In the past few years it has become clear that the demographic shift towards increasing maternal age for pregnancies in developed nations, with its associated risks of fetal chromosomal anomalies, will necessitate a change in current strategies for prenatal screening and detection of fetal aneuploidies (Refs Reference Hahn, Jackson and Zimmermann1, Reference Bianchi2, Reference Ferguson-Smith and Bianchi3, Reference Lo4). This is illustrated by a recent analysis of the English and Welsh National Down Syndrome (DS) Cytogenetic Register over the period 1989–2008 (Ref. Reference Morris and Alberman5), which reported a 71% increase in the number of diagnosed cases with DS, with no comparable increase in birth rate. The increase in the number of cases with DS was largely attributed to the concomitant increase in maternal age, as more than 20% of pregnancies now occur in mothers older than 35 years (Ref. Reference Morris and Alberman5).

This development is likely to add additional costs to already strained healthcare budgets, because positive cases from initial noninvasive screening based on ultrascans and maternal serum biochemical analysis need to be verified by invasive analysis such as amniocentesis or chorionic villous sampling, which are labour-intensive examinations requiring highly skilled personnel (Ref. Reference Philip6). To break this spiralling cost, it would be advantageous for any future noninvasive method to be so accurate that it could function as a ‘stand-alone’ test and not require further invasive verification. Naturally, it will also need to be considerably cheaper than current invasive practices. An alternative scenario might be to use such a noninvasive test to reduce the cost and risk associated with the invasive analysis of the high number of false-positive cases resulting from current screening practice (Refs Reference Chiu7, Reference Jackson and Pyeritz8). In either case, strenuous efforts will need to be undertaken in order to ensure that cost-effective noninvasive prenatal diagnosis of DS cases finally becomes a reality (Ref. Reference Jackson and Pyeritz8).

Has research in the field of noninvasive prenatal diagnosis become ‘mature’?

Since the discovery of fetal cell-free DNA (cf-DNA) in maternal plasma or serum in 1997 (Ref. Reference Lo9), more than 1000 papers have been published on this topic. A cursory review of publications listed in public depositories such as PubMed (http://www.ncbi.nlm.nih.gov/pubmed) suggests that the most prolific period was probably the latter half of the previous decade. In this period, several large-scale studies were initiated to test the efficacy of this new-found tool for the noninvasive determination of fetal genetic traits such as the fetal RhD (Rhesus D) gene or gender (Refs Reference van der Schoot, Hahn and Chitty10, Reference Chitty11, Reference Finning12, Reference Gautier13). These data indicated that this approach was indeed sufficiently robust that widespread clinical application could be safely implemented, and as a result concerted efforts were undertaken by multicentre consortia, such as the European Union (EU)-funded SAFE Network, to standardise these assays (Refs Reference Chitty11, Reference Maddocks14). These data also indicated that the analysis of cf-fetal DNA could form the secure and sound basis for subsequent developments, such as the noninvasive detection of fetal aneuploidies (Ref. Reference Lo15).

In recent years this prolific output in publications seems to have largely tapered off, although the number of reviews dealing with the field has increased tremendously. Hence the scenario appears very similar to economic models of the ‘product life cycle’, where a particular item has progressed through development, introduction and growth, and is settling into a pattern of maturity and subsequent decline (Ref. Reference Krentz and Pilskaln16). Although this simplistic view might lead to the impression that research in this field has waned, this is far from the truth, for this punctuated modicum in published reports indicates that the field has finally entered a phase where quality, using a quantum leap in technological development, rather than quantity is the prevailing trend. This facet will become very evident in this review.

The question of fetal cell-free DNA fragment size and concentration

It is necessary to reiterate that the major problem still hampering the use of fetal cf-DNA for the noninvasive detection of fetal genetic loci that are not distinct from maternal loci is that fetal sequences constitute only 5–10% of the total cf-DNA in maternal plasma, and even less in maternal serum (Refs Reference Lo17, Reference Zimmermann18, Reference Sikora19).

Groundbreaking research performed in the middle of the past decade on the biophysical properties of cf-DNA indicated that fetal cf-DNA was more fragmented and had a shorter size than comparable maternal cf-DNA fragments (Refs Reference Chan20, Reference Li21). Interestingly, this topic still continues to be the focus of intense research efforts and considerable debate (Refs Reference Hahn and Zimmermann22, Reference Fan23, Reference Lo24). The fragmentation was exploited for the selective enrichment of fetal cf-DNA sequences (to up to 50% of total cf-DNA), thereby permitting the detection of otherwise masked fetal genetic loci such as short tandem repeats, single-nucleotide polymorphisms (SNPs) and paternally inherited point mutations, such as those involved in β-thalassaemia or achondroplasia (Refs Reference Li21, Reference Li25, Reference Li26, Reference Li27, Reference Li28). These experiments, however, used conventional agarose gel electrophoresis for the physical-size-based separation of fetal and maternal cf-DNA species – a cumbersome, labour-intensive, inefficient procedure that is prone to contamination. For the widespread applicability of this approach in the clinic, new approaches such as microfluidics will be required.

Another key development during this period was verification of the long-standing suspicion that fetal cf-DNA was of placental origin, and was not the result of the demise of trafficking fetal cells. This was confirmed by a number of different approaches, such as analysis of (1) placental mosaicism, where key fetal loci such as the Y chromosome were missing both in the placenta and in fetal cf-DNA, but not in the male fetal karyotype (Ref. Reference Flori29), (2) molar pregnancies, where cf-DNA specific for the molar karyotype (46 XY) could be detected (Ref. Reference Alberry30), and (3) epigenetic markers specific for placental tissues, such as the hypomethylated maspin (SERPINB5) or the hypermethylated RASSF1 genetic loci (Refs Reference Chim31, Reference Chim32, Reference Chiu33).

The latter markers could prove to be useful as gender-independent tools to verify the presence of fetal cf-DNA in ambiguous diagnostic cases, such as when determining fetal gender or RhD status (Ref. Reference van der Schoot, Hahn and Chitty10). It might also be possible to use these as tools to quantify fetal cf-DNA levels (Ref. Reference Ehrich34). Reliable, accurate assessment of fetal cf-DNA levels, however, appears to require the use of high-copy sequences, such as DYS14 on the Y chromosome (Ref. Reference Zimmermann18).

Indirect methods for the noninvasive detection of fetal aneuploidy using cell-free nucleic acids

Digital PCR: first hint at direct noninvasive aneuploidy detection

Most of the studies examining fetal cf-DNA to date have used a form of real-time PCR and Y-chromosome-specific sequences (Refs Reference Lo17, Reference Zimmermann43). These studies indicated that the amount of fetal cf-DNA increases during gestation, and disappears rapidly from the maternal circulation following delivery. Increases were also observed in several pregnancy-related disorders or conditions such as preeclampsia, preterm delivery and trisomy 21, suggesting that this phenomenon could serve as the basis for a new generation of screening tests.

During this period, studies had indicated that real-time PCR could be used for the rapid determination of chromosomal ploidy on pure fetal genetic material obtained by invasive means (Ref. Reference Zimmermann44). Because real-time PCR is not well suited for the detection of less than twofold differences in template copy numbers, special conditions had to be introduced to detect the 1.5-fold difference in template concentration occurring in trisomy 21. It was, however, very clear that this approach would not be suited for the noninvasive determination of DS, because of the overwhelming presence of maternal cf-DNA fragments.

Consequently, a different tack had to be taken, which was provided by ‘digital PCR’ (Ref. Reference Vogelstein and Kinzler45) and the advent of microfluidic devices (Ref. Reference Zimmermann46). Unlike real-time PCR, where an ‘analogue’ signal of the entire PCR reaction containing the entire input template is obtained, in ‘digital PCR’ the PCR reaction is split into thousands of minute individual reactions, with each individual reaction containing at most a single template copy. A quantitative assessment of the concentration of input template in the sample examined is then made by counting the number of individual positive PCR reactions (Fig. 1).

Figure 1. Detection of fetal aneuploidy using digital PCR. In this procedure, fluorescent PCR specific for sequences on chromosomes 12 and 21 is carried out in individual microreaction chambers. The amount of input template is titrated in such a fashion that each microreaction vessel contains >1 copy. After the plateau phase has been reached (approximately 40 cycles), the PCR reaction is terminated, and the number of positive reactions for each locus is counted. In euploid cases the ratio of blue (chromosome 12) to red (chromosome 21) signals should be 1, whereas in cases with Down syndrome the ratio of blue (chromosome 12) to red (chromosome 21) signals should be 1.5 (illustrated in more detail in Ref. Reference Zimmermann46). This method has to date not been successfully used for the detection of fetal aneuploidy using cell-free DNA in maternal plasma.

This procedure has been demonstrated to permit a much more accurate quantification than more conventional approaches such as real-time PCR, but, more importantly, it permits the detection of very small changes in input DNA. Thus, ‘digital PCR’ could perhaps indicate whether a fetus was affected by DS, simply by counting the number of chromosome-21-specific target sequences in comparison to a similar locus on an unaffected chromosome (Refs Reference Zimmermann46, Reference Fan and Quake47, Reference Fan48, Reference Lo49) (Fig. 1).

A strategy in which the amyloid gene locus on chromosome 21 and the GAPDH (glyceraldehyde 3-phosphate dehydrogenase) gene locus on chromosome 12 were co-examined by digital PCR was shown to be able to detect DS when examining pure fetal genetic material obtained by invasive means (Ref. Reference Fan and Quake47) (Fig. 1). The report indicated that the method might also be suitable for the analysis of cf-DNA, in that ‘DS cases’ could be detected using artificial mixtures involving only 10% DS material, which is very similar to the concentration of fetal cf-DNA in maternal plasma. A crucial limitation of these observations, however, was that they would hold true only when 10 000 or more individual PCR reactions were monitored.

Next-generation sequencing and the possible advent of noninvasive DS detection

Although reports from two research groups indicated that digital PCR might offer success for the noninvasive detection of DS by the analysis of cf-DNA (Refs Reference Fan and Quake47, Reference Fan48, Reference Lo49), these studies also made it clear that these analyses would be conducted at the limits of current digital PCR platforms, which at the time offered some 12 000 individual reaction events. Hence, if this strategy were to be pursued, then its successful transition would require an even greater level of ‘individual event analysis’ to permit the necessary degree of discrimination between normal and DS cases.

This was offered by the advent of ‘next-generation sequencing (NGS)’, also termed ‘deep sequencing’, whereby the entire genomic template is fragmented, sequenced in short reads and then reassembled through complex bioinformatic comparison with a genomic database (Ref. Reference Ansorge50). In this manner, subtle mutations having key roles in tumour initiation and progression could be ascertained (Refs Reference Jones51, Reference Leary52).

Apart from offering unprecedented detail with regard to genomic alterations, the NGS approach also offered a unique opportunity to overcome the current limitation of digital PCR, by providing information concerning tens of thousands of sequence reads per chromosome; more than 60 000 reads might be recorded for chromosome 21, and many millions over the entire genome.

In the case of a DS fetus, there would be a small increase in the number of chromosome 21 reads, whereas there should not be any comparable alteration across the other chomosomes. As such, in a manner akin to digital PCR, by simply counting the number of chromosome 21 reads, and then comparing these to a much larger number of sequence reads over the entire genome, it should be possible to determine the ploidy of chromosome 21 (Fig. 2). This indeed was the case, and it was clearly demonstrated that even minute quantitative alterations in the number of chromosome 21 reads could lead to an unparalleled discrimination between pregnancies bearing a fetus with trisomy 21 or those with a normal karyotype (Refs Reference Chiu53, Reference Fan55). Furthermore, because this analysis was not restricted to chromosome 21, it was also possible to detect two cases with ES (trisomy 18) and one case with Patau syndrome (trisomy 13), thereby indicating the possible widespread applicability of this technology (Ref. Reference Fan55).

Figure 2. Detection of fetal aneuploidy using next-generation sequencing. In this procedure the cell-free DNA fragments in maternal plasma are isolated, and a library with special sequence tags is then made. These tags permit subsequent multiplex analysis. The library is examined by next-generation sequencing, which determines the sequence of each and every fragment. By bioinformatic analysis these sequences are ascribed to chromosomal locations. Following this, the number of sequence reads for each chromosome is counted. For chromosome 21 this is typically of the order of several thousand reads, which can then be compared with several million reads spread across the genome. If the fetus is affected by Down syndrome, then slightly more reads will be recorded for chromosome 21 compared with those from a euploid fetus. By comparing these data with a bank of reference samples, and by the use of predetermined cut-off values (Z score), the ploidy of the sample being examined can be determined (described in more detail in Refs Reference Chiu53, Reference Chiu, Cantor and Lo54).

The downside of this approach is the prohibitively high cost per sample and the length of time taken for sample preparation, sequencing and subsequent bioinformatic analysis, which occupies the better part of several days per sample (Ref. Reference Lo and Chiu56).

Can NGS be simplified?

The above results have indicated that the use of complex cutting-edge genomic tools such as NGS can finally offer the long-sought dream of noninvasive detection of fetal aneuploidy (Refs Reference Hahn, Jackson and Zimmermann1, Reference Lo and Chiu56). However, this comes at a very high cost and unacceptably long analysis. A possible strategy to overcome these disadvantages is the targeting of only those chromosomes or regions of interest such as those on chromosome 21 for the detection of DS cases. Although several different strategies exist that permit some form of target enrichment prior to the subsequent sequencing step, they have to date been limited to examination of intact genomic DNA samples, and not fragmented DNA species such as those found in cf-DNA (Ref. Reference Liao57). Furthermore, these procedures, such as on-array capture, require rather large concentrations of input template DNA (up to 7.5 µg per sample). Hence, an approach that permitted targeted sequencing of the small quantities of fragmented DNA in maternal plasma had to be sought.

Very recently, a solution hybrid selection method using ultralong oligonucleotides, commercially marketed by Agilent, USA, has been used for targeted enrichment (Ref. Reference Liao57) (Fig. 3). A capture library specific for the X chromosome was hybridised with 500 ng of an amplified plasma DNA library, and the selected capture targets were pulled down using a combination of biotinylated oligonucleotide probes and streptavidin-coated magnetic beads. The enriched target DNA was then subjected to a further 12 rounds of PCR with specially tagged primers necessary for the subsequent sequence analysis. In the examination of 12 maternal plasma samples, this targeted capture procedure led to a mean enrichment of 213-fold. This enrichment was also reflected in the increased ability to detect fetal-specific loci on the X chromosome, which changed from 3.5% in the un-enriched samples to 95.9% in the samples subjected to a targeted enrichment step.

Figure 3. Schematic representation of a targeted sequencing approach using the SureSelect^TM Target Enrichment System. In this procedure the cell-free DNA fragments are isolated and a library is generated as per the standard next-generation sequencing protocol. Prior to sequencing, however, this library is hybridised to the SureSelect^TM Oligo Capture Library, which is manufactured in such a manner that it will recognise a specific chromosome, such as chromosome 21. These oligo sequences contain magnetic particles to permit their retrieval in a magnetic field. Hence, following hybridisation (65°C, 24 h), captured sequences are selected by magnetic selection, and unselected sequences are washed away. The bound fragments are then purified and prepared for sequencing and examined as described for Figure 2. This procedure was recently shown to permit a 213-fold enrichment of the targeted X chromosome (Ref. Reference Liao57).

A concern with any enrichment approach is that it might lead to bias, by preferential enhancement of select sequences, including the preferential accumulation of maternal cf-DNA sequences over fetal ones. However, in this study the proportion of fetal to maternal sequences was similar in un-enriched loci (~16–30% from first to third trimester) and in enriched X-chromosome regions (~15–32%) (Ref. Reference Liao57). These results are encouraging, because they could facilitate targeted enrichment of more crucial fetal loci, such as those on chromosome 21, thereby paving the way for a less complex noninvasive test for DS.

The question of intellectual property and how to optimise procedures

An important concern that needs to be addressed is how intellectual property and commercialisation will affect future research and applications (Ref. Reference Szczepura, Osipenko and Freeman58). For the best possible service to reach the patient, clarity is essential; otherwise, conflicting reports and views might prevent superior products from being developed. In the case of fetal cf-DNA, such an issue was raised by the commercialisation of a noninvasive test for fetal RhD determination in the EU, and subsequently in the USA (Ref. Reference Szczepura, Osipenko and Freeman58). Although it was clear to leading researchers in the field that the tests initially marketed might be flawed or were not state of the art, it was felt that it would be a weary process to convince the commercial parties involved to change their standard of practice and to adopt more modern effective approaches. This issue was further highlighted by the near-simultaneous reports concerning the application of digital PCR or NGS, leading to conflicting reports in the lay media as to who the main patent claimants were (Refs Reference Hahn, Jackson and Zimmermann1, Reference Hahn59). It is currently also unclear what business model or strategies should be used in translating this research from the bench to the clinic. Hence, it might be worthwhile to echo previous concerns that such uncertainty might hinder or restrict further research in the field (Ref. Reference Daniels, Finning and Martin60).

Are fetal cells in maternal blood still worth pursuing?

Despite the enormous strides that have been made with the determination of fetal aneuploidies through the use of cf-DNA, the latest reports have indicated that these will be very labour intensive, and require extremely high-tech devices and bioinformatic services. Hence it is unclear how quickly these developments will be translatable into the clinic.

For this reason there has been a resurgent interest in the isolation of fetal cells from maternal blood. This have been fostered by the development of (1) sophisticated high-throughput automated scanning devices, which permit the rapid identification of putative target fetal cells among a large pool of maternal cells (Refs Reference Ntouroupi61, Reference Seppo62), and (2) efficient microfluidic systems and electronic micromanipulation systems, which permit the effective retrieval of individual trafficking fetal cells (Refs Reference Hahn59, Reference Lee63, Reference Huang64).

Even though this process will most likely also be laborious, it has several advantages over cf-DNA analysis, in that fetal cells offer a pure source of fetal genomic material. Furthermore, widely standardised procedures commonly used in many diagnostic laboratories, such as FISH (fluorescence in situ hybridisation), can be applied for the very rapid detection of chromosomal anomalies (Ref. Reference Evans65). In addition, use can now be made of a wide body of experience in the analysis of single cells, a routine practice in many in vitro fertilisation clinics offering preimplantation genetic diagnosis (PGD) (Ref. Reference Wells66).

A possible further advantage is that the intellectual property situation in this area is not as complex as that of cf-DNA, in that many of the original patents are close to expiry, and no commercial entity holds a wide portfolio of key elements. As such, it will be of interest to monitor progress in this sphere of research, because it might yet have a few surprises in store.

The challenge of proteomics

The development of the first-trimester DS screening test using a combination of ultrasound measurement of nuchal translucency, serum analyte analysis and maternal age has led to a dramatic increase in the ability to detect pregnancies at-risk of bearing an aneuploid fetus (85% sensitivity) when compared with the previous generation of second-trimester screening tests (65% sensitivity), which relied solely on serum analyte analysis in combination with maternal age (Ref. Reference Avgidou67). This facet is reflected in the report on the English and Welsh National Down Syndrome Cytogenetic Register analysis, which indicated that the vast majority of DS cases were detected by current screening practice (Ref. Reference Morris and Alberman5).

The downside of this startling development is that the first-trimester screen is still hampered by a rather high false-positive rate of 5–8% (Refs Reference Falcon68, Reference Cowans69). This implies that a large number of healthy pregnancies are being subjected to unnecessary invasive procedures – a healthcare risk for mother and unborn child, as well as a considerable financial burden for healthcare services. Furthermore, almost 15% of cases remain undetected and, because these occur in the group of pregnant women not judged to be at a higher risk (<35 years old), could result in undesired live births.

Therefore, increasing the efficiency of this test has been proposed, perhaps by the addition of further biochemical analytes, such as members of the activin family (Refs Reference Spencer70, Reference Spencer71). The inclusion of further highly specific blood-based analytes might be possible as a result of the development of quantitative proteomic technologies as part of the Human Proteome Association, especially the affiliated plasma proteome project (Ref. Reference Schmidt and Aebersold72). The placenta in DS cases exhibits structural alterations, which are most likely the result of alterations in protein expression, and so it is possible that these changes will be reflected in the maternal plasma proteome (Ref. Reference Choolani73). In this regard, the use of quantitative isobaric labelling has been shown to permit the detection of significant alterations in the levels of several placenta-derived peptides in DS cases when compared with controls (Ref. Reference Kolla74). Analogous observations have been made by other research groups using a variety of different proteomic approaches (Refs Reference Kolialexi75, Reference Tsangaris76, Reference Lopez77, Reference Nagalla78). It now remains to explore how useful these potential biomarkers will be in increasing the detection rate of the current first-trimester screen.

Summary and conclusion

Tremendous strides have been made with the analysis of fetal cf-DNA in maternal blood, in that we now stand on the threshold of a new generation of noninvasive diagnostic tools, which might even permit a detailed analysis of the entire fetal genome or a full karyotype. However, it remains to be seen how rapidly these can be translated into the clinic, in a manner that is cost effective and accessible to all. In this context it will be interesting to see which system, namely NGS, digital PCR or enrichment of methylated fetal cf-DNA fragments, passes the final hurdle to enter clinical service. Although the isolation and analysis of trafficking fetal cells might remain a peripheral test, it could be useful in those centres already offering FISH or PGD services. Should multiparameter mass spectroscopy and quantitative proteomics finally come of age, it might yield a set of biomarkers with unparalleled specificity, not only for DS screening but also for other pregnancy-related disorders such as preeclampsia.