The putative tumor suppressor gene EphA7 is a novel BMI-1 target

Abstract

Bmi1 was originally identified as a gene that contributes to the development of mouse lymphoma by inhibiting MYC-induced apoptosis through repression of Ink4a and Arf. It codes for the Polycomb group protein BMI-1 and acts primarily as a transcriptional repressor via chromatin modifications. Although it binds to a large number of genomic regions, the direct BMI-1 target genes described so far do not explain the full spectrum of BMI-1-mediated effects. Here we identify the putative tumor suppressor gene EphA7 as a novel direct BMI-1 target in neural cells and lymphocytes. EphA7 silencing has been reported in several different human tumor types including lymphomas, and our data suggest BMI1 overexpression as a novel mechanism leading to EphA7 inactivation via H3K27 trimethylation and DNA methylation.

Keywords: Bmi1; DNA methylation; EphA7; neural stem cells.

Figure 1. Bmi1 overexpression increases proliferation and self-renewal of postnatal NSP cells in vitro

(A) Frequency of neurosphere initiation of empty vector, pCMMP-Bmi1 and pCMMP-Bmi1-FLAG transduced cells assessed 7 days post plating for 8 passages (n = 3). Error bars represent standard deviations. (B) Box plots representing neurosphere diameters determined for empty vector, pCMMP-Bmi1 and pCMMP-Bmi1-FLAG transduced NSP cells at passage 2 (50 spheres were investigated in 3 independent cultures). Whiskers represent the 10–90th percentiles. Results of unpaired t-tests in (A) and (B) are marked as follows: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ns: not significant (p > 0.05). (C) Fluorescent micrographs of empty vector and Bmi1-FLAG transduced neurospheres. (D) Total cell numbers measured at 7–9 days post plating over 8 passages (n = 3). Mean values with standard deviation and linear regression lines are shown. Linear regression analysis showed a significant difference between empty vector and pCMMP-Bmi1 and pCMMP-Bmi1-FLAG transduced NSP cultures (ANCOVA, p < 0.0001).

Figure 2. Identification of direct BMI-1 targets by ChIP

Representative agarose gel electrophoresis images of PCR-amplified Ink4a, EphA7, Ndn, Rps6ka6, Trp53bp2 and Tssc4 genomic regions using material from ChIP samples and input controls as template (n = 3). ChIP was performed with empty vector (empty) and pCMMP-Bmi1-FLAG-transduced NSP cells, applying the anti-FLAG antibody M2. A matched IgG1 isotype antibody was used as negative control and post-sonication cell lysate served as input control. Tssc4 PCR results are shown as negative example.

Figure 3. Derepression of EphA7 isoforms in Bmi1^−/− cells and downregulation in Bmi1-overexpressing cells

Gene expression levels determined by qPCR analyses are shown. p16ink4a and p19arf served as positive controls. (A) Upregulation of candidate BMI-1 target genes in embryonic day E14.5 neurosphere (NSP) cells isolated from Bmi1^−/− animals in comparison to cells isolated from Bmi1^+/+ animals (n = 8). Note the increased transcript levels of the full length (FL), truncated (T1) and secreted (S) EphA7 isoforms (B) Upregulation of EphA7 secreted isoform in postnatal spleen tissue from adult Bmi1^−/− animals in comparison to tissue from Bmi1^+/+ mice (n = 3). Downregulation of EphA7 isoform transcripts in postnatal neurosphere cells (C) and in postnatal spleen B cells (D) in Bmi1-overexpressing cells in comparison to empty vector control cells (n = 3). Note that upon Bmi1 overexpression in spleen B cells, transcripts of the secreted EphA7 isoform were not detectable (nd) anymore. Significant results of unpaired t-tests are marked as follows: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Mean values with standard deviation are shown.

Figure 4. EphA7 deletion does not rescue postnatal neurosphere and cerebellar defects in Bmi1^−/− mice

(A) Neural stem/progenitor cells grown as neurospheres (passage 1). Scale bars 100 μm. (B) Comparison of sphere diameters (n = 10–13). Mean values with standard deviation are shown. In Bmi1^−/− and Bmi1^−/−/EphA7^−/− cultures, most of the cells were single cells and did not form aggregates. (C) Sagittal sections of paraffin-embedded cerebelli stained with hematoxylin-eosin. Scale bars 800 μm. Cells and tissue were isolated from six week-old mice. (D) Granular and molecular layer thickness, mean values with standard deviation. 3 mice per genotype, 3–6 measurements per mouse. Significant results of unpaired t-tests are marked as follows: n.s. not significant (p > 0.05), **p ≤ 0.01, ***p ≤ 0.001.

Figure 5. Deletion of EphA7 increases the number of proliferating cells in the LVW dorsolateral corner of Bmi1^−/− mice

(A) Coronal anterior brain sections of 6–7 week-old mice stained with an antibody against Ki67 (red) and DAPI (blue). Scale bars: 100 μm. (B) Number of Ki67-positive nuclei per DAPI-positive nuclei shown as percentages (Wild-type: n = 10 ventricles, EphA7^−/− n = 10, Bmi1^−/− n = 8, double knockout n = 6). Results of unpaired t-tests are marked as follows: *p ≤ 0.05, ***p ≤ 0.001, ns: not significant (p > 0.05). Mean values with standard deviation are shown. Cartoons depict anterior (left hand side) and posterior (right hand side) forebrain sections. Arrows point to the dorsolateral corner (red). (C) Coronal anterior brain section of a 6 week-old wildtype animal stained with antibodies against Ki67 (green), the neuroblast marker Dcx (red), and the neural progenitor cell marker Mash1 (purple). Arrows point to Mash1-positive cells. Scale bar: 50 μm.

Figure 6. Overexpression of Bmi1 leads to increased DNA methylation of a genomic sequence within intron 1 of the EphA7 locus

(A) Mouse EphA7 genomic sequence including exon 1 (upper case) and the beginning of intron 1 (lower case). The start codon is marked by a box. Relative position of the CpG island is depicted on top. Underlined sequences indicate primers used to amplify three regions (B–D). CpGs are highlighted in grey. (B–D) Bisulfite sequence analysis of the three regions. Each row represents the sequence of an individual sequenced clone (open circle: unmethylated CpG site, filled circle: methylated CpG site). Clones from three independent biological samples of Bmi1- overexpressing and of empty control neurospheres cultured for 6 weeks after transduction were analyzed. The frequency of methylated CpGs was calculated for each clone and compared between Bmi1-overexpressing and empty control samples for each of the three regions. p-values of unpaired t-tests: region B p = 0.3351, region C p = 0.5611, region D p = 0.0029. (E) qRT-PCR of Bmi1-overexpressing NSPs cultured for 6 weeks after transduction shows an up-regulation of EphA7 secreted isoform upon AZA treatment. Results of unpaired t-tests (n = 3) are marked as follows: *p ≤ 0.05, ns: not significant (p > 0.05). (F) Agarose gel electrophoresis images of PCR-amplified EphA7 genomic regions (primer sets 1 and 2) using material from ChIP and input control samples as template. ChIP was performed with empty vector (empty) and pCMMP-Bmi1-FLAG-transduced NSP cells 2 and 6 weeks after transduction, applying an anti-H3K27me3 antibody. A matched IgG1 isotype antibody was used as negative control. Post-sonication cell lysate served as input control. Note the higher relative levels of bound H3K27me3 (H3K27me3 relative to input control) in Bmi1-overexpressing cells compared with empty control cells. The primer set 1 region is located 925 bp upstream of the EphA7 translation start site, the primer set 2 region is 341 bp upstream of this site.