1. Neuroscience
Download icon

CB1-receptor-mediated inhibitory LTD triggers presynaptic remodeling via protein synthesis and ubiquitination

  1. Hannah R Monday
  2. Mathieu Bourdenx
  3. Bryen A Jordan
  4. Pablo E Castillo  Is a corresponding author
  1. Dominick P. Purpura Department of Neuroscience, Albert Einstein College of Medicine, United States
  2. Department of Developmental and Molecular Biology, Albert Einstein College of Medicine, United States
  3. Institute for Aging Studies, Albert Einstein College of Medicine, United States
  4. Department of Psychiatry and Behavioral Sciences, Albert Einstein College of Medicine, United States
Research Article
  • Cited 0
  • Views 615
  • Annotations
Cite this article as: eLife 2020;9:e54812 doi: 10.7554/eLife.54812

Abstract

Long-lasting forms of postsynaptic plasticity commonly involve protein synthesis-dependent structural changes of dendritic spines. However, the relationship between protein synthesis and presynaptic structural plasticity remains unclear. Here, we investigated structural changes in cannabinoid-receptor 1 (CB1)-mediated long-term depression of inhibitory transmission (iLTD), a form of presynaptic plasticity that involves a protein-synthesis-dependent long-lasting reduction in GABA release. We found that CB1-iLTD in acute rat hippocampal slices was associated with protein synthesis-dependent presynaptic structural changes. Using proteomics, we determined that CB1 activation in hippocampal neurons resulted in increased ribosomal proteins and initiation factors, but decreased levels of proteins involved in regulation of the actin cytoskeleton, such as ARPC2 and WASF1/WAVE1, and presynaptic release. Moreover, while CB1-iLTD increased ubiquitin/proteasome activity, ubiquitination but not proteasomal degradation was critical for structural and functional presynaptic CB1-iLTD. Thus, CB1-iLTD relies on both protein synthesis and ubiquitination to elicit structural changes that underlie long-term reduction of GABA release.

Introduction

Synaptic plasticity, the ability of synapses to change their strength in response to activity or experience, underlies information storage in the brain. While presynaptic forms of plasticity, that is long-term synaptic strengthening (long-term potentiation or LTP) and weakening (long-term depression or LTD) due to long-lasting increase and decrease in neurotransmitter release, respectively, are widely expressed in the brain, their mechanism remains poorly understood (Castillo, 2012; Monday and Castillo, 2017; Monday et al., 2018; Yang and Calakos, 2013). A good example of a ubiquitous form of long-lasting reduction of transmitter release is type-1 cannabinoid receptor (CB1)-mediated LTD (Araque et al., 2017; Castillo et al., 2012; Heifets and Castillo, 2009). Here, endogenous cannabinoids (eCBs) are released upon activity and travel in a retrograde manner to bind presynaptic CB1, a Gi/o-coupled receptor, resulting in CB1-LTD at both excitatory and inhibitory synapses. Induction of long-term eCB-mediated plasticity requires extended (minutes) CB1 activation (Chevaleyre and Castillo, 2003; Chevaleyre et al., 2007; Ronesi, 2004). Although the presynaptic changes downstream CB1 that suppress transmitter release in a long-term manner remain unclear, there is evidence that presynaptic protein synthesis is required (Yin et al., 2006; Younts et al., 2016), but what proteins are synthesized and the precise role of these proteins in CB1-LTD is unclear.

Proteostatic mechanisms, the cellular processes that balance protein synthesis and degradation, are vital for neuronal function and synaptic plasticity (Biever et al., 2019; Birdsall and Waites, 2019; Cohen and Ziv, 2019; Liang and Sigrist, 2018; Wang et al., 2017). In postsynaptic forms of plasticity, such as NMDA receptor-dependent LTP, local protein synthesis has been tightly correlated with both consolidation of LTP and structural changes (Bosch et al., 2014; Ostroff et al., 2002; Tanaka et al., 2008; Tominaga-Yoshino et al., 2008; Yang et al., 2008). In particular, synthesis of β-actin, AMPA receptors, and CaMKII proteins may be critical for the increase in dendritic spine volume and synapse strength associated with LTP (Bramham, 2008; Nakahata and Yasuda, 2018; Rangaraju et al., 2017). Concurrent regulation of protein degradation through the proteasome and lysosome is also required for activity-dependent pre- and postsynaptic changes in synapse strength (Biever et al., 2019; Cohen and Ziv, 2017; Hegde, 2017; Monday et al., 2018). We and others have recently provided evidence for rapid (<30 min) presynaptic protein synthesis under basal conditions and during plasticity (Hafner et al., 2019; Younts et al., 2016), but whether these newly synthesized proteins participate in CB1-LTD by regulating presynaptic structural changes is unknown.

Presynaptic structure and function are controlled by actin polymerization and depolymerization (Cingolani and Goda, 2008; Nelson et al., 2013). Branched actin filaments in the presynaptic compartment provide a scaffold for synaptic vesicles and the active zone (Michel et al., 2015; Rust and Maritzen, 2015). Moreover, structural changes of the presynaptic terminal are associated with altered synapse strength (Gundelfinger and Fejtova, 2012; Matz et al., 2010; Meyer et al., 2014; Monday and Castillo, 2017), and the size of the presynaptic terminal and active zone has been correlated with the postsynaptic response (Bartol et al., 2015; Bourne et al., 2013; Meyer et al., 2014). There is evidence that CB1 activation alters the ultrastructural vesicle distribution in CB1-expressing (CB1+) boutons on short time scales (García-Morales et al., 2015; Ramírez-Franco et al., 2014) and leads to retraction of growth cones in developing axons (Roland et al., 2014). However, it is not known whether CB1-LTD is associated with morphological changes in presynaptic structure in the mature mammalian brain.

Here, to gain insights into the expression mechanisms of CB1-LTD, we examined potential structural changes in CB1-mediated LTD of inhibitory transmission (CB1-iLTD) in the hippocampus (Chevaleyre and Castillo, 2003). Using high-resolution microscopy in acute rat hippocampal slices we found that this form of plasticity was associated with a reduction of presynaptic bouton volume that required protein synthesis. To test how protein synthesis could alter presynaptic structure during CB1-iLTD, we used an unbiased proteomics approach to identify CB1 activation-mediated changes in the proteome of hippocampal neuron cultures. CB1 activation elicited an increase in proteins involved in protein synthesis, processing and degradation, whereas presynaptic and actin cytoskeletal proteins, including ARPC2 and WASF1/WAVE1 were decreased. CB1-iLTD involved actin remodeling, Rac1 and the actin branching protein complex Arp2/3. Lastly, ubiquitination of proteins but not proteasomal degradation was necessary for both structural and functional CB1-iLTD.

Results

Induction of CB1-iLTD is associated with a reduction in CB1+ bouton size

Diverse forms of long-lasting synaptic plasticity require translation-dependent structural remodeling (Bailey et al., 2015; Bramham, 2008; Nakahata and Yasuda, 2018; Rangaraju et al., 2017). To test whether CB1-iLTD is associated with structural plasticity, we examined changes in the morphology of CB1+ boutons following induction of CB1-iLTD. To accurately measure individual bouton volume in acute hippocampal slices, we utilized high-resolution and high yield Airyscan confocal microscopy combined with 3D reconstruction (Figure 1A). To induce CB1-iLTD the CB1 agonist WIN 55,212–2 (5 µM) was bath applied for 25 min, and induction was confirmed by monitoring extracellular field inhibitory postsynaptic potentials (fIPSPs), which allows non-invasive, stable long-term assessment of inhibitory synaptic transmission (Heifets et al., 2008; Younts et al., 2016; Figure 1B). This LTD not only mimics synaptically induced iLTD (Chevaleyre and Castillo, 2003; Chevaleyre et al., 2007; Heifets et al., 2008), but also allows us to shortcut the synthesis and release of eCBs, thereby excluding potential effects of pharmacological inhibitors (see below) on these processes.

Figure 1 with 2 supplements see all
Induction of CB1-iLTD is associated with a reduction in CB1+ bouton size.

(A) Left, representative stitched widefield confocal of hippocampal CB1 immunolabeling. White boxes indicate where high-magnification images were acquired (as seen in center panel). Center, High-resolution Airyscan confocal maximum projection used for 3D reconstruction of individual boutons in CA1 stratum pyramidale. Right, 3D reconstruction of Airyscan confocal image shown in center panel. (B) Top, representative extracellular field inhibitory postsynaptic potentials (fIPSPs) recorded in the CA1 pyramidal cell body layer in acute hippocampal slices before and after WIN treatment (5 µM, 25 min). Bottom, summary time-course plot showing WIN-induced depression; n = 3 slices, three animals. For all electrophysiology figures, representative traces correspond to the gray-shaded areas and in the summary time-course plots (averaged summary data expressed as normalized change from baseline ± S.E.M.). Shaded boxes in all electrophysiology figures correspond to when plasticity was analyzed with respect to baseline and when representative traces were collected and averaged. (C) Quantification of mean bouton volume per slice normalized to Control. Activation of CB1 receptors with 5 µM WIN for 25 min led to decreased bouton volume that remained 1 hr after WIN treatment and was blocked by treatment with cycloheximide (CHX, 80 µM, applied throughout the experiment). Control: 1.0 ± 0.02 vs. WIN: 0.92 ± 0.02 vs. WIN + 1 hr: 0.89 ± 0.02 vs. CHX + WIN: 0.97 ± 0.02 (Mean ± S.E.M.); F[3,22]=8.682; p=0.0005, one-way ANOVA with post-hoc Tukey test for multiple comparisons. Control vs. WIN: CI[0.005,0.14], p=0.03. Control vs. WIN 1 hr: CI[0.05,0.18], p=0.0005. WIN 1 hr vs. WIN + CHX: CI[−0.11,0.03], p=0.012. n = number of slices (three images/slice, 1–2 slices/rat, six rats/condition). For all structural plasticity figures, data are presented as points representing mean bouton volume per slice with a horizontal black line representing the mean per condition and the vertical line indicating 95% confidence interval (CI). 95% CIs are given as CI[lower CI, upper CI]. p-Values are exact. Refer to Figure 1—figure supplement 1A for distribution of individual bouton values. (D) Representative 3D reconstruction widefield images of Bassoon labeling inside CB1+ boutons. (E) Representative single CB1 boutons with Bassoon reconstructed in 3D. (F) Quantification of mean Bassoon volume per slice normalized to Control after 25 min WIN treatment revealed a reduction in active zone volume as measured using Bassoon immunolabeling that was blocked by treatment with cycloheximide (CHX, 80 µM, applied throughout the experiment). Control: 1.0 ± 0.01 vs. WIN: 0.89 ± 0.04 vs. CHX + WIN: 1.0 ± 0.01 vs. CHX: 0.98 ± 0.03 (Mean ± S.E.M.); F[3,12]=4.11, p=0.032, one-way ANOVA with post-hoc Tukey test for multiple comparisons. Control vs. WIN: CI[0.004,0.22], p=0.042. n = number of slices (three images/slice, 1–2 slices/rat, three rats/condition). Refer to Figure 1—figure supplement 1D for distribution of individual Bassoon values.

Using CB1 immunolabeling, which accurately approximates bouton volume (Dudok et al., 2015), we found that CB1-iLTD is associated with a significant decrease of CB1 bouton volume (Figure 1C). This structural change was long-lasting as it persisted for 60 min after WIN treatment (Figure 1C), and it was blocked by concurrent bath application with protein synthesis inhibitor cycloheximide (CHX, 80 µM), demonstrating a requirement for protein synthesis (Figure 1C,E, Figure 1—figure supplement 1A). CHX alone did not affect basal bouton volume (Figure 1—figure supplement 1B). This overall shrinkage by WIN treatment was driven by an increase in the proportion of the small CB1 boutons and a trend toward a decrease in large boutons (Figure 1—figure supplement 1C). This effect was specific because volume of parvalbumin (PV+) boutons in the CA1 pyramidal layer was not altered by WIN application (Figure 1—figure supplement 2A,B), as assessed by immunolabeling for PV boutons (Younts et al., 2016), which do not express CB1 receptors (Glickfeld and Scanziani, 2006; Figure 1—figure supplement 2C). As a complementary approach, we used Bassoon immunolabeling to assess the size of the presynaptic active zone. Similar to the total bouton volume, Bassoon size within CB1+ boutons was also significantly decreased following WIN application and this effect was blocked by CHX treatment (Figure 1D–F, Figure 1—figure supplement 1D). We also found that the volume of vesicular GABA transporter (VGAT), a well-established marker of inhibitory synapses, was markedly reduced in CB1+ boutons by WIN treatment (Figure 1—figure supplement 2D). These results strongly suggest that CB1-iLTD is associated with a protein-synthesis-dependent shrinkage of CB1+ boutons, which may contribute to the long-lasting reduction in neurotransmitter release observed in this form of plasticity. Along with our previous study (Younts et al., 2016), our findings indicate that protein synthesis is required for both structural and functional presynaptic changes involved in CB1-iLTD.

CB1 activation alters the abundance of proteins linked to protein synthesis, synaptic structure/function and energy metabolism

To glean insights into the mechanism(s) underlying structural and functional CB1-iLTD, we sought to identify proteins synthesized upon CB1 activation. We previously showed CB1-dependent increases in protein synthesis were evident after brief CB1 activation in cultured hippocampal neurons (Younts et al., 2016). To identify and quantitate changes in the neuronal proteome, we used Stable Isotope Labeling of Amino Acids in Cell Culture (SILAC) coupled with tandem mass spectrometry (MS/MS) (Jordan et al., 2006; Zhang et al., 2011; Zhang et al., 2012). Two populations of cultured rat hippocampal neurons (‘medium’ and ‘heavy’) were labeled using distinct combinations of ‘medium’ or ‘heavy’ stable-isotope weight variants of Arginine and Lysine. The two groups were treated with WIN (25 min, 5 µM) or Vehicle as before (Figure 1) then rapidly lysed and harvested (Figure 2—figure supplement 1A,B). Samples were combined and simultaneously analyzed by tandem MS/MS to identify and quantify changes induced by CB1 receptor activation. To strengthen the robustness of findings, we performed a replicate ‘reverse’ experiment where ‘heavy’ neurons were treated with WIN and observed a high degree of correlation between replicates (Figure 2—figure supplement 1C).

We found significant changes across the protein landscape. Examples of these proteins grouped by their suggested function are shown in Figure 2A (see Supplementary file 1 for all proteins). Consistent with previous studies of axonal mRNAs, components of the protein synthesis machinery were upregulated (Hafner et al., 2019; Shigeoka et al., 2016), as well as the protein degradation machinery. A number of presynaptic proteins were downregulated following CB1 activation. Notably, two key regulators of the actin cytoskeleton, Actin-related protein 2/3 complex subunit 2 (ARPC2) and Wiscott-Aldrich Associated Protein family (WASF1/WAVE1) were significantly downregulated, suggesting these proteins could be implicated in the reduction of neurotransmitter release and presynaptic volume associated with CB1-iLTD. Using Gene Set Enrichment Analysis (GSEA) (Subramanian et al., 2005), we identified enriched functional gene ontology (GO) terms (Figure 2B and Supplementary file 2). To reduce redundancy, we clustered closely related GO terms using network analysis (Merico et al., 2010), where edge length corresponds to the number of overlapping proteins in the GO term, node size indicates the number of proteins belonging to the term, and color represents the enrichment score (Figure 2C and Supplementary file 2). In accordance with translational upregulation following CB1 activation (Younts et al., 2016), the top cluster represented upregulated GO terms related to ‘Protein synthesis and processing’ (Figure 2C). The second cluster was composed of GO terms relating to ‘Neuronal projections’, suggestive of the structural change associated with CB1-iLTD. The third cluster was GO terms associated with ‘Energy metabolism’ which may be representative of the previously reported CB1-mediated decrease in cellular respiration (Hebert-Chatelain et al., 2016; Mendizabal-Zubiaga et al., 2016; Figure 2C). Examples of GO terms identified in each cluster are provided in Figure 2—figure supplement 2. Ingenuity pathway analysis (IPA) also identified pathways related to the processes outlined above, including eIF2 signaling, mitochondrial dysfunction, and actin cytoskeleton signaling (Figure 2—figure supplement 1D). Similarly, analyses using SynGO (Koopmans et al., 2019), an expert-curated tool to identify GO terms associated with synaptic function, linked our results to regulation of synaptic protein synthesis (Figure 2—figure supplement 1E). Among the differentially expressed proteins, we found that 43 proteins were upregulated and 56 proteins were downregulated by CB1 activation (Figure 2D). Together, these results suggest that both protein synthesis and coincident degradation of structural and presynaptic proteins occur downstream of CB1 activation, and could therefore be implicated in CB1-iLTD.

Figure 2 with 2 supplements see all
CB1activation alters the abundance of proteins linked to protein synthesis, synaptic structure/function and energy metabolism.

(A) Examples of proteins that were identified in enriched GO terms and were significantly altered by CB1 activation (p<0.05). Proteins are grouped by proposed biological function and average log2 fold change is plotted. (B) List of enriched GO terms and normalized enrichment scores (NES) as identified by GSEA. Positive NES reflects overall upregulation of proteins associated with the GO term whereas negative values indicate the opposite. (C) Cluster analysis of enriched/depleted GO terms from GSEA revealed four distinct biological processes that were consistently up- or downregulated by CB1 activation. Each node represents a single GO term. Node size represents magnitude of enrichment and edge length gives degree of overlap between 2 GO terms. Color represents up (red) or downregulation (blue) of proteins associated with that GO term. See Figure 2—figure supplement 2 for examples. (D) Volcano plot of differentially expressed proteins between vehicle and CB1 activation. Red dots: differentially expressed proteins showing upregulation (adj. p<0.05 and log2 fold change >0). Blue dots: differentially expressed proteins showing downregulation (adj. p<0.05 and log2 fold change <0). six select top hits are highlighted: ECHA: trifunctional enzyme subunit alpha (mitochondrial); SNX3: Sorting nexin-3; RPN2: Subunit of the oligosaccharyl transferase; SV2A: Synaptic vesicle glycoprotein 2A; ANXA1: Annexin A1; SYPH: Synaptophysin.

CB1-iLTD involves actin remodeling via Rac1 and Arp2/3

CB1 directly interacts with Rac1 and members of the WAVE regulatory complex (WRC), which includes actin branching modulators WASF1/WAVE1 and Arp2/3 (Njoo et al., 2015), and these proteins are downregulated in hippocampal neurons following CB1 activation (Figure 2A). Therefore, regulation of the abundance of these proteins may represent a mechanism underlying structural and functional presynaptic changes involved in CB1-iLTD. For example, CB1 activation could reduce the presynaptic terminal volume by favoring actin depolymerization. To test this possibility, we first examined whether actin cytoskeletal dynamics were required for CB1-iLTD induced structural plasticity (Figure 1). Using the same high-resolution microscopy and 3D reconstruction as Figure 1, we activated CB1 in the presence of jasplakinolide (JSK, 250 nM), a reagent that promotes actin polymerization (Holzinger, 2009). We found that JSK application blocked the WIN-induced decrease in presynaptic bouton volume (Figure 3A, Figure 3—figure supplement 1A). These results indicate that actin dynamics likely underlie the structural changes following CB1 activation. We next tested the functional requirement for actin remodeling in CB1-iLTD (as in Figure 1). Similar to the effects on structural plasticity, bath application of JSK impaired CB1-iLTD (Figure 3B), whereas JSK application alone had no effect on basal synaptic transmission (Figure 3—figure supplement 1B). These results strongly suggest that actin remodeling is critical for structural and functional CB1-iLTD.

Figure 3 with 1 supplement see all
CB1-iLTD involves actin remodeling via Rac1 and Arp2/3.

(A) Left, representative single boutons reconstructed in 3D. Right, quantification of mean bouton volume per slice normalized to control. Activation of CB1 receptors with WIN for 25 min led to decreased bouton volume that was blocked by treatment with jasplakinolide (JSK, 250 nM). Summary data expressed as normalized change from Control. WIN + 1 hr: 0.89 ± 0.02 vs. WIN + JSK: 0.97 ± 0.02 vs. JSK: 0.95 ± 0.02 (Mean ± S.E.M.); F[2,16]=5.56, p=0.015, one-way ANOVA with post-hoc Tukey test for multiple comparisons. WIN vs. WIN + JSK: CI[−0.16,–0.018], p=0.014. n = number of slices (three images/slice, 1–2 slices/rat, six rats/condition). For all structural plasticity figures, data are presented as points representing mean bouton volume per slice with a horizontal black line representing the mean per condition and the vertical line indicating 95% confidence interval (CI). 95% CIs are given as CI[lower CI, upper CI]. p-Values are exact. Refer to Figure 3—figure supplement 1A for distribution of individual bouton values. (B) CB1-iLTD is impaired by bath application of actin-stabilizing drug, jasplakinolide (JSK, 250 nM). Extracellular field inhibitory postsynaptic potential (fIPSP) were recorded in the CA1 pyramidal cell body layer in acute hippocampal slices. Control: 61.4 ± 4% vs. JSK: 80.5 ± 4%; p<0.05, unpaired t-test. Unless otherwise specified, n = number of slices (s), number of animals (a). (C) CB1-iLTD was blocked by acute bath application of the Rac1 inhibitor NSC (30 µM). Control: 68.8 ± 6% vs. NSC23766: 89.7 ± 4%; p<0.05, unpaired t-test. (D) NSC (30 µM) bath application reversibly depressed basal transmission. NSC: 98 ± 2%, one sample t-test, p>0.05. (E) CB1-iLTD is enhanced by acute bath application of the Arp2/3 inhibitor CK-666 (100 µM). Control: 63.8 ± 4% vs. CK-666: 45.2 ± 4%, p>0.05, unpaired t-test. (F) Left, representative Western blots of staining for Arpc2 and Ponceau loading control in vehicle or WIN-treated hippocampal cultures. Right, Arpc2 was downregulated in hippocampal neuron cultures after CB1 activation with WIN (5 µM, 25 min). Arpc2 (Fold of Veh): 0.851 ± 0.06, U = 16, Z = 2.31, * indicates p<0.05, Mann-Whitney. Dots represent individual values for four independent experiments. Data in the bar plot represent mean ± S.E.M. (G) Proposed model of CB1-iLTD pathway and mechanism of action of pharmacological reagents. CB1 activation triggers protein synthesis (not shown) and leads to inhibition of Rac1 which causes disassembly of the Arp2/3-WASF1 complex. Arp2/3 is degraded leading to actin remodeling. Actin dynamics are required for CB1-iLTD. NSC inhibits Rac1-GEF interaction. CK-666 stabilizes the inactive conformation of Arp2/3, preventing it from binding actin filaments. JSK stabilizes actin filaments and promotes polymerization.

The Rac1 GTPase is one of the principal regulators of actin polymerization via WASF1/WAVE1 and Arp2/3 activity (Derivery and Gautreau, 2010; Stradal and Scita, 2006). To test the role of this pathway, we inhibited Rac1 activity using NSC 23766 (NSC), an inhibitor of Rac1-GEF interaction (Gao et al., 2004). CB1-iLTD was impaired by application of NSC (30 µM, 25 min) during induction (Figure 3C). NSC alone transiently suppressed inhibitory transmission (Figure 3D), unlike excitatory transmission (Hou et al., 2014), and this effect was associated with a decrease in PPR (Figure 3—figure supplement 1C), suggesting Rac1 activity regulates GABA release. To directly test the role of Arp2/3 in CB1-iLTD we utilized CK-666 (100 µM), a compound that inhibits Arp2/3-mediated actin assembly by stabilizing the inactive conformation of Arp (see Figure 3G; Basu et al., 2016; Hetrick et al., 2013). CK-666 bath application enhanced CB1-iLTD (Figure 3E) suggesting that Arp2/3 participates in CB1-iLTD. Unlike NSC, CK-666 had no effect on basal inhibitory transmission (Figure 3—figure supplement 1D), presumably because the inhibitor stabilizes the inactive (unbound) Arp2/3, but does not affect the Arp2/3 bound to actin filaments. ARPC2 protein, an essential component of the Arp2/3 complex, was reduced upon CB1 activation in hippocampal neuron cultures (Figure 3F). We speculate that during normal CB1-iLTD, CB1 activation-mediated Rac1 inhibition leads to removal of Arp2/3 from actin branches, and the subsequent degradation of Arp2/3 (Figure 3G). The enhancement of CB1-iLTD by CK-666 application probably occurs because the unbound Arp2/3 that has not been degraded following CB1 activation becomes inhibited and cannot maintain actin branches, thereby resulting in further depolymerization. Together, our findings suggest that Rac1 signaling and loss of Arp2/3 likely underlie the actin remodeling required for functional and structural CB1-iLTD (Figure 3G).

CB1-iLTD requires ubiquitination, but not degradation by the proteasome

The simplest interpretation of our findings is that CB1-induced degradation of ARPC2 and WASF1/WAVE1 led to impaired actin remodeling and reduced presynaptic bouton size (see Figure 2). Congruent with this idea, presynaptic release machinery and cytomatrix proteins were consistently downregulated (Figure 4A), whereas proteins involved in the ubiquitin/proteasome system (UPS) were upregulated (Figure 4B). We confirmed that presynaptic proteins identified in the SILAC screen, Munc18-1, Synapsin-1, and α-Synuclein, were significantly reduced by WIN (25 min, 5 µM) in hippocampal cultures (Figure 4C), suggesting rapid protein degradation upon CB1 activation. To test whether presynaptic proteins are downregulated locally in acute hippocampal slices, we prevented anterograde and retrograde axonal transport by incubating slices in nocodazole (1 hr, 20 µM), an agent that depolymerizes axonal microtubules (Barnes et al., 2010; Younts et al., 2016). We found that CB1 activation with WIN reduced Synapsin-1 puncta intensity in CB1+ boutons despite blockade of axonal transport (Figure 4D), as measured by immunostaining and quantitative Airyscan microscopy, consistent with local downregulation. These results suggest that CB1 activation elicits rapid downregulation of presynaptic proteins in culture and in acute slices which likely contributes to the reduction in GABA release associated with CB1-iLTD and may be mediated by degradation by the UPS.

Presynaptic proteins are rapidly reduced following CB1 activation.

(A) Combined score plot of differentially expressed proteins between vehicle vs. CB1 activation with WIN. Blue dots are manually selected presynaptic cytomatrix and release machinery proteins. SYPH: Synaptophysin; SV2A: Synaptic vesicle glycoprotein 2A; Synaptosome associated protein 91: SNAP91; WASF1: Wiskott-Aldrich syndrome protein family member 1; NSF: Vesicle-fusing ATPase; SNG1: synaptogyrin; MUNC18: mammalian homologue of UNC-18; SYUA: alpha-synuclein; SYN1: Synapsin-1; ARPC2: Arp complex subunit 2; STX1A-B: Syntaxin1A-1B; SYUB: beta-synuclein; NCKP1: Nck-associated protein 1. (B) Combined score plot of differentially expressed proteins between vehicle vs. CB1 activation with WIN. Yellow dots are manually selected ubibquitin-proteasome system-related proteins. Note that the only downregulated protein is a de-ubiquitinase. UCHL1: Ubiquitin carboxyl-terminal hydrolase isozyme L1; PSB 2,3,4,5,6: Proteasome subunit beta type-2,4,5,6; PSA 1,6: Proteasome subunit alpha type 1,6; SUMO2: Small ubiquitin-related modifier 2; PSME1: Proteasome activator complex subunit 1. (C) Left, representative Western blot images of staining for presynaptic proteins Munc18-1, Synapsin-1, and α-Synuclein and Ponceau loading control in vehicle vs. WIN-treated hippocampal cultures (5 µM, 25 min). Right, quantification of three experimental replicates normalized to Vehicle revealed a decrease in all three proteins consistent with SILAC. Munc18-1: 0.78 ± 0.09, p<0.05; Synapsin-1: 0.60 ± 0.11, p<0.01; α-Synuclein: 0.86 ± 0.01, p<0.0001, unpaired t-test, n = number of cultures. (D) Top, Airyscan confocal representative images of CB1+ boutons in acute hippocampal slices in CA1 pyramidal layer showing colocalization of CB1+ boutons (green) and Synapsin-1 (magenta). Bottom, Gardner-Altman estimation plot showing the mean difference between Control and WIN of intensity of Synapsin-1 puncta within CB1+ boutons per slice was significantly diminished by WIN application (5 µM, 25 min). Both groups are plotted on the left axes; the mean difference is plotted on a floating axes on the right as a bootstrap sampling distribution. The mean difference is depicted as a dot; the 95% confidence interval is indicated by the ends of the vertical error bar. Control: 1.0 ± 0.09, WIN: 0.73 ± 0.06 (Mean ± S.E.M.), n = number of slices (10 slices, four rats/condition). The unpaired mean difference between Control and WIN is −0.267 [95.0%CI −0.494, −0.0819]. Two-sided permutation t-test, p=0.0234.

Next, we assessed the overall contribution of the UPS to CB1-iLTD. First, to dynamically assess the UPS pathway, we measured K48-linked ubiquitinated proteins, the canonical form of ubiquitin linkage (Dantuma and Bott, 2014), following induction of CB1-iLTD in acute rat hippocampal slices in presence or absence of the specific proteasome inhibitor, MG-132. We found that both net flux, that is the amount of ubiquitinated proteins degraded by the proteasome (difference between UbK48 level when proteasomal degradation is blocked and UbK48 level under normal conditions), and the rate of degradation, (measured by the ratio of UbK48 levels between blocked and basal conditions), were significantly increased. These results suggest both a larger pool of protein to degrade as well as a faster turnover rate (Figure 5A). However, to our surprise, CB1-iLTD was unaffected by application of the proteasomal inhibitor MG-132 (5 µM) during the baseline and induction (Figure 5B). MG-132 alone had no lasting effect on basal transmission either (Figure 5—figure supplement 1A). As a positive control, MG-132 application in interleaved slices resulted in accumulation of ubiquitinated proteins (Figure 5—figure supplement 1B). Therefore, while UPS activity is increased downstream of CB1 activation, proteasomal degradation is not necessary for CB1-iLTD.

Figure 5 with 1 supplement see all
CB1-iLTD requires ubiquitination, but not degradation by the proteasome.

(A) Left, representative Western blot images of K48 polyubiquitin in hippocampal slices treated with Veh v. WIN or Veh v. WIN and MG-132. Right, top: Levels of K48 polyubiquitin following CB1 activation with WIN (5 µM, 25 min). Bottom left: UPS net flux [difference between basal (none) and proteasome blockade (MG-132) condition] is significantly increased upon CB1 activation. Control: 1.0 ± 0.04 vs. WIN: 1.74 ± 0.04, unpaired t-test, ****=p < 0.0001, n = 5 animals. Bottom right: UPS rate [ratio between MG-132 and basal condition] of K48 polyubiquitinated protein degradation is significantly increased after WIN. Control: 1.65 ± 0.08 vs. WIN: 2.31 ± 0.12, unpaired t-test, **=p < 0.01, n = 5 animals. (B) Blockade of the proteasome by bath application of MG-132 (5 µM) had no effect on iLTD. Control: 77.6 ± 4% vs. MG-132: 73.3 ± 5%; p>0.05, unpaired t-test. For all electrophysiology figures, averaged summary data expressed as normalized change from baseline ± S.E.M. and n = number of slices (s), number of animals (a). (C) Inhibiting ubiquitination with ziram (25 µM) or PYR-41 (50 µM) fully blocked iLTD. Control: 66 ± 5 vs. ziram: 99 ± 4 vs. PYR-41: 93 ± 6; F[2,19]=10.22; p<0.05, one-way ANOVA. (D) Blockade of E1 ubiquitin ligase function with ziram (25 µM, 25 min) rescued the volumetric decrease associated with CB1 activation by WIN (5 µM, 25 min). Top, representative inhibitory boutons immunolabeled with CB1 and reconstructed in 3D. Bottom, quantification of normalized mean CB1 bouton volume per slice. Control: 1.0 ± 0.04 vs. WIN: 0.81 ± 0.03 vs. WIN + ziram: 0.95 ± 0.03 vs. ziram: 1.0 ± 0.03 (mean ± S.E.M.); F[3,30]=8.11, p=0.0004, one-way ANOVA with post-hoc Tukey test for multiple comparisons. Control vs. WIN: CI[0.07,0.32], p=0.0008; WIN vs. WIN + ziram: CI[−0.27,–0.025], p=0.014; WIN vs. ziram: CI[−0.32,–0.063], p=0.0018. n = number of slices (three images/slice, one slices/rat, nine rats/condition for Control, WIN, WIN + ziram, seven rats/condition for ziram). For all structural plasticity figures, data are presented as points representing mean bouton volume per slice with a horizontal black line representing the mean per condition and the vertical line indicating 95% confidence interval (CI). 95% CIs are given as CI[lower CI, upper CI]. p-Values are exact. Refer to Figure 5—figure supplement 1E for distribution of individual bouton values. (E) Schematic summary of proposed mechanism of CB1–iLTD 1. CB1 activation rapidly engages presynaptic protein synthesis (Younts et al., 2016). 2. CB1–iLTD increases components of the ubiquitin/proteasome system and leads to enhanced proteasomal activity (Figure 4A,B and (A)) 3. Ubiquitination is required for concomitant decrease in the volume of the bouton (D) and reduction in neurotransmitter release (C). 4. CB1–iLTD involves signaling via Rac1 and Arp2/3 and actin dynamics.

Ubiquitination not only targets proteins for degradation, but can also affect their localization and function (Hamilton and Zito, 2013). We analyzed ubiquitination sites on a subset of proteins that were decreased by CB1 activation and found that most ubiquitination sites (~60%) were located in protein-protein or protein-membrane interaction domains (Figure 5—figure supplement 1C; Akimov et al., 2018), indicating that ubiquitination of these proteins could impact their function. We hypothesized that perhaps protein ubiquitination itself, independent of degradation, may play a role in CB1-iLTD. Using two structurally and mechanistically distinct E1 Ubiquitin ligase inhibitors, ziram and PYR-41 (Rinetti and Schweizer, 2010), we directly tested whether ubiquitination was required for CB1-iLTD and found that bath application of ziram or PYR-41 blocked CB1-iLTD (Figure 5C) but had no significant effect on basal transmission (Figure 5—figure supplement 1D). Moreover, inhibition of ubiquitination also blocked the CB1-mediated decrease in CB1+ bouton volume (Figure 5D, Figure 5—figure supplement 1E). In summary, CB1-iLTD leads to increases in UPS proteins and is associated with functional increases in proteasomal activity (Figure 5E). Ubiquitination is required for protein synthesis-dependent structural and functional changes of CB1+ boutons. CB1-iLTD is associated with decreases in presynaptic and cytomatrix proteins, including ARPC2 and WASF1/WAVE1, and relies on actin dynamics. However, while proteasomal activity increases by CB1 activation, only protein ubiquitination is required for structural and functional CB1-iLTD.

Discussion

We discovered that CB1-iLTD involves structural changes of the presynaptic bouton that require protein synthesis. We identified the proteins that are up- and downregulated following CB1 activation. Increased proteins are implicated in protein synthesis, processing and degradation, whereas decreased proteins are implicated in presynaptic structure, including ARPC2 and WASF1/WAVE, and function. CB1-iLTD involved actin remodeling, Rac1 and Arp2/3 signaling. Unexpectedly, we found that protein ubiquitination, but not proteasomal degradation, is responsible for structural and functional CB1-iLTD. Together, these findings point to a mechanism by which inhibitory presynapses can control their strength in response to CB1 activation via rapid proteostatic regulation of presynaptic structural change.

Presynaptic structural changes in CB1-iLTD

While structural changes are part and parcel of postsynaptic forms of plasticity (Bramham, 2008; Nakahata and Yasuda, 2018), and changes associated with plasticity are thought to be coordinated across the synaptic cleft, the involvement of structural changes of the presynaptic terminal in forms of long-term presynaptic plasticity are less clear. Here, we provide evidence for long-term structural changes at mature CB1+ terminals associated with CB1-iLTD. Previous work showed CB1 receptor activation can trigger ultrastructural changes in vesicle distribution associated with short-term CB1-mediated plasticity (García-Morales et al., 2015; Ramírez-Franco et al., 2014), collapse of axonal growth cones (Berghuis et al., 2007), and inhibitory bouton formation in response to strong postsynaptic excitation (Hu et al., 2019). Our data show that transient activation of CB1 receptors leads to a long-term reduction of the presynaptic CB1+ compartment volume in somatic synapses onto CA1 pyramidal cells. Our findings (Figure 3) are consistent with a previous study showing that CB1 receptors regulate actin dynamics in growth cones by directly interacting with Rac1 (Njoo et al., 2015), a Rho GTPase (Mattheus et al., 2016). By directly binding CB1, Rac1 can localize the WRC which consists of WASF1/WAVE1, Cyfip1, Nap1, Abi and HSP300, at the plasma membrane (Chen et al., 2010; Eden et al., 2002). The WRC is intrinsically inactive at rest (Derivery et al., 2009), but upon recruitment to the membrane by Rac1 the cytoplasmic side is opened for binding to Arp2/3 and actin (Chen et al., 2010; Eden et al., 2002), leading to dissociation of WASF1. Although the exact mechanism of degradation has not been shown for Arp2/3 or WASF1/WAVE1, the non-neuronal, structurally homologous isoform WAVE2 was demonstrated to undergo activation-dependent dissociation from the WRC ubiquitination and proteasomal degradation (Joseph et al., 2017). Therefore, Rac1 could be required for the degradation of WASF1/WAVE1, but its direct involvement in presynaptic degradation of Arp2/3 is not as clear. Our strategy of bath application of pharmacological inhibitors does not allow us to rule out an extra-presynaptic effect.

Other studies have suggested alternative signaling pathways by which structural changes occur downstream of CB1 activation. For instance, atypical coupling of CB1 to G12/13 proteins reportedly engages Rho-associated kinase (ROCK) signaling to the actomyosin cytoskeleton (Roland et al., 2014), whereas in another example, β-integrin signaling to ROCK has been implicated in a cannabinoid-mediated form of LTP (Wang et al., 2018a). Therefore, the signaling pathways and structural changes involved downstream of CB1 receptors seem to be synapse and cell-type-dependent (Steindel et al., 2013). Whether CB-LTD at other synapses shares similar mechanisms remains unclear.

We showed that CB1-iLTD involves a rapid reduction in the size of CB1 boutons and the volume of active zones that requires both actin dynamics and protein synthesis (Figure 1). CB1 bouton size was measured by immunolabeling the CB1 receptor, which, due to the extremely high density and of CB1 receptors expressed on presynaptic boutons in the CA1 region of the hippocampus, has been shown to be a reliable approximation of bouton volume (Dudok et al., 2015). Although studies in cultured neurons suggest CB1 internalization occurs (Coutts et al., 2001; Hsieh et al., 1999; Jin et al., 1999; Tappe-Theodor et al., 2007), little is known about the internalization of CB1 receptor in the intact brain, which may require longer timescales than the brief CB1 activation (25 min) we have used here (Dudok et al., 2015; Thibault et al., 2013). It is unlikely that CB1 internalization can account for the volumetric change since the high density of CB1 means that a large proportion of receptors would have to be internalized to affect measurement of the bouton size (Thibault et al., 2013). Moreover, the CB1-mediated presynaptic shrinkage is blocked by protein synthesis inhibition whereas internalization is not protein synthesis-dependent (Hsieh et al., 1999) and is also detectable at the sub-presynaptic compartment level with Bassoon and VGAT labeling (Figure 1; Figure 1—figure supplements 1 and 2), two well-established markers of presynaptic structure. Intriguingly, VGAT size showed a marked reduction in CB1+ boutons, which could reflect synaptic vesicle redistribution, consistent with previous reports showing short-term CB1 activation-induced vesicle redistribution and changes in numbers of docked/primed vesicles (García-Morales et al., 2015; Ramírez-Franco et al., 2014). While the exact functional implications of presynaptic bouton and active zone shrinkage are yet unclear, associated changes in neurotransmitter release may result from altered Ca2+ channel coupling distance with the active zone (Nakamura et al., 2015), less area for vesicular release, shift from multi- to univesicular release mode (Aubrey et al., 2017), or reorganization of transsynaptic nanocolumns (Chen et al., 2018; Glebov et al., 2017).

Protein synthesis in presynaptic CB1-iLTD likely regulates diverse cellular processes

We have recently reported that protein synthesis is required for CB1-iLTD (Younts et al., 2016). Using a well-established unbiased proteomics approach in primary hippocampal neuron cultures (Figure 2; Jordan et al., 2006; Zhang et al., 2011; Zhang et al., 2012), we identified what proteins are synthesized upon CB1 activation and thus can mediate CB1-iLTD. Despite the fact that we used whole neuron lysates, our data revealed results highly consistent with other studies that isolated presynaptic mRNAs (Bigler et al., 2017; Hafner et al., 2019; Ostroff et al., 2019; Shigeoka et al., 2016), specifically, an enrichment of mRNAs encoding for initiation factors and ribosomal proteins. We found that CB1 activation significantly upregulated proteins involved in protein synthesis and processing pathways (Figure 2). This increase in initiation factors and ribosomal proteins suggests that plasticity likely triggers an enhanced translational capacity which is consistent with our previous findings using Fluorescent Noncanonical Amino acid Tagging (FUNCAT) (Younts et al., 2016). mRNAs for Arp2/3, WAVE1, and β-actin have been detected in axonal preparations and their synthesis may be important for axon formation (Donnelly et al., 2013; Spillane et al., 2012; Wong et al., 2017). However, in CB1-iLTD, protein synthesis seems to mediate the change in presynaptic structure via the UPS, rather than direct synthesis of actin regulators. What then is being synthesized to mediate the change in structure and function? Our finding of downregulation of many presynaptic proteins and upregulation of components of the UPS suggested to us that perhaps this upregulation represents an activity-dependent synthesis of regulatory elements in the ubiquitination and proteasomal degradation pathway.

Although such a mechanism requires greater coordination, it would also reduce energy expenditure over time, that is if a presynaptic terminal will not be releasing neurotransmitter for an extended period of time (hours to days) it makes sense to degrade and recycle the release machinery, to reduce energy production, and to shrink the terminal to make space for new growth. Protein synthesis is likely necessary for the coordination and engagement of these structural, metabolic, and degradative processes. We measured changes in the protein landscape that occur fairly rapidly after CB1-iLTD induction (25 min) given that CB1-iLTD was dependent on protein synthesis during this time window (Younts et al., 2016). It is likely that additional ‘plasticity-related’ proteins are synthesized or degraded in the hours that follow iLTD induction.

Protein degradation and presynaptic function

Proteomic analysis revealed a population of downregulated proteins involved in presynaptic function and structure, as well as energy metabolism (Figures 1 and 4). In contrast, components of the protein degradation pathway, including proteasomal subunits, E2 ubiquitin ligases, and degradative enzymes, were upregulated perhaps reflecting on-demand synthesis which could regulate fast, local presynaptic protein degradation. While activity-dependent local presynaptic synthesis of UPS machinery has never been demonstrated, there is strong evidence that presynaptic function can be regulated by expression of specific E3 ubiquitin ligases and degradation of select presynaptic proteins. For example, SCRAPPER an E3 ligase has been shown to target presynaptic proteins like RIM1, synaptophysin, and Munc18 (Yao et al., 2007), amongst other proteins with high similarity to those identified in our SILAC dataset. Moreover, SCRAPPER KO has been shown to impair neurotransmitter release, short-term plasticity and presynaptic LTP (Koga et al., 2017; Yao et al., 2007). The tight coupling of translation and degradation in the context of synaptic plasticity was described previously (Klein et al., 2015) and is presumed to occur widely in the brain as a means of maintaining proteostasis over the course of plastic changes (Biever et al., 2019; Dong et al., 2008; Hanus and Schuman, 2013; Wang et al., 2017). This rapid activity-dependent degradation could be mediated by the recently discovered neuron-specific proteasome complex (NMP) (Ramachandran et al., 2018), although this complex is believed to target non-ubiquitinated substrates. Protein degradation has also been shown to regulate presynaptic silencing, specifically by degradation of presynaptic proteins such as RIM1 and Munc13 (Jiang et al., 2010) and CB1-mediated suppression of transmission at excitatory synapses via degradation of Munc18 (Schmitz et al., 2016; Wang et al., 2018b). Homeostatic plasticity in response to neuron silencing has also been associated with degradation of select presynaptic proteins (Lazarevic et al., 2011). However, our data strongly support the idea that protein degradation by the UPS is not directly required for CB1-iLTD, but likely does occur quickly after CB1 activation, as indicated by rapid loss of presynaptic proteins measured with both SILAC and western blot (Figure 4). This rapid reduction in protein levels presumably occurs as a consequence of enhanced UPS activity (Figure 5), but we cannot exclude that a different mechanism could be involved including reduced synthesis (Dörrbaum et al., 2020) or other protein degradation pathways, such as autophagy (Liang and Sigrist, 2018).

We demonstrated that ubiquitination is required for CB1-iLTD, likely by controlling the trafficking, interactions, or the activity of its substrates, upstream of degradation (Hamilton and Zito, 2013). A previous study showed that inhibition of protein ubiquitination and degradation increased miniature EPSCs/IPSCs in cultured neurons, suggesting an important role for these processes in maintaining normal neurotransmitter release (Rinetti and Schweizer, 2010). However, in our hands, proteasomal inhibitor MG-132 had no significant effects on basal synaptic transmission and neither did E1 ubiquitin ligase inhibitors, ziram and PYR-41. To our knowledge, our study is the first to describe a mechanism of long-term presynaptic structural and functional plasticity that relies on ubiquitination. The regulation of presynaptic ubiquitination is likely achieved through the targeted expression of different E2 and E3 ubiquitin ligases (Hallengren et al., 2013; Koga et al., 2017; Yao et al., 2007) or via presynaptic cytomatrix proteins themselves (Chen et al., 2003; Ivanova et al., 2016; Waites et al., 2013). This raises the possibility that presynaptic structural dynamics and UPS activity are tightly linked.

Potential relevance in the normal and diseased brain

CB1 activation via eCBs, as well as exogenous cannabinoids like Δ9- tetrahydrocannabinol (THC), the primary psychoactive ingredient in marijuana, can influence cognition, goal-directed behaviors, sensory processing and other critical brain functions (Araque et al., 2017; Augustin and Lovinger, 2018; Häring et al., 2012; Heifets and Castillo, 2009; Hoffman and Lupica, 2013; Zlebnik and Cheer, 2016). Cannabinoid signaling has also been implicated in several brain disorders (Zou and Kumar, 2018). Autism is broadly associated with changes in synaptic protein levels, but also disruption of CB1-LTD (Busquets-Garcia et al., 2014; Chakrabarti et al., 2015). In particular, in a mouse model of Fragile X Syndrome (FXS), the most common monogenic cause of autism, where RNA-binding protein, Fragile X mental retardation protein (FMRP) is deleted (Bagni and Zukin, 2019), eCB-mediated plasticity in the hippocampus, striatum, prefrontal cortex is impaired (Jung et al., 2012; Maccarrone et al., 2010; Martin et al., 2017; Wang et al., 2018b; Zhang and Alger, 2010). Although changes in the eCB mobilization in FXS may explain some of the impairment (Jung et al., 2012; Maccarrone et al., 2010; Zhang and Alger, 2010), the role of FMRP in the regulation of local presynaptic protein synthesis may also play a role (but see Jung et al., 2012), although this remains to be tested (Busquets-Garcia et al., 2013). Many neurodegenerative diseases are characterized by imbalanced proteostasis, including Alzheimer’s disease (AD) and Parkinson’s disease (PD), resulting in pathological accumulations of misfolded proteins (Klaips et al., 2018). WIN treatment in animal models of AD and PD has been shown to be neuroprotective and to alleviate cognitive and motor symptoms (Basavarajappa et al., 2017), potentially through the ability of the CB1 receptor to regulate synaptic proteostasis.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional
information
Biological sample wildtype, Sprague Dawley, Rattus norvegicus, male and femalePrimary hippocampal neuron culturesCharles RiverIsolated from DIV 20–21 pups
Biological sample wildtype, Sprague Dawley, Rattus norvegicus, male and femaleAcute hippocampal slicesCharles RiverIsolated DOB 18–25 rats
AntibodyCB1 (rabbit polyclonal)ImmunoGenesCat# CB1, RRID:AB_28138231:1000
AntibodySynapsin 1 (mouse monoclonal)Synaptic SystemsCat# 106 011C2, RRID:AB_108051391:1000
AntibodyvGAT (mouse monoclonal)Synaptic SystemsCat# 131 011C3, RRID:AB_8878681:500
AntibodyBassoon (mouse
monoclonal)
Enzo Life SciencesCat# ADI-VAM-PS003, RRID:AB_106187531:1000
AntibodyParavalbuminSigma-AldrichCat# P3171, RRID:AB_23138041:1000
Antibodyα-SynucleinBD BiosciencesCat# 610787, RRID:AB_3981081:1000
AntibodyMunc-18–1Synaptic SystemsCat# 116 002, RRID:AB_8877361:1000
AntibodyArp2/3NovusCat# NBP1-88852, RRID:AB_110404641:1000
AntibodyUbiquitin K48MilliporeCat# 05–1307, RRID:AB_15875781:1000
SoftwareIgor ProIGOR ProRRID:SCR_000325

Immunohistochemistry and microscopy

Request a detailed protocol

Acute rat hippocampal slices were made as described below for electrophysiological recordings and allowed to recover for at least 1 hr after slicing. Slices were incubated in beakers containing ACSF and drug treatments described in Results and underwent constant oxygenation. Slices were fixed immediately after treatments in 4% PFA in PBS overnight at RT. Slices were washed twice in PBS then incubated in blocking buffer (4% BSA in PBS + 0.1% Tx-100) for 1 hr at RT. Primary antibodies (CB1, 1:1000, Immunogenes (Budapest, Hungary)); Synapsin-1 1:1000 Synaptic Systems (Goettingen, Germany); Bassoon, 1:1000, Enzo Life Sciences (Farmingdale, NY); Paravalbumin, 1:1000, Sigma Aldrich; VGAT, 1:500, Synaptic Systems were diluted directly into the blocking buffer and floating slices were incubated overnight at 4C. After four washes with PBS, slices were incubated in secondary antibodies (Invitrogen) diluted in blocking buffer overnight at 4°C. Slices were washed 5X with PBS, then mounted. Images were acquired on a Zeiss LSM 880 with Airyscan using a Plan-Apochromat 63x/1.4 Oil DIC M27 and 1.8X zoom. Images were Airyscan processed prior to analysis. Pixel width and height was 0.049 µm and voxel depth was 0.187 µm. Imaris 9.2 software was used to reconstruct boutons in 3D using the Surface function. Threshold, laser power, and gain were kept constant for each experiment. CB1 boutons were screened after 3D reconstruction to ensure correct identification. Only boutons that fell between 0.05–5 µm3, did not touch the image border, and had a sphericity value above 0.3 were considered. For Bassoon (Figure 1F) and VGAT (Figure 1—figure supplement 2D), FIJI was used to remove all Bassoon signal that did not overlap with CB1 labeling by creating a dilated binary mask of CB1 labeling then using the Image Calculator ‘AND’ function to create a mask for the non-CB1 channel, then this mask was used to isolate signal in the non-CB1 channel, then Imaris was used to measure the volume of the non-CB1 channel. FIJI was used to analyze synapsin puncta (Figure 4D) inside CB1 boutons by creating a dilated binary mask of CB1 labeling then using the Image Calculator ‘AND’ function to create a mask for the non-CB1 channel, followed by the ‘Analyze Particles’ function to determine the intensity or percent overlap of the two channels. All imaging and analysis were performed blind to treatment group.

SILAC

Request a detailed protocol

Primary hippocampal neurons were prepared from E18-19 rat brains and grown on poly-D-lysine coated 15 cm plates at a density of 3.5 million cells/ 15 cm plate in DMEM media without l-arginine or l-lysine (Cambridge Isotope Laboratories, Tewksbury, MA, USA, Cat# DMEM-500), with pen/strep and B-27 supplement (Invitrogen, Carlsbad, CA, USA). 84 mg/L of l-arginine 13 C6 (Sigma Aldrich, St. Louis, MO) and 146mg/L of l-lysine D4 (Thermofisher, Waltham, MA) was supplemented for ‘medium’ labeled media and 84 mg/L of l-arginine 13 C615N4 and 146 mg/L of l-lysine 13 C615N4 was added to ‘heavy’ labeled media. Neurons were grown in this media for 15 days, which results in >90% incorporation of labeled amino acids into the cellular proteomes (Zhang et al., 2011; Zhang et al., 2012). For treatment, neuron cultures DIV 16 received 15 ml fresh media containing either vehicle or WIN (5 µM) for 25 min. Neurons were washed 3X with ice-cold PBS without Mg2+ or Ca2+ (0.01 M, pH = 7.4). After three washes, cells were harvested and lysed in SDS lysis buffer containing: (50 mM Tris, 2% SDS, 2 mM EDTA) for 30 min at RT. Lysates were then sonicated briefly, allowed to incubate for another 30 min at RT, and centrifuged for 5 min at 15,000 g to remove insoluble debris. 10 µg of lysate from ‘medium’ cells treated with vehicle were mixed with 10 µg ‘heavy’ WIN-treated cells (Forward sample). Separately, 10 µg of lysate from ‘medium’ cells treated with WIN were mixed with 10 µg ‘heavy’ vehicle-treated cells (Reverse sample). These mixtures (20 µg total protein) were loaded onto a 10% Bis-Tris gel and subjected to SDS-PAGE. The gel was stained with Coomasie for 1 hr and protein lanes were cut up into 12 equal sized portion in order to improve protein coverage (Jordan et al., 2004). Mass spectrometry was performed in collaboration with the Einstein Proteomic Facility using the Orbitrap Velos. In Mascot the Quantitation Method (SILAC K+4 K+8 R+6 R+10) was used with each SILAC modification in exclusive mode. Those listed as variable were for determining incomplete protein labeling: 2H(4) (K); 13C(6) (R); 13C(6)15N(2) (K); and 13C(6)15N(4) (R). The raw data files were first processed using precursor ion quantitation of the Quantitation toolbox of Mascot Distiller (Matrix Science Ltd; version 2.7). Mascot was then used to search the rat databases (SwissProt and NCBInr along with a decoy database to obtain FDRs) using the following parameters: trypsin; product ion mass tolerance of 0.40 Da; precursor ion tolerance of 50 PPM; carbamidomethyl Cys - fixed modification and variable modifications of: deamidated Asn and Gln; label:2H(4) of Lys; label:13C(6) of Arg; label:13C(6)15N(2) of Lys; label:13C(6)15N(4) of Arg and oxidation of Met. The result files obtained from Distiller and the Mascot searches were then uploaded to Scaffold Q+S (Proteome Software Inc; version 4.9) using between-subjects, log2 ratio-based analysis of unique peptides against a reference to obtain the protein’s mean quantitative values and t-test p-values with Benjamini-Hochberg correction. Peptide identifications were accepted if they could be established at greater than 95.0% probability by the Scaffold Local FDR algorithm. Protein identifications were accepted if they could be established at greater than 99.0% probability and contained at least two identified peptides. Protein probabilities were assigned by the Protein Prophet algorithm (Nesvizhskii et al., 2003). Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were grouped into clusters. The mass spectrometry proteomics data have been deposited to the ProteomeX change with identifier Consortium via the PRIDE [1] partner repository with the dataset identifier PXD020008 and 10.6019/PXD020008.

Gene ontology analysis

Request a detailed protocol

SILAC results were ranked according based on fold change and submitted to a GSEA Preranked analysis in GSEA (v. 4.0.2) with 1000 permutations. Terms smaller than 15 genes or bigger than 500 were discarded as previously reported (Merico et al., 2010). The enrichment map was generated in Cytoscape (3.7.1) (Kucera et al., 2016) using Enrichment map plugin (3.2.0) (Merico et al., 2010) using the following thresholds: p value < 0.05, FDR < 0.001. The overlap coefficient was set to 0.5. For confirmation, we also performed Gene Ontology analysis using two other tools. First, filtered lists (|log2 fold change| > 0.5) were analyzed through the use of IPA (QIAGEN Inc, Hilden, Germany, https://www.qiagenbioinformatics.com/products/ingenuitypathway-analysis). Then, we performed ontology enrichment using a recently published expert-curated knowledge database for synapses (Koopmans et al., 2019). Terms were selected with a FDR < 0.01. The parental term ‘Synapse’ was discarded as not being informative (e.g. to general).

Electrophysiology slice preparation and recording

Request a detailed protocol

Experimental procedures adhered to NIH and Albert Einstein College of Medicine Institutional Animal Care and Use Committee guidelines. Acute transverse slices were prepared from young adult male and female Sprague Dawley rats (P18-27). The cutting solution contained (in mM): 215 sucrose, 20 glucose, 26 NaHCO3, 4 MgCl2, 4 MgSO4, 1.6 NaH2PO4, 2.5 KCl, and 1 CaCl2. The artificial cerebral spinal fluid (ACSF) recording solution contained (in mM): 124 NaCl, 26 NaHCO3, 10 glucose, 2.5 KCl, 1 NaH2PO4, 2.5 CaCl2, and 1.3 MgSO4. After ice-cold cutting, slices recovered at RT (in 50% cutting solution, 50% ACSF) for <30 min and then at room temperature (RT) for 1 hr in ACSF. All solutions were bubbled with 95% O2 and 5% CO2 for at least 30 min. Although the form of long-term inhibitory synaptic plasticity studied here (i.e. iLTD) is present under physiological recording conditions at 37°C (Younts et al., 2013), inhibitory synaptic transmission is less stable at this temperature, and therefore we conducted our experiments at 25.5 ± 0.1°C.

For extracellular field recordings, a single borosilicate glass stimulating pipette filled with ACSF and a glass recording pipette filled with 1M NaCl were placed approximately 100 µm apart in stratum pyramidale. To elicit synaptic responses, paired, monopolar square-wave voltage or current pulses (100–200 μs pulse width) were delivered through a stimulus isolator (Isoflex, AMPI) connected to a broken tip (∼10–20 μm) stimulating patch-type micropipette filled with ACSF. Typically, stimulating pipettes were placed in CA1 stratum pyramidale (150–300 μm from the putative apical dendrite of the recorded pyramidal cell, 150–200 μm slice depth). Stimulus intensity was adjusted to give comparable magnitude synaptic responses across experiments less than ∼0.6 mV. Inhibitory synaptic transmission was monitored in the continuous presence of the NMDA receptor antagonist d-(-)−2-amino-5-phosphonopentanoic acid (d-APV; 25 μM), the AMPA/kainate receptor antagonist 2,3-dihydroxy-6-nitro-7-sulfonyl-benzo[f]quinoxaline (NBQX; 5 μM), and the μ-opioid receptor agonist, [D-Ala2, N-MePhe4, Gly-ol]-enkephalin (DAMGO, 50 nM). To elicit chemical-iLTD (Heifets et al., 2008) the CB1 agonist WIN 55,212–2 (WIN; 5 µM) was bath applied for 25 min, and 5 stimuli at 10 Hz were delivered at 0.1 Hz during the last 10 min of WIN. WIN was chased with the CB1 inverse agonist/antagonist SR 141716 (5 µM) or AM251 (5 µM) to halt CB1 signaling. Baseline and post-induction synaptic responses were monitored at 0.05 Hz during iLTD. Stimulation and acquisition were controlled with IgorPro 7 (Wavemetrics). Shaded boxes in figures correspond to when plasticity was analyzed with respect to baseline and when representative traces were collected and averaged. Summary data (i.e. time-course plots and bar graphs) are presented as mean ± standard error of mean (S.E.M.). PPR was defined as the ratio of the amplitude of the second EPSC (baseline taken 1–2 ms before the stimulus artifact) to the amplitude of the first EPSC. The magnitude of LTD was determined by comparing 20 min baseline responses with responses 80–100 min post-LTD induction.

Western blotting

Request a detailed protocol

Protein concentration was determined using the Lowry method with bovine serum albumin as a standard (Lowry et al., 1951). Primary hippocampal neurons or hippocampal slices were solubilized on ice with RIPA buffer (1% Triton X-100, 1% sodiumdeoxycholate, 0.1% SDS, 0.15MNaCl, 0.01Msodium phosphate, pH7.2) followed by sonication. Immunoblotting was performed after transferring SDS-PAGE gels to nitrocellulose membrane and blocking with 5% low-fat milk for 1 hr at room temperature. The proteins of interest were visualized after incubation with primary antibodies (α-synuclein 1:1000 BD Biosciences (San Jose, CA) #610787; Synapsin-1 1:1000 Synaptic System (Goettingen, Germany) #106001; Munc18–1 1:1000 Synaptic System #116 002; Arp2/3 1:1000 Novus Biologicals (Centennial, CO) # NBP188852; Ubiquitin K48 1:1000 EMD Millipore (Burlington, MA) #05–1307) by chemiluminescence using peroxidase-conjugated secondary antibodies in LAS-3000 Imaging System (Fujifilm, Tokyo, Japan). Densitometric quantification of the immunoblotted membranes was performed using ImageJ (NIH). All protein quantifications were done upon normalization of protein levels to Ponceau staining. Ponceau normalization was chosen over comparison to actin as our work and others showed that CB1-iLTD induces modification of actin cytoskeleton.

Ubiquitination sites analysis

Request a detailed protocol

Ubiquitination sites were identified using Ubisite, a publicly available resource for ubiquitination site prediction (Akimov et al., 2018). To minimize false positive rate, confidence level was set on high. When available, functional domains were annotated using the uniport.org database (UniProt Consortium, 2019).

Data analysis, statistics and graphing

Request a detailed protocol

Analysis and statistics were carried out in OriginPro 2015 (OriginLab) and Graphpad Prism 7.02. Significance (p<0.05) was assessed with one-way ANOVA (means comparison with post hoc Tukey test), Student’s paired and unpaired t-tests, Wilcoxon matched-pairs signed rank test, Mann Whitney U test, or Pearson’s correlation coefficient, as indicated. All electrophysiology experiments were performed in an interleaved fashion –that is control experiments were performed every test experiment. Unless stated otherwise, ‘n’ represents number of field recordings in slices. All experiments include at three animals. Plotting of SynGO results was made using matplotlib (3.0.3)(Hunter, 2007) in Python (3.7. 3)(Oliphant, 2007) environment. Figure 4D used Garder-Altman estimation plots to represent effect size. Statistics and graphing were performed using estimationstats.com (Ho et al., 2019). Supplementary figures include Superplots to represent individual bouton values color-coded by slice that were created using Python (3.7.3) (Lord et al., 2020) and informed by plot design from the Superplots app (https://huygens.science.uva.nl/SuperPlotsOfData/). Slice numbers were balanced across conditions in some Superplots (in Figure 1—figure supplement 1A, Figure 3—figure supplement 1A, and Figure 5—figure supplement 1E) to make statistical comparison across similar sized groups (Lord et al., 2020), and to make it easier for the reader to compare the distribution of individual bouton values across conditions. To balance the conditions in an unbiased manner, if a condition had an unequal number of slices, the slice that had the most different number of ‘n’ was removed. Importantly, balancing the conditions did not change the interpretation of the data as plotted in the main figures and all key observations were reproduced.

Reagents

Stock reagents were prepared according to the manufacturer’s recommendation in water, DMSO (<0.01% final volume during experiments), or phosphate buffered saline (PBS), stored at −20°C, and diluted into ACSF or intracellular recording solutions as needed. CNQX, D-APV, SR 141716, and WIN 55,212–2 were acquired from the NIMH Chemical Synthesis and Drug Supply Program; salts for making cutting, ACSF, ziram, and intracellular recording solutions from Sigma Aldrich (St. Louis, MO); AM251, NSC-23766, MG-132, DAMGO, cycloheximide from Tocris Bioscience (Bristol UK); jasplakinolide, anisomycin, PYR41 from Cayman Chemical (Ann Arbor, MI); CK-666 from EMD Millipore. Reagents were either acutely bath applied, diluted into the intracellular recording solution, or preincubated with slices/cultures, as indicated in Results.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
    New Tricks for an Old Dog: Proteomics of the PSD
    1. BA Jordan
    2. BD Fernholz
    3. TA Neubert
    4. EB Ziff
    (2006)
    In: J. T Kittler, S. J Moss, editors. The Dynamic Synapse: Molecular Methods in Ionotropic Receptor Biology.  CRC Press/Taylor & Francis. pp. 37–50.
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
    Protein measurement with the folin phenol reagent
    1. OH Lowry
    2. NJ Rosebrough
    3. AL Farr
    4. RJ Randall
    (1951)
    The Journal of Biological Chemistry 193:265–275.
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84
  85. 85
  86. 86
  87. 87
  88. 88
  89. 89
  90. 90
  91. 91
  92. 92
  93. 93
  94. 94
  95. 95
  96. 96
  97. 97
  98. 98
  99. 99
  100. 100
  101. 101
  102. 102
  103. 103
  104. 104
  105. 105
  106. 106
  107. 107
  108. 108
  109. 109
  110. 110
  111. 111
  112. 112
  113. 113
  114. 114
  115. 115
  116. 116
  117. 117
  118. 118
  119. 119
  120. 120
  121. 121
  122. 122
  123. 123
  124. 124
  125. 125
  126. 126
  127. 127
  128. 128
  129. 129

Decision letter

  1. Gary L Westbrook
    Senior and Reviewing Editor; Oregon Health and Science University, United States
  2. Erin M Schuman
    Reviewer; Max Planck Institute for Brain Research, Germany

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This study reveals structural plasticity of presynaptic terminal bouton sizes and underlying molecular mechanisms for endocannabinoid induced long-term depression of inhibitory synapses (iLTD) in the hippocampus. The results establish that the well-described postsynaptic plasticity-related spine growth/shrinkage can also exist in presynaptic compartments in presynaptic forms of plasticity and provide insights into this relatively less understood aspect of synaptic plasticity.

Decision letter after peer review:

Thank you for submitting your article "CB1 receptor-mediated inhibitory LTD triggers presynaptic remodeling via protein synthesis and ubiquitination" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Gary Westbrook as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Erin M Schuman (Reviewer #1).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

The reviewers were positive about the manuscript but had a robust discussion of points that they though required additional input. The summary includes the main points that arose in the discussion. The original reviews follow. Please address both the summary and the reviews in your revised manuscript.

Summary of reviewer discussion:

1) Immunostaining for the CB1 receptor was used as a proxy for terminal size. This could be OK, but the bouton volume data should be checked critically. For example, the shrinkage of the terminal as measured by Bassoon is smaller than that measured by CB1. The authors use an agonist of CB1 throughout the paper – this will certainly change the configuration of the receptor (dimerization, clustering, internalization) that could artificially inflate the shrinkage they claim. This point should be discussed.

2) The MS data needs the tolerance value and stats. The link between the requirement for ubiquitination and protein synthesis is not as strong as it could be. The authors could add experimental data or should at least include a working model in their discussion addressing such questions as: Are both required? Do they operate in parallel or is new synthesis needed for ubiquitination?

3) Drug alone controls are missing for several conditions, particularly CHX and JSK.

4) The conclusions RE Rac and Arp need to be modified because the cellular site of action is not established in this study. The manipulations are pharmacological, and although the narrative and proteomics wants us to think about a pre-terminal action, one could imagine postsynaptic effects as these proteins are well-known for postsynaptic roles. The authors need to modify their conclusions or use tools that exist to cell specifically manipulate these proteins and establish the site of action.

Reviewer #1:

This manuscript explores a functional role for presynaptic protein synthesis in modulating structural plasticity of inhibitory presynaptic boutons during LTD. The authors provide data for protein synthesis dependent shrinkage of CB1+ boutons during endocannabinoid mediated LTD. Using proteomics, they identify 33 upregulated and 27 downregulated proteins during LTD. They narrow in and explore a role for actin remodeling during LTD, focusing on the role of Rac1 and Arp2/3, specifically the downregulation of both, for the manifestation of CB1+ boutons shrinkage. Lastly, they argue that while proteasomal degradation is induced during LTD, only the ubiquitination itself is important for the structural and functional manifestation of LTD.

The data presented on presynaptic structural plasticity are exciting but there are missing key control experiments and validations – as indicated below.

Figure 1:

While the detailed description of the analysis of the boutons was appreciated, the data presented in 1C are not convincing in their present state. The authors provide no controls to ensure that it's not endocytosis/remodeling of the CB1 receptors themselves, which would also result in a decreased CB1 volume. Additional inhibitory presynaptic markers (such as VGAT) should also be assessed and scored. Given the strict cutoffs the authors use in the analysis (ie sphericity value within the Imaris analysis), it could be that they're simply detecting a subset of smaller CB1 puncta resulting from local changes in the receptor pool; while the boutons themselves remain largely unchanged. The use of this CB1 volumetric analysis through the paper is problematic, as in almost all cases the authors do not test the effects of the pharmacological treatment alone (CHX, JSK, CK-666, NSC and ziram). Most if not all of the conclusions could be substantially undercut if these treatments in control conditions show the same change as when paired with WIN treatment. The study of structural plasticity in the postsynapse has largely relied on live cell imaging, so the compartment can be analyzed before, during and post treatment. While challenging, the authors should consider a similar approach, even if only to demonstrate their CB1 volumetric approach is valid. Using either an inhibitory neuron restricted GFP expression or expression of a tagged inhibitory presynaptic protein to allow live imaging of the bouton volume before and after WIN application, the authors could the perform fixation and staining for CB1 receptors and analyze how this compartment changed during the course of the treatments. This would bring their approach more in line with the standard in the field and allow for much more current and compelling analysis for structural plasticity to be performed.

Specific comments:

– Examples (similar as in 1A) should be provided of CB1 boutons with WIN/WIN+CHX treatment.

– CHX treatment alone should be assessed for C.

– CHX and WIN+CHX should be assessed for D.

– Did the authors look at what happens to the corresponding inhibitory postsynaptic compartments? Presumably these also may decrease in size as well.

– For Figure 1—figure supplement 1C, the indicated mean for the WIN group for the graph provided should be checked. Given the large increase in the low volume population seen in the WIN group vs control group; and the similar size of the groups (540 vs 602) it's hard to understand how both means can be right around 1.

Figure 2:

The experimental design (multiplexing of fully medium-heavy and heavy labeled cell cultures) is nice, but there are several potential issues regarding data processing and interpretation.

Protein identification and quantification:

– The accepted parent mass tolerance of 50 PPM is very high and might lead to false identifications. When acquiring the data with an MS1 resolution of 60k on a well-calibrated Orbitrap LTQ instrument, a MS1 mass tolerance of ~5 PPM should be used.

– Why are the SILAC labels (Lys4/Arg6 and Lys8/Arg10) specified as variable modifications instead of “SILAC pairs”? Please describe in more detail how the medium-heavy and heavy counterparts were quantified and how the fold-changes were calculated. Were only peptides species considered that were quantified in the medium heavy and heavy form? Why not using software that provide well-established workflows for SILAC quantification (such as the freeware MaxQuant)?

– From Supplementary file 1, we understand that each ~700 proteins were identified in the forward and the reverse samples. The overlap between the two samples is 391 proteins (main text), indicating that the reproducibility is fairly low between the measurements.

Statistics:

– Please perform proper statistics to identify differentially regulated proteins (e.g. t-test and multiple testing correction). A simple fold-change cut-off (as described in Figure 2C) is not state-of-the-art.

Specific comments:

– The supplementary data should be improved to enable the reader to understand the proteomics data and inspect all quantified proteins. Please include one table that presents the quantitative results for all proteins in all biological replicates (forward and reverse), so that the reader can get a sense for the biological variability. Please include a column with gene names. In Supplementary file 1, why does the table contain proteins from rattus norvegicus, mus musculus and human?

– All mass spec raw data and database search results should be uploaded to the PRIDE repository prior to publication and the PRIDE identifier should be included in the Materials and methods section.

Data interpretation:

– Protein up-regulation is often interpreted as increased protein synthesis and protein down-regulation is often interpreted as increased degradation in the manuscript. However, based on the SILAC-MS data one cannot draw these conclusions. Protein up-regulation could result from more synthesis or less degradation and protein down-regulation could result from less synthesis or more degradation. Based on the presented data there is no evidence for one or the other.

– Based on the proteomics data one cannot distinguish proteins from different subcellular compartment or different cell types. The following statement from the Discussion sections is hence misleading: "[…] why ribosomal proteins would be synthesized locally if ribosomes are exclusively assembled in the nucleolus is unclear."

Figure 3:

The authors refer to JSK as an "actin stabilizing drug", whereas in fact it's a drug that promotes actin polymerization. This fact complicates the data presented in this figure as essential controls (such as JSK treatment alone for 3A) are missing. If JSK treatment alone increases the bouton size, is this really demonstrating that the structural remodeling requires actin depolymerization?

Specific comments:

– 3A, the examples provided with the black background are harder to examine. The authors should use the white background used in Figure 1 and Figure 1—figure supplement 1 examples.

– 3B, what explanation do the authors have for why the +JSK/WIN treated curve stabilizes around 80%?

– Additionally, the authors should test the effect of the Arp2/3 inhibitor and Rac inhibitor with and without WIN treatment on the bouton volume.

Figure 5:

Specific comments

– 5D, examples would be better with white backgrounds.

– Analysis for the effect of ziram alone needs to be done.

Reviewer #2:

In this study by Monday and colleagues, the investigators reveal structural plasticity of presynaptic terminal bouton sizes and underlying molecular mechanisms for iLTD, a form of endocannabinoid induced long-term depression of inhibitory synapses. Their discovery establishes that the well-described post-synaptic phenomenon of spine growth/shrinkage with plasticity can also exist in presynaptic compartments in presynaptic forms of plasticity. This finding is demonstrated by volumetric bouton measurements of CB1+ terminals in hippocampal slices. The remainder of the study involves a series of elegantly and rigorously designed experiments to provide insights to the molecular mechanisms for this structural plasticity. Insights begin with a proteomic analysis performed on cultured hippocampal neurons, and then use pharmacology and Western analysis to test a main hypothesis that changes in the abundance and activity of actin regulators underlie the structural plasticity. The study continues to reveal that despite their hypothesis that proteasomal degradation could drive those shifts, experiments instead support the idea that ubiquitination, in the absence of degradation (assessed by MG-132) appears to underlie the functional differences required for iLTD.

Overall, this study provides an important extension to the Castillo lab's recent work demonstrating the requirement for local protein synthesis for this form of plasticity (Younts et al., 2016). Strengths of this study include the mechanistic depth, general rigor in approaches, and novel insights to a relatively less understood aspect of synaptic plasticity.

I have two major concerns that should be addressed.

1) Boutons simply should not be considered individual n's for statistics. The authors are welcome to argue for what the independent biological unit is. Please also clarify, for a given condition (such as Veh or WIN), were slices from the same animal used for multiple conditions, or were the slices from one animal used for one condition at a time? If the latter, then an animal is the biological unit, with the possibility of using nested data statistical approaches. Alternatively, the mean value/slice could be considered. Issues such as animal health, temperature fluctuations, buffer osmolarity et cetera could otherwise exist and generate 100s of values per slice that are a little bigger or smaller due to technical variables. This is a foundational observation for the study and needs to be robust. As presented, it looks small and only visible when n's are in the 1000s. Likewise, although it was good to consider involvement of other inhibitory synapses (like PV+), a figure showing no significant difference that is potentially grossly underpowered is not so useful (n's of 500).

2) In general, the bioinformatics approach was rigorously and thoroughly presented. It is also nice that the authors further discuss results from other programs. One item to clarify – since the GO analyses are the major data presented: please review and clarify the extent to which the same protein is annotated with multiple functional properties such that it is populating (and driving) multiple terms on those lists.

Reviewer #3:

In this manuscript the authors show that CB1-LTD involves presynaptic structural plasticity: bouton volumes are reduced, which occurs at least partially via actin reorganization.

Mass spec data showed results that may have been expected (downregulation of synaptic proteins), but also some unexpected results were found (upregulation of protein synthesis as well as protein degradation system). Finally, the authors show that activation of the ubiquitination system is required for CB1-LTD, however not by promoting protein degradation by the proteasome.

The manuscript contains an interesting data set and the involvement of the ubiquitination system in CB1-LTD is highly intriguing. However, after reading the manuscript there are many “lose ends”, and it is still not clear what happens during CB1-LTD. Indeed, the authors themselves describe these events as “coordinated engagement of multiple cellular processes”, which does not clarify much. I would like to encourage the authors to try and formulate a more complete hypothesis of the cellular events that underlie CB1-LTD.

I have two major points:

1) It is not easy to link the proteomics findings to changes in the actin cytoskeleton in presynaptic boutons. For instance, the authors show that WIN slightly reduces Arp2c protein levels, but that inactivating Arp2/3 actually enhances CB1-LTD. I am wondering if the explanation given by the authors the most plausible interpretation of the data. Is it known how CB1-LTD is expressed (fewer vesicles? reduced release probability?) and is it possible to link this to specific changes in the actin cytoskeleton?

2) The proteomics data show that multiple processes are activated during CB1-LTD. The manuscript does not provide a model or hypothesis for how these processes are linked. For instance, are ubiquitination and protein synthesis independently regulated by CB1? What is the link between ubiquitination and presynaptic actin remodeling? Would it be possible to link CB1-LTD to a specific E2 or E3 ligase in the presynapse?

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your revised article " CB1 receptor-mediated inhibitory LTD triggers presynaptic remodeling via protein synthesis and ubiquitination" to eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by Gary Westbrook as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Erin M Schuman (Reviewer #1).

The reviewers have discussed the reviews with one another and the Senior Editor. There are some remaining concerns raised by one of the reviewers and by me that we would like you to address as below. These are concerns about the data presentation as embodied in comment 1 of reviewer 2 in the original review. The issue concerns what constitutes "n" in the bouton counts. We suggest a compromise in which supplemental data are shown by slice for full data transparency. Overall, we suspect that artifact is not the basis of the findings, but the data presentation does not allow that assessment and therefore lowers the rigor and potential reproducibility.

Requested revisions

1) The authors in their rebuttal argue that standard power calculations are "unreasonable" because n of 145 slices would be required. At the most fundamental level, if this were true, this question and conclusion are not appropriate given the constraints of the methodology and one should focus on powerable questions. However, a statistical analysis by a statistician might indicate other analytical approaches. The authors show how close to being powered they may be with the slice means alone (not including nesting) showing significance. If the authors want to show the data as is in the main figure, the authors could address this concern by creating supplemental figures that show bouton size distributions for each slice and also figures with slice means and exact p values.

2) Bassoon puncta were to serve as an independent measure of bouton size to support the CB1 observations. However, the methods describe first making a binary mask of CB1 signal, then using the "and" function to isolate the bassoon pixels that overlap with CB1 pixels. As described, it is almost impossible to reach any other conclusion than reproducing the CB1 measurement results. Please clarify methods substantially to remove this concern.

3) In working to propose a solution for data presentation, the authors raise an important point, supplemental data should include an analysis that addresses whether the plasticity effects preferentially involve a particular size subset, smaller or large.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article " CB1 receptor-mediated inhibitory LTD triggers presynaptic remodeling via protein synthesis and ubiquitination" for consideration by eLife. Your article has been reviewed by one peer reviewer, and the evaluation has been overseen by Gary Westbrook as the Senior Editor. Before final acceptance, please edit the text and Materials and methods to indicate your response to the final reviewer comments as below. I would appreciate it if you indicate in the text where the revisions have been made. The final submission will not go back to the reviewers.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

Reviewer #2:

1) The supplementary figures are highly appreciated and allow critical review of the nature of the experimental data and robustness of the conclusions.

2) The authors' response indicating that data was removed to balance slice numbers needs to be clarified. One should not have to remove data. If slices were removed from supplementary figure analyses, what guided those decisions? Do the removed slices remain in the main figure data? I would suggest all the data that were believed to be valid in the original submission be included in the main and supplementary figures. (Or was there a situation of grossly unequal n's in the control/win condition and the authors are trying to make statistical comparisons across similarly sized groups post hoc? In principle, a similar number of interleaved controls should exist for each additional test condition.) Please clarify.

https://doi.org/10.7554/eLife.54812.sa1

Author response

Summary of reviewer discussion:

1) Immunostaining for the CB1 receptor was used as a proxy for terminal size. This could be OK, but the bouton volume data should be checked critically. For example, the shrinkage of the terminal as measured by Bassoon is smaller than that measured by CB1. The authors use an agonist of CB1 throughout the paper – this will certainly change the configuration of the receptor (dimerization, clustering, internalization) that could artificially inflate the shrinkage they claim. This point should be discussed.

The bouton volume and active zone data has been thoroughly checked. We have addressed the reviewers concerns about using CB1 as a proxy for bouton volume in the following ways: 1. While CB1 internalization has been reported, it does not require protein synthesis, so the fact that we see a blockade of the structural change with cycloheximide would suggest we are measuring a change in volume, not internalization of CB1. 2. Based on previous work, it seems unlikely that our protocol would cause significant redistribution of CB1 (Dudok et al., 2015; Thibault et al., 2013). In particular, the very high density of CB1 on interneuron boutons makes it such that even a partial shift of CB1 to endosomes would not affect the labeling of the bouton membrane (Thibault et al., 2013). 3. We have quantified the colocalization of CB1 and endosomal marker, Rab7 (Feng et al., 1995), a way of estimating internalized CB1 (Howlett et al., 2010), which was not significantly altered by CB1-iLTD induction (see Author response image 1 in response to reviewer 1 below) 4. We do not anticipate that clustering or dimerization could artificially inflate the shrinkage because these processes would not necessarily pull receptors towards the bouton centerpoint, but would more likely stochastically change the distribution of CB1 on the surface of the membrane which, given our 3D analysis, would still provide an accurate measure of the bouton volume. Moreover, dimerization/clustering should not be sensitive to protein synthesis inhibition. 5. Finally, and as suggested by reviewer 1, we have added new data demonstrating that VGAT, an additional CB1-independent marker of inhibitory boutons, is also robustly reduced by CB1 (Figure 1—figure supplement 1E). 6. Lastly, we have now added a thorough discussion of the research that exists on CB1 internalization.

In addition, as requested, we have added new data in which we analyzed the Bassoon size following cycloheximide treatment (See Figure 1F). In this dataset, which now includes more “n’s”, the magnitude of the reduction in active zone size following WIN treatment is similar to that measured with CB1. However, it is not clear whether all components of the presynaptic bouton shrink proportionately. For example, we see a fairly large change in VGAT volume, consistent with previous reports of short-term CB1 activation-induced vesicle redistribution and changes in numbers of docked/primed vesicles (Garcia-Morales et al., 2015; Ramirez-Franco et al., 2014).

2) The MS data needs the tolerance value and stats. The link between the requirement for ubiquitination and protein synthesis is not as strong as it could be. The authors could add experimental data or should at least include a working model in their discussion addressing such questions as: Are both required? Do they operate in parallel or is new synthesis needed for ubiquitination?

In response, we now include the tolerance value (added in Materials and methods section) and results of statistical comparison between vehicle and treatment conditions. We have also included our proposed model for the molecular mechanisms underlying CB1-iLTD (Figure 5E). Our data demonstrate that CB1 activation triggers protein synthesis, protein degradation and structural changes. Our SILAC data clearly show an increase in several proteins that are involved in proteasomal function, suggesting that protein synthesis is upstream of protein degradation, and not in parallel.

3) Drug alone controls are missing for several conditions, particularly CHX and JSK

The drug alone controls have now been included for CHX and JSK (Figure 1—figure supplement 1A, Figure 1E, 3A) and also for ziram (Figure 5D).

4) The conclusions RE Rac and Arp need to be modified because the cellular site of action is not established in this study. The manipulations are pharmacological, and although the narrative and proteomics wants us to think about a pre-terminal action, one could imagine postsynaptic effects as these proteins are well-known for postsynaptic roles. The authors need to modify their conclusions or use tools that exist to cell specifically manipulate these proteins and establish the site of action.

The model of the mechanism of action of Rac1 and Arp2/3 has been modified (Figure 3G) to reflect this criticism. We have also modified the Discussion to acknowledge that these drugs may not only be acting presynaptically. While these proteins have been well studied postsynaptically, they also have established roles in presynaptic function (Gokhale et al., 2016; Njoo et al., 2015; Roesler et al., 2019). Given that our induction protocol relies only on presynaptic receptor activation (e.g. CB1 receptors are expressed at GABAergic terminals but not CA1 pyramidal neurons), which triggers a presynaptic structural change that relies on actin dynamics, a presynaptic target is the most likely explanation for our findings.

Reviewer #1:

[…]

Figure 1:

While the detailed description of the analysis of the boutons was appreciated, the data presented in 1C are not convincing in their present state. The authors provide no controls to ensure that it's not endocytosis/remodeling of the CB1 receptors themselves, which would also result in a decreased CB1 volume. Additional inhibitory presynaptic markers (such as VGAT) should also be assessed and scored. Given the strict cutoffs the authors use in the analysis (ie sphericity value within the Imaris analysis), it could be that they're simply detecting a subset of smaller CB1 puncta resulting from local changes in the receptor pool; while the boutons themselves remain largely unchanged. The use of this CB1 volumetric analysis through the paper is problematic, as in almost all cases the authors do not test the effects of the pharmacological treatment alone (CHX, JSK, CK-666, NSC and ziram). Most if not all of the conclusions could be substantially undercut if these treatments in control conditions show the same change as when paired with WIN treatment. The study of structural plasticity in the postsynapse has largely relied on live cell imaging, so the compartment can be analyzed before, during and post treatment. While challenging, the authors should consider a similar approach, even if only to demonstrate their CB1 volumetric approach is valid. Using either an inhibitory neuron restricted GFP expression or expression of a tagged inhibitory presynaptic protein to allow live imaging of the bouton volume before and after WIN application, the authors could the perform fixation and staining for CB1 receptors and analyze how this compartment changed during the course of the treatments. This would bring their approach more in line with the standard in the field and allow for much more current and compelling analysis for structural plasticity to be performed.

We thank the reviewer for her constructive criticism.

Protein synthesis inhibition blocks the change in bouton volume, whereas receptor internalization or dimerization is not dependent on protein synthesis. The parameters used in Imaris include a lower size limit which was meant to exclude small CB1 puncta. Our experiment measuring Bassoon labeling within CB1+ boutons was designed as a control that is independent of changes in CB1. We have now also added data demonstrating that VGAT labeling is also reduced by CB1-iLTD induction (Figure 1—figure supplement 1E).

In response to concerns about internalization, we have performed an experiment to compare the subpopulation of CB1 receptors that colocalize with endosomal marker Rab7, following CB1-iLTD induction which showed no increase in CB1 internalization with WIN.

Author response image 1
Left, Airyscan confocal images of Rab7/CB1 labeling.

Right, % overlap of CB1 and Rab7 within CB1+ boutons was not significantly altered by 25 min WIN treatment when quantified as average per field of view nor by animal. CTRL: 0.09 ± 0.01 v. WIN : 0.06 ± 0.004, U = 18111; Z = 0.746, n.s. indicates p > 0.05; Mann-Whitney. n = number of fields of view (3 images/ slice, 1 slices/rat, 5 rats/CTRL, 4 rats/WIN). Data are presented as box plots (left) and data points (right) where box represents the 25th and 75th percentile of data range, mean is represented with a square, and median with a line inside the box.

In addition, we have added an in-depth discussion wherein we argue that it is very unlikely that receptor internalization and degradation could account for our findings.

We performed all drug only controls with JSK, CHX, and ziram alone and found they have no effect on basal bouton volume.

We attempted using CCK-Cre mice to test the viability of a Cre-dependent viral strategy to label CB1 interneurons with GFP, but we found this line to be leaky in excitatory neurons and we struggled to resolve boutons clearly given the background in the acute slices with 2-photon laser microscopy. We then attempted live imaging strategies with this interneuron population, by patch-loading CB1+ interneurons with morphological dyes as we did previously (Younts et al., 2016). Unfortunately, in our hands, the tiny caliber of the axon restricts dye loading and is difficult to resolve with the 2P to the extent needed to detect small changes in bouton volume.

Specific comments:

– Examples (similar as in 1A) should be provided of CB1 boutons with WIN/WIN+CHX treatment.

We have now added representative boutons for WIN and WIN + CHX and additional widefield images in supplemental information to illustrate the boutons selected.

– CHX treatment alone should be assessed for C.

See Figure 1—figure supplement 1A

– CHX and WIN+CHX should be assessed for D.

See new Figure 1F

– Did the authors look at what happens to the corresponding inhibitory postsynaptic compartments? Presumably these also may decrease in size as well.

This is a very interesting question. Given that there are multiple reports of concurrent presynaptic changes resulting from postsynaptically-expressed forms of plasticity, we agree that there may be postsynaptic changes resulting from presynaptic plasticity. Testing this possibility is not straightforward given that inhibitory synapses mainly occur directly on the dendritic shaft, or in this case, on the cell soma, which makes structural analysis very difficult. In any event, we feel this issue is beyond the scope of our paper.

– For Figure 1—figure supplement 1C, the indicated mean for the WIN group for the graph provided should be checked. Given the large increase in the low volume population seen in the WIN group vs control group; and the similar size of the groups (540 vs 602) it's hard to understand how both means can be right around 1.

We have now added an additional technical replicate to this dataset (Figure 1—figure supplement 1C). Even with the addition of more “n”, there is still no significant difference between control and WIN-treated slices. We admit that while it appeared that the distributions were different in the previous dataset, we carefully rechecked the graph, and the means were accurate. We think that the population of larger boutons was also increased in the WIN group, but this was more difficult to detect by eye.

Figure 2:

The experimental design (multiplexing of fully medium-heavy and heavy labeled cell cultures) is nice, but there are several potential issues regarding data processing and interpretation.

Protein identification and quantification

– The accepted parent mass tolerance of 50 PPM is very high and might lead to false identifications. When acquiring the data with an MS1 resolution of 60k on a well-calibrated Orbitrap LTQ instrument, a MS1 mass tolerance of ~5 PPM should be used.

The instrument error during acquisition was ± 15 PPM and was collected prior to calibrating the mass spectrometer. Though the precursor tolerance was ± 50 PPM the false discovery rate was adjusted to below 1% at the protein and peptide levels. The MS/MS of a large sampling of proteins were validated manually as well.

– Why are the SILAC labels (Lys4/Arg6 and Lys8/Arg10) specified as variable modifications instead of “SILAC pairs”? Please describe in more detail how the medium-heavy and heavy counterparts were quantified and how the fold-changes were calculated. Were only peptides species considered that were quantified in the medium heavy and heavy form? Why not using software that provide well-established workflows for SILAC quantification (such as the freeware MaxQuant)?

In Mascot the Quantitation Method (SILAC K+4 K+8 R+6 R+10) was used with each SILAC modification in exclusive mode. The raw data files were first processed using precursor ion quantitation of the Quantitation toolbox of Mascot Distiller (Matrix Science Ltd; version 2.7). Mascot was then used to search the rat databases (SwissProt and NCBI nr along with a decoy database to obtain FDRs) using the following parameters: trypsin; product ion mass tolerance of 0.40 Da; precursor ion tolerance of 50 PPM; carbamidomethyl Cys – fixed modification and variable modifications of: deamidated Asn and Gln; label:2H(4) of Lys; label:13C(6) of Arg; label:13C(6)15N(2) of Lys; label:13C(6)15N(4) of Arg and oxidation of Met. The result files obtained from Distiller and the Mascot searches were then uploaded to Scaffold Q+S (Proteome Software Inc.; version 4.9) using between-subjects, log2 ratio-based analysis of unique peptides against a reference to obtain the protein’s mean quantitative values and t-test p-values with Benjamini-Hochberg correction.

– From Supplementary file 1, we understand that each ~700 proteins were identified in the forward and the reverse samples. The overlap between the two samples is 391 proteins (main text), indicating that the reproducibility is fairly low between the measurements.

Statistics

We thank the reviewer for pointing out the inconsistency. After careful re-examination of our results, we identified a misprint in the initial version of the manuscript that has now been corrected. The quantitation software Scaffold gives the Mascot search results showing 68% overlap (see Venn diagram in Author response image 2) which is quite good for MS/MS: 754 Proteins at 99.0% Minimum, 0.1% Decoy FDR with 2 unique peptides; 84,231 Spectra at 95.0% Minimum and 0.35% Decoy FDR. We have also updated Supplementary file 1 and Figure 2—figure supplement 1C accordingly.

Author response image 2

– Please perform proper statistics to identify differentially regulated proteins (e.g. t-test and multiple testing correction). A simple fold-change cut-off (as described in Figure 2C) is not state-of-the-art.

We have now performed a new analysis of the dataset using statistical testing and multiple test correction (Benjamini-Hochberg) following the reviewer’s recommendation. We identified a total of 99 differentially expressed proteins (43 upregulated and 56 downregulated). Results of the analysis have been included in a volcano plot in Figure 2D as well as in Supplementary file 1.

Specific comments:

– The supplementary data should be improved to enable the reader to understand the proteomics data and inspect all quantified proteins. Please include one table that presents the quantitative results for all proteins in all biological replicates (forward and reverse), so that the reader can get a sense for the biological variability. Please include a column with gene names. In Supplementary file 1, why does the table contain proteins from rattus norvegicus, mus musculus and human?

We have now edited the Supplementary file 1 following the reviewer’s request to include values for all biological replicates, gene name. The initial table had not been properly filtered by species. However, all analysis was conducted on filtered data from exclusively Rattus norvegicus proteins, and the table has been updated to reflect this.

– All mass spec raw data and database search results should be uploaded to the PRIDE repository prior to publication and the PRIDE identifier should be included in the Materials and methods section.

The mass spectrometry proteomics data have been deposited to the ProteomeX change with identifier Consortium via the PRIDE [1] partner repository with the dataset identifier PXD020008 and 10.6019/PXD020008.

Data interpretation:

– Protein up-regulation is often interpreted as increased protein synthesis and protein down-regulation is often interpreted as increased degradation in the manuscript. However, based on the SILAC-MS data one cannot draw these conclusions. Protein up-regulation could result from more synthesis or less degradation and protein down-regulation could result from less synthesis or more degradation. Based on the presented data there is no evidence for one or the other.

The reviewer is correct that changes in protein levels could result from more or less synthesis/degradation and that these cannot be disentangled at the level of the SILAC-MS. We have now edited the manuscript to reflect this point. We do, however, provide evidence that ubiquitin-proteasome system activity and the degradation of ubiquitinated proteins is enhanced by CB1 activation in hippocampal slices (Figure 5A). Moreover, cycloheximide treatment blocks the reduction in bouton volume suggesting that protein synthesis is required.

– Based on the proteomics data one cannot distinguish proteins from different subcellular compartment or different cell types. The following statement from the Discussion sections is hence misleading: "[…] why ribosomal proteins would be synthesized locally if ribosomes are exclusively assembled in the nucleolus is unclear."

We have now removed this statement so as not to mislead the reader. We also clearly state in the Discussion that our results reflect protein isolated from whole neurons.

Figure 3:

The authors refer to JSK as an "actin stabilizing drug", whereas in fact it's a drug that promotes actin polymerization. This fact complicates the data presented in this figure as essential controls (such as JSK treatment alone for 3A) are missing. If JSK treatment alone increases the bouton size, is this really demonstrating that the structural remodeling requires actin depolymerization?

New results indicate that JSK alone had no effect on the structure of boutons on its own (Figure 3A) suggesting that the blockade of structural change we see with WIN + JSK is due to the drug preventing an activity-dependent remodeling process. While some in vitro data supports the notion that JSK promotes actin nucleation, the effect of JSK is highly dependent on the presence of monomeric actin, other actin binding proteins, and the concentration and duration of JSK exposure (Bubb et al., 2000). There is also evidence for a paradoxical decrease in the amount of F-actin with JSK in some parts of the cell wherein actin remodeling was prevented via a sequestration of monomeric actin (Bubb et al., 2000; Spector et al., 1999). Overall, the net effect of JSK may be hard to predict, but all reports agree that it interferes with the normal polymerization of actin. For this reason and since we cannot measure exact F- or G-actin levels at these particular boutons, we have modified our conclusions with JSK accordingly to indicate only that CB1-iLTD relies on “actin dynamics” and not polymerization specifically.

Specific comments:

– 3A, the examples provided with the black background are harder to examine. The authors should use the white background used in Figure 1 and Figure 1—figure supplement 1 examples.

Good suggestion; thank you. We have now made these changes.

– 3B, what explanation do the authors have for why the +JSK/WIN treated curve stabilizes around 80%?

We do not know but can speculate on two possibilities. The most likely explanation is that the mechanism of iLTD may rely only partially on the stabilization of filamentous actin. This is somewhat borne out by the structural data which is also not completely restored to baseline and may be due to the additional requirement of Arp2/3 or other actin binding proteins in iLTD. Alternatively, this could be a result of incomplete stabilization of actin polymers due to the inability of the drug to permeate deep enough into the tissue or other changes associated with JSK treatment such as sequestration of monomeric actin (discussed in more detail above).

– Additionally, the authors should test the effect of the Arp2/3 inhibitor and Rac inhibitor with and without WIN treatment on the bouton volume.

We attempted to perform this experiment but were interrupted by the COVID-19-related shutdown of the lab which resulted in the loss of the samples we had collected. Repeating these experiments involves a significant investment of time that is current magnified by the fact that our animal colonies, facilities, and laboratory access are all still operating at reduced capacity due to COVID-19 precautionary measures. However, given the strong correlation that is observed with all structural and functional manipulations (i.e. CHX, JSK, and ziram), our electrophysiology results with the Rac1 and Arp2/3 inhibitors strongly suggest concomitant structural changes or blockade thereof.

Figure 5:

Specific comments:

– 5D, examples would be better with white backgrounds.

We have now changed the representative boutons accordingly.

– Analysis for the effect of ziram alone needs to be done.

Agreed; please, see new Figure 5D.

Reviewer #2:

[…]

1) Boutons simply should not be considered individual n's for statistics. The authors are welcome to argue for what the independent biological unit is. Please also clarify, for a given condition (such as Veh or WIN), were slices from the same animal used for multiple conditions, or were the slices from one animal used for one condition at a time? If the latter, then an animal is the biological unit, with the possibility of using nested data statistical approaches. Alternatively, the mean value/slice could be considered. Issues such as animal health, temperature fluctuations, buffer osmolarity et cetera could otherwise exist and generate 100s of values per slice that are a little bigger or smaller due to technical variables. This is a foundational observation for the study and needs to be robust. As presented, it looks small and only visible when n's are in the 1000s. Likewise, although it was good to consider involvement of other inhibitory synapses (like PV+), a figure showing no significant difference that is potentially grossly underpowered is not so useful (n's of 500).

In these figures, we have used slices from the same rats for multiple conditions. We absolutely agree that slices are vulnerable to factors such as temperature fluctuations, changes in buffer osmolarity/oxygenation, etc. For this reason, each experiment and all conditions were performed at the same time on the same day with the same buffer using slices from the same animal. In order to further balance the conditions, slices were matched by location in the dorsalventral axis across the conditions. We also image at 3 specific locations in the CA1 pyramidal layer to tile the area between CA2 and subiculum. However, even with all these variables controlled for, it is still highly expected that boutons will not be uniform in size. We feel that the intrinsic variability in volume present in the population of boutons is biologically relevant and should be represented for the reader as has been previously reported (Dudok et al., 2015; Lees et al., 2019; Meyer et al., 2014). Moreover, while we fully acknowledge that the percent change we see in bouton volume is small, it is clear that small changes in the nervous system are no less functionally important. According to our power analysis, in order to achieve 80% power, given the standard deviation of bouton volume per slice was averaged at ~0.45, we would require 142 slices per group for an effect size of 10% (see Author response image 3). Such a high number would be next to impossible to achieve with any standard animal protocol.

Author response image 3

As for the analysis of inhibitory synapses established by PV+ neurons, we would anticipate a similar “n” is necessary to see a WIN-induced change and therefore n’s of 500 are well-within range. However, to make the power of the PV control experiment more comparable to those performed on CB1+ boutons we have now added additional “n” to this experiment. We still see the WIN treatment has no effect on PV bouton volume. We also should note that although all datasets presented in this paper represent multiple technical replicates pooled together, we routinely see the WIN induced shrinkage in each individual technical replicate with much lower ‘n’s and the WIN effect is evident if we average the bouton volume values from Figure 1C per slice (see Author response image 4).

Author response image 4

2) In general, the bioinformatics approach was rigorously and thoroughly presented. It is also nice that the authors further discuss results from other programs. One item to clarify – since the GO analyses are the major data presented: please review and clarify the extent to which the same protein is annotated with multiple functional properties such that it is populating (and driving) multiple terms on those lists.

We agree with the reviewer regarding the annotation redundancy in Gene Ontology (GO) analysis. In order to overcome that issue, we have attempted a two-pronged approach: (i) we used several annotation tools (GSEA, IPA and SynGO) and (ii) we performed Enrichment map analysis of GSEA (Figure 2C). Here, each node represents a GO term and edges represent overlap between gene sets. This analysis highlighted 4 dense clusters (e.g. protein synthesis and processing, neuronal projections, energy metabolism and extracellular space). The observation of distinct clusters of enriched GO terms demonstrates the low overlap between gene-sets and confirms that the associated functional properties are populated by distinct gene-sets. Within a given cluster, there is indeed a higher level of similarity as indicated by the proximity between the nodes (see also Supplementary file 2).

Reviewer #3:

[…]

The manuscript contains an interesting data set and the involvement of the ubiquitination system in CB1-LTD is highly intriguing. However, after reading the manuscript there are many “lose ends”, and it is still not clear what happens during CB1-LTD. Indeed, the authors themselves describe these events as “coordinated engagement of multiple cellular processes”, which does not clarify much. I would like to encourage the authors to try and formulate a more complete hypothesis of the cellular events that underlie CB1-LTD.

We thank the reviewer for their comments about our study. To clarify our hypothesis of the molecular events that underlie CB1-LTD, we have now modified the text and summarized the molecular mechanisms underlying CB1-iLTD in a diagram (Figure 5E). Our SILAC data clearly show an increase in several proteins that are involved in proteasomal function, raising the possibility that protein synthesis is upstream of protein degradation, and not in parallel.

I have two major points:

1) It is not easy to link the proteomics findings to changes in the actin cytoskeleton in presynaptic boutons. For instance, the authors show that WIN slightly reduces Arp2c protein levels, but that inactivating Arp2/3 actually enhances CB1-LTD. I am wondering if the explanation given by the authors the most plausible interpretation of the data. Is it known how CB1-LTD is expressed (fewer vesicles? reduced release probability?) and is it possible to link this to specific changes in the actin cytoskeleton?

CB1-iLTD causes a reduction in neurotransmitter release, i.e. reduced release probability (Chevaleyre and Castillo, 2004; Heifets et al., 2008), however the mechanism(s) driving this phenomenon are, as yet, unknown. We have attempted to address this question here, but more work is still needed to understand the exact molecular pathways involved. Early research suggested the mechanism of CB1-iLTD might rely on dephosphorylation of the active zone protein RIM1α (Chevaleyre et al., 2007), but phosphorylation of RIM1α by PKA was later shown not to be involved (Kaeser et al., 2008). CB1 receptor activation has also been suggested to control release via change in vesicular release mode (Aubrey et al., 2017) or change in the distribution of vesicles around the active zone (Garcia-Morales et al., 2015; Ramirez-Franco et al., 2014), but these studies were performed at different synapses that may not express the same form of CB1-iLTD. In general, it is well-established that the size of the presynaptic bouton and active zone is correlated with the postsynaptic response (Bartol et al., 2015; Bourne et al., 2013; Meyer et al., 2014). Here, we suggest that a smaller bouton and active zone in combination with ubiquitination and subsequent degradation of presynaptic release machinery proteins could explain the reduction in GABA release associated with iLTD.

The finding that Arp2/3 inactivation enhances CB1-iLTD may seem a bit counterintuitive but in fact fits well with literature demonstrating that loss of interaction between WASF1/WAVE1 and Arp2/3 causes a loss of functional presynapses (Wegner et al., 2008) and CB1 activation drives growth cone collapse in a Rac1 dependent manner (Njoo et al., 2015) Our SILAC and WB data show that CB1-iLTD triggers loss of ARPC2 and WASF1/WAVE1 which functionally would have the effect of reducing neurotransmitter release (i.e. LTD). Presumably, when we apply CK-666 we are further reducing the pool of Arp2/3 available to participate in the maintenance of actin branches, on top of the Arp2/3 protein that is presumably being degraded, and that is why we see enhanced LTD (i.e. further reduction in neurotransmitter release).

2) The proteomics data show that multiple processes are activated during CB1-LTD. The manuscript does not provide a model or hypothesis for how these processes are linked. For instance, are ubiquitination and protein synthesis independently regulated by CB1? What is the link between ubiquitination and presynaptic actin remodeling? Would it be possible to link CB1-LTD to a specific E2 or E3 ligase in the presynapse?

We have now provided a proposed model in Figure 5E. Based on two pieces of data, the most likely scenario seems to be that protein synthesis triggered by CB1 activation via mTOR (Younts et al., 2016) is upstream of ubiquitination and structural change. CB1 activation triggers an upregulation of several proteasomal subunits and E2 ligases (Figure 2, 4), which suggests these components may be synthesized on demand to regulate protein degradation locally. While we cannot exclude that their upregulation comes from a reduction in their degradation, the long half-life of proteasomal subunits makes it unlikely that we could see such a strong increase from reduction in degradation alone in such a short time window (25 min)(Dorrbaum et al., 2018). Structural change requires protein synthesis. Our data suggests that CB1 activation induces rapid reduction of ARPC2, WASF1/WAVE1 and other components of the WAVE regulatory complex (WRC), and that this pathway is required for iLTD. We propose that ubiquitination of these proteins underlies structural remodeling.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Requested revisions:

1) The authors in their rebuttal argue that standard power calculations are "unreasonable" because n of 145 slices would be required. At the most fundamental level, if this were true, this question and conclusion are not appropriate given the constraints of the methodology and one should focus on powerable questions. However, a statistical analysis by a statistician might indicate other analytical approaches. The authors show how close to being powered they may be with the slice means alone (not including nesting) showing significance. If the authors want to show the data as is in the main figure, the authors could address this concern by creating supplemental figures that show bouton size distributions for each slice and also figures with slice means and exact p values.

Excellent point. In response, we have created new supplemental figures (Superplots) that show the bouton size distribution color-coded by slice with the slice means plotted on top and the overall mean of slices overlaid with a vertical line representing the 95% confidence interval (CI) to clearly represent the effect size (Lord et al., 2020). For these figures, we reanalyzed the data and balanced the number of slices/animals across conditions and performed one-way ANOVA with post hoc Tukey test for multiple comparisons. Exact p-values and confidence intervals are included in each supplemental figure legend. Importantly, our key observations were reproduced using slice as the independent variable.

2) Bassoon puncta were to serve as an independent measure of bouton size to support the CB1 observations. However, the methods describe first making a binary mask of CB1 signal, then using the "and" function to isolate the bassoon pixels that overlap with CB1 pixels. As described, it is almost impossible to reach any other conclusion than reproducing the CB1 measurement results. Please clarify methods substantially to remove this concern.

We apologize for the confusion. A dilated CB1 mask was used to isolate Bassoon in the vicinity of and inside CB1 boutons, the size of Bassoon is free to vary inside the presynaptic terminal since Bassoon labeling is considerably smaller than the CB1 terminal volume (Average control volume of bassoon: ~ 0.51 μm3 v. CB1: 1.85 μm3). We acknowledge the limitations of our method and have removed mentions of the CB1-independent nature of this measurement. Of note, our findings are consistent with a large body of previous work showing the active zone size is highly correlated to the overall presynaptic size (Matz et al., 2010; Maus et al., 2020; Meyer et al., 2014).

3) In working to propose a solution for data presentation, the authors raise an important point, supplemental data should include an analysis that addresses whether the plasticity effects preferentially involve a particular size subset, smaller or large.

We have now included a histogram showing binned bouton distribution per slice to address this question (Figure 1—figure supplement 1C). These data suggest that iLTD results in a reduction in the number of large boutons and an increase in the fraction of small boutons.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #2:

1) The supplementary figures are highly appreciated and allow critical review of the nature of the experimental data and robustness of the conclusions.

2) The authors' response indicating that data was removed to balance slice numbers needs to be clarified. One should not have to remove data. If slices were removed from supplementary figure analyses, what guided those decisions? Do the removed slices remain in the main figure data? I would suggest all the data that were believed to be valid in the original submission be included in the main and supplementary figures. (Or was there a situation of grossly unequal n's in the control/win condition and the authors' are trying to make statistical comparisons across similarly sized groups post hoc? In principle, a similar number of interleaved controls should exist for each additional test condition.) Please clarify.

We have made the requested clarification points in the manuscript. The reviewers suggested we include all slices and use “slice” as our independent variable. Following their advice, we went ahead and included plots showing average bouton volume per slices in the main figures (similar to the plots we included in our response to the reviewers) and kept the superplots in supplementary figures.

References

Bubb, M.R., Spector, I., Beyer, B.B., and Fosen, K.M. (2000). Effects of jasplakinolide on the kinetics of actin polymerization. An explanation for certain in vivo observations. J Biol Chem 275, 5163-5170.

Dorrbaum, A.R., Kochen, L., Langer, J.D., and Schuman, E.M. (2018). Local and global

Feng, Y., Press, B., and Wandinger-Ness, A. (1995). Rab 7: an important regulator of late endocytic membrane traffic. J Cell Biol 131, 1435-1452.

Gokhale, A., Hartwig, C., Freeman, A.H., Das, R., Zlatic, S.A., Vistein, R., Burch, A., Carrot, G., Lewis, A.F., Nelms, S., et al. (2016). The Proteome of BLOC-1 Genetic Defects Identifies the Arp2/3 Actin Polymerization Complex to Function Downstream of the Schizophrenia Susceptibility Factor Dysbindin at the Synapse. J Neurosci 36, 12393-12411.

Howlett, A.C., Blume, L.C., and Dalton, G.D. (2010). CB(1) cannabinoid receptors and their associated proteins. Curr Med Chem 17, 1382-1393.

Kaeser, P.S., Kwon, H.B., Blundell, J., Chevaleyre, V., Morishita, W., Malenka, R.C., Powell, C.M., Castillo, P.E., and Sudhof, T.C. (2008). RIM1alpha phosphorylation at serine-413 by protein kinase A is not required for presynaptic long-term plasticity or learning. Proc Natl Acad Sci U S A 105, 14680-14685.

Lees, R.M., Johnson, J.D., and Ashby, M.C. (2019). Presynaptic Boutons That Contain Mitochondria Are More Stable. Front Synaptic Neurosci 11, 37.

Maus, L., Lee, C., Altas, B., Sertel, S.M., Weyand, K., Rizzoli, S.O., Rhee, J., Brose, N., Imig, C., and Cooper, B.H. (2020). Ultrastructural Correlates of Presynaptic Functional Heterogeneity in Hippocampal Synapses. Cell Rep 30, 3632-3643 e3638.

Roesler, M.K., Lombino, F.L., Freitag, S., Schweizer, M., Hermans-Borgmeyer, I., Schwarz, J.R., Kneussel, M., and Wagner, W. (2019). Myosin XVI Regulates Actin Cytoskeleton Dynamics in Dendritic Spines of Purkinje Cells and Affects Presynaptic Organization. Front Cell Neurosci 13, 330.

Spector, I., Braet, F., Shochet, N.R., and Bubb, M.R. (1999). New anti-actin drugs in the study of the organization and function of the actin cytoskeleton. Microsc Res Tech 47, 18-37.

Wegner, A.M., Nebhan, C.A., Hu, L., Majumdar, D., Meier, K.M., Weaver, A.M., and Webb, D.J. (2008). N-wasp and the arp2/3 complex are critical regulators of actin in the development of dendritic spines and synapses. J Biol Chem 283, 15912-15920.

https://doi.org/10.7554/eLife.54812.sa2

Article and author information

Author details

  1. Hannah R Monday

    Dominick P. Purpura Department of Neuroscience, Albert Einstein College of Medicine, Bronx, United States
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3969-492X
  2. Mathieu Bourdenx

    1. Department of Developmental and Molecular Biology, Albert Einstein College of Medicine, Bronx, United States
    2. Institute for Aging Studies, Albert Einstein College of Medicine, Bronx, United States
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Funding acquisition, Visualization, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Bryen A Jordan

    1. Dominick P. Purpura Department of Neuroscience, Albert Einstein College of Medicine, Bronx, United States
    2. Department of Psychiatry and Behavioral Sciences, Albert Einstein College of Medicine, Bronx, United States
    Contribution
    Supervision, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Pablo E Castillo

    1. Dominick P. Purpura Department of Neuroscience, Albert Einstein College of Medicine, Bronx, United States
    2. Department of Psychiatry and Behavioral Sciences, Albert Einstein College of Medicine, Bronx, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Writing - original draft, Writing - review and editing
    For correspondence
    pablo.castillo@einsteinmed.org
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9834-1801

Funding

National Institute of Mental Health (F31MH114431)

  • Hannah R Monday

National Institute of Mental Health (R01-MH081935)

  • Pablo E Castillo

National Institute on Drug Abuse (R01-DA17392)

  • Pablo E Castillo

National Institute of Neurological Disorders and Stroke (R01-NS113600)

  • Pablo E Castillo

National Institute on Aging (R01-AG039521)

  • Bryen A Jordan

Rainwater Charitable Foundation

  • Mathieu Bourdenx

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. Thomas Younts, Dr. Matthew Klein, and Dr. Ana Maria Cuervo for helpful discussions. We thank Dr. Kostantin Dobrenis, Kevin Fisher, and Vladimir Mudragel of the Einstein Neural Cell Engineering and Imaging Core (supported by The Rose F Kennedy Intellectual Disabilities Research Center) for their advice and assistance with Airyscan confocal microscopy acquisition and analysis. We thank Edward Nieves of the Einstein Proteomics Facility for assistance performing and analyzing proteomic data. This research was supported by the National Institutes of Health: F31MH114431 to HRM, R01-MH081935, R01-DA17392, R01-NS113600 and a pilot grant through NICHD U54 HD090260 to PEC, a shared instrument grant (1S10OD25295) to KD, R01-AG039521 to BAJ. MB was supported by the Rainwater Charitable Foundation. HM and MB also received Junior Investigator Neuroscience Research Award (JINRA) from Dominick P Purpura Department of Neuroscience at Albert Einstein College of Medicine for this project.

Ethics

Animal experimentation: Experimental procedures adhered to NIH and Albert Einstein College of Medicine Institutional Animal Care and Use Committee guidelines as approved by protocol #00001047.

Senior and Reviewing Editor

  1. Gary L Westbrook, Oregon Health and Science University, United States

Reviewer

  1. Erin M Schuman, Max Planck Institute for Brain Research, Germany

Publication history

  1. Received: December 30, 2019
  2. Accepted: September 8, 2020
  3. Accepted Manuscript published: September 9, 2020 (version 1)
  4. Version of Record published: September 28, 2020 (version 2)

Copyright

© 2020, Monday et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 615
    Page views
  • 137
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Genetics and Genomics
    2. Neuroscience
    Marina Kovalenko et al.
    Research Article Updated

    Somatic expansion of the Huntington’s disease (HD) CAG repeat drives the rate of a pathogenic process ultimately resulting in neuronal cell death. Although mechanisms of toxicity are poorly delineated, transcriptional dysregulation is a likely contributor. To identify modifiers that act at the level of CAG expansion and/or downstream pathogenic processes, we tested the impact of genetic knockout, in HttQ111 mice, of Hdac2 or Hdac3 in medium-spiny striatal neurons that exhibit extensive CAG expansion and exquisite disease vulnerability. Both knockouts moderately attenuated CAG expansion, with Hdac2 knockout decreasing nuclear huntingtin pathology. Hdac2 knockout resulted in a substantial transcriptional response that included modification of transcriptional dysregulation elicited by the HttQ111 allele, likely via mechanisms unrelated to instability suppression. Our results identify novel modifiers of different aspects of HD pathogenesis in medium-spiny neurons and highlight a complex relationship between the expanded Htt allele and Hdac2 with implications for targeting transcriptional dysregulation in HD.

    1. Computational and Systems Biology
    2. Neuroscience
    Pedro J Gonçalves et al.
    Research Article Updated

    Mechanistic modeling in neuroscience aims to explain observed phenomena in terms of underlying causes. However, determining which model parameters agree with complex and stochastic neural data presents a significant challenge. We address this challenge with a machine learning tool which uses deep neural density estimators—trained using model simulations—to carry out Bayesian inference and retrieve the full space of parameters compatible with raw data or selected data features. Our method is scalable in parameters and data features and can rapidly analyze new data after initial training. We demonstrate the power and flexibility of our approach on receptive fields, ion channels, and Hodgkin–Huxley models. We also characterize the space of circuit configurations giving rise to rhythmic activity in the crustacean stomatogastric ganglion, and use these results to derive hypotheses for underlying compensation mechanisms. Our approach will help close the gap between data-driven and theory-driven models of neural dynamics.