A hidden threshold in motor neuron gene networks revealed by modulation of miR-218 dose

Abstract

Disruption of homeostatic microRNA (miRNA) expression levels is known to cause human neuropathology. However, the gene regulatory and phenotypic effects of altering a miRNA's in vivo abundance (rather than its binary gain or loss) are not well understood. By genetic combination, we generated an allelic series of mice expressing varying levels of miR-218, a motor neuron-selective gene regulator associated with motor neuron disease. Titration of miR-218 cellular dose unexpectedly revealed complex, non-ratiometric target mRNA dose responses and distinct gene network outputs. A non-linearly responsive regulon exhibited a steep miR-218 dose-dependent threshold in repression that, when crossed, resulted in severe motor neuron synaptic failure and death. This work demonstrates that a miRNA can govern distinct gene network outputs at different expression levels and that miRNA-dependent phenotypes emerge at particular dose ranges because of hidden regulatory inflection points of their underlying gene networks.

Keywords: amyotrophic lateral sclerosis; gene dosage; gene networks; haploinsufficiency; microRNA-218; motoneuron; neurodevelopment; neuromuscular junction; neuropathology; single cell RNA sequencing.

Figure 1.

miR-218 is reduced in ALS patient spinal tissue. (A) miR-218 in situ hybridization signal in ventrolateral motor neurons of post-mortem adult human spinal cords. Scale bar, 1mm. (B) Venn diagram of miRNAs significantly reduced in profiling studies of human post-mortem neural tissue (either dissected spinal tissue (red) or laser captured motor neurons (blue) from sporadic (sALS), familial (fALS) and C9ORF72 ALS patients versus controls (Butovsky et al., 2015; Campos-Melo et al., 2013; Emde et al., 2015; Figueroa-Romero et al., 2016; Reichenstein et al., 2019). Only miR-218 was significantly reduced in patients versus controls across all studies, quantified in (C) where miR-218 reduction was between 30% and 90%.

Figure 2.

A miR-218 dose threshold for neuromotor control. (A) miR-218 is expressed via alternative motor neuron-specific promoters (green) encoded within the Slit2 and Slit3 genes. (B) Motor neuron-specifying transcription factor binding sites targeted by CRISPR/Cas9 (Rhee et al., 2016). (C) Mouse oocytes were injected with Cas9 mRNA and gRNAs to generate a 264bp deletion (ΔILO). (D) Overlayed RNAseq reads of control (black) and 218-2^ΔILO/ΔILO (red) motor neurons. (E) miR-218-2 knockout (2KO) mice have a partially penetrant phenotype of early post-natal death. (F) Neurofilament M (NF) and synaptophysin (SYN) staining of pre-synaptic motor axons and alpha-bungarotoxin (α-BTX) staining of post-synaptic acetylcholine receptors on E18 muscle tissue. Large regions of 1KO;2^−/ΔILO muscle (white dashed line) lack motor axon innervation. Scale bar, 50μm. (G) Neuromuscular apposition of synaptophysin/NF and α-BTX in limb muscles at E18. SEM, one-way ANOVA, Holm-Sidak multiple comparison test. (H) 1KO;2^−/ΔILO mice (right) die after E18 cesarean section due to lack of respiration.

Figure 3.

Wide variance in miR-218 target gene dose responses. (A) Hb9::gfp+ motor neurons from each respective genotype were FACS-isolated from microdissected embryonic spinal cords in triplicate. (B) miR-218 expression normalized against U6 and the pan-neuronal miR-124 by Taqman qPCR. SEM, one-way ANOVA, Sidak multiple comparison test. (C) miR-218 expression level inversely correlates with phenotypic severity. (D) Identification of 316 high confidence mRNA targets in motor neurons. (E) As a group, miR-218 target mRNAs exhibit stepwise de-repression with decreasing miR-218 expression across genotypes (Repeated measures ANOVA, Holm-Sidak’s multiple comparisons test). (F) Hierarchical clustering of target mRNA expression by genotype reveals differences in miR-218 dose responses (Pearson correlation). (G-J) Semi-log and exponential regulons have distinct miR-218 dose-responses. (G) Representative semi-log and (I) exponential regulon target mRNAs (semi-log and exponential regressions with 95% CI, SEM). Double normalized dose-response curves of (H) semi-log and (J) exponential regulon target mRNAs versus miR-218 level, phenotype, and genotype.

Figure 4.

miR-218 dose responsive regulons are distinguished by 3’UTR characteristics. (A) Magnitude of fold change and (B) bioinformatic prediction of miR-218 targeting efficiency are not significantly different in semi-log and exponential regulons. (C) Semi-log regulon mRNAs have longer 3’UTRs and (D) carry more miR-218 miRNA response elements (MREs) than exponential regulon mRNAs. (E) Exponential regulon mRNAs have higher abundance and (F) are more biased towards the axonal compartment than semi-log regulon mRNAs. (G) GO term enrichment of semi-log and exponential regulons. (H) miR-218 dose proportionally modulates the expression of semi-log regulon mRNAs, which are characterized by longer 3’UTRs and more MREs per mRNA. Exponential regulon mRNAs are highly sensitive to miR-218 dose at low expression levels and their repression is saturated at high levels, resulting in a miR-218 dose-dependent inflection point in target gene de-repression. Mann-Whitney test, multiple hypothesis testing-adjusted p-values reported (Bonferroni-Holm method).

Figure 5.

miR-218 indirectly activates a peripheral neuronal gene signature within motor neurons. (A) 64 mRNAs were significantly de-activated by greater than 50% in DKO motor neurons. (B) These mRNAs, termed the indirect regulon, are indirectly activated by miR-218, potentially as a consequence of direct miR-218-mediated repression of unknown transcriptional repressors or downstream gene network effects. (C) Indirect regulon genes exhibited stepwise de-activation with decreasing miR-218 levels across genotypes. (D) The indirect regulon is associated with synaptic and vesicular processes and a gene signature of dorsal root ganglion (DRG) peripheral sensory neurons. (E) Representative images of the spatial expression pattern of indirect regulon mRNAs in the Allen Brain database in P4 mice. Strong and specific signal is seen in the DRG (magenta) and spinal motor neurons (MNs, green) but not spinal interneurons or glia. Image credit: Allen Institute. (F) Prph RNA is reduced in DKO motor neurons but unchanged in DRG. Scale bar 150μm. (G) We dissected DRG and performed FACS of Wnt1:Cre; Rosa:LSL:tdt embryos in duplicate prior to RNA sequencing. (H) PCA demonstrates separation of spinal MNs, spinal interneurons (INs), and neural crest-derived DRG reflecting their highly distinct transcriptomic identities. In contrast, individual V1, V2a and V3 interneuron subtypes cluster together. (I) Fold-change versus fold-change plot of cellular mRNA enrichment. Indirect regulon genes are enriched in the top right quadrant, representing higher expression in both DRG and MNs versus interneurons.

Figure 6.

Motor neuron subtype identity acquisition is not affected by miR-218. (A) Motor neuron spinal diversity along the rostro-caudal axis by motor columns and divisions. (B) Droplet-based scRNA sequencing of FACS-isolated Hb9::gfp+ motor neurons from microdissected WT and DKO spinal cords. Motor neurons were identified by choline acetyltransferase (ChAT) expression. (C) Guided identification of motor neuron subtypes via (D) expression patterns of known transcription factor markers. (E) Rostro-caudal identity by Hox gene expression. (F) Pseudo-time plot of pMN, immature and MMC motor neurons aligns cells along a neurodevelopmental differentiation timeline and does not segregate by presence or absence of miR-218. (G) UMAP does not segregate motor neurons by genotype. (H) Subtype identity is not significantly affected by the presence or absence of miR-218. pMN, motor neuron progenitors; MMC, medial motor column; P/HMC, phrenic and hypaxial motor column; LMC, lateral motor column – (l) lateral and (m) medial divisions; PGCa/b, preganglionic motor column – (a/b) divisions.

Figure 7.

miR-218 dose-dependent effects differ in magnitude across somatic and visceral motor neuron subtypes. (A-D) Hypergeometric enrichment analysis of 8bp 3’UTR miRNA binding site motifs in ranked gene lists. (A) miR-218’s seed sequence (AAGCACAA) is enriched in genes expressed higher in DKO versus WT motor neurons (B) but not DKO versus WT interneurons. (C) miR-218’s seed sequence is enriched in genes expressed higher in WT interneurons versus WT motor neurons (D) but not versus DKO motor neurons. (E) Motor neurons of the same genotype, subtype, and replicate were combined in silico into individual pseudo-bulk samples. Hierarchical clustering identified divisions in somatic versus visceral motor neurons and further divisions between WT and DKO motor neurons (Pearson correlation). (F) PCA identifies divisions between visceral and somatic motor neurons along PC1 and genotype along PC2. (G) Differential expression of bioinformatically predicted miR-218 targets identifies significant de-repression of targets in DKO versus WT motor neurons within a given subtype. (H) miR-218-mediated effects are highly correlated in motor neuron subtype. (I) Fold-change versus fold-change plot (DKO versus WT) of direct and indirect miR-218 targets by subtype. (J and K) Fold-change (DKO versus WT) of mRNAs within (J) exponential and semi-log and (K) indirect regulons versus pri-miR-218-1/2 expression level, by motor subtype (error bars: SEM; exponential and indirect regulons: one-phase association regression; semi-log regulon: semi-log regression; 95% CI).