1 INTRODUCTION

The fetal demise of a cotwin (i.e., spontaneous reduction to a singleton pregnancy) is common, with an estimated prevalence of 6.2% for all twin pregnancies.1,2 This prevalence might be underestimated, since fetal demise occurring during the first trimester, also known as the “vanishing twin syndrome,” may affect as many as 30% of pregnancies.3-6 Chromosomal anomaly appears to be an important reason for fetal demise, given the high rate of chromosome aneuploidy detected in early singleton miscarriages.7-9 The deceased fetus has been reported to release residual cell-free DNA (cfDNA) continually into the maternal circulation, which leads to the false-positive results of fetal aneuploidy and incorrect fetal sex on noninvasive prenatal testing (NIPT).10-12 These results lead to queries of testing accuracy and counseling difficulties concerning the use of NIPT for twin pregnancies or pregnancy with an undetected fetal death. The current clinical guidelines on how long NIPT should be performed after fetal demise are based on limited clinical evidence, since the longitudinal changes in the residual cfDNA of the deceased fetus in the maternal circulation have not been extensively studied. We aimed to investigate dichorionic diamniotic (DCDA) twin pregnancies undergoing selective fetal reduction for an aneuploid cotwin as a model of natural fetal demise to determine the sequential changes in the residual cfDNA of the deceased fetus, as detected by NIPT.

2 METHODS 2.1 Study population and study design

This was a prospective observational study of women with a DCDA twin pregnancy who opted for selective reduction of a cotwin in the Department of Fetal Medicine and Prenatal Diagnosis, the Third Affiliated Hospital of Guangzhou Medical University, China. The inclusion criteria were as follows: pregnant woman who had undergone in vitro fertilization (IVF) consisting of a double-embryo transfer, older than 18 years, with a gestational age older than 8 weeks. Each woman had been previously confirmed to carry a dichorionic twin pregnancy by ultrasound, and opted for fetal reduction because of a cotwin with aneuploidy diagnosed by chorionic villus sampling (CVS) or amniocentesis. Before reduction, each woman was given a detailed explanation of the study and signed informed consent. The exclusion criteria were as follows: either parent with a chromosomal abnormality; or the mother had undergone blood transfusion, organ transplantation, cell therapy, or immunotherapy.

A transabdominal intracardiac injection of potassium chloride was administered for fetal reduction, according to the standard of practice.13 Zygosity of each pregnancy was confirmed using the chorionic villus or amniotic fluid obtained from the invasive procedures for prenatal diagnosis by a quantitative fluorescent polymerase chain reaction (PCR) assay (Devyser, Sweden). Immediately before and approximately twice a month after the reduction until the birth, a blood sample from each woman was collected for NIPT to detect fetal trisomy 21 (T21), trisomy 18 (T18), and trisomy 13 (T13). The parents' ages, fetal gestational age at the time of blood sampling, ultrasound scanning results, prenatal diagnosis, and other clinically relevant information were recorded. This study was approved by the Ethics Committee of the Third Affiliated Hospital of Guangzhou Medical University (no. 2019096) and the Institutional Review Board of BGI-Shenzhen (BGI-IRB 19081).

2.2 Plasma cfDNA extraction and sequencing

For each blood sample collection, 5 ml of peripheral blood was drawn from the participating mother into an ethylene diamine tetra-acetic acid-anticoagulated tube. Plasma was separated within 8 hours by a two-step centrifugation protocol.14 Briefly, blood was centrifuged twice at 4°C, and cfDNA was extracted from the plasma by a Micro DNA Kit (Tiangen Biotech, China), following the manufacturer's instructions. The extracted cfDNA was end-repaired and then ligated with adapters for multiplex sequencing following a commercial NIPT protocol in a clinical laboratory certified by ISO15189 and ISO17025. Briefly, the ligated products were added to the Kapa HIFI hotstart ready master mix (Kapa Biosystems, US) and subjected to 12 cycles of amplification. PCR amplification products were then normalized, and then the MGIEasyTM DNA Library Prep Kit (MGI, China) was used to process the products for circularization.15 Briefly, the PCR products were heat-denatured and incubated at 37°C to create single-stranded DNA (ssDNA) circles. The ssDNA circles underwent rolling circle amplification (RCA) to generate DNA nanoballs (DNBs).16 After RCA and the formation of DNBs, the final product was quantified by a QubitTM ssDNA Assay Kit (Invitrogen, US) and loaded on a DNBSEQ-500 platform (MGI, China) for multiplex sequencing,17 with a strategy of single-end 35 base pairs. The data that support the findings of this study have been deposited in the sequencing archive of the China National GeneBank Database (https://db.cngb.org/cnsa/).

2.3 Calculation of fetal fraction

The following three different methods were used to calculate the fetal fractions for different purposes: an artificial neural network model (FF-QuantSC),18 a Y-chromosome method,19 and the relative coverage of the trisomy chromosome.

FF-QuantSC employs a neural network model to identify differential genomic features between the fetal and maternal genomes to calculate fetal fraction. Importantly, FF-QuantSC does not rely on genomic information from the fetal Y chromosome, allowing estimation of the fetal fraction from female fetuses. A fully connected neural network model with a single hidden layer and 128 neurons was trained by the Vapnik–Chervonenkis dimension20 and trained on more than 100,000 male pregnancies. Sequencing reads were partitioned into continuous genomic windows of 60 kb for selection of features. Standardization by within-sample z-score transformation was then used to generate a final feature matrix. After training, the FF-QuantSC model was tested on six groups of about 240,000 pregnancies, including 36,000 twin pregnancies and 80,000 singleton female pregnancies, and showed high concordance with the Y-chromosome method.18

The Y-chromosome method uses the unique reads of chromosome Y to calculate fetal fractions after the sequencing reads are partitioned into continuous genomic windows of 60 kb, as follows: Fetal fraction = 2 × where represents the mean of unique reads on chromosome Y. represents the mean of unique reads on autosomal chromosomes.

In pregnancies with a trisomy cotwin, the fetal fraction was calculated from the relative coverage of the trisomy chromosome with the use of the sequencing reads of the trisomy chromosome, typically T13, T18, or T21, as follows: i = chromosome 13, 18, or 21, where represents the fetal fraction estimate by chromosome ; represents the mean of unique reads on chromosome ; represents the mean of unique reads on autosomal chromosomes. Similarly, sequencing reads were partitioned into continuous genomic windows of 60 kb.

In this study, the FF-QuantSC method was used to calculate the overall fetal fraction of both cotwins in each pregnancy. In pregnancies with one male cotwin, the Y-chromosome method was used to calculate the fetal fraction of the male fetus. In pregnancies with two male fetuses, the Y-chromosome method was used to calculate the fetal fraction equivalent to the overall fetal fraction, and was compared with the results of the FF-QuantSC method. The relative coverage of the trisomy chromosome was specifically used to calculate the fetal fraction of the trisomy cotwin that was terminated by the selective reduction.

2.4 Calculation of chromosome trisomy risk

Since each pregnant woman had been found to carry a cotwin with chromosomal aneuploidy, the risk scores of fetal trisomy before and after the selective fetal reduction were analyzed using a previously described method.19 Briefly, a binary hypothetical t test and logarithmic likelihood L-score ratio between the two t tests were used to classify whether the fetuses had T21, T18, or T13. A T-score greater than three was used as the cut-off value for identifying fetal trisomy, and an L-score greater than one was used to determine the confidence level.

3 RESULTS 3.1 Patients

From January 2017 to January 2018, five women carrying DCDA twins, with one cotwin affected by fetal aneuploidy, were enrolled in this study (Table 1). The median maternal age was 38.8 years, and 80% (4/5) of the mothers were aged older than 35 years. All the women became pregnant by IVF and embryo transfer (IVF-ET), and each woman underwent transplantation of two embryos. At 12 weeks of gestation, an ultrasound examination confirmed that both implanted embryos were viable in each participant and the lack of empty yolk sacs.

TABLE 1. Characteristics and clinical details of the study participants and their pregnancies Samples Patient 1 Patient 2 Patient 3 Patient 4 Patient 5 Maternal age (years) 42 40 31 44 37 Pregnancy history G1P0 G3P0 G1P0 G2P1 G2P1 Chorionicity DCDA DCDA DCDA DCDA DCDA ART method IVF-ET IVF-ET IVF-ET IVF-ET IVF-ET Number of embryos transferred 2 2 2 2 2 Gestational age by ultrasound (week) 13 14 12 + 5 12 + 6 12 + 5 Ultrasound findings Fetus 1: NT 1.5 mm; Fetus 1: NT 2.0 mm; Fetus 1: NT 1.8 mm; Fetus 1: NT 2.7 mm, nasal bone loss, tricuspid regurgitation, single umbilical artery; Fetus 1: NT 3.3 mm; Fetus 2: NT 4.3 mm, nasal bone loss, tricuspid regurgitation, single umbilical artery Fetus 2: NT 3.5 mm, nasal bone loss, omphalocele Fetus 2: NT 2.0 mm, nasal bone loss Fetus 2: NT 2.5 mm Fetus 2: NT 2.6 mm, nasal bone loss, omphalocele Prenatal diagnosis Fetus 1: 46,XX Fetus 1: 46,XY Fetus 1: 46,XY Fetus 1: 47,XX,+18 Fetus 1: 46,XX Fetus 2: 47,XX,+18 Fetus 2: 47,XY,+18 Fetus 2: 47,XX,+21 Fetus 2: 46,XY Fetus 2: 47,XY,+18 Gestational age at reduction surgery (weeks) 14 + 3 14 + 0 17 + 4 15 + 5 14 + 5 Delivery One female without apparent abnormalities One male without apparent abnormalities One male without apparent abnormalities One male without apparent abnormalities One female without apparent abnormalities Gestational age at birth (weeks) 38 + 6 36 + 5 39 + 2 37 + 1 38 + 5 Abbreviations: ART, artificial reproductive technology; DCDA, dichorionic diamniotic; IVF-ET, in vitro fertilization and embryo transfer; NT, nuchal translucency.

All the women had at least one risk factor for fetal aneuploidy, including advanced maternal age, and the abnormal findings during routine prenatal ultrasound examination, such as increased thickness of fetal nuchal translucency and structural abnormalities (Table 1). Fetal aneuploidies were confirmed by CVS or amniocentesis, as follows: four women had a cotwin with T18 and one had a cotwin with T21. Each pregnancy was confirmed to be dizygotic using chorionic villus or amniotic fluid obtained from the invasive procedures.

The median gestational age of the fetuses at the reduction procedure was 15 weeks. After reduction, each woman voluntarily donated her blood samples every 2–5 weeks until delivery. Patients 1, 2, and 5 donated blood seven, five, and six times after reduction, respectively. Patients 3 and 4 both donated blood nine times after reduction. Approximate 10–20 million sequencing reads were obtained from each blood sample.

3.2 Fetal fractions

Figure 1 shows the dynamic changes in fetal DNA fractions, obtained by three different methods, over time in all pregnancies. Detailed data on the fetal fractions are shown in Table S1.

Fetal fractions at multiple time points before and after fetal reduction in five twin pregnancies. The overall fetal fraction was calculated by FF-QuantSC. The fetal fraction of the deceased cotwin was calculated based on the relative coverage of the aneuploidy chromosome. The fetal fraction of male fetus(es) was calculated based on Chromosome Y.

In Patient 1, both fetuses were female, and the fetal fractions could not be calculated based on Y-chromosome signals. However, the fetal fraction of the reduced fetus with T18 was calculated based on trisomy signals. The fetal fraction decreased from 3.3% to 2.0% immediately after reduction and increased slightly to 3.4% at Week 8 after reduction. Subsequently, the deceased T18 fetus gradually stopped releasing cfDNA at Week 16 after reduction, based on a T18 fetal fraction of less than 1%. In contrast, after a transient decline, the overall fetal fraction maintained stability for several weeks, and then increased when the terminated fetus showed reduced release of cfDNA at Week 8 after reduction.

In Patient 2, who had a male euploid fetus and a male T18 fetus, the fetal fraction of the terminated T18 fetus increased from 1.8% to a maximum of 11.7% at Week 9 after reduction and then rapidly decreased to almost 0% at Week 12 after reduction. Since both fetuses were male, the fetal fractions calculated by the Y-chromosome method and the overall fetal fractions calculated by FF-QuantSC almost overlapped, and showed peak levels at Week 9 after reduction following the quick decline afterward. At Week 12 after reduction, when the terminated fetus stopped releasing cfDNA, the fetal fractions of the euploid fetus showed a rapid rebound and increased steadily until birth. At Weeks 16 and 20 after reduction, the overall fetal fractions calculated by FF-QuantSC were slightly higher than the fetal fractions calculated by the Y-chromosome method.

Both Patients 3 and 4 had twins with discordant sex, consisting of a trisomy female fetus and a male euploid fetus. The fetal fractions of the terminated female fetuses showed an increase to peak levels at Weeks 7–9 after reduction. Subsequently, the cfDNA from the terminated fetus of Patient 3 gradually decreased to undetectable levels at Week 20 after reduction, while the fetal fraction of the terminated fetus of Patient 4 decreased to almost 0% at Week 15 after reduction. Once the fetal fractions of the terminated female fetuses began to decline in both patients, the fetal fractions of the euploid male fetuses rapidly increased, especially when the fetal fractions of the terminated female fetuses were almost 0%. While the terminated female fetuses in Patients 3 and 4 still released substantial amounts of cfDNA into the maternal circulation, the overall fetal fractions were higher than the fetal fractions of each of the cotwin. However, after the fetal fractions of the terminated female fetuses fell to relatively low levels, the total fetal fractions were closely correlated with the fetal fractions of the euploid male fetuses.

In Patient 5, there was a male T18 fetus and a female euploid fetus. After the male T18 fetus was terminated, the fetal fractions based on the relative coverage of the trisomy chromosome and the Y chromosome were almost identical. Both measurements showed a transient decrease shortly after fetal reduction and later increase to peak levels at Week 7 after reduction. Finally, they decreased to almost 0% at Week 14 after reduction. Before Week 7 after reduction, the overall fetal fraction was two-to-four-fold greater than the fetal fraction of the terminated male fetus. When the terminated fetus stopped releasing cfDNA at Week 14 after reduction, the overall fetal fraction calculated by FF-QuantSC began to increase rapidly.

3.3 NIPT results

Because of the persistence of a high fetal fraction from each deceased trisomy cotwin in this study, the NIPT results of all five pregnancies suggested they were at high-risk for aneuploidy for 8–16 (median 9.5) weeks after fetal reduction (Figure 2A). Detailed NIPT results of each time point are shown in Table S1.

Noninvasive prenatal testing results according to T-scores at each sample timepoint in five twin pregnancies undergoing selective reduction of a cotwin. (A) T-scores in five twin pregnancies. (B) Incidental finding of unexpected trisomy 13 (T13) along with disappearing T18 signals in Patient 2. (C) Sequencing confirmation of placental T13 mosaicism identified in placental tissue samples obtained from the live cotwin of Patient 2. A copy ratio of 1.0 symbolizes euploidy

The T-score for chromosome 18 in Patient 1 decreased from 7.3 before reduction to 4.5 at Week 3 after reduction. It then increased to 8.2 at Week 8 after reduction. Her T-score subsequently declined to 1.2 at Week 16 after reduction, and her NIPT results then remained low risk for T18.

The T-score for Chromosome 18 in Patient 2 increased dramatically from 4.2 at Week 2 after reduction to 24.6 at Week 9 after reduction. It then decreased rapidly to 2.1 at Week 12 after reduction, and then remained at low risk for T18. Interestingly, her T-score for Chromosome 13 increased to 3.3 at Week 16 after reduction, and then increased to 5.7 at Week 20 after reduction, indicating a high risk of T13 (Figure 2B). This finding was consistent with the results of her fetal fractions, showing that at Weeks 16 and 20 weeks after reduction, the overall fetal fraction as calculated by FF-QuantSC was slightly greater than the fetal fractions of both male fetuses as calculated by the Y-chromosome method (Figure 1). At delivery, DNA extracted from the placental tissue of each cotwin was sequenced, and the results confirmed the placental mosaicism of T13 in the placenta of the live cotwin but not in the placenta of the deceased cotwin (Figure 2C).

The T-score of Chromosome 21 in Patient 3 increased slightly during the first 2 weeks after reduction, from 10.0 to 13.9. Her T-score peaked at 36.6 at Week 7 after reduction, and then quickly decreased to 8.8 at Week 12 after reduction. Since then, her NIPT results suggested a high risk for T21 until Week 20 after reduction.

The T-score for Chromosome 18 in Patient 4 initially increased slightly and then increased rapidly to 17.2 at Week 7 after reduction. The T-score then decreased gradually to 1.9 at Week 15 after reduction. Subsequently, her NIPT results showed a low risk for T18.

The T-score of Chromosome 18 in Patient 5 decreased from 12.4 before reduction to 4.8 at Week 3 after reduction, and then increased to 16.0 at Week 7 after reduction. The T-score then continued to decrease, and at Week 14 after reduction, her NIPT results showed a low risk for T18.

3.4 Outcome

All five women delivered a healthy baby without congenital abnormalities at 37–39 gestational weeks.

4 DISCUSSION

Using selective reduction of a cotwin with aneuploidy to simulate the natural vanishing twin, we demonstrated in this study that the fetal fraction of the terminated fetus underwent dynamic changes throughout pregnancy, which led to NIPT results indicating fetal aneuploidy over a long period of time.

We discovered that two of the five women with DCDA twin pregnancies showed a slight increase in their peripheral blood of cfDNA from the deceased cotwin. In contrast, two of the other women with DCDA twin pregnancies showed a short initial decrease at Weeks 2–3 after fetal reduction. All cases subsequently demonstrated rapid increases in the fetal fractions of the deceased cotwin until 7–9 weeks after reduction. Boyd recently described the placenta's histopathological changes at the intrauterine demise of a fetus.21 After fetal demise, maternal perfusion and fetal blood pressure to the placenta became undetectable, which led to compromised integrity of the villous capillaries and endothelial karyorrhexis, eventually resulting in fibrosis of the avascular villi and the deposition of solid fibrin in the intervillous spaces. Villous edema and intravillous hemorrhage accompanied by maternal inflammation were also seen to occur with a resulting increase in the placenta's autolytic changes.21 Intravascular karyorrhexis and multifocal villous obliteration have been thought to occur soon after fetal demise, whereas the extensive obliteration of stem villi and increased level of villous fibrosis may take 14 days or more.21,22 Thus, we speculate that the initial discordance between the fetal fractions of deceased co-twins at 2 to 3 weeks after reduction reflects the complexity of placental autolysis occurring immediately after the cessation of fetal blood flow and maternal perfusion of the placenta. The later increase in residual cfDNA from the deceased cotwins reflects the increased rate of autolysis of placental tissue as the result of extensive villous fibrosis and stem villi obliteration occurring 14 days after fetal demise. Subsequently, the rate of placental tissue autolysis decreases, leading to decrease in residual cfDNA from the deceased cotwins.

The fetal demise of a cotwin has been known as an important factor in association with false-positive NIPT results.11,23-26 Substantial proportions of false-positive NIPT results ranging from an estimated 15% to 33% might be caused by unrecognized fetal demise.27-29 Critical questions of NIPT clinical practice include how long the residual cfDNA from a deceased cotwin remains in the maternal circulation and if NIPT should be used when the fetal demise of a cotwin has been identified. Two previous studies have provided some intermediate evidence. Curnow et al.23 demonstrated detectable fetal cfDNA from the deceased fetus at 8 weeks after demise using the NIPT data obtained from five vanishing twins, with the fetus deceased at 7–8 gestational weeks. Bevilacqua et al.30 reported seven cases of fetal reduction of a trisomy cotwin at 12–13 gestational weeks. They showed that the aneuploidy risk scores remained positive in two cases at 8–13 weeks after fetal reduction. However, one of these studies did not provide follow-up data, while the other study performed an insufficient follow-up. Thus, neither study revealed any pattern reflecting the risk of aneuploidy or provided longitudinal information on the fetal fractions of residual cfDNA. By contrast, our study consecutively monitored fetal fraction changes after fetal reduction quantified by different methods of estimation. We observed a consistent pattern for the release of residual cfDNA from the demised cotwin that, to the best of our knowledge, had not been previously reported. Niles et al.31 recently reported a sporadic case of NIPT false-positive results that showed discordant fetal sex and aneuploidy risk at 15 weeks after the death of a cotwin. In comparison with this single-case study, our study investigated five clinical cases with a longer follow-up, and revealed a longer duration of residual cfDNA from the deceased fetus in maternal blood samples.

A limitation of this study was that the selective reductions were performed at 14–17 gestational weeks, limiting this fetal demise model to the early second trimester of pregnancy. This window of time is later than the time when NIPT is administered as a first-tier screening test during the first trimester in some countries. Thus, the results of this study would have reduced clinical significance for predicting the results of NIPT when the death of the cotwin occurs early in the first trimester. However, previous reports on vanishing twins during early pregnancy, although lacking consecutive follow-up, showed false-positive results on NIPT at 8 weeks after fetal demise.23,30 Hence, the clinical implications of this study could reasonably be extended to early pregnancies with the fetal demise of a cotwin. Furthermore, in countries such as China, where NIPT is provided each year to a large number of pregnant women early in the second trimester as a second-tier screen test, the mean gestational age of the fetuses of mothers receiving NIPT may be 18–19 weeks.32 Therefore, this study is important to clinicians and genetic counselors who should understand that undiagnosed fetal demise may be an important factor leading to false-positive NIPT results. Our results also emphasize the role of ultrasound before NIPT to investigate a suspected occurrence of vanishing twin.

Other limitations of this study include the small sample size and restriction to only have women pregnant with DCDA twins, in which a trisomy cotwin was selectively reduced to simulate the natural demise of a cotwin. To facilitate the differential analysis of the fraction from each cotwin, we used dizygotic twins only in our study. Whether the same findings would occur in monozygotic twins, especially monochorionic twins with a significantly higher risk of fetal demise is unclear.33,34 Moreover, the results from the reduced trisomy cotwin might not fully represent the natural demise of a cotwin, which can be caused by many other reasons with or without the presence of chromosomal abnormalities.7,35 Last, because of logistical difficulties during clinical practice, the time points of sample collection from each woman were not identical. In one case (Patient 2), a blood sample was not taken before the reduction.

5 CONCLUSIONS

We observed temporary increases over time in the residual cfDNA in the maternal plasma from selectively reduced cotwins aged 14–17 gestational weeks, which ultimately led to the identification of consistent patterns in the changes in the fetal fractions and suggested a significant risk of fetal trisomy for up to 16 weeks after fetal reduction. Thus, our data do not support the use of NIPT in pregnancies in which fetal death has been identified at early second trimester.

ACKNOWLEDGMENTS

The study was supported by the National Key Research and Development Program of China (2018YFC 1004104), the National Natural Science Foundation of China (no. 81671470), and Shenzhen Municipal Government of China (JCYJ20180703093402288 and JCYJ20170412152854656). The authors would like to thank all the participating women for their cooperation in this study, and Dr Tze Kin Lau for his valuable suggestions on the revision of the manuscript.

CONFLICT OF INTERESTS

The authors declare that there are no conflict of interests.

ETHICS STATEMENT

This study was approved by the Ethics Committee of the Third Affiliated Hospital of Guangzhou Medical University (no. 2019096) and the Institutional Review Board of BGI-Shenzhen (no. 19081).