The Role of Long Noncoding RNA MALAT1 in Diabetic Polyneuropathy and the Impact of Its Silencing in the Dorsal Root Ganglion by a DNA/RNA Heteroduplex Oligonucleotide

Diabetic polyneuropathy (DPN) causes sensory disturbances in the distal parts of the extremities (1–3). DPN is a complex multifactorial disease, and the degeneration of axons linked to dorsal root ganglion (DRG) sensory neurons may be its pathogenetic mainstay (4–7). Experimental diabetic mice demonstrate pathological features of human DPN, which is characterized by the dying-back degeneration of DRG sensory axons and accompanying morphological alterations of neuronal cell bodies and nuclear architecture, including axon terminals in the skin, among others (5,7–10). The DRGs in experimental diabetic mice also demonstrate global gene expression profile shifts of mRNA and miRNA, including increases in HSP27, receptor for AGE (RAGE), phosphatase and tensin homolog (PTEN), and CWC22, and declines, among others, in miRNA let-7i (4,7,10–13). Counteracting these gene expression changes in sensory neurons through gene therapy may be effective in treating human DPN.

The knockdown of target genes related to neural functions in DRG sensory neurons by RNA interference has been reported in the treatment of pain control and in nerve regeneration (14–17). Oligonucleotides such as siRNA and RNase H-dependent antisense oligonucleotide (ASO) are known to bind and to cleave the complementary RNA sequences to downregulate expression (18). We developed DNA/RNA heteroduplex oligonucleotide (HDO), a novel nonviral gene delivery tool, which differs in structure from siRNAs of double-stranded RNA or ASOs of single-stranded DNA (19). The HDO, which is conjugated by α-tocopherol (Toc-HDO) to the complementary RNA sequence at the 5′ end, can much more effectively inhibit the target gene expression than parent ASO in the central nervous system (CNS) as well as liver (19,20). Here, Toc-HDO was used as considered for a specific form of DPN treatment.

Long noncoding RNAs (lncRNAs) are noncoding transcripts with ≥200 nucleotides (21,22); they have recently been recognized as diagnostic and therapeutic targets for human diseases such as diabetes (23,24). Metastasis-associated lung adenocarcinoma transcript 1 (MALAT1), originally identified as a prognostic marker for non-small cell cancer, is expressed in almost all cell types and localized in nuclear speckles, which are subnuclear membraneless structures enriched in pre-mRNA splicing factors (24–28); it also regulates alternative splicing by modulating the phosphorylation and distribution of serine/arginine splicing factors (SRSFs) (29). MALAT1 is upregulated in target tissues, such as the retina, kidney, and heart, during diabetes-related complications (24,30,31). Its upregulation has been implicated in pathogenic responses, including for example, endothelial cells with decreases in antioxidant defense genes through enhancement of Kelch-like ECH-associated protein 1 (Keap1)–nuclear factor-erythroid 2-related factor 2 (Nrf2) signaling in diabetic retinopathy (31,32). It may also induce a protective response against neurodegeneration processes, including retinal ganglion cell survival and glial cell activation, through continuous activation of CREBP signaling after optic nerve transection (33). It remains unknown whether the role of MALAT1 is pathogenic or protective in DPN.

We identified a significant upregulation of MALAT1 in DRGs of diabetic mice, which was effectively inhibited by the systemic injection of Toc-HDO. MALAT1 silencing was unexpectedly associated with an exacerbation of the neuropathic deficits of DPN and by changes in nuclear architecture, specifically speckles, which contain splicing factors for DRG sensory neurons. Our study indicates that Toc-HDO is highly effective for gene silencing in DRGs and that MALAT1 likely provides intrinsic neuroprotection, in part by regulating the distribution of splicing factors during the development of DPN.

All animal experiments were performed adhering to the guidelines for animal experiments for Tokyo Medical and Dental University and approved by the Tokyo Medical and Dental University Animal Experiment Committee. Slc:ICR male mice (4 weeks of age, 22–25 g) were used. Diabetes was induced by the intraperitoneal injection of 85, 70, and 55 mg/kg streptozotocin (STZ) (Sigma-Aldrich, St. Louis, MO) for 3 consecutive days and defined as a fasting glucose level of ≥250 mg/dL 2 weeks after STZ injection. The maximum value of the glucometer (OneTouch UltraVue; Johnson and Johnson KK, Tokyo, Japan) was ≥600 mg/dL.

A PBS control solution, ASO, and Toc-HDO–targeting MALAT1 (25 mg/kg in PBS) (Gene Design, Osaka, Japan) were intravenously injected through the tail vein 4 months after STZ injection three times once a week (Fig. 2C). For immunohistochemistry labeling of anti-phosphorothioate (PS) antibody to map Toc-HDO targeting, Toc-HDO–targeting MALAT1 (50 mg/kg in PBS) was intravenously injected through the tail vein. The sequences of ASO and the Toc-complementary RNA (cRNA) targeting the MALAT1 RNA were as follows: 16-mer gapmer ASO: 5′-C*T*A*g*t*t*c*a*c*t*g*a*a*T*G*C-3′, where lowercase letters represent DNA, uppercase letters represent locked nucleic acids (LNA; capital C denotes LNA methylcytosine), and asterisks represent PS linkages; 16-mer Toc-cRNA, 5′-g*c*a*UUCAGUGAAC*u*a*g-3′, where uppercase letters represent RNA, lowercase letters represent 2-O-methyl sugar modification, and asterisks represent PS linkages. The shuffle sequences of the ASO and Toc-cRNA targeting MALAT1 RNA were as follows: 16-mer gapmer MALAT1 ASO: 5′-T*A*C*a*t*a*t*g*c*g*c*t*a*C*T*G-3′, where lowercase letters represent DNA, uppercase letters represent LNA (capital C denotes LNA methylcytosine), and asterisks represent PS linkages; 16-mer MALAT1 Toc-cRNA: 5′-c*a*g*UAGCGCAUAU*g*u*a-3′, where uppercase letters represent RNA, lowercase letters represent 2-O-methyl sugar modification, and asterisks represent PS linkages. α-Tocopherol was covalently bound to the 5′ ends of Toc-cRNA.

One day after the first administration of Toc-HDO, the harvested DRGs (fourth and fifth lumber [L4 and L5] vertebrae), gastrocnemius muscles, and footpad skins of 1-month-old nondiabetic mice were placed in 4% paraformaldehyde and 30% sucrose solution. After being embedded in optical cutting temperature compound (Miles Laboratories, Elkhart, IN), 16-μm-thick (DRG neurons and skins) and 10-μm-thick (muscles) sections were placed on poly l-lysine–coated glass slides and fixed in 10% formalin neutral buffer solution. The sections were pretreated using antigen retrieval with proteinase K (Dako, Santa Clara, CA) and incubated in endogenous peroxidase and alkaline phosphatase blocking solution (Vector Laboratories, Richmond, CA). The sections were blocked using Background Buster (Innovex Biosciences, Richmond, CA). A polyclonal rabbit anti-ASO Ab53 that recognizes the PS-modified backbone containing ASO was applied at a dilution ratio of 1:4000 (diluted in 10% normal goat serum). Subsequently, the sections were incubated with secondary goat anti-rabbit IgG antibody biotinylated (Vector Laboratories) at a 1:1000 dilution and incubated using the ABC kit (Vector Laboratories). After washing, the sections were mounted within 1 h, depending on the 3,3′-diaminobenzidine–nickel–H₂O₂ reaction. Tissue images were acquired using an Olympus BX53 microscope (Olympus, Tokyo, Japan).

Five months after diabetic induction, the DRGs and the sciatic nerves were harvested, snap-frozen in liquid nitrogen, and stored at −80°C until RNA extraction. To detect the mRNA, the RNA (2 μg) was reverse transcribed using the Transcriptor Universal cDNA Master (Roche Diagnostics, Tokyo, Japan). To detect short oligonucleotides, including the parent ASO strand, quantitative real-time PCR analysis was performed using the TaqMan MicroRNA Reverse Transcription Kit (Invitrogen, Carlsbad, CA) and a Light Cycler 480 Real-Time PCR Instrument (Roche Diagnostics). The primers and probes for the mouse MALAT1 (NR_002847), GAPDH (4352932E), and ASO strands were designed by Invitrogen. The relative MALAT1 RNA concentrations were compared with the GAPDH mRNA concentrations, which were used as internal controls.

Four and five months after diabetic induction, the harvested DRGs and sciatic nerves were fixed with 2.5% glutaraldehyde, stained with osmic acid, embedded in EPON resin, and used for 1-μm sections. An Olympus BX53 microscope (Olympus) was used to observe semithin sections with toluidine blue staining and measure neuron and myelin caliber. To evaluate DRG neuronal atrophy in DPN, neurons of the L5 ganglia with distinguishable nuclei were counted, and their areas were measured and averaged. All images were analyzed using ImageJ software (National Institutes of Health, Bethesda, MD) in a blinded manner. DRG neuronal atrophy in the 5-month diabetic mice were evaluated by frozen sections stained with neurofilament 200 (NF200) (Sigma-Aldrich). The total number of neurons analyzed in the DRG per a mouse was at least 100 neurons. For sural nerves, at least two pictures were randomly taken from each sample to quantify the myelin thickness and axon diameter. The total number of myelinated fibers in the sciatic nerves per mouse was at least 300.

Five months after diabetes induction, the harvested DRGs were fixed with 4% paraformaldehyde and embedded in optimal cutting temperature compound, and 10-μm-thick sections were placed on poly l-lysine–coated glass slides. The primary antibodies used were rabbit anti-NF200 (1:100; N4142, Sigma-Aldrich), mouse anti-SC35 (1:100; AB11826, Abcam, Cambridge, U.K.), and mouse anti-SF2/ASF (1:500; sc-33652, Santa Cruz Biotechnology, Dallas, TX). The secondary antibodies were Alexa Fluor 488 goat anti-rabbit (1:200; A-27034, Invitrogen) and Alexa Fluor 647 goat anti-mouse antibody (1:200; A-21244, Invitrogen). The sections were mounted with VECTASHIELD (Vector Laboratories) and observed under a Nikon A1R MP+ multiphoton confocal microscope (Nikon, Tokyo, Japan). The quantitative analysis of number of nuclear speckles in DRG neurons was performed in randomly selected 10-μm sections of double-stained DRG counted in blinded fashion by direct observation of nuclear focal planes at 40× for at least 100 nuclei/mouse of DRG sensory neurons. Each group contained five mice.

Footpad skin samples were harvested using a skin punch and processed 5 months after diabetes induction (34). The footpad skin samples were fixed with 4% paraformaldehyde, cryoprotected in 20% glycerol 0.1 mol/L Sorenson phosphate buffer, and then placed on the slide in 25-μm thickness. Primary rabbit anti-protein gene product (PGP)9.5 (1:200; ADI-905-520-1, Enzo Life Sciences, Farmingdale, NY) and goat anti-collagen type IV (1:400; AB769, Millipore, Billerica, MA) were applied, followed by Alexa Fluor 647 donkey anti-rabbit (1:200; A-31573, Invitrogen) and Alexa Fluor 488 donkey anti-goat (1:200; A-32814, Invitrogen) secondary antibodies. Images were captured using a Nikon A1R MP+ multiphoton confocal microscope (magnification ×100 and step size of 1 μm). Epidermal fibers labeled with PGP9.5 crossing the dermal-epidermal junction were counted in a blinded manner, and the intraepidermal nerve fiber density (IENFD) was calculated by dividing the number of counted fibers by the area of epidermis. All analyses were conducted by an examiner blinded to the identities of the samples being studied.

Motor and sensory nerve conduction recordings were performed in sciatic-tibial nerves in mice anesthetized with isoflurane at a near-nerve subcutaneous temperature of 37°C. An electroneuromyography device (Nuromatic 2000, Dantec, Tonsbakke, Denmark) was used for the electrophysiological recordings. The methods for measuring sensory nerve conduction velocity (SNCV) and motor nerve conduction velocity (MNCV) were adapted from previously described methods (34).

Mechanical sensitivity was measured by applying a series of calibrated von Frey filaments (0.02, 0.04, 0.07, 0.16, 0.4, 0.6, 1, 1.4, 2, 4, 6, and 8 g) (Muromachi Kikai, Tokyo, Japan) to the plantar of the hind paw. Each filament was applied once to each mouse. Beginning with the 1-g filament, each filament was applied perpendicularly to the hind paw for 4–6 s. A brisk withdrawal of the hind paw indicated a positive response, and a lack of withdrawal indicated a negative response (35).

Responses to noxious heat were determined using a hot plate (NISSIN, Saitama, Japan). The mice were placed in a transparent plastic chamber on 50°C and 55°C metal hot plates to measure the latency of paw flinching, licking, or withdrawal. A maximum cutoff of 30 s was used to prevent tissue damage. A 5 min interval between consecutive stimulations of the same hind paw was used. The testing of the left lateral plantar hind paw was performed three times, and the withdrawal latencies were averaged (35).

Data are expressed as the mean ± SEM unless specified. Prism 7 (GraphPad Software, San Diego, CA) was used to perform the Student t test for comparing two conditions or one-way ANOVA, followed by the Tukey post hoc test, for multiple comparisons. The size-frequency histograms of the DRG neuronal area and the myelinated axon diameter in the sciatic nerve were compared by the χ² test. The Wilcoxon rank sum test was used to compare ranked size distributions between the groups. The differences were considered statistically significant at P < 0.05.

The data sets analyzed during the current study are available from the corresponding author upon reasonable request. The anti-PS antibody may be available from Ionis Pharmaceuticals upon reasonable request and with permission of Ionis Pharmaceuticals.

Elevated blood glucose levels were maintained in diabetic mice at 4 months after diabetes onset (133 ± 4.6 mg/dL in the nondiabetic group [n = 18] vs. 596 ± 3.2 mg/dL in the 4-month diabetic group [n = 25]; P < 0.0001). The size frequency distributions of the DRG neuronal area and the myelinated axon diameter in the sciatic nerve for the nondiabetic mice differed with high significance from those for the 4-month diabetic mice (P < 0.0001 and P < 0.0001, respectively) (Fig. 1A and B). In the nondiabetic mice, the size frequency histogram of the DRG neuronal area and the myelinated axon diameter in the sciatic nerve was biphasic with two peaks (700–800/1,200–1,300 μm² and 3–4/5–6 μm). In the 4-month diabetic mice, the frequency of large neurons and myelinated fibers was decreased, while that of small neurons and myelinated fibers increased compared with those in the nondiabetic mice, suggesting that atrophy of sensory neurons and axons occurred in diabetic mice. The g-ratio, a measurement of myelin sheath thickness, in the 4-month diabetic mice significantly increased compared with the nondiabetic group (P < 0.0001) (Fig. 1C). For a few degenerating profiles, the number of myelin ovoids per transverse section in the sciatic nerves in the 4-month diabetic mice (4 ± 0.4; n = 5) was significantly augmented compared with the nondiabetic mice (0.4 ± 0.2; n = 5, P < 0.0001) in the nondiabetic group. These findings indicated that the 4-month diabetic mice demonstrated neuronal and axonal atrophy with demyelinating features, which is consistent with previous reports (10,36).

Five months after the onset of diabetes, we examined MALAT1 expression by reverse transcription quantitative PCR (RT-PCR) in the DRG neurons of the diabetic mice. MALAT1 in the DRG was significantly (∼1.5 times) upregulated in the 5-month diabetic mice (n = 12) compared with that of the nondiabetic mice (n = 18; P = 0.0402) (Fig. 2A). The upregulation of MALAT1 was not observed in the sciatic nerves and the spinal cords of the 5-month diabetic mice (n = 12; P = 0.7086) (Fig. 2A).

We tested whether the expression of MALAT1 in the DRG of the 5-month diabetic mice can be inhibited by once-weekly intravenous injections of Toc-HDO targeting MALAT1 for 3 weeks (Fig. 2C). The PS backbones within Toc-HDO were stained with 3,3′-diaminobenzidine, and strong PS-positive signals were detected in the nuclei of DRG neurons 1 day after the first injection, identifying successful delivery to the DRG neurons, whereas muscle and skin, targets for sensory neuron innervation, had lesser uptake of Toc-HDO (Fig. 2D). MALAT1 expression in the DRG and the sciatic nerve of the diabetic mice (n = 12), compared with the diabetic mice that did not receive Toc-HDO injections (n = 10), was effectively inhibited by >70–80% after once-weekly Toc-HDO injections for 3 weeks (P < 0.0001) (Fig. 2B).

We determined the impact of MALAT1 silencing in the DRG by Toc-HDO injection on the phenotypes of experimental DPN. The 4-month diabetic mice, compared with the nondiabetic mice, showed significantly decreased MNCVs (n = 13, diabetic 25.9 ± 2.1 m/s; P < 0.0001) (Fig. 3A) and SNCVs (n = 6, diabetic 26.3 ± 3.8 m/s; P < 0.001) (Fig. 3B) in the sciatic-tibial nerve (n = 18, nondiabetic MNCVs 52.2 ± 1.1 m/s and n = 9, nondiabetic SNCVs 46.8 ± 3.5 m/s). The 5-month diabetic mice that received once-weekly Toc-HDO injections for 3 weeks (diabetes+Toc-HDO group), compared with the diabetic mice that did not (diabetes+PBS group), showed a significant further decrease in the SNCVs (n = 6, 21.6 ± 2.8 m/s vs. n = 8, 12.8 ± 0.4 m/s; P = 0.048) (Fig. 3B), although there was no significant change in the MNCVs.

To determine the effect of MALAT1 silencing on nociceptive responses, we assessed the thermal and mechanical sensitivities using the hot plate tests at 50°C and 55°C and a series of calibrated von Frey hairs (Fig. 3C–E). On the hot plate at 50°C and 55°C, the 4-month diabetic mice, compared with the nondiabetic mice, showed a significant decrease in the paw withdrawal response latencies, suggesting thermal hypersensitivity in the diabetic mice (n = 13, nondiabetic 19.5 ± 0.7 s vs. n = 18, diabetic 29.0 ± 0.4 s, P < 0.0001; n = 13, 5.6 ± 1.1 s vs. n = 18, 11.1 ± 0.4 s, P < 0.0001) (Fig. 3C and D). The paw withdrawal latencies of the 5-month diabetic mice that received once-weekly Toc-HDO injections for 3 weeks (diabetes+Toc-HDO group) on the hot plate at 50°C had a rise in their sensory threshold, indicating loss of sensation, whereas those that did not (diabetes+PBS group) remained hypersensitive (n = 10, Toc-HDO diabetic 27.8 ± 0.8 s vs. n = 13, control diabetic 19.6 ± 0.3 s, P < 0.0001) (Fig. 3C). Furthermore, the paw withdrawal latencies of the diabetes+Toc-HDO group, compared with those of the nondiabetic and diabetic+PBS mice, at 55°C were significantly prolonged, indicating additional sensory loss induced by MALAT1 inhibition (n = 10, Toc-HDO diabetic 18.3 ± 1.1 s vs. n = 13, diabetic 11.4 ± 0.3 s, P < 0.0001) (Fig. 3D). During the von Frey test, the mechanical withdrawal threshold of the 4-month diabetic mice, compared with that of the nondiabetic mice, showed a significant increase (n = 13, diabetic 4.2 ± 0.1 s vs. n = 18, nondiabetic 3.1 ± 0.1 s, P < 0.0001) (Fig. 3E), indicating mechanical hyposensitivity in the diabetic mice. The mechanical withdrawal threshold was significantly higher in the diabetes+Toc-HDO group than in the diabetes+PBS group at 5 months (n = 10, Toc-HDO diabetic 4.7 ± 0.1 s vs. n = 13, control diabetic 4.2 ± 0.1 s, P < 0.001) (Fig. 3E). These results showed that MALAT1 inhibition in the DRG induced additional sensory nerve dysfunction in DPN. The body weight of both diabetic groups was significantly reduced compared with that of the nondiabetic control group (n = 18, nondiabetic 52.7 ± 1.07 g vs. n = 13, diabetic 44.4 ± 1.41 g and n = 12, control diabetic 40.5 ± 0.96 g, P < 0.0001) (Fig. 3F). There was no significant difference at 5 months after STZ injection between the body weight of the diabetes+PBS group and the diabetes+Toc-HDO group, which indicated that Toc-HDO and MALAT1 knockdown did not influence the body weight in diabetic mice.

We identified an impact of MALAT1 silencing on neuronal and axonal caliber of DRG neurons in diabetic mice. The 5-month diabetic control mice showed neuronal atrophy in frozen sections, which was similar to those of the 4-month diabetic mice as shown in Fig. 1 (Fig. 4A–C). Moreover, after MALAT1 silencing, the size frequency of neuronal area tends to be increased in small neurons (<500 μm²) and decreased in large neurons (>1,000 μm²), including statistically significant reduction of the size frequency in 1100–1200 μm² of large neurons, compared with those in diabetic control mice (P < 0.05) (Fig. 4A). However, there were no additional measures to indicate changes of axonal atrophy and myelin thickness after MALAT1 knockdown in diabetic mice compared with diabetic control mice (Fig. 4B and C). These findings indicated that MALAT1 knockdown may have further exacerbated the DRG neuronal atrophy in the diabetic mice without an impact on myelin thinning.

We determined the impact of MALAT1 silencing on IENFD of the footpads where sensory nerves terminate. IENFD was significantly reduced in the diabetic mice (diabetes+PBS group, 28.8 ± 1.2 fibers/mm, n = 6) compared with nondiabetic mice (nondiabetic control group, 58.8 ± 2.1 fibers/mm, n = 9, P < 0.0001) (Fig. 5A and B). In the diabetes+Toc-HDO group, compared with the diabetes+PBS group, IENFD was further reduced (19.3 ± 0.7 fibers/mm, n = 7, P = 0.0028) (Fig. 5A and B), indicating that MALAT1 inhibition promotes additional degenerative axon loss in DPN.

MALAT1 is localized to the nucleus, specifically the nuclear speckles enriched in SRSFs (25,29) We confirmed that SRSF1/2 is localized in the nuclear speckles of the DRG sensory neurons in nondiabetic mice (Fig. 6 and Supplementary Fig. 1). In addition, the nuclear speckles in the diabetic and nondiabetic mice showed no structural changes. However, quantitative analysis revealed SRSF1/2-positive nuclear speckles in most DRG neurons in nondiabetic mice (n = 7, 93.7 ± 1.8%; n = 7, nondiabetic 83.7 ± 3.4% per neuron) (Fig. 6B and Supplementary Fig. 1B). They were observed in significantly fewer DRG neurons in diabetic mice (n = 5, diabetic percentage of SRSF1-positive signal in DRG neurons was 73.6 ± 4.4%, and percentage of SRSF2-positive signal in DRG neurons was 60.8 ± 4.6%; P < 0.0001) (Fig. 6B and Supplementary Fig. 1B). Furthermore, we examined the effect of MALAT1 silencing on the distribution of nuclear speckles in the DRG neurons of the diabetic mice. Most of the sensory neurons in diabetic mice with MALAT1 inhibition seemed to lose their nuclear speckles. In a small portion of sensory neurons, SRSF1/2 appeared to be sparsely distributed within the cytoplasm, and not within the nuclei, suggesting a mislocalization of splicing factors (Fig. 6A and Supplementary Fig. 1A). The percentage of DRG neurons with SRSF1/2-positive nuclear speckles was dramatically reduced in the diabetes+Toc-HDO group compared with the diabetic mice+PBS group (n = 5, percentage of SRSF1-positive signal in DRG neurons was 15.2 ± 6.4% and the percentage of SRSF2-positive signal in DRG neurons was 14.4 ± 3.3%, P < 0.0001) (Fig. 6B and Supplementary Fig. 1B). Linear regression showed a significant positive correlation between IENFD and the percentage of DRG neurons with SRSF1/2-positive nuclei, and this suggested nerve terminal loss correlated with the abnormal assembly of nuclear speckles in DRG neurons (n = 8, P = 0.0028 and P = 0.0002, respectively) (Fig. 6C). These results suggest that MALAT1 inhibition disturbs the formation of nuclear speckles in DRG neurons and that this may participate in the sensory neuron degeneration in DPN.

lncRNAs are acquiring greater recognition for their roles in the development of diabetes and its complications (22–24). The current study identified the upregulation of lncRNA MALAT1 in the DRG of diabetic mice. The administration of Toc-HDO targeting MALAT1 effectively inhibited its expression in the DRG of diabetic mice. MALAT1 inhibition, however, promotes sensory nerve dysfunction and degeneration accompanied by the disruption of nuclear speckles in DPN. These findings suggest that MALAT1 expression in DRG neurons limits the progression of nerve degeneration in DPN.

The mouse model of DPN has characteristic features, including nerve conduction velocity deficits, epidermal nerve fiber loss, and sensory loss. Previous studies involving similar animal models of chronic type 1 diabetes have reported the following (9,10,14,36,37): the DRG neurons of the 5-month diabetic mice had smaller areas (neuronal atrophy); motor and sensory conduction velocities reduced 1–3 months after the STZ injection; thermal and mechanical hypersensitivities occurred, often during the early stage of DPN, and decreased responses and hyposensitivity were observed usually later; and the footpad epidermal nerve fiber density was reduced after 3 months of diabetes. Sensory behavior may vary with the strain and species, sometimes with persistent hyperalgesia or earlier hypoalgesia. Our results are consistent with these generally accepted features of DPN in mice. However, our unique outcomes in the mice reported here showed that our diabetic mice showed thermal hypersensitivity when exposed to thermal nociceptive stimuli at 50°C, but this was resolved, and the sensitivity declined within a month after the MALAT1 inhibition. At 55°C, the diabetic mice also showed hypersensitivity, but the MALAT1 inhibition drastically reversed their hypersensitivity to hyposensitivity, rendering them less sensitive than nondiabetic controls. The difference between the thermal nociceptive responses at 50°C and 55°C may be attributed to the temperature properties dependent on the classes of nociceptors. Transient receptor potential vanilloid (TRPV) 1 acts as a moderate-threshold temperature (≥43°C) sensor, whereas TRPV2 is a high-threshold temperature (>52°C) sensor (38). TRPV1 and TRPV2 are mainly respectively distributed in C-fibers and Aδ-fibers that innervate the skin, joints, and muscles to carry information about noxious stimuli (38). Therefore, the difference between the thermal behaviors at 50°C and 55°C, especially the hyposensitivity at 55°C induced by MALAT1 inhibition in diabetic mice, may reflect the severe loss of Aδ-fibers rather than C-fibers, together with the evidence of the further reduction of SNCV and IENFD after MALAT1 inhibition.

Local delivery approaches have advantages, such as less systemic exposure, for achieving gene knockdown in DRG neurons, whereas intranasal or intrathecal administration may provide direct and rapid routes for drugs to enter the brain and cerebrospinal fluid through the olfactory epithelium and the subarachnoid space (39). However, these approaches are limited by several factors, including short-lasting effects and the requirement for more invasive procedures. ASOs are novel options for systemic gene delivery, but they have not been delivered well to DRG and cannot reach the CNS through the blood-brain barrier by systemic administration.

We recently developed a novel gene delivery technology, “DNA/RNA HDO,” for the highly efficient gene delivery to the DRG (H. Kaburagi, T.N., M. Enomoto, T. Hirai, M. Ohyagi, K. Ihara, K. Yoshida-Tanaka, S. Ebihara, K. Asada, H. Yokoyama, A. Okawa and T.Y., unpublished observations) and the CNS in HDO conjugated by lipid ligands through systemic injections compared with the parent single-stranded ASO (20). Systemic administration of Toc-HDO could be toxic, but we have reported no treatment-related adverse effect of Toc-HDO (19,20), and >90% silencing of MALAT1 expression in DRG of healthy mice by intravenous injection did not also impact pain behavior assessed by the acetone test, von Frey test, and the hot plate test (H. Kaburagi, T.N., M. Enomoto, T. Hirai, M. Ohyagi, K. Ihara, K. Yoshida-Tanaka, S. Ebihara, K. Asada, H. Yokoyama, A. Okawa and T.Y., unpublished observations). In the current study, we also demonstrated that Toc-HDO was apparently delivered to DRG neurons with less uptake in the targets for sensory neuron innervation (Fig. 2D) and very efficiently reduced their MALAT1 expression (Fig. 2B), resulting in the progressive neuronal atrophy and loss of nerve endings in DPN (Figs. 4 and 5), which suggests that MALAT1 silencing within DRG may contribute to a “dying-back” phenomenon, a retrograde degeneration following damage to neuronal cell bodies (Fig. 4). Moreover, previous reports (40–42) showed that MALAT1-knockout mice were healthy and fertile, showing no apparent phenotypes. Thus, MALAT1 knockdown targeting DRG may be a causative factor of further nerve dysfunction and IENF loss, but we cannot completely exclude other forms of targeting. This is also the first example we are aware of in which a systemic nucleotide therapy had a direct impact on DRG neurons with a functional consequence in DPN.

MALAT1 expression has been recently reported to be upregulated in the retina, kidney, and myocardium during diabetic complications and in the blood samples of the patients with DPN (24,30,31,43–45). Several previous studies have examined the effect of modulating MALAT1 in diabetic complications. MALAT1 knockdown diminished vascular leakage in the retina of diabetic mice and reduced diabetes-induced retinal inflammatory cytokines (43). In diabetic nephropathy, MALAT1 knockdown induced a protective effect on functional integrity of podocytes in vitro (44). Cardiomyocyte apoptosis was reduced with MALAT1 knockdown in diabetic rats, leading to an improvement in left ventricular function (30). Here, we first identified the upregulation of MALAT1 expression in the DRG in an experimental DPN model. MALAT1 has been known to be a stress response-related molecule that is upregulated by the exposure of cells to hypoxia (46,47). Hypoxia, which is a central pathophysiological phenomenon in diabetes-targeted organs, induces the transactivation of the MALAT1 promoter through the enhanced activity of hypoxia-inducible factor-1α (HIF-1α) in vitro (47). MALAT1 was also upregulated in in vivo blood samples of diabetic mice, which enhanced wound healing in diabetic mice through the activation of the HIF-1α signaling pathway (48). Thus, the upregulation of MALAT1 may be an essential stress response in various tissues, including DRG neurons, under diabetic conditions. Furthermore, we found that MALAT1 silencing promoted sensory neuron degeneration in DPN, although we reported that MALAT1 silencing by Toc-HDO in wild-type mice was not toxic to DRG neurons and it did not alter their pain behavior (H. Kaburagi, T.N., M. Enomoto, T. Hirai, M. Ohyagi, K. Ihara, K. Yoshida-Tanaka, S. Ebihara, K. Asada, H. Yokoyama, A. Okawa and T.Y., unpublished observations); the upregulation of MALAT1 in DPN may be a compensatory and protective response to diabetic stresses in DRG sensory neurons.

MALAT1 is a nuclear-retained lncRNA that contributes to synapse formation and neurite outgrowth through the activation of the extracellular signal-regulated kinase/mitogen-activated protein kinase (ERK/MAPK) signaling pathway in neurons (49,50). MALAT1 seems to have a critical physiological neuronal function; however, MALAT1-knockout mice did not demonstrate any abnormal phenotype, and MALAT1 deletion in mice embryo fibroblasts did not also alter the structures of nuclear speckles (42). However, under pathological conditions, MALAT1 has often been reported to have a protective role against neuronal stress (33,51,52). MALAT1 promotes neuronal survival after injury by mediating alterative splicing of protein kinase C-δ in vitro (52). MALAT1 silencing during ischemic stroke in vivo induced endothelial cell death through the upregulation of a proapoptotic BCL1 family protein (Bim) and proinflammatory cytokines, such as MCP-1, interleukin 6, and E-selectin, leading to a larger brain infarct and worsened sensorimotor function (51). Yao et al. (33) reported that MALAT1 was upregulated in the retina after optic nerve injury. They demonstrated that MALAT1 knockdown reduced reactive gliosis, Müller cell activation, and retinal ganglion cell survival through CREB signaling, which ultimately led to retinal neurodegeneration. Therefore, our results provide further evidence that MALAT1 has a neuroprotective role. On the other hand, the upregulation of MALAT1 in diabetic retinopathy has been reported to promote oxidative stress by impeding the transcription of antioxidant defense genes in retinal endothelial cells, and its knockdown serves as a potential target for antiangiogenic therapy in diabetic retinopathy (31,32). These results are at variance with the present work but might reflect the tissue-dependent effects related to the differences in pathogenesis between diabetic retinopathy and DPN. Given the findings here, it would be of interest to investigate whether overexpression of MALAT1 in diabetic mice would offer protection from neuropathy. However, this was beyond the scope of the current study.

The nucleus contains membraneless structures termed nuclear bodies, such as nucleoli, nuclear speckles, Cajal bodies (CBs), promyelocytic leukemia nuclear bodies, and paraspeckles (26). Nuclear bodies contribute to genomic functions, and the dysfunction of the nuclear body assembly has been reported to be involved with neurodegenerative diseases. For instance, Alzheimer disease and other tauopathies are related to the aberrant assembly of nuclear speckles, which causes the alternation of pre-mRNA splicing of tau exon 10 and results in tau accumulates (53). Our previous report also indicated that dysregulated CBs in DRG neurons may cause aberrant pre-mRNA splicing associated with the overexpression of CWC22, a splicing factor, and this leads to sensory neurodegeneration (5,7,10). CBs and nuclear speckles have been recently shown to offer a nuclear environment involved in the storage, assembly, and modification of small nuclear ribonucleoproteins and splicing factors during the process of mRNA splicing (26). Nuclear speckles also contain noncoding RNAs, including MALAT1, which regulates alternative splicing by modulating the phosphorylation states and their distributions of SR proteins (25,29,54). In the current study, diabetes moderately and significantly reduced the number of nuclear speckles in DRG neurons. Furthermore, MALAT1 silencing resulted in the evident loss of nuclear speckles in the DRG accompanied by a reduction in IENFD, and this suggested the aberrant assembly of nuclear speckles in the DRG neurons may be involved in sensory neurodegeneration. El Bassit et al. (52) reported that the recruitment of SRSF2 into the nuclei by MALAT1 was important for promoting the alternative splicing of the protein kinase C isoform, which is the activator of survival and proliferation in neurons after injury. Taken together, the disruption of nuclear speckles following the absence of MALAT1 in the DRG neurons may cause aberrant mRNA splicing of genes involved in axon maintenance and outgrowth that contributes to loss of terminal innervation and sensory neurodegeneration.

Taken together, we demonstrated that lncRNA MALAT1 was significantly upregulated in the DRGs of diabetic model mice as reported in the target organs for other diabetic complications and that the systemic administration of Toc-HDO targeting lncRNA MALAT1 facilitated a substantial neurodegenerative change in of DRG sensory neurons and their axons in chronic diabetic mice accompanied by the loss of nuclear speckles that contain splicing factors. Therefore, lncRNA MALAT1 may have neuroprotective roles during the development of DPN; it is possible that its overexpression in DRG neurons may be explored as a therapeutic strategy for DPN.

Acknowledgments. The authors thank Dr. Punit Seth, Ionis Pharmaceuticals for providing anti-PS antibody.

Funding. This research was supported in part by the Japan Agency for Medical Research and Development (AMED) under grant numbers 18am0301003h0005, 20am0401006h0002, and 19ek0610013h0003 to T.Y.

Duality of Interest. T.Y. collaborates with Daiichi Sankyo Company, Ltd, Mitsubishi Tanabe Pharma Corporation, Ono Pharmaceutical Company, Ltd, Rena Therapeutics Inc, Takeda Pharmaceutical Company, Ltd, Nanocarrier Pharmaceutical Company, Ltd, and Toray Industries, Inc, and serves as an academic adviser for Rena Therapeutics Inc. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. A.M. wrote the initial drafts of the paper. A.M. and M.K. designed the study, performed the experimental work, analyzed the data, and provided input into all drafts of the paper. M.K. wrote the initial drafts of the paper. S.I., T.N., and T.Y. guided the experiments using Toc-HDO. A.C. performed experiments and analyzed data. D.W.Z. offered intellectual input and editing to the drafts of the paper. T.Y. supervised the overall project, helped to collate and analyze data, and drafted the final drafts of the paper. All authors read and approved the manuscript. T.Y. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.