Evaluation of non-invasive prenatal RHD genotyping of the fetus

Venous blood samples were prospectively collected from 140 RhD-negative women presenting for routine obstetric care at two tertiary maternity hospitals in Sydney and Brisbane between November 2006 and April 2008. There were 26 samples taken at 12–16 weeks’ gestation, 61 samples at 17–27 weeks’ gestation, and 53 samples at ≥ 28 weeks’ gestation. Nine women were isoimmunised to RhD.

The study was approved by ethics committees of Mater Health Services, Royal North Shore Hospital and the Australian Red Cross Blood Service (ARCBS). All women gave informed consent.

Blood samples (18 mL) were collected in tubes containing EDTA anticoagulant and transported to the tertiary ARCBS laboratory within the timeframes used for screening blood donors for labile viral RNA markers.17 Plasma was separated from the cellular component using a two-step centrifugation protocol and stored at − 70°C. Plasma was frozen within 24 hours of blood sampling, except for three samples that were frozen within 46, 100 and 170 hours.

During pregnancy, women who are RhD-negative are identified by routine serological screening. From a clinical perspective, women who have “partial” RhD antigen expression, lacking some but not all of the D epitopes, should be managed according to protocols devised for RhD-negative women. Many hospital laboratories therefore use screening assays that have a lower sensitivity and specificity and do not distinguish between RhD-negative women and those who have an RhD variant. As a common standard, we ensured that all maternal RhD blood group testing was performed according to blood donor screening protocols that are optimised to detect partial and weak D.18 Although not present in this cohort, this included testing for the extreme form of the weak D antigen, DEL, by red cell antibody absorption and elution tests on samples with an RhCe haplotype.

We recognised that the presence of a parental RHD variant may cause errors in fetal genotyping.19-21 To assess the impact of these variants, the study protocol included a procedure for further analysis of samples that gave discrepant or indeterminate results. To allow later comparison with maternal genotype, maternal genomic DNA (mgDNA) was extracted and stored from the white cell “buffy” coat at the time of DNA extraction from maternal plasma.

Cell-free DNA, including both free maternal DNA and ffDNA, was extracted from maternal plasma, with each sample extracted in duplicate and the resultant two eluates pooled for same-day fetal RHD genotyping.

Initially, two methods using different QIAamp kits (QIAGEN, Melbourne, Vic) were evaluated with nine samples. The first method, using a QIAamp DNA Blood Mini Kit, isolated DNA from 800 μL of plasma, eluting DNA in a final volume of 55 μL.10 The second, using a QIAamp MinElute Virus Spin Kit, involved a protocol recommended by Sequenom Inc (San Diego, Calif, USA), isolating DNA from 1000 μL plasma and eluting DNA in 55 μL.

Comparison of the two DNA isolation methods showed lower cycle threshold (Ct) values for fetal markers (RHD exons 4, 5 and 10, and SRY) using the QIAamp MinElute Virus Spin Kit (P < 0.001, χ2 goodness-of-fit test). This result is consistent with yielding higher levels of ffDNA, and this method was adopted for the remainder of the study.

The presence of fetal RHD sequences was determined with two separate duplex RT-PCR assays, with each test performed in quadruplicate.10,11 One duplex amplified RHD exon 4 and 10 sequences, and the other amplified RHD exon 5 and the Y chromosome-located SRY gene.10,11 RT-PCR was performed using the Rotor-Gene 3000 (Corbett Research, Sydney, NSW), and results were analysed with quantification software.22

The primer and probe designs for exons 4 and 5 do not permit amplification of the non-functional RHD pseudogene associated with RhD-negative black Africans. Such samples will therefore type as exon 4- and 5-negative and exon 10-positive. In addition, other rearranged RHD-CE-D genes can generate discordance between RT-PCR exon typing and either be non-functional or generate partial D antigens. For example, the gene responsible for the clinically significant RhDVI antigen will also type as exon 4- and 5-negative and exon 10-positive, and be indistinguishable from the pseudogene on the primary RT-PCR test. A follow-up supplemental test algorithm, including RT-PCR using pseudogene primers and probes specific for exons 4 and 5, and also RT-PCR for exon 7, was performed to further resolve such samples. A pseudogene DNA control was provided by Dr Kirstin Finning (International Blood Group Reference Laboratory, Bristol, UK).

The quality of the primers, probes and RT-PCR reagents was assessed using a defined quantity of purified human male genomic DNA (Promega, Madison, Wis, USA) in a twofold dilution series from 0.02 ng/μL down to 0.001 ng/μL. The RT-PCR duplex assays detected down to 1.5 copies of target genomic RHD and SRY sequence, as reported elsewhere.9 For fetal RHD genotyping, three water (no template), four RhD-negative and three RhD-positive controls were included in each PCR run to monitor for contamination. For each sample, total (maternal and fetal) cell-free DNA yield was quantified by amplifying the chemokine receptor gene (CCR5).10,11 Inclusion of the CCR5 assay helps identify samples with an excess of maternal DNA (eg, from maternal white cell lysis) where there may be interference with detection of ffDNA sequences. In addition, comparison of RHD exon 10 RT-PCR Ct values compared with the CCR5 Ct may signify the presence of a non-functional maternal RHD gene and trigger further mgDNA studies.10,11

The SRY gene served as an internal control marker to confirm the presence of fetal DNA. To guard against false-negative RHD results, in samples where neither RHD nor SRY sequences were detected, further analysis to show that fetal DNA was present was performed by testing for the presence of a hypermethylated RASSF1A gene. The promoter of the RASSF1A gene, a tumour suppressor gene, is hypermethylated in DNA derived from the placenta and hypomethylated in DNA derived from the mother.16 The test involves digestion of 17.5 μL of plasma-derived DNA with 100 U of BstUI, a methylation-sensitive restriction enzyme, at 60°C for 16 hours. This effectively removes the maternal RASSF1A gene, and the fetal gene can then be detected by RT-PCR. Amplification was performed in tandem with beta-actin, which served as an internal control to test for the completeness of the digestion.16 Tests were performed in triplicate, with each run also including DNA extracted from whole blood derived from an RhD-negative non-pregnant woman as a negative control to further show that DNA digestion was complete.

Criteria established by Finning and colleagues were used to interpret results.10,11 A fetus was assessed as being RHD-positive if two of the four replicates for each of the 4, 5 and 10 RHD exons plus an additional three replicates from any of the RHD exons gave Ct values < 42. A fetus was assessed as being RHD-negative if at least 11 of 12 RHD replicates gave no Ct values (ie, no amplification) and there was a positive signal for the SRY (≥ 2/4 replicates) or the RASSF1A (≥ 2/3 replicates) genes. Samples not classified as either positive or negative were described as indeterminate, and further testing was performed according to the supplemental algorithm. The accuracy of predicted fetal RHD status was assessed by comparison with the infants’ RhD serotype determined from cord blood after delivery.

Statistical analysis was performed using GraphPad Prism, version 4.03 (GraphPad Software, San Diego, Calif, USA).

Extended donor testing of maternal serotype showed one of the 140 apparently RhD-negative mothers to be partial RhD-positive, carrying an RhDVI antigen. This mother, who was white, had an RhD-negative C+c+e+ phenotype. Genotyping confirmed that she was RHD exon 4- and 5-negative and exon 7- and 10-positive, consistent with an RhDVI variant.7,20,21

In the first round of analysis, 135 of the 140 RhD-negative women had informative results predicting fetal RHD status and, of these, there was 100% concordance with the infant RhD serotype. Ninety-five fetuses were predicted to be RHD-positive, and an example of the molecular analysis for such a case is shown in Box 1.

Fetal RHD status could not initially be determined for five of the 140 women. These indeterminate results triggered the supplemental test algorithm, which produced further information about D gene variants in two cases, but failed to resolve the issues of low replicate amplification in the other three. Two of these three samples, collected at 12 and 15 weeks’ gestation, gave positive amplification signals in 7/12 RHD exon replicates, rather than the 9/12 required to state that the fetus was likely to be RHD-positive. Both of these neonates did, however, prove to be RhD-positive in subsequent serotyping. The third indeterminate sample consistently showed low-level signal detection (4/12 across all exons), and the baby’s serotype was found to be RhD-negative.

For the two samples revealing D gene variants, initial testing had shown no amplification of exons 4 and 5, but exon 10 gave a positive signal in all four replicates (Box 2). In one sample, supplemental testing amplified exons 4 and 5 using the pseudogene-specific primers (data not shown). Further assessment of mgDNA showed that the mother was RHD-negative for all three exons. These findings were indicative of paternal inheritance of the pseudogene, consistent with the ethnic origin of the father, and the neonate serotyped RhD-negative as anticipated.

In contrast, supplemental testing of the second variant sample showed no evidence of a pseudogene, and there was no amplification signal for exon 7. Given that the mother serotyped as RhD-negative, and that analysis of mgDNA also showed an exon 4-, 5- and 7-negative but exon 10-positive result, we concluded that this was not a functional RHD gene and the neonate would serotype as RhD-negative, which was confirmed after delivery.

A third RHD variant detected in the study cohort was seen in the mother reclassified as having an RhDVI partial antigen variant. In this case, the fetus had 4/4 positive signals for RHD exons 4, 5 and 10, and 4/4 positive signals for SRY. As exons 4 and 5 were not detected in the mother, we predicted that the fetus was an RHD-positive male, and this was confirmed after birth (Box 2).

In the 40 women predicted to have an RHD-negative fetus, the presence of fetal DNA was verified by detection of SRY signals for 24, which correlated correctly with delivery of male infants. The presence of fetal DNA for the remaining 16 women with negative SRY signals was verified by detection of the fetal-associated hypermethylated RASSF1A gene (Box 3), and female sex was correctly predicted in all 16 cases.

Using SRY as an internal control to confirm the presence of fetal DNA, post-delivery correlation of predicted fetal sex was correct in 137/140 cases, including all of the 40 predicted RHD-negative cases. The three discrepant cases included one where no assignation could be made for either RHD or SRY. In another case, plasma preparation was delayed until 170 hours after collection and, although an SRY signal was not detected, a male infant was born. Significantly higher total (maternal and fetal) free DNA levels were noted in this sample (2200 ng/mL; mean, 288 ng/mL). The final sample was genotyped as an RHD-positive male, whereas the infant was an RhD-positive female. We were not able to explain this discrepancy.

RASSF1A enzyme digestion and amplification, required as an internal control in the 16 RHD- and SRY-negative samples, was performed on a total of 40 samples (20 RHD-positive and 20 RHD-negative) to further evaluate its robustness. There was good correlation of the Ct values for the RASSF1A and RHD assays for the 20 RHD-positive samples (R2 = 0.9), suggesting both sequences were of fetal origin.16 Dilution studies comparing the sensitivity of the RASSF1A and RHD assays (to determine the risk of a false-negative result) have shown that amplification fails with the RASSF1A probe at the same dilution as the RHD probes, which is important if it is to be relied on as an internal control for non-invasive RHD genotyping.