miR expression profiling data from CC mice maps QTL regions containing candidate regulatory genes
miR expression profiling data indicate genetic diversity in CC mice
Upstream regulation of miRs may involve multiple factor/s, so we needed to determine whether the CC was a suitable model for studying miR regulation. We performed qPCR expression analysis of three miRs: miR-9, miR-29a and miR-107. In brief, we utilized the hippocampi of male mice from 54 genetically diverse CC strains at two ages (5–6 weeks, and 30–40 weeks; Table S1). We found differential expression of miR-9, miR-29a and miR-107 between the CC strains. We observed an upregulation of 1.2±0.5 and downregulation of 0.65±0.17 in miR-9/29a/107 expression within the 5–6-week-old CC strains (Figure 1A, Figures S1 A-B). Similarly, we observed an upregulation of 5.1±2.7 and downregulation of 2.75±0.5 in miR-9/29a/107 expression within the 30–40-week-old CC strains used in this study (Figure 1B, Figures S1 C-D).
To further study diversity in gene expression between CC lines, we tested expression of the Bace1 gene, a downstream target for miR-9/29a/107 14; 15; 16. Upregulation of miR should downregulate its targets, including Bace1, and vice versa. We checked whether in CC mice upregulation of miRs’ expression downregulates Bace1. We observed this only for 50% CC strains in miR-9 whereas for 45% CC strains in miR-29a/107 in young mice hippocampus (Figure 1C). Similarly, for the hippocampus of older mice, only 63% CC lines in miR-29a and 53% in miR-9/107 respectively showed this change (Figure 1D). Therefore, these data confirm that CC mice offer the potential to identify upstream genetic factors in miR regulation.
QTL analysis and siRNA knockdown identify upstream candidates
The Gene Miner and HAPPY software packages were used to analyse miR expression data to map QTLs to specific genome locations. Here, we calculated the fold-change expression of miR-9/29a/107 and mapped the QTLs associated with this trait for each miR. We considered logarithm of the odds (LOD) scores >3 as significant to link the loci to the miR trait of interest. We identified QTLs on Chr 19 (LOD>3.78), Chr 15 (LOD>6.32), Chr 5 (LOD>4.24), Chr 7 (LOD>12.6), Chr 12 (LOD>11.2), Chr 17 (LOD>3.68), Chr 1 (LOD>12.5). QTLs were considered at the 50th and 95th percentile significance with a total of 131 genes resided within the mapped QTLs regulating miR-9/29a/107 expression (Figures S1 E-J). Of these genes, 40 were associated with the miR-9 regulatory QTL, 17 with the miR-29a QTL, and 88 were linked to the miR-107 QTL (Table 1).
In order to make a shortlist of the most promising candidates for further study, we short interfering RNAs (siRNAs) to knockdown the expression of all 131 genes individually (knockdown efficiency >70% was achieved, Figures S1 K-L). We performed this experiment using a mouse neuroblastoma cell line (Neuro-2a) and compared the fold-change (FC) regulation of miR-9/29a/107 to the scrambled siRNA negative control using qPCR (oligonucleotides listed in Table S2). We pooled the genes with FC>1 and identified 23 (of 131 candidates) that affected levels of miR-9/29a/107 (Figure 1E). Candidate gene prioritisation was performed employing a bioinformatics platform for gene ontology (GO) analysis and disease segregation. We used Enrichr, DAVID, DisGeNET and EBI to further reduce the candidate list by identifying modules enriched in processes related to neurogenesis, gene regulation and neurodegeneration. We also looked at the gene–disease association score to identify ontology classes related to neurodegenerative disorders (Figures S1 M-N). Thirteen (of 23) candidate genes – Ahctf1, Mgea5, Polr1c, Trabd, Creld2, Lzts2, Mrpl14, Sccpdh, Pdcd11, Cuedc2, Ephx1, Nt5c2 and Panx2 were matched to the desired GO terms (Figure 1F). Taken together, these experiments demonstrated that CC is a valuable tool for studying the genetic basis of miR regulation.
Panx2, Polr1c and Mgea5 alter neuronal differentiation through miR-9 in mouse neural progenitor cells
Panx2, Polr1c and Mgea5 are involved in neuronal differentiation
miR-9 plays a prominent role in neuronal differentiation 5. Therefore, in order to reduce the list of 13 candidate genes, we conducted a neuronal differentiation assay. Mouse neural progenitor cells (mNPCs) were isolated from the adult (30–40-week-old) mouse hippocampi to study the molecular mechanisms controlling hippocampal neurogenesis. To determine if the 13 candidate genes played any role in neuronal differentiation, mNPCs were treated with siRNAs for the 13 candidate genes, then allowed to differentiate for three days. Neuronal differentiation was quantified by the βIII-tubulin marker using Image J software. The number of βIII-tubulin positive cells changed significantly only after knockdown of four candidate genes, namely Panx2, Polr1c, Mgea5 and Ephx1 (Figures S2 A-D). Panx2 increased, while both Polr1c and Mgea5 decreased the number of neuronal-like cell phenotypes after knockdown (Figures 2A-B). Ephx1 knockdown increased neuronal differentiation in mNPCs (Figures S2 C-D). To confirm these results, we performed rescue experiments to overexpress Panx2, Polr1c and Mgea5 using a lentiviral system and determined if this reversed the effect of knockdown. We found that overexpression of Panx2 in mNPCs decreased neuronal differentiation, whereas overexpression of Polr1c and Mgea5 led to increased neuronal differentiation compared to the effect of single knockdowns and control. (Figures S2 E-H). Therefore, Panx2, Polr1c, Mgea5 and Ephx1 were selected as the best candidates for further study.
miR-9 inhibition decreased neuronal differentiation
Since all the candidates identified resided in the QTL regions controlling miR-9/107 expression, we performed further study of the neuronal differentiation of miR-9/107 in mNPCs. We used miR inhibitors (miR-in) to inhibit miR function in mNPCs. After neuronal differentiation, we observed that miR-9-in reduced the number of βIII tubulin-positive cells significantly compared to the miR negative control (Figures 2C-D). miR-107 did not change neuronal differentiation significantly (Figures S2 I-J). To confirm the role of miR-9 in neuronal differentiation, we overexpressed miR-9 using lentivirus and found that neuronal differentiation was increased (Figures S2 K-L). These results demonstrate that miR-9 decreased neuronal differentiation in adult mNPCs.
Panx2, Polr1c and Mgea5 regulate neuronal differentiation through miR-9
To further test if Panx2, Polr1c, Mgea5 and Ephx1 altered neuronal differentiation through miR-9, we performed double knockdown experiments using siRNAs, together with miR-9-in. We found that neuronal differentiation was dramatically lower in Panx2, Polr1c and Mgea5 than in the individual knockdowns and the scrambled negative control. However, Ephx1-transfected mNPCs did not show any significant change in neuronal differentiation after inhibition with miR-9 (Figures S2 M-N). Compared to the single knockdown of Panx2, double knockdown using siPanx2 and miR-9-in reduced 92% of the βIII tubulin-positive cells (Figures 2E, F). Similarly, double knockdown using siPolr1c and miR-9-in (Figures 2E, G), siMgea5 and miR-9-in (Figures 2E, H) decreased 57% and 80% of the βIII tubulin-positive cells, respectively. Furthermore, we tried to rescue the miR-9 mediated decrease in neuronal differentiation of Panx2, Polr1c and Mgea5 by using miR-9 overexpressed lentiviral mNPCs. The knockdown of Panx2 in miR-9 overexpressed mNPCs significantly increased 86% of βIII tubulin-positive cells (Figures S2 O, P). Similarly, miR-9-overexpressed mNPCs increased 89% and 53% of βIII tubulin-positive cells after knockdown using siPolr1c (Figures S2 O, Q) and siMgea5 (Figures S2 O, R) respectively. Thus, the miR-9-in-mediated decrease in neuronal differentiation of Panx2, Polr1c and Mgea5 could be reversed by miR-9 overexpression. Overall, these data suggest that Panx2, Polr1c and Mgea5 regulated neuronal differentiation through miR-9.
Distribution of the Panx2, Polr1c and Mgea5 binding peaks on the miR-9 locus
Panx2, Polr1c and Mgea5 proteins bind to the miR-9 locus
miR-9 is located within the 34kb long miR-9 host gene (miR-9hg) on chromosome 7. To determine how Panx2, Polr1c and Mgea5 regulate miR-9 expression, we performed ChIP-seq (in duplicates) in mNPCs. The datasets were mapped to the miR-9 genomic locus using MACS3 software for peak calling. All three proteins encoded by these genes were found to have peaks enriched on the miR-9 locus in two independent ChIP-seq experiments. The ChIP-seq peaks were viewed using the Integrative Genomics Viewer 17. Panx2 (Figure 3A), Polr1c (Figure 3B) and Mgea5 (Figure 3C) peaks are displayed in colour against the input control (shown in grey). Similarly, we performed ChIP-qPCR in wild-type mNPCs and validated the binding of Panx2, Polr1c and Mgea5 on the miR-9 genomic region (Figure 3D, oligonucleotides listed in Table S2). This experiment was compared using rabbit IgG as negative control, which showed no enrichment, and H3K4me3 (as positive control), which showed high enrichment on the miR-9 locus. Thus, our findings indicate that Panx2, Polr1c and Mgea5 bound to the miR-9 genomic locus.
In order to determine if the Panx2, Polr1c and Mgea5 binding regions were located on the regulatory region of miR-9, we examined whether they overlapped with cis-acting elements. Using the UCSC genome browser to identify the distribution of Encyclopedia of DNA Elements (ENCODE) sites and CpG islands, the region of interest overlapped with the sites that were highly enriched in ENCODE peaks and CpG islands (Figures S3 A-C). Therefore, these data suggest that Panx2, Polr1c and Mgea5 were recruited to the regulatory region of the gene encoding miR-9.
Panx2, Polr1c and Mgea5 binding sites correlate with acetylation
Acetylation of H3 at Lys9/14 is associated with transcriptionally active chromatin, whereas methylation at H3K9 is a hallmark for silent chromatin 18. Histone mark H3K4me3 is also associated with higher transcriptional activity. We next examined the histone-specific modification of Panx2, Polr1c and Mgea5 binding sites on the miR-9 locus. Using ChIP-qPCR in wild-type mNPCs, we studied histone modification pattern of H3K4me3, H3K9/14ac, H3K9me2 and H3K9me3 on the 34kb locus of miR-9. We found that H3K9/14ac was predominantly enriched in all the binding sites of Panx2, Polr1c and Mgea5 on the miR-9 locus. We saw high enrichment of H3K4me3 in PoBS-1/2 and the miR-9 promoter region. MgBS2 also coincided with H3K9me3, which marks methylation. Overall, the level of acetylation was higher throughout the binding sites of Panx2, Polr1c and Mgea5 on the miR-9 locus in the wild-type mNPCs (Figure 3E).
Panx2, Polr1c and Mgea5 are associated with each other
In order to determine if Panx2, Polr1c and Mgea5 form a complex, we performed co-immunoprecipitation experiments (Co-IP). Wild-type mNPC lysates were individually immunoprecipitated with rabbit antibodies specific for each of Panx2, Polr1c and Mgea5 and probed using antibodies to the other two proteins. We used rabbit IgG (as negative control) and input control to confirm the specificity of the antibody. Polr1c and Mgea5-specific bands were found in the fraction immunoprecipitated with Panx2 antibody (Figures 3F-G). Mgea5 also precipitated Panx2 and Polr1c from mNPCs (Figures 3H-I). However, we did not observe any Panx2 and Mgea5-specific bands in the fraction immunoprecipitated with Polr1c antibody (possibly because anti-Polr1c binds an epitope that disrupts the interaction via steric interference or direct competitive binding) (Figures 3J-K). Thus, we demonstrated that Panx2/Mgea5 existed in a complex with each other and Polr1c. In this way, we determined an association between Panx2, Polr1c and Mgea5. A schematic of Co-IP interactions is shown in Figure S3 D.
Panx2, Polr1c and Mgea5 as a miRSome acting upstream of miR-9
Overall, we confirmed that Panx2, Polr1c and Mgea5 bind to the regulatory region of miR-9 locus and demonstrated a physical association with each other. We termed this intracellular complex, made of chromosomal DNA bound by the proteins, which serves as the site for regulating the expression of the gene encoding specific miR, called “miRSome”. The miRSome regulating miR-9 expression is referred to as “miR9Some”.
Nuclear signaling by Panx2, Polr1c and Mgea5
Cellular localization of Panx2, Polr1c and Mgea5
To further explore the functions of Panx2, Polr1c and Mgea5, we examined the localization of these three proteins as reported in the Uniprot protein database, that is, Panx2 in the cytoplasm, Polr1c in the nucleus and Mgea5 in both the cytoplasm and nucleus. To confirm this, we performed immunostaining in mNPCs using antibodies to Panx2, Polr1c and Mgea5. Panx2 was located in the cytoplasm, while Polr1c was localized to the nucleus. Mgea5 was present both in the nucleus and cytoplasm (Figure S4 A). Next, we performed immunoblotting in the cytoplasmic and nuclear extract using the respective antibodies. We found a 70KDa band in the cytoplasmic extract for Panx2, but this was absent from the nuclear extract. For Polr1c, a 40KDa band was identified in the nuclear extract, but was absent from the cytoplasm. In contrast, for Mgea5 a band at 130KDa was present in both the cytoplasmic and nuclear extracts (Figures S4 B-D).
Cleavage of Panx2 C-terminal fragment
Panx2 is a membrane protein with four transmembrane domains, two extracellular loops, one cytoplasmic N/C-terminal and one intracellular loop (Figure 4A). To determine whether the intracellular domain of Panx2 was cleaved and released from the cells, we performed a shedding assay. We tested the shedding of endogenous Panx2 from mNPCs, as well as from mNPCs stably expressing Panx2 protein. Western blot analysis of whole cell lysates using antibody directed against the C-terminal domain of Panx2 allowed detection of the cleaved C-terminal fragment (CTF). Panx2 CTF migrated as a lower molecular weight fragment than the full-length Panx2 (70KDa). The cleavage of membrane proteins can be increased by agents that activate protein kinase C 19. In order to investigate if Panx2 cleavage was mediated by a similar mechanism, we used phorbol ester PMA, which is an activator of protein kinase C. mNPCs were treated with PMA for 2 hours at 37oC, and this resulted in an increased Panx2 CTF. We detected the presence of two fragments that were enriched after PMA treatment, which we termed CTF25 and CTF17 (Figure 4B). We also detected CTF25/CTF17 in mNPCs stably expressing Panx2, and overall showed a similar pattern as endogenous Panx2 as detected by the western blot analysis (Figure 4C). We transfected Panx2 CTF into mNPCs and found enrichment of CTF25/CTF17 fragments (Figure 4D). We also tested the effect of inhibition on the cleavage of Panx2 using DAPT (γ-secretase inhibitor) and metalloprotease inhibitor GM6001. The presence of either agent reduced the levels of CTF25/CTF17 compared to the PMA treatment alone but was the same as treatment with the vehicle (DMSO) (Figures 4B-D). Overall, these data suggested that Panx2 cleavage was mediated by PMA and γ-secretase/metalloprotease inhibitors, similar to other membrane proteins 20; 21; 22; 23. Finally, we identified a nuclear localization signal (NLS) in the CTF (amino acid: 341-343) of Panx2. We transfected the mutated NLS Panx2 c-terminal fragment into mNPCs and, using western blotting, found that CTF25/CTF17 was not formed at all (Figure 4E).
Nuclear translocation of Panx2
Next, we determined whether the cleaved Panx2 would be translocated to the nucleus. We treated mNPCs with PMA followed by nuclei isolation and immunoprecipitation with antibody to the Panx2 c-terminal domain or control rabbit IgG. There was a drastic increase in the levels of CTF25 relative to the control. Upon overexposure, we also detected CTF17 in the nucleus (Figure 4F). To visualize endogenous Panx2 in the nucleus, we stained the mNPCs under experimental conditions (unstimulated, PMA treatment, mutated Panx2-CTF-△341-343 and vehicle) with Panx2 c-terminal antibody. We saw higher nuclear signals in the PMA treated mNPCs than in the controls (Figure 4G and Figures S4 E).
We also analyzed the cellular localization of endocytosed CTF by performing live antibody uptake experiments. mNPCs incubated with Panx2 antibody (directed against the c-terminal fragment) at 4oC (to prevent endocytosis) showed surface labelling of cells. Following binding, the cells were incubated at 0 minutes or 2 hours at 37°C after PMA treatment. At 0 minutes, the antibody was still surface bound, but after 2 hours, the vast majority of the surface-bound antibody was endocytosed and translocated to the nucleus. The DMSO vehicle was sued as a treatment control with only secondary antibody to rule out non-specific binding (Figure S4 F). We also performed live cell imaging using a similar method to determine if the surface-bound CTF Panx2 antibody was transported to the nucleus. PMA-treated Panx2 showed internalization into the nuclear region (Figure S4 Movie G), whereas the control remained unaffected (Figure S4 Movie H). These results suggest that Panx2 was cleaved and the CTF was transported to the nucleus.
Panx2, Polr1c and Mgea5 regulate a chromatin-associated loop of the miR-9 locus
Panx2, Polr1c and Mgea5 form chromatin associated loop across miR-9 locus
Next, we explored the potential role of the miR9Some in higher-order chromatin loop organization in the 34kb miR-9 locus (Figure 5A). We hypothesized that if two or more remote sequences are physically close to each other, they may be confined to the same restriction enzyme-digested fragments of the crosslinked chromatin. Two candidate gene binding site (CGBS) sequences were extracted from Panx2, Polr1c and Mgea5 binding on the miR-9 genomic region as determined by the ChIP-seq data. We designed CGBS-specific primers to test if any of the binding sites were brought near each other (oligonucleotides listed in Table S2). If they were trapped in the same fragment, then it could be ligated and amplified by qPCR. To determine the loop, we used a chromosome conformation capture (3C) assay on crosslinked chromatin 24. We also used a ChIP-loop assay, which combines the 3C assay and immunoprecipitation with a specific antibody. This method helped to reveal the Panx2, Polr1c and Mgea5-specific loops and understand the histone-specific modifications related to transcriptionally active chromatin state associated with acetylation or transcriptionally silent chromatin configuration linked to methylation. We detected amplified signals using the primer sets F2 and F4 in the Panx2/Polr1c immunoprecipitated chromatin. The F2–F4 combination gave rise to an 8.4kb loop at the miR-9 locus. We detected a similar 8.4kb F2-F4 associated loop on the crosslinked chromatin isolated from the wild-type mNPCs without immunoprecipitation (3C assay). In contrast, this loop was not formed on the chromatin isolated from Panx2/Polr1c-null chromatin (Figures 5B-C, red box). Another 11.2kb loop was detected in the chromatin isolated from Mgea5 null mNPCs by the primers F3-F5. This loop was also seen after immunoprecipitation with Mgea5 antibody in the Mgea5 null mNPCs. But the wild-type mNPCs immunoprecipitated with Mgea5 antibody and wild-type mNPCs (3C assay) did not show the F3–F5 mediated 11.2kb loop (Figure 5D, magenta box). These data indicate that Panx2/Polr1c organised the chromatin into an 8.4kb loop, whereas Mgea5-null organised the chromatin into an 11.2kb loop of the miR-9 locus.
Panx2/Polr1c chromatin represents active chromatin and Mgea5 null chromatin represents silent chromatin
Next, we performed a ChIP-loop assay on the chromatin isolated from wild-type and Panx2/Polr1c/Mgea5 null mNPCs after immunoprecipitation with antibodies against acetylated H3-Lys9/Lys14 and dimethylated H3-Lys9. Using acetylated H3-Lys9/Lys14, we detected an 8.4kb F2-F4 loop on the wild-type mNPCs (Figure 5E, red box). We specifically detected the 11.2kb loop using the dimethylated H3-Lys9 in the Mgea5 null mNPCs (Figure 5F, blue box). The 8.4kb loop was closely associated with the miR-9 promoter and showed the enrichment of acetylation, whereas the 11.2kb loop was formed away from the miR-9 promoter region and was methylated. Therefore, the 8.4kb loop represented the transcriptionally active chromatin state and the 11.2kb loop formed the transcriptionally silent chromatin.
A Bacterial Artificial Chromosome (BAC) clone (RP23-207N5) containing the 34kb miR-9 locus was used to test the primer sets and normalize the PCR data (Figure 5G). Non-immune serum and genomic DNA were used as negative controls (Figures 5H-I). C1 and C2 were used as internal primer controls. A1 and A2 indicate PCR products derived from β-actin locus primers (Figure 5J).
Overall, according to our data, the higher-order chromatin structure was mediated by Panx2/Polr1c/Mgea5 binding to the miR-9 locus. A model based on our data is shown in Figure 5K. Panx2 and Polr1c mediated the 8.4kb loop at the miR-9 promoter and represented the active chromatin marked by acetylation. Mgea5 mediated the 11.2kb loop and depicted the silent chromatin marked by methylation. These data confirm that the long-range interactions formed by the miR9Some forms the miR-9 genomic region into a chromatin loop structure.
The miR9Some acts upstream to regulate miR-9 expression and drives neuronal differentiation
The miR9Some regulates miR-9 expression
We determined how the miR9Some regulates miR-9 expression in wild-type mNPCs and in Panx2-, Polr1c-, Mgea5- null mNPCs. There was a basal level of miR-9 expression in wild-type mNPCs, while miR-9 expression increased (approximately 1.5-fold change) in the individual Panx2/Polr1c/Mgea5 null mNPCs (Figure 6A). Next, we tested the combined effect of the miR9Some on miR-9 expression. We knocked down all three miR9Some genes in mNPCs and found that expression of miR-9 increased dramatically (approximately a threefold change) compared to the wild-type mNPCs and individual Panx2-, Polr1c- and Mgea5- null mNPCs (Figure 6B). To confirm that the miR9Some was essential to trigger luciferase activity in miR-9 promoter, we transfected Panx2, Polr1c and Mgea5 siRNAs with miR-9 and a mutant miR-9 lentiviral system in mNPCs. We found that Panx2, Polr1c and Mgea5 strongly triggered luciferase activity. The luciferase activity of miR-9 promoter was significantly increased upon downregulation of individual Panx2, Polr1c, Mgea5 and miR9Some compared to the miR-9 wild-type and scrambled negative control (Figure 6C). In the mutant miR-9 promoter, luciferase activity was unchanged (Figure 6D). Taken together, these data suggest that miR9Some upregulated miR-9 transcription by directly binding to the miR-9 promoter.
A regulatory pathway comprising miR9Some, miR-9 and Bace1
Bace1 is a downstream target of miR-9. We hypothesized that miR9Some should affect Bace1 expression in the same way as for miR-9, because miR-9 targets Bace1. We measured Bace1 expression in wild-type mNPCs, individual Panx2, Polr1c, Mgea5 null mNPCs and miR9Some null mNPCs and found that Bace1 levels were increased in individual Panx2-, Polr1c- and Mgea5- null mNPCs compared to the wild-type mNPCs (approximately 1.5-fold change) (Figure 6E). Furthermore, the miR9Some-null mNPCs showed approximately 30% greater reduction in the levels of Bace1 than in mNPCs with individual knockdown of Panx2, Polr1c and Mgea5. Next, we compared the expression of Bace1 and miR-9 after miR9Some knockdown and found reduction in Bace1 expression was probably due to increased miR-9 expression (Figure 6F). These results demonstrated that miR9Some regulated the expression of both miR-9 and Bace1 forming a regulatory pathway.
miR9Some acts in synergy to modulate neuronal differentiation
Because the miR9Some regulates miR-9 expression, we questioned whether it synergistically modulates neuronal differentiation of mNPCs. To perform this experiment, we generated mNPCs with different combinations of the double and triple knockdown cells Panx2/Polr1c, Panx2/Mgea5, Polr1c/Mgea5, miR9Some and each of these alongside miR-9-in. All the above experimental conditions were compared to the individual Panx2, Polr1c and Mgea5 null mNPCs. We quantified the number of βIII tubulin-positive cells to determine neuronal differentiation. Panx2 knockdown showed a higher increase in neuronal differentiation than all other experimental conditions (Figure 6G and Figure S5 A). The increase in neuronal differentiation was mainly mediated by Panx2 and Mgea5. In Polr1c null mNPCs, Panx2 and Mgea5 increased the βIII tubulin-positive cells significantly compared to the single Polr1c knockdown (Figure 6H and Figure S5 A). In Mgea5 null mNPCs, we observed a Panx2-mediated increase in neuronal differentiation of mNPCs compared to the single Mgea5 knockdown (Figure 6I and Figure S5 A). The miR9Some knockdown also showed increased neuronal differentiation. Likewise, miR9Some in combination with miR-9-in did not significantly decrease neuronal differentiation (Figures 6G-I). These experiments suggest that the miR9Some knockdown increased neuronal differentiation.