Genome‐wide non‐invasive prenatal screening for all cytogenetically visible imbalances

Non-invasive prenatal testing (NIPT) through the analysis of cell-free DNA (cfDNA) in maternal plasma has been shown to be highly effective in screening for trisomies 21, 18 and 131-3. Based on limited validation data, the testing has also been applied to screening for fetal sex chromosome abnormalities and establishing fetal sex3. NIPT is also available to screen for a specific set of submicroscopic recurrent microdeletions associated with potentially severe clinical phenotypes4, 5. A number of proof-of-principle studies, two recent clinical reports and a retrospective analysis of a large set of NIPT referrals suggest that this testing can be further extended to a broad range of other chromosome abnormalities that would be detectable through karyotyping of a chorionic villus sample (CVS) or amniotic fluid cells6-18. This includes unbalanced chromosome translocations, rare autosomal trisomies (RATs), other large imbalances and supernumerary marker chromosomes (SMCs). Individually, each of these imbalances is rare, but collectively they are relatively common, particularly in prenatal diagnostic specimens19. The clinical consequences of these abnormalities may include spontaneous fetal loss, major fetal anatomic abnormalities and pregnancy complications. However, many chromosome abnormalities may also be clinically inconsequential, because the abnormal cell lineage is confined to the placenta (or present at a negligible frequency in fetal tissues), the chromosome abnormality is present in a phenotypically normal parent or the cytogenetic abnormalities do not involve a gene imbalance that materially affects phenotype. Consequently, the clinical utility of providing NIPT for all these additional imbalances in a genome-wide assessment is questionable20, 21. The fetal component of cfDNA is primarily derived from cytotrophoblasts and the same lineage of cells is routinely analyzed in prenatal cytogenetic testing of direct preparations (or short-term cultures) of chorionic villi22. The identification of rare cytogenetic imbalances or abnormalities that are confined to the placenta is often associated with complex genetic counseling, additional follow-up testing and uncertain pregnancy outcome23. Although NIPT and karyotype analysis of cytotrophoblasts may not necessarily be fully equivalent in their ability to identify cytogenetic abnormalities, it is useful to consider the specific chromosome abnormalities detected in cytotrophoblasts as an initial theoretical projection for the type of additional studies, clinical interpretation issues and counseling burden that will arise with the implementation of genome-wide NIPT. We have therefore reviewed a large dataset of CVS analyses under the assumption that NIPT will identify a similar spectrum of cytogenetically unbalanced abnormalities with the same sensitivity as karyotyping of cytotrophoblasts. Our analysis focuses only on those cytogenetically visible imbalances that are not included in current NIPT standard protocols (i.e. excluding trisomies 21, 18 and 13, and sex chromosome imbalances). We reviewed previously published cytogenetic results from 41 782 CVS analyses performed at a single laboratory (TOMA, Varese, Italy) in women in the first trimester24. The results were combined with an overlapping set of 45 867 CVS specimens from the same laboratory, in which chromosome mosaicism in cytotrophoblasts and/or mesenchyme and, if necessary, confirmation in amniotic fluid had been documented25. A detailed description of these two datasets is provided in Appendix S1. All cases of unbalanced chromosome abnormality identified in cytotrophoblasts were assumed to be detectable by genome-wide NIPT, including those that showed any level of mosaicism. Chromosome abnormalities that would potentially be detectable through NIPT were classified into the following groups: (1) non-mosaic RATs that involve any autosome, except trisomies 21, 18 and 13, that would be expected to be associated with first-trimester pregnancy loss26; (2) mosaic RATs with separate consideration of those with a risk for clinically significant uniparental disomy (UPD) (i.e. those involving chromosomes 6, 7, 11, 14, 15 and 20)27 and also separate consideration of trisomy 16 owing to the relatively high risk for fetal abnormality, intrauterine growth restriction (IUGR) and pre-eclampsia28; (3) other non-mosaic segmental imbalances including unbalanced translocations, derivative chromosomes and SMCs; and (4) mosaic segmental imbalances. Any mosaic imbalance was considered to be detectable by NIPT if the abnormal cell line was present in cytotrophoblasts, regardless of whether the abnormal cell line was also present in villus mesenchyme or amniotic fluid cells (i.e. confined placental mosaicism (CPM) types I and III and true fetal mosaicism (TFM) types IV and VI; Appendix S1). Rates of chromosome abnormalities were expressed as the number present in 100 000 women undergoing testing. The probability that a cytogenetic abnormality present in trophoblasts would be detected in amniotic fluid cells (TFM) was calculated25. We also determined which cases would require additional testing and estimated the proportion that were likely to be clinically significant based on available information as referenced below. Under the assumption that NIPT is fully equivalent to karyotyping of cytotrophoblasts, with no additional sources of false-positive or false-negative findings, the expected findings of a genome-wide NIPT are summarized in Figure 1. For every 100 000 women undergoing genome-wide NIPT testing, 473 (0.5%) tests would be positive for a RAT and would therefore have an indication for follow-up testing, preferably based on amniocentesis. Of the 473 cases, 72 (0.07%) would be non-mosaic. Based on current NIPT assays, the non-mosaic cases would not be distinguishable from those with mosaicism. However, the non-mosaic RATs and a low proportion of the mosaic RATs would result in first-trimester spontaneous fetal loss26, 29-31. Provision of CVS would be a suboptimal follow-up test for women with RAT-positive NIPT results because the abnormality had already been detected in cytotrophoblasts, and confirmation in amniotic fluid cells would be a consideration regardless of the finding in CVS mesenchymal cells25. At most, 401 (0.4%) mosaic RAT pregnancies would survive into the second trimester, and these would include seven cases in which TFM would be expected to be confirmed based on an analysis of amniotic fluid cells25. These seven cases would not necessarily show an abnormal phenotype in the abortus or newborn; only approximately two cases would show fetal abnormalities and only approximately two would be confirmed using conventional follow-up cytogenetic studies32. In approximately half (n = 214 pregnancies) of the surviving RAT-positive cases, the additional chromosome identified by NIPT would be 6, 7, 11, 14, 15 or 20 and, although the finding would not be confirmed by amniocentesis, these cases would be at risk for a clinically significant imprinted gene disorder arising through UPD27. Follow-up testing would therefore require single nucleotide polymorphism-based chromosome microarray analysis or another approach that would identify UPD. Approximately four (1.9%) of the 214 cases would be expected to have a UPD25. A risk for fetal UPD is present in all cases in which trisomy corrects to disomy and, theoretically, this could lead to the expression of an autosomal recessive disorder through a reduction to homozygosity of a recessive allele inherited from one parent. Although, in routine prenatal diagnosis, UPD is usually excluded only when abnormality involves a chromosome with imprinted genes, follow-up of all RAT-positive cases with UPD studies might be a consideration for all cases, regardless of the chromosome involved. The risk for an autosomal recessive disorder has not been quantified but can be expected to differ depending on whether the trisomy arises primarily through meiotic or mitotic error33, the gene region involved and the prevalence of carriers for the specific disorders. The 401 RAT-positive cases would include 22 cases in which the RAT would involve an additional chromosome 16. In these cases, the risk for IUGR, fetal abnormality and pre-eclampsia is high and the risk is not entirely removed if a subsequent amniocentesis shows only normal cells28, 34, 35. A reliable estimate of these risks is not available. For all RATs involving chromosomes other than 16, there is also a possibility that these pregnancies are at increased risk for adverse outcome owing to placental insufficiency caused by the abnormal placental cell populations. However, evidence of a clear association between CPM and IUGR or pregnancy complications has been conflicting and this issue remains controversial33, 36-43. For these cases, increased surveillance to detect IUGR, fetal anatomic abnormalities and pregnancy complications seems to be prudent. In 100 000 women undergoing genome-wide NIPT testing, 300 (0.3%) cases would be positive for a segmental chromosome imbalance. This would include 50 cases with non-mosaic unbalanced structural rearrangements, 22 with non-satellited de novo SMCs, five cases of dicentric chromosomes and five complex unbalanced karyotypes. For these 82 cases, confirmation of the imbalance through invasive testing would be indicated, and generally studies on the parents would be indicated to establish whether they are carriers of a balanced chromosome rearrangement. In some cases, notably those with smaller or de novo imbalances, chromosome microarray analysis might be indicated to better delineate the imbalance. With the exception of some of the cases with SMCs, all cases with unbalanced karyotypes would be expected to be associated with an abnormal fetus and many would be expected to show ultrasound abnormalities. Of the 22 cases with de novo SMCs, approximately seven (32%) would be expected to be associated with fetal abnormality44. The 300 cases with segmental chromosome imbalance would also include 46 cases with a true mosaic unbalanced karyotype (15 with a de novo unbalanced translocation and 31 with a marker chromosome). The phenotype in these cases would be unpredictable, since it would depend on the gene content and the proportion and distribution of abnormal cells in the fetus. The remaining 172 of the 300 cases with segmental imbalance would show a normal result on amniocentesis, and the partial chromosome imbalance would therefore be interpreted as a confined placental mosaicism. For these cases, the risk for UPD should be considered very low and the risk for placental insufficiency is also probably very low. As discussed above for mosaic RATs, some additional surveillance of the pregnancy might be warranted. In summary, based on the modeled data presented here, an equivalent genome-wide NIPT that identified all these abnormalities would be expected to have an additional screen-positive rate of approximately 0.8% beyond that of the current NIPT, which focuses on trisomies 21, 18 and 13 and sex chromosome abnormalities. This includes approximately 0.5% of cases with a RAT and 0.3% of those with a segmental aneuploidy. Of these cases, approximately 0.1% would be associated with early fetal loss owing to a non-mosaic RAT. Of the remainder (0.7%) cases, following amniocentesis, approximately 0.06% would be associated with an unambiguous diagnosis of fetal abnormality (non-mosaic unbalanced chromosome rearrangement or clinically significant UPD). All other cases would be associated with either a mosaic chromosome abnormality with a highly variable risk for fetal abnormality or an unconfirmed result that would still retain some degree of residual risk, even after amniocentesis. The main reason for the uncertainty is mosaicism. Chromosome mosaicism is a common finding in prenatal diagnosis and is widely recognized as being highly problematic from a genetic counseling perspective. Even for the commonly encountered mosaicisms, there is considerable diversity in phenotype, and the proportion of abnormal cells detected on CVS or in amniotic fluid cell samples is a poor guide to the proportion present in fetal tissues45. For each specific cytogenetic abnormality that is encountered as a CPM, the risk for UPD and associated likelihood of a single-gene disorder are poorly defined. There is substantial evidence that a placental trisomy 16 cell line can result in IUGR and pre-eclampsia, but for all other specific findings in CPM the association is plausible but unproven. The essential difficulty is the paucity of data that allows clinical interpretation of many of the specific abnormalities seen in cytotrophoblasts. It is important to recognize that our projections for the performance of genome-wide NIPT only serve as an illustration of the issues arising in the testing and do not necessarily reflect actual performance of the test. Rates of abnormality will depend on the prior risks of the referral population. In addition, NIPT is probably reflective of a broad distribution of apoptotic trophoblasts in the maternal circulation, while cytogenetic analysis of trophoblasts reflects the cells present in a discrete biopsy obtained at an invasive procedure22. Site-specific variation in the proportion of abnormal cells has been documented46-48. Our modeled performance of expanded NIPT also does not consider false-negative findings. Reasons for false-negative results could include cases of mosaicism in which the abnormal cell line is present in the fetus but not detected in cytotrophoblasts (TFM type 5) or low-level mosaicism not sufficiently represented in the cytotrophoblast to be detected with current technologies (TFM types 4 and 6)49. The sensitivity of NIPT for smaller imbalances and lower levels of mosaicism has not been established across all genomic regions and for the full range of fetal DNA fractions typically encountered at the time of NIPT. Two clinical studies that present genome-wide data show large differences in the proportion of additional positive tests, suggesting that detection may well depend on the specific methodology applied by each laboratory16, 17. A genome-wide NIPT can also be expected to be associated with additional false-positive results compared with NIPT for a limited set of chromosomes. This would include false-positives attributable to the presence of a vanished twin. Given the high rate of twin conception and early loss of aneuploid conceptions, the number of such cases could be considerable50. Expanded NIPT that interrogates for a broad set of subchromosomal regions can also be expected to identify additional maternal imbalances, some of which may be clinically significant51. These include constitutive as well as acquired abnormalities that can be associated with malignancy52. Is there clinical utility in genome-wide NIPT? Can genome-wide NIPT provide clinically actionable indicators for pregnancy complications? It is clear that this testing can potentially detect some additional clinically significant unbalanced chromosome abnormalities, which would otherwise be undetectable except through an invasive test or, perhaps, ultrasound abnormalities. However, this needs to be balanced against a much larger number of cases in which the outcome is uncertain, even when follow-up invasive testing and ultrasound are offered. This differs fundamentally from offering NIPT for the common autosomal trisomies and recurrent microdeletions, in which phenotypes are understood and follow-up CVS or amniocentesis nearly always provides a definitive diagnosis. A crucial question is also whether genome-wide NIPT would lead to additional terminations of normal pregnancies. In a recent review of 89 817 referrals for NIPT, Pertile et al.18 noted 399 (0.4%) cases with findings consistent with an additional imbalance beyond those involving chromosome 21, 18, 13, X or Y. Of a subset of 57 cases with outcome data, 24 were associated with pregnancy loss (missed abortion, miscarriage or intrauterine fetal demise). Of the 33 continuing pregnancies, there were four cases of single copy number variants (phenotypes not described), one dysploid result (not confirmed later in pregnancy), five RAT TFMs confirmed at amniocentesis (pregnancy outcomes or phenotypes not described), one case with UPD15 mat (Prader–Willi Syndrome), one case with congenital abnormality at birth (trisomy 9) and two cases of IUGR. The actual incremental gain in unambiguous, clinically actionable findings associated with an abnormal fetus was therefore low. The use of NIPT to screen for RATs has been discouraged on the basis that detection of lethal abnormalities results in unnecessary diagnostic and termination procedures53. Cost and availability of resources to provide pre-test counseling and manage women with positive screening results should also be considered with respect to clinical utility. Based on these considerations, it has been recommended that more clinical trials are carried out before genome-wide NIPT is implemented into clinical practice20, 54. P.B. is a consultant for, and holds stock options in, Natera, Inc. F.R.G. is a full-time employee of TOMA laboratory without ownership shares. She is an expert panel member for Roche. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.