Abstract

   Circular RNAs (circRNAs) represent a class of non-coding RNAs that are
   involved in transcriptional and posttranscriptional gene expression
   regulation and associated with different kinds of human diseases.
   However, the characterization and function of circular RNAs in
   peripheral nerve injuries remain elusive. Here, we established a rat
   sciatic nerve injury model and identified at least 4,942 distinct
   circular RNA candidates and a series of circular RNAs that were
   differentially expressed in injured nerve tissues compared with matched
   normal tissues. We characterized one frequently downregulated circular
   RNA, circRNA.2837, and further investigated its function in sciatic
   nerve injury. We found that circRNA.2837 regulated autophagy in neurons
   in vitro and in vivo, and downregulation of circRNA.2837 alleviated
   sciatic nerve injury via inducing autophagy in vivo. Mechanistically,
   knockdown of circRNA.2837 may protect neurons against neurological
   injury by acting as a sponge for members of miR-34 family. Our findings
   suggested that differentially expressed circular RNAs were involved in
   the pathogenesis of sciatic nerve injury, and circular RNAs exerted
   regulatory functions in sciatic nerve injury and might be used as
   potential targets in sciatic nerve injury therapy.

   Keywords: circular RNAs, sciatic nerve injury, microRNA, autophagy

Introduction

   Peripheral nerve injury (PNI) is a common clinical problem mainly
   caused by trauma and surgical intervention, which has a heavy social
   burden in terms of both long-term disability and economic costs.[33]1,
   [34]2 It is known that the injured peripheral nerve has an intrinsic
   capacity for axons to regenerate, but the functional recovery was
   poor.[35]^3 Thereby, growing efforts have been dedicated to the
   development of effective treatment for peripheral nerve injury.
   Currently, pre-clinical research of peripheral nerve injury is mainly
   carried out in animal models, and a sciatic nerve injury (SNI) model is
   the most used experimental paradigm for investigating peripheral nerve
   injury.[36]^4 Many experiments have been carried out using the SNI
   model with an attempt to increase nerve regeneration and functional
   recovery after peripheral nerve injury. However, existing therapeutic
   approaches for nerve injuries seem to be inadequate, partly
   attributable to an insufficient understanding of post-injury cellular
   and molecular events. Thus, it is still important to improve the
   understanding of the molecular and intermolecular interactions
   post-peripheral nerve injury.

   Circular RNAs (circRNAs) are a novel type of endogenous noncoding RNA
   characterized by forming circular structures with neither 5′-to-3′
   polarity nor a polyadenylated tail.[37]^5 Recently, it was reported
   that circular RNAs performed numerous biological functions, including
   promoting rolling-circle translation, controlling transcription of
   parent genes, helping to form alternatively spliced mRNA, and acting as
   microRNA (miRNA) sponges.[38]5, [39]6, [40]7 Among these functions, a
   growing number of studies reported that circular RNAs acted as sponges
   to interact with miRNAs and regulate gene expression, thus playing a
   significant role in the pathogenesis and diagnosis of human
   diseases.[41]6, [42]8, [43]9 However, the roles of circular RNAs in
   peripheral nerve injuries are largely unknown.

   In this study, we generated ribo-minus RNA sequencing data of sciatic
   nerve from the rat SNI model and identified approximately 4,942
   circular RNA candidates (at least two unique back-spliced reads). We
   identified a small number of significantly differentially expressed
   circular RNAs in injured sciatic nerve tissues and characterized one
   circular RNA, circRNA.2837, which is frequently downregulated in
   injured sciatic nerve tissues, and further examined the functions and
   mechanisms of circRNA.2837. We found that circRNA.2837 induced
   autophagy in neurons in vitro and in vivo. Down- regulation of
   circRNA.2837 alleviated sciatic nerve injury via inducing autophagy
   in vivo. Moreover, luciferase assay revealed that circRNA.2837 could
   bind to miR-34 family members, including miR-34a, miR-34b, and miR-34c.
   Furthermore, using miR-34a as a representative of this miRNA family, we
   found that silencing of circRNA.2837 induced autophagy in neurons by
   directly targeting miR-34a. Taken together, we concluded that
   circRNA.2837 acts as a ceRNA (competing endogenous RNA) to regulate
   neuronal autophagy by sponging the miR-34 family. Our findings
   indicated that differentially expressed circular RNAs were involved in
   the pathogenesis of sciatic nerve injury, and circular RNAs exerted
   regulatory functions in SNI and can be used as potential targets in SNI
   therapy.

Results

Identification of Circular RNAs by RNA-Seq Analyses in Rat Sciatic Nerve
Injury Model

   We first characterized circular RNA transcripts using RNA sequencing
   (RNA-seq) analysis of ribosomal RNA-depleted total RNA from two paired
   rat injured sciatic nerve tissues and matched normal tissues
   ([44]Figure 1A). RNA samples were sequenced on an Illumina HiSeq and
   mapped to the rat reference genome by BWA-MEM. circular RNAs were
   identified by analyzing the unmapped reads via computational pipelines.
   In all, among the 7,222 distinct circular RNA candidates that were
   found in these tissues (GEO: [45]GSE115315), 4,942 circular RNAs
   contained at least two unique back-spliced reads ([46]Figure 1B). The
   total number of circular RNAs and junction reads detected in different
   sciatic nerve tissue was shown in [47]Figure 1C. We further calculated
   the SRPBM (spliced reads per billion mapping) of each circular RNA to
   quantitate their expression level. Each normal and injured sciatic
   nerve tissue sample expressed approximately 2,000 circular RNAs with
   SRPBM that ranged from 0.1 to 62.6 ([48]Figure 1D). The genomic loci
   distributions of circular RNAs in rat chromosomes were depicted in
   [49]Figure 1E, and we found that the circular RNAs were transcribed
   from all chromosomes, but the distribution was not uniform among
   different chromosomes.

Figure 1.

   [50]Figure 1
   [51]Open in a new tab

   Identification of Circular RNAs by RNA-Seq Analyses in a Rat Sciatic
   Nerve Compression Model

   (A) Workflow for RNA-seq analysis of circular RNAs in two paired rat
   compressed sciatic nerves and matched normal sciatic nerve. (B) The
   total number of circular RNAs and back-spliced reads that were
   identified in two paired rat compressed sciatic nerves and matched
   normal sciatic nerve. (C) The number of circular RNAs and back-spliced
   reads detected in four samples. (D) The number of circular RNAs that
   were identified from two paired rat compressed sciatic nerves and
   matched normal sciatic nerve using different spliced reads per billion
   mapping (SRPBM) cut-off. (E) Circos plot showing the distribution of
   circular RNAs in different chromosomes.

Differentially Expressed Circular RNAs in Rat Sciatic Nerve Injury Model

   On the basis of expression of circular RNA analysis, we found 131
   circular RNAs to be significantly (p ≤ 0.05, fold change ≥2)
   differently expressed in injured sciatic nerve tissues compared with
   matched normal tissues ([52]Figure 2A). All these circular RNAs are
   derived from 21 rat chromosomes, and the distribution of differentially
   expressed circular RNAs on chromosomes are shown in [53]Figure 2B.
   Among these differentially expressed circular RNAs, 83 were upregulated
   and 48 were downregulated. There were 54 circular RNAs transcribed from
   the exonic region, 10 from the intronic region, and 19 from the
   intergenic region among the upregulated circular RNAs, while for the
   downregulated circular RNAs, 35 were transcribed from the exonic
   region, 2 from the intronic region, and 11 from the intergenic region
   ([54]Figure 2C). Since it has been reported that circular RNAs had a
   highly conserved expression pattern in mammals, we conducted analyses
   to detect conserved expression between human and rat based on our
   profile data. We found only 10 circular RNAs could not be mapped to the
   human genome using liftOver among these differently expressed circular
   RNAs; over 90% of them were detected to be 3′ or 5′ slice sites, or
   even both slice sites were utilized by human circular RNAs, indicating
   that circular RNAs were highly conserved between human and rat
   ([55]Figure 2D).

Figure 2.

   [56]Figure 2
   [57]Open in a new tab

   Differentially Expressed Circular RNAs in a Rat Sciatic Nerve
   Compression Model

   (A) Clustered heatmap of the differentially expressed circular RNAs in
   two paired rat compressed sciatic nerves and matched normal sciatic
   nerve. Rows represent circular RNAs, while columns represent tissues.
   The circular RNAs were classified according to the Pearson correlation.
   (B) The distribution for differentially expressed circular RNAs on rat
   chromosomes. (C) The classification of differentially expressed
   circular RNAs based on the genomic origin. (D) Venn diagrams of
   detection results of conserved expression between human and rat. rno3′
   to has3′ represented the 3′ slice site of rat circular RNA and was
   utilized by circular RNAs in human and mapped to the 3′ slice site of
   human circular RNA. Similarly, rno3′ to has5′, rno5′ to has3′, and
   rno5′ to has5′ represented slice sites of rat circular RNA and were
   mapped to certain slice sites of human circular RNA.

Delineation of Gene Ontology and KEGG Pathway for the Host Genes of Circular
RNAs

   To investigate the potential function of differently expressed
   circular RNAs, we performed gene ontology (GO)-enrichment analysis of
   those host genes. The 30 most significant functional annotations are
   shown in [58]Figure 3A. The major enriched and meaningful GO terms in
   biological process (BP) were neurogenesis, neuron differentiation, and
   nervous system development. In terms of cellular component (CC), the
   majority of the altered circular RNA-related mRNAs were found to be
   cell-intracellular membrane-bounded and intracellular organelles. As
   for molecular function (MF), the most enriched molecular function terms
   were protein binding, DNA binding, and organic cyclic compound binding.
   We then employed the Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathway enrichment analysis to further understand their biological
   functions and molecular interactions. The top 30 most significant KEGG
   pathways are listed [59]Figure 3B. Of these, the major enriched and
   meaningful pathways were the axon guidance, vascular endothelial growth
   factor (VEGF) signaling, phosphatidylinositol 3-kinase (PI3K)-Akt
   signaling, and mitogen-activated protein kinase (MAPK) signaling
   pathway.

Figure 3.

   [60]Figure 3
   [61]Open in a new tab

   Gene Ontology and Kyoto Encyclopedia of Genes and Genomes Pathways

   Shown are the top 30 GO (A) and KEGG (B) terms of parental genes from
   which differentially expressed circular RNAs were uncovered. Rich
   factor refers to the ratio of the number of genes located in the GO
   entry to the total number of genes in the GO entry. The larger the rich
   factor, the greater the enrichment. The q value is the p value after
   the multiple hypothesis test correction, where smaller values indicate
   the greater significance.

Validation of Differentially Expressed Circular RNAs by qRT-PCR in Rat
Sciatic Nerve Injury Model

   We confirmed the RNA-seq results in rat injured sciatic nerve tissues
   by qRT-PCR analysis. According to our RNA-seq results, we selected the
   10 most differentially upregulated and 10 most differentially
   downregulated circular RNAs for further validation ([62]Figure 4A).
   Their junction regions were amplified using specific qPCR primers
   ([63]Figure 4B), and Sanger sequencing was performed to verify the
   amplified PCR products with specific circular RNA junctions
   ([64]Figure 4C). Relative expression levels of those circular RNAs
   normalized by SRPBM from our circular RNA-seq analysis are shown in
   [65]Figure 4D. As shown in [66]Figure 4E, our qRT-PCR results indicated
   that the expression patterns of up- and downregulation of the selected
   20 circular RNAs were consistent with the sequencing results,
   suggesting relatively high reliability of this circRNA-seq profile.

Figure 4.

   [67]Figure 4
   [68]Open in a new tab

   Real-Time qPCR Validation of Putative Circular RNA Differentially
   Expressed in a Rat Sciatic Nerve Compression Model

   (A) Clustered heatmap of the top 10 differentially up- and
   down-expressed circular RNAs. (B) Schematic view illustrating the
   design of primers for circular RNAs used in qPCR. (C) The junction
   sequences of selected circular RNAs were validated by Sanger
   sequencing. (D) Relative expression levels of circular RNAs normalized
   by SRPBM after sequencing. SRPBM = number of circular reads/number of
   mapped reads (units in billion)/read length. (E) Relative expression
   levels of circular RNAs detected by qRT-PCR analysis. The qRT-PCR
   results were consistent with the sequencing results. Data represent the
   mean ± SD of three independent experiments. *p < 0.05.

Knockdown of circRNA.2837 Induces Autophagy in Primary Spinal Neurons

   To further investigate the function of differently expressed circular
   RNAs in nerve injury, we focused on the most downregulated circular
   RNA-circRNA.2837 with an expression level downregulated 6.49-fold
   change in injured sciatic nerve tissues compared to matched normal
   tissues from those selected circular RNAs. We noted that circRNA.2837
   is derived from rat Slc25a16 gene locus and located at chromosome
   20:27278618-27279990(-). We used OGD to mimic an injured condition
   in vitro and found that the expression level of circRNA.2837 in spinal
   neurons was also downregulated after OGD treatment for 6 hr
   ([69]Figure 5A). Interestingly, upregulation of LC3-II expression
   level was detected in spinal neurons under OGD in a time-dependent
   manner (data not shown), and the relative fluorescence intensity of
   LC3-II was most significantly elevated under OGD after 6 hr (2.63-fold
   the control, p < 0.001) ([70]Figures 5B and 5C). Meanwhile, the
   formation of numerous autophagosomes with double-membraned structures
   was also observed in rat injured sciatic nerve tissue ([71]Figure 5D),
   indicating autophagy was induced in spinal neurons and sciatic nerve
   under injured conditions. We then explored whether autophagy was
   inhibited by circRNA.2837 in primary spinal neurons. We used siRNAs
   against circRNA.2837 (si-circRNA) to suppress circRNA.2837 expression,
   and circRNA.2837 overexpression vector (pcDNA-circRNA) to upregulate
   circRNA.2837 expression, and the effect of downregulation and
   upregulation of circRNA.2837 on the expression of LC3-II and p62 in
   primary spinal neurons were determined by western blot. qRT-PCR
   revealed that, compared with the negative control group, the expression
   of circRNA.2837 was significantly downregulated by si-circRNA
   transfection and upregulated by pcDNA-circRNA transfection in neurons
   under OGD ([72]Figure 5E). Western blotting showed that upregulation of
   circRNA.2837 dramatically decreased the level of LC3-II compared with
   the OGD group, but the upregulated the level of p62 in neurons
   ([73]Figure 5F). However, knockdown of circRNA.2837 markedly increased
   the expression level of LC3-II while reducing the level of p62
   ([74]Figure 5G). Taken together, these data indicated that circRNA.2837
   negatively regulates autophagy in primary spinal neurons under injured
   conditions.

Figure 5.

   [75]Figure 5
   [76]Open in a new tab

   Knockdown of circRNA.2837 Induces Autophagy in Primary Spinal Neurons

   (A) Expression level of circRNA.2837 in spinal neurons after OGD for
   6 hr was determined by qRT-PCR. (B) Immunofluorescence staining for
   LC3-II in spinal neurons under OGD for 6 hr and normal culture
   condition. Scale bars, 20 μm. (C) The relative fluorescence intensity
   of LC3-II was shown in the bar graphs. (D) Electronmicrographs of the
   compressed sciatic nerve and matched normal sciatic nerve from rats.
   Autophagosomes containing cytoplasmic components with double membrane
   structures were observed (arrow). Scale bars, 1 μm. (E) Primary spinal
   neurons were transfected with siRNAs against circRNA.2837 (si-circRNA)
   or circRNA.2837 overexpression vector (pcDNA-circRNA) or their negative
   control (si-NC and pcDNA-NC) and then treated OGD for 6 hr. Expression
   levels of circRNA.2837 were determined by qRT-PCR assay. (F and G) The
   protein levels of LC3-II and p62 were determined by western blot in
   spinal neurons after pcDNA-circRNA (F) or si-circRNA (G) transfection.
   β-actin was used as an endogenous control. The relative protein levels
   of LC3-II and p62 were shown in the bar graphs. Error bars represent ±
   SD. *p < 0.05, #p < 0.05.

Downregulation of circRNA.2837 Alleviated Sciatic Nerve Injury by Inducing
Autophagy In Vivo

   We next examined the effects of inhibiting circRNA.2837 on neuronal
   autophagy in vivo. The sciatic nerve injury model was established in
   Sprague-Dawley rats ([77]Figure 6A); lentiviruses encoding the
   circRNA.2837 antagonist sequence or control lentivirus were injected
   into the sciatic nerve using a 33G fine needle, and tissues were
   harvested for further examination after 1 week. Knockdown of
   circRNA.2837 by lentiviruses in sciatic nerve was confirmed by qRT-PCR
   ([78]Figure 6B). We found that administration of circRNA.2837 inhibitor
   (lenti-inhibitor circular RNA) markedly reinforced neuronal autophagy
   induced by SNI. Western blotting revealed that downregulated
   circRNA.2837 level markedly upregulated the level of LC3-II but
   decreased the expression of p62 compared with the SNI group
   ([79]Figure 6C). Correspondingly, immunofluorescent staining of sciatic
   nerve tissue showed LC3-II expression in neuronal axons significantly
   increased after lenti-inhibitor circular RNAtransfection
   ([80]Figure 6D). Histological analysis was also performed to evaluate
   the effect of circRNA.2837 downregulation on target muscles after
   sciatic nerve injury. The gastrocnemius muscles of the operated side
   were harvested and subjected to H&E staining, and we found that the SNI
   caused significant reduction on the percentage of muscle fiber area
   (Pm). However, the percentage of muscle fiber area significantly
   increased after the lenti-inhibitor circular RNAinjection at 1 week
   after surgery ([81]Figures 6E and 6F). The motor function recovery was
   assessed by the walking track analysis. The SFI values in the
   LV-inhibitor circular RNAgroup were significantly higher than those in
   the SNI group ([82]Figure 6G). These results indicated that
   downregulation of circRNA.2837 alleviated sciatic nerve injury via
   inducing autophagy in vivo.

Figure 6.

   [83]Figure 6
   [84]Open in a new tab

   Downregulation of circRNA.2837 Alleviated Sciatic Nerve Injury by
   Inducing Autophagy In Vivo

   (A) A flow diagram of the experiments in vivo. A total of 24 rats were
   randomly divided into four groups: rats received sham surgery (sham
   group), rats received SNI surgery (SNI group), rats received SNI
   surgery and the injured sciatic nerve was treated with lentiviruses
   that expressed an antagonist sequence of circRNA.2837 (LV-inhibitor
   circular RNA group), and rats received same operation but were treated
   with negative control viruses (LV-inhibitor NC group) (n = 6 in each
   group). (B) Expression level of circRNA.2837 in sciatic nerve tissue of
   different groups were determined by qRT-PCR. (C) The protein levels of
   LC3-IIand p62 were determined by western blot. (D) Immunofluorescent
   staining of NeuN (green), LC3-II (red), and DAPI (blue) of sciatic
   nerve tissue in different groups. Scale bars, 50 μm. (E) Light
   micrographs of the transverse-sectioned gastrocnemius muscle following
   H&E staining for the operated side of different groups; scale bars,
   100 μm. (F) The percentage of muscle fiber area (Pm) for different
   groups. (G) The SFI of rats in different groups at 1 week after
   surgery. Error bars represent ± SD. *p < 0.05, #p < 0.05.

circRNA.2837 Serves as a miRNA Sponge for the miR-34 Family

   Given that circular RNAs could regulate miRNA target genes by acting as
   miRNA sponges, we speculated that circRNA.2837 could also target a
   specific miRNA and modulate its downstream functions. According to the
   public databases miRanda and TargetScan, circRNA.2837 was observed to
   act as sponge for 25 miRNAs in which the putative binding sites in the
   3′ UTR, where the circRNA.2837 sequence exists. To determine whether
   these miRNAs can directly target the 3′ UTR of circRNA.2837, we
   performed luciferase reporter assays for those miRNAs. MiRNA mimic was
   co-transfected with the luciferase reporters into HEK293 T cells, and
   then the luciferase intensity was detected. Results showed that
   luciferase intensity was reduced by more than 30% when mimics of
   miR-34a, miR-34b, miR-34c were transfected ([85]Figure 7A). As an
   example, the potential binding site between the alignment of miR-34a
   and the 3′ UTR of circRNA.2837 is illustrated in [86]Figure 7B.
   Co-transfection of luciferase reporters containing a 3′ UTR sequence
   and miR-34a mimics into HEK293T cells reduced over 40% of the
   luciferase intensity, but co-transfection of miR-34a mimics and the
   mutated luciferase reporter had no significant effect on luciferase
   activity ([87]Figure 7C), thus confirming the direct interaction
   between miR-34a and circRNA.2837. Additionally, qRT-PCR results
   revealed that the expression of miR-34a was upregulated in primary
   spinal neurons after OGD treatment and rat sciatic nerve compressed
   tissues, which was negatively correlated with the expression of
   circRNA.2837 ([88]Figure 7D). Then the functional relationship between
   miR-34a and circRNA.2837 was further investigated. We constructed over-
   or under-expression vector of miR-34a (miR-mimic and miR-inhibitor) and
   explored their effect on autophagy in spinal neurons. The efficiency of
   miR-mimic and miR-inhibitor was verified by qRT-PCR after transfection
   ([89]Figure 7E). The western blot results of LC3-II and p62 showed that
   the effect on neuronal autophagy of miR-34a was also negatively
   correlated with circRNA.2837 ([90]Figure 7F). Taken together, this
   evidence suggested that circRNA.2837 served as a sponge for the miR-34
   family and regulated the function of miR-34a on autophagy in spinal
   neurons.

Figure 7.

   [91]Figure 7
   [92]Open in a new tab

   circRNA.2837 Serves as a miRNA Sponge for the miR-34 Family

   (A) Luciferase reporter assay for the luciferase activity of
   LUC-circRNA.2837 in HEK293 T cells transfected with a library of 25
   miRNA mimics to identify miRNAs that were able to bind to the
   circRNA.2837 sequence. Three miRNAs (miR-34a, miR-34b, miR-34c) that
   inhibited luciferase activity by 30% are indicated by red dots. (B) A
   schematic drawing shows the putative binding sites of miR-34a with
   respect to circRNA.2837. (C) A luciferase reporter assay was used to
   detect the luciferase activity of LUC-circRNA.2837 or the
   LUC-circRNA.2837 mutant in primary spinal neurons co-transfected with
   miR-34a mimics. (D) Expression level of miR-34a in compressed sciatic
   nerve and primary spinal neurons after OGD for 6 hr was determined by
   qRT-PCR. (E) Primary spinal neurons were transfected with over- or
   under-expression vector of miR-34a (miR-mimic and miR-inhibitor), and
   then treated OGD for 6 hr. Expression levels of miR-34a were determined
   by qRT-PCR assay. (F) The protein levels of LC3-II and p62 were
   determined by western blot in spinal neurons after miR-mimic or
   miR-inhibitor transfection. β-actin was used as an endogenous control.
   The relative protein levels of LC3-II and p62 were shown in the bar
   graphs. Error bars represent ± SD. *p < 0.05, #p < 0.05.

Silencing of circRNA.2837 Induces Autophagy in Primary Spinal Neurons by
Targeting miR-34a

   We then explored the involvement of miR-34a in neuronal autophagy
   regulated by circRNA.2837. Primary spinal neurons were transfected with
   si-circRNA or miR-34a inhibitor or a combination of si-circRNA and
   miR-34a inhibitor and treated with OGD for 6 hr, then the protein
   expression levels of LC3-II and p62 were determined by qRT-PCR and
   western blot analysis. Results suggested that the upregulation of
   LC3-II caused by transfection with si-circRNA was reversed when neurons
   were pre-treated with the miR-34a inhibitor, whereas the miR-34a
   inhibitor restored the production of p62, which was reduced by
   knockdown of circRNA.2837 ([93]Figures 8A and 8B). Likewise,
   immunofluorescence assay revealed that downregulation of circRNA.2837
   caused by si-circRNA transfection markedly increased the relative
   fluorescence intensity of LC3-II, and this phenomenon was also
   neutralized when neurons were co-transfected with the miR-34a inhibitor
   ([94]Figures 8C and 8D). Taken together, these results suggested that
   silencing of circRNA.2837 induced autophagy in primary spinal neurons
   by targeting miR-34a.

Figure 8.

   [95]Figure 8
   [96]Open in a new tab

   Silencing of circRNA.2837 Induces Autophagy in Primary Spinal Neurons
   by Targeting miR-34a

   Primary spinal neurons were transfected with si-circRNA or miR-34a
   inhibitor or a combination of si-circRNA and miR-34a inhibitor and then
   treated with OGD for 6 hr. (A) The protein levels of LC3-IIand p62 were
   determined by western blot. β-actin was used as an endogenous control.
   (B) The relative protein levels of LC3-II and p62 were shown in the bar
   graphs. (C) The expression of LC3-II was detected by
   immunofluorescence. Scale bars, 20 μm. (D) The relative fluorescence
   intensity of LC3-II was shown in the bar graphs. (E) The schematic
   diagram shows the mechanism underlying circRNA.2837 as a ceRNA for
   miR-34a. Error bars represent ± SD. *p < 0.05, #p < 0.05.

Discussion

   In brief, the present study reports the expression profile and
   regulatory function of circular RNAs on autophagy in sciatic
   nerve injury. A large number of circular RNAs were aberrantly
   expressed in SNI nerve tissue compared with matched normal tissue,
   suggesting that these circular RNAs may exert a potential role in SNI.
   We further characterized one of the most downregulated circular RNAs
   and proved that circRNA.2837 played an important role in sciatic nerve
   injury by modulating autophagy via sponging the miR-34 family.

   A great number of studies have examined the molecular mechanisms
   underlying peripheral nerve injury with a sciatic nerve injury model,
   and accumulating evidence indicates that non-coding RNAs (ncRNAs),
   especially miRNAs, and long non-coding RNAs (lncRNAs), significantly
   affect the biological behaviors of neurons during the processes of
   nerve injury.[97]10, [98]11 For example, miR-340 regulates cell debris
   removal and axonal regrowth after sciatic nerve injury.[99]^12 MiR-9
   regulates mammalian axon regeneration by regulating FoxP1 expression in
   sensory neurons after peripheral nerve injury,[100]^13 and miRNA-210
   promotes sensory axon regeneration in vivo and in vitro.[101]^14
   Additionally, lncRNA uc.217 played an important role in peripheral
   nerve regeneration by regulating neurite outgrowth following sciatic
   nerve injury,[102]^15 and lncRNA [103]BC088327 plays a synergistic role
   with heregulin-1β in repairing peripheral injury.[104]^16 Overall,
   these findings showed the potential for treatment of peripheral nerve
   injury by targeting miRNAs and lncRNAs. However, research regarding the
   potential functions of circular RNAs, a new type of ncRNAs, in the
   pathophysiologic process of peripheral nerve injury has not been
   reported.

   It is becoming increasingly evident that circular RNAs played an
   important role in the pathogenesis of human diseases, such as heart
   failure, osteoarthritis, diabetes, and cancer.[105]17, [106]18,
   [107]19, [108]20, [109]21 More importantly, studies revealed that
   circular RNAs are abundant, conserved, and expressed in neuronal
   tissues of mammals and closely related to the development and
   progression of neurodegenerative diseases.[110]22, [111]23, [112]24
   Zhao et al.[113]^25 found that dysregulated expression of ciRS-7
   contributes to the accumulation of amyloid and the formation of senile
   plaque deposits in Alzheimer’s disease by targeting
   ubiquitin-conjugating enzyme E2A. Junn et al.[114]^26 reported that
   ciRS-7 also promoted the progression of Parkinson’s disease by
   modulating the α-synuclein protein aggregation pattern. Besides that, a
   study showed that ectopic circular RNAexpression participated in the
   pathological process of amyotrophic lateral sclerosis.[115]^27
   Recently, two independent groups discovered that circular RNAs were
   differentially expressed in traumatic brain injury, indicating that
   circular RNAs also have potential functions in traumatic nerve
   injury.[116]28, [117]29 In this study, we identified abundant
   expression of circular RNAs in peripheral nerve and found a substantial
   fraction of circular RNAs were also highly conserved and differentially
   expressed after SNI. It is known that the circular RNAfunction was
   related to the known function of the host gene,[118]30, [119]31 so we
   performed GO and KEGG biological pathway analysis and preliminary
   predicted potential functions of these differentially expressed
   circular RNAs. At this point, we only have limited evidence that
   circular RNAs significantly altered after SNI, and they may be
   correlated with post-SNI pathophysiology; however, further studies are
   required to expound upon whether these circular RNAs are really major
   regulators in SNI.

   Mounting studies have focused on the treatment of peripheral nerve
   injury by protecting neurons and promoting nerve regeneration.
   Neurotrophins are reported to be pivotal cytokines promoting peripheral
   nerve regeneration. Nerve growth factor (NGF), the first discovered
   member of neurotrophin family, contributes to the development and
   phenotype maintenance of the peripheral nervous system.[120]^32 Stem
   cells, such as neural stem cells and mesenchymal stem cells, have
   neuroprotective roles in peripheral nerve injury.[121]33, [122]34,
   [123]35 Attempts to enhance nerve regeneration and functional recovery
   with chemical compounds after SNI were also carried out in many
   studies.[124]36, [125]37 In addition, studies have shown that
   artificial nerve conduits along with neurotrophins or stem cells can be
   used to regenerate peripheral nerves after injury as well.[126]38,
   [127]39, [128]40, [129]41 On the basis of our observation of several
   circular RNAs differentially expressed after SNI, we hypothesized that
   further investigation of circular RNAs may lead to the discovery of
   specific functions involved in this process. We characterized one
   circular RNA, circRNA.2837, and found that it could regulate neuronal
   autophagy in primary spinal neurons under OGD, a condition we used to
   mimic nerve injury in vitro. Additionally, administration of
   circRNA.2837 inhibitor markedly enhanced neuronal autophagy and
   alleviated sciatic nerve injury via inducing autophagy in vivo. Our
   results provided a new insight into circular RNA regulation of
   peripheral nerve protection and suggest a potential therapy for
   peripheral nerve injury. Autophagy is an intracellular degradative
   process that recycles cytoplasm to maintain cellular survival and
   stability,[130]^42 and evidence revealed that neuroprotective effects
   of autophagy existed in neurological injury, including spinal cord
   injury and brain injury.[131]43, [132]44, [133]45, [134]46 A recently
   study revealed that a novel indentified circular RNA, circHECTD1, may
   play important roles in cerebral ischemic stroke by inhibiting
   astrocyte activation via autophagy.[135]^47 Here, our experiment showed
   that autophagy was induced both in spinal neurons and sciatic nerve
   tissue under injured conditions, and circRNA.2837-induced autophagy
   protected neurons from injury, indicating that autophagy was activated
   in sciatic nerve after injury and activation of neuronal autophagy
   might be a strategy for the treatment of SNI.

   Mechanistically, circular RNAs are reported to regulate the
   pathophysiological process in eukaryocytes by functioning as miRNA
   sponges, sponging RNA-binding proteins, acting as transcriptional
   regulators, and even translating proteins.[136]6, [137]48, [138]49,
   [139]50, [140]51, [141]52 Among these mechanisms, the most studied
   function of circular RNAs is that they act as miRNA sponges and repress
   their functions. CiRS-7 was reported to be a tumor-related molecule in
   gastric cancer and colorectal cancer by absorbing miR-7[142]^5. The
   circular RNA SRY participates in colorectal cancer and ovarian cancer
   by interacting with miR-138-5p[143]^6. Another circular RNA, circHIPK3,
   which is derived from exon 2 of the HIPK3 gene, was observed to act as
   a sponge for miR-124 in human cells.[144]^53 Similarly, in our study,
   we found that circRNA.2837 functioned as a miR-34a sponge to modulate
   autophagy in neurons. There were several lines of evidence implicating
   circRNA.2837 as a sponge of miR-34a family members against neurological
   injury via regulating autophagy. First, bioinformatic analyses showed
   that the 3′ UTR of circRNA.2837 contains binding sites for miR-34a
   family members. Second, luciferase reporter assays verified this
   prediction. Third, the expression of miR-34a was upregulated in spinal
   neurons after OGD treatment and rat injured sciatic nerve, which was
   negatively correlated with the expression of circRNA.2837. Finally,
   inhibition of miR-34a reversed the effect of circRNA.2837 knockdown on
   neurological autophagy. All of the above results suggested that
   circRNA.2837 could protect neurons against neurological injury by
   sponging miR-34 family members. Besides that, miR-34 was reported to
   regulate cell autophagy in retinoblastoma and asthma. Yin
   et al.[145]^54 found that miR-34 may contribute to airway inflammation
   and fibrosis by modulating IGFBP-3-mediated autophagy activation. Liu
   et al.[146]^55 revealed that miR-34a regulates autophagy and apoptosis
   by targeting HMGB1 in the retinoblastoma cell. In this study, we also
   found that miR-34a modulates neuronal autophagy under OGD treatment,
   indicating miR-34a might play pivotal role in neuron injury via
   autophagy regulation, and it could also be a potential therapeutic
   target in SCI therapy.

   Taken together, our study provided an overview of circular RNA
   expression in sciatic nerve injury, which may serve as a starting point
   for in-depth investigation into the roles played by those
   differentially expressed circular RNAs. We characterized a
   significantly downregulated circular RNA, circRNA.2837, and found
   circRNA.2837 could protect neurons against injury by inducing autophagy
   via sponging miR-34 family members ([147]Figure 8E). However, our
   results call for further investigation into the roles played by other
   differentially expressed circular RNAs in sciatic nerve injury, and
   more attention should be paid to the functional relationship between
   circular RNAs and nerve regeneration after sciatic nerve injury.
   Additionally, further specific studies are needed to decipher whether
   circRNA.2837 plays a role in some other pathological process in
   neurological injury in addition to autophagy. Revealing the role of
   circular RNAs will be critical for understanding sciatic nerve injury
   pathogenesis and offering a novel insight into the identification of
   new biomarkers or new potential therapeutic targets for sciatic nerve
   injury.

Materials and Methods

Animals and Establishment of the Rat Sciatic Nerve Injury Model

   All animal experiments strictly followed the guidelines of the Animal
   Ethics Committee of the Second Military Medical University (Shanghai,
   China). Male Sprague-Dawley rats that weighed between 200 and 230 g
   were housed at a temperature of 21°C to 24°C. A 12-hr light/dark cycle
   was maintained throughout the experimental period. Four rats were
   divided into two groups randomly. The SNI model was established as
   previously described.[148]^56 In brief, rats were anesthetized with
   sodium pentobarbital (1%, 40 mg/kg, intraperitoneal injection). After
   disinfection, the right sciatic nerve was exposed, and a 3-mm-long
   sciatic nerve was crushed two times (15 s each time, 3 s interval) with
   hemostatic forceps. Sham operation (exposure of the nerve) was
   performed in the other group. The rats were sacrificed, and a 3-mm-long
   crushed sciatic nerve, together with both nerve ends (1 mm long), was
   harvested 24 hr after surgery for RNA-seq analysis.

RNA-Seq Analysis

   Total RNA was isolated using TRIzol reagent (Life Technologies,
   Carlsbad, US). Approximately 3 mg of total RNA from each sample was
   subjected to the RiboMinus Eukaryote Kit (QIAGEN, Valencia, CA) to
   remove ribosomal RNA prior to the construction of RNA-seq libraries.
   Strand-specific RNA-seq libraries were prepared using the VAHTS Total
   RNA-Seq (H/M/R) Library PrepKit for Illumina (Vazyme, Nangjing, China).
   In brief, ribosome-depleted RNA samples was purified and fragmented
   with RNA purification beads and fragment mix, then the purified RNA was
   used for first- and second-strand cDNA synthesis with random hexamer
   primers. A-Tailing Mix was used to adenylate 3′ ends of cDNA fragments,
   which were finally ligated to adapters by ligation mix and RNA adaptor
   index. The ligated products were purified and treated with uracil DNA
   glycosylase (UDG) to remove the second-strand cDNA. Purified
   first-strand cDNA was subjected to 13–15 cycles of PCR amplification,
   followed by analysis of the libraries with a Bioanalyzer 2100 (Agilent,
   Santa Clara, CA, USA); the cDNA was then sequenced with an Illumina
   HiSeq X10 system according to the manufacturer’s instructions on a
   150-bp paired-end run.

Identification and Quantification of Circular RNAs

   Raw reads were cleaned using Seqtk then mapped to the rat reference
   genome (Rattus_norvegicus.Rnor_6.0) using BWA-MEM
   ([149]http://bio-bwa.sourceforge.net/). Reads that aligned contiguously
   to the genomes were discarded, and the unmapped reads were then used to
   identify circular RNAs by CIRI.[150]^57 The genomic regions mapping to
   inferred circular RNA were annotated according to the RefSeq
   databases.[151]^58 We determined the parental genes of circular RNAs
   via a custom script, and the corresponding gene of transcript fragment
   whose boundaries (5′ end or 3′ end) exactly matched both ends of a
   circular RNA was defined as parental gene. Then we detected the
   conservation of circular RNA expression between human and rat by the
   method that was previously described.[152]59, [153]60 In brief, by
   using the UCSC liftOver tool, the 5′ and 3′ flank coordinates of each
   differently expressed circular RNA from our profile were converted to
   human genome coordinates. Splice sites were considered homologous when
   they were detected in ± 2 nucleotide intervals around the putative
   human sites.

GO and KEGG Pathway Analyses of the Host Genes

   Gene ontology (GO) analysis, covering three different aspects,
   including biological process, cellular component, and molecular
   function, was performed to explore potential functions of the host
   genes in terms of the circular RNAs. KEGG pathway analysis was carried
   out to detect the involvement of the host genes in the biological
   pathways. The top 30 enriched GO terms and pathways among the two
   groups were ranked by enrichment score (−log10 (p value)) identified by
   the Database for Annotation, Visualization, and Integrated Discovery
   (DAVID; [154]https://david.ncifcrf.gov/).

Neuron Isolation and Culture

   Primary spinal neurons were isolated according to previously described
   methods.[155]^61 In brief, embryos were obtained from pregnant female
   Sprague-Dawley rats. The spinal cord was removed from the vertebral
   canal and cut into small pieces. The pieces were then digested with
   0.05% trypsin (Invitrogen, Carlsbad, CA, USA) in culture medium for
   20 min, then the supernatant was removed after centrifugation, and the
   tissue was resuspended and re-digested according to the procedures
   described above. The supernatant was collected and plated into a
   poly-L-lysine-coated (PLL; Sigma, St. Louis, MO, USA) culture plate for
   4 hr, the medium was replaced with serum-free neurobasal medium
   (Invitrogen, Carlsbad, CA, USA) supplemented with 2% B27 supplement
   (Invitrogen, Carlsbad, CA, USA) and 2 mM glutamine (Invitrogen,
   Carlsbad, CA, USA), and then neurons were cultured at 37°C in a 5%
   CO[2]incubator.

Real-Time qRT-PCR

   Quantification of circular RNAand mRNA was performed using SYBR Premix
   Ex Taq (Takara Bio) in a real-time detection system (ABI7500), and
   miRNA concentrations were determined using the Bulge-Loop miRNA qRT-PCR
   Starter Kit (RiboBio, Guangzhou, China). Before calculation using the
   2^−ΔΔCT method, the levels of glyceraldehyde-3-phosphate dehydrogenase
   (GAPDH) were used to normalize the relative expression levels of
   circular RNA and mRNA, and the levels of small nuclear U6 were used to
   normalize the miRNA expression levels. The primers for circRNA.2837 are
   as follows: forward, 5′-GATCCCAGCTCTTTCACCCG-3′; reverse,
   5′-CAACCAGCUAAGACACUGCGAAA-3′. The primers for miR-34a are as follows:
   forward, 5′-UGGCAGUGUCUUAGCUGGUUGUU-3′; reverse,
   5′-CAACCAGCUAAGACACUGCGAAA-3′.

   The primers of the other 19 circular RNAs were listed in [156]Table S1.

Immunofluorescence Analysis of LC3-II

   Neurons were fixed with 4% paraformaldehyde for 30 min and incubated
   with anti-LC3-II antibody (Proteintech, Chicago, USA) at 4°C overnight
   and then with secondary antibody at 37°C for 1 hr. DAPI was used to
   counterstain nuclei, and then cultures were examined using a
   fluorescence microscope. For immunofluorescence staining of tissue, the
   harvested sciatic nerve was fixed in buffered 4% paraformaldehyde
   overnight and embedded in paraffin. Consecutive serial sections (4 μm)
   were collected for immunostaining. The sections were deparaffinized,
   rehydrated, blocked in PBS with 3% BSA, incubated with anti-LC3-II
   antibody (Abcam, Cambridge, USA) and anti-NeuN (Cell Signaling, Boston,
   MA, USA), and diluted in PBS overnight at 4°C. After rinsing with PBS,
   the sections were incubated with secondary antibody for 2 hr at room
   temperature and examined using an inverted microscope after labeling
   the nuclei with DAPI.

Western Blot Analysis

   The proteins that were extracted from primary spinal neurons and the
   sciatic nerve tissues were electrophoresed on 12% SDS-polyacrylamide
   gel and then transferred to nitrocellulose membranes (Millipore,
   Carrigtwohill, Ireland), blocked with nonfat milk, and probed with
   primary antibodies at 4°C overnight; the membranes were washed three
   times and immunoblotted with secondary antibodies at room temperature
   for 2 hr. Finally, blots were detected by enhanced chemiluminescence
   (ECL) (Pierce, Rockford, IL, USA), and protein bands were quantitated
   with ImageJ software (NIH) using β-actin as an internal control.
   Primary antibodies were as follows: anti-LC3-II (Proteintech, Chicago,
   USA), anti-p62 (Novus, Littleton, Colorado, USA), and anti-β-actin
   (A5441, Sigma-Aldrich Chemicals, St. Louis, MO, USA).

Lentiviral Infection

   We purchased lentiviruses that expressed knockdown or overexpression
   constructs of circRNA.2837 and miR-34a, as well as their corresponding
   negative controls (Gene-Pharma, Shanghai, China) to reinforce or
   silence circRNA.2837 and miR-34a expression in vitro and in vivo.
   According to the manufacturer’s protocol, neurons were incubated with
   lentivirus for 48 hr, and cells were harvested for RNA and protein
   analysis. All experiments were performed in duplicates and repeated at
   least three times independently. For lentiviral injection, 24 rats were
   randomly divided into four groups of six rats each. They were treated
   with an injection of lentivirus-incorporated circRNA.2837 inhibitor or
   corresponding control lentivirus (MOI of both injected lentiviruses was
   1 × 10^9 plaque-forming units [PFU] in a total volume of 5 μL)
   immediately and 4 days after the SNI surgery. Injection was performed
   at the center of the crushed sciatic nerve using a 33G needle and a
   micro-syringe.

Luciferase Reporter Assay

   A luciferase reporter assay was used to detect the direct binding
   between circRNA.2837 and miRNAs. Cells were seeded in 96-well plates at
   a density of 5 × 10^3 cells per well 24 hr before transfection. The
   miRNA mimics were obtained from GenePharma (Shanghai, China) and
   pmiR-RB-Report vector (Saierbio, Tianjin, China) containing the 3′ UTR
   sequence of circRNA.2837, or the sequence with mutated binding sites
   was applied in this experiment. After 48 hr of co-transfection, the
   luciferase activity was measured with a dual luciferase reporter assay
   system (Promega, Madison, WI). The relative fold change of luciferase
   activity normalized with negative control (NC).

Histological Analysis of Targeted Muscle

   The gastrocnemius muscles of the operated side were harvested 1 week
   after surgery, fixed in buffered 4% paraformaldehyde, and subjected to
   H&E staining. For each sample, the cross-sectional area of muscle
   fibers was measured by photographs taken from four random fields and
   analyzed with a Leica software package. The extent of the atrophy of
   target muscles was evaluated by the percentage of muscle fiber area
   calculated according to the formula percentage of muscle fiber area =
   Am/At × 100%, where Am represents the area of muscle fibers in each
   field (scale bars, 100 μm) and At represents the total area including
   muscle fibers and other tissue, such as collagen fibers, in the same
   field as Am.

Behavioral Analysis

   The sciatic functional index (SFI) proposed by de Medinaceli
   et al.[157]^62 and modified by Hare et al.[158]^63 was used to evaluate
   motor function. Walking track analysis was performed on rats at 1 week
   after surgery, and the SFI was calculated according to our previously
   described methods.[159]^64

Statistical Analysis

   The means ± SD was used to show quantitative data. One-way ANOVA was
   applied to compare differences between the two groups. A value of p ≤
   0.05 indicated that the difference was statistically significant.

Author Contributions

   Z.-b.Z., Y.-l.N., and G.-x.H. contributed equally to this work.
   Z.-b.Z., Y.-l.N., and G.-x.H. conducted the experiments. J.-j.L. helped
   with data acquisition. Z.-b.Z., A.C., and L.Z. designed the
   experiments. Z.-b.Z., G.-x.H., and L.Z. wrote the paper.

Conflicts of Interest

   The authors declare no conflict of interest.

Acknowledgments