Fourth Annual DNA Grantees' Workshop
Tuesday, June 24, 2003
AFTERNOON SESSION
Research Update Briefings: Ongoing Projects
William T. Vosburgh, Moderator
Biography
DR. VOSBURGH: I think we've had a lot of stimulating talks and great crossover between those practitioners of forensics and those who do the research here.
My name is Bill Vosburgh. I'm the director of the Forensic Services Division of Prince George's County [ Maryland ]. For those of you who don't know where that is, it's the eastern suburb of Washington, D.C. We are the 32nd largest police department in the United States.
Assessment of Damaged DNA Templates
Jack Ballantyne
Biography
(Note: The PowerPoint presentation that supplements Dr. Ballantyne's transcript is not available.)
DR. VOSBURGH: It's my distinct pleasure today to introduce two people who are very well known in the field of forensic science. First up, I'd like to introduce Jack Ballantyne. Jack will be speaking today about damaged DNA, something which he personally knows a great deal about.
DR. BALLANTYNE: Today I want to talk about the assessment and in vitro repair of damaged DNA templates.
First of all, I'd like to thank the person who's done all of the work for this, Ashley Hall. She's a Ph.D. student in my lab and literally all the work that I will talk about today has been done by her. Yesterday, John Butler talked about the Y papers that have just come out. It's going to be interesting to now correlate the Y sequence with this Y genetic map.
For convenience purposes, when I'm talking about damage to DNA, I'm referring to a change in the covalent structure of DNA. That can either be degraded DNA—that is, damage brought on by double-strand breaks, which causes DNA to migrate on a gel to a high molecular weight—or damage caused by other covalent modifications.
Why bother with this? Obviously in certain cases, DNA may be intractable to analysis. Maybe there's a problem, perhaps you get a sample and try to isolate the DNA, but you can't get a result. What causes this? Well, there are several possibilities:
- The stains could be insoluble. There may be alternative extraction methods that could be used for that.
- You could have coextracting inhibitors. There may also be remedies for that.
- It could be just a low copy number of DNA templates, either starting template or an effectively low copy number based on other reasons. Again, there may be remedies, like whole genome analysis, for that low copy number down the road.
But one other problem is that the DNA could just be damaged. At the present time, there are no remedies to deal with just damaged DNA. The objective of the project that I'm going to talk about, then, is to ascertain the types of damage that may be encountered in forensic stains and eventually repair the damage that's been done.
In the past few years there have been a lot of advances in the field of DNA repair. For example, the Human Genome Project has shown that there are approximately 150 DNA repair enzymes in the human genome, and nature has found many ways to repair DNA. The concept is to try and use in vitro biochemistry to repair any damage to the DNA. But first of all, we want to find out what kind of damage is done, other than just the double-strand breaks that we see quite often.
So I just want to talk briefly about what DNA damage is, what types of damage take place in DNA, the experimental strategy for DNA damage assessment, results so far, and the future. I will talk about some preliminary data we have on the actual repair of DNA damage.
First, I want to summarize the types of DNA damage that can take place. First of all, you can have oxidative damage. When oxidative damage takes place—often as a result of the presence of reactive oxygen species formed through a variety of mechanisms—you have a variety of products formed from amidol primidines, 8-hydroxy guanine, and others. Thiamine glycol also is very common. There's actually a very large number of oxidative products that can be formed.
Ultraviolet (UV) damage results in the formation of photo products. These are the pyrimidine dimers and 6,4-pyrimidine primidone products (6,4-PPs) that I'll talk more about later.
We also have hydrolytic products, either depurination or deamination. Depurination results in the loss of the base, the so-called apurinex side, and chain cleavage. With deamination, the amine group is deaminated, for example, a cytosine will go to a uracil.
There are other types of damage. You have chain breaks, either double-strand breaks or single-strand breaks. You also have cross-linking that could take place, with either DNA-DNA crosslinks or DNA-protein crosslinks. And the final type of damage is adducts. This happens when bulky groups, such as methyl groups and alkyl type groups, are commonly added to the bases.
Of these types of damage that could take place, which are relevant to forensic science?
Initially, we decided to concentrate on the area of DNA damage done by UV light, because sunlight is a common environmental insult. The biochemistry of UV damage is well characterized in model systems. Based on the relevance of sunlight to skin cancer, quite a lot of work has been done in this area. So, as a proof of concept, we decided to start with UV damage: Can we detect UV damage and, if so, can we repair it?
DNA exposed to UV light can either form the pyrimidine dimer commonly found in textbooks, thiamine dimers, or cyclobutane pyrimidine dimers (CPDs), not necessarily only two TTs. There can be other primidines. The other product is a 6,4-PP isomer. Structurally, 6,4-PP contains a sigma bond in a C6–C4 arrangement. The cyclobutane ring distorts the helix. These balky lesions will cause DNA polymerase to stall and perhaps fail in terms of a polymerase chain reaction (PCR). In the literature, there's also some indication that UV radiation can cause strand chain breaks.
The experimental strategy was to concentrate on UV damage. We're going to look for photo products, strand breaks, and also crosslinks if possible. The method we're going to employ is lesion-specific endonucleases. There are endonucleases that we can obtain that are very specific, that nature has designed, and that various organisms use to repair DNA. These endonucleases can recognize various lesions, and I'll talk about that in a second.
We can use gel electrophoresis when looking for single-strand breaks or double-strand breaks. Double-strand breaks can be detected with just a native gel, and we can use denaturing agarose gels to detect single-strand breaks. Also, we're employing genome white scans using an ALU type system. Approximately 10 percent of the human genome comprises ALU sequences. Therefore, any damage detected in the ALU would be representative of general DNA damage. Of course, we can look at typability of the sample because STRs (short tandem repeats) are widely dispersed on different chromosomes.
The strategy of using lesion-specific endonucleases is as follows. We can damage the DNA and have no endonuclease treatment or have a lesion-specific endonuclease treatment. Once we do that, we can run the samples with and without treatment on the gel. We can look for double-strand breaks on native gel electrophoresis and single-strand breaks using alkaline gel electrophoresis and determine whether or not these are caused by the photo products.
If lesion-specific endonucleases—in this case specific for CPDs or 6,4-PPs—are present, then the endonucleases should cut the DNA, and you should get a difference between these samples. You should be able to see the formation of single-strand breaks because the endonuclease cuts at the lesion.
Also, these endonucleases have lysis activity that cause single-strand breaks. We can detect them by agarose gel. If a break is just caused by UV light, then we just look at the no-endonuclease treatment to see what damage has been done, because the damage may not only be from photo products.
Now, there are a variety of lesion-specific enzymes available. The UV ones are shown here and we're going to be using these. There are ones specific for CPDs. One is this yeast enzyme UV-damaged endonuclease, which is specific for both CPDs and 6,4-PPs. Using this one tells us something about the presence of the 6,4-PPs.
We also have oxidative-specific enzymes that we can use to determine damage, and although we've tried some, I'm not going to talk about them today. Enzymes are also available for hydrolysis that, if they can be incorporated into proper assays, can be used for any type of DNA-damage products.
So let's look at the damage that UV light can inflict on DNA. Here we have what we call naked DNA exposed to UV light. Basically, we take DNA that's in solution and expose it to UV light, run the DNA on a native gel and an alkaline gel, and compare the typability using autosomal STRs.
What you see here is the native agarose gel and the time course, which goes from zero minutes up to 48 hours, for exposure to UV light. You have high molecular weight DNA up to 25 minutes and then you start to see degradation. These are double-strand breaks beginning to appear. So you have these so-called high molecular weight DNA bands and then double-strand breaks start to appear.
The profile is obtained after 1 minute, after which you start to lose the profile. The partial profile is indicated by parentheses. This is using the Profiler STR system.
After 16 minutes the profile is gone. Now, that 16-minute sample, if you look at it, appears to have a high molecular weight band present. If you're running the DNA on native gel, it would look like a nice DNA sample. But when you type it, it doesn't work. There's no typability.
When you look at the same samples of naked DNA on an alkaline gel, you immediately see (i.e., at zero minutes) the high molecular weight band. But you also start to see the formation of single-strand breaks with increasing significance as time progresses.
Although this 16-minute sample apparently has high molecular weight DNA, when you look at on the alkaline gel you can see there's a very large number of single-strand breaks, and the particular flux is given here. From this, you can basically see that the DNA is progressively damaged in a short period of time by UV light and you lose a profile. That loss of profile appears to be correlated with the presence of single-strand breaks. But if you were to extract DNA and run a U gel (i.e., the old-fashioned way), you would see a high molecular weight band and think that the DNA was reasonably intact, when in fact it has large numbers of single-strand breaks. Here's a profile at zero time point. At 4 minutes, you start to lose the profile, and after 16 minutes there's no profile.
Now, let's go back a minute. If you look at this with an alkaline gel, you see, even after the shortest time period, some DNA fragments up here [indicating] above the region where you would expect the high molecular weight DNA, but they eventually disappear. That's a characteristic of crosslinked DNA. It would have an aberrant mobility on the gel, especially if you see it in high molecular weight compared with your standard. Obviously, there may be some crosslinks down here [indicating], but you can't detect them. Up here [indicating], however, it's highly indicative of crosslinked DNA.
So the low UV doses cause single-strand breaks and interstrand crosslinks, and these may or may not contribute to the loss of the DNA profile. Once you get beyond an hour or so, the double-strand breaks form. Now, of course, there's not enough energy in UV photons to cause double-strand breaks directly. We think, and it's consistent with the data, that the number of single-strand breaks is sufficiently large enough on both strands that they actually appear as double-strand breaks.
Now, that's the story on naked DNA. If you take DNA of a blood stain and you do the same thing (expose it to ultraviolet light, extract it, and run it on the gels), then this is what you get.
On the left-hand side is the native gel and here's the alkaline gel [indicating]. You see that the profile is still obtained up to a much longer period of time (8 hours). Then you get a partial profile that eventually disappears after 102 hours. The DNA looks relatively intact as high molecular weight DNA. You can see some degradation, but nevertheless it's still high molecular weight, despite the fact you've lost the profile. So with this 102-hour sample, you've lost the profile, but it still looks intact.
When you then look at it using the alkaline gel, single-strand breaks appear very, very quickly. Here are the single-strand breaks at 4, 8, 12, 24, and 48 hours. We didn't show the later time points. A lot of single-strand breaks appear after just a few hours with stains.
Stained DNA or double-strand breaks are not present in significant quantity with UV exposure up to 102 hours. There was a small but not a tremendous amount of smearing, not a lot of double-strand breaks.
The single-strand breaks gradually increase between 4 and 48 hours. Crosslinks may be present, and a partial genetic profile appears and remains constant during this timeframe but disappears at 102 hours. In the stained form, the kinetics of the process are different which means it takes longer to get a damaged product. So there is some protection afforded by the stained DNA as opposed to the naked DNA.
Just out of interest, then, we wanted to see whether the dehydrated nature of the stain is what's protecting the DNA from damage. We took the same DNA, dehydrated it, and did the same experiment. We exposed it to UV light, brought it up in solution, and then ran it on gels. This time, we found that you could get a full profile up to 12 hours but then the profile would be lost. You could see the high molecular weight DNA, but then things would degrade significantly from there.
Again, when you look at alkaline gel, single-strand breaks appear very quickly. So, single-strand breaks are evident very early. Double-strand breaks don't appear until later, at which time the DNA appears highly degraded. The crosslinks are not as evident, but the profile is not lost until 24 hours.
With this early data, the genetic profile is totally lost in naked DNA after 16 minutes, in dehydrated DNA after 24 hours, and in stained DNA after 102 hours. Therefore, DNA is protected from damage by the presence of cellular constituents and significantly by dehydration. Of course, DNA is not just dehydrated. It's also in a nuclear protein complex and chromatin, which will obviously afford some protection, or at least there's some evidence that it affords some degree of protection. The fact that dehydration and nuclear protein afford protection is good news from a forensics standpoint.
So that's just the general kinetics of when you treat a sample with UV light. What about the photo products? We found single-strand breaks appear in very large numbers, while double-strand breaks are not a factor. Single-strand breaks mainly appear very early and appear to be correlated with the loss of profile.
Next, we wanted to look at the bipyrimidine photo products. For this we used lesion-specific endonucleases like this particular endonuclease from the Chlorella virus. It recognizes CPDs. Here you have samples with and without the presence of the enzyme. This is an alkaline gel to show the presence of the single-strand breaks that would occur. If the enzyme cuts at a lesion, we can detect that lesion by the presence of single-strand breaks. Eventually, single-strand breaks appear in the enzyme-treated sample, indicating that you have in fact formed CPDs. They do occur quite early, for example, 5, 10, 20, and 30 seconds and so on. The further you go, the less difference you see among time points. So CPDs are being formed very early on in naked DNA, which is consistent with previously published data.
You can do and see the same thing with another endonuclease, T4 endonuclease V. There's a big difference between this set of data and this set, but the T4 recognizes the same lesion, so it's consistent.
So CPDs are formed within 5 seconds and saturate around 30 seconds. We can see much more activity after 30 seconds. The 6,4-PPs do increase linearly with Dewars using the UVDE enzyme.
Now, when we do the same thing with stained DNA, we don't see a lot of differences among the enzymes from 4 to 79 hours in both the untreated and enzyme-treated control. The same is true, for the most part, with T4 endonuclease V. There might be a slight difference, perhaps a lower molecular weight, but it's not significant in terms of the resolution of this gel. Similarly with UVDE, there's no significant difference visually or on the gels, to show that the formation of these photo products is taking place at any rate in stained DNA.
Therefore, we say that there is no evidence of significant quantities of UV photo products from the stains exposed up to 79 hours. That's not to say they're not being formed, but we don't detect them at this gross level compared with the detection of single-strand breaks and crosslinks. The typability normally is affected after 12 hours and disappears after 102 hours, so we cannot rule out or exclude the possibility that CPDs are having an effect, as they probably do. In stains, however, it appears that single-strand breaks play a much greater role. There's also no evidence of significant quantities of UV photo products formed in dry, naked DNA. Again, the dehydration may protect the DNA against the formation of photo products.
So basically, there are not a lot of photo products being formed in stained DNA when compared with naked DNA. We've also seen in the dehydrated state that there's been some protection afforded to typability in the previous experiments.
We think that the structural form of DNA changes when it's dehydrated. It goes from the B form to the A form. The A form is a different structure, a different tilt, a different axial rise, thus, different coordinates. Perhaps the structure is sufficiently changed such that the absorption of photons from UV light is more efficiently transmitted, allowing the UV photons to be absorbed by the DNA, normally UVC and type B, and CPDs are formed. It's likely that the structural integrity in the A form is such that CPDs are formed less efficiently. That's our hypothesis for what's going on there.
We've also looked at various storage conditions in environmental samples, for example, dried blood stains in a paper envelope and plastic bag; wet blood stains in a plastic bag; different temperatures, 4, 25, and 30 degrees; and different lengths of exposure, from 1 day to 30 months. In terms of the ability to type, there were no significant differences in the profiles of samples stored at any of these conditions. However, these samples are not exposed to the three things that we don't like in DNA: heat, light, and humidity. In the lab setting, we used UV light but didn't expose the samples to heat, humidity, or bacterial influences. It's likely that adding to these conditions could change things.
The gel-based assays that we developed are relatively insensitive, as you can imagine. But we're also developing assays that are more sensitive. I'll talk about that quickly.
We've also looked at environmental samples: exposing dried blood stains to the environment in a glass tank to prevent rain up to 30 months and using autosomal STRs and other stains on clothing exposed for months, in this case using Y–STRs because they were available. You can lose a profile, for example, at 6 months.
This is a sample that was exposed to the environment in a glass tank, which is basically just UV light, heat, and humidity and no rain. You get partial loss of the profile at 3 to 6 months and complete loss of the profile after 30 months. It takes a long time under these conditions to lose the profile. Trying to ascertain the damage to DNA under these conditions using these gel-based assays is very difficult because you need very large quantities of DNA to run these gels.
So we have been working, of course, on a more sensitive assay—the ALU-mediated detection assay. ALU is present in 10 percent of the human genome. Structurally speaking, the ALU is composed of several T regions that could form thymidine dimers. If you could ascertain damage to the ALU, then basically you could come up with a genome-wide damage assessment scheme.
The strategy is, if you can use a lesion-specific endonuclease on a DNA sample that will create a single-strand break, then you can use S1 nuclease to cut the other strand to form double-strand breaks. Then, if you were to do normal PCR (polymerase chain reaction) amplification, you would get a reduction in the product formation. So in terms of the PCR product, you would get a reduction in these products if the lesion was present. Linear amplification is another way of doing it, but we found it to be too insensitive. We were hampered by the fact that we didn't have real-time PCR. For us, we had to do all of this basic work just to get the assay working, so we could then transfer it to the real-time PCR system.
For example, here's a naked DNA sample and here is the ALU signal. As you increase exposure to UV, you get a reduction in the ALU signal. After 4 minutes you start to get a reduction, and at 60 minutes, it's gone. You get the same thing when you look at dehydrated DNA.
We didn't see a lot of reduction in the ALU sequence in stained DNA, and we think that's because of its sensitivity, so we bypass that. With the ALU, the lesions are stalling the polymerase, making it difficult to compare the ALU peaks before and after. We need the real-time PCR, which we're about to install. We're trying to alter the lesion bypass properties of DNA. There are polymerases that will help bypass the lesions. We can also use lesion bypass enzymes to solve this problem, and it's called trans-lesion synthesis.
Over the past few years, a number of DNA polymerases have been developed to directly bypass lesions, either with some error or with no error. These are going to be very useful in forensic science. So now, I want to talk about the preliminary data that we have on DNA repair.
If you take naked DNA, treat it with UV light, and look for the ALU, you get no signal. If you then—using typical TAQ gold—add a distributive, not a processive, thermostable trans-lesion polymerase, the polymerase will keep going until it falls off. Trans-lesion enzymes have evolved to come and help when there's a problem. They help the polymerase bypass the lesion. Once passed, they get off and let the real polymerase take over. If not, the lesion would create too high of an error rate.
If we take DNA that is damaged by UV light and add it to a PCR reaction that contains a thermostable translesion polymerase and enzyme, then you will get a signal.
Some organisms use photolyase—a way in which nature has decided to repair UV damage and other types of damage in direct reversal to that of humans. The system catalytically breaks the CPDs and 6,4-PPs and actually repairs the lesion directly. It's significant, and we're very excited by the fact that it can happen because this is an alternative way of looking at repair.
In the future, we want to expand the use of lesion-specific endonuclease to detect different types of damage, especially oxidative damage, which I didn't describe, and refine this ALU assay because it's got to be more sensitive than the gel methods. We need the real-time PCR for that, of course. We need to look at the effects of UVA and UVB (two types of UV light), which are more physiologically relevant.
We also want to look at comprehensive storage environmental studies to detect all possible lesions. Well, we have the samples, now it's just a matter of applying the more refined assay to them.
And of course, on the repair side we want to use the photolyase system. Single-strand break repair can be done by DNA polymerase and T4 ligase. However, some single-strand breaks don't have the appropriate structure. If the break is caused by depurination, the 3' end is not the appropriate structure. So we may have to then use NER (nucleotide excision repair), BER (base excision repair), or other methods or enzymes to polish off the ends. And of course, we're interested in trans-lesion synthesis because it's simpler for UV damage, interstrand crosslinks, and adducts.
The concept is to develop a simple system, by using the various combinations, that is going to be better. The way it looks at the moment, maybe the trans-lesion synthesis system is going to be the best because it's going to be quite simple. But that remains to be seen.
With that, I'd like to thank you.
DR. VOSBURGH: We'll hold the questions until after our next speaker.