrum, and raw abundance values are listed in Supplementary Table S2. A wide variety of lipids have been reported in plasma samples, including acylcarnitines, N-acyl amino acids, sulfolipids, glycerolipids, mostly diacylglycerols (DAGs) and TAGs, and a heterogeneous range of phospholipids, lysophospholipids, and sphingolipids, mostly represented by ceramides and sphingomyelins (Supplementary Table S2). Moreover, sterol derivatives were also reported in all samples, including cholic acids, vitamin A and D esters, fatty acid esters, and acylhexosyl derivatives of plant sterols, as shown for brassicasterol, campesterol, and sitosterol, whose levels were constant in all samples.

Supplementary Table S2. Raw dataset for the lipidomic profile of plasma samples from the individuals participating in the study.

Hierarchical cluster analysis naively suggests a complex interaction between factors on the plasma lipidomic profile

First, to determine the importance of each factor (Treatment, Time, Groups, and their interactions), an unsupervised HCA was performed (Fig. 3). The combination of the fold-change-based map with the hierarchical dendrogram indicates a clear discrimination between the two groups of the trial, grouping apart their lipid profiles that are not derived from the intervention, but rather from the children’s own characteristics. In this sense, considering group A that began with CSC intake, the treatment played a differential role, showing two different subgroups between the data after the CSC and PESC interventions. In contrast, in group B, sampling time played a more differential role than treatment, as subclusters preferentially discriminated between t1 and t2, while CSC and PESC clustered together within each time point (Fig. 3).

Fig. 3.

Hierarchical cluster analysis (HCA) on the lipidomic profile of plasma samples (Euclidean distance, Ward’s linkage rule). CSC, control spreadable cream; PESC, phytosterols-enriched spreadable cream

Supervised modeling of plasma lipid profile outlines the group-depending effect of PESC administration

Supervised AMOPLS analysis was performed to decipher the influence of each factor and their interaction on plasma lipid profiles. The results from the AMOPLS modeling considering all factors are displayed in Table 3 and Fig. 4. Data indicate that three factors were found to be statistically significant: “Time” and the interactions “Group x Treatment,” and “Time x Treatment” (p < 0.05; Table 3). To gain insight into the influence of each factor, the RSS and RSR reveal that the interaction "Group x Treatment" contributes with a 14.8% in explaining the variability reported by the model as well as provides the highest RSR value (1.47), whereas “Time” represents 9.0%, and the rest of the factors exhibit a marginal contribution (RSS < 2%). In addition, treatment played a significant role in the interaction with time, reporting an RSS of 1.2% for the factor “Time x Treatment.” Moreover, a high contribution associated with the residuals (RSS = 70.9%) was observed, suggesting the significant involvement of additional non-considered factors in the modulation of the lipidomics plasma profile (Table 3).

Table 3.

AMOPLS modeling on the lipidomics profile of plasma samples for all effects and interactions, featuring the markers of the projected discrimination

Effect	RSS	RSR	p-value	Tp1	Tp2	Tp3	Tp4	Tp5	Tp6	To
Group	1.8%	1.01	0.45	3.3%	3.6%	31.2%	13.1%	12.8%	13.1%	15.3%
Time	9.0%	1.26	0.01	2.7%	79.7%	10.0%	10.5%	10.0%	10.5%	12.3%
Treatment	1.3%	1.01	0.09	3.4%	3.6%	12.5%	19.6%	13.1%	13.1%	15.4%
Group × Time	1.2%	1.01	0.22	3.4%	3.6%	12.6%	13.4%	20.2%	13.4%	15.4%
Group × Treatment	14.8%	1.47	0.01	80.5%	2.5%	8.4%	9.0%	8.6%	8.8%	10.6%
Time × Treatment	1.2%	1.01	0.04	3.3%	3.6%	12.6%	13.3%	13.0%	19.9%	15.4%
Residuals	70.9%	1.00	NA	3.4%	3.6%	12.7%	13.2%	13.0%	13.3%	15.5%

Principal VIP2 markers according to the effect “Time × Treatment”
Metabolite name	Class	Score	LogFC (t1-PESC vs t1-CSC)	p	LogFC (t2-PESC vs t2-CSC)	p	LogFC (t2-CSC vs t1-CSC)	p	LogFC (t2-PESC vs t1-PESC)	p
Cer 12:0;2O/14:0	Ceramide	7.56	−0.28	n.s	1.69	*	−1.92	**	0.04	n.s
ST 29:1;O	Brassicasterol	7.17	−0.37	n.s	0.02	n.s	−0.82	**	−0.43	n.s
SPB 17:0;3O	Phytosphingosine	7.06	−0.53	n.s	0.002	n.s	−0.70	*	−0.17	n.s
PS 45:5	Phosphatidylserine	6.89	−1.06	*	0.20	n.s	−0.61	n.s	0.64	n.s
VAE 20:3	Vitamin A fatty acid ester	6.70	−0.12	n.s	0.32	n.s	−0.70	**	−0.26	n.s
SPB 23:0;3O	Phytosphingosine	6.57	0.01	n.s	0.39	n.s	−0.43	n.s	−0.05	n.s
PE P-15:2_17:4	Ether-linked phosphatidylethanolamine	6.36	−0.56	n.s	1.58	*	−2.13	**	0.01	n.s
SPB 26:0;2O	Sphinganine	6.20	0.31	n.s	-0.22	n.s	0.19	n.s	−0.33	n.s
Phosphocreatine	Carnitine	5.81	−0.17	n.s	0.20	n.s	−0.42	*	−0.06	n.s
LPC 9:0	Lysophophatidylcholine	5.73	0.46	n.s	-0.31	n.s	0.37	n.s	−0.40	*
DAG 45:3	Diacylglycerol	5.63	−0.40	**	0.33	n.s	−0.66	n.s	0.07	n.s
SPB 19:0;3O	Phytosphingosine	5.58	−0.43	n.s	0.03	n.s	-0.44	n.s	0.02	n.s
DAG 29:3	Diacylglycerol	5.47	−0.57	n.s	0.12	n.s	−0.28	n.s	0.41	n.s
PI 41:2	Phosphatidylinositol	5.45	−0.23	n.s	1.01	n.s	−1.33	*	−0.09	n.s
NAOrn 22:6;O(FA 26:6)	N-acyl ornithine	5.41	−0.26	n.s	0.67	**	−0.36	n.s	0.57	*
BMP 12:0_28:7	Bismonoacylglycerophosphate	5.28	0.10	n.s	-0.76	*	0.25	n.s	−0.60	n.s
NAGlySer 19:2;O	N-acyl glycyl serine	5.24	−0.42	n.s	0.16	n.s	−0.73	**	−0.14	n.s
Cer 12:0;3O/8:0;(2OH)	Ceramide	5.22	0.44	n.s	−0.14	n.s	0.24	n.s	−0.34	n.s
VAE 6:0	Vitamin A fatty acid ester	5.04	−0.06	n.s	0.41	*	−0.36	n.s	0.11	n.s
DAG 33:6	Diacylglycerol	5.03	−0.37	n.s	1.26	*	−1.23	*	0.40	n.s

AMOPLS analysis of variance-multiblock orthogonal projection to latent structures discriminant analysis; BMP bismonoacylglycerophosphate; Cer ceramides; CSC control spreadable cream; DAG diacylglycerols; logFC logarithm of fold change; LPC lysophosphatidylcholine; n.s non significant; NAOrn N-acyl ornithine; PE phosphatidylethanolamine; PESC phytosterols-enriched spreadable cream; PE phosphatidylethanolamine; PI phosphatidylinositol; PS phosphatidylserine; RSR residual structure ratio; RSS relative sum of squares; SPB sphinganine; ST sterol; To orthogonal component; Tp predictive component; VAE vitamin A fatty acid esters; VIP variable importance in projection

Statistically significant differences were reported by Volcano analysis: *, p < 0.05; **, p < 0.01

Fig. 4.

Results from the AMOPLS modelling of lipidomics data considering all the study factors

The most important variables in the projection of the AMOPLS model were represented by VIP² metabolite markers, which were specifically obtained for the interaction between “Time x Treatment”, indicating their influence in the projection and the calculated logFC and statistical probability, according to t-test analysis for each pairwise comparison (Table 3). The identified VIP² markers included a large proportion (20%) of phospholipids and lysophospholipids, showing an accumulation of phospholipids for PESC at t2 (logFC = 0.20–1.58), with PE P-15:2/17:4 being significantly accumulated (p < 0.05), whereas the lysophospholipid LPC 9:0 accumulated in CSC at t2 (logFC = −0.31). Furthermore, VIP² markers also showed a large proportion of sphingolipids, characterized by a non-significant accumulation of phytosphingosines (logFC = 0.20–1.58) reported for PESC at t2 with respect to CSC, combined with a downregulation of the sphinganine SPB 26:0;2O (log FC = −0.22). At the same time, PESC promoted the accumulation of DAG at t2 with respect to the control (logFC = 0.12–1.26), and DAG 33:6 showed a significant accumulation (p < 0.05). Concerning the EVOO phytoconstituents, PESC promoted a slight accumulation of sitosterol (ST29:1;O, logFC = 0.02), as well as vitamin A fatty acid esters (VAEs) (logFC = 0.32–0.41), which was statistically significant in the case of VAE 6:0 (p < 0.05).

Discussion

The present study has demonstrated that daily supplementation with 0.03 – 0.06 g plant sterols/kg of body weight/day provided into an EVOO-based spreadable cream for 8 weeks may decrease serum cholesterol (TC and LDLc) and modify lipidic metabolites in children with untreated hypercholesterolemia. Phytosterols can be used as an important tool for the primary prevention of hypercholesterolemia and may also contribute to the prevention of cardiovascular events in patients with high-to-moderate serum levels of LDLc (Fontané et al., 2023; Gao et al., 2023; Wang et al., 2024). However, for severe hypercholesterolemia and familial hypercholesterolemia, it is necessary to combine dietary treatment with pharmacological treatment mainly statins (Turini et al., 2022). Pharmacotherapy in children and adolescents is only recommended from the age of 10 years when they do not respond positively to lifestyle modifications, particularly for those with LDLc serum levels ≥ 190 mg/dL (Jellinger et al., 2017). Previously, a daily intake of 2–3 g of phytosterols has been demonstrated to reduce LDLc levels by approximately 10–15% in adults (Musa-Veloso et al., 2011; Ruuth et al., 2020). In accordance with our results, an umbrella meta-analysis has confirmed that the daily administration of phytosterol in adults significantly reduces LDLc and TC, mainly at doses of ≥ 2 g/day and treatment duration > 8 weeks, suggesting its possible application as a complementary therapy for cardiovascular risk reduction (Wang et al., 2024). These authors also described a lower but significant beneficial effect on serum TAG that we did not find in our volunteers, probably because of the low levels found in the children included in the present study (Simin et al., 2020). On the other hand, another recent umbrella review based in meta-analyses has confirmed the LDLc-lowering effects of phytosterols but not the effect on plasma TAG and HDLc concentrations in adults, and highlight the need for long-term follow-up, well-designed clinical trials with large sample size to demonstrate the benefits of phytosterols on health (Liu et al., 2024).

Although our results corroborate the efficient plasma cholesterol-lowering effect of plant phytosterols, we did not find a significant accumulation of esters and acylhexosyl-derivatives. It is worth highlighting that these compounds were solely annotated based on their ms1, which raises the possibility of incorrect assignments. Nonetheless, sitosterol, annotated at ms2 level, was found in the plasma of children treated with the phytosterol-enriched cream. This fact is probably because the main mechanism of action of phytosterols is intestinal competition with cholesterol for micellization, transport, and absorption (Li et al., 2022), and they scarcely reach the bloodstream. Borel reported that the effectiveness of the absorption of sitosterol and campesterol is approximately 0.51–1.9% (Borel, 2003), and Li et al. agreed that the intestinal absorption rate of phytosterols is less than 5%, much lower than that of cholesterol (50–60%), and depends on many genetic, physiological, and nutritional factors (Li et al., 2022). In fact, some authors have suggested that the hypocholesterolemic effect of phytosterols may be mediated by changing the structure of the gastrointestinal microbiota (Lv et al., 2023). One limitation of the present study is that we have not analyzed microbiota changes induced by the intake of the phytosterol-enriched cream, nor metabolites derived from its metabolism. Therefore, more studies are needed to identify these potential microbiota-derived hypocholesterolemic mechanisms.

Given the complexity of the lipid profile provided by the lipidomics approach, the application of multivariate chemometrics analyses is essential to determine the biological significance of annotated metabolites and to better interpret the effect of phytosterol intake on lipid metabolism modulation in the children included in the present study. Thus, the experimental design included three factors: Group (A or B), Treatment (PESC or CSC), and sampling Time (before and after the intervention, t1 and t2, respectively), which were expected to intricately influence the lipid profile in a complex manner. The results derived from HCA support the complex interaction between all the factors involved in this study, as well as the interactions that could eventually occur among them. This evidence was mainly supported by the supervised AMOPLS model, which showed that treatment played a significant role in the plasma lipidomic profile in the interactions “Group x Treatment” and “Time x Treatment”. Indeed, the results from AMOPLS modeling show differences associated with the intervention groups (A or B), which is in line with the differences observed in the biochemical parameters abovementioned, including differences in HDLc and LDLc. In fact, it has been shown that HDL and LDL cholesterol contain their own lipidome (Kontush et al., 2013) predominantly constituted by several species of lysophosphatidilcholine, phosphatidylcholyne, and phosphatidylethanolamine among others. Lipids corresponding to such classes were among the top metabolites in the Group x Treatment interaction (Table 3).

One of the main strengths of AMOPLS is that it has individual predictive components, each focused on a specific effect or interaction. As a result, regardless of the variability explained by each factor or the significance of other factors, it allows for the independent biological interpretation of each independent factor. Therefore, we extracted that the main discriminant metabolites associated with the interaction “Time x Treatment” were phospholipids and lysophospholipids (20%), sphingolipids (20%), as well as DAGs and VAEs. The results showed the accumulation of phospholipids, phytosphingosines, DAGs, and VAEs at the end of the intervention in the PESC-treated compared with the control group. This modulation provided by PESC at the end of the intervention involves phospholipids and DAGs, two of the major constituents of lipoproteins, as well as sphingolipids, which participate in the recognition of their cell receptors (Mikhailova et al., 2019). These findings suggest the regulation of lipoprotein metabolism by PESC by interfering with their assembly and mobilization and may also be implicated in the previously described better functionality of HDLc induced by phytosterols that may contribute to the cholesterol efflux and therefore for plasma cholesterol reduction (Soltani Saba et al., 2021).

Indeed, our lipidomic results are barely significant and should motivate further research to link the plasma lipidomic profile with the observed biochemical parameters involved in the evaluation of hypercholesterolemia. To the best of our knowledge, no other study has investigated changes in lipid metabolites in humans, either children or adults, with hypercholesterolemia after supplementation with plant phytosterols. One limitation of the present study is the scarce number of children (only 23) who complied with more than 70% of the nutritional treatment. This might be reflected in the Group x Treatment significant effect as several of the participants initially assigned to groups A and B were not included in the final analysis. Furthermore, the initial values of various plasma biomarkers had a bias that made them incomparable between the two groups. Consequently, novel interventions with a large number of subjects and longer intervention time should be considered to study the effects of phytosterol-enriched diet supplementation in individuals with hypercholesterolemia.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ALT

Alanine aminotransferase

ANOVA

Analysis of variance

AMOPLS

Analysis multi-block orthogonal partial least squares

AST

Aspartate aminotransferase

AGC Target

Automatic gain control target

BMI

Body mass index

CVDs

Cardiovascular diseases

EVOO

Extra virgin olive oil

GGT

Gamma-glutamyl transferase

HESI

Heated electrospray ionization

HCA

Hierarchical cluster analysis

HDLc

High-density lipoprotein cholesterol

IT

Injection time

LC-HRMS

Liquid chromatography coupled with high-resolution mass spectrometry

logFC

Logarithm of fold change values

LDLc

Low-density lipoprotein cholesterol

QC

Quality control

PESC

Spreadable cream based on EVOO enriched in phytosterols

CSC

Spreadable cream without phytosterols

RSS

Relative sum of squares

RSR

Residual structure ratio

TC

Total cholesterol

TAG

Triacylglycerols

UHPLC

Ultra-high-pressure liquid chromatography

VIP

Variable importance in the projection

Author contributions

MGC, AG, CMA and MDM designed research; JMMC, BPV, MGC, LAVA, GR and LL conducted research; PGP, ESR and ODRH analyzed data or performed statistical analysis; PGP, LAVA, ESR, ODRH and MDM wrote paper; PGP, ODRH and MDM had primary responsibility for final content. All authors read and approved the final manuscript.

Funding

The MEDKIDS Study was supported by the Center for Industrial Technological Development (CDTI; file number: IDI-20190983). Primary Care Interventions to Prevent Maternal and Child Chronic Diseases of Perinatal and Developmental Origin (RICORS), RD21/0012/0008, Instituto de Salud Carlos III, Madrid, Spain. We acknowledge the University of Guadalajara (Mexico) for the predoctoral scholarship and also to Erasmus Santander for the international mobility scholarship of Laura Alejandra Vazquez Aguilar.

Data availability

Data described in the manuscript, code book, and analytic code will be made available upon request pending.

Declarations

Competing interests

The authors declare no competing interests.