Clinical Utility of SNP Array Analysis in Prenatal Diagnosis: A Cohort Study of 5000 Pregnancies

Background Single nucleotide polymorphism array (SNP-array) has been introduced for prenatal diagnosis. We aimed to evaluate the clinical value of SNP-array in the diagnosis of fetal chromosomal anomalies. Methods A retrospective study was conducted on 5000 cases tested by SNP-array, and the results of 4022 cases analyzed by both karyotyping and SNP-array were compared. Results SNP-array analysis of 5000 samples revealed that the overall abnormality detection rate by SNP-array was 12.3%, and the overall detection rate of clinically significant copy number variations (CNVs) by SNP-array was 2.6%. SNP-array identified clinically significant submicroscopic CNVs in 4.5% fetuses with anomaly on ultrasonography, in 1.6% of fetuses with advanced maternal age (AMA), in 2.5% of fetuses with abnormal result on maternal serum screening, in 2.9% of fetuses with abnormal non-invasive prenatal testing (NIPT) results and in 3.0% of fetuses with other indications. Of the 4022 samples analyzed by both karyotyping and SNP-array, SNP-array could identify all the aneuploidy and triploidy detected by karyotyping but did not identify balanced structural chromosomal abnormalities and low-level mosaicism detected by karyotyping. Conclusion SNP-array could additionally identify clinically significant submicroscopic CNVs, and we recommend the combination of SNP-array analysis and karyotyping in prenatal diagnosis.


INTRODUCTION
In the past few years, prenatal diagnosis has expanded from karyotyping and fluorescence in situ hybridization (FISH) to chromosomal microarray analysis (CMA). Standard G-banded karyotype analysis is the conventional cytogenetic technique used in prenatal diagnosis, which can detect chromosomal aneuploidies, polyploidies, mosaicism, and structural abnormalities such as balanced or unbalanced translocations or inversions, and deletions or duplications (>5-10 Mb). However, it has several disadvantages such as low resolution, long turnaround time and high rate of culture failure. The development of CMA allows us to identify chromosomal aneuploidy as well as submicroscopic imbalance on the whole genome with short reporting time and higher resolution, and it is recommended as a first-tier approach in prenatal diagnosis for detection of copy number variations (CNVs) in fetuses with structural anomalies observed by ultrasound (American College of Obstetricians and Gynecologists Committee on Genetics, 2013; Society for Maternal-Fetal Medicine et al., 2016). There are two types of CMA: array-based comparative genomic hybridization (aCGH) and single nucleotide polymorphism array (SNP-array). aCGH demonstrates the relative amounts of DNA from various areas of the genome by comparing the test DNA sample with a normal reference DNA sample hence could not detect triploidy, while SNP-array can identify triploidy as well as loss of heterozygosity (LOH) by hybridizing the test sample to the array platform and analyzing the signal intensity of SNP probes (Levy and Wapner, 2018).
Several meta-analyses and systematic reviews have reported that CMA detected clinically significant submicroscopic CNVs in 5.2-10% of fetuses with abnormal fetal ultrasound findings and a normal karyotype (Hillman et al., 2011(Hillman et al., , 2013Callaway et al., 2013). However, little is known about the prevalence of clinically significant CNVs in fetuses with structural anomaly in a single anatomical system or with abnormality of ultrasonic soft markers. Furthermore, for pregnant women who chose invasive prenatal diagnosis due to other indications such as advanced maternal age, abnormal result on maternal serum screening and abnormal non-invasive prenatal testing (NIPT) results, there is little information about the diagnostic benefits of SNP-array analysis (Hay et al., 2018).
In this study, we retrospectively reviewed a cohort study of 5000 pregnancies to evaluate the clinical utility of SNP-array analysis in prenatal diagnosis, and explore the improvement of diagnostic yield by CMA over karyotyping. In addition, the distribution of chromosomal abnormalities were also analyzed according to indications for invasive prenatal diagnosis.

Samples
This study was approved by the institutional Ethics Committee of The Affiliated Suzhou Hospital of Nanjing Medical University. All pregnant women choosing to participate in this study received genetic counseling and provided written informed consent before the invasive prenatal test. A total of 5000 samples including chorionic villi, amniotic fluid and cord blood were collected and analyzed successfully at the center for reproduction and genetics, The affiliated Suzhou hospital of Nanjing medical university, Suzhou, Jiangsu, China from September 2015 through February 2020. Table 1 listed the characteristics and indications for invasive prenatal diagnosis of all the 5000 samples. The mean age of the pregnant women was 32.04 years (range: 18-49 years), and the mean gestational age was 21.28 weeks (range: 9-34 weeks). Parental blood samples were also collected.
The received amniotic fluid samples were centrifuged immediately to collect amniocytes, and amniocytes were sent for culture if contaminating blood was visualized. 225 (4.6%) of the 4941 amniotic fluid samples were cultured. Chorionic villi were rinsed by saline solution three times and separated using needles under a dissecting microscope. Then genomic DNA of chorionic villi and amniotic fluid were extracted by QIAamp DNA Mini Kit (Qiagen GmbH, Hilden, Germany), and genomic DNA of cord blood was extracted by QIAamp DNA Blood Mini Kit (Qiagen GmbH, Hilden, Germany). Maternal cell contamination were ruled out for all the 5000 samples by short tandem repeat (STR) profiling using Microreader TM 21 (Direct) ID System (Microread, Suzhou, China), which could amplify 20 STR loci and the amelogenin gender marker simultaneously. In addition, 4022 pregnant women chose to perform SNP-array and G-banded karyotyping simultaneously, and 978 pregnant women chose to perform SNP-array only. G-banded karyotyping was performed for 4022 (80.4%) samples according to the principle of ' An International System for Human Cytogenetic Nomenclature, ISCN2013' as described previously .
This was a selective cohort of pregnant women during the study period of Sep 2015 to Feb 2020 who received SNP-array ± karyotyping only. There were women who had karyotyping only during the study period which were not included. There was no detailed information on the number of patients who had maternal serum screening or NIPT in the cohort.

SNP-Array Analysis
The SNP-array analysis was conducted on the Affymetrix CytoScan platform (Affymetrix, Santa Clara, CA, USA) according to the manufacturer's protocol. 250 ng genomic DNA was digested, ligated, PCR amplified, purified, fragmented, labeled and hybridized to the Affymetrix Cytoscan 750K array, which includes 550,000 CNV markers and 200,000 SNP markers. After washing, staining and scanning of arrays, raw data were analyzed by Chromosome Analysis Suite (ChAS) 3.2 (Affymetrix, Santa Clara, CA, USA). CNVs were called at a minimum length of 50 Kb containing at least 20 contiguous markers, interpreted and classified as pathogenic (P), likely pathogenic (LP), variants of uncertain significance (VOUS), likely benign (LB) or benign (B), according to the standards and guidelines released by the American College of Medical Genetics (Kearney et al., 2011). Some databases were used as reference resources, including ClinGen Dosage Sensitivity Map 1 , ClinVar 2 , Database of Genomic Variants (DGV) 3 , DECIPHER 4 , Online Mendelian Inheritance in Man (OMIM) 5 , and literatures in PubMed 6 . Microdeletions less than 200 Kb or microduplications less than 500Kb in size and benign CNVs were not reported unless they contained genes known to be associated with disease. We also Cord blood (no.) 13 2 0 0 1 15 a Other indications for prenatal testing include history of adverse pregnancy, parental genetic abnormalities, in vitro fertilization, medication use or toxic exposure during pregnancy, consanguineous marriages and parental anxiety. b As more than one indication may be associated with each sample, the total number of listed five groups is more than 5000. c Plus-minus values are means ± SD. a As more than one indication may be associated with each sample, the total number of listed five categories is more than 5000. b 2 cases of 69,XXX triploidy were identified in the group of anomaly on ultrasonography. c CNVs were classified into four groups according to interpretation: P, pathogenic; LP, likely pathogenic; VOUS, variation of uncertain significance; LB, likely benign. For 28 cases with more than one CNV, 10 cases with 2 CNVs including a pathogenic CNV and a VOUS CNV were listed in the P group, and the rest 16 cases including 13 cases with 2 P CNVs, 3 cases with 2 VOUS CNVs, 1 case with 2 LB CNVs and 1 case with 3 LB CNVs. d Other indications for prenatal testing include history of adverse pregnancy, parental genetic abnormalities, in vitro fertilization, medication use or toxic exposure during pregnancy, consanguineous marriages and parental anxiety.
reported mosaicism greater than 30% and loss of heterozygosity (LOH) of more than 10 Mb in size. If SNP-array detected a CNV in prenatal samples, the parents may choose to perform SNP-array to identify the origin of the CNV. And if a VOUS CNV unrelated to the clinical features was confirmed to be inherited from an unaffected parent, the classification of this CNV will be changed to be likely benign after precluding special considerations listed by the standards and guidelines released by the American College of Medical Genetics (Kearney et al., 2011).

Anomaly on Ultrasonography
A total of 1055 samples showed fetal anomalies on ultrasound scan, and SNP-array analysis detected 165 samples with chromosome abnormalities (165/1055, 15.6%), including 66 cases with numerical chromosome anomalies (66/1055, 6.3%), 2 cases with LOH (2/1055, 0.2%) and 97 cases with CNV (97/1055, 9.2%) ( Table 3 and Figure 1A). The 1055 samples with anomaly on ultrasonography were further divided into five subgroups based on ultrasound characteristics (Edwards and Hui, 2018): 38 cases (38/1055, 3.6%) with structural anomaly  (Table 3). Overall, the rate of chromosomal abnormalities detected by SNP-array in the group with structural anomaly in multiple systems (20/38, 52.6%) was the highest of the five subgroups, and the rate of chromosomal abnormalities in the group with abnormality of a single ultrasonic soft marker (55/458, 12.0%) was the lowest (Table 3 and Figure 1B). In the subgroup with structural anomaly in a single system, anomaly in cardiovascular system was the most common anomaly (206/411, 50.1%), of which SNP-array detected 14.6% (30/206) chromosomal abnormalities, followed by anomaly in genitourinary system (62/411, 15.1%) with 11.3% (7/62) chromosomal abnormalities (  4.1%), 10 cases with LOH (10/1784, 0.6%) and 109 cases with CNV (109/1784, 6.1%) ( Table 4 and Figure 1A). The 1784 samples with AMA were further divided into seven subgroups according to the maternal age (Table 4 and Figure 2A), and the relationship between maternal age and chromosome abnormality rate was examined. As shown in Figure 2B, the rate of aneuploidy significantly increased with age, while there was no such trend in the rate of CNVs.

Abnormal NIPT Results
Currently, whole-genome sequencing-based NIPT could be used for detecting aneuploidy as well as genome-wide CNVs (> 10 Mb) (Liu et al., 2016

Other Indications
The rest (709 samples) had other indications for prenatal testing, such as history of adverse pregnancy, parental genetic abnormalities, in vitro fertilization, medication use or toxic exposure during pregnancy, consanguineous marriages and parental anxiety (  by karyotyping, 6 cases were not detected by SNP-array including 3 cases of low-level mosaic aneuploidy and 3 cases of mosaic balanced structural rearrangement, while SNParray additionally detected 1 case of LOH and 1 case of microduplication (Table 7 and Supplementary Figure 1). Among 185 cases with structural rearrangement revealed by karyotyping, 147 cases of structural rearrangement including balanced structural rearrangement, chromosomal heteromorphisms and marker chromosomes were not detected by SNP-array as expected, while SNP-array additionally detected 1 case of mosaic trisomy 22 and 10 cases of microduplication/microdeletion ( Table 7).

DISCUSSION
In this study, we investigated the clinical value of SNP-array in prenatal diagnosis in a cohort of 5000 pregnancies. The detection rate of abnormalities by SNP-array was 12.3%, including 4.1% of cases with numerical chromosome anomalies, 0.4% with LOH and 7.8% with CNV (Table 2 and Figure 1A).And the detection rate of clinically significant CNVs (i.e., CNVs classified as P or LP) by SNP-array was 2.6% (Table 2), which is similar to a previous systematic review that overall 295/12362 (2.4%) have CNVs with associated clinical significance (Callaway et al., 2013) and a recent retrospective analysis of 10-year data reported a CNVs were classified into four groups according to interpretation: P, pathogenic; LP, likely pathogenic; VOUS, variation of uncertain significance; LB, likely benign. For cases with 2 CNVs including a pathogenic CNV and a VOUS CNV were listed in the P group. b AFP and freeβ-HCG levels in maternal serum were converted to MoM (multiples of median). a CNVs were classified into four groups according to interpretation: P, pathogenic; LP, likely pathogenic; VOUS, variation of uncertain significance; LB, likely benign. For cases with 2 CNVs including a pathogenic CNV and a VOUS CNV were listed in the P group. b Other autosomes: autosomes except chr21, chr13, and chr18.
that the abnormality detection rate of pathogenic CNVs by aCGH is 2.59% for prenatal cases (Chai et al., 2019). Moreover, the 5000 samples were categorized into five groups according to the indications for invasive prenatal testing: anomaly on ultrasonography, advanced maternal age, abnormal result on maternal serum screening, abnormal NIPT results and other indications. The rate of chromosomal abnormalities detected by SNP-array in the group with abnormal NIPT results was the highest (190/515, 36.9%), followed by the group with anomaly on ultrasonography (165/1055, 15.6%), and the rate of chromosomal abnormalities in the group with other indications was the lowest (63/709, 8.9%) ( Table 2 and Figure 1A).
In the group of anomaly on ultrasonography, SNP-array identified 4.5% (47/1055) clinically significant submicroscopic CNVs in fetuses with abnormal prenatal ultrasound findings ( Table 2), which is lower than previous meta-analyses that CMA identified 5.2%-10% additional chromosome aberrations over conventional karyotyping in fetuses with a structural malformation on ultrasound (Hillman et al., 2011(Hillman et al., , 2013Callaway et al., 2013). This might result from the addition of samples with ultrasonic soft marker abnormalities into this group. A systematic review revealed that a causative submicroscopic CNV could be identified in 3.1-7.9% of fetuses with ultrasound anomaly restricted to one system and in 9.1% of fetuses with multiple ultrasound anomalies (de Wit et al., 2014). In our study, the fetuses with structural anomaly in a single system had a 4.4% (18/411) chance of carrying clinically significant CNV, and the chance increased to 15.8% (6/38) for fetuses with structural anomaly in multiple systems and to 13.2% (10/76) for fetuses with structural anomaly combined with soft marker abnormalities (Table 3). In the subgroup of structural anomaly in a single system, clinically significant CNVs most commonly associated with cardiovascular (11/206, 5.3%) and genito-urinary systems (3/62, 4.8%) ( Table 3), which is in accordance with a previous report that the renal and cardiac systems were significantly associated with other-than-common benign CNVs (Donnelly et al., 2014). The detection rates of clinically significant submicroscopic CNVs in fetuses with abnormality of a single ultrasonic soft marker and multiple ultrasonic soft markers were 2.2% (10/458) and 4.2% (3/72) respectively (Table 3), which is consistent with a previous report that clinically significant genomic alterations were identified in 2.6% of cases with anomalies of ultrasonic soft markers (Shaffer et al., 2012). Recently, a research revealed a small but statistically insignificant increase in odds of clinically relevant CNV in fetuses/children with one or more ultrasound soft markers, which is in consistency with our result that no statistically significant difference was noted between rates were not detected by SNP-array, while SNP-array additionally detected 1 case of LOH and 1 case of microduplication (Supplementary Figure 1). c 147 cases of structural rearrangement including balanced structural rearrangement, chromosomal heteromorphisms and marker chromosomes were not detected by SNP-array, while SNP-array additionally detected 1 case of mosaic trisomy 22 and 10 cases of microduplication/microdeletion. of clinically significant submicroscopic CNVs in fetuses with abnormality of a single ultrasonic soft marker and multiple ultrasonic soft markers (2.2% vs. 4.2%, chi-squared test, P = 0.55) (Angras et al., 2020).
In the group of AMA, clinically significant submicroscopic CNVs were detected by SNP-array in 1.6% (29/1784) of fetuses with AMA (Table 2), which is similar to the result of a previous multicenter study by the National Institute of Child Health and Human Development (NICHD) that a submicroscopic pathogenic aberration was detected prenatally in 1.7% (95% CI 1.2-2.4) of fetuses tested due to AMA (Wapner et al., 2012). And our results demonstrated that the rate of numerical chromosome anomalies significantly increased with age, while there was no such trend in the rate of CNVs for fetuses with AMA, which is consistent with previous studies (Grati et al., 2015;Shi et al., 2019).
In the group of abnormal result on maternal serum screening, SNP-array identified clinically significant submicroscopic CNVs in 2.5% (30/1199) of fetuses with abnormal result on maternal serum screening ( Table 2), which is consistent with the result of a previous NICHD multicenter study that 1.6% (95% CI: 0.9-2.9) of fetuses tested abnormal on Down's syndrome screening had clinically relevant findings on microarray that were not detected on karyotyping (Wapner et al., 2012).
In the group of abnormal NIPT results, SNP-array analysis detected clinically significant submicroscopic CNVs in 2.9% (15/515) of fetuses with abnormal NIPT results, while in the group of other indications, clinically significant submicroscopic CNVs were detected by SNP-array in 3.0% (21/709) of fetuses with other indications ( Table 2). In this study, abnormal NIPT results included chromosome aneuploidies and CNVs of all 24 chromosomes detected by NIPT. NIPT has been widely applied for detecting fetal chromosome trisomy 13, 18, and 21 (T13, T18, and T21) and sex chromosome aneuploidies. However, the performance of NIPT for screening other fetal chromosome aneuploidies and CNVs is still limited. In 2019, we have evaluated the clinical practical performance of NIPT to analyze all 24 chromosome aneuploidies among 57,204 pregnancies in our center, and concluded that NIPT showed good performance for detecting T13, T18, and T21, but the accuracy rate of NIPT for detecting rare fetal chromosome aneuploidies was insufficient . In this study, only cases with abnormal NIPT results were analyzed by SNP-array, while the SNP-array or karyotyping results of cases with normal NIPT results were unknown, and the abnormal NIPT results of 516 cases were given by different NIPT platforms in our center or even other centers. Therefore it is difficult to calculate the false positive rate of NIPT in this study.
We also compare the results of karyotype and SNP-array analysis on 4022 samples (4022/5000, 80.4%) of our cohort. As SNP-array can identify submicroscopic CNVs and LOH not detected on karyotyping, 286 abnormal results (286/4022, 7.1%) were additionally revealed by SNP-array analysis over G-banded karyotyping. Apart from 2 cases of mosaic 45, X, SNP-array analysis detected clinically significant submicroscopic CNVs in 2.9% (64/3665, 1.7%) of fetuses with a normal karyotype. And the discrepant results between karyotyping and SNP-array analysis for samples with mosaicism or structural rearrangement detected by karyotyping could be due to direct (uncultured) analysis by SNP-array, and structural rearrangement (including balanced structural rearrangement, chromosomal heteromorphisms and marker chromosomes) and low-level mosaicism (<10-15%) not detectable by SNP-array. As conventional karyotyping is still valuable in the identification of balanced structural chromosomal abnormalities and low-level mosaicism, we recommend the combined application of karyotyping and SNP-array analysis in prenatal diagnosis.
In this cohort, the detection rate of VOUS CNVs by SNParray was 2.9% ( Table 2), which is higher than that of a previous NICHD multicenter study (1.6%) (Wapner et al., 2012) and lower than that of another cohort of 5026 pregnancies (4.6%) . These differences may result from different reporting criteria and interpretation deviations. Moreover, reporting VOUS and LB CNVs may increase patient anxiety and the genetic counseling workload in clinical practice.
In summary, a retrospective analysis was performed on a cohort of 5000 pregnancies, and the detection rate of chromosome abnormalities by SNP-array was 12.3%, including 4.1% of cases with numerical chromosome anomalies, 0.4% with LOH and 7.8% with CNVs. The overall detection rate of clinically significant CNVs by SNP-array was 2.6%, and we recommend the combined application of karyotyping and SNParray analysis in prenatal diagnosis. According to the indications for invasive prenatal testing, SNP-array identified clinically significant submicroscopic CNVs in 4.5% fetuses with anomaly on ultrasonography, in 1.6% of fetuses with AMA, in 2.5% of fetuses with abnormal result on maternal serum screening, in 2.9% of fetuses with abnormal NIPT results and in 3.0% of fetuses with other indications.

DATA AVAILABILITY STATEMENT
The genotyping data for this article are not publicly available to assure patient confidentiality and participant privacy.
Requests to access the datasets should be directed to TW, biowt@njmu.edu.cn.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the institutional Ethics Committee of The Affiliated Suzhou Hospital of Nanjing Medical University. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.