Name: Dagan Wells
Title: Senior Fellow, University of Oxford
Professional background: 2007-present, senior fellow, Nuffield department of obstetrics and gynecology, University of Oxford, UK; 2007-present, laboratory director, PGD services, Reprogenetics UK, Oxford, UK; 2003-2007, assistant professor, department of obstetrics and gynecology, Yale University, New Haven, Conn.; 2000-2003, researcher, Reprogenetics, Livingston, NJ
Education: 1997 — PhD, genetics, University College London
Dagan Wells has been performing preimplantation genetic screening of embryos for nearly two decades, beginning as a student at University College London in the early 1990s. Now, as a professor of obstetrics and gynecology at the University of Oxford and a laboratory director for the company Reprogenetics UK, Wells has seen the technologies used to test embryos for chromosomal abnormalities evolve from fluorescence in situ hybridization to array comparative genomic hybridization and, more recently, SNP genotyping.
Wells has co-authored four papers so far this year that detail the use of array-CGH in pre-implantation genetic screening. Last month, he and fellow colleagues published a paper in Human Reproduction in which they discussed the first successful births in the UK following preimplantation genetic diagnosis of structural chromosome abnormalities using array-CGH (BAN 4/5/2011).
Wells is one of a growing number of researchers adopting arrays for use in PGS and array companies have been keen to make their products available for such purposes. In addition to Cambridge, UK-based BlueGnome, which offers its 24sure array products and services for research use in PGS, a number of other array vendors, such as Affymetrix and Illumina, have seen their chips used for preimplantation screening.
To gain more insight into how arrays are being adopted for use in PGS, BioArray News spoke with Wells this week. Below is an edited transcript of that interview.
You've been doing preimplantation genetic screening for nearly two decades. What technologies were in use before people started using arrays?
Initially, [preimplantation genetic diagnosis] was envisaged as an alternative for prenatal diagnosis for patients with a very high risk of passing on an inherited disorder to their children. And certainly it is widely used in that way today, but the majority of cases we do are actually not for patients who have a familial predisposition for a certain disorder, but for regular [in vitro fertilization] patients with no particular inherited problem.
IVF patients use these genetic technologies in order to help guide the doctors and embryologists to which of the embryos produced has the greatest chance of making a baby. That stems from the fact that a huge proportion of the eggs ovulated and embryos produced are chromosomally abnormal. For a woman over 40, you would expect more than half of the eggs that are retrieved during an IVF cycle to be abnormal.
The difficulty they have in the IVF lab is choosing which of these embryos to transfer. The ideal thing is to transfer a single embryo, but they may have six or seven to choose from. The choice is usually made based on embryo morphology, looking down a microscope to gauge which embryos seem to be growing the best. While that does have some value, morphology is unable to tell the embryologist anything about the chromosomal status of the embryos. You could end up with two identical-looking embryos, one of which is chromosomally normal and one of which is abnormal with no chance of producing a baby. Chromosome problems are usually lethal a little later in development, causing miscarriage, but during the first few days of life, the time when embryologists assess the embryos visually, chromosome abnormalities have little effect. The lack of a visible effect of such chromosome problems may be due to the fact that, at this extremely early stage, the embryos aren't yet expressing their own genes. They are surviving on raw materials that were provided from the egg itself and therefore can get away with significant genetic abnormalities. Of course those disorders end up having a very extreme effect just a few days or weeks later. Currently, technologies such as arrays offer the only possibility of telling the difference between normal and abnormal embryos before they are transferred to the uterus.
When we started doing these tests, the only thing available to us was FISH, and to begin with, people were looking at about five chromosomes and as technologies and methodologies improved, they moved up to about 12 chromosomes. Still, that was only looking at about half of the chromosomes that existed in the embryos. This means that, using those old methods, some of the embryos we thought were normal were probably not. They were abnormal for chromosomes that weren't being tested. One of the things that microarrays have shown us is that while it is true that some chromosomes are more often abnormal than others, really any chromosome can be affected in human embryos, so you do need to look at them all.
There were a few other technical issues associated with PGD using FISH. You are doing a diagnosis in most cases on a single cell, and so you can imagine, if by bad luck you happen to have two of the FISH signals overlapping, you may wrongly score that cell. Also, the process of spreading and fixing the nucleus of the cell onto a microscope slide can stretch the chromatin, causing little splits in the DNA, so you sometimes see double signals that should have been one signal. Also, if a cell had just undergone DNA replication, you may get two signals when really there's only essentially one chromosome there. There are all sorts of technical issues with the FISH approach that have largely been overcome with arrays. We are now able to see all of the chromosomes and you are not reliant on spreading the cell on a slide; rather, it is placed whole into a microcentrifuge tube.
[ pagebreak ]
When did you personally start using arrays?
We were doing FISH since about 1992 and then there was a slow evolution into microarrays. It didn't go immediately to the microarray. First the number of FISH probes increased. Then we introduced conventional CGH — we were applying labeled embryo DNA to normal metaphase chromosomes on a slide, allowing us to see the copy number of each chromosome. We introduced CGH in a research context in 1996, but clinical application didn’t follow until a few years later.
Technologies like CGH and microarrays are DNA hungry, requiring a couple of hundred nanograms of DNA, whereas a single cell only has about 5-10 picograms of DNA. The trick to it is to do a whole-genome amplification. Eventually, we managed to optimize a whole genome amplification method so that it would work in combination with CGH. But we still couldn't apply it clinically at that time. The actual CGH procedure just took too long. There was the amplification, then the labeling, and then the real killer was the hybridization, which took a couple of days; that wasn't compatible then with testing embryos. These embryos are on a pretty defined schedule. They have to be transferred to the uterus no later than five days after fertilization of the egg, because that is the point in their development when they are ready to implant. It's a fairly narrow window of time. You can't delay the embryo’s transfer to the uterus or it may never implant.
We've been using [microarrays] for about two years now. Yes, they were available in other areas of cytogenetics well before they became available in this area, and that was again really because of the need for whole-genome amplification. … There were a couple of false starts. We tried a number of platforms and whole-genome amplification methods, really beginning around 2001. But most of those did not work out well enough to be used clinically.
The real goal of getting a clinically applicable method came about three years ago. That was largely pioneered by BlueGnome in Cambridge, [UK]. Their technology [which is marketed for research use] is little different to any other microarray providers [because they use] bacterial artificial chromosome probes. But they were the ones who took the time and effort to try out whole-genome amplification methods in combination with a large number of different probes and find out which ones are compatible with each other. What you find with all these whole-genome amplification methods is that, in reality, none of them truly amplify the whole genome. There are always gaps and parts of the genome that amplify much better than other parts. So, the thing is to find which of the probes on your microarray you can actually trust. You probably start with [a] huge number of probes and you throw out some because they don't amplify, others because they over-amplify, and some more because they under-amplify. In the end, you wind up with some that you can pretty much rely on. BlueGnome has a pretty good product for that, and that has been a massive area of growth for them.
Have you been using their products since that time? I understand you have also had some interaction with Oxford Gene Technology.
Although we have primarily been using the BlueGnome array, and it works well, we are always keen to identify new technologies that might benefit our patients. OGT has been developing an array specifically designed for PGD, which may offer some unique advantages. It is not commercially available yet, but I understand it is in the final stages of validation. We have been testing it out in a research context, and it appears to work very well.
Ultimately, I suspect that both the BlueGnome and OGT platforms will probably have similar accuracy in terms of screening for chromosome abnormalities. It will then be a matter of which is more economical and whether the oligonucleotide array that OGT provides has any benefits over and above those offered by a BAC array platform. It is possible that an oligonucleotide array might allow for superior consistency of manufacturing, quality assurance, and might also allow simultaneous analysis of specific DNA polymorphisms or mutations, but this remains to be seen.
One of the key things for us when we are doing a PGD test is the price. Of course, that is true for any researcher or anyone offering any kind of clinical test, but for us the effect of cost is particularly acute. That is because one patient does not equal one test. One patient might produce two or three embryos or they might produce 20 embryos. Microarrays are not the cheapest things you can use in molecular biology. So you are talking about using as many as 20 for a single patient. That's a huge cost. What we do in practice is absorb those expenses, losing money on patients with large numbers of eggs and making up for it on patients with small numbers of eggs or embryos.
The ideal is to have a relatively inexpensive test. That makes it much more accessible to patients. For patients, the decision to have chromosome screening as part of an IVF cycle involves a kind of cost-benefit evaluation. They know the array will cost X amount, the IVF cycle will cost Y amount, and it is a question of whether the benefit of the array is sufficient to warrant its cost. If it costs next to nothing or if the benefit is shown to be huge, then I think every patient would have screening. However, if it were expensive, or if the benefits are only small, then many patients might think it better to save their money and use it to fund another IVF cycle if they do not become pregnant the first time around.
Right now an IVF costs about £4,000 ($6,500) or £5,000. A microarray adds another 50 percent to the cost of the IVF cycle. Is it worth that? I think for some patients it absolutely is; for others perhaps not.
[ pagebreak ]
Why is that? Why do some patients benefit from array-based screening over others?
Not all patients have an equal chance of producing chromosomally abnormal embryos. In the past, the tradition has been to offer these kinds of tests for four primary groups of patient. One is patients who are considered to be of advanced maternal age, which is, biologically speaking, around the mid-thirties. There are also patients who have had several failed IVF cycles. These are a very heterogeneous group of patients, but an elevated rate of chromosome abnormalities seem to be involved in the lack of IVF success in some cases. A third group are those who have recurrent miscarriage. Miscarriage is not always caused by chromosomally abnormal embryos, but more than two-thirds are. The final group is those who have had a previous aneuploid pregnancy. Some of them have it because they don't want to go through a miscarriage or termination again. Those are the major groups that have been tested.
Are you aware of any array-based offerings in addition to BlueGnome and OGT that have been used in PGS?
There are some researchers who have used a CGH array sold by PerkinElmer, but that array has not been customized for PGS and may have some detection issues, particularly involving the Y chromosome and chromosome 19. On the SNP array side, there are two major players, Affymetrix and Illumina, and both of those are being used in a PGS context at the moment. Gene Security Network uses an Illumina-based SNP array for PGD, as does Bill Kearns' group at Shady Grove [Fertility Center in Maryland] and [Bridge Fertility Center Scientific Director] Alan Handyside, who is working with BlueGnome, while Nathan Treff's group [at Reproductive Medicine Associates of New Jersey] has focused on the Affymetrix platform. The nice thing about the SNP arrays is that they provide DNA polymorphism information as well as copy number. It's been argued that you can use these polymorphisms to combine chromosome analysis with detection of single-gene disorders via linkage analysis. Additionally, if you find an abnormal embryo, in some cases you can work out whether the affected chromosome came from the mother or the father, which you can't currently do with array-CGH. So that is the potential advantage.
The disadvantage of trying to assign parental origin to a chromosome abnormality is that many of the errors seen in embryos are actually not meiotic at all. They are mitotic in origin. Chromosome errors are very common in embryos during the first few mitotic divisions. You get a high degree of mosaicism that occurs at random, so if you say an aneuploidy came from the father or the mother, it may not actually be true; it's more likely to be a random mitotic error. Additionally, with most PGD methods based upon SNP arrays, you have to test the parents as well as the embryos. This adds to the expense and makes the test less flexible than the array-CGH approach, as patients have to plan in advance and make sure that DNA is collected from both parents. The most significant issues for the SNP arrays are that they are more expensive that the equivalent arrays used for CGH and the protocols take longer to perform. As we have already discussed, the length of time taken for the PGD procedures and the expenses are critical. Most SNP platforms seem to have a hard time getting the protocol down to a 24-hour period, which is problematic and may involve staff working into the evenings, which adds further to costs. However, these arrays are coming down in price all the time, so SNP approaches may well turn out to be the way of the future.
If you want to use SNP arrays to look at single-gene disorders and chromosomal abnormalities simultaneously, it is definitely doable, but you have to have a good high density of SNPs to do that with any kind of accuracy. Also, you need to test additional family members in order to work out which SNP alleles are associated with the mutation. Unfortunately, about a third of the patients requesting PGD for a single-gene disorder don't have any additional family members available for testing, so this approach is not universally applicable. There are definitely some theoretical pluses with the SNP arrays, but whether those pluses are actually valuable at this time remains to be seen.
You are affiliated with Reprogenetics. What array do they use?
Reprogenetics UK primarily uses the BlueGnome arrays and offers a range of genetic testing services related to IVF and PGD. Reprogenetics in the US, which is a much bigger company and probably does about half the PGS that is done over there, uses the BlueGnome array also.
Can you put the use of arrays in PGS in perspective? Are they increasing the rates of successful pregnancies?
It's the million-dollar question, really. At this point in time, there is no randomized, controlled trial that proves that they really do offer the benefits that we think they probably do. What we can say at this time is that the underlying theory is very solid. It is absolutely the case that a large proportion of the eggs produced by women undergoing IVF have an abnormal number of chromosomes. Those eggs will produce embryos with the wrong number of chromosomes in all of their cells, and those embryos are most often either going to fail to implant or miscarry. If we can identify the embryos with the correct number of chromosomes during IVF, we should be able to lower miscarriage rates, improve pregnancy rates, and reduce instance of Down [syndrome] and other syndromes related to an incorrect chromosome number.
That is the theory. What we can say practically is that we now have technologies that tell you about the chromosome copy number in individual cells with high accuracy. We have the genetic tools; now we just have to assess the clinical benefits.
The area where there's still an open question is on the biology and embryology side of things. To do the test, we have to take material away from the embryo and that may have an impact in some cases on the embryo's chances of making a baby. There is also the issue of chromosomal mosaicism. Sometimes, no matter how good the test is, the cell you have taken from the embryo isn't representative of the embryo it came from. The embryo may have started from the fertilization of a normal egg with a normal sperm, divided normally for the first division, but then subsequently, one of those daughter cells divided its chromosomes incorrectly and ended up with a couple of abnormal cells. So now you have an embryo that is a mixture of abnormal and normal cells. Depending on which cell you biopsy, you'll either call it normal or abnormal. They can't both be right. There is still that biological question that leads to a clinical question of whether it really works. The preliminary data is extremely promising, indicating substantial increases in pregnancy rates and reduced miscarriage rates for IVF patients and patients with a history of recurrent miscarriage, but this needs to be verified in randomized studies.