Final Report Summary - RNA-DIREC (RNA-DIRECTED EPIGENETIC REMODELLING IN CANCER)
Prostate cancer is the most commonly detected cancer in men in Europe and the US and the second-leading cause of cancer death. Importantly, due to the highly heterogeneous behavior of the cancer, over-diagnosis and overtreatment are of major concern. Clearly, any progress made in the further characterization of the molecular subtypes of prostate cancer will be of great benefit to both patients and society.
Dr. Coolen and his group study the mechanisms of gene (de)regulation in prostate cancer, in particular via systems that do not imply genetic changes, but rather concern an affected use of the genomic information; a phenomenon termed epigenetics. Through his interest, Dr. Coolen has established a vast knowledge on the prostate cancer epigenome and transcriptome both within his group and through (inter)national scientific networks with the aim to understand and ultimately better diagnose and target the prostate cancer subtypes. Recently, he has revealed that epigenetic changes in prostate cancer are not intrinsically localized to individual genes, but may encompass large regions resulting in silencing or activation of neighboring genes (Long Range Epigenetic Silencing or Activation, LRES or LREA, respectively). Through LRES and LREA, a combination of epigenetic marks concordantly suppress or activate multiple neighboring genes within the genome. A key question that remained was how such regions or domains are being formed, and potentially also reverted back to a more normal state.
With the current grant, Dr. Coolen and his group set out to investigate the potential involvement of RNA molecules in the formation of epigenetic domains, in particular domains that are repressed in prostate cancer by a specific signature: trimethylation of histone H3 (H3K27me3).
Firstly, high resolution maps of the genome-wide distribution of the H3K27me3 mark in various healthy and cancer prostate cells were generated via so-called ChIP-seq experiments, which revealed vastly different patterns between cancer and healthy cells. The differences were present as large cancer-specific domains of strong H3K27me3 signals, which could not be explained by copy number variations or rearrangements in the DNA. Attempts to identify these large H3K27me3 domains using existing computational algorithms, failed to detect these domains as uninterrupted regions. Therefore, the group of Dr. Coolen has set up a novel combinatorial gene-centered approach to identify contiguous domains of H3K27me3 marked genes. Using this approach, 1,050 H3K27me3 domains were identified in the prostate cancer genome that span a median of 5 genes (range: 3-64) and with a median size of 250 kb (range: 1.8 kb - 5.1 Mb).
Next, the transcriptional state of genes in these H3K27me3 domains were investigated using nuclear RNA sequencing experiments on prostate cancer and healthy prostate epithelial cells. To this end, RNA-seq data were statistically interrogated and merged with information on the H3K27me3 domains, in order to identify all cancer-specific H3K27me3 domains that were transcriptionally repressed. This analysis yielded 37% of H3K27me3 domains (386 out of 1,050) harbor genes that are significantly repressed in prostate cancer cells, while the other 664 domains were transcriptionally inactive in both cancer and normal cells. Both types of domains were further studied in detail to identify unique genomic and epigenomic features. Importantly, many of the large H3K27me3 domains harbor tumor suppressor genes and are recurrently repressed in clinical prostate cancer as apparent from The Cancer Genome Atlas (TCGA) database. These experiments have allowed for the first time an in-depth analysis of the occurrence and features of repressive H3K27me3 domains in prostate cancer. These findings are currently being written up and both the approach and the results will soon be submitted for publication.
After the demarcation of transcriptionally repressed H3K27me3 domains in prostate cancer, Dr. Coolen’s group set out to identify any RNA molecules that are putatively involved in the formation of these domains. The hypothesis for the involvement of RNA molecules was partly based on the evidence that in healthy cells, specific genomic regions that harbor clusters of so-called imprinted genes are being repressed by a locally synthesized long non-coding RNA (lncRNA). Such a lncRNAs can act as a guide for epigenetic enzymes that are responsible for the repression of these genes. To test this hypothesis, two lines of investigation were set up. In the first line, the formation of (novel) non-coding RNA molecules within the vicinity of the H3K27me3 domains was investigated. To this end, a bioinformatic algorithm was developed that was able to detect (novel) transcripts based on a combination of epigenetic marks that demarcates transcriptional start sites (such as H3K4me3) and nuclear RNA sequencing data. Results of this algorithm were overlaid with information on known genes (GENCODE) to allow the identification of both known and novel transcripts. This approach yielded important information on 86 novel transcripts detected within or near the cancer-specific H3K27me3 domains. For example, a novel lncRNA within a H3K27me3 domain on 3p12 encompassing the tumor suppressor gene ROBO1, was prostate cancer specific and importantly was found to strongly bind to the polycomb repressive complex 2 (PRC2). The PRC2 complex is responsible for the deposition of the H3K27me3 mark. Using antisense oligonucleotide manipulation studies, the levels of the lncRNA could be efficiently reduced, while conversely the nearby repressed genes ROBO1 and GBE1 within the H3K27me3 domain were transcriptionally elevated, and local levels of H3K27me3 were decreased. Similar results were obtained for a novel lncRNA in the 4p15 locus, harboring the tumor suppressor gene SLIT2. These data provide functional evidence that prostate cancer specific H3K27me3 domains are regulated by locally synthesized lncRNAs, involved in guiding the PRC2 complex to their target regions. Currently, the group of Dr. Coolen is writing up these data into a publishable format.
A second line of approach, to investigate the role of RNA molecules in guiding the PRC2 complex to its target regions and the subsequent formation of H3K27me3 domains, involved the global characterization of RNA molecules that were able to interact with the PRC2 complex, via a technique called RNA-immunoprecipitation (RNA-IP). This challenging technique revealed many RNA molecules to be able to bind to the PRC2 complex and confirmed the findings for the locally synthesized lncRNAs described above. In addition and to their surprise, both lncRNAs and protein-coding mRNAs appeared to bind to the PRC2 complex with similar binding efficiencies. These results could be independently confirmed and were recently published in Cancers (PMID: 24216986), along with a novel hypothesis that mRNAs in general can have diverse regulatory functions besides their protein coding role. This is a clear dogma change and may have vast implications for any conclusions drawn for disease-associated mRNAs, also outside the field of prostate cancer. The group of Dr. Coolen is currently further investigating these important findings.
In conclusion, the group of Dr. Coolen has with the help of this grant, been able to prove the causative role for RNA molecules in the formation of large repressed domains in the prostate cancer genome. These findings form highly interesting leads to further study the possibilities for manipulating this mechanism in vivo in order to reactivate tumor suppressor genes, present in these domains. Ultimately, this could lead to novel treatment strategies for prostate cancer, which will obviously have great beneficial impact on both patients and society. In addition, this grant has highlighted the putative regulatory roles for protein-coding transcripts, which may have vast impact on the phenotypic attributions to protein-coding genes in general and which is clearly an area deserving further attention. More specifically, we will proceed to investigate: (1) the role of protein-coding transcripts in the targeting of epigenetic enzymes to their genomic targets, and (2) the role of protein-coding transcripts as a sponge or decoy for microRNAs, thereby affecting the available pool of microRNAs in a cell. This is of particular importance in our understanding of the complete range of consequences of transcript (de)regulation in both health and disease.
Dr. Coolen and his group study the mechanisms of gene (de)regulation in prostate cancer, in particular via systems that do not imply genetic changes, but rather concern an affected use of the genomic information; a phenomenon termed epigenetics. Through his interest, Dr. Coolen has established a vast knowledge on the prostate cancer epigenome and transcriptome both within his group and through (inter)national scientific networks with the aim to understand and ultimately better diagnose and target the prostate cancer subtypes. Recently, he has revealed that epigenetic changes in prostate cancer are not intrinsically localized to individual genes, but may encompass large regions resulting in silencing or activation of neighboring genes (Long Range Epigenetic Silencing or Activation, LRES or LREA, respectively). Through LRES and LREA, a combination of epigenetic marks concordantly suppress or activate multiple neighboring genes within the genome. A key question that remained was how such regions or domains are being formed, and potentially also reverted back to a more normal state.
With the current grant, Dr. Coolen and his group set out to investigate the potential involvement of RNA molecules in the formation of epigenetic domains, in particular domains that are repressed in prostate cancer by a specific signature: trimethylation of histone H3 (H3K27me3).
Firstly, high resolution maps of the genome-wide distribution of the H3K27me3 mark in various healthy and cancer prostate cells were generated via so-called ChIP-seq experiments, which revealed vastly different patterns between cancer and healthy cells. The differences were present as large cancer-specific domains of strong H3K27me3 signals, which could not be explained by copy number variations or rearrangements in the DNA. Attempts to identify these large H3K27me3 domains using existing computational algorithms, failed to detect these domains as uninterrupted regions. Therefore, the group of Dr. Coolen has set up a novel combinatorial gene-centered approach to identify contiguous domains of H3K27me3 marked genes. Using this approach, 1,050 H3K27me3 domains were identified in the prostate cancer genome that span a median of 5 genes (range: 3-64) and with a median size of 250 kb (range: 1.8 kb - 5.1 Mb).
Next, the transcriptional state of genes in these H3K27me3 domains were investigated using nuclear RNA sequencing experiments on prostate cancer and healthy prostate epithelial cells. To this end, RNA-seq data were statistically interrogated and merged with information on the H3K27me3 domains, in order to identify all cancer-specific H3K27me3 domains that were transcriptionally repressed. This analysis yielded 37% of H3K27me3 domains (386 out of 1,050) harbor genes that are significantly repressed in prostate cancer cells, while the other 664 domains were transcriptionally inactive in both cancer and normal cells. Both types of domains were further studied in detail to identify unique genomic and epigenomic features. Importantly, many of the large H3K27me3 domains harbor tumor suppressor genes and are recurrently repressed in clinical prostate cancer as apparent from The Cancer Genome Atlas (TCGA) database. These experiments have allowed for the first time an in-depth analysis of the occurrence and features of repressive H3K27me3 domains in prostate cancer. These findings are currently being written up and both the approach and the results will soon be submitted for publication.
After the demarcation of transcriptionally repressed H3K27me3 domains in prostate cancer, Dr. Coolen’s group set out to identify any RNA molecules that are putatively involved in the formation of these domains. The hypothesis for the involvement of RNA molecules was partly based on the evidence that in healthy cells, specific genomic regions that harbor clusters of so-called imprinted genes are being repressed by a locally synthesized long non-coding RNA (lncRNA). Such a lncRNAs can act as a guide for epigenetic enzymes that are responsible for the repression of these genes. To test this hypothesis, two lines of investigation were set up. In the first line, the formation of (novel) non-coding RNA molecules within the vicinity of the H3K27me3 domains was investigated. To this end, a bioinformatic algorithm was developed that was able to detect (novel) transcripts based on a combination of epigenetic marks that demarcates transcriptional start sites (such as H3K4me3) and nuclear RNA sequencing data. Results of this algorithm were overlaid with information on known genes (GENCODE) to allow the identification of both known and novel transcripts. This approach yielded important information on 86 novel transcripts detected within or near the cancer-specific H3K27me3 domains. For example, a novel lncRNA within a H3K27me3 domain on 3p12 encompassing the tumor suppressor gene ROBO1, was prostate cancer specific and importantly was found to strongly bind to the polycomb repressive complex 2 (PRC2). The PRC2 complex is responsible for the deposition of the H3K27me3 mark. Using antisense oligonucleotide manipulation studies, the levels of the lncRNA could be efficiently reduced, while conversely the nearby repressed genes ROBO1 and GBE1 within the H3K27me3 domain were transcriptionally elevated, and local levels of H3K27me3 were decreased. Similar results were obtained for a novel lncRNA in the 4p15 locus, harboring the tumor suppressor gene SLIT2. These data provide functional evidence that prostate cancer specific H3K27me3 domains are regulated by locally synthesized lncRNAs, involved in guiding the PRC2 complex to their target regions. Currently, the group of Dr. Coolen is writing up these data into a publishable format.
A second line of approach, to investigate the role of RNA molecules in guiding the PRC2 complex to its target regions and the subsequent formation of H3K27me3 domains, involved the global characterization of RNA molecules that were able to interact with the PRC2 complex, via a technique called RNA-immunoprecipitation (RNA-IP). This challenging technique revealed many RNA molecules to be able to bind to the PRC2 complex and confirmed the findings for the locally synthesized lncRNAs described above. In addition and to their surprise, both lncRNAs and protein-coding mRNAs appeared to bind to the PRC2 complex with similar binding efficiencies. These results could be independently confirmed and were recently published in Cancers (PMID: 24216986), along with a novel hypothesis that mRNAs in general can have diverse regulatory functions besides their protein coding role. This is a clear dogma change and may have vast implications for any conclusions drawn for disease-associated mRNAs, also outside the field of prostate cancer. The group of Dr. Coolen is currently further investigating these important findings.
In conclusion, the group of Dr. Coolen has with the help of this grant, been able to prove the causative role for RNA molecules in the formation of large repressed domains in the prostate cancer genome. These findings form highly interesting leads to further study the possibilities for manipulating this mechanism in vivo in order to reactivate tumor suppressor genes, present in these domains. Ultimately, this could lead to novel treatment strategies for prostate cancer, which will obviously have great beneficial impact on both patients and society. In addition, this grant has highlighted the putative regulatory roles for protein-coding transcripts, which may have vast impact on the phenotypic attributions to protein-coding genes in general and which is clearly an area deserving further attention. More specifically, we will proceed to investigate: (1) the role of protein-coding transcripts in the targeting of epigenetic enzymes to their genomic targets, and (2) the role of protein-coding transcripts as a sponge or decoy for microRNAs, thereby affecting the available pool of microRNAs in a cell. This is of particular importance in our understanding of the complete range of consequences of transcript (de)regulation in both health and disease.