Scientist Stories: Mike Snyder, Noncoding Regions in Cancer and Silencers

Hello, my name is Michael Snyder, and I'm here to tell you about some of our recent work to characterize silencers in the human genome. This is a very understudied area in genomics, and the work I'm going to tell you about today is the work of a talented postdoc in the lab, Baoxu Pang, who now has his own lab in Leeds. So, as I'm sure you're all aware by now, most of the human genome is not protein coding. In fact, only about 1.5% of the genome encodes protein coding genes, and the rest is thought to be non-coding elements, including regulatory sequence. There's been considerable attention paid to trying to characterize the activators of gene expression. Specifically, there's a lot of work to characterize enhancers, such as distal enhancers that might lie far away from genes, enhancers that lie near the promoter regions, and promoter regions themselves. So there's been lots and lots of work to study these. Many— actually, now millions of elements have been identified that are involved in activating gene expression, or at least thought to be activating gene expression. But if you think about it, there's lots of the genome that's not expressed, and some of it's actively shut off. Some of it might be just through more general mechanisms, heterochromatin, but probably a lot of it's regulated regions that are actually characterized by proteins such as silencers. And so we set out to carry out a systematic study of silencers in the human genome. The way we did this is kind of complicated. I'm going to give you— try a simple explanation. I know it's a complicated slide, but we decided to actually look for open regions of the genome that might encode silencing. So you can actually isolate them by a method called FAIRE, where you can pull down the regions, the open chromatin regions of the human genome that will contain both activators and silencers. And we would take these fragments, fairly small, they're only about 200 kb, and we would clone them upstream of a selection system. And this is where it gets complicated, but the way the selection system is set up is that if we turn off a downstream gene caspase, we can actually basically stop apoptosis. So we made this giant library, a very large library of short fragments, and actually integrated it into the genome. And then what we do is we actually are able to induce induction of gene expression. And if you have a silencer, you actually keep that gene off, and those are the cells that survive. They don't undergo apoptosis. And so you actually kill all the regions that aren't silenced, and what you can do is select for cells that actually have silencer fragments. And then what we do is we actually sequence those silencer fragments using a PCR method, and then we compare it to the starting library, and wherever we've cloned the silencer, there'll be a significant enrichment in those fragments upstream of this region that's turning off the apoptosis gene. So the bottom line is we can select for cells that live because they've had this negative selection, this apoptosis gene shut off. And so we did this and we actually carried out several times, and what we came up with was that fairly high stringency of roughly about 2,600 silencers. These are all done in a cell line you've probably heard of a lot by now, K562. It's used a lot in the ENCODE project, and basically we found over 2,000 silencers, and they're scattered throughout the human genome. And in fact, it turns out many of them lie upstream of genes or in intergenic regions. A fair number are also in introns, and interestingly, they're in other places of the genes as well, including in genes themselves in some occasions. So the point is we actually find silencers, they're generally throughout the non-coding region of the human genome. We've actually validated whether they're truly silencers. We take them, we clone them upstream, with a very active promoter. It's called PGK. It's upstream of a luciferase construct, and this is in a transient transfection. And the silencers for the most part will actually shut off gene expression in this reporter assay, and occasionally there'll be one that fails, but most of them, I think there were 90% of them, succeed. We've also deleted them from the genome itself. So in a more endogenous assay, you can go in, use CRISPR-Cas9, actually remove the silencers, and basically these are 3 regions that are really strongly shut off, at least these 2, this one's actually expressed some. When you go in and delete them, these are 3 independent clones, basically the genes can activate, in some cases well over 100-fold relative to background. So the point is that these are in fact true silencers, they're shutting off the genes in cells. We actually start characterizing them. What's special about these silencers? Well, first of all, the ones in K562, you look over here, they're actually enriched for interesting histone modification called H4K20 monomethylation. It's actually not very well studied and its role is thought to be complex. It's not 100% clear what's going on here, but it's been implicated in silencing and activation.. But clearly in our assay, the silencers we have are highly enriched for this interesting mark, and there's some other interesting marks as well. We look at the motifs that are enriched by these silencers, we find things you might expect. We actually find AP2, which has been implicated in silencing in other studies. Same with this other one, KLF12 has been implicated in silencing. But then we actually found a new motif, so there's other silencers to be discovered out there, we think, as well. And so there's a lot of interest in doing that. We've also looked to see whether what it looks like for silencers in other cell types. So we actually took the exact same library, even though it was isolated from open chromatin in K562 cells, and we transfected it into HEPT2 cells. And we in fact pulled out another set of silencers, and the overlap's very, very small. So in fact, what we think is that the silencers are very tissue-specific. If they were the same silencers over and over again, they would get used many times. Very, very few that are shared. We can show statistically this is highly enriched for being tissue-specific. We've looked at the pathways these silencers are involved in, and let's just use HepG2 as an example. It turns out that the silencers are preferentially rich up here in neuronal genes and things like that. These are genes that would not be expected to be expressed in HepG2 cells. There are also, in the case of K562 cells, most of these pathways are not expected to be expresses well, actin cytoskeletal, and so on. Well, I have to admit, this cancer one I might have expected, so that was a little bit of a surprise, but it must be a very specific set of cancer genes. So the point is, these silencers are upstream of regions you don't expect to be expressed. We actually wanted to get a little more insight into the functional role of silencers, so we looked at two of them that were upstream of these drug transporters, ABCC2, and ABCG2 are two drug transporters that are actually known to remove anti-cancer drugs. They're actually involved in cancer resistance. And so, we discovered that in fact we had silencers that were upstream of each of these genes. And so, what we did was we went in and knocked them out. First, we tested them in luciferase assays, and sure enough, they were silencers. They do shut off gene expression. We also tested them in knockout experiments, so we knocked them out.. And what we find is that they will induce gene expression, the expression of neighboring gene. If you knock them out, this is for the ABCC2, this is for the ABCG2, in both cases when we knock out the silencer upstream of them, the genes become highly active. We've also then looked at sensitivity to drugs, and sure enough, it turns out that in each case you will see increased survival, meaning the transporter is not as shut off, and so actually when you knock them out, the genes become active, and so they'll actually crank out the drug, they'll knock it out. This is doxorubicin, well, a cancer drug, and here's two others. So the point is, in each of these cases, when you knock out the silencer, the gene becomes active and you can transport out the drug, and the cells become more resistant. So actually, this is really important then for functional assays for cancer resistance, it turns out. It turns out that silencers, we think, work across domains. So just drilling in on this one a little further, this is the ABCC2 gene, and here's the silencer here. And if we knock it out, we could show that not only does this gene go on, but then we actually see some other genes. Here's this one, CP1. Actually, so here's ABCC2, jumps way up in expression when you knock it out. Here's CPN1, it also jumps up in expression. Even this one jumps up a little bit. So all these genes in the vicinity of the silencer, they all go on. So we think it's having a general effect. And even this one next door goes on a little bit as well. So again, we think silencers actually operate over long regions. Not only that, they can work over long regions of the genome as well. Here's a case where for 3 different silencers that can act at long distances, we showed— I'll just show you one example here. There's a silencer here, and it turns out when we knock it out, it's thought to have 3D interactions with interactions up here, over here, which again, this is a long distance away, hundreds of KB, and it turns out this is upstream of this norexin gene NRXM2. Here's another one, this RAS GRPB2 gene. And when you knock out the silencer, these genes, this one here and this one here, which lie way upstream, will actually now go on in expression. So silencers, we think, can operate over long distances. It's true for this locus. We actually showed it for two other loci as well. So the bottom line is we think silencers are pretty cool. We set up a method to identify them. We showed there are thousands of them throughout the human genome, and they're in intergenic regions, introns, and other places. Like enhancers, we think many of them are tissue-specific. There may be general repressive ones as well, like heterochromatin, that we have not necessarily identified from this. These are open chromatin gene turning-off silencers. We validated many of them through the surfrase assays and knockouts. We can show that they actually are important for regulating, for example, drug resistance, and that's really relevant for cancer. We can also show they operate over long distances. Once again, I want to impress upon you that this is the work of one postdoc, Baoshu Pang, who actually did this beautiful analysis of these silencers, a relatively uncharacterized region of the human genome.