Reduced Vrk2 expression is associated with higher risk of depression in humans and mediates depressive-like behaviors in mice

In 23andMe sample (75,607 cases and 231,747 controls) [2], the definition of depression was based on responses to web-based surveys, in which case individuals were required to self-report as having received a clinical diagnosis or treatment for depression. Due to restrictions on data sharing policy, the 23andMe GWAS did not publicly release genome-wide summary statistics but only provided the summary statistics of SNPs spanning the VRK2 region (chr2:57917222–58482646 (hg19)) in the Supplementary Materials of their GWAS [2].

In UK Biobank sample (127,552 cases and 233,763 controls) [15], broad definition of depression was utilized, in which cases were required to answer “yes” to the questions “Have you ever seen a general practitioner for nerves, anxiety, tension or depression?” or “Have you ever seen a psychiatrist for nerves, anxiety, tension or depression?” Individuals with self-reported bipolar disorder, schizophrenia, or personality disorder were excluded.

In PGC2 samples (43,204 cases and 95,680 controls) [4], a clinical diagnosis of lifetime major depressive disorder was applied to define cases, and most controls randomly selected from local populations were screened for the absence of lifetime major depressive disorder.

GWAS meta-analysis summary statistics of UK Biobank and PGC2 samples were publicly released [3]. We used PLINK v1.9 to combine the summary statistics (e.g., effect allele, odds ratio (OR), standard error (SE), P value) of the SNPs spanning the VRK2 region (chr2:57917222–58482646 (hg19)) across 23andMe, UK Biobank, and PGC2 samples and conduct the meta-analysis [16].

The CONVERGE Consortium has conducted a depression GWAS in Han Chinese individuals, which included 5303 cases and 5337 non-psychiatric controls [8]. As described in the original study, cases were diagnosed using the Composite International Diagnostic Interview (WHO lifetime version 2.1; Chinese version) according to the DSM-IV criteria. Controls were recruited either from local communities or from the patients who underwent minor surgical procedures at the general hospitals.

The regional association results of VRK2 SNPs were plotted using LocusZoom [17]. Haploview (version 4.1) [18] was used to examine LD between SNPs using the r² algorithm based on the SNP data from the 1000 Genomes Project [19]. We applied ENCODE [20] and HaploReg v4.2 [21] for SNP functional prediction analyses, which is primarily based on open-chromatin peaks defined by DNase I hypersensitivity and ChIP-Seq peaks of histone modifications such as H3K4me1 in brain hippocampal tissues and multiple types of cells (SK-N-SH, bipolar neurons originated from GM23338, neural progenitor cells differentiated from H9). Detailed information about the assays and methods in ENCODE project can be found online. We also used a web-based tool GWAVA (genome-wide annotation of variants), which aims to prioritize regulatory elements according to evolutionary conservation and GC-content as well as experimentally epigenetic features in ENCODE/GENCODE, to further predict whether the functional elements overlapped with the tested SNPs [22].

The eQTL results in the CommonMind datasets (209 schizophrenia patients, 52 affective disorder cases, and 206 controls) were examined [23]. In brief, the gene expression was measured using polyA⁺ RNA-sequencing, followed by normalization via adjusting for 20 surrogate variables (SVs), RNA integrity number (RIN), RIN2, diagnosis, sex, institution, PMI, clustered LIB, and age of disease onset. The eQTL analysis was performed according to the formula: adjusted gene expression ~ SNP + ancestry vectors + diagnosis.

We also obtained RiboZero RNA-seq eQTL results in the postmortem caudate nucleus tissues of 444 individuals (154 schizophrenia patients, 44 bipolar patients, and 246 controls) from Lieber Institute for Brain Development [24]. The eQTL analyses were performed using Matrix eQTL with log₂ transformed RPKM adjusted for diagnosis, sex, SNP PCs, and expression PCs.

A third Braineac eQTL dataset comprising of independent individuals of European ancestry was also utilized [25]. In this dataset, the UK Brain Expression Consortium (UKBEC) has generated genotype (using Illumina Omni1-quad and Immunochip arrays) and expression data (using Affymetrix Human Exon 1.0 ST arrays) for brain tissues from 134 control individuals of European descent, and we retrieved the eQTL results in the frontal cortex for the present study. As described in the original study, the eQTL analysis was conducted using Matrix eQTL for each normalized expression profile against SNP.

Jansen et al. [26] previously examined mRNA expression in the peripheral venous blood of remitted depression (present in lifetime but not at the current interview, n = 635) or current depression (within the 6 months prior to interview, n = 882) and control subjects with no lifetime history of depression or anxiety disorders (n = 331), using Affymetrix U219 arrays and the GeneTitan System. For differential expression analysis between depression patients and controls, P value and effect size were computed for the associations between depression and gene expression based on a linear model consisting of empirical covariates.

Knockdown of mouse Vrk2 was achieved by short hairpin RNA (shRNA). Three shRNA sequences targeting Vrk2 were designed: 5’-GATGCAAGACATGTCATAA-3’ (shRNA1), 5’-CCGTCCTACAACTTGGCAT-3’ (shRNA4), and 5’-GGGCCATAATGGGACAATA-3’ (shRNA5), and a non-specific target sequence (5’-GATTTGCTGTTCGCCCAAG-3’) on mouse genome was employed as negative control. The shRNA sequences were constructed into pAAV-CAG-tdTomato (Addgene, #59,462). To overexpress Vrk2, the full-length murine Vrk2 coding sequence with the C-terminal Flag-tag was constructed into the pAAV-CAG-tdTomato vector (Addgene, #59,462), and the empty vector plasmid was utilized as the control. Sanger sequencing was used to ensure the correct construction.

Recombinant plasmids were co-transfected with packaging plasmids (AAV-DJ and pHelper) into HEK293T cells by linear polyethylenimine 40 kDa (Polysciences) [27]. Seventy-two hours post-transfection, cells were collected, resuspended (0.15 M NaCl; 20 mM Tris–HCl) and subjected to 4 freeze–thaw cycles in liquid nitrogen and 37 °C water bath to release AAV virus. Benzonase treatment (Sigma, #E1014) was used to remove extraviral DNA contamination at 37 °C for 30 min and the reaction mixture was centrifuged at 8000 × g for 90 min at 4 °C. The supernatant containing AAV virus was collected and then added to 15%, 25%, 40%, 54% iodixanol (Sigma-Aldrich) step gradient and centrifuged at 60,000 × g for 3 h at 4 °C. Finally, 100–200 μL AAV virus were purified by Amicon ultrafiltration tube (EMD Millipore).

Wild-type C57BL/6 J male mice and Vrk2-knockout (KO) mice received stereotaxic injection at 7–8 weeks of age. The mice were fixed on the brain stereotactic apparatus horizontally (RWD Life Science Co., Ltd., Shenzhen, China) and anesthetized with isoflurane gas. A longitudinal incision was made to expose the skull surface. Set the anterior fontanelle as the origin and drill holes according to the ventral hippocampal coordinates (AP – 3.08 mm; ML ± 3.20 mm; DV − 4.0 mm/ − 3.5 mm) with a skull drill (RWD Life Science Co., Ltd., Shenzhen, China). A micro-syringe (RWD Life Science Co., Ltd., Shenzhen, China) was positioned above the drilled hole according to the coordinates and was used to take in the AAV virus (titer = 5 × 10¹² IU/mL), then the injection needle was slowly lowered through the hole into the target brain area. Five minutes later, a syringe pump controller was used (RWD Life Science Co., Ltd., Shenzhen, China) to perform a bilateral injection of 1 µl per side at a speed of 2 nL/s. Ten minutes after injection, the needle was slowly withdrawn to prevent the virus from leaking out. The scalp was sutured and the mice were then returned to the cage after awakening.

Total RNAs from fresh hippocampal tissues were extracted with TRIzol Reagent (Life Technologies). Complementary DNA (cDNA) was synthesized from 1 μg of RNA using the RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific). The mRNA expression of Vrk2 was quantified through real-time quantitative PCR (RT-qPCR) by using FastStart Universal SYBR Green Master (Roche) and the reaction was performed on the CFX96 real-time PCR system (Bio-Rad, USA). Gapdh was used as an internal control and the relative gene expression values were presented as the means of 2^{^–ΔΔCt}. Primer sequences for RT-qPCR are as follows: Gapdh (Forward: 5’-ATGGTGAAGGTCGGTGTGAA-3’, Reverse: 5’-CAATCTCCACTTTGCCACTGC-3’) and Vrk2 (Forward: 5’- GTTCATGGTGATATAAAAGCCGCA-3’, Reverse: 5’-TGGCCCTTTCTGGGATCTTC-3’).

The hippocampus tissues were homogenized by a tissue homogenizer in RIPA lysate with protease and phosphatase inhibitors. The mixture was centrifuged and the supernatant was subjected to protein quantification using a BCA protein assay kit (Thermo Scientific, USA). After being denatured at 95 °C with a fivefold loading buffer, protein samples were separated on a 10% SDS-PAGE gel and transferred to polyvinylidene difluoride (PVDF) membranes (EMD Millipore, USA). The transferred membranes were blocked with 5% BSA and then incubated with the primary antibody overnight at 4 °C. After that, the secondary antibody was incubated for 60 min at room temperature with a peroxidase-conjugated secondary antibody. Finally, the blots were reacted with the Tanon ECL Chemiluminescent reagent (#180-5001W) and the images were captured by Tanon 5200 imaging system. The primary antibodies were as follows: anti-Vrk2 (#12946–1-AP, Proteintech, USA) and anti-Gapdh (#60004–1-Ig, Proteintech, USA).

The open-field box (40 cm L × 40 cm W × 40 cm H) floor was virtually divided into 9 equal squares, with the central square considered as the central area and the other 8 outer squares designated as the peripheral area. Each mouse was placed in the same corner of the open field box, facing the wall, and permitted to explore the environment freely for 5 min. Their activity tracks were recorded by a video camera. The time and distance in the central area were analyzed by SuperMaze software (RWD Life Science Co., Ltd., Shenzhen, China) to evaluate their anxiety. At the end of each experiment, the open-field box was cleaned with 75% alcohol.

The murine behavioral data were analyzed using GraphPad Prism software and were presented as mean ± SD. For normally distributed data evaluated by Shapiro–Wilk test, two-tailed t-test was used to assess the difference between two groups, and one-way analysis of variance (ANOVA) was performed for more than two groups. For all the behavioral analyses, we only used male mice, as according to previous studies, female mice are more susceptible to individual variances brought by the estrous cycle, which might lead to unreliable results [28].

The elevated plus maze (EPM) is composed of a pair of open arms (33 cm L × 8 cm W) and a pair of closed arms (30 cm L × 6 cm W). The mice were put into the central area facing one of the open arms. SuperMaze software (RWD Life Science Co., Ltd., Shenzhen, China) was used to record the number and time of entering the open arm and the closed arm within 6 min [29, 30]. Their activity tracks were recorded by a video camera. The percentage of time and distance that the mice entered the open arm were analyzed. At the end of each experiment, the elevated plus maze was cleaned with 75% alcohol.

The general motor function was evaluated on the rotarod test. Mice were placed on the rotarod (RWD Life Science Co., Ltd., Shenzhen, China) and trained to walk on the rotarod at a constant speed of 10 rpm for 5 min. After 2 days of training, the motor capacity of mice was tested using the accelerating task (from 4 to 40 rpm in 5 min). The experiment ended when the mice fell or gripped and spun around for one complete revolution. Each mouse underwent three trails with an intertrial interval of 60 min. The average latency time to fall for the three trials was used as a statistical indicator.

TST was used to detect desperation behavior in mice [31]. The mouse tail tip was taped about 1 cm on the hook of the suspension tail test chamber (20 cm L × 20 cm W × 32 cm H). The time of each activity state was recorded using SuperTST software (RWD Life Science Co., Ltd., Shenzhen, China). According to previous studies [32] and the analytical software, the behavior states of mice were categorized according to activity levels: (1) immobility: the mouse was suspended without any activity behavior, (2) swinging: when a mouse repeatedly moved its paw or significantly moved its body in a vertical position, and (3) curling: the mice’s bodies make active twisting movements. The experiment process for a total of 6 min, and the time duration of each behavior state in the last 4 min was analyzed.

SPT is used to detect the absence of pleasure in experimental animals, which is one of the main indicators of depression [33]. First, the mice are housed in a single cage, and two water bottles, one with clean water and the other with 1% sucrose solution, are placed on the cage and acclimated for 48 h, with the position of the two water bottles changing every 24 h. Mice were deprived of drinking water for 24 h at the end of the acclimatization period. During the test, the water bottle containing clean water and the 1% sucrose solution remained on the mouse cage, and the test lasted a total of 12 h, with the positions of the two water bottles changing every 6 h, and the amount of clean water and the 1% sucrose solution consumed was recorded. Sucrose preference % = 1% sucrose solution consumption / (1% sucrose solution consumption + clean water consumption) × 100%.

Individual neurons’ local dendritic morphology was recreated by sparse and highly visible tagging of neurons with fluorescent proteins carried by adenovirus (Brain Case, #BC-SL001) to precisely clarify the quantity and form of their dendritic spines. By stereotaxic brain injection, the adenovirus was administered to the ventral hippocampus (AP – 3.08 mm; ML ± 3.20 mm; DV − 3.8 mm) of 7–8-week-old mice (Vrk2^+/+ and Vrk2^−/−) in a volume of 200 nL on each side. Four weeks after surgery, mice were perfused with PBS and 4% paraformaldehyde, and undamaged brain tissue was extracted and fixed overnight in 4% paraformaldehyde at 4 °C, followed by dehydration in 30% sucrose solution at 4 °C. The dehydrated brain tissues were embedded with OCT and placed on a frozen section machine to create 50-μm coronal brain slices. The slices were sealed with anti-quenching sealer and imaged using a laser confocal microscope at high magnification. Secondary or tertiary dendrites in the hippocampal pyramidal neurons were used as photographic objects, and statistical analysis was carried out using NeuronStudio [34] and Image-J software. Dendritic spines can be divided into mushroom-shaped spines (head-neck ratio ≥ 1.5), thin and long spines (head-neck ratio < 1.5, length/head diameter ≥ 2), and short and thick spines (head–neck ratio < 1.5, length/head diameter < 2), and the number of spines in a dendrite of 10 μm length was used as its density, and the density of different morphological dendritic spines was analyzed and counted. Before statistical analysis, the data were evaluated whether they were normally distributed by Shapiro–Wilk test, and two-tailed t-test was used to compare the total number of dendritic spines, and the two-way ANOVA was used to determine the various morphologies. All experimental data are presented as mean ± SD.

Hippocampus tissues (n = 3, each group, Vrk2^+/+ and Vrk2^−/−) were isolated and the homogenized tissues were subjected to protein extraction by SDT (4% SDS, 100 mM Tris–HCl, 1 mM DTT, pH 7.6) buffer. The protein concentrations were determined by the BCA (Bio-Rad, USA) method. A suitable amount of protein from each sample was trypsinized using the filter-aided proteome preparation (FASP) method [35], and the peptides were desalted using a C18 Cartridge, vacuum centrifuged, and dissolved in 40 l of 0.1% formic acid solution. The peptide content was determined using UV spectroscopy at 280 nm. Each peptide solution was vacuum-lyophilized and then enriched with the High-SelectTM Fe-NTA Phosphopeptides Enrichment Kit (Thermo Scientific). The enriched phosphorylated peptides were concentrated under vacuum and diluted in 20 L of 0.1% formic acid solution before being analyzed by mass spectrometry. The LC–MS/MS analysis was carried out using a Bruker TimsTOF Pro mass spectrometer. The MS raw data for each sample were combined and searched using the MaxQuant 1.6.14 software for identification and quantitation analysis. The differentially phosphorylated proteins were selected on the basis of the combination of |log₂(fold change)|≥ 1 and P ≤ 0.05. The differentially phosphorylated proteins then underwent enrichment analyses of Kyoto Encyclopedia of Genes and Genomes (KEGG), Gene Ontology (GO), Reactome pathway [36], and Monarch Human Phenotype Ontology (HPO) [37] using the STRING dataset (version 11.5) [38].

We retrieved the summary statistics of VRK2 SNPs in 23andMe, UK Biobank, and PGC2 samples [2, 3], and 256 SNPs region were available in all three GWAS datasets (a total of 246,363 cases and 561,190 controls). An overview about the strategy and logic of choosing SNPs within the VRK2 region is provided in Figure S1 (Additional file 2). Meta-analysis found that 71 SNPs within Chr2:57,917,222–58,482,646 (hg19) region showed genome-wide significant associations with risk of depression (P < 5.00⨯10^–8, Fig. 1). Among these SNPs, the lead SNP was rs1568452 in the upstream of VRK2 (P = 8.12⨯10^–12; Fig. 1A; Additional file 1: Table S1); however, the previously reported SNP rs1518395 did not achieve genome-wide significance in this meta-analysis (P = 1.23⨯10^–6).

To test whether these 71 significant SNPs within Chr2:57,917,222–58,482,646 (hg19) region also showed evidence of associations with depression in Han Chinese population, we then retrieved them from a Han Chinese depression GWAS (5303 cases and 5337 controls) [8], and 47 SNPs (including rs1568452) also showed nominal associations (P < 0.05; Fig. 1A; Additional file 1: Table S1) and were remained for the following population genetic analyses and functional predictions. We constructed the LD map of these 47 SNPs in Europeans and Han Chinese populations using genotype data from 1000 Genomes Project [19]. The LD structure was similar between Europeans and Han Chinese (Additional file 2: Figure S2). These 47 SNPs constituted up to 3 LD blocks in Europeans and Han Chinese, and most of the SNPs in each block were in strong LD.

The above 47 SNPs were located in the noncoding regions spanning VRK2, and we therefore carried out the functional predictions using epigenome data (GSE96018, GSE123202, GSE100991) from ENCODE project [20] by comparing their locations with regions of open chromatin, with regions of active histone H3 lysine modifications, and with regions of ChIP-Seq data for CTCF transcription factor, we found that the genomic region spanning rs2678907 exhibited large spatial overlap with enhancer regulatory marks (e.g., H3K4me1) in multiple human cell lines (SK-N-SH, neuron, neural progenitor cells) and brain tissues (Fig. 1A), and the DNA region was demonstrated to bind CTCF and other transcription factors.

We verified the results of functional predictions using HaploReg v4.2 [21], which also utilized data primarily from ENCODE projects. In this dataset, rs2678907 was predicted to be located in the regulatory enhancer region in brain tissues (assessed by DNase I, H3K4me1 and H3K27ac marks), and ChIP-Seq experiments demonstrated that DNA regions spanning rs2678907 bound 8 proteins, including CTCF, BCLAF1, RAD21, SMC3, YY1, ZNF143, SRF, and TAL1 (Additional file 2: Figure S3). Further functional prediction using GWAVA [22] also indicated that rs2678907 was likely a functional SNP, as the prediction scores was higher than or near 0.5 (prediction scores from three different versions of the classifier (Region score, TSS score, Unmatched score) range 0–1, and higher scores suggest a greater likelihood of functionality) (Additional file 2: Figure S4).

Rs2678907 was located 1.3 kb 5’ upstream of VRK2, and showed genome-wide significant association with risk of depression in our meta-analysis of European samples (P = 7.40⨯10^–10), and also exhibited nominal associations with in the Han Chinese GWAS (P = 4.70⨯10^–3) (Fig. 1B; Additional file 1: Table S1), further strengthening its link with the illness. According to the genotype data in 1000 Genomes Project, rs2678907 was in moderate to strong LD with rs1568452 (r² > 0.7), but both SNPs showed weak LD with rs1518395 (r² < 0.4) (Fig. 1C).

We then tested if the DNA fragment containing rs2678907 showed enhancer activity. The DNA fragments in the 5’ upstream of VRK2 gene containing either the G allele or the A allele at rs2678907 were constructed into pGL3-promoter plasmid and the relative enhancer activities were measured by dual-luciferase reporter assay in human SK-N-SH, U251, and HEK293T cell lines. The enhancer activity containing the A allele was significantly higher than that of the G allele in SK-N-SH (P = 0.0001), U251 (P = 0.0098), and HEK293T (P = 0.0047) cell lines (Fig. 2A). These data indicated that the depression risk G-allele at rs2678907 decreased the DNA enhancer activity.

To understand whether rs2678907 was associated with mRNA expression of nearby genes, we examined the eQTL data in human brain tissues from several independent datasets, and notably, the depression risk G-allele at rs2678907 consistently predicted lower mRNA level of VRK2 in three independent samples (P = 3.98⨯10^–3 in 467 individuals of CommonMind dataset [23]; P = 7.02⨯10^–3 in 444 individuals of Caudate eQTL dataset [24]; P = 0.014 in 134 individuals of Braineac dataset [25]; Fig. 2B). The rs2678907 G-allele also predicted reduced mRNA expression of FANCL in one dataset (P = 2.89⨯10^–4 in 467 individuals of CommonMind dataset [23]; Additional file 2: Figure S5). Therefore, reduced mRNA levels of VRK2 and FANCL were likely associated with an increased genetic risk of depression.

We then investigated the mRNA expression of VRK2 and FANCL in patients with depression and healthy controls. Intriguingly, we found that expression of VRK2 was significantly reduced in the blood of either 882 current depression patients or 635 remitted depression patients compared with 331 healthy controls (P = 0.00888 for current depression versus control; P = 0.023 for remitted depression versus control; P = 0.0074 for current + remitted depression versus control) [26]. The diagnostic analysis of VRK2 mRNA expression is consistent with the results of eQTL analysis and further confirms the involvement of VRK2 in depression. However, FANCL did not alter significantly between depression patients and healthy controls (P = 0.199 for current depression versus control; P = 0.22 for remitted depression versus control; P = 0.17 for current + remitted depression versus control) [26] and was thus dropped out from further analyses.

We conducted behavioral analyses using Vrk2^−/− and wild-type (Vrk2^+/+) mice. Both types of mice were male littermates 8–10 weeks old, and Vrk2 expression was nonvisible in the Vrk2^−/− mice (Fig. 3A; Additional file 2: Figure S6). In our results, Vrk2^−/− mice did not show significantly higher exploratory activity in OFT than Vrk2^+/+ mice (Vrk2^+/+, 54,388 ± 10,253 mm distance; Vrk2^−/−, 52,148 ± 9558 mm distance, df = 24, t = 0.576, P = 0.570, n = 12–14; Fig. 3B), and there were no differences between experimental groups for the time spent in the center of arena (Vrk2^+/+, 14.09 ± 14.57 s in center arena; Vrk2^−/−, 8.87 ± 4.71 s in center arena, df = 24, t = 1.269, P = 0.217, n = 12–14; Fig. 3B). Rotarod test was used to determine whether deficiency of Vrk2 affects motor ability; however, no significant difference was observed (Vrk2^+/+, 202.50 ± 25.36 s in latency time; Vrk2^−/−, 201.70 ± 21.07 s in latency time, df = 24, t = 0.0867, P = 0.932, n = 12–14; Fig. 3C). In EPM, no significant differences were found in the percentage of time (Vrk2^+/+, 14.83 ± 8.70% time in open arms; Vrk2^−/−, 10.84 ± 8.81% time in open arms, df = 24, t = 1.159, P = 0.258, n = 12–14; Fig. 3D) and distance (Vrk2^+/+, 11.89 ± 6.85% distance in open arms; Vrk2^−/−, 7.16 ± 5.57% distance in open arms, df = 24, t = 1.942, P = 0.0640, n = 12–14; Fig. 3D) that the mice entered the open arms. There was no difference in sucrose preference between Vrk2^−/− and Vrk2^+/+ mice (Vrk2^+/+, 73.44 ± 19.51% sucrose preference; Vrk2^−/−, 64.91 ± 16.16% sucrose preference, df = 21, t = 1.146, P = 0.2648, n = 11–12; Fig. 3E). In TST, although no difference in the immobility time was observed between genotypic groups (Vrk2^+/+, 164.50 ± 19.89 s in immobility time; Vrk2^−/−, 171.80 ± 24.58 s in immobility time, df = 22, t = 0.790, P = 0.438, n = 11–13; Fig. 3F), the Vrk2^−/− mice exhibited significantly reduced time in curling (Vrk2^+/+, 22.53 ± 12.63 s in curling time; Vrk2^−/−, 8.28 ± 6.15 s in curling time, df = 22, t = 3.606, P = 0.0016, n = 11–13; Fig. 3F), a more active escape-oriented behavior. These results suggest that Vrk2^−/− mice might exhibit depressive-like behaviors compared with Vrk2^+/+ mice.

Subsequently, we investigated whether overexpressing Vrk2 could rescue the depressive-like behaviors in Vrk2^−/− mice. We respectively injected Vrk2 overexpressing AAV and control virus into the ventral hippocampus of Vrk2^−/− and Vrk2^+/+ mice at 8-weeks old and performed rotarod test and TST four weeks later (Fig. 3G). The chosen ventral hippocampus for microinjection is based on previous studies, in which they demonstrated that the ventral hippocampus is more connected to stress, mood, and emotion than the dorsal hippocampus [39, 40] and seems to be more closely linked to the processing of emotional information [41, 42]. We observed no change in motor ability between the experimental groups(Vrk2^+/+ + ctrl, 184.30 ± 14.96 s in latency time, n = 8; Vrk2^−/− + ctrl, 176.60 ± 16.43 s in latency time, n = 10; Vrk2^+/+ + Vrk2, 180.20 ± 20.62 s in latency time, n = 7; Vrk2^−/− + Vrk2, 189.20 ± 21.04 s in latency time, n = 10; Vrk2^+/+ + ctrl vs. Vrk2^−/− + ctrl, df = 31, t = 0.891, P = 0.380; Vrk2^−/− + ctrl vs. Vrk2^−/− + Vrk2, df = 31, t = 1.535, P = 0.135; multiple comparisons using Fisher’s LSD in one-way ANOVA; Fig. 3H); however, we found that the curling time of Vrk2^−/− mice after overexpressing Vrk2 in the ventral hippocampus was significantly enhanced (Vrk2^+/+ + ctrl, 17,51 ± 17.04 s in curling time, n = 8; Vrk2^−/− + ctrl, 4.07 ± 4.72 s in curling time, n = 10; Vrk2^+/+ + Vrk2, 12.38 ± 9.99 s in curling time, n = 7; Vrk2^−/− + Vrk2, 21.78 ± 12.95 s in curling time, n = 10; Vrk2^+/+ + ctrl vs. Vrk2^−/− + ctrl, df = 31, t = 2.394, P = 0.0229; Vrk2^−/− + ctrl vs. Vrk2^−/− + Vrk2, df = 31, t = 3.346, P = 0.0022; multiple comparisons using Fisher’s LSD in one-way ANOVA; Fig. 3I). These results imply that overexpression of Vrk2 in the ventral hippocampus of Vrk2^−/− mice may have antidepressant effect.

To further validate the behavioral phenotypes observed in the Vrk2^−/− mice, we specifically knocked down Vrk2 in the ventral hippocampus of wild-type C57BL/6 J mice using adeno-associated virus (AAV) packaged shRNA (Fig. 4A). We designed three shRNA sequences respectively targeting the Vrk2 gene and examined their knockdown efficiencies, and the shRNA1 showing the strongest suppressive effects was selected (Additional file 2: Figure S7). Four weeks after bilateral microinjection of AAV into the ventral hippocampus, the Vrk2 knockdown efficiency was verified (Fig. 4B) and then behavioral tests were carried out. In OFT, no significant difference was observed in the total distance (control, 51,464 ± 11,418 mm distance; Vrk2-KD, 45,080 ± 11,837 mm distance, df = 38, t = 1.736, P = 0.0907, n = 20–20; Fig. 4C) and central exploration time (control, 11.00 ± 8.22 s in center arena; Vrk2-KD, 8.03 ± 6.76 s in center arena, df = 38, t = 1.249, P = 0.219, n = 20–20; Fig. 4C) between shRNA-mediated Vrk2 knockdown (Vrk2-KD) and the mice injected with control shRNA AAV (control). The proportion of time (control, 16.67 ± 11.26% time in open arms; Vrk2-KD, 11.19 ± 6.72% time in open arms, df = 38, t = 1.866, P = 0.0698, n = 20–20; Fig. 4D) and distance (control, 8.67 ± 6.16% distance in open arms; Vrk2-KD, 7.17 ± 5.05% distance in open arms, df = 38, t = 0.843, P = 0.405, n = 20–20; Fig. 4D) that the mice entered the open arms had no influence in EPM, and there was also no difference in motor ability between the groups in rotarod test (control, 173.00 ± 23.49 s in latency time; Vrk2-KD, 164.10 ± 28.02 s in latency time, df = 38, t = 1.087, P = 0.284, n = 20–20; Fig. 4E). In SPT, Vrk2-KD mice showed significantly lower sucrose preference than control mice (control, 86.60 ± 11.19% sucrose preference; Vrk2-KD, 62.54 ± 18.39% sucrose preference, df = 38, t = 4.998, P < 0.0001, n = 20–20; Fig. 4F). In TST, the Vrk2-KD mice spent less time in curling behavior compared with control mice (control, 10.67 ± 10.71 s in curling time; Vrk2-KD, 3.34 ± 4.70 s in curling time, df = 38, t = 2.804, P = 0.0079, n = 20–20; Fig. 4G). These results indicate that decreased Vrk2 in the ventral hippocampus leads to depressive-like behaviors in mice and also in line with the previous study showing that activation of the ventral hippocampal projections to the nucleus accumbens projections increases sensitivity to stress while decreasing the sucrose preference [43].

The results of neurobiological and histopathological analyses have demonstrated that depression is associated with synaptic dysfunction and loss of dendritic spines in brain regions associated with emotional and cognitive regulation, such as the hippocampus [44, 45]. Dendritic spines are the postsynaptic sites of most excitatory synapses, which are the basic units of connection and information exchange between neurons and are intimately associated with emotional and cognitive functions [46, 47]. Notably, alteration in the morphogenesis of dendritic spines is one of the key pathological characteristics of depression [48]. Therefore, we evaluated the effect of Vrk2 deficiency on the density and morphology of dendritic spines in mice. The neuronal sparse labeling results of pyramidal neurons in the ventral hippocampus showed no difference in the density of total dendritic spines between Vrk2^+/+ and Vrk2^−/− mice (Vrk2^+/+, 14.030 ± 2.023 spines per 10 μm; Vrk2^−/−, 13.840 ± 1.944 spines per 10 μm, df = 75, t = 0.436, P = 0.664, two-tailed t-test; Fig. 5A). The morphological analysis revealed that compared to wild-type mice, deletion of Vrk2 induced a significant decrease in the density of mushroom spines (Vrk2^+/+, 3.868 ± 1.007 spines per 10 μm; Vrk2^−/−, 3.313 ± 0.924 spines per 10 μm, df = 225, t = 2.233, P = 0.0265, multiple comparisons using Fisher’s LSD in two-way ANOVA; Fig. 5A), the most mature type of spines that play vital roles in synaptic transmission and plasticity, and a significant increase in the density of thin spines (Vrk2^+/+, 7.348 ± 1.380 spines per 10 μm; Vrk2^−/−, 8.073 ± 1.420 spines per 10 μm, df = 225, t = 2.916, P = 0.0039, multiple comparisons using Fisher’s LSD in two-way ANOVA; Fig. 5A). Our results demonstrated that Vrk2 knockout can cause a loss of mushroom spines and an increase in thin spines in the ventral hippocampus.

To investigate the molecular mechanisms underlying the depressive-like behaviors in Vrk2^−/− mice, we performed phosphorylated 4D-label-free quantitative proteomics analyses of the ventral hippocampal proteins in Vrk2^−/− and Vrk2^+/+ mice (n = 3 each). After stringent quality control, 6658 quantifiable phosphorylated peptides were identified in 2847 phosphorylated proteins. The distribution of phosphorylation modification sites revealed that serine (S, 79.9%) and threonine (T, 14.85%) were most commonly phosphorylated. A total of 98 differentially phosphorylated peptides with an absolute fold change (FC) > 2 (|log2(FC)|> 1) at P < 0.05 between Vrk2^−/− and Vrk2^+/+ mice were identified (Additional file 1: Table S2). Among them, 56 (mapped to 50 proteins) and 42 peptides (mapped to 38 proteins) had significantly lower and higher phosphorylation levels, respectively, following the deletion of Vrk2.

For 38 proteins with highly phosphorylated peptides, KEGG pathway and GO analyses did not reveal any significant enrichment pathways or biological processes. However, for the 50 proteins with lower phosphorylated peptides, KEGG pathway analysis revealed significantly enriched signals in glutamatergic synapse. Reactome pathway analysis demonstrated that these were significantly enriched in the pathways of “transmission across chemical synapses”. Furthermore, Monarch HPO analysis revealed that those proteins were enriched in terms including “abnormal synaptic physiology,” “abnormal behavior,” and “abnormal neuron morphology” (Fig. 5B). Additionally, gene set enrichment analyses (GSEA) indicated that these proteins were strongly enriched in the GO terms “maintenance of synapse structure,” “postsynapse organization,” “regulation of synaptic plasticity,” and “structural constituent of synapse” etc. (Fig. 5B). These 50 proteins also formed dense protein–protein interaction (PPI) networks (enrichment P < 1.00⨯10⁻¹⁶, Fig. 5C), and notably, 11 proteins were located at postsynaptic density, including Pkp4, Map 1a, Map 1b, Dlg2, Dbn1, Psd3, Shank1, Grin2b, Syn1, Rims1, and Pclo. In conclusion, our findings show that the deletion of Vrk2 primarily affects synapse-associated proteins and processes.

In addition to highlighting essential pathways, phosphoproteomic analyses revealed several proteins, such as Pclo and Itpr2, whose involvement in depression has been supported by strong evidence. Pclo (encoding piccolo) has been implicated as a depression susceptibility gene in the early GWAS and follow-up replication studies [49‐51], and mice overexpressing the C2A domain of piccolo exhibited depressive-like behaviors in forced swim and tail suspension tests [52]. Itpr2 encodes the inositol 1,4,5-trisphosphate-type-2 receptor. Mice lacking Itpr2 exhibit deficiencies in astrocytic ATP release, abnormal rsfMRI functional connectivity (rsFC) in depression-related networks, and depressive-like behaviors, and which can be rescued by the administration of ATP [53, 54]. Pclo and Itpr2 phosphorylation at serine 1563 and 937, respectively, were significantly downregulated after the deletion of Vrk2 (P = 0.049 for Pclo and P = 0.010 for Itpr2, Fig. 5D). This study provides additional evidence suggesting pivotal roles of Pclo and Itpr2 in depression pathogenesis.