Blood hsa-miR-122-5p and hsa-miR-885-5p levels associate with fatty liver and related lipoprotein metabolism—The Young Finns Study

MicroRNAs are involved in disease development and may be utilized as biomarkers. We investigated the association of blood miRNA levels and a) fatty liver (FL), b) lipoprotein and lipid pathways involved in liver lipid accumulation and c) levels of predicted mRNA targets in general population based cohort. Blood microRNA profiling (TaqMan OpenArray), genome-wide gene expression arrays and nuclear magnetic resonance metabolomics were performed for Young Finns Study participants aged 34–49 years (n = 871). Liver fat status was assessed ultrasonographically. Levels of hsa-miR-122-5p and -885-5p were up-regulated in individuals with FL (fold change (FC) = 1.55, p = 1.36 * 10−14 and FC = 1.25, p = 4.86 * 10−4, respectively). In regression model adjusted with age, sex and BMI, hsa-miR-122-5p and -885-5p predicted FL (OR = 2.07, p = 1.29 * 10−8 and OR = 1.41, p = 0.002, respectively). Together hsa-miR-122-5p and -885-5p slightly improved the detection of FL beyond established risk factors. These miRNAs may be associated with FL formation through the regulation of lipoprotein metabolism as hsa-miR-122-5p levels associated with small VLDL, IDL, and large LDL lipoprotein subclass components, while hsa-miR-885-5p levels associated inversely with XL HDL cholesterol levels. Hsa-miR-885-5p levels correlated inversely with oxysterol-binding protein 2 (OSBPL2) expression (r = −0.143, p = 1.00 * 10−4) and suppressing the expression of this lipid receptor and sterol transporter could link hsa-miR-885-5p with HDL cholesterol levels.


Materials and Methods
The Young Finns Study. YFS is a multicenter follow-up study on cardiovascular risk from childhood to adulthood in Finland. The YFS was launched in 1980, when 3,596 children and adolescents (3-18 years old) participated in the baseline study 35 . The subjects were randomly selected from Finnish national registry among the chosen age groups and the five study districts. Thereafter, the subjects have been followed up (in 1986, 2001, 2007 and 2011) with several examinations including comprehensive risk factor assessments. The 30-year follow-up was performed in 2011, with 2,063 adults, aged 34-49 years, participating in the study. The examinations included physical measurements, blood tests, and questionnaires. Participants in the follow-up studies have been found to be more often women and older than those who dropped out, but no significant differences in risk factors have been found 35 . The present study has been approved by the 1st ethical committee of the Hospital District of Southwest Finland on September 21st, 2010 and by local ethical committees. All study subjects gave an informed consent and the study was conducted according to the principles of the Declaration of Helsinki. YFS samples for miRNA analysis (n = 992) were selected independently form liver status from individuals having the most comprehensive data on risk factor, metabonomic, transcriptomics, genome wide genotyping and other phenotyping in the follow-up studies performed in 1986, 2001, 2007 and 2011. After quality control, study population comprises of 871 individuals with successful miRNA profiling (demographics in Table 1).
Clinical and biochemical measurements. Weight and height were measured and BMI calculated as weight(kg)/(height(m)) 2 . Waist circumference was measured to the nearest 0.1 cm. Blood pressure was measured with a random zero sphygmomanometer.
Venous blood samples were drawn from the right antecubital vein after an overnight fast. For blood count analysis, whole blood was anticoagulated with EDTA. Blood cell parameters were measured by flow cytometric particle counting (cells) and photometry (Hb) using Sysmex XE-5000 and XT-2000i analyzers (Sysmex Corporation) with reagents provided by the manufacturer (Cellpack and Sulfolyser).
For the biochemical measurements, serum was separated, aliquoted and stored at − 70 °C until analysis. Serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GT), glucose, cholesterol, and triglyceride concentrations were measured with ALT, AST, GT, Glucose, Cholesterol, and Triglycerides System Reagent, (Beckman Coulter Biomedical). Apolipoprotein A1 (ApoA1), apolipoprotein B (ApoB), and C-reactive protein (CRP) were determined immunoturbidimetrically (ApoA1 and B assay reagent, Orion Diagnostica and CRP Latex reagent, Beckman Coulter Biomedical). The serum triglyceride concentration was assayed using the enzymatic glycerol kinase-glycerol phosphate oxidase method (Beckman Coulter Biomedical). Serum total cholesterol levels were measured by the enzymatic cholesterol esterase-cholesterol oxidase method (Beckman Coulter Biomedical). The same reagent was used for estimating HDL cholesterol levels after the precipitation of low-density lipoprotein (LDL) and very low-density lipoprotein (VLDL) with dextran sulfate-Mg 2+ . Serum glucose concentrations were determined by the enzymatic hexokinase method (Beckman Coulter Biomedical). All the above-mentioned assays were performed on an AU400 instrument (AU400, Olympus).
Glycated hemoglobin (HbA1c) fraction in whole blood was measured by an Abbott Architect ci8200 analyzer (Abbott Laboratories). The concentration of total hemoglobin was first determined colorimetrically, after which the concentration of HbA1c was measured immunoturbidimetrically using the microparticle agglutination inhibition method (Fisher Diagnostics). These two concentrations were used to calculate the HbA1c percentage. All the above-mentioned methods besides ALT, AST, and GT quantifications were accredited by the Finnish Accreditation Service (FINAS).

NMR metabolomics.
A high-throughput serum NMR metabolomics platform was used for absolute quantification of serum lipids and metabolites, including lipoprotein subclass distributions, fatty acids, and various small molecules such as amino acids and glycolysis precursors 36,37 (Supplementary Table 1). The analyzed 14 lipoprotein subclasses were defined based on particle size. These detailed lipoprotein subclass measurements, together with standard total lipids and apolipoproteins, provide a good basis for studying the lipid-and lipoprotein-related metabolic pathways. The NMR-based metabolic profiling has previously been used in various epidemiological and genetic studies 37 , and details of the experimentation have been described [36][37][38] . Data was available from all the individuals with successful miRNA profiling.
Liver ultrasonography. Ultrasound imaging and evaluation of hepatic fat in hepatic ultrasound scan images (taken in 2011) of the liver was performed using a validated protocol 39 and Sequoia 512 ultrasound mainframes (Acuson) with 4.0 MHz adult abdominal transducers. The diagnostic evaluation of hepatic steatosis was performed visually using a standard system by a highly trained sonographer according to liver-to-kidney contrast, parenchymal brightness, deep beam attenuation, and bright vessel walls 40 . According to these criteria, the presence of hepatic steatosis (= fatty liver, FL All ) was assessed and the participants were further classified into subjects with clearly identifiably fatty liver (FL C ) or mild fatty liver (FL M ), and normal liver (NL).
RNA isolation and quality control. Whole blood (2.5 ml) was collected into PaXgene Blood RNA Tubes (PreAnalytix). The tubes were inverted 8-10 times then stored at room temperature for at least 2 hours. The tubes were frozen (− 80 °C) and thawed overnight before RNA isolation (both miRNA and total RNA) with a PAXgene Blood microRNA Kit (Qiagen) including the DNase Set using the QiaCube. The concentrations and purity of the RNA samples were evaluated spectrophotometrically (BioPhotomer, Eppendorf). The RNA isolation process was validated by analyzing the integrity of several RNAs with the RNA 6000 Nano Chip Kit (Agilent Primary data analysis was performed with Expression Suite Software version 1.0.1. As recommended by the manufactures of the miRNA panels, RNU6, RNU44, and RNU48 were used as housekeeping small RNAs. Assays with Amplification score > 1 and Cq Confidence > 0.7 were accepted. Ninety-five samples were excluded due to a low number of miRNAs expressed (≤ 200 miRNAs per sample), and in further analysis, 243 miRNAs that were expressed in at least 2/3 of the samples were included (number of miRNAs present in detectable levels in majority of the blood samples is well in line with previous similar analysis of blood tissue 41 ). The RNA quality and functionality of the TaqMan OpenArray microRNA expression panels have been validated previously 42 . After quality control and removal of outlier miRNAs, profiling was successful on 871 samples. To correct for batch effects, the principal component analysis was performed for the miRNA expression data. The data was adjusted for 10 of the first 20 principal components from the principal component analysis.

Genome-wide expression analysis (transcriptomics).
The expression levels were analyzed with an Illumina HumanHT-12 version 4 Expression BeadChip (Illumina Inc.). Utilizing the same RNA sample for both mRNA and miRNA expression profiling, 200 ng of RNA was reverse-transcribed into cDNA and biotin-UTP-labeled using the Illumina TotalPrep RNA Amplification Kit (Ambion); 1,500 ng of cDNA was then hybridized to the Illumina HumanHT-12 v4 Expression BeadChip. The BeadChips were scanned with the Illumina iScan system. Raw illumina probe data was exported from Genomestudio and analyzed in R (http:// www.r-project.org/) using the Bioconductor (http://www.bioconductor.org/) packages. The expression data was processed using nonparametric background correction, followed by quantile normalization with control and expression probes, using the neqc function in the limma package and log2 transformation. Data processing described in more detailed in ref. 43. The expression analysis was successful in 743 of the 871 samples with a miRNA expression profile. Statistical analysis. MicroRNA expressions were compared over individuals with NL, FL M and FL C (using one-way ANOVA for normally distributed miRNAs). In order to take account of the multiple testing Bonferroni corrected p-values (p c -value) were calculated and p c -value < 0.05 (= p < 0.00021) was considered significant. For dysregulated miRNAs fold changes (FCs) were calculated for each individual sample in comparison to the average of all individuals with NL. FL index was calculated as previously described 44 and the levels of discovered miRNAs were correlated with this index using Spearman's rank-order correlation and independent association was evaluated with linear regression model adjusted with age and sex.
The independent predictors of fatty liver status (FL All vs. NL and FL C vs. NL) were researched using stepwise Akaike information criterion (AIC) logistic regression. Three different models were used as follows: Model 1: FL status (FL All /FL C vs. NL) predicted with discovered miRNAs only, (one by one forced in to the model); Model 2: age, sex and BMI added among explanatory variables of model 1; and fully adjusted Model 3 including all explanatory variables known to be associated with FL in YFS study 6 i.e., age, sex, BMI, alcohol consumption, waist circumference, age, apoB levels, triglycerides, insulin levels, systolic blood pressure, smoking, and physical activity index, (excluding liver enzymes due to high correlation with studied miRNAs). Only independent predictors of liver status remained in the final model. Similar analysis were done separately for normal weight (BMI < 25) and overweight or obese individuals (BMI > 25) as the explanatory parameters associated to FL have been shown to differ in these subpopulations 6 and also separately for men and women and individuals with or without excess of alcohol consumption (> 1.67 standard drinks per day; corresponding to 20 g of pure ethanol). In stratified analysis, only models predicting FL All vs NL were possible, due to low number of individuals with FL C .
The sensitivity and specificity of the discovered miRNAs to detect individuals with FL All or FL C was analyzed by the receiver operator curve (ROC) analysis and areas under curve (AUC) were compared between miR-NAs and the liver enzymes ALT, AST and GT. To evaluate the incremental predictive value of miRNA levels in comparison to parameters known to be associated with FL in YFS participants 6 (liver enzymes, sex, BMI, waist circumference, age, apoB levels, triglycerides, insulin levels, systolic blood pressure, smoking, alcohol consumption, and physical activity index), a continuous net classification improvement (NRI) was calculated using reclassification function in the PredictABEL R package. Only parameters with independent association to liver status were included in the model (ALT, GT, waist circumference, insulin levels and systolic blood pressure). All continuous variables were inverse-normalized. A p-value < 0.05 for continuous NRI was considered significant.
The correlation between miRNAs and metabolite levels as well as physiological features previously associated with metabolic dysfunction (list of metabolites and other characteristics shown in Supplementary Table 1) was analyzed using Spearman rank-order correlation. The independent predictors value of the FL associated miRNAs was analyzed by applying the stepwise AIC regression model including FL associated miRNAs (one by one forced into model) and sex, age, BMI, and liver status as explanatory variables when analyzing liver enzymes, and also with ALT, AST and GT when analyzing other phenotypes. All continuous variables were inverse-normalized, and individuals with ALT, AST or GT levels over the Finnish reference ranges were discarded from the analysis when other phenotypes were used as an dependent variable in the model.
The predicted mRNA target expressions were included in the analysis if they were recognized as a target of the miRNA of interest by at least two target prediction programs in miRGator v.3.0 45 . Spearman's rank-order correlations between the FCs of the miRNAs of interest and the expression their targets mRNA in transcriptomics analysis with a p < 0.05 are reported. An independent association was assessed with a same model as with metabolites. The pathways enriched by the down-regulated (p < 0.05) predicted target of these microRNAs were discovered by computing overlaps in a molecular signature database (http://www.broadinstitute.org/gsea/msigdb/annotate.jsp).
Also hsa-miR-885-5p significantly predicted the existence of FL All in stepwise logistic regression model adjusted with age, sex and BMI (OR = 1.41 95%CI = 1.13-1.77, p = 0.002). In the fully adjusted model (Model 3) hsa-miR-885-5p predicted FL All when all other risk factors were added to the model (including alcohol consumption), but the addition of insulin levels abolished the significance of the predictive value of this miRNA (OR = 1.23 95%CI = 0.94-1.60, p = 0.131) ( Table 2). In the sex, weight and alcohol usage stratified analysis, hsa-miR-885-5p was an independent predictor of FL All in the model including age, BMI and sex (no included in the sex stratified analysis) but not in the fully adjusted model, with the exception of the group of subjects with excess alcohol usage where the results remained significant (OR = 2.08, 95%CI = 1.18-3.94, p = 0.016) (Supplementary Table 2).
Associations of hsa-miR-122-5p and -855-5p with serum liver enzymes. Hsa-miR-122-5p and 855-5p both correlated with the levels of ALT, AST and GT (p c < 3.05 * 10 −6 for all). Also in stepwise linear regression models including age, sex, BMI and liver status, the individual FL associated miRNAs (hsa-miR-122-5p and -885-5p one by one) predicted significantly all liver enzyme levels ( Table 3). -miR-122-5p and -855-5p over traditional fatty liver risk factors. When comparing the utility of FL associated miRNAs in identifying individuals FL C or with FL All to commonly utilized liver enzyme levels (ALT, AST and GT), the AUC of the hsa-miR-122-5p levels from the was comparable to ALT and GT ( Fig. 2A-C). Moreover, hsa-miR-122-5p outperformed AST when detecting individuals with FL C ( Fig. 2A and C). In this ROC analysis when comparing AUCs of hsa-miR-885-5p levels to those of liver enzymes in detecting FL All all liver enzymes performed better than hsa-miR-885-5p ( Fig. 2B and C).
Correlation and adjusted (age, sex, BMI, liver status and liver enzyme levels) linear regressions between hsa-miR-122-5p and -885-5p and in silico predicted mRNA target expression levels from transcriptomics analyses are shown in Table 5. Hsa-miR-885-5p levels correlated significantly with the levels of 48 predicted mRNA targets (Supplementary Table 5). The correlations with AHNAK nucleoprotein (AHNAK), trophinin (TRO) and oxysterol binding protein like 2 (OSBPL2) were significant after multiple testing correction (p c < 0.05 i.e. p < 0.00021). In linear regression adjusted for age, sex, BMI, liver status, ALT, AST and GT, hsa-miR-885-5p predicted inversely (indicating a possible direct miRNA-induced target gene expression reduction) with GABA(A) receptor-associated protein (GABARAP), OSBPL2, arylsulfatase A (ARSA), and radial spoke 3 homolog (Chlamydomonas) (RSPH3) ( Table 5), OSBPL2 being the only mRNA with statistically significant negative Spearman correlation with hsa-miR-885-5p levels (after multiple testing correction), of which levels were independently predicted by hsa-miR-885-5p levels in the regression model. The majority of the 77 correlations (Spearman correlation p < 0.05) between hsa-miR-122-5p levels and its predicted target mRNA expressions levels were direct (Supplementary Table 6) and none were significant after multiple testing correction (p c < 0.05 i.e. p < 0.00021). In linear regression model adjusted with age, sex, BMI, liver status, ALT, AST and GT, hsa-miR-122-5p inversely predicted eukaryotic translation initiation factor 1A, the X-linked (EIF1AX) and zinc finger CCCH-type containing 10 (ZC3H10) expressions, indicating their possible direct miRNA-induced target gene expression reduction (Table 5). In pathway analysis, miRNA targets with negative association to hsa-miR-122-5p and/or -855-5p were not enriched in any specific pathways.

Discussion
We found that elevated blood levels of hsa-miR-122-5p and -885-5p were associated with ultrasonographically detected FL in individuals with and without excess alcohol consumption in a large, population-based cohort of adults. These miRNAs highly correlated with serum liver enzyme levels and FL index. In the prediction of FL C , the levels of hsa-miR-122-5p were comparable to most commonly used liver enzyme levels and both hsa-miR-122-5p and -885-5p together improved risk stratification beyond established risk factors. In the metabolomic analysis, hsa-miR-122-5p levels were found to independently predict apoB, small VLD, IDL, and large LDL component levels, while hsa-miR-885-5p levels were inversely associated with very large HDL total lipid, phospholipid and cholesterol content. Finally, the levels of hsa-miR-885-5p correlated inversely with its predicted target mRNA OSBPL2 expression, presenting a possible link between hsa-miR-885-5p and both HDL levels and it lipid content.
The association of hsa-miR-122-5p blood levels with FL, FL index and liver enzyme levels independently of age, sex, and BMI was shown here in a large clinically asymptomatic population. Levels of this miRNA remained as a significant predictor of FL when adjusting with known traditional risk factors of FL and the predictive value could be seen in normal weight and overweight individuals independently. This confirms the previous result where elevated miR-122 serum/plasma levels in individuals with NAFLD in comparison to those without has been detected small case-control profiling studies (n < 50) 18,20,34 and two large expression analyses with selected miRNAs (n~400 each; 4-5 miRNAs quantified) 18,28 . Our study also extends the data applicability to general population. Hsa-miR-122-5p is almost solely expressed in the liver, and it is the most expressed miRNA in the liver, constituting approximately 70% of liver miRNA expression 46 . During the progression of NAFLD, liver miR-122 levels have been shown to be down-regulated in mice 47,48 and humans 16 . There have also been conflicting results on the effects of alcohol consumption on miR-122 liver expression 49,50 , but it seems certain that the development of ALD also affects miR-122 expression in the human liver 17 . Even modest amount of alcohol consumption has been show to elevate serum miR-122 levels 51 . Pathological states in the liver [NAFDL 18,28 , ALD 23 cirrhosis 52 , cancer 29 , toxicity 53 and infection 31 ] have all been shown to be associated with altered serum/plasma levels of miR-122. In support to our results associating hsa-miR-122-5p levels with liver enzyme levels, the serum levels of miR-122 have been previously shown to be associated with ALT and total cholesterol in hepatitis C patients receiving a  nucleic acid-based miR-122 inhibitor 54 . Priola et al. also showed significant correlation between miR-122 and ALT, AST and GT and their results indicate that miR-122 may even regulate ALT activity 34 .
The increased levels of hsa-miR-885-5p were associated with FL. Levels of this miRNA remained an independent predictor of FL when adjusting with age, sex, BMI and most of the risk factors of FL, including alcohol consumption, but the significance was abolished when insulin levels were added to predicting model. Previous studies have indicated that hsa-miR-885-5p expression is relatively liver tissue enriched 55 but also peripheral blood mononuclear cells have been thought to express this miRNA 56 , indicating possible function in liver and in blood stream. In line with our results, up-regulation of plasma miR-885-5p levels has been reported in paracetamol induced liver toxicity 53 . In contrast to a previous finding of increased levels of miR-885-5p in liver cirrhosis and hepatocellular carcinoma in a small case control study 55 , we observed an independent and direct association between miR-885-5p and ALT, AST and GT concentrations in our wide population-based sample. In addition, we found for the first time that hsa-miR-885-5p levels were inversely correlated with large HDL cholesterol independently of age, sex, BMI, measured liver enzyme levels and liver fat status.
Blood hsa-miR-122-5p levels were comparable to serum liver enzyme levels when identifying individuals with FL. Adding hsa-miR-122-5p and -885-5p to a model containing traditional risk factors and biomarkers of FL improved the risk stratification of FL. Although the improvement was statistically significant, it was minor and, unlike in the study by Tan et al. 18 or Pirola et al. 34 in more select populations, our results do not support blood miRNA levels having considerable novel clinical value in the diagnostics of FL in general population with no clinical symptoms of FL.
Our results of the metabolomic analysis showed that hsa-miR-122-5p levels had a direct association with small VLDL, IDL, large LDL subfractions, and apoB levels, indicating that hsa-miR-122-5p is connected with cholesterol levels in a viral hepatitis-free human population. These results are well supported by animal studies, where knock-out, knock-down, and overexpression studies on rodents and non-human primates have indicated that miR-122 expression modifies serum cholesterol and triglyceride levels by controlling cholesterol synthesis and VLDL secretion, and that suppressing miR-122 expression lowers LDL levels and increases HDL levels 47,[57][58][59] . Hsa-miR-122-5p has also been detected in HDL and LDL particles 14 , and thus transport in lipoproteins may contribute to the direct associations observed between hsa-miR-122-5p and lipoprotein subfraction levels. In the present study, hsa-miR-122-5p levels did not correlate with the leucocyte count, suggesting that the measured concentrations do originate from serum. We did not find any predicted target mRNAs, which would have correlated negatively and significantly (after multiple testing correction) with hsa-miR-122-5p levels, indicating that this miRNA functions mainly in the liver.  Table 5. Correlation and adjusted linear regressions between hsa-miR-122-5p and -885-5p and in silico predicted mRNA target expression levels from transcriptomics analysis. *Statistical model: In silico predicted gene target (gene ID) expression predicted with FL associated miRNAs (one by one forced in the model) adjusted with age, sex, BMI, liver status, alanine aminotransferase (ALT), aspartate aminotransferase (AST) and gamma-glutamyltransferase (GT). Note: Individuals with ALT, AST or GT levels over the Finnish reference ranges were removed from the analysis.
In our transcriptomics analyses, we observed a significant inverse correlation between hsa-miR-885-5p and its target OSBPL2 expression. The OSBPL protein family has been shown to act as either lipid transporters or sterol sensors that control lipid metabolism by being involved in reverse cholesterol transport, vesicle transport, and cell signaling and by suppressing hepatic lipogenesis and VLDL production 60 . OSBPL2 has also been previously shown affect the metabolism of neutral lipids, possibly integrating the cellular metabolism of triglycerides with that of cholesterol 61 . Furthermore, OSBPL2 overexpression has been shown to be associated with increased cell efflux of cholesterol, apoA1, and phosphatidyl choline vesicles to serum 62 . Suppressing the expression of OSBPL2 may be hypothesized to be the biological pathway through which elevated hsa-miR-885-5p is associated with decreased levels of large HDL cholesterol.
A limitation of our study is that profiling miRNAs from peripheral blood poses a challenge in identifying the origin of the analyzed miRNAs, as blood contains miRNAs from circulatory cells but also vesicle-and protein-bound miRNA originating from various tissues. Whole blood was selected to enable gene-expression analysis from the same sample. Although clinical diagnosis of ADL or NAFLD was not available, we had detailed information about patients lifetime alcohol consumption, liver enzyme levels, disease status and ultrasonic assessment of FL. This data was used to separate subjects with FL without significant alcohol use and FL with alcohol use. The detection of fatty liver relies on ultrasonic assessment performed as a part of a scientific follow-up study, and liver biopsy (histology) was not available due to ethical reasons. According to the Finnish Red Cross, the prevalence of hepatitis B and C are 0.03% and 0.05%, respectively, among Finnish blood donors and even though we have excluded individuals with ALT, AST and GT levels over the Nordic reference ranges from transcriptomics and metabolomics analysis, we cannot exclude that some of the study subject may have had steatohepatitis, which could affect the these results. In addition, majority of individuals with FL in our population are men. Even though sex has been included in statistical analyses as covariate and sex stratified analyses were performed, possible features specific for FL in women could have been missed due to low number of women with FL. The results cannot therefore be directly generalized to older populations with equal amounts of women and men with FL. The strengths of this study are the large, well-phenotyped population-based cohort and the availability of comprehensive molecular phenotyping by serum NMR metabolomics, which enables detailed analysis of specific biological processes through which miRNAs may exert their effects. As this work is essentially descriptive, more research is needed to shed light on the mechanism how miRNAs are released into the blood and to validate the interaction of the discovered miRNAs and the their targets, for example In Vitro studies confirming the connection between hsa-miR-885-5p and OSBPL2 are needed.
In summary, we found that hsa-miR-122-5p and -885-5p are significantly up-regulated in individuals with ultrasonographically defined FL in a large Finnish population-based study cohort (YFS). The blood levels of hsa-miR-122-5p and -885-5p slightly improved FL prediction beyond established risk factors. In the metabolomics analysis, hsa-miR-122-5p levels were associated directly with levels of small VLDL, IDL, and large LDL components and hsa-miR-885-5p levels were associated inversely with extra-large HDL particle lipids and cholesterol levels. Furthermore, we found significant and inverse association between hsa-miR-885-5p and OSBPL2, a protein known to affect efflux of cholesterol.