Repeat-element RNAs integrate a neuronal growth circuit

Summary

Neuronal growth and regeneration are regulated by local translation of mRNAs in axons. We examined RNA polyadenylation changes upon sensory neuron injury and found upregulation of a subset of polyadenylated B2-SINE repeat elements, hereby termed GI-SINEs (growth-inducing B2-SINEs). GI-SINEs are induced from ATF3 and other AP-1 promoter-associated extragenic loci in injured sensory neurons but are not upregulated in lesioned retinal ganglion neurons. Exogenous GI-SINE expression elicited axonal growth in injured sensory, retinal, and corticospinal tract neurons. GI-SINEs interact with ribosomal proteins and nucleolin, an axon-growth-regulating RNA-binding protein, to regulate translation in neuronal cytoplasm. Finally, antisense oligos against GI-SINEs perturb sensory neuron outgrowth and nucleolin-ribosome interactions. Thus, a specific subfamily of transposable elements is integral to a physiological circuit linking AP-1 transcription with localized RNA translation.

Graphical abstract

Highlights

• •GI-SINEs are a distinct subset of B2-SINE RNAs that promote neuronal growth
• •GI-SINEs are induced upon injury in sensory but not retinal ganglion neurons
• •GI-SINE expression is regulated by AP-1 transcription factors
• •GI-SINEs interact with ribosomal proteins and nucleolin to regulate translation

SINEs of recovery: A subset of small RNAs encoded by repetitive genomic elements are induced by nerve lesion to regulate protein synthesis and enhance axonal growth.

Introduction

Neuronal growth and regeneration are regulated by RNA localization and local translation.^1,2 We previously described an intrinsic neuronal-growth-regulating mechanism based on axonal transport of the RNA-binding protein (RBP) nucleolin and local translation of key mRNA cargos, including importin β1 and mTOR^3,4,5 (Figure S1A). Local translation of these and other mRNAs at the cell periphery and retrograde transport of the resulting proteins is thought to set up a length-dependent oscillatory signal that regulates neuronal growth rates.^6,7 Indeed, perturbation of the mechanism by sequestering importin β1 mRNA or nucleolin away from axons significantly enhances neuronal growth.^3,5,6

Figure S1.

Polyadenylated B2-SINEs in DRG neurons and N2a cells, related to Figure 1

(A) Model of intrinsic neuron growth regulation by motor-driven RNA localization. Anterograde transport of mRNAs on nucleolin (Nucl) and kinesin (K), calcium-induced local translation of nucleolin-associated mRNAs, including mTor and importin β1 (β), and retrograde transport of the locally synthesized proteins on dynein (D) and importin α (α), regulates protein synthesis and growth.

(B) Volcano plots of sciatic-nerve-injured/naive (72 h) differential expression of B2-SINE RNAs in L4/L5 DRG neurons (data from Figure 1A). n = 19,279 elements. Red elements are p < 0.05 (marked by dotted line).

(C) Examples of polyadenylation site (PAS) mapping reads in L4/L5 DRG neurons. (1) Genome browser tracks showing reads mapping to 3′ end of Mrpl37. Injury-induced PAS reads were also detected in a Mrpl37 intron, mapping on a SINE. (2) A chromosome 14 intergenic locus with a highly regulated SINE-PAS (closed triangle) alongside a modestly induced SINE-PAS (open triangle).

(D) In vitro removal of B2 RNA polyadenylated tails. Total Neuro2a cell RNA was treated with RNase H in the presence of (dT)₂₀, degrading A-rich RNA RNA-DNA duplexes, collapsing the smeared B2 signal to a single band. 7SL was blotted and shown as loading control.

(E) Volcano plots of optic nerve injury/naive differential expression of B2-SINE RNAs at 72 h after injury in RGC (data from Figure 1F). n = 8,028 elements. Red elements are p < 0.05 (marked by dotted line).

Computational modeling of this intrinsic mechanism postulates the existence of a negative feedback loop for periodic resetting of the signal.^6,7 Because the mechanism is critically dependent on RNA localization to axons, we considered how this might be regulated. RNA localization motifs are often located within 3′ UTRs,^2,8 and 3′ UTR length can be regulated by alternative polyadenylation.^9,10,11,12 We therefore examined the possibility that shortening of 3′ UTRs by alternative polyadenylation might regulate injury-induced growth of peripheral sensory neurons. This led to the unexpected identification of a subfamily of B2-SINE non-coding repeat-element (RE) RNAs as key regulators of a physiological growth circuit.

B2-SINEs are non-coding RNAs transcribed by RNA polymerase III (Pol III) from short interspersed nuclear elements (SINEs), which are high copy number transposable elements in the mouse genome.^13,14 B2-SINEs are often polyadenylated,^15,16 and although mostly transcriptionally repressed in somatic cells, they can be upregulated upon cellular stress.^17,18,19,20 The few studies on B2-SINE functions in the nervous system focused on stress-regulated effects on expression or transcript localization of associated coding genes.^{21,22,23,24,25} Non-coding RNAs such as microRNAs have been reported to regulate nerve regeneration,²⁶ but there is no information to date on specific upregulation of B2-SINEs in neuronal injury and no known connection of B2-SINEs to nerve growth regulation.

Results

B2-SINEs are upregulated in peripheral but not central neuron injury

We tested whether changes in mRNA 3′ UTRs might regulate neuronal growth by 3′ end-targeted RNA sequencing (RNA-seq) of dorsal root ganglia (DRG) neurons after sciatic nerve injury. We did not find evidence for extensive alternative polyadenylation or mRNA 3′ UTR changes in the resulting datasets but unexpectedly observed significant induction of non-coding RNAs from the B2-SINE RE family in DRG neurons following sciatic nerve injury (Figures 1A and S1B).

Figure 1.

B2-SINEs are upregulated in DRG but not RGC neurons after nerve injury

(A) Polyadenylated B2-SINE RNA expression determined by Quantseq REV, 75 bp single-end read RNA-seq from L4/L5 DRG neurons 24 and 72 h after sciatic nerve injury. Injury-dependent log₂ fold changes (log₂FC) of uniquely mapped B2 elements. Boxplot midline shows medians, and whiskers indicate 10th and 90th percentiles. RE (non-B2)—all repeat-element-mapped RNAs excluding B2-SINEs. ^∗∗∗∗p < 0.0001, Mann-Whitney test, n = 23,212, 13,686, 19,279, and 9,121 elements for B2 (24 h), RE (24 h), B2 (72 h) and RE (72 h), respectively.

(B) Northern blot with consensus B2-SINE probe on total L4/L5 DRG RNA 72 h after sciatic nerve injury. Representative blot shown on the left, 7SL RNA probed as loading control. Right: B2/7SL ratios from quantification of all bands visualized relative to the mean of naive controls. n = 4 mice; ^∗p < 0.05, t test, means ± SEM.

(C) qRT-PCR on the same samples as (B) with primers recognizing consensus mouse B2-SINE sequence. 18S RNA was used for normalization, data are relative to the mean of naive controls. n = 4 mice; ^∗∗p < 0.01, t test, means ± SEM.

(D) Analysis of B2-SINE RNA expression by whole transcriptome sequencing of DRG neurons from an independent sciatic nerve injury time course experiment. Data presentation as in (A), n = 18,905.

(E) All non-B2 repeat-element RNAs from the experiment described in (D), n = 7,034.

(F) B2-SINE RNA expression by whole transcriptome sequencing of retinal ganglion cells (RGCs) after optic nerve injury. Data presentation as in (A), n = 8,028.

(G) All non-B2 repeat-element RNAs from the experiment described in (F), n = 12,319.

(H) Statistical comparison of time courses shown in (D)–(G). Fold changes were modeled with a linear mixed effects model, with time, group and set as fixed factors, and a random intercept per element. ^∗∗∗p < 0.001 between DRG B2 and RGC B2.

(I) Proportions of reads mapping to different repeat-element classes in DRG and RGC datasets shown in (E) and (G) (summing reads from all data points). All repeat-element-mapping reads were taken as 100%.

See also Figure S1 and Table S1.

We first observed robust induction of individual loci of polyadenylated B2-SINE RNA in DRG neurons 72 h after sciatic nerve lesion (Figure S1B). Next, we examined ratios of B2-SINE RNAs and all other RE RNAs and found that numerous B2 elements were specifically upregulated at 24–72 h after injury, compared with all RE RNAs (Figure 1A). Northern blot and quantitative reverse-transcriptase PCR (RT-qPCR) analyses validated polyadenylated B2-SINE upregulation in DRG after sciatic nerve injury (Figures 1B, 1C, S1C, and S1D). A time course of injury-induced B2-SINE RNA expression in DRG revealed peak upregulation at 72 h post injury (Figure 1D). Expression of other RE RNAs were largely unaffected (Figure 1E), indicating specific injury-induced B2-SINE transcription over a similar time course as injury-induced mRNAs.^27,28 We then asked whether B2-SINEs are induced by injury in CNS neurons. RE RNAs were profiled in retinal ganglion cells (RGCs) 24–96 h after optic nerve crush. In contrast to DRG, B2-SINE expression was not altered in injured RGCs (Figures 1F and S1E) nor was any other RE RNA (Figure 1G). Comparison of B2-SINE and all RE time courses in sciatic and optic nerve lesions shows a marked difference between B2-SINE dynamics in DRG and RGCs (Figure 1H) and a much higher proportion of B2-SINEs in DRG compared with RGC neurons (Figure 1I). Thus, B2-SINE RNAs are upregulated in peripheral DRG neurons upon nerve injury but not in central RGC neurons.

B2-SINE overexpression enhances growth of injured peripheral and central neurons

CNS and peripheral nervous system (PNS) neurons differ markedly in their capacity for regeneration after nerve injury.^26,29,30 Injury-induced upregulation of B2-SINEs only in regeneration-competent neurons might indicate a role in neuron outgrowth. We validated adeno-associated virus (AAV) vectors for expression of a B2-SINE consensus sequence (Figures S2A and S2B) and tested effects on neuronal growth. To this end, we adapted an adult DRG culture system³¹ to spot culture injury format³² and compared outgrowth in B2-SINE-transduced cultures vs. controls. AAV-B2-SINE-transduced sensory neurons indeed reveal enhanced processive growth after axotomy (Figures 2A and 2B; Videos S1 and S2).

Figure S2.

Validation of exogenous B2-SINE expression, related to Figure 2

(A and B) B2-SINE consensus northern blot on human HEK293 cells transfected with AAV-B2 vs. untransfected HEK293 and mouse N2a cells. RNA size marker on left. No B2-SINE signal in untransfected HEK293, whereas transfection generates both non-polyadenylated and polyadenylated B2-SINE.

(C) FISH of exogenous B2-SINE in retinal ganglion cells (RGCs) transduced with AAV-B2. B2-SINE FISH in white, YFP co-expressed from AAV-B2 in magenta.

(D) Quantification of (C), n = 7 (control), 13 (B2) transduced retinas (72 and 189 cells), means ± SEM, ^∗∗p < 0.01, Mann-Whitney test.

(E and F) Representative images of whole retinas injected with AAV-B2, AAV-control, or not injected, confirming specific B2-SINE overexpression. Scale bars, 100 μm.

Figure 2.

Exogenous B2-SINE expression promotes axonal growth

(A) DRG neuron spot cultures transduced with B2-SINE AAV (AAV-B2) or control vector (AAV-control) underwent axotomy and axon growth was tracked by live imaging of AAV-expressed YFP (yellow fluorescent protein). Representative images of axon outgrowth at the site of axotomy at indicated time points following axotomy. Scale bar, 100 μm.

(B) Processive growth rates across imaging sequences for individual axons are shown based on distance extended in perpendicular to the axotomy line. n = 23 (AAV-control) or 26 (AAV-B2) axons from three independent cultures after ROUT (Robust regression and Outlier removal) outlier analysis with Q = 1%; mean ± SEM, ^∗∗∗∗p < 0.0001, Mann-Whitney test.

(C) Experiment timeline for transduction of AAV2 for exogenous expression of consensus B2-SINE in mouse retinal ganglion cells (RGCs). Crush lesions of the optic nerve were done 2 weeks after intraocular injection with AAV-B2 or AAV-control. 12 days later, mice received intraocular injections with the CTB-555 axonal tracer (cholera toxin B subunit) and 2 days later animals were sacrificed for histology.

(D) Representative images of axonal growth after optic nerve crush injury, visualized with CTB-555. Scale bar, 100 μm.

(E) RGC axon numbers at indicated distances from the crush site, normalized to control at 400 μm, n = 9 mice. Mean ± SEM, ^∗∗p < 0.01, ^∗∗∗p < 0.001, t test.

(F) Experiment timeline in corticospinal tract injury model. AAVs were injected into the mouse motor cortex 1 week before dorsal column spinal cord lesions at T1 level (SC lesion). Two weeks later an anterograde AAV2 axonal tracer expressing tdTomato was injected into the motor cortex and 2 weeks after that animals were sacrificed for histology.

(G) Representative tdTomato tracer images of axons growing into gray matter above the lesion. Horizontal lines mark dorsal-to-ventral distance from tract-gray matter border. Scale bar, 100 μm.

(H) Axonal growth was normalized to axon numbers in main tract and calculated as percent of mean control axon growth. Data points show mean percentages at 100 μm intervals over 0–500 μm from the border. Results are from 3 (AAV-control) or 4 (AAV-B2) mice. ^∗∗∗p < 0.001 for effect of AAV-treatment on outgrowth by two-way ANOVA.

See also Figure S2 and Videos S1 and S2.

Video S1. Time-lapse of injured DRG axon growth under control conditions, related to Figure 2

Time-lapse imaging of regrowth of DRG neurons transduced with control AAV-Php.S after injury in spot cultures.

Video S2. Time-lapse of injured DRG axon growth upon B2-SINE overexpression, related to Figure 2

Time-lapse imaging of regrowth of DRG neurons transduced with B2-SINE-AAV-Php.S after injury in spot cultures.

We then examined B2-SINE effects in injured RGC neurons. Intraocular transduction of AAV2-B2-SINE followed by fluorescence in situ hybridization (FISH) confirmed B2-SINE overexpression in transduced RGC neurons (Figures S2C–S2F). Analyses of axonal regeneration after AAV transduction and optic nerve crush lesion³³ (Figure 2C) showed that B2-SINE overexpression enhanced axon outgrowth beyond the crush site (Figures 2D and 2E). To further test the effects of exogenous B2-SINE expression on injured CNS neurons, we conducted a corticospinal tract (CST) lesion³⁴ (Figure 2F). As in RGCs, B2-SINE overexpression enhanced axonal growth in injured CST neurons (Figures 2G and 2H). Thus, B2-SINE RNAs act as pro-regenerative factors to enhance growth after injury across different neuron subtypes.

Specific transcriptional regulation of growth-inducing B2-SINEs

We next examined B2-SINE transcriptional induction by nerve injury, using 150 bp paired-end library sequencing to unambiguously map B2-SINE-encoded RNAs to specific B2 genomic loci. We found that nerve-injury-induced upregulation of B2-SINEs is mainly confined to the evolutionary younger B2-mm1 and B2-mm2 SINE subfamilies³⁵ (Figure 3A). B2-SINE upregulation peaked at 72 h post injury (Figures 1D and S3A), and 453 individual B2-SINE RNAs mapped to unique loci were significantly changed at this time point, 447 of which were upregulated (Table S1). These loci were mainly intergenic or in intronic regions (Figure 3B), suggesting that these elements are not co-transcriptionally regulated with coding genes. Non-regulated B2-SINE loci, namely the 1,887 elements detected as RNAs that were not regulated by sciatic nerve injury, were less prevalent in intergenic regions (Figures S3A and S3B). A correlation analysis for the minority of regulated B2-SINEs that are within intragenic regions showed that they are not co-regulated with associated coding genes in sciatic nerve injury (Figure S3C). Finally, analysis of other Pol III-transcribed ncRNA genes shows that sciatic nerve injury does not upregulate global Pol III transcription in DRG (Figure S3D). We therefore define this distinct subset of nerve-injury-regulated B2-SINEs and their genomic loci as growth-inducing SINEs (GI-SINEs).

Figure 3.

GI-SINES are a specifically regulated subset of B2 elements

(A) Enrichment of B2-SINE subtypes in naive and injured DRG at 24 and 72 h after sciatic nerve crush. Paired-end, 150 bp reads total RNA-seq was carried out and B2-SINEs were mapped using RSEM (RNA-seq by Expectation Maximization) algorithm for injury/naive differential expression analysis. Boxplot whiskers show 10th and 90th percentiles. n = 4,205, 2,282, 2,559, and 886 elements for B2mm1, B2mm2, B3, and B3A subfamilies, respectively.

(B) Genomic loci distribution of B2 elements significantly upregulated with fold change > 2 at 72 h after injury (GI-SINEs). n = 447 individual loci.

(C) Profile of GI-SINEs association with Tn5-accessible loci by ATAC-seq analysis. Loci of GI-SINEs and non-significantly regulated B2-SINEs expressed 72 h after injury were aligned to ATAC-seq datasets of either sciatic-nerve-crushed or sham-operated DRG neurons (NeuN+). Mean ± SEM are shown for ATAC-seq read densities across the span of the B2 metagene loci for each paired alignment.

(D) Transcription-factor-binding site (TFBS) predictive analysis using TOBIAS to identify enrichment of TFBS for GI-SINE loci. Data points in green and red are TFBS significantly enriched in injury and sham datasets, respectively. Yellow points depict a cluster of AP-1 TFBS enriched in GI-SINE loci after injury.

(E) B2-SINE expression in cultured DRG neurons transduced with the AP-1 dominant-negative A-Fos, GFP in PHP.S AAV, or left untransfected (Unt.) were taken from differential expression analysis of RNA-seq (150 bp reads, paired-end sequencing). Fold change ratios of B2-SINE RNA that match GI-SINE loci (GI-SINE RNA), plotted in boxplots, as described in (A). n = 73 GI-SINE loci, from 9 independent cultures. ^∗∗∗p < 0.0001, Kruskal-Wallis with Dunn’s multiple comparison test.

(F) Differential expression of GI-SINE RNA in sciatic-nerve-injured vs. naive DRG in sensory neuron conditional knockout of ATF3 (Advillin-Cre⁺/ATF3-exon3-flox^+/+) vs. wild-type (WT) mice. RNA was collected 72 h after injury for RNA-seq and expression analysis was as described in (A). B2-SINEs with injury/naive fold change > 2, p < 0.05 in ATF3-wt mice were annotated as GI-SINEs. Sciatic-injured and naive DRG RNA were each taken from 9 (ATF3-wt) and 8 mice (ATF3-cKO). Data are plotted in boxplots, as in (A), n = 311 GI-SINE loci, ^∗∗∗∗p < 0.0001, Mann-Whitney test.

(G) Volcano plots of injury/naive differential expression of GI-SINE RNA. Data are from the experiment described in (F). Red data points are elements with p < 0.05 (marked by dotted line). Dashed line marks injury/naive fold change = 2 (fold change criteria for GI-SINE), n = 311 elements.

See also Figure S3 and Table S2.

Figure S3.

Supporting data for specific regulation of GI-SINES, related to Figure 3

(A) Volcano plot of B2-SINE RNAs expression in L4/L5 DRG 72 h after injury, as described in Figure 3A. Red elements are p < 0.05 (dotted line). Dashed line is 2× threshold for GI-SINE definition.

(B) Genomic region annotation of B2-SINE loci not regulated in injured DRG neurons (left, 1,887 elements) and distribution of annotation categories across mm10 mouse genome (right).

(C) Correlation of 72 h injury/naive expression changes from intragenic GI-SINEs with their associated coding genes (n = 144). Pearson’s correlation linear fit with 95% confidence intervals.

(D) Injury/naive fold changes of Pol III-transcribed ncRNA genes and SINE subfamilies (group mean) in DRG after sciatic nerve injury.

(E) RGC ATAC-seq association profiles of GI-SINEs identified in injured DRG. Loci of significantly upregulated GI-SINEs and non-significantly regulated B2-SINEs from DRG data were aligned to ATAC-seq datasets of RGC 3 days following optic nerve injury or sham operation, as previously described.³⁷ Note the lack of association of SINE loci with accessible chromatin in DRG.

(F) Mean expression fold changes of injury-induced and regeneration-associated genes, n = 9 for AAV-PHP.S-A-Fos dominant negative, control GFP transduced, and untransduced control (A-Fos/GFP and GFP/Unt.), n = 4 sciatic-nerve-injured mice (injury/naive 24 h and 72 h).

(G and H) Mean expression fold changes and p values of mapped B2-SINE loci from Figure 3E, compared between A-Fos vs. GFP or GFP vs. untransduced. n = 9. Red elements are p < 0.05 (dotted line).

(I) Mean expression fold change of injury-induced and regeneration-associated genes between ATF3-wt and sensory neuron ATF3-cKO (Advillin-Cre⁺/ATF3-exon3-fl^+/+) mice. n = 8 (ATF3-cKO) and n = 9 (ATF3-wt).

Further insights on the loci-specific upregulation of GI-SINEs were sought from spatial correlations of regulated GI-SINEs with ATAC-seq of NeuN-positive DRG nuclei after sciatic nerve injury.³⁶ GI-SINEs, but not non-regulated B2-SINEs, were specifically enriched at accessible chromatin sites following injury, with peak accessibility at the 5′ end, suggesting that GI-SINE upregulation by sciatic nerve injury is transcriptionally regulated via dedicated promoters (Figure 3C). In contrast, alignment to RGC ATAC-seq datasets³⁷ did not show any enrichment of GI-SINE loci in accessible chromatin of injured RGC (Figure S3E). Transcription-factor-binding-site analyses³⁸ revealed significant enrichment of AP-1 family transcription-factor-binding sites upstream of GI-SINE loci in injured DRG neurons (Figure 3D; Table S2). Transduction of cultured DRG neurons with AAV expressing A-Fos, a dominant-negative Fos construct,³⁹ downregulated a battery of injury-induced mRNA targets of AP-1 transcription factors (Figure S3F) and also led to robust downregulation of GI-SINEs (Figures 3E, S3G, and S3H). Thus, GI-SINE transcription is specifically regulated by AP-1 transcription factors.

Several AP-1 family members have been implicated in nerve injury responses,^26,30 and ATF3 has been identified as a prominent AP-1 co-factor in regenerative growth.^40,41 We took advantage of a conditional ATF3 allele that generates nuclear-import-deficient ATF3⁴¹ to generate a sensory-neuron-specific conditional mouse knockout (cKO). Sciatic nerve injury was carried out on ATF3 wild-type and cKO mice, and L4/L5 DRG neurons were extracted 3 days afterward for RNA-seq. Functional ATF3 KO was confirmed from expression profiles of known ATF3-responsive regeneration-associated genes (Figure S3I). Paired-end 150-bp sequencing revealed that over two-thirds of the GI-SINEs are not upregulated upon injury in ATF3 cKO DRG neurons (Figures 3F and 3G), clearly confirming a key role for this AP-1 co-factor in GI-SINE regulation. The residual upregulation of a minority of GI-SINEs in ATF3 cKO neurons may indicate involvement of AP-1 complexes lacking ATF3 or other transcription factors, albeit to a lesser extent.

GI-SINEs integrate mRNA localization and translation

We then used pull-downs with biotinylated B2-SINE RNA to identify intracellular targets of GI-SINEs. Sciatic nerve axoplasm was used for the pull-downs because it can be extracted without using detergents,⁴² hence preserving interaction complexes. We used two B2-SINE RNA baits, comprising 5′ (1–75 nt) and 3′ (75–166 nt) consensus sequences (Figure 4A) predicted to retain their respective structures⁴³ (Table S3). Strikingly, one of the top hits in this proteomics analysis was nucleolin (Figure 4B). In addition, the dataset was remarkably enriched with numerous ribosomal subunits and translation-associated proteins pulled down with both B2-SINE baits (Figures 4B–4D). We examined the specificity of these interactions by comparison with a previously published dataset of similar pull-downs with four additional RNA baits⁴⁴ (Figure S4A). Nucleolin (Ncl) was identified as one of the strongest interactors of two of these baits (GAP43 and Hmgb1), a weaker interactor of another (Nrn1), and was not significantly enriched for the Actb bait. Notably, ribosomal subunit proteins are not specifically co-precipitated by any of these four baits.

Figure 4.

B2-SINE RNA interacts with the protein translation machinery

(A) Schematic of B2-SINE interactome analysis (consensus structure model based on Espinoza et al.⁴³ and Ponicsan et al.⁴⁵), blue and red dashed lines delineate the two biotin-conjugated constructs used for pull-downs.

(B) Top 20 enriched proteins co-precipitated with B2-SINE baits. Nucleolin (Ncl) is highlighted in magenta and ribosomal proteins are in green. Enrichment scores over mock (biotin only) pull-downs were calculated for each protein hit. Heatmap depicts mean log₂FC (vs. biotin-only control) of three independent pull-down experiments for B2_5′ and B2_3′ pull-downs.

(C) GO term analysis of cellular processes and subcellular component enrichment in B2-SINE pull-downs (n = 3). GO terms presented are terminal nodes with enrichment FDR (false discovery rate) adjusted p < 0.05.

(D) Heatmap of B2-SINE-bound proteins enriched in the GO term “ribosome cellular component.” Mean log₂ of PSM fold change over control, n = 3 independent pull-downs.

(E) Western blot validation of B2-SINE binding to nucleolin in axoplasm. Representative blot from three independent experiments.

(F) Nucleolin-B2-SINE interaction requires the nucleolin GAR domain. DRG extracts were taken from nucleolin-GAR^+/Δ mice for B2-SINE RNA bait pull-downs. Representative nucleolin western blots of pull-down and input samples from three independent experiments are shown. Arrowheads mark bands of nucleolin WT (full length) and nucleolin-ΔGAR.

(G) RT-qPCR analysis of RNA co-immunoprecipitation (RIP) of endogenous B2-SINE and nucleolin in DRG extracts. Nucleolin IP/input ratio for each transcript is normalized to mean IgG IP/input ratio. Mean ± SEM from 3 RIP experiments on extracts from 2 mice each.

(H) B2-SINE upregulation in ribosome-associated RNA from DRG neurons 72 h after sciatic nerve injury. Advillin-RiboTag mouse DRG were analyzed by Quantseq REV 75 bp read paired-end RNA-seq of RiboTag HA-RPL22 immunoprecipitation (IP) and input RNA extracts. B2-SINEs were mapped in injured and naive samples. Boxplot midlines show median fold change, and whiskers indicate 10th and 90th percentile ranges of injury/naive of IP and naive DRG from three independent sciatic nerve crush experiments. n = 10,644 B2 elements.

(I) Enhanced upregulation of B2-SINE in RiboTag-IP-bound RNA compared with total RNA (input). RiboTag RPL22 IP vs. input of significantly regulated B2-SINEs injury/naive expression (log₂FC injury/naive in IP or input with p < 0.05). Diagonal line depicts the trend of equal IP/input fold change ratio. n = 379 elements.

See also Figure S4 and Table S3.

Figure S4.

B2-SINE vs. U1 RNA-interacting proteins from sciatic nerve axoplasm, related to Figure 4

(A) Re-analysis of RNA bait pull-downs from a previous dataset,⁴⁴ comparing protein spectral counts. Top 20 enriched hits (mean RNA/biotin ratio) with p < 0.05, two-tailed t test from three independent pull-downs. Nucleolin (Ncl) in magenta and ribosomal proteins in green.

(B) Proteins co-precipitated from sciatic nerve axoplasm with biotinylated B2_5′ (nt1–75) vs. U1 (SL3 + 4) baits. Protein hits identified by TMT mass spectrometry before filtering based on enrichment in either B2 or U1 against biotin-only control at p < 0.05 from four independent pull-down experiments. The plot shows mean ratio of B2/U1 protein abundance fold change (log₂ scale). Hits with negative x-axis values are de-enriched for B2, whereas those with positive values are enriched in B2.

(C) Mean B2/U1 enrichment for the top 20 B2_5′ enriched and de-enriched proteins, ordered by ratio.

(D) Mean B2/U1 enrichment of ribosome-associated protein interactors, filtered by p < 0.05 in either B2 or U1 sample vs. biotin control.

(E) Western blots for RBP interactors with B2_5′, U1 (SL3 + 4) and Actb baits from sciatic nerve axoplasm. n = 6 (B2, U1, and biotin-only), n = 2 (Actb).

(F) Quantified normalized mean integrated density of nucleolin and PurB bands from the experiment of (E), mean ± SEM, n = 6. ^∗∗p < 0.01, ^∗∗∗∗p < 0.0001, two-tailed t test.

(G) Nucleolin-B2-SINE interaction requires the nucleolin GAR domain. B2-SINE RNA bait pull-downs from HEK cell lysates transfected with the indicated HA-nucleolin constructs, representative HA blots from four independent experiments.

We further compared axoplasm pull-downs with biotinylated B2_5′ (nt1–75) vs. a segment of non-coding U1 stem loops 3 and 4 (SL3 + 4) RNA of similar length and GC content. U1 was chosen because it is the small nuclear RNA component of the spliceosome and, as such, has strong secondary structure (as do B2-SINEs) but as a nucleus-restricted RNA should interact with a different subset of proteins than the GI-SINEs. Quantitative analyses confirm specificity in the pull-downs, with significant enrichment of B2-SINE co-precipitated proteins over U1 (Figures S4B and S4C; Table S4). Notably, ribosomal components are enriched for B2-SINE over U1 RNA (Figure S4D). Nucleolin did not show B2-SINE over U1 enrichment, although it was enriched for both vs. biotin control, perhaps because nucleolin has both nuclear and cytoplasmic roles in cells.

Finally, we validated B2-SINE-nucleolin interaction by western blots on axoplasm pull-downs with B2-SINE baits, in comparison with U1 and Actb baits, with additional readouts for four other RBPs (Figures 4E, S4E, and S4F). We found strong binding of B2-SINE to nucleolin and PurB and weak or nonexistent binding to the other three RBPs tested. U1 shows lesser binding than B2-SINE overall, with some binding to PurB, and Actb shows a different specificity profile. Comparative pull-downs from transfected HEK293 cells suggested that B2-SINE-nucleolin interaction is dependent on the C-terminal nucleolin GAR (glycine-arginine rich) domain (Figure S4G), which is essential for axonal transport.⁵ We further validated the role of the GAR domain by pull-downs of DRG neuronal lysates from heterozygous nucleolin-GAR-deletion mice.⁵ The B2-SINE baits pulled down endogenous full length, but not GAR-deleted, nucleolin from DRG neuron lysates (Figure 4F). In reciprocal pull-downs we used RT-qPCR to quantify significant enrichment of endogenous B2-SINE RNA, but not of U1 RNA, in nucleolin immunoprecipitates from sensory neurons (Figure 4G). B2-SINE interactions with ribosomes were validated in sensory-neuron-specific RPL22 RiboTag⁴⁶ pull-downs from mice after sciatic nerve injury. We found that ribosome-associated B2-SINEs are markedly upregulated in injured DRG compared with total (input) DRG RNA (Figures 4H and 4I). Thus, injury-induced B2-SINEs associate with the RBP nucleolin and with ribosomes in neurons.

Previous studies have reported interactions of nucleolin and ribosomes in nuclei and nucleolar complexes.^47,48 In this context, the above findings suggest that GI-SINEs may regulate formation of a nucleolin-ribosome complex. To test this, we transduced mouse DRG by intrathecal injection of AAV vectors for B2-SINE or control and cultured DRG neurons 14 days later. Proximity ligation assays (PLAs) revealed markedly increased cytoplasmic association between nucleolin and the 60S ribosomal subunit RPL11 in B2-SINE-transduced as compared with control cultures (Figures 5A, 5B, S5A, and S5B). There was no significant increase in nuclear PLA signal for association of RPL11 and nucleolin, suggesting that GI-SINE effects are mainly on cytoplasmic complexes. Previous work has shown that growth-enhancing perturbations of nucleolin in neurons cause subcellular shifts in protein synthesis foci away from axon tips.^3,5 We therefore sought to examine whether B2-SINE overexpression affects local protein synthesis by puromycin pulse-labeling of DRG cultures. B2-SINE overexpression significantly reduced puromycin incorporation in axon tips (Figures 5C–5E, S5C, and S5D). Thus, GI-SINEs enhance association of nucleolin and the translation machinery in neuronal somata, while attenuating protein synthesis at axon tips.

Figure 5.

B2-SINE affects protein synthesis in DRG neurons

Neurons were cultured from mouse L4/L5 DRG 2 weeks after intrathecal injection of either PHP.S-AAV-B2 or PHP.S-AAV-control vectors.

(A) Representative images of YFP-positive (transduced) neurons from 48 h cultures, fixed and processed for proximity ligation assay (PLA) between nucleolin (Ncl) and RPL11, with indicated markers. Scale bar, 20 μm.

(B) Quantification of PLA signal density in the nucleus and somatic cytosol of AAV-control (n = 49) vs. AAV-B2-transduced (n = 57) neurons. Mean ± SEM, ^∗p < 0.05, Mann-Whitney test.

(C) Cultures pulsed with puromycin for 10 min and processed as indicated. Puromycin and merged channels masked to show signal only in YFP-positive (AAV-transduced) regions. Scale bar, 50 μm.

(D) High-contrast zoomed-in images of puromycin signal in axon tips. Yellow lines show axon tip boundaries based on YFP channel (AAV transduction reporter). Scale bar, 5 μm.

(E) Quantification of puromycin intensities, mean ± SEM, n = 26 (soma, AAV-control), 57 (soma, AAV-B2), 101 (axon tips, AAV-control), 226 (axon tips, AAV-B2) from 3 (AAV-control) or 4 (AAV-B2) independent cultures, after ROUT outlier analysis with Q = 0.1%. ^∗p < 0.05, Mann-Whitney test.

See also Figure S5.

Figure S5.

Supporting data for puromycin and proximity ligation assay, related to Figure 5

(A) Untransduced DRG neurons cultured for 48 h before imaging as indicated. Scale bar, 10 μm.

(B) Quantification of proximity ligation assay (PLA) signal density in somata, as indicated. Mean ± SEM, n = 22 (PLA) or 21 (negative control) somata, ^∗∗∗∗p < 0.0001, Mann-Whitney test.

(C) Representative NFH-masked images of untransduced DRG neurons pulsed with either puromycin alone or pre-incubated with anisomycin followed by addition of puromycin. Scale bar, 50 μm.

(D) Comparison of puromycin intensity in somata and axon tips between puromycin-only and anisomycin + puromycin-treated DRG neurons. Mean ± S.E.E., Soma n = 29 or 40, Axon tips n = 115 or 162 for puromycin or anisomycin+puromycin, respectively, ^∗∗p < 0.01, ^∗∗∗∗p < 0.0001, Mann-Whitney test.

Specific targeting of GI-SINEs

We sought to identify GI-SINE-specific sequence motifs for selective targeting without general perturbation of all B2-SINEs. To this end, we compared GI-SINE vs. non-injury-regulated B2-SINE RNA sequences using STREME motif search⁴⁹ and identified two GI-SINE consensus motifs (Figures 6A, S6A, and S6B). Antisense oligonucleotides (ASOs) were designed to target the GI-SINE motifs (Figure 6A), with locked nucleic acid (LNA) bases at both ends to enhance avidity for the target sequence.²⁵ FISH probes for GI-SINE-specific sequences effectively visualized endogenous GI-SINE expression in DRG neuronal cell bodies (Figures 6B and 6C) and axons (Figures 6D and 6E). The FISH also confirmed reduction of GI-SINE expression upon ASO treatment. To verify this further and ensure no overlap between ASOs and FISH probes, we repeated the experiment with a more restricted mix of three ASOs, targeted to nucleotides 10–30 in GI-SINEs. The FISH probe is designed to interact with nucleotides 39–78 in GI-SINEs, thus the sequence segments are well separated. FISH revealed clear downregulation of GI-SINE expression in both cell bodies (Figures S6C and S6D) and axons (Figures S6E and S6F) with the restricted ASO mix. Moreover, independent quantification of this experiment by ddPCR confirmed downregulation of GI-SINE expression by the restricted ASO mix (Figure S6G).

Figure 6.

Disrupting GI-SINE inhibits nucleolin-ribosome complexes and axon growth

(A) GI-SINE consensus sequence obtained by motif analysis. Colored highlights show motifs 1 and 2 by STREME motif search that have the highest GI-SINE specificity. Black boxed sites mark conserved box A and box B domains of the internal RNA Pol III promoter. Red dashed boxes and black dashed underline show GI-SINE ASO mix and GI-SINE RNA-FISH probe target sites, respectively.

(B) GI-SINE RNA-FISH in LNA-ASO-transfected DRG neurons. Neurons were transfected with ASO mix (5 ASOs) targeting GI-SINE or with ASO control and probed with GI-SINE RNAscope probe or negative control probe (bacterial DapB mRNA) and counterstained with NFH (neurofilament heavy chain). Scale bar, 10 μm.

(C) Mean RNA-FISH probe intensity in DRG neuronal somata. Mean ± SEM of n = 28, 31, and 38 somata for control probe, ASO control and ASO GI-SINE, respectively, from 3 independent cultures. ^∗∗∗∗p < 0.0001, one-way ANOVA, Tukey’s multiple comparison test.

(D) Representative images of axonal RNA-FISH with NFH counterstain from the experiment described in (B).

(E) RNA-FISH mean signal in axon segments. Mean ± SEM of n = 58, 135, and 150 axon segments for control probe, ASO control and ASO GI-SINE, respectively, from 2 independent cultures. Scale bar, 5 μm, ^∗∗∗∗p < 0.0001, one-way ANOVA, Tukey’s multiple comparison test.

(F) Heatmap of B2-SINEs expression ratio in GI-SINE-ASO-transfected DRG neurons. Data are based on differential expression analysis from RNA-seq (150 bp reads, paired-end sequencing) of GI-SINE-ASO and control-ASO-transfected DRG cultures extracted 2 days after transfection. Fold-change expression ratios of significantly regulated (fold change p < 0.05) GI-SINE-matched elements and all other B2-SINE are shown (n = 21 GI-SINE and 323 other B2-SINE elements).

(G) Representative images of LNA-ASO-transfected and untransfected (Unt.) neurons labeled with NFH and imaged for axon outgrowth analysis, scale bar, 50 μm.

(H) Ratio of axon-growing vs. non-growing NFH-positive neurons. Neurons with longest axons < 2× their soma diameters were classified as non-growing. Lines depict mean ± SEM of n = 4 independent culture experiments. ^∗p < 0.05, one-way ANOVA and Tukey’s multiple comparisons test.

(I) Total axonal outgrowth per NFH-positive neuron, mean ± SEM of n = 3,730, 3,632, 3,044, and 3,432 neurons in untransfected (Unt.), ASO control, GI-SINE ASO, and ASO B2_146–166 groups, respectively, pooled from 4 separate culture experiments. ^∗∗∗∗p < 0.0001, one-way ANOVA, Kruskal-Wallis with Dunn’s multiple comparisons test.

(J) Cultured DRG neurons transfected with either GI-SINE ASO mix, control ASO or untransfected. Neurons were fixed 48 h after transfection and nucleolin-RPL11 PLA was carried out, followed by NFH and DAPI staining. Representative z-projected confocal images. Scale bar, 10 μm.

(K) Density of PLA spots was quantified in the cytosol. Data shown are mean ± SEM of n = 21, 29, 31, and 37 neurons in untransfected (Unt.), ASO control, GI-SINE ASO, and ASO B2_146–166 groups, respectively, pooled from 3 independent culture experiments. ^∗∗∗p < 0.001, one-way ANOVA, Kruskal-Wallis with Dunn’s multiple comparisons test.

See also Figures S6 and S7.

Figure S6.

ASOs targeting GI-SINE expression, related to Figure 6

(A) Top two GI-SINE-specific motifs from STREME motif search results, mapping on 391 GI-SINEs vs. 6,665 non-regulated B2-SINEs.

(B) STREME motif search results for top 8 GI-SINE sequence motifs used to assemble the consensus motif shown in Figure 6A.

(C) GI-SINE RNA-FISH on DRG neurons transfected with LNA-ASO mix targeted to a GI-SINE sequence stretch distinct from that recognized by the FISH probe. Neurons were transfected with ASO control or with GI-SINE ASOs targeting GI-SINE consensus nucleotides 10–30. FISH was conducted with GI-SINE RNAscope probe matching nucleotides 39–78 (Figure 6) or a negative control bacterial DapB mRNA probe. Scale bar, 10 μm.

(D) RNA-FISH probe intensity in DRG neuronal somata, mean ± SEM, n = 71, 31, and 39 somata for control, ASO control, and ASO GI-SINE, respectively. ^∗∗∗∗p < 0.0001, one-way ANOVA, Tukey’s multiple comparison test.

(E) As in (C) for axon segments. Scale bar, 5 μm.

(F) RNA-FISH probe intensity in axon segments, mean ± SEM of n = 71, 107, and 100 axon segments for control, ASO control and ASO GI-SINE, respectively. ^∗∗∗∗p < 0.0001, one-way ANOVA, Tukey’s multiple comparison test.

(G) Droplet digital PCR (ddPCR) analysis of GI-SINE RNA expression in cultured DRG neurons transfected with GI-SINE ASO (nt10–30) compared with ASO control. GI-SINE RNA expression is normalized to Hmgb1. n = 3, mean ± SEM, ^∗∗p < 0.01, two-tailed t test.

We then used RNA-seq to further evaluate the effects of ASOs on differential expression of B2-SINEs. GI-SINEs with statistically significant differential expression between ASO- and control-treated cultures were largely downregulated, whereas other B2-SINEs were either unchanged or regulated in both directions (Figure 6F). Based on the previous findings, effective loss of function of GI-SINEs is expected to affect neuronal growth and nucleolin-ribosome interactions. Indeed, axonal outgrowth was significantly reduced in GI-SINE ASO-treated cultures compared with controls (Figures 6G–6I). PLA revealed significantly reduced cytoplasmic nucleolin-RPL11 association upon treatment with anti-GI-SINE ASOs as compared with control ASO (Figures 6J and 6K). Similar results were observed for both axon outgrowth (Figures S7A–S7C) and nucleolin-RPL11 interaction (Figures S7D and S7E) upon ASO treatment of DRG neurons after conditional lesion of the sciatic nerve.

Figure S7.

Characterization of B2-SINE ASOs, related to Figure 6

(A) L4/L5 DRG neurons cultured 72 h after sciatic nerve crush. ASO transfections and outgrowth analyses as for Figures 6G–6K. Scale bar, 50 μm.

(B) Longest axon ratios > 2× soma diameter from all NFH-positive somata, n = 5.

(C) Total outgrowth, mean ± SEM, n = 1,577 (ASO control), 1,000 (ASO GI-SINE). ^∗∗∗∗p < 0.0001, one-way ANOVA, Kruskal-Wallis with Dunn’s multiple comparison test.

(D) Ncl-RPL11 PLA in cultured DRG after conditioning sciatic nerve injury and ASO transfections as in Figure 6J. Scale bar, 10 μm.

(E) PLA spot density in cytosol, mean ± SEM, n = 60 (ASO control), 61 (ASO GI-SINE), ^∗∗∗∗p < 0.0001, one-way ANOVA, Kruskal-Wallis with Dunn’s multiple comparisons test.

(F) B2-SINE secondary structure model,^43,45 blue and red dashed lines delineate the two biotin-conjugated constructs used for pull-downs. Overlaying lines mark B2-antisense oligonucleotide (ASO) position.

(G) Nucleolin western blot after biotinylated-B2_5′ (nt1–75) and -B2_3′ (nt75–166) RNA pull-down from sciatic nerve axoplasm. Biotin-B2 RNA segments pre-annealed with DNA ASO targeting different B2-RNA sequence sites to test effects on B2-nucleolin interaction.

(H) Nucleolin integrated density in pull-down samples of B2_3′ pre-annealed with ASO-146-166 normalized to B2_3′ alone. Mean ± SEM, n = 3.

Finally, we sought additional interfering agents for GI-SINEs by performing a screen for ASOs that competitively interfere with GI-SINE functions, testing a series of 20-mer ASOs tiled along the GI-SINE consensus sequence for inhibition of SINE binding to nucleolin (Figure S7F). ASO_146–166 targeted to the 3′ region of the SINE sequence (hence distinct from the internal promoter region; Figure 6A) effectively reduces nucleolin pull-down by the B2-SINE 3′ region bait (Figures S7G and S7H). DRG neurons treated with this ASO revealed reduced neuronal outgrowth (Figures 6G–6I) and nucleolin-ribosome interaction (Figures 6J and 6K) at comparable levels as those seen with the previously tested ASO GI-SINE. Thus, targeting of GI-SINEs by two distinct ASOs reveals loss-of-function effects on both neuron growth and nucleolin-ribosome interactions.

Discussion

The findings above establish the GI-SINEs as intrinsic axon growth regulators. This unique subset of B2-SINE RNAs integrates mRNA localization and translation to enhance sensory neuron growth after axon injury. Exogenous expression of B2-SINEs also enhances growth in CNS neurons that do not upregulate endogenous SINE elements upon nerve injury. The GI-SINEs are induced in response to retrograde injury signals via activation of AP-1 transcription factors and modulate growth and protein synthesis (Figure 7). Thus, in addition to the well-documented stochastic, genome-modifying, and deleterious effects of transposons,^{50,51,52,53,54} our study reveals embedding of specific REs as intrinsic components of a physiological circuit, expanding the roles of such elements in mammalian genomes.

Figure 7.

GI-SINEs are intrinsic to a physiological neuronal growth circuit

Upper: Axon-soma communication in neuronal growth is mediated by kinesin (K)-dependent anterograde transport of mRNAs on nucleolin (Nucl), including mTor and importin β1 (β), and retrograde transport of their encoded proteins in a complex with dynein (D) and an importin α (α) after local protein synthesis. The GI-SINEs are induced by a retrograde injury signal (S) through AP-1 transcription and then regulate nucleolin-ribosome interactions and local translation in the cytoplasm.

Lower: GI-SINEs are embedded in a physiological circuit that responds to a canonical transcription factor complex to impact protein synthesis. Such circuits may control other functions beyond neuronal growth, and functionally analogous circuits may exist for related repeat elements in other species.

Our findings show that nerve-injury-induced regulation of the GI-SINEs is restricted to ⁓0.3% of the over 150,000 B2-SINE loci³⁵ in the mouse genome. Moreover, these loci share specific conserved sequence motifs. This high specificity, together with the lack of GI-SINE upregulation in injured CNS neurons, suggests that selective mechanisms control GI-SINE expression. B2-SINE transcription is thought to require Pol III activity on the integral SINE promoter, and additional RNA Pol II-related complexes for polyadenylation of the B2 transcript.^16,55,56 More generally, Pol II activity has been suggested to facilitate Pol III transcription through enhancer sites and chromatin remodeling.^57,58 GI-SINE loci are specifically associated with AP-1 transcription-factor-binding sites, and GI-SINE expression is repressed by dominant-negative Fos or ATF3 KO, implicating AP-1 transcription factors in GI-SINE expression specificity. Indeed, AP-1 and other specific transcription-factor-binding sites are enriched within individual RE subfamilies,^59,60 and AP-1 sites can act as enhancers in diverse transcriptional networks.^61,62 AP-1 transcription factors and their co-factor ATF3 are robustly activated in nerve injury,^41,63,64 consistent with AP-1 sites providing a likely determinant for specific activation of GI-SINEs. AP-1 activation is observed in many physiological paradigms⁶⁵; hence, GI-SINEs might have diverse biological roles beyond neuronal growth regulation.

Other studies have suggested that SINE RNAs, including B2-SINEs, can act as repressors or enhancers of proximal gene transcription.^23,52,66 Although we cannot rule out such mechanisms contributing to GI-SINE activities, we found no overlap between GI-SINEs and the known enhancer SINEs,²³ and most GI-SINE loci are in intergenic regions of the genome, suggesting other modes of action. Indeed, proteomics data and cellular analyses suggest that GI-SINEs bind both nucleolin and ribosomes in the cytoplasm, enhancing their association to modify local translation in neurons. Previous studies have shown that B2-SINEs have ribosome binding capacity, and B2-SINE domains in SINEUP RNAs enable enhanced translation of specific transcripts.^67,68 The proposed mode of action of GI-SINEs—binding an RBP and thereby associating its mRNA cargos with ribosomes—may prioritize translation of nucleolin-associated mRNAs and shift translation away from axon tips. Notably, GI-SINEs require the GAR domain of nucleolin for effective binding, and this domain is required for axonal targeting.⁵ GAR domain perturbation and subcellular translation shifts were previously linked with enhanced growth,^3,5 and we have shown above that exogenous expression of a consensus B2-SINE sequence can elicit growth in two types of CNS neurons. Thus, exogenous expression of GI-SINE-related elements might be useful as a general growth-enhancing strategy in diverse neuron subtypes.

SINEs are among the most prolific mobile genomic elements in eukaryotes, emerging on multiple occasions by de novo evolution from non-protein-coding genes such as tRNAs and 7SL.⁶⁹ Alu elements are perhaps the closest functionally related non-coding RNAs to B2-SINEs in primate genomes, and Alu monomers are structurally and functionally equivalent to SINE domains in human SINEUP RNAs.^70,71 Moreover, Alu-related elements can associate directly and indirectly with both ribosomes^72,73 and nucleolin.^74,75 Indeed, examples of convergent evolution of SINE-related RNAs to the same function in rodent and primate lineages have already been described.^76,77 Thus, it is likely that elements with similar functions to GI-SINEs will be found in different eukaryotic lineages.

To summarize, this study shows that the GI-SINE subfamily of non-coding RNA REs act as intrinsic regulators of axon-soma communication mechanisms regulating neuronal growth. The mechanism exemplifies how non-coding RNA elements can be embedded within an intrinsic intracellular transport, signaling, transcription, and translation circuit to regulate fundamental aspects of cell growth. Identification of GI-SINEs as effectors linking a canonical transcription factor complex to localized translation suggests that similar mechanisms may function in a broad range of physiological paradigms. It will be interesting to determine which other functions, beyond neuronal growth, are controlled by such circuits.

Limitations of the study

The neuron growth assays conducted in this study do not suffice for comprehensive understanding of GI-SINE roles in nerve regeneration, and future studies in multiple nerve injury paradigms in vivo will be required to clarify these issues. In addition, it is important to note that our findings on GI-SINE interactions with nucleolin and ribosomes do not exclude the possibility of additional modes of action of GI-SINEs, and direct testing of other genomic and transcriptional mechanisms will be of interest in the future. Finally, although we show that GI-SINEs require the GAR domain for interaction with nucleolin, and our previous work has shown that the GAR domain determines subcellular localization of nucleolin, we still lack direct evidence for GI-SINE effects on nucleolin transport or localization. Future structural characterization of these RNA-protein complexes may shed light on this issue.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to, and will be fulfilled by, the lead contact, Mike Fainzilber (mike.fainzilber@weizmann.ac.il).

Materials availability

Materials generated in this study will be available upon reasonable request.

Data and code availability

RNA-seq datasets are available from the NCBI GEO database under accession numbers GEO: GSE231596 and GEO: GSE279906. Proteomics data are available from the ProteomeXchange PRIDE partner repository under dataset identifier PRIDE: PXD060976. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request. This paper does not report original code.

Acknowledgments

We thank Clifford Woolf for generously providing ATF3 conditional allele mice, Vladimir Kiss and Reinat Nevo for help and advice on imaging analyses, Ron Rotkopf for guidance on statistical tests, Dalia Gordon for helpful discussions, Oded Singer and Yiming Zhang for AAV production, and Qing Wang for sample processing for RNA-seq. We gratefully acknowledge funding from the following sources: Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (M.F., A.L.B., Z.H., J.L.T., M.H.T., and R.K.), European Research Council (ERC advanced grant, GrowthSINE, 101141522 to M.F.), Weizmann Center for Research on Injury and Regeneration (M.F.), National Institutes of Health grant R01-NS117821 (J.L.T.), Estonian Research Council grants MOBTP192 and PRG2206 (I.K.), Merkin Peripheral Neuropathy and Nerve Regeneration (MPNNR) Center (P.K.S.), Israel Academy of Sciences and Humanities Postdoctoral Fellowship (E.E.Z.), Azrieli International Postdoctoral Fellowship (AB), HFSP Long Term Fellowship (LT001195/2012-L to C.A.A.), Chaya Professorial Chair in Molecular Neuroscience (M.F.), and South Carolina Smartstate Chair in Childhood Neurotherapeutics (J.L.T.).

Author contributions

Hypotheses and concepts were developed by I.K., E.E.Z., C.A.A., and M.F. RNA-seq was performed and analyzed by R.K.; ATAC-seq by R.B.-T.P., I.U., and Y.C.; bioinformatics analyses by R.K., S.B.-D., E.F., D.L., and Y.R.; and protein mass spectrometry by J.A.O.-P. Experiments were carried out by I.K., E.E.Z., A.B., A.M., S.H., R.J.D., I.D.C., J.L., P.K.S., E.v.N., R.B.-T.P., E.F., O.A., N.S., N.O., S.A., and S.M. Research supervision was performed by M.F., I.R., I.U., A.L.B., Z.H., J.L.T., and M.H.T. The manuscript was written by M.F., E.E.Z., and I.K. and was revised and edited by all the authors.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-nucleolin	Cell Signaling Technology	Cat. # 14574; RRID:AB_2798519
Mouse anti-nucleolin	Santa-Cruz	Cat. # sc8031; RRID:AB_670271
Rabbit anti-nucleolin	abcam	Cat. # ab50279; RRID:AB_881762
Goat anti-GFP	Abcam	Cat. # ab6673; RRID:AB_305643
Rabbit anti-GFP antibody	Abcam	Cat. # ab290; RRID:AB_303395
Chicken anti-NFH	Abcam	Cat. # ab72996; RRID:AB_2149618
Mouse anti-puromycin [12D10]	Millipore	Cat. # MABE343; RRID:AB_2566826
Rabbit anti-HA	Sigma Aldrich	Cat. # H6908; RRID:AB_260070
Rat anti-HA	Sigma Aldrich	Cat. # 11867423001; RRID:AB_390918
Rabbit anti-Rpl11	Abcam	Cat. # ab79352; RRID:AB_2042832
Rabbit anti-RFP	Rockland	Cat. # 600-401-379z; RRID:AB_2209751
Mouse anti-Neurofilament marker (SMI-312)	BioLegend	Cat. # 837904; RRID:AB_2566782
Mouse anti-β-III-tubulin (Tuj1)	BioLegend	Cat. # 801213; RRID:AB_2313773
Chicken anti- β-III-tubulin	Millipore	Cat. # AB9354; RRID:AB_570918
Mouse anti-HnrnpK	Abcam	Cat. # ab39975; RRID:AB_732981
Rabbit anti-Purb	Abcam	Cat. # ab130999; RRID:AB_11155516
Rabbit anti-HnrnpH1	Abcam	Cat. # ab10374; RRID:AB_297112
Rabbit ant-Elavl1 (HuR) [4C8]	Abcam	Cat. # ab136542; RRID: N/A
Isotype control Rabbit IgG (DA1E)	Cell Signaling Technology	Cat. # 3900; RRID:AB_1550038
Donkey anti-rabbit-Alexa594	Thermo Fisher Scientific	Cat. # A-21207; RRID:AB_141637
Mouse anti-digoxigenin-Cy3	Jackson ImmunoResearch	Cat. # 200-162-156; RRID:AB_2339025
Goat anti-mouse-Cy3	Jackson ImmunoResearch	Cat. # 115-165-003; RRID:AB_2338680
Donkey anti-chicken-Cy5	Jackson ImmunoResearch	Cat. # 703-175-155; RRID:AB_2340365
Goat anti-rabbit-FITC	Jackson ImmunoResearch	Cat. # 111-095-003; RRID:AB_2337972
Donkey anti-mouse-FITC	Jackson ImmunoResearch	Cat. # 715-095-150; RRID:AB_2340792
Goat anti-chicken-Alexa488	Jackson ImmunoResearch	Cat. # 103-545-155; RRID:AB_2337390
Donkey anti-chicken-Alexa488	Jackson ImmunoResearch	Cat. # 703-545-155; RRID:AB_2340375
Goat anti-rabbit-HRP	Jackson ImmunoResearch	Cat. # 111-035-144; RRID:AB_2307391
Mouse anti-mouse-HRP	Jackson ImmunoResearch	Cat. # 115-035-062; RRID:AB_2338504
Bacterial and virus strains
PHP.S-U6-B2_EF1a-YFP	This paper	N/A
PHP.S-U6-shControl_EF1a-YFP	This paper	N/A
AAV2-U6-B2_EF1a-YFP	This paper	N/A
AAV2-U6-shControl_EF1a-YFP	This paper	N/A
AAV2-Rcomet (tdTomato)	This paper	N/A
Chemicals, peptides, and recombinant proteins
Streptavidin-HRP	Abcam	Cat. # ab7403
Phalloidin-Rhodamine	Thermo Fisher Scientific	Cat. # R415
DAPI	Thermo Fisher Scientific	Cat. # 62248
Papain	Sigma-Aldrich	Cat. # P4762
Collagenase-II	Roche	Cat. # 11179179001
Dispase-II	Roche	Cat. # 04942078001
Poly-L-lysine	Sigma-Aldrich	Cat. # P4832
Laminin	Invitrogen	Cat. # 23017-015
Primocin	InvivoGen	Cat. # ant-pm-1
Streptavidin Myone-Dynabeads	Thermo Fisher Scientific	Cat. # 65002
Dynabeads Protein G	Invitrogen	Cat. # 10003D
Puromycin	Sigma-Aldrich	Cat. # P8833
Anisomycin	Sigma-Aldrich	Cat. # A9789
D-biotin	Sigma-Aldrich	Cat. # B4501
Cytosine β-D-arabinofuranoside (AraC)	Sigma-Aldrich	Cat. # C1768
JetPEI DNA transfection reagent	Polyplus	Cat. # 101000053
Dharmfect-4 transfection reagent	Dharmacon	Cat. # T-2004-03
Visikol HISTO-1 & HISTO-2 kit	Visikol	Cat. # HH-10
Fluoromount-G	Southern Biotechnology	Cat. # 0100-01
EDTA-free protease inhibitor cocktail	Sigma-Aldrich	Cat. # P8340
cOmplete Protease Inhibitor cocktail	Roche	Cat. # 11836170001
Phosphatase inhibitor cocktail 3	Roche	Cat. # 200-664-3
RNAsin Ribonuclease Inhibitor	Promega	Cat. # N2111
Ribonucleoside vanadyl complexes	Sigma-Aldrich	Cat. # R3380
RNeasy mini kit	Qiagen	Cat. # 74104
RNeasy micro kit	Qiagen	Cat. # 74004
RNeasy mini plus kit	Qiagen	Cat. # 74134
TRI Reagent	Sigma-Aldrich	Cat. # T9424
Zymo-Spin IC columns	Zymo	Cat. # R1013
RiboRuler Low Range RNA Ladder	Thermo Fisher Scientific	Cat. # SM1831
SuperScript™ III First-Strand Synthesis System	Thermo Fisher Scientific	Cat. # 18080051
Sensifast cDNA synthesis kit	Meridian Biosciences	Cat. # BIO-65053
SYBR green FastMix	Quanta Biosciences	Cat. # 95072
Digoxigenin succinimide ester	Roche	Cat. # 11333054001
E. coli tRNA	Roche	Cat. # B1107540
ProLong Gold Antifade	Thermo Fisher Scientific	Cat. # P36930
TMTpro™ 16plex Label Reagent Set	Thermo Fisher Scientific	Cat. # A4452
Critical commercial assays
NaveniFlex MR- Proximity Ligation Kit	Navinci	Cat. # NF.MR.100.1
NaveniFlex Detection Reagents	Navinci	Cat. # NF.100.2
RNAscope® Assay for Adherent Cells Cultured on Coverslips	Advanced Cell Diagnostics	N/A
BaseScopeTM v2 Assay	Advanced Cell Diagnostics	N/A
Droplet digital PCR Assay	Bio-Rad	N/A
Truseq Stranded Total RNA Library Prep Gold	Illumina	Cat. # 20020599
Illumina Stranded Total RNA Prep, ligation with Ribo-Zero Plus	Ilumina	Cat. # 200405297
QuantSeq 3′ mRNA-seq Library Prep (REV)	Lexogen	Cat. # 225.96
Deposited data
RNA sequencing	This paper	GEO: GSE231596, GSE279906
Mass-spectrometry proteomics	This paper	PRIDE: PXD041633
Experimental models: Cell lines
HEK293 (human)	ATCC	CRL-1573; RRID:CVCL_0045
N2a (Neuro-2a) (mouse)	ATCC	CCL-131; RRID:CVCL_0470
Experimental models: Organisms/strains
C57BL6/OlaHSD mouse	Envigo	N/A
Advillin-Cre mouse	da Silva et al.78	N/A
RiboTag (loxP-Rpl22HA) mouse	Sanz et al.79	N/A
Atf3-exon3-flox mouse	Renthal et al.41	N/A
Oligonucleotides
B2-SINE consensus sequence	Table S5	N/A
Biotinylated B2-SINE & U1 RNA	Table S5	N/A
Antisense Oligonucleotides (ASOs)	Table S5	N/A
qPCR primers and probes	Table S5	N/A
RNA-FISH probes	Table S5	N/A
Northern blot probes	Table S5	N/A
Recombinant DNA
PHP.S plasmid (AAV envelope)	Chan et al.80	N/A
AAV2/2 plasmid (AAV envelope)	Boston Children’s Hospital Viral Core / Salk Viral Vector Core Facility	N/A
AAV-U6-B2_EF1a-YFP plasmid	This paper	N/A
AAV-U6-shControl_EF1a-YFP plasmid	Addgene	Plasmid # 85741; RRID:Addgene_85741
pAAV-hSyn-A-Fos-WPRE plasmid (AAV-A-Fos)	Marvaldi et al.81	N/A
AAV-hSyn-EGFP-WPRE plasmid (AAV-GFP control)	Marvaldi et al.81	N/A
HA-Dendra2-nucleolin plasmid	Doron-Mandel et al.5	N/A
HA-Dendra2-nucleolin_ΔGAR plasmid	Doron-Mandel et al.5	N/A
Software and algorithms
Fiji (ImageJ)	Schindelin et al.82	https://imagej.net/software/fiji/; RRID:SCR_002285
Imaris 9.9.1	Oxford Instruments (Bitplane)	https://imaris.oxinst.com/; RRID:SCR_007370
ImageExpress- Neurite Outgrowth module	Molecular Devices	https://www.moleculardevices.com/products/cellular-imaging-systems/acquisition-and-analysis-software/metaxpress; RRID:N/A
STREME 5.5.5 (MEME suite)	Bailey49	https://meme-suite.org/meme/tools/streme; RRID:SCR_001783
STAR aligner	Dobin et al.83	https://github.com/alexdobin/STAR; RRID:SCR_004463
UTAP 2.0	Kohen et al.84	; RRID: N/A
Bioconductor DESeq2 v1.36	Love et al.85	https://bioconductor.org/packages/release/bioc/html/DESeq2.html; RRID:SCR_015687
BEDtools suite 2.26.0	Quinlan and Hall86	https://bedtools.readthedocs.io/en/latest/index.html; RRID:SCR_006646
Bowtie2	Langmead and Salzberg87	https://bowtie-bio.sourceforge.net/bowtie2/index.shtml; RRID:SCR_016368
NGSplot	Shen et al.88	https://github.com/shenlab-sinai/ngsplot; RRID:SCR_011795
TOBIAS algorithm	Bentsen et al.38	https://github.com/loosolab/TOBIAS; RRID: N/A
Oligo software (Mac v7)	Molecular Biology Insights	https://www.oligo.net/; RRID: N/A
RSEM algorithm	Li and Dewey89	https://github.com/deweylab/RSEM; RRID:SCR_000262
Bioconductor edgeR	Robinson et al.90	https://bioconductor.org/packages/release/bioc/html/edgeR.html; RRID:SCR_012802
RepeatMasker (mm10)	RepeatMasker Open-3.0. 1996-2010	http://www.repeatmasker.org; RRID:SCR_012954
BEDOPS toolkit	Neph et al.91	https://bedops.readthedocs.io/en/latest/; RRID:SCR_012865
UCSC genome browser (mm10 assembly)	Karolchik et al.92	https://genome.ucsc.edu/; RRID:SCR_005780
HOMER algorithm	Heinz et al.93	http://homer.ucsd.edu/homer/motif/; RRID:SCR_010881
Protein Prospector	Chalkley et al.94	https://prospector.ucsf.edu/prospector/mshome.htm; RRID:SCR_014558
GOrilla gene ontology (v.Mar_2013)	Eden et al.95	https://cbl-gorilla.cs.technion.ac.il; RRID:SCR_006848
Prism 10	Graphpad	https://www.graphpad.com/; RRID:SCR_002798
Other
EASY-Spray PepMap RSLC C18 column	Thermo Fisher Scientific	Cat. # ES906
C18 100 μl OMIX tip	Agilent	Cat. # A57003100
ZipTip with 0.6 μL C18 resin	Milipore	Cat. # ZTC18S096

Experimental model and study participant details

All animal experiments were reviewed and approved by the Institutional Animal Care and Use Committees at the Weizmann Institute of Science, the University of South Carolina, University of California, San Diego and Harvard University, respectively. Adult C57BL6/OlaHSD mice were purchased from Harlan Laboratories (Envigo, Israel). For RiboTag experiments, Advillin Cre mice⁷⁸ were crossed with the RiboTag line,⁷⁹ as previously described.⁴⁶

Reagents and antibodies

Culture media and serum were from Thermo Fisher Scientific. Puromycin (P8833), anisomycin (A9789) and D-biotin (B4501) were purchased from Sigma. 5-biotinylated RNA was purchased from IDT. The following antibodies were used in this study: rabbit anti-nucleolin (Cell Signaling Technology, #14574, 1:1000 for WB), mouse anti-nucleolin (Santa-Cruz, sc8031, 1:50 for Immuno-PLA), goat anti-GFP (Abcam, ab6673, 1:1000 for IF), chicken anti-NFH (Abcam, ab72996, 1:1000 for IF), mouse anti-puromycin (Millipore, 12D10 clone, MABE343, 1:5000 for IF), rabbit anti-HA (Sigma, H6908, 1:1000 for WB), rabbit anti-Rpl11 (Abcam, ab79352, 1:100 for Immuno-PLA), rabbit anti-RFP (Rockland, cat#600-401-379z, 1:1000), donkey anti-Rabbit (Invitrogen, cat#A21207, 1:1000). Phalloidin-Rhodamine was from Thermo Scientific (R415, 1:500 for IF).

Cell cultures

HEK293 and Neuro-2a cell lines were purchased from ATCC and grown in high-glucose DMEM (Gibco), supplemented with 10% fetal bovine serum (Gibco), 100 U/mL penicillin and 100 μg/mL streptomycin.

Method details

Sciatic nerve crush injury model

Unilateral sciatic nerve crush was done in male C57B/6J mice at 8-12 weeks of age as follows, unless otherwise noted. Mice were first anesthetized by IP injection of Ketamine (100 mg/kg) and Xylazine (20 mg/kg), and a single 1 cm skin incision was performed along the hip towards the thigh to expose the muscles above the femur. The muscle or the femur and lateral thigh muscle were gently separated with scissors to expose the sciatic nerve. Close 1 mm curved serrated forceps were used to lift the sciatic nerve without pinching and then to spread the nerve segment to allow access of the crushing tweezers. Crush was done by pinching the nerve with straight #4 Dumont-style tweezers for 15 seconds holding the tweezers in flat parallel to body followed by additional 15 seconds of pinching in the same position, holding the tweezers in a perpendicular angle. After crush, skin incision was closed by clipping or suturing and mice were injected with Carprofen (SC, 20 mg/kg) for analgesia. Lakripos gel was applied to the eyes to prevent dryness during anesthesia.

Neuron cultures

Unless otherwise noted DRG neuron cultures were prepared as previously described.⁵ Briefly, DRGs were dissected from adult mice and dissociated for 20 min with 100 U of papain (P4762, Sigma) followed by 20 min treatment with 1 mg/ml collagenase-II (11179179001, Roche) and 1.2 mg/ml Dispase-II (04942078001, Roche) for additional 25-30 minutes. The ganglia were mechanically triturated by a fire-polished Pasteur pipette in HBSS supplemented with 10 mM glucose and 5 mM HEPES (pH 7.35). Cells were then laid on a 20% Percoll cushion in Leibovitz L15 medium and recovered through centrifugation at 1000 g for 8 min. Neurons were plated on glass coverslips pre-coated with poly-L-lysine (P4832, Sigma) and laminin (23017-015, Invitrogen) in F12 supplemented with 10 % Fetal Bovine Serum and Primocin (InvivoGen). 10μM Cytosine β-D-arabinofuranoside (AraC) was added to the medium 24 hours after plating to prevent growth of mitotic cells.

DRG neuron cultures for RNA-FISH were prepared as follows: DRGs were harvested into Hybernate-A medium (BrainBits) and then dissociated as described.⁹⁶ After centrifugation and washing in DMEM/F12 (Life Technologies), dissociated ganglia were resuspended in DMEM/F12, 1 x N1 supplement (Sigma), 10% Fetal Bovine Serum, and 10 μM AraC. Dissociated DRGs were plated immediately on poly-L-lysine /laminin coated glass coverslips for RNAscope and immunostaining.

B2-SINE consensus sequence

The B2-SINE consensus sequence used in this study is based on Mus musculus B2-SINE sequences in the Repbase repository (version 3, 21/08/2021): gggctggaga gatggctcag tggttaagag cacctgactg ctcttccaga ggtcctgagt tcaattccca gcaaccacat ggtggctcac aaccatctgt aatgagatct gatgccctct tctggtgtgt ctgaagacag ctacagtgta cttacatata ataaataaat aaataaataa atcttaaaaa aaaaaaaaag aaagaaaaa

DNA constructs for viral transduction

For B2-SINE overexpression, the mouse B2-SINE consensus sequence was synthesized as a gBlock gene fragment by Integrated DNA Technologies (IDT). 44 nucleotides of 3′ end A-rich sequence (TAAATAAATAAATAAATCTTAAAAAAAAAAAAAAGAAAGAAAAA) were excluded due to difficulties in homopolymeric tract amplification in gene synthesis. Thus, the cloned B2-SINE consensus ends with a full AATAAA polyadenylation signal. In addition, a (T)₇ tract was added to the 3’ end as a Pol III terminator sequence. Finally, BamHI and XbaI sites were added for cloning to the 5’ and 3’ ends, respectively. The B2-SINE DNA fragment was cloned into an AAV shRNA expression vector under the U6 promoter (AAV-shRNA-ctrl, Addgene plasmid #85741), which also encodes a YFP-reporter under a separate promoter. The B2 consensus sequence was inserted instead of the control shRNA sequence. The original shControl plasmid was used to package AAV vectors for transduction control conditions.

B2-SINE expression and polyadenylation from this construct was validated by transfection into HEK293 cells using JetPEI (Polyplus), followed by Northern blot analysis in comparison with un-transfected HEK293 cells and N2a cells (Figure S2A). Note that since HEK293 are human cells, they do not express endogenous B2-SINEs.

The A-Fos dominant negative is based on previously described construct.³⁹ A-Fos and GFP control construct were cloned into a neuron-specific AAV backbone as previously described.⁸¹

Anti-Sense Oligos (ASO) transfection into cultured DRG neurons

To perturb GI-SINE expression in DRG neurons, a mix of five antisense sequences was designed based on STREME searches revealing two top enriched motifs in GI-SINE transcript sequences compared to other B2-SINE transcripts expressed in DRG. Perturbation of B2-nucleolin interaction was done using a B2 antisense oligo targeting nt146-166. Sequences of all ASOs are provided in Table S5. LNA-DNA ASOs for transfection were purchased from IDT as Affinity-Plus®, phosphorothioate DNA oligonucleotides with locked nucleic acid flanks (known also as GAPmers).

LNA-ASO were transfected into cultured DRG neurons using Dharmafect-4 reagent. Primary DRG neurons suspension was plated in antibiotics-free, FBS supplemented F-12 media in 24 or 12-well plates for 1 hour prior to transfection. ASOs and Dharmafect-4 were mixed in F-12 media (serum and antibiotics free, 100 μl for 24-well plate & 200 μl for 12-well plates) for 20 minutes before applying on cells. ASOs were used in final concentration of 50 nM (10 nM each for individual ASO in the anti-GI-SINE pool), Dharmafect-4 was diluted 1:200 in the final growth volume: 400 μl for 24-wells & 800 μl for 12-well. Media was changed to complete F-12 after 24 hours.

Generation of AAV-PHP.S vectors and transduction of DRG neurons

AAV-PHP.S⁸⁰ vectors for expression of B2, shControl, GFP and A-Fos constructs in DRG neurons were produced using AAVpro HEK293 cells. AAV titer was measured using DNA qPCR (for AAV ITR) and equilibrated at a concentration of 1-2x10¹⁰ gc/μl (genomic copies/μl) in PBS, aliquoted and stored at -80°C. For in-vivo transduction of DRG neurons, mice were first anesthetized by IP injection of Ketamine (100 mg/kg) and Xylazine (20 mg/kg), followed by surgical exposure of a small segment of back skin to allow access to the lumbar area. AAV suspension at volumes of up to 5 μl was injected into the sub-dura between L4-L5 or L5-L6 using a 10 μl Hamilton syringe. For transduction of dissociated DRG neurons in cell-culture, AAV suspension was mixed with whole-mouse DRG cell-suspension before plating (15 μl of AAV suspension per 1 mouse).

Optic nerve injury model

To perform optic nerve crush in anesthetized mice, the optic nerve was exposed intra-orbitally and crushed with fine forceps (Dumont #5 FST) for ∼2 seconds, approximately 1 mm behind the optic disc. Eye ointment was applied post-operation to protect the cornea, and buprenorphine was used as a postoperative analgesic. Intravitreal injection of 2 μl CTB conjugated with Alexa-647 (Life Technology) was performed 48 hours before sacrifice. Two weeks after the optic nerve crush, mice were perfused transcardially with ice-cold PBS followed by 4% paraformaldehyde (PFA, Sigma). Mouse heads were post-fixed for 3 hours in 4% PFA and optic nerves were dissected out. The meninges surrounding the nerve were removed, and the nerves were cleared using the reagents and protocol provided by Visikol®. Briefly, the nerves were dehydrated with 100% methanol for 4 minutes and then transferred into Visikol Histo-1 solution for overnight incubation at 4^oC. The nerves were then incubated in Visikol Histo-2 solution for 2 hours before mounting them in Visikol Histo-2 solution. Optic nerves were imaged using an LSM710 confocal microscope. Z-stack scanning was used to capture all regenerated axons. The number of CTB+ axons was quantified along the nerve at different distances away from the crush site.

Intravitreal injections of AAVs carrying B2-SINE overexpression vector or control were performed in anesthetized mice (ketamine 100-120 mg/kg and xylazine 10 mg/kg) two weeks before the optic nerve injury. A pulled-glass micropipette was inserted into the vitreous chamber (near the peripheral retina behind the ora serrata) to take out 2 μl of the vitreous humor. Then, 2 μl of the virus solution was injected using another glass micropipette into the vitreous chamber through the same opening. Antibiotic ophthalmic ointment was applied to the eyes and mice were warmed on a heating pad until fully awake. Serotype AAV2/2 was generated by the Boston Children’s Hospital Viral Core and a concentration of ∼5 x 10¹² gc was used for the intravitreal injections.

Corticospinal tract (CST) injury model

All surgeries were done under deep anesthesia using isoflurane gas. Euthanasia for tissue collection was performed by injection of an overdose amount of ketamine and xylazine in accordance with AVMA Guidelines for Euthanasia of Laboratory Animals. For anterograde labeling of CST neurons projecting into the spinal cord, wild type C57Bl/6 mice were injected with AAV2-Rcomet (tdTomato reporter) bilaterally into the cortical parenchyma spanning the motor cortex as previously described.³⁴ Viral titers were: AAV2-B2 OE and AAV2-Control 0.585 x 10¹² gc and AAV2-Rcomet 0.5 x 10¹² gc per hemisphere. Two weeks later, dorsal column lesions were performed as previously described,³⁴ and three weeks after lesion animals were euthanized for histological analysis.

For CST histology, spinal cords and brains were removed from the vertebrae and skull and serially sectioned in the sagittal plane at 35 μm intervals. Every sixth section was used for antibody labelling. All steps were performed at room temperature unless otherwise noted. Sections were washed three times in PBST (0.25% TX-100 in PBS) for 10 minutes and blocked in PBST+ 5% donkey serum for 1 hour, then incubated overnight in PBS + 5% donkey serum with primary antibodies at 4^oC. Sections were washed three times in PBST and incubated 1 hour with secondary antibodies, then washed three times in PBST for 10 minutes, and mounted on glass slides with fluoromount G (Southern Biotechnology). For quantification, percent of axons in the grey matter was normalized to corticospinal tract tracing in the main tract, relative to the mean in the control condition.

RNA sequencing (RNA-seq)

Purified RNA was quality-checked on either Bioanalyzer RNA chips or TapeStation RNA tapes (Agilent). RNA Integrity Numbers (RINs) were 7 or above. The RNA-seq libraries were prepared with ribosomal-RNA-depletion based methods (Truseq Stranded Total RNA Library Prep Gold or Illumina Stranded Total RNA Prep, ligation with Ribo-Zero Plus (Illumina)) depending on the initial RNA amounts. At least 100 ng RNA was used for Truseq total RNA prep, and 8 ng RNA was used for Illumina total RNA prep. Libraries were indexed and pooled for sequencing on HiSeq 2000 or NovaSeq 6000 (Illumina) for paired-end 2 x 75 or 2 x 150.

QuantSeq libraries were prepared using QuantSeq 3′ mRNA-seq Library Prep (REV) (Lexogen). At least 50 ng RNA was used in the preparation. The libraries were indexed and pooled for sequencing on NextSeq v2 (Illumina) for single-end 1 x 75 bp with the custom primer CSP. The sequencing was performed in the Lexogen facility.

Analysis of 3’-end RNAseq and bulk-RNAseq data

Sequencing data using QuantSeq-REV were obtained as fastq files and processed as follows. First, adaptors were trimmed using BBDuk with k=13, forcetrimleft=0 mink=5 trimq=5 ktrim=r qtrim=r minlength=20 and useShortKmers=t. Trimmed reads were aligned to mm10 mouse genome using STAR aligner⁸³ with outFilterType=BySJout, alignSJoverhangMin=8, outFilterMismatchNmax=999, alignIntronMin=20, alignMatesGapMax=1000000, outFilterMultimapNmax=20, alignSJDBoverhangMin=1, outFilterMismatchNoverLmax=0.6, alignIntronMax=1000000, outSAMattributes=NH HI NM MD. The resulting BAM files were combined and used to generate unique clusters with HOMER⁹³ findpeaks algorithm (cluster size 24bp). For each bam file, aligned reads were separated and converted to bed files for forward and reverse strand alignments to the genome. Read counts for each cluster were generated separately using bedmap script in BEDOPS.⁹¹ Coordinates for reference mouse SINEs were obtained from UCSC genome browser from mm10 RepeatMasker track (repClass=SINE). Finally, read counts for forward and reverse strands were combined and used for differential expression for each class of repetitive elements.

To identify outlier samples, quality control was performed on base qualities and nucleotide composition of sequences, mismatch rate, mapping rate to the whole genome, repeats, chromosomes, key transcriptomic regions (exons, introns, UTRs, genes), insert sizes, AT/GC dropout, transcript coverage and GC bias to identify problems in library preparation or sequencing.

For bulk-RNAseq, 75 bp paired-end read sequences were generated for DRG neurons and RGC. For more precise detection of SINEs, 150 bp paired-end sequences were generated for DRG injury, A-Fos dominant negative against AP-1 and ASO experiments. Reads were aligned to mm10 using STAR aligner with default setting. Read counts for SINEs are derived using HTSeq with mm10 RepeatMasker track as reference. Repeat elements (RE) with more than 5 read counts for at least 1/3 of total samples were kept for downstream analysis. Differential expression analysis was conducted using the edgeR⁹⁰ package.

GI-SINE identification and genomic distribution

RNA-seq of DRG with 150bp paired-end sequences was used for improved accuracy of short RE. SINE and additional RE coordinates and sequences were extracted using the RepeatMasker (http://www.repeatmasker.org) track of the UCSC Table browser, genome version mm10 (https://genome.ucsc.edu/).⁹² The sequences of the GENECODE version 23 transcripts were also extracted and used later for mapping the reads.

Mapping of the reads to the transcripts and transcript abundance were calculated using RSEM⁸⁹ with the Bowtie2 parameter. Differentially expressed elements were identified using UTAP 2.0.⁸⁴

Normalization of the counts and differential expression analysis was performed using the Bioconductor package DESeq2⁸⁵ v1.36.0 with the parameters: betaPrior=True, cooksCutoff=FALSE, independentFiltering=FALSE. Raw P values were adjusted for multiple testing using the procedure of Benjamini and Hochberg.⁹⁷ Elements were considered as expressed and introduced into the analysis if the average of their normalized count values, taken over all samples, was at least 3. Elements were assessed as GI-SINEs if expressed at Injury/Naive (72 hours after injury) fold change > 2 & FDR-adjusted p < 0.05 (447 elements). Elements were assessed as non-regulated expressed B2-SINEs if they were up or down regulated with fold-change <2 & p>0.05 (1887 elements). For the ATF3 cKO sciatic nerve crush experiment, differential expression analysis was performed using DESeq2 paired-model, to account for the individual mice effect and test the genotype-specific effect of crush, applied with parameters: cooksCutoff=FALSE, independentFiltering=FALSE.

Genomic distribution of the GI-SINE elements overlapping the following genomic features: promoter, downstream, 5’ UTR, 3’ UTR, intron and intergenic regions was obtained using the CEAS web tool (Enrichment on chromosome and annotation, version 1.0.0) at the Cistrome Analysis Pipeline website (http://cistrome.org/ap/).

GI-SINE specific sequence motif identification

In order to determine the precise genomic start and end coordinates of B2-SINE based on RNAseq reads, the bamCoverage tool from deepTools suite, version 3.5.1⁹⁸ was used. Briefly, the bam files produced by Bowtie2 were merged, separated by strand and the bamCoverage tool was applied for each strand separately using default parameters. All areas covered by reads were then intersected with the 447 GI-SINEs RepeatMasker database coordinates using the intersectBed tool from BEDTools suite, version 2.26.0.⁸⁶ Matched bamCoverage loci coordinates were used for further GI-SINE analysis (391 elements). The rest of the mapped B2-SINE elements were classified as non-GI-SINE (6665 elements) for the control set. Small differences in the number of identified GI-SINE transcripts by RSEM and bamCoverage are due to differences in the underlying algorithms used by each tool.

The sequences of GI-SINEs and non-GI-SINE B2 elements (6665 sequences) were then used as input for STREME⁴⁹ (meme-suite.org) version 5.5.5. STREME search was carried with the default parameters with the following exceptions: pattern length (8-30). Due to sample size constraints, STREME sampled 3888 out of the 6665 sequences of the control set (non-GI-SINE).

ATAC-seq

ATAC-seq was performed as described⁹⁹ with minor adjustments for DRG tissue.⁸⁵ Briefly, DRG tissue was extracted in 500 μl Nuclear extraction buffer (10 mM Tris, 10 mM NaCl, 3 mM MgCl₂, 0.1% Igepal, 0.1% Tween, protease inhibitor cocktail) for 5 minutes on ice, then a 21 g needle on a 1 ml syringe was used to shear the tissue through the needle 5 times. NeuN-positive nuclei were separated by fluorescence-activated cell sorting. Libraries were sequenced under 50 bp paired-end mode on NovaSeq6000. Reads were mapped to the mouse genome with Bowtie2⁸⁷ with a ‘-X 2000’ parameter and metagene plots were prepared using NGSplot.⁸⁸ TOBIAS³⁸ was run with default parameters on the BAM files generated by Bowtie2 and on the BED file containing the B2 elements that were significantly upregulated.

ATF3 conditional knockout mice for DRG RNAseq after sciatic nerve injury

To obtain conditional knockout of ATF3 in DRG sensory neurons, female mice with homozygous loxP insertions flanking Atf3 exon3⁴¹ were crossed with male Advillin-Cre⁺ (heterozygous) / Atf3(exon3)-flox^+/wt (heterozygous) mice to yield Advillin-Cre+/ Atf3(exon3)-flox^+/+ (ATF3-cKO) mice. Control mice were age-matched mice from the same colony that were either Advillin-Cre null or Atf3(exon3)-flox null (ATF3-wt). Left sciatic nerve crush was performed on 8 cKO (4 males & 4 females) & 9 wt (6 females & 3 males) mice. Ipsilateral (injured) and contralateral (naive) L4/L5 DRG were harvested 72 hours after injury and processed for RNA extraction.

RiboTag

Pulldown of ribosome-associated RNA from L4/L5 DRG (Adv-Cre x RiboTag mice) was performed as described previously.⁴⁶ Briefly, three days after unilateral sciatic nerve crush injury L4/L5 DRG were homogenized on ice in 600 μl supplemented homogenization buffer [50 mM Tris (pH 7.0), 100 mM KCl, 12 mM MgCl₂, 1% NP-40, and 1 mM DTT, EDTA-free protease inhibitor cocktail (P8340 Sigma), 300 units/ml RNasin (Promega), 150 μg/ml cycloheximide, and 10 mM ribonucleoside vanadyl complexes (RVC), R3380 Sigma] using a 2 ml Potter-Elvehjem tissue homogenizer. After homogenization, lysates were transferred to Eppendorf tubes and centrifuged at 4°C for 10 minutes at 10000 x g. Supernatants were used for immunoprecipitation and 10% of each sample was set aside as input; 5 μg of rat anti-HA antibody (11867423001, Sigma) was added to each sample and incubated on rotator for 4 hours at 4°C. 80 μl Protein G Dynabeads Magnetics beads were washed with the homogenization buffer, added to lysates containing the HA antibody and rotated overnight at 4°C. After this, beads were washed 3 x 5 min at 4°C with high salt buffer [50 mM Tris (pH 7.0), 300 mM KCl, 12 mM MgCl₂, 1% NP-40, and 1 mM DTT, protease inhibitor cocktail (P8340 Sigma), 200 units/ml RNasin, and 150 μg/ml cycloheximide]. After washes, RNA was eluted from beads and purified using the RNeasy Mini kit (Qiagen) including on-column DNase.

Northern blot

Total RNA was isolated from HEK293, N2a cells or DRG with RNeasy Mini kit (Qiagen), and 1-3 μg of RNA was separated in an 8% polyacrylamide-8M urea denaturing gel using RiboRuler Low Range RNA Ladder (Thermo Fisher) as a size marker. After visualizing bands with EtBr (1 μg/mL) on a UV table, RNA was electro-transferred in 0.5 x TBE buffer to Hybond-N+ nylon membranes at 4°C for 3 hours at 200 mA. After UV-crosslinking (0.12 J/cm²), membranes were hybridized overnight with gentle agitation at 42°C in hybridization buffer (200 mM phosphate buffer pH 7.0, 7% SDS, 1 mM EDTA) with 50 nM biotinylated probes. For B2, 3’-end biotinylated antisense probe to nucleotides 127-150 in mouse B2 consensus (TACACTGTAGCTGTCTTCAGACA, from(5)⁵) was used. After hybridization, membranes were washed 2 x 5 minutes in 1x SSC-0.1% SDS and incubated at room temperature with streptavidin-HRP (ab7403, Abcam, 1:20000) in 5% SDS-PBS for 15 minutes. Membranes were washed in 5% SDS-PBS for 10 minutes and in 0.1% SDS-PBS for 2 x 10 minutes and visualized using ECL substrate (Radiance ECL, Azure). For re-probing with biotinylated anti-7SL, membranes were stripped with 0.5% SDS-PBS brought to boil (2 x 10 minutes). Chemiluminescence signal intensities were quantified using ImageQuant TL software.

Quantitative reverse-transcriptase PCR (RT-qPCR)

For the measurement of B2-SINE expression in DRG after sciatic nerve injury, cDNA was prepared from 10-1000 ng of purified RNA, using SuperScript™ III First-Strand Synthesis System (Thermo Scientific, 18080051). Quantitative real-time PCR was performed using the SYBR green FastMix on the ViiA-7 system (Thermo Fisher Scientific), according to manufacturer’s instructions. Primers were designed against the mouse B2 consensus sequence (retrieved from https://www.girinst.org/repbase/) using Primer3 Plus and used at 60°C annealing temperature. Target amplicon identity was verified by cloning and Sanger sequencing the qPCR product. 18S target was used for normalization. GI-SINE expression in ASO-transfected DRG neuron cultures was measured as follows: RNA was extracted using RNeasy Plus kit. 300ng of RNA were used for cDNA preparation with Sensifast cDNA library kit. Droplet digital PCR (ddPCR) was carried using BioRad ddPCR assay kit with primers and probes noted in Table S5. GI-SINE expression was normalized to Hmgb1 RNA.

FISH analysis of AAV-transduced B2-SINE expression

For detection of B2-SINE expression in RGC, mouse retinas were fixed for 4 hours in 2% PFA in 1x PBS followed by cryoprotection in 30% sucrose for 1 day. Tissues were cryosectioned at 12 μm thickness and adhered to poly-D-lysine coated glass slides. Sections were stored at -20°C until used. Digoxigenin labeled antisense oligonucleotides were used for FISH, based on the methods of Bassell and colleagues,¹⁰⁰ with modifications due to the limit of two 50 nucleotide segments given the short B2 consensus sequence (Repbase). Antisense oligonucleotide probes were designed using MAC Oligo© software. Probes were verified for specificity by BLASTN against GenBank for murine refseq RNA entries. These were synthesized with the 5′-amino modifier C6 at four thymidines per oligonucleotide (IDT), and then chemically labeled with digoxigenin succinimide ester ([Digoxigenin-3-O-methylcarbonyl-ε-aminocaproic acid-N-hydroxysuccinimide] ester from Roche).

FISH with the digoxigenin-labeled probes was performed on cryosections of mouse retina as described¹⁰⁰ using the two probes. Briefly, tissue sections were warmed to room temperature and then incubated with 20 mM glycine (3 x 10 minutes) followed by 0.25 M NaBH₄ (3 x 10 minutes) to quench the autofluorescence. Sections were then incubated with 0.1 M Triethylamine (TEA; Sigma) plus 0.25 % acetic anhydride (Sigma) for 10 minutes, followed by wash with 2x saline-sodium citrate buffer (SSC; Sigma). Samples were dehydrated in 70, 95 and 100 % ethanol followed by chloroform treatment to delipidate the retinas. For each retina section, probe plus carrier mixture was prepared by mixing 50 ng of each probe with 9 μg denatured salmon sperm DNA (Sigma) and 10 μg E. coli tRNA (Roche; B1107540). The probe mixture was dehydrated to near dryness using a speed vac and then re-suspended with 15 μl 80 % formamide in 1x SSC for 5 minutes at 95°C, before mixing with 15 μl of hybridization buffer (10 % Ribonucleoside Vanadyl Complex (RVC; Sigma), 2x SSC, 20% bovine serum albumin, and 20 mM sodium borate). After rehydration, slides were incubated with the probe mixture for 5 hours at 42°C. After washing with 40 % formamide in 2X SSC at 37°C for 15 minutes, sections were incubated with Cy3-conjugated mouse monoclonal anti-digoxigenin (1:100; Jackson ImmunoResearch, 200-162-156) and primary chicken anti-βIII tubulin (1:100; Millipore; AB9354) antibodies overnight. The next day, sections were incubated for 2 hours at room temperature with rabbit anti-GFP antibody (1:100; Abcam; AB290). Following this, sections were incubated with secondary antibody Cy3-conjugated anti-mouse, Cy5-conjugated anti-chicken, and FITC-conjugated anti-rabbit antibodies (1:100 each; Jackson ImmunoResearch.; 115-165-003, 703-175-155, and 111-095-003, respectively) diluted in 0.3% Triton X-100 in PBS plus 10x blocking buffer (Roche) for 1 hour at room temperature. Samples were mounted using ProLong™ Gold Antifade Mountant (Thermofisher). Probes were imaged with the same acquisition parameters using an oil immersion 63x/1.4 NA objective with a Leica SP8X confocal microscope. Acquisitions were identically post-processed with Leica Lightning deconvolution and analyzed as outlined below. For FISH signal quantification, z planes of the xyz tile scans from 3-5 random fields along each retina section were analyzed using ImageJ. β-III-tubulin labeling was used to focus on neurons and AAV-transduced cells were identified by anti-GFP IF. RNA signals were quantified for each z-stack of each individual cell. RNA signal intensity per cell was averaged for all image locations in each biological replicate. Prism software (GraphPad) was used for statistical analyses.

RNA FISH analysis of endogenous GI-SINE expression

For detection of GI-SINE RNA in cultured DRG neurons, RNAscope was performed following the “RNAscope® Assay for Adherent Cells Cultured on Coverslips” and the BaseScopeTM v2 Assay” protocols from Advanced Cell Diagnostics, Inc. GI-SINE probe sequence, designed based on the GI-SINE motif consensus sequence: 5’-CTGCTCTTCCAGAGGTCCTGAGTTCAATTCCCAGCAACCAC-3’.

Negative control probe (1ZZ, Cat. No 701021) is targeted against bacterial DapB (positions 821-862 of GenBank: EF191515.1). DRG neurons were cultured and transfected with ASOs promptly after plating as described. Neurons were fixed 60 hours after plating in 4% PFA for 15 min and dehydrated in 50%, 70 % and 100 % ethanol. After incubation for 10 min with hydrogen peroxide, 1:15 dilution of Protease III has been applied to the samples for 10 min. Samples have been incubated with the hybridization probes and then processed for signal amplification and detection steps. After completing the RNAscope protocol, samples have been incubated with a mix of mouse anti-Neurofilament marker (SMI 312; 1:200) and mouse anti-β-III-tubulin (TUBB3) (Biolegend; 1:200) primary antibodies for labeling neurons FITC-conjugated donkey anti-mouse (1:200, Jackson ImmunoResearch.) was used for secondary antibody and detection. Imaging and analysis of GI-SINE signal in DRG somas was carried out as described for B2-SINE detection above. The same microscopy setup was used for imaging of axonal GI-SINE and analysis was done with ImageJ software. Axon segments averaging 60 μm, at distances at least 100 μm from the soma, were delineated manually using 10 pixel thick segmented line tool to measure mean RNA-FISH intensity per axon segment.

Protein pulldown with biotinylated RNA probes

B2-SINE neuronal interactome analysis was conducted from mouse sciatic nerve axoplasm, extracted as previously described.⁴² Streptavidin Myone-Dynabeads (Thermo Fisher, 65002) were used as resin to bind biotinylated B2-SINE RNA probes. 300 μl of beads were pre-washed as follows: twice in Buffer A (100 mM NaOH, 50 mM NaCl), once in Buffer B (50 mM NaCl), once in 2x binding buffer (10 mM Tris-HCl (pH 7.5), 2 M NaCl, 1 mM EDTA) and once in 1x binding buffer (5 mM Tris-HCl pH 7.5, 1 M NaCl, 0.5 mM EDTA). Axoplasm from 60 mouse sciatic nerves was prepared in nuclear transport buffer¹⁰¹ (20 mM HEPES pH 7.3, 110 mM potassium acetate, 5 mM magnesium acetate) supplemented with Complete Protease Inhibitors (Roche), Phosphatase Inhibitors #3 (Roche) and RNasin (Promega). Axoplasm was incubated for 1 hour at 4°C with half of the bead suspension (150 μl of stock bead slurry) for pre-clearing of biotinylated proteins. For binding of biotin-RNA to Streptavidin beads, 400 pmoles of biotinylated B2_5’ (nt 1-75 of the B2-SINE consensus sequence), B2_3’ (nt 75-166) and biotin-only control were added separately to 400 μl of bead suspension in binding buffer and incubated for one hour at 4°C before washing once with binding buffer and twice with 1 ml low-salt buffer (10 mM HEPES pH 7.4, 3 mM μCl₂, 14 mM NaCl, 1 mM DTT, and 5% glycerol). In the experiment testing the interference of B2-antisense oligonucleotides (ASOs) to the interaction with nucleolin, biotinylated-B2 RNA was pre-incubated with 800 pmoles of B2-ASO in 100μl of RNA-DNA annealing buffer (100mM NaCl, 5mM MgCl₂, 20mM Tris-HCL-pH 7.5) for 1 min at 95°C and then for 30 min at room temperature. Controls were incubated without addition of ASO and the RNA-DNA ASO mixture was then added into the bead suspension in 400 μl of binding buffer. Cleared axoplasm was equally divided to the bound-beads tubes (B2_5, B2_3’ and control) and incubated for 1 hour at 4°C. Beads were then washed five times with high-salt wash buffer (10 mM HEPES pH 7.4, 3 mM MgCl₂, 250 mM NaCl, 1 mM DTT, and 5% glycerol). For the comparison between B2_5’ and U1 (stem loops 3+4) axoplasm interactomes and nucleolin co-precipitation, samples were processed identically except for the final bead washout steps, which were carried in high-salt buffer supplemented with 0.5% NP-40 to reduce non-specific protein binding. For preparation of samples for mass-spectrometry analysis, beads were washed in ultra-pure water and frozen at -80°C until shipping for analysis. For Western blot (WB) analysis, proteins were eluted using Laemmli buffer supplemented with 100 mM DTT by 10 minutes incubation at 65°C.

Pull-down of nucleolin with B2-RNA was carried on neuron-enriched DRG tissue from heterozygous GAR domain-deleted mice⁵ (Ncl-GAR^+/Δ), extracted as described for nucleolin RIP experiments. Total DRG from 2 mice per each pulldown condition were lysed in 200 μl lysis buffer with 5% kept as input sample. Lysates were centrifuged and diluted for pre-clearing and split for incubation with B2-RNA-bound beads in 250 μl of lysis buffer. Pulldown wash and elution was done as described above for sciatic nerve axoplasm samples for Western blot analysis. Full length and GAR-deleted nucleolin were detected by anti-nucleolin antibody (Cell Signaling Technology, 14574). For pull-down of overexpressed nucleolin with B2-SINE RNA, HEK293T cell lysate was used. 300,000 cells/well were plated in 6-well plates and transfected after 24 hours with either full-length tagged nucleolin (HA-Dendra2-ncl) or ΔGAR-nucleolin (HA-Dendra2-ncl_ΔGAR) constructs (3 μg/well), using JetPEI (Polyplus Transfection). The nucleolin constructs were as previously described.⁵ After 24 hours, proteins were extracted using RIPA buffer (50 mM Tris–HCl pH 8.0, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with Complete EDTA-free protease inhibitor cocktail (1187358000, Roche). Lysates were centrifuged at 12,000 x g to pellet debris, and 10% of each lysate was set aside to probe for nucleolin expression in the input. The remainder was used for pull-down with B2-SINE RNA as described for axoplasm samples. Nucleolin was detected by Western blot with HA antibody.

Tryptic digestion and tandem mass tag (TMT) labelling for protein mass spectrometry

Beads were washed three times with water and stored at -20^oC until tryptic digestion. For digestion, beads were thawed, and liquid-free beads resuspended in 15 μl of 10 mM Tris(2-carboxyethyl)phosphine (TCEP) in 20 mM triethylammonium bicarbonate (TEAB) pH 8.5, and incubated for 30 minutes at 56^oC. After this, iodoacetamide was added to a final concentration of 15 mM and samples incubated at room temperature for 60 additional minutes. 0.5 μg of sequencing grade trypsin (Promega) was added to each sample and incubated at 37^oC overnight. Supernatants were recovered, and beads were digested again using 0.5 μg trypsin in 20 mM TEAB for 4 hours at 37^oC. Peptides from both consecutive digestions were mixed and acidified with 1% (V/V) formic acid and recovered by solid phase extraction using C18 ZipTips (Millipore), then eluted in 2 x 7 μl aliquots of 50% acetonitrile (MeCN)/0.1% formic acid, dried and resuspended in 8 μl 100 mM TEAB in preparation for labeling with TMT reagents. For TMT labelling, samples were labeled according TMTPro 16plex kit instructions (ThermoFisher Scientific), with minor modifications. Briefly, TMT reagents were resuspended in acetonitrile at 25 μg/μl, and 4 μl of this solution were added to the individual samples to be labelled. After incubating for 1 hour at 22^oC, reactions were quenched by adding 1 μl 5% hydroxylamine and incubated for an additional 15 minutes. After that, the labelling reactions were combined and diluted with 1692 μl 0.1% formic acid and desalted using a C18 100 μl OMIX tip (Agilent Technologies) as indicated by the manufacturer. Peptides were eluted in 2 x 45 μl aliquots of 50% MeCN/0.1% formic acid, and dried and resuspended in 2.5 μl 0.1% formic acid for mass spectrometry analysis.

Mass spectrometry analysis, peptide and protein identification and TMT quantification

Labelled peptide digests were subjected to chromatographic separation using a 2 μm, 75μm ID x 50 cm PepMap RSLC C18 EasySpray column (Thermo Scientific). 3-hour MeCN gradients (2–25% in 0.1% formic acid) were used to separate peptides, at a flow rate of 200 nl/min, for analysis in a Orbitrap Exploris 480 (Thermo Scientific) in positive ion mode. MS spectra were acquired between 375 and 1500 m/z with a resolution of 120000. For each MS spectrum, multiply charged ions over the selected threshold (2E4) were selected for MSMS in cycles of 3 seconds with an isolation window of 0.7 m/z. Precursor ions were fragmented by HCD using stepped relative collision energies of 30, 35 and 45 in order to ensure efficient generation of sequence ions as well as TMT reporter ions. MSMS spectra were acquired in centroid mode with resolution 60000 from m/z=120. A dynamic exclusion window was applied which prevented the same m/z from being selected for 30 seconds after its acquisition.

Peak lists were generated using PAVA in-house software.¹⁰² All generated peak lists were searched against the mouse subset of the SwissProt database (SwissProt.2019.07.31), using Protein Prospector^94,103 with the following parameters: Enzyme Specificity” was set as Trypsin, and up to two missed cleavages per peptide were allowed. Carbamidomethylation of cysteine residues, and TMT 16plex labeling of lysine residues and N-terminus of the protein were allowed as fixed modifications. N-acetylation of the N-terminus of the protein, loss of protein N-terminal methionine, pyroglutamate formation from peptide N-terminal glutamines, oxidation of methionine were allowed as variable modifications. Mass tolerance was 4 ppm in MS (with systematic error correction of 0.5 ppm) and 30 ppm in MS/MS. The false positive rate was estimated by searching the data using a concatenated database which contains the original SwissProt database, as well as a version of each original entry where the sequence has been randomized. A 1% FDR was permitted at the protein and peptide level. For quantification only unique peptides were considered; peptides common to several proteins were not used. Relative quantification of peptide abundance was performed via calculation of the intensity of reporter ions corresponding to the different TMT labels, present in MS/MS spectra. Intensities were determined by Protein Prospector. For each spectrum, relative abundances of proteins in the 6 RNA pulldowns were calculated as ratios vs the matching biotin pulldown. For peptide and then total protein relative levels, peptide spectral matches and then peptide ratios were aggregated to the peptide and protein levels using median values of the log₂ ratios. Statistical significance was calculated at the protein levels by comparing the values of B2_5’ (nt1-75)/biotin and B2_3’ (nt75-166)/biotin (three replicates) or B2 5 vs. U1(SL3+4) (four replicates) in the TMT experiment with a 2-tailed t-test.

Re-analysis of a previously published dataset⁴⁴ was done as follows: peptide counts from RNA bait interactors were taken from Table S2 in Ponicsan et al.⁴⁵. A count of 0.5 was added to all values to allow calculating ratios where measured counts = 0. Mean RNA bait/biotin-only ratios and p-values (two tailed t-test) were calculated from the three independent repeats.

Gene ontology (GO) analysis

GO analysis was carried on the 1-75/biotin and 75-166/biotin sets of protein hits. Individual proteins were selected for analysis only if enriched over biotin-only control with FDR < 0.01 for which three or more unique peptides were detected. After selection, hits were ranked by descending mean fold-change over biotin-only control from three independent experiments. The two ranked sets of hits were separately analyzed using GOrilla⁹⁵ Mus musculus was selected as target organism, and gene name lists were analyzed for component and processes ontologies in single ranked mode with GO term p-value < 0.001 threshold.

Nucleolin RNA co-immunoprecipitation (RIP)

Neurons isolated from total DRGs from 8-12 week old mice were dissociated and purified by Percoll gradient as described above. Cell pellets were washed with ice-cold PBS and lysed by gentle pipetting on ice for 30 minutes in lysis buffer (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 0.5% NP-40) supplemented with protease inhibitors and RNasin Plus (Promega #N2611, 400U/ml). Lysates were cleared by centrifugation (20,000x g, 10 mins, 4°C), and 5% of the supernatant was retained as input. Cleared lysates split into tubes containing 50 μl Dynabeads Protein G (Invitrogen #10003D), pre-coupled in lysis buffer for three hours at 4°C with 7 μg rabbit anti-nucleolin (Abcam #ab50279) or isotype control IgG (Cell Signaling Technology, #3900), and incubated overnight with rotation at 4°C. Beads were washed three times with lysis buffer at 4°C, after which RNAs were extracted by phenol-chloroform extraction (TRI Reagent, Sigma-Aldrich #T9424). RNA clean-up was performed by mixing the RNA aqueous phase with a 1x volume of 100% ethanol and loading onto Zymo-Spin IC columns, with on-column DNase treatment (Zymo # R1013). Input RNA was generated using the same kit according to manufacturer’s instructions. cDNA was prepared with SuperScript III (Thermo Fisher #18080051) and qPCR was performed with PerfeCTa SYBR green FastMix (Quanta Biosciences #95073).

Proximity ligation assay (PLA) in DRG cultures

Proximity ligation assay (PLA)¹⁰⁴ was used to detect spatial coincidence of nucleolin and RPL11 proteins. PLA was performed according to manufacturer's protocol using NaveniFlex (Navinci; NaveniFlex MR Species: Mouse & Rabbit NF.MR.100.1 and NaveniFlex Detection Reagents NF.100.2). After the PLA protocol, cells were stained with chicken anti-NFH antibody in non-transduced samples or goat anti-GFP in AAV-transduced samples. DAPI (1:1000) and Phalloidin (1:200) staining was performed. YFP positive (reporting AAV-B2/control transduction) neurons were imaged using LSM700 or FV10i (Zeiss) confocal microscope with 60x oil immersion or water immersion objectives. Phalloidin staining was used to detect and omit non-neuronal cells from the analysis. PLA spots were counted using Imaris 9.9.1 (Bitplane) software as follows: YFP positive neurons were selected and the PLA channel was used for spot object detection using default quality parameters. The nucleus was delineated in 3D using surface object detection on the DAPI channel with default quality parameters. The number of PLA spots co-localized with nuclei surface object (distance=0) or outside the nucleus were then counted. Max-projection images of YFP and DAPI were used for measuring cytosol and nucleus areas by manually tracing ROIs using Fiji (ImageJ) software.

Protein synthesis assays with puromycin

L4/L5 DRGs were extracted from mice two weeks after intrathecal injection of PHP.S-B2, PHP.S-shControl or untransduced mice. Puromycin incorporation, staining, imaging and analysis of soma and axon-tip signal was carried out as previously described.⁴ Briefly, after 48 hours in cell culture, cells were pulsed with either 9.8 μM puromycin for 10 minutes or with 40 μM anisomycin for 30 minutes followed by puromycin for 10 minutes. Cells were then washed once with PBS and fixed in PFA 4% in PBS. Samples were processed for immunofluorescent (IF) labeling for puromycin, YFP (GFP antibody) & counter-stained for F-Actin with phalloidin-rhodamine. Samples were imaged in a semi-automated tile-scanning Nikon Ti2 epifluorescent microscope with 40x objective to capture the entire axon outgrowth. Phalloidin staining was used to detect and omit non-neuronal cells from the analysis. Mean puromycin signal was measured in manually traced regions of interest (ROI) of AAV-transduced (YFP channel) or NFH-positive somata and axons using Fiji (ImageJ) software.

Measurement of axon outgrowth in fixed DRG cultures

DRG neurons plated on 13mm coverslips were fixed after 48 hours in culture and labelled using anti-NFH immunofluorescence. Imaging was carried using automated ImageXpress micro at 10x magnification for whole-sample acquisition. Outgrowth was analyzed using MetaExpress software Neurite Outgrowth tool. Parameters for detection of NFH-positive soma and neurites were set manually for selected images from each experiment. Longest neurite, total outgrowth length and cell area per each soma were used for post-analysis. Neurons were classified as growing if the longest neurite was >2x longer than soma diameter. Cell diameter was calculated based on soma area, assuming cells are perfect circles. Mean total outgrowth (per growing neuron) and percentage of growing neurons from total NFH-positive cells (per experiment) were then calculated.

DRG neuron spot culture axotomy and live re-growth assay

DRG were dissociated in collagenase for 30 min in tissue-culture incubator followed by trituration 10–15 times with a 1000 μl pipette tip and additional 10 min incubation. Suspensions were further triturated in F-12+10% FBS media containing 1x penicillin/streptomycin solution (F-12 complete). Cells were centrifuged for 10 min at 700x g and pellet was resuspended in F-12 complete media and plated at a density of approximately 2.8 DRGs in 7 μl spots on poly-D-lysine/laminin-coated chambered coverslips. Spotted DRGs were incubated for 7 min and 0.5ml of media containing 2x10⁶ gc/ml PHP.S AAV-B2 or AAV-Control virus. Half of the media volume was freshly changed at days 1 and 5 after plating. Axotomy was performed at 7-8 days after plating as follows: axons were cut on one side of the spot under a stereomicroscope using a surgical blade (Fine Science Tools, 10035-10) at distance approximately equal to the radius of the spot and axons distal to the cut site were gently scraped away. Regeneration of axons in axotomized spot cultures was visualized by tile scans on a Leica Stellaris confocal microscope fitted with an environmental chamber, taken at 10 min intervals for 7-11 hours post axotomy. Axon growth in X and Y dimensions was traced across the image sequences using the Fiji/ImageJ⁸² Trackmate plugin (National Inst. Health, Bethesda, MD). Prism software was used for statistics with outlier testing by ROUT.

Quantification and statistical analysis

Descriptive and analytical statistics used in this study are detailed in the relevant results and legends sections. Significance difference was determined if statistical test yielded p < 0.05 for all experiments, except for RNA-seq differential expression analysis in which false discovery rate (FDR)-adjusted p < 0.05 was used as significance threshold. No statistical method was used to determine sample sizes.

Published: May 16, 2025

Supplemental information

Table S1. Expression levels of individual B2-SINE RNAs after nerve injury, related to Figure 3

DRG neuron differential expression (fold changes) of B2-SINEs mapped to unique loci at 24 or 72 hours after sciatic nerve injury.

Table S2. Transcription factor binding-site analyses for GI-SINE loci in injured DRG neurons, related to Figure 3

GI-SINE loci were analyzed for enrichment of transcription factor binding sites using the footprinting algorithm TOBIAS.

Table S3. B2-SINE interactome in sciatic nerve axoplasm, related to Figure 4

Protein identifications by mass spectrometry of B2-SINE bait pull-downs from sciatic nerve axoplasm.

Table S4. B2-SINE versus U1 RNA interactome in sciatic nerve axoplasm, related to Figure 4

Protein identifications by quantitative mass spectrometry of B2-SINE versus U1 RNA bait pull-downs from sciatic nerve axoplasm.

Table S5. B2-SINE consensus, probes, and oligonucleotide sequences, related to STAR Methods

Sequences of B2-SINE reagents used in this study, including (1) the consensus sequence used for overexpression, (2) biotinylated bait sequences used for pull-downs, (3) primers and probes used for qPCR, Northern blot, and FISH, and (4) antisense oligonucleotides used for functional perturbations.