Changes in Dclk1 expression in response to psychotropic drugs
First, we reexamined the expression of alternatively spliced transcripts from the Dclk1 locus based on the previously reported dataset describing the effects of various psychotropic drugs on the mouse striatum [2]. The results of the analysis are shown in Fig. 1. Panel A shows a schematic representation of the main transcripts of the mouse Dclk1 gene based on the NCBI37/mm9 mouse genome release. Regions corresponding to probes from the MouseWG-6 v2 BeadChip are marked with blue or red symbols. Probes correspond to nonoverlapping transcripts and neither detected the full-length canonical Dclk1 transcript. As shown in the reanalysis of the array data presented in panel B, levels of transcripts detected by the probes were differentially affected by treatments with psychotropic drugs.
The mRNA levels measured using the first microarray probe (ILMN_1259689, presented as a red star in Fig. 1; drug P value = 4.14 × 10−16, time P-value = 9.9 × 10−15, interaction P-value = 4.2 × 10−5) indicated increased expression after the administration of mianserin (1, 2 and 4 h after the injection), risperidone (1 and 2 h) and, to a lesser extent, haloperidol (2 h) treatments (Fig. 1b). The second probe (ILMN_2434274, blue star; drug P-value = 4.68 × 10−16, time P-value = 1.2 × 10−4, interaction P-value = 2.5 × 10−7) revealed a different pattern of regulation, with an increase in mRNA abundance levels 8 h after the tranylcypromine treatment. Analysis of the array profiling results showed isoform-specific regulation of Dclk1 expression by psychotropic drugs. However, the interpretation was also confounded by ambiguous detection of the Dclk1 transcripts.
Transcriptome profiling of the effects of psychotropic drugs using RNA sequencing
We used next-generation sequencing to comprehensively examine drug-induced Dclk1 gene expression at the level of specific transcriptional units. Sequencing was performed on ribo-depleted RNA samples derived from the mouse nucleus accumbens septi (NAc) and prefrontal cortex (PFCx) 2 h after treatment with antidepressants (venlafaxine and mianserin), antipsychotics (haloperidol and risperidone), a psychostimulant (methamphetamine) or a psychotomimetic (ketamine).
We measured normalized transcript abundance levels (FPKM) for all transcripts annotated in the GRCm38.p5 genome release using the Cufflinks package. A total of 110,327 different transcripts corresponding to 45,935 annotated genes were detected in the PFCx or NAc at the threshold of a mean FPKM ≥ 0.1.
The overall differences in drug-induced gene expression between the NAc and PFCx were assessed using a one-way ANOVA for drug factor performed separately in each tissue. We found 90 transcripts regulated by the treatment only in the NAc and 246 transcripts altered in the PFC (at FDR < 0.0001). The examples of regulated genes are Map6 and Cdkn1a in the NAc, and Bhlhe40 and Fkbp5 in the PFC. We also identified 26 transcripts regulated in both the analyzed tissues above the threshold, including Homer1, Sgk1 and Fosb (Additional file 1: Figure S1).
Drug-regulated changes in transcript levels were classified by biotypes and alternatively spliced events. Notably, 78% of regulated transcripts in the NAc and 65% of the transcripts in the PFCx encoded proteins, compared with 56% of the transcripts expressed in the NAc or PFCx under basal conditions (Additional file 2: Figure S2). Among the drug-induced alterations, an enrichment of protein-coding variants was observed, although the majority of noncoding, regulated transcripts were classified as intron-retained transcripts.
We used a two-way ANOVA with the drug and tissue as factors to identify drug-regulated changes in transcript levels (arbitrary cutoff at treatment effect FDR < 0.001; Additional file 3: Figure S3). Figure 2 shows the hierarchical clustering of the top 113 transcripts, grouped into four clusters. Transcripts from the four main branches were examined for overrepresented putative transcription factor binding sites in their promoter regions using seqinspector [20].
Transcripts from the first pattern were upregulated by mianserin in both brain regions. Their expression also increased in the NAc in response to risperidone and haloperidol treatments. Examples of transcripts clustered in this group included Sgk1, Nfkbia and the long noncoding RNA Neat1. Promoters of these transcripts exhibit a significant overrepresentation of the ChIP-seq signal for several transcriptional factors, including GR (P = 9.7 × 10−6, t-test with Bonferroni’s correction, track GEO accession: GSM686976), E2F1 (P = 1.1 × 10−6, GSM881056) and NFKB1 (P = 1 × 10−5, GSM88115). Haloperidol, methamphetamine and, to a lesser extent, risperidone induced the expression of the greatest number of transcripts (pattern 2) in the NAc. The strongest induction in the PFCx was observed after methamphetamine treatment. Pattern 2 included several genes involved in the molecular control of neuronal plasticity, such as Fosb, Arc, Junb or Homer1. The promoters of these transcripts contained a different set of putative transcriptional regulator binding sites, including SRF (P = 1.2 × 10−22, GSM530190) and EGR2 (P = 4.7 × 10−12, GSM881094). The expression of the third group of transcripts (pattern 3, e.g., Celf2 and Dclk1) was induced by mianserin, with stronger effects observed in the NAc. We only identified one overrepresented potential transcriptional regulator of these genes, TBP (P = 3.5 × 10−7). The TBP binding motif was present in 18 of the 28 total genes. The Dclk1 transcript was clustered into pattern 3. Notably, at the statistical threshold used to analyze the whole transcriptome, we observed significant changes in the levels of the transcript for only one Dclk1 isoform, a noncoding variant with a retained intron (ENSMUST00000198757). Finally, the expression of transcripts from pattern 4, including Clk1 and Hes5, decreased after mianserin treatment in both brain regions. Significantly overrepresented TFB sites were not present in the upstream promoter regions of transcripts from this cluster.
The profile of the drug-specific transcripts corresponding to the Dclk1 locus was obtained by comparing the results of two high-throughput gene profiling methods-microarray (Fig. 1) and RNA-sequencing (Fig. 2).
Isoform-specific regulation of Dclk1 expression
Next, we performed a detailed analysis of the sequencing results mapped to the Dclk1 locus. Based on the assignments from Cufflinks using the GRCm38.p5 mouse genome release, sequencing reads corresponded to 12 transcripts expressed from the Dclk1 locus (at FPKM > 1), with 8 highly abundant isoforms (FPKM > 5). Three isoforms were not detected in the NAc or PFCx (ENSMUST00000198821, ENSMUST00000197870, and ENSMUST00000196745). The majority of the sequencing reads corresponded to Carp (ENSMUST00000199585), Dcl (ENSMUST00000167204) and Cpg16 (ENSMUST00000198437). Two of the transcripts assembled by Cufflinks contained retained introns and had not previously been described (Additional file 4: Figure S4). Relative changes in the abundance of the transcripts are shown in Fig. 3.
The expression of the Carp transcript in the NAc was induced in response to mianserin (log2FC = 1.78, P = 0.0002), risperidone (log2FC = 1.07, P = 0.007) and venlafaxine (log2FC = 1.01, P = 0.009). Carp expression in the PFCx was only regulated only mianserin (log2FC = 1.57, P = 0.0027). The level of the Cpg16 transcript was not significantly altered in any of the analyzed brain regions upon drug treatment. The venlafaxine treatment downregulated the expression of the Dcl isoform in the NAc (log2FC = − 0.74, P = 0.078). However, the level of the Dcl transcript in the PFCx increased in response to the mianserin treatment (log2FC = 0.46, P = 0.0009). The expression of the full-length Dclk1 transcript (ENSMUST00000054237) was below the level of detection in the NAc, and no significant changes were observed in the PFCx. The expression of the intron-retained isoform (RI) was substantially upregulated by mianserin (log2FC = 0.9, P = 2.8 × 10−6) and risperidone (log2FC = 0.49, P = 0.002) in the NAc and by mianserin (log2FC = 0.69, P = 0.0003) in the PFCx.
Furthermore, when we mapped individual reads to the Dclk1 locus, a considerable number of fragments did not match known transcripts and corresponded to a fragment annotated as an intronic region (Fig. 4b, red arrow). Exon-level analysis of the data suggested the presence of a novel transcript, referred to as Dclk1-m, with an alternative transcription start site (first exon), termination codon (last exon) and a sequence corresponding to the sixth intron of the canonical Dclk1 transcript (ENSMUST00000054237). The detected retention of intron 6 is unlikely to correspond to an unspliced pre-mRNA. The level of intron coverage calculated as the median read coverage for each position at each intron was significantly higher (P = 0.002) for this intron compared with that of other introns (Fig. 4a).
Based on the frequency of the sequence reads corresponding to the Dclk1-m variant, the expression of this isoform was upregulated in the PFCx by the mianserin treatment (log2FC = 0.679, P = 0.0000253) and in the NAc by mianserin and risperidone (log2FC = 0.862, P = 3.8 × 10−8; log2FC = 0.542, P = 0.00079, respectively; Fig. 5 and Additional file 5: Figure S5). Notably, the ILMN_1259689 probe shown in Fig. 1 hybridized with Dclk1-m. In conclusion, next-generation sequencing analysis confirmed the drug-specific induction of transcription from the Dclk1 locus and showed that psychotropic drugs affected the transcription of a short region close to the sequence encoding the serine-proline rich domain. Most importantly, changes in transcription might actually represent a novel transcript, Dclk1-m.
Validation of drug-induced expression of Dclk1 isoforms
We designed a series of isoform-specific probes for qPCR to confirm which of the transcripts exhibited increased expression in response to drug treatments. The analysis was performed on a new set of samples obtained from mice euthanized 2 h after the i.p. injection of mianserin (20 mg/kg), risperidone (0.5 mg/kg), haloperidol (1 mg/kg) or saline (n = 8). Five fluorescent probe assays were used to analyze changes in the expression of Dclk1 transcripts; two of the assays were specifically designed to detect the isoform containing the retained intron (Fig. 6, 5′ region of intron 6 and 3′ region of intron 6).
The mianserin treatment increased the levels of transcripts containing intron 6 in the NAc (approximately 1.5-fold). The level of Cpg16 was slightly decreased, whereas levels of Dcl/Carp were not different from those of the saline-treated control group. Risperidone significantly induced the expression of the Dclk1/Dcl isoform (approximately 1.2-fold), as well as transcripts containing intron 6 (approximately 1.6- to 2.1-fold). The haloperidol treatment had no effect on the expression of Dclk1 isoforms in the NAc. In the PFCx, the mianserin treatment upregulated the expression of Dcl/Carp isoforms (1.5-fold), as well as transcripts containing intron 6 (more than twofold). Risperidone did not regulate the expression of Dclk1 isoforms in the PFCx. The haloperidol treatment slightly decreased the expression of a variant containing part of intron 6 in the PFCx (Fig. 6 and Additional file 6: Figure S6).
We also performed two additional experiments and measured the transcript levels 4 h after a single treatment and 2 h after 5 days of treatment. The mianserin treatment significantly increased levels of Dclk1-m in the NAc. The effect of mianserin on Dclk1 expression was not detected 5 days after repeated drug treatment (Additional file 7: Figure S7). Together, these data confirm that Dclk1-m expression is acutely increased in response to the mianserin treatment. Levels of other Dclk1 transcripts were not increased in response to the mianserin treatment, with the possible exception of Carp.
Functional features of the newly detected Dclk1 variant
Finally, we used mass spectrometry to determine whether protein products of the Dclk1-m transcript were detected. Dclk1-m and Carp sequences overlap on the 5′ side of the putative protein-coding sequences; only the C-termini differ. Both sequences include the proline- and serine-rich (SP-rich) region that interacts with other proteins. Mass spectrometry analysis of the SDS-PAGE-purified protein fraction with a molecular weight less than 10 kDa confirmed the presence of the SP-rich fragment (Fig. 7). Shotgun LC–MS/MS analysis, performed in addition to PRM, did not show a presence of other fragments derived from either CARP or DCLK1-M. Thus, we were only able to conclude that either or both the DCLK1-M and CARP proteins are translated.