Okay. Welcome to the ACMG Genomics Case Conference. This series has been planned by the ACMG Educational Committee and rotates among labs each month. This month is hosted by Connective Tissue Gene Test, and our topic is Prenatal Molecular Diagnosis of Skeletal Dysplasias Using Targeted NGS Panels. So this activity is supported by an educational grant from QIAGEN. It is intended for educational purposes only. I am Keri Brown. I'm the Associate Medical Director at Connective Tissue Gene Test. I will be presenting the webinar today. I have one disclosure. I am an employee of Connective Tissue Gene Test, which is a fee-for-service lab which performs clinical genetic testing. Here are the disclosures for the planning committee and ACMG staff for this conference. This session is designated for one Category 1 CME credits. If you registered for credits, you'll receive an email at the end of the presentation. Here are the learning objectives for the overall conference series. Today will be a little bit different because I will not be speaking about exome or genome sequencing. We were actually asked to speak about a specific group of disorders, skeletal dysplasias. So I will be telling you about our laboratory's experience with testing of skeletal dysplasias. In this session, the learning objectives are discuss the utility of targeted panels for molecular diagnosis of skeletal dysplasias, identify the genes most commonly responsible for lethal skeletal dysplasias, recognize the importance of gene-specific guidelines for sequence variant interpretation, and describe the value of continuously evaluating test results to improve the clinical sensitivity of targeted panels. If you have a question during the talk, you can type it into the box on the lower left corner, and these questions will be addressed at the end of the presentation. So just a quick introduction to the talk. As I mentioned, we'll be talking about skeletal dysplasias. This is a very broad group of disorders. So for the purposes of this presentation, I'm focusing on one panel that we've run and on the prenatal cases that we have run on this particular panel. So first, we're going to go over the overall results for this panel, and the value of targeted testing for prenatal skeletal dysplasias. Then we'll speak about the genes most commonly responsible for prenatal skeletal dysplasias and classification of variants in these most common genes. There will be a presentation of several cases, and we'll finish with the need for updating panels based on test results and new information. So the panels that I'm going to talk to you about today were specifically designed to diagnose prenatally identifiable, so lethal and severe, skeletal dysplasias. The panel is actually divided into two subsets. There's the core panel, which is 10 genes, and that includes the genes most commonly responsible for prenatally identified skeletal dysplasias. And then there's the extended panel, which includes an additional 19 genes. Those genes are also responsible for prenatally identifiable skeletal dysplasias, but they are much more rare than those on the core panel. So when we initially designed this gene, we envisioned that people would run the core panel, and if it was negative, then they could reflect to the extended panel. That's actually not ended up being what happened. So people pretty much either just ordered the core panel or they ordered the core and extended to be run concurrently. There were actually very few people who did reflex from the one panel to the other. So this slide shows the genes on these panels. So on the left are the genes on the core panel. On the right are the genes on the extended panel. So version one of this test included only six genes on the core panel, so call zone 1A1, 1A2, 2A1, FGFR3, SLC26A2, and SOX9. And those are known from the literature and from testing we've done in the past as being the most common genes that come up in prenatal skeletal dysplasias. The other genes were added on in other subsequent versions of the panel, but ALPL I'll talk about a little bit later. The three at the bottom are actually more rare causes of skeletal dysplasia that were put on that panel for other reasons. And then the genes on the extended panel we came up with in consultation with several clinical geneticists who we had consulted about this panel. And another thing to note is that while these genes can cause severe phenotypes, they also can cause much more mild phenotypes. So there's a whole range of phenotypes that you can get testing the genes on this panel. So we actually use this panel for prenatal and postnatal testing, and it works very well in a postnatal setting as well. So just to briefly tell you about our methodology, we use an Agilent SureSelect custom-designed library to capture these genes. So this is a hybridization-based capture. And then the genes are sequenced by Illumina MiSeq. We also do detection of large deletions and duplications. So the CMV detection is currently being done based off of our next-gen data using Golden Helix VarSeq CMV software. We've been running this method for about a year, and prior to that, we were doing CMVs using a custom-designed high-density targeted microarray, so basically an exon array. So this is a full-coverage panel. All exons and intron-exon boundaries are sequenced, and any exons that are not fully covered by the MGS are filled in by Sanger. All variants classified as pathogenic, likely pathogenic, or uncertain significance are confirmed by Sanger and included on the report. So here's a summary of our results. I pulled out the 500 most recent cases that we had tested in the lab, and from those 500 cases, we determined that 384 of them we would consider prenatal. So most of these are fetuses, so amniotic fluid samples or CVS samples. Some of the cases were actually from newborns, but these were newborns who had a fetal dysplasia identified prenatally, but genetic testing was not done until the baby was actually born. So of these 384 cases, 210 were positive, and I'm defining positive as cases with a pathogenic or likely pathogenic variant that we think is responsible for this disorder. So about 55% of the cases were positive. Another 10% were inconclusive. There were two different reasons for being inconclusive. Six of these cases had one pathogenic or likely pathogenic variant, but were in genes that were expected to have autosomal recessive inheritance, and then another 8% of cases were inconclusive because the variants identified were all classified as uncertain significance. And then there were about 36% of the cases were negative. So just a note about the negative cases. The cases that were included in this group included all the prenatal cases that were run on the core panel. So about 48% of these cases were only run on the core panel, and then the other 52% were run on both the core and extended. So for 82 of the negative cases, about 60% were only run on the core panel. So potentially there could be a gene on the extended panel responsible for the disorder, but those genes were not actually run. There are also five negative cases that we now know are due to genes that are not included on this panel. So these were cases that the same patient was tested later on, either further along in the pregnancy or after birth, and they were tested for different panels, and a causative variant was found on a different panel. Another reason why some of them may be negative, well, some of these cases are definitely cases of lethal severe skeletal dysplasias. The phenotypes actually cover a wide range of phenotypes. So some of the clinical information that we've gotten for these cases, the baby was just reported to have short, long bones, or the baby was just reported to have IUGR. In some cases, we just got that they have increased NT on testing. So it is possible that some of these cases are actually not really cases of skeletal dysplasia. Particularly in a prenatal setting, we're very concerned about the rate of variants of unknown significance. But in this setting, parental testing is very important. So what we're seeing is mostly fetuses who have severe disorders with parents who seem to be unaffected. So in that scenario, you would expect the disease to either be recessive or to be caused by a de novo variant. So when we get a VUS back in a dominant gene, this VUS can usually be resolved by performing parental testing. So you can see the table at the bottom. Initially, I told you there were 31 cases with variants of unknown significance. Ten of those cases were in genes with autosomal dominant inheritance. In eight of those 10 cases, the parents were tested, and in all cases, the parents were negative. So in this case, we don't actually test for the identity of the parents, but we would presume that these individuals are de novo. So based on the fact that we think they're de novo, we would elevate these variants up to likely pathogenic. So after testing patients in these cases, we ended up with 218 patients that were positive, so about 57%. The parental testing also decreased the rate of VUSs in these cases to 6%. So the table on top shows our overall results, but again, we said that we specifically designed this for severe and lethal skeletal dysplasias. So if we picked out just those subsets that we know are suspected to be severe or lethal, how does that affect our results? So there were 96 of these cases where it was specifically listed in their clinical history that a lethal skeletal dysplasia was suspected. Out of those 96 cases, 81 were positive, so almost 85%. And the rate of inconclusives in that population was 4%, and the negatives were 11%. So in our population that the test was specifically designed for, we have a very high clinical sensitivity, almost 85%, and the rate of variants of unknown significance are pretty low, about 4%. So now just to tell you, what are these genes that are responsible for these positive cases? As expected, the core panel genes were responsible for 94.5% of cases. The extended panel was responsible for 5.5% of cases. And there were four genes that were much more prevalent than all the others. So FGFR3 was the most common, and as expected, collagen 2A1, 1A1, and 1A2 were the other really common genes. Then there seems to be, like, a second tier of genes, ALPL, SLC26, A2, and SOX9, which, while they're not nearly as common as those top four, we do see them repeatedly, and we would consider them to be moderate causes of skeletal dysplasia. And then the third category is basically all of the rest of the genes. So other than those seven, everything else only had, you know, there was one or two cases here or there of other genes causing the phenotype. And if you look at the distribution of genes of variants of uncertain significance, it's very different than those with pathogenic variants. So the extended panel contains 73.9% of the variants of uncertain significance, the core panel only 26%. You can see there's 1VUS and COL2A1 1VUS in FGFR3. Those could probably be resolved if the parents were tested, but the parents were never sent for testing in those cases. Another interesting note is that the HSPG2 gene had eight cases with variants of unknown significance. This we can attribute pretty much entirely to the size of this gene. This gene is very large. It's 97 exons. It has 13,000 coding nucleotides. So it's basically just a big target. And then there were two cases where multiple variants of unknown significance were identified. In one of those cases, the mother was reported to be Hispanic. In the other case, the mother was reported to be Filipino. So there's a good chance that those are actually benign variants that are specific to those ethnic backgrounds, and we just don't have the controlled data to be able to classify those. So to sum up this section, targeted panels are very useful in this sort of situation where a small number of genes is responsible for a large proportion of the disorders. So with a targeted panel, you can get full coverage of all of those very important genes, and you limit the number of variants of uncertain significance, which can be important in a prenatal setting. In general, you also have a quicker turnaround time. The average turnaround time for this panel is 17 days. So it's, you know, a little over two weeks. 27% of our cases were actually finished in less than two weeks, and that's the time to final report. At this point, the cost of targeted panels is also lower than the cost of exome or genome sequencing. And another benefit is that parental testing is immediately – parental and familial testing is immediately available, which it may or may not be with exome or genome sequencing. All right. So now I'm going to talk about those – some of those really common genes and the types of variants that we see in those genes. So first, our most common gene, FGFR3. FGFR3 is a receptor tyrosine kinase. As you can see in the picture on the left side, in the presence of the FGF ligand, two FGFR3 molecules will dimerize, and that causes activation of the kinase. And FGFR3 related skeletal dysplasias are caused by gain-of-function variants. So these are variants that will cause activation of the kinase in the absence of ligand. On this slide, I'm showing you all of the FGFR3 variants that we saw in this cohort. So what's probably immediately obvious is that there's not very many of them. These are recurrent pathogenic variants. So we tend to see the same variants or at least the same types of variants over and over again. In this particular cohort, we had 13 cases of achondroplasia, and all the rest of the FGFR3 cases were thanatophoric dysplasia cases. You can see that the – on the top line, the arginine 24815 change was seen in 32 cases. This is known to be the most common cause of thanatophoric dysplasia type 1. You can also see the first five variants are all variants where it's a change from amino – some amino acid to a cysteine. So study of this gene has shown that when you insert cysteines into certain regions of the extracellular domain, it can cause dimerization of – dimerization of FGFR3 in absence of the ligand. So in this case, the cysteines are very important for disulfide bonds, and you need two paired cysteines to make a bond. And when you introduce in unpaired cysteines, you can get this dimerization in absence of the ligand. If you look at the next five variants, they all involve alterations of the stop codon. So basically, destruction and read-through of the stop coding codon, leading to an elongated protein. And this is another – the elongated protein also causes dimerization of the FGFR3 ligand – or FGFR3 molecule in absence of the ligand. And then the lysine 650 glutamic acid change is the change that causes thanatophoric dysplasia type 2, and we saw that in six cases. That is a variant that occurs in the activation domain, and it's thought to cause activation of the – or the activation loop. So that's thought to cause activation of the kinase in the absence of dimerization. And the very last line of the slide, there was – there's the one variant that we call the VUS. It's in the intron. It's a minus 3 change. So in some cases, a minus 3 change can have an effect on splicing. From what we know about this particular disorder, we actually think that this is probably a benign variant. It doesn't really fit in with the mechanism of pathogenic FGFR3 mutations, but we just don't have enough evidence about – on that variant to be able to make a call, and the parents were not tested in this particular case. So the other three very common genes are all fibrillar collagens. So collagens 2A1, 1A1, and 1A2. So the fibrillar collagens have a very typical triple helical structure. So three peptides come together and form a helix. In the type of – in the case of type 2 collagen, this is three collagen 2A1 peptides that come together. In type 1 collagens, it's two collagen 1A1 peptides and one collagen 1A2 peptides that form the triple helix. And these collagens have very specific structures. There's an N-terminal region. There's a very large triple helical region, and then there's a D-terminal region. And it's essential that – so in the triple helical region, it's essential that there's a glycine at every third amino acid. So the sequence of this triple helical region will be glycine XY, glycine XY, glycine XY, where the X and Y can be basically any amino acid. But having that every third glycine is essential. Another important thing is that the helix forms from the C-terminal end. And the C-terminal domain is important for association of the three chains and beginning of the formation of the helix. So these collagens have been studied extensively, and it's been realized that the severe and lethal phenotypes are due to variants that disrupt the structure of the triple helix. Loss of function variants in these genes also cause phenotypes, but they cause non-lethal phenotypes. For example, in collagen 2A1, loss of function typically causes Sickler syndrome, and in col-1A1, loss of function typically causes OI type 1, the most mild form of OI. So first I'm going to show you the col-2A variants that we identified in this cohort. You can basically split these variants into three distinct groups. So the first group, the most common pathogenic variant in col-2A1, are glycine substitutions in the triple helix. So those account for about 70% of the pathogenic cases of collagen 2A1. The thing that's different from FGFR3 is these are not recurrent mutations. So I've listed all 34 mutations there. The ones that have a 2 behind them were seen in two individuals, actually I guess 28 mutations, because four of them were seen in two individuals. All the rest were only seen in one individual. Some of these changes have been published, but many of them have not been published. But we will still call them pathogenic just based on how important the glycine is to the structure of the protein. The second category of variants that you see in col-2A1 are small in-frame deletions within the triple helix. So this accounts for 12.2% of the cases. And you can get these small deletions in a couple of different ways. So the first two variants listed are standard small deletions, a deletion of 18 nucleotides and a deletion of 6 nucleotides. The third variant listed, the 905C to T, was seen in two patients. This variant appears on the surface to just be a missing change, an alanine change to a valine change, an amino acid 302. But this change is actually the most common change causing yeast dysplasia and has been studied intensively. And we know experimentally that the real mechanism of this change is that it generates a cryptic donor splice site in exon 14. And this leads to deletion of the last seven amino acids of that exon. So even though it looks like a missense, this is an in-frame deletion. And then the last sort of group of variants that cause these in-frame deletions are splicing changes. So we had three of those in this cohort. The top one you can see is minus 2 delay. So that's at the minus 2 invariant position of the splice acceptor site. And then the two other changes were ones, they were in, they're in the donor splice site region. We weren't actually sure what they did. So both of these two at the bottom were called variants of uncertain significance. They were both tested in parents and both, the parents were negative. So these were assumed to be de novo. And based on that, we reclassified them as likely pathogenic. In many cases, a splicing variant can cause a loss of function mutation. In this case, that wouldn't really be consistent with this type of disease. So what we think is happening here is that you're actually getting skipping of the adjacent exons. And the skipping of the exon is causing an in-frame deletion. So for the top change, you could potentially get a 15 amino acid inframe deletion if exon 17 was skipped, and you could get a 18 amino acid inframe deletion if exon 22 was skipped, and you can also get an 18 amino acid inframe deletion if exon 31 was skipped. So, we think that's probably the likely mechanism in those changes. And then, the third type of variance that you see in COLS 2A1 are variants in the C-terminal domain. And, if you remember, the C-terminal domain is very important for association of the three peptides and the start of the formation of the helix. There's two specific disorders that tend to be caused by these C-terminal domain variants. One is platyspondic cell dysplasia torrents type. The other is spondyloperforal dysplasia. Both of these genes have been published, and they've been published to be caused by nonsense mutations, frameshift mutations, 16 substitutions, and inframe deletions in the C-terminal domain. So, you can see the list of the variants found in our cohort, and you can see there's a nonsense, there's three frameshifts, there's two 16 substitutions, there's one inframe deletion, and then, at the bottom, we had two changes that we called VUSs. One is an inframe duplication in the C-terminal domain. We would expect that this would probably be pathogenic because it would have the same sort of mechanism as a deletion, an inframe deletion in the C-terminal domain. But none of these small duplications have actually, in the C-terminal domain, have actually been published that we're aware of, so we called this a VUS. These were not tested for this case, so we were not able to resolve this particular VUS. And then, the last change was a missense change in the C-terminal domain. This one, we also called a VUS. It's not entirely clear how this change is causing disease, but it was not seen in the parents, so it does seem to be de novo, so we did request by this one as likely pathogenic. The aspartic acid at 1436 is very highly conserved, and it's conserved in the C-terminal domains of other fibrillar collagens. How exactly it causes disease is not clear, so... So, just a word about interpretation of what we call 2A1 variants. To do... Since so much is known about the structure of the variant, you can...you really need to have gene-specific guidelines for this gene. So, any glycine substitution in the triple helix, we'll call it 2A1, we call pathogenic. Any small end frame deletions and duplications in the triple helix, we call pathogenic. Frame shifts and nonsense variants anywhere in the gene, we call pathogenic. Ones in the C-terminal would be expected to cause severe phenotypes. Ones in the other region of the genes would be suspected to cause tickler syndrome. Cysteine substitutions in the C-terminal domain, we call likely pathogenic, and end frame deletions in the C-terminal domain, we call likely pathogenic. And we're able to do this just because the gene is so well-studied and so much is known about the structure and how these types of changes affect the structure. One more comment on COL2A1. So, COL2A1 is responsible for a large number of disorders. Tickler syndrome, as we mentioned, are loss-of-function variants, but pretty much all of the other disorders are caused by disruption of the triple helix. There are some genotype-phenotype correlations. So, the C-terminal domain tends to be the platyspondial dysplasia torrents type and spondopropyl dysplasia. Mase dysplasia tends to be caused by end frame deletions. But in general, all of these other than the tickler are due to disruption of the triple helix, and you can't really tell based on the variant which kind of phenotype you're going to end up with. So, while we call these pathogenic variants, we usually call them pathogenic variants causing a type 2 collagenopathy, and in most cases, we cannot actually tell you which specific disorder they will cause. So, collagen 1A1 and 1A2 has a very similar structure to collagen 2A1, and so the variants that we've seen in those genes also fall into – fall mainly into these three categories. Substantial substitutions in the triple helix are, again, the most common with about 70 percent of cases. Small end frame deletions and duplications in the triple helix are also seen. In this cohort, five of those were just standard end frame deletion duplications, and eight of them were splicing changes that were presumed to cause end frame deletions by exome skipping. And then there were four variants that – in the C-terminal domain. One nonsense, one frame shift, one substantial substitution. We also did see one end frame duplication in the C-terminal domain, which, again, we called VUF. And then there were two variants in collagen 1A1 and 1A2 that didn't really seem to fit exactly into any of these categories. So, now I'm going to talk in a little more depth about the end frame duplication in the C-terminal domain, and then the two that don't seem to fit well into the three main categories. The patient case 1 is a fetus at 22 weeks, 6 days. The sample that we got were cultured cells derived from amniotic fluid. They had requested skeletal dysplasia core panel with sequencing and deletion duplication testing. The clinical information that we received included short, long bones, abnormal curvature of bones, hypomineralized cranium, small thorax, abnormal appearing ribs, pericardial effusion, polyhydramnios, and a karyotype had already been done, and the fetus was 46XX. So, as I mentioned, we found that the NGS uncovered a variant of uncertain significance in collagen 1A1. This was in the C-terminal domain. It was a nine-nucleotide duplication, led to the insertion of three amino acids. It was not seen in any of the population databases or reported, but we did suspect the duplication in the C-terminal domain would be pathogenic, but there just wasn't enough evidence. So, it was called a VUS. We recommended parental testing. Both parents were unaffected, and they were both negative for this variant. So, we presumed that this was de novo in the fetus, and reclassified it as likely pathogenic. Some time later, a second fetus was sent to us from these same parents, and they requested targeted testing just for this COL2A1 variant, and the second fetus was also heterozygous for the COL1A1 variant, and the second fetus was also heterozygous for the COL1A1 variant. So, you can see the Sanger sequencing traces there for the two fetuses, and the arrow marks where the double sequence marking the duplication starts. And then, if you look at the traces for the father and mother, which are shown underneath, there is no sign of any sort of duplication in the blood samples from either the father or the mother. So, this, we believe, is a case of germline mosaicism. It was the only case in this cohort that we are aware of where germline mosaicism exists, although it is kind of difficult to detect germline mosaicism, because you would need to have a second fetus from the family that was affected. So, germline mosaicism does happen in this sort of cohort. It's pretty rare, but it does happen. So, now, the next case is one of those where the variant doesn't fall into one of those three main categories for brule ecologens. So, this was a fetus at 25 weeks, 5 days. The sample we received was amniotic fluid, or cultured cells from amniotic fluid. We had requested skeletal dysplasia core testing, sequencing, and deletion duplication. The clinical information that we received was that the fetus had short boned, long bones in the lower extremities, and the upper extremities were within normal limits. So, this actually seemed like a milder end phenotype, not necessarily a lethal. We ran the NGS, and it was negative. And when this case came in, we were still doing the exon array. And on the exon array, we found a 4.7 kb heterozygous deletion of exons 25 to 31 of COL1A1. So, you can sort of see the results from deletion duplication on the screen. And the area with all of the red that's marked in red are the probes of lower intensity that show the heterozygous deletion. So, generally, these large deletions were thought to be a loss-of-function mutation, which is not necessarily consistent with this phenotype. But if you actually look at the sequence of exons 25 to 31, they actually are...it's a deletion of 153 amino acids in-frame. So, potentially, this could be an in-frame deletion in the triple helix. So, this change could disrupt the triple helix and cause disease due to that. So, we tried to follow up on all of our prenatal cases and were not successful in all of them. But in this particular case, we did get some follow-up information. So, prenatally, they were suspecting OI type 2 or 3 as fractures were identified prenatally. When the baby was born, a postnatal skeletal survey showed healing fractures of the left fibula, right fibula, and right tibia. The baby was discharged from the NICU on day 3 of life with no respiratory problems and no new fractures noted in the newborn period. So, they concluded that this was probably an OI type 3 case. The baby was doing pretty well about a month after birth, and pamadronate infusion therapy was planned for this child. So, this was actually the only large deletion duplication that was identified in this cohort. And when you look, I've listed the genes again where we saw mutations. And if you look at the list of genes, it kind of makes sense because the four most common genes are not caused by loss-of-function mutations. So, the phenotypes that we see in these four genes are typically not caused by loss-of-function mutations. So, it's not that unusual that you wouldn't see a large...you would only see a very small number of large deletions and duplications. Of note, a lot of the other genes on the panel are due to loss-of-function mutations. So, you could get large deletions and duplications in those genes. They're just much more rare. So, going back to this slide with our three categories of mutations, I can add a fourth category, which is a large C and V. One case had a large C and V presumed to cause an in-frame deletion. And then we still had one case that did not fit into these categories. So, that case was a terminated fetus. We received a DNA sample derived from amniotic fluid. They had requested skeletal dysplasia core and extended panel with sequencing and deletion duplication. We got very limited clinical information. They just said they suspected campylic dysplasia and that the father of the baby had joint hyperlapsity. So, we did the NGS, and we ended up finding two pathogenic variants in COL1A1. So, the first variant was an exon 7. This was a glycine 188 sphere missense mutation. This glycine substitution was near the N-terminal end of the triple helix. So, these variants near the N-terminal end tend to have milder phenotypes. And variants in this particular area often had more of an EDS phenotype than an OI phenotype. So, there's been several...there's been a number of variants in this region that have been published that cause either an overlapping phenotype between EDS and OI, or where they cause mostly EDS features. And a change at the same codon, a glycine 188 aspartic acid change, had been previously published in an individual who had an overlapping phenotype between EDS and OI. The other variant was in exon 42, and this was a nonsense variant that had been previously published as causing OI type 1. So, we tested the parents in this case for both of these COL1A1 variants. The father had the glycine 188 spherine change, again, a change which might cause an EDS-OI overlapping phenotype. And the father was reported to us as having joint hyperlapsity, so it seems like that might fit with his phenotype. The mother was negative for both of the variants. So, we ended up having a paternally inherited glycine substitution in the triple helix, and then we had this de novo nonsense variant. So, it raises the question of, is this an autosomal recessive disorder caused by COL1A1? So, you have two variants that individually cause very mild phenotypes, and then we had this fetus who had a very severe phenotype. So, unfortunately, we weren't able to get any other information from this family. The fetus was terminated, so we can only speculate, you know, is this an autosomal recessive manifestation of COL1A1? So, in the last part of the talk, I'm just going to talk a little bit about the importance of continuously updating your targeted panels. So, this is needed as new genes are discovered, which cause the disorder, but it's also helpful in reviewing your own data to see if there's any other genes that you think should be included on the panel. So, you know, as you run these tests, sometimes you discover that, oh, maybe I should have put this gene on the panel, or maybe this gene is not that relevant and we should take it out. And making these changes to the panel can increase the clinical sensitivity of your test. So, case four sort of illustrates this point, and this is a two-day-old female. They had requested skeletal dysplasia core NGS testing and ALPL Sanger sequencing. So, in version one of the core panel, ALPL was not included in the skeletal dysplasia panel version one. So, they requested the core panel plus the ALPL Sanger. And the clinical information for this case was prenatal ultrasound showed severe micromelia, bilateral club feet, undermineralization of cranium and spine, platyfondly, and short ribs. So, we ran the core panel, the skeletal dysplasia core panel, and it was negative. And then we did the Sanger sequencing for ALPL, which causes hypophosphatasia, and we found one pathogenic variant and one likely pathogenic variant. So, the pathogenic variant was a missense change that had previously been reported causing infantile hypophosphatasia, highly conserved, rare. We had also previously seen it in our lab, combined hendrozygous with other pathogenic variants. So, this change was pathogenic. And then the other change was an in-frame deletion of one amino acid. This was in the alkaline phosphatase domain. It was not reported in any of the population databases. It's seen with this other change, which we know is pathogenic, and it fits with the patient's phenotype. So, we called this variant likely pathogenic, and we think that this baby does have a hypophosphatasia phenotype. So, shortly after seeing this case, we did version two of our panel, and we added ALPL onto the core panel as part of version two of the panel. And since then, we have found six more cases of prenatal skeletal dysplasias that are due to ALPL mutations. So, severe hypophosphatasia is an autosomal recessive disease, so you would expect to see, you know, two variants, two pathogenic or likely pathogenic variants for these cases. And, you know, I've listed on the left is one of the variants. On the right is the other variant in the seven cases that we identified. Four of those were confirmed to be compound heterozygous, and then there were four additional cases where we found one pathogenic variant in ALPL, but no second variant was found. And it's possible that the patients are just carriers for these variants, or it could have something to do with their phenotype. The glutamic acid 191 lysine change was seen as a heterozygote in two individuals. That's actually the most common ALPL pathogenic variant. It's seen in the NOMAD database with an overall frequency of 0.26%, and in the finished population, it's actually seen in over 1% of individuals. So, the baby could potentially just be a carrier for those changes. The other two changes that were found in a heterozygous state were pretty rare but known ALPL mutations. In the one case, they didn't request deletion duplication testing, so it's possible that, you know, they could have missed out. There could be a large deletion duplication in that case that, you know, wasn't detected because they didn't do that testing. But our sort of conclusions from this is that we definitely think that ALPL should be included on the core panel. In our cohort, it's as common as SLC26A2 and SOX9, so it was a little bit, while we know that ALPL causes, you know, these prenatal skeletal dysplasias, it was a little bit of a surprise to us that it actually, you know, came up in this many individuals in our cohort. And adding this to our panel actually increased its clinical sensitivity by about 2%. And then one final case, this was a fetus. We got cultured cells from amniotic fluid. They had requested the skeletal dysplasia core and extended panel with deletion and duplication, and the clinical information we received was just significantly shortened in both long bones and small chest circumference. So the results for that panel were negative. We didn't find any potential pathogenic variants. But about four months later, the same patient was sent to us for testing. At this point, the patient was a 6-day-old female, and now they were requesting our skeletal dysplasia celiopathy panel sequencing and deletion duplication. So this panel basically tests for asphyxiating thoracic dystrophy, short-rib polydactyly syndrome, June syndrome, so those genes that are known to be skeletal celiopathies. And there are some features that can distinguish that from the other skeletal dysplasias that we were previously talking about. But sometimes, especially if the babies don't have polydactyly, it can be difficult to distinguish them. So the clinical information we got for the newborn baby was skeletal dysplasia was concerning for June syndrome. The baby had handlebar clavicles, short ribs, trident pelvis, shortened tubular long bones, metaphyseal irregularities, premature ossification of the capital of seminal epitheces, and there was no polydactyly. So the NGS for the skeletal celiopathy panel showed two pathogenic variants in the DYMC2H1 gene. One of these variants was nonsense mutation. The other variant was a known pathogenic mutation that had been previously published. So this baby did, in fact, have a skeletal celiopathy. And unfortunately, the baby died at 29 days of life with severe respiratory distress caused by pulmonary dysplasia. And not long after we saw that case, we got another case. The only information we got about this case was that it was a fetus with skeletal dysplasia. They requested the skeletal dysplasia core panel sequencing and deletion duplication. And if that was negative, they wanted to reflex to the skeletal dysplasia celiopathy panel sequencing and deletion duplication. In this case, we, again, saw two variants in the DYMC2H1 gene, one pathogenic and one likely pathogenic. So this case was also a skeletal celiopathy. So based on these two cases, we actually recently started offering a new panel that includes the 29 kind of general skeletal dysplasia genes plus the 19 skeletal celiopathy genes. And this would be for cases where, you know, you think it could possibly be a celiopathy or you just want to be more comprehensive and include, you know, all those celiopathy genes as well. We're also considering adding DYMC2H1 to the skeletal dysplasia extended panel. This particular gene accounts for about 50% of these skeletal celiopathies. So you know, we think putting that on the extended panel, we still think it's probably a pretty rare cause of prenatal skeletal dysplasias, but, you know, we think putting that on the panel could increase the clinical sensitivity a little bit more, especially in those cases where the babies don't actually have polydactyly. So just to sum up the presentation, our target panel has a clinical sensitivity of almost 85% for severe lethal prenatal skeletal dysplasias. It has a low rate of variance of uncertain significance, which is important in this prenatal setting. There were seven genes that caused 95% of the cases of prenatal skeletal dysplasias. These were FGFR3, COL2A1, COL1A1, COL1A2, ALPL, SLC26A2, and SOX9. And when you're testing these genes, particularly the collagen genes, you really need to have gene-specific guidelines for variant interpretation. If you want to, you know, be able to efficiently categorize these variants. And then we ended with talking about how evaluating, updating your targeted panels can be helpful for increasing your clinical sensitivity. So here are some of the references that I, you know, specifically noted in the slides. And then we just wanted to thank all the patients, families, physicians, and genetic counselors who have used our services, and also some of the staff at CTGT who had worked on these cases and, you know, made this work possible. So now we can move into the question portion. And it looks like there have been some questions that were asked. Okay, so the first question that I'm seeing here, the question was, were there other issues in the 13 cases of achondroplasia that made them lethal? So, we don't expect the cases of achondroplasia to be lethal. They were identified prenatally, but we expect them to just sort of have a normal achondroplasia presentation and, you know, clinical history. But it has been published that you can, you do sometimes, you can uncover achondroplasia in the prenatal setting, and that was the case for those particular cases. So, there's a second case, there's a second question that's can you identify skeletal dysplasias using noninvasive prenatal screening? So, we actually do not do that technology here, but I would say that the answer is yes. I think there actually is a panel that at least has SGFR3 on it, and there probably might be some other genes that are being done with noninvasive testing as well. But, I mean, definitely the answer is yes, especially for those, especially for changes that you would expect to be de novo. So, the third question is, in patient case one, was fetus two observed to have a skeletal dysplasia versus ultrasound? So, this was the case where we think there's germline mosaicism. So, the collagen 1A1 variant was found in the first baby, and then the second baby was sent for testing. We did not get any clinical information for that second child. I'm assuming it was probably sent pretty early. So, it may have been before features would have been present, but I actually, you know, can't tell you for sure because we didn't get any follow-up information. We didn't get any specific clinical information when the test was sent in, and we didn't get any follow-up information on that fetus. There's another question here that's, how do you confirm your large NGS-detected deletion or duplication? There's a couple. It kind of depends on the situation. There's a couple of different methods that we use. It's a really convincing deletion or duplication. So, some of these that you get off of, some of these deletion duplications that you call off the next gen are very clear. You can be pretty convinced that they actually are real deletions and duplications. So, many of those we basically just confirm by a repeat NGS test. So, it's basically confirming it just by the same method again. Some of our cases we have actually confirmed by array CGH. We do have these exon arrays that we had previously run, and we've confirmed many of them using that technology. A few, we have had a few very small, like one-exon deletions that we were able to confirm by PCR. That's actually very dependent on how the gene is set up and basically how far the exons are away from each other, but we have been able to confirm a few just by PCR. And then we have one more question. Do you determine your VUS rates at the specimen case level or at the overall variance level? For the purposes of this particular talk, we were considering the VUSs at the case level. So, if a case had a definite mutation and a VUS, I considered it a positive case, and it wasn't considered in the VUS category. And there were a few cases like that. So, the ones that were considered as VUS were cases where that was always found, just the VUS. One more question came in, what metrics do you use to determine that a CVS call is likely to be real? You obviously validate your technology and you kind of base your criteria on, you set your parameters and base it on your set of deletions that you know are real and then how you can call that. So I can tell you that this particular CMV caller that we used is, in our validation, it detected 100% of all heterozygous variants. It did have some trouble with mosaic variants, so we had some mosaic variants in our validation set too. It only detected one of the three mosaic variants that we had in our set, but I mean you basically have to pick your criteria based on your validation set and how well you can validate things that you have already found and know to be real. And so there's a follow-up about sensitivity to single exons on validation. So this is the one area where this caller does have some issues, so we do see some false positives sometimes when it's only a single exon that is called. And we use some metrics to try and weed out some of those false positives, and some of them are just, you know, to see if there is a heterozygous variant in the region. So if there's a heterozygous variant, there's obviously not going to be a deletion of that region. Another thing is you can actually, like, look at your reads and see if you have split reads and breaks in your reads to see, you know, if it looks like there's a deletion there or not. And we do actually – all of the single exons that we actually have called have been confirmed by another method. So for things larger than a single exon, the caller has actually performed very well for us. It's the single exons that we do, when we see those come up, we do worry about them a little bit, but we do confirm all of those single exons before we actually report it on our court. Thank you. So, that looks like all of the questions that we have. So, ACMG just has asked me to announce that they are accepting abstracts right now for the 2019 ACMG Clinical Genetics Meeting. So, please submit your abstracts. And then, again, just want to acknowledge QIAGEN for their support of this conference.

Connective Tissue Gene Test

Keri Brown

ACMG Genomics Case Conference

skeletal dysplasias

targeted NGS panels

FGFR3

COL2A1

ALPL

QIAGEN