Investigation on combined copy number variation sequencing and cytogenetic karyotyping for prenatal diagnosis

Background We aimed to evaluate the clinical value of copy number variation-sequencing (CNV-Seq) in combination with cytogenetic karyotyping in prenatal diagnosis. Methods CNV-Seq and cytogenetic karyotyping were performed in parallel for 9452 prenatal samples for comparison of the diagnostic performance of the two methods, and to evaluate the screening performance of maternal age, maternal serum screening, fetal ultrasound scanning and noninvasive prenatal testing (NIPT) for fetal pathogenic copy number variation (CNV). Results Among the 9452 prenatal samples, traditional karyotyping detected 704 cases (7.5%) of abnormal cytogenetic karyotypes, 171 (1.8%) chromosome polymorphism, 20 (0.2%) subtle structural variations, 74 (0.7%) mutual translocation (possibly balanced), 52 (0.6%) without karyotyping results, and 8431 (89.2%) normal cytogenetic karyotypes. Among the 8705 cases with normal karyotype, polymorphism, mutual translocation, or marker chromosome, CNV-Seq detected 63 cases (0.7%) of pathogenic chromosome microdeletion/duplication. Retrospectively, noninvasive prenatal testing (NIPT) had high sensitivity and specificity for the screening of fetal pathogenic CNV, and NIPT combining with maternal age, maternal serum screening or fetal ultrasound scanning, which improved the screening performance. Conclusion The combined application of cytogenetic karyotyping and CNV-Seq significantly improved the detection rate of fetal pathogenic chromosome microdeletion/duplication. NIPT was recommended for the screening of pathogenic chromosome microdeletion/duplication, and NIPT combining with other screening methods further improved the screening performance for pathogenic fetal CNV.


Background
Different levels and types of genetic variation exist in the human genome, ranging from single nucleotide mutations to structural or numerical chromosome abnormalities. One or more genetic variations may exist in an individual, and some genetic variations cause severe congenital malformations or death. In addition to triploid and numerical chromosome abnormalities, pathogenic chromosome microdeletion/duplication also leads to poor fetal prognosis. For example, Wolf-Hirschhorn syndrome mostly results in developmental retardation, unusual faces and structural abnormalities, and Miller-Dieker syndrome can be complicated by pachygyria; 22q11.21 microdeletion syndrome often has various degrees of cardiac malformations [1][2][3][4][5][6]. If pathogenic chromosome microdeletion/duplication can be diagnosed prenatally, the births of children with such severe congenital defects can be avoided.
However, the target diseases of traditional maternal serum screening are limited to common aneuploidies. Fetal ultrasound scanning is mainly used to monitor fetal growth and development, and to find structural fetal abnormalities and soft markers. Noninvasive prenatal testing (NIPT) using maternal plasma cell-free fetal DNA has made prenatal screening for pathogenic chromosome microdeletion/ duplication possible.
Traditional cytogenetic karyotyping has been used as the gold standard diagnosis of chromosome abnormalities for decades. However, it is time-consuming and labor-intensive, largely dependent on cell culture, and has a low chromosome resolution of 5~10 Mb. In recent years, the application of high-resolution chromosome micro-array analysis (CMA), which can detect abnormal chromosome number, micro-deletions/duplication, uniparental disomy, has revolutionized the testing methodology of prenatal diagnosis. It has been suggested in some studies that CMA can be solely used instead of cytogenetic karyotyping in prenatal diagnosis laboratories with limited human resources [7]. Nextgeneration sequencing (NGS) now offers an alternative methodology to CMA, named copy number variation sequencing (CNV-Seq) with a resolution of 0.2 Mb for the detection of clinically significant chromosomal abnormalities. CNV-Seq has uniform sequencing coverage and relatively low price and has been gradually used in prenatal diagnosis [8]. However, more studies are required to further verify the efficiency of CNV-Seq in prenatal diagnosis.
Traditional karyotyping has characteristics of low-cost and covering the whole genome, including abnormal chromosome number and structural variation of specific regions, such as euchromatic and heterochromosomal regions. It highly depends on the experience of technicians to recognize these regions using different banding techniques under the microscope, which provide information about the frequency and location of these variations. Due to the morphologic similarity between chromosomes, karyotyping is difficult to accurately distinguish subtle structural variations. Genome copy number variation (CNVs) refers to structural variations of DNA sequence of more than 0.2 Mb. Karyotyping by conventional chromosome banding technology cannot distinguish these subtle variations. Compared with karyotyping, CNV-Seq which based on next-generation sequencing technology and comparative genomics has high resolution, high throughput, and simple laboratory operations. However, CNV-Seq also has limitations such as short read lengths and not covering the whole genome, and it cannot detect balanced translocations, polymorphism, marker chromosomes, and other genetic variations out of the detection range, and it cannot accurately detect polyploidy and low-proportion chromosome mosaic. Karyotyping and CNV-Seq are two different technologies, and the combination of them in prenatal diagnosis may make up for each other's shortcomings and verifie each other's results to improve the accuracy of prenatal diagnosis. Therefore, in this study, we comparatively analyzed the difference between cytogenetic karyotyping and CNV-Seq for the same fetal samples, evaluated the value of adding CNV-Seq in traditional prenatal diagnosis, analyzed the performances of maternal age, maternal serum screening, NIPT and fetal ultrasound scanning for the screening of pathogenic fetal CNV and investigated whether combined application of these prenatal screening methods could improve the sensitivity and specificity for fetal pathogenic CNV.

Study patients
All methods were carried out in accordance with relevant guidelines and regulations.

Prenatal screening
Four types of prenatal screenings were involved in this study: (1) combined screening in 11~13 + 6 weeks that comprised of fetal NT measurement + maternal serum screening using pregnancy associated plasma protein-A (PAPP-A), placental growth factor (PLGF) and free human chorionic gonadotropin beta unit (fβ-HCG), with or without NIPT; (2) maternal serum screening in 162 0 + 6 weeks using alpha fetoprotein (AFP), fβ-HCG and unconjugated estriol (uE 3 ), with or without NIPT; (3) NIPT only if gestational weeks at screening ≥21 weeks, and (4) all study patients had fetal ultrasound scanning in our department. The cases with high risk of Down Syndrome (DS), high risk of Edwards Syndrome (ES), or high risk of both by maternal serum screening were all classified as high-risk cases. Fetal ultrasound scanning was classified into five grades based on the severity of abnormalities: grade 0: without abnormal findings; grade 1: fetuses with subtly ultrasound abnormalities other than grade 2, e.g. gallbladder was not detected; grade 2: soft markers that were closely associated with chromosome aneuploidy, such as thickened nuchal fold (NF), nuchal translucency (NT) ≥ 3.0 mm, absence and/or dysplasia of nasal bone, mild to moderate ventriculomegaly, aberrant subclavian arteries, fetal growth restriction (FGR), short limb bones length, micrognathia, and acromphalus; grade 3: mild to moderate structural fetal malformations; grade 4: severe structural fetal malformations or lethal abnormalities. Maternal age ≥ 35 years at the expected date of confinement was defined as advanced maternal age.

Invasive prenatal diagnosis
Three types of surgeries for invasive prenatal diagnosis were used in this study. Amniocentesis: Twenty mL of amniotic fluid was collected by aspiration for cell culture and cytogenetic karyotyping, and 5 mL for CNV-Seq. If amniotic fluid was contaminated by maternal blood, adherent amniocytes after cell culture were used for CNV-Seq. Cordocentesis: Five mL of amniotic fluid was collected at first for CNV-Seq, and then 1.5 mL of cord blood for cell culture and cytogenetic karyotyping, and 0.5 mL of cord blood for hemoglobin electrophoresis to exclude maternal blood contamination. In prenatal diagnosis, the standard method to exclude maternal blood contamination should be linkage analysis of DNA polymorphism. We had used the method of STR polymorphism linkage analysis. In recent years, we used hemoglobin electrophoresis instead because STR polymorphism linkage analysis was much more timeconsuming and labor-intensive. Chorion villus sampling: a small amount of villous tissue was sampled for CNV-Seq directly. Before 2019, we were inexperienced in villus cell culture techniques. Therefore, the cases received chorion villus sampling were those with severe fetal structural abnormalities. No villus cell culture was applied for those cases, and only CNVs was provided.

Laboratory testing
Cell culture and cytogenetic karyotyping Amniotic fluid and umbilical blood samples were set up for cell culture following the standard protocols. Chromosome preparations were G-banded using trypsin-Giemsa staining for cytogenetic karyotyping after a series of standard protocols including colchicine treatment, hypotonic treatment, fixation and centrifugation. Karyotypes were diagnosed according to the international system for human cytogenetic nomenclature  (16). A total of 52 cases had only CNV-Seq results, but no karyotyping results. Among them, 44 cases who received chorion villus sampling, since our center cannot provide villus cell culture during that time; 8 cases encountered amniotic fluid cell culture failure. The maximum and minimum gestational weeks of amniotic fluid cell culture failure were 31 and 20 weeks, respectively. It should be noted that none of the 52 cases with missing results were used in the data analysis of this study.

CNV-Seq and result interpretation
Nextseq 550AR platform (Illumina, San Diego, CA) was used for DNA sequencing, with an average sequencing depth of 0.08×, following the Q30 sequencing quality standard. The amount of fetal DNA used for CNV-Seq was 10~50 ng for each prenatal sample. AnnoroadPD software (Annoroad Gene Technology Co., Ltd., Beijing, China) was applied to analyze the sequencing data referring to the human reference genome GRCh37/hg19. The identified fetal CNV were interpreted [11] and classified into five categories: pathogenic (P-), likely pathogenic (LP-), uncertain significance (VUS-), likely benign (LB-) and benign (B-), according to the standards and guidelines that were jointly developed by the American College of Medical Genetics and Genomics (ACMG), the Association for Molecular Pathology (AMP) and the College of American Pathologists (CAP) in 2015. To conveniently show the CNV-Seq results, we used "P-"as abbreviation for pathogenic chromosome microdeletion/duplication, "None" for no copy number variation found, "auto A" for autosomal aneuploidy, "sex A" for abnormal sex chromosomes number, "auto AM" for autosomal aneuploidy mosaic, and "sex AM" for abnormal sex chromosomes number mosaic. Cytogenetic karyotyping was the diagnostic method for numerical and structural chromosome abnormalities, and high-throughput sequencing for CNV. For LPand VUS-, family (parents and fetuses) CNV-Seq tests, fluorescence in-situ hybridization (FISH) or multiplex ligation probe amplification (MLPA) were used for further verification.

Statistical analysis
The data were analyzed statistically using IBM SPSS Statistics (version 22.0, IBM Corp., Armonk, NY, USA). Continuous variables (for example, age and gestational weeks at prenatal diagnosis) were expressed as "median [lower quartile, upper quartile]", and analyzed using Kruskal-Wallis one-way analysis of variance. Categorical variables are represented by "n (%)" and analyzed using Chi-square test for two-way disordered R × C table. Calculation for sensitivity and specificity: sensitivity = true positive / (true positive + false negative) *100%; specificity = true negative / (true negative + false positive) *100%. Paired chi-square test was used to test the difference between CNV-Seq and various prenatal screening methods and prenatal diagnosis results (P < 0.01 was considered statistically significant). Missing items were not applied in data analysis.

Basic characteristics of study patients
Basic information of 9452 cases of prenatal diagnosis was listed and statistically analyzed in Table 1. Among study patients, 9452 (100%) had received one to two times of fetal ultrasonography in our center, 5688 (60.2%) had maternal serum screening, 1409 (14.9%) had NIPT, 551 (0.58%) had both NIPT and maternal serum screening, and 3142 cases (33.2%) were in advanced maternal age. The results showed that only 1165 (12.3%) of patients received invasive prenatal diagnosis due to high risk of NIPT. Other indications for prenatal diagnosis included advanced maternal age, abnormal fetal ultrasound scanning, high risk of maternal serum screening, adverse reproductive history, family history of singlegene genetic diseases, or others.
The results of CNV-Seq in Tables 2, 8,354 fetuses with CNV-Seq findings were included as None, B-, and LB-, cytogenetic karyotyping showed that except for 2 cases of triploid, the rest 271 cases of abnormal karyotypes had good prognosis. A total of 530 cases of fetal aneuploidies (DS, ES, PS, XXY, XYY) were diagnosed, and the results of karyotyping and CNV-Seq were consistent. The details for 60 cases of pathogenic microdeletion/duplication detected by CNV-Seq were shown in Table 3. CNV-Seq detected 1 case of chromosomal aneuploidy and 2 cases of mosaic in fetuses with normal cytogenetic karyotypes. Furthermore, 2 cases (No.29 and 30) of pathogenic microdeletion/duplication were detected in 9 fetuses with marker chromosomes, and 1 case (No.38) of pathogenic microdeletion/duplication were detected in fetuses with mutual translocations (Seen in Table 4). Therefore, we may conclude that the combination of the two methodologies significantly improved the accuracy of prenatal diagnosis for fetal pathogenic CNV and was helpful to assess fetal prognosis. Due to its detection limitations, for example, two cases of triploid by karyotyping had normal CNV-Seq results, CNV-Seq could not replace karyotyping at present stage but might be an effective complement.
Performances of maternal age, maternal serum screening, NIPT and fetal ultrasound scanning for pathogenic CNV-Seq results and pathogenic karyotypes The target diseases of maternal serum screening are common chromosomal aneuploidies. For fetuses with pathogenic microdeletion/duplication, the prognosis is mostly poor. NIPT is the ideal prenatal screening method for pathogenic microdeletion/duplication, but the cost may limit its clinical use to a certain extent. If NIPT was unavailable, we wondered whether other screening methods could recognize pathogenic microdeletion/duplication. In this study, we retrospectively analyzed the results of maternal age, maternal serum screening, NIPT, and fetal ultrasound scanning for women with pathogenic fetal CNV-Seq results, including pathogenic microdeletion/duplication and Auto A, Sex A, Auto AM, Sex AM, as shown in Table 5. NIPT missed one case of Auto AM (CNV-Seq) and Triploid (karyotypes), shown in Table 6. In maternal serum screening, 55.2% of pathogenic microdeletion/ duplication (CNV-Seq) and 42.1% of unbalanced fragment deletion/duplication (karyotypes) showed high risks results. As to abnormal ultrasound findings (grades 2-4), there was significant difference between CNV-Seq of auto A and pathogenic microdeletion/ duplication (73.6% versus 43.2%, P < 0.001). Regarding to pathogenic karyotypes, NIPT missed one case of triploid, whose maternal serum screening and fetal ultrasound were abnormal. Only 42.2% of severe chromosomal abnormalities (aneuploidy, unbalanced fragment deletion/ As seen in Table 7, for pathogenic fetal CNV-Seq results, NIPT had the highest sensitivity of 1.    Table 8. Single indication had the highest sensitivity and the lowest specificity. In a similar trend, four indicators had the lowest sensitivity and the highest specificity. Accordingly, to achieve the optimal sensitivity and specificity, a prenatal screening program that combines two methods could be considered.

Discussion
Combination of cytogenetic karyotyping and CNV-Seq can prenatally diagnose more fetal pathogenic microdeletion/duplication and provide comprehensive prenatal information In addition to traditional cytogenetic karyotyping, CMA and CNV-Seq have gradually been used in prenatal diagnosis. CNV-Seq is worth applying in prenatal diagnosis due to its lower cost and uniform sequencing coverage. Fetuses with normal karyotype/chromosome polymorphism usually have good prognosis, but pathogenic microdeletion/duplication cannot be excluded. In this study, among fetuses with normal karyotypes, CNV-Seq diagnosed 60 (0.6%) cases of pathogenic CNV, and fortunately the births of 44 fetuses with poor prognosis were avoided. A complex case was also prenatal diagnosed with CNV-Seq result of 47,XXY and karyotyping result of 46,XX. SRY gene detection indicated a male gender, but all 6 loci on AZF gene were missing. Fetal ultrasound showed male external genitalia. Taking fetal ultrasound findings together, the fetus might have sexual reversal and risks of abnormal reproductive system development in puberty. The parents were fully informed the advantages, disadvantages and limitations of karyotyping and CNV-Seq, and to make clear that the two results cannot be denied by each other but be mutually complementary. The couple chose to continue the pregnancy. The boy is now 1.5 years old and is generally healthy. Follow-up and etiological examination were recommended. On the other hand, we should make clear that not all fetuses with structural chromosomal abnormalities by karyotyping have poor prognosis. For example, mutual translocation (paternal/maternal/de novo) with normal CNV-Seq and fetal ultrasound is very much likely that the fetuses have good prognosis. The combination of karyotyping and CNV-Seq enables mutual verification of the results in prenatal diagnosis and helps to avoid misdiagnosis and provide more information for comprehensive evaluation of fetal prognosis. However, the combined application of CNV-Seq and karyotyping may lead to increased economic burden. It needs further verification about whether the cost-effectiveness is worth promoting. We need to choose an appropriate prenatal diagnosis program based on our own characteristics.
Combined several prenatal screening significantly improves the specificity but reduces the sensitivity for fetal pathogenic CNV Maternal serum screening detects 70~80% of DS, at a false positive rate of 5% [12]. It was reported that a small  portion of sex chromosome abnormalities showed abnormal findings in maternal serum screening [13]. It is unknown whether maternal serum screening is abnormal for pathogenic CNV. The findings of this study indicated that maternal serum screening can detect 55.2% of fetal pathogenic chromosomal microdeletion/ duplication in fetuses with normal or abnormal karyotype. The intrauterine phenotype of fetuses with pathogenic microdeletion/ duplication lacks specificity, so prenatal ultrasound scanning is difficult to identify. In this study, fetuses with autosomal aneuploidy had the most severe ultrasound abnormalities, followed by pathogenic microdeletion/duplication with a large variability that some fetuses had completely normal ultrasonography. Fetuses with abnormal sex chromosome number and structure and the mosaic had mild ultrasound abnormalities. Fetal ultrasound scanning had higher sensitivity for autosomal aneuploidy. However, if other prenatal screening    limited. NIPT has been widely used as a first-line prenatal screening method [14]. It was reported that the sensitivities of NIPT for Trisomy 21, 18, 13 are 99.1, 98.2 and 100%, respectively [14]. The detection rates of NIPT for fetal aneuploidy and CNV > 20 Mb were  reported to be 100% [15]. NIPT was an effective method for prenatal screening of fetal CNV ranging from 1 to 129 Mb, with a sensitivity of 84.2% [16]. The findings of this study indicated that NIPT was a reliable method for prenatal screening of fetal pathogenic microdeletion/duplication. However, NIPT cannot detect fetal polymorphism, polyploids, balanced translocations and other fetal structural abnormalities. Therefore, the combination of maternal age, maternal serum screening, fetal ultrasound scanning and NIPT had been recommended for prenatal screening in some studies [17]. Our findings were partially in agreement with this perspective. For pathogenic CNV, the sensitivity of single prenatal diagnosis indication was 0.99 (0.98, 0.99), but the specificity was 0.10 (0.09, 0.11). When prenatal diagnosis indications increased from two to four, the sensitivity was decreased to 0.02 (0.01-0.03) and the specificity increased to 1.00 (1.00-1.00). According to our data, the combination of two screening method was possibly to achieve a maximal summation of sensitivity and specificity. Moreover, each combined screening program had its own advantages and limitations, which require comprehensive consideration by the couples and doctors. In short, we would recommend that combined at least two kinds of prenatal screening could be used as the efficient program if medical resources for prenatal care are sufficient and the couples are willing to receive the screening.

Conclusions
Combination of cytogenetic karyotyping and CNV-Seq significantly improves the detection rate of fetal pathogenic chromosome microdeletion/duplication. NIPT was recommended for the screening of pathogenic chromosome microdeletion/duplication, and NIPT combining with other screening methods further improved the screening performance for pathogenic fetal CNV.