Protein composition of axonal dopamine release sites in the striatum

  1. Lauren Kershberg
  2. Aditi Banerjee
  3. Pascal S Kaeser  Is a corresponding author
  1. Department of Neurobiology, Harvard Medical School, United States

Abstract

Dopamine is an important modulator of cognition and movement. We recently found that evoked dopamine secretion is fast and relies on active zone-like release sites. Here, we used in vivo biotin identification (iBioID) proximity proteomics in mouse striatum to assess which proteins are present at these sites. Using three release site baits, we identified proteins that are enriched over the general dopamine axonal protein content, and they fell into several categories, including active zone, Ca2+ regulatory, and synaptic vesicle proteins. We also detected many proteins not previously associated with vesicular exocytosis. Knockout of the presynaptic organizer protein RIM strongly decreased the hit number obtained with iBioID, while Synaptotagmin-1 knockout did not. α-Synuclein, a protein linked to Parkinson’s disease, was enriched at release sites, and its enrichment was lost in both tested mutants. We conclude that RIM organizes scaffolded dopamine release sites and provide a proteomic assessment of the composition of these sites.

Editor's evaluation

Using a smart proximity labeling approach, the protein composition of dopaminergic neurotransmitter release sites was determined in striatal axons. Using mice in which release sites were disrupted as control, the authors identified not only established components of the secretory machinery but also many new proteins whose function awaits further characterization. The datasets provided are of very high quality and provide an important foundation for studies on the dopaminergic exocytotic machinery.

https://doi.org/10.7554/eLife.83018.sa0

Introduction

Dopamine is a critical neuromodulator and regulates target cells through G-protein-coupled receptor (GPCR) signaling (Berke, 2018; Liu et al., 2021a; Surmeier et al., 2014). In the vertebrate brain, midbrain dopamine neurons form a crucial regulatory system. The dopamine neuron somata are located in the substantia nigra pars compacta (SNc) and the ventral tegmental area (VTA), and one prominent target area of their axons is the striatum. In the striatum, dopamine is thought to act as a volume transmitter because only a small percentage of dopamine varicosities are directly apposed to postsynaptic specializations (Descarries et al., 1996; Wildenberg et al., 2021) and because dopamine receptors are found mostly extrasynaptically (Liu et al., 2021a; Rice et al., 2011; Sesack et al., 1994; Uchigashima et al., 2016; Yung et al., 1995). Hence, dopamine likely signals by diffusing through the extracellular striatal space before initiating responses in target cells. While much is understood about the mechanisms and molecules involved in the synaptic release of classical neurotransmitters, the machinery mediating the release of dopamine is less well known.

Spatial and temporal precision of vesicular exocytosis at classical synapses is established by the active zone, a conserved molecular machine that docks synaptic vesicles at the presynaptic plasma membrane close to voltage-gated Ca2+ channels, primes the vesicles for release, and aligns these processes with postsynaptic receptor clusters (Biederer et al., 2017; Südhof, 2012). Several lines of evidence point to the existence of active zone-like protein complexes for the control of striatal dopamine release. First, dopamine release is rapid, has a high release probability, and occurs at small hotspots (Banerjee et al., 2022; Banerjee et al., 2020; Beyene et al., 2019; Liu et al., 2018; Marcott et al., 2014; Patriarchi et al., 2018; Silm et al., 2019; Wang et al., 2014; Zych and Ford, 2022), indicating the need for vesicle tethering close to Ca2+ channels before the stimulus arrives. Second, at least some dopamine receptors respond rapidly to local, high concentrations of dopamine (Beckstead et al., 2004; Condon et al., 2021; Courtney and Ford, 2014; Gantz et al., 2013; Marcott et al., 2014), which likely necessitates synchrony of vesicular release. Third, dopamine neuron activity and secretion often correlate with behavior on fast time scales (Bova et al., 2020; Chaudhury et al., 2013; da Silva et al., 2018; Hamilos et al., 2021; Hollerman and Schultz, 1998; Howe and Dombeck, 2016; Jin and Costa, 2010; Schultz et al., 1997), suggesting the need for precise signaling mechanisms. Taken together these properties indicate the presence of protein machinery for rapid and efficient release of dopamine.

Indeed, our recent work has found that proteins for the control of spatial and temporal precision of synaptic vesicle exocytosis are important for evoked dopamine release. These proteins include the active zone scaffolds RIM and Liprin-α, the priming protein Munc13, and the fast Ca2+ sensor Synaptotagmin-1 (Syt-1) (Banerjee et al., 2022; Banerjee et al., 2020; Lebowitz et al., 2022; Liu et al., 2018; Robinson et al., 2019). Conversely, the active zone scaffolds RIM-BP and ELKS are dispensable for axonal dopamine release, even though at least ELKS is present at these sites (Banerjee et al., 2022; Liu et al., 2018). Striatal dopamine release only partially depends on voltage-gated Ca2+ channels of the CaV2 family, channels that are required for stimulated release at most synapses (Brimblecombe et al., 2015; Held et al., 2020; Liu et al., 2022; Luebke et al., 1993; Takahashi and Momiyama, 1993). Thus, there appear to be both similarities and differences in the secretory machines in dopamine varicosities compared with synapses. However, only a handful of proteins important for secretion are known to mediate dopamine release, and many pieces of the underlying secretory machine remain unidentified.

We here characterized the composition of dopamine release sites using an unbiased proteomic approach. We adapted in vivo biotin identification (iBioID), a method that allows for purifying proteins in close proximity to a marker protein (Kim et al., 2014; Roux et al., 2012; Uezu et al., 2016). Using three different release site bait proteins, we found that 527 proteins were present at these sites when we applied a threshold of ≥2.0-fold enrichment over soluble axonal proteins. A total of 190 proteins were enriched in multiple bait conditions and 41 proteins with all three baits. Many proteins with previously known functions at classical synapses were identified, including active zone, Ca2+ regulatory, and synaptic vesicle proteins, as well as proteins that presently do not have defined roles in axonal neurotransmitter secretion. We also tested whether structural scaffolding and dopamine release are important for the composition of these secretory sites. Conditional, dopamine neuron-specific knockout of the presynaptic organizer protein RIM strongly decreased the number of detected proteins, while that of the Ca2+ sensor Syt-1 to abolish synchronous dopamine release did not. We conclude that dopamine release sites are organized structures controlled by the scaffolding protein RIM, and we provide a proteomic assessment of the content of these release sites.

Results

Proximity proteomics to assess dopamine release site composition in the mouse striatum

With the overall goal to generate a comprehensive proteome of release sites in striatal dopamine axons, we adapted a proximity proteomic approach. We fused a promiscuous version of the biotinylase BirA (also called BirA*) (Roux et al., 2012; Uezu et al., 2016) to several proteins associated with dopamine release sites and expressed these bait proteins Cre-dependently using AAVs specifically in dopamine neurons of DATIRES-Cre mice (Bäckman et al., 2006). To locally biotinylate proteins in vivo, we provided excess biotin through subcutaneous injection for 7 days. We then performed affinity purification of the biotinylated proteins from striatal homogenates and identified them by mass spectrometry (Figure 1A). The striatal dopamine axons are particularly well suited for this approach because they are elaborately branched in the striatum and can be easily separated from their midbrain somata and dendrites during tissue dissection, limiting confounds arising from the co-purification of somatodendritic proteins.

Figure 1 with 2 supplements see all
In vivo biotin identification (iBioID) for release site proteins in dopamine axons of the mouse striatum.

(A) Schematic of the experiment with AAVs for Cre-dependent BirA fusion protein expression (AAV2/5-DIO-BirA-fusion), followed by in vivo biotinylation, affinity purification, and analyses by mass spectrometry. (B) Overview of BirA fusion proteins expressed Cre-dependently with AAVs in midbrain dopamine neurons of DATIRES-Cre mice, P: proline-rich motif, H: hemagglutinin (HA) tag. Each mouse expressed one of the three BirA baits (RIMPPCP-BirA, ELKS2β-BirA, and CaVβ4-BirA) or BirA-tdTomato (to generate a proteome for normalization). (C) Sample confocal images of striatal slices of DATIRES-Cre mice expressing the BirA fusion proteins shown in (B), slices were stained with anti-TH antibodies and anti-HA antibodies to label dopamine axons and BirA fusion proteins, respectively. Images are from an experiment in which all constructs were imaged in the same session and with the same settings, and adjusted identically for display, except for ELKS2β-BirA. Images for ELKS2β-BirA were acquired in a separate experiment and with its own control; these images were adjusted for brightness and contrast slightly differently to match overall appearance, and adjustments were identical to those made to its own control acquired at the same time and shown in Figure 1—figure supplement 1. A sample image area is shown from 3 to 5 overview images per mouse and condition, each experiment was repeated in ≥3 mice. (D–F) Protein enrichment in BirA bait conditions over BirA-tdTomato. Log2 fold change values are plotted as frequency histograms. Values at or below 0 represent proteins that are equal to or lower than in the BirA-tdTomato condition (light colors), values >0 represent proteins that are higher than in the BirA-tdTomato condition (saturated colors), and values ≥1 represent hits with ≥2.0-fold enrichment; (D) 1306 total proteins identified, 659 proteins with log2 fold change >0, and 269 proteins with log2 fold change ≥1 (hits); (E) 1496/805/382; (F) 1168/354/114. The gene encoding the protein that was used as a bait is shown as a dot and labeled individually in each panel; (D) 4 independent repeats (12 striata each); (E) 4 (12/12/12/10); (F) 4 (12/12/12/10). For sample images of the negative control for ELKS2β-BirA, see Figure 1—figure supplement 1; for pilot iBioID experiments and assessment of self-biotinylation, see Figure 1—figure supplement 2; for a table of all proteins identified, see Source data 1A; for cDNA sequences used to generate the fusion proteins, see Source data 1B.

We first generated a series of AAVs to express bait proteins in a Cre-dependent manner. In these bait proteins, BirA was fused to RIM or ELKS, active zone proteins associated with dopamine release sites (Banerjee et al., 2022; Liu et al., 2018), or to the β4 subunit of voltage-gated Ca2+ channels (CaVβ4), a component of CaV Ca2+ channel complexes that are present at release sites (Brimblecombe et al., 2015; Dolphin, 2003; Held et al., 2020; Tan et al., 2022b). Full-length active zone proteins fused to BirA exceed the packaging limit of AAVs. Considering this limitation, the following constructs were selected from initial trial experiments (Figure 1B): RIMPPCP-BirA, containing the central region of RIM (PDZ and C2A domains with endogenous linker sequences and proline-rich motifs) that is important for its active zone localization, with BirA inserted at the C-terminus (Kaeser et al., 2011; Tan et al., 2022a; Tang et al., 2016; Wu et al., 2019); ELKS2β-BirA, a short version of ELKS with BirA inserted between the CCC and CCD domains (Held et al., 2016; Kaeser et al., 2009; Liu et al., 2014); and CaVβ4-BirA, a CaV subunit with BirA inserted at the C-terminus (Dolphin, 2003; Held et al., 2020; Tan et al., 2022b). In all experiments, we used BirA fused to tdTomato (BirA-tdTomato) that localizes throughout the axon for calculating enrichment. All constructs also contained hemagglutinin (HA)-tags for identification with HA antibodies.

To assess the axonal presence of these BirA fusion proteins, we injected the AAVs into the midbrain of DATIRES-Cre mice. After 6 weeks, we perfused the mice and stained striatal brain slices with antibodies against HA (to label the fusion proteins) and antibodies against tyrosine hydroxylase (TH, to label dopamine axons) (Figure 1C, Figure 1—figure supplement 1). The fusion proteins were detected within TH-labeled axons. RIMPPCP-BirA and ELKS2β-BirA were present in small punctate structures consistent with release site localization (Liu et al., 2018). CaVβ4 appeared somewhat more widespread, similar to Ca2+ entry in these axons (Pereira et al., 2016), and BirA-tdTomato was broadly overlapping with TH, likely reflecting a soluble axonal localization. We previously performed superresolution microscopy to assess the localization of select endogenous dopamine axonal proteins (Banerjee et al., 2022; Banerjee et al., 2020; Liu et al., 2018). High-quality super-resolved images could not be obtained here, likely because of the specific antibodies and conditions that were needed. In vivo biotinylation was next confirmed for the BirA fusion proteins with or without biotin injections (for 7 days) followed by pilot biotin-affinity purifications and Western blotting of the purified fractions (Figure 1—figure supplement 2 and 'Materials and methods'). This established that biotinylation and protein purification are efficient, and that background biotinylation in the absence of biotin is limited.

To determine dopamine release site composition, we performed iBioID using these three BirA bait proteins and BirA-tdTomato expressed in DATIRES-Cre mice (Figure 1A). After 4–6 weeks of expression, biotin injections were done on seven consecutive days, and 10–12 striata per condition and repeat were dissected and homogenized. Biotinylated proteins were then isolated using affinity purification. The eluates were depleted with antibodies against pyruvate carboxylase (PC), an endogenously biotinylated protein, before assessment of protein content with mass spectrometry. In total, four independent biological repeats per BirA fusion protein were performed (i.e., four times 10–12 striata per condition, conditions: RIMPPCP-BirA, ELKS2β-BirA, CaVβ4-BirA, BirA-tdTomato, also see 'Materials and methods'). The four repeats were run as two large experiments for in vivo biotinylation, biotin affinity purification, and mass spectrometry; two repeats were run in parallel in each experiment, and the two experiments were run approximately 1 year apart from one another (the second experiment also contained analyses of conditional knockout mice described below).

Peptides from 1969 proteins were identified in at least one of the BirA conditions. A total of 300 (15%) of those proteins are mitochondrial proteins, which is common in iBioID, and these proteins were removed from further analyses unless noted otherwise, as has been done before (Loh et al., 2016; Uezu et al., 2016). To identify proteins enriched in the immediate vicinity of the bait proteins (RIMPPCP-BirA, ELKS2β-BirA, and CaVβ4-BirA), the number of peptides identified for each protein and bait was normalized to the number of peptides found for the same protein in the BirA-tdTomato condition, effectively generating a ratio of enrichment over soluble dopamine axonal protein content as fold change. For proteins that were not detected with all BirA fusions, we assigned an average value of 0.5 peptides so that we could calculate fold change values for proteins that otherwise had a 0 in the denominator and log2 of fold change for all detected proteins. After producing log2 of the fold change, values >0 represent proteins detected at levels higher than in the BirA-tdTomato condition (Figure 1D–F, saturated colors), while proteins <0 (light colors) were below. For the main analyses, 'hits' were proteins enriched ≥2.0-fold and have log2 fold change values of ≥1 (Figure 1D–F). These proteins are included in the analyses presented in Figure 2. Analyses with alternate enrichment thresholds are shown in Figure 2—figure supplements 1 and 2. These applied cutoffs are similar to previous studies using iBioID (Takano et al., 2020; Uezu et al., 2016).

Figure 2 with 4 supplements see all
Protein composition of release sites in dopamine axons.

(A) Venn diagram listing genes that encode protein hits enriched at least 2.0-fold over BirA-tdTomato with RIMPPCP-BirA, ELKS2β-BirA, or CaVβ4-BirA iBioID baits. Bait-encoding genes are shown outside the Venn diagram in the color of the corresponding part of the diagram. Enrichment is shown as the average of 4 independent repeats run in 2 mass spectrometry sessions (repeats 1 and 2 in the first session, repeats 3 and 4 in the second) and was calculated as described in the 'Materials and methods.' Circle size reflects average enrichment across repeats (key on the bottom right). For proteins enriched in multiple bait conditions, circle size corresponds to the bait condition with the largest enrichment. (B) Hits from (A) that have cellular compartment annotations in the SynGO database (Koopmans et al., 2019) are colored in brown and are classified into increasingly specific SynGO subcategories, subcategorizations are not mutually exclusive; proteins used as baits are not included. For a list of hits, see Source data 1C; for Venn diagrams constructed with 1.5- and 2.5-fold enrichment thresholds, see Figure 2—figure supplements 1 and 2; for a comparison of SynGO cellular compartment annotations across thresholds, see Figure 2—figure supplement 3; and for assessment of Neuroplastin (Nptn) localization in striatal synaptosomes, see Figure 2—figure supplement 4.

For RIMPPCP-BirA (Figure 1D) and ELKS2β-BirA (Figure 1E), 50% and 54% of the detected proteins were higher than with BirA-tdTomato, and 20% and 25% were above the 2.0-fold enrichment threshold, respectively. Only 30% of all proteins detected in the CaVβ4-BirA condition were higher compared to BirA-tdTomato, and 9% were enriched ≥2.0-fold. These observations generally align with the morphological data (Figure 1C), where CaVβ4-BirA appeared more widely expressed than RIMPPCP-BirA and ELKS2β-BirA. They are also consistent with functional analyses that revealed that Ca2+ entry in dopamine axons is widespread (Pereira et al., 2016), while active zone proteins (Banerjee et al., 2022; Liu et al., 2018) and release (Pereira et al., 2016) are only detected in 20–30% of the varicosities. In each condition, the protein that was used as a BirA bait was robustly enriched (individually labeled proteins in Figure 1D–F). This further establishes the approach, as self-biotinylation of the BirA fusion proteins should be high compared to other proteins if the labeling radius is small, which has been estimated to be 10–50 nm (Kim et al., 2014; Roux et al., 2012).

Assessment of the protein composition of release sites in dopamine axons

We next constructed Venn diagrams with proteins that were above the 2.0-fold enrichment threshold to assess the overlap of hits between the different bait conditions. In total, there were 527 proteins ≥2.0-fold enriched (Figure 2A). The enrichment compared to BirA-tdTomato indicates that these proteins have a localization that is restricted compared to general dopamine axonal proteins. Of the 527 proteins, 190 were enriched in more than one condition and 41 were enriched with all three baits, RIMPPCP-BirA, ELKS2β-BirA, and CaVβ4-BirA.

Given that the BirA baits are proteins with roles in transmitter secretion and are expressed in dopamine axons, it is expected that proteins with known roles in neurotransmitter release are enriched in the iBioID dataset. We assessed the hits by evaluating their cellular compartment annotations in SynGO, an expert-curated database that assigns localization and function of synaptic genes (Koopmans et al., 2019). Of the 527 enriched proteins, 194 (37%) have one or multiple synaptic cellular compartment annotations in SynGO (Figure 2B), reflecting that their synaptic localization has been established. SynGO does not distinguish between synapse or neurotransmitter type, but instead broadly determines whether there is evidence to support the presence or function of a given protein at synapses in general (Koopmans et al., 2019).

To assess how the enrichment threshold of ≥2.0 influenced this overall assessment of the release site proteome, we constructed Venn diagrams with alternate enrichment thresholds of ≥1.5 (Figure 2—figure supplement 1) or ≥2.5 (Figure 2—figure supplement 2). Naturally, the overall number of proteins that reached threshold varied, with 991 proteins being ≥1.5-fold and 348 proteins ≥2.5-fold enriched, compared to the 527 proteins detected with the ≥2.0-fold enrichment threshold. Changing enrichment thresholds did not increase the percentage of synaptic proteins as identified via SynGO (31% for ≥1.5 and 33% for ≥2.5 thresholds, respectively, Figure 2—figure supplement 3), justifying the use of ≥2.0 as the threshold for further analyses.

We next categorized the proteins based on whether they are known to be localized pre- or postsynaptically in order to assess which types of synaptic proteins were enriched. Proteins with SynGO annotations may have multiple annotations and can be both pre- and postsynaptic, but do not require a more specific annotation. A total of 102 proteins (19% of the entire dataset, 53% of the SynGO synaptic proteins) have a presynaptic annotation with many having additional specific assignments to the active zone or synaptic vesicles, while 87 (17% of the entire dataset, 45% of the SynGO synaptic proteins) have postsynaptic annotations (Figure 2B). Also, 43 proteins (8% of the entire dataset, 22% of the SynGO synaptic proteins) are annotated both pre- and postsynaptically.

If the iBioID approach used here enriched proteins at sites of dopamine release, it would be expected that proteins previously shown to localize to dopamine release sites are enriched. Indeed, RIM1 (Rims1), ELKS2 (Erc2), and Bassoon (Bsn), proteins that can be used as markers for dopamine release sites (Banerjee et al., 2022; Liu et al., 2018) and that were shown to be associated with synaptic vesicle docking sites using a different proteomic approach (Boyken et al., 2013), were all enriched across multiple conditions (Figure 2A).

Proteomic assessment of release site composition, like the one described here, can serve to identify new putative components. Of the 527 enriched proteins with the ≥2.0-fold threshold, 333 (63%) do not have a SynGO annotation (Figure 2B), and 46 proteins are annotated broadly as 'synaptic' without more exact specification. Additionally, for the characterization of the dopamine axon secretory machinery, proteins known to be associated with classical synaptic release that have not been implicated in dopamine release can also be identified. One example of the latter category is Neuroplastin (Nptn), a transmembrane protein with pre- and postsynaptic roles at synapses (Beesley et al., 2014; Boyken et al., 2013; Schmidt et al., 2017; Smalla et al., 2000). Neuroplastin is strongly enriched across conditions (Figure 2A), but its association with dopamine release sites has – to our knowledge – not been described. To assess Neuroplastin localization with a secondary approach, we prepared striatal synaptosomes as we described before (Banerjee et al., 2022; Banerjee et al., 2020; Liu et al., 2018) and stained them with antibodies against TH to mark dopamine synaptosomes, Bassoon to label release sites, and Neuroplastin. In particles labeled with TH, Neuroplastin antibody staining intensity was significantly higher when Bassoon was present (Figure 2—figure supplement 4). These results suggest that Neuroplastin is enriched in proximity to release sites of dopamine axons and provide an independent approach that helps validate iBioID for this specific hit.

Ablation of RIM, but not of Synaptotagmin-1, strongly decreased the number of proteins detected in the dopamine release site proteome

We next asked whether removing proteins important for dopamine release affects the composition of the dopamine release site proteome. Previous work has established genetic strategies to abolish (1) evoked dopamine release by removing the presynaptic scaffolding protein RIM (Banerjee et al., 2022; Liu et al., 2018; Robinson et al., 2019) or (2) synchronous dopamine release by removing the fast Ca2+ sensor Syt-1 (Banerjee et al., 2020; Lebowitz et al., 2022). We crossed mice with floxed alleles for RIM1 and RIM2 or for Syt-1 to DATIRES-Cre mice (Bäckman et al., 2006; Kaeser et al., 2011; Banerjee et al., 2020; Kochubey et al., 2016; Liu et al., 2018; Zhou et al., 2015) to remove these proteins from dopamine neurons (Figure 3A, RIM cKODA or Syt-1 cKODA, respectively). We then performed iBioID analogous to the proteome described in Figures 1 and 2 (called 'control proteome' from this point forward), with the modification that in each mutant we completed two independent repeats instead of four (i.e., two times 10–12 striata for each of the bait conditions in RIM cKODA and Syt-1 cKODA mice) due to the genetic complexity and volume of the experiment.

Figure 3 with 2 supplements see all
Enrichment of release site proteins after conditional ablation of RIM or of Synaptotagmin-1 from dopamine neurons.

(A) Strategy for ablation of RIM1 and RIM2 (RIM cKODA) or Synaptotagmin-1 (Syt-1 cKODA) from dopamine neurons using conditional mouse genetics (Banerjee et al., 2020; Liu et al., 2018). (B) The average number of peptides per protein in bins of five detected with BirA-tdTomato. The control is from data in Figures 1 and 2 and is the average peptide count of 4 repeats, the proteomes from RIM cKODA and Syt-1 cKODA are an average peptide count of 2 repeats each. The x-axis is cut at a peptide count of 50 covering >99% of the detected proteins. Average number of detected proteins: control, 1195; RIM cKODA, 1155; Syt-1 cKODA, 1011. (C–E) Protein enrichment in BirA bait conditions over BirA-tdTomato in RIM cKODA mice. Log2 fold change values are plotted as frequency histograms. Values at or below 0 represent proteins that are equal to or lower than in the BirA-tdTomato condition (light colors), values >0 represent proteins that are higher than in the BirA-tdTomato condition (saturated colors), and values ≥1 represent hits with ≥2.0-fold enrichment; (C) 1089 total proteins identified, 334 proteins with log2 fold change >0, and 149 proteins with log2 fold change ≥1 (hits); (D) 1003/123/62; (E) 1008/130/65. The gene encoding the protein that was used as a bait is shown as a dot and labeled individually in each panel: (C) 2 independent repeats (10/12 striata); (D) 2 (12/10); (E) 2 (10/12). (F–H) Same as (C–E), but for Syt-1 cKODA mice; (F) 1017 total proteins identified, 424 proteins with log2 fold change >0, and 199 proteins with log2 fold change ≥1 (hits); (G) 1103/619/327; (H) 1016/544/246, (F) 2 independent repeats (10/12 striata); (G) 2 (12/12); (H) 2 (12/12). Two-way ANOVA was used in (B) (genotype: p>0.6, peptide count: p<0.001). For a table of all peptides identified in RIM cKODA, see Source data 1F; for a table of all peptides identified in Syt-1 cKODA, see Source data 1G; and for Venn diagrams of release site protein enrichment, see Figure 3—figure supplement 1 for RIM cKODA and Figure 3—figure supplement 2 for Syt-1 cKODA.

We first compared the BirA-tdTomato condition in control, RIM cKODA, and Syt-1 cKODA and found that both the number of identified proteins and the peptide counts were similar across experiments (Figure 3B). Hence, removal of RIM or Syt-1 from dopamine neurons did not cause strong disruptions in the overall dopamine axon protein content in the striatum. The presence of similar amounts of axonal material is consistent with previous work that found TH axon density to be similar to control mice in these mutants (Banerjee et al., 2020; Liu et al., 2018).

We next assessed fold change over tdTomato for each of the BirA baits and each mutant. In RIM cKODA mice, 31% (RIMPPCP-BirA), 12% (ELKS2β-BirA), and 13% (CaVβ4-BirA) of the identified proteins were higher than with BirA-tdTomato, and 14% (RIMPPCP-BirA), 6% (ELKS2β-BirA), and 4% (CaVβ4-BirA) reached the ≥2.0-fold enrichment threshold to be considered hits (Figure 3C–E). These percentages are overall lower than in the control proteome (Figure 1D–F). The protein enrichment over BirA-tdTomato in Syt-1 cKODA mice was more similar to the control proteome (Figure 3F–H, RIMPPCP-BirA: 42% higher than with BirA-tdTomato and 20% ≥2.0-fold enriched; ELKS2β-BirA: 56% and 30%, CaVβ4-BirA 53% and 24%). We conclude that in RIM cKODA mice, release site protein enrichment close to the baits is disrupted compared to control or Syt-1 cKODA mice. This suggests that RIM removal disrupts mechanisms important for release site scaffolding. Abolishing synchronous dopamine release by Syt-1 cKODA has no strong effects on the overall number of proteins detected at release sites in dopamine axons with iBioID.

We next assessed hits in more detail by generating and analyzing Venn diagrams for each mutant. The RIM cKODA dataset contained a total of 198 hits (Figure 3—figure supplement 1), compared to the 527 hits in the control proteome (Figure 2). RIM1 enrichment was absent from all conditions except for RIMPPCP-BirA, which likely reflects self-biotinylation of the bait. It is noteworthy that 49% of all hits (98 out of 198) in the RIM cKODA dataset came from the RIMPPCP-BirA bait only (Figure 3—figure supplement 1, yellow circle), indicating that expression of RIMPPCP, the central region of RIM containing its scaffolding domains (Figure 1B), may restore some scaffolding deficits caused by RIM cKODA. In Syt-1 cKODA mice, 450 hits were detected with the 2.0-fold enrichment threshold (Figure 3—figure supplement 2), similar to the 527 hits in the control dataset, and 104 (23%) of the hits were at least 2.0-fold enriched for all three BirA baits, including RIM (Figure 3—figure supplement 2, center). Changing enrichment thresholds to ≥1.5-fold (RIM cKODA: 268 hits; Syt-1 cKODA 602 hits) or ≥2.5-fold (RIM cKODA: 82 hits; Syt-1 cKODA 252 hits) substantiated the conclusion that RIM but not Syt-1 ablation strongly decreased the overall hit number.

To characterize which proteins were depleted from the release site proteome in the RIM cKODA mice, we first used SynGO to categorize all identified proteins in each dataset into synaptic proteins (Figure 4A, gray bars), presynaptic proteins (Figure 4B), and active zone proteins (Figure 4C). We then plotted the log2 of the fold change of the average of all annotated proteins in each category. In the control dataset (gray bars), synaptic proteins were enriched as expected (Figure 4A), and the extent of enrichment had a tendency to increase as the SynGO annotation became more specific (to presynaptic, Figure 4B, and to active zone, Figure 4C). These effects not only disappeared in RIM cKODA mice, but instead reverted (Figure 4A–C, green bars), and for ELKS2β-BirA and CaVβ4-BirA, proteins in all three SynGO categories appeared depleted (log2 fold change values <0). The depletion is absent in the RIMPPCP-BirA condition in RIM cKODA mice, supporting that release site scaffolding is partially restored. In contrast to RIM cKODA, the proteins found in the Syt-1 cKODA dataset show an enrichment that is overall relatively similar to the control dataset with all conditions having log2 fold change values >0 (Figure 4A–C, maroon bars), supporting that abolishing synchronous dopamine release does not strongly disrupt overall release site protein content. These effects were overall similar for proteins annotated as synaptic vesicle or postsynaptic proteins in SynGO (Figure 4—figure supplement 1).

Figure 4 with 1 supplement see all
RIM cKODA disrupts the protein composition of release sites in dopamine axons.

(A) The average log2 fold change of identified proteins over the same proteins in the BirA-tdTomato conditions, all proteins with a synaptic localization annotation in SynGO for each bait and genotype are included. Positive values represent release site enrichment and negative values represent depletion relative to axonal protein content assessed with BirA-tdTomato. Total number of proteins detected with a SynGO annotation: control, 540; RIM cKODA, 412; Syt-1 cKODA, 438. (B) Same as (A), but for proteins with a presynaptic SynGO annotation. Total number of proteins: control, 257; RIM cKODA, 208; Syt-1 cKODA, 213. (C) Same as (A), but for proteins with an active zone SynGO annotation. Total number of proteins: control, 42; RIM cKODA, 37; Syt-1 cKODA, 36. (D) Proteins enriched in the RIM cKODA dataset mapped onto the control Venn diagram shown in Figure 2A. Proteins present in the control proteome but not in the RIM cKODA dataset are shown in light gray. Enriched proteins are colored based on relative enrichment (key on the right). Proteins in the RIM cKODA dataset that are only enriched with RIMPPCP-BirA are outlined in dark yellow independent of which bait enriched them in the control dataset. (E) Same as (D), but for the Syt-1 cKODA dataset and without outlining a specific condition. Data in (A–C) are shown as mean ± SEM, and significance is presented as *p<0.05, **p<0.01, and ***p<0.001. Two-way ANOVA was used in (A–C) (A: genotype ***, bait ***, interaction ***; B: genotype ***, bait **, interaction ***; C: genotype ***, bait *, interaction not significant), and Bonferroni post-hoc tests (p-values indicated in figure) were used to compare each genotype to control for each BirA bait. For analyses as in (A–C) but for proteins with a synaptic vesicle or postsynaptic SynGO annotation, see Figure 4—figure supplement 1.

We next mapped the ≥2.0-fold enriched proteins detected in each mutant onto the Venn diagram of the control proteome shown in Figure 2. Only 15% of proteins in the control dataset were also enriched in the RIM cKODA dataset (Figure 4D), and 49% of those proteins stem from the RIMPPCP-BirA condition (Figure 4D, circles with yellow outline). In contrast, 37% of the proteins in the control dataset were also enriched in the Syt-1 cKODA dataset (Figure 4E). The observation that there is no full overlap between the control and Syt-1 cKODA datasets is likely due to a combination of factors. Examples are that the approach may not be saturating and may not detect all release site proteins, that the mutant data were derived from fewer repeats, and that Syt-1 cKODA may influence the exact composition but not the overall extent of the release site proteome.

Finally, to assess the organization of the proteins and their interactions in the various release site proteomes, we used the STRING database (Snel et al., 2000; Szklarczyk et al., 2019). STRING combines both theoretical predictions and empirical data to generate maps of protein–protein functional and physical interactions. We selected the proteins that were ≥2.0-fold enriched and assigned as synaptic proteins in SynGO (Figure 2B, Figure 3—figure supplement 1B, Figure 3—figure supplement 2B, 'synaptic'), and analyzed each dataset (Figure 5A–C). In the control dataset, the proteins formed an integrated network with functional nodes classified as active zone proteins, synaptic vesicle proteins, Ca2+ regulatory proteins, and synaptic ribosomes (Figure 5A). The same nodes were detected in the Syt-1 cKODA mice (Figure 5C), substantiating the overall similarities of these datasets. In contrast, these networks were disrupted in RIM cKODA mice (Figure 5B), even when hits from the RIMPPCP-BirA bait (lighter colors in Figure 5B) were included. Only the synaptic ribosome node remained, suggesting that while release site disruption was strong in RIM cKODA mice, other protein complexes near the bait proteins remained intact.

STRING diagrams illustrate functional categories of release site proteins and their disruption in RIM cKODA mice.

(A) STRING diagram of the enriched proteins that have a synaptic SynGO annotation in the control dataset. Physical or functional interactions determined by either empirical data or predictive modeling are illustrated as lines between proteins (Snel et al., 2000; Szklarczyk et al., 2019). The BirA bait proteins are color-coded. Functional categories identified by STRING analyses are labeled close to the corresponding clusters. (B) Same as (A), but for the RIM cKODA dataset. The proteins that are enriched only in the RIMPPCP-BirA bait are shown in lighter gray. (C) Same as (A), but for the Syt-1 cKODA dataset. Number of proteins: (A) 194; (B) 79; (C) 163.

Disrupted recruitment of α-synuclein to release sites in dopamine axons after impairing dopamine release

Loss of dopamine neurons underlies the motor symptoms characteristic of Parkinson’s disease (Poewe et al., 2017), and recent human genetic studies establish that mutations in RIMS genes increase the risk of Parkinson’s disease and are key drivers of disease progression (Liu et al., 2021b; Nalls et al., 2019). Monogenic forms of Parkinson’s disease exist and genetic associations with variable penetrance have been described (Blauwendraat et al., 2020; Day and Mullin, 2021; Marras et al., 2016). Our control dataset contained five of these genes: Snca, Park7, Dnajc6, Synj-1, and Vps35 (Figure 6A). Three were enriched in the release site proteome with at least one of the BirA baits, and Snca, which encodes the synaptic vesicle associated protein α-synuclein, was ≥2.0-fold enriched in each condition (Figures 2A and 6A). In both RIM cKODA and Syt-1 cKODA mice, α-synuclein enrichment was reduced and fell below the 2.0-fold threshold (Figure 6B, Figure 3—figure supplements 1A and 2A). α-Synuclein is associated with synaptic vesicles and regulates synaptic vesicle clustering, SNARE complex formation, and endocytosis (Burré et al., 2010; Diao et al., 2013; Murphy et al., 2000). RIM or Syt-1 knockout causes a loss of docked vesicles at synapses (Chang et al., 2018; Kaeser et al., 2011). If these proteins have similar roles in dopamine axons, the loss of α-synuclein enrichment might be explained by reduced vesicle docking.

Genes associated with Parkinson’s disease in the dopamine release site proteomes.

(A) Enrichment of proteins associated with monogenic forms of Parkinson’s disease (Blauwendraat et al., 2020; Day and Mullin, 2021; Marras et al., 2016) in the control dataset. Hits above the 2.0-fold enrichment threshold are shown in saturated colors, and those below the threshold in light colors; control, 4 independent repeats per bait condition. (B) Enrichment of the five proteins shown for each bait and in control, RIM cKODA and Syt-1 cKODA; control, 4 independent repeats (as described in Figure 1), RIM cKODA, 2 (as described in Figure 3); Syt-1 cKODA, 2 (as described in Figure 3).

Discussion

We used iBioID to assess release site composition in striatal dopamine axons using three bait proteins. Applying a ≥2.0-fold enrichment threshold, we identified 527 proteins, of which 190 were enriched with multiple baits. This proteome contains known secretory machinery, including active zone, Ca2+ regulatory, and synaptic vesicle proteins, and additional proteins not previously associated with vesicular exocytosis. Dopamine neuron-specific knockout of RIM, but not of Syt-1, strongly decreased the number of enriched proteins. We conclude that dopamine axons contain active zone-like sites with secretory machinery for rapid vesicular exocytosis and present an assessment of the protein composition of these sites. Each newly detected protein will require validation for a definitive assignment to dopamine release sites. Our findings establish that the integrity of these active zone-like sites strongly depends on the scaffolding protein RIM.

Proximity proteomics establish a scaffolding role for RIM in dopamine axons

In previous studies, we found that several active zone proteins are present in release site-like structures of striatal dopamine axons, including Bassoon, RIM, Munc13, and ELKS. Using dopamine-neuron-specific knockout, we established that RIM and Munc13 are essential for evoked axonal dopamine release, while ELKS and RIM-BP are dispensable. RIM removal also induced deficits in Bassoon clustering in striatal dopamine axons. We proposed the model of active zone-like release sites for action potential-evoked dopamine release (Banerjee et al., 2022; Liu et al., 2021a; Liu et al., 2018; Liu and Kaeser, 2019). We here used iBioID to generate a list of putative components of these sites. We found that active zone proteins, synaptic vesicle proteins, Ca2+ regulatory proteins, and many additional proteins are present. RIM was enriched across datasets, perhaps indicating a role for RIM in dopamine active zone assembly. RIM ablation from dopamine axons indeed resulted in a strong reduction in the number of proteins identified as release site components with iBioID. This loss of material was observed more strongly with ELKS2β-BirA and CaVβ4-BirA than with RIMPPCP-BirA, suggesting that RIMPPCP-BirA partially restored release site composition. Together with previous data, these findings establish a key scaffolding role for RIM at dopamine release sites. The overall reduced number of hits in RIM cKODA mice also suggests that many hits in the control proteome are specific.

Reductions in hit numbers in RIM cKODA mice might at least in part arise from mislocalized BirA baits in their dopamine axons. However, aberrant bait localization likely does not explain all effects. Twenty-two proteins were enriched with all three baits in RIM cKODA mice, indicating overlapping labeling radii with a resulting proteome relying on bait colocalization (Figure 3—figure supplement 1A, Figure 5). In a previous study, overall dopamine axon morphology was not strongly disrupted in RIM cKODA mice apart from changes in Bassoon clustering (Liu et al., 2018). Here, the general axonal proteome identified by BirA-tdTomato was similar between RIM cKODA and control mice (Figure 3B), arguing for a generally intact axonal arbor in these mice. Together, these points make it unlikely that bait mislocalization fully accounts for the disruption observed in RIM cKODA mice. Even if some of the bait protein is mislocalized, the main conclusion that RIM ablation disrupts release site structure in dopamine axons is supported through this mislocalization.

An active zone-like site for dopamine release

Two recent studies have assessed the composition of dopamine axons or of dopamine varicosities using different methodologies (Hobson et al., 2022; Paget-Blanc et al., 2022). Neither of the studies was designed to enrich for release site proteins over other proteins in dopamine axons. One used mass spectrometry to analyze dopaminergic synaptosomes obtained through fluorescent sorting (Paget-Blanc et al., 2022) and identified 57 proteins associated with these synaptosomes, which contain release sites, vesicles, mitochondria, cytoskeletal elements, and many other components. A second study used proximity proteomics with APEX2 to identify axonal proteins (Hobson et al., 2022), similar to the BirA-tdTomato condition used for normalization in our experiments. Indeed, there is good overlap between the axonal proteins identified with BirA-tdTomato in our experiments with those in Hobson et al., 2022, with 65% of the proteins identified here also present with the APEX2-based method.

We here used an approach to determine enrichment at release sites over these axonal proteomes. RIM1 (Rims1), Bassoon (Bsn), and ELKS1 (Erc1), ELKS2 (Erc2), and P/Q-type (CaV2.1, Cacna1a) and N-type Ca2+ channels (CaV2.2, Cacna1b) were all detected, consistent with previous studies that assessed roles and/or localization of these proteins in dopamine neurons (Banerjee et al., 2022; Brimblecombe et al., 2015; Daniel et al., 2009; Ducrot et al., 2021; Liu et al., 2022; Liu et al., 2018; Uchigashima et al., 2016), and overlapping with a proteomic study that determined release site composition at synapses via purifying proteins associated with docked synaptic vesicles (Boyken et al., 2013). The active zone proteins Liprin-α and RIM-BP were not present in the 2.0-fold enriched dataset. We found that Liprin-α2 and -α3 knockout has less severe effects on dopamine release compared to RIM knockout, while removal of RIM-BP leaves dopamine release unimpaired (Banerjee et al., 2022; Liu et al., 2018). Thus, the presence of active zone proteins in iBioID generally correlates with their known functional roles in dopamine release. One exception to this correlation is Munc13. Striatal dopamine release is strongly impaired after Munc13 ablation (Banerjee et al., 2022), but only enriched in some cases with iBioID (Figures 2 and 5, Figure 3—figure supplements 1 and 2). At synapses, some Munc13 may be more broadly distributed than just at active zones (Tan et al., 2022a; Tan et al., 2022b), but this does not explain its nonenrichment here because in the BirA-tdTomato dataset, Munc13 was not strongly detected either. Hence, Munc13 may either be overall sparse despite its requirement for evoked dopamine release or may be difficult to detect, for example, because it is poorly biotinylated or not easily affinity-purified. Notably, Munc13 was also absent in the release site proteome assessed via docked vesicles (Boyken et al., 2013). Overall, while a saturated proteome is challenging to obtain, many of the proteins previously shown to mediate dopamine release were detected with iBioID.

Dopamine neurons corelease glutamate and GABA (Hnasko et al., 2010; Stuber et al., 2010; Tritsch et al., 2012). While GABA is loaded into vesicles via VMAT2 (Melani and Tritsch, 2022; Tritsch et al., 2012), glutamate corelease may or may not originate from the same vesicular compartment (Hnasko et al., 2010; Silm et al., 2019; Zhang et al., 2015) and it is currently not possible to determine whether some iBioID hits might be from dedicated release sites for glutamate. Overall, however, our results fit with candidate approaches on dopamine release (Banerjee et al., 2022; Liu et al., 2018), suggesting that the proteome we describe might in general represent dopamine release sites.

Postsynaptic, unknown, and unexpected proteins in the dopamine release site proteome

There were 87 proteins with postsynaptic SynGO annotations in the release site proteome. Apart from experimental noise, multiple possibilities may account for this. First, SynGO may not have enough empirical evidence to designate any given protein as presynaptic. Some proteins with only postsynaptic annotations may be genuinely present in the dopamine release machinery. Second, dopamine axons receive input from cholinergic interneurons (Cachope et al., 2012; Threlfell et al., 2012; Zhou et al., 2001), and activation of nicotinic receptors triggers dopamine axon action potential firing (Liu et al., 2022). The structure and organization of this cholinergic input are commonly considered nonsynaptic (Chang, 1988; Jones et al., 2001), but synaptic-like transmission exists (Kramer et al., 2022). Hence, dopamine varicosities may in principle contain postsynaptic scaffolds, although nicotinic acetylcholine receptors were not detected with iBioID. Third, the estimated labeling radius of iBioID is up to 50 nm (Kim et al., 2014; Roux et al., 2012). Striatal dopamine axons make sparse synaptic-like connections with medium spiny neurons that contain the GABAergic postsynaptic scaffold Gephyrin (Uchigashima et al., 2016; Wildenberg et al., 2021). Gephyrin (Gphn) was indeed enriched with ELKS2β-BirA and RIMPPCP-BirA (Figure 2). Biotinylation across membranes may be common in iBioID, and detection of intramitochondrial proteins and of synaptic vesicle proteins with postsynaptic baits has been observed (Uezu et al., 2016). It is noteworthy that in our experiments dopamine receptors were not enriched, consistent with their localization outside of the postsynaptic scaffolds in target neurons (Uchigashima et al., 2016).

The proteome presented in Figure 2 contains 63% proteins that are not annotated in SynGO. Some of these proteins may be real hits that are not included in SynGO, while others may be experimental noise. Cend1, for example, a protein associated with cell cycle exit, appears both in our data and a previous vesicle docking proteome, and electron microscopic studies suggested that it might be present at the presynaptic plasma membrane (Boyken et al., 2013; Patsavoudi et al., 1995; Politis et al., 2007). It is surprising that Hist1h1d and Hist1h4a, genes encoding for the histone proteins H1.3 and H4, were robustly enriched (Figure 2A). These hits might be entirely unspecific or their co-purification could be due to biotinylation of H1 and H4 proteins (Stanley et al., 2001). It is also possible that there are unidentified synaptic functions of some of the unexpected proteins. An example for recently identified synaptic functions of nuclear proteins are those of the kinetochore complex (Zhao et al., 2019).

Some hits might arise because the BirA baits are not exclusively at release sites. Localization away from release sites is in part expected because a fraction of any release site protein will be away from its endogenous target localization. It might also be caused or amplified by the approach, for example, through viral bait expression and/or modification of protein sequences through bait design. The baits are made in the soma and transported along the axon. Some hits could reflect activity during protein trafficking that occurs within the 7-day time window for biotinylation. Transport complexes may be distinct between the baits and BirA-tdTomato and thus show as enriched. Indeed, several kinesins (Kif21a, Kif21b, Kif5a, Kif5b) were enriched, possibly reflecting transport packets. The synaptic ribosomes (Figure 5) that reached the 2.0-fold threshold across conditions may also reflect hits by non-active zone-localized bait proteins.

Summary of conclusions and limitations

Altogether, we present an assessment of dopamine release site content. Each protein will require validation through morphological and functional characterization before an unequivocal assignment to dopamine release sites is possible. As discussed, some hits may be noise, viral expression of tagged and modified bait proteins may influence the results, and thresholding strongly affects the overall number of proteins included in the release site proteome. We note that iBioID does not identify direct interaction partners. Instead, the defining property is proximity to the bait. The estimated ~50 nm labeling radius goes beyond the radius of protein interactions for most proteins. Despite these limitations, we have made progress. The presented data support a model of active zone-like dopamine release sites for rapid and efficient dopamine secretion. The disruption of the iBioID-generated proteome when RIM is removed establishes a scaffolding role for RIM. This role appears more pronounced in dopamine axons compared to classical synapses. Although similar proteomic approaches were not performed, RIM ablation does not severely disrupt release site structure at classical synapses when assessed with microscopy, but redundant scaffolds might maintain the sites instead (Acuna et al., 2016; de Jong et al., 2018; Emperador-Melero and Kaeser, 2020; Kaeser et al., 2011; Kushibiki et al., 2019; Oh et al., 2021; Tan et al., 2022a; Tan et al., 2022b; Wang et al., 2016). Overall, our work supports the model that dopamine release site scaffolding may have less redundancy and rely heavily on RIM (Banerjee et al., 2022; Liu et al., 2018).

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Genetic reagent (Mus musculus)B6.SJL-Slc6a3tm1.1(Cre)Bkmm/J (DATIRES-Cre)Bäckman et al., 2006RRID:IMSR_JAX:006660
Genetic reagent (M. musculus)Rims1tm3Sud/J
(RIM1αβfl/fl)
Kaeser et al., 2008RRID:IMSR_JAX:015832
Genetic reagent (M. musculus)Rims2tm1.1Sud/J
(RIM2αβγfl/fl)
Kaeser et al., 2011RRID:IMSR_JAX:015833
Genetic reagent (M. musculus)C57BL/6Ntac-Syt1tm1a(EUCOMM)Wtsi/WtsiCnrm
(Syt-1fl/fl)
Zhou et al., 2015RRID:IMSR_EM:06829The identifier refers to the line before flp recombination
Cell line (Homo sapiens)HEK293T cellsATCCCat#: CRL-3216; RRID:CVCL_0063
Recombinant DNA reagentpAAV2/5-syn-DIO-RIM-PRM-PDZ-PxxP-C2A-BirAThis studyLK18005; lab plasmid code (LPC) p864This reagent was used to generate AAV viruses and can be obtained from the corresponding author
Recombinant DNA reagentpAAV2/5-syn-DIO- ELKS2β-BirAThis studyLK17008; LPC p857This reagent was used to generate AAV viruses and can be obtained from the corresponding author
Recombinant DNA reagentAAV2/5-syn-DIO-CaVβ4-BirAThis studyLK19004; LPC p868This reagent was used to generate AAV viruses and can be obtained from the corresponding author
Recombinant DNA reagentAAV2/5-syn-DIO-BirA-tdTomatoThis studyLK17011; LPC p860This reagent was used to generate AAV viruses and can be obtained from the corresponding author
AntibodyAnti-HA (rabbit polyclonal)Cell Signaling TechnologyCAT# 5017; RRID:AB_10693385,
lab antibody code (LAC) A40
Immunofluorescence (IF) (1:500)
AntibodyAnti-HA (mouse monoclonal)BioLegendCAT# 901501;
RRID:AB_2565006,
LAC A12
IF (1:500)
Western blot (WB) 1:500
AntibodyAnti-tyrosine hydroxylase (guinea pig polyclonal)SySyCAT# 213 104;
RRID:AB_2619897,
LAC A111
IF (1:1000)
AntibodyAnti-Bassoon (mouse monoclonal)MilliporeCAT# ADI-VAM-PS003-F; RRID:AB_11181058, LAC A85IF (1:500)
AntibodyAnti-Neuroplastin (goat polyclonal)R&D SystemsCAT# AF5360;
RRID:AB_2155920,
LAC A253
IF (1:500)
AntibodyAnti-red fluorescent protein (rabbit polyclonal)RocklandCAT#600-401-379; RRID:AB_2209751, LAC A81WB (1:1000)
AntibodyAnti-β-actin (mouse monoclonal)Sigma-AldrichCAT# A1978;
RRID:AB_476692,
LAC A127
WB (1:10,000)
AntibodyAnti-pyruvate carboxylase (rabbit polyclonal)NovusCAT# NBP1-49536G; RRID:AB_11016707, LAC A252This antibody was used for depletion as described in the 'Materials and methods'
Software, algorithmPrism9GraphPadRRID:SCR002798;
https://www.graphpad.com/scientific-software/prism
Software, algorithmCytoscape v 3.8.2CytoscapeRRID:SCR003032;
http://cytoscape.org/
Software, algorithmSEQUEST PROThermo Fisherhttps://scicrunch.org/resolver/SCR_014594
Software, algorithmMATLAB code for object analysisLiu, 2021https://github.com/kaeserlab/3DSIM_Analysis_CL

Mouse lines

Request a detailed protocol

Expression via AAVs and ablation of active zone proteins in dopamine neurons was performed in mice with Cre recombinase specifically expressed in these neurons (Bäckman et al., 2006) (Jackson Laboratories; RRID:IMSR_JAX:006660, B6.SJL-Slc6a3tm1.1(Cre)Bkmm/J, also called DATIRES-Cre mice). iBioID to generate the control proteome was performed in mice heterozygote for this allele. Conditional ablation of RIM in dopamine neurons (RIM cKODA) was performed as described (Liu et al., 2018). Mice with floxed alleles for Rims1 (to remove RIM1α and RIM1β, RRID:IMSR_JAX:015832, Rims1tm3Sud/J) (Kaeser et al., 2008) and Rims2 (to remove RIM2α, RIM2β and RIM2γ, RRID:IMSR_JAX:015833, Rims2tm1.1Sud/J) (Kaeser et al., 2011) were crossed to DATIRES-Cre mice. Conditional ablation of Syt-1 in dopamine neurons (Syt-1 cKODA) was performed as described (Banerjee et al., 2020). Mice with floxed alleles for Syt-1 (RRID:IMSR_EM:06829, C57BL/6Ntac-Syt1tm1a(EUCOMM)Wtsi/WtsiCnrm) (Skarnes et al., 2011) were flp-recombined and analyzed before (Kochubey et al., 2016; Zhou et al., 2015). They were crossed to DATIRES-Cre mice as described in Banerjee et al., 2020. For generating cohorts of mice for iBioID, the floxed alleles were homozygote in both parents and one parent contained a heterozygote DATIRES-Cre allele. Male and female mice were used in all experiments irrespective of sex. All animal experiments were approved by the Harvard University Animal Care and Use Committee (protocol number IS00000049).

Plasmids, cell lines, and AAVs

Request a detailed protocol

AAVs were generated to express BirA fusion proteins in dopamine neurons. The following cDNAs were produced by standard techniques (for sequences, see source data tables): LK18005-AAV2/5-syn-DIO-RIMPPCP-BirA (also called LK18005-AAV2/5-syn-DIO-RIM-PRM-PDZ-PxxP-C2A-BirA, p864), LK17008-AAV2/5-syn-DIO-ELKS2β-BirA (p857), LK19004-AAV2/5-syn-DIO-CaVβ4-BirA (p868), and LK17011-AAV2/5-syn-DIO-BirA-tdTomato (p860). A cDNA encoding a promiscuous version of BirA (also called BirA*) was provided by S. Soderling and A. Uezu (Roux et al., 2012; Uezu et al., 2016). p864 to express RIMPPCP-BirA was generated based on a RIM full-length cDNA (de Jong et al., 2018; Kaeser et al., 2011; Tan et al., 2022b), and it contained endogenous proline-rich motifs (PRMs, also called PxxP motifs), the PDZ and C2A domains, and corresponding endogenous linkers, with BirA flanked by HA-tags added at the C-terminus. The RIMPPCP-BirA protein had the following sequence (numbered in subscript according to amino acid sequences of XM_039084276.1 for RIM and of UVH36278.1 for BirA): M-RIM1,524RPSP...GSIEQ1120-AAAYPYDVPDYA-BirA,154DNTV...SAEK472-AYPYDVPDYA. p857 to express ELKS2β-BirA was produced based on previously published N-terminally tagged ELKS2β (Kaeser et al., 2009) with removal of most of the N-terminal tag and insertion of BirA flanked by HA sequences between the coiled-coil regions CCC and CCD. The ELKS2β-BirA protein had the following sequence (numbered in subscript according to amino acid sequences of NM_170787.3 for ELKS2 and of UVH36278.1 for BirA): MGAALNRSVQTCFCSRMPCEQQICSH-ELKS2,367ELHRR…NIED657-AAAYPYDVPDYA-BirA,154DNTV….SAEK472-AYPYDVPDYAISRWRA-ELKS2,658DSRMN….GIWA921. p868 to express CaVβ4-BirA was produced from a previously used plasmid that was generated from a cDNA obtained from Annette Dolphin through Addgene (Addgene #107426; https://www.addgene.org/107426/; RRID:Addgene_107426; Brodbeck et al., 2002; Tan et al., 2022b) with addition of a linker sequence and an HA-flanked BirA protein at the C-terminus. The CaVβ4-BirA protein had the following sequence (numbered in subscript according to amino acid sequences of NM_001399143.1 for CaVβ4 and of UVH36278.1 for BirA): CaVβ4,1MSSS...RHRL519-VYNPAHNIEDAAAYPYDVPDYA-BirA,154DNTV...SAEK472-AYPYDVPDYA. p860 to express BirA-tdTomato was generated by fusing the BirA protein with tdTomato connected with a linker sequence that included an HA-tag. The BirA-tdTomato protein had the following sequence (numbered in subscript according to amino acid sequences of UVH36278.1 for BirA and of AJP62580.1 for tdTomato): M-BirA,154DNTV...SAEK472-AYPYDVPDYAGAPAS-tdTomtato,2VSKG...ELYK476. AAVs were made in HEK293T cells (purchased mycoplasma free from ATCC, CRL-3216, RRID:CVCL_0063, immortalized human cell line of female origin) using Ca2+ phosphate transfection and were of the serotype 2/5. Three days after transfection, HEK293T cells were collected and stored in freezing buffer (150 mM NaCl, 20 mM Tris-Cl, 2 mM MgCl2, pH 8.0) at –20°C until the virus was purified. For purification of AAVs, cells were lysed by three freeze-thaw cycles with dry ice/ethanol and a 37°C incubator. After 1 hr benzonase nuclease treatment at 37°C, cells were loaded onto an iodixanol gradient (5 ml each, 15%, 25%, 40%, 60%) and ultra-centrifuged at 208,000 × g for 4 hr. Viral particles were then purified from the 40% layer of the gradient. Quantitative reverse transcriptase PCR was used to determine viral titers, and viruses were used at concentrations ranging from 4.0 × 1011 to 9.6 × 1012 viral genome copies/ml.

Stereotaxic surgery and biotin injections

Request a detailed protocol

Mice (at postnatal days 30–55) were anesthetized in a 5% isoflurane induction chamber and then mounted on a stereotaxic frame; anesthesia was maintained with 1.5–2% isoflurane for the length of the surgery with a nose cone. The scalp was cut open and a hole was drilled in the skull and 1 µl of AAV viral solution was injected in the substantia nigra pars compacta (right or bilaterally depending on the experiment, 0.6 mm anterior, ±1.3 mm lateral of Lambda and 4.2 mm below the surface of the brain) using a microinjector pump (PHD ULTRA, Harvard Apparatus) at 100 nl/min. Mice were treated with post-surgical analgesia and were allowed to recover for at least 28 days prior to biotin injections (for iBioID) or transcardial perfusion (for morphological experiments). Biotin injections were started 4–6 weeks after stereotaxic AAV injection. Mice were subcutaneously injected for seven consecutive days with 500 µl of 5 mM biotin dissolved in phosphate-buffered saline (PBS).

Immunostaining and confocal imaging of brain sections

Request a detailed protocol

At least 4 weeks after stereotaxic injection of BirA viruses, mice (58–100 days old) were deeply anesthetized with isoflurane. Transcardial perfusion was performed with 30–50 ml ice-cold PBS followed by 50 ml of 4% paraformaldehyde in PBS (4% PFA) at 4°C. Brains were then dissected out and incubated in 4% PFA for overnight at 4°C. Fixed brains were sliced on a vibratome (Leica, VT1000s) at 100 μm thickness. Coronal sections containing the midbrain and striatum were collected in ice-cold PBS. Sections were blocked in PBS containing 0.25% Triton X-100 and 10% goat serum (PBST) for 1 hr at room temperature. Slices were incubated overnight in primary antibody in PBST at 4°C. The following primary antibodies were used: rabbit polyclonal anti-HA (1:500, A40, RRID:AB_10693385) or mouse monoclonal anti-HA (1:500, A12, RRID:AB_2565006), and guinea pig polyclonal anti-TH (1:1000, A111, RRID:AB_2619897). Slices were washed three times in PBST followed by 2 hr incubation with secondary antibody in PBST for 2 hr at room temperature in the dark. The following secondary antibodies were used: goat anti-rabbit IgG Alexa 488 (1:500, S5, RRID:AB_2576217), goat anti-mouse IgG1 Alexa 488 (1:500, S7, RRID:AB_2535764), and goat anti-guinea pig IgG Alexa 633 (1:500, S34, RRID:AB_2535757). Slices were washed again three times with PBST before being mounted on glass slides with Fluoromount-G (Southern Biotech 0100-01). Stained slices were then imaged on an Olympus FV1000 confocal microscope with a 60× objective. Images were pseudo-colored in ImageJ for display. All image acquisition was done in comparison to an uninfected control imaged at the same time to assess background fluorescence. Representative images were brightness and contrast adjusted to facilitate inspection, and these adjustments were made identically for images within the same experiment. All images in Figure 1C except for those of mice expressing ELKS2β-BirA were taken at the same time. Images of mice expressing ELKS2β-BirA were taken in a separate session and compared to their own uninfected control to confirm signal specificity (Figure 1—figure supplement 1). Three to five images per virus condition were taken during each imaging session, and the experiment was repeated in at least three mice per condition.

Biotin affinity purification

Request a detailed protocol

Biotin affinity purifications were adapted from previously established methods (Uezu et al., 2016), 2 (for pilot experiments, up to 4) to 12 mice (for mass spectrometry analyses) were used per experiment and condition. The mice were 65–100 days old and previously injected with AAVs for BirA fusion protein expression and subjected to subcutaneous biotin injections. They were deeply anesthetized with isoflurane and decapitated. Brains were collected in ice-cold PBS and striata were dissected out. Except for those used in pilot experiments, striata were then flash frozen and stored at –80°C until further processing. For pilot experiments, 2–4 dissected striata were immediately homogenized. For mass spectrometry, 5–6 dissected striata were homogenized at a time. Homogenization was performed using a glass-Teflon homogenizer in 1 ml of homogenizing buffer (50 mM HEPES, 150 mM NaCl, 1 mM EDTA+ mammalian protease inhibitor cocktail, Sigma Cat# P8340) with 30 slow strokes on ice. An appropriate volume of 5× lysis buffer (1% SDS, 5% Triton X-100, 5% deoxycholate in homogenizing buffer) was added to the homogenized tissue (working concentration 0.2% SDS, 1% Triton X-100, 1% deoxycholate) and incubated while rotating at 4°C for 1 hr. The mass spectrometry samples were split into two batches of 5–6 striata per condition for the initial homogenization, and the batches were combined after lysis into a single tube. Lysed samples were then sonicated twice for 10 s each at 4°C with a Branson Sonifier 450. Sonicated samples were centrifuged at 15,000 × g for 15 min at 4°C in a table-top centrifuge. The cleared supernatant was removed and added to open-top polycarbonate tubes (Beckman Cat# 343778), and centrifuged in a table-top ultracentrifuge (Beckman Rotor TLA120.2) for 1 hr at 100,000 × g. After ultracentrifugation, SDS from 0.4% and/or 5% SDS stock solutions was added to the cleared supernatant to adjust to an SDS concentration of 1% and a volume of 1.5 ml. The sample was boiled at 95°C for 5 min and allowed to cool to room temperature. Neutravidin agarose beads (Thermo Cat# 29200) were washed three times in binding buffer (1% SDS, 1% Triton X-100, 1% deoxycholate in homogenizing buffer). 20 µl washed neutravidin beads were added to each sample and the samples were incubated for 16 hr at 4°C on a rotator. After spinning at 500 × g for 2 min at 4°C, the supernatant was removed and the beads were transferred to Protein Lo-bind tubes (Eppendorf Cat# 022431081). Beads were washed twice with 500 µl 2% SDS in H2O, twice with 500 µl 1% Triton X-100, 1% deoxycholate, 25 mM LiCl in H2O, and twice in 500 µl 1 M NaCl. For each washing step, beads were pelleted by spinning at 500 × g for 2 min in a table-top centrifuge at 4°C. Bead pellets were then washed five times in 500 µl 50 mM ammonium bicarbonate in water with spinning at 500 × g for 2 min between steps. After the final wash, the bead pellet was stored at –20°C until next steps (either Western blot or PC removal).

Western blot after biotin pulldown

Request a detailed protocol

Neutravidin bead pellets were incubated in 60 µl of 1× SDS-PAGE loading buffer and boiled for 10 min at 95°C. The sample was spun at 13,000 × g in a table-top centrifuge for 1 min to pellet the beads. 15 µl of the supernatant were loaded on an SDS-PAGE gel, and 1% of the total input used for the binding reaction (cleared lysate just before addition of the neutravidin beads) was also loaded onto the gel. After gel electrophoresis, proteins were transferred onto a nitrocellulose membrane. Membranes were blocked in 10% milk, 5% goat serum in Tris-buffered saline + 0.1% Tween-20 (TBST) for 1 hr at room temperature and then incubated overnight at 4°C in primary antibody diluted in antibody binding solution (blocking solution diluted 1:1 with TBST). Primary antibodies used: mouse monoclonal anti-HA (1:500, LAC A12, RRID:AB_2565006), rabbit polyclonal anti-red fluorescent protein (RFP) (1: 500, LAC A81, RRID:AB_2209751), rabbit polyclonal anti-pyruvate carboxylase (1:500, LAC A252, RRID:AB_11016707), and mouse monoclonal anti-β-actin (1:10,000, LAC A127, RRID:AB_476692). Membranes were washed three times with TBST and incubated in HRP-conjugated secondary antibodies for 1 hr at room temperature. Membranes were washed three times in TBST and enhanced chemiluminescence followed by exposure to film was used to visualize protein bands.

Pyruvate carboxylase depletion

Request a detailed protocol

For samples being submitted to mass spectrometry, frozen neutravidin bead pellets were thawed on ice. Proteins were eluted by incubation in 500 µl RapiGest elution buffer (0.1% RapiGest, Waters Cat# 1866001861, in 2 mM biotin, 50 mM ammonium bicarbonate) for 2 hr at 60°C, while shaking. Neutravidin beads were pelleted by spinning at 18,000 × g for 5 min at 4°C and the supernatant was moved to a new tube. To prepare anti-pyruvate carboxylase (PC) antibody-conjugated beads, Protein G Sepharose beads (GE Healthcare Cat#17-0618-01) were washed three times in RapiGest elution buffer and conjugated to anti-PC antibodies (Novus, A252, RRID:AB_11016707) by incubating 3 μl of antibody per 20 µl of beads for 1 hr at 4°C on a rotator. Conjugated beads were spun down at 1000 × g for 2 min, the supernatant was removed, and beads were diluted 1:1 with fresh RapiGest elution buffer to make a 50% slurry. 20 µl sepharose beads (40 µl of a 50% slurry) were added to the 500 µl of eluted protein solution of the biotin affinity purification and incubated for 1 hr at 4°C on a rotator. PC-conjugated beads were pelleted by spinning the sample at 18,000 × g for 5 min. The proteins in the PC-depleted supernatant were then precipitated with trichloroacetic acid (TCA). One volume of 100% TCA was added to four volumes of the PC-depleted supernatant and inverted several times before incubation on ice for 10 min. Tubes were then spun at 20,000 × g for 10 min at 4°C. The supernatant was removed leaving behind a protein pellet. The pellet was air dried and stored at –20°C until analysis by mass spectrometry. In pilot experiments, we assessed the efficiency of PC depletion. Before PC depletion, 28% and 10% of the detected peptides were from PC and the related protein propionyl-CoA carboxylase (PCCA), respectively. After depletion, 3% and 1.7% of the detected peptides were from PC and PCCA, respectively.

Mass spectrometry

Request a detailed protocol

Liquid chromatography tandem mass spectrometry (LC-MS/MS) was performed by the Taplin Mass Spectrometry Facility of the Department of Cell Biology at Harvard Medical School. Either the neutravidin bead pellet (some pilot experiments) or the TCA-precipitated protein pellet after PC depletion (all experiments to assess release site composition) were submitted to the facility. The pellet was subjected to 5 ng/µl trypsin digest overnight at 37°C and dried until further analysis. On the day of analysis, the samples were reconstituted in 2.5% acetonitrile, 0.1% formic acid, and loaded via a Famos auto sampler (LC Packings, San Francisco, CA) onto the column after column equilibration. Peptides were eluted with increasing concentrations of 97.5% acetonitrile and 0.1% formic acid. As peptides were eluted, they were subjected to electrospray ionization and added to an LTQ Orbitrap Velos Pro ion-trap mass spectrometer (Thermo Fisher Scientific). Peptides were detected, isolated, and fragmented to produce a tandem mass spectrum of specific fragment ions for each peptide. Peptide sequences (and hence protein identity) were determined by matching protein databases with the acquired fragmentation pattern by the SEQUEST software program (Thermo Fisher Scientific). The data were filtered to establish a false discovery rate between 1% and 2% using a database of mouse protein sequences made up of half real protein sequences and half reversed sequences. Independent of whether the results were run against a mouse or a rat database, the number of peptides identified for bait proteins (which were made from rat cDNA) was similar.

Analyses of mass spectrometry data

Request a detailed protocol

The control dataset consists of four repeats that stem from two separate mass spectrometry sessions (repeats 1 and 2 were run first, repeats 3 and 4 were run ~1 year later). Each mutant dataset (RIM cKODA and Syt-1 cKODA) consists of two repeats that were assessed in the same mass spectrometry session together with repeats 3 and 4 of the control dataset. Mitochondrial proteins were removed from the analyses using MitoCarta3.0 (Pagliarini et al., 2008; Rath et al., 2021), except for Figures 3B and 6. PC and PCCA were also removed from the analyses along with neutravidin and IgGs. Peptide counts for each protein in each BirA condition (RIMPPCP-BirA, ELKS2β-BirA, CaVβ4-BirA, BirA-tdTomato) were first averaged for the two repeats that were acquired in the same mass spectrometry session. If the peptide count averaged from the two repeats was zero, it was assigned a value of 0.5 such that we could calculate circle size for Figure 2 and more generally enrichment or depletion for all analyses. This assignment warranted that for a protein to be included, it had to be detected more than once with a single peptide. The average peptide count for a given protein for each of the three BirA baits was divided by the average peptide count for the same protein in the BirA-tdTomato condition. This is referred to as the fold change (FC) value. For the control dataset, the average fold change values of each mass spectrometry session were then averaged together to generate a final fold change value for every protein shown in Figure 2 and Figure 2—figure supplements 1 and 2; in the mutant datasets, this final averaging was not performed as all data stem from a single mass spectrometry session. Proteins with at least 2.0-fold enrichment over BirA-tdTomato in any release site-BirA condition were considered 'hits' in Figure 2 and alternate thresholds of 1.5-fold and 2.5-fold enrichment are shown in Figure 2—figure supplements 1 and 2. The Venn diagrams were constructed using Cytoscape (v3.8.2) (Shannon et al., 2003) to map the circle size to fold change value. Circle size reflects average enrichment across repeats of an individual bait. For proteins enriched in multiple bait conditions, circle size corresponds to the condition with the largest average enrichment value. The log2 of the fold change was calculated for any protein that had at least a single peptide found for any of the BirA fusion proteins. To sort proteins for synaptic localization, either the enriched proteins (Figures 2B and 5, Figure 2—figure supplement 3, Figure 3—figure supplements 1 and 2) or all proteins (Figure 4A–C, Figure 4—figure supplement 1) were run through the SynGO database (https://www.syngoportal.org; Koopmans et al., 2019) and selected by their cellular component annotation. For Figure 4A–C and Figure 4—figure supplement 1, the self-biotinylated protein was removed from its own release site-BirA condition (e.g., RIM was removed from analysis in the RIMPPCP-BirA dataset but not in the CaVβ4- or ELKS2β-BirA datasets). For Figure 5, the enriched proteins that possessed SynGO annotations were run through the STRING database (v11) (https://string-db.org; Szklarczyk et al., 2019) to generate a STRING network diagram of enriched synaptic proteins.

Synaptosome preparation and staining

Request a detailed protocol

Synaptosome experiments were performed according to previously established protocols (Banerjee et al., 2022; Liu et al., 2018). Mice (30–60 days old) were deeply anesthetized with isoflurane and decapitated. The brain was collected in ice-cold PBS, and the striatum was dissected out. The striatal sections were homogenized in 1 ml of synaptosome homogenizing buffer (320 mM sucrose, 4 mM HEPES, 1× Sigma protease inhibitor cocktail [Cat# P8340], pH 7.4), with 12 slow strokes using a glass-Teflon homogenizer. One ml of homogenizing buffer was then added to the homogenate. The homogenate was spun at 1000 × g for 10 min at 4°C, the supernatant was pipetted into a new tube, and spun again at 12,500 × g for 15 min at 4°C. The pellet ('P2 fraction') was resuspended in 1 ml homogenizing buffer and homogenized again with six slow strokes. An additional 1 ml of homogenizing buffer was added, and the sample (~1.5 ml) was loaded onto a sucrose gradient made up of 5 ml 1.2 M sucrose on the bottom and 5 ml of 0.8 M sucrose at the top in thin wall ultracentrifugation tubes (Beckman Coulter, Cat# 344059). The loaded gradient was ultracentrifuged at 69,150 × g for 1 hr at 4°C (SW 41 Ti Swinging-Bucket Rotor, Beckman Coulter, Cat# 331362) and the synaptosome fraction was harvested from the interface between the two sucrose layers. The synaptosome fraction was diluted 20- to 40-fold in homogenizing buffer and 1 ml was added to a poly-D-lysine-coated coverslip (neuVitro Cat# GG-18-1.5) in a 12-well plate and spun for 15 min at 4000 × g. The buffer was removed and the synaptosomes adhering to the coverslips were fixed with 4% PFA in PBS for 10 min. The PFA was removed, and a solution with 3% bovine serum albumin and 0.1% Triton X-100 in PBS (TBP) was used for blocking and permeabilization for 1 hr at room temperature. The following primary antibodies were used (diluted in TBP) overnight at 4°C: goat polyclonal anti-NPTN (1:500, A253, RRID:AB_2155920), mouse monoclonal IgG2a anti-Bassoon (1:1000, A85, RRID:AB_11181058), and guinea pig polyclonal anti-TH (1:1000, A111, RRID:AB_2619897). After primary antibody incubation, the coverslips were washed three times in TBP and then incubated in secondary antibodies for 2 hr at room temperature in the dark. Secondary antibodies used were donkey anti-goat Alexa 488 (1:500, S6, RRID:AB_2534102), donkey anti-mouse Alexa 555 (1:500, S48, RRID:AB_2534013), and donkey anti-guinea pig Alexa 647 (1:500, S59, RRID:AB_2340476). The stained synaptosomes were washed three more times in TBP before mounting on glass slides with Fluoromount-G (Southern Biotech 0100-01).

Confocal microscopy and image analyses of stained synaptosomes

Request a detailed protocol

Coverslips with fixed synaptosomes were imaged with an oil immersion 60× objective and a 1.5× optical zoom on an Olympus FV1000 confocal microscope. Raw confocal images were analyzed in a custom MATLAB program as described before (Banerjee et al., 2022; Liu et al., 2018) and the code was deposited on GitHub (https://github.com/kaeserlab/3DSIM_Analysis_CL; Liu, 2021). 1000–2000 synaptosomes per image were detected using Otsu intensity thresholds and size thresholds (0.2–1 µm2 for TH and 0.15–2 µm2 for Neuroplastin and Bassoon). These threshold settings were identical for every image analyzed and used to detect Bassoon-positive (Bassoon+) ROIs, TH-positive (TH+) ROIs, and double-positive ROIs (Bassoon+/TH+). TH+ ROIs that had a Bassoon signal less than 1× the average intensity of all pixels in the image were designated as Bassoon-negative (Bassoon-). Neuroplastin antibody signal intensities within Bassoon+ TH+ or Bassoon- TH+ ROIs were quantified and a frequency histogram was plotted. The experimenter was blind to which group a specific particle belonged to during image acquisition and analyses.

Statistics

Statistics were performed in GraphPad Prism 9. Data are displayed as individual data points, mean ± SEM, and/or frequency histograms. Significance is presented as *p<0.05, **p<0.01, and ***p<0.001. Sample sizes were determined based on previous studies, no statistical methods were used to predetermine sample size, and no outliers were excluded. An unpaired two-tailed Student’s t-tests was used in Figure 2—figure supplement 4B, a Kolmogorov–Smirnov test was used in Figure 2—figure supplement 4C, a two-way ANOVA test was used in Figure 3B, and two-way ANOVA tests followed by Bonferroni post-hoc tests were used in Figure 4A–C and Figure 4—figure supplement 1. In all figures, sample sizes and the specific tests used are stated in the figure legends.

Materials, data and code availability

Request a detailed protocol

Plasmids will be shared upon request within the limits of respective material transfer agreements. Mouse mutant alleles are publicly available as outlined in the Key Resources Table. The previously published code used for analyses of synaptosomes has been deposited to GitHub (https://github.com/kaeserlab/3DSIM_Analysis_CL; Liu, 2021) and is publicly available as listed in the Key Resources Table. All data generated or analyzed in this study are included in the figures and the source data tables. Source data files are provided for Figures 13, Figure 1—figure supplement 2, Figure 2—figure supplements 1 and 2, and Figure 3—figure supplements 1 and 2. Requests for materials should be directed to the corresponding author (kaeser@hms.harvard.edu).

Data availability

All data generated or analyzed in this study are included in the figures and the source data tables. Source data files are provided for Figs. 1 to 3, Fig. 1 - figure supplement 2, Fig. 2 - figure supplements 1 and 2, and Fig. 3 - figure supplements 1 and 2.

References

Decision letter

  1. Reinhard Jahn
    Reviewing Editor; Max Planck Institute for Biophysical Chemistry, Germany
  2. Gary L Westbrook
    Senior Editor; Oregon Health & Science University, United States
  3. Reinhard Jahn
    Reviewer; Max Planck Institute for Biophysical Chemistry, Germany
  4. Matthijs Verhage
    Reviewer; VU University Amsterdam, Netherlands

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Protein composition of axonal dopamine release sites in the striatum" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Reinhard Jahn as Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Gary Westbrook as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Matthijs Verhage (Reviewer #3).

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

All reviewers agree that the approaches chosen to identify proteins localized to dopaminergic release sites are excellent and cutting-edge. However, the reviewers also agree that while the data provide a valuable and useful resource for future work, both analysis of the data and their interpretation should be carefully reconsidered and revised accordingly:

1. It is not clear to the reviewers, how robust the findings are with respect to the thresholds used for the enrichment analysis (see comments by Ref. 3). For instance, how sensitive are the major trends towards increasing or decreasing the threshold values by e.g. 0.5 points?

2. The discussion should be reconsidered. As pointed out by Ref. 3, it is partially redundant whereas other relevant issues (e.g. completeness of the release site proteome, "oddities" in the data) are not mentioned. Moreover, as pointed out by Ref. 1, a more critical appraisal of the limitations of the chosen labeling procedure should be provided.

Reviewer #1 (Recommendations for the authors):

In this study, the protein composition of exocytotic sites in dopaminergic neurons is investigated. While extensive data are available for both glutamatergic and GABA-ergic synapses, it is far less clear which of the known proteins (particularly proteins localized to the active zone) are also required for dopamine release, and whether proteins are involved that are not found in "classical" synapses. The approach used here uses proximity ligation to tag proteins close to synaptic release sites by using three presynaptic proteins (ELKS, RIM, and the beta4-subunit of the voltage-gated calcium channel) as "baits". Fusion proteins containing BirA were selectively expressed in striatal dopaminergic neurons, followed by in-vivo biotin labelling, isolation of biotinylated proteins and proteomics, using proteins labelled after expression of a soluble BirA-construct in dopaminergic neurons as reference. As controls, the same experiments were performed in KO-mouse lines in which the presynaptic scaffolding protein RIM or the calcium sensor synaptotagmin 1 were selectively deleted in dopaminergic neurons. To control for specificity, the proteomes were compared with those obtained by expressing a soluble BirA construct. The authors found selective enrichments of synaptic and other proteins that were disrupted in RIM but not Syt1 KO animals, with some overlap between the different baits, thus providing a novel and useful dataset to better understand the composition of dopaminergic release sites.

Technically, the work is clearly state-of-the-art, cutting-edge, and of high quality, and I have no suggestions for experimental improvements. On the other hand, the data also show the limitations of the approach, and I suggest that the authors discuss these limitations in more detail. The problem is that there is very likely to be a lot of non-specific noise (for multiple reasons) and thus the enriched proteins certainly represent candidates for the interactome in the presynaptic network, but without further corroboration it cannot be claimed that as a whole they all belong to the proteome of the release site.

For instance, in Figure 2 it is shown that a surprisingly high number of postsynaptic proteins are labelled in all three datasets, showing that labelling is not as specific as hoped for (I acknowledge that the authors try to explain this with the presence of neighbouring postsynaptic sites but are they really this close?). The same applies to other proteins that are not expected to be enriched. One of the reasons may be the rather long labelling periods (7 days) which are likely to increase the number of labelled proteins. Recent work from the Rizzoli lab has shown that the half-life of synaptic proteins is around 7-10 days, which means that between 30 and 40 % of the proteins used as baits are turned over during the labelling period. Thus, it is to be expected that labelling is not restricted to the neighbours in the active zone but includes proteins interacting with the baits during their life cycle after biosynthesis and folding at the ribosome, during axonal transport and also during removal from the active zone and transport towards degradative pathways. In this respect, the comparison to the RIM-KO dataset becomes important (Figure 4). In this experiment: what happens to the postsynaptic proteins, to proteins of the presynaptic plasma membrane? Are they still present in the enriched samples?

Reviewer #2 (Recommendations for the authors):

The Kaiser lab has been on the forefront in understanding the mechanism of dopamine release in central mammalian neurons. assessing dopamine neuron function has been quite difficult due to the limited experimental access to these neurons. Dopamine neurons possess a number of unique functional roles and participate in several pathophysiological conditions, making them an important target of basic research. This study here has been designed to describe the proteome of the dopamine release apparatus using proximity biotin labeling via active zone protein domains fused to BirA, to test in which ways its proteome composition is similar or different to other central nerve terminals. The control experiments demonstrating proper localization as well as specificity of biotinylation are very solid, yielding in a highly enriched and well characterized proteome data base. Several new proteins were identified and the data base will very likely be a very useful resource for future analysis of the protein composition of synapse and their function at dopamine and other synapses.

The authors find that loss of RIM leads to major reduction in the number of synaptically enriched proteins, while they did not see this loss of number of enriched proteins in the Syt1-KO's, arguing for undisrupted synaptome. Maybe I missed this, but which fraction of proteins and synaptic proteins are than co-detected both in the Syt1 and control conditions when comparing the Venn diagrams of Figure 2 and Figure 3 Suppl. 2? This analysis may provide an estimate of the reliability of the method across experimental conditions.

Reviewer #3 (Recommendations for the authors):

In this study Kershberg et al. use three novel in vivo biotin-identification (iBioID) approaches in mice to isolate and identify proteins of axonal dopamine release sites. By dissecting the striatum, where dopamine axons are, from the substantia nigra and VTA, where dopamine somata are, the authors selectively analyzed axonal compartments. Perturbation studies were designed by crossing the iBioID lines with null mutant mice. Combining the data from these three independent iBioID approaches and the fact that axonal compartments are separated from somata provides a precise and valuable description of the protein composition of these release sites, with many new proteins not previously associated with synaptic release sites. These data are a valuable resource for future experiments on dopamine release mechanisms in the CNS and the organization of the release sites. The BirA (BioID) tags are carefully positioned in three target proteins not to affect their localization/function. Data analysis and visualization are excellent. Combining the new iBioID approaches with existing null mutant mice produces powerful perturbation experiments that lead and strong conclusions on the central role of RIM1 as central organizers of dopamine release sites and unexpected (and unexplained) new findings on how RIM1 and synaptotagmin1 are both required for the accumulation of α-synuclein at dopamine release sites.

It is not entirely clear how certain decisions made by the authors on data thresholds may affect the overall picture emerging from their analyses. This is a purely hypothesis-generating study. The authors made little efforts to define expectations and compare their results to these. Consequently, there is little guidance on how to interpret the data and how decisions made by the authors affect the overall conclusions. For instance, the collection of proteins tagged by all three tagging strategies (Figure 2) is expected to contain all known components of dopamine release sites (not at all the case), and maybe also synaptic vesicles (2 TM components detected, but not the most well-known components like vSNAREs and H+/DA-transporters), and endocytic machinery (only 2 endophilin orthologs detected). Whether or not a more complete collection the components of release sites, synaptic vesicles or endocytic machinery are observed might depend on two hard thresholds applied in this study: (a) "Hits" (depicted in Figure 2) were defined as proteins enriched {greater than or equal to} 2-fold (line 178) and peptides not detected in the negative control (soluble BirA) were defined as 0.5 (line 175). How crucial are these two decisions? It would be great to know if the overall conclusions change if these decisions were made differently.

Given the good separation of the axonal compartment from the somata (one of the real experimental strengths of this study), it is completely unexpected to find two histones being enriched with all three tagging strategies (Hist1h1d and 1h4a). This should be mentioned and discussed.

It would also help to compare the data more systematically to a previous study that attempted to define release sites (albeit not dopamine release sites) using a different methodology (biochemical purification): Boyken et al. (only mentioned in relation to Nptn, but other proteins are observed in both studies too, e.g. Cend1).

The Discussion section could be improved. It is rather long (7.5p) with some repetition from the Introduction, some issues already demonstrated convincingly before (2.5p on RIM being a scaffold in DA neurons) and a rather speculative discussion on RIM being associated with Parkinson's disease through dopamine release site recruitment of α-synuclein, possibly via protein interactions or through vesicle docking (but how is release site recruitment of α-synuclein important for protein interactions, vesicle docking or Parkinson symptoms?). Instead some of the more pressing questions (at least to this reviewer) are not discussed, such as the expectations mentioned above: why were expected components of release sites not detected, why only some but not all expected components of synaptic vesicles/endocytic machinery and why are two histones in the axonal compartment?

https://doi.org/10.7554/eLife.83018.sa1

Author response

1. It is not clear to the reviewers, how robust the findings are with respect to the thresholds used for the enrichment analysis (see comments by Ref. 3). For instance, how sensitive are the major trends towards increasing or decreasing the threshold values by e.g. 0.5 points?

We agree that thresholding criteria are critical decisions, both for our work and for papers using similar techniques. To illustrate why we chose 2.0 as an enrichment threshold beyond justification through literature 1,2, we have included new figures showing analyses of the main release site proteome at varying thresholds. Figure 2 presents the 2.0 enrichment threshold, while Figure 2 —figure supplements 1 and 2 present the same data with 1.5 and 2.5 thresholds, respectively. Naturally, the overall size of the release site proteome varies considerably with 527 hits at 2.0, 991 hits at 1.5, and 348 hits at 2.5 enrichment thresholds. We then took each thresholded proteome and analyzed it with SynGO to determine the fraction of proteins previously associated with synapses (Figure 2 —figure supplement 3). SynGO-annotated proteins were 37% (194 of 527) for 2.0, 31% (304 of 991) for 1.5, and 33% (116 of 348) for 2.5. Based on these analyses, and consistent with literature on other iBioID experiments, we think it is appropriate to select 2.0 as the threshold for further analyses and comparisons. The data that led to this decision are now fully incorporated in the manuscript, including the figures (Figure 2 —figure supplements 1 to 3) and a new paragraph in the Results section on lines 208-215. We further have added a summary of limitations together with conclusions in the last section of the discussion that includes the point of thresholding. This section is captioned “Summary of conclusions and limitations” (lines 501-518).

To complement the presentation of the control proteome with various thresholds, we also determined the number of hits in the release site proteomes of RIM cKODA and Syt-1 cKODA mice at these thresholds. The major finding that RIM ablation disrupts this proteome was independent of thresholding, with 268 hits at 1.5, 198 at 2.0 and 82 hits at 2.5 enrichment thresholds. The overall number of hits in Syt-1 cKODA mice remained higher, with 602, 450 and 252 at the 1.5, 2.0 and 2.5 enrichment thresholds, respectively. This is included in the revised manuscript on lines 298-301. Overall, the main findings are not affected by thresholding, but the numbers for proteins in each proteome vary based on threshold.

2. The discussion should be reconsidered. As pointed out by Ref. 3, it is partially redundant whereas other relevant issues (e.g. completeness of the release site proteome, "oddities" in the data) are not mentioned. Moreover, as pointed out by Ref. 1, a more critical appraisal of the limitations of the chosen labeling procedure should be provided.

We agree, and we have made the requested adjustments. Overall, we have shortened the discussion from 2400 words to 1900 words. We have removed redundancies between the discussion and the Results section and have included additional discussion of unexpected hits (for example: histones, lines 482-485; transport proteins, lines 493-497). We have also included a section on summarizing limitations, together with key conclusions, at the end of the discussion (lines 501-518). We specifically include a more in-depth discussion of noise, of oddities, of thresholding and of the fundamental point that the proteome is based on proximity, not on interaction.

Reviewer #1 (Recommendations for the authors):

In this study, the protein composition of exocytotic sites in dopaminergic neurons is investigated. While extensive data are available for both glutamatergic and GABA-ergic synapses, it is far less clear which of the known proteins (particularly proteins localized to the active zone) are also required for dopamine release, and whether proteins are involved that are not found in "classical" synapses. The approach used here uses proximity ligation to tag proteins close to synaptic release sites by using three presynaptic proteins (ELKS, RIM, and the beta4-subunit of the voltage-gated calcium channel) as "baits". Fusion proteins containing BirA were selectively expressed in striatal dopaminergic neurons, followed by in-vivo biotin labelling, isolation of biotinylated proteins and proteomics, using proteins labelled after expression of a soluble BirA-construct in dopaminergic neurons as reference. As controls, the same experiments were performed in KO-mouse lines in which the presynaptic scaffolding protein RIM or the calcium sensor synaptotagmin 1 were selectively deleted in dopaminergic neurons. To control for specificity, the proteomes were compared with those obtained by expressing a soluble BirA construct. The authors found selective enrichments of synaptic and other proteins that were disrupted in RIM but not Syt1 KO animals, with some overlap between the different baits, thus providing a novel and useful dataset to better understand the composition of dopaminergic release sites.

Technically, the work is clearly state-of-the-art, cutting-edge, and of high quality, and I have no suggestions for experimental improvements.

We thank the reviewer for this summary and for pointing out the high quality of the work.

On the other hand, the data also show the limitations of the approach, and I suggest that the authors discuss these limitations in more detail. The problem is that there is very likely to be a lot of non-specific noise (for multiple reasons) and thus the enriched proteins certainly represent candidates for the interactome in the presynaptic network, but without further corroboration it cannot be claimed that as a whole they all belong to the proteome of the release site.

We fully agree with the reviewer. Most importantly, we have changed the final section from “Conclusions” to “Summary of conclusions and limitations” (lines 501-518) to summarize the limitations with equal weight to the conclusions. In the revised manuscript, we also included many specific additional points in this respect throughout the discussion and the results: many hits could be noise (lines 458, 478-479), thresholding affects the inclusion of proteins in the release site dataset (lines 208-215), the seven-day time window could deliver interactors from the soma to the synapse (lines 493-495), specific oddities (for example histones, lines 482-485), iBioID does not deliver an interactome per se but is simply based on proximity (lines 505-508), and several more. We also clearly state that each specific hit needs follow-up studies (lines 501-503: “Each protein will require validation through morphological and functional characterization before an unequivocal assignment to dopamine release sites is possible.”), and a similar statement was added on lines 374-375.

For instance, in Figure 2 it is shown that a surprisingly high number of postsynaptic proteins are labelled in all three datasets, showing that labelling is not as specific as hoped for (I acknowledge that the authors try to explain this with the presence of neighbouring postsynaptic sites but are they really this close?). The same applies to other proteins that are not expected to be enriched. One of the reasons may be the rather long labelling periods (7 days) which are likely to increase the number of labelled proteins. Recent work from the Rizzoli lab has shown that the half-life of synaptic proteins is around 7-10 days, which means that between 30 and 40 % of the proteins used as baits are turned over during the labelling period. Thus, it is to be expected that labelling is not restricted to the neighbours in the active zone but includes proteins interacting with the baits during their life cycle after biosynthesis and folding at the ribosome, during axonal transport and also during removal from the active zone and transport towards degradative pathways. In this respect, the comparison to the RIM-KO dataset becomes important (Figure 4). In this experiment: what happens to the postsynaptic proteins, to proteins of the presynaptic plasma membrane? Are they still present in the enriched samples?

We fully agree with the reviewer and have added these points to the results, discussion, and limitations sections mentioned above.

The RIM cKODA and control datasets reveal that axonal proteins assessed with BirA-tdTomato have high overlap, with 620 out of the 671 found in control also present in RIM cKODA. This suggests that the overall protein content is very similar. Importantly, in the release site dataset, presynaptic proteins decrease from 99 hits in the control to 43 hits in RIM cKODA. Synaptic vesicle proteins decreased from 30 to 11, active zone proteins from 14 to 4, and postsynaptic proteins from 87 to 37. We now added these numbers to the legend of Figure 3 —figure supplement 1 and discuss postsynaptic proteins on lines 457-475. We have now also added analyses of postsynaptic proteins identical to how we assessed synaptic, presynaptic and active zone proteins. As we observed for the presynaptic and active zone annotations, RIM cKODA depleted postsynaptic proteins. This supports that enrichment of these proteins is specific and does not simply reflect noise. These new analyses are now included in Figure 4 —figure supplement 1.

These analyses support that the hits in the control dataset are generally specific and that the postsynaptic hits are not just noise; if they were noise, they should be similar between the control proteome and the RIM cKODA proteome. These proteins could either be detected because they are also present in dopamine axons, and/or because iBioID biotinylates proteins across membranes as long as they are close enough. We suspect that both factors contribute. We think that the latter point also matches with data from others: if PSD proteins are used as iBioID baits, synaptic vesicle and active zone proteins are generally detected 2. Ultimately, we think that this may reflect that the pre- and postsynaptic compartments are within the labeling radius of each other and that reactive biotin can pass through lipid bilayers. The latter point is also vividly demonstrated by the notion that our and published proteomes contain intra-mitochondrial proteins 2,3.

Reviewer #2 (Recommendations for the authors):

The Kaiser lab has been on the forefront in understanding the mechanism of dopamine release in central mammalian neurons. assessing dopamine neuron function has been quite difficult due to the limited experimental access to these neurons. Dopamine neurons possess a number of unique functional roles and participate in several pathophysiological conditions, making them an important target of basic research. This study here has been designed to describe the proteome of the dopamine release apparatus using proximity biotin labeling via active zone protein domains fused to BirA, to test in which ways its proteome composition is similar or different to other central nerve terminals. The control experiments demonstrating proper localization as well as specificity of biotinylation are very solid, yielding in a highly enriched and well characterized proteome data base. Several new proteins were identified and the data base will very likely be a very useful resource for future analysis of the protein composition of synapse and their function at dopamine and other synapses.

We thank the reviewer for this positive assessment of our work.

The authors find that loss of RIM leads to major reduction in the number of synaptically enriched proteins, while they did not see this loss of number of enriched proteins in the Syt1-KO's, arguing for undisrupted synaptome. Maybe I missed this, but which fraction of proteins and synaptic proteins are than co-detected both in the Syt1 and control conditions when comparing the Venn diagrams of Figure 2 and Figure 3 Suppl. 2? This analysis may provide an estimate of the reliability of the method across experimental conditions.

We thank the reviewer for proposing to be clear in the comparison of the control and Syt-1 cKODA data. A direct comparison of hit numbers is included on lines 323-324, with 37% overlap between control and Syt-1 cKODA (vs. 15% between control and RIM cKODA). A direct mapping of this overlap is included in Figure 4E. We think that this direct comparison is complicated by a number of factors, as outlined below, and did our best to include these complications in the discussion, including the last section (lines 501-518).

First, to assess overall similarity, the initial comparison should be to assess axonal proteins identified in the BirA-tdTomato samples. These datasets are quite similar, with 671 (control) and 793 (Syt-1 cKODA) proteins detected, and a high overlap of 601 proteins. We think that this indicates that the experiment per se is quite reproducible. The comparison of the release site proteome between control and Syt-1 cKODA is more complicated. We think that the main point of this comparison is that the overall number of hits is quite similar, with 450 hits in the Syt-1 cKODA proteome and 527 hits in the control proteome, and we now show that this similarity holds across multiple thresholds (lines 298-301; ≥ 1.5: Syt-1 cKODA 602 hits, control 991, ≥ 2.0: 450/527, ≥ 2.5: 252/348). Detailed analyses of overlap reveals that known active zone proteins such as Bassoon, CaV2 channels, RIMs, and ELKS proteins are present in both proteomes, but the overlap is partial and incomplete with 191 proteins found in both proteomes. As discussed throughout and summarized on lines 501-518, the reasons for this partial overlap may be manifold. Trivially, it could be explained by noise or non-saturation (“incompleteness”) of the proteome. We also think that the Syt-1 proteome is not expected to be identical because there is a strong release deficit in these mice. If Syt-1 has a dopamine vesicle docking function (which it does at conventional synapses 4), this could influence the proteome. We note that protein functions in the dopamine axon are not well established, but inferred from studies of classical synapses.

We have scrutinized the manuscript to not express that the control and Syt-1 cKODA proteomes are identical; we know they are not and discuss the example of α-synuclein specifically (Figure 6, lines 347-362). Rather, the striking part is that the extent of the proteomes with high hit number, much higher than RIM cKODA, are similar. Specific hits have to be assessed in a detailed way, one hit at a time, in future studies, as expressed unequivocally on lines 501-503).

Reviewer #3 (Recommendations for the authors):

In this study Kershberg et al. use three novel in vivo biotin-identification (iBioID) approaches in mice to isolate and identify proteins of axonal dopamine release sites. By dissecting the striatum, where dopamine axons are, from the substantia nigra and VTA, where dopamine somata are, the authors selectively analyzed axonal compartments. Perturbation studies were designed by crossing the iBioID lines with null mutant mice. Combining the data from these three independent iBioID approaches and the fact that axonal compartments are separated from somata provides a precise and valuable description of the protein composition of these release sites, with many new proteins not previously associated with synaptic release sites. These data are a valuable resource for future experiments on dopamine release mechanisms in the CNS and the organization of the release sites. The BirA (BioID) tags are carefully positioned in three target proteins not to affect their localization/function. Data analysis and visualization are excellent. Combining the new iBioID approaches with existing null mutant mice produces powerful perturbation experiments that lead and strong conclusions on the central role of RIM1 as central organizers of dopamine release sites and unexpected (and unexplained) new findings on how RIM1 and synaptotagmin1 are both required for the accumulation of α-synuclein at dopamine release sites.

We thank the reviewer for assessing our paper, for summarizing our main findings, and for expressing genuine enthusiasm for the approach and the outcomes.

It is not entirely clear how certain decisions made by the authors on data thresholds may affect the overall picture emerging from their analyses. This is a purely hypothesis-generating study. The authors made little efforts to define expectations and compare their results to these. Consequently, there is little guidance on how to interpret the data and how decisions made by the authors affect the overall conclusions. For instance, the collection of proteins tagged by all three tagging strategies (Figure 2) is expected to contain all known components of dopamine release sites (not at all the case), and maybe also synaptic vesicles (2 TM components detected, but not the most well-known components like vSNAREs and H+/DA-transporters), and endocytic machinery (only 2 endophilin orthologs detected). Whether or not a more complete collection the components of release sites, synaptic vesicles or endocytic machinery are observed might depend on two hard thresholds applied in this study: (a) "Hits" (depicted in Figure 2) were defined as proteins enriched {greater than or equal to} 2-fold (line 178) and peptides not detected in the negative control (soluble BirA) were defined as 0.5 (line 175). How crucial are these two decisions? It would be great to know if the overall conclusions change if these decisions were made differently.

We agree with the reviewer that the thresholding decisions are important and have now better incorporated the rationale for these decisions in the manuscript.

Two-fold enrichment threshold. As outlined in the response to point 1 in the editorial decision letter, we now include figure supplements to illustrate the composition of the control proteome if we apply 1.5- or 2.5-fold enrichment thresholds (Figure 2 —figure supplements 1 and 2) instead of the 2.0-fold threshold used in Figure 2. This leads to more or less hits (991 and 348, respectively) compared to the 2.0-fold threshold (527 hits). It is noteworthy that the SynGO-overlap is the highest with the 2.0 threshold (37% vs. 31% at 1.5 and 33% at 2.5, Figure 2 —figure supplement 3), justifying this threshold experimentally in addition to what was done in previous work 1,2. These data are now described on lines 208-215 of the manuscript. When we apply these different thresholds to RIM and Syt-1 cKODA datasets, the finding that RIM ablation disrupts release site assembly persists. The following hit numbers were observed in the mutants at the 1.5, 2.0 and 2.5 enrichment thresholds, respectively: RIM cKODA 268, 198 and 82 hits; Syt cKODA 602, 450 and 252 hits. Hence, the extent of the release site proteome remains much smaller after RIM ablation independent of the enrichment threshold, bolstering the conclusion that RIM is an important scaffold for these release sites. This is included in the revised manuscript on lines 298-301.

Undetected peptides in BirA-tdTomato. We did not express this well enough in the manuscript. The undetected proteins were set to 0.5 such that a protein that was detected with a specific bait but not with BirA-tdTomato could be illustrated with a specific circle size, not to determine inclusion in the analyses. If the average peptide count across repeats with a specific bait was 1, this resulted in inclusion in Figure 2 and consecutive analyses with the smallest circle size. Hence, this decision was made to define circle size. It did not affect inclusion in Figure 2 beyond the following two points. If one were to further decrease it, this might result in including peptides that only appeared once as a single peptide for some of the experiments, which we wanted to avoid. If one would set it higher (to 1), this artificial threshold would be equal to proteins that were actually detected experimentally multiple times, which we wanted to avoid as well. We have now clarified this on lines 165-167 and lines 1119-1121.

Expected proteins. In general, interpreting our dataset with a strong prior of expected proteins is difficult. The literature on release site proteins specifically characterized for dopamine is limited. We have found Bassoon, RIM, ELKS and Munc13 to be present using 3D-SIM super-resolution microscopy 5,6, and we indeed found these proteins in the data as discussed on lines 227-232 and lines 423-445 in the revised manuscript. The prediction for vesicular and endocytic proteins is complicated. Release sites are sparse 5,7, and vesicle clusters are widespread in the dopamine axon, in some cases filling most of the axon (for example, see extended vesicle clusters filling much of the dopamine axon in Figure 7E of 5). Furthermore, docking in dopamine axons has not been characterized, and it is unclear how frequently vesicles are docked. Hence, it is not clear whether vesicular proteins should be concentrated at release sites compared to the rest of the axon (the BirA-tdTomato proteome we use for normalization). Similar points can be made for proteins for endocytosis and recycling of dopamine vesicles. Within the dopamine system, it is unclear whether the recycling pathway is close to the exocytic sites. One consistent finding across functional studies is that depletion after activity is unusually long-lasting in the dopamine system, for tens of seconds, even after only mild stimulation 5,8–13. Hence, endocytosis and RRP replenishment might be very slow in these axons. It is not certain that endocytic factors are predeployed to the plasma membrane, and if they are, it is unclear how close to release sites they would be. As such, we agree with the reviewer that the proteome we describe is a hypothesis-generator. With the limited knowledge on dopamine release, predictions beyond the previously characterized proteins in dopamine axons are difficult to make.

We thank the reviewer for suggesting to include a better analysis of different thresholds and for giving us the opportunity to clarify the other points that were raised.

Given the good separation of the axonal compartment from the somata (one of the real experimental strengths of this study), it is completely unexpected to find two histones being enriched with all three tagging strategies (Hist1h1d and 1h4a). This should be mentioned and discussed.

We agree with the reviewer and have addressed this point in the manuscript. This could either reflect noise, or there could be more specific reasons behind it. The manuscript now states on lines 482-485: “It is surprising that Hist1h1d and Hist1h4a, genes encoding for the histone proteins H1.3 and H4, were robustly enriched (Figure 2A). These hits might be entirely unspecific, or their co-purification could be due to biotinylation of H1 and H4 proteins (Stanley et al., 2001). It is also possible that there are unidentified synaptic functions of some of the unexpected proteins.” Ultimately, we do not know why these proteins are enriched, and we state clearly in the section “Summary of conclusions and limitations” that each new hit has to be validated in future studies (lines 501-503).

It would also help to compare the data more systematically to a previous study that attempted to define release sites (albeit not dopamine release sites) using a different methodology (biochemical purification): Boyken et al. (only mentioned in relation to Nptn, but other proteins are observed in both studies too, e.g. Cend1).

We agree with the reviewer that Boyken et al., 2013 14 is an important resource for our paper and for the assessment of the proteomic composition of release sites. We have now introduced links and citations to this paper multiple times (for example, on lines 231, 241, 430, 443, 481) and have expanded the discussion of overlap between these proteomes, including on Cend1 (lines 479482).

We think that a systematic comparison with Boyken et al., 2013 14 is complicated because (1) so little is known about dopamine release mechanics and (2) because the approach is very different between the two papers. In respect to (1), most prominently, it is not certain how frequently vesicles are docked in the dopamine axon. Only ~25% of the varicosities contain these release sites, and vesicle docking has not been characterized in striatal dopamine axons to the best of our knowledge. Hence, how a docking site at a classical synapse compares to a dopamine release site remains unclear at the outset. For point (2), the key difference is that “within dataset normalizations” are very different in these two studies. In our iBioID dataset, we normalize to soluble proteins defined as proximity to BirA-tdTomato. In ref. 14, the authors express enrichment over “light”, regular synaptic vesicles purified with the same approach. This has a major impact on the proteome that strongly influences a direct comparison of hits, because there are large differences in the normalization. While each normalization makes sense for the respective paper, it complicates direct comparison.

With these points in mind, we have compared hits across both datasets class-by-class. For some classes, the datasets have reasonable overlap for ≥ 2-fold enriched proteins: for example for active zone proteins (3 of 7 hits in 14 appear in our control proteome) and adhesion and cell surface proteins (8 of 18). For other classes, the overlap is limited: for example for nucleotide metabolism/protein synthesis (0 of 16 hits in 14 appear in our dataset) and cytoskeletal proteins (5 of 29). We hope the reviewer agrees, that given these factors, the analyses and discussion needed for a systematic comparison goes beyond the scope of our paper. We have instead added a number of references to Boyken et al., 2013 14, as outlined above, when direct comparison is meaningful.

The Discussion section could be improved. It is rather long (7.5p) with some repetition from the Introduction, some issues already demonstrated convincingly before (2.5p on RIM being a scaffold in DA neurons) and a rather speculative discussion on RIM being associated with Parkinson's disease through dopamine release site recruitment of α-synuclein, possibly via protein interactions or through vesicle docking (but how is release site recruitment of α-synuclein important for protein interactions, vesicle docking or Parkinson symptoms?). Instead some of the more pressing questions (at least to this reviewer) are not discussed, such as the expectations mentioned above: why were expected components of release sites not detected, why only some but not all expected components of synaptic vesicles/endocytic machinery and why are two histones in the axonal compartment?

We have re-written the discussion. Specifically, we have (1) shortened it from ~2400 to ~1900 words, (2) cut down the RIM scaffolding section from 2.5 pages to 1 page, (3) removed the repeated discussion of α-synuclein as it is redundant with the Results section, (4) added more discussion on unexpected components (for example histones on lines 482-485, trafficking components on lines 493-497, and ribosomes on lines 497-498), (5) expanded the section on predicted active zone components (lines 423-445), and (6) more clearly described the limitations of the method throughout the manuscript (for example on lines 458, 478-479 for noise, on lines 208-215 for thresholding) and summarize this point in the last section (lines 501-518).

It is difficult to make specific predictions on vesicular and endocytic proteins, and we think those components are not necessarily expected to be enriched over their general axonal levels within 50 nm of the baits at rest. Vesicular proteins may be high within the rest of the axon because the axon is densely packed with vesicles and release sites are sparse 5,7,15, and it is unclear how frequent dopamine vesicle docking is. It is also unclear whether endocytic machinery is predeployed to sites of exocytosis. Replenishment is astonishingly slow in dopamine axons (tens of seconds to a minute) 5,8–13, it is uncertain where endocytosis happens relative to release sites, and it is unclear whether endocytic factors are concentrated at the release site compared to their average axonal protein content (detected with BirA-tdTomato). Hence, we think it is better to not describe the proteome with a prior, but to interpret it as it is. We hope that the reviewer agrees in light of these explanations and with the numerous adjustments that we have made to the discussion. There are strong predictions based on previous dopamine release site analyses what active zone proteins should be found at these sites, and we discuss them in detail (lines 423-445). Beyond these proteins, predictions are difficult to make.

References

1. Takano, T.; Wallace, J. T.; Baldwin, K. T.; Purkey, A. M.; Uezu, A.; Courtland, J. L.; Soderblom, E. J.; Shimogori, T.; Maness, P. F.; Eroglu, C.; Soderling, S. H. ChemicoGenetic Discovery of Astrocytic Control of Inhibition in vivo. Nature 2020, 588, 296–302.

2. Uezu, A.; Kanak, D. J.; Bradshaw, T. W. A.; Soderblom, E. J.; Catavero, C. M.; Burette, A. C.; Weinberg, R. J.; Soderling, S. H. Identification of an Elaborate Complex Mediating Postsynaptic Inhibition. Science 2016, 353, 1123–1129.

3. Loh, K. H.; Stawski, P. S.; Draycott, A. S.; Udeshi, N. D.; Lehrman, E. K.; Wilton, D. K.; Svinkina, T.; Deerinck, T. J.; Ellisman, M. H.; Stevens, B.; Carr, S. A.; Ting, A. Y. Proteomic Analysis of Unbounded Cellular Compartments: Synaptic Clefts. Cell 2016, 166, 12951307.e21.

4. Chang, S.; Trimbuch, T.; Rosenmund, C. Synaptotagmin-1 Drives Synchronous ca2+Triggered Fusion by C2B-Domain-Mediated Synaptic-Vesicle-Membrane Attachment. Nature Neuroscience 2018, 21, 33–40.

5. Liu, C.; Kershberg, L.; Wang, J.; Schneeberger, S.; Kaeser, P. S. Dopamine Secretion Is Mediated by Sparse Active Zone-like Release Sites. Cell 2018, 172, 706-718.e15.

6. Banerjee, A.; Imig, C.; Balakrishnan, K.; Kershberg, L.; Lipstein, N.; Uronen, R.-L.; Wang, J.; Cai, X.; Benseler, F.; Rhee, J. S.; Cooper, B. H.; Liu, C.; Wojcik, S. M.; Brose, N.; Kaeser, P. S. Molecular and Functional Architecture of Striatal Dopamine Release Sites. Neuron 2022, 110, 248-265.e9.

7. Pereira, D. B.; Schmitz, Y.; Mészáros, J.; Merchant, P.; Hu, G.; Li, S.; Henke, A.; LizardiOrtiz, J. E.; Karpowicz, R. J.; Morgenstern, T. J.; Sonders, M. S.; Kanter, E.; Rodriguez, P. C.; Mosharov, E. V; Sames, D.; Sulzer, D. Fluorescent False Neurotransmitter Reveals Functionally Silent Dopamine Vesicle Clusters in the Striatum. Nature Neuroscience 2016, 19, 578–586.

8. Banerjee, A.; Lee, J.; Nemcova, P.; Liu, C.; Kaeser, P. S. Synaptotagmin-1 Is the ca2+ Sensor for Fast Striatal Dopamine Release. eLife 2020, 9.

9. Wang, L.; Shang, S.; Kang, X.; Teng, S.; Zhu, F.; Liu, B.; Wu, Q.; Li, M.; Liu, W.; Xu, H.; Zhou, L.; Jiao, R.; Dou, H.; Zuo, P.; Zhang, X.; Zheng, L.; Wang, S.; Wang, C.; Zhou, Z. Modulation of Dopamine Release in the Striatum by Physiologically Relevant Levels of Nicotine. Nature communications 2014, 5, 3925.

10. Brimblecombe, K. R.; Gracie, C. J.; Platt, N. J.; Cragg, S. J. Gating of Dopamine Transmission by Calcium and Axonal N-, Q-, T- and L-Type Voltage-Gated Calcium Channels Differs between Striatal Domains. The Journal of Physiology 2015, 593, 929– 946.

11. Marcott, P. F.; Mamaligas, A. A.; Ford, C. P. Phasic Dopamine Release Drives Rapid Activation of Striatal D2-Receptors. Neuron 2014, 84, 164–176.

12. Patriarchi, T.; Cho, J. R.; Merten, K.; Howe, M. W.; Marley, A.; Xiong, W.-H.; Folk, R. W.; Broussard, G. J.; Liang, R.; Jang, M. J.; Zhong, H.; Dombeck, D.; Zastrow, M. von; Nimmerjahn, A.; Gradinaru, V.; Williams, J. T.; Tian, L. Ultrafast Neuronal Imaging of Dopamine Dynamics with Designed Genetically Encoded Sensors. Science 2018, 360, eaat4422.

13. Wang, L.; Zhang, X.; Xu, H.; Zhou, L.; Jiao, R.; Liu, W.; Zhu, F.; Kang, X.; Liu, B.; Teng, S.; Wu, Q.; Li, M.; Dou, H.; Zuo, P.; Wang, C.; Wang, S.; Zhou, Z. Temporal Components of Cholinergic Terminal to Dopaminergic Terminal Transmission in Dorsal Striatum Slices of Mice. The Journal of Physiology 2014, 592, 3559–3576.

14. Boyken, J.; Grønborg, M.; Riedel, D.; Urlaub, H.; Jahn, R.; Chua, J. J. Molecular Profiling of Synaptic Vesicle Docking Sites Reveals Novel Proteins but Few Differences between Glutamatergic and GABAergic Synapses. Neuron 2013, 78, 285–297.

15. Wildenberg, G.; Sorokina, A.; Koranda, J.; Monical, A.; Heer, C.; Sheffield, M.; Zhuang, X.; McGehee, D.; Kasthuri, B. Partial Connectomes of Labeled Dopaminergic Circuits Reveal Non-Synaptic Communication and Axonal Remodeling after Exposure to Cocaine. eLife 2021, 10.

https://doi.org/10.7554/eLife.83018.sa2

Article and author information

Author details

  1. Lauren Kershberg

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Resources, Formal analysis, 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:" 0000-0002-5676-4160
  2. Aditi Banerjee

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Contribution
    Resources, Formal analysis, Investigation, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2016-0717
  3. Pascal S Kaeser

    Department of Neurobiology, Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Formal analysis, Supervision, Funding acquisition, Writing – original draft, Writing – review and editing
    For correspondence
    kaeser@hms.harvard.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1558-1958

Funding

National Institute of Neurological Disorders and Stroke (R01NS103484)

  • Pascal S Kaeser

National Institute of Neurological Disorders and Stroke (R01NS083898)

  • Pascal S Kaeser

National Institute of Neurological Disorders and Stroke (F31NS105159)

  • Lauren Kershberg

Harvard Medical School (Lefler Foundation)

  • Pascal S Kaeser

Harvard Medical School (Brooks Fellowship)

  • Aditi Banerjee

Harvard Medical School (Quan Fellowship)

  • Lauren Kershberg

Harvard Medical School (Dean's Innovation Grant)

  • Pascal S Kaeser

Harvard Medical School

  • Pascal S Kaeser

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

Acknowledgements

We thank J Wang and C Qiao for technical support, current and former members of the Kaeser laboratory and D Brann for insightful discussions and feedback, M Feany and H Nyitrai for comments on the manuscript, R Tomaino and the Taplin Mass Spectrometry Facility of the Department of Cell Biology at Harvard Medical School for help with mass spectrometric analyses, and S Soderling and A Uezu for sharing the BirA plasmid and for advice on iBioID early in the project. LK is currently an employee of Mass General Brigham (Boston, MA, USA). This work was supported in part by the NIH (R01NS103484, R01NS083898 to PSK; F31NS105159 to LK), the Lefler foundation (to PSK), a Dean’s Innovation grant (PSK), a Quan fellowship (LK), an Alice and Joseph Brooks fellowship (AB), and Harvard Medical School. We acknowledge the Neurobiology Imaging Facility (supported by a P30 Core Center Grant P30NS072030).

Ethics

All animal experiments were performed according to institutional guidelines of Harvard University, and were in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All animal experiments were approved by the Harvard University Animal Care and Use Committee (protocol number IS00000049).

Senior Editor

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

Reviewing Editor

  1. Reinhard Jahn, Max Planck Institute for Biophysical Chemistry, Germany

Reviewers

  1. Reinhard Jahn, Max Planck Institute for Biophysical Chemistry, Germany
  2. Matthijs Verhage, VU University Amsterdam, Netherlands

Version history

  1. Received: August 26, 2022
  2. Preprint posted: August 31, 2022 (view preprint)
  3. Accepted: December 22, 2022
  4. Accepted Manuscript published: December 29, 2022 (version 1)
  5. Version of Record published: February 17, 2023 (version 2)

Copyright

© 2022, Kershberg 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

  • 1,751
    Page views
  • 350
    Downloads
  • 4
    Citations

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

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

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

  1. Lauren Kershberg
  2. Aditi Banerjee
  3. Pascal S Kaeser
(2022)
Protein composition of axonal dopamine release sites in the striatum
eLife 11:e83018.
https://doi.org/10.7554/eLife.83018

Share this article

https://doi.org/10.7554/eLife.83018

Further reading

    1. Biochemistry and Chemical Biology
    2. Cell Biology
    Kristian Davidsen, Jonathan S Marvin ... Lucas B Sullivan
    Research Article

    Intracellular levels of the amino acid aspartate are responsive to changes in metabolism in mammalian cells and can correspondingly alter cell function, highlighting the need for robust tools to measure aspartate abundance. However, comprehensive understanding of aspartate metabolism has been limited by the throughput, cost, and static nature of the mass spectrometry (MS)-based measurements that are typically employed to measure aspartate levels. To address these issues, we have developed a green fluorescent protein (GFP)-based sensor of aspartate (jAspSnFR3), where the fluorescence intensity corresponds to aspartate concentration. As a purified protein, the sensor has a 20-fold increase in fluorescence upon aspartate saturation, with dose-dependent fluorescence changes covering a physiologically relevant aspartate concentration range and no significant off target binding. Expressed in mammalian cell lines, sensor intensity correlated with aspartate levels measured by MS and could resolve temporal changes in intracellular aspartate from genetic, pharmacological, and nutritional manipulations. These data demonstrate the utility of jAspSnFR3 and highlight the opportunities it provides for temporally resolved and high-throughput applications of variables that affect aspartate levels.

    1. Cell Biology
    2. Immunology and Inflammation
    Chinky Shiu Chen Liu, Tithi Mandal ... Dipyaman Ganguly
    Research Article

    T cells are crucial for efficient antigen-specific immune responses and thus their migration within the body, to inflamed tissues from circulating blood or to secondary lymphoid organs, plays a very critical role. T cell extravasation in inflamed tissues depends on chemotactic cues and interaction between endothelial adhesion molecules and cellular integrins. A migrating T cell is expected to sense diverse external and membrane-intrinsic mechano-physical cues, but molecular mechanisms of such mechanosensing in cell migration are not established. We explored if the professional mechanosensor Piezo1 plays any role during integrin-dependent chemotaxis of human T cells. We found that deficiency of Piezo1 in human T cells interfered with integrin-dependent cellular motility on ICAM-1-coated surface. Piezo1 recruitment at the leading edge of moving T cells is dependent on and follows focal adhesion formation at the leading edge and local increase in membrane tension upon chemokine receptor activation. Piezo1 recruitment and activation, followed by calcium influx and calpain activation, in turn, are crucial for the integrin LFA1 (CD11a/CD18) recruitment at the leading edge of the chemotactic human T cells. Thus, we find that Piezo1 activation in response to local mechanical cues constitutes a membrane-intrinsic component of the ‘outside-in’ signaling in human T cells, migrating in response to chemokines, that mediates integrin recruitment to the leading edge.