Bioinformatic analysis and validation of microRNA‐508‐3p as a protective predictor by targeting NR4A3/MEK axis in pulmonary arterial hypertension

Abstract Pulmonary arterial hypertension (PAH) featured a debilitating progressive disorder. Here, we intend to determine diagnosis‐valuable biomarkers for PAH and decode the fundamental mechanisms of the biological function of these markers. Two mRNA microarray profiles (GSE70456 and GSE117261) and two microRNA microarray profiles (GSE55427 and GSE67597) were mined from the Gene Expression Omnibus platform. Then, we identified the differentially expressed genes (DEGs) and differentially expressed miRNAs (DEMs), respectively. Besides, we investigated online miRNA prediction tools to screen the target gene of DEMs. In this study, 185 DEGs and three common DEMs were screened as well as 1266 target genes of the three DEMs were identified. Next, 16 overlapping dysregulated genes from 185 DEGs and 1266 target gene were obtained. Meanwhile, we constructed the miRNA gene regulatory network and determined miRNA‐508‐3p‐NR4A3 pair for deeper exploring. Experiment methods verified the functional expression of miR‐508‐3p in PAH and its signalling cascade. We observed that ectopic miR‐508‐3p expression promotes proliferation and migration of pulmonary artery smooth muscle cell (PASMC). Bioinformatic, dual‐luciferase assay showed NR4A3 represents directly targeted gene of miR‐508‐3p. Mechanistically, we demonstrated that down‐regulation of miR‐508‐3p advances PASMC proliferation and migration via inducing NR4A3 to activate MAPK/ERK kinase signalling pathway. Altogether, our research provides a promising diagnosis of predictor and therapeutic avenues for patients in PAH.

developed, which deliver more tangible benefits to patients in PAH.
In recent years, microarray has been reported effective in detecting the complex network during the process of PAH, and in screening biomarkers for PAH diagnosis and prognosis. Gene Expression Omnibus (GEO) is an online database containing millions of gene profiles in various diseases, and the GEO datasets could be used to determine differentially expressed genes (DEGs) and to establish miRNA-mRNA regulatory networks. 8 In this article, the mRNA microarray datasets were processed for further exploring. Dysregulated genes in PAH samples were examined to screen vital predictors. They were investigated via bioinformatic tools: Gene Ontology (GO) annotation, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment, protein-protein interaction (PPI) network and gene set enrichment analysis (GSEA). Finally, miR-508-3p and its responsive gene NR4A3 represent the highest interaction relation, implying miR-508-3p-NR4A3 acts as an essential regulator in the progression of PAH.
miRNA, a type of non-coding RNA, is a length of approximately 22 nucleotides, which negatively regulates the expression of posttranscriptional genes by inhibiting the translation of target mRNA or by partially binding to the mRNA's 3′ untranslated region (UTR) to degrade the mRNA. 9,10 Currently, miRNAs are widely recognized to associate with biological progression in PASMCs and endothelial cells. Studies also report that the abnormally expressed miRNAs are engaged in the pathogenesis of PAH by functioning as a pro-PAH or anti-PAH factor. [11][12][13][14][15] However, the detailed mechanism of miRNA in PAH remains largely unformulated. To this end, we examined the miRNA gene inter-regulation network by using bioinformatic methods and found that miR-508-3p was down-regulated in PAH tissues. The role of miR-508-3p in PAH, however, remains relatively unknown.
It is exciting and vital that the functional expression of miR-508-3p in PAH and whether miR-508-3p could bind and regulate NR4A3 and what are the biological consequence of miR-508-3p-NR4A3 axis. For now, there are more research needed to answer these questions. In this study, we demonstrate that miR-508-3p is significantly down-regulated in PAH. Furthermore, miR-508-3p acts as a potential anti-PAH-miR that inhibits PASMC proliferation and migration by targeting NR4A3.

| Preparation and procession of raw datasets
Series matrix files of GSE70456, GSE117261, GSE55427 and GSE67597 were downloaded from the GEO (http://www.ncbi.nlm. nih.gov/geo) website. 16 Two mRNA expression profiling include 99 PAH patients and healthy controls. The microRNA (miRNA) microarray data consist of 27 PAH patients and healthy controls. All the samples we screened from the microarray are human-original sequenced tissue profiling (the detailed information about four microarray profiling see Table 1). Then, the combined data were possessed by SVA package in R (4.0.2) to correct batch effects. 17 A limmabased rule was operated to screen DEGs from two mRNA microarrays and differentially expressed miRNAs (DEMs) from two miRNA datasets. 18,19 We regarded an adjusted P < .05 and |log FC| > 1.5 as statistically significance. To make this paper more clearly, we illustrated data processed procedure in the workflow plot ( Figure 1).

| GO annotation and KEGG pathway enrichment analysis
Gene Ontology (GO) analysis was applied to deeper exploring the potential biological process (BP), molecular function (MF) and cellular component (CC) of DEGs. 20 The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of the DEGs was submitted to R to conduct functional and signalling pathway analysis. 21,22 Gene count >2 and P < .05 were considered to be statistically significant.

| Protein-protein interaction (PPI) network and Hub module identification
To systematically analyse the biological functions of DEGs, online tool STRING database (https://strin g-db.org/) was employed with a combined score >0.9 was set as the cut-off point. Then, the Cytoscape were set as follows: MCODE score ≥ 4, degree cut-off = 2, node score cut-off = 0.2, max depth = 100 and k-score = 2.

| Construction of the mRNA-miRNA interaction network
The common DEMs were acquired from two miRNA datasets, respectively. To construct and analyse the predominant miRNA-mRNA regulatory network, we investigated the online RNA-prediction tools: TargetScan (http://www.targe tscan.org/vert_72/), miRDB (http://mirdb.org/) and miWalk (http://mirwa lk.umm.unihe idelb erg. de/) to predict the target mRNAs of common DEMs. 9 Then, we combined DEGs with miRNA target genes and obtained the overlapping genes. Finally, we used Cytoscape to construct and visualize the interaction network of mRNAs and related miRNAs.

| Cell proliferation assay
Cell proliferation assays were measured using a cell counting kit-8

| Wound healing assays
For migration assays, cells were grown in a six-well plate. As the cells reached 70%-80% confluence, a scratch was performed with a micropipette tip followed by wash three times, used sterile phosphatebuffered solution (PBS), and cultured with 10% FBS with Nocodazole (HY-13520, MCE, USA) in DMEM. Then, the scratch width was observed at 0 and 48 hours and photographed by a microscope.
Image J was used to analyse scratch closure per cent.

| Western blot analysis
The total protein was harvested and lysed in RIPA buffer and was meas- at room temperature, and protein bands were analysed by Image J.

| Dual-luciferase reporter assay
The putative miR-508-3p wild-type and mutant sequences in the 3′ UTR of NR4A3 gene were cloned into a pmirGLO-firefly luciferase reporter vector (Genechem) for NR4A3. The established report vectors coupled with renilla vector, and miR-508-3p mimics or

| Animal model
Male Sprague-Dawley rats weighing (180-200 g) were purchased from Shandong University and divided into two groups (n = 5 for each group) as TA B L E 2 The primer sequences for qRT-PCR designed in this study Gene symbol Sequence (5′ to 3′)

GAPDH-R TGGTGAAGACGCCAGTGGA
Note: GAPDH was viewed as an internal control to calculate the relative gene expression. U6 was used as a reference control to calculate the relative miR-508-3p term; F, forward; R, reverse.
follows: the control group and monocrotaline (MCT) group. MCT-induced PAH model was developed by a single subcutaneous injection of MCT (60 mg/kg; C2401, Sigma-Aldrich) in a consecutive 2 weeks. 25 Control group rats received saline. Our research was conducted abided by the principles of the Laboratory Animal Ethics Committee of Shandong University.

| Statistical analysis
Data are presented as mean ± standard. Student's t test was used to analyse the differences between two groups, and more than two groups were performed with one-way ANOVA followed by a Tukey-Kramer post hoc test. All of the statistical analyses and plots were performed by GraphPad Prism 8.0. software. Image J was used to analyse cell migration and protein expression level. P-value < .05 was considered a statistically significant.

| DEGs and DEMs identified from the raw datasets
The four microarrays profiling were subjected to merge and normalize; then, DEGs and DEMs were screened using limma algorithm. In total,  (Figures 2A-D). The common intersect of DEM between the two miRNA expression profiles was shown using a Venn diagram ( Figure 2E).

| GO enrichment and KEGG pathway analysis of DEGs
To determine biological features of DEGs, GO annotation and KEGG pathway analysis were accomplished by R as follows (Table 3) KEGG pathway enrichment analysis suggested that this group of genes was enriched in glycolysis and gluconeogenesis ( Figure 3B).

| Protein-protein inter-regulatory network and miRNA-mRNA network construction
To unveil the underlying interaction relations among DEGs, the STRING online tool was applied and visualized via Cytoscape ( Figure 4A). We have identified ten critical gene signatures in the network, named IL6, CCL7, CCL20, IL1B, ICAM1, CX3CL1, CXCL1, CXCL8, NR4A3 and VEGFA. Also, we determined two crucial modules in the PPI network using MCODE, which contained 34 genes ( Figure 4C-D). Next, online bioinformatic platforms miRWalk, miRDB and TargetScan were used to predict target genes of the three commons miRNA, and a total of 1266 overlapping molecule was screened. Then, we have combined 185 DGEs and 1266 miRNA target genes to identify overlapping genes. Finally, 16 overlapping dysregulated genes were presented, and the internetwork linked these genes, and three common miRNAs were portrayed via Cytoscape ( Figure 4B). The outcome indicated that miR-508-3p plays a crucial role in the regulatory network, and NR4A3 was  (Table 4). Finally, hsa-miR-298 was up-regulated, and hsa-miR-508-3p and hsa-miR-632 were downregulated in response to PDGF stimulation, which is in agreement with bioinformatic analysis ( Figure 4E). Besides, miR-508-3p was found to be highly enriched in h-PASMC as well as its expression was markedly down-regulated, in a time-dependent manner, in response to PDGF-BB treatment. Therefore, these findings indicate that miR-508-3p may be a dominant regulator for h-PASMC proliferation. In consequence, the bio-function and complicating mechanisms of miR-508-3p-NR4A3 pair in PAH were warrant more in-depth exploring.

| Down-expression of miR-508-3p enhance PASMC proliferation and migration
To substantiate the role of miR-508-3p in PAH, we constructed PASMC/ HUVEC with overexpressing or down-regulating miR-508-3p by transfecting miR-508-3p mimics or inhibitor and analysing cell proliferation by CCK-8 assay ( Figure 5A-B). Quantitative RT-PCR showed that miR-508-3p is overexpressed in cells treated with miR-508-3p mimics and down-regulated in cells transfected with inhibitor when compared with their corresponding control cells ( Figure 5C-D). The immunocytofluorescence assay results indicated that overexpression of miR-508-3p inhibited the proliferation of PASMC ( Figure 5E-F). Besides, we examined whether down-regulating miR-508-3p would promote PASMC migration by performing wound healing assay in PASMC. The result proved that the wound healing was significantly accelerated in cells with miR-508-3p inhibitor when compared with the control cells ( Figure 5G-H).
CCK-8 assay exhibited that the down-regulation of miR-508-3p markedly promoted proliferation of PASMC, whereas the up-regulation of miR-508-3p inhibited cell growth. These data indicated that downregulation of miR-508-3p could advance the proliferation and migration of PASMC.

| NR4A3 is the target of miR-508-3p in PAH
To search for the possible target gene signatures of miR-508-3p, the online resources miMAP (https://mirmap.ezlab.org/) and miRDB were investigated. Bioinformatic evidence suggested that one seed sequence is available between the 3′UTR of NR4A3 and miR-508-3p ( Figure 6A). To demonstrate whether miR-508-3p can bind to 3′UTR of NR4A3 via the seed sequences, we constructed wild-type and mutated seed sequences in the 3′UTR of NR4A3 mRNA. The 3′UTR sequences were cloned into the Firefly luciferase reporter plasmid. Luciferase reporter experiment displayed that when PASMC/ HUVEC were cotransfected with miR-508-3p mimics and the wildtype 3′UTR, the cells had significantly lower relative luciferase intensity than the cells with miRNA-NC and the Mut-type 3′UTR, which demonstrate that wild-type miR-508-3p can bind to the seed sequences of the 3′UTR of NR4A3 mRNA to inhibit translation and intensity of luciferase ( Figure 6B). To this purpose, we performed qRT-PCR and Western blot to examine whether miR-508-3p inhibited the mRNA and protein expression of NR4A3. The result reveals that the cells overexpressing miR-508-3p show a pronounced down-regulation of NR4A3 mRNA and protein whereas the cells less expressing miR-508-3p shows up-regulated levels of NR4A3 mRNA and protein when compared with control cells in PASMC/ HUVECs ( Figure 6C-D). To further corroborate our research, MCT-PAH rat pulmonary artery tissues were used to test the mRNA and protein level of NR4A3 in vivo (n = 5), as it did align in cell experiments, the results unveil that NR4A3 represents higher expression than the control group (n = 5; Figure 6E). Next, we examined the was down-regulated as the dose of PDGF-BB was increased from 10, 50 to 100 nmol/L ( Figure 6F). Altogether, we confirmed that NR4A3 is a direct regulated target of miR-508-3p.

| miR-508-3p down-expression promotes PASMC proliferation contribute to PAH via NR4A3/ MEK pathway
To further explore the biological function of the miR-508-3p-NR4A3 axis, we used small-interfering RNA to knock down NR4A3 expression. By comparing the knockdown efficiency at the protein and mRNA level, siNR4A3#2 was selected for further study ( Figure 7A). Meanwhile, we observed mRNA and protein expression of NR4A3 by qRT-PCR and Western blot when PASMC or HUVEC were treated with miR-508-3p inhibitor and siNR4A3#2.
The result indicates that the mRNA and protein expression of NR4A3 was increased as the cells were treated with the miR-508-3p inhibitor, whereas they were decreased when these cells were transfected with siNR4A3#2 ( Figure 7C). To unveil the mechanism underlying miR-508-3p promotion of PASMC proliferation,

Slug and ERK phosphorylation (p-ERK) was increased in PASMC/
HUVEC with miR-508-3p down-expression and decreased in these cells with overexpression of miR-508-3p when compared with control cells, which is suggesting that miR-508-3p advances migration  Figure 7D). Then, the matter question is whether miR-508-3p mediates MEK signalling by inducing NR4A3. To further test our hypothesis that miR-508-3p activates MEK pathways via promoting NR4A3, we performed a rescue experiment: as merely miR-508-3p was down-regulated by miR-508-3p inhibitor, Slug expression and p-ERK were enhanced; when only NR4A3 was knockdown by siNR4A3#2, Slug expression and p-ERK were decreased; when miR-508-3p and NR4A3 were down-regulated sequentially, Slug expression and ERK phosphorylation were also inverted ( Figure 7E). Our findings suggest that NR4A3 activates MEK pathway, regulating ERK phosphorylation and Slug expression. In summary, our results demonstrate that down-regulated miR-508-3p activates MEK pathway by inducing NR4A3 to promote migration and proliferation of PASMC.

F I G U R E 4
Protein-protein interaction and mRNA-miRNA regulatory network construction. A, The interconnection of protein-protein and 'green' stand for down-expression genes while 'red' represents up-expression genes, 'triangle' shape symbolized the top ten genes. B, MicroRNA-mRNA interacts network. 'triangle' shape represented microRNA, and 'circle' shape indicated target genes. C, The first module identified in PPI, which 'green' indicated down-expression genes while 'red' represents up-expression genes, and 'triangle' shape symbolized the top ten genes. D, The second module screened in PPI, which 'triangle' shape suggested the top ten genes. E, The expression of three miRNAs in PASMC stimulated by PDGF at different time-points (n = 3, normalized to U6, **P < .01, ***P < .001, compared to 0 h. ns, no difference). PPI, protein-protein interaction and was examined via experiments in vivo and vitro method.

| D ISCUSS I ON
Previous researches have reported that miR-508-3p show tumour-suppressive roles in different types of cancers. [26][27][28][29][30] As an anti-tumour molecule, miR-508-3p is down-regulated in gastric cancer (GC) and suppressed GC cell proliferation and invasion by targeting NF-κB pathway; miR-508-3p expression also is reduced in ovarian cancer and targets MAPK signalling pathways. In this research, the screened miR-508-3p presents low expression in PAH samples. In vitro, the expression pattern of miR-508-3p in HPASMC or HUVEC was measured by qRT-PCR, with the condition that these cells were treated with PDGF-BB. As expected, miR-508-3p is downexpression in both PASMC and HUVEC treated with PDGF-BB, suggesting that miR-508-3p serves as a potential biomarker diagnosis and prognosis of PAH. In the present study, we demonstrated that less-expression of miR-508-3p promotes HPASMC proliferation and migration by gain-and loss-of-function experiment in vitro. Notably, the CCK-8 assay and scratch wound healing assay indicate that downexpression of miR-508-3p promotes proliferation and migration of HPASMC, whereas overexpression of miR-508-3p shows the inverse results. More importantly, we confirmed that miR-508-3p promotes migration and proliferation of PASMC or HUVEC through targeting NR4A3. Initially, the ectopic expression of miR-508-3p decreases the mRNA and protein levels of NR4A3 in vitro; meanwhile, the downregulated miR-508-3p enhances this gene expression. Secondly, miR-508-3p bind to the wild-type 3′UTR of NR4A3 as consolidated by dual-luciferase reporter assay and bioinformatic evidence.

F I G U R E 4 (Continued)
F I G U R E 5 miR-508-3p enhances proliferation and migration of PASMC. A-B, In CCK-8 assay, PASMC with down-expressing miR-508-3p represents high proliferation rate while PASMC transfected miR-508-3p mimics showed a significantly lower cell proliferation rate compared with the control cell (n = 3). C-D, qRT-PCR indicated that miR-508-3p represents higher or lower expression in PASMC and HUVEC transfected with miR-508-3p mimics or inhibitor compared with the control cells (n = 3). E-F, In immunocytofluorescence assay, PASMC treated with miR-508-3p NC or miR-508-3p mimics/inhibitor and cultured for 24 h, then, Ki67, a marker of cell proliferation, as shown in red and DAPI was in blue. Image-Pro Plus 6.0 was operated to measure the number of DAPI and Ki67 + nuclei. The ratio of Ki67 staining was defined as the percentage of positive nuclei within the total (DAPI staining plus Ki67 + staining) number of nuclei (n = 3). G-H, In scratch wound healing assay, PASMC with transfection of miR-508-3p inhibitor or mimics showed a significantly accelerated wound healing in down-expressed miR-508-3p cells compared with the control cells (n = 3). *P < .05, **P < .01 and ***P < .001 F I G U R E 6 miR-508-3p targets NR4A3 in PASMC/HUVEC. A, The interaction model of miR-508-3p and NR4A3. The binding seed sequence (in red) of miR-508-3p in 3′UTR of NR4A3 mRNA as predicted by the TargetScan online tool, and the mutated seed sequence (in green) in the same 3′UTR, and the seed sequence in miR-508-3p (in red). B, In dual-luciferase report assay, PASMC/HUVEC were transfected with a plasmid vector, which containing wild-type 3′UTR of NR4A3 and miR-508-3p mimics, showed a significant reduction in luciferase intensity compared with the control cells. In contrast, the cells were transfected with a plasmid vector coupled with the mutated 3′UTR of NR4A3 and miR-508-3p mimics. The luciferase signal intensity showed no reduction compared with the control cells (n = 3). From qRT-PCR and protein expression level, after PASMC (C) and HUVEC (D) transfected with miR-508-3p mimics, the expression levels of NR4A3 shown significantly low; on the contrary, the cells were transfected with miR-508-3p inhibitor, and the expression level of NR4A3 was enhanced compared with the control cells (n = 3, normalized to GAPDH). E, In MCT-PAH rat pulmonary artery tissues, NR4A3 was higher expressed at mRNA and protein level compared with the NC group (n = 5, each group, normalized to GAPDH). F, In the qRT-PCR assay, miR-508-3p was a lower expression in PASMC treated with a various dose of PDGF-BB compared with the control cells (n = 3, normalized to U6). The values labelled under protein bands represent the relative intensity of the bands. (*P < .05, **P < .01 and ***P < .001)
NR4A3 embraces different characterize of biological functions upon specific pathogenesis of diseases. [33][34][35][36][37] Based on the two mRNA datasets, NR4A3 is up-regulated in PAH samples, consistent with our findings. In vivo experiment, we found that NR4A3 was up-regulated in MCT-PAH rat model tissues and in vitro experiment. This gene was up-regulated in PASMC/HUVEC treated with miR-508-3p inhibitor, whereas NR4A3 represented downregulated cells subjected to miR-508-3p mimics compared with responsive control cells. Besides, knockdown of miR-508-3p promotes PASMC/HUVEC migration and miR-508-3p overexpression blocks these cells migration compared with responsive control cells. Therefore, we hypothesized that inhibition of miR-508-3p promotes migration and proliferation of PASMC/HUVEC by targeting NR4A3. Next, to unveil the underlying mechanism, we conducted GSEA analysis, which showed that enriched genes of KEGG datasets in PAH group mainly involving in MAPK/ERK pathway, of which MEK is reported to act as an essential mediator related to the progress of PAH. Therefore, we primarily focused on whether miR-508-3p could activate MEK signalling in PASMC/HUVEC. The outcome indicates that miR-508-3p can regulate both Slug expression F I G U R E 7 miR-508-3p down-expression contributes aggressive PASMC proliferate in PAH via activation of NR4A3/MEK pathway. A, To knock down NR4A3 expression, we tested the efficiency of candidate siRNA target NR4A3 at the mRNA and protein level and selected siRNA#2 as the most effective siRNA for further study (n = 3, compared to control). Lipo2000 means lipo-transfection reagent 2000. B, GSEA analysis showed that PAH groups were primarily involved in activating MAPK, Toll-like receptor and mTOR signalling pathways. NES: normalized enrichment score. C, PASMC and HUVEC treated with miR-NC and siRNA# 2 against NR4A3 compared with that transfected with miR-NC and siRNA-NC, the expression of NR4A3 was decreased at mRNA and protein level; as these cells were transfected with miR-508-3p inhibitor and siRNA-NC, NR4A3 expression was increased; when these cells were sequentially knockdown miR-508-3p by inhibitor and NR4A3 by siNR4A3#2, NR4A3 expression was down-regulated again (n = 3, compared to the first two groups). D, Western blotting analysis indicated that transfection of miR-508-3p mimics in PASMC and HUVEC decreased the expression level of p-ERK and Slug. However, these cells transfected with miR-508-3p inhibitors showed the reverse results compared with the control cell (n = 3, compared to NC). E, The Western blotting revealed that the protein levels were decreased when PASMC and HUVEC were knockdown of NR4A3; as merely miR-508-3p was down-regulated by miR-508-3p inhibitor, the protein levels of p-ERK and Slug were increased; when these cells were treated by knockdown of miR-508-3p and NR4A3, the p-ERK and Slug were decreased (rescued) again (n = 3, compared to the first two groups). (ns: no difference, *P <.05, **P <.01 and ***P <. promoter to boost cyclin A2 production, thereby promoting PASMC proliferation and resulting in PAH. 38 Meanwhile, miR-508-3p mimics inhibited ovarian cancer cell proliferation, migration by directly targeting the 3′-UTR of CCNA2. 26 Therefore, whether miR-508-3p could mediate CCNA2 to alleviate PAH required more investigation.

Synaptotagamin 5 (SYT5) is a Weibel-Palade bodies (WPBs)-associated
Ca 2+ -sensor, which exert its biological progress in regulating exocytosis and elevating of intracellular free Ca 2+ concentration. 39 The increased Ca 2+ influx via TRPV4 contributed to the contractile, hyperproliferative and anti-apoptotic phenotype of PASMCs. 40  condition and enabling EC to accommodate pro-angiogenic stimuli by accessing to more ATP than oxidative. Pulmonary artery ECs exhibit a further tendency to lactate synthesis and aerobic glycolysis. 45 In summary, we demonstrate that miR-508-3p expression is linked with diagnosis and prognosis value in PAH patients with substantiated our findings by in vivo and in vitro experiments.
Furthermore, we formulate that miR-508-3p low-expressed promotes PASMC migration and proliferation via targeting NR4A3 to activate MEK signalling pathway. Altogether, our study offers evidence that miR-508-3p is a promising biomarker for therapeutic target in patients with PAH.

ACK N OWLED G EM ENTS
The present study was supported by the grants from National Natural Wang. We thank Dr Sheng-Qiang Li for his kindly suggestions.

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The raw data that support the findings of this study are available from the corresponding author.