Abstract

   Neuronal development and function are orchestrated by a plethora of
   regulatory mechanisms that control the abundance, localization,
   interactions, and function of proteins. A key role in this regard is
   assumed by post-translational protein modifications (PTMs). While some
   PTM types, such as phosphorylation or ubiquitination, have been
   explored comprehensively, PTMs involving ubiquitin-like modifiers
   (Ubls) have remained comparably enigmatic (Ubls). This is particularly
   true for the Ubl Nedd8 and its conjugation to proteins, i.e.
   neddylation, in nerve cells. In the present study, we generated a
   conditional Nedd8 knock-out mouse line and examined the consequences of
   Nedd8-deletion in cultured post-mitotic glutamatergic neurons. Our
   findings reveal that Nedd8-ablation in young glutamatergic neurons
   causes alterations in the expression of developmental transcription
   factors that control neuronal differentiation, ultimately leading to
   defects in the development of a mature glutamatergic neuronal
   phenotype. Apparent manifestations of these defects include increased
   vGlut2 expression levels, reduced vGlut1 and endophilin1 expression
   levels, reduced dendrite complexity, and increased transmitter release
   probability. Collectively, our results highlight a pivotal role for
   neddylation in controlling the fate of glutamatergic neurons and
   excitatory synaptic transmission.

   Subject terms: Synaptic transmission, Molecular neuroscience
     __________________________________________________________________

   This study highlights a pivotal role for Neddylation in controlling the
   fate of glutamatergic neurons and excitatory synaptic transmission.

Introduction

   Neurons are post-mitotic cells that usually survive for the lifespan of
   their host organism. To allow for such long lifetimes, neuronal protein
   homeostasis is dynamically regulated and tightly controlled to support
   optimal functional efficacy and adaptability to changes in their
   environment. Synapses between neurons represent the prime communication
   units in the brain. They determine the spatiotemporal organization of
   neuronal inputs and outputs, thereby defining the layout and function
   of neuronal circuits, and their plasticity is a key determinant of
   memory processes. Accordingly, as with neurons in general, synaptic
   protein homeostasis and function are under tight dynamic
   control^[46]1–[47]3.

   Post-translational protein modifications (PTM) are essential regulatory
   mechanisms in this regard, beyond processes that control gene
   transcription and mRNA translation. They expand the proteome by
   diversifying protein structure and function. Moreover, PTMs are at the
   core of many signaling pathways that enable cells to develop,
   differentiate, and rapidly respond or adapt to changes in their
   physiological status and environment^[48]4–[49]8.

   The spectrum of known PTMs is broad, ranging from proteolytic
   processing and chemical modification of amino acid side chains to
   protein conjugation. Likewise, the functional consequences of PTMs are
   very diverse, including, among other aspects, alterations of protein
   interactions, activities, localization, or half-life^[50]4,[51]8,[52]9.
   As concerns neuronal development and function, an extensive number of
   studies demonstrated fundamental roles of multiple PTMs, such as
   phosphorylation and ubiquitination, in neuronal physiology and
   pathophysiology^[53]9,[54]10. Indeed, the PTM of key transcription
   factors orchestrates the acquisition of cell type identity, neuronal
   differentiation, and synapse diversity^[55]11–[56]14, and the PTM of
   synaptic proteins regulates every step of the synapse fate, from
   synaptogenesis, via synapse maturation and plasticity, to synapse
   maintenance and elimination^[57]6,[58]8,[59]15–[60]17.

   PTMs involving ubiquitin-like modifiers (Ubls, e.g., SUMO) have emerged
   as a novel protein regulatory principle of fundamental importance in
   many cell biological processes, operating at a level of complexity akin
   to phosphorylation or ubiquitination^[61]4,[62]8,[63]10. For instance,
   we and others showed that the sumoylation of multiple transcriptional
   regulators affects neuronal development and function, supporting the
   important notion that the PTM of key transcriptional regulators is
   essential for the determination of cell type identity, neuronal
   differentiation, and synapse diversity^[64]18–[65]23.

   Beyond SUMOs, Nedd8 (neuronal precursor cell expressed developmentally
   down-regulated 8) is arguably the second best characterized Ubl^[66]24.
   Like all other Ubls, Nedd8 is conjugated to lysine residues in target
   proteins via a three-step enzymatic cascade^[67]4,[68]25, involving an
   E1 (Nae1, heterodimer of Appb1 and Uba3), an E2 (Ubc12), and several E3
   enzymes (RBX1/2, DCN1, MDM2, Rnf111), whereas several peptidases
   (Nedp1/Senp8, Uchl3, Csn5) ensure dynamics of this PTM^[69]4. Nedd8
   knock-out mice, as well as mice lacking Nae1 or Ubc12, die during
   embryogenesis, establishing neddylation as an essential PTM^[70]4.

   As with other Ubls, neddylation has mainly been studied in highly
   proliferating cells, where it regulates a plethora of signaling events
   involved in the maintenance of DNA integrity or
   proteostasis^[71]26,[72]27. In contrast, its role in post-mitotic
   cells, such as neurons, remains enigmatic. Several functional
   perturbation studies linked neddylation to neurite outgrowth, spine
   growth, synapse density, and synaptic transmission^[73]28–[74]33.
   However, the Nedd8 targets at the basis of these processes are unknown.
   The most abundant Nedd8 targets in all cell types are the cullins,
   components of the large multi-protein cullin-RING E3 ubiquitin
   ligases^[75]34,[76]35 that regulate a myriad of cellular processes,
   including the development and dendrite arborization of
   neurons^[77]36–[78]38. So far, only very few neuronal non-cullin Nedd8
   targets have been described^[79]39, including Cofilin, PSD95, and
   mGlu7^[80]28,[81]29,[82]32, whose neddylation is thought to partially
   account for the morphological and functional defects seen in studies on
   the perturbation of neddylation^[83]28–[84]33.

   However, a definitive and systematic understanding of the role of
   neddylation in neuronal development and synaptic function is lacking,
   not least because corresponding earlier approaches to perturb
   neddylation were either of unknown specificity (i.e., pharmacological
   inhibition of Nae1 with MLN4924), indirect (i.e., KO of Nedd8
   peptidases), or targeting late neurodevelopmental stages, making data
   interpretation and generalizations difficult^[85]28–[86]30,[87]33.

   To obtain stringent and comprehensive insight into the role of
   neddylation in neurons, we generated a conditional Nedd8 mouse line and
   analyzed the impact of Nedd8 loss on neuronal development and synaptic
   transmission. We systematically assessed the morphology and function of
   Nedd8-deficient neurons in autaptic culture, and observed a key role
   for neddylation in controlling the fate of glutamatergic neurons and
   excitatory synaptic transmission.

Results

   Using CRISPR/Cas9, we generated a conditional Nedd8 knock-out
   (Nedd8cKO) mouse line by inserting LoxP sites before the second and
   after the third Nedd8 exons (Fig. [88]1A). Western blot analysis of
   lysates from primary hippocampal neurons infected with a CRE-expressing
   virus confirmed the loss of Nedd8, as compared to neuronal cultures
   expressing red fluorescence protein (RFP) as a control (Fig. [89]1B,
   C). For all subsequent experiments, primary hippocampal neurons
   obtained from P0 Nedd8cKO pups were infected after one day in vitro
   (DIV1) with CRE-expressing virus to ablate Nedd8 expression (Nedd8-KO),
   RFP-expressing virus or no virus as control (CTRL), and were analyzed
   at DIV9–12.

Fig. 1. Nedd8 deficiency alters dendritic branching but not synapse number.

   [90]Fig. 1
   [91]Open in a new tab

   A On scale schematic view of the Nedd8 mouse gene. LoxP sites (red
   triangles, not on scale) were inserted before the second exon and after
   the third exon using a CRISPR/Cas9 strategy. B Total protein stain
   (right) Anti-Nedd8 (top left) and RFP (bottom left) Western blot
   analysis of DIV12 primary hippocampal culture lysates induced with
   virus at DIV1 as indicated on top. The line and stars indicate
   Nedd8-specific signal. Molecular weight marker is indicated on the left
   (kDa). C Bar graph showing normalized Nedd8 protein levels calculated
   from Western blot analysis as depicted in (B). Nedd8 signal (black line
   and stars) was normalized to total protein stain (N = 3, n[RFP] = 3,
   n[CRE] = 3). Bars represent mean ± SEM. Data were compared using a
   one-tailed Mann-Whitney comparison test, where P = 0.05. D DIV13
   Nedd8cKO autaptic hippocampal neurons infected at DIV1 with lentivirus
   expressing RFP or CRE-recombinase-RFP (CRE) were fixed and
   immunostained for MAP2. Neurons without lentiviral infection were used
   as infection controls (CTRL). Images were acquired using a confocal
   microscope, and representative binary images of MAP2-immunostained
   neurons are shown. Scale bar = 20 μm. E Line plots showing the number
   of dendritic intersections as a function of distance from the soma
   obtained by Sholl analysis (N = 3; n[CTRL] = 22, n[RFP] = 26,
   n[CRE] = 20 cells). F Bar graph showing the number of crossing
   dendrites between 50 and 100 μm from the soma, calculated from Sholl
   analysis (N = 3; n[CTRL] = 22, n[RFP] = 26, n[CRE] = 20 cells). Data
   was compared using Tukey’s multiple comparisons test, where
   P[CTRL-RFP] = 0.72, P[CTRL-CRE] = 0.14, P[RFP-CRE] = 0.020. G Bar
   graphs showing the total number of crossing dendrites, calculated from
   Sholl analysis (N = 3; n[CTRL] = 22, n[RFP] = 26, n[CRE] = 20 cells).
   Data was compared using Dunn’s multiple comparisons test, where
   P[CTRL-RFP] > 0.99, P[CTRL-CRE] = 0.53, P[RFP-CRE] = 0.21. H Bar graph
   showing the dendrite length sum (N = 3; n[CTRL] = 22, n[RFP] = 26,
   n[CRE] = 20 cells). Data was compared using Dunn’s multiple comparisons
   test, where P[CTRL-RFP] > 0.99, P[CTRL-CRE] = 0.35, P[RFP-CRE] = 0.11.
   I Immunostaining of autaptic hippocampal neurons with antibodies
   against MAP2 (cyan), synapsin1 (green), and PSD95 (magenta). Scale
   bar = 50 μm. The white square indicates a closer view depicted in the
   right column of panels. Scale bar = 10 μm. J Bar graph showing the
   total number of synapsin1 (presynaptic) counts. (N = 3; n[CTRL] = 28,
   n[RFP] = 31, n[CRE] = 27 cells). Data was compared using Dunn’s
   multiple comparisons test, where P[CTRL-RFP] = 0.37,
   P[CTRL-CRE] = 0.063, P[RFP-CRE] > 0.99. K Bar graph showing the total
   number of PSD95 (postsynaptic) counts (N = 3; n[CTRL] = 28,
   n[RFP] = 31, n[CRE] = 27 cells). Data was compared using Dunn’s
   multiple comparisons test, where P[CTRL-RFP] > 0.99,
   P[CTRL-CRE] > 0.99, P[RFP-CRE] = 0.65. L Bar graph showing the total
   number of synapsin1counts colocalized with PSD95 signal (N = 3;
   n[CTRL] = 28, n[RFP] = 31, n[CRE] = 27 cells). Data was compared using
   Dunn’s multiple comparisons test, where P[CTRL-RFP] = 0.73,
   P[CTRL-CRE] = 0.94, P[RFP-CRE] > 0.99. M Bar graphs showing the total
   number of PSD95 counts colocalized with synapsin1 (N = 3; n[CTRL] = 28,
   n[RFP] = 31, n[CRE] = 27 cells). Data was compared using Dunn’s
   multiple comparisons test, where P[CTRL-RFP] = 0.77,
   P[CTRL-CRE] > 0.99, P[RFP-CRE] > 0.99. In graph (J–M), dots correspond
   to the total number of synaptic counts obtained from one individual
   neuron. For all graphs, error bars represent mean ± SEM. Data were
   compared using a Mann-Whitney, Kruskal-Wallis, Dunn´s multiple
   comparison test, and a Tukey’s multiple comparisons test, where *
   p < 0.05.

Nedd8-KO neurons show morphological defects but no change in synapse number

   Many studies described defects in neuronal development, morphology and
   synapse number upon neddylation blockade^[92]28,[93]32,[94]33. Thus, we
   analyzed these parameters in Nedd8-deficient neurons as compared to RFP
   or non-infected control neurons (Fig. [95]1D–M). Sholl analysis
   revealed that the complexity of the neuronal dendritic tree was
   slightly altered upon Nedd8-deficiency, with a minor, but significant,
   reduction in the number of crossing dendrites within 50–100 μM from the
   soma (Fig. [96]1D–F). However, the total number of crossing dendrites
   and the total dendrite length were not reduced (Fig. [97]1G, H),
   indicating a moderate, but significant, impact of Nedd8 depletion on
   neuronal morphology.

   Next, we quantified the number of excitatory synapses via
   immunolabeling using Synpasin1 as a pre-synaptic marker and PSD95 to
   label excitatory post-synapses (Fig. [98]1I–M); MAP2 was used to
   visualize overall neuronal processes. The number of pre- and
   post-synaptic puncta was similar in Nedd8-KO as compared to the
   controls, e.g., RFP or non-infected CTRL neurons (Fig. [99]1J, K),
   leading to a similar number of pre- and post-synaptic colocalized
   puncta (Fig. [100]1L, M). Taken together, our data indicate that loss
   of Nedd8 expression alters neuronal morphology only moderately, without
   any impact on the number of synapses.

Altered synaptic transmission in Nedd8-KO neurons

   Next, we examined how Nedd8 depletion affected glutamatergic
   neurotransmission in autaptic hippocampal neurons (Figs. [101]2 and
   [102]3). While Nedd8-deficient autaptic neurons displayed no change in
   evoked excitatory post-synaptic currents (EPSCs) as compared to RFP or
   non-virus infected neurons (CTRL, Fig. [103]2A), CRE-infected neurons
   showed a 30% decrease in the readily-releasable pool (RRP) of synaptic
   vesicles (SV) as assessed by hypertonic sucrose stimulation
   (Fig. [104]2B), accompanied with a significant increase in synaptic
   transmitter release probability (P[vr], Fig. [105]2C). Importantly,
   there was no change in spontaneous miniature EPSC (mEPSC) amplitude
   (Fig. [106]2D), while a non-significant trend towards a reduction in
   mEPSC frequency was observed (Fig. [107]2E). Importantly, currents
   induced by fast superfusion of neurons with glutamate or GABA were not
   different between Nedd8-KO and the CTRL and RFP controls (Fig. [108]2F,
   G), indicating no differences in the surface expression of functional
   glutamate receptors. Thus, Nedd8 depletion does not profoundly alter
   post-synaptic function in a profound fashion.

Fig. 2. Nedd8 knock-down alters synaptic release probability.

   [109]Fig. 2
   [110]Open in a new tab

   Autaptic excitatory hippocampal neurons from Nedd8cKO animals were
   infected at DIV1 with lentivirus expressing RFP or CRE-recombinase-RFP
   (CRE) and analyzed at DIV11-13. Neurons without lentiviral infection
   were used as infection controls (CTRL). A Representative traces (left)
   and bar graph (right) of evoked EPSC amplitudes of autaptic neurons
   infected as indicated below the graphs (N = 5; n[CTRL] = 41,
   n[RFP] = 47, n[CRE] = 47 cells). Data was compared using Dunn’s
   multiple comparisons test, where P[CTRL-RFP] = 0.22,
   P[CTRL-CRE] > 0.99, P[RFP-CRE] > 0.18. B Bar graph depicting the total
   charge transferred by the release of the RRP obtained after treatment
   with hypertonic sucrose solution (N = 5; n[CTRL] = 36, n[RFP] = 44,
   n[CRE] = 42 cells). Data was compared using a Dunn’s multiple
   comparisons test, where P[CTRL-RFP] = 0.10, P[CTRL-CRE] > 0.99,
   P[RFP-CRE] = 0.011. C Bar graph showing the probability of vesicular
   release (P[vr]) of individual neurons (N = 5; n[CTRL] = 36,
   n[RFP] = 44, n[CRE] = 42 cells). P[vr] is calculated by dividing the
   charge transfer during an evoked EPSC by the charge transfer during the
   sucrose response, and then expressed as a percentage. Data was compared
   using a Tukey’s multiple comparisons test, where P[CTRL-RFP] = 0.99,
   P[CTRL-CRE] = 0.0055, P[RFP-CRE] = 0.0047. D Representative traces
   (left) and bar graph (right) depicting spontaneous mEPSC amplitudes of
   autaptic neurons treated with 300 nM TTX (N = 5; n[CTRL] = 33,
   n[RFP] = 43, n[CRE] = 39 cells). Data was compared using a Dunn’s
   multiple comparisons test, where P[CTRL-RFP] = 0.33,
   P[CTRL-CRE] > 0.99, P[RFP-CRE] = 0.34. E Bar graph showing the
   frequency of mEPSC measured in autaptic neurons in the presence of
   300 nM TTX (N = 5; n[CTRL] = 33, n[RFP] = 43, n[CRE] = 39 cells). Data
   was compared using a Dunn’s multiple comparisons test, where
   P[CTRL-RFP] = 0.99, P[CTRL-CRE] = 0.86, P[RFP-CRE] = 0.091. F Bar graph
   showing the amplitude of the peak current generated by the superfusion
   of 100 μM Glutamate (N = 5; n[CTRL] = 20, n[RFP] = 34, n[CRE] = 29
   cells). Data was compared using a Dunn’s multiple comparisons test,
   where P[CTRL-RFP] > 0.99, P[CTRL-CRE] > 0.99, P[RFP-CRE] > 0.99. G Bar
   graph showing the amplitude of the peak current generated by the
   superfusion of 3 μM GABA (N = 5; n[CTRL] = 21, n[RFP] = 32, n[CRE] = 28
   cells). Data was compared using a Dunn’s multiple comparisons test,
   where P[CTRL-RFP] = 0.053, P[CTRL-CRE] = 0.62, P[RFP-CRE] = 0.72. H
   Line plots showing the change in EPSC amplitudes during a 10 Hz
   stimulation train (N = 5; n[CTRL] = 31, n[RFP] = 41, n[CRE] = 37
   cells). Data were normalized to the first response of the respective
   train and shown as mean +/− SEM. I Line plots showing the change in
   EPSC amplitudes during a 40 Hz stimulation train (N = 5; n[CTRL] = 29,
   n[RFP] = 42, n[CRE] = 37 cells). Data were normalized to the first
   response of the respective train and shown as mean +/– SEM. J Bar graph
   showing paired-pulse ratios obtained from 10 Hz stimulation trains
   (N = 5; n[CTRL] = 31, n[RFP] = 41, n[CRE] = 37 cells). Data was
   compared using a Dunn’s multiple comparisons test, where
   P[CTRL-RFP] = 0.088, P[CTRL-CRE] > 0.99, P[RFP-CRE] = 0.013. K Bar
   graph showing paired-pulse ratios obtained from 40 Hz stimulation
   trains (N = 5; n[CTRL] = 29, n[RFP] = 42, n[CRE] = 37 cells). Data was
   compared using a Dunn’s multiple comparisons test, where
   P[CTRL-RFP] = 0.23, P[CTRL-CRE] = 0.47, P[RFP-CRE] = 0.0017. L Bar
   graphs showing the RRP size as estimated by back extrapolation of the
   cumulative EPSC after 40 Hz stimulus trains (N = 5; n[CTRL] = 28,
   n[RFP] = 40, n[CRE] = 35 cells). Data was compared using a Dunn’s
   multiple comparisons test, where P[CTRL-RFP] > 0.99,
   P[CTRL-CRE] = 0.48, P[RFP-CRE] = 0.039. M Bar graphs showing the slope
   of the back extrapolated cumulative EPSC curve of the 40 Hz trains
   (N = 5; n[CTRL] = 28, n[RFP] = 39, n[CRE] = 35 cells). Data was
   compared using a Dunn’s multiple comparisons test, where
   P[CTRL-RFP] > 0.99, P[CTRL-CRE] = 0.65, P[RFP-CRE] = 0.12. N Bar graph
   showing the ratio between the slope of the back extrapolated cumulative
   EPSC curve (shown in M), divided by the RRP[40Hz] (shown in L), (N = 5,
   n[ctrl] = 32, n[rfp] = 39, n[cre] = 34 cells). Data was compared using
   a Dunn’s multiple comparisons test, where P[CTRL-RFP] > 0.99,
   P[CTRL-CRE] > 0.99, P[RFP-CRE] > 0.99. In each graph, dots correspond
   to measurements recorded from individual neurons and bars represent
   mean ± SEM. Data were compared using a Kruskal-Wallis, Tukey’s, Dunn´s
   multiple comparison test, where *p < 0.05, **p < 0.01.

Fig. 3. VGCC currents are unchanged in Nedd8-KO neurons.

   [111]Fig. 3
   [112]Open in a new tab

   Analysis of calcium currents obtained from DIV9-13 Nedd8cKO autaptic
   excitatory hippocampal neurons infected at DIV1 with lentivirus
   expressing RFP, CRE-recombinase-RFP (CRE) or non-infected neurons
   (CTRL). A Current-voltage relationship of HVA calcium channels (N = 3;
   n[CTRL] = 23, n[RFP] = 21, n[CRE] = 23 cells). B Bar graph showing the
   change in evoked EPSC amplitudes after treatment with 0.2 µM ω-Agatoxin
   (Aga) (N = 2; n[CTRL] = 14, n[RFP] = 15, n[CRE] = 16 cells). Data was
   compared using a Dunn’s multiple comparisons test, where
   P[CTRL-Aga] > 0.99, P[RFP-Aga] > 0.99, P[CRE-Aga] > 0.99. C Bar graph
   showing the inhibition ratio of EPSC amplitudes after treatment with
   0.2 µM ω-Agatoxin (N = 2; n[CTRL] = 14, n[RFP] = 15, n[CRE] = 16
   cells). Data was compared using a Dunn’s multiple comparisons test,
   where P[CTRL-Aga] = 0.60, P[RFP-Aga] = 0.90, P[CRE-Aga] > 0.99. D Bar
   graph depicting the total charge transferred by the release of the RRP
   upon application with hypertonic sucrose solution after treatment with
   0.2 µM ω-Agatoxin (Aga) (N = 2; n[CTRL] = 14, n[RFP] = 15, n[CRE] = 16
   cells). Data was compared using a Dunn’s multiple comparisons test,
   where P[CTRL-Aga] > 0.99, P[RFP-Aga] > 0.99, P[CRE-Aga] > 0.99. E Bar
   graph showing the probability of vesicular release of individual neuron
   after treatment with 0.2 µM ω-Agatoxin (Aga) (N = 2; n[CTRL] = 14,
   n[RFP] = 14, n[CRE] = 16 cells). Data was compared using a Dunn’s
   multiple comparisons test, where P[CTRL-Aga] = 0.69, P[RFP-Aga] > 0.99,
   P[CRE-Aga] > 0.99. F Bar graph showing the change in EPSC amplitudes
   after treatment with 10 µM Nimodipine (Nimo) (N = 2; n[CTRL] = 16,
   n[RFP] = 15, n[CRE] = 16 cells). Data was compared using a Tukey’s
   multiple comparison test, where P[CTRL-Nimo] = 0.36,
   P[RFP-Nimo] = 0.58, P[CRE-Nimo] = 0.99. G Bar graph showing the
   inhibition ratio of EPSC amplitudes after treatment with 10 µM
   Nimodipine (N = 2; n[CTRL] = 16, n[RFP] = 15, n[CRE] = 16 cells). Data
   was compared using a Tukey’s multiple comparisons tests, where
   P[CTRL-RFP] = 0.61, P[CTRL-CRE] = 0.42, P[RFP-CRE] = 0.083. H Bar graph
   of the total charge transferred by the release of the RRP upon
   application with hypertonic sucrose solution after treatment with 10 µM
   Nimodipine (Nimo) (N = 2; n[CTRL] = 16, n[RFP] = 15, n[CRE] = 16
   cells). Data was compared using a Dunn’s multiple comparisons test,
   where P[CTRL-Nimo] > 0.99, P[RFP-Nimo] > 0.99, P[CRE-Nimo] > 0.99. I
   Bar graph showing the probability of vesicular release of individual
   neurons after treatment with 10 µM Nimodipine (Nimo) (N = 2;
   n[CTRL] = 16, n[RFP] = 15, n[CRE] = 16 cells). Data was compared using
   a Tukey’s multiple comparison test, where P[CTRL-Nimo] = 0.99,
   P[RFP-Nimo] = 0.82, P[CRE-Nimo] = 0.95. In each graph, dots correspond
   to measurements recorded from individual neurons and bars represent
   mean ± SEM. Data were compared using an unpaired t test, Mann-Whitney
   test, a Kruskal-Wallis and Dunn´s multiple comparison test.

   We next studied synaptic short-term plasticity by measuring the
   synaptic response to high-frequency stimulation (HFS) (Fig. [113]2H–N).
   We found that the EPSC amplitude of CRE-infected neurons depressed
   faster than that of RFP or non-infected CTRL neurons (Fig. [114]2H, I).
   In addition, and in accord with the decreased RRP size as assessed by
   hypertonic sucrose stimulation (Fig. [115]2B), the paired pulse ratio
   (Fig. [116]2J, K) and the RRP size (RRP[40Hz]) as estimated by back
   extrapolation of the cumulative EPSC during 40 Hz stimulus trains were
   decreased in Nedd8-deficient neurons as compared to RFP or non-infected
   CTRL neurons (Fig. [117]2L). Finally, the refiling rate of the RRP –
   i.e., the slope of the back extrapolated cumulative EPSC curve of the
   40 Hz trains – showed a trend towards reduction that was not
   significant (Fig. [118]2M). When we normalized the RRP refiling rate to
   the RRP[40Hz,] no difference between the tested group were detectable
   (Fig. [119]2N). Thus, the more pronounced short-term depression
   observed in Nedd8-deficient neurons (Fig. [120]2H, I) is primarily
   attributable to a reduction in RRP size and increase P[vr], rather than
   to an aberrant activity-dependent RRP refiling process.

Normal VGCC activity but reduced SV-VGCC coupling in Nedd8-KO neurons

   Regarding the changes in short-term plasticity observed in
   Nedd8-deficient neurons, we next examined the possibility of
   alterations in the properties of voltage-gated Ca^2+ channels (VGCCs).
   Given the serious challenges associated with studying VGCCs in the
   small presynaptic terminals of cultured hippocampal neurons, we
   investigated the VGCC properties in the soma of proximal dendrites.
   Initially, we assessed the overall density and voltage-dependence of
   VGCCs by measuring the current-voltage relationships, and found that
   Nedd8-loss had no discernible effect (Fig. [121]3A). We expanded our
   analyses by employing VGCC subtype-specific blockers to determine
   whether the influence of different presynaptic VGCCs in the presynaptic
   terminal on EPSC amplitude, RRP size, and P[vr] is altered by
   Nedd8-loss, by administering ω-Agatoxin and nimodipine, which
   specifically inhibit P/Q-type and L-type VGCCs, respectively
   (Fig. [122]3B–I). We found no significant differences between the
   groups tested, indicating that the altered RRP and P[vr] in Nedd8-KO
   neurons cannot be attributed to modifications in VGCCs properties.

   In light of the results outlined above, particularly the P[vr] increase
   without changes in VGCC function, we next explored the possibility that
   Nedd8-loss selectively affects the priming and fusion of SVs that are
   more readily available for synaptic transmitter release due to
   VGCC-coupling. We performed electrophysiological recordings in the
   presence of EGTA-AM, a slow Ca^2+ chelator used to perturb VGCC
   coupling to SV Ca^2+ sensor^[123]40,[124]41 (Fig. [125]4A). EGTA-AM
   reduced the P[vr] to a similar extent in Nedd8-KO neurons as compared
   to RFP infected neurons (Fig. [126]4B–D). Strikingly, however, EGTA-AM
   reversed synaptic depression during trains of action potentials in
   RFP-infected neurons (Fig. [127]4E, [128]G) but not in CRE-expressing
   Nedd8-deficient cells (Fig. [129]4F, [130]H). These data are compatible
   with the notion that a pool of ready-releasable SVs present near VGCCs
   is altered in Nedd8-deficient neurons (Fig. [131]4A).

Fig. 4. Nedd8 regulates the number of synaptic vesicles coupled to VGCC.

   [132]Fig. 4
   [133]Open in a new tab

   Recordings were performed using DIV9-13 Nedd8cKO autaptic excitatory
   hippocampal neurons infected at DIV1 with lentivirus expressing RFP,
   CRE-recombinase-RFP (CRE) or non-infected neurons (CTRL). A Schematic
   representation of the effect of EGTA-AM on synaptic vesicles coupled to
   calcium channels. The size of proteins and synaptic vesicles is not on
   scale. This image was generated with the help of Biorender. (i)
   Following an action potential, voltage-gated calcium channels (VGCC)
   open and Ca^2+ ions access the intracellular compartment. The red cloud
   indicates the area and number of vesicles impacted by the entry of
   calcium. (ii) After treatment with the permeable ester, EGTA-AM,
   hydrolyzation of intracellular esterases leads to the production of
   EGTA, which lowers the intracellular calcium concentration (reduction
   in the size of the red cloud), except in the vicinity of VGCC, leading
   to a decrease in the number of released vesicles. Ultimately, EPSC
   amplitude and P[vr] are reduced, but not the size of the RRP, as
   measured by hypertonic sucrose treatment. (iii) During a high-frequency
   train of action potentials (HFS), the effect of EGTA is restored due to
   the accumulation of calcium ions in the intracellular compartment. (iv)
   Working model: Nedd8-deficient neurons present fewer synaptic vesicles
   capable of being released. B Bar graph showing EPSC amplitudes, with
   and without treatment with 100 µM EGTA-AM (N = 3; n[RFP] = 14,
   n[RFP+EGTA-AM] = 16, n[CRE] = 20, n[CRE+EGTA-AM] = 15 cells). Data was
   compared using a Mann-Whitney test, where P[RFP-CRE] = 0.53. C Bar
   graph of the total charge transferred by the release of the RRP upon
   application of hypertonic sucrose solution, with and without treatment
   with 100 µM EGTA-AM (N = 3; n[RFP] = 14, n[RFP+EGTA-AM] = 17,
   n[CRE] = 19, n[CRE+EGTA-AM] = 15 cells). Data was compared using a
   Mann-Whitney test, where P[RFP-CRE] = 0.24. D Bar graph showing the
   impact of EGTA-AM treatment on the probability of vesicular release of
   individual neuron (N = 3; n[RFP] = 14, n[RFP+EGTA-AM] = 17,
   n[CRE] = 20, n[CRE+EGTA-AM] = 15 cells). Data was compared using an
   unpaired t test, where P[RFP-CRE] = 0.24. E Line plots showing the
   change in EPSC amplitudes during a 10 Hz stimulation train in neurons
   infected with RFP virus, with and without application of 100 µM EGTA-AM
   (N = 3; n[RFP] = 15, n[RFP+EGTA-AM] = 16 cells). F Line plots showing
   the change in EPSC amplitudes during a 10 Hz stimulation train in
   neurons infected with CRE virus, with and without application of 100 µM
   EGTA-AM (N = 3; n[CRE] = 19, n[CRE + EGTA-AM] = 14 cells). G Line plots
   showing the change in EPSC amplitudes during a 40 Hz stimulation train
   in neurons infected with RFP virus, with and without application of
   100 µM EGTA-AM (N = 3; n[RFP] = 15, n[RFP+EGTA-AM] = 16 cells). H Line
   plots showing the change in EPSC amplitudes during a 40 Hz stimulation
   train in neurons infected with CRE virus, with and without application
   of 100 µM EGTA-AM (N = 3; n[CRE] = 19, n[CRE+EGTA-AM] = 14 cells). In
   each graph, dots correspond to measurements recorded from individual
   neurons and bars represent mean ± SEM. Data were compared using an
   unpaired t test and a Mann-Whitney test.

Normal ultrastructural organization of Nedd8-KO synapses at rest

   We conducted a 3D ultrastructural analysis of synapses in cultured
   neurons to investigate possible ultrastructural correlates of the
   presynaptic functional deficits observed in Nedd8-KO neurons. We
   performed dual-axis TEM tomography to quantify the size and
   distribution of SV  pools in presynaptic terminals (Supplementary
   Fig. [134]1). Tomographic analyses of high-pressure frozen and
   freeze-substituted primary hippocampal neurons infected with either RFP
   as a control or with CRE to deplete Nedd8 revealed no major
   morphological differences (Supplementary Fig. [135]1A–L). The
   cumulative frequency distribution (Supplementary Fig. [136]1E) and
   total number of SVs within 0–200 nm of the AZ (Supplementary
   Fig. [137]1F) were very similar between conditions. Likewise, an
   analysis of nearest-neighbor distances for all SVs within 0–200 nm of
   the AZ did not reveal any statistically significant differences in SV
   cluster characteristics between conditions (Supplementary
   Fig. [138]1G). Morphological analyses were subsequently restricted to
   vesicles localized in close proximity to the AZ membrane (0–40 nm from
   the AZ, Supplementary Fig. [139]1H–J). No statistically significant
   differences were observed in either the spatial distribution of
   vesicles (Supplementary Fig. [140]1H), mean AZ areas (Supplementary
   Fig. [141]1H, Insert), density of docked vesicles per unit AZ
   (Supplementary Fig. [142]1I), or inter-vesicular nearest neighbor
   distances (Supplementary Fig. [143]1I, J). Finally, SV diameters were
   not distinguishable between Nedd8-deficient and RFP control neurons
   (Supplementary Fig. [144]1K, L). No difference in the number of
   morphologically docked SVs, which serve as a proxy for the number of
   molecularly primed SVs comprising the RRP^[145]42, was observed between
   genotypes. Altogether, Nedd8-deficient synapses do not display any
   observable alteration in organization at AZ release sites at rest. This
   indicates that the functional defects described above manifest without
   incurring structural changes detectable by our methodology and
   originate from deeper and molecular Nedd8-dependent mechanisms.

Altered expression of genes encoding key pre-synaptic proteins in Nedd8-KO
neurons

   In a first attempt to explore the molecular pathways by which
   Nedd8-loss causes the functional phenotypes we observed, we performed
   RNA-sequencing and differential expression analysis of CRE-expressing
   neurons vs. RFP-expressing neurons (Fig. [146]5A, Supplementary
   Fig. [147]2–[148]4 and Supplementary Data [149]2). The underlying
   motivation was to not only determine direct transcriptional changes
   upon Nedd8-loss, but also to assess indirect compensatory mechanisms.
   Beyond a general analysis, we put a special focus on genes encoding
   synaptic proteins by using the SynGO database^[150]43.

Fig. 5. Nedd8 depletion alters the expression of synaptic proteins.

   [151]Fig. 5
   [152]Open in a new tab

   A RNA was extracted from DIV13 Nedd8cKO primary hippocampal neurons
   infected at DIV1 with lentivirus expressing RFP or CRE-recombinase-RFP
   (CRE) and sequenced. Each dot represents a single gene (blue dots).
   Volcano plot depicts each gene based on their Log2 fold change (LogFC)
   and the reverse log10 of their p-value (-log10p). Red dots correspond
   to genes with a LogFC above 1,3 and -log10p above 6. Green dots
   correspond to genes found in SynGO. B, C Bar graphs showing the raw
   gene expression levels of Slc17a7 (vGlut1, B) and Slc17a6 (vGlut2, C)
   (N = 3, n[RFP] = 11, n[CRE] = 12). Data was compared using an unpaired
   t test, where P[Slc17a7] < 0.0001, and a Mann-Whitney test, where
   P[Slc17a6] = 0.011. D, E Bar graphs showing the normalized mRNA
   expression levels of Slc17a7 (vGlut1, D) and Slc17a6 (vGlut2, E)
   (N = 3, n[RFP] = 11, n[CRE] = 11). Data was compared using a
   Mann-Whitney test, where P[Slc17a7] = 0.0006, P[Slc17a6] = 0.0010. F, G
   Bar graph depicting normalized protein levels of vGlut1 and vGlut2, as
   assessed by Western blot showed in (H). Data was compared using a
   Mann-Whitney test, where P[vGlut1] = 0.0012, P[vGlut2] = 0.013. H Total
   protein stain (left panel), and anti-vGlut1 (middle panel) and vGlut2
   (right panel) Western blot analysis of primary hippocampal Nedd8cKO
   neurons lysates infected with lentivirus expressing RFP or
   CRE-recombinase-RFP (CRE). Molecular weight is indicated on the left
   side (kDa). For all bar graphs, data represent mean ± SEM. Data were
   compared using an unpaired t test or a Mann-Whitney comparison test,
   where *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

   Upon stringent filtering, the expression of 566 genes remained
   significantly altered in Nedd8-deficient neurons (Fig. [153]5A; Log2FC
   of 1.3; -log10 p-value of 6), with 77% (436) up- and 23% (130)
   down-regulated. Among these were 66 SynGO genes that remained
   significantly changed after filtering. Interestingly, half of these
   were up- and the other half down-regulated (Fig. [154]5A), which
   strongly contrasts with the remodeling of the rest of the transcriptome
   induced by Nedd8-deficiency, indicating that Nedd8-loss causes a
   general increase in gene expression but a more pronounced
   downregulation of genes expressing synaptic proteins.

   SynGO pathway analysis showed that Nedd8-deletion leads to decreased
   expression of multiple mRNAs encoding pre-synaptic components, in
   particular proteins involved in the SV cycle (Supplementary
   Fig. [155]3) and to the up-regulation of genes related to the
   extracellular matrix or protein translation (Supplementary
   Fig. [156]4). Strikingly, vGlut1 (Slc17a7) was strongly down-regulated
   upon Nedd8-loss, while vGlut2 was up-regulated (Slc17a6,
   Fig. [157]5A–C). This result was further validated via RT-qPCR and at
   the protein level (Fig. [158]5D–H). Given the known correlation between
   P[vr] and the resident isoform of vGlut expressed, with
   vGlut2-expressing neurons exhibiting higher P[vr]^[159]44,[160]45, we
   further studied this phenomenon.

Reduced density of vGlut1-containing synapses in Nedd8-KO neurons

   We next compared vGlut1 immunoreactivity in Nedd8-deficient neurons
   compared to RFP-expressing controls (Fig. [161]6). Whereas the total
   number of synapses was unchanged (Fig [162]1I–M), the total number of
   vGlut1 puncta was significantly reduced in Nedd8-deficient neurons
   (Fig. [163]6A–C). Importantly, vGlut1 signal intensity was not altered
   in vGlut1-containing mutant synapses (Fig. [164]6D, [165]E). STED
   imaging confirmed that the intensity of vGlut1 puncta along dendrites
   was unchanged and showed further that the sub-synaptic distribution of
   vGlut1 is similar between Nedd8-deficient and control neurons
   (Fig. [166]6F–H). Our data indicate that the number of synapses
   containing robust levels of vGlut1 is decreased in Nedd8-KO neurons,
   whereas the synaptic enrichment of vGlut1 is unchanged in synapses that
   still contain vGlut1.

Fig. 6. vGlut1 levels are altered upon Nedd8 depletion.

   [167]Fig. 6
   [168]Open in a new tab

   DIV13 Nedd8cKO autaptic hippocampal neurons infected at DIV1 with
   lentivirus expressing RFP or CRE-recombinase-RFP (CRE) were fixed and
   immunostained with antibodies as indicated. Neurons without lentiviral
   infection were used as infection controls (CTRL). A Representative
   confocal images of autaptic neurons stained with antibodies against
   MAP2 and vGlut1. Scale bars = 80 (top panels) and 10 μm (bottom
   panels). B Bar graph showing the total number of vGlut1 counts (N = 2;
   n[RFP] = 17, n[CRE] = 19 cells). Data was compared using a Mann-Whitney
   test, where P = 0.0020. C Bar graph showing the number of somatic
   vGlut1 counts (N = 2; n[RFP] = 17, n[CRE] = 19 cells). Data was
   compared using an unpaired t test, where P = 0.026. D Bar graph showing
   the total vGlut1 intensity mean (N = 2; n[RFP] = 17, n[CRE] = 19
   cells). E Bar graph showing the somatic vGlut1 intensity mean (N = 2;
   n[RFP] = 17, n[CRE] = 19 cells). F Representative confocal (left column
   of panels) and STED images of autaptic hippocampal neurons labeled with
   nanobodies against vGlut1 (green) and PSD95 (magenta). Scale
   bars = 10 μm (left column of panels) and 5 μm (other panels). G Line
   plots showing the change in vGlut1 signal intensity against distance
   (N = 3; n[CTRL] = 13, n[RFP] = 10, n[CRE] = 13 cells). H Line plots
   showing the change in PSD95 signal intensity against distance (N = 3;
   n[CTRL] = 13, n[RFP] = 10, n[CRE] = 13 cells). AU arbitrary unit. In
   each graph, dots correspond to the total number of synaptic counts
   obtained from one individual neuron. Bars represent mean ± SEM. Data
   were compared using a Kruskal-Wallis, Dunn´s multiple comparison test,
   unpaired t test, or Mann-Whitney test, where *p < 0.05, **p < 0.01.

P[vr] defects in Nedd8-KO neurons are not reverted by vGlut2 knock-down or
vGlut1 overexpression

   Our data show that vGlut2 expression is up-regulated by 50% and vGlut1
   down-regulated by 60–70% upon Nedd8-loss (Fig. [169]5). This finding
   could indicate a delayed differentiation of glutamatergic neurons,
   which switch from vGlut2 to vGlut1 expression as they mature, or a
   complex scenario of vGlut1 depletion compensated by vGlut2
   expression^[170]46. Previous studies showed that the vGlut1:vGlut2
   ratio co-determines synaptic transmitter release probability, with
   vGlut2-expressing neurons exhibiting a higher release
   probability^[171]45. This observation could (partly) explain the
   release probability phenotype of Nedd8-KO neurons (Fig. [172]2). To
   test this possibility, we performed rescue experiments and either
   re-expressed vGlut1 or knocked-down vGlut2 (Supplementary Fig. [173]5A)
   in the Nedd8-KO background in order to restore a normal vGlut1:vGlut2
   ratio, and studied the impact on the RRP and P[vr]. However, neither
   overexpression of vGlut1 nor knockdown of vGlut2 rescued the RRP and
   P[vr] phenotype of Nedd8-KO neurons (Fig. [174]7A–F). Crucially,
   endophilin1 levels were also shown to correlate with the reduced
   release probability associated with vGlut1 expression in
   neurons^[175]45. Strikingly, Nedd8-KO causes a 50% reduction in
   endophilin1 expression that is detectable at both mRNA and protein
   levels (Fig. [176]7G–L). Moreover, neither re-expression of vGlut1 nor
   knockdown of vGlut2 rescued endophilin1 levels in Nedd8-KO cells
   (Fig. [177]7I–L). Thus, it is possible that restoration of vGlut1 or
   vGlut2 levels failed to rescue synaptic defects in Nedd8-KO neurons
   because of reduced endophilin1 level, which cannot be compensated by
   changing vGlut1:vGlut2 ratios.

Fig. 7. Over expression of vGlut1, or knock-down of vGlut2 do not rescue the
defects in release probability.

   [178]Fig. 7
   [179]Open in a new tab

   Hippocampal neurons from Nedd8cKO animals were infected at DIV1 with
   lentivirus expressing RFP, CRE-recombinase-RFP (CRE), co-infection with
   CRE and a virus expressing vGlut1 (+ vGLUT1), co-infection with CRE and
   a virus expressing shRNA to knock down the expression of vGlut2 (vGLUT2
   KD), or co-infection with CRE and a virus expressing scrambled RNA
   sequences (Scr). Neurons were analyzed at DIV12. A Bar graph showing
   evoked EPSC amplitudes of autaptic excitatory neurons infected with
   lentivirus expressing RFP, CRE-recombinase-RFP (CRE), or
   CRE-recombinase-RFP (CRE) and a virus re-expressing vGlut1 (+ vGlut1)
   (N = 2, n[RFP] = 32, n[CRE] = 26, n[vGlut1] = 27). Data was compared
   using a Dunn’s multiple comparisons test, where P[RFP-CRE] = 0.91,
   P[RFP-vGlut1] = 0.33, P[CRE-vGlut1] > 0.99. B Bar graph showing the
   total charge transferred by the release of the RRP upon application of
   a hypertonic sucrose solution in autaptic excitatory neurons infected
   with lentivirus expressing RFP, CRE-recombinase-RFP (CRE), or
   CRE-recombinase-RFP (CRE) and a virus re-rexpressing vGlut1 (+ vGlut1)
   (N = 2, n[RFP] = 32, n[CRE] = 27, n[vGlut1] = 28). Data was compared
   using a Dunn’s multiple comparisons test, where P[RFP-CRE] = 0.23,
   P[RFP-vGlut1] = 0.13, P[CRE-vGlut1] > 0.99. C Bar graph showing the
   probability of vesicular release of autaptic excitatory neurons
   infected with lentivirus expressing RFP, CRE-recombinase-RFP (CRE), or
   CRE-recombinase-RFP (CRE) and a virus re-rexpressing vGlut1 (+ vGlut1)
   (N = 2, n[RFP] = 32, n[CRE] = 27, n[vGlut1] = 28). Data was compared
   using a Dunn’s multiple comparisons test, where P[RFP-CRE] > 0.99,
   P[RFP-vGlut1] > 0.99, P[CRE-vGlut1] > 0.99. D Bar graph showing evoked
   EPSC amplitudes of autaptic excitatory neurons infected with lentivirus
   expressing RFP, CRE-recombinase-RFP (CRE), CRE-recombinase-RFP and a
   virus to knock down the expression of vGlut2 (vGlut2 KD) or with a
   control virus (Scr) (N = 3, n[RFP] = 15, n[CRE] = 19, n[vGlut2] = 16,
   n[scramble] = 15). Data was compared using a Tukey’s multiple
   comparisons test, where P[RFP-CRE] = 0.95, P[RFP-vGlut2] > 0.99,
   P[RFP-Scr] = 0.65, P[CRE-vGlut2] = 0.94, P[CRE-Scr] = 0.88,
   P[vGlut2-Scr] = 0.61. E Bar grap of the total charge transferred by the
   release of the RRP upon application of a hypertonic sucrose solution in
   autaptic excitatory neurons infected with lentivirus expressing RFP,
   CRE-recombinase-RFP (CRE), CRE-recombinase-RFP and a virus to
   knock-down the expression of vGlut2 (vGlut2 KD) or with a control virus
   (Scr) (N = 3, n[RFP] = 15, n[CRE] = 19, n[vGlut2] = 16,
   n[scramble] = 15). Data was compared using a Dunn’s multiple
   comparisons test, where P[RFP-CRE] > 0.99, P[RFP-vGlut2] > 0.99,
   P[RFP-Scr] > 0.99, P[CRE-vGlut2] > 0.99, P[CRE-Scr] > 0.99,
   P[vGlut2-Scr] > 0.99. F Bar graph showing the probability of vesicular
   release in autaptic excitatory neurons infected with lentivirus
   expressing RFP, CRE-recombinase-RFP (CRE), CRE-recombinase-RFP and a
   virus to knock down the expression of vGlut2 (vGlut2 KD) or with a
   control virus (Scr) (N = 3, n[RFP] = 15, n[CRE] = 19, n[vGlut2] = 16,
   n[scramble] = 16). Data was compared using a Dunn’s multiple
   comparisons test, where P[RFP-CRE] = 0.29, P[RFP-vGlut2] > 0.99,
   P[RFP-Scr] = 0.014, P[CRE-vGlut2] > 0.99, P[CRE-Scr] > 0.99,
   P[vGlut2-Scr] = 0.51. G Bar graphs showing the raw gene expression
   levels of Sh3gl2 (endophilin1, N = 3, n[RFP] = 11, n[CRE] = 11). Data
   was compared using a Mann-Whitney test, where P = 0.0002. H Normalized
   mRNA levels of Sh3gl2 (endophilin1) as assessed by RT-qPCR (N = 3,
   n[RFP] = 11, n[CRE] = 11). Data was compared using an unpaired t test,
   where P < 0.0001. I, J Bar graph (I) showing normalized expression
   levels of endophilin1 as assessed by Western blot, as shown in (J)
   (N = 2, n[CRE] = 2, n[vGlut1] = 2). Molecular weight is indicated on
   the left side (kDa). K, LBar graph (K) showing normalized expression
   levels of endophilin1 as assessed by Western blot as shown in (L)
   (N = 2, n[CRE] = 3, n[scramble] = 2, n[vGlut2] = 2). Molecular weight
   is indicated on the left side (kDa). In graphs (A–F), dots correspond
   to measurements recorded from individual neurons. Bars represent
   mean ± SEM. Data were compared using a Kruskal-Wallis and Dunn´s
   multiple comparison test, unpaired t test, or Mann-Whitney test, where
   *p < 0.05, ***p < 0.001, ****p < 0.0001.

Neddylation regulates excitatory neurodevelopment

   We show that Nedd8 ablation leads to minor morphological changes but
   strong alterations in presynaptic transmitter release probability and
   the cycling of synaptic vesicles. To explore the biological processes
   underlying these morphological and functional defects, we examined
   changes in the neuronal transcriptional landscape of mature (DIV12)
   neurons upon Nedd8 deletion (Fig. [180]5). Our transcriptome analysis
   has revealed that Nedd8-KO alters the expression of genes encoding
   synaptic proteins involved in the release and cycling of synaptic
   vesicles, consistent with the functional deficits we observed. These
   phenotypes are likely to be the culmination of a more profound, as yet
   unknown, altered Nedd8-mediated signaling earlier in neuronal
   development. The most abundant Nedd8 targets are the cullins, which are
   large multiprotein cullin-RING E3 ubiquitin ligases^[181]34,[182]35
   that regulate a variety of cellular processes, including neuronal
   development and dendritic arborization of neurons^[183]36–[184]38. We
   therefore hypothesized that an impaired degradation of key regulatory
   factors was at the origin of the transcriptional and functional changes
   we observed in Nedd8-deficient neurons. For instance, our transcriptome
   analysis of matured neurons (DIV12) revealed a strong alteration in the
   expression profile of transcriptional regulators that are essential for
   the development and maturation of the glutamatergic neuron phenotype,
   e.g., Pax6, Sox2, Neurogenin2, NeuroD1, NeuroD2, or Tbr1 (Supplementary
   Fig. [185]5B–G); changes that were further validated by RT-qPCR
   analyses (Supplementary Fig. [186]5H–M). Thus, it is likely that
   Nedd8-depletion alters, directly or indirectly, the protein expression
   levels of such regulators during the course of neuronal maturation.
   Taken together, the failure to maintain adequate protein expression
   levels of key regulators of glutamatergic cell fate in Nedd8-deficient
   neurons likely explains the aberrant expression of synaptic genes
   required for the final expression of a mature glutamatergic identity
   and the aberrant synaptic transmission.

Discussion

   PTMs are essential throughout the lifetime of a nerve cell, from early
   development and differentiation to synaptic signaling and plasticity.
   The present study focused on neddylation, an enigmatic, Ubl-based PTM
   that has recently attracted major attention due to studies linking
   neddylation to neuronal development and synaptic
   transmission^[187]28–[188]33. For optimal stringency in analyzing
   effects of perturbed neddylation, we employed a conditional Nedd8-KO
   mouse line and studied the impact of Nedd8 deletion on neuronal
   development and synaptic transmission in hippocampal autaptic neurons.
   We found that Nedd8-deficient neurons (i) have a moderately reduced
   dendrite complexity but normal numbers of synapses (Fig. [189]1) and
   normal synapse ultrastructure (Supplementary Fig. [190]1), (ii) exhibit
   an increase in synaptic release probability, partly due to altered
   coupling between SVs in the RRP and VGCCs (Figs. [191]2 and [192]4),
   accompanied by a reduced paired-pulse ratio and more profound synaptic
   depression (Fig. [193]2H–L) and (iii) show no change in the RRP
   refiling rate normalized to the RRP[40Hz] size (Fig. [194]2N) or in
   VGCC function (Fig. [195]3).

   On the other hand, Nedd8 deletion does not cause apparent postsynaptic
   defects. Some of these results, particularly the observation that
   postsynaptic features are unaffected by Nedd8-loss, are at odds with
   previous findings^[196]28,[197]30,[198]31. This is likely due to
   differences in the neuron culture models used, the methodologies
   applied, and the developmental time point at which neddylation was
   perturbed. Whereas our approach exploited a genetic Nedd8-KO strategy
   to abolish neddylation during early neurodevelopment, other studies
   were either based on pharmacological perturbation of neddylation using
   the Nae1 inhibitor MLN-4924, which has known off-target
   effects^[199]47,[200]48, or on the ablation of Nae1 at a later
   developmental stage^[201]28, thus precluding an assessment of the role
   of neddylation during earlier developmental timepoints.

   Strikingly, our transcriptome analysis revealed that Nedd8-loss leads
   to a down-regulation of many genes encoding synaptic proteins. Most
   notably, we observed a strong decrease in vGlut1 expression, but a
   concomitant increase in vGlut2 expression upon Nedd8-deletion
   (Fig. [202]5 and Supplementary Data [203]2). This is consistent with
   previous studies linking reduced vGlut1 levels to neddylation
   defects^[204]28,[205]32,[206]33, and resonates well with the
   correlation of synaptic vGlut1 and vGlut2 levels with respective lower
   and higher transmitter release probability^[207]44,[208]45. While the
   total number of synapses was not changed in Nedd8-KO neurons, the
   60–70% reduction in vGlut1 levels (Fig. [209]5) was paralleled by a
   substantial decrease in the number of vGlut1-containing synapses,
   without changes in the sub-synaptic distribution of the remaining
   vGlut1 (Fig. [210]6). The vGlut1-loss in Nedd8-KO neurons is offset by
   a 50% increase in vGlut2 levels (Fig. [211]5), which likely compensates
   for the loss of vGlut1, explaining why mEPSC amplitudes, which are
   co-determined by SV filling, are not affected by Nedd8-loss
   (Supplementary Figs. [212]1 and 2).

   In view of the fact that high synaptic vGlut2 levels are linked to
   increased P[vr]^[213]44,[214]45, we tested the hypothesis that the
   increased vGlut2-vGlut1 ratio might be at the basis of the increased
   P[vr] in Nedd8-KO neurons. This hypothesis was favored because VGCC
   function was unchanged in Nedd8-KO neurons (Fig. [215]3), and only
   slight changes in SV-VGCC coupling were observed (Fig. [216]4).
   However, neither re-expression of vGlut1 nor knock-down of vGlut2 in
   Nedd8-KO cells reverted the altered P[vr] (Fig. [217]7). We currently
   attribute this to reduced endophilin1 levels in Nedd8-KO neurons, which
   are not restored by vGlut1 re-expression or vGlut2 knock-down
   (Fig. [218]7), because the effect of vGlut1 expression levels on the
   P[vr] is known to be dependent on the presence of
   endophilin1^[219]45,[220]49. In accord with this notion, the
   endocytotic regulator endophilin1 operates in release site
   clearance^[221]50 and has been linked to RRP maintenance and the
   support of sustained exocytosis^[222]51, which are compromised in
   Nedd8-KO neurons (Fig. [223]2).

   To obtain deeper insight into the molecular mechanisms underlying
   deficits in Nedd8-KO neurons, we analyzed the neuronal transcriptome
   upon Nedd8 deletion. This approach was chosen to directly evaluate the
   impact of Nedd8-loss on the neuronal transcriptional landscape and to
   infer signaling events that can be linked to the physiological changes
   we have observed^[224]52. RNA-seq and differential expression analysis
   revealed an unexpected consequence of Nedd8 loss, characterized by a
   strong decrease in the expression of genes encoding synaptic proteins,
   especially when considering the overall upregulation of the rest of the
   transcriptome. After stringent filtering, 77% of all transcriptomic
   changes reflected an upregulation of gene expression upon Nedd8-loss
   (Fig. [225]5A). The subset of affected SynGO genes behaved very
   differently, with half upregulated and the other half downregulated,
   indicating that Nedd8-deletion strongly reduces the expression of genes
   encoding synaptic proteins, whereas the rest of the transcriptome is
   mainly upregulated.

   Gene ontology and pathway analysis showed that upregulated genes
   represent a broad range of pathways that are not necessarily
   interconnected (Supplementary Fig. [226]4). In contrast, genes
   downregulated upon Nedd8-depletion represented a more restricted
   functional spectrum, with glutamatergic synapses, glutamatergic
   signaling, neuronal development, synaptic vesicle cycle, and synaptic
   plasticity featuring prominently (Supplementary Fig. [227]3). In
   particular, we found an imbalance in the vGlut1:vGlut2 expression
   ratio, which reflects the maturation of glutamatergic neurons. As
   neuronal morphology was only moderately affected while glutamatergic
   synaptic features showed strong changes upon Nedd8-loss, it appears
   that blocking neddylation impacts glutamatergic neurons quite
   specifically rather than neurodevelopment in general. Overall, our data
   reveal a previously unknown role of neddylation in regulating the
   transcriptional landscape in neurons, with a strong influence on genes
   involved in excitatory synaptic signaling.

   Given the substantial transcriptomic changes caused by
   Nedd8-depeletion, the phenotypes we observed in Nedd8-KO neurons,
   including morphological changes and increased transmitter release
   probability, likely represent the final manifestations of a defined but
   complex Nedd8-dependent remodeling of the transcriptional landscape.
   This, in turn, is most likely linked to the function of cullins, which
   are the predominant Nedd8 substrates, part of the major class of
   cullin-RING E3 ubiquitin ligases, and involved in a broad range of
   biological processes, including cell growth, development, signal
   transduction, transcriptional control, and genome
   integrity^[228]53,[229]54.

   As mentioned above, the differentiation of glutamatergic neurons
   entails a switch from vGlut2 to vGlut1 expression as the neurons
   mature^[230]55. In view of this, the increased vGlut1:vGlut2 ratio we
   observed in glutamatergic Nedd8-KO neurons might reflect defects in the
   differentiation and maturation program of these cells. Indeed, our
   transcriptome analysis revealed elevated levels of transcription
   factors found in neural progenitors (Pax6, Sox2, Supplementary
   Fig. [231]5), and decreased levels of markers for neural commitment
   (NeuroD1, NeuroD2, NeuroG2, Trb1, Supplementary
   Fig. [232]5)^[233]11,[234]12. In general accord with the notion above,
   neddylation was reported to contribute to the regulation of
   neurogenesis and cell fate identity. For instance, Nae1 ablation in
   neuronal progenitor cells leads to severe morphological defects, likely
   due to impaired neurogenesis and neuronal differentiation^[235]56.
   Furthermore, Nae1 deletion in Schwann cells causes severe alterations
   in differentiation pathways, ultimately leading to peripheral
   neuropathies^[236]57.

   Altogether, our study shows that neddylation plays a key role in the
   differentiation and function of glutamatergic neurons. Our data
   indicate that Nedd8-ablation in young, post-mitotic, glutamatergic
   neurons causes alterations in the coordinated expression, activation,
   or homeostasis of developmental transcription factors that control
   neuronal differentiation, leading to aberrant transcriptional control
   and defects in the development of a mature glutamatergic neuronal
   phenotype. This becomes manifest in multiple ways, including increased
   vGlut2 expression level, reduced vGlut1 and endophilin1 expression
   levels, reduced dendrite complexity, and an increased probability of
   neurotransmitter release, partly due to altered coupling between SVs in
   the RRP and VGCCs.

Methods

Materials availability

   The material described in this study is available upon request. This
   includes antibodies, DNA constructs or mouse lines.

Animals

   Generating Nedd8 mouse mutant using CRISPR/Cas9 gene editing.

   The Nedd8cKO knock-out mouse line was generated by site-directed
   CRISPR-Cas9 mutagenesis (LAVES Animals Protocol Permit
   33.19-42502-04-19/3265). Superovulated FVB/N females were mated with
   FVB/N males, and fertilized eggs collected. In-house prepared CRISPR
   reagents (hCas9_mRNA, sgRNAs, preformed Cas9_sgRNA RNP complexes, and
   the dsDNA (HDR fragment) used as a repair template with the required
   two loxP insertions) were microinjected into the pronucleus and the
   cytoplasm of zygotes at the pronuclear stage using an Eppendorf
   Femtojet. Importantly, all nucleotide-based CRISPR-Cas9 reagents
   (sgRNAs and hCAS9_mRNA) were used as RNA molecules and were not
   plasmid-encoded, reducing the probability of off-target effects, due to
   the short live of RNA-based reagents^[237]58,[238]59. The sgRNAs
   targeting the Nedd8 Intron1 (201 bp upstream Exon2) and Intron3 (127 bp
   downstream Exon3) were selected using the guide RNA selection tool
   CRISPOR^[239]60,[240]61. The correct site-specific insertion of the HDR
   fragment was confirmed by localization PCR with primers upstream and
   downstream of the HDR sequence, followed by cloning of the obtained PCR
   products and sequencing of the obtained clones. The new line was
   designated Nedd8^em1Bros. All oligonucleotide sequences are listed in
   Supplementary Data [241]1.

   Mice were bred in the animal facility of the Max Planck Institute for
   Multidisciplinary Sciences (MPI-NAT, City Campus). Nedd8cKOs were
   classified as unburdened and showed no signs of pathogens, according to
   standard ethical guidelines. Animals were killed via cervical
   dislocation. We have complied with all relevant ethical regulations for
   animal use and received ethical approval from the Niedersächsisches
   Landesamt für Verbraucherschutz und Lebensmittelsicherheit (LAVES).
   Mouse maintenance and breeding were performed with permission of the
   LAVES. Animals were kept in ventilated cages (TYPE III, 800 cm2) at
   21 ± 1 °C, 55% relative humidity and under a 12 h/12 h light dark cycle
   in specific pathogen-free conditions. Cages were changed once a week.
   Animals received ad libitum access to autoclaved food and water and
   were provided with bedding and nesting material. Animals’ health was
   controlled daily by caretakers and a veterinarian. Health monitoring
   was done quarterly according to FELASA recommendations, with NMRI
   sentinels or animals from the respective colony. The sex of neonatal
   animals used for experiments was not checked.

Genotyping strategy

   For genotyping, genomic DNA (gDNA) was isolated from tail biopsies
   using a genomic DNA isolation kit (Nexttec, #10.924).

   For the location PCR, 20 μL reactions were prepared using 1 μL clean
   gDNA (15–80 ng), 4 µL PrimerSet (4 pmol final each), 4 µL 5X Reaction
   Buffer (Finnzymes #F-524), 0.4 µL PhireHot-Start II Taq DNA Polymerase
   (Finnzymes #F-122L), 1 µL 10 mM dNTPs (Bioline #DM-515107), 4 µL
   Hi-Spec Additive (Bioline #HS-014101) and 5.6 µL H2O. Thermocycler
   parameters: 98 °C for 5 min, (98 °C for 45 s, 64 °C for 30 s, 72 °C for
   60 s) repeated for 34 cycles, final step at 72 °C for 10 min.

   For the diagnostic routine genotyping PCR, 20 μL reactions were
   prepared using 1 μL clean gDNA (15–80 ng), 4 µL PrimerSet (4 pmol final
   each), 4 µL 5X Reaction Buffer (Biozym #331620XL), 0.2 µL Hot-Start Taq
   DNA Polymerase (Biozym #331620XL), 1 µL 50 mM MgCl2 (AGCTLab stock) and
   9.8 µL H2O. Thermocycler parameters: 96 °C for 3 min, (94 °C for 30 s,
   62 °C for 60 s, 72 °C for 60 s) repeated for 32 cycles, final step at
   72 °C for 7 min.

   Primer sequence and fragment size information: Location PCR to confirm
   the site-specific insertion

   Primer information: 38438 5’-CTAGGCCAAGTGTGGTGTTTGTGGT-3’;
   s_Up_HDR1_Nedd8-cKO location PCR; 38445
   5’-AATCCCAGCAGACAAATCCATGTGA-3’; as_Down_HDR1_Nedd8-cKO location PCR

   Fragment sizes: Nedd8_WT = 2983 bp (38438_38445_2983bp);
   Nedd8_cKO_floxed = 3051 bp (38438_38445_3051bp)

   Routine genotyping diagnostic PCR:

   Primer information: 38447 5’-CGTATTTAATTTGGAGCTTACCGTT-3’;
   Sense2_Nedd8_UpExon2; 38448 5’-GCAAAGATGGCGTAAGTCCCACA-3’;
   Asense3_Nedd8_UpExon2; 38450 5’-GGCACCTGTACTCAAGCGTA-3’;
   Asense_Nedd8_DownExon3

   Fragment sizes: Nedd8_UpEx2_WT = 302 bp (38447_38448_302bp;
   Nedd8_cKO_LoxP1 = 336 bp (38447_38448_336bp); Nedd8_KO = 388 bp
   (38447_38450_388bp), after Cre recombination

Cell lines

   HEK293-FT were kept in DMEM medium (Gibco) containing 10% FBS and
   antibiotics (Gibco, 100 μ/ml Penicillin, 100 μg/ml streptomycin, and
   500 µg/ml gentamycin) at 37 °C with 5% CO[2].

Primary hippocampal neuron culture

   Continental cultures of primary hippocampal neurons were performed as
   described^[242]19. Hippocampi were dissected from homozygous Nedd8cKO
   P0 mice. The tissue was digested in a solution containing 1 mM CaCl[2],
   0.5 mM EDTA, 1.65 mM l-cysteine in DMEM (Gibco) for 1 h at 37 °C under
   agitation (450 rpm). Digestion was stopped via a 15 min incubation at
   37 °C in DMEM supplemented with 10% heat-inactivated fetal bovine serum
   (FBS, Gibco), 2.5 mg albumin, and 2.5 mg/ml trypsin inhibitor. The
   tissue was washed twice with NBA medium (NBA, Gibco) containing, 2% B27
   (Gibco), 2mM Glutamax (ThermoFisher), 100 U/ml of penicillin and 100
   mg/ml of streptomycin (ThermoFisher). Careful trituration was performed
   manually in 100 µL of the same medium. Once the tissue was dissociated,
   the supernatant was plated onto 60 mm Petri dishes (CytoOne) previously
   coated with Poly-L-Lysine (Sigma) at an approximate density of 200.000
   cells/well (two hippocampi). The medium was replaced the next day.
   Cultures were kept at 37 °C with 5% CO[2].

Primary autaptic culture

   Primary autaptic neuronal cultures and astrocyte micro-island cultures
   were performed as previously described^[243]62. Briefly, hippocampi
   were dissected from homozygous Nedd8cKO P0 animals. Tissue digestion,
   washing and trituration steps were performed as described above for
   primary continental cultures. Neurons were plated on 6-well astrocyte
   micro-island plates at a density of 4.000 cells/well. Plates were kept
   in the NBA media containing Neurobasal A medium (Gibco), 2% B27
   (Gibco), 2mM Glutamax (ThermoFisher), 100 U/ml of penicillin, at 37 °C
   and 5% CO[2].

Virus production

   Lentivirus production was performed as previously
   described^[244]63–[245]66. In brief, HEK293-FT (Thermo) cells grown in
   DMEM medium containing 10% FBS and antibiotics (100 µ/ml Penicillin,
   100 μg/ml streptomycin, and 500 µg/ml gentamycin) with an 80–90%
   confluency were transfected with 20 µg of pVSVG packaging and pCMV
   delta R8.9 envelope constructs, and 40 µg of lentivirus construct
   expressing under the Synapsin1 promoter under the Synapsin1 promoter
   either red fluorescent protein (RFP) or CRE-recombinase tagged with an
   RFP protein (CRE). Transfection was performed using Lipofectamine 2000
   (Thermo Scientific). After 6 h, the medium was changed to Optimem
   (Gibco) supplemented with 2% FBS, penicillin 100 µ/ml, Streptomycin
   100 µ/ml, 2mM of Glutamax and 10 mM sodium butyrate. After 46–48 h, the
   cell supernatant was centrifuged at 2.000 rpm for 5 min at 4 °C and
   then filtered using a 0.45 µm filter (Merck). Viral particles were
   filtered and washed with Optimem once and twice with 1x tris-buffered
   saline (TBS) using a 100 kDa Amicon centrifugal filter (Millipore) via
   a series of centrifugation at 3.500 rpm for 10 min at 4 °C. Viral
   particles were concentrated to a final volume of 500 µL, and aliquots
   of 50 µL were snap frozen in liquid nitrogen. vGlut1 and vGlut2 rescue
   viral constructs, as well as empty control constructs, were provided by
   the Viral Core Facility of Charité Universitätsmedizin Berlin.

Electrophysiology

   Whole-cell patch-clamp recordings were performed in autaptic excitatory
   hippocampal neurons at DIV10-12. All recordings were done using a HEKA
   EPC-9USB amplifier controlled by PatchMaster software (Version 2×90.3,
   HEKA electronics), the recording rate was 10 kHz, and R[series] up to
   10 MΩ with 50% compensation. The membrane potential of excitatory
   neurons was set to − 70 mV in voltage-clamp configuration and
   depolarized to 0 mV to induce action potentials. For synaptic
   properties, the extracellular solution contained (mM); 140 NaCl, 2.4
   KCl, 10 HEPES, 10 Glucose, 4 CaCl[2], 4 MgCl[2], pH 7.6.
   Microelectrodes were fabricated from borosilicate glass pipettes using
   a micropipette puller (P-27, Sutter Instrument) and filled with
   intracellular solution containing (mM) 136 KCl, 17.8 HEPES, 1 EGTA, 4.6
   MgCl[2], 4 NaATP, 0.3 Na[2]GTP, 15 creatine phosphate, and 5 µ/ml
   phosphocreatine kinase osmolarity 315–320 mOsmol/L, pH 7.4. Pipette
   resistances were around ∼ 2–4 MΩ. Excitatory post-synaptic currents
   (EPSC) were recorded using frequencies of 0.2, 10 and 40 Hz. The
   release of vesicles from the readily releasable pool (RRP) was
   triggered by application of a hypertonic 0.5 M sucrose solution.
   Estimation of the size of the RRP was obtained from the hypertonic
   sucrose treatment and back extrapolation of the cumulative EPSC from
   40 Hz trains of action potential^[246]67. Miniature EPSC (mEPSC) were
   recorded under 300 mM tetrodotoxin (TTX, Tocris) treatment, and for
   post-synaptic responses, 100 µM Glutamate (Sigma) or 3 µM
   γ-aminobutyric acid (GABA) (Sigma) solutions were used. The
   paired-pulse ratio was calculated from the 40 Hz train stimulus and is
   the ratio of the peak amplitudes of the second and first responses. For
   calcium channel sensitivity studies, the extracellular solution
   contained (mM) 140 TEA-Cl, 2.5 CsCl, 10 CaCl[2], 1 MgCl[2], 10 HEPES,
   10 glucose, Osmolarity 310 mOsm, pH 7.3; and the intracellular solution
   contained 130 CsCl, 10 HEPES, 2 CaCl[2], 10 EGTA, 5 MgATP, Ph 7.3.
   10 µM Nimodipine (Alomone Labs) or 0.2 µM ω-Agatoxin (Alomone Labs)
   (L-type and P/Q-type calcium channel blocker, respectively) were used
   to block specific calcium channels. For synaptic vesicle coupling
   studies, 100 µM EGTA-AM (Invitrogen) was used. Infected neurons were
   distinguished by a nuclear RFP signal that was visualized using an
   inverted microscope (Zeiss). Data were analyzed using AxoGraph software
   (Version 1.5.4., Axograph Scientific).

Western blot

   Neuronal cultured were lysed in lysis buffer containing 20 mM Tris
   Base, 150 mM NaCl, 1% Triton X-100 (Roche), and protease inhibitors
   (0.5 µg/ml aprotinine, 0.2 mM phenylmethylsulfonyl fluoride, 1 µg/ml
   leupeptine, 20 mM N-ethylmaleimide), pH 7.4–7.6. For storage, samples
   were snap-frozen in liquid nitrogen and kept at − 80 °C. SDS-PAGE was
   performed using self-made 10% acrylamide gels, or commercial,
   pre-casted 4–12% Bis-Tris gels (Thermo Scientific). Samples were
   sonicated, and protein concentration was determined using the Bradford
   assay (Biorad). Samples were heated for 5 min at 95 °C, or for 20 min
   65 °C, shortly before loading. Gel migration was performed in a buffer
   containing 5 mM MOPS, 5 mM Tris Base, 0.01% SDS, 0.1 mM EDTA, pH 7.7.
   Proteins were transferred to nitrocellulose membranes (Cytiva) using a
   buffer containing 192 mM Glycine, 25 mM Tris Base, and 20% Methanol.
   Total protein content was visualized using the Memcode reversible
   protein staining kit (Thermo Scientific). Primary and secondary
   antibody incubation of the membrane were performed in 1x PBS
   supplemented with 5% milk (Frema) and 0.1% Tween 20 (PBS-T) following
   standard procedure. Signal was revealed using enhanced chemiluminescent
   (GE Healthcare), and signals were detected with an ECL Chemostar Imager
   (Intas Science Imaging) and analyzed with Fiji (NIH). Quantification
   was performed by normalizing the signal in each lane with its
   respective Memcode signal.

RNA isolation, RNA-sequencing, and analysis

   The continental culture of primary hippocampal neurons aged for 13 days
   were subjected to TRIzol-RNA extraction following the manufacture
   protocol (Zymo). RNA quality was assessed by measuring the RNA
   integrity number (RIN) using a Fragment Analyzer HS Total RNA Kit
   (Advanced Analytical Technologies, Inc.). Library preparation for
   RNA-Seq was performed in the STAR Hamilton NGS automation using the
   Illumina Stranded mRNA Prep, Ligation (Cat. N°20040534 and the Illumina
   RNA UD Indexes Set A, Ligation, 96 Indexes, 96 Samples Cat.
   N°20091655), starting from 300 ng of total RNA. The size range of the
   final cDNA libraries was determined by applying the SS NGS Fragment 1-
   to 6000-bp Kit on the Fragment Analyzer (average 340 bp). Accurate
   quantification of cDNA libraries was performed by using the
   QuantiFluor™ dsDNA System (Promega). cDNA libraries were amplified and
   sequenced by using an S4 flow cell NovaSeq6000; 300 cycles, 25 Mio
   reads/sample from Illumina.

Raw read & quality check

   Sequence images were transformed with BaseCaller Illumine Software to
   BCL files and demultiplexed to fast files with bcl2fastq v2.20.0.422.
   The sequencing quality was asserted using FastQC v.0.11.5 software
   ([247]http://www.bioinformatics.babraham.ac.uk/projects/fastqc/)^[248]6
   8.

Mapping and normalization

   Sequences were aligned to the reference genome Mus musculus (GRCm39
   version 110, [249]https://www.ensembl.org/Mus_musculus/Info/Index)
   using the RNA-Seq alignment tool (version 2.7.8)^[250]69, allowing for
   2 mismatches within 50 bases. Subsequently, read counting was performed
   using featureCounts^[251]70. Read counts were normalized in the
   R/Bioconductor environment (version 4.3.1) using the DESeq2^[252]71
   package version 1.40.2.

PCA and other quality checks

   Principal component analysis (PCA) was performed using the sklearn
   Python library (version 1.3.1) (Supplementary Fig. 2B. [253]1). The
   total gene count box plot displays gene expression levels in each of
   the samples analyzed Supplementary Fig. [254]2B. The horizontal line
   inside each box represents the median gene expression. The box itself
   encompasses the interquartile range, covering from the 25th to the 75th
   percentile. The whiskers extend to the 10th percentile on the lower end
   and the 90th percentile on the higher end. Any points outside the
   whiskers, particularly those above the 90th percentile, are shown as
   individual circles. The plot uses color to differentiate between
   conditions, with red indicating the ‘CRE’ condition and green
   representing the ‘RFP’ condition. After PCA and gene count analysis,
   two samples were removed from the differential analysis: Nedd8Cre1-3
   and Nedd8Cre3-3.

Differential expression and pathway enrichment analysis

   Differential expression analysis was performed using PyDESeq2, a Python
   implementation of DESeq2 method^[255]71,[256]72 (version 0.4.4).
   Pathway enrichment was done with GSEA method^[257]73 using blitzgsea
   library (version 1.3.42)^[258]74. The calculation of the score in
   pathway analysis was done as previously described^[259]75.

Real time quantitative PCR (RT-qPCR)

   Continental cultures of DIV13 primary hippocampal neurons were
   subjected to TRIzol-RNA extraction according to the manufacturer’s
   protocol (Zymo). RNA extraction and purification were performed using
   the Direct-zol RNA microprep Kit (Zymo) according to the manufacturer’s
   instructions. Reverse transcription was performed with 500 ng of total
   RNA using SuperScript III Reverse Transcriptase (Invitrogen) according
   to the manufacturer’s instructions. qPCR was performed with a Mic qPCR
   Cycler (Bio Molecular Systems). All genes were analyzed using Power
   SYBR™ Green PCR Master Mix (Applied Biosystems). For each reaction,
   1 µL of cDNA, 2 pmol/µL of primers, and 16 µL of SYBR Green Master Mix
   were used. All genes were tested in duplicate. For the analysis, all
   the Ct values were standardized to the Ct of Gapdh and Ubc, and
   normalized to the levels of the RFP condition. The list of primers used
   for RT-qPCR is available in Supplementary Data [260]1.

Immunostaining and confocal microscopy

   DIV12-13 autaptic hippocampal neurons were fixed with 4%
   paraformaldehyde (PFA) prepared in 1x phosphate-buffered saline (PBS)
   for 10 min. All steps were done at room temperature. Neurons were
   washed three times with 1x PBS and quenched for 10 min with 50 mM
   Glycine in 1x PBS. Neurons were permeabilized in 1 x PBS containing
   0.1% Triton X-100 (Roche), 2.5% goat serum (GS) (Gibco) for 30 min.
   Cells were washed in 1 x PBS supplemented with 2.5% GS and incubated
   with primary antibodies diluted in the same solution for 1 h. Neurons
   were washed and incubated with secondary antibodies for 45 min in
   darkness. Coverslips were washed in 1x PBS and mounted on slides using
   Aqua-Poly Mount (Polysciences) medium and stored at 4 °C until image
   acquisition. Images were obtained with a SP8 confocal Leica microscope
   (Leica Microsystems), 63x immersion objective (NA 1.4), 12-bit image
   depth, step size of 0.3 µm, and resolution of 1232 × 1232 corresponding
   to a pixel size of 149.9 nm. Synapse counting, Sholl analysis and total
   dendrite length were calculated using Imaris software (Version 9.9.1,
   Oxford Instruments). Representative images were processed using Fiji
   (Version 2.14.0/1.54 f, NIH).

STED microscopy and analysis

   DIV12 autaptic hippocampal neurons were fixed with 4% PFA for 45 min
   and then quenched using 100 mM NH[4]Cl in 1 x PBS for 20 min. All steps
   were done at room temperature. Cells were washed twice in 1 x PBS and
   incubated in a blocking solution containing 2.5% bovine serum albumin
   (BSA, Biomol) and 0.1% Triton X-100 (Roche) in 1 x PBS. Neurons were
   incubated with nanobodies against vGlut1-AZD568 (Nanotag),
   PSD95-STAR635 (Nanotag), and Synaptotagmin1-ATTO488 (Nanotag) in a
   1:100 dilution in blocking solution. Coverslips were washed twice in
   1 x PBS and mounted on slides using Mowiol (Carl Roth). Image
   acquisition was done using an Abberior Expert line setup (Abberior
   Instruments) with an IX83 microscope (Olympus), with a 100x immersion
   objective (1.4 NA). Excitation of STAR635 was performed with a 640 nm
   laser (5% of max. power) and detected with an avalanche photodiode
   (APD) laser (range 650–720 nm). For AZD568, excitation was done using a
   580 nm laser (30% of max. power) and detected with an APD laser (range
   605–625 nm). For both STAR635 and AZD568, depletion was achieved with a
   775 nm depletion laser (20% and 60%, respectively). Excitation of
   ATTO488 was achieved with a 488 nm laser (4% of max. power) and
   detected with an APD laser (range 525–575 nm). For ATTO488, a
   solid-state 595 nm depletion laser (10% of max. power) was used. For
   each neuron, 2–5 regions of interest were selected and imaged using
   confocal and STED microscopy. Pixel sizes were 200 and 20 nm for
   confocal and STED, respectively, dwell times 8–10 µs, and lines
   accumulation of 3–5 for STED. The analysis was performed using MATLAB
   (Version R2022b, The Mathworks, Inc.). Synapses were identified
   automatically, using a band-pass filter procedure, which revealed the
   synapse-sized spots in the different channels. Spots of less than 9
   pixels were removed, being considered to be noise events. All synapse
   images from one experiment were then automatically rotated and oriented
   to display the postsynaptic density in the center, and the presynapse
   to the left. Line scans were generated across the synapses, and the
   intensities in the different channels were measured and displayed.
   Representative images were processed using Fiji (Version 2.14.0/1.54 f,
   NIH).

Ultrastructural analysis of Nedd8-deficient synapses

Primary hippocampal neuron culture

   For electron microscopy experiments, neuronal monolayer cultures from
   P0 Nedd8-cKO mice were grown directly on carbon- and PDL-coated 6 mm
   sapphire disc freezing substrates (Wohlwend). Sapphire discs were
   mounted on 18 mm glass coverslips using Matrigel and placed in a
   12-well plate for UV sterilization for 1 h at RT. Hippocampi were
   prepared as described above in this section. Cultured neurons were
   vitrified by high-pressure freezing and processed for ultrastructural
   analysis at DIV13.

High pressure freezing, freeze-substitution, and ultramicrotomy

   Sapphire discs with neurons were infected at DIV1 with lentiviral
   constructs designed to express Cre-recombinase and RFP, or RFP alone
   (negative control). At DIV 13, infected neurons underwent three washing
   steps in pre-warmed (37 °C) conditioned medium before rapid
   cryofixation using an EM ICE high-pressure freezer (Leica) and
   subsequent storage in liquid nitrogen. Semi-automated
   freeze-substitution was performed using an AFS2 device (Leica) equipped
   with customized aluminum sapphire disc revolvers, and samples were
   subsequently embedded in epoxy resin according to previously published
   protocols^[261]76,[262]77. Sapphire discs were separated from
   polymerized samples with a razor blade, and resin block-faces were
   trimmed with an EM TRIM high-speed milling device (Leica) in
   preparation for ultramicrotomy. A 35° Ultra-Semi diamond knife
   (Diatome) mounted on a UC7 ultramicrotome (Leica) was used to cut 250
   nm-thick sections onto formvar-coated copper parallel line grids
   (Gilder; G100P-Cu). Mounted sections were coated on both surfaces with
   Protein A-conjugated 10 nm gold fiducial particles (Cell Microscopy
   Core Products, University Medical Center Utrecht, The Netherlands) in
   preparation for 3D ultrastructural analysis using transmission electron
   tomography.

Electron microscopy

   Synapses were identified and selected for 3D ultrastructural analysis
   in tiled overviews acquired at 5000 x magnification using a 200 kV
   Talos F200C G2 scanning/transmission electron microscope equipped with
   a 16 MP Ceta CMOS camera (Thermo Scientific) and SerialEM software
   (IMOD Software Package^[263]78). Tilt series ( ± 60° tilt range; 1°
   tilt increments) were acquired at 36000 x (unbinned pixel
   size = 0.4 nm) magnification from orthogonal axes using a Model 2040
   dual-axis specimen holder (Fischione). Tomograms were generated from
   tilt-series using Gaussian filtering and a weighted back-projection
   algorithm implemented using Etomo software (IMOD software
   package^[264]79). Presynaptic vesicles were initially segmented using
   SynapseNet^[265]80 as perfect spheres and then proof-read using 3dmod
   software (IMOD Software PackMastronadeage^[266]79). Active zones were
   segmented manually on every 3^rd tomographic slice. Segmented data were
   analyzed using mtk and imodinfo programs in conjunction with a
   customized Python script to quantify the relative spatial distribution
   of presynaptic vesicles (i.e., closest approach to active zone release
   sites), their size (i.e., SV diameter) and clustering characteristics
   (i.e., intervesicular nearest-neighbor distances), respectively.

Statistical analysis and reproducibility

   All statistical analyses were performed using GraphPad Prism software
   (Version 10.3.0., GraphPad). Normal distribution was determined using
   D´Agostino-Pearson or Shapiro-Wilk normality tests. Statistical
   analysis was performed using a Kruskal-Wallis and Dunn´s multiple
   comparison test, unpaired t test, Welch’s test, or Mann-Whitney tests.
   For graphs, data are shown as mean ± SEM. For all figures,
   statistically significant differences are denoted in graphs with *
   standing for p-value, where * refers to p < 0.05, ** to p < 0.01, ***
   to p < 0.001, and **** to p < 0.0001. N refers to the number of animals
   or biological replicates, and n corresponds to the number of cells. N
   and n are indicated in the legends.

Reporting summary

   Further information on research design is available in the [267]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [268]Supplementary Information^ (3.2MB, docx)
   [269]42003_2025_8680_MOESM2_ESM.docx^ (14.4KB, docx)

   Description of additional supplementary file
   [270]Supplementary data 1^ (28.8KB, docx)
   [271]Supplementary data 2^ (3MB, xlsx)
   [272]Supplementary data 3^ (74.9KB, xlsx)
   [273]Supplementary data 4^ (396.7KB, xlsx)
   [274]Supplementary data 5^ (17.4KB, xlsx)
   [275]Supplementary data 6^ (216.4KB, xlsx)
   [276]Supplementary data 7^ (4.5MB, xlsx)
   [277]Reporting Summary^ (1.6MB, pdf)
   [278]Transparent Peer Review file^ (170.9KB, pdf)

Acknowledgements