Upregulation of microRNA-96-5p is associated with adolescent idiopathic scoliosis and low bone mass phenotype

Bone densitometry revealed low bone mass in patients with adolescent idiopathic scoliosis (AIS) and its prognostic potential to predict curve progression. Recent studies showed differential circulating miRNAs in AIS but their diagnostic potential and links to low bone mass have not been well-documented. The present study aimed to compare miRNA profiles in bone tissues collected from AIS and non-scoliotic subjects, and to explore if the selected miRNA candidates could be useful diagnostic biomarkers for AIS. Microarray analysis identified miR-96-5p being the most upregulated among the candidates. miR-96-5p level was measured in plasma samples from 100 AIS and 52 healthy girls. Our results showed significantly higher plasma levels of miR-96-5p in AIS girls with an area under the curve (AUC) of 0.671 for diagnostic accuracy. A model that was composed of plasma miR-96-5p and patient-specific parameters (age, body weight and years since menarche) gave rise to an improved AUC of 0.752. Ingenuity Pathway Analysis (IPA) indicated functional links between bone metabolic pathways and miR-96-5p. In conclusion, differentially expressed miRNAs in AIS bone and plasma samples represented a new source of disease biomarkers and players in AIS etiopathogenesis, which required further validation study involving AIS patients of both genders with long-term follow-up.


Results
Clinical and demographic characteristics of participants. Our initial discovery cohort was composed of four severe AIS patients and four age-and ethnically matched non-scoliotic subjects who underwent spinal fusion surgery and orthopaedic bone related reconstructive surgery, respectively. Both groups were recruited for collection of iliac crest bone tissues to identify miRNA candidates by miRNA microarray (Table 1 and Supplementary Table 1). This selection strategy for non-scoliotic subjects with low bone mass being the control group here was pivotal to distinguish miRNAs that were associated with AIS from those linked solely to low BMD. In validation cohort, one hundred AIS girls (with Cobb angles ranging from 15° to 80°) and fifty-two healthy girls were recruited. Anthropometric information, aBMD from DXA and bone quality parameters from HR-pQCT were summarized in Table 2. These two groups were similar to each other in terms of age, body height, arm span, sitting height, Tanner stage of pubic hair and year since menarche. AIS subjects were shown to have significantly lower body weight and slightly earlier Tanner breast stage. DXA results showed significantly lower aBMD at femoral necks in AIS girls than healthy girls. At distal radius, HR-pQCT analysis showed significantly lower cor- Differential miRNA profile in AIS bone tissues. A volcano plot ( Fig. 1) was used to identify differentially expressed miRNAs in AIS bone. Thirty most differentially up-or down-regulated miRNAs in AIS bone tissues were summarized in Supplementary Table 2. Taken into consideration the fold change and statistical significance, miR-96-5p was selected due to its highest expression level among candidates with a 45.5-fold upregulation when compared to non-scoliotic subjects. To get more insights into the mechanistic roles of our identified miRNAs in the molecular functions and their possible physiological roles in AIS pathogenesis, systematic gene pathway and network analyses were performed. Our analyses using the Ingenuity Pathway Analysis (IPA) software suggested that miR-96-5p targeting genes were playing more roles in abnormal morphology of bone and morphogenesis of skeletal system, and they were speculated to be associated with etiopathogenesis of scoliosis (Fig. 2).
Higher plasma level of miR-96-5p in AIS. The expressions of miR-96-5p in plasma of 100 patients with AIS and 52 healthy girls were compared. Plasma levels of miR-96-5p were found to be significantly higher in the AIS group (p = 0.001) (Fig. 3a). The power of Mann-Whitney U test in detecting the difference between AIS and healthy girls was estimated to be 0.85. Area under the curve (AUC) analysis demonstrated discriminating potency of plasma miR-96-5p between AIS and healthy girls, suggesting reasonably good diagnostic potential of miR-96-5p with AUC of 0.671 (95% CI, 0.579-0.763) shown in Fig. 3b.
Logistic regression models in distinguishing AIS patients from healthy controls. Logistic regression revealed that miR-96-5p along with body weight, femoral neck aBMD and total vBMD were significant predictors for the development of AIS (Table 3). The AUC for each variable was shown in Fig. 4. Given body weight, age and year since menarche are associated with curve progression and likely with onset of AIS 42-44 , these variables were included in model 0 as the default model for further comparisons (Model 0 in Table 3 and AUC in Fig. 4h). Model 1 was composed of bone quality parameters and classical risk factors (body weight, age www.nature.com/scientificreports/ and year since menarche) and explained 0.437 (Negelkerke R 2 ) of the variances (Model 1 in Table 3 and AUC in Fig. 4i). Model 2 containing miR-96-5p and variables in Model 1 showed the highest Negelkerke R 2 of 0.461 (Model 2 in Table 3) with AUC being 0.817 (95% CI 0.722-0.912, Fig. 4j). The addition of miR-96-5p increased the Negelkerke R 2 of Model 0 with classical risk factors from 20.9% to 27.5% in Model 3 (Table 3)

Discussion
Significantly, higher expression levels of the miR-96-5p in AIS were shown in plasma samples from a retrospective case-control cohort. After excluding irrelevant or redundancy information that was leading to overfitting, a composite model composed of plasma miR-96-5p level, age, years since menarche and body weight was established and showed to have an AUC of 0.752 (95% CI 0.660-0.844) without bone qualities scan by DXA and HR-pQCT, indicating its potential to serve as clinically useful minimal invasive diagnostic biomarker for AIS. The pathomechanism underlying the abnormal bone qualities has been one of the major research areas to understand the pathogenesis of AIS 1 . The findings of current genetic studies are far from reaching a consensus on biological contribution of individual single nucleotide polymorphisms (SNPs) to AIS pathogenesis. It appears that either single or multiple SNPs can only explain onset of AIS in small proportion of patients 45,46 . In view that AIS is a complex disease, the potential of epigenetics to explain its complexity and clinical heterogeneity represents a largely unexplored research field. Previous studies which relied on whole blood cells for the identification of prognostic biomarker candidates for AIS showed limited success 28,47 . Furthermore, given that DNA methylation patterns are often tissue specific, investigations focusing on DNA methylation alterations in whole blood cell of AIS patients are unlikely to be successful to decipher AIS pathogenesis [48][49][50] . Indeed, a large cohort study on 5515 individuals indicated a lack of association between epigenetic changes in whole blood and bone mineral density 51 . Therefore, we developed a more direct approach by investigating bone tissues in our discovery phase to identify miRNA candidates, which were likely to be more biologically relevant to the abnormal bone qualities in AIS.
Emerging evidence suggested that aberrant expression of several non-coding RNAs in AIS impaired cellular activities of mesenchymal stromal cells and osteoblasts/osteocytes 41,52,53 , which may account for the observed phenotypic changes affecting bone metabolism and bone qualities in AIS. In contrast to lncRNA, the stability of miRNA in biofluids renders its superiority for being biomarkers to reflect molecular alterations in complex diseases such as AIS 54,55 . The miR-96-5p identified in the present study had controversial effects on osteogenic activity. Yang et al. reported that miR-96 promoted the osteogenic ability of murine osteoblasts and bone marrow-derived mesenchymal stem cells 56 , and inhibits the proliferation of vascular smooth muscle cells 57 . Another publication showed that higher levels of miR-96 were associated with age-related bone loss 58 . Intravenous injection of agomiR-96 to 3-month-old mice induced defective osteogenesis by targeting osterix gene and the mice manifested low bone mass 58 . This in vivo evidence was supportive of observation in our current study that overexpression of miR-96-5p was associated with low bone mass and AIS in young Chinese adolescents. Our group previously reported that miR-145-5p impaired osteocytes activities and differentiation in AIS 41 and it was a significant factor in a predictive model to predict whether the curve progressed to a Cobb angle of 40°5 9 . There are several limitations of our current study. First, our validation cohort was cross-sectional by design. Our participants were skeletally immature, and we did not follow them until their skeletal maturity to determine the predictive potential of our miRNAs and models in prediction of the risk of spinal disease progression among a symptomatic population at an early stage. Therefore, a new prospective study with long-term follow-up is warranted. Secondly, expression of miR-96-5p was reported to be regulated by oestrogens (E2), implying the likelihood of a sexual dimorphism on this regard 60 . Indeed, one could rea sonably spe culate that the elevation of miR-96-5p was influenced by the hormonal rise, which could explain why girls are more predisposed to curve progression around puberty than boys. The present study only validated the abnormal expression of miR-9 6-5p in girls and future validations in both females and males are required to clarify whether expression of miR-96-5p is abnormal in male AIS patients and whether it is differe nt between genders. Third, although abnormal miRNAs that were identified in bone tissue in AIS have been validated in our previous work (miR-145-5p) and the present study (miR-96-5p), a c ombination of multiple biomarkers from different omics layers is warranted in future work to achieve an optimal effectiveness in early detection or prognostication for AIS.
In summary, this study proposed a predictive model for AIS using pathology-associated clinical features such as low bone mass and aetiology-associated circulating miR-96-5p levels, which possessed the ability to distinguish AIS patients from healthy controls. This study addressed the outstanding clinical question for a reliable diagnosis Predicted gene pathways and network of miR-96-5p. The miRNA is represented in blue; the genes that were predicted to interact with each other are in green; the diseases that were predicted to be associated with the miRNA or genes are in light pink; and the molecular and physiological functions are in yellow. The hybrid approach of IPA and manual curations was applied to construct the network of miR-96-5p. www.nature.com/scientificreports/ system of AIS at the early presentation of AIS, which has important clinical implications for clinical decisions to prevent onset and to avoid over-treatment and frequent radio-exposure.

Methodology
Study populations. Patients undergoing surgery were recruited for microarray assay. Trabecular bone tissues were intraoperatively collected from iliac crests of AIS patients who underwent spinal fusion surgery as a part of taking autograft to enhance bony fusion. Trabecular bone tissues from non-scoliotic controls were collected from trauma cases who needed iliac crest bone tissues during orthopedics surgery. To investigate the clinical implication of circulating miRNAs, patients were recruited in a case-control cohort. Anthropometrical parameters including standing height, sitting height, body weight and arm span were measured with standard protocols. AIS was diagnosed clinically by at least two senior orthopedic surgeons and radiologically with standing full-spine posteroanterior (PA) X-ray. HR-pQCT measurement, DXA measurement and full-spine PA X-ray for Cobb angle of the major curve were performed within a month before or after the date when blood samples were taken. In the case-control cohort, one hundred AIS girls were recruited from the scoliosis clinic in Prince of Wales Hospital, Hong Kong. Fifty-two aged matched heathy girls were recruited randomly from local secondary school as control. Exclusion criteria included subjects with congenital deformities, neuromuscular diseases, autoimmune disorders, endocrine disturbances or other medical conditions that affecting bone metabolism. E thical approval in accordance with Declaration of Helsinki was obtained from Clinical Research Ethics Committee of the University and Hospital (CREC No. 2016.567). Written informed consent was ob tained from all the subjects an d their legal guardians before the examinations and measurements.

BMD and bone qualities assessment. aBMD of bilateral femoral necks was measured by DXA (XR-46;
Norland Medical Systems, Wisconsin, USA). Due to the bias caused by axial vertebral rotation in AIS, aBMD of the spine measured by DXA should not be used in AIS 61 . The detail of DXA measurement was presented in our previous study 32 . The 3-dimentional assessment of bone qualities in the non-dominant d istal radius was measured with HR-p QCT (XtremeCT I, Scanco Medical, Brüttisellen, Switzerland) according to the standard protocols 62 . Reference line was marked at the most proximal point of the inner aspect of the growth plate. Onehundred-ten computed tomography slices with a resolution of 82 μm were obtained from a segment spanning 9.02 mm starting from 5 mm proximal to the reference line.
miRNA microarray on bone biopsies. Total RNA was extracted from iliac crest bone biopsies from 4 AIS and 4 non-scoliotic subjects with TRIZOL (Life technologies, California, USA) based on manufacturer's instruction. RNA integrity was tested with Agilent Bioanalyzer 2100 (Ag ilent Technologies, Palo Alto, California, USA) to ens ure satisfactory RNA integ rity and quali ty before microarray a ssay. Differentially expressed miRNAs were identified by Agilent expression array-Human miRNA 8 × 60k (Agilent Technologies, Palo Alto, CA, USA) with 100 ng of total RNA. Agilent m icroarray could ident ify all matu re miRNA sequences i n the latest miRbas e (version 21.0) w ith hi gh sensiti vity and wide coverag e. Data processing was conducted with GeneSpring station (version 12.6).

Construction of gene pathways and networks targeted by dysregulated miRNAs in AIS.
The potential targets of miRNAs of interest, includ ing genes, molecular and physiological functions in AIS patho-

Statistical analysis.
The normality of data was tested by Shapiro-Wilk test. Normal data was presented as mean ± SD and the other was presented as median (lower quartile, upper quartile). To compare the parameters between AIS and non-scoliotic controls, independent sample t test was used for the normal data, and the Mann- www.nature.com/scientificreports/ Whitney U test was us ed for the skewed data. Analysis of covariance (ANCOVA) analysis was used to compare miRNA levels between AIS and healthy girls with adjustment to age. Backward stepwise regression was used to determine whether the plasma miRNA is a important predictor in classifying AIS from non-scoliotic subjects with adjustment for bone maturity, body size, bone qualities by HR-pQCT and aBMD by DXA. In the logistic regression, plasma miRNA levels were transformed with natural log transformation. The factor was dichotomised with criteria shown as follows: 1 as logistic score ≥ 0.5 and 0 as logistic score < 0.5. ROC curves and the binary classification tables were used to study sensitivity and specificity. Variable importance was estimated using a method described by Gevery et al. 63 P-values < 0.05 (two-tailed) indicated statistical significance. Data collection and analyses were performed by two blinded investigators independently. Analyses were per formed using R (version 4.0.0, The R Foundation, Vienna, Austria). Effect size of AIS status on expression of miR-96-5p was calculated by dividing the difference between means pertaining to AIS and healthy girls by SD of the whole validation cohort. Power of statistical testing were estimated by G*Power (ve rsion 3.1.9.6, Kiel University, Germany) with α being 0.05.

Ethics declarations.
Ethical approval in accordance with Declaration of Helsinki was obtai ned from Clinical Research Ethics Committee of the University and Hospital (CREC No. 2016.567). Written informed consent was obtained from all the subjects and their legal guardians before the examinations and measurements.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.