Funding sources: Some of the work described in this manuscript presents independent research funded by the National Institute for Health Research (NIHR) under the Programme Grants for Applied Research programme (the ‘RAPID’ project) (RP-PG-0707-10107) and the Central and East London NIHR Comprehensive Local Research Network. Professor Lyn S. Chitty is partially funded by the Great Ormond Street Hospital Children's Charity and the NIHR Biomedical Research Centre at Great Ormond Street Hospital. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.
To date, the main clinical application of non-invasive prenatal diagnosis (NIPD) for single-gene disorders has been for severe X-linked conditions affecting male fetuses, for example Duchenne muscular dystrophy, where it has been shown to reduce the need for invasive testing by identifying male bearing pregnancies.[1, 2] In addition, fetal sex determination may be used to identify female fetuses in pregnancies at risk of congenital adrenal hyperplasia, as genital virilisation in affected female fetuses can be significantly reduced by the antenatal administration of dexamethasone therapy. On the basis of detecting the presence or absence of Y-chromosome sequences in maternal plasma, fetal sex determination does not present a significant technical challenge. These tests are widely practised, clinically cost-effective and over the last few years have gradually been implemented into clinical genetic practice across Europe (Figure 1).[4, 5]
This has been followed by the development of NIPD of autosomal dominant disorders where the mutation is carried on the paternal allele, some de novo mutations and recessive conditions where the parents carry different mutations, and diagnosis is based on detection or exclusion of the paternal allele. The principal approach in these examples is generally polymerase chain reaction (PCR)-based, using a relatively straightforward procedure to determine the presence or absence of the mutant allele in the maternal plasma (see Supplementary Table 1). More recently, attention has focussed on recessive disorders, particularly the haemoglobinopathies (sickle-cell anaemia and thalassemia)[7, 8] and definitive diagnosis of X-linked conditions such as haemophilia where the mother carries the mutant allele and where there is already an existing and significant demand for conventional, invasive prenatal diagnosis. The diagnosis in these situations presents a number of technical challenges, as there is a high circulating load of mutant allele from the carrier mother. Nonetheless, a number of studies based on digital PCR and next-generation DNA sequencing technologies have now been published that can discriminate affected from unaffected fetuses[7-9] (see Supplementary Table 1).
In the UK, as in other countries, most prenatal diagnostic tests are requested because of risk of aneuploidy, and in 2010–2011, only 2600 of 16 500 were performed because of a risk of a genetic condition. The indication for genetic testing can be divided into two main groups with the majority being carried out in pregnancies where the mutation is established, either because of a known family history or following carrier screening in early pregnancy. The other group comprise pregnancies where fetal ultrasound findings suggest an underlying genetic aetiology. For the latter, molecular testing is often required to confirm the diagnosis. Until recently, definitive molecular diagnosis within a timescale, which could influence pregnancy management, was not often achievable because any one of a number of different mutations, often in several different genes, was potentially responsible for the sonographic findings. For example, although the identification of short long bones in the third trimester may be indicative of a skeletal dysplasia, the underlying diagnosis is not always clear. In this situation, prenatal testing for achondroplasia has been possible via invasive testing for many years and was one of the first tests developed using NIPD (see Supplementary Table 1), as this condition is caused by a single mutation in FGFR3 in around 98% of cases. However, if NIPD is negative, there remains a range of other conditions with similar features that cannot easily be excluded, for example hypochondroplasia, acromesomelic dysplasia, etc, without further time-consuming and costly testing. There are other situations involving many more and often very large genes, for example, the fetus presenting with large echogenic kidneys where aetiology includes autosomal dominant and recessive polycystic kidney disease, Beckwith–Wiedemann and Laurence–Moon–Barbet–Biedl to name but a few. In these situations, knowledge of the underlying diagnosis would be invaluable in prenatal counselling, but definitive molecular diagnoses in pregnancy in the absence of a positive family history remain a challenge.
There are a number of fundamental technical challenges that need to be addressed before NIPD can be applied more widely to genetic disorders. These predominantly relate both to the size of cell-free fetal DNA (cffDNA) fragments, which are known to be shorter, on average, than maternal cell-free DNA (cfDNA) fragments, and the relative abundance of maternal cfDNA. Interrogation of fetal DNA sequences would be greatly simplified if there was a reliable, cost-effective method for selectively enriching and isolating short fragment cffDNA from maternal plasma. Although some progress has been made with the development of polymeric microsystems that combine electrokinetic trapping, isotachophoresis and capillary electrophoresis, these technologies still remain relatively far-removed from routine implementation within a diagnostic service laboratory.
Size of cffDNA
The size of circulating DNA has been studied by next-generation sequencing (NGS), and fragments have been shown to be around 160 bp in size with a small percentage ranging to 340 bp. Y-chromosome sequences were generally <150 bp in length, suggesting that fetal-free DNA is shorter than maternal-free DNA. This means that there are size constraints for the design of NIPD assays to detect a given fetal DNA target. When analysing a single-point mutation, the constraint of the small fetal fragments can be accommodated easily by designing small amplicons, for example <150 bp, to ensure efficient amplification of target mutation. However, for some diseases, it may not be possible to use NIPD to directly analyse the disease-causing mutation, as in the case of fragile X syndrome where the disease can involve large (CGG)n trinucleotide expansions of >1 kb upstream of the FMR1 gene. NIPD for Huntington disease (HD) expansions have been reported; however, reliable detection will be limited inevitably by the size of the polymorphic (CAG)n trinucleotide repeat in exon 1 of the HTT gene.[28, 29, 31] Triplet repeats ≥40 will always give rise to HD, and whereas the lower expansion sizes (40–60) have potential to be detected, larger expansions may be undetectable because of the small size of the cffDNA fragments. In cases such as HD, it may be possible to overcome this problem by using alternative approaches such as extensive haplotype analysis (using small amplicons) to map the inheritance of the affected chromosome.
Presence and quantification of cffDNA
A critical factor for any non-invasive prenatal test is ensuring that cffDNA is present in the sample being tested. Currently, clinical applications of NIPD require the detection of the paternally inherited (or de novo) allele to make a definitive diagnosis. Failure to detect the paternal target sequence or de novo mutation is either indicative of a true-negative result or could be due to the lack of amplification of the sequence due to low concentrations of cffDNA or the complete absence of cffDNA in the sample. Amplification of a fetal-specific marker that confirms the presence of cffDNA allows a negative result to be more accurately interpreted as either a true-negative or false-negative result. The use of several fetal markers has been reported including Y-chromosome sequences for male pregnancies, panels of common polymorphic short tandem repeats, single-nucleotide polymorphism (SNPs) or insertion/deletion markers,[51, 52] and epigenetic markers such as hypermethylated RASSF1A promoter.[53, 54]
Y-chromosome sequences (e.g. DYS14 and SRY) can be used to confirm the presence of cffDNA but only in male pregnancies. DYS14 is a multicopy sequence present in the TSPY1 gene and has been used as a target for fetal sex determination, fetal marker and for quantification of cffDNA. The multicopy nature of the target does mean that detection is improved at low fetal DNA concentrations, but some studies have shown that DYS14 can be amplified at very low levels in female samples. Therefore, it has been suggested that it should not be used as a sole marker for fetal sex determination.[51, 54] The multicopy nature of DYS14 also means that the sequence is not optimal for fetal DNA quantification. SRY has been used successfully to confirm the presence of fetal DNA and for fetal sex determination. An alternative approach, also applicable to female pregnancies, is to analyse panels of SNPs or insertion/deletions for paternally inherited sequences. Scheffer et al. were able to confirm the presence of fetal DNA in 87% of samples tested by using 24 biallelic insertion/deletion polymorphisms or paternally inherited blood group antigens. In 2011, Tynan et al. used a universal multiplexed SNP genotyping method. Restriction digestion of one allele of 92 SNPs allowed the detection of at least four paternal alleles with 98% sensitivity and 96% specificity. However, this approach is time-consuming and potentially costly, as it requires large panels of markers to ensure amplification of at least one unique paternal allele. Ideally, a universal fetal marker independent of paternally inherited sequences should be used. RASSF1A has been shown to be hypermethylated in the placenta and hypomethylated in maternal blood. Restriction enzyme digestion of the hypomethylated maternal signal allows the fetal-specific hypermethylated RASSF1A targets to be amplified using simple real-time PCR assays. RASSF1A has been used successfully as a fetal marker in several studies.[54, 57-59]
With the development of highly quantitative single-molecule counting techniques such as digital PCR and NGS, it is now possible to use these technologies to assess the under-representation or over-representation of fetal alleles (relative mutation dosage) in cfDNA samples for maternally inherited mutations and recessive disorders. For these tests, it is not only important to confirm the presence of cffDNA in the sample but it is also critical to accurately quantify the percentage of cffDNA present in the sample. Determination of the under-representation or over-representation of the mutated fetal allele using a statistical calculation (sequential probability ratio testing) relies heavily on the accurate quantification of total cffDNA (fetal load) in the sample. Any variation in the estimated fetal load has the potential to result in both false-positive and false-negative results. Techniques that have been used to quantify the proportion of cffDNA in samples include methylation-based DNA discrimination using matrix-assisted laser desorption/ionisation-time of flight mass spectrometry (MALDI-TOF MS) analysis, digital PCR[7, 8] and NGS analysis of polymorphic sequences. The MALDI-TOF MS and NGS approaches are more amenable to high-throughput testing and have been used extensively in non-invasive prenatal testing for aneuploidy where a cffDNA fraction of >4% has been suggested as an acceptable proportion of cffDNA in a sample to issue a reliable result. For low-throughput testing of single-gene disorders, digital PCR or an NGS approach that analyses a highly polymorphic locus may be a more appropriate strategy for determining fetal load.
Sample collection and time to processing
Several factors relating to blood sample collection have been shown to affect the proportion of cffDNA present in a sample. Studies using blood collected from pregnant women have shown that there is an increase in total cfDNA over time (due to lysis of maternal cells) but that the absolute quantity of cffDNA remained constant. Therefore, the proportion of cffDNA reduces over time making the time to processing of samples a significant factor when carrying out NIPD. Some studies have used blood collection tubes that contain cell-stabilising solutions (e.g. cfDNA BCT tubes (StreckTM)) to preserve the original proportion of cffDNA when the time to processing of the samples is greater than 8 h,[64, 65] whereas others have shown that the cffDNA fraction is stable for up to 5 days and suitable for use in fetal sexing and RHD assays. Where an absolute quantification of cffDNA is needed, the use of cell-stabilising agents may be beneficial, but for less complex testing that requires confirmation of the presence or absence of an allele not present in the mother, provided the assay includes a control for the presence of fetal DNA, collection into K3EDTA and timely processing may be sufficient.
Other issues relate to the presence of multiple pregnancies. NIPD can be used in multiple pregnancies, but unless there are sonographic signs to indicate which fetus is affected, invasive testing will be required to determine whether one or more fetuses are affected in the presence of a positive NIPD result. Of particular significance when performing NIPD in early pregnancy is the potential for an ‘empty sac’ to give rise to erroneous results, as the placenta continues to shed fetal DNA into the maternal circulation after demise of the fetal pole. This emphasises the need for careful ultrasound scanning before NIPD, not just to confirm the gestational age but also to look for evidence of multiple pregnancies.
METHODS USED FOR NIPD FOR SINGLE-GENE DISORDERS
A multitude of methods have been utilised for NIPD of genetic disorders (Supplementary Table 1). For the detection of alleles that are not present in the mother, the majority are based on the detection of a single-base substitution, insertion or deletion and can be either probe-mediated or primer-mediated, using PCR to produce a specific amplicon. The discrimination of an amplicon containing the mutant allele versus an amplicon containing the wild-type allele can be based on fragment length(s), fluorescence detection (via the incorporation of different florescent labels) or by direct DNA sequencing of the PCR amplicon (Table 1). Translation into routine clinical practice for many of these approaches is not easy; for example, qualitative assays based on agarose gel electrophoresis lack sensitivity, are prone to contamination and, in cases where enzyme digest is also used, are reliant on the restriction efficiency remaining constant, and furthermore, analysis is subjective with results being recorded ‘by eye’. New quantitative technologies with a higher sensitivity and throughput are emerging, bringing NIPD closer to clinical practice.
Table 1. Methods used for non-invasive prenatal diagnosis of single gene disorders
PCR, polymerase chain reaction.
PCR (end point)
Size fractionation of amplicons
PCR (end point)
Restriction enzyme digestion
Quantitative fluorescent PCR (end point)
Quantitative real-time PCR
Taqman probes; MGB probes
Primer-mediated or probe-mediated discrimination of alleles
Peptide nucleic acid clamp PCR
Primer/primer competitive hybridisation
Differential melt characteristics of heteroduplexes and homoduplexes
Denaturing high performance liquid chromatography
Differential melt characteristics of heteroduplexes and homoduplexes
Sequence discrimination based on mass
Arrayed primer (single base) extension
Single-molecule counting combined with probe based assays
Digital PCR allows for a more sensitive approach to the detection of low abundance sequences by the discrete counting of mutant and wild-type alleles present in a sample. Sensitivity is only limited by the number of molecules that can be analysed and the false-positive rate of the mutation detection assay. Digital PCR relies on the ability to perform each reaction with an average of one template molecule. Digital PCR can be performed by the spatial separation of PCR reactions in capillary systems. The first commercial platform available for digital PCR was the Fluidigm microchip that utilises nanolitre compartments defined by pneumatic valves enabling the simultaneous analysis of up to 39 960 reactions per run with the 48.770 Digital Array integrated fluidic circuit.
The use of this platform has been reported for the NIPD of thalassaemia, sickle-cell anaemia and haemophilia (Supplementary Table 1), using relative mutation dosage. This approach requires accurate determination of the proportion of cffDNA in order to calculate the ratio of mutant to wild-type allele present in maternal plasma and thereby predict genetic status of the fetus. Whereas this is readily achieved in X-linked conditions, such as haemophilia, using Y-chromosome sequences, its application clinically in female fetuses is limited by the requirement for other of highly polymorphic markers.
Droplet digital PCR
The QX100TM droplet digital PCR (ddPCR) system (Bio-Rad, USA) is able to generate 20 000 individual PCRs in a single experiment using a conventional TaqMan® quantitative PCR (qPCR) assay. Using a duplex TaqMan® assay on this platform, the BRAF V600E mutation was detected down to 0.001% mutant in a wild-type background. The same group used ddPCR with an assay for the detection of hypermethylated RASSF1A in cffDNA to quantify fetal load and have demonstrated the utility of this technology in a variety of clinical samples with low levels of target sequence.
The RainDropTM ddPCR system (Raindance Technology, USA) is able to perform digital PCR in millions of picolitre droplets. Genomic DNA is compartmentalised in droplets at a concentration of less than one genome equivalent per droplet together with two TaqMan® probes, one specific for the mutant allele and one specific for the wild-type allele. After PCR, the ratio of mutant to wild type is determined by counting the ratio of green to red droplets. This technology is able to detect one mutant molecule in 250 000 and can multiplex up to ten tests on the same sample. Pekin et al. demonstrated the utility of the RainDropTM ddPCR system for the detection of the six most common KRAS mutations, quantitating a single mutation in a background of 200 000 wild-type molecules using 1 000 000 partitioned reactions.
Digital PCR may offer a more sensitive approach to NIPD, not just for X-linked and recessive conditions where the mother carries the mutant allele but also for the detection of paternally inherited alleles or those arising de novo. In our UK National Health Service diagnostic laboratory, we have experience in several conditions including NIPD for achondroplasia and thanatophoric dysplasia. This is now the standard of care in pregnancies at risk of these conditions in the UK. Our initial approach was to use standard PCR-based approaches,[13, 38, 39, 43] but in the presence of low levels of cffDNA when PCR gave inconclusive results, we have found digital PCR is more sensitive (Figure 2). This was also clearly demonstrated in the analysis of a cfDNA from a pregnancy where the fetus was found to have autosomal recessive polycystic kidney disease following the sonographic presentation at 36 weeks' gestation with anhydramnios and very large echogenic kidneys. In this case, the mother carried a p.Ala1254fs mutation (c.3761_3762delCCdupG) in PKHD1; the father has not been found to carry a mutation in this gene, but DNA analysis after birth showed that the fetus also had a c.9374C>T mutation, which is predicted in silico to affect protein function, and has most probably arisen de novo. Analysis of cfDNA extracted from frozen maternal plasma (taken at 34 weeks' gestation) using qPCR for the detection of the de novo c.9374C>T mutation in PKHD1 gave an inconclusive result, although the detection of the paternally inherited insertion/deletion polymorphism MID836a confirmed the presence of cffDNA. Subsequently, digital PCR using the same sets of hydrolysis probes and primers confirmed that neither parent carried the c.9374C>T mutation but clearly demonstrated that the fetus was a carrier of this mutation (Table 2, Figure 1). Analysis of cfDNA taken at 12 weeks' gestation in the next pregnancy did not detect the mutant allele, indicating that the fetus was unaffected (Table 2, Figure 1).
Table 2. Estimated target molecules for wild-type and mutant alleles for the Fraser syndrome and ARPKD cases using digital PCR and NGS (MiSeq, Illumina Inc)
Two samples were tested for the second Fraser syndrome pregnancy because the first sample was tested early at 9 weeks, the second at 12 weeks. The estimated number of single copies present in the sample is calculated by the software based on the number of FAM positive reaction chambers detected, using a Poisson-based correction of the data.
The potential for increased sensitivity of digital PCR over PCR is clear as it is challenging to distinguish one mutant molecule in 1000 using PCR, but for digital PCR, each molecule is partitioned, and so it should be possible to amplify mutated sequences at 0.1%. Digital PCR is also thought to be less sensitive to inhibition; although it is possible that inhibitors will delay the amplification of a target molecule over a target threshold in a qPCR reaction, this does not affect the final count for digital PCR. One of the main disadvantages of digital PCR using the Fluidigm Biomark is the cost involved in running a sample. Each 12.765 digital array contains 12 panels; taking into account the fact that a minimum of two panels are required for each sample to be tested, two panels for each maternal, paternal and normal control gDNA, and two no template control panels, ten panels are required and, realistically, one chip per test is required. Designing a set of primers and hydrolysis probes for each family-specific mutation is time-consuming and costly, but as they are exactly the same primers and probes that would have been used for PCR, there is no disadvantage in comparison with this more traditional method. Whether transferring these analyses to platforms based on ddPCR technology renders NIPD for these conditions more efficient and cost-effective requires further evaluation. However, given the improved sensitivity and ability to multiplex increased numbers of samples, these platforms do require further consideration.
Reliable, reproducible and cost-effective NGS chemistries now make it feasible to move it out of the research setting into routine diagnostic laboratories. A number of different bench-top platforms are now available that are well-suited to a clinical diagnostic laboratory service, particularly when using amplicon sequencing approaches to mutation detection, as they are cost-effective and have short run times. It is likely that they will provide a more sensitive and flexible approach to NIPD for single-gene disorders. It is possible to design panels that can screen for multiple mutations in a single assay, and samples can be multiplexed to analyse samples from different patients in a single run.
We have explored the use of NGS for NIPD in a number of conditions, including skeletal dysplasias due to FGFR3 mutations, Fraser syndrome and the case of autosomal recessive polycystic kidney disease described earlier, and have identified a number of advantages of using NGS for the detection of paternal or de novo mutations over both digital PCR or qPCR. Firstly, the number of sequence reads obtained using NGS is much higher, allowing more certainty of a positive result. Secondly, whereas for each mutation tested using digital PCR, a new pair of probes specific for the wild-type and mutant alleles is needed, using NGS, a whole gene can be sequenced using the same set of overlapping amplicons. The cost involved in designing a family-specific primer/probe set for digital PCR is considerably greater than for NGS. Furthermore, a large number of samples can be multiplexed using desktop sequencers such as the MiSeq (Illumina Inc.), reducing the cost per sample for a run. A digital PCR reaction can be turned around quickly (less than 4 h including data analysis), and a MiSeq run takes around 8 h; although both of these technologies are slower and more costly than PCR, the benefit of increased sensitivity may greatly outweigh these disadvantages.
To date, the implementation of NIPD for genetic disorders into routine clinical practice has been slow. It has been limited by the rarity of the disorders and the availability of suitable technical platforms. Whereas the former problem is only likely to be overcome with time and banking of samples and collaborative working, service laboratories increasingly have access to the platforms required to deliver NIPD for a wide range of single-gene disorders, a situation that will be welcomed by patients and health providers alike.[71, 72] However, for successful implementation, it is necessary to look beyond technology development and address issues such as costs, service regulation and stakeholder needs. Until now, research specifically addressing the ethical and psychosocial concerns and issues for service delivery for NIPD for single-gene disorders has been very limited, although some work is now emerging.[71-75] Concerns have been raised regarding the potential ease of access and hence ‘routinisation’ of testing, with health professionals and service users feeling strongly that NIPD should be offered through existing specialist services such as genetics units[71, 72, 74] to ensure appropriate pre-test and post-test counselling. The development of policy and guidelines will be critical to ensure high quality and equitable service provision. In addition, costs will have a major impact on how testing for single-gene disorders is implemented. Hall et al. have raised concerns that the costs of setting up tests for individual disorders may limit which conditions are tested for and how many laboratories are able to offer tests. To promote equity of access and standardised service delivery, it will be important to seek formal approval for new tests, explore cost-effectiveness and encourage further development.
WHAT'S ALREADY KNOWN ABOUT THIS TOPIC?
Non-invasive prenatal testing of cell-free fetal DNA in maternal plasma is widely used for aneuploidy detection, fetal RHD typing and fetal sex determination in pregnancies at high risk of sex-linked disorders.
WHAT DOES THIS STUDY ADD?
We describe the current limited use of cell-free fetal DNA analysis for prenatal diagnosis of single-gene disorders and discuss the application of new technologies to aid implementation.