Dr. Schweiger: I'm going to talk about a story. We've been working now over the last few years in prostate cancer

Prostate cancer is the second leading cause of death in--for men in the United States. And it develops typically over benign form and then the high-grade prostate intraepithelial and then coming to the cancer into metastasis

And for prostate cancer, it's in particular interesting that there are two subgroups, which are characterized by a specific translocation. And this translocation is apparent in approximately 50 percent of all prostate cancers

It's a temporal translocation, which is also directed by an androgen promoter. And via this oncogenic translocation, those subgroup of prostate cancers develops

However, for the other cancers, which don't have a translocation, the pathogenesis is not known so far. And we tried to figure out what could be the underlying reason for those cancers

So, what we did in the laboratory, we thought one of the possibilities is that there is a genome-wide differential methylation alteration. And we did use a MeDIP-seq approach, where we used--where we shared the DNA. Then we performed an adapter ligation

The methylated DNA immunoprecipitation means that we used an antibody which targets the 5-methylcytosine and then can enrich those regions which are highly methylated. Then in amplification, finally, we did all the sequencing

We used more than 100 probes. Half of them were benign prostate samples, and the other half were tumor samples

And we achieved more than 1 billion uniquely mappable reads. And using a bioinformatics analysis, we first visualized what we found. And we looked here on the GSE-pi-1 [sp] region. This is a biomarker region, which is already in clinical trials for diagnostics

And you find here on the top four lines, these are the normal samples and down here the tumor samples. And you can easily see those distinct differentially methylated regions in the tumor samples

So, but, we didn't do just a single region analysis, but we did a whole genome analysis

And if we look on all the--on the complete genome, we find more than 150,000 cancer-related differentially methylated regions. And this number is enormous

And we can see in which areas of the genome they are located, and the hypermethylation we find predominantly in the promoters, and the hypomethylation down here is widely dispersed over the complete genome

We then did a principle component analysis. This means that we used a statistical analysis to use all the methylation data we have and look if there are similarities between the samples

And in blue, you can see the normal samples, whereas in red and yellow, these are our tumor samples. And not really surprising, we see that the normal samples are distinct from all the other samples

But, if you look even further, you can see that the yellow samples cluster together, and the red ones are more disperse

And you have to know that yellow stands for the translocation of fusion-positive tumors and red for the translocation-negative samples

This means that we have--coming from the methylation level, we have a distinct difference between the fusion-gene positive and negative samples

This is just more enlargement of this finding, only looking on the methylation of those tumor samples

And we can use also the differential methylation, pick out two regions, and then completely discriminate between those two tumor subgroups

We now wanted to know, what are the reasons that we really can distinguish between those two subgroups? And we counted the number of differential methylations occurring in those tumors

And surprisingly, we found this pattern. We found that normal tumors harbor certain number of differential methylations and fusion positives some more

But, surprisingly, we really see a significant increase in differential methylation events in the fusion-gene negative tumors

And in particular interesting is that the fusion-gene positives are quite similar to the normal samples, and only the fusion-gene negative tumors are outstanding. And this is even true if we deplete for copy number areas. It still remains the same

So, for validation, we then use the bisulfite conversion assay, which is the MeDIP-seq kit or assay, where we've got early access to

The advantage here is that you can really have a nucleotide-wise resolution of your bisulfite-treated DNA. In the MeDIP-seq experiment, you can have broad genome-wise look at your differential methylation. But, you don't get an information about single CPG regions

So, what we did here is we also shared the DNA and repaired it and added some ace [sp]. We then ligated the methylated adapters to our fragments, performed the hybridization that's similar to the enrichment, targeted enrichment assay we have also used quite a lot

Then after the enrichment, you perform the bisulfite conversion before you do a library quantification with PCR and then do a sequencing

For this protocol, we have used now the Illumina sequencing to be sure that we have a completely independent technology to prove our findings

And here, the outline of these assays, you're looking on--we are looking on quite a large region of the complete genome for 84 million base pairs are covered. And you see that there are lots of promoters covered and differential methylated regions which have been already published, mainly by iritsariat [sp] or--and also CPG islands, which are contained in this bisulfite-specific assay

So, we now didn't use again the prostate samples, but we thought we'd use a different system to validate our data. And we focused here on the--on two cell lines

One cell line, the VCaP cell lines are fusion-positive prostate cancer cells. And the DU145 cell lines are fusion-negative prostate cancer cells

And now, to confirm our previous data, we asked if we find a differential--an increase in differential methylation in the DU145 cells. Those would represent our fusion-gene negative tumor samples

And that's the output we had achieved from the sequencing. So, for the VCaP cells, we had overall 119 million sequence--base pair sequence and a large proportion, which was uniquely mappable. And on target, we were in the range of approximately above 95 percent

This is true for--we have subsequently more methyl-seq studies. And that's the average we achieved

If we are looking now on the regions which are covered, we --again, more than 90 percent of those regions, which are contained in the kit covered was a sufficient high coverage

This is a standard plot which we always used to see--to look on the uniformity of the coverage, of the targeted regions

And the only thing you have to take out here is we look if the different samples are--the lines are close together. This means that we have a high reproducibility and a high overall coverage over the samples. And this is quite nice

One cell is a little bit less well distributed. But, this is due because our sequencing capacity was not as high as it was for the others

We--to answer our question, if we really find more differential methylations in the fusion-gene negative cancers, we now investigated more in detail the differentially methylated regions

For example, here we looked on the--at the 1 kb region downstream of the transcription start site, 500 base pairs upstream of the transcription start sites and the CPG islands

And the level of methylation you find down here, and how many regions we find with the certain degree of methylation is depicted on the Y axis

You can see that we mainly get either no methylation at all or 100 percent methylation. But, the most important for us is that the red line is the DU145, meaning the fusion-gene negative samples

And since they are much higher here in the highly methylated regions, you can see that those cell lines have much more differential methylation or more methylation occurring than the fusion-gene positive cells

And this underlines our finding, which we've--which I first presented with the MeDIP-seq on primary tumors. We were now able to validate this with the methyl-seq on gene lines--cell lines, sorry

Looking on specific target regions here, we looked on the EZH2 target genes. You will understand why we looked on the EZH2 in a moment

And here, on the promoter region, and again, you see a much higher peak for the fusion-gene negative cell lines than compared to the fusion-gene positives

If we--of course, everybody wants to have a look on how one has to visualize or how one can imagine how the methylation looks like, and this is the UCSC browser picture. And for those who are not so familiar, down here, you see the location of the gene

Up here, there's the CPG island. Visualize this other probes from the Agilent kit. And down here, we have either visualized the DU145 cell lines or down here the VCaP cell lines

And this region as well as this region very nicely illustrates that you have for the gene-fusion negative cell lines much higher methylation rates than for the gene--fusion-gene positive cell lines

And the interesting is now to see if--how the MeDIP-seq approach, which I introduced in the beginning and the bisulfite sequencing compares. And therefore, we--I have included this figure here

Let me explain it for a second here. Down here, you have the same region as visualized with the UCSC picture. It's the chromosomal localization

And the gene localization we have for the HOXA9 gene. And those lines represent the methylation level after the MeDIP-seq experiment

And green is the DU145 cell line. This is another fusion-gene negative cell line, the LNCaP cell lines. And down here, the VCaP cells, just the fusion-gene positive cell line is down here

And the higher the peak, the more enriched fragments we find. This means the more methylation we have in this region. And you see that this region is very highly methylated in the MeDIP-seq experiment

And it's also represented in the methyl-seq experiment with the difference that we have here, an overall impression of the region, whereas here, we can pin it down to single CPGs

Okay. Well, did--this is right now where we are. We are increasing this emphasis and going into more detail to find out which regions are characteristic for the fusion-gene negative cells

And just doing a plain David [sp] pathway analysis brings up quite interesting gene terms, like homeobox genes proliferation, all things which are quite understandable and interesting to follow up in this setting

So, I'm coming now--first in the mechanistic region and more into the basic research. We have performed this project

We asked, "What is the reason for the higher differential methylation in the fusion-gene negative tumors?" And we looked on different methyl-transferases or chromatin-modifying enzymes

And we came up with EZH2, where we really see a significant increase in the gene expression level of the fusion-gene negative samples

We have performed on the same samples G-narase [sp]. So, this is the most significant gene coming up for the fusion-gene negative tumors

Now, the question arises, what is the reason for this high level of EZH2 in the fusion-gene positive but the even further increased EZH2 level in the fusion-gene negative tumors

Well, for the fusion-gene positive tumors, it's quite clear. It's the increase in the ERG expression. This is the fusion-gene TMPRSS ERG. It also drives the EZH2 expression

But, for the fusion-gene negative tumors, it was not clear. And we looked on the micro-RNA patterns, which we have acquired by co-PCR experiments. We looked in detail on those micro-RNAs which are known to target EZH2

And looking into detail, we find the pattern for micro-RNA 26A, which shows a significant decrease in this micro-RNA fusion-gene negative tumors

Interestingly, in this region, we find other significant differential methylation in the fusion-gene negative tumors. This led us to the assumption that the high methylation of this region leads to a down-regulation of the micro-RNA

And we proved this also with the 5-atsa [sp] treatment in the way that we treated the cells with increasing concentrations of 5-atsa. And we see an increase of the micro-RNA 26A after the treatment and the corresponding decrease of EZH2

We also performed an in vitro methylation assay, where we used this region which we thought is responsible for the down-regulation of micro-RNA 26A, cloned it in front of the luciferase gene, did in vitro methylation, and then performed the luciferase assay

And here, you see the unmethylated vector in comparison to the methylated vector. And the reporter expression goes significantly down if you have a methylation on your promoter

Okay. So, with this, I'm already coming to the summary. We have--for the fusion-gene negative tumors, we think that we can explain the appearance of the differential methylation that was by a decrease of the micro-RNA 26A expression due to a methylation increase of this locus. And this decrease of the 26A leads to an increase in the EZH2 expression

This is just to give you some colors and to model. The normal situation is that EZH2 is expressed in a certain amount. For the fusion-gene positive tumors, ERG drives the EZH2 expression, which is increased

But, in the fusion-gene negative tumors, the methylation of the micro-RNA 26A leads to a decrease of the micro-RNA 26A, which alleviates then the EZH2 expression, which then in turn can modulate the DNA methylation of the genome

Okay. And this is already the story I have provided to you. And this--most of the work is due to Stefan Boerno. He did the experimental side, and Martin Kerick and Axel Fischer, who were the bioinformatics people behind the project. And this was in collaboration with the Department of Pathology and Urology in Hamburg-Eppendorf as well as the DKFZ in the Heidelberg and the University Hospital in Innsbruck as well as the pathology and bioinformatics people in Connell University

Thank you very much