1. Introduction
Pathological substance use disorders are a group of recalcitrant, relapsing and remitting conditions that have deleterious effects on the patient, their family, and society at large. While there have been attempts made to mitigate the prevalence of substance abuse disorders, the incidences of illicit substance abuse and misuse has remained steady or increased since 1990 [
1], and the economic burden created by substance use disorders is tremendous with a societal cost of over 500 billion dollars per year in the United States alone [
2]. Of these conditions, pathological use of psychostimulants such as cocaine and amphetamine account for a significant portion of the morbidity and mortality. However, there are currently no FDA-approved pharmacological treatments for cocaine use disorder [
3,
4]. Previous drug discovery attempts in this arena have generally failed due to lack of efficacy, intolerable side effects, or both [
5,
6,
7].
In recent years there has been growing interest in the role that neuroimmune interactions play in the development of psychiatric illness, including addictive disorders [
8,
9,
10]. This raises the intriguing possibility that targeting neuroimmune signaling pathways may be a viable translational treatment strategy to reduce the persistence of pathological substance use disorders. Our lab recently discovered granulocyte-colony stimulating factor (G-CSF) as a cytokine that is up-regulated both centrally and peripherally after chronic cocaine treatment [
11]. Peripheral injections of G-CSF potentiated the development of locomotor sensitization, conditioned place preference, and self-administration of cocaine, and blockade of G-CSF function in the mesolimbic dopamine system abrogated the formation of conditioned place preference.
While the behavioral effects of G-CSF on cocaine-induced behavioral plasticity are known, the cellular and molecular mechanisms underlying these effects remain to be identified. We have recently found that acute treatment with G-CSF enhances release of dopamine from the ventral tegmental area (VTA) into the nucleus accumbens (NAc) [
12]. Previous work has found that the G-CSF receptor is robustly expressed on dopamine expressing neurons of the midbrain [
13,
14]. G-CSF has been found to be a potent neurotrophic and neuroprotective factor in response to stroke or other insults [
15,
16,
17]. Importantly, G-CSF is also neuroprotective in the midbrain where treatment with G-CSF reduces neuronal death in the MPTP model of Parkinson’s disease [
18]. Additionally, within these midbrain neurons, G-CSF has been found to induce activity of the immediate-early gene
Cfos and acute treatments upregulate tyrosine hydroxylase—the rate limiting step in dopamine synthesis [
13]. Moving forward, it will be critical to determine the molecular signaling cascades that control the effects of G-CSF on behavior.
Given the known effects of G-CSF within the midbrain and the importance of the VTA in the development and persistence of substance use disorders [
19,
20] we characterized the effect of G-CSF and its interaction with cocaine on the proteomic makeup of the VTA. Via an unbiased quantitative proteomics approach, we identified and characterized the regulation pattern of more than two thousand proteins in the VTA. We found that G-CSF treatment on its own regulated many of the same signaling pathways that are regulated by cocaine and induced numerous factors important for neurite and dendritic spine plasticity. Specifically, we found significant regulation of proteins predicted to be downstream from Fragile X mental retardation (FMRP) and mammalian target of rapamycin (mTOR). Additionally, we report multiple intracellular signaling cascades that are differentially regulated by combined cocaine and G-CSF treatment, suggesting future targets for study on the effects of G-CSF on the behavioral response to cocaine.
2. Materials and Methods
2.1. Animals and Drug Treatments
Male C57BL/6J mice (7 weeks old ~20–25 g; Jackson Laboratories, Bar Harbor, ME, USA) were housed in the animal facilities at Icahn School of Medicine at Mount Sinai. Mice were maintained on a 12:12 h light/dark cycle with lights on at 0700 and lights off at 1900. Mice had food and water available ad libitum throughout the experiments. Drug treatments were performed in a 2 × 2 design with the first group receiving phosphate buffered saline vehicle, followed by saline (PBS/Sal), the second group was injected with G-CSF 50 μg/kg (GenScript Biotech, Piscataway, NJ—G-CSF/Sal) followed by saline, the third group was injected with PBS followed by cocaine hydrochloride 7.5 mg/kg (NIDA—PBS/Coc), and the fourth group with both G-CSF and cocaine (G-CSF/Coc). Injections were performed once daily for 7 days and the animals were euthanized 24 h after the final injection. All animals were maintained according to the National Institutes of Health guidelines in Association for Assessment and Accreditation of Laboratory Animal Care accredited facilities. All experimental protocols were approved by the Institutional Animal Care and Use Committee at Mount Sinai.
2.2. Protein Preparation
For each mouse the VTA was dissected from fresh tissue on ice using a reference brain atlas and anatomical landmarks to guide dissection. Tissue from each animal was then sonicated into 50 μL of ice-cold RIPA buffer (50 mM Tris [pH 8.0], 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 110 mM NaCl & Halt Protease and Phosphatase Inhibitor Cocktails [Fisher]). Protein concentrations were determined by Bradford colorimetric assay according to manufacturer protocols (Thermo Fisher, Waltham, MA, USA). For these analyses tissue from individual animals was used as distinct data points. There was no pooling of samples between animals other than to make the master mix for cross-assay normalization as described below.
2.3. Tandem Mass Tag (TMT) Labeling
TMT samples were prepared according to the manufacturer’s instructions. Briefly, 50 µg proteins per condition were reduced by incubating the samples with TCEP (tris(2-carboxyethyl)phosphine) at 55 °C for 1 h and alkylated by incubating with iodoacetamide at room temperature in the dark for 30 min. The proteins were precipitated by the acetone precipitation, resuspended in 25 mM Triethyl ammonium bicarbonate (TEAB) and digested with trypsin at 37 °C overnight. The peptide concentrations of the tryptic digests were measured by Amino Acid Analysis method using a Hitachi L-8900 Amino Acid Analyzer. Equal amount (30 µg) of peptides were labeled with TMT reagents from the TMT-10plex kit (ThermoFisher Scientific). The samples were labeled by distributing them into two experimental groups. Each TMT experimental setup has two TMT tags (126 and 129N) that labeled the two pooled samples, which were created by collecting and combining an equal amount of peptides from each sample. The pooled samples served as a global internal standard for normalizing the data across the two experimental setups and is henceforth referred to as the Master Mix. The remaining 8 TMT reagents in each experimental setup were used for labeling the two biological replicates for each of the four conditions. The TMT labels carried by each sample and the mixing design is shown in
Figure 1B. For labeling, the peptides were incubated with TMT reagents for 1 h at room temperature. The labeling reaction was quenched by adding 5% hydroxylamine to the sample and incubating for 15 min. Before combining the labeled samples for mass spectrometry analysis, an aliquot was combined and analyzed by LC-MS/MS to ensure the labeling was complete and also that the mixing generated a ratio of 1. Eventually, all ten labeled samples were combined and fractionated offline by high pH reversed-phase fractionation.
2.4. High-pH Reversed-Phase C18 Peptide Fractionation
High-pH reversed-phase C18 peptide fractionation was performed on an ACQUITY UPLC H-class system (Waters Corporation, Milford, MA, USA) on ACQUITY UPLC BEH C18 column, 1.7 µm, 2.1 mm × 50 mm. Elution was performed at a flowrate of 0.4 mL/min using a gradient of mobile phase A (10 mM ammonium acetate) and B (10 mM ammonium acetate in 90% acetonitrile). The gradient extended from 2% to 37% mobile phase B in 17.6 min and then to 75% mobile phase B in another 8.8 min. The collected pooled 10 fractions were dried in a speed-vac centrifuged and reconstituted in buffer A (0.1% formic acid in water); 400 ng digests from each fraction were used for reversed-phase liquid chromatography-tandem mass spectrometry (RP-LC-MS/MS/MS) analysis.
2.5. SPS-MS3 TMT Data Acquisition on an Orbitrap Fusion Tribrid Mass Spectrometer
RP-LC-MS/MS/MS was performed using a nanoACQUITY UPLC system (Waters Corporation, Milford, MA, USA) connected to an Orbitrap Fusion Tribrid (ThermoFisher Scientific, San Jose, CA, USA) mass spectrometer. After injection, samples were loaded into a trapping column (nanoACQUITY UPLC Symmetry C18 Trap column, 180 µm × 20 mm) at a flowrate of 5 µL/min and separated with a C18 column (nanoACQUITY column Peptide BEH C18, 75 µm × 250 mm). The compositions of mobile phases A and B were 0.1% formic acid in water and 0.1% formic acid in acetonitrile, respectively. Peptides were eluted with a gradient extending from 6% to 20% mobile phase B in 120 min and then to 40% mobile phase B in another 50 min at a flowrate of 300 nL/min and a column temperature of 37 °C. The data were acquired with the mass spectrometer operating in a top speed data-dependent mode with multinotch synchronous precursor selection (SPS)-MS3 scanning for TMT tags. The full scan was performed in the range of 380–1580 m/z at an Orbitrap resolution of 120,000 at 200 m/z and automatic gain control (AGC) target value of 2 × 105, followed by selection of ions above an intensity threshold of 5000 for collision-induced dissociation (CID)-MS fragmentation in the linear ion trap with collision energy of 35%. The isolation width was set to 1.6 m/z. The top 10 fragment ions for each peptide MS2 were notched out with an isolation width of 2 m/z and co-fragmented with higher-energy collision dissociation (HCD) at a collision energy of 65% to produce MS3 scans which were analyzed in the Orbitrap at a resolution of 60,000.
2.6. Protein Identification and Quantification
Raw data from the Orbitrap Fusion were processed using Proteome Discoverer software (version 2.1, ThermoFisher Scientific, San Jose, CA, USA). MS2 spectra were searched using Sequest HT which was set up to search against the SwissProt mouse database (downloaded on 06292017). The search criteria included 10 ppm precursor mass tolerance, 0.6 Da fragment mass tolerance, trypsin enzyme and maximum missed cleavage sites of two. Static modification included carbidomethylation (+57.02146 Da) on cysteine and TMT labels (+229.16293 Da) on lysine and peptide N-terminus. Dynamic modifications included oxidation (+15.99492 Da) on methionine, deamidation (+0.98402 Da) on asparagine and glutamine, and acetylation (+42.01057 Da) on protein N-terminus. Peptide spectral match (PSM) error rates were determined using the target-decoy strategy coupled to Percolator modeling of true and false matches [
21]. Reporter ions were quantified from MS3 scans using an integration tolerance of 20 ppm and the most confident centroid as the integration method in the Reporter Ions Quantifier node.
2.7. Mass Spec Data Analysis
Scaffold Q+ (version Scaffold_4.8.5, Proteome Software Inc., Portland, OR, USA) was used for label-based TMT10-plex quantitation of peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 95.0% probability by the Scaffold Local FDR algorithm. Protein identifications were accepted if they could be established at greater than 99.0% probability and contained at least 2 identified peptides. Peptide probabilities were calculated by the Scaffold Local FDR algorithm, and protein probabilities were assigned using the Protein Prophet algorithm [
22]. Proteins identified with fewer than two peptides were excluded from quantitation. Proteins sharing redundant peptides were grouped into clusters. Normalization was performed iteratively (across samples and spectra) on intensities, as previously described [
23]. After setting the minimum dynamic range to 5%, removing spectra that were missing a reference value and those that arose from degenerate peptides that match to more than one protein, the remaining log-transformed spectra were weighted by an adaptive intensity weighting algorithm. Of 71,507 spectra in the experiment, 59,861 (84%) met the threshold criteria and were included in quantitation. Statistical testing was performed using uncorrected Student’s t-test between groups.
p-values < 0.05 were considered statistically significant. Volcano plots were created using GraphPad Prism version 7 (La Jolla, CA, USA). Pathway analyses to determine specifically regulated pathways were created using Ingenuity Pathway Analysis software from Qiagen. The network diagrams depicted in Figure 3 were created using significantly regulated proteins from our dataset that were predicted to be directly downstream of the hub genes, and then up to 5 genes predicted to be downstream of each of those was added to the outer layer. There were no additional filters applied. Predicted targets downstream from activity-dependent transcription factors was performed using the Enrichr analysis suite (
http://amp.pharm.mssm.edu/Enrichr/). Full methodology for the Enrichr analyses is described in detail in the original Chen et al. paper [
24]. Heatmaps were created using the freely available Morpheus software from the Broad Institute (
https://software.broadinstitute.org/morpheus).
2.8. Western Blot Analysis
For Western blot analysis animals were treated identically to those above, and VTA tissue was fresh dissected and frozen on ice until further processing. Samples were thoroughly sonicated into SDS lysis buffer (1% SDS, 50 mM Tris [pH 8.0], 130 mM NaCl, 5 mM EDTA, 50 mM NaF, 1 mM PMSF, protease and phosphatase inhibitor cocktails from ThermoFisher) according to previously published procedures [
25]. Sample concentrations were determined using a Bradford colorimetric assay (ThermoFisher) according to manufacturer protocols, and 10μg of protein was run on a 4–12% gradient gel. Proteins were transferred to PVDF membranes using standard techniques. Membranes were blocked using LiCor blocking buffer with TBS based mixed 1:1 with standard TBS for one hour at room temperature. Primary antibodies were incubated with mixing at 4 °C overnight with constant agitation. Primary antibodies used were tyrosine hydroxylase (AbCam #ab112, 1:1000), Mecp2 (Cell Signaling #3456, 1:1000) & actin (Cell Signaling #3700, 1:10,000). Membranes were washed with TBS + Tween-20 before incubation with secondary antibodies raised against the appropriate species (LiCor, 1:10,000) for one hour at room temperature. Membranes were then washed with TBS + Tween-20, rinsed with TBS without Tween, and imaged using a LiCor Fluorescent imager. Image quantification was performed using freely available ImageJ software. Representative images shown in Figure 8 were flipped horizontally to achieve representative bands in the correct order but were not otherwise altered or retouched.
4. Discussion
We have recently identified G-CSF as a key mediator of neuronal and behavioral plasticity in response to cocaine [
11]. In this manuscript an unbiased proteomics analysis is employed to identify protein changes induced in the VTA by G-CSF, both on its own and in combination with cocaine. In our original studies G-CSF signaling in the NAc was found to play a key role in the behavioral effects of G-CSF. Given that dopamine release from VTA terminals in the NAc is a crucial substrate of reward learning and the attribution of salience to rewarding stimuli, understanding changes in protein expression in the VTA is critical for understanding the neuroplasticity that occurs in response to drugs of abuse. Additionally, since the publication of our initial study we found that peripheral injections of G-CSF are capable of modulating dopamine signaling by enhancing release from VTA terminals in the NAc [
12]. Given this and the fact that G-CSF receptors are densely expressed in the VTA [
28] lead to these proteomic analyses of the VTA.
Review of the literature demonstrates that the exact intracellular signaling mechanisms of G-CSF in the brain are not fully clear and may be complex. G-CSF treatment has variously been shown to induce activity of the Jak-Stat, Erk, and CREB1 signaling cascades among others [
28,
39,
40,
41,
42]. These results demonstrate that treatment with G-CSF decreased signaling in the CREB1 transcription factor signaling cascades, as well as the cAMP and PKA pathways which are well known to be upstream of CREB1 (
Figure 2B) [
43]. Increased expression of CREB1 in the NAc and in subregions of the VTA has been shown to decrease cocaine reward, and inhibition of CREB1 in these regions has been shown to enhance reward in a region-specific manner [
35,
44]. Analysis of significantly upregulated proteins in the G-CSF treatment group found that CREB1 was predicted to be one of the transcription factors driving gene expression (
Figure 7). This apparent discrepancy in
Figure 2 and
Figure 7 may be due to the fact that the IPA analysis looks at networks of proteins based on literature review, while the Enrichr software looks only at those proteins predicted to be directly downstream of the transcription factor. Since G-CSF enhances cocaine intake and place preference and alters CREB1-related signaling, it is possible that the behavioral effects of G-CSF are at least partially mediated through the CREB1 pathway.
We also observed regulation of proteins related to the maintenance of synapses and other cell-cell contacts in our G-CSF treated groups (
Figure 2B and
Figure 4). This is of particular interest as numerous studies have demonstrated that changes in synapse density are induced by cocaine and are important for the behavioral response to drugs of abuse [
45,
46]. While most of these studies have focused on the NAc, there is also evidence for synaptic remodeling in the VTA [
47,
48]. These findings raise the possibility that G-CSF may participate in neurite remodeling, and may prime animals for further changes in synaptic structure in response to cocaine, thus leading to the potentiation of behavioral response induced by G-CSF [
11].
The G-CSF-treated animals displayed significant changes in signaling cascades that are related to initiation of mRNA translation. IPA analyses predicted that one of the most up-regulated canonical signaling pathways is the eukaryotic initiation factor 2 (EIF2) pathway which is a critical mediator of protein translation initiation and has been implicated in synaptic plasticity and memory [
49] (
Figure 2C). Interestingly, EIF2 signaling has been shown to be inhibited by PKA signaling which is found to be decreased in our G-CSF-treated animals (
Figure 2B). EIF2 is also known to be activated by the mTOR pathway which was predicted to be a key upstream regulator of the altered proteins in our dataset [
50] (
Figure 3B). Indeed, the two most highly predicted upstream regulators, mTOR and FMRP (
Figure 3), have been shown to be critical regulators of translation of synaptic mRNAs and play key roles in synaptic plasticity [
51].
There is a growing literature demonstrating the importance of regulators of synaptic translation regulators in the neuronal and behavioral plasticity in response to cocaine. Recently, an elegant study by the Wolf lab demonstrated increased protein translation during cue-induced drug seeking, and inhibition of mTOR or EIF2 could significantly attenuate cocaine seeking [
52]. Studies of FMRP have shown that it is also critical for the rewarding effects of cocaine and changes in synapse structure in response to cocaine [
53]. A number of studies have found roles for mTOR-mediated intracellular signaling cascades in NAc in response to cocaine [
54,
55,
56]. Behaviorally it has been demonstrated that inhibition of mTOR with rapamycin can reduce locomotor sensitization, conditioned place preference, and cocaine seeking [
57,
58,
59]. The role of mTOR in the VTA was recently interrogated by Liu and colleagues who found that deletion of mTOR reduced VTA dopamine release and decreased conditioned place preference for cocaine [
60].
When examining the number of proteins that were significantly altered between the different treatment groups, it was found that treatment with G-CSF alone leads to changes in the largest number of proteins (
Figure 5 and
Figure 6). This may be due to the fact that activation of the G-CSF receptor has been coupled to direct activation of transcription factors [
39,
40,
61]. In contrast, cocaine directly leads to changes in multiple neurotransmitter systems, but its effects on gene expression are tightly coupled with context and behavior [
62,
63,
64]. It is interesting that the combination of G-CSF and cocaine lead to the smallest number of regulated proteins of the three treatment groups (
Figure 5B). This suggests the possibility that there are interactions between signaling pathways after G-CSF and cocaine in the two that temper changes in protein expression in the VTA.
One of the more surprising findings from these studies was the similarity in changes between treatment groups. Pathways that were regulated by G-CSF, Cocaine, or the combination were largely the same (
Figure 7A and
Table 3) despite some differences. Given that G-CSF enhances the behavioral effects of cocaine [
11] and enhances dopamine release from the VTA [
12] one might have suspected that the effects of G-CSF and cocaine on protein expression in the VTA would have been additive. Comparisons of levels of proteins relative to Saline revealed only 42 proteins in which Saline <G-CSF <Cocaine <G-CSF + Cocaine and 107 in which Saline > G-CSF > Cocaine > G-CSF + Cocaine (
Table S10). This raises the possibility that the behavioral and physiological responses potentiated by G-CSF may be owing in part to this smaller subset of proteins, or, more likely, that the changes induced by G-CSF are complex and dependent on the function and response of multiple brain regions. Further examination of these clusters of regulated proteins will be important for understanding interactions between G-CSF and cocaine.
While these results have provided new and interesting findings related to the effects of G-CSF and cocaine on proteomic expression in the midbrain, there are important caveats to their interpretation. This study was designed as a discovery analysis to identify G-CSF and cocaine interactions in a 2 × 2 design, and while this allowed us to investigate effects and interactions it lead to a study with low power in terms of sample size (N = 4/group). While we were able to perform successful Western blot validation of several regulated targets (
Figure 8) the low sample size and decision not to correct
p values leads to a high likelihood that some of the reported changes are indeed false positives. Additionally, while the use of network and pathway analyses (IPA, GO, Enrichr) are very useful for the identification of potentially regulated pathways, it is important to note that none of these software packages are built on a comprehensive review of the entire scientific knowledge base, but rather large cross-sections of data that are available to be mined. Additionally, most of these software packages pool data across tissues to increase statistical power in the analyses. While this has utility, it is important to note that regulation of intracellular pathways in other tissues, or even in other brain regions, is likely to be different from that seen in the VTA and has the potential to lead to spurious conclusions.
In sum, we have identified G-CSF as a neuroimmune factor that significantly influences the behavioral and neuronal response to cocaine [
11]. While this initial study established the possibility that G-CSF may be a translationally-relevant target for the treatment of cocaine abuse, there remains much to be done to establish its mechanism of action in the brain. Here we present an unbiased proteomic analysis of the VTA animals treated with G-CSF, cocaine, or both. This study identified key intracellular signaling pathways that are altered by systemic G-CSF treatment and lays the groundwork for future mechanistic studies into the effects of G-CSF in brain reward structures.