The long noncoding RNA Arrl1 inhibits neurite outgrowth by functioning as a competing endogenous RNA during neuronal regeneration in rats

The intrinsic regeneration ability of neurons is a pivotal factor in the repair of peripheral nerve injury. Therefore, identifying the key modulators of nerve regeneration may help improve axon regeneration and functional recovery after injury. Unlike for classical transcription factors and regeneration-associated genes, the function of long noncoding RNAs (lncRNAs) in the regulation of neuronal regeneration remains mostly unknown. In this study, we used RNA-Seq–based transcriptome profiling to analyze the expression patterns of lncRNAs and mRNAs in rat dorsal root ganglion (DRG) following sciatic nerve injury. Analyses using the lncRNA-mRNA co-expression network, gene ontology enrichment, and Kyoto Encyclopedia of Genes and Genomes pathway databases indicated that the lncRNA Arrl1 decreases neurite outgrowth after neuronal injury. shRNA-mediated Arrl1 silencing increased axon regeneration both in vitro and in vivo and improved functional recovery of the sciatic nerve. Moreover, inhibiting an identified target gene of Arrl1, cyclin-dependent kinase inhibitor 2B (Cdkn2b), markedly promoted neurite outgrowth of DRG neurons. We also found that Arrl1 acts as a competing endogenous RNA that sponges a Cdkn2b repressor, microRNA-761 (miR-761), and thereby up-regulates Cdkn2b expression during neuron regeneration. We conclude that the lncRNA Arrl1 affects the intrinsic regeneration of DRG neurons by derepressing Cdkn2b expression. Our findings indicate a role for an lncRNA-microRNA-kinase pathway in the regulation of axon regeneration and functional recovery following peripheral nerve injury in rats.

The intrinsic regeneration ability of neurons is a pivotal factor in the repair of peripheral nerve injury. Therefore, identifying the key modulators of nerve regeneration may help improve axon regeneration and functional recovery after injury. Unlike for classical transcription factors and regeneration-associated genes, the function of long noncoding RNAs (lncRNAs) in the regulation of neuronal regeneration remains mostly unknown. In this study, we used RNA-Seq-based transcriptome profiling to analyze the expression patterns of lncRNAs and mRNAs in rat dorsal root ganglion (DRG) following sciatic nerve injury. Analyses using the lncRNA-mRNA co-expression network, gene ontology enrichment, and Kyoto Encyclopedia of Genes and Genomes pathway databases indicated that the lncRNA Arrl1 decreases neurite outgrowth after neuronal injury. shRNA-mediated Arrl1 silencing increased axon regeneration both in vitro and in vivo and improved functional recovery of the sciatic nerve. Moreover, inhibiting an identified target gene of Arrl1, cyclin-dependent kinase inhibitor 2B (Cdkn2b), markedly promoted neurite outgrowth of DRG neurons. We also found that Arrl1 acts as a competing endogenous RNA that sponges a Cdkn2b repressor, microRNA-761 (miR-761), and thereby upregulates Cdkn2b expression during neuron regeneration. We conclude that the lncRNA Arrl1 affects the intrinsic regeneration of DRG neurons by derepressing Cdkn2b expression. Our findings indicate a role for an lncRNA-microRNA-kinase pathway in the regulation of axon regeneration and functional recovery following peripheral nerve injury in rats.
Peripheral nerve injury (PNI) is a common clinical issue due to the resulting dysfunction of sensory and motor nerves. Contrary to the central nervous system (CNS), which has poor regeneration ability, the peripheral nervous system (PNS) has a restricted regeneration power following the injury (1). The differential neuronal intrinsic regeneration ability and regenerative microenvironment are two major pivotal factors in axon regeneration and functional recovery (2,3). The signal transduction and transcription response in the initiation of regeneration are well-understood, and many regeneration-associated genes (RAGs) are identified during the PNI (4). However, the molecular mechanisms of axon regeneration have focused mainly on transcriptional factors or RAGs; whether long noncoding RNAs (lncRNAs) have vital roles in this process remains largely unknown.
Genomic research has revealed that only 2% of genes encode coding RNA, whereas the others transcriptionally generate noncoding RNAs, especially lncRNAs, most of which until recently were still largely unknown (5). With the development of RNA-sequencing (RNA-Seq) technology, an increasing number of lncRNAs with a relatively low expression level have been identified to have crucial functions in many physiological and pathological processes (6 -8). Previous studies have demonstrated that lncRNAs are involved in the occurrence of neurological disorders (9), such as Alzheimer's disease (10) and neuropathic pain (11). However, the role of lncRNAs in peripheral nerve injury repair remains to be investigated.
The transcriptome analyses based on the RNA-Seq and microarray have revealed the altered level of lncRNAs in the mouse and rat PNI models (12,13). Our previous study indicated that the level of lncRNAs was changed following sciatic nerve injury (SNI) in rat (14,15), and recent research has reported that lncRNAs impacted the neuronal outgrowth by regulating RAG expression (16). Nonetheless, the function of lncRNAs in PNI repair and the mechanism in axon regeneration are still little understood. In this study, we identified the co-expression regulation network between lncRNAs and mRNAs following PNI. Moreover, we determined the critical role of a new lncRNA, Arrl1, in axon regeneration and functional recovery of injured sciatic nerve. Finally, we explored the mechanism underlying the regulation of Arrl1 in neuronal intrinsic regeneration.

The different expression profiles of lncRNAs and mRNAs following SNI
SNI is a common model for studying peripheral nerve regeneration. To characterize the transcriptome changes in peripheral nerve regeneration, we implemented RNA-Seq of DRG at different time points (0 h, 1 h, 3 h, 12 h, 1 day, 4 days, and 7 days) following SNI in rats. Then unsupervised cluster analysis was conducted to uncover the mRNA and lncRNA expression pat-This article contains supporting information. ‡ These authors contributed equally to this work. * For correspondence: Bin Yu, yubin@ntu.edu.cn.

cro ARTICLE
terns. We identified 17 differentially expressed lncRNAs (Fig.  1A). Besides, we also identified 895 differentially expressed mRNAs (Fig. 1B), including RAGs: Atf3 (17), GAP43 (18), Sox11 (19), and Gadd45a (20), the detail data of differentially expressed lncRNAs and mRNAs was available in Table S1 in the supporting information. The expression changes of the differentially expressed lncRNAs (Fig. S1) and RAGs (Fig. S2) were validated by qRT-PCR. Next, we performed co-expression analysis of lncRNAs and mRNAs to search for the key regulators and found that nine of the differentially expressed lncRNAs were co-expressed or potentially interact with their target genes (Fig. 1C). Furthermore, gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed that these differential co-expressed mRNAs were Figure 1. The transcriptome sequence and bioinformatic analysis of the rat DRG after sciatic nerve injury. A and B, identification of the significantly differentially expressed lncRNAs and mRNAs at 0 h, 1 h, 3 h, 12 h, 1 day, 4 days, and 7 days following sciatic nerve injury. The red rectangular frame in A indicated NONRATT032301.1. C, co-expression network of the differentially expressed lncRNAs-mRNAs in A and B. D, gene ontology and KEGG pathway analysis for differential co-expression genes from C.

Arrl1 regulates neurite outgrowth
enriched in neuron development or axon regeneration-related processes (Fig. 1D), such as the Jak-STAT (21), cAMP (22), or mitogen-activated protein kinase signaling pathway (23,24). The expression of candidate genes involved in those pathways were also validated by qRT-PCR (Fig. S2). According to the validated lncRNA expression patterns and lncRNA feature analysis with the coding potential and ORF prediction analysis (Fig. S3), we chose lncRNA NONRATT032301.1, which was named as axon regeneration-related lncRNA 1 (Arrl1), for further study.

Expression and distribution of Arrl1 in DRG neuron post-SNI
QRT-PCR results indicated that Arrl1 expression was dramatically decreased in rat DRG post-SNI ( Fig. 2A). Fluorescent in situ hybridization (FISH) experiments further validated that the level of Arrl1 was reduced on days 1 and 4 following the SNI and mainly distributed in the cytoplasm (Fig. 2B). The analysis of the Arrl1 sequence in the UCSC Genome Browser (RRID: SCR_005780) showed that no overlapping annotated genes were identified, and Arrl1 was located on chromosome 17p12 (Fig. S3). To examine the Arrl1 distribution in DRG neuron, cytoplasmic and nuclear RNA was isolated and then analyzed by qRT-PCR. The result indicated that Arrl1 was mainly distributed in the cytoplasm (Fig. 2C). The predominantly cytoplasmic distribution of Arrl1 was further validated by a FISH experiment. Arrl1 was co-localized with Tuj1, a neuronal marker, in the cytoplasm (Fig. 2D).

Inhibiting Arrl1 promotes neurite outgrowth in DRG neurons in vitro
To investigate the role of Arrl1 in axon regeneration in vitro, two Arrl1-specific shRNAs were used to disturb Arrl1 expression. DRG neurons infected with adeno-associated virus (AAV) containing Arrl1-specific shRNA (KD1 or KD2) exhibited a reduced level of Arrl1, compared with those infected with AAV containing negative control shRNA (NC) (Fig. 3A). The immunostaining of axons with Tuj1 indicated that knockdown of Arrl1 significantly promoted axon regeneration of DRG neurons in vitro (Fig. 3, B-D). As shown in Fig. 3C, two batches of Arrl1 knockdown increased the total length of axons by 107.1 Ϯ 12.2% and 90.7 Ϯ 25.1%, compared with the control, respectively. Moreover, the length of the longest neurite was also increased in DRG neurons infected with AAV containing Arrl1-specific shRNA (Fig. 3D). Collectively, these data suggested that Arrl1 has a negative role in the regrowth of DRG neuron in vitro.

Arrl1 regulates sciatic nerve regeneration and behavioral recovery
As Arrl1 knockdown promoted neurite outgrowth in DRG neurons in vitro, we further explored its role in sciatic nerve regeneration and behavioral recovery in vivo. First, we intrathecally injected rats with AAV containing Arrl1-specific or negative control shRNA. Arrl1-specific shRNA can knock down Arrl1 in DRG in vivo (Fig. S4). Then the rats infected with the virus were subjected to SNI surgery. Finally, we collected the injured sciatic nerve at 3 days post-surgery and used SCG10 to label regenerated axons (Fig. 4A). The regeneration ability of sciatic nerve was significant enhanced by Arrl1 knockdown (Fig. 4, A-C). To assess the function of Arrl1 on the behavioral recovery, behavioral assays in rats were performed to quantify the latency of heat-or mechanical force-induced response. The rats in the control and Arrl1-knockdown groups exhibited similar response latency in a mechanical force-evoked test at 7 days post-SNI, whereas the rats in the Arrl1-knockdown group presented a significantly better functional recovery at 10 and 18 days post-SNI, compared with those in the control group (Fig.  4D). However, in the assay of heat-induced response, the rats in Arrl1-knockdown group exhibited a moderate thermal sensory recovery at 7 days post-SNI, when compared with the control group (Fig. 4E). Taken together, these results demonstrated that Arrl1 knockdown promotes sciatic nerve regeneration and improves behavioral recovery.

miR-761 is the target molecule of Arrl1
lncRNAs were reported to participate in the development of diseases in different manners (25,26), including the competing endogenous RNAs (ceRNA) mode (27). The target genes of Arrl1 were predicted by MRE enrichment analysis (28), and 61 candidate genes and related miRNAs were found. Meanwhile, RNA-Seq analysis was performed following the knockdown of Arrl1 in DRG neurons. Compared with the control group, 229 genes with 1.5-fold down-regulation in the Arrl1-knockdown

Arrl1 regulates neurite outgrowth
group were selected to conduct GO analysis, which showed that most of those genes were involved in neuronal development or maturation. Then we obtained the overlap between the target genes predicted by MRE enrichment analysis and the downregulated gene analyzed by RNA-Seq and constructed an Arrl1regulating ceRNA network (Fig. 5A). Based on this, we selected three candidate target genes of Arrl1 (Cda, Cdkn2b, and Kcnh4) potentially involved in neurite outgrowth and four related miRNA (miR-1956-5p, miR-761, miR-25-3p, and miR-185-5p) (Fig. 5B). Cda and Cdkn2b, but not Kcnh4, were dramatically decreased following Arrl1 knockdown (Fig. 5C), and interfering Arrl1 significantly increased the miR-761 expression and moderately enhanced miR-185-5p and miR-25-3p expression, whereas it had no effect on miR-1956-5p expression (Fig. 5D). To further explore the direct relationship between Arrl1, related miRNAs, and target genes, we constructed WT and miRNA binding site-mutated Arrl1 (MUT1 and MUT2) luciferase plasmid and performed luciferase assays. The result demonstrated that miR-761 and miR-25-3p significantly reduced the luciferase activity of Arrl1 compared with control miRNA mimic (NC), but miR-185-5p or miR-1956-5p had no notable effect. However, the mutated form (MUT1 and MUT2) of miR-761 or miR-25-3p with the miRNA-binding sites mutated lost their suppressive effect on the luciferase activity, suggesting that the ability of miR-761 and miR-25-3p to regulate Arrl1 expression depends on them binding with Arrl1 (Fig. 5E). Together, these data suggested that miR-761 and miR-25-3p are the target microRNA of Arrl1.
As shown in Fig. 5B, Cdkn2b has a high align score with both Arrl1 and miR-761/miR-25-3p, and Arrl1 knockdown significantly increased the Cdkn2b expression, so we proposed Cdkn2b as the target gene of Arrl1 for subsequent study. We constructed the WT and 3Ј-UTR-mutated Cdkn2b (MUT) plasmid for luciferase assays. Compared with control miRNA mimic (NC), miR-761 significantly reduced the luciferase activity of Cdkn2b, whereas miR-25-3p had no significant effect (Fig.  5F). However, the miR-761 had no effect on 3Ј-UTR-mutated Cdkn2b (MUT), suggesting that miR-761 regulates Cdkn2b expression through binding with its 3Ј-UTR domain (Fig. 5F). Moreover, miR-761 also suppressed the Cdkn2b expression (Fig. 5G). Taken together, these data suggested that Cdkn2b is the direct downstream target gene of miR-761, and miR-761 is the target molecule of Arrl1. In brief, Arrl1 works as a sponge for miR-761 targeting Cdkn2b.

Arrl1 impacts axon regeneration through the Arrl1/miR-761/ Cdkn2b axis in ceRNA mode
Because miR-761 is a key target molecule of Arrl1, we investigated the role of miR-761 in axon regeneration. We found that DRG neurons transfected with miR-761 exhibited marked increased axon outgrowth (Fig. 6A), with the total and maximum length of neurite increased by 136.5 Ϯ 27.8% and 119.2 Ϯ 8.4%, respectively (Fig. 6, B and C). This result suggests that miR-761 promotes axonal regeneration.
We next explored the function of Cdkn2b, a downstream target gene of miR-761, in axonal growth. The Cdkn2b-specific siRNA (Si-1 or Si-2) was transfected into DRG neurons to knock down the expression of Cdkn2b (Fig. 6D). DRG neurons with Cdkn2b down-regulation exhibited significant increase in the total and maximum length of neurite compared with the negative control (Fig. 6, E-G), indicating that Cdkn2b has a negative effect on axonal growth.
As the miR-761 is the target microRNA of Arrl1, we assumed that the suppressive activity of Arrl1 on axonal growth depends

Arrl1 regulates neurite outgrowth
on miR-761. To test the hypothesis, we transfected DRG neurons with vehicle control AAV or with Arrl1-specific shRNA-2 AAV(KD2), KD2 and miR-761 inhibitor (miR-I), then performed the neurite outgrowth assay in vitro (Fig. 7A). As expected, the total and maximum length of neurites were all increased in Arrl1-knockdown cells (Fig. 7, A-C). However, inhibiting miR-761 in Arrl1-knockdown cells abrogated the promoting effect of Arrl1 knockdown on axon growth (Fig. 7, B and C), suggesting that Arrl1 regulates axonal growth through miR-761. To further test the hypothesis in vivo, the miR-761 inhibitor, miR-761 antagomir (miR-A), was injected into DRG following Arrl1 knockdown in vivo. The sciatic nerve regeneration analysis was performed at day 3 after sciatic nerve crush. Consistent with the in vitro study, Arrl1 knockdown in DRG improved the sciatic nerve regeneration, whereas miR-761 antagomir eliminated the promoting effect of Arrl1 knockdown on sciatic nerve regeneration (Fig. 7, D and E). This result further demonstrates that Arrl1-regulated axon regeneration depends on miR-761.
Collectively, these data demonstrate that Arrl1 works as a sponge for miR-761 targeting Cdkn2 and regulates axonal regeneration through the Arrl1/miR-761/Cdkn2b axis in a ceRNA mode following SNI in rats (Fig. 8).

Discussion
Despite the PNS having the intrinsic regeneration ability, severe PNI also induces axonal regeneration delay and a functional recovery barrier (29). Intrinsic regeneration ability is the main mechanism of nerve repair after injury (30), so it is necessary to deeply explore the mechanism of axon

Arrl1 regulates neurite outgrowth
regeneration. Identifying novel regulators in axon regeneration could provide new potential targets for treating PNI. Here, we demonstrated that a novel long noncoding RNA, Arrl1, is decreased upon sciatic nerve injury and functions as a sponge molecule to regulate DRG axonal growth and sensory function recovery.
Previous reports mainly focused on illustrating the effects of transcriptional factors or cytoskeleton-associated genes Arrl1 regulates neurite outgrowth (31), such as Atf3 (32) and Klf4 (33). Only a few studies demonstrated the action of lncRNAs in nerve injury and repair. lncRNA, a multifunctional factor, participates in transcriptional modulation and post-transcriptional modification (34). In the present study, the RNA-Seq analysis revealed the differential expression pattern of lncRNA and

Arrl1 regulates neurite outgrowth
mRNA in DRG neurons following SNI. The lncRNA-mRNA co-expression analysis suggested an important role of lncRNAs in nerve injury repair. The functional analysis of mRNAs in the co-expression network indicates that the expression of mRNAs involved in axonal regeneration pro-cess might be regulated by some co-expressed lncRNAs. The differentially expressed lncRNAs following SNI might be involved in earlier repair of nerve injury. Exploring the function of differentially expressed lncRNAs might help to identify new therapeutic targets for neuroregeneration.

Arrl1 regulates neurite outgrowth
In this study, we identified Arrl1, which was decreased during SNI, as a pivotal negative regulator on axon regeneration. Suppressing Arrl1 expression in DRG neurons promoted the robust neurite outgrowth both in vitro and in vivo, suggesting that Arrl1 might be an intrinsic regeneration inhibitor in SNI. A recent study (16) has uncovered a conserved lncRNA, Silc1, which plays a positive role in regulating mice sciatic injury repair. Knocking out Silc1 impaired axon regeneration, but the gain function of Silc1 in axon regeneration was not uncovered (16). Our previous study identified another conserved lncRNA, uc.217, as a negative regulator in DRG neuron neurite outgrowth in vitro (14). Although we found a comparable increase in axon regeneration following Arrl1 knockdown, the effect of Arrl1 overexpression in neurite outgrowth remains to be determined.
The distribution of lncRNAs may be related to their functions. lncRNAs localized in the nucleus can act as cis-regulatory elements or trans-acting elements to regulate gene expression, chromatin state, or protein/RNA molecules (35,36), whereas lncRNAs expressed in the cytoplasm can perform as a molecular sponge of miRNA and function in ceRNA mode (37). In this study, Arrl1 was mainly expressed in DRG's cytoplasm, and we proposed that a ceRNA-modulated mode may underlie the mechanism of Arrl1 regulating axon regeneration. Moreover, the following bioinformatic analysis and experimental verifica-tion identified Cdkn2b as the potential downstream gene of Arrl1, which was involved in axon regeneration moderation. Previous reports indicated that the dysregulated expression of Cdkn2b impacts neurite extension in retinal ganglion cell degeneration diseases (38). In the present work, we found that Cdkn2b knockdown dramatically increased the axon outgrowth in DRG neurons. Our result demonstrated that Cdkn2b is a new potential target gene in nerve repair post-PNL.
The function of lncRNAs has been widely reported; lncRNAs directly interact with DNA, RNA, and proteins or perform as an miRNA sponge. In our work, we found a significantly increased expression of miR-761 after knocking down Arrl1. miR-761 was first reported in the regulation of the mitochondrial network via affecting the mitochondrial fission factor (39). Previous research has illustrated that miR-761 is involved in tumorigenesis (40) and synaptic plasticity in hippocampal neurons (41). However, the role of miR-761 in axon regeneration is still unknown. In our study, the transfection of miR-761 significantly increased the axonal outgrowth. Meanwhile, overexpression of miR-761 reduced the Cdkn2b level, and the inhibitor of miR-761 eliminated the promoting effect of Arrl1 knockdown on axon outgrowth. Thus, we proposed that Arrl1 regulates peripheral nerve regeneration in a ceRNA mode. Collectively, we explored the downstream targets of Arrl1 and found that miR-761 is the target miRNA of Arrl1, whereas Cdkn2b is the downstream target of miR-761. Both miR-761 and Cdkn2b play pivotal roles in neurite outgrowth. Arrl1 regulates peripheral nerve damage repair through the Arrl1/miR-761/Cdkn2b axis, but the molecular mechanism of Arrl1 differential expression post-PNI remains to be further studied.
In conclusion, our study constructed the lncRNA-mRNA coexpression network following the peripheral nerve injury and identified a novel lncRNA, Arrl1, which can regulate the intrinsic regeneration ability of DRG neurons through sponging miR-761 to modulate Cdkn2b expression. Our finding provides a new mechanism of lncRNA in axon regeneration and functional recovery following peripheral nerve injury.

Experimental procedures
Animals SPF degree male Sprague-Dawley (SD) rats (180 -220 g) were provided by the Experiment Animal Center of Nantong University. All of the experimental procedures involving animals were conducted in accordance with institutional animal care guidelines and approved ethically by the Administration Committee of Experimental Animals, Jiangsu Province, China.  , 1000 m (D). ImageJ was applied to quantify the normalized fluorescence intensity of SCG10 from the crush site toward the distal end (E), and quantification of the maximum regenerated axon length from the injury site in D (F). All of the data are shown as mean Ϯ S.D. (error bars) (unpaired, two-tailed Student's t test; n ϭ 3). *, p Ͻ 0.05; **, p Ͻ 0.01 between the NC and KD2 groups; #, p Ͻ 0.05; ##, p Ͻ 0.01 between the KD2 and KD2 ϩ miR-A groups or KD2 ϩ miR-I groups.

Arrl1 regulates neurite outgrowth Sciatic nerve crush and sample preparation
Twenty-eight SD rats were randomly divided into seven groups, and a 2-cm incision was made in the skin at the left thigh perpendicular to the femur following the intraperitoneal injection of compound anesthetic (10 mg/kg body weight). The muscle tissue was bluntly dissected to expose sciatic nerves. Then the sciatic nerves were crushed at 1 cm proximal to the bifurcation of tibial and fibular nerves using the fine forceps three times at 54 newtons of force. The incision was sutured after surgery. The L4-L5 DRGs were collected at 0 h, 1 h, 3 h, 12 h, 1 day, 4 days, and 7 days following sciatic nerve injury.

RNA extraction and RNA-Seq analysis
Total RNAs were extracted by TRIzol regent (Invitrogen) following the manufacturer's instructions, and RNA-Seq analysis was performed by Shanghai Biotechnology Corp. The differential expression profiles of lncRNAs and mRNAs were determined by bioinformatics analysis as reported previously (42). In hierarchical clustering, the Z score was calculated from the expression of lncRNAs, and the Euclidean distance measure was used to compute the distance (dissimilarity) in lncRNA and the time. The gene co-expression network was built according to the normalized signal intensity of specific expression lncRNAs and mRNAs. GO and KEGG pathway enrichment analyses were performed to elucidate biological processes and signaling pathways associated with the correlated target genes of specific lncRNAs.

qRT-PCR
The cDNA samples were prepared by the Prime-Script RT reagent kit (TaKaRa, Dalian, China) according to the manufacturer's instructions, and qRT-PCR was performed with SYBR Premix Ex Taq (TaKaRa) on an ABI system (Applied Biosystems, Foster City, CA) according to standard protocols. The primers in Table 1 and Table S2 were used to validate candidate lncRNAs and genes, and the primers for target gene detection are presented in Table 2. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as an internal reference. The primers in Table 2 are for Arrl1 distribution analysis. The relative expression of miR-185-5p, miR-761, miR-25-3p and miR-1956-5p was quantified by a commercial qRT-PCR primer set designed by RiboBio (Guangzhou, China) and normalized with the U6 expression level.

Adult rat DRG neuron culture
DRG neurons were separated and maintained in vitro as reported previously (43). In detail, DRGs were dissociated sterilely and incubated with 0.25% trypsin (Invitrogen) for 10 min with intervals of trituration followed by 0.3% collagenase type I (Sigma) for 90 min at 37°C. The supernatant was purified through 15% BSA and planted on the cell culture plate coated with poly-D-lysine and laminin (Sigma) in Neurobasal medium (Invitrogen) with B27 supplement.

AAV infection, siRNA and miRNA mimics, or inhibitor transfection
DRG neurons were replanted on the poly-D-lysine-and laminin-pretreated coverslips and incubated for 18 h after infection with AAV containing control or Arrl1-specific shR-NAs for 5 days or transfection with the Cdkn2b siRNA and miR-761 mimic or inhibitor for 48 h. The targeting site of Arrl1specific shRNAs and Cdkn2b siRNAs was present in Table 3.

Neurite outgrowth assay
Cells were fixed and immunostained with anti-␤-tubulin III antibody (R&D Systems (Minneapolis, MN), AB_2313773), Tuj1, also defined as a class III ␤-tubulin. Neurite length was measured and quantified by ImageJ.

Sciatic nerve regeneration assay
Sciatic nerves were crushed as mentioned above following intrathecal injection with the indicated AAV for 14 days. The sciatic nerves and L4-L5 DRGs were collected at 3 days post-SNI. The sciatic nerves were fixed and immunostained with anti-SCG10 antibody (1:100; Abcam (Cambridge, UK), catalog no. ab66155). Regenerated axon was measured and quantified by ImageJ.

FISH
DRG sections or DRG neurons were fixed by 4% paraformaldehyde (Sigma) and hybridized with 1 g/ml digoxigenin-labeled anti-Arrl1 probe (RiboBio) at 42°C overnight and then

Arrl1 regulates neurite outgrowth
incubated with anti-digoxigenin antibody (RiboBio) and secondary antibody (RiboBio). The slides were observed and photographed using fluorescent microscopy.

lncRNA distribution assay
Cytoplasmic and nuclear RNAs from DRG neurons were isolated by the PARIS TM kit (Life Technologies, Inc.) following the manufacturer's protocols. The expression levels of Arrl1 in the cytoplasm and nucleus were measured by qRT-PCR. To perform quantitative PCR, SYBR Green Mix was used (Takara) with validated primers, which are listed in Table 2.

Luciferase reporter assay
The full-length sequence of Arrl1, the 3Ј-UTR sequence of Cdkn2b, and the mutations of Arrl1 and Cdkn2b were constructed into the pmirGlo vector. The indicated mutations were generated by direct DNA synthesis (GenScript, Nanjing, China). The luciferase reporter assay was performed at 48 h after the reporter vectors co-transfected with the indicated miRNAs into HEK293T cells. Renilla luciferase reporter was used as an internal control, and the relative luciferase activity was normalized to Renilla luciferase activity measured by a Dual-Luciferase reporter assay system (Promega, Madison, WI). To predict the binding sites for rat miRNAs, miRanda software (RRID:SCR_017496) was used.

DRG injection
The lumbar (L)4/L5 intra-DRG injection was performed in male SD rats (180 -220 g) at 12 days after infection with Arrl1knockdown AAV. In brief, the left L4 and L5 DRGs were exposed by removing the lamina of vertebra and opening the epineurium lying covered on the DRG. In the process of DRG injection, the glass needle was inserted into the ganglion, and the miR-761 antagomir (2 nmol in 4 l) was injected over a period of 5 min into the L4/L5 DRGs through the indwelling catheter attached to a 10-l Hamilton syringe. After a delay of 2 min, the needle was removed. After 2 days, the left sciatic nerve was crushed and obtained for analysis 3 days later.

Western blotting
Protein extracts were prepared from primary cultured DRG neurons. Equal amounts of protein were electrophoresed on 12% SDS-PAGE and then transferred onto a nitrocellulose membrane (Bio-Rad). Blots were probed with antibody against Cdkn2b (OriGene, catalog no. TA312926; 1:1000). ImageJ was used to quantify the results of Western blotting.

Behavioral assessment
The behavioral assessment was performed as reported previously (44). For the hotplate test, rats were placed on the glass surface for an acclimation period in 5 min. A focused heat light source linked to a timer (model 33 Analgesia Meter; IITC/Life Science Instruments, Woodland Hills, CA) was used to radiate the plantar surface of hind paw, and quick hind paw raising action was considered to be a positive response. Each paw was tested five individual times, and the mean value was calculated.
For the von Frey test, all experimental animals were adapted to the environment for 30 min before the test, and the plantar surfaces of the left hind paws were vertically stimulated by a series of von Frey filaments (Stoelting, Wood Dale, IL). The filament was bent for 5 s with a sufficient force on the central plantar surface, and paw flinching or brisk withdrawal was considered a positive response. Every test was repeated two times in each rat, and the mean value was calculated.

Statistical analysis
The experimental data statistics were analyzed by unpaired, two-tailed Student's t test or analysis of variance with Tukey's post hoc test, and p Ͻ 0.05 was considered statistically significant. All quantitative data are expressed as mean Ϯ S.D.

Data availability
All of the data are contained within the article.