Hypercholesterolaemia is one of the major causes of CVD (cardiovascular disease). It is associated with enhanced oxidative stress, leading to increased lipid peroxidation which in turn determines endothelial dysfunction and susceptibility to coronary vasoconstriction and atherosclerosis. Different miRNAs are involved in the pathogenesis of CVD and play an important role in inflammatory process control, therefore, together with atherogenic factors, they can stimulate atherosclerotic degeneration of the vessel walls of arteries. miR-33a and miR-33b play a pivotal role in a variety of biological processes including cholesterol homoeostasis, HDL (high-density lipoprotein)-cholesterol formation, fatty acid oxidation and insulin signalling. Our study aimed to determine whether circulating miR-33a and miR-33b expression was altered in familial hypercholesterolaemic children. Total RNA was extracted from plasma, and miR-33a and miR-33b were measured by quantitative real-time PCR. We found that miR-33a and miR-33b were significantly up-regulated in the plasma of 28 hypercholesterolaemic children compared with 25 healthy subjects (4.49±0.27-fold increase, P<0.001, and 3.21±0.39-fold increase, P<0.05 respectively), and for both miRNAs, a positive correlation with total cholesterol, LDL (low-density lipoprotein)-cholesterol, LDL-cholesterol/HDL-cholesterol ratio, apolipoprotein B, CRP (C-reactive protein) and glycaemia was found. OLS (ordinary least squares) regression analysis revealed that miR-33a was significantly affected by the presence of FH (familial hypercholesterolaemia), glycaemia and CRP (P<0.001, P<0.05 and P<0.05 respectively). The same analysis showed that miR-33b was significantly related to FH and CRP (P<0.05 and P<0.05 respectively). Although it is only explorative, the present study could be the first to point to the use of miR-33a and miR-33b as early biomarkers for cholesterol levels in childhood, once validated in independent larger cohorts.

CLINICAL PERSPECTIVES

  • miR-33a and miR-33b have been demonstrated to play a pivotal role in a variety of biological processes including collaboration with SREBP in cholesterol homoeostasis, HDL-cholesterol formation, fatty acid oxidation and insulin signalling.

  • In the present study, we found that miR-33a and miR-33b are significantly up-regulated in the plasma of hypercholesterolaemic children and positively correlate with the levels of TC, LDL-cholesterol, ApoB, CRP and glycaemia.

  • Thus miR-33a and miR-33b up-regulation could play an active role in the pathogenesis of CVD that has been already described in these children, opening a new window for a therapeutic intervention for this disease. miR-33a and miR-33b could also be employed as new prognostic markers and/or as effective therapeutic targets for CVD associated with paediatric hypercholesterolaemia.

INTRODUCTION

Cholesterol is a major component of the plasma membrane in mammalian cells. In addition to its structural requirement, cholesterol is important for other cell functions such as cell proliferation, bile acid and hormone biosynthesis [1]. Abnormal levels of cholesterol have been shown to trigger a number of pathological processes, such as atherosclerosis, and are considered to be a target of therapy in patients with CVD (cardiovascular disease) and diabetes-associated cardiometabolic diseases [2]. Risk factors for the development of CVD include dyslipidaemia, hypertension, diabetes mellitus, obesity, cigarette smoking and a family history of CVD [3].

Atherosclerosis is a chronic degenerative process that could be already present during childhood and can evolve over time to clinically complicated CVD [4]. The cardiovascular risk factors may act at any age, even in utero [5]. In children, early atherosclerosis is often asymptomatic and the only sign is an increase in TC (total cholesterol), of LDL (low-density lipoprotein)-cholesterol and of TG (triacylglycerol) and a decrease in HDL (high-density lipoprotein)-cholesterol [68]. Certainly, one of the major challenges in CVD is the identification of reliable biomarkers. miRNAs are 21–23-nt-long non-coding RNA molecules that modulate the stability and/or the translational efficiency of target mRNAs. miRNAs have been demonstrated to be involved in most biological processes, including differentiation, proliferation development, migration and apoptosis [9,10], and are also involved in the pathogenesis of a number of CVDs [1113]. miRNAs not only are intracellular molecules, but also are detectable outside the cells in body fluids (e.g. in plasma, serum, saliva, urine and milk) [14]. They are protected from RNase degradation since they are contained in small membranous vesicles (e.g. exosomes, exosome-like vesicles, apoptotic bodies and microparticles), packaged within HDL-cholesterol or linked to RNA-binding proteins [14]. It has been suggested that extracellular miRNAs have specific physiological functions, and are necessary for cell-to-cell communication [14]. Given the robust stability of miRNAs in plasma, circulating miRNAs have been used as excellent potential biomarkers for complex systemic diseases, and many studies have shown that miRNAs from biofluids can be used as risk markers for cardiovascular complications [11,15]. Different miRNAs have been implicated in the regulation of lipid metabolism and their therapeutic inhibition may lead to new approaches to treat cardiometabolic diseases, including atherosclerosis and the metabolic syndrome [16]. Among others, miR-33a and miR-33b have been shown to regulate cholesterol homoeostasis, and their chromosomal location is embedded within intronic sequences of SREBPs (sterol-regulatory-element-binding proteins), therefore the transcriptional up-regulation of SREBPs induces miR-33a and miR-33b expression [17]. SREBP transcription factors are critical regulators of cholesterol/lipid and glucose homoeostasis, since they control the expression of many cholesterogenic, lipogenic and glucose homoeostasis genes [18]. miR-33a and miR-33b have multiple targets, they have been shown to directly target the ABCA1 (ATP-binding cassette transporter A1) and ABCG1 (ATP-binding cassette transporter G1), proteins implicated in the efflux of cholesterol to generate HDL-cholesterol [19]. Moreover, miR-33a and miR-33b target IRS2 (insulin receptor substrate 2) and different fatty acid oxidation enzymes [19]. Therefore miR-33a and miR-33b up-regulation leads to a decrease in HDL-cholesterol levels, a reduction in insulin signalling and in cellular β-oxidation [19]. Since miR-33 plays a pivotal role in cholesterol and lipid metabolism, different studies evaluated miR-33 inhibition in animal studies. It has been shown that inhibition of miR-33a raised plasma HDL-cholesterol levels and protected against atherosclerosis in mice [20]; similarly, miR-33 inhibition in non-human primates, increased hepatic expression of ABCA1 and induced an increase in plasma HDL-cholesterol levels and a marked suppression of the plasma levels of VLDL (very-low-density lipoprotein)-associated TG [21]. In keeping with this, miR-33a and miR-33b were markedly increased in human carotid atherosclerotic plaques, suggesting that these miRNAs could play a pivotal role in advanced atherosclerosis in humans [22].

Given the importance of these miRNAs in cholesterol homoeostasis, the present study aimed to determine whether circulating miR-33a and miR-33b expression was altered in FH (familial hypercholesterolaemia) in children. To date, this is the first study known to evaluate circulating miRNA expression in FH in paediatric age.

MATERIALS AND METHODS

Subjects selection and classification

For the present study we considered 28 children (13 males, 15 females) 5–15 years of age affected by FH and referred to the Center of Clinic Lipid Research, Department of Pediatrics, Sapienza University of Rome [23]. The phenotype of the subjects was classified as FH on the basis of the presence of a first degree relative with hypercholesterolaemia (TC>95th age and sex-specific percentile).

Over the same study period, a control group consisting of 25 healthy children (11 males, 14 females) with BMI (body mass index) appropriate for age and gender, matched for age with the hypercholesterolaemic group, was recruited from the well-child care of Center of Clinic Lipid Research, Department of Pediatrics, Sapienza University of Rome. Subjects with acute or chronic disease, use of medication potentially affecting growth and development, nutritional status, or dietary intake, and history of alcohol consumption and smoking (where appropriate) were excluded. The existence of hypothyroidism, renal disease and malignancy, treatment with immunosuppressive drugs, connective tissue disease and acute illness represent exclusion criteria. Subjects were also excluded if they had current or recent systemic or localized infections and if they were taking non-steroidal anti-inflammatory drugs, lipid-lowering drugs and/or vitamin supplements. Clinical evidence of cardiovascular disease (assessed by clinical history, physical and instrumental examinations), diabetes mellitus, hypertension or the metabolic syndrome represented additional exclusion criteria.

At first visit, anthropometric data were measured (body weight, waist, hip and arm circumferences). The weight was measured using an electronic scale (Soehnle), and the standing height was measured with the Harpenden Stadiometer (Holtain). Systolic and diastolic blood pressure was measured by using a random zero sphygmomanometer (Hawksley and Sons); the mean of three measurements was used in the analysis. BMI was calculated as weight/height2 (kg/m2). The stage of sexual development was determined in all study participants using the grading system by Tanner for breast development in girls and genital status in boys [24].

Plasma samples and blood analyses

Venous blood samples (10 ml) were collected in EDTA-containing tubes. Blood was then centrifuged at 1200 g for 10 min at 4°C. Supernatant was collected and centrifuged at 12000 g for 10 min at 4°C. Plasma samples were stored at −80°C and were thawed on ice before use.

Plasma concentrations of lipoprotein, apolipoproteins, blood glucose and insulin were determined as described previously [23,25]. LDL-cholesterol was calculated by using Friedewald's equation. Age- and sex-specific percentiles for TC and TG reported by the Lipid Research Clinics were used as reference values in children [26]. Plasma high-sensitive CRP (C-reactive protein) concentrations were analysed by particle-enhanced turbidimetric immunoassay (C-Reactive Protein Latex High Sensitive Test, Roche Diagnostics) following the manufacturer's instructions.

Informed written consent was obtained from all subjects. The study conformed to the ethical guidelines of the Declaration of Helsinki and was approved by the Ethical Committee of Sapienza University of Rome.

RNA extraction from plasma

RNA extraction was performed from 200 μl plasma samples. miRNAs were isolated using a Total RNA Purification Plus kit (Norgen Biotek) according to the manufacturer's protocol. Briefly, 3 volumes of lysis solution and 6 μl of 2-mercaptoethanol were added to each 200 μl of plasma. In order to ameliorate miRNA extraction from cholesterol-rich plasma, 30 μl of lysis additive (Norgen) were added to all samples during extraction. As an internal control, 10 fmol of cel-miR-39a was spiked into each plasma sample after lysis, in order to prevent cel-miR-39a degradation by RNase activity present in plasma, as recommended previously [27,28]. We then followed the manufacturer's protocol to extract RNA.

miRNA qRT-PCR

miRNA levels were analysed using TaqMan qRT-PCR (quantitative real-time PCR) and quantified with an ABI Prism 7000 SDS instrument (Applied Biosystems). Primers for miR-33a, miR-33b, cel-miR-39a and the reagents for reverse transcription and quantitative PCR were all obtained from Applied Biosystems. miRNA expression levels in each sample were normalized to cel-miR-39a. Relative expression of miR-33a and miR-33b in Figure 1(A) was calculated using the comparative threshold cycles values (CT) method (2−ΔΔCT) [29]. In Figures 24, all clinical parameters were correlated to miR-33a and miR-33b ΔCTCTmiR-33a/miR-33b=CTmiR-33a/miR-33bCTcel-miR-39a). Given the logarithmic nature of the qRT-PCR CT, a decrease in ΔCT corresponds to an increase in miRNA levels in the analysed samples. Therefore data were expressed as −ΔCT, in order to obtain a positive correlation between miRNA levels and clinical parameter values.

Circulating miR-33 levels in HC compared with HS

Figure 1
Circulating miR-33 levels in HC compared with HS

(A) miR-33a and miR-33b expression up-regulation in HC compared with HS (*P<0.05 and ** P<0.001, Student's t test, results are means±S.E.M.). (B) Correlation between miR-33a and miR-33b in plasma of HS and HC (Spearman's correlation test).

Figure 1
Circulating miR-33 levels in HC compared with HS

(A) miR-33a and miR-33b expression up-regulation in HC compared with HS (*P<0.05 and ** P<0.001, Student's t test, results are means±S.E.M.). (B) Correlation between miR-33a and miR-33b in plasma of HS and HC (Spearman's correlation test).

Correlation analysis of miR-33a with serum lipids

Figure 2
Correlation analysis of miR-33a with serum lipids

Correlation of miR-33a with (A) TC, (B) LDL-cholesterol (‘LDL’), (C) ApoB, (D) LDL-cholesterol/HDL-cholesterol ratio (‘LDL/HDL’), (E) TC/HDL-cholesterol (‘TC/HDL’), (F) HDL-cholesterol (‘HDL’), (G) TG (Spearman's correlation test).

Figure 2
Correlation analysis of miR-33a with serum lipids

Correlation of miR-33a with (A) TC, (B) LDL-cholesterol (‘LDL’), (C) ApoB, (D) LDL-cholesterol/HDL-cholesterol ratio (‘LDL/HDL’), (E) TC/HDL-cholesterol (‘TC/HDL’), (F) HDL-cholesterol (‘HDL’), (G) TG (Spearman's correlation test).

Correlation analysis of miR-33b with serum lipids

Figure 3
Correlation analysis of miR-33b with serum lipids

Correlation of miR-33b with (A) TC, (B) LDL-cholesterol (‘LDL’), (C) ApoB, (D) LDL-cholesterol/HDL-cholesterol ratio (‘LDL/HDL’), (E) TC/HDL-cholesterol (‘TC/HDL’), (F) HDL-cholesterol (‘HDL’), (G) TG (Spearman's correlation test).

Figure 3
Correlation analysis of miR-33b with serum lipids

Correlation of miR-33b with (A) TC, (B) LDL-cholesterol (‘LDL’), (C) ApoB, (D) LDL-cholesterol/HDL-cholesterol ratio (‘LDL/HDL’), (E) TC/HDL-cholesterol (‘TC/HDL’), (F) HDL-cholesterol (‘HDL’), (G) TG (Spearman's correlation test).

Correlation analysis of miR-33 with glycaemia, insulin and CRP

Figure 4
Correlation analysis of miR-33 with glycaemia, insulin and CRP

Correlation of miR-33a with (A) glycaemia (GLI), (B) insulin (INS), (C) CRP (Spearman's correlation test). Correlation of miR-33b with (D) glycaemia (GLI), (E) insulin (INS), (F) CRP (Spearman's correlation test).

Figure 4
Correlation analysis of miR-33 with glycaemia, insulin and CRP

Correlation of miR-33a with (A) glycaemia (GLI), (B) insulin (INS), (C) CRP (Spearman's correlation test). Correlation of miR-33b with (D) glycaemia (GLI), (E) insulin (INS), (F) CRP (Spearman's correlation test).

Statistical analyses

Each quantitative variable was checked for normality distribution using the D’Agostino and Pearson omnibus normality test. Since most of the variables had a non-normal distribution, comparisons between groups were carried out by a performing Wilcoxon–Mann–Whitney rank test. Correlation analysis was carried out using a Spearman test. In order to measure the impact of all independent variables on the variables miR-33a and miR-33b, four different OLS (ordinary least squares) regression models were estimated applying a stepwise variable selection criterion. Models 1a and 1b took into account as dependent variable miR-33a and miR-33b respectively and the following variables as predictors: FH, BMI, sex, age, HDL-cholesterol, TG, LDL-cholesterol, GLI (glycaemia), CRP and INS (insulin). ApoB (apolipoprotein B), TC, TC/HDL-cholesterol, LDL-cholesterol/HDL-cholesterol and TG/HDL-cholesterol were excluded for multi-collinearity issues. Stepwise variables selection criteria led to restricted models as reported in the first two columns of Table 2. Models 2a and 2b were obtained in the same way or excluding the dummy variable FH. For each of the four models, we checked all of the OLS assumptions. Statistical significance was defined at P<0.05. Statistical analysis was performed in the R statistical computing environment (http://www.R-project.org/) using GraphPad Prism 5.0.

RESULTS

The clinical characteristics of hypercholesterolaemic and normocholesterolaemic children are reported in Table 1. Serum levels of cholesterol were 151.54±3.141 mg/dl in children with normocholesterolaemia and 254.90±13.3 mg/dl in those with hypercholesterolaemia (P<0.001).

Table 1
Clinical characteristics and laboratory parameters of children with and without FH

BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; HOMA-IR, homoeostasis model assessment of insulin resistance; CRP, C-reactive protein. *P<0.05, **P<0.01 and *** P<0.001 between groups. Comparisons between two groups were carried out by performing Wilcoxon–Mann–Whitney rank test for all variables with the exception of the variable ‘Sex’, for which a difference in proportion test was performed.

CharacteristicControl children (n=25) (mean±S.E.M.)Children with FH (n=28) (mean±S.E.M.)Comparison between two groups (P value)
Age (years) 8.38±0.564 8.86±0.635 0.79 
Sex (% of males) 0.44±0.099 0.46±0.094 0.86 
BMI (kg/m218.91±0.777 18.56±0.660 0.60 
SBP (mmHg) 102.65±2.974 111.30±2.548 0.06 
DBP (mmHg) 65.00±2.100 65.93±2.233 0.98 
Waist circumference (cm) 60.71±2.569 61.07±1.749 0.85 
Hip circumference (cm) 67.00±3.283 66.93±2.228 0.76 
Arm circumference (cm) 20.20±1.043 20.44±0.699 0.88 
Plasma lipids (mg/dl)    
 Total cholesterol 151.54±3.141 254.90±13.271 0.00*** 
 HDL-cholesterol 50.63±2.084 58.03±1.724 0.01** 
 Triacylglycerols 56.17±2.635 61.79±3.711 0.20 
 LDL-cholesterol 89.71±3.201 182.02±13.393 0.00*** 
 Apolipoprotein B 64.00±3.918 110.72±7.277 0.00 
Blood glucose (mg/dl) 73.79±2.348 83.11±1.581 0.00 
Insulin (μ-units/ml) 7.33±0.869 7.16±0.808 0.77 
HOMA-IR 1.39±0.206 1.56±0.310 0.39 
CRP (mg/l) 0.60±0.146 0.73±0.116 0.02 
CharacteristicControl children (n=25) (mean±S.E.M.)Children with FH (n=28) (mean±S.E.M.)Comparison between two groups (P value)
Age (years) 8.38±0.564 8.86±0.635 0.79 
Sex (% of males) 0.44±0.099 0.46±0.094 0.86 
BMI (kg/m218.91±0.777 18.56±0.660 0.60 
SBP (mmHg) 102.65±2.974 111.30±2.548 0.06 
DBP (mmHg) 65.00±2.100 65.93±2.233 0.98 
Waist circumference (cm) 60.71±2.569 61.07±1.749 0.85 
Hip circumference (cm) 67.00±3.283 66.93±2.228 0.76 
Arm circumference (cm) 20.20±1.043 20.44±0.699 0.88 
Plasma lipids (mg/dl)    
 Total cholesterol 151.54±3.141 254.90±13.271 0.00*** 
 HDL-cholesterol 50.63±2.084 58.03±1.724 0.01** 
 Triacylglycerols 56.17±2.635 61.79±3.711 0.20 
 LDL-cholesterol 89.71±3.201 182.02±13.393 0.00*** 
 Apolipoprotein B 64.00±3.918 110.72±7.277 0.00 
Blood glucose (mg/dl) 73.79±2.348 83.11±1.581 0.00 
Insulin (μ-units/ml) 7.33±0.869 7.16±0.808 0.77 
HOMA-IR 1.39±0.206 1.56±0.310 0.39 
CRP (mg/l) 0.60±0.146 0.73±0.116 0.02 

The expression of circulating miR-33a and miR-33b was assayed in plasma samples of 25 paediatric healthy subjects (HS) and 28 hypercholesterolaemic children (HC).

A significant increase in miR-33a in HC compared with HS was found (4.49±0.27-fold increase, P<0.0001) (Figure 1A, left-hand panel). miR-33b was also higher in HC than in HS (3.21±0.39-fold increase, P<0.05) (Figure 1A, right-hand panel). As expected, a very high significant correlation was found between miR-33a and miR-33b (Rs=0.91, P<0.001) (Figure 1B).

We then performed correlation analyses of miR-33a and miR-33b −ΔCT with relevant blood analysis parameters.

miR-33a correlated positively with TC (Rs=0.44, P<0.001), LDL-cholesterol (Rs=0.45, P<0.001) and ApoB (Rs=0.34, P<0.001), but it did not correlate significantly with HDL-cholesterol and TG (Figures 2A–2C, 2F and 2G). Furthermore, it correlated positively with LDL-cholesterol/HDL-cholesterol ratio (Rs=0.27, P<0.05) and the correlation with TC/HDL-cholesterol ratio was positive and nearly significant (Rs=0.26, P =  0.059) (Figures 2D and 2E).

miR-33b, similarly to miR-33a, correlated positively with TC (Rs=0.35, P<0.01), LDL-cholesterol (Rs=0.39, P<0.01) and ApoB (Rs=0.30; P<0.05), but it did not correlate significantly with HDL-cholesterol and TG (Figures 3A–3C, 3F and 3G). miR-33b, likewise, correlated positively with the LDL-cholesterol/HDL-cholesterol ratio (Rs=0.26, P<0.05) and with the TC/HDL-cholesterol ratio, although this correlation was nearly significant (Rs=0.25, P =  0.07) (Figures 3D and 3E). A positive correlation of miR-33a with glycaemia (GLI) was also found (Rs=0.45, P<0.001), but not with insulin (INS) (Figures 4A and 4B). Similarly, miR-33b correlated positively with GLI (Rs=0.46, P<0.001), but not with INS (Figures 4D and 4E). Finally, both miR-33a and miR-33b correlated positively with CRP (Rs=0.43, P<0.05, Figure 4C; Rs=0.48, P<0.05, Figure 4F).

The outcomes of the regression analysis are reported in Table 2, where only the variables retained by the restricted models are shown. All of the models are highly significant (see F statistics in Table 2, all with P<0.001). Results indicate that FH and CRP are statistically significant positive-signed regressors both for miR-33a and miR-33b, indicating a positive relationship with miRNA expression levels (P<0.001 and P<0.05 respectively for both Model 1a and Model 1b), whereas GLI seems to be significant only for miR-33a (P<0.05). If we exclude the FH variable from the regression analysis (Models 2a and 2b), the significant predictors of the dependent variable miR-33a (Model 2a) are LDL-cholesterol, GLI and CRP (P<0.05, P<0.01 and P<0.01 respectively, Table 2), all of them positive-signed, whereas the significant predictors of the dependent variable miR-33b (Model 2b) are GLI and CRP (P<0.01 and P<0.057 respectively).

Table 2
OLS regression analysis output

*P<0.05, **P<0.01, ***P<0.001. α, Type I error rate used for multiple regression statistical power calculation.

Model 1aModel 1bModel 2aModel 2b
Dependent variable…miR-33amiR-33bmiR-33amiR-33b
Independent variables     
 Intercept −13.671±1.492*** −17.264±2.924*** −17.105±1.711*** −21.563±2.222*** 
 FH 1.532±0.432*** 1.361±0.666*   
 BMI  −0.15906±0.117   
 HDL-cholesterol    0.038±0.020 
 LDL-cholesterol   0.007±0.003* 0.006±0.004 
 GLI 0.046±0.020* 0.051±0.032 0.060±0.019** 0.081±0.029** 
 INS  0.145±0.096   
 CRP 0.704±0.300* 1.310±0.505* 0.885±0.317** 1.123±0.466* 
R2 0.45 0.36 0.43 0.2951 
Adjusted R2 0.42 0.29 0.38 0.2519 
F statistic (degrees of freedom) 13.39 (3; 49)*** 5.19 (5; 47)*** 8.90 (4; 48)*** 6.84 (3; 49)*** 
Observed statistical power (α=0.01) 0.999 0.952 0.996 0.925 
Model 1aModel 1bModel 2aModel 2b
Dependent variable…miR-33amiR-33bmiR-33amiR-33b
Independent variables     
 Intercept −13.671±1.492*** −17.264±2.924*** −17.105±1.711*** −21.563±2.222*** 
 FH 1.532±0.432*** 1.361±0.666*   
 BMI  −0.15906±0.117   
 HDL-cholesterol    0.038±0.020 
 LDL-cholesterol   0.007±0.003* 0.006±0.004 
 GLI 0.046±0.020* 0.051±0.032 0.060±0.019** 0.081±0.029** 
 INS  0.145±0.096   
 CRP 0.704±0.300* 1.310±0.505* 0.885±0.317** 1.123±0.466* 
R2 0.45 0.36 0.43 0.2951 
Adjusted R2 0.42 0.29 0.38 0.2519 
F statistic (degrees of freedom) 13.39 (3; 49)*** 5.19 (5; 47)*** 8.90 (4; 48)*** 6.84 (3; 49)*** 
Observed statistical power (α=0.01) 0.999 0.952 0.996 0.925 

DISCUSSION

Our results show that the levels of both miR-33a and miR-33b were significantly higher in HC plasma than in HS. Indeed, miR-33a and miR-33b correlated positively with the levels of TC, LDL-cholesterol and ApoB. Both miRNAs correlated positively with the LDL-cholesterol/HDL-cholesterol ratio and with TC/HDL-cholesterol, albeit in the latter case statistical significance was only nearly significant for both miRNAs. Finally, our data also show a positive correlation of both miRNAs with GLI and CRP.

Regression analyses showed that FH, GLI and CRP are good predictors (significant and correctly signed) of miR-33a levels and that FH and CRP are good predictors of miR-33b levels. Once the FH variable was excluded from the analyses, miR-33a correlated with LDL-cholesterol, GLI and CRP, whereas miR-33b correlated only with GLI and CRP. These results suggest that circulating miR-33a reflects serum levels of LDL-cholesterol. In contrast, miR-33b levels are more sensitive to changes in glycaemia and CRP.

Finely controlled mechanisms regulate the intracellular cholesterol levels, through its endogenous synthesis and uptake from plasma lipoproteins; SREBP transcription factors are the major actors in this process [30]. The SREBP family comprises three subtypes, SREBP-1a and SREBP-1c, generated by alternative splicing, and SREBP-2 [30]. These proteins sense the intracellular cholesterol concentration and plasma nutrients, such as insulin concentration, and transcriptionally regulate the expression of many cholesterogenic and lipogenic genes such as HMG-CoA (3-hydroxy-3-methylglutaryl-CoA) reductase, HMG-CoA synthase, PCSk9 (proprotein convertase subtilisin/kexin-type 9) and LDL receptor. Moreover, SREBPs are self-regulated by a positive transcriptional feedback loop [30].

In humans, miR-33a is located in intron 16 of the SREBP2 gene on chromosome 22, and miR-33b is located in intron 17 of the SREBP1 gene on chromosome 17. As a consequence SREBP mRNA induction up-regulates miR-33 expression levels. Therefore the circulating increase in these miRNAs in children with FH can be ascribed to SREBP-1 and SREBP-2 transcriptional induction. A SREBP transcriptional increase has already been reported in many animal and human studies in pathological conditions, such as obesity and diabetes and also in fibroblasts derived from patients with genetic FH [18,31]. Notably, an increase in miR-33a was also reported in a previous study that analysed miRNA expression levels in adult human hyperlipidaemic sera [32].

In HS, a decrease in cholesterol and/or an increase in insulin are considered as positive stimuli for SREBP activation and transcription [18]. The cohort of HC evaluated in the present study was selected on the basis of the presence of a first degree relative with hypercholesterolaemia and a TC>95th age- and sex-specific percentile. The LDL receptor mutations were not analysed, but we cannot rule out their presence. Certainly, it is possible to exclude that the SREBP/miR-33 increase can be due to insulin increase, since these children are not obese, and do not display insulin resistance.

We can speculate that in children that display FH, the hepatocytes with mutated LDL receptors are not able to take up LDL, which consequently increases in plasma, determining TC level up-regulation. Hepatocytes, in contrast with macrophages, do not display scavenger receptors and, as a consequence, the cholesterol content of hepatocytes is low, which should stimulate SREBP gene expression and, in turn, the expression of miR-33a and miR-33b. Although the origin of plasma miR-33a and miR-33b is unclear, presumably it might quantitatively derive from the liver. Indeed, in a previous study, it was shown that in cultured fibroblasts derived from genetic FH subjects displaying LDL receptor mutations, the expression of SREBP-regulated genes such as those encoding HMG-CoA reductase, LDL receptor, SREBP-2 and SREBP-1 was higher [31].

Although CRP in this cohort of children is only modulated slightly, the regression analyses underlined a positive correlation of miR-33a and miR-33b with CRP. Since children with FH display a higher inflammation grade compared with normocholesterolaemic children [33], our results raise the possibility that the up-regulation of miR-33a and miR-33b in FH children could be due to inflammatory processes induced by hypercholesterolaemia. In keeping with this, SREBPs have also been shown to be induced by inflammatory stimuli [34]. In particular, it has been shown that SREBP-1a was highly expressed in cells of the immune systems, such as macrophages and dendritic cells, and its promoter was activated by NF-κB (nuclear factor κB) as part of the pro-inflammatory phase of the innate immune response [34].

Moreover, oxidative stress is able to induce SREBP activity and to induce a molecular pathway that finally leads to endothelial dysfunction [35]. Furthermore, in a previous study, we showed a gp91phox-mediated oxidative stress increase in these children [6]. We can therefore hypothesize that, together with inflammation, oxidative stress also might contribute to SREBP/miR-33 up-regulation in HC.

An important aspect that should be considered is that, together with LDL-cholesterol levels, cholesterol efflux capacity represents an important cellular function that affects the toxicity of these lipoproteins and determines the incidence of cardiovascular events [36]. Cholesterol efflux capacity of cells depends, along with other factors, on ABCA1 and ABCG1 membrane levels, and miR-33a and miR-33b are potent inhibitors of their expression on hepatocytes and macrophages [19]. These cells are the principal actors in systemic cholesterol homoeostasis and, in particular, in the process defined atherosclerotic plaque delipidation. Interestingly, a recent study showed that miR-33a and miR-33b are markedly up-regulated in human carotid atherosclerotic plaques compared with normal arteries [22], and could be predictive of atherosclerotic plaque formation and rupture. However, whether elevated circulating miR-33 expression levels are able to alter the cholesterol efflux capacity of macrophages in FH patients deserves further studies.

Since miR-33a and miR-33b regulate pathways controlling three of the risk factors for CVD, namely levels of HDL-cholesterol, TG and insulin signalling, [19], the circulating up-regulation of these miRNAs could be considered carefully for future studies aiming to demonstrate that their expression levels can be used as a reliable biomarker for CVD. It has already been demonstrated that miRNAs can be used as non-invasive biomarkers for early detection of asymptomatic coronary atherosclerosis in obese children with the metabolic syndrome [37]; this raises the possibility that miRNAs could be used also for the diagnosis and possible therapeutic target of CVD associated with paediatric hypercholesterolaemia in an asymptomatic stadium (or stage), although further studies should be carried out to support this hypothesis.

It must be underlined that the present study is an association study and, consequently, has some limitations. Our results should be confirmed in independent and larger cohorts of children, moreover a full screening of miRNA expression could determine whether miR-33 is the best circulating miRNA that describes FH. A possible limitation could arise from the fact that we did not identify an endogenous circulating miRNA to be used as a solid reference control. Our results were normalized to cel-miR-39a spiked-in, which significantly correlated with total RNA concentrations (P<0.01 using Spearman's correlation test). Therefore we felt confident of the reliability of cel-miR-39a as a reference control. Although the cohort of children we used was not very large, setting a Type I error rate at 0.01, we obtained a very good observed regression statistical power calculation for Models 1a and 2a (power>0.996) and acceptable for Models 1b and 2b (power>0.925). Therefore, we could exclude Type II error severe issues for all the Models analysed.

To our knowledge, this is the first study to evaluate the expression of miRNAs in plasma of FH in paediatric age; although it is only an explorative study, it could be the first evidence to point to the use of miR-33a and miR-33b as early biomarkers for cholesterol levels, once validated in independent larger cohorts.

AUTHOR CONTRIBUTION

Francesco Martino participated in patient enrolment and in plasma collection and processing, provided the study concept and participated in writing the paper and analysis interpretation. Fabrizio Carlomosti extracted RNA from plasma and performed the qRT-PCR of miR-33a and miR-33b. Marcello Arca and Anna Montali participated in plasma collection and processing, and performed the phenotypic classification of patients. Mario Picozza provided technical support. Daniele Avitabile helped in writing the paper and participated in statistical analysis. Luca Persico performed statistical analysis and helped with writing the paper. Francesco Barillà, Eliana Martino and Cristina Zanoni participated in plasma collection and processing, and performed the clinical analysis and patient enrolment. Sandro Parrotto participated in analysis interpretation and paper revision. Alessandra Magenta provided the study concept and designed the experiments, helped in RNA extraction and qRT-PCR, participated in statistical analysis, writing the paper and analysis and interpretation of the data.

We thank Dr Armando Felsani for useful technical advices.

FUNDING

This study was supported by Ministero della Salute [grant number GR-2010-2309531 (to A.M.)], and European Union Framework Programme FP7-PEOPLE-2011-CIG [grant number 294176 (to D.A.)].

Abbreviations

     
  • ABCA1

    ATP-binding cassette transporter A1

  •  
  • ABCG1

    ATP-binding cassette transporter G1

  •  
  • ApoB

    apolipoprotein B

  •  
  • BMI

    body mass index

  •  
  • CRP

    C-reactive protein

  •  
  • CVD

    cardiovascular disease

  •  
  • FH

    familial hypercholesterolaemia

  •  
  • HC

    hypercholesterolaemic children

  •  
  • HDL

    high-density lipoprotein

  •  
  • HMG-CoA

    3-hydroxy-3-methylglutaryl-CoA

  •  
  • HS

    healthy subjects

  •  
  • LDL

    low-density lipoprotein

  •  
  • OLS

    ordinary least squares

  •  
  • qRT-PCR

    quantitative real-time PCR

  •  
  • SREBP

    sterol-regulatory-element-binding protein

  •  
  • TC

    total cholesterol

  •  
  • TG

    triacylglycerol

References

References
1
Fernández
 
C.
Lobo
 
M.
del
 
V.T.
Gómez-Coronado
 
D.
Lasunción
 
M.A.
 
Cholesterol is essential for mitosis progression and its deficiency induces polyploid cell formation
Exp. Cell Res.
2004
, vol. 
300
 (pg. 
109
-
120
)
[PubMed]
2
Ikonen
 
E.
 
Mechanisms for cellular cholesterol transport: defects and human disease
Physiol. Rev.
2006
, vol. 
86
 (pg. 
1237
-
1261
)
[PubMed]
3
Upadhyay
 
R.K.
 
Emerging risk biomarkers in cardiovascular diseases and disorders
J. Lipids
2015
, vol. 
2015
 pg. 
971453
 
[PubMed]
4
Kusters
 
D.M.
Vissers
 
M.N.
Wiegman
 
A.
Kastelein
 
J.J.P.
Hutten
 
B.A.
 
Treatment of dyslipidaemia in childhood
Expert Opin. Pharmacother.
2010
, vol. 
11
 (pg. 
739
-
753
)
[PubMed]
5
Palinski
 
W.
Napoli
 
C.
 
The fetal origins of atherosclerosis: maternal hypercholesterolemia, and cholesterol-lowering or antioxidant treatment during pregnancy influence in utero programming and postnatal susceptibility to atherogenesis
FASEB J.
2002
, vol. 
16
 (pg. 
1348
-
1360
)
[PubMed]
6
Martino
 
F.
Loffredo
 
L.
Carnevale
 
R.
Sanguigni
 
V.
Martino
 
E.
Catasca
 
E.
Zanoni
 
C.
Pignatelli
 
P.
Violi
 
F.
 
Oxidative stress is associated with arterial dysfunction and enhanced intima-media thickness in children with hypercholesterolemia: the potential role of nicotinamide-adenine dinucleotide phosphate oxidase
Pediatrics
2008
, vol. 
122
 (pg. 
e648
-
e655
)
[PubMed]
7
McNeal
 
C.J.
Dajani
 
T.
Wilson
 
D.
Cassidy-Bushrow
 
A.E.
Dickerson
 
J.B.
Ory
 
M.
 
Hypercholesterolemia in youth: opportunities and obstacles to prevent premature atherosclerotic cardiovascular disease
Curr. Atheroscler. Rep.
2010
, vol. 
12
 (pg. 
20
-
28
)
[PubMed]
8
McGill
 
H.C.
McMahan
 
C.A.
Herderick
 
E.E.
Malcom
 
G.T.
Tracy
 
R.E.
Jack
 
P.
 
Origin of atherosclerosis in childhood and adolescence
Am. J. Clin. Nutr.
2000
, vol. 
72
 
Suppl.
(pg. 
1307S
-
1315S
)
[PubMed]
9
Bartel
 
D.P.
 
MicroRNAs: target recognition and regulatory functions
Cell
2009
, vol. 
136
 (pg. 
215
-
233
)
[PubMed]
10
Winter
 
J.
Jung
 
S.
Keller
 
S.
Gregory
 
R.I.
Diederichs
 
S.
 
Many roads to maturity: microRNA biogenesis pathways and their regulation
Nat. Cell Biol.
2009
, vol. 
11
 (pg. 
228
-
234
)
[PubMed]
11
Olson
 
E.N.
 
MicroRNAs as therapeutic targets and biomarkers of cardiovascular disease
Sci. Transl. Med.
2014
, vol. 
6
 pg. 
239ps3
 
[PubMed]
12
Fasanaro
 
P.
D’Alessandra
 
Y.
Magenta
 
A.
Pompilio
 
G.
Capogrossi
 
M.C.
 
MicroRNAs: promising biomarkers and therapeutic targets of acute myocardial ischemia
Curr. Vasc. Pharmacol.
2015
, vol. 
13
 (pg. 
305
-
315
)
[PubMed]
13
Magenta
 
A.
Greco
 
S.
Gaetano
 
C.
Martelli
 
F.
 
Oxidative stress and microRNAs in vascular diseases
Int. J. Mol. Sci.
2013
, vol. 
14
 (pg. 
17319
-
17346
)
[PubMed]
14
Zhu
 
H.
Fan
 
G.-C.
 
Extracellular/circulating microRNAs and their potential role in cardiovascular disease
Am. J. Cardiovasc. Dis.
2011
, vol. 
1
 (pg. 
138
-
149
)
[PubMed]
15
Creemers
 
E.E.
Tijsen
 
A.J.
Pinto
 
Y.M.
 
Circulating microRNAs: novel biomarkers and extracellular communicators in cardiovascular disease?
Circ. Res.
2012
, vol. 
110
 (pg. 
483
-
495
)
[PubMed]
16
Condorelli
 
G.
Latronico
 
M.V.G.
Dorn
 
G.W.
 
MicroRNAs in heart disease: Putative novel therapeutic targets?
Eur. Heart J.
2010
, vol. 
31
 (pg. 
649
-
658
)
17
Najafi-Shoushtari
 
S.H.
Kristo
 
F.
Li
 
Y.
Shioda
 
T.
Cohen
 
D.E.
Gerszten
 
R.E.
Näär
 
A.M.
 
MicroRNA-33 and the SREBP host genes cooperate to control cholesterol homeostasis
Science
2010
, vol. 
328
 (pg. 
1566
-
1569
)
[PubMed]
18
Shao
 
W.
Espenshade
 
P.J.
 
Expanding roles for SREBP in metabolism
Cell Metab.
2012
, vol. 
16
 (pg. 
414
-
419
)
[PubMed]
19
Fernández-Hernando
 
C.
Moore
 
K.J.
 
MicroRNA modulation of cholesterol homeostasis
Arterioscler. Thromb. Vasc. Biol.
2011
, vol. 
31
 (pg. 
2378
-
2382
)
[PubMed]
20
Rayner
 
K.J.
Sheedy
 
F.J.
Esau
 
C.C.
Hussain
 
F.N.
Temel
 
R.E.
Parathath
 
S.
van Gils
 
J.M.
Rayner
 
A.J.
Chang
 
A.N.
Suarez
 
Y.
, et al 
Antagonism of miR-33 in mice promotes reverse cholesterol transport and regression of atherosclerosis
J. Clin. Invest.
2011
, vol. 
121
 (pg. 
2921
-
2931
)
[PubMed]
21
Rayner
 
K.J.
Esau
 
C.C.
Hussain
 
F.N.
McDaniel
 
A.L.
Marshall
 
S.M.
van Gils
 
J.M.
Ray
 
T.D.
Sheedy
 
F.J.
Goedeke
 
L.
Liu
 
X.
, et al 
Inhibition of miR-33a/b in non-human primates raises plasma HDL and lowers VLDL triglycerides
Nature
2011
, vol. 
478
 (pg. 
404
-
407
)
[PubMed]
22
Karunakaran
 
D.
Thrush
 
A.B.
Nguyen
 
M.-A.
Richards
 
L.
Geoffrion
 
M.
Singaravelu
 
R.
Ramphos
 
E.
Shangari
 
P.
Ouimet
 
M.
Pezacki
 
J.P.
, et al 
Macrophage mitochondrial energy status regulates cholesterol efflux and is enhanced by anti-miR33 in atherosclerosis
Circ. Res.
2015
, vol. 
117
 (pg. 
266
-
278
)
[PubMed]
23
Campagna
 
F.
Martino
 
F.
Bifolco
 
M.
Montali
 
A.
Martino
 
E.
Morrone
 
F.
Antonini
 
R.
Cantafora
 
A.
Verna
 
R.
Arca
 
M.
 
Detection of familial hypercholesterolemia in a cohort of children with hypercholesterolemia: results of a family and DNA-based screening
Atherosclerosis
2008
, vol. 
196
 (pg. 
356
-
364
)
[PubMed]
24
Tanner
 
J.M.
 
Growth and maturation during adolescence
Nutr. Rev.
1981
, vol. 
39
 (pg. 
43
-
55
)
[PubMed]
25
Robciuc
 
M.R.
Maranghi
 
M.
Lahikainen
 
A.
Rader
 
D.
Bensadoun
 
A.
Öörni
 
K.
Metso
 
J.
Minicocci
 
I.
Ciociola
 
E.
Ceci
 
F.
, et al 
Angptl3 deficiency is associated with increased insulin sensitivity, lipoprotein lipase activity, and decreased serum free fatty acids
Arterioscler. Thromb. Vasc. Biol.
2013
, vol. 
33
 (pg. 
1706
-
1713
)
[PubMed]
26
National Heart, Lung, and Blood Institute
The Lipid Research Clinics Population Studies Data Book, U.S. Department of Health and Human Services, Public Health Service, National Institutes of Health, Bethesda
1980
27
Zampetaki
 
A.
Willeit
 
P.
Drozdov
 
I.
Kiechl
 
S.
Mayr
 
M.
 
Profiling of circulating microRNAs: from single biomarkers to re-wired networks
Cardiovasc. Res.
2012
, vol. 
93
 (pg. 
555
-
562
)
[PubMed]
28
Kroh
 
E.M.
Parkin
 
R.K.
Mitchell
 
P.S.
Tewari
 
M.
 
Analysis of circulating microRNA biomarkers in plasma and serum using quantitative reverse transcription-PCR (qRT-PCR)
Methods
2010
, vol. 
50
 (pg. 
298
-
301
)
[PubMed]
29
Livak
 
K.J.
Schmittgen
 
T.D.
 
Analysis of relative gene expression data using real-time quantitative PCR and the 2−ΔΔCT method
Methods
2001
, vol. 
25
 (pg. 
402
-
408
)
[PubMed]
30
Bengoechea-Alonso
 
M.T.
Ericsson
 
J.
 
SREBP in signal transduction: cholesterol metabolism and beyond
Curr. Opin. Cell Biol.
2007
, vol. 
19
 (pg. 
215
-
222
)
[PubMed]
31
Soufi
 
M.
Ruppert
 
V.
Kurt
 
B.
Schaefer
 
J.R.
 
The impact of severe LDL receptor mutations on SREBP-pathway regulation in homozygous familial hypercholesterolemia (FH)
Gene
2012
, vol. 
499
 (pg. 
218
-
222
)
[PubMed]
32
Simionescu
 
N.
Niculescu
 
L.S.
Sanda
 
G.M.
Margina
 
D.
Sima
 
A.V.
 
Analysis of circulating microRNAs that are specifically increased in hyperlipidemic and/or hyperglycemic sera
Mol. Biol. Rep.
2014
, vol. 
41
 (pg. 
5765
-
5773
)
[PubMed]
33
Narverud
 
I.
Ueland
 
T.
Nenseter
 
M.S.
Retterstøl
 
K.
Telle-Hansen
 
V.H.
Halvorsen
 
B.
Ose
 
L.
Aukrust
 
P.
Holven
 
K.B.
 
Children with familial hypercholesterolemia are characterized by an inflammatory imbalance between the tumor necrosis factor α system and interleukin-10
Atherosclerosis
2011
, vol. 
214
 (pg. 
163
-
168
)
[PubMed]
34
Im
 
S.-S.
Yousef
 
L.
Blaschitz
 
C.
Liu
 
J.Z.
Edwards
 
R.A.
Young
 
S.G.
Raffatellu
 
M.
Osborne
 
T.F.
 
Linking lipid metabolism to the innate immune response in macrophages through sterol regulatory element binding protein-1a
Cell Metab.
2011
, vol. 
13
 (pg. 
540
-
549
)
[PubMed]
35
Chen
 
Z.
Wen
 
L.
Martin
 
M.
Hsu
 
C.Y.
Fang
 
L.
Lin
 
F.M.
Lin
 
T.Y.
Geary
 
M.J.
Geary
 
G.G.
Zhao
 
Y.
, et al 
Oxidative stress activates endothelial innate immunity via sterol regulatory element binding protein 2 (SREBP2) transactivation of microRNA-92a
Circulation
2015
, vol. 
131
 (pg. 
805
-
814
)
[PubMed]
36
Khera
 
A.V.
Cuchel
 
M.
de la Llera-Moya
 
M.
Rodrigues
 
A.
Burke
 
M.F.
Jafri
 
K.
French
 
B.C.
Phillips
 
J.A.
Mucksavage
 
M.L.
Wilensky
 
R.L.
, et al 
Cholesterol efflux capacity, high-density lipoprotein function, and atherosclerosis
N. Engl. J. Med.
2011
, vol. 
364
 (pg. 
127
-
135
)
[PubMed]
37
Omran
 
A.
Elimam
 
D.
He
 
F.
Peng
 
J.
Yin
 
F.
 
Potential role of blood microRNAs as non-invasive biomarkers for early detection of asymptomatic coronary atherosclerosis in obese children with metabolic syndrome
Med. Hypotheses
2012
, vol. 
79
 (pg. 
889
-
893
)
[PubMed]