RNF8/UBC13 ubiquitin signaling suppresses synapse formation in the mammalian brain

Abstract

Although ubiquitin ligases have been implicated in autism, their roles and mechanisms in brain development remain incompletely understood. Here, we report that in vivo knockdown or conditional knockout of the autism-linked ubiquitin ligase RNF8 or associated ubiquitin-conjugating enzyme UBC13 in rodent cerebellar granule neurons robustly increases the number of parallel fiber presynaptic boutons and functional parallel fiber/Purkinje cell synapses. In contrast to the role of nuclear RNF8 in proliferating cells, RNF8 operates in the cytoplasm in neurons to suppress synapse differentiation in vivo. Proteomics analyses reveal that neuronal RNF8 interacts with the HECT domain protein HERC2 and scaffold protein NEURL4, and knockdown of HERC2 or NEURL4 phenocopies the inhibition of RNF8/UBC13 signaling on synapse differentiation. In behavior analyses, granule neuron-specific knockout of RNF8 or UBC13 impairs cerebellar-dependent learning. Our study defines RNF8 and UBC13 as components of a novel cytoplasmic ubiquitin-signaling network that suppresses synapse formation in the brain.

Fig. 1

The E3 ubiquitin ligase RNF8 suppresses presynaptic differentiation and synapse formation in the cerebellum in vivo. a Left: a representative image of a cerebellum from a P12 rat pup electroporated with the GFP expression plasmid 8 days earlier. Granule neurons have descended to the internal granule layer (IGL) and axonal parallel fibers reside in the molecular layer (ML). Arrowheads denote varicose structures along the parallel fibers that represent presynaptic boutons. Right: P4 rat pups were electroporated with the RNF8 RNAi or control U6 plasmid together with a GFP expression plasmid and sacrificed 8 days later. The cerebellum was removed, sectioned, and subjected to immunohistochemistry using a GFP antibody. Knockdown of RNF8 increased the density of presynaptic boutons in the cerebellar cortex in vivo (***p < 0.05, t test, n = 5 rats). b P9 RNF8 ^loxP/loxP mice were electroporated with the Cre expression plasmid or control vector together with the GFP expression plasmid and boutons along granule neuron parallel fibers were analyzed as in a. Cre-induced knockout of RNF8 in granule neurons increased presynaptic bouton density (***p < 0.001, t test, n = 6 mice). c P9 RNF8 conditional knockout (RNF8 cKO) and control RNF8 ^loxP/loxP mice were electroporated with the GFP expression plasmid and analyzed at different stages of synapse development as in a. Left: representative axons at P17. Right: quantification of the number of presynaptic boutons at different stages of development. Little or no difference in presynaptic bouton number was observed at P13. By contrast, the number of presynaptic boutons was increased at P17 upon conditional knockout of RNF8 (***p < 0.001, t test, n = 3 mice), and this difference was maintained at P21 (**p < 0.01, t test, n = 4–5 mice). Scale bars represent 10 μm. d The cerebellum from P24 RNF8 cKO and control mice were subjected to electron microscopy analyses. Left: representative electron micrographs of the molecular layer of the cerebellar cortex. Parallel fiber/Purkinje cell synapses are denoted by asterisks. Scale bar = 500 nm. Right: quantification of the density of parallel fiber/Purkinje cell synapses in RNF8 cKO and control mice. The density of parallel fiber/Purkinje cell synapses was increased in RNF8 cKO mice compared to control mice (*p < 0.05, t test, n = 3 mice)

Fig. 2

RNF8 suppresses granule neuron to Purkinje cell neurotransmission in the cerebellum. a Acute sagittal cerebellar slices from P20–25 RNF8 cKO and control mice were subjected to electrophysiological analyses. Evoked excitatory postsynaptic currents (EPSCs) were recorded in Purkinje neurons in response to stimulation of parallel fibers with increasing intensity (30, 40, and 50 μA). Representative current traces (left) and quantification of the amplitude of evoked EPSCs (right) are shown. The amplitude of evoked EPSCs in Purkinje neurons was increased in RNF8 cKO mice compared to control mice (*p < 0.05 at 30 and 40 μA, ***p < 0.005 at 50 μA, ANOVA followed by Fisher’s PLSD post hoc test, n = 14–15 neurons, four mice). b Acute coronal cerebellar slices were prepared as in (a), and parallel fiber axons were stimulated at sites 400 μm away from an extracellular recording electrode. A representative trace of the stimulus-evoked presynaptic waveform before and after the application of tetrodotoxin is shown (left). The stimulus artifact was removed for clarity. On the right, quantification of presynaptic volley amplitude is shown. Conditional knockout of RNF8 had little or no effect on the presynaptic volley amplitude. c, d Acute sagittal cerebellar slices were prepared as in (a) and Purkinje cell miniature EPSCs (mEPSCs) were recorded. Representative traces of mEPSCs from RNF8 cKO and control mice are shown (c, left). Quantification of the mean (c, right) and cumulative distribution (d) of the mEPSC frequency and amplitude are shown. The frequency of mEPSCs was increased in RNF8 cKO mice compared to control mice (*p < 0.05, t test, n = 18–20 neurons, three mice). Conditional knockout of RNF8 had little or no effect on the amplitude of mEPSCs in Purkinje neurons

Fig. 3

RNF8 operates in the cytoplasm to suppress presynaptic differentiation in vivo. a Immunoblotting analyses of fractionated lysates of P8 rat cerebellum revealed abundant endogenous RNF8 in the cytoplasmic fraction in the cerebellum. b P4 rat pups were electroporated with an expression plasmid encoding an RNF8–GFP fusion protein together with a mCherry expression plasmid. After 8 days, rat pups were sacrificed and the cerebellum was subjected to immunohistochemistry using GFP and DsRed (mCherry) antibodies. RNF8 localizes predominantly in the cytoplasm. c P4 rat pups were electroporated with the control U6 plasmid or RNF8 RNAi plasmid together with an expression plasmid encoding NES(nuclear export signal)-RNF8Res fusion protein, NLS(nuclear localization signal)-RNF8Res fusion protein or the control vector and with the GFP expression plasmid and analyzed as in Fig. 1a. Expression of NES-RNF8Res, but not NLS-RNF8Res, reversed the RNF8 RNAi-induced phenotype of increased number of parallel fiber presynaptic boutons in the cerebellum in vivo (*p < 0.05, ANOVA followed by Fisher’s PLSD post hoc test, n = 3–4 rats). Scale bars = 10 μm

Fig. 4

The ubiquitin-conjugating E2 enzyme UBC13 operates with RNF8 to suppress presynaptic differentiation in vivo. a P4 rat pups were electroporated with the control U6 plasmid or RNF8 RNAi plasmid together with an expression plasmid encoding RNF8Res, RNF8Res in which C405 was mutated to serine (RNF8ResC405S), or the control vector and with the GFP expression plasmid and analyzed as in Fig. 1a. RNF8Res, but not RNF8ResC405S, reversed the RNF8 RNAi-induced phenotype of increased presynaptic bouton number in vivo (*p < 0.05, ANOVA followed by Fisher’s PLSD post hoc test, n = 3-4 rats). b P4 rat pups were electroporated with the UBC13 RNAi, UBCH8 RNAi, or control U6 plasmid and analyzed as in Fig. 1a. Knockdown of UBC13, but not knockdown of UBCH8, increased the density of presynaptic parallel fiber boutons in the cerebellar cortex in vivo (***p < 0.005, t test, n = 3 rats). c P4 rat pups were electroporated with the control U6 plasmid or RNF8 RNAi together with a RNF8Res mutant in which isoleucine 407 was replaced with alanine (RNF8ResI407A) or the control vector and with the GFP expression plasmid and analyzed as in Fig. 1a. Expression of RNF8ResI407A reversed the RNF8 knockdown-induced phenotype on presynaptic boutons (*p < 0.05, ANOVA followed by Fisher’s PLSD post hoc test, n = 3–6 rats). For representative images of axons see Supplementary Fig. 6c. d P6/7 UBC13 ^loxP/loxP mice were electroporated with the Cre expression plasmid or control vector together with the GFP expression plasmid and analyzed as in Fig. 1a. Cre-induced knockout of UBC13 in granule neurons increased presynaptic bouton density in vivo (***p < 0.001, t test, n = 6 mice). Scale bars = 10 μm. e The cerebellum from P25 UBC13 conditional knockout mice and control UBC13 ^loxP/loxP mice were subjected to electron microscopy analyses. Left: representative electron micrographs of the molecular layer of the cerebellar cortex. Parallel fiber presynaptic bouton/Purkinje cell synapses are denoted by asterisks. Scale bar = 500 nm. Right: quantification of the density of granule neuron parallel fiber/Purkinje cell synapses in UBC13 cKO and control mice. Knockout of UBC13 induced an increase in the density of parallel fiber/Purkinje cell synapses compared to control mice (*p < 0.05, t test, n = 3 mice)

Fig. 5

UBC13 suppresses granule neuron to Purkinje cell neurotransmission in the cerebellum. a Acute sagittal cerebellar slices from P23–26 UBC13 cKO and control mice were subjected to electrophysiological analyses. Evoked excitatory postsynaptic currents (EPSCs) were recorded in Purkinje neurons in response to stimulation of parallel fibers with increasing intensity (30, 40, and 50 μA). Representative current traces (left) and quantification of the amplitude of evoked EPSCs (right) are shown. The amplitude of evoked EPSCs in Purkinje neurons was increased in UBC13 cKO mice compared to control mice (***p < 0.005 at 40 μA, *p < 0.05 at 50 μA, ANOVA followed by Fisher’s PLSD post hoc test, n = 25–29, six mice). b Acute coronal cerebellar slices were prepared as in (a), and parallel fiber axons were stimulated at sites 400 μm away from an extracellular recording electrode. A representative trace of the stimulus-evoked presynaptic waveform before and after the application of tetrodotoxin is shown (left). The stimulus artifact was removed for clarity. On the right, quantification of presynaptic volley amplitude is shown. Conditional knockout of UBC13 had little or no effect on the presynaptic volley amplitude. c, d Acute sagittal cerebellar slices were prepared as in (a) and Purkinje cell mEPSCs were recorded. Representative traces of mEPSCs from UBC13 cKO mice and control mice are shown (c, left). Quantification of the mean (c, right) and cumulative distribution (d) of the mEPSC frequency and amplitude are shown. The frequency of mEPSCs was increased in UBC13 cKO mice compared to control mice (*p < 0.05, t test, n = 32 neurons, six mice). Conditional knockout of UBC13 had little or no effect on the amplitude of mEPSCs in Purkinje neurons

Fig. 6

RNF8 interacts with HERC2 and NEURL4 in the cerebellum and thereby suppresses parallel fiber presynaptic differentiation in vivo. a Lysates of Neuro2A mouse neuroblastoma cells constitutively expressing FLAG-RNF8 were subjected to immunoprecipitation followed by immunoblotting. Coimmunoprecipitation of FLAG-RNF8 with endogenous HERC2 and NEURL4. b The cytoplasmic fraction of lysates of P7 rat cerebellum was subjected to immunoprecipitation followed by immunoblotting. Endogenous coimmunoprecipitation of RNF8 and HERC2 in the cytoplasmic fraction of the cerebellum. c P4 rat pups were electroporated with the NEURL4 RNAi, HERC2 RNAi, or control U6 plasmid and analyzed as in Fig. 1a. Knockdown of NEURL4 and HERC2 increased the density of presynaptic parallel fiber boutons in the cerebellum in vivo (NEURL4 RNAi, ***p < 0.005, t test, n = 3–8 rats; HERC2 RNAi, ***p < 0.001, t test, n = 6–9 rats). d P4 rat pups were electroporated with the control U6 plasmid or RNF8 RNAi together with an RNF8Res mutant in which arginine 42 was replaced with alanine (RNF8ResR42A) or the control vector and with the GFP expression plasmid and analyzed as in Fig. 1a. Expression of RNF8ResR42A failed to reverse the RNF8 knockdown-induced presynaptic bouton phenotype in the cerebellum in vivo. Scale bars = 10 μm

Fig. 7

Cerebellar-dependent learning is impaired in RNF8 or UBC13 conditional knockout mice. a Schematic of delayed eye-blink conditioning assay depicting a mouse trained with a conditioned stimulus (CS: blue LED) paired with an eyeblink-eliciting unconditioned stimulus (US: periocular air puff). Depletion of RNF8 (b) or UBC13 (c) impaired performance on the eye-blink conditioning task (***p < 0.005, *p < 0.05, ANOVA followed by Fisher’s PLSD post hoc test)