Chromosome Analysis and Postnatal Follow-Up of Fetuses With Abnormal Ultrasound Findings

Objective—To analyze the correlation between copy number variations (CNVs) and prenatal ultrasound abnormalities, characterize the CNVs in diverse prenatal phenotypes, and provide detailed follow-up results for prenatal diagnosis. Method—The results obtained in 1152 fetuses referred for chromosome testing due to different ultrasound anomalies were analyzed. CNV detection was performed by chromosomal microarray analysis (CMA) and copy number variation sequencing (CNV-seq). The postnatal outcomes of the fetuses were followed up. Result—Chromosomal abnormalities were detected in 149 of 1152 pregnancies (12.9%), including 84 (7.3%) macroscopic and 65 (5.6%) submicroscopic anomalies. The highest proportion of chromosomal defects was found in fetuses with hydrops (38.5%), followed by fetuses with multiple abnormalities (32.1%). The most common CNV number variation was 22q11 deletion (0.7%), followed by 17q12 deletion (0.4%) and 15q11.2 deletion (0.4%). Bilateral hyperechogenic kidneys were identied in all fetuses with the 17q12 deletion, and all cases of 15q11.2 deletion were positive only for soft markers. On postnatal follow-up of 362 live births, 2 cases of esophageal atresia and 2 cases of craniosynostosis were found that were not accurately diagnosed by prenatal ultrasound. Conclusion—we identied chromosomal abnormalities in approximately one eight of the pregnancies with abnormal ultrasound ndings. More emphasis should be placed on the improvement of prenatal detection of esophageal atresia and craniosynostosis. 4, 5 have been increasingly adopted in prenatal diagnosis to detect submicroscopic pathogenic copy number variations (PCNVs). The addition of CMA testing to karyotyping increases the frequency of detection of chromosomal abnormalities by 3–5% 2, 3 . Certain CNVs correlate with specic prenatal phenotypes 6–9 . The frequency of chromosomal anomalies varies with the specic structural anomaly and the number of ultrasound-detected defects. However, more data are needed to clarify the correlation between CNVs and prenatal abnormalities. Moreover, the follow-up studies of the cases with a congenital abnormality are frequently restricted to a single disease 10–12 ; therefore, these results may not be generalizable. In our study, samples of 1152 fetuses with various ultrasound ndings were analyzed by CMA and CNV-Seq to determine the correlation between CNVs and prenatal ultrasound abnormalities, including soft makers and structural anomalies. Additionally, we conducted a detailed follow-up of fetuses with abnormal ultrasound ndings to gain information for prenatal diagnosis. age at birth, age, height, weight, physical exam results, large motor skills development, and other positive results. The correspondence between the postnatal exam and prenatal ultrasound was dened as an agreement (same diagnosis), additional (additional ndings undetected by ultrasound), or unconrmed (false positive and false negative ultrasound results). Categorical data are presented as numbers and proportions. Comparison between groups was performed using the Chi-square test. A P-value of <0.05 was considered statistically signicant. All statistical analyses were done using the SPSS 16.0 software (SSPS, Chicago, IL, USA).


Introduction
Fetal structural abnormalities are found in approximately 3% of pregnancies 1 . Ultrasonography can identify fetal abnormalities ranging from a single minor defect to severe multisystem anomalies. They may be associated with chromosomal aneuploidy or copy number variants (CNVs). Chromosomal microarray (CMA )2, 3 , and CNV sequencing (CNV-Seq) 4,5 have been increasingly adopted in prenatal diagnosis to detect submicroscopic pathogenic copy number variations (PCNVs). The addition of CMA testing to karyotyping increases the frequency of detection of chromosomal abnormalities by 3-5% 2, 3 .
Certain CNVs correlate with speci c prenatal phenotypes [6][7][8][9] . The frequency of chromosomal anomalies varies with the speci c structural anomaly and the number of ultrasound-detected defects. However, more data are needed to clarify the correlation between CNVs and prenatal abnormalities. Moreover, the follow-up studies of the cases with a congenital abnormality are frequently restricted to a single disease [10][11][12] ; therefore, these results may not be generalizable. In our study, samples of 1152 fetuses with various ultrasound ndings were analyzed by CMA and CNV-Seq to determine the correlation between CNVs and prenatal ultrasound abnormalities, including soft makers and structural anomalies. Additionally, we conducted a detailed follow-up of fetuses with abnormal ultrasound ndings to gain information for prenatal diagnosis.

Patients and Ultrasonography
Our study included 1152 pregnant women who received fetal chromosome examination by CMA or CNV-seq tests due to abnormal ultrasonographic ndings, including positive soft markers and structural anomalies. The women were admitted to the Xiangya Hospital of Central South University and Hunan Jiahui Genetic Hospital between January 2013 and June 2019. All pregnant couples provided written informed consent. The mean gestational age of mothers was 28.2 years (range, 22-45 years), and the mean age of fetuses at the invasive procedure was 20.3 weeks (range, 11-37 weeks). The study was approved by the Medical Ethics Committee of Xiangya Hospital of Central South University and Hunan Jiahui Genetic Hospital.
All methods were performed in accordance with Medical Ethics Committee of Xiangya Hospital of Central South University and Hunan Jiahui Genetic Hospital.
Ultrasound scanning of fetuses was performed between 11 and 35 weeks of gestation. The positive soft markers included screening for thickened nuchal translucency or nuchal fold (greater than 95th percentile), absence or dysplasia of nasal bone, short long bones (more than 2 SDs below the mean), hyperechogenic bowel, tricuspid regurgitation, mild ventricular expansion (10-15mm), intracardiac echogenic focus, choroid plexus cyst, single umbilical artery, enlarged cisterna magna (10-15mm), hydronephrosis (>4mm before 28 weeks, >7mm after 28 weeks), persistent left superior vena cava, and aberrant right subclavian artery. Fetal structural anomalies were categorized into 10 phenotypic classes based on their location, including the cardiovascular system, central nervous system, genitourinary system, musculoskeletal, gastrointestinal, respiratory, facial, abdominal wall system, hydrops fetalis, and multiple malformations involving two or more systems [Supplement 1].

Sample Collection and Preparation
Fetal DNA was obtained from amniotic uid, cord blood, or fetal skin. The genomic DNA was prepared using the DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany). The quality and concentration of genomic DNA in the samples were assessed by agarose gel electrophoresis and a NanoDrop spectrophotometer (Thermo Fisher Scienti c, Wilmington, DE, USA).

SNP Array Analysis and CNV-seq
Chromosome CNV analysis was used before 2017 and was performed using the HumanCytoSNP-12 BeadChip array (Illumina, San Diego, CA, USA), which comprises about 300,000 SNP probes with an average marker spacing of approximately one probe every 10 kb. CNV-seq was used from 2017 and utilized the HiSeq 2000 platform (Illumina). All sequencing procedures were performed as previously described 4,13,14 .
Copy number variations (CNVs) were called at an effective minimal resolution of 100 kb involving at least 10 contiguous probes.
The CNVs were classi ed based on their size. CNVs ≥10 Mb were de ned as chromosomal macroscopic anomalies, which could be identi ed by traditional karyotyping. By contrast, CNVs <10 Mb were de ned as chromosomal microscopic anomalies (microdeletions/microduplications).
Detected CNVs were evaluated based on a review of scienti c literature and the following public databases: Database of Genomic Variants (http://projects.tcag.ca/variation/), DECIPHER (http://decipher.sanger.ac.uk/), UCSC (http://genome.ucsc.edu/), and OMIM (http://www.ncbi.nlm.nih.gov/omim). Variants of uncertain signi cance (VOUS), and pathogenic CNVs. CNVs coinciding with known polymorphic CNVs reported in multiple peer-reviewed publications or databases as benign variants were considered benign. Conversely, CNVs were de ned as pathogenic if 1) the CNV was documented as clinically signi cant in multiple peer-reviewed publications, regardless of its penetrance and expressivity, or 2) the CNV overlapped a smaller interval with clearly established clinical signi cance, or 3) the CNV was >5 Mb long and no well-established cytogenetic heteromorphisms were present in this region. CNVs that did not t any of the above criteria were considered as VOUS.

Parental study
If CNVs were detected in the fetus, parental samples were subsequently analyzed by karyotyping or FISH, according to the CNV result. Routine G-banding karyotyping was performed using standard methods. The FISH analysis was performed using commercially available subtelomeric speci c probes (VYSIS, Inc., Downers Grove, IL, USA) according to the manufacturer's protocols.

Postnatal Follow-up and Statistical Analysis
Follow-up was conducted by phone calls. The median time of follow-up was 10 months (range 3-24). The follow-up content based on medical reports included gestational age at birth, age, height, weight, physical exam results, large motor skills development, and other positive results. The correspondence between the postnatal exam and prenatal ultrasound was de ned as an agreement (same diagnosis), additional (additional ndings undetected by ultrasound), or uncon rmed (false positive and false negative ultrasound results). Categorical data are presented as numbers and proportions. Comparison between groups was performed using the Chi-square test. A P-value of <0.05 was considered statistically signi cant. All statistical analyses were done using the SPSS 16.0 software (SSPS, Chicago, IL, USA).

Chromosome Analysis
Fetal chromosome examination using CMA or CNV-seq tests was performed in 1152 fetuses with abnormal ultrasonographic ndings, including 405 positive soft markers and 747 structural anomalies. The analysis employed CMA in 367 cases and CNV-seq in 785 cases. Fetal DNA was obtained from amniotic uid in 880 (76.4%) fetuses, skin samples after pregnancy termination in 212 (18.4%) fetuses, and cord blood in 60 (5.2%) fetuses.
Chromosomal abnormalities were detected in 149 (12.9%) samples, involving 84 (7.3%) macroscopic anomalies and 65 (5.6%) submicroscopic anomalies ( Table 1). The frequency of chromosome macroscopic anomalies was signi cantly higher in hydrops fetuses (35.7%) than in the remaining cohort. While multiple abnormalities (13.9%) are most common in copy number variations, followed by the central nervous system (7.9%) and musculoskeletal system (7.3%). Among the 65 pathogenic and likely pathogenic CNVs, 22q11 deletion (9 cases, 0.7%) was the most common, followed by 17q12 deletion (5 cases, 0.4%) and 15q11.2 deletion (5 cases, 0.4%). Of note, bilateral hyperechogenic kidneys were found in all 17q12 deletion fetuses, and in all 15q11.2 deletion cases only soft markers were positive (Table 2). However, both ultrasound manifestations (hyperechogenic kidneys and soft markers) corresponded to benign phenotypes. The ve cases of 15q11.2 deletion syndrome included 4 cases of the termination of pregnancy for 15q11.2 deletion and 1 case that was found to have craniosynostosis after birth with signi cant global developmental delay. The rate of chromosomal abnormalities in the soft markers group was 8.1% (4.4% corresponded to macroscopic anomalies and 3.7% to submicroscopic anomalies), which was signi cantly lower than in the structural abnormalities group (15.5%, P < 0.001). Fetal structural anomalies were categorized into 10 phenotypic classes based on their location. The number of cases in these classes ranged from 16 (respiratory) to 192 (cardiovascular). The rate of chromosomal abnormality varied by the anatomical systems involved: the greatest fraction of chromosomal abnormalities were found in fetuses with hydrops (38.5%), multiple malformations (32.1%), cardiovascular (11.4%), and central nervous system (10.1%) anomalies, followed by the genitourinary (8.5%), skeletal (7.3%), and facial (5.7%) anomalies; the other systems were negative.

Follow-up Results
Pregnancy outcomes were available for 533 (46.3%) of the 1152 fetuses. Of these, 165 (30.9%) were pregnancies that parents opted for termination and 6 (1.1%) died in the uterus. Additionally, there were 12 (2.3%) postnatal deaths, and 350 (65.7%) live births. Postnatal ndings con rmed prenatal ultrasound in 96.9% fetuses (351/362), provided additional information in 4 (1.1%) fetuses, and did not con rm prenatal ultrasound in 7 (1.9%) fetuses. The additional ndings undetected by ultrasound include 2 cases of esophageal atresia (1 with cleft palate) and 2 cases of prenatal craniosynostosis. The missed ndings were 4 ventricular septal defects, and only soft markers were detected by prenatal ultrasound. The instances of misdiagnosis included 1 case of clubfoot (found to be normal after birth), 1 abdominal lymph cyst (diagnosed postnatally as intestinal atresia), and 1 duplex kidney (found to be normal after birth) ( Table 3).

Discussion
In our study, chromosomal abnormalities were detected in 12.9% of pregnancies, 7.3% were macroscopic anomalies and 5.6% were submicroscopic anomalies. The rate of chromosomal abnormalities associated with structural defects (15.5%) is consistent with previous studies, which identi ed signi cant chromosomal abnormalities in 16% of fetuses in which ultrasonography indicated at least one anomaly 15 . The frequency of chromosomal abnormalities in fetuses with soft markers (8.1%) was higher than reported in other studies (approximately 4%). This difference may be due to the following reasons: 1) at our referral center, many pregnant women with positive soft markers are of advanced age and have a complicated pregnancy, and 2) our study included fetuses with persistent left superior vena cava and aberrant right subclavian artery. Although the rate of chromosomal abnormalities in the soft markers group was lower than in the structural abnormalities group, identifying an additional 4.4% pathogenic or likely pathogenic CNV can avert the possibility of birth of children with a chromosome microdeletion or microduplication syndrome. Certainly, the risk of pregnancy loss or other post-amniocentesis complications should be considered, even if currently the risk related to amniocentesis is approximately 1 in 800 16 .
The rate of chromosomal abnormalities varied among the anatomical systems involved. We detected chromosomal abnormalities more frequently in association with fetal hydrops, multiple malformations, cardiovascular and central nervous anomalies, and, to a lesser extent, abnormalities of the genitourinary and musculoskeletal systems. In a recent study using WES to identify genetic variants in fetuses with structural anomalies, the greatest proportion of variants was found in cases of hydrops 17 . Their relative frequency can give insight into the proportion of anomalies, such as hydrops fetalis, that potentially have a detectable underlying genetic cause.
Certain CNVs are associated with speci c phenotypes. For example, cardiovascular abnormalities represent one of the pillars of clinical diagnosis of 22q11.2 deletion (DiGeorge syndrome). In our study, 50% of cases with 22q11 deletion syndrome exhibited cardiovascular defects. Bilateral hyperechogenic kidneys were found in all fetuses with 17q12 deletion, which is in line with previous studie s18 . The 15q11.2 BP1-BP2 deletion has been associated with variable phenotypes and low penetrance. Our results tend to suggest that the deletion confers milder phenotypes, as only soft markers were positive prenatally. A previous investigation documented that the 15q11.2 deletion is associated with congenital cardiovascular malformations 19 , but no such case was present in our study. There are some controversies regarding the classi cation of 15q11.2 deletion in prenatal management. Marshall et al. 20 recommended classifying the deletion as "pathogenic of mild effect size, while Maya et al 21 . recommended that 15q11.2 deletion should not be considered pathogenic and should not be reported to the parents in the prenatal setting. In our laboratory, this deletion is classi ed as pathogenic variation with incomplete penetrance. Unfortunately, among the 5 cases of 15q11.2 deletion syndrome, in 4 instances the termination of pregnancy was chosen based on the report of pathogenic copy number variant, and 1 case was found after birth to have craniosynostosis with a signi cant global developmental delay. 15q11.2 BP1-BP2 structural variation is associated with changes in in brain morphology and cognition 22 , possibly indicating that this CNV is more commonly associated with neurocognitive impairment than with structural anomalies. Larger studies of prenatal and postnatal phenotypic groups will be required to provide useful information for interpreting and consulting cases with 15q11.2 deletions.
Follow-up results demonstrated a high agreement between prenatal ultrasound and living birth. Not surprisingly, the most frequently missed prenatal diagnosis was the ventricular septal defect. Also, esophageal atresia and craniosynostosis are di cult to diagnose by ultrasound prenatally. Analysis of the prenatal ultrasound images of these patients revealed that 2 cases of craniosynostosis had abnormal head morphology (1 occipital attening, 1 navicular head). However, since both fetuses had multiple malformations, the doctors focused on the obvious structural abnormalities and ignored the changes in cranial morphology. Therefore, alterations in cranial morphology provide an important clue that requires consideration to exclude the possibility of craniosynostosis.
The antenatal detection rate of esophageal atresia has been extremely poor even in expert hands, with a sensitivity ranging from 40 to 60% 23,24 . Remarkably, in our study, 2 cases in which esophageal atresia was missed in prenatal ultrasound had co-existent esophagotracheal stula, making a correct prenatal diagnosis more challenging. More recently, Dall'Asta et al. 25 demonstrated that the detection of esophageal atresia is possible when three-dimensional ultrasound and Crystal Vue rendering techniques are used. Thus, employing the most current ultrasound methods may increase the accuracy of prenatal detection of esophageal atresia.
Limitations of our study include a high rate of lost follow-up (telephone number errors or refusal to answer). Additionally, many pregnant women do not consent to the autopsy of the fetus due to cultural norms, and the prenatal structural abnormalities could not be con rmed. The absence of data regarding the inheritance (i.e., parental vs. de novo) of each alteration and the presence of many variants of unknown signi cance increased the di culty of genetic counseling.
In summary, we identi ed chromosomal abnormalities in approximately one eight of the pregnancies with abnormal ultrasound ndings.. Follow-up is recommended in the management of fetal structural anomalies since it con rms, modi es, or adds information to prenatal diagnosis.