Comparison of two high-throughput semiconductor chip sequencing platforms in noninvasive prenatal testing for Down syndrome in early pregnancy

Recently, early stage prenatal screening to detect fetal aneuploidy has become common for pregnant women [1, 2]. Women over the age of 35 have an increased risk of giving birth to an abnormal baby; hence, accurate prediction-based tests for fetuses are required. Several prenatal screening methods are in use. Common first-trimester screening comprises a combination of ultrasound and maternal serum markers [3]. Women at high risk for fetal chromosome abnormalities have the option to undergo invasive prenatal diagnostic tests such as chorionic villus sampling (CVS) at 10–13 weeks gestational age or amniocentesis at 15–18 weeks gestational age [4, 5]. However, these tests are reported to be associated with iatrogenic pregnancy loss [6].

Decades ago, researchers discovered that cancer DNA could be detected as both circulating tumor cells and cell-free DNA (cfDNA) in human blood. Using this observation, researchers proposed that fetal DNA could be detected using molecular assays and cell-free fetal DNA (cffDNA) was indeed detected in maternal blood [7]. Today, women can choose noninvasive prenatal testing (NIPT) of cffDNA [8]. CffDNA is generally produced from apoptotic trophoblasts in the placenta [7] during pregnancy, and the amount of cffDNA increases with gestational age. During weeks 10–22 of gestation, approximately 10 % of free DNA in the mother’s plasma is estimated to be cffDNA. There is a very high level of variance in the fraction and amount of cffDNA among individuals [9]. Therefore, applying molecular testing requires an extremely accurate detection approach or a massive amount of data to overcome this detection difficulty. Our previous work [10] showed that it is possible to detect fetal chromosome abnormalities for pregnant women in weeks 12–21 of gestation by amplifying and producing a large number of DNA fragments for quantitative analyses. The percentage of cffDNA is proportional to gestational age; therefore, it is important to perform the test at the right time. At the early stage of pregnancy, the fraction of fetal DNA is perhaps the most important factor for NIPT, because the common aneuploidies are very difficult to detect for pregnant women in the early weeks of gestation [11] because of the lack of fetal DNA.

To carry out efficient NIPT, it is necessary to understand the limitations and characteristics of cffDNA. For example, cffDNA is normally only around 150 base pairs (bp) [12]. Also, cffDNA has a short half-life of about 16 min [13, 14].

NIPT technologies have been well accepted because of two critical clinical benefits: there is no risk of pregnancy loss and NIPT can be used as an early pregnancy test compared with amniocentesis; however, discordant NIPT data resulting from placental or maternal cell mosaicism requires full karyotyping, which is the gold standard for aneuploidy tests, to confirm positive outputs [15–17].

Nowadays, next-generation sequencing (NGS) is widely used for NIPT [18–24]. Ion Torrent PGM from Life Technologies, a semiconductor-based sequencing platform assessed here for NIPT and compared with its later and larger-capacity version of Ion Torrent Proton, enables a reduced turnaround time of sequencing data to within 2 to 4 h for a clinical sequencing service. A recent study described the outcomes for noninvasive detection of common fetal trisomy 13, 18, and 21 using the ion semiconductor platform, Ion Proton, which produced greater than 98 % sensitivity and specificity [23]. However, there are many different types and versions of sequencers, making it necessary for granting authorities to evaluate the generality and robustness of such sequencers for NIPT.

As far as we know, there has been no direct comparison between PGM and Proton sequencers for NIPT with large sample sizes. Another reason to compare PGM and Proton sequencers is that PGM is faster. The turnaround time for clinical applications is very important, and ion semiconductor sequencers, such as PGM, have been widely accepted because of their simplicity and speed. Prenatal testing using PGM has become feasible for the noninvasive detection of fetal aneuploidy [25]. PGM machines are small and inexpensive; therefore, comparing the benefits and overall accuracy of both PGM and Proton in terms of data produced could be important clinically for doctors who would like to use NIPT. The experimental cost of the bench-top PGM sequencer is much lower than other common sequencers, although PGM’s general sequencing accuracy is lower. From Ion 314/316/318 chips of PGM, 70 Mbp to 2 Gbp of raw sequencing data can be produced, and 85 % of the reads usually reaches Q20, a common sequencing quality threshold. With the Ion Proton platform, nearly 10 Gbp of raw sequencing data are produced with approximately 75 % of them reaching Q20. This amount is sufficient to determine the quantitative variation caused by chromosomal abnormality. A recent study confirmed that high-throughput ion semiconductor sequencing was feasible in noninvasive prenatal testing of fetal aneuploidies [10, 23]. In that study, only eight maternal plasma DNA samples were used, comprising four normal pregnancies and four with trisomy 21 fetuses, which were sequenced on Ion Torrent 314/316/318/PI chips [26]. Separately, using different samples, Ion Proton sequencer-based results have also demonstrated successfully that ion semiconductor sequencing is suitable for NIPT [10, 23].

Here, we compared the Ion Torrent PGM and Proton platforms for NIPT for fetal trisomy 21 directly using PGM and Proton simultaneously for the same set of samples to provide some general applicability of ion semiconductor-based sequencers. We were also interested in demonstrating whether any sequencers could discriminate fetal aneuploidy from normal chromosomes in blood, even if the fetal DNA fraction is not sufficient for NIPT. In doing so, we tried to detect fetal trisomy 21 for the pregnant women, most of whom (69.79 %) were at weeks 11 to 13 of gestation.

Previous studies detected no quality differences between Ion Torrent sequencer and the nucleotide synthesis-based Illumina sequencers by comparing PGM data vs. HiSeq2000- and MiSeq-derived data, although the error rate of PGM data was relatively higher than those of the HiSeq2000 and MiSeq data [31, 32]. Chen et al. [33] evaluated the performances of Proton and MiSeq systems for extremely low-coverage sequencing, and showed that both sequencers detected aneuploidies correctly. Wang et al. concluded that the data quality of Ion Torrent PGM was generally better than that of the Proton system [26]. However, they showed that both PGM- and Proton-based semiconductor high-throughput sequencing was feasible in the noninvasive prenatal testing of fetal aneuploidies [26], although it was carried out with a very small number of samples. In our previous study, we showed that the detection of fetal T18 and T21 could be carried out using the Ion Torrent Proton system [10]. A large-scale clinical study by Liao et al. also showed that NIPT using an Ion Proton sequencer could be successful [23]. In addition, the Ion Proton system can generate about 80 million raw reads in 3–4 h, allowing chromosomal aneuploidy detection in 2 working days. Hence, it could be suitable for operations that require fast and accurate turnaround times [33].

Conventional prenatal screening is not perfect and its detection accuracy for chromosomal aneuploidy is below 100 %. Screens using maternal serum markers and ultrasound have been approved, but have lower accuracy. Henry et al. [34] analyzed Down syndrome births after routine noninvasive screening, based on conventional, non-sequencing tests, which have mostly replaced age-related invasive processes. They found that despite the increase in prenatal screening, newborn children with Down syndrome increased in women over the age of 35 years, because the mothers believed that a single blood test would be sufficient to detect trisomy 21.

A positive NIPT result should include a follow-up test with an invasive prenatal method to confirm the fetal chromosomal aneuploidies. Accurate detection of common chromosomal aneuploidies, particularly T18 and T21, has been consistently reported in previous works of NIPT using cfDNA [18–23]. The current price of a plasma cfDNA test is $800–2000 in the US and $500–1500 in some other countries [8]. Two previous studies evaluated the cost of a cfDNA test in women with positive prevalent screening outcomes and concluded that the use of the cfDNA test was associated with a net price reduction compared with conventional CVS or amniocentesis [35, 36]. One limitation of our current study is that it could not detect additional positive cases for other common chromosomal aneuploidies, such as T18 and T13.

In terms of the time to perform NIPT, a previous study [10] reported that for high sensitivity and specificity, the samples should be collected during weeks 14–21 of gestation. It is advantageous to find a chromosomal abnormality at an early stage of pregnancy; therefore, it is crucial to investigate whether a similar accuracy could be achieved in early pregnancy to expand the clinical utility of NIPT. Our current results indicated that both semiconductor-based platforms are sufficiently sensitive and effective for pregnant women, most of which were at 11 to 13 weeks of gestation.

Although both the PGM and Proton platforms yielded similar results for NIPT, using the minimal z-scores thresholds identified here for classification cannot be a standard measure, because there is usually a grey zone where the classification is not certain and further investigation is required. Therefore, a larger sample size is required to confirm the reliability and validity of our results.

We showed that fetal trisomy 21 could be detected successfully by two different ion semiconductor sequencers (Ion Torrent PGM and Proton) and confirmed that both Ion chips are suitable for cfDNA screening for pregnant women at an early stage. In Korea, PGM is approved as a medical instrument for cfDNA testing, while the Proton awaits approval. Therefore, our investigation provided evidence of PGM’s applicability to public clinical tests. As PGM has been used for NIPT for some years, these results could be used in other countries where NIPT is provided to pregnant women.