Healthy adult male Sprague-Dawley rats (weight, about 250 g; Animal Center of Chongqing Medical University) were used in the present study. All the rats were individually housed in standard laboratory conditions (55 ± 5% relative humidity, 12/12 h light/dark cycle, 21–22°C) with ad libitum feeding. The study protocol was approved by the local Ethics Committee (2017013). All animals were treated according to the National Institutes of Health protocols for the use and care of laboratory animals.

As previously described (Tang et al., 2019), the 8-weeks CMS protocol was employed to build the rat model. Following exposure to the CMS, the stressed rats were divided into the three groups: 1) Dep-Sus group [assessed by sucrose preference (SP) test and forced swimming (FS) test]; 2) Anx-Sus group [assessed by elevated plus-maze (EPM) test]; and 3) Insus group. Additional non-handled rats acted as the control (Ctrl) group. For a more detailed description, please refer to our previous study (Tang et al., 2019).

After the behavioral assessment, the animals were anesthetized and decapitated and their whole brains were carefully removed on ice. The pituitary tissue was isolated from the rat brain and frozen rapidly in liquid nitrogen and then stored at −80°C in a refrigerator prior to use. For protein extraction, a sample of the pituitary of each animal was added to an SDT buffer composed of 4% SDS, 0.1 M dithiothreitol, 0.1 M Tris–HCl, pH 8.0, and protease inhibitors. The tissues were homogenized and lyzed, the extracted proteins were boiled for 5 min. After centrifugation at 4°C and 40,000 × g for 15 min, the supernatants were collected and the protein concentrations were quantified using Pierce bicinchoninic acid assay kit.

Following our previously described procedure (Gong et al., 2021), the protein samples were in parallel digested using filter-aided sample preparation (FASP). In this method, an ultrafiltration filter (10 kD cutoff) was used for effective digestion. In brief, UA buffer (8 M urea, 0.15 M Tris-HCl, pH 8.0) was added to each sample. The sample was transferred to an ultrafiltration centrifuge tube and then centrifuged, and washed again with UA buffer. Subsequently, 0.05 M iodoacetamide in UA buffer was added to the filter. The protein mixture was incubated and alkylated for 30 min at room temperature in the dark. The filter unit was centrifuged and then washed twice with UA buffer. Finally, trypsin solution was added and digested at 37°C overnight. The resulting peptides were collected as a filtrate and then dried in a Speed Vac.

The tryptic peptides were labeled with eight-plex iTRAQ reagents according to the protocol of the manufacturer. The reagents 113–121 were used to label the eight samples from the three stressed and the Ctrl groups, as depicted in Figure 1A. Each used sample was obtained from 2 to 3 rats in each group (Lenselink et al., 2015). Subsequently, the eight labeled samples were pooled and preliminarily separated using high-pH RPLC. Briefly, the peptides were dissolved with buffer A (5% acetonitrile, 0.01 M ammonium formate, pH 10.0) and fractionated through linear elution in a gradient of 5–38% buffer B (90% acetonitrile (ACN), 0.01 M ammonium formate, pH 10.0) for 80 min at 300 μL/min. A total of sixteen fractions were collected, desalted and dried for the subsequent LC-MS/MS analysis.

The peptides in each fraction were re-dissolved in 0.1% formic acid, and delivered into Thermo Scientific Easy-nLC 1200 system coupled with a nanoViper C18 trap column (3 μm, 100 Å). The peptide mixtures were trapped and then desalted using 100% solvent A (0.1% formic acid). Afterward, the peptides were eluted with 8–38% solvent B (80% ACN/0.1% formic acid) for 50 min, and separated with an analytical column (50 μm × 150 mm, 3 μm-C18 100 Å). Q-Exactive Orbitrap mass spectrometer equipped with a Nano Flex ion source (ThermoFisher) was used for the MS analysis (interface heater temperature, 275°C; ion spray voltage, 1.9 kV). The tandem MS data were acquired through the use of a data-dependent acquisition mode along with full MS scans. The acquisition range was 350–1,200 m/z for the MS1 and 110–1,200 m/z for the MS2. For the information acquisition, survey scans were acquired in 250 ms and up to 14 product ion scans (50 ms) were collected. Those MS spectra along with charge state 2–4 were selected and subjected to fragmentation using higher-energy collision dissociation, and dynamic exclusion for selected precursor ions was set to 25 s.

Raw files were processed and searched using the Sequest HT search engine embedded into Proteome Discoverer software 2.1 (ThermoFisher) against the UniProt Rat database. The following search parameters were set: monoisotopic mass values, fragment mass tolerance at 0.05 Da and precursor mass tolerance ± 10 ppm, trypsin as the enzyme, and allowing up to 2 missed cleavages. Fixed modifications were defined as iTRAQ labeling and carbamidomethylation of Cys; Oxidation on Met, acetylation on protein N-term, deamidation on Asn and Gln, and Pyro-Glu were specified as a variable modification. The decoy database pattern was set as the reversed version of the target database. All reported data were based on 99% confidence for peptide identification as determined by a false discovery rate (FDR) of lower than 1%. Relative ratios of identified peptides among labeled samples were computed using relative peak intensities of released iTRAQ reporter ions in each of the MS/MS spectra, and introduced into Excel spreadsheet for manual treatment. Then, the ratios of all identified proteins were analyzed via a two-tailed Student’s t-test. Those proteins with 1.2-fold expression alterations and p-values lower than 0.05 could be considered as significantly different. The raw data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the iProX partner repository with the dataset identifier PXD025429 (Ma et al., 2019).

GO analyses including biological processes (GO-BP), molecular functions (GO-MF), and cellular components (GO-CC) were conducted through the use of the OmicsBean tool (http://www.omicsbean.cn/). KEGG (http://www.genome.jp/kegg/) was used to identify the significant pathways with p-values of lower than 0.1 following the previously described procedure (Yu et al., 2017). Moreover, the STRING database and Cytoscape were used to construct PPI networks following the previously reported protocol (Gong et al., 2021).

Following the iTRAQ-based proteomics experiment, extraction and digestion of the pituitary proteins were performed. The resulting peptides were analyzed using the Q-Exactive Orbitrap mass spectrometer. A normalized collision energy of 28 was used for the fragmentation of the peptides, and the resulting fragments were analyzed at a resolution of 35,000. The acquired raw data were analyzed via the Proteome Discoverer tool. The MS data were further processed using the analysis software Skyline 19.1 (ThermoFisher). The statistical analysis of the data were performed using Student’s t tests of SPSS software. The data were presented as means ± standard error (SE). The difference was considered to be statistically significant when p-values lower than 0.05.

In the present work, our used pituitary tissue samples were from the identical batch of the stressed animals in our recently published paper (Tang et al., 2019). Briefly, the stress-induced depressive-like behavior including anhedonia and behavioral despair were firstly assessed through the use of the SP and FS tests. Meanwhile, we also utilized the EPM test for indexing the anxious-like symptom. Based on these testing data, a subset of the Dep-Sus, Anx-Sus, Insus, and Ctrl groups was finally obtained. Overall, these results indicated that we could effectively utilized the CMS model to investigate the neurobiological processes associated with the resistance and vulnerability of stress-related anxious or depressive disorders.

Next, we investigated the effects of CMS on the expression of the rat pituitary proteins through the use of iTRAQ-based quantitative proteomics analyses (Figure 1A). In this experiment five animals per group were used, and the pituitary proteins from 2 to 3 rats were equally pooled for each sample (Lenselink et al., 2015). Matching to the UniProt database, within the Ctrl, Dep-Sus, Anx-Sus and Insus groups, totally 3,601 non-redundant proteins were identified and quantified based on the FDR lower than 0.01. The iTRAQ-based protein expressions that changed greater than 1.2-fold and p-values lower than 0.05 versus the values for the Ctrl group were deemed to be significantly different. Overall, 197 proteins were found to exhibit a significant differential expression in the three groups (Supplementary Table S1). Here, the proteome profile of the pituitary was contrasted with that of the hypothalamus from our previous work (Gong et al., 2021) (Supplementary Figure S1). Despite the profile of hypothalamus and pituitary was similar based on a 60–70% overlap of the total quantified proteins, the differential protein sets in each area exhibited considerably divergent. This suggested that there were different proteome responses to stress in these two areas.

The pituitary site-specific proteome signature of the CMS-exposed rats unraveled 37 downregulated and 32 upregulated proteins in the Dep-Sus group, 44 downregulated and 26 upregulated proteins in the Anx-Sus group, and 44 downregulated and 68 upregulated in the Insus group (Figure 1B). In the two stress-susceptible cohorts, 30 proteins were seen to be similarly deregulated, potentially representing the commonality of stress-induced anxiety and depression. Among the susceptible and the insusceptible groups, 27 similarly deregulated proteins were seen and might sever as a consequence of stress exposure (Figure 2A). To sum up, as many as 78% of these deregulated proteins were uniquely connected with the three phenotypes, which demonstrated that the three stressed cohorts had specific protein expression disturbances as a response of stress. Further, based on the unsupervised hierarchical clustering analysis, the expression profile of the 197 deregulated proteins were divided into three different units, to some extent suggesting the three specific CMS responses (Figure 2B).

We carried out GO classification and pathway enrichment of the deregulated proteins through the use of the OmicsBean software, for a better understanding of the significant protein functions and biological pathways correlated with the three behavioral phenotypes. The 69 deregulated proteins in the Dep-Sus group were subjected to enrichment analyses of the GO and KEGG pathways. Total 377, 85, 82, and 9 terms in the GO-BP, GO-CC, GO-MF, and KEGG pathways were significantly overrepresented (Supplementary Table S2). The ten top enriched GO terms are shown in Figure 3A. According to the GO-BP annotations, many deregulated proteins were associated with acute-phase and inflammatory responses, tRNA modification and processing, interferon-alpha, type 1 interferon and protein secretion and regulation. The GO-CC annotations showed that these differential proteins were mainly located in blood microparticle, extracellular region and organelle, and membrane-bounded organelle and vesicle. According to the GO-MF annotations, most of proteins were involved in enzyme inhibitor, peptidase inhibitor and regulator activity, and RNA binding. In pathway enrichment analyses, the deregulated proteins were mainly related to complement and coagulation cascades, RNA transport, mRNA surveillance pathway, synthesis and degradation of ketone bodies, metabolism, apoptosis and SNARE interactions in vesicular transport (Figure 3B).

At the same time, enrichments of GO annotations and KEGG pathways of the 70 deregulated proteins in the Anx-Sus group were carried out. There were 388 GO-BP, 84 GO-CC, 82 GO-MF, and 9 KEGG pathway terms overrepresented. The top 10 enriched GO terms are indicated in Figure 3C. The GO-BP annotations displayed that the majority of proteins were associated with coagulation, hemostasis, and amino acid and glutathione metabolic processes. According to the GO-CC annotations, the differential proteins were mainly found in membrane-bounded and intracellular organelle, protein, supraspliceosomal and macromolecular complex, and nucleoplasm and cytoplasm parts. The GO-MF annotations indicated that most proteins were involved in enzyme activity, protein and thyroid hormone receptor binding. According to pathway enrichment analyses, the deregulated proteins were primarily implicated in complement and coagulation cascades, metabolism and biosynthesis, cytosolic DNA-sensing pathway, and endocytosis (Figure 3D).

Afterwards, GO annotation and KEGG pathway enrichments of the 115 deregulated proteins in the Insus group were also conducted. There were 516 GO-BP, 116 GO-CC, 104 GO-MF, and 10 KEGG pathway terms overrepresented. The top ten enriched GO terms are shown in Figure 3E. According to the GO-BP annotations, most of the differential proteins were involved in vesicle-mediated, cytosolic and endosomal transport and regulation, localization and metabolic process, and GO-CC category analysis showed that the majority of the deregulated proteins located in the cytoplasmic and intracellular parts, endomembrane system, organelle and endosome. The GO-MF annotations predicted that most of the proteins were engaged in vitamin, protein and ubiquitin binding, and enzyme and peptidase regulator, SNAP receptor activities. The pathway enrichment analyses uncovered that the deregulated proteins were mainly enriched in SNARE interactions in vesicular transport, endocytosis, ECM-receptor interaction, focal adhesion, metabolism and biosynthesis, and GABAergic synapse (Figure 3F).

Interestingly, of these significantly-enriched KEGG pathways, there were one shared terms among the three cohorts (Figure 3G). Meanwhile, we could see the two common pathways between the two susceptible cohorts. Importantly, the 6, 6 and 7 pathways were seen to be uniquely related to the Dep-Sus, Anx-Sus and Insus groups, respectively, potentially suggesting the three different neurobiological response to the identical CMS.

Furthermore, we also focused on the proteome-inferred PPI networks in the Dep-Sus, Anx-Sus and Insus groups, as shown in Figures 4A–C. The PPI network maps of the three stressed groups were built through the use of the deregulated proteins correlated with the significant pathways. 29, 36, and 71 deregulated proteins were identified to be several important factors based on the unified conceptual framework of the three networks from the Dep-Sus, Anx-Sus and Insus groups, respectively. As expected, these networks unraveled close relationships between the deregulated proteins and the significantly enriched pathways, thereby furnishing a useful and valuable interactome unit connected with the three different behavioral phenotypes.

In this work, the PRM technique was used to further independently validate nineteen abnormally-expressed proteins of interest involved in the significant biological functions and pathways. On the whole, the PRM data mirrored the iTRAQ results (Supplementary Figure S2). As illustrated in other proteomics work (Abdi et al., 2006; Cheng et al., 2011; Xu et al., 2012; Wu et al., 2019), some discrepancies existed between the iTRAQ and PRM data. Except as the assay difference of these two approaches, another probable reason was the additional mixing step in the iTRAQ experiment (Xu et al., 2012; Wu et al., 2019). Compared with the Ctrl group, the expressions of Oxct1, Ppp1cb, Dock1, and Coq3 were significantly down-regulated while Sec24c was up-regulated in the Dep-Sus group; the expressions of Gapdh was significantly down-regulated whereas Lama1, Glb1, Sccpdh, and Renbp were up-regulated in the Anx-Sus group; the expressions of Sephs1, Spp1, and Srm were significantly down-regulated whereas Nup188 and Prodh1 were up-regulated in the Insus group (Figure 4D). In addition, the expression level of Ifih1 was displayed to be significantly reduced in both the Dep-Sus and Anx-Sus groups as compared to the Ctrl group. The reduced level of Vamp7 in both the Dep-Sus and Insus groups, and the elevated level of Arfgap3 and Ubqln4 in the three stressed groups were observed as contrasted with the Ctrl group.