Founder Stories: Antonio Capalbo, PGT and Embryo Selection in the era of NGS

Hi everyone, and let me first of all thank the scientific committee of the European Society of Human Genetics for this very much appreciated invitation. The next 30 minutes I'm going to walk you through an update on the current status of knowledge about the origin and implication of aneuploidy in human preimplantation embryos, and what we can do to manage this issue by the use of next-generation sequencing in in the context of preimplantation genetic testing for aneuploidies. So to start with, I would like to quickly recap the paramount importance of meiotically derived aneuploidies in human reproduction. Very recently, in a study that was led by Professor Eva Hofmann, we got a, a direct observation from a large dataset of human oocytes donated from research in women across the broad spectrum of reproductive lifespan showing that aneuploidies can affect between 20 to 85% of human eggs, and that this rate of aneuploidies, as you can see here, follows a U-curve that shapes and perfectly matches the natural fertility curve of women between the menarche and the menopause, clearly highlighting how meiotically derived aneuploidies are for sure the main determinant factor for reproduction in humans and also the main cause of unsuccessful IVF treatment cycles, particularly for women of advanced reproductive age. In the same study, we could also gather some evidence about distinct age-dependent error types that shape this U-curve of aneuploidy in human oocytes, with meiotic I nondisjunction error being far more prevalent in young ages, while, uh, premature separation of sister chromatids, um, at meiosis I, uh, explaining the rise in aneuploidies that we commonly observe with advancing female age. Similar findings have been also observed in human embryo samples at the blastocyst stage, where as you can see in this figure, aneuploidies at the blastocyst stage of development can range between 20 and 80-90%, and further confirming and supporting the fact that aneuploidies also in embryos are primarily inherited and derived from aneuploidies that originate during female meiosis. Furthermore, observation at the blastocyst stage of human development suggested and showed another important finding, that is that the whole spectrum of aneuploidies is compatible with development to this stage, and that therefore all the design of PGTA strategies should account for this information, developing methodologies that are capable of detecting both monosomies and trisomies for whole chromosome as well as segmental imbalances. Meiotically derived aneuploidies are expected to be present throughout the embryos, throughout all the cells making up the developing preimplantation human embryo, and are not supposed to be self-corrected during the preimplantation development. Sometimes we have here in our field a false common belief that meiotically inherited aneuploidies frequently going through self-correction event during the preimplantation development. However, this is not supported by the evidence, because this is a measurable outcome that we can appreciate by looking at the frequency of uniparental disomy events that we can observe at the blastocyst stage of development. This large study, very large study, was performed by a group led by Nathan Treff back in 2014, showing that instances of uniparental disomies could be observed in as low as 0.6% of human embryos at the blastocyst stage, supporting the fact that meiotically derived aneuploidies are fixed through development and are also associated with certain outcomes. Very elegant recent studies, the group by Richard Scott and colleagues have provided very robust evidence that when transferred, uniformly aneuploid embryos do not make babies. In the context of a non-selection trial where the blastocyst stage human embryos were biopsied and where the samples were saved for blinded aneuploid testing, the embryos were transferred prior to any genetic analysis. So both euploid and aneuploid embryos were transferred without knowing their aneuploidy status, and this allowed a very accurate estimation of the positive and negative predictive value. In particular, what the authors found out was that none of the 102 embryos diagnosed as uniformly aneuploid resulted in a sustained implantation or delivery, while the positive predictive value of the PGTA blastocyst stage was extremely high and outstanding, showing that 64-65% of them resulted, of the euploid embryos, resulted in a delivery. In the same context, in the same study, they also could gather some evidence about the fact that trophectoderm biopsy has not detectable adverse impact on sustained implantation by comparing the study group with an age-matched control group where biopsy was not performed. And as you can see here in this figure, both the biopsied and the untested embryos show exactly the same overall sustained implantation rate, supporting the concept that when properly performed, trophoblast biopsy is not expected to lower the reproductive potential of human embryos. And in support of this data, we have today a broad clinical experience from several different settings and countries consistently showing that euploid embryo transfer can result in a higher live birth rate, in lower pregnancy loss risk, as well as in a reduced risk of having a chromosomally abnormal pregnancy. In some settings, PGT-A at the blastocyst stage has been also shown to be cost-effective as compared to standard care, and at present, even though we still lack prospective randomized trial power for this outcome, the cumulative live birth rate in all large retrospective studies has been shown to be similar between standard IVF and PGT-A treatment, thus providing reassuring evidence that the cumulative impact on biopsy and the potential level of false positive results that we get from PGT-A blastocyst stage when properly performed is not clinically relevant. To get to this point, we and the field of PGTA went through a significant evolution of the technology, not just at the embryo biopsy stage but also in the genetic laboratory, where we started with fluorescence in situ hybridization, looking for few chromosomes in interphase nuclei with very questionable technical accuracy. And then we move through more comprehensive chromosome testing technologies able to capture the whole spectrum of aneuploidies in human embryos, like microarrays or quantitative PCR. And finally, we incorporate next-generation sequencing technologies also in our practice. While I think that the most objective advantage that NGS has provided is in the field of PGTA so far has been the increased throughput and the consequent decrease in analytical cost per sample that has allowed today many more couples and patients undergoing IVF to benefit from their embryo being tested for aneuploidies. So the standard workflow of a typical NGS experiment into the context of PGTA consists of barcoding and multiplexing hundreds of samples into the same reaction, that is then followed by a very shallow next-generation sequencing, where we use 100 to 200k mapped reads to make the reporting of all chromosome aneuploidies and segmental imbalances down to 2 to 5 megabases of resolution. An interesting evolution and implementation of NGS methodologies in the recent years has been to try incorporating some level of genotyping information into the sequencing analysis. And this made possible to expand the application range of NGS toward the detection of contamination, for instance, or embryo fingerprinting in case we have some mismatch between the PGT and prenatal diagnosis. As well as we could also evaluate the ploidy constitution of developing human embryos, which in some cases can result in very important information for, um, the inclusion of genotyping data into our standard NGS workflow has made possible to have new opportunity for treatment in the IVF setting, particularly where we can today assess directly the ploidy configuration of embryos that that are generated from abnormally fertilized oocytes that are supposed to bring ploidy abnormality, and that they are usually and commonly discarded from clinical use because of this concern. Today we can assess directly the ploidy configuration of blastocyst stage embryos derived from these abnormally fertilized oocytes, and we show that a significant proportion of them are indeed biparental diploid can be used to generate normal and healthy live births, providing a very nice example, I think, of how the development of new genetic technologies can directly improve the treatment of IVF patients, in this case by providing additional embryos that are available for transfer following genetic testing that otherwise would be discarded in the absence of any improved genetic testing technology. Remarkably, next-generation sequencing has been also suggested as an improved technology for the diagnosis of chromosomal mosaicism in trophoblast biopsies, leading to a massive controversy and debate in our field, particularly about the capability of this new development to bring any level of clinical utility for patients undergoing a PGT-A treatment cycle.

The capability of detecting mosaicism in trophoenterobiosis started from the assumption that intermediate chromosome copy number values can unequivocally predict mosaicism in blastocysts. And this assumption has been verified on experimental models where colleagues in the field start taking and looking at fibroblast cell lines that are very stable and reproducible, and they start mimicking mosaicism by mixing different proportion normal and abnormal cells. There's no doubt that when we apply next-generation sequencing technologies and copy number analysis to these very stable and reproducible models, we observe a linear trend and correlation between the chromosome copy number value and the expected mosaicism rate in these fibroblast cell lines. However, the translation of algorithms that were developed on these reproducible experimental models has not been as effective and straightforward for deriving clear diagnostic criteria when applied to embryo biopsy placement. Indeed, the use of this validation system combined with the use of subjective threshold for mosaicism diagnosis has generated a critical inconsistency in detection rates reported across laboratories, with values that range from 2% and even up to 40% of mosaicism being reported in human blastocyst stage. There's no doubt that the experimental variability is significantly different from this, from the research to the clinical setting, where in our clinical setting we used to experience TE biopsies coming from embryo of different quality. They may be of different cellularity, from 2 to 20 cells, and we also have changing condition over time, such as the use of different reagent batches, buffer volumes. Also, shipping conditions may vary over time, as well as at the end of our diagnostic workflow, we may have some different interpretation that is due to the human factor, where I may consider this plot and designate this plot as non-informative, while other colleagues may interpret this profile plot as a mosaic embryo. But probably the most important parameter that is explaining the inconsistent data reporting that we have experienced in our practice about the mosaicism rates at the blastocyst stage is due to the use of different thresholds for mosaicism classification. We can be more stringent and report mosaicism when the chromosome copy number value falls between the 30 to 70% of variation, and in this case, we experience 4 to 5% of human blastocysts being reported as mosaic. Or, in an attempt to increase the analytical sensitivity toward mosaicism, best guess classification, we may apply less stringent parameters for mosaicism designation and reporting mosaicism when the chromosome copy number values fall between 20 to 80%. With this second, less stringent approach, for sure, our analytical sensitivity toward mosaicism, the true mosaicism cases, increases, but as a consequence, we may also expect a significant overcalling of mosaicism, and in particular, we may expect to experience a lower accuracy for distinguishing between the uniformly euploid and aneuploid embryos, which is the primary objectives, as we discussed at the beginning of my presentation, of a PGTA approach, that is, deselecting uniformly aneuploid embryos from transfer. In particular, indeed, when we see the reporting of centers using this threshold, we observe that mosaicism is reported in a range that follows between 20 to 40% of human embryos being mosaic. And no need to say that there is not a biological difference. This is not a biological condition that is changing between one center to another, but in this case, mosaicism is actually a consequence of different laboratory procedures and criteria that are used for diagnosis. As you can see in this table, also the use of different threshold for MOSAICS classification can severely impact the prediction of the ICM outcome in comparison with the trophoectoderm diagnosis. Indeed, when low stringency criteria are used for MOSAICS classification, like the 20-80% thresholds, then we have observed in several studies, independent studies, poor performance in predicting the ICM chromosomal constitution, while the use of more stringent criteria or mosaicism classification at all in different and independent studies have consistently resulted in a high performance, high accuracy of prediction of the ICM chromosomal constitution in a very reproducible way. That means that mosaicism in embryos is likely to be present in a range between 2 to 4% of human preimplantation embryos at the blastocyst stage, and is often overrepresented in some studies using weak criteria for X classification. Different considerations are required for segmental imbalances that seems to originate primarily from mitotic segregation error in early embryogenesis. Indeed, as we observe in this multifocal embryo biopsy model, we observed that around 70% of segmental aneuploidies were indeed associated with mitotic consistent patterns and configuration, while only 30% of them show a meiotic configuration where the same abnormalities was consistently detected across all the 5 biopsies collected from the same embryo. And here it comes, my first take-home message, where meiotic aneuploidies are expected to be for sure the main abnormality that found in preimplantation embryos and can be effectively detected by next-generation sequencing protocols using proper criteria for aneuploidy classification. Mosaicism is expected to account for a minority of embryos at the blastocyst stage, around 2 to 4%, but it can be very easily overestimated by the application of inaccurate methods and weak criteria for mosaicism classification. Also, we need to acknowledge that there are some levels of technical and biological variation that may challenge the proper diagnosis and reporting of mosaicism, that in a way that probably the best way to report these findings that are intermediate chromosome copy number value should be as to report as putative mosaic. So the next question that comes now is whether these mosaicism consistent findings are providing any level of clinical utility for infertile patients in the context of a PGTA treatment cycle at the blastocyst stage. So first of all, we have to acknowledge that when we report mosaicism back to patients in PGTA cycle, this comes with really significant consequences. Particularly, we have to acknowledge that the couples will need to undergo an additional session of genetic counseling, they will have to evaluate the replacement of these embryos in the context of increased anxiety, distress. They will be, in case they transfer these embryos and they get pregnant, they will, this will increase the adoption of invasive prenatal procedures such as amniocentesis. But the worst outcome, I think, is the potential wastage of perfectly normal and viable embryos, because these embryos are usually perceived as chromosomally abnormal and always they are discarded from, and from clinical use. Indeed, what we've seen from one of the largest experience of reporting mosaicism with less stringent classification criteria by Mooney and colleagues was that less than 3% of the embryos designated as mosaic were then used for replacement and embryo transfer. So is this practice based on evidence and on clinical utility? Well, so far we only have very few studies that have assessed the reproductive performance of putative mosaic embryos in the context of a retrospective analysis and showing a slightly lower implantation and live birth rate with the use of these embryos as compared to euploid ones. However, these studies are all limited by the retrospective nature. Indeed, particularly there is a common selection bias when assessing the outcomes of these studies, that is that the mosaicism performance and reproductive outcomes have been assessed in very selected and poor prognosis patient population because these embryos are usually transferred in patients who fail with previous euploid embryo transfer or who produce all aneuploid embryos. Thus, there is a selection, a strong selection for poor prognosis patients where we are going to measure the outcomes of putative mosaic embryos. Furthermore, it still remains a possibility that some of these embryos, these mosaic embryos, particularly high-range mosaic ones, are misclassified as mosaic and they are indeed uniformly aneuploid embryos with a little bit of technical variation, making them to be called as high-range mosaic and explaining why these resulted in a lower implantation rate as compared to the uniformly euploid embryos. So in order to avoid selection bias in the assessment of reproductive performance of putative low and medium-grade mosaic embryos, we have recently completed a prospective double-blind and non-selection trial where, to avoid selection bias, Embryos showing low and medium grade mosaicism were blindly reported as euploid and were transferred using single embryo transfer based on standard morphological features. So all these embryos, both the euploid and the mosaic, were reported as a single category back to the patients and the physicians. Embryos were blindly allocated soon after the embryo transfer procedure occurred in 3 distinct categories: euploid, low grade mosaic, and medium grade mosaic. And then reproductive outcomes were assessed soon after. As a first step in our study before running the clinical trial, we have been disaggregating that to 5 different portions. 91 human blastocysts donated for research, they were disaggregated in 4 different trophectoderm equally sized portions and the ICM, and we've been assessing what was the predictive value of the reference TE outcome with respect to the whole embryo, we find out that the low and medium mosaic embryos were almost always confined to a single trophoblast portion not affecting the ICM, while the high-range mosaicism, the ones above 50% of chromosome copy number value, were often affecting the vast majority or the whole embryo including the ICM. That's why in our clinical trial we decided to report as normal the embryos that were showing a putative mosaic finding up to 50% of chromosome copy number value. And see the main outcome of our clinical trial that is clearly showing that both the low putative mosaic and the moderate putative mosaic category performed similarly in terms of clinical outcome as compared to the euploid embryos. In particular, they showed a similar positive pregnancy test, biochemical pregnancy loss miscarriage rate, and at the end, the same live birth rate with a non-inferiority endpoint that was met as expected. The logistic regression model did not find any association between the PGTA category and the main outcome measure, as well as the neonatal outcomes were similar between the three groups. Next, we have evaluated if during the course of the pregnancy we could find any instance of mosaicism, and we have followed up all the cases of miscarriages and subsequent cytogenetic analysis of the product of conception, as well as of invasive prenatal diagnostic procedures performed in the three groups. The mosaic embryos show no evidence of mosaicism, in the follow-up, while the only instance of confined placental mosaicism was detected for one case in the euploid embryo transfer group. We had also the opportunity to enroll 38 families in a more detailed cytogenetic analysis by SNP array of the newborns and the parental DNA in order to investigate any potential instance of uniparental disomy and mosaicism in babies derived from mosaic embryo transfer., and in all the 38 newborns assessed, we didn't find any evidence of chromosomal abnormalities. And at the end of the trial, we modeled the situation where mosaic embryos were excluded from clinical use because perceived as chromosomally abnormal. As you can see in the pie chart, Low and medium-grade mosaic embryos account for around 15 to 20% of the whole embryos produced during the course of an IVF treatment cycle, and if we exclude the live births obtained from the transfer and the use of these embryos, you can see that 1 in 3 babies will not be born for any gain of clinical utility. So our findings are supported and corroborated by the fact that mosaicism is not increased in IVF-derived pregnancies as compared to spontaneous conception. This has been extensively demonstrated in several studies. Also, the positive predictive value of a mosaicism diagnosis in the general clinical experience of PGTA is extremely low. Today we have up to or over 2,500 mosaic embryo transfer procedure performed worldwide, and so far only one case has been confirmed in the newborn with the same mosaic finding that was previously detected at the embryo level. And on the top of that, it's worth mentioning that this case, this baby, had also no phenotype associated with this condition. And thus, our study is also supported by evidence in animal model, where chromosomally abnormal cells in chimeric mouse model of mosaicism have been shown to become progressively depleted after blastulation. More recently also, the analysis of phylogeny of human placenta has shown the existence of early developmental bottlenecks that are able to normalize postzygotic mosaic aneuploidies in human embryos and in the early post-implantation stages.

So to conclude, I think we can reasonably state that meiotically derived aneuploidies are for sure the most common type of aneuploidy in human preimplantation embryos, and NGS in this context can provide a very powerful and particularly cost-effective technology for detecting meiotic aneuploidies and improving the IVF outcomes particularly in high-risk women. For what concerns the mosaicism issue, I think there is enough evidence today supporting that low-medium mosaic embryos are either artefacts generated by the increased analytical sensitivity of NGS, or they are extremely confined to a single trophoectoderm portion without ICM involvement and any major implication for the pregnancy. And accordingly, Today, reporting low, medium, rich mosaicism seems to have no elements of clinical utility, based also on the prospective non-selection study, while it can result in a significant wastage of potentially viable and otherwise healthy embryos. I think that, as it's been always performed and currently done in the practice of prenatal diagnosis, also in PGT mosaicism should just be acknowledged as a biological limitation of the this type of analysis, considering that it will not have a major impact in the clinical outcomes of PGTA cycles if properly managed and handled. And final claim I think is worth it for all the new developments of algorithms, criteria for aneuploidy classification in PGTA that should first provide evidence, a class I evidence, of preclinical validity and clinical utility before being incorporated into the clinical practice. This is just to avoid for the future the same issue and challenging that we faced with incorporation of poorly validated mosaic classification schemes in the practice of PGTA.

And what we can expect for the future? For sure, we can expect to expand the applications of next-generation sequencing, always in in the light of improved clinical utility for infertile couples undergoing IVF. In this context, we can think that the next development will involve the detection of de novo conditions that are associated to embryonic lethal genes or to fatal structural defects that can be anticipated in preimplantation genetic analysis. Or we may also consider the application and development of multi-omic profilings TE biopsies in order to capture additional level of genetic features that can be assessed in parallel with aneuploidies and can provide an improved embryo selection for our patients. However, it needs to be mentioned that these applications are at present still challenged by the outside errors in whole genome amplification of samples from single cells. Another interesting area of development is the application of next-generation sequencing for developing non-invasive frameworks for embryo selection based on cell-free DNA analysis from embryonic stem culture media that hold the potential to be also a revolutionary technology in our field if properly validated and shown to be effective in prospective clinical trials. So that's all on my side, and I wanted to thank you all for your attention and looking forward to answer to any question you may have during the live Q&A session. Thank you.