Transcriptome analysis of fasudil treatment in the APPswe/PSEN1dE9 transgenic (APP/PS1) mice model of Alzheimer’s disease

Alzheimer's disease (AD) is the most common cause of progressive dementia. In the present study, we showed hippocampal tissue transcriptome analysis in APPswe/PSEN1dE9 (APP/PS1, AD model) mice treated with fasudil (ADF) and compared with AD mice treated with saline (ADNS) and wild type mice (WT). The competing endogenous RNA (ceRNA) network was constructed and validated the differential expression of mRNA, lncRNA, miRNA, and circRNA. Our study showed differentially expressed mRNAs (DEMs) between WT and ADNS, while enriched in cell growth and death and nervous system pathways. DEMs between ADNS-ADF were enriched in the nervous system, glycosaminoglycan biosynthesis-keratan sulfate (KS) and Quorum sensing pathways. We validated four genes with RT-PCR, whereas enrichment of Acyl-CoA Synthetase Long Chain Family Member 4 (Acsl4, ENSMUST00000112903) in Quorum sensing pathways, and BTG anti-proliferation factor 1 (Btg1, ENSMUST00000038377) in RNA degradation pathways were conducted. Expression of these two genes were higher in ADNS, but were significantly reduced in ADF. Histone H4 transcription factor (Hinfp, ENSMUST00000216508) orchestrate G1/S transition of mitotic cell cycle and co-expressed with mmu-miR-26a-2-3p-mediated ceRNA and mmu-miR-3065-5p-mediated ceRNA; Wnt family member 4 (Wnt4, ENSMUST00000045747) was enriched in mTOR, Hippo and Wnt signaling pathway. Expression of these two genes were significantly lower in ADNS, and fasudil treatment reverse it. The present studies demonstrated four genes: Acsl4, Btg1, Hinfp, Wnt4 could be potential biomarkers of AD and the targets of fasudil treatment. These results will pave a novel direction for future clinic studies for AD and fasudil treatment.


Results
Fasudil treatment altered the expression of Aβ and p-Tau in hippocampus of APPswe/ PSEN1dE9 mice. Aβ and p-Tau are biomarkers of Alzheimer's disease 29,30 . Expression of Aβ and p-Tau in the hippocampus of WT, ADNS, and ADF mice were evaluated by immunostaining. Our results showed that ADNS induced Aβ 1-42 plaques (1016.59) in the hippocampus; these were significantly reduced in ADF (437.8; p < 0.01), suggesting that fasudil reduced Aβ burden (p < 0.05) (Fig. 1A). Tau phosphorylation is a key factor playing an important role in the pathogenicity of AD 31 . The expression of p-Tau (S404) in hippocampal neurons of ADNS was significantly increased as compared to WT (5029.2) (p < 0.01). However, it was reversed by fasudil (2353.7) as shown by immunofluorescent intensity quantification (p < 0.05), suggesting the protective effect of fasudil in terms of decreasing p-Tau formation in AD (Fig. 1B).
Transcriptome Analysis of hippocampal tissues of APPswe/PSEN1dE9 mice treated with fasudil. Protein-coding (mRNAs) and noncoding RNAs (lncRNAs, miRNAs, circRNAs) associated with fasudil treatment in APPswe/PSEN1dE9 mice were analyzed using samples of hippocampal tissues from WT, ADF and ADNS (n = 3 in each group), as described below.
Further, we identified the biological functions of the DEMs between WT versus ADNS and ADNS versus ADF by gene ontology (GO) analysis. In both WT versus ADNS and ADNS versus ADF, our results showed that the most important molecular functions are binding and catalytic activity. In biological processes, the dominant GO terms were cellular process, single organism process, and biological regulation. The key cellular components involved were cell, cell part, organelle and membrane (Fig. 2C,D; Supplementary Table 3).
Analysis of differentially expressed protein-coding transcripts (mRNA) in fasudil-treated APPswe/PSEN1dE9 mice. In the KEGG pathway analysis, DEMs between WT-ADNS were enriched in the cell growth and death pathways including the cell cycle and apoptosis pathway; 11 genes were associated with the above pathways ( Fig. 2E; Supplement Table 4). DEMs between ADNS-ADF were enriched in the nervous system pathways including the neurotrophin signaling pathway, glutamatergic synapse, cholinergic synapse, long-term potentiation, retrograde endocannabinoid signaling, long-term depression, dopaminergic synapse, glycosaminoglycan biosynthesis-keratan sulfate (KS) pathways. A total of 35 genes were associated with the above pathways. DEMs between ADNS-ADF were enriched in Glycosaminoglycan biosynthesis-keratan sulfate pathways and Quorum sensing pathways ( Fig. 2F; Supplementary Table 4). Acyl-CoA Synthetase Long Chain Family Member 4 (Acsl4, ENSMUST00000112903) were enriched in Quorum sensing pathways. BTG anti-proliferation factor 1 (Btg1, ENSMUST00000038377) was enriched in RNA degradation pathways and significantly reduced in ADF. Wnt family member 4 (Wnt4, ENSMUST00000045747) was enriched in signal transduction pathways, including mTOR, Hippo and Wnt signaling pathway (Supplementary Table 4).
Analysis of lncRNA changes in fasudil-treated APPswe/PSEN1dE9 mice. To identify the functions of lncRNAs, the upstream and downstream genes of lncRNAs were analyzed. We observed 14,906 lncRNA transcripts, including 4311 annotated lncRNAs and 10,595 novel lncRNAs. The lncRNA transcripts corresponded to 4,311 annotated lncRNA genes and 10,595 novel lncRNA genes (Supplementary Table 5). Overall, 46 ADF. Immunofluorescent intensity quantification of anti-Aβ 1-42 showed significant increased intensity in ADNS group when compared with WT (p < 0.01) and the reverse in the Fasudil-treated mice, suggesting the protective effect on Aβ burden and therapeutic potential for clearance of Aβ plaques in AD with Fasudil treatment. (B) Tau phosphorylation at S404 was checked in the hippocampus of the mice in each group. (a-d) WT group, (e-h) ADNS group, (i-l) ADF group. Images (a-c, e-h, i-l) were taken at 10× (200 μm). The box area is enlarged, and the images (d, h, l) were taken at 60× (20 μm). DAPI (blue) was used for nuclei staining. Immunofluorescent intensity quantification of anti-Phospho-Tau (S404) showed a significant increase in intensity in the ADNS group when compared with WT (p < 0.01); the reverse was observed in the Fasudil-treated mice. Data is presented as Mean ± S.E.M. *p < 0.05, **p < 0.01 versus WT; #p < 0.05, ##p < 0.01 versus ADNS. The Dunnett's test were used for statistical analysis. www.nature.com/scientificreports/   Table 6). There were enriched signaling pathways, such as styrene degradation, sesquiterpenoid and triterpenoid biosynthesis, phenylpropanoid biosynthesis in WT versus ADNS. Peroxiredoxin 6 (PRDX6) attenuates Aβ pathology in phenylpropanoid biosynthesis pathway 32 . Our results showed LNC_000304, which in ADNS mice was significantly higher than WT, co-expressed with mRNA of Peroxiredoxin 6 (Prdx6, ENSMUST00000051925).
In ADNS versus ADF, most of genes were involved in the phagosome, synaptic vesicle cycle and adipocytokine signaling pathways ( Fig. 3G; Supplementary Table 6). Also, nitrotoluene degradation and phenylpropanoid biosynthesis are enriched signaling pathways in ADNS versus ADF. Our results demonstrated that LNC_002258, which was significantly lower in ADF than ADNS, was co-expressed with mRNA (Novel_000141) in the phenylpropanoid biosynthesis pathway.
Next, biological functions of the differentially expressed circRNAs between WT versus ADNS and ADNS versus ADF were identified by gene ontology (GO) and pathway enrichment analysis. Differentially expressed circRNAs between WT versus ADNS were significantly enriched in 521 GO terms (p < 0.05), and differentially expressed circRNAs between ADNS versus ADF were significantly enriched in 1226 GO terms (p < 0.05) (Supplementary Table 9). Our results showed that the important biological process GO terms were cellular process, single organism process and biological regulation in both WT versus ADNS and ADNS versus ADF. The cellular components involved were cell, cell part, organelle, and membrane; the molecular functions consisted of binding, catalytic activity and the molecular function regulator (Fig. 4G,H and Supplementary Table 9).
Construction of the competing endogenous RNA (ceRNA) network. Construction analysis of the ceRNA network of all the miRNA, mRNA, lncRNA, and circRNA identified the top 300 ceRNA network, was shown in Fig. 5A and Supplementary Table 10. We also analyzed the ceRNA network for differential expression of miRNA, mRNA, lncRNA, and circRNA. The results indicated 7 miRNA-mediated ceRNA crosstalk between 9 lncRNAs, 14 mRNAs and 3 circRNA ( Fig. 5B; Supplementary Table 10).
Compared to ADNS, mmu_circ_0005875 and Hinfp (ENSMUST00000216508) were both significantly higher in ADF (Fig. 6A1,A2,B1,B2). Compared to WT, Wnt4 (ENSMUST00000045747) was significantly or tended to be lower in ADNS, which was reversed in ADF (Fig. 6D1,D2). Expression of Acsl4 (ENSMUST00000112903) was higher in ADNS, but were significantly reduced in ADF (Fig. 6E1,E2). Compared to WT, mmu-miR-200c-3p was significantly lower in ADNS, which was prevented in ADF (Fig. 6F1,F2). A similar case was observed for LNC_009699 (Fig. 6H1,H2). These data were consistent with the results obtained by the transcriptome sequencing and implied RNAs regulatory roles in AD development and fasudil treatment. We performed the RT-PCR in order to explore the function of LNC_007479 and mmu-miR-183-5p regulation of the expression of Btg1. Btg1 tended to be increased in ADNS relative to WT, but was significantly reduced in ADF (Fig. 6C1,C2). Compared to WT, mmu-miR-183-5p was significantly lower while LNC_007479 higher in ADNS; all of which were www.nature.com/scientificreports/ reversed in ADF (Fig. 6G1,G2,I1,I2), suggesting that LNC_007479 is negatively related to mmu-miR-183-5p and positively to Btg1.

Discussion
Fasudil is the most commonly used Rho kinase inhibitor, as well as an effective calcium channel antagonist and vasodilator. It crosses the blood brain barrier and clinically used in the treatment of cerebral artery spasm after subarachnoid haemorrhage 34,35 . Our previous published study showed that fasudil improve cognitive function in Morris water maze test and further we used same batches of mice, where feces were used for gut microbiota and metabolites, and hippocampus for transcriptome study 36 . APPswe/PSEN1dE9 mice treated with fasudil showed significant decreased in latency to locate the platform, latency of first entrance into target zone and increased time spent in target zone. Morris water maze test suggested the reversal of learning and memory impairment of AD mice 36 . Fasudil retains the BBB integrity by up-regulating expression of tight junction proteins ZO-1 and occludin has been reported by our studies [37][38][39] . Activation of ROCK in astrocytes causes retraction and affects the BBB, where ROCK-2 is predominant isoform and acts as the key factor in the function and maintenance of BBB 39 . Our previous studies have revealed the potent effect of fasudil in the CNS, preventing BBB leakage and Aβ deposition 40 . We also reported protective effect of fasudil by clearing Aβ or p-Tau in primary neurons 41 . In this study, fasudil treatment ameliorated Aβ and p-Tau in the hippocampus of APPswe/PSEN1dE9 mice and transcriptome changes in AD mice were similar to vehicle treatment control. We also observed that proteincoding genes, miRNAs, circRNA, and lncRNA transcripts regulates the pathogenesis of AD. Recent studies have shown that lncRNAs, circRNAs, miRNAs, and mRNAs form large-scale ceRNA crosstalk networks, which have exciting implications in gene regulation at the transcriptional level during physiological and pathophysiological processes [42][43][44][45][46] . Although lncRNAs and circRNAs have been identified and reported in several diseases, however, their functions remain largely unknown 47,48 . We performed a full transcriptome analysis of lncRNAs, circRNAs, miRNAs, and mRNAs by using high-throughput sequencing to identify their functions. DEMs in ADNS versus ADF were enriched in glycosaminoglycan biosynthesis-keratan sulfate (KS) pathways. KS is an extracellular glycan in the brain and plays an important role in AD and neuritis plaque formation 49 . The maturation of N-glycan requires modification of GlcNAc by catalysis of Fucosyltransferase 8 (Fut8) and the action of Beta-1,4-Galactosyltransferase 4 (B4galt4). GlcNAc6ST1 regulates microglial functions in Alzheimer's pathology 50 . The increased Fut8 expression during aging could make the IGF-1 signaling pathway more sensitive in an older organism 51 . Our results showed that B4galt4 (ENSMUST00000023482) was significantly higher in ADF compared to ADNS, suggesting that fasudil enhances the regulation of B4galt4 in the maturation process of N-glycan. In addition, the Fut8 transcript (ENSMUST00000062804) was significantly lower in ADF compared to ADNS (− 0.8-fold); in contrast, the other Fut8 transcript (ENSMUST00000171770) was significantly higher in ADF compared to ADNS (11.6-fold). FUT8 changed differentially, which is inconsistent with previous studies 51 . This may be related to polymorphism, but needs further verification.
DEMs between ADNS-ADF were enriched in Quorum sensing pathways. Acsl4 (Ensmust00000112903) was found to be significantly increased in subarachnoid hemorrhage (SAH). Inhibition of Acsl4 expression reduces inflammatory responses, blood-brain barrier (BBB) damage, oxidative stress, and increases the number of surviving neurons after SAH 52 . Acsl4 significantly increases protein damage, lipid peroxidation markers and related signaling molecules in hearts from AD mice 53 . Our results showed significantly low Acsl4 (ENS-MUST00000112903) in ADF as compared to ADNS, suggesting that fasudil alleviates the impairment caused by Acsl4 in AD mice. The transcriptome sequencing quantitative results are consistent with RT-PCR. It is worth discussing as candidate biomarkers that are positively altered in AD mice after treatment.
We investigated the function of the lncRNAs and circRNAs in response to fasudil treatment in AD mice. Our results demonstrated that circRNAs were differentially expressed in the hippocampus, including 17 up and 17 down in ADNS versus WT, and 27 up and 35 down in ADNS versus ADF. The recent identification of several circRNAs as vital regulators of miRNAs underlines the increasing complexity of ncRNA-mediated regulatory networks 56 . LncRNAs are mainly involved in epigenetic modifications that regulate gene expression such as DNA methylation, histone acetylation, and transcriptional activation, with positive regulatory effects on target genes 44,57 . A gene pair-based method incorporated with lncRNA-mediated gene regulatory networks has been established to identify module biomarkers associated with different brain regions and comprehensively develop biomarker prediction methods and therapeutic strategies for the treatment of AD patients 58 . We demonstrated that protein-coding genes, miRNAs, circRNA, and lncRNA transcripts are candidate regulators for development of AD. Further studies are needed to verify this.
We also constructed a ceRNA network by integrating competing relationships between protein-coding transcript-miRNA and lncRNA transcript-miRNA pairs. Btg1 (ENSMUST00000038377), a member of the cell cycle inhibitory gene family 59 , can inhibit cell proliferation, promote cell apoptosis, and regulate cell cycle progression and differentiation [60][61][62] . Btg1 is expressed in the developing and adult brain 63 , but its function in neural tissues is unclear. Our results demonstrated that the high level of Btg1 in ADNS was significantly reduced in ADF. Similarly, the high level of LNC_007479 in ADNS, which is positively correlated with Btg1, was significantly reduced in ADF. mmu-miR-183-5p had decreased expression in ADNS than WT and ADF and was negatively correlated with Btg1. There is a competitively inhibitory effect between LNC_007479 and mmu-miR-183-5p. www.nature.com/scientificreports/ More specifically, mmu-miR-183-5p negatively regulates the Btg1 mRNA; LNC_007479 binding to mmu-miR-183-5p weakens the inhibitive effect of the miRNA. Upregulation of miR-183-5p along with the downregulation of miR-206-3p and miR-381-3p may serve as putative biomarkers of 4,4′-methylene diphenyl diisocyanate (MDI) exposure 58 . Several miRNAs, including mmu-miR-183-5p, are significantly decreased in a few diseases. Our results demonstrated that while few genes could be involved in the PI3K/Akt pathway, Btg1 appears to be one of these genes. Dysregulation of the PI3K/Akt pathway is implicated in a number of human diseases including cancer, diabetes, cardiovascular disease, and neurological diseases. Our previous results demonstrated that fasudil reversed Aβ 1-42 -induced apoptosis that could be related to Btg1 41 . Therefore, we believe that LNC_007479, miR-183-5p and Btg1 could be potential candidate biomarkers which are positively altered in AD mice, and reverse after fasudil treatment.
Hinfp is the only known transcription factor required for histone H4 gene expression, which binds directly to a unique H4 promoter-specific element to regulate histone H4 transcription at the G1/S phase transition 64,65 . mmu-miR-3065-5p-mediated ceRNAs crosstalk between mmu_circ_0005875 and Hinfp (ENSMUST00000216508). Our sequence result analysis showed that mmu_circ_0005875 of ADF was significantly higher compared to ADNS, and Hinfp in ADF was significantly higher than that in ADNS, and RT-PCR validated it same, suggesting a positive relationship among them. In addition, mmu_circ_0005875 was positively correlated with Hinfp. Therefore, Hinfp could be a candidate biomarker that is positively altered in AD mice after treatment.
Wnt signaling pathway could be a potential target for the treatment of AD due to its close association in pathogenesis of AD. Loss of Wnt signaling enhance neuronal susceptiblity to Aβ-induced apoptosis, while activation of Wnt signaling prevent Aβ induced neuronal death and behavioural deficit 66,67 . Our results showed Wnt4 (ENSMUST00000045747) mRNAs in ADF are significantly higher than in ADNS. A recent study in mouse model of AD reported that dickkOPF-1 (DKK-1), an inhibitor of a typical Wnt signaling pathway, is induced by Aβ and promotes the atypical Wnt-PCP pathway 68 . Fasudil antagonize the Wnt-PCP pathway, restored balance between typical and atypical Wnt signaling pathways and reduces Aβ dependent synaptic toxicity 69 . Regulation of Wnt signaling pathways in central nervous system is expected as powerful tools to control neural immune homeostasis. We also observed mmu-miR-200c-3p is ADNS significantly lower than WT, higher expression of LNC_009699 in ADF and WT as compared to ADNS. There is a probable competitive inhibitory effect between LNC_009699 and mmu-miR-200c-3p; mmu-miR-200c-3p negatively regulates Wnt4 (ENSMUST00000045747) mRNAs expression. These candidate biomarkers are positively altered in AD mice after treatment.
In conclusions, we discovered mRNAs, miRNAs, circRNAs and lncRNAs transcripts that are candidate regulators of development of AD. We also constructed a ceRNA network by integrating competing relationships between protein-coding transcript-miRNA and lncRNA transcript-miRNA pairs. Further, we identified several lncRNA transcripts predicted to regulate the Acsl4, Btg1, Hinfp and Wnt4. The RT-PCR expression patterns were similar to high-throughput sequencing results, which verified the reliability of sequencing results. These four genes (ENSMUST00000112903, ENSMUST00000038377, ENSMUST00000216508 and ENSMUST00000045747) could be potential biomarkers of AD and the targets of fasudil treatment.

Materials and methods
Animals and treatment. Animal Ethics Committee of Shanxi Datong University approved all procedures.
All the experiments were performed in compliance with the guidelines and regulations of the Administration Office of the International Council for Laboratory Animal Science and ARRIVE guidelines. Male APPswe/ PSEN1dE9 mice and the C57BL/6 background (8 months old) were obtained from Beijing Huafukang Bioscience CO., LTD (HFK, Beijing, China). Mice were placed in the pathogen free facilities at the Institute of Brain Science, Shanxi Datong University, pen), separated in a temperature and humidity controlled room with 12-h light/12-h dark cycle. Animals had ad libitum access to food and water. The three groups of mice were selected: (1) vehicle-treated mice (ADNS) were administered normal saline (volume was adjusted similar to fasudil treatment); (2) fasudil (ADF) mice were injected daily with fasudil (Tianjin Chase Sun Pharmaceutical Co., Ltd.), 25 mg/kg/day, i.p. for 16 weeks; (3) age-and gender-matched C57BL/6 (WT) mice (8 months) were injected with the same volume of normal saline (n = 6 per group). 9 mice (n = 3) were sacrificed by cervical dislocation at the end of treatment, and the hippocampal tissues were rapidly collected, fixed with 4% cold paraformaldehyde and embedded in paraffin for immunofluorescence staining. Another, 9 mice (n = 3, biological replicate) were sacrificed at the end of treatment, and snap-frozen in liquid nitrogen and stored at − 80 °C for RNA extraction and RNA-sequencing.
RNA extraction and sequencing. The hippocampal samples were extracted using Trizol reagent (Thermo Fisher) (n = 3 mice per group). Purity and concentration was determined using the NanoPhotometer® spectrophotometer (IMPLEN, CA, USA). The sample quality was assessed using Agilent Bioanalyzer 2100 and sequencing libraries were prepared using the concentration of total RNA (3 μg www.nature.com/scientificreports/ mRNA and lncRNA-Seq analysis. Raw data in fastq format was processed first by removing potential rRNA contamination by mapping to known rRNA gene sequences. Next, an in-house tool was used to remove low quality read pairs, which includes reads with adapter sequences, reads with consequtive Q value < 20 at the ends, and reads less than 50 bp. QC of the sequencing data was assessed using FastQC. Hisat2 was used to map reads to the reference genome. Reference genome and gene model annotation files were downloaded from the genome website directly 70 . The mapped reads of each sample were assembled and convert to FPKM by StringTie (v1.3.1). We used three tools (CNCI, CPC2 and PLEK) to predict coding potential of each transcript. Transcripts predicted without coding potential were selected as candidate lncRNAs and the remaining transcripts were deemed as mRNA. Transcripts not overlapping with known gene annotations were considered as "novel transcripts". The Ballgown suite includes functions for interactive exploration of the transcriptome assembly, visualization of transcript structures and feature-specific abundance for each locus and post hoc annotation of assembled features to annotated features 71 . DESeq2 was used to analyse differential gene expression for significance analysis (p < 0.05 after Benjamin-Hochberg correction) 72 . Cuffdiff provides statistical routines for determining differential expression in digital transcript or gene expression data using a model based on the negative binomial distribution 73 . We searched co-located (the screening range was less than 100 K between lncRNA and mRNA) and co-expressed (the correlation coefficient between lncRNA and mRNA was greater than 0.95) target genes of lncRNAs, and analyzed their functional annotations. Gene Ontology (GO) 74 and KEGG 75 enrichment analysis of differentially expressed genes or lncRNA target genes were implemented by the GOseq R package and KOBAS software (http:// www. genome. jp/ kegg/). The terms with corrected p-values less than 0.05 were considered significantly enriched by differentially expressed genes 75-77 . circRNA-Seq analysis. circRNA-seq analysis was performed by combining the results of two software, find-circ and CIRI2 to reduce false positives [78][79][80] . Only the results predicted by both tools were used for subsequent analyses. Quantification was carried out using SRPBM (Spliced Reads per Billion Mapping) 81 . Differential circRNA expression was carried out using DESeq2 (p < 0.05). Functional annotation of circRNA was based on GO and KEGG annotation. Enrichment analysis was performed as described above in the methodology.
miRNA-Seq analysis. The miRNA library was constructed using NEBNext® Multiplex Small RNA Library Prep Set for Illumina® (NEB, USA). Index codes were added to attribute sequences for each sample as per the manufacturer's recommendations. Briefly, the cDNA library size was separated using PAGE gel and small RNA between 18 and 40 bp was excised and purified. Bowtie2 was used to compare miRNAs to the reference genome, and the distribution of miRNAs were analyzed 82 .  www.nature.com/scientificreports/ Construction of competing endogenous RNA (ceRNA) network. miRNAs, the target mRNAs, lncRNAs, and circRNAs with Pearson's correlation coefficient < − 0.6 & p < 0.05 were selected for analysis. Based on the data of differentially expressed protein-coding transcripts, miRNA, and lncRNA transcripts, we used Miranda (v3.3a) to identify the biological targets of each miRNA from the protein-coding and lncRNA transcripts that showed a significant negative correlation with miRNA expression 83 . We subsequently obtained the protein-coding transcript-miRNA and lncRNA transcript-miRNA pairs and further constructed the competing endogenous RNA (ceRNA) network 78 .
Quantitative real-time quantitative PCR. To confirm the transcriptomic analysis results, circ_0005875, Hinfp, Btg1, Wnt4, Acsl4, miR-200c-3p, miR-183-5p, LNC_009699, LNC_007479 were subjected to real-time PCR. Total RNA was isolated using the RNA Extraction Kit (Servicebio, China), and cDNA was quantified by RT-PCR using ServicebioRT First Strand cDNA Synthesis Kit (Servicebio, China). The fold change in gene expression was calculated using the 2 -ΔΔCt method with the house keeping genes, U6 as the miRNA internal control and GAPDH as the internal control of mRNA, lncRNA, and circRNA. The primer sequences we used are shown in Table 1.
Statistical analysis. SPSS software (International Business Machines Corporation, IBM, USA) was used for statistical analysis. All data was expressed as means ± SEM. Dunnett's test was used to compare two groups, using ADNS as a control, WT and ADF were compared with ADNS. We used Dunnett's post hoc test for analysis of immunofluorescent intensity quantification and quantitative RT-PCR analysis. A value of p < 0.05 was considered statistically significant.