Original Article

Effects of Maternal and Fetal Characteristics on Cell-Free Fetal DNA Fraction in Maternal Plasma

Reproductive Sciences 1-7 ª The Author(s) 2015 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/1933719115584445 rs.sagepub.com

Yi Zhou, MD1, Zhongyi Zhu, BMed2, Ya Gao, PhD2, Yuying Yuan, BS2, Yulai Guo, BS2, Lijun Zhou, BS2, KaiKai Liao, BMed2, Jun Wang, MSc2, Bole Du, BS2, Yumei Hou, BS2, Zhonglin Chen, BS2, Fang Chen, MSc3,4, Hongyun Zhang, MSc2, Cong Yu2, Lijian Zhao, BMed2, T. K. Lau, PhD5, Fuman Jiang, MSc2, and Wei Wang, PhD2,3,4

Abstract Objective: To study factors that influence the concentration of cell-free fetal DNA (fetal fraction) using a large clinical data set of pregnancies with male fetus. Method: A retrospective analysis of 23 067 pregnancies that received noninvasive prenatal testing from January 2012 to October 2013, including 22 650 normal singleton pregnancies (control group) and 417 pregnancies with aneuploidy, twin pregnancy, or various maternal conditions including preexisting hypertension, preexisting diabetes, hyperthyroidism, and carrier of the surface antigen of the hepatitis B virus (HBsAg; study group). Multiples of the median (MoM) analysis was performed in the control group to derive gestation and body mass index (BMI)-corrected fetal fraction. The effects of study group conditions on fetal fraction were examined by calculating the ratio of MoM (RMoM) values. Results: Fetal fraction showed a positive correlation with gestational age (r2 ¼ .10, P < .001) and increased rapidly after the 21 weeks of gestation (r2 ¼ .26, P < .001). Negative association with maternal BMI was found with fetal fraction (r2 ¼ .04, P < .001). In study group, fetal fraction was higher among pregnant women with a trisomy 21 fetus (RMoM ¼ 1.24, P < .001) and lower among trisomy 18 (RMoM ¼ 0.84, P < .001). A 1.6-fold incensement of fetal fraction was observed in twin fetuses comparing to singleton pregnancy (RMoM ¼ 1.62, P < .001). Women with preexisting hypertension had significantly lower fetal fraction (RMoM ¼ 0.85, P ¼ .02). Preexisting diabetes, hyperthyroidism, or carrier of HBsAg did not affect fetal fraction. Conclusion: The fetal fraction was affected by fetal aneuploidy, maternal BMI, and the number of gestation. Maternal preexisting of hypertension appeared to reduce fetal fraction. Keywords fetal fraction, gestational age, body mass index, preexisting hypertension, twin pregnancy, multiples of the median.

Introduction Cell-free fetal DNA (cffDNA) was first discovered in maternal plasma in 1997.1 Since then, cffDNA has been used successfully in detecting fetal aneuploidies and even for fetal whole genome genotyping. There have been continuous efforts trying to enrich cffDNA in maternal plasma during the development of noninvasive prenatal testing (NIPT) because low fetal fraction can impair the test efficacy, potentially leading to a false-negative result or a higher test failure rate.2 Therefore, identifying factors that affect fetal fraction may help to optimize protocol of NIPT bioinformatics analysis, and the presence of these influential factors may also be taken into consideration during genetic counseling of pregnant women. In recent years, several factors affecting fetal fraction have been reported. Previous researchers have found that fetal

fraction increases with gestational age and decreases with maternal weight.3,4 Other studies have found that fetal fraction was significantly higher in pregnant women with preeclampsia

1 Fetal Medicine Center, First Affiliated Hospital of Sun Yat-sen University, Guangzhou, Guangdong Province, China 2 Clinical Laboratory of BGI Diagnostics, BGI-Shenzhen, Shenzhen, China 3 BGI-Research, BGI-Shenzhen, Shenzhen, China 4 Shenzhen Birth Defect Screening Project Lab, BGI-Shenzhen, Shenzhen, China 5 Fetal Medicine Centre, Paramount Medical Centre, Hong Kong

Corresponding Authors: Fuman Jiang and Wei Wang, Clinical Laboratory of BGI Diagnostics, Shenzhen, China. Emails: [email protected]; [email protected]

Downloaded from rsx.sagepub.com at SUNY BINGHAMTON on May 13, 2015

2

Reproductive Sciences

or fetal growth restriction.5-7 In addition, levels of serum pregnancy-associated plasma protein A and free b-human chorionic gonadotropin have been found to correlate with fetal fraction in maternal plasma.8 On the other hand, no correlation was detected between fetal fraction and smoking status, ethnicity, or the use of in vitro fertilization.9,10 However, most of these previous researches had small sample size. There is also lack of conclusive data in twin pregnancy to show the concentration of fetal fraction per fetus and the comparison to singleton pregnancies. Moreover, controversial results were reported regarding the correlation between fetal aneuploidy and fetal fraction. Three studies reported increased fetal fraction in pregnant women carrying fetuses with Down syndrome and trisomy 13.11,12 However, some other studies showed no significant increment of fetal fraction in pregnancy with Down syndrome fetus,3,8,13 whereas others reported the decreased level of fetal fraction in trisomy 18, 13, and monosomy X.14-16 The objective of this study was to validate the previous findings of fetal fraction with gestational age, maternal weight, and fetal aneuploidy using large clinical data of pregnancy with male fetus previously collected in our clinical laboratory. Relationship of fetal fraction with twin pregnancy and various maternal conditions was explored to add new evidence to the understanding of factors affecting fetal fraction, which is important for clinical application of NIPT.

pregnancy; (2) those with maternal conditions, including hyperthyroidism, carrier of the surface antigen of the hepatitis B virus (HBsAg), preexisting diabetes, and preexisting hypertension; and (3) fetal aneuploidy including trisomy 21 and trisomy18.

Fetal Fraction Estimation in Male Pregnancy Fetal fraction of male pregnancy was estimated using the following formula: ei;Y ¼

0 cri;Y  cri;Y ;f ; 0 0 cri;Y ;m  cri;Y ;f

where ei;Y represents the fetal fraction estimate by chromosome 0 Y of sample i. Specifically, cri;j;m ¼ fj;m ðGCi;j Þ ðj ¼ X ; Y Þ indicates the fitted relative k-mer coverage from a regression of an 0 adult male data set, and cri;j;f ¼ fj;f ðGCi;j Þ ðj ¼ X ; Y Þ indicates the fitted relative k-mer coverage from a regression of a fetal female dataset. More detailed information about this estimation method can be obtained in the previously published MPS-based test methods article.17

Accuracy Evaluation of Fetal Fraction Estimated by MPSBased Test

Materials and Methods Study Population This study retrospectively analyzed maternal plasma samples of male pregnancies submitted to the clinical laboratories of BGI Diagnostics (Shenzhen, China) between January 2012 and October 2013. Written informed consent was obtained from each participant. Ethical approval was obtained from the institutional review board of BGI before recruiting at each obstetric unit. For each sample, 5 mL of maternal peripheral blood was collected for MPS-based NIPT followed by the testing pipeline17 and quality control procedure18 as described previously. Samples included in this study should meet the following criteria: (1) confirmation of male pregnancy by karyotyping or postnatal follow-up (pregnancy outcome). For twin pregnancy, only confirmed male/male and male/female pregnancies were included; (2) samples passed quality control and completed the NIPT test; and (3) samples with full information of gestational age, maternal age, body mass index (BMI), and medical history. Totally, 23 067 cases fulfilling the abovementioned criteria remained for further analysis. These cases were then stratified into the control group (n ¼ 22 650) and study group (n ¼ 417). Cases in the control group were singleton pregnancies with natural conception, without known medical diseases and with normal fetal karyotype or normal phenotype at birth if karyotyping was not performed. All remaining cases were categorized to the study group, which was further divided into the following 3 subgroups: (1) twin

In order to evaluate the accuracy and repeatability of our fetal fraction estimation protocol, a series of mixed DNA samples with fetal fraction of 2%, 4%, 8%, 10%, and 13% were artificially constructed. For each concentration, 4 replicates were made, resulting in a total of 20 samples with ‘‘theoretical fetal fractions.’’ Subsequently, each artificial sample was subjected to the MPS experiment in a blinded manner, using the same MPS-based NIPT protocol and fetal estimation method mentioned earlier, and the results were called ‘‘estimated fetal fractions.’’ The accuracy of MPS-based fetal fraction estimation was studied by comparing the theoretical and estimated fetal fractions by Pearson product-moment coefficient (R) and by the pair differences between these 2 measurements.

Statistical Analysis Using the 22 650 control cases, quadratic and linear models were used to calculate the expected multiples of the median (MoM) of fetal fraction corrected by gestational age and maternal BMI. Gestation-corrected MoM and BMIcorrected MoM (BcMoM) values of each case in the control group and study group were calculated based on abovementioned correction model (Supplemental Method). The deviation between the value of mean BcMoM in each study group and theoretical value ‘‘1’’ was defined as the ratio of MoM (RMoM), and the corresponding significance was tested by 1-sample t test.

Downloaded from rsx.sagepub.com at SUNY BINGHAMTON on May 13, 2015

Zhou et al

3 Table 1. Characteristics of the Study Population.a,b Maternal Characteristic

Figure 1. A scatter plot of the estimated fetal fraction by massively parallel sequencing against the theoretical fetal fraction. Pearson product moment coefficient (r) was .996 (P < .001).

Results Accuracy Validation of Fetal Fraction Estimation Through Artificial Samples The fetal fraction estimated by the MPS-based method using artificial samples is shown in Figure 1. At the designed concentration of 2%, 4%, 8%, 10%, and 13%, a strong linear correlation was found between the estimated fetal fractions and the theoretical fraction (R > .99, P < .001). The absolute difference between the estimated and theoretical fraction was on average 0.47% (Supplemental Table S1). These results showed high concordance between the estimated and theoretical fetal fraction, which demonstrates that our method of fetal fraction estimation is accurate and consistent between replicates. Based on this result, our method of fetal fraction estimation was used in real clinical samples for the following analysis.

Fetal Fraction with Gestational Age, Maternal Age, and BMI A total of 23 067 cases that fulfilled the recruitment criteria were included in this study. Basic characteristics of the study population are shown in Table 1. In the control group, fetal fractions at different gestational weeks were estimated, and the correlations between fetal fraction and maternal BMI as well as maternal age are shown in Figure 2. Specifically, in Figure 2A, a nonlinear correlation was observed between fetal fraction and gestational age (r2 ¼ .10, P < .001). The median fetal fraction at 10th gestational week was about 9%, and the level maintains relatively stable until 21st gestational week. The total increment of median fetal fraction from 10th to 21st gestational week was about just 1%. After 21st week, fetal fraction

Age, years Weight, kg Height, m BMI Gestational age 10 þ 0 to 12 þ 6 weeks 13 þ 0 to 16 þ 6 weeks 17 þ 0 to 21 þ 6 weeks 22 þ 0 to 27 þ 6 weeks 28 þ 0 to 31 þ 6 weeks 32 weeks and above Karyotype Trisomy 21 Trisomy 18 Twins Male–female twins Male–male twins Preexisting hypertension Preexisting diabetes Hyperthyroidism HBsAg positives

Samples (n ¼ 23 067) 30.9 (16-50) 58.8 (40-80) 1.61 (1.38-1.85) 22.5 (13.7-34.2) 424 4229 14 719 3231 355 98

(1.84) (18.34) (63.84) (14.01) (1.54) (0.43)

108 (0.47) 38 (0.17) 28 31 42 41 62 67

(0.10) (0.11) (0.18) (0.18) (0.27) (0.29)

Abbreviations: MoM, multiples of the median; BMI, body mass index; HBsAg, surface antigen of the hepatitis B virus. a Values are median (range) or n (%). b The correlations between fetal fraction and maternal characteristics (eg, gestational age, maternal age, and maternal BMI) and the MoM adjust approach.

increased rapidly with gestation, at a rate of almost 1% per week (r2 ¼ .26, P < .001). After 32nd week, the fetal fraction appeared still increasing, although the sample sizes were inadequate for reliable statistical analysis. On the other hand, destiny distribution of fetal fraction in the control group was also presented in Figure 2A. It showed skew normal distributions, with a median fetal faction of 9.31% (mean ¼ 10.02%). Of theses samples, 2% were lower than cutoff value 3.5%, and 0.9% of samples have a fetal fraction higher than 25%. As expected, fetal fraction in this population showed negative correlation with maternal BMI (Figure 2B, r2 ¼ .04, P < .001) and no correlation with maternal age (Figure 2C). Based on this result, curve fits were generated using 22 650 control group cases to drive correction model for gestational age and BMI (Supplemental Figure 1). Corresponding output correction equation was used for further analysis.

Fetal Fraction With Twin Pregnancy, Maternal Conditions, and Fetal Aneuploidy Mean BcMoM value of the 22 650 cases in the control group was calculated as 1.001, which showed no statistical difference to the standard value 1 (P ¼ .957). For studying the fetal fraction of fetal aneuploidy, twin pregnancy, and maternal conditions in the study group, mean fetal fraction, BcMoM value, and RMoM value are shown in Table 2. Results showed that fetal fraction was significantly higher among women carrying a trisomy 21 fetus (RMoM ¼ 1.24, P < .001) and significantly

Downloaded from rsx.sagepub.com at SUNY BINGHAMTON on May 13, 2015

4

Reproductive Sciences

Figure 2. A, The density distribution of fetal fraction and its relationship with gestational weeks. The boxes represent 22 650 samples of control group. The upper and lower whiskers represent the 5th and 95th percentiles. The upper, middle and lower bars represent the 25th, 50th, and 75th percentiles, respectively. The open circles represent outlier samples of control group. B, Relationship between the fetal fraction and pregnancy BMI. C, Relationship between the fetal fraction and maternal age. BMI indicates body mass index.

lower among those with a trisomy 18 fetus (RMoM ¼ 0.84, P < .001). This implies that pregnancy with fetal trisomy 21 can result in higher fetal fraction than euploid singleton pregnancy, whereas pregnancy with fetal trisomy 18 leads to reduced fetal fraction. In twin pregnancies, male–male twins had a significantly higher fetal fraction than euploid singleton pregnancy, as shown by a RMoM of 1.62 (P < .001). In contrast, the RMoM in the male–female twin pregnancies had no statistical

difference from euploid singleton pregnancy (RMoM ¼ 0.89, P ¼ .36), which was expected because only sequencing tags from Y-chromosome was analyzed and the contribution of only male fetus was measured. Detail information about twin samples and chronicity is shown in Supplemental Table S2. Among the pregnancies with maternal medical conditions, pregnancies with preexisting hypertension had a significantly lower fetal fraction, with a RMoM of 0.85 (P ¼ .02), while preexisting

Downloaded from rsx.sagepub.com at SUNY BINGHAMTON on May 13, 2015

Zhou et al

5

Table 2. Comparison of the Fetal Fraction Values of Cases in the Study Group. Fetal Fraction Study Groups Trisomy 21 Trisomy 18 Hyperthyroidism HBsAg positive Preexisting diabetes Preexisting hypertension Male–female twins Male–male twins

Estimation Based on Y-Chromosome 12.35 8.03 9.78 9.91 8.82 7.29 9.70 18.06

(4.8-31.7) (3.7-14.7) (3.1-19.3) (4.3-28.9) (3.7-22.3) (2.5-16.2) (2.5-23.3) (7.2-38.6)

Statistical Values BcMoM Values 1.24 0.84 0.99 0.96 0.93 0.85 0.89 1.62

(0.5-2.7) (0.3-1.5) (0.4-2.1) (0.5-2.0) (0.5-1.7) (0.2-1.7) (0.3-2.7) (0.7-2.9)

Cases No.

RMoM

P Values

108 38 62 67 41 42 28 31

1.24 0.84 0.99 0.96 0.93 0.85 0.89 1.62

.00a .00a .84 .35 .19 .02a .36 .00a

Abbreviations: BcMoM, body mass index-corrected multiples of the median; HBsAg, surface antigen of the hepatitis B virus; RMoM, the ratio between mean adjust multiples of the median value and theoretical ‘‘one’’; case no., the total case number of each study groups. a P values calculated with the 1-sample t test of MoM values (P value