A functional screen implicates microRNA-138-dependent regulation of the depalmitoylation enzyme APT1 in dendritic spine morphogenesis

Abstract

The microRNA pathway has been implicated in the regulation of synaptic protein synthesis and ultimately in dendritic spine morphogenesis, a phenomenon associated with long-lasting forms of memory. However, the particular microRNAs (miRNAs) involved are largely unknown. Here we identify specific miRNAs that function at synapses to control dendritic spine structure by performing a functional screen. One of the identified miRNAs, miR-138, is highly enriched in the brain, localized within dendrites and negatively regulates the size of dendritic spines in rat hippocampal neurons. miR-138 controls the expression of acyl protein thioesterase 1 (APT1), an enzyme regulating the palmitoylation status of proteins that are known to function at the synapse, including the alpha(13) subunits of G proteins (Galpha(13)). RNA-interference-mediated knockdown of APT1 and the expression of membrane-localized Galpha(13) both suppress spine enlargement caused by inhibition of miR-138, suggesting that APT1-regulated depalmitoylation of Galpha(13) might be an important downstream event of miR-138 function. Our results uncover a previously unknown miRNA-dependent mechanism in neurons and demonstrate a previously unrecognized complexity of miRNA-dependent control of dendritic spine morphogenesis.

Figure 1. Expression profiling of miRNAs in synaptososomes

A) Western blot analysis of protein extracts derived from P15 rat forebrain (FB), low-speed centrifugation supernatant (S1), high-speed centrifugation supernatant (S2), high-speed centrifugation pellet (membrane, P2) or synaptosomes (SYN). Whereas pre- (synaptophysin) and postsynaptic proteins (GluR1) are enriched in SYN, the cytoplasmic EEA-1 is depleted from P2 and SYN. B) List of miRNAs that displayed an at least twofold change in expression between synaptosomes and forebrain in three different experiments. C) Hierarchical clustering of miRNAs displaying significantly different expression between whole forebrain and synaptosomes in three different experiments. Colour coding: enrichment in synaptosomes (syn) vs. whole forebrain (fb) in logarithmic scale, yellow=enriched in synaptosomes, blue=depleted in synaptosomes.

Figure 2. A subset of neuronal miRNAs is localized in the synaptodendritic compartment

A) Validation of microarray results by Northern blot. RNA was extracted from P15 rat forebrain or P15 rat synaptosomes, separated on 15% PAGE and probed for the presence of miR-218, miR-138, miR-124 or U6 snRNA. Numbers represent fold enrichment in synaptosomes. Note the absence of premiR-124 from synaptosomes. B) Subcellular localization of synaptically enriched miRNAs in hippocampal neurons. In situ hybridization was performed on hippocampal neurons at 18 DIV using FITC-coupled LNA probes directed against miR-9, miR-218 and miR-124, a DIG-coupled LNA probe directed against miR-138 or a DIG-coupled control LNA probe. Asterisks: Prominent staining of miR-124 in neuronal cell bodies. Arrowheads: Dendritic staining of miR-134, miR-218 and miR-138. Scale bar = 20 µm. Lower right panel: Quantification of ISH signal. Values are expressed as ratio between dendritic and somatic signal intensities. Dendrite/soma ratios of miR-138, miR-218 and miR-9 are significantly higher than the ratio of the somatic miR-124. * p<0.05 (t-test, n=3).

Figure 3. The synaptically enriched miR-138 regulates dendritic spine size

A) Flowchart of the functional screen for the identification of miRNAs controlling dendritic spine size. Rat hippocampal neurons (10 DIV) were transfected with GFP together with miRNA inhibitors (40 nM) or 134 duplex RNA (20 nM), and processed at 18 DIV for confocal microscopy. Fluorescence intensity of the freely diffusible GFP within 3D projections was used to calculate relative spine volume. B) Quantitative analysis of relative spine volume of hippocampal neurons transfected with the indicated double-stranded (ds) RNA (20 nM) or antisense 2’-O-methyl (2’O-me) oligonucleotides (40 nM). Values represent the mean ± S.D. (n=3; at least 750 spines from 15 neurons). *p<0.05 (compared to 2’O-methyl control conditions). C) Representative hippocampal neurons (18 DIV) transfected with GFP together with either 2’O-me control (left panel) or 2’O-me 138 (40 nM). Boxed insets display spines at higher magnification. Scale bars: 20 µm (overview); 8 µm (inset). D) LNA-mediated inhibition of miR-138 increases spine volume. Average spine volume of neurons treated with the indicated LNA antisense oligonucleotides (100 nM). Values were normalized to the GFP-only condition and represent the mean ± S.D. (n=3; at least 1200 spines from 15 neurons). p=0.001 (ANOVA). P-value of pairwise Student`s t-test is indicated. E) Overexpression of miR-138 in hippocampal neurons reduces spine volume. Duplex RNAs representing the full-length version of miR-138, a 3’ truncated miR-138 (miR138 sh) or miR-134 were transfected as in B and relative spine volume was determined. Values were normalized to the GFP-only condition and represent the mean ± S.D. (n=3; at least 1200 spines from 15 neurons). P=0.029 (ANOVA). P-values of pairwise Student’s t-tests are indicated. F) Restoring miR-138 function rescues increased spine size caused by miR-138 inhibition. Average spine volume was determined from neurons transfected with 2’O-me-138 (40 nM) together with synthetic miR-138 duplex RNA (20 nM) or control duplex RNA (20 nM, let-7c). Values were normalized to the GFP-only condition and represent the mean ± S.D. (n=3; at least 1200 spines from 15 neurons). P=0.0067 (ANOVA). P-values of pairwise Student’s t-tests are indicated.

Figure 4. miR-138 negatively regulates miniature synaptic transmission at excitatory synapses

A) Original recordings of mEPSCs in cultured hippocampal neurons in the presence of TTX and gabazine. Cells were voltage clamped at −70 mV. Traces show representative data from cells expressing GFP only, GFP + miR-138 duplex RNA, and control duplex RNA, respectively. B) Reduction of median mEPSC amplitudes in miR138 expressing neurons. Bars show the mean value and standard deviation of medians for each group (GFP n = 14; miR-138 n = 16; control duplex RNA n = 12). p-values (pairwise Student’s t-test) are indicated above bar graph. C) Cumulative amplitude distributions of mEPSCs recorded from neurons transfected as in A). Data pooled from 14 cells (GFP), 16 cells (mir-138), and 12 cells (control duplex RNA). Note leftward shift of cumulative curve for miR-138 (p < 0.001, Kolmogorov-Smirnov test). D) GluR2 surface staining of cells transfected as in A). Upper panel: Representative dendrites of neurons transfected with either miR-138 or control duplex RNA and stained for GluR2 (red) and synapsin (blue). Arrowheads depict large GluR2 clusters in spine heads of control neurons and small or absent GluR2 clusters in protrusions of miR-138 expressing neurons. Scale bar: 10 µm. Lower panel: The cumulative distribution of GluR2 surface cluster sizes is plotted for the indicated conditions. KS-test: p=0.001; D: 0,0927 (miR-138 vs. control transfected cells).

Figure 5. APT1 is expressed in neuronal dendrites during synapse development

A) Validation of predicted miR-138 target mRNAs by dual-luciferase reporter assay in cortical neurons (5 DIV). 3’UTR regions of the indicated genes were cloned downstream of the firefly luciferase coding region within pGL3, and the resulting vectors were transfected into neurons together with miR-138 duplex RNA (20 nM). Percent firefly luciferase activity, normalized to the internal Renilla luciferase control, of miR-138 expressing cells compared to control-transfected cells is plotted for each individual reporter construct. Error bars represent the mean ± S.D. (empty vector: n=5; APT1: n=6), each performed in duplicate. * p<0.05 (pairwise Student’s t-test; individual pvalues are indicated above each bar) B) APT1 mRNA is expressed in mouse hippocampus. In situ hybridization (ISH) was performed in adult mouse hippocampal slices using a probe perfectly complementary to mouse APT1 mRNA (red signal, left panel) or a mismatch control probe (right panel). Cell nuclei were visualized by Hoechst counterstain (blue signal). DG: dentate gyrus. C) APT1 mRNA localizes to dendrites of primary hippocampal neurons. ISH was performed in 9–10 DIV primary dissociated hippocampal neurons as described in B. Note the presence of multiple dendritic APT1 mRNA-containing granules along the entire length of dendrites (arrowheads). Scale bar: 20 µm. D) Quantification of ISH signals obtained in C. Data represent the mean ± S.D. for each of the indicated ISH probes (n=6 (APT1, control); n=8 (MAP2)). E) APT1 protein expression is upregulated during the development of primary cortical neurons in culture. Whole cell extracts were prepared from the indicated stages of cortical neuron development and simultaneously probed with a rabbit antiserum against APT1 (lower band) and a mouse monoclonal anti-β-Actin antibody as a loading control (upper band). F) APT1 protein is present at low levels in synaptosomal extracts. P15 rat brains were fractionated into cytosolic fractions (S1, S2), membranous fraction (P2) and synaptosomes (SYN), and fractions were probed for the presence of PSD-95, APT1 and TuJ1 as a loading control. Protein levels in whole forebrain before fractionation (WB) are shown for comparison.

Figure 6. APT1 is a miR-138 target mRNA in neurons

A) Schematic representation of the mouse miR-138-APT1 3’UTR duplex as determined by RNAfold. G:C base pairs are indicated by red, A:U and G:U base pairs by blue dots. Lower panel: Sequence conservation of the miR-138 target site within the APT1 3’UTR across the indicated vertebrate species. Note the extensive conservation of nucleotides at both ends of the duplex that are engaged in base-pairing, and the less extensive conservation in the internal bulge structure (adopted from UCSC genome browser). Nucleotide substitutions in the APT1 mutBS construct are indicated. B) Importance of the “seed” region within the APT1 3’UTR for miR-138 mediated downregulation in primary rat cortical neurons (5 DIV). Luciferase assay was performed with APT1 wt and mut 3’UTR constructs in the presence of the indicated concentrations of miR-138 duplex RNA. Values are expressed relative to the internal Renilla luciferase activity and normalized to the activity of the APT1 wt reporter under basal conditions. Bars represent the mean ± S.D. (10 nM: n=6; 20 nM: n=18; APT1 mut: n=8). *p<0.00001 (pairwise Student’s t-test between corresponding values of wt and mut reporter). C) Endogenous miR-138 inhibits APT1 luciferase reporter gene expression. APT1 wt or mut luciferase reporter genes were co-transfected together with increasing amounts of 2’O-me-138 AS oligonucleotide (50–200 nM) into hippocampal neurons (16 DIV) and luciferase activity was measured two days later. Results represent the mean ± S.D. (n=5). Activity of the APT1 wt and mut reporters under basal conditions was arbitrarily set to one. *p<0.01 (pairwise Student’s t-test between corresponding values of wt and mut reporter; p-values are shown above bar graphs). D) Endogenous miR-138 inhibits accumulation of APT1 protein. Cholesterol-modified 2’O-methyl antisense oligonucleotides directed against miR-138 or an unrelated sequence (scrambled) were introduced into hippocampal neurons at 13 DIV at a concentration of 1 µM, and protein extracts were prepared for Western blotting at 18 DIV. The average fold increase of the APT1 signal is indicated (n=3; 2’O-me control = 1).

Figure 7. APT1 activity is required for the control of dendritic spine size

A) shRNA-mediated knockdown of endogenous APT1 in primary neurons. 2 µg of pSuper shRNA vectors were electroporated into cortical neurons, and expression of endogenous APT1 and β-Actin were determined by Western blotting (7DIV). B) Interfering with APT1 expression in neurons reduces dendritic spine volume. Representative hippocampal neurons at 18 DIV transfected with eGFP and either control shRNA (left panel) or pSuper-APT1-shRNA-3 are shown. Scale bar = 20 µm. C) shRNA-mediated knockdown of endogenous APT1 in neurons reduces dendritic spine volume. Rat hippocampal neurons (14 DIV) were transfected with GFP alone or together with pSuper expressing shRNAs directed against APT1 mRNA (APT1 shRNA-1, 3 and 4) or an unrelated sequence (control shRNA, 5 ng/ml each). Values were normalized to the GFP-only condition and represent the mean ± S.D. (n=3; at least 1800 spines from 18 neurons). p<0.0001 (ANOVA). P-values of pairwise student’s t-tests are indicated. D) Pharmacological inhibition of APT1 in neurons reduces dendritic spine volume. Hippocampal neurons (18 DIV) were treated for 6 h with either inactive control compound (RB020) or specific APT1 inhibitors (FD196, FD253, 10 µM each) before analyzing spine volume. Values were normalized to the DMSO-only condition and represent the mean ± S.D. (n=3; 1800 spines from 18 neurons). p=0.044 (ANOVA). P-values of pairwise Student’s t-tests are indicated. E) shRNA-mediated knockdown of APT1 suppresses the increase in spine volume caused by miR-138 inhibition. Rat hippocampal neurons (14 DIV) were transfected with the indicated 2’O-me antisense oligonucleotides (40 nM) together with APT1 shRNA-3 or control shRNA (5 ng each). Values represent the mean ± S.D. (n=4; at least 2400 spines of 24 neurons). p=0.0075 (ANOVA). P-values of pairwise Student’s t-tests are indicated. F) APT1 expression rescues miR-138 mediated spine shrinkage. Rat hippocampal neurons (10 DIV) were transfected with miR-138 duplex RNA (20 nM) in combination with APT1 wt or APT1 mutBS (250ng/ml each). Values were normalized to the GFP-only condition and represent the mean ± S.D. (n=3; at least 1800 spines from 18 neurons). p=0.029 (ANOVA). P-values of pairwise Student’s t-tests are indicated.

Figure 8. miR-138 regulated membrane-association of Gα13 is required for the control of dendritic spine size

A) Western blot of total lysates from HEK293 cells transfected with control shRNA or APT1 shRNA 1–4 using anti-β-actin (top) or anti-APT1 (bottom) antibodies. A representative blot is shown (n=3). Note that only the APT1 shRNA-3 target site is conserved in human and rat. B) shRNA-mediated knockdown of endogenous APT1 increases the fraction of membrane-bound myc-Gα13 in HEK293 cells. Western blot of membrane (M) or cytosolic (C) lysates expressing myc-Gα13 together with APT1 shRNA-3 or a control shRNA using anti-myc (top), anti-calnexin (middle) or anti-beta-actin (bottom) antibodies. C) Quantification of signals obtained by Western blot (Fig. 8B). *p<0.05 (n=3; pairwise Student’s t-test). D) Expression of miR-138 increases the fraction of membrane-bound myc-Gα13 in HEK293 cells. Western blot of membrane (M) or cytosolic (C) lysates expressing myc-Gα13 with or without synthetic miR-138 duplex (40 nM) using anti-myc (top) anti-calnexin (middle) or anti-β-actin (bottom) antibodies. E) Quantification of signals obtained by Western blot (Fig. 8D). *p<0.05 (n=4; pairwise Student’s t-test). F) Expression of miR-138 decreases endogenous APT-1 protein levels in HEK293 cells. Western blot of whole cell lysates transfected with miR-134, miR-138 or control duplex RNA (10 nM) using anti-APT-1 (top) or anti-β-actin (bottom) antibodies. Quantification of independent blots is shown at the bottom (n=3; control duplex RNA = 1). Note the modest reduction in endogenous hAPT1 protein levels due to low co-transfection efficiency of miR-138 duplex RNA (ca. 50–60%). G) Expression of membrane-targeted Gα13 rescues spine enlargement caused by miR-138 inhibition. Rat hippocampal neurons (10 DIV) were transfected with the indicated LNA oligonucleotides (100 nM) together with Gα13 expression constructs (250 ng/ml each). Values were normalized to the GFP-only condition and represent the mean ± S.D. (n=3; at least 1800 spines from 18 neurons). p=0.019 (ANOVA). P-values of pairwise Student’s t-tests are indicated. H) Expression levels and subcellular distribution of Gα13 variants. Western blot of membrane (M) or cytosolic (C) lysates from HEK293 cells expressing the indicated Gα13 mutants.