Broken Chromosome Repair by Homologous Recombination
Transcript of Part 3: Mutations Arising during Repair of a Broken Chromosome
00:00:07.21 Hi. 00:00:08.21 My name is Jim Haber. 00:00:09.21 I'm a professor of biology at Brandeis University. 00:00:13.16 I have been talking about how cells repair broken chromosomes, and I want to talk specifically, now, 00:00:19.19 about how the repair of broken chromosomes is actually fraught with mutagenic danger. 00:00:26.25 It turns out that although repair of breaks by using homologous recombination is 00:00:34.08 far more conservative than nonhomologous end-joining 00:00:37.12 -- that would just join random segments of DNA together, and translocations that you see frequently in tumor cells -- 00:00:44.00 it turns out not to be as accurate as normal DNA replication. 00:00:48.15 And understanding that has led to a great deal of understanding of exactly how repair 00:00:54.02 is taking place at all. 00:00:57.22 Okay. 00:00:59.05 So, the normal process of DNA repair that I've talked about in the previous videos is 00:01:06.12 illustrated on the left side of this screen. 00:01:09.11 A double-strand break is made; the ends of the double-strand break are resected by exonucleases; 00:01:16.13 Rad51 recombination protein is recruited; 00:01:20.15 and then a search for homology takes place, where base pairing is finally accomplished 00:01:25.15 between the invading recipient DNA and the donor. 00:01:31.20 And that base pairing facilitates, then, the recruitment of the new DNA polymerase. 00:01:37.09 And the polymerase starts to copy this template in order to patch up the double-strand break. 00:01:44.19 But this repair process is different from normal DNA replication, in that 00:01:51.00 the newly synthesized DNA is being unwound as it's being synthesized, 00:01:55.06 much in the way that RNA is unwound from DNA during transcription. 00:01:59.23 And the consequence of that is that this strand sometimes can be completely liberated. 00:02:05.16 In fact, it often is liberated so that it will pair up with the other end of the double-strand break. 00:02:11.06 That makes for an efficient repair event, where just a patch of new DNA synthesis has taken place. 00:02:17.06 But it's a dangerous point. 00:02:19.08 Because at the point where this has been liberated from its template, it's possible that 00:02:24.17 this broken segment of DNA will go somewhere else and participate in an illegitimate recombination event, 00:02:31.20 leading to some hybrid or mutagenic outcome that was not part of the original repair process. 00:02:39.28 And I'm gonna talk about events that basically hijack the system to lead to certain kinds 00:02:46.17 of errors in DNA repair. 00:02:51.27 Okay. 00:02:52.21 To learn about this in some detail, we used the mating-type switching system that 00:02:57.12 I've previously described. 00:02:59.12 This is a system in which a site-specific double-strand break made by a nuclease called HO... 00:03:03.25 analogous to the way Cas9 enzymes make double-strand breaks... 00:03:09.22 that it will cut one place in the genome, and then use a template to repair that break. 00:03:17.24 In this case, the template is a sequence called HMR, and the sequences that are normally inside HMR, 00:03:24.21 that would be used in mating-type switching, had been removed and replaced by a copy 00:03:30.17 of a Uracil-3 gene that comes from a different yeast called Kluyveromyces lactis. 00:03:36.15 And these sequences, although they're intact, are not expressed because the donor 00:03:41.03 in this situation is kept in a silent heterochromatic form, 00:03:45.20 which is the normal donor state of the HMR locus in Saccharomyces. 00:03:52.04 So, even though these sequences could be expressed, they're not expressed. 00:03:57.00 And the cells, as a consequence, are Uracil -. 00:04:00.19 If we turn on the OH endonuclease, 00:04:02.21 then these sequences are used and copied as the template sequences to go into the MAT locus. 00:04:09.01 And now these sequences are expressed, because they're Ura+, and there is... 00:04:13.04 they're no longer heterochromatic. 00:04:14.24 So, the cells can be easily screened to go from Ura- to Ura+. 00:04:18.24 They've successfully done this mating-type switch, or this analogous switching process. 00:04:25.05 But that means we could select for mutations. 00:04:28.04 And it's very easy to select for Uracil- mutations, because they become resistant to a drug called 00:04:33.28 5-fluoroorotic acid. 00:04:36.08 And so we could select for your Ura- outcomes that would be the result of some event, 00:04:42.17 switching event, that had a mutation. 00:04:45.08 So, we found a lot of these. 00:04:48.12 And a reviewer came along and said, but how do you know that these mutations didn't arise 00:04:54.08 already in the silent copy? 00:04:56.15 Maybe heterochromatic experience for the Uracil-3 gene would lead to a high level of mutation? 00:05:02.18 And we could show that wasn't the case, because if we inhibit this silencing 00:05:08.05 -- and you can do that by adding a drug called nicotinamide, 00:05:11.16 which inhibits the SIR2 histone deacetylase -- 00:05:13.27 we would unsilence this locus to become Ura... expressed, and these Ura- cells became Ura+. 00:05:22.15 Because this copy is mutant, arising during the repair process, but the donor template 00:05:28.17 is still intact and is Ura+. 00:05:31.02 So all the mutations that we are looking at arose during recombination. 00:05:35.20 So, we sequenced a lot of these. 00:05:39.23 Not surprisingly, a significant fraction -- more than half of the mutations -- 00:05:43.22 were base pair substitutions. 00:05:45.03 Here, they're illustrated above the DNA sequence line. 00:05:49.21 And they are single base pair changes. 00:05:53.25 The ones in black represent chain terminating, or nonsense mutations, that stop the translation, 00:06:01.00 but there are lots of base pair substitutions. 00:06:04.02 But then there were a lot of other mutations that were... that were quite different. 00:06:10.06 One very simple one is just that there are two copies of a sequence in a row, ATCATC. 00:06:15.04 And this becomes a deletion of three bases. 00:06:18.09 Somehow, the polymerases ignored or stepped over one copy of this triplet. 00:06:23.08 But there were a lot of these kinds of frameshift mutations. 00:06:29.04 About 30% of all the events were -1 frame shifts. 00:06:34.00 And that's interesting, first of all, because they all occurred in very specific contexts. 00:06:39.15 They weren't just random losses of one nucleotide. 00:06:43.07 Almost all of these mutations occurred in homonucleotide runs. 00:06:48.08 For example, here are 3 G's that become 2 G's. 00:06:52.07 And here there are 4 C's that become 3 C's. 00:06:56.02 So, what was happening frequently is that the polymerase was copying in a... in a homonucleotide run, 00:07:02.12 and just simply dropping out a base. 00:07:05.21 It's interesting also that there are no +1's. 00:07:08.08 And it turns out that this family of polymerases... and it turns out, it's DNA polymerase delta, 00:07:13.15 as I'll show you, that's doing all of this... this family of polymerases very frequently 00:07:18.06 will step over a base, making a -1 mutation, but it doesn't make +1 mutations. 00:07:26.16 But there were other, far weirder events than what I've just described. 00:07:31.07 And the two that I'm gonna focus on are the ones in the red boxes, 00:07:35.05 which are called quasipalindrome mutations, 00:07:37.16 and these long red lines that I will talk about some more. 00:07:42.09 The quasipalindrome mutations are a reflection of what I told you at the beginning, 00:07:47.18 which is that the newly synthesizing DNA is gonna fall off from its template. 00:07:52.03 And here, when it falls off from its template, it's capable of base pairing with itself. 00:07:59.02 Obvious... these are so-called quasipalindrome mutations because they're not entirely perfect, 00:08:05.06 as is illustrated here. 00:08:07.11 But if this folding over happens, and then you copy the newly synthesized strand, 00:08:12.24 you are effectively perfecting the palindrome. 00:08:15.05 And in the course, introducing complex mutations. 00:08:17.25 And we found a number of these quasipalindrome mutations, as a reflection of the fact that 00:08:23.26 the DNA polymerase that is copying the template is falling off, pairing with itself, 00:08:29.09 and then eventually getting back to doing its... finishing the job so that we could recover these events. 00:08:35.10 So, quasipalindromes are yet another reflection of the instability of this copying process. 00:08:42.20 And then there were the very weird events, where it turns out that a significant fraction 00:08:46.27 of the sequence didn't come from the original template. 00:08:50.22 And it turns out that it came from another copy of Uracil-3 on a different chromosome. 00:08:57.17 So, as I said at the beginning, we were using a copy of Uracil-3 that did not come from cerevisiae, 00:09:03.07 but comes from Kluyveromyces. 00:09:05.24 It turns out to be 72% identical to the Uracil-3 that Saccharomyces has, which is, 00:09:12.19 first of all, located on a different chromosome, and was not expressed because it is interrupted 00:09:17.13 by a transposable element. 00:09:19.06 And in a sense, we were lazy, because we didn't get rid of these sequences. 00:09:24.02 We didn't think anything would really happen. 00:09:26.16 And because we were lazy, we were a bit lucky, because otherwise we probably 00:09:30.27 never would have seen these interchromosomal template switches, which is what is going to happen 00:09:36.00 in these cases. 00:09:37.00 So, what's happening here is that, again, the newly synthesized strand is dissociating 00:09:42.02 from its template. 00:09:43.27 But here it is somehow going to a completely different chromosome, interacting with sequences 00:09:48.20 that it is only 72% identical with, copying some of that information, 00:09:53.16 and then in order to be complete, and for us to recover the events, it has to actually jump back. 00:09:58.20 So, it requires two jumps in order to be able to do this, and each one of them 00:10:04.21 involving only a 72% identical sequence between these two templates. 00:10:09.24 So, we call these interchromosomal template switches. 00:10:13.26 And they turned out to be surprisingly frequent. 00:10:18.27 Okay. 00:10:20.17 It turns out that all of these errors that I just described are the consequence 00:10:25.15 not of some special DNA polymerase... yeast cells, like mammalian cells, have a number of 00:10:31.11 so-called translesion or bypass polymerases, which they use in situations such as confronting photodimers 00:10:40.28 that are generated by UV, and in other circumstances. 00:10:44.15 But these are not the polymerases that are generating these mistakes. 00:10:48.21 The polymerase that's generating these mistakes is DNA polymerase delta itself. 00:10:54.00 And we found this out by taking advantage of a proofreading-defective mutant, 00:10:59.22 which I'll describe in just a moment as to what that does to DNA polymerase delta. 00:11:04.13 But in the absence of this proofreading activity, what we found were many more base pair substitutions. 00:11:10.22 But all those things below the line -- the -1 frame shifts, the quasipalindromes, 00:11:16.17 and the interchromosomal template switches -- are virtually gone. 00:11:19.20 So, it is the polymerase delta, the normal replicative polymerase, which is in fact 00:11:26.03 the agent of most of these mutational events. 00:11:32.00 Okay. 00:11:32.15 So, I have to say a little bit about how polymerase delta really works. 00:11:36.16 It turns out that polymerase delta actually does... has two importantly different conformations. 00:11:44.06 Every time it adds a base, it then rearranges itself to check whether the base it added 00:11:49.19 is correct. 00:11:51.00 And it will do that... and if it finds that the base is incorrect, if that base is 00:11:55.27 not properly paired, an exonuclease activity of the polymerase will then chew it away, 00:12:01.16 it'll go back to the original conformation and resynthesize the base. 00:12:05.26 So, every step in DNA replication has the potential to go back and forth between 00:12:10.22 these two conformational states. 00:12:13.03 If we delete the proofreading activity of this enzyme, it stays, we think, 00:12:19.25 locked in one of these two conformations. 00:12:21.28 And we think that that... this conformation is more processive and doesn't fall off 00:12:28.18 the DNA the way that we see in the wild type enzyme. 00:12:32.01 And as a consequence, it doesn't make frameshift mutations, and it doesn't make 00:12:36.17 quasipalindrome mutations, but it makes many, many more base pair substitutions 00:12:40.22 because it doesn't have a proofreading domain anymore. 00:12:43.21 And the beautiful experiment done by Peter Burgers and Dimitri Gordenin, which illustrates this, 00:12:48.14 is shown here. 00:12:50.06 So, here's a template. 00:12:52.20 It has a priming oligonucleotide to which the polymerase can bind. 00:12:57.12 And once the polymerase binds, it starts synthesizing DNA down the DNA. 00:13:01.24 But when it hits this blocking oligonucleotide, which is here, the replication stops. 00:13:07.05 So, we get this one very clear block in the synthesis. 00:13:12.27 The proofreading-defective mutation just keeps on going. 00:13:16.09 It goes all the way to the end of the template, which means it somehow can displace 00:13:20.22 the blocking oligonucleotide. 00:13:23.03 And I think that's a very clear indication that the reason that we don't see 00:13:29.20 quasipalindrome mutations, or frameshifts, or interchromosomal switches is simply that 00:13:35.03 this mutated enzyme stays on its original template and doesn't dissociate in the same way that we see 00:13:42.05 in these mutagenic events. 00:13:45.07 Okay. 00:13:45.23 So, we've studied these events in a little bit more detail. 00:13:48.13 Olga Tsaponina, who was a postdoc in my lab, created a 32-base pair deletion inside 00:13:55.07 the donor locus. 00:13:56.08 So, if this template is used in a mating-type switching event, which happens most of the time, 00:14:01.02 it produces a Uracil- cell, because the 32-base pair deletion isn't corrected. 00:14:07.12 And the only way to get a Ura+ plus cell out of this arrangement is to do 00:14:11.20 this interchromosomal switch and create a chimeric protein, 00:14:15.27 which is partly Kluyveromyces and partly Saccharomyces, 00:14:19.16 where the 32-base pair mutation has been corrected by copying from this 72%-identical sequence. 00:14:30.04 One thing that Olga learned by doing these experiments was to answer a question that had... 00:14:36.18 we'd been wondering about for some time. 00:14:38.23 We wondered whether the frequent events that we were seeing, where the polymerase 00:14:44.11 fell off the DNA, if you will, was because the donor locus was held in this very highly heterochromatic state, 00:14:52.05 where it uses very highly positioned nucleosomes across this region to 00:14:57.22 prevent its transcription. 00:14:59.06 And we wondered whether polymerase banging into these highly positioned nucleosomes 00:15:03.24 was the reason that the... that we were seeing such a high level of these mutational events. 00:15:08.23 But that turns out not to be the case. 00:15:11.02 Because here, we could unsilence the locus using a mutation that prevents the silencing 00:15:16.03 and the well-positioning of these nucleosomes, and the result was that the frequency 00:15:20.20 didn't change. 00:15:21.20 So, it isn't that the polymerase is being displaced by the highly positioned nucleosomes. 00:15:27.03 It's an intrinsic instability of the polymerase itself. 00:15:32.12 Okay. 00:15:32.26 So, these sequences are only 72% identical. 00:15:36.25 And the question is, when the jumps happen, where do they go in and where do they come out? 00:15:42.16 And so we measured these questions in some detail. 00:15:47.03 The amount of DNA that's being copied turns out to be about, on average, 00:15:53.21 250 to 300 bases from the... from the second template. 00:15:57.24 But it's not... it doesn't start at the same place, and it doesn't always stop at the same place. 00:16:03.10 It's pretty well normally distributed. 00:16:05.13 And if we look more carefully at these events, as to where the strand invasion goes into 00:16:13.03 the second copy on the other chromosome, you can see that it goes into many different places. 00:16:19.14 And if you choose one of those places where it goes in to the template, then the question is, 00:16:24.21 how does it come out on the other side of the 32-base pair deletion 00:16:29.08 in order to do the correction? 00:16:30.14 And the answer is it comes out all over the place. 00:16:32.28 So, there's no preferential place where it starts, and there's no preferential place 00:16:37.05 where it ends. 00:16:38.25 And that raises a question as to what exactly is being recognized by this polymerase. 00:16:45.02 And so one thing that we... we wanted to know is, well, that's... we had been studying these 00:16:50.26 with only 72%-identical sequences. 00:16:53.28 What would happen if we made them identical? 00:16:56.13 And what we discovered is that the rate goes up 10,000-fold. 00:17:00.03 And now, 3 out of 1000 of the... of the switching events that we recover have actually done 00:17:06.22 this interchromosomal switch. 00:17:09.04 And what that means to me is that this polymerase is very frequently falling off the DNA. 00:17:15.11 Now, we have facilitated it finding a partner, because the template is 100% identical, 00:17:21.15 and then it will go back. 00:17:23.02 And again, the template is 100% identical. 00:17:25.15 And under these circumstances, we see this very high rate of these events, which I think 00:17:30.01 means that the repair polymerase is dissociating very, very frequently from its template. 00:17:40.24 So, let me just say one more thing about these microhomologies. 00:17:44.20 So, whenever it goes in or out, we could define exactly where it goes in and out by 00:17:50.23 the matches between the sequences that it was copying and the sequences where it continued. 00:17:57.06 And some of... and so these define what are called regions of microhomology between 00:18:02.14 the divergent sequences that are being used in the repair event. 00:18:06.04 And some of these sequences are being used very, very frequently. 00:18:09.28 But other sequences, which are almost as long, are being used very infrequently. 00:18:15.09 And we don't yet understand the rules as to exactly what sequences are being used to go 00:18:21.16 in or out of these sequences. 00:18:24.09 But we think it's a fundamentally important question... is to understand exactly how microhomology 00:18:30.20 is being used in this repair event. 00:18:33.13 And one of the reasons that I think it's very important to understand microhomology-mediated events 00:18:39.06 is a phenomenon in human cells called chromothripsis. 00:18:43.24 So, I... if you've seen karyotypes of tumor cells, you can... you can stain chromosomes 00:18:51.02 and see where different segments of DNA came from. 00:18:55.23 But the surprising thing is that in some cases 00:18:58.24 there's no obvious rearrangement of the chromosomal material, 00:19:02.11 because everything comes from one chromosome. 00:19:05.07 But this phenomenon, called chromothripsis, was discovered by Peter Campbell's lab 00:19:11.07 only in 2011. 00:19:12.07 It seems astonishing that something as fundamental as this could be not known. 00:19:15.18 But it could not be known until DNA sequencing was sufficiently robust to be able to 00:19:20.21 really see these kinds of events. 00:19:22.27 And what Campbell's lab first described was chromothripsis, which means chromosome shattering. 00:19:29.28 And what was happening was that one chromosome was being broken apart into dozens, 00:19:34.27 if not hundreds of pieces and rearranged, but all within the same chromosome. 00:19:39.21 And this of course could cause translocations, deletions, all sorts of crazy events 00:19:47.21 that we did not appreciate happened at all. 00:19:50.06 And the bottom part of this just illustrates this with one chromosome. 00:19:54.05 This happens to be the one highly, highly rearranged chromosome in the... in the chromosomes 00:20:00.01 of Henrietta Lacks, the woman whose HeLa cells have been used in probably 100,000 publications. 00:20:09.28 And it turns out that Lacks' chromosome 11... one copy of her chromosome 11 shows examples 00:20:16.25 of this incredible rearrangement of chromothripsis. 00:20:20.10 So, what we would like to know is, how do these events occur? 00:20:25.07 One easy way to explain these events is that they're just being joined together by taking 00:20:30.15 all the shattered pieces and putting them back together by nonhomologous end-joining. 00:20:35.04 But another possibility, which has been suggested by Phil Hastings and Jim Lupski, 00:20:42.10 and also by Gilles Fisher and Bernard Dujon, 00:20:48.11 is that there is something called microhomology-mediated template switching. 00:20:52.23 And the idea of these events is that... that it is the repair DNA polymerase which 00:20:58.21 starts to copy some region, and then jumps, sometimes a megabase away or even further, 00:21:05.10 to another place, picking up some sequence, and then going to another place and going to another place. 00:21:10.17 And stitching together a rearranged chromosome, which, again, has very small junctions, 00:21:15.24 microhomology junctions, but is a replicatively dependent process. 00:21:21.00 And we think, in fact, that the DNA repair events that we are looking at fall into this category. 00:21:27.15 I think that understanding microhomology-mediated template switching 00:21:31.03 is a fundamentally important question. 00:21:33.18 And the yeast system that I've described provides insights into what is really happening 00:21:40.00 during these template-switching events. 00:21:42.20 I've spent a lot of time talking about the individual steps in DNA repair. 00:21:47.11 We've done this mostly in the budding yeast Saccharomyces cerevisiae, 00:21:51.16 a simple unicellular organism. 00:21:53.24 But the principles of repair that we have learned about turn out to be shared by 00:21:59.09 all eukaryotic organisms. 00:22:00.26 There are differences in details 00:22:02.27 -- mammalian cells have some different proteins from those yeast has -- 00:22:07.11 but the fundamental processes are really remarkably conserved. 00:22:11.13 And it becomes even more important to understand some of these steps in detail, 00:22:18.07 because all of the new gene editing techniques using site-specific nucleases 00:22:22.18 -- Cas9 in particular -- 00:22:25.09 also rely, basically, on these very similar kinds of repair mechanisms. 00:22:30.17 So, understanding them, and the dangers of associated mutations associated with 00:22:36.02 these processes will turn out to be a very important thing to follow up. 00:22:41.05 Thanks very much.