Motor neuron-derived microRNAs cause astrocyte dysfunction in amyotrophic lateral sclerosis

Abstract

We recently demonstrated that microRNA-218 (miR-218) is greatly enriched in motor neurons and is released extracellularly in amyotrophic lateral sclerosis model rats. To determine if the released, motor neuron-derived miR-218 may have a functional role in amyotrophic lateral sclerosis, we examined the effect of miR-218 on neighbouring astrocytes. Surprisingly, we found that extracellular, motor neuron-derived miR-218 can be taken up by astrocytes and is sufficient to downregulate an important glutamate transporter in astrocytes [excitatory amino acid transporter 2 (EAAT2)]. The effect of miR-218 on astrocytes extends beyond EAAT2 since miR-218 binding sites are enriched in mRNAs translationally downregulated in amyotrophic lateral sclerosis astrocytes. Inhibiting miR-218 with antisense oligonucleotides in amyotrophic lateral sclerosis model mice mitigates the loss of EAAT2 and other miR-218-mediated changes, providing an important in vivo demonstration of the relevance of microRNA-mediated communication between neurons and astrocytes. These data define a novel mechanism in neurodegeneration whereby microRNAs derived from dying neurons can directly modify the glial phenotype and cause astrocyte dysfunction.

Figure 1

MiR-218 directly regulates EAAT2 expression through 3′ UTR binding and is sufficient for EAAT2 loss. (A) Schematic of the mouse EAAT2 3′ UTR indicating six putative miR-218 binding sites (targetscan.org) and the sequence conservation in select mammals. (B) Luciferase assays were performed to assess the relative strength of miR-218 binding sites in the EAAT2 3′ UTR as compared to empty vector and negative control Slc1a2/EAAT2 3′ UTR. n = 3 replicates/trial and 2–9 trials/condition. (C) Transient transfection of primary astrocytes with miR-218, but not scrambled miRNA results in EAAT2 loss. n = 4. The monomer EAAT2 band (∼60 kDa) is shown. (D) Quantification of westerns from four independent trials of C. Density is represented relative to GAPDH for the corresponding trial. Values represented as mean ± SEM. One-way ANOVA with multiple comparisons (B).

Figure 2

Wild-type and ALS astrocytes from adult mice do not express miR-218. (A) MiRAP data from Hoye et al. (2017) were plotted to demonstrate that miR-218 is specifically enriched in ChAT+ motor neurons as compared to other CNS cell types in adult spinal cord tissue (ChAT = choline acetyltransferase, cholinergic neurons; Syn = synapsin 1, all neurons; Lyz2 = lysozyme M, microglia; GFAP = glial fibrillary acidic protein, astrocytes), including GFAP+ astrocytes. n = 3/cell type. (B–E) MiR-218 in situ hybridizations indicate miR-218 is enriched in ChAT + motor neurons as compared to background (D; no probe) and EAAT2+, GFAP+ and ALDH1L1+ astrocytes in mouse spinal cord tissue. Scale bar = 25 μm. (F) MiRAP of tAgo2 derived from GFAP+ astrocytes indicates miR-218 is not upregulated in astrocytes from the spinal cords of 70- and 140-day ALS SOD1G93A mice as compared to astrocytes from littermate LSL-tAgo2, GFAP Cre control mice. n = 3 (control), 3 (70-day SOD1G93A) and 6 (140-day SOD1G93A). Values represented as mean ± SEM. One-way ANOVA with multiple comparisons (A and F).

Figure 3

Astrocytes take up extracellular, motor neuron-derived miR-218. (A) An miR-218 uptake reporter based on the cumate gene switch, where RNAi binding sites (black rectangles) for miR-218 or siRNA to GFP (negative control) are placed downstream of CymR repressor protein. In the absence of miR-218/siRNA, CymR binds to the cumate operon (CuO) and represses downstream firefly luciferase (Fluc). When miR-218/siRNA are present, they repress CymR, thus allowing Fluc transcription. The Fluc has been inverted and floxed, such that it is Cre dependent. (B) HEK 293T cells transiently transfected with Cre and the siRNA to GFP-regulated sensor respond to increasing amounts of siRNA. n = 3 technical replicates/condition. (C) To model extracellular miR-218, media from lysed NSC-34 cells was applied to primary astrocytes. EAAT2, but not EAAT1, is selectively lost in a dose-dependent fashion. The monomer EAAT2 band (∼60 kDa) is shown. (D and E) Primary astrocytes expressing the miR-218-regulated sensor, but not siRNA-regulated sensor, respond to lysed NSC-34 media in a Cre-dependent manner. n = 2–4 biological replicates with 3–4 technical replicates/assay. Values represented as mean ± SEM. One-way ANOVA with multiple comparisons (B and D).

Figure 4

Extracellular, motor neuron-derived miR-218 is protein-bound and its activity can be blocked using ASOs. (A) ALS end-stage rat model CSF was treated with RNase A, proteinase K and Triton™ X-100. Treatment with RNase A only degrades free miR-218 (∼15% of total). Treatment with proteinase K followed by RNase A degrades free and protein-bound miR-218 (∼80% of total). Treatment with Triton^™ X-100, to solubilize vesicles, followed by proteinase K and RNase A has no further effect on the miR-218 signal. Conversely, treatment of rat serum with proteinase K followed by RNase A degrades only ∼50% of let-7a, whereas pretreatment with Triton™ X-100 results in 100% loss of the let-7a signal. n = 3–4/condition. (B) Biotinylated anti-miR-218 oligonucleotides can precipitate miR-218, but not control miRNA 103a-3p, from ALS rat model CSF. n = 2/condition. Each trial was normalized to the corresponding no oligonucleotide control. (C) Treatment of primary astrocytes with media from sporadic ALS patient (sALS) iPSC-derived motor neurons (iMNs) results in loss of EAAT2 protein, but can be mitigated by addition of anti-miR-218 ASOs. n = 3. (D) Primary astrocytes expressing the miR-218 uptake sensor also respond to iPSC-derived motor neuron media, and this response is mitigated by anti-miR-218 ASOs. n = 3 Biological replicates with 2–4 technical replicates/assay. Values represented as mean ± SEM. One-way ANOVA with multiple comparisons (D).

Figure 5

Blocking miR-218 mitigates EAAT2 loss in ALS model mice. (A) SOD1G93A mice were treated with miR-218 ASOs or saline via ICV injection at 110 days, which is post-disease onset. n = 6–7/group (two cohorts of 3–4 mice/group). (B and C) Polysomes were isolated from the brainstems of treated mice at 140 days to check for miR-218 inhibition. Polysome-containing fractions from miR-218 ASO-treated mice have less miR-218, indicating it is excluded from these fractions as compared to saline-treated mice, consistent with miR-218 inhibition. n = 4/treatment. (D–G) Mice treated with miR-218 ASOs have preserved EAAT2 as compared to saline-treated mice in both the lumbar (D and E) and cervical (F and G) spinal cords, as assessed by immunoblot (D and E) and immunofluorescence (F and G), respectively. n = 6–7/group (two cohorts of 3–4 mice/group) (D and E) and n = 3–5/group with 6–8 sections/mouse (F and G). The monomer EAAT2 band (∼60 kDa) is shown in (D). Images in (F) were taken such that no saturation occurred for quantitation and then equivalently brightened for demonstrative purposes. Scale bar = 100 μm. Values represented as mean ± SEM. Student’s unpaired, two-tailed t-test (E and G) with Benjamini-Hochberg correction for multiple comparisons (FDR < 0.1). Adjusted P-values: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

Figure 6

Putative miR-218 targets that are translationally downregulated in ALS astrocytes are de-repressed upon miR-218 inhibition. RT-qPCR of RNA obtained from polysome profiles of SOD1G93A mouse brainstems treated with saline or anti-218 ASOs indicates that putative miR-218 targets are de-repressed following anti-218 ASO treatment. n = 4/group. Conversely, non-miR-218 targets are not de-repressed upon anti-218 ASO treatment. n = 4/group. Student’s unpaired, two-tailed t-test with Benjamini-Hochberg correction for multiple comparisons (FDR < 0.1). Adjusted P-values: *P ≤ 0.1.

Figure 7

Blocking miR-218 mitigates astrogliosis in ALS model mice. (A–H) Mice treated with miR-218 ASOs have reduced expression of both Cx43 and GFAP, markers of astrogliosis, as compared to saline-treated mice in both the lumbar (A and B) and cervical (C–H) spinal cords, as assessed by immunoblot (A and B) and immunofluorescence (C–H), respectively. n = 6–7/group (two cohorts of 3–4 mice/group) (A and B) and n = 3–5/group with 8–10 sections/mouse (C–H). Images in (C, D, F and G) were taken such that no saturation occurred for quantitation and then equivalently brightened for demonstrative purposes. Scale bar = 100 μm. Values represented as mean ± SEM. Student’s unpaired, two-tailed t-test (B, E and H) with Benjamini-Hochberg correction for multiple comparisons (FDR < 0.1). Adjusted P-values: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.