Reliable detection of subchromosomal deletions and duplications using cell‐based noninvasive prenatal testing

Abstract Objective To gather additional data on the ability to detect subchromosomal abnormalities of various sizes in single fetal cells isolated from maternal blood, using low‐coverage shotgun next‐generation sequencing for cell‐based noninvasive prenatal testing (NIPT). Method Fetal trophoblasts were recovered from approximately 30 mL of maternal blood using maternal white blood cell depletion, density‐based cell separation, immunofluorescence staining, and high‐resolution scanning. These trophoblastic cells were picked as single cells and underwent whole genome amplification for subsequent genome‐wide copy number analysis and genotyping to confirm the fetal origin of the cells. Results Applying our fetal cell isolation method to a series of 125 maternal blood samples, we detected on average 4.17 putative fetal cells/sample. The series included 15 cases with clinically diagnosed fetal aneuploidies and five cases with subchromosomal abnormalities. This method was capable of detecting findings that were 1 to 2 Mb in size, and all were concordant with the microarray or karyotype data obtained on a fetal sample. A minority of fetal cells showed evidence of genome degradation likely related to apoptosis. Conclusion We demonstrate that this cell‐based NIPT method has the capacity to reliably diagnose fetal chromosomal abnormalities down to 1 to 2 Mb in size.

copy number variants (CNVs). cfDNA-based NIPT is currently only recommended for common fetal aneuploidies but not for screening for microdeletions/duplications in statements from professional societies. 1,2 During a normal pregnancy, only 5% to 20% of the total cfDNA pool is of fetal origin, referred to as the fetal fraction. 3 The current NIPT methodology thus relies on identifying a chromosomal abnormality in an amalgamation of maternal and fetal DNA fragments, which can lead to false positive results, and its performance can be affected by a below average fetal fraction (<4%). cfDNA-based NIPT is also potentially influenced by maternal chromosomal mosaicism or maternal malignancies. 4 It thus remains a screening test requiring diagnostic testing for confirmation of positive results. Since the clinical implementation of cfDNA-based NIPT, the number of Chorionic villus sampling (CVS)/amniocentesis procedures performed has decreased substantially over recent years. [5][6][7] While this reduces the procedure-related risk for pregnancy loss, it also leads to failure to diagnose clinically significant subchromosomal abnormalities such as deletion and duplication syndromes, easily detectable with chromosomal microarray (CMA), the current standard diagnostic test of DNA extracted from amniotic fluid or chorionic villi.
In contrast, cell-based NIPT offers a more attractive alternative if it can be performed reproducibly and at reasonable cost. Although cell-based NIPT also has limitations such as the risk of too few cells recovered, the specific isolation of multiple individual fetal cells from the maternal circulation offers the advantage of providing pure fetal DNA, free of maternal DNA contamination. As such, the fetal genome can be analyzed at a higher resolution, allowing for the detection of CNVs as small as 1 to 2 Mb in size. This would thus allow for increased accuracy and improved positive and negative predictive values compared with cfDNA-based NIPT in detecting microdeletion syndromes that are responsible for a range of rare conditions including some cases of autism and intellectual disability and can be detected in up to 1.7% of amniotic fluid or CVS samples from pregnancies without fetal anomalies. 8 Additionally, the analysis of multiple individual fetal cells from one sample yields data replicates, creating the potential for a higher test result confidence and to identify two different fetal genotypes in case of confined placental mosaicism.
Multiple recent publications [9][10][11][12] substantiate the feasibility of this approach and show concordant results with the corresponding microarray and karyotype data from invasive diagnostic testing, including a case in which the fetus was affected with a 2.7-Mb deletion of chromosome 15. 9 This was the first indication that cell-based NIPT has the capacity to perform at the resolution required for a clinically diagnostic prenatal test. The published methods describe the enrichment of fetal trophoblastic cells, by either depletion of maternal cells 9 or specific fetal cell positive enrichment, [10][11][12] for downstream genome-wide CNV analysis.
These publications, along with earlier reports of a high failure rate when attempting to isolate fetal nucleated red blood cells (fnRBCs), 13 and our own unpublished failed fnRBC attempts guided our decision to focus initially on circulating trophoblasts. Fetal nRBCs would certainly be attractive and avoid confined placental mosaicism, if successful and consistent recovery and analysis can be achieved, but this has not been demonstrated so far. In this follow-up report, we present multiple additional cases processed with the aforementioned depletion protocol. In this method, fetal trophoblasts are isolated and analyzed, after depletion of maternal white blood cells (WBCs), immunostaining, high-resolution scanning, visual verification of target cells, subsequent whole genome amplification (WGA) and low coverage (0.3-0.6X, 100 BP paired-end reads) single cell next-generation sequencing (NGS).

| Sample collection
Blood samples from pregnant women were collected after informed consent, under a protocol approved by the Institutional Review Boards of Baylor College of Medicine and Columbia University. The study subjects were recruited following routine prenatal genetic counseling, and in many cases also underwent CVS or amniocentesis followed by conventional chromosome analysis and/or CMA. Approximately 30 mL maternal venous blood was drawn into blood collection tubes containing a proprietary preservative (RareCyte) for trophoblast enrichment.
An additional 4 mL was collected in Ethylene-Diamine-Tetra-acetic acid (EDTA) tubes for maternal genomic DNA (gDNA) extraction and fetal cfDNA collection for fetal sex determination. When possible, a 2-mL blood sample in EDTA or saliva (Oragene) from the father was also collected. The tubes for fetal cell isolation were kept overnight at room temperature or shipped by overnight carrier at ambient temperature until further processing the next day. Maternal gDNA extraction (and paternal when available), fetal cfDNA extraction, and quantitative polymerase chain reaction (PCR) for fetal sex determination were performed as previously described. 9 Table 1 summarizes this sample series' characteristics, more detailed information is available in the Table S1.

| Trophoblast enrichment and isolation
Trophoblasts were enriched and stained as previously described, 9 with the inclusion of a maternal WBC depletion step for all samples. The What is already known about this topic?
• Fetal trophoblastic cells can be isolated from maternal blood and be used for the detection of fetal aneuploidies and copy number variants. The data on the detection of subchromosomal delestions and duplications is currently limited.
What does this study add? amount of depleted maternal WBCs ranged from 79.6% to 99.3%, with an average of 91.5% based on cell counting. After the depletion step, all nucleated cells were separated based on density centrifugation, fixed, and stained. All samples were spread on CyteSlides (RareCyte, 800 000 cells/well) and scanned on the CyteFinder instrument (RareCyte), as previously described. 9 All putative trophoblasts meeting the internally specified criteria (specific nuclear morphology, cytokeratin-positive staining in a defined pattern, and WBC marker CD45-negative) were picked as single cells and deposited in 2 uL of PBS in a PCR tube and stored at −80°C until further processing.
We applied this protocol to 125 samples from 122 women (three of whom had two blood draws), including 15 cases with a fetal aneuploidy and five in which a fetal subchromosomal abnormality was diagnosed.

| Whole genome amplification and genotyping
Single cells were processed with the PicoPLEX WGA kit (Rubicon, now Takara Bio) as previously described. 9 All purified WGA products were stored at −20°C until further use. fetal when the corresponding WGA products show two or more polymorphic alleles not present in the maternal gDNA; one such allele is scored as likely fetal. Based on data from male pregnancies, the prior probability that a female cell from a female pregnancy is fetal is 89%, and using a Bayesian calculation, the presence of one or two nonmaternal alleles gives a very high probability that the cell is fetal.

| CNV analysis
Sequencing for genome-wide copy number detection and subsequent analysis was done as previously described. 9 In short, library preparation was started from 300 ng of WGA product after which pairedend, whole genome sequencing was performed on a HiSeq platform A more detailed methods description is provided in the Data S1.

| Fetal trophoblast yield
A range of 0 to 38 trophoblastic cells per sample was identified by microscopy, corresponding to an average of 4.17 cells/sample or 0.18 cells/mL of maternal blood (Table 2). For the four twin pregnancy samples, 12,9,14, and 15 trophoblastic cells were found, corresponding to 0.50, 0.50, 0.58, and 0.63 cells/mL respectively. For 27 samples, some fetal cells appeared in a cluster of two to five cells (Table S1).
These grouped cells were mostly picked as one cluster and also analyzed as such. All other trophoblasts undergoing CNV assessment were analyzed as single cells.   Panel A in Figure 1 shows   showed a similar result to the gain observed in the CMA data. The serum screening for this case, performed as part of the patient's clinical care, had indicated a one in five risk for Down syndrome and a first trimester NT of 2.8 mm. Figure 4 illustrates the whole genome plot from a male fetus with a 6.0-Mb loss of chromosome 1. The mother was of advanced maternal age and ultrasound examination showed fetal cystic hygroma. A CVS sample was collected at 13 weeks and 6 days, as well as a maternal blood sample 1 hour after procedure for cell-based NIPT. As we could not recover any fetal trophoblastic cells from that sample, a redraw was done at 15 weeks and 4 days, and NGS data obtained from a trophoblast doublet from that sample were concordant with the CMA result obtained earlier. Figure 5 shows a case in which a 1.2-Mb gain of the X chromosome was detected in both the male fetus and his carrier mother. This sample was collected at 12 weeks and 1 day, before CVS. cfDNAbased NIPT was done for this pregnancy as well but gave no reportable result because, as noted by the reported laboratory, "due to technical or sample-related issues the data failed to meet the quality standards for interpretation (the patient reportedly has a large fibroid which is likely interfering with the analysis of the fetal DNA)."

| Mosaicism
We encountered a case of possible confined placental mosaicism in this data set. Study subject NIPT733 first underwent cfDNA-based NIPT, which indicated a "possible partial or full monosomy of chromosome 13." The patient then underwent a CVS procedure with CMA analysis on DNA directly extracted from the CVS sample, without prior culture, which showed a normal result. Simultaneously, we enrolled this subject for our study and were able to isolate three fetal trophoblastic cells. One of those clearly showed two small losses (4 Mb and 2.6 Mb, respectively) on chromosome 13 ( Figure 6), which we interpret as likely related to the cfDNA-based NIPT result and not coincidental. The other cells did not show these deletions, implying that confined placental mosaicism led to the false positive cfDNAbased NIPT result. This highlights the potential of cell-based NIPT to demonstrate two different fetal genotypes in case of mosaicism, which cannot be achieved by cfDNA-based testing.

| Quality of single cell NGS data
In general, most isolated fetal cells yield NGS plots of adequate quality. Occasionally, however, we obtain NGS data of inferior quality, usually associated with fetal cells for which only a weak DAPI signal is observed during microscopic validation. Figure 7 illustrates the whole genome plots for four cells isolated from the same sample: cell G200 has a lower DAPI signal and shows extreme copy number loss going to near zero copies for multiple large chromosome segments (chr1, 9) or entire chromosomes (chr7, 8,13,15). As discussed below, this was interpreted as apoptosis based on prior experience on such profiles. [14][15][16] Panel B in this figure illustrates the impact of pooling the data of an apoptotic (G200) and a high-quality cell (G1127): although the pooled data show a less extreme profile compared with G200 by itself, the overall data are unusable for CNV analysis.

| DISCUSSION
The results presented here further support the feasibility of cell-based NIPT as a clinical test, in agreement with other publications. 9-12 Our data set represents the largest to date showing successful, reproducible analysis of single trophoblasts by NGS-based CNV analysis with sample. The goal of this study was primarily to develop and refine an optimal protocol; thus, the diagnostic results are still exploratory and preliminary, and further clinical validation is needed. We acknowledge that the protocol is still evolving. Our ultimate goal for clinical testing is to have excellent NGS data on three to five cells, but we have observed cases for which a single cell provided valuable information.
Although 18.4% and 19.2% of the samples described here resulted in zero or one cell recovered, previously published reports on immunomagnetic bead enrichment indicate that fetal cell yield can be further increased. 10,11 Our own recent experience with a new protocol for positive trophoblast selection at present being explored in our lab, supports this (unpublished data). Any manipulation before or during the blood draw that could increase the number of trophoblasts in the maternal circulation, such as physical activity before blood collection, would also be helpful. However, exercise on a stationary bicycle provided a small increase in cell number. 17 Our data suggest that trophoblast recovery is unlikely to be influenced by the maternal BMI that is known to be a cause for cfDNA-based NIPT failure because of its lowering effect on the fetal fraction. The effect of a CVS/amniocentesis procedure on fetal cell yield for blood samples collected postprocedure has been described for fnRBCs. 18 In our series the effect seems limited, but more samples are needed for statistical comparison.
Although the isolated fetal cells have a distinct cytokeratin staining pattern that has been used to validate their fetal origin, 11  should be possible to obtain reliable independent data from both fetuses, assuming robust genotyping to distinguish cells from the FIGURE 7 Segmental loss of copy number, including homozygous loss for entire chromosomes suggestive of apoptosis. The whole genome plots for four cells from a male fetus compared with female fetal reference are shown in Panel A. The plot for cell G200 is extremely noisy while that of the other three cells shows a normal male profile. The DAPI staining for this cell is faint compared with the other fetal (white arrows) and maternal cells. The highly segmental loss of copy number including homozygous loss for entire chromosomes or arms of chromosomes as seen in cell G200 is interpreted to represent apoptosis (see Section 4). Panel B shows the substantial impact of pooling the data of an apoptotic cell (G200) with the data of a cell of high quality (G1127): Even though the copy number changes seen for G200 are somewhat compensated by G1127, the overall profile remains useless for copy number variant (CNV) analysis nevertheless. This next-generation sequencing (NGS) plot was generated by comparing this pool to a reference pool of two normal fetal cells [Colour figure can be viewed at wileyonlinelibrary.com] two twins. In case of monozygous twins with identical CNV findings, testing a larger number of cells offers increased statistical confidence that both fetuses have been studied. For rare cases in which only one of monozygous twins is carrying a de novo CNV, the two CNV profiles should be distinguishable if enough cells are tested.
If more than one fetal cell is recovered from a maternal blood sample, there is the option to analyze the cells individually, or as a pool of cells. We prefer to analyze them as individual cells, even though this slightly increases WGA and sequencing costs. Occasionally, trophoblast clusters (≥2 fetal cells together) are identified, and we speculate that these may be recently released from the placenta and be daughter cells from a recent mitosis. When these clusters remain attached during the picking procedure, they are analyzed as one. As WGA product quality varies from cell to cell, we do not wish to risk compromising the NGS profile of one good quality cell by pooling with another cell of inferior quality. For instance, pooling cells G200 and G1127 shown in Figure 7 give an unusable result, although cell G1127 alone is useful. Since data from multiple single cells can also be pooled post NGS and data analysis, we recommend against pooling cells before WGA. Single trophoblast data can help to address mosaicism, as is suggested by our current and earlier data, and they are also preferred for multiple pregnancies.
Single cells with severe and often homozygous whole or segmental chromosome loss as illustrated in Figure 7 are interpreted to represent fetal cells undergoing apoptosis. Every apoptotic cell has its own unique pattern of copy number changes (often to zero copies), affecting multiple entire chromosomes. In contrast, a true CNV is seen in multiple cells of confirmed fetal origin from the same sample. Kolialexi et al summarized extensive evidence of apoptosis in fetal cells in the maternal circulation including much higher apoptosis rates in Down syndrome and Turner syndrome pregnancies. 14 The apoptotic cells were suggested to be a source of cfDNA in mother's plasma. Bártová used FISH and TUNEL methods to demonstrate that "chromosomal territory segmentation precedes the formation of nuclear apoptotic bodies." 15 Particularly the chromosomal territory images shown in their publication 15 lead us to interpret these fetal cells as apoptotic. Similar large homozygous deletions were described in cancer cells and termed chromazemic cells that might represent dying cells. 16 NoTUNEL or Annexin V staining was done in the study presented in this manuscript.