This is part two of a two-part interview. Part one appeared in last week's issue.
Name: Sharon Plon
Position: Professor, Department of Pediatrics and Department of Molecular and Human Genetics; member, Human Genome Sequencing Center, Baylor College of Medicine
Chief of Cancer Genetics Clinic, Texas Children's Hospital
Experience and Education:
Fellow in medical genetics, Fred Hutchinson Cancer Research Center, University of Washington, 1990-1993
Postdoctoral fellow, National Cancer Institute, 1988-1990
Residency training, internal medicine, University of Washington, 1987-1988
MD and PhD from Harvard University, 1987
SB degrees in chemistry and chemical engineering, Massachusetts Institute of Technology, 1980 and 1981
As a member of Baylor College of Medicine's Whole Genome Laboratory management committee, Sharon Plon has been a leader in establishing Baylor's clinical exome sequencing diagnostic test, which was launched last November.
She is also one of two principal investigators for Baylor's grant under the National Human Genome Research Institute's Clinical Sequencing Exploratory Research Project program (CSN 2/1/2012), which will explore exome sequencing for childhood cancer patients.
Since Baylor's WGL started offering its clinical exome test a year ago (CSN 11/16/2011), it has received more than 600 samples and issued about 400 clinical reports.
At the Personal Genomes and Medical Genomics meeting at Cold Spring Harbor Laboratory last month, Plon gave a report of the lab's experience after its first year of testing. Clinical Sequencing News caught up with her during the conference and the following is part two of an edited version of the interview, which address how the lab handles interpretation and reporting for the test. Part one, which provided an overview of the test, appeared last week.
What types of results do you report?
Our focused report has five components. The first is, did we identify a mutation that explains the clinical problem? Is there a mutation in a disease gene that's known to be associated with that phenotype? Is the mode of inheritance correct? That's really the most important result, and we report that out in about 25 percent of patients.
We also include variants of unknown significance if they are rare or novel alleles in genes that are known to be related to the phenotype. We provide all the information that’s known about those mutations, and it's really up to the judgment of the clinician whether they think that's responsible for the phenotype. Or over time, more information might cause one to think that those are the responsible alleles.
We also report out other medically actionable findings. We do not use a pre-assigned list because we really think you have to explore what's known about individual genes and whether there clearly is medical management, either increased screening or intervention. There are obviously genes we've all agreed upon, but you may find an allele in a very rare disease that didn't make your list, but when you actually read the article about that gene, you discover that medically, it is important.
We also will report out autosomal recessive carrier status, but only if it's a mutation that's clearly associated with a recessive condition. And we do allow patients and parents to opt out of that part of the report because there are some families that don't believe in prenatal testing and are really offended by the idea of carrier status testing.
The last category is pharmacogenetic loci. We focus on those loci where there are clear FDA recommendations for testing for drug dosing. Some other platforms are focused on a much larger number of polymorphisms that have been associated with drug metabolism in the literature, but our pharmacogenetic aspect is really focused on very clear-cut examples.
We also offer an expanded report; it does require additional consent. We want the physician to get the focused report, decide whether they got their question answered, and then discuss with the patient whether they want to get the additional information. The expanded report has to do with other disease genes that are not medically actionable, for example neurologic conditions for which there is not a treatment; a larger set of variants in autosomal recessive genes, for example missense variants that are not well characterized; and any loss of function mutations in genes, whether disease-associated or not.
In terms of the receiving physicians, it's been extremely variable what they want. That was somewhat unexpected. There has been a range of requests from 'I don't want variants of uncertain significance in genes' to 'I want to know every variant, even if it's in an uncharacterized gene.' Many of the people ordering the test are obviously early adaptors of technology, and many of them want every mutation. They want to go online, they want to search all these variants, and they want to make a decision.
I am particularly interested in physician interpretation of genetic test results. I think what we've learned from the cancer field is that it's actually quite difficult to prove disease association of rare alleles, and that one has to be cautious about that. I'm not sure all physicians are as aware of that, and some of them really want to delve into the literature on their own to make those decisions for their patients.
How often do you find medically actionable incidental findings?
We report out what we consider an actionable finding in about every five to six tests. Those are findings that are unrelated to why the patient's sample was sent to begin with. There have been several patients who had mutations in cardiomyopathy genes. There is a very large number of cardiomyopathy genes, and there is incomplete penetrance, so not everyone who carries one of those mutations is going to have a heart problem. But if they have a mutation that's been characterized by the field as being associated with cardiomyopathy, we do report that out and we recommend that the patient see a cardiologist.
One of our patients had a Marfan syndrome mutation that doesn't give you the full spectrum of Marfan syndrome. But that child in fact did have an aortic root problem that's one of the most serious complications of Marfan syndrome, and the child is now being followed carefully by a cardiologist, which would not have happened except for that it was an incidental finding.
I also had a pediatric patient who had a colon cancer mutation, [a disease] that generally only appears in adults. There has been a lot of discussion about calling out incidental findings in children when they are adult-onset disorders. We actually feel strongly about doing that because the child normally has parents, and it's typically to the benefit of the child for their parents to be healthy. For example, for my own patient, when I asked the father and the mother whether they had any history of the disease — I'd asked them that previously, when I first saw the patient — the father mentioned that he actually had part of his colon removed when he was a young adult, but he did not know why. He in fact carries that mutation, so we very quickly referred him back to a gastroenterologist to see if he in fact has this condition and [should] be screened regularly. So I do think it makes sense to call out actionable findings in pediatric patients because the vast majority of them are inherited from their parents.
Is there any type of result you will not report?
We will not report mutations like Alzheimer's or Huntington's. Obviously, the exome is not very sensitive to some of those because they are triplet repeats, but unless there is a medical indication to do Huntington's testing, in which case there is a better test anyway, we don't report it out.
One of the very interesting questions that has arisen over the first year is, 'Who is owed what data?' We offer a CLIA test, and there are certain guidelines; you have to say what your test is and what you report out. But there are a number of people who want the BAM file, the VCF file, the reads. I've been charged to get our group together to decide on our policy because we have gone through a whole lot of validation of our clinical process, and there is no clear guidance about giving out raw data from a clinical test to have a physician re-interpret it.
I don't think there is any problem if someone is in a research study, we don't find the cause, and the researcher doesn't want to resequence the same person but wants to use our data. There is a charge for that, but I can understand that. But there are a number of clinicians who want just the raw data. I think it's an interesting question, it's not a solved question, even from a regulatory perspective, because that kind of thing would violate our CLIA license for this test. We have had at least five to 10 requests for various forms of the data. We decided this is something we didn't really think about when we were working to set up the test but we really have to decide what's appropriate. Most radiologists don't give the raw findings from an MRI machine; they go through very complicated software that has to be approved, and I think it's similar [here].
For those patients where the exome test provided a diagnosis, was there a recurrent theme?
Some of them are syndromes that as clinical geneticists, we are all familiar with, like Noonan syndrome, and some of them are very rare disorders.
One of the interesting things is, there are two genes that were only associated with intellectual disability over the last couple of years, and we have already had three patients with mutations in those genes, so it may certainly be that as we do exome sequencing, we discover that some genes are more common causes of intellectual disability than we previously realized.
We have also diagnosed Cornelia de Lange syndrome in several different patients. There have been three different patients with that disorder; they each have mutations in different genes, and one of those genes was only discovered to be associated with the disease last year.
How do you incorporate newly discovered disease mutations, and do you re-analyze patient data in light of new findings?
One of the recurring themes at this meeting is the need for a database that rapidly accumulates information on clinically relevant mutations. We all use similar databases to try to find the existing data, but none of these are updated as frequently as one would like. And also, because it's a CLIA-certified test, we can only change our analytic pipeline so frequently.
We have a group of people who literally are just reading the literature. And every month, as new genes are identified, there is essentially a list of new genes in the lab that the data is compared against, and a number of diagnoses have come out of that list. This whole Mendelian discovery process that's going on around the world is kind of walking hand in hand with the clinical diagnostic process.
We have re-analyzed patient data, predominantly when the new finding is within a few months of a negative report. Like any lab, we tend to hold onto the data for the negatives as long as we can because we really want to find an answer, and there have certainly been cases where we thought we would have to report it out as negative, and then based on a new finding, for example a missense change that we thought was suspicious and has now been associated with a disease, we can now say, this is really a recessive condition, here are the two mutations, and there is now data on both of the mutations.
With regard to reanalyzing data on a regular basis, this is something that's been an issue in genetic testing for many years and will obviously be heightened by the exomes. Generally, our recommendation is that physicians, especially if we have not made a diagnosis, request a reanalysis probably a year after the first time it was done.
I think once you get out more than a year, if you are not in a circumstance where you don’t have the opportunity to resequence, it's going to make sense to resequence as opposed to reanalyze. Our test is better a year later than it was a year ago. It's certainly true that a reanalysis might yield the result, but I think it's probably more likely that a better test is going to yield the result.
How rare is it that the diagnostic mutation leads to changes in treatment?
I think it's turning out to be not as rare as people may have thought. One child we analyzed had a congenital myasthenia gravis syndrome, that's very much a treatable condition. That patient had seen numerous specialists, that patient's sister had died of the disorder, and it's a known, treatable disorder. It's very rare, and even for this very rare disease, there are many different genes that can cause it, so even if someone had thought about it, whether they would have tested the right gene is not clear. I think we are going to have more of these cases.
I think it is correct that we should not think everyone with a diagnosis is going to be treatable. But I think it's too early to say what proportion actually change treatment. In addition, patients are no longer being treated for the wrong diagnosis. And I think that is actually a very important point that gets underestimated.
The WGL also started cancer exome sequencing?
Baylor has one of the NHGRI Clinical Sequencing Exploratory Research Project grants, for which Will Parsons and I are co-PIs. We are focused on newly diagnosed childhood cancer patients. For that study, we are doing germline exome and tumor exome, and we are asking questions about how often we find a cancer susceptibility mutation, and how often we find that in a gene that we would not have predicted.
In childhood cancer, the vast majority of children are put on existing clinical trials, so we are not impacting initial treatment decisions, but we are monitoring if one of the patients in the study recurs, how often the oncologist uses the sequencing information to help decide what to do next. That’s a clinical study; all of the results are going in the medical record. The clinicians are part of the study, too, they're subjects; we survey them with each patient, whether the exome gave them information that would change what they would do if the patient recurred, just theoretically, and then what actually happens when patients do recur.
For the cancer exome, we do not confirm all variants by Sanger. That is the case because physicians are making decisions constantly about these patients. We do report out all of the somatic mutations that meet our quality metrics. We confirm the most important ones by another methodology, but we will report them all.
You mentioned that you also did exome sequencing on fetal samples.
We are not doing it as a prenatal test, so we are currently not offering, for example, testing of amniocentesis samples for reproductive decision making. However, there are fetuses that are lost to miscarriage or termination of pregnancy where traditionally, geneticists will inspect the fetus and try to make a diagnosis. They have submitted DNA for testing of the most likely diagnosis, and we have done exome sequencing in that circumstance.
It has huge advantages for the family. For example, in one case, the parents had been told it could be one of several disorders. One of the likely disorders was recessive, in which case the parents would have a 25-percent risk of having a second severely affected child; the other likely disorders were dominant. We were able to show that it was one of the dominant disorders, that it was a de novo mutation.
Are you considering switching to whole-genome sequencing anytime soon?
I don't think it's very far off. What we've generally done is formed a subcommittee of people to work on the problem and come back to the bigger group that manages the Whole Genome Lab when they are ready for launch. I think it will obviously involve patients who failed whole-exome sequencing.
The key is, there is an enormous amount of variation outside of the coding regions, and the variation you can actually put into a clinical report with an interpretation is very limited right now. We really view this as a clinical function, and we really want to be able to give physicians results that have clinical meaning. So if we have thousands of variants in non-coding regions that we don't know what they mean, I don't think it's going to be very useful clinically. That's really the biggest problem with whole-genome sequence right now — not the cost, which is certainly one thing, but being able to discern what you can report out outside of what you would get in an exome anyway.