Dynamic proteomic and phosphoproteomic atlas of corticostriatal axons in neurodevelopment

Version of Record

Accepted for publication after peer review and revision.

Download
Cite
Share
CommentOpen annotations (there are currently 0 annotations on this page).

Version of Record published: November 2, 2022 (This version)
Accepted Manuscript published: October 14, 2022 (Go to version)
Accepted: October 12, 2022
Received: March 22, 2022
Preprint posted: March 22, 2022 (Go to version)

1. Of interest
Chronic intermittent hypoxia reveals role of the Postinspiratory Complex in the mediation of normal swallow production

Alyssa D Huff, Marlusa Karlen-Amarante ... Jan-Marino Ramirez

Research Advance Apr 24, 2024
Further reading

Abstract
Editor's evaluation
Introduction
Results
Discussion
Materials and methods
Data availability
References
Article and author information
Metrics

Abstract

Mammalian axonal development begins in embryonic stages and continues postnatally. After birth, axonal proteomic landscape changes rapidly, coordinated by transcription, protein turnover, and post-translational modifications. Comprehensive profiling of axonal proteomes across neurodevelopment is limited, with most studies lacking cell-type and neural circuit specificity, resulting in substantial information loss. We create a Cre-dependent APEX2 reporter mouse line and map cell-type-specific proteome of corticostriatal projections across postnatal development. We synthesize analysis frameworks to define temporal patterns of axonal proteome and phosphoproteome, identifying co-regulated proteins and phosphorylations associated with genetic risk for human brain disorders. We discover proline-directed kinases as major developmental regulators. APEX2 transgenic reporter proximity labeling offers flexible strategies for subcellular proteomics with cell type specificity in early neurodevelopment, a critical period for neuropsychiatric disease.

Editor's evaluation

Knowledge of the protein composition of defined subcellular compartments is of key importance for the characterization of protein machines that mediate defined cellular functionalities. The current paper presents a novel mouse line that will serve as a tool of fundamental value in this context – a Cre-inducible APEX2 reporter mouse line for acute ex-vivo proximity biotinylation. The authors provide compelling evidence documenting the usefulness of the novel reporter line, describing circuit-specific proteomes and phosphoproteomes in the corticostriatal system of the mouse brain during development. The biological insights deduced from bioinformatic analyses of the proteomic data are convincing. The new APEX2 reporter mouse line will be of substantial interest to researchers in many fields of mammalian biology.

https://doi.org/10.7554/eLife.78847.sa0

Introduction

Neurons are morphologically diverse cells that compartmentalize cellular signaling and information processing in cell bodies, dendrites, and axons, enabling spatiotemporal control over synaptic transmission. While these subcellular domains are maintained and regulated by both local and distal cues, local proteome regulation is particularly important for fine-tuning neural connectivity during circuit development and in the context of plasticity (Gonzalez-Lozano et al., 2016; Poulopoulos et al., 2019; Schanzenbächer et al., 2018). During the embryonic and early postnatal stages, axons rapidly grow and navigate to target regions across the brain (Winnubst et al., 2019). Developing axons require local protein synthesis and dynamics to coordinate axon outgrowth and growth cone collapse (Lin and Holt, 2007). Although axons contain machinery for local protein synthesis (Hafner et al., 2019), a large fraction of the neuroproteome is synthesized in the soma and trafficked throughout the cell, as evidenced by generally greater abundance of transcripts in the soma, compared to neurites (Glock et al., 2021). Thus, proteomic rather than transcriptomic techniques are necessary to capture the functional state of axons; this is especially important for axon guidance signaling that is highly regulated by post translational modifications such as protein phosphorylation (Costa-Mattioli et al., 2009). In addition, neuronal cell types are characterized by distinct patterns of gene expression (Saunders et al., 2018; Zeisel et al., 2018), implying biodiversity of their somatic, dendritic, and axonal proteomes. Therefore, it is important that measurements of the microdomain proteomes also contain cell-type specific information.

The postnatal proteome of developing axons remains largely unknown for any neuronal type. Mapping axonal proteomes from specific neurons directly in the vertebrate brain remains extremely challenging. Axonal compartments, such as axon growth cones, are typically isolated by tissue dissociation and sorting techniques, resulting in substantial losses of axon components (Chauhan et al., 2020; Poulopoulos et al., 2019). Axons from different circuits and cell types are intermingled, and isolation techniques that lack genetic targeting likely mask neuronal class specific features. Several genetically encoded proteomics strategies have been developed to profile cell-type-specific proteomes in the mouse brain, including metabolic labeling (Alvarez-Castelao et al., 2017; Krogager et al., 2018) and proximity labeling strategies (Dumrongprechachan et al., 2021; Hobson et al., 2022; Rayaprolu et al., 2021; Spence et al., 2019; Uezu et al., 2016). We and others have previously shown that APEX-based proximity labeling in the acute brain sections is highly efficient and robust, providing sufficient material with both cell-type and subcellular compartment specificity such that there is no need for tissue dissociation, biochemical fractionation, or subject pooling (Dumrongprechachan et al., 2021; Hobson et al., 2022). APEX2 was engineered to be highly efficient for rapid protein biotinylation within seconds to minutes in the presence of hydrogen peroxides (Lam et al., 2015), providing superior labeling speed over other proximity labeling methods (Hung et al., 2016; Lobingier et al., 2017). Our prior work also demonstrates that rapid labeling speed of virally transduced APEX is capable of taking proteomic snapshots and capturing changes in the proteome within 3–4 hr time windows (Dumrongprechachan et al., 2021), but it is incompatible with applications that require early postnatal time points. Although APEX-based proximity labeling is a promising candidate for accessing the proteomic landscape of neurodevelopment, significant modifications to the overall strategy are needed in order to successfully map genetically targeted axonal proteomes in the mouse brain during postnatal development.

In this study, we focus on the postnatal development of corticostriatal projections. Corticostriatal systems represent a powerful model for studying axon development and axon guidance. In rodents, corticostriatal projections arise in the embryonic period and continue to innervate the striatum postnatally (Nisenbaum et al., 1998; Sheth et al., 1998; Sohur et al., 2014), where growth cones start to disappear around postnatal day P7, based on the reduction of growth-associated protein 43 (GAP43) immunostaining (Dani et al., 1991). As axon outgrowth declines after P7, the maturation of striatal neuron physiology becomes more pronounced, with high levels of dendritic spinogenesis, synaptogenesis, and tuning of intrinsic excitability until reaching maturity (P8 – P30) (Kozorovitskiy et al., 2012; Kozorovitskiy et al., 2015; Krajeski et al., 2019; Lieberman et al., 2018; Peixoto et al., 2016). Therefore, the dynamic phases of axonal development in the striatum raise a question about how the local axonal proteome changes across the first three postnatal weeks. Further understanding of typical neurodevelopment provides insight into how these processes break down in disease, as corticostriatal circuits are implicated in a number of neurodevelopment disorders such as autism spectrum disorder and schizophrenia. Because the striatum receives excitatory inputs from several brain regions (Hunnicutt et al., 2016), conventional approaches without a genetic targeting strategy cannot distinguish axons from different input types. To date, no genetically targeted proteome or phosphoproteome of axons across development has been reported. Phosphorylation state information, along with proteomics measurements, would provide an unparalleled window into local protein dynamics critical for neurodevelopment. Current APEX-based proteomics approaches primarily use Cre-dependent adeno-associated viral transduction (AAV) to deliver APEX reporters. This workflow restricts possible experimental timelines, because optimal expression of Cre-dependent AAV may take 1–3 weeks, rendering the early postnatal proteome intractable.

To address both technical limitations and biological questions in this study, we created a Cre-dependent APEX2 reporter mouse line using CRISPR knock-in. We demonstrated that the APEX reporter line can be used with multiple Cre-driver lines and is suitable for proteomics applications, including in the early postnatal period. We mapped the temporal expression patterns of the corticostriatal projection proteome during postnatal development. Combining APEX-based proximity labeling with phosphopeptide enrichment enabled a characterization of the local phosphoproteome in axons, revealing dynamic regulation of proline-directed kinases and phosphosites in corticostriatal projection during development.

Results

Generation of a Cre-dependent APEX2 reporter mouse line

We created a Cre-dependent APEX reporter mouse line for broad applications in neuroscience and biology. A monomeric APEX2 variant was chosen because of its small size and efficient activity for proximity labeling of proteins (Lam et al., 2015). A double-floxed inverted orientation of the APEX transgene (DIO-APEX2-P2A-EGFP) was introduced into the Gt(ROSA)26 locus under the EF1a promoter using CRISPR/Cas9 strategy (Figure 1a). Zygotes were microinjected with enhanced specificity eSpCas9, HDR template and crRNA:tracrRNA duplex, using a previously validated gRNA sequence targeting the Gt(ROSA)26 locus (Chu et al., 2016). We identified 8 out 112 founder mutants by genotyping PCR (Figure 1b, Supplementary file 1). We confirmed germline transmission and mendelian inheritance by crossing selected founders to C57BL/6 wildtype mice over two generations. The reporter line was bred to homozygosity by crossing F2 heterozygotes (Figure 1c). We evaluated transgene integration using long-range PCR amplification. Using primers inside the transgene and outside the left or right homology arms, respectively, we found that the size of PCR amplicons matched the predicted sequences, suggesting that transgene integration was site-specific (Figure 1d). In addition, we PCR amplified and sequenced top eight predicted off-target sites with NGG PAM across 6 founders. We found a small number of off-targets that were affected by Cas9 activity on non-coding protein regions (Figure 1—figure supplement 1a-c, Figure 1, Supplementary file 2). We did not observe noticeable phenotypes in either heterozygous or homozygous animals. From this point onward, we chose to work with founder line #83 (Figure 1—figure supplement 1a-c). To evaluate conditional expression of the APEX transgene, the APEX reporter line was crossed to Vglut2^Cre (Slc17a6^Cre) or Vgat^Cre (Slc32a1^Cre) mice (Vong et al., 2011). We performed direction-specific PCR to assess Cre-dependent inversion of the APEX transgene using genomic DNA that was directly extracted from the cortex. Inverted transgene was only detected in Cre-expressing brain tissues (Figure 1e), confirming Cre-dependent recombination. We also inspected the pattern of EGFP reporter expression across multiple brain regions in both Vglut2^Cre and Vgat^Cre crosses (Figure 1f–g). Dense reporter expression was observed in the cortex for Vglut2^Cre and in the striatum for Vgat^Cre, respectively. In the hippocampus, the granule cell layer in the dentate gyrus was clearly visible in Vglut2^Cre cross, with smaller numbers of GABAergic cells present in the hilus of Vgat^Cre cross. In the cerebellum, Purkinje cell layer was observed in Vgat^Cre and absent in Vglut2^Cre mice crossed to the APEX reporter. Both genetic and histological evidence demonstrates Cre-dependent control of the APEX transgene.

Figure 1 with 1 supplement see all

Download asset Open asset

Generation of a Cre-dependent APEX2 reporter mouse line.

(A) Design of a Cre-dependent APEX reporter mouse line. A Cre-dependent APEX transgene under the control of EF1a promoter is targeted for knock-in at the *Gt(ROSA)26* locus in the mouse genome. Inverted APEX-NES (nuclear exporting sequence) is flanked by lox sites in a double floxed inverted orientation (DIO). (B) CRISPR/Cas9 strategy for mouse line generation. Enhanced specificity Cas9 (ESPCas9), targeting vector, and crRNA::tracrRNA duplex were used in pronuclear injection to generate founders. Agarose gel electrophoresis shows founders containing APEX transgene. (C) PCR amplification across the APEX transgene. Gray arrows: 317 bp WT loci and ~4 kb APEX transgene. (D) PCR amplification across the homology arms evaluating genomic integration. Blue arrow, left homology arm (HA-L), 2586 bp. Red arrow, right homology arm (HA-R), 5409 bp. (E) PCR amplification of genomic DNA extracted from the brain for the presence of Cre, the APEX transgene before and after Cre-mediated recombination in Cre^-/-, *Vgat*^Cre+/- and *Vglut2*^Cre+/- animals. (F) Cre-dependent expression of EGFP reporter in the prefrontal cortex (PFC), striatum (STR), hippocampus (HC), and cerebellum (CB) of *Vglut2*^Cre mice crossed to APEX reporter line (immunostaining using anti-GFP antibody). Scale bars, 500 µm for PFC and STR, and 200 µm for HC and CB. (G) same as (F) but for *Vgat*^Cre.

Figure 1—source data 1 Raw and uncropped gel images for Figure 1b, c, d and e.: https://cdn.elifesciences.org/articles/78847/elife-78847-fig1-data1-v2.zip
Download elife-78847-fig1-data1-v2.zip

Ex vivo biotinylation workflow for cell-type-specific proteomics

To broadly distribute APEX inside neurons, we included the nuclear exporter sequence (NES) at the C-terminus of the APEX protein. We confirmed APEX localization using anti-FLAG immunofluorescent staining (Figure 2a). APEX localizes in the cytosolic compartment compared to untargeted EGFP structural markers, separated by a P2A linker. To demonstrate the utility of the APEX reporter line for proteomics, we used biotin phenol (BP) and H₂O₂ to induce biotinylation of proteins in APEX expressing neurons in the Vglut2^Cre and Vgat^Cre crosses. We increased BP incubation time to 2 min from previously published APEX-mediated biotinylation protocols to optimize protein labeling in thick brain tissue (Dumrongprechachan et al., 2021), since APEX has been largely characterized and applied in cell culture (Hung et al., 2016; Lobingier et al., 2017; Paek et al., 2017) and small organisms (Chen et al., 2015; Reinke et al., 2017). Our prior work and other studies have demonstrated that AAV-mediated APEX expression and biotinylation in the mouse brain is sufficient for cell-type specific mass spectrometry-based (MS) proteomics (Dumrongprechachan et al., 2021; Hobson et al., 2022). To label proteins in APEX-expressing tissues, acute brain slices of 250 µm thickness were prepared from Vglut2^Cre;APEX animals and incubated in ACSF supplemented with 500 µM BP up to 1 hr. Following BP incubation, slices were briefly washed in ACSF and biotinylation was induced by 0.03% H₂O₂ treatment for 2 min and quenched with Na-ascorbate ACSF (Figure 2b). Streptavidin dot blot analysis of cortical protein lysates shows increasing amounts of biotinylated proteins over 60 min (Figure 2—figure supplement 1a). Western blot analysis of total prefrontal cortex (PFC) lysates at 60 min BP incubation shows reduced protein abundance in Vgat^Cre compared to Vglut2^Cre, with negligible labeling in the Cre-negative APEX controls (Figure 2—figure supplement 1b). As expected, the level of biotinylation in Vglut2^Cre PFC sample was greater than that of Vgat^Cre, consistent with a greater number of glutamatergic neurons relatively to GABAergic neurons present in the neocortex, since under ~20% of cortical neurons are GABAergic (Keller et al., 2018). Using the 1 hr-long incubation protocol, we confirmed that biotinylation is both Cre and APEX-dependent (Figure 2c). We observed some labeling in Vglut2^Cre+ samples during 1 hr BP incubation without H₂O₂, and labeling efficiency was greatly enhanced in the presence of H₂O₂. To streamline sample preparation for MS-based proteomics, we evaluated the enrichment method using streptavidin magnetic beads (Figure 2d). Successful enrichment was verified by the depletion of streptavidin signal in flowthrough fractions and the presence of high streptavidin signal in the enrichment output, where the corresponding total amount of enriched proteins is shown by silver staining. Altogether, these experiments establish a method for selective proteome labeling, which we next apply to profile the proteome of excitatory cortical inputs to the striatum across postnatal development.

Figure 2 with 1 supplement see all

Download asset Open asset

APEX-mediated biotinylation in brain tissue.

(A) APEX subcellular localization in the mPFC of *Vglut2*^Cre;APEX and *Vgat*^Cre;APEX (immunostained EGFP 488, anti-FLAG 647 for APEX, scale bars 20 µm). (B) Sample preparation schematic for protein analysis. Acute slices were incubated in carbogenated artificial cerebrospinal fluid (ACSF) supplemented with 500 µM biotin phenol for 1 hr. Sections were briefly rinsed in ACSF. APEX was activated by 0.03% H₂O₂ ACSF for 2 min and quenched in Na ascorbate ACSF solution. Protein lysates were prepared from dissected tissues. Streptavidin beads were used to enrich tagged proteins for analysis. (C) Cell-type-specific biotinylation in Cre-negative, *Vgat*^Cre, and *Vglut2*^Cre crosses. Western blot analysis of cortical protein lysates. Biotinylated proteins and total protein loading control were detected by streptavidin-CW 800 and REVERT 700 stain, respectively. (D) Validation of streptavidin bead enrichment. *Left*, detection of biotinylated proteins in flow-through and enriched fractions. *Right*, silver stain gel of enrichment output proteins. Differential enrichment outputs reflect differences in GABAergic and glutamatergic cell proportions in the cortex.

Figure 2—source data 1 Raw and uncropped blots for Figure 2c and d.: https://cdn.elifesciences.org/articles/78847/elife-78847-fig2-data1-v2.zip
Download elife-78847-fig2-data1-v2.zip

Selective biotinylation of corticostriatal axons across postnatal development

Glutamatergic inputs into the striatum develop rapidly across embryonic and early postnatal stages (Kuo and Liu, 2019; Peixoto et al., 2016), yet deep knowledge of the overall proteome dynamics in developing corticostriatal projections has remained out of reach for technical reasons. To capture axonal proteome dynamics during this process, we crossed the Rbp4^Cre mouse line to our APEX reporter (Gerfen et al., 2013). The Rbp4^Cre mouse line expresses Cre recombinase under the control of the retinol binding protein 4 (Rbp4) gene promoter. A previous characterization of this line shows Cre-mediated reporter expression, enriched in the accessory olfactory bulb, layer 5 pyramidal tract neurons throughout the neocortex, and hippocampal granule cells in the dentate gyrus (Harris et al., 2014). As expected, Rbp4^Cre;APEX expression is primarily restricted to layer 5 neurons which also send corticostriatal projections (Figure 3a). We showed that EGFP reporter expression is observed in the striatum throughout postnatal time period P5-P40 (Figure 3a). Western blot analysis after ex vivo biotinylation detects labeled proteins in both cortical and striatal tissues above the negative control, including at the earliest P5 time point (Figure 3b, Figure 3—figure supplement 1a). This is especially important because viral based genetically encoded proteomics is limited for early development applications (Figure 3—figure supplement 1b). No somatic expression was observed in the striatum; therefore, biotinylated proteins in the striatal lysates represent corticostriatal axonal proteins. Having validated axonal targeting strategy, we used mass spectrometry-based bottom-up proteomics to identify and quantify the axonal proteome of the striatum. In addition, to rigorously evaluate the extent of axonal enrichment for each protein, we included a somatic compartment control in the experiment. We used adeno-associated viral transduction to express nuclear localized H2B.APEX broadly in the cortex of Rbp4^Cre mice. Immunofluorescence (Figure 3—figure supplement 1c, d) and western blot analysis (Figure 3—figure supplement 1e, f) show that protein biotinylation in H2B.APEX-expressing Rpb4^Cre animals is restricted to the cortex and absent in the striatum. Thus, H2B.APEX presents one compelling option for somatic compartment control because its labeling is primarily restricted to the nucleus and soma. Although this comparison is appropriate for this study, some observed proteome differences from H2B.APEX cortex vs. Rbp4^Cre;APEX mouse line comparison could arise from factors such as viral expression vs. Gt(ROSA)26-locus expression levels, or cytosolic vs. histone-fused APEX.

Figure 3 with 3 supplements see all

Download asset Open asset

Genetic targeting approach for proteomics analysis of corticostriatal axonal compartment.

(A) Cre-dependent APEX.EGFP expression in excitatory cortical inputs to the striatum. Expression pattern of APEX.EGFP (immunostained EGFP) in *Rbp4*^Cre+;APEX-EGFP+/+animals across postnatal days 5, 15, 20, and 40. Coronal brain section, scale bar: 1 mm. (B) Biotinylation of proteins in corticostriatal axons across development. Western blot analysis of striatal lysates prepared from biotinylated acute brain slices. Acute slices were prepared and biotinylated from *Rpb4*^Cre+;APEX animals at the indicated age. Cre-negative APEX reporter line at P18 was used as non-labeling control. *Left*, streptavidin blot. *Right*, total protein loading control. Lower amount of protein biotinylation at early age (P5) is consistent with developing innervation by axonal projections (see (A) ). (C) Sample preparation workflow for proteomics analysis. Acute slices were prepared and biotinylated from *Rbp4*^Cre+;APEX or *Rbp4*^Cre+ injected with H2B.APEX at the indicated time points. Streptavidin beads were used to enrich biotinylated proteins for the bottom-up proteomics workflow. Samples were labeled with tandem mass tags (TMT) and combined. TMT mixtures were enriched for phosphopeptide analysis using Fe-NTA tips. The unbound fractions were fractionated using high pH-reverse phase resin for protein abundance analysis. Synchronous precursor selection (SPS) MS3 method was used for proteome analysis. MS2 with high energy collisional dissociation (HCD) method was used for phosphopeptide analysis. (D) Data analysis flow chart. Raw MS data were analyzed by Proteome Discoverer with false discovery rate (FDR) set at 1% for peptide spectrum matches (PSMs). Protein-level summarization was performed using MSstatsTMT using unique peptides. Striatal (STR) and cortical (CTX.H2B) samples prepared from P18 were compared to determine the axon-enriched proteome. Proteins with missing values were removed prior to this comparison. (E) Log2-fold change (FC) plot for STR–CTX comparison. Proteins were ranked by their log2FC. Red and gray dots, proteins with or without presynaptic or axon gene ontology annotations, respectively. (F) Volcano plot for the STR–CTX comparison. Blue dots, axon-enriched proteins with log2FC >0 and q-value <0.05 (horizontal dotted line). Gray dots, somatic or nuclear proteins that did not pass axon enrichment cutoff. Red border, proteins with presynaptic or axon gene ontology annotations. (G) Proportion of proteins with GO axon/presynaptic annotation for somatic/nuclear and axon protein lists. Using STR–CTX comparison, the proteome is classified into somatic/nuclear and axon-enriched proteins. (H) Over-representation analysis using gene ontology (GO) cellular component database for somatic and nuclear enriched protein list. Dot plot ranked by fold enrichment. Dot size and color correspond to the number of genes, and -log10(FDR), respectively. (I) Same as (H), but for the axon-enriched protein list.

In our experimental design (Figure 3c), we biotinylated and enriched axonal proteins from the striatum across 4 time points (~P5, P11, P18, and P50, n=5 biological replicates) (Supplementary file 3). We targeted these early time points, because striatal circuits start to transition from axon innervation to synaptogenesis, which lead to maturation of striatal neurons as animals enter adolescence and adulthood (Kuo and Liu, 2019). Additionally, we included samples prepared from the Rbp4^Cre+ cortex expressing H2B.APEX as the somatic compartment control (n=3) from P18 animals. Non-labeling controls were collected from Cre-negative cortical and striatal samples that underwent the same ex vivo biotinylation procedure (n=3, 4 for cortex and striatum replicates, respectively). Enriched biotinylated proteins were digested on-bead, and peptides were barcoded with tandem mass tag (TMT) for mass spectrometry analysis. TMT-labeled peptides were mixed equally and enriched for phosphopeptides. The unbound flowthrough was fractionated using the high-pH reverse fractionation to measure protein abundance. To validate the application of flowthrough for protein abundance analysis, in a separate set of samples, we split digested TMT-labeled peptides into two equal volumes: 50% unenriched (UE) and 50% enriched for phosphopeptides (PE) to generate flowthrough (FT) for analysis (n=3 replicates) (Supplementary file 3). We compared three pairs of UE and FT samples at the peptide and protein levels using log2-transformed unnormalized intensities (Figure 3—figure supplement 3a, b). We found that UE and FT are highly correlated at the peptide level (Spearman correlation, R>0.92) (Figure 3—figure supplement 3c). Only 252 out of 13,905 peptide groups (1.8%) are significantly changed after phosphoenrichment (123 decrease, and 129 increase in intensities with q-value <0.05) (Figure 3—figure supplement 3d). In addition, there is no apparent relationship between the number of acidic residues and log2 FC (Figure 3—figure supplement 3e, f). We obtained even higher correlation between UE and FT after protein summarization (R>0.98) (Figure 3—figure supplement 3g-k). This suggests that relative protein abundance in FT is comparable to UE. Only 4 out of 2088 proteins (~0.2%) changed in abundance after phosphoenrichment. Altogether, peptide loss in the flowthrough after phosphoenrichment has limited impact on protein-level summarization in our dataset, supporting the use of this approach.

We adapted our previous analysis workflow using MSstatsTMT for protein summarization, and comparison with Cre-negative samples for removing non-specific binding contaminants (Dumrongprechachan et al., 2021). MSstatsTMT was used to summarize peptide intensity into protein abundance (Huang et al., 2020). We identified 5582 proteins and quantified 2276 out of 2709 proteins without missing values across conditions (Figure 3d Supplementary file 4). Boxplot of normalized protein abundance shows aligned medians across samples (Figure 3—figure supplement 2a). Principle component analysis (PCA) shows a distinct separation of sample groups with minimal TMT plex effects (Figure 3—figure supplement 2b). To evaluate the extent of axonal enrichment for each protein, we used moderated t-tests implemented in MSstatsTMT to compare P18 striatal samples to the P18 H2B.APEX nuclear/somatic controls. We rank proteins by their log2 fold change (STR–CTX.H2B) (Figure 3e; Supplementary file 5). Proteins with greater log2FC are more likely to have prior axon/presynaptic gene ontology annotations, consistent with a greater GO-axon term annotation rate as rank increases (Figure 3—figure supplement 2c). Proteins with q-value <0.05 and log2FC >0 were considered axonally enriched, depicted in the volcano plot (Figure 3f). Using this criterion, we loosely classified axonal proteins in the dataset into 1129 axon-enriched proteins and 1577 somato-nuclear proteins, with 30% and 17.4% having GO-axon annotations, respectively (Figure 3g). Overrepresentation analysis using the cellular component GO database shows an enrichment of cell projection-related terms for the axon-enriched proteome (e.g. axon, neuron projection, synapse) and nucleus and mitochondrion-related terms for the somato-nuclear proteome (e.g., cytosolic ribosomal subunits, mitochondria, nucleoplasm) (Figure 3h–i; Supplementary file 6). In addition, we also found postsynaptic density proteins such as PSD95 in our axon-enriched proteome. Together, rank based analysis confirms that our approach successfully enriched axonal proteome and GO analysis breaks down the overall subcellular compartments present in the axons. This demonstrates a workflow for mapping axonal proteomes using the Cre-dependent APEX reporter mouse line.

Temporal trajectories of the corticostriatal axonal proteome

Next, we evaluate how the axonal proteome changes across postnatal development. We stringently include only the proteins that were quantified in all biological replicates in this analysis. First, we globally assess whether protein is significantly altered across time, using the least-squares linear regression implemented in the maSigPro R package (Conesa et al., 2006). Neonate (P5) was chosen as the reference time point. maSigPro reported 2274 proteins with corresponding adjusted p-values (Supplementary file 7). As expected, the majority of the detected proteome was highly dynamic, with 1981 out of 2274 proteins significantly changing (adjusted p-values < 0.05). To better observe protein expression patterns, we performed hierarchical clustering of protein abundance, based on time points, revealing 8 general developmental trajectories (Supplementary file 7). The heatmap (Figure 4a) shows individual protein abundance normalized to the neonatal time point for each cluster, reflecting the dataset quality with overall CV <10% (Figure 4—figure supplement 1a). For each cluster, we compute and plot the mean protein abundance (Figure 4b). Clusters 1, 2, 4, and 5 show increase protein expression over time, while clusters 3 and 8 are characterized by decreased protein expression over time. Clusters 6 and 7 fluctuate around the baseline with negative and positive peaks during the early postnatal time point, respectively. To track the overall maturation of presynaptic proteins, we mapped our data to SynGO presynapse annotation. For visualization, we plotted heatmaps divided into seven categories: presynapse, cytosol/cytoskeleton/endosome, presynaptic membrane, active zone, endocytic zone, ribosome and ER, and neuronal dense core vesicles (Figure 4c, Figure 4—figure supplement 1b, Supplementary file 7). Majority of presynaptic proteins shows increased expression over development, while presynaptic ribosomal and ER proteins decreased. For quality control, we plotted several examples of known developmentally regulated proteins (Gonzalez-Lozano et al., 2016), including doublecortin (DCX), neurofilament medium polypeptide (NEFM), synaptophysin (SYP), Vesicular glutamate transporter 1 (VGLUT1), and vesicle marker Ras-related protein Rab-3A (RAB3A) (Figure 4—figure supplement 1c). As expected, an immature neuronal marker like microtubule-associated protein DCX rapidly decreases during development, while mature neuronal markers such as medium intermediate filament (NEFM) and other synaptic and vesicle markers (SYP, VGLUT1, and RAB3A) increase over time. In addition, we performed a meta-analysis comparing our findings with previously reported developmental proteomics studies. We selected three studies for the comparison: cortical synaptic membranes (Gonzalez-Lozano et al., 2016) for P70–P9 vs. P50–P5 in this study (adult), forebrain growth cone membrane and particulates (Chauhan et al., 2020) for P9–P3 vs. P12–P5 in this study (earlypostnatal), and striatal synaptosomes (Peixoto et al., 2019) for P18–P8 vs. P18–P5 in this study (preweanling). We examined the trend of log2FC (e.g. positive or negative) for the overlapping proteome among datasets. We found that >65% of the overlapping proteome between any pair of studies change in the same direction (85% for adult, 65% for earlypostnatal, and 69% for preweanling comparison, respectively) (Supplementary file 7). Given extensive methodological differences, sample types, and variance in age across these studies, we consider the results to be a high level of agreement, confirming protein abundance trajectories over development across systems.

Figure 4 with 2 supplements see all

Download asset Open asset

Temporal trajectories of the corticostriatal axonal proteome.

(A) Heat map showing temporal expression across development. Protein clusters were defined using a time course regression analysis implemented in maSigPro R package. The row dendrogram shows hierarchical clustering of proteins in each protein cluster. Log2 protein abundance was normalized to the mean of the neonatal time point. All biological replicates are shown in columns. (B) Temporal trajectories of protein expression in each cluster. Log2 protein abundance was normalized to the mean of the neonatal time point. Red lines represent the average expression of all proteins in a cluster. Gray lines represent individual protein trajectories. Neonate (P5), early postnatal (P11), preweanling (P18–20), and adult (P50). (C) SynGO analysis of presynaptic protein maturation, annotated by gene names. Proteins mapped to SynGO presynapse ontology term were grouped into seven categories: presynapse, cytosol-cytoskeleton-endosome, presynaptic membrane, presynaptic active zone, presynaptic endocytic zone, presynaptic ribosome and ER, and neuronal dense core vesicle. Heatmap shows log2 protein abundance normalized to neonatal time point. Three categories, presynaptic membrane, presynaptic active zone, and presynaptic endocytic zone are shown. Remaining categories are in Figure 4—figure supplement 2. (D) Over-representation analysis of protein clusters for central nervous system (CNS) traits and disorders. Dot size indicates the number of genes enriched for a particular disease. Color density indicates the degree of significance, –log(pvalue). Only significantly enriched diseases are shown. AD (Alzheimer’s disease), EP (Epilepsy), BP (bipolar disorders), and ASD (Autism spectrum disorders). (E) Reactome pathway enrichment analysis for protein clusters that are associated with CNS traits and disorders (clusters 2, 4, 7, and 8). Top 5 Reactome pathway terms are plotted. Gray bar indicates –log(p value). (F) KEGG axon guidance pathway analysis, annotated by gene names. Heatmap of normalized Log2 protein abundance for proteins mapped to the KEGG axon guidance pathway (mmu04360). *Left*, protein abundance for kinases and phosphatases clustered by family. *Right*, other proteins in the pathway.

Abnormal neurodevelopmental processes including axonal development have been implicated in many psychiatric diseases, with numerous genetic variations and mutations associated with specific disorders (Werling et al., 2020). While transcript levels in brain tissues from multiple species have been well cataloged across development (Colantuoni et al., 2011; Werling et al., 2020), the temporal patterns of protein and protein phosphorylation levels in neuronal subtypes are not fully mapped (Carlyle et al., 2017; Gonzalez-Lozano et al., 2016), which limits our understanding of how disease risk genes contribute to disease etiology. This question is especially important for our understanding of neuropsychiatric diseases of polygenic origin, where protein networks are believed to serve as the point of convergence for multiple genetic risk factors. To determine whether protein clusters are enriched for disease risk genes, we compiled a list of risk genes associated with 11 disorders: autism spectrum disorders (ASD), Alzheimer’s disease (AD), Parkinson’s disease (PD), epilepsy (Ep), developmental delay (DD), Schizophrenia (Scz), multiple sclerosis (MS), major depressive disorder (MDD), bipolar disorders (BP), attention deficit hyperactivity disorder (ADHD), and glioma (Supplementary file 8 Abrahams et al., 2013; Chang et al., 2017; Trubetskoy et al., 2022; Study, 2017; Demontis et al., 2019; Fu et al., 2021; Heyne et al., 2018; International Multiple Sclerosis International Multiple Sclerosis Genetics Consortium, 2019; Rice et al., 2016; Stahl et al., 2019; Wightman et al., 2021; Wray et al., 2018). We used hypergeometric testing implemented in Webgestalt (Wang et al., 2017) to statistically evaluate risk gene enrichment using the 2107 maSigPro quantified proteins as the background. Clusters with p-value < 0.5 and FDR < 0.3 were considered significant. We found that clusters 2, 3, 5, and 6 were enriched in ASD, BP, EP, and AD risk genes, respectively (Figure 4d, Supplementary file 9). In addition, we performed Reactome pathway enrichment analysis (Jassal et al., 2020) to look at which cellular pathways associate with each protein cluster (Figure 4e, Supplementary file 9). We examine the relationship between risk genes and biological pathways in the cluster using STRING-DB interactome analysis (Figure 4—figure supplement 2a). For example, we found that epilepsy risk genes are significantly enriched in protein cluster 5 (p-value = 0.022), which is composed of proteins with increasing expression over time. This implies that postnatal increase in these gene products is important for proper neural circuit development. Cluster 5 also contains genes enriched in α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) trafficking and plasticity (p=1.1e-9), linking epilepsy risk genes in this cluster to excitatory glutamatergic transmission. In fact, reduced corticostriatal excitatory transmission to fast-spiking interneurons was found to trigger absence seizures in Stxbp1 haplodeficient mouse model and this effect was mitigated by pharmacological activation of AMPAR in striatum (Miyamoto et al., 2019). Altogether, this analysis framework generates an overview of how cellular pathways, developmental trajectories, and genetic risk for neuropsychiatric diseases relate to one another in the context of normal axon development in the striatum.

Pathway-centric analysis for the axon guidance system

To gain deeper insights into the regulation of axon development, we performed pathway-centric analyses focusing on signaling pathways during axon guidance. We mapped our proximity labeling axon proteome data to the KEGG axon guidance pathway (Kanehisa et al., 2017). Because kinases and phosphatases are the major regulators of axonal development, we classified mapped genes into two categories: kinases/phosphatases, and axon-guidance related genes. Temporal trajectories of KEGG genes were visualized in heatmaps with row clustering, separating kinases and phosphatases by their families (Figure 4f). We found that the majority of kinases increase in expression over time, such as CAMK and AGC families. However, some kinases (e.g. PAK2, PAK3, PAK5, GSK3B, SSH2, SRC, and FYN) show elevated expression during neonatal and early postnatal time points, but decrease in expression over time, suggesting a functional relevance for early axon guidance. In addition, we detected kinases that are known for their crucial roles in neurodevelopment, including the MAPK family, GSK3B, CDK5, and SRC/FYN. MAPK, GSK3B, and CDK5 are proline-directed kinases while SRC and FYN are non-receptor tyrosine kinases. As for other axon guidance related KEGG genes, we can group them by their functional similarities including ligand-guidance receptor systems, GTPases and their regulators, and cytoskeleton remodeling proteins. In particular, small GTPases including RHOA, RAC1, and CDC42 are important actin cytoskeleton regulators that control axon growth cone expansion during development (Shekarabi et al., 2005).

As an example of mining the datasets generated here, we show the Netrin1 (NTN1)-DCC subnetwork in the KEGG axon guidance pathway to investigate whether corticostriatal projections follow a canonical model of axon development (Figure 4—figure supplement 2b). Dcc is a netrin receptor that expresses in the axon (Vosberg et al., 2020). Activation of the DCC pathway by netrin promotes axon outgrowth by recruiting regulators of actin cytoskeleton (Shekarabi et al., 2005). This process depends on FYN-mediated phosphorylation of DCC (Meriane et al., 2004) as well as SRC-mediated phosphorylation of PLCG1 (Kang et al., 2018). Our data show that DCC expression decreases over time, reflected by negative log2 fold change in the early postnatal and adult samples compared to the neonate. Similarly, SRC and FYN kinases are upregulated during the early postnatal period and downregulated in the adulthood. The timing of SRC and FYN upregulation in the early postnatal time coincides high DCC expression, implicating SRC and FYN as primary regulators of DCC signaling in the neonatal and early postnatal time windows, but not in the adult with matured striatal circuits. Our data for corticostriatal systems are in an agreement with previous reports examining axonal DCC signaling in Xenopus laevis retinal ganglion cells (Meriane et al., 2004) and in mouse midbrain dopaminergic neurons (Kang et al., 2018), validating these data as an important resource for evaluating axonal proteomes.

Phosphosite-centric analysis of kinase substrate interactions

Measuring protein levels in time and space offers many clues as to how a biological system matures across development and how it can break down in neuropsychiatric disease. However, it cannot measure protein state (e.g. activity) which is regulated in large part by post translational modifications. Even the abundance of protein kinases alone is insufficient to infer their activity and their constellation of downstream phosphorylation; thus, a direct measurement of protein phosphorylation is essential. This is especially important in the context of axon development, where coordinated phosphorylation events control axon pathfinding, growth cone elongation, and growth cone collapse. Therefore, during the study design, we aimed to create a workflow that enables quantification of both protein and phosphopeptide abundance from the same samples (Figure 3c). During sample preparation, on-bead digested peptides were labeled with TMTPro reagents, desalted, and then phospho-enriched. We found that TMTPro multiplexing provides sufficient phosphopeptide signal after enrichment (Figure 5a). The unbound fraction can then be used to measure relative protein abundance, as confirmed by the analysis in previous sections (Figures 3–4).

Figure 5 with 1 supplement see all

Download asset Open asset

Temporal trajectories of phosphopeptide abundance and kinase-substrate interaction network.

(A) Phosphopeptide analysis workflow. Raw MS data were analyzed by Proteome Discoverer with false discovery rate (FDR) set at 1% for peptide spectrum matches (PSMs). Only phosphopeptide groups with identified phosphorylated residues were used in the analysis. A unique phosphopeptide is defined by a combination of phosphopeptide and its corresponding protein. Entries without missing abundance values across replicates were used in the timecourse and clustering analysis. (B) Heat map showing temporal expression of phosphorylation across development. Peptide clusters were defined using a time-course regression analysis implemented in maSigPro R package. The row dendrogram shows hierarchical clustering of peptide in each cluster. Log2 peptide abundance was normalized to the mean of the neonatal time point. All biological replicates are shown in columns. (C) Temporal trajectories of phosphorylation in each peptide cluster. Log2 peptide abundance was normalized to the mean of the neonatal time point. Red lines represent the average expression of all proteins in a cluster. Gray lines represent individual protein trajectories. Neonate (P5), early postnatal (P11), preweanling (P18–20), and adult (P5). (D) Kinase-substrate interaction network downstream of FYN kinase. PHOsphorylation Networks for Mass Spectrometry (PHONEMeS) was used to model phosphosites using known prior knowledge network from PhosphoSitePlus database. Nodes, kinase or phosphosite. Node color: dark gray, detected, light gray, inferred by the model. Edges, kinase-substrate interactions. Edge thickness shows the degree of importance inferred for each kinase-substrate interaction. Edge color: blue, green, red for earlypostnatal–neonate, preweanling–neonate, and adult–neonate comparison, respectively.

Here, we only focused on phosphopeptides that were identified with high confidence (1% FDR), with unambiguous phosphosites, and with quantified proteins. After stringently remove peptides with any missing TMT intensity in any sample, we quantified 2353 unique phosphopeptides that were mapped to 708 proteins (Supplementary file 10). Batch effect was corrected with internal reference standard normalization. Global median normalization was used to align sample medians (Figure 5—figure supplement 1a). Many proteins have more than one quantified phosphopeptide, with a median of 2 phosphopeptides per protein (Figure 5—figure supplement 1b). Using the same maSigPro analysis used for proteomic analysis, we divided phosphopeptides into 8 clusters. Clusters 3, 4, and 7 show overall increase in phosphorylation, while clusters 1, 2, 6, and 8 show overall decrease in phosphorylation over development (Figure 5b–c). Next, we used motif analysis to look for phosphopeptide sequences that are enriched in our dataset. Motif-x implementation in R (i.e. rmotifx R package) (Wagih et al., 2016) was used to identify over-represented motifs against the PhosphoSitePlus as the background database (Hornbeck et al., 2015). We found that the top 3 enriched motifs are sequences that can be phosphorylated by proline-directed kinases (Figure 5—figure supplement 1c). This is consistent with our analysis of the axon guidance pathway and previous reports implicating proline-directed kinases such as MAPK family, GSK3B and CDK5, in neurodevelopment (Igarashi and Okuda, 2019).

To incorporate phosphorylation data in the context of axon development, we aim to build a signaling model that can infer upstream kinases in the context of the data, while taking into account developmental timepoint to identify functional kinase-substrate interactions. We used PHOsphorylation Networks for Mass Spectrometry (PHONEMeS) analysis tool to model our data (Gjerga et al., 2021). PHONEMeS reconstructs phosphorylation signaling networks from known kinase-substrate interaction database (i.e. prior knowledge network, PKN), taking into account ‘perturbed’ (differentially detected) phosphosites across time points. With respect to a chosen upstream target kinase, PHONEMeS models phosphorylation signaling propagation from the chosen target node to reach the perturbed phosphosites. Our prior analysis of the NTN1-DCC pathway points towards SRC and FYN kinases as key regulators of the axon guidance signaling. To better understand the axon guidance pathway, we focused on FYN as our target kinase that is significantly upregulated in the earlier time windows Figure 5—figure supplement 1d. For PHONEMeS input, phosphosites that are differentially detected from neonate were considered as perturbed nodes for that time point (e.g. early postnatal – neonate). PHONEMeS allows users to configure background network to a species other than human. This is important because exact phosphorylation residues are not readily mapped across species. We used OmnipathR to obtain known mouse kinase-substrate interactions (Türei et al., 2016). We relied on Cytoscape for visualizing PHONEMeS output (Figure 5d). The model assigned edge color and thickness based on their significance for that time point, evidenced by the data. From this model, we observed that FYN is important for phosphorylation events in the early postnatal time point, revealed by a greater proportion of blue edges (early postnatal) compared to green (preweanling) and red (adult). We also found that FYN can directly or indirectly regulate proline-directed kinases such as MAPK1 (MK_01), CDK5, and GSK3B, whose phosphorylation motifs are enriched in our dataset (Figure 5—figure supplement 1e). Thus, the result is in agreement with our pathway-centric analysis (Figure 4f) and motif analysis (Figure 5—figure supplement 1c), supporting the model that FYN is one of the key regulators in the axon guidance pathway in the corticostriatal systems. Altogether, phosphosite-centric analysis by PHONEMeS provides a useful roadmap to extract kinase-substrate interactions that are contextualized to the data, as well as inferences about intermediate kinases and their potential roles in the signaling pathways.

Correlation analysis of phosphopeptide and protein abundance

Phosphorylation plays a crucial role in protein function, including control over enzymatic activity and protein-protein interactions. As protein expression changes, phosphorylation (as a percentage of total protein) may remain steady or change as levels or activity of kinase and phosphatases fluctuate. Our dataset highlights dramatic reshaping of axonal proteome levels across development, which raises the question of how protein activity is altered on top of changing protein levels across development. To address this question, we computed Spearman rank correlations between phosphopeptides and proteins across developmental time (Figure 6a, Supplementary file 11). Values associated with correlation coefficients from 2109 peptide-protein pairs were corrected, with 10% FDR as the cutoff for statistical signficance. We found that the majority of phosphopeptide intensities correlate with protein abundance, illustrated by a left-skewed histogram (Figure 6b). Out of 309 phosphopeptides with a negative correlation to protein level, 59 decorrelated peptides passed the FDR cutoff (Figure 6c). To determine the directionality of changes between protein and peptide abundance across time, we calculated log2 fold change for the corresponding protein between adult and neonate for each peptide. Comparing protein log2 fold change with correlation coefficients, we found 17 decorrelated peptides with negative protein log2 fold change, indicating that phosphorylation increases over time while protein abundance decreases. Analogously, 42 decorrelated peptides with positive protein log2 fold change indicate that phosphorylation decreases over time as protein increases.

Figure 6

Download asset Open asset

Correlation analysis of phosphorylated peptides and proteins implicates Tsc2 Ser664 in MTOR regulation in corticostriatal axon development.

(A) Correlation analysis workflow. Spearman rank correlation is calculated for each protein and phosphopeptide pair across time points. One protein may have multiple phosphopeptide-protein pairs, yielding multiple correlation values. False-discovery rate was controlled by the Benjamini-Hochberg approach.(B) Distribution of protein-phosphopeptide correlations. Left-skewed histogram (bin width = 1). Most phosphopeptides correlate with their respective protein expression across time. (C) Classification of protein-phosphopeptide correlations. Cutoff for correlated peptides, Spearman correlation ≥0. Of 1800 phosphopeptides, 1344 are positively correlated with proteins by their temporal expression at 10% false discovery rate (FDR). A small proportion of peptides (59 of 309 phosphopeptides) are decorrelated (inversely correlated) at 10% FDR. 42 phosphopeptides decrease in abundance as protein levels increases, while 17 phosphopeptides show the opposite pattern. (D) TSC2 and TSC2_Ser664 expression. Log2 abundance for TSC2 (open circles) and TSC2_Ser664 (solid circles) were normalized to the neonatal time point (Spearman rank correlation: –0.570, FDR = 0.009. n=5 for each time point. Error bars represent SEM).(E) Expression heatmap for proteins and phosphosites in the TSC2-RHEB-MTOR signaling pathway. Log2 protein expression was normalized to the neonate time point. Log2 phosphopeptide signal was normalized to the corresponding protein abundance and to the neonate time point. (F) Model for MTOR-mediated growth cone collapse.

Next, we investigate whether decorrelated phosphosites are important for axonal development. To narrow down the list, we chose to examine decorrelated phosphopeptides on proteins that are associated with disease risk genes (Supplementary file 11). Among decorrelated peptides, 15 phosphosites are mapped to disease risk genes with 12 out of 15 sites conserved in humans. Particularly, we found TSC2_Ser664 in our PHONEMeS model under the regulation of MAPK1 (Figure 5d, Figure 5—figure supplement 1e). We found that TSC2 protein moderately increases throughout development, while TSC2_Ser664 rapidly drops off after the neonatal period (Figure 6d, corr = –0.570, FDR = 0.009). TSC2 is one of the key regulators of the MTOR pathway (Ma et al., 2005), and is implicated in autism spectrum disorder. Activation of TSC1/TSC2 complex decreases MTOR activity by inhibiting RHEB, a positive MTOR regulator. When TSC2 is phosphorylated by MAPK1 at Ser664, TSC1/TSC2 complex is destabilized, allowing RHEB to activate MTOR. This interaction between MAPK1 and TSC2_Ser664 during the earlier time point is emphasized by the PHONEMeS model. To evaluate this regulation more closely, we generated a simplified model of TSC2-RHEB-MTOR signaling and plotted relative protein and phosphopeptide abundance normalized to the neonatal time point (Figure 6e). In our dataset, relative phosphorylation of TSC2 at Ser664 in the corticostriatal axons is highest around P5 and decreases over time. Elevated TSC2_Ser664 phosphorylation inhibits TSC1/TSC2 activity, in turn, activating MTOR. We inferred the activity of MTOR by examining its substrate phosphorylation. We detected one MTOR substrate, WIPI2 and WIPI2_Ser395. Consistent with the model, WIPI2_Ser395 phosphorylation is higher in the neonate, supporting the hypothesis that the overall MTOR activity is greater during axon innervation (Figure 6f). As neural circuits mature, axonal outgrowth stops and growth cone collapses. A general mechanism for growth cone collapse mediated by ephrin-Eph receptor signaling was studied in the retinal ganglion cells (Nie et al., 2010). The activation of EPH receptor leads to inhibition of MAPK1, in turn, allowing TSC2 to inhibit RHEB and MTOR. In the striatum, it has been reported that the growth cones collapse after P7. As expected, we found EPHA4 and EPHB3 Eph receptor expression increase from P5 to P11, coinciding with reduction in TSC2_Ser664 and WIPI2_Ser395 (Figure 6f). Therefore, our data support the general mechanism that MTOR activity is elevated during axon outgrowth and reduced during growth cone collapse, and that this process is conserved in corticostriatal projections.

In summary, we demonstrate a sensitive proximity labeling workflow involving a tandem enrichment of biotinylated proteins and phosphopeptides. We use a newly generated and validated APEX reporter line to map the axonal proteome in the corticostriatal system and establish an analysis framework to examine the subcellular proteome and phosphoproteome across development, generating broadly applicable resources for studying the proteomic landscape of mammalian neural circuits.

Discussion

Genetically encoded proximity labeling is an effective way to perform cell-type and subcellular compartment specific proteomics in the mouse brain (Dumrongprechachan et al., 2021; Hobson et al., 2022; Rayaprolu et al., 2021; Sun et al., 2022). Here, we generated a new Cre-dependent APEX2 reporter line to capture snapshots of the neuroproteome with cell type specificity, illustrated by several genetic crosses including Vglut2^Cre, Vgat^Cre, and Rbp4^Cre. In comparison with biotin ligase-based reporters (e.g. BioID, TurboID), APEX-mediated labeling occurs within 1 hr ex vivo in acute brain sections, while in vivo BioID/TurboID labeling takes several days up to weeks, via continuous biotin supplementation (Rayaprolu et al., 2021; Spence et al., 2019; Sun et al., 2022; Uezu et al., 2016). Therefore, the APEX-based approach is more suitable for biological questions that require short temporal windows, while BioID/TurboID can measure in vivo steady-state proteome spanning the duration of biotin administration. Importantly, short H₂O₂ treatment during labeling induction may result in undesired oxidative stress signaling, which could be reduced through several additional steps. Transcardial perfusion during acute slice preparation may help remove endogenous catalases and peroxidases in blood vessels. Specialized protective perfusing solution, such as the N-methyl-D-glucamine-based solution (Hobson et al., 2022), could provide additional neuroprotection. Transcardial delivery of biotin phenol was found sufficient to induce biotinylation in cardiac tissues (Liu et al., 2020), but whether biotin phenol or other APEX substrates can penetrate the blood brain barrier effectively remains to be investigated. These additional steps must be tested for cell classes and tissue types of interest. We speculate that the blood brain barrier could be an obstacle for perfusion-based biotinylation in the central nervous system.

Profiling the axonal proteome or transcriptome is challenging, particularly from complex tissues composed of multiple intermingled cell types. Canonical approaches utilize cell culture models in compartmentalized devices to ensure directional growth (Cagnetta et al., 2018; Chuang et al., 2018), laser capture microdissection of axon terminals (Zivraj et al., 2010), manual dissection of nerve segments (Michaelevski et al., 2010), synaptosome sorting from hippocampal mossy fibers (Apóstolo et al., 2020), or growth cone (GC) sorting from developing brain (Chauhan et al., 2020; Poulopoulos et al., 2019). However, no single approach can flexibly measure the overall axonal proteome across a wide age range, with high efficiency, genetic targeting, and pre-synaptic specificity. In this work, we leverage the early expression of APEX in the transgenic reporter cross to map the proteomic landscape of corticostriatal axons from neonate (P5) to young adult (P50). Enabled by temporal precision of APEX-mediated biotinylation, we demonstrate that APEX-based proximity labeling detects dynamically changing proteome of corticostriatal axons across multiple subcellular compartments including axoplasm, synapses, and mitochondria, without the need for biochemical fractionation, or sample pooling beyond P5. It is unexpected that postsynaptic proteins, such as the postsynaptic density 95 protein PSD95, were detected in our axon dataset. One possibility is that such proteins exist in high abundance and are strongly associated with presynaptic partners during streptavidin enrichment, resulting in co-purification. Although it is unlikely, we cannot completely exclude the possibility of leakiness of the APEX2 reporter expression, or the transfer of APEX2 across compartments. Another explanation is that corticostriatal projections form axo-axonic synapses among excitatory inputs (e.g. cortical-cortical or cortico-thalamic axo-axonic synapses), since PSD95 in axons implies glutamatergic presynaptic partners. Indeed, prior ultrastructural observations identified putative glutamate-dopamine axo-axonic contacts in the striatum (Cover and Mathur, 2021). Dual color fluorescent tracing also showed rare appositions among thalamic, cortical, and basolateral amygdala inputs onto dendritic spines of spiny projection neurons in nucleus accumbens (Xia et al., 2020). Therefore, further ultrastructural studies are needed to determine whether there are glutamaterigic axo-axonic synapses on corticostriatal axons in the dorsal striatum.

Furthermore, we have designed a workflow to add additional steps during the sample preparation for enriching phosphorylated peptides from the same samples, as the ex vivo biotinylation yields sufficient materials. The tandem enrichment workflow enables measurement of both protein and phosphopeptide abundance using TMTPro-based multiplexing and quantification, similar to a recent report (Zhang et al., 2022). Although some peptide loss was observed in this strategy, protein quantification in the flowthrough was highly correlated with the unenriched samples. This is crucial for our study, because investigating phosphosite abundance alone without protein abundance can lead to inaccurate interpretation, especially in a system where protein abundance is changing (Wu et al., 2011). While our approach profiles phosphosites on biotinylated proteins, Qin et. al., and colleagues recently developed a sequential protein enrichment strategy that combines streptavidin enrichment with functional protein enrichment (e.g. phase separation for RNA-binding proteins, IMAC resin for phosphoproteins) to discover novel functions of a protein subclass in a specific subcellular compartment in HEK cells, highlighting the utility of tandem enrichment in the context of APEX-based proteomics (Qin et al., 2021).

Using our workflow, we interrogate axonal proteome of the developing corticostriatal system. We obtained sufficient profiling of the overall landscape. Despite stringent removal of proteins with missing measurements, we were able to perform pathway-centric and phosphosite-centric analyses with good coverage. We showcase an important set of systems-level analyses, encompassing a mapping of protein and phosphopeptide temporal trajectories, disease risk gene association, contextualized kinase-substrate interactions, and correlations between protein and phosphopeptide levels across time, enabling interrogation of complex signaling pathways such as the axon guidance. Molecular level mechanisms of axon guidance systems in the brain have been investigated in a limited number of models. Several examples including Netrin-DCC signaling in dopaminergic (Kang et al., 2018; Vosberg et al., 2020), thalamocortical (Castillo-Paterna et al., 2015) and commissural axon projections (Varadarajan et al., 2017), and Ephrin-EPH receptor signaling in retinal ganglion cell projections (Nie et al., 2010), callosal axon navigation (Nishikimi et al., 2011) and hippocampal mossy fibers (Xu and Henkemeyer, 2009). Our pathway-centric analysis reveals that both Netrin-DCC signaling and EPHA signaling are conserved in corticostriatal projections, demonstrating an analysis framework that synthesizes prior knowledge from different neural circuits. Beyond this example, many signaling pathways remains to be explored and our datasets can be used as a general resource to generate testable models for axon development in the context of neurodevelopmental disorders for further mechanistic studies.

Our example of phosphorylation network and correlation analysis emphasizes developmental roles of MAPK1, a proline-directed kinase (PDK), in the MTOR-dependent regulation of axon growth cones. Other developmentally regulated PDKs include CDK5 and GSK3B, well-known regulators of the axon guidance pathway. Deletion of neuron-specific CDK5 activator protein, p35, disrupts callosal axons assimilation into the corpus callosum in mice and affects cortical neuron migration and axodendritic orientations (Chae et al., 1997; Kwon et al., 1999). In contrast, inactivation of GSK3B promotes axon extension (Zhou et al., 2004). Deletion of Gsk3b in cortical neurons disrupts dendritic orientations, but spares axon elongation into the corpus callosum. Density and arborization of axons in Gsk3b–/– neurons remain to be investigated (Morgan-Smith et al., 2014). Altogether, proper PDK activity is essential for axonal development.

In this study, we identified FYN as a regulator of PDKs via the kinase-substrate map. Fyn knockout (Fyn KO) shows cortical, hippocampal, and cerebellar abnormalities in neuronal migration, leading to partial Reeler phenotype and behavior (Grant et al., 1992; Kuo et al., 2005). In addition, inversion of cortical lamination was observed in Fyn KO mice (Yuasa et al., 2004). Therefore, we predict that dysregulated cortical stratification in Fyn KO mice likely affects cortical projection patterns and striatal function. FYN is also known to regulate CDK5, a PDK that controls actin cytoskeleton remodeling (Sasaki et al., 2002). Fyn KO resulted in abnormal axodendritic branching of cortical neurons due to decrease in CDK5 tyrosine phosphorylation and activity (Sasaki et al., 2002). The same CDK5 phosphosite was found in our kinase-substrate network, highlighting Fyn as a regulator of cell migration and subcellular morphology development. However, the specific role of Fyn in the development of corticostriatal connectivity is understudied, and to what extent Fyn deletion affects other PDK functions in axons remains unclear. Generation of conditional Fyn KO in corticostriatally projecting neurons would be informative to elucidate its role in striatal development. Altogether, our analyses show signaling crosstalk among PDKs and their regulators in developing brain.

Data generated by this work serve as a general reference for overall corticostriatal afferent development at the proteomic and phosphoproteomic levels. Notably, cortical afferents are classified into pyramidal tract (PT) and intertelencephalic tract (IT) projections (Shepherd, 2013). PT projections innervate the striatum in the embryonic period, while IT projections arise postnatally, around P3–P4 (Nisenbaum et al., 1998; Sheth et al., 1998; Sohur et al., 2014). Both types of cortical projections to the striatum are topographically organized, forming functionally and spatially distinct striatal regions to process and integrate cortical inputs (Hooks et al., 2018). With appropriate Cre-driver lines (e.g. Tlx3^Cre_PL56, Sim1^Cre_KJ18 for IT and PT neurons, respectively) (Gerfen et al., 2013) and timing, analogous approaches can be applied to examine the molecular basis of regional input specific proteome from PT or IT projections during early postnatal development. As a proof of principle in this study, we focus on corticostriatal axons during normal development, but axon maturation requires both presynaptic and postsynaptic interactions. The timely formation of striatal circuits including the maturation of striatal neurons depends on proper levels of synaptogenesis and dendritic spinogenesis, regulated by recurrent network activity, neuromodulation, and experience (Kozorovitskiy et al., 2012; Kozorovitskiy et al., 2015). Disruption of axon development by disease associated mutations or absence of appropriate dopamine modulation can result in impaired striatal function (Lieberman et al., 2018; Peixoto et al., 2016). Thus, our work positions the field for genetically targeted proteomics that can be combined with pharmacological, genetic, or engineered effectors to reveal the principles of neural circuit formation between pre- and post-synaptic neurons.

In sum, genetically encoded proteomics tools are rapidly evolving. Together with advances in mass spectrometry, we are starting to obtain sufficient depth of genetically targeted proteome from low microgram samples, with lower rates of missing measurements and with post-translational modification information. With more proteomic datasets of this type, numerous features including cell type, subcellular compartment, developmental trajectory, and phosphorylation can facilitate more accurate theoretical modeling of signaling pathways in the brain during development and in disease contexts. Altogether, methodological and conceptual advancements, along with the datasets generated by this work, provide a new avenue for interrogating the proteomic landscapes of any genetically targeted neural system in the mouse brain.

Share this article

Cite this article

Generation of a Cre-dependent APEX2 reporter mouse line.

Figure 1—source data 1

APEX-mediated biotinylation in brain tissue.

Figure 2—source data 1

Genetic targeting approach for proteomics analysis of corticostriatal axonal compartment.

Temporal trajectories of the corticostriatal axonal proteome.

Temporal trajectories of phosphopeptide abundance and kinase-substrate interaction network.

Correlation analysis of phosphorylated peptides and proteins implicates Tsc2 Ser664 in MTOR regulation in corticostriatal axon development.

Author details

Vasin Dumrongprechachan

Contribution

Competing interests

Ryan B Salisbury

Contribution

Competing interests

Lindsey Butler

Contribution

Competing interests

Matthew L MacDonald

Contribution

For correspondence

Competing interests

Yevgenia Kozorovitskiy

Contribution

For correspondence

Competing interests

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)

Categories and tags

Research organism

Further reading