ETV2 mediates endothelial transdifferentiation of glioblastoma

Glioblastoma multiforme (GBM) is characterized by extensive endothelial hyperplasia. Recent studies suggest that a subpopulation of endothelial cells originates via vasculogenesis by the transdifferentiation of GBM tumor cells into endothelial cells (endo-transdifferentiation). The molecular mechanism underlying this process remains poorly defined. Here, we show that the expression of ETS variant 2 (ETV2), a master regulator of endothelial cell development, is highly correlated with malignancy. Functional studies demonstrate that ETV2 is sufficient and necessary for the transdifferentiation of a subpopulation of CD133+/Nestin+ GBM/neural stem cells to an endothelial lineage. Combinational studies of ChIP-Seq with gain-of-function RNA-Seq data sets suggest that ETV2, in addition to activating vascular genes, represses proneural genes to direct endo-transdifferentiation. Since endo-transdifferentiation by ETV2 is VEGF-A independent, it likely accounts for the observed resistance of GBM tumor cells to anti-angiogenesis therapy. Further characterization of the regulatory networks mediated by ETV2 in endo-transdifferentiation of GBM tumor cells should lead to the identification of more effective therapeutic targets for GBM.


ChIP-seq
Binding of the mCherry antibody (rabbit anti-DS Red polyclonal, Clontech) to Etv2-mCherry was tested by whole mount fluorescent immunostaining before performing ChIP-seq experiments on progeny from a hsp70l:etv2-mCherry heterozygote line outcrossed to a wild type line, producing half a clutch with the inducible etv2-mCherry transgene and half without. The primary antibody was tested at 1:500 followed by 1:1000 of the Alexa Fluor 488 anti-rabbit secondary IgG (A11008 Life Technologies).
For ChIP, homozygous heat shock hsp70l:etv2-mCherry fish were out-crossed with wild-type fish, exclusively producing embryos with ubiquitous heat inducible Cterminal mCherry tagged Etv2, and equal numbers of wild type in-crosses were processed in parallel as controls. Transgenic expression was stimulated by floating embryos in Petri dishes at 38.5° C for 45 minutes, followed by re-incubation for 1 hour at 28.5° C before harvesting. ChIP was performed with the Protein A ChIP kit (Millipore) according to the manufacturer's instructions with the following modifications prior to sonication. Cells from 200 embryos per ChIP were first deyolked with Ringer's solution containing 0.5M EDTA with 2 cycles of centrifugation at 1,000Xg between changes of Ringer's buffer before fixation in 1% formaldehyde for 10 minutes. Samples were quenched with 0.125M Glycine, followed by 3 washes in PBS with protease inhibitors (Roche) and stored at -80°C. Cells were lysed in cell lysis buffer (10mM Hepes pH7.9, 0.5% NP40, 1.5mM MgCl2, 10mM KCl, 0.5mM DTT), followed by a high salt wash (50mM Hepes pH7.9, 0.1% SDS, 1% Triton X-100, 0.1% deoxycholate, 1mM EDTA, 500mM NaCl), and a final lysis in nuclear lysis buffer (1% SDS, 50mM Tris-HCl pH8, 20mM EDTA). Each lysis step was performed by passing samples through a 200 μl pipette tip and a 10 minute incubation on ice followed by a 10 minute centrifugation step (2,000Xg) at 4°C. Thereafter, lysates were sonicated with a standard Bioruptor (Diagenode) at a high frequency setting for 2 cycles, each cycle lasting 10 minutes with each minute divided into 30 seconds of sonication followed by 30 seconds of rest. Ice was changed between cycles to prevent overheating and denaturation of samples, producing a final sheared protein bound-DNA range of 200-500bp. The anti-DS Red polyclonal antibody (Clontech) was used at 5 ug per ChIP. A total of 1200 embryos were used for each ChIP-seq library. 2ng of each sample was used to generate each library with the Ovation Ultralow IL Multiplex System (Nugen) according to manufacturer's instructions. The final libraries were gel extracted with mini-elute columns (QIAGEN) at a range of 300-550 bp. Libraries were processed on an Illumina Hi-seq 2000 sequencer to generate single end 50bp reads.
Each ChIP sample was resuspended in 40μl elution buffer (QIAGEN) after chloroform extraction, and 2μl of each was tested by qPCR with SYBR green (Roche) using the primers listed in table $$1. Enrichment was determined by the ΔΔ Ct method and standardized with rhodopsin primers that were published previously 1 .

RNA-seq
The Illumina HiSeq 2000 platform yielded approximately 150 ~ 215 million single end short reads (50bp) per sample (Tables SM1 and SM2). Bowtie (version 0.12.8) (Langmead et al., 2009) was used to align the reads to the zebrafish genome Zv9 (danRer7) with up to 2 mismatches allowed. Each sample yielded mappable reads in excess of 80% (Tables SM1 and SM2), however, only uniquely mapped reads were used in subsequent gene differential expression analysis. Read counts per gene were summarized by the R package, GenomicRanges coupled with the UCSC danRer7 gene annotation table ensGene. A standard pair-wise comparison of two samples was performed by DESeq (Anders and Huber, 2010) with default parameters.

ChIP-seq
ChIP-seq DNA samples from wild type and hsp701:etv2-mCherry heterozygous zebrafish were sequenced separately in 2 lanes on the Illumina Hiseq 2000 platform.
Overall, 82,311,932 short reads (50bp) were obtained from the wild type control sample and 55,560,402 short reads were obtained from etv2-mCherry. The R package, FastQC, was utilized to perform a sequencing quality check. The first 4 bases corresponding to the inline barcodes were trimmed from the raw reads before mapping. Bowtie (--sam --best -n 2 -m 1) (Langmead et al., 2009) was used to map the trimmed raw reads to the zebrafish genome Zv9. Approximately 17 million and 13 million uniquely mapped reads for wild type and etv2-mCherry samples were identified respectively (Table   SM2). MACS (version 1.4) (Zhang et al., 2008) was then used to identify the corresponding read enriched regions on the genome with the following parameters: -t etv2_uniquly_mapped_reads -c control_uniquely_mapped_reads -g 1.5e9 -s 46 -w --single-profile --verbose 1. 10,029 peak regions were obtained with p-values < 1.00e- Sites(TSSs) were utilized in ChipPeakAnno to perform the annotation step. 6,950 Ensembl zebrafish TSSs were found to be neighboring the aforementioned 10,029 peak regions (Supplementary Table 5). Among these, 9,687 peak regions were within 10Kb of the nearest TSS, and 6,856/6,950 TSSs were within 10 KB of a peak region. A +/-10Kb flanking distance distribution to these TSSs indicated that most Etv2-mCherry binding sites were within several hundred base pairs of the TSSs, and a bimodal distribution flanking these TSSs can also be observed (below).
Genomic sequences +/-100bp from the summits of peaks enriched by MACS were extracted and submitted to DREME (Bailey, 2011) to investigate possible motifs in our Etv2-mCherry ChIP-seq data.