This study was approved by the Shaoxing Maternal and Child Health Hospital (approval n. 277). A total of 76 RhD-negative pregnant women underwent polymerase chain reactions with sequence-specific primers (PCR-SSP), 10 RHD exons Sanger sequencing, RHD zygosity detection, and mRNA-seq between November 2018 and June 2020.

The PCR-SSP results indicated 63.2% (48/76) RHD deletion homozygous, 15.8% (12/76) RHD 1227A homozygous, 14.5% (11/76) RHD-CE(2-9)-D, 2.6% (2/76) RHD VI III, and 3.9% (3/76) genotypes could not be identified by PCR-SSP (Table 1).

The first case was RHD*10.08 partial D type (exon 3 c.340C>T) (Figure 1, Sample a). The second was RHD*01EL.02 type (exon 1 c.3G>A) (Figure 1, Sample b), and a mutation at the splicing site was found as RHD*01N.25 type (intron 2 c.336-1G>A) (Figure 1, Sample c) in the third case.

Two other RhD-negative family members (III-1, III-3) were identified in the family of the patient with RHD*01N.25 (intron 2 c.336-1G>A) (Figure 2). PCR-SSP and RHD zygosity analysis confirmed the proband's mother (II-4) and uncle-in-law (II-1) had the hybrid Rhesus box detected, with RHD+/RHD- genotype (Supplementary Figures S1–S2). In comparison, proband's father (II-3), aunt (II-2) (Supplementary Figure S3) have the same genotype RHD+/336-1G>A, which speculated the zygosity of the proband and her cousin (III-1) as RHD-/336-1G>A. RhD serological results and RHD genotype of this family are shown in Table 2.

The sequencing results (Figure 3) verified 336-1G>A mutation site and revealed the multiple splicing products of different lengths with the intron 2 residue. In addition, the number of reads in intron 2 was significantly higher than other introns (Supplementary Figures S4–S6), suggesting that intron 2 was improperly spliced after c.336-1G>A mutation.

Fetal RhD status in 28 cases of maternal homozygous RHD deletion and 3 cases of maternal RHD-CE(2-9)-D were predicted (Table 3). All predicted results were consistent with the newborn RhD status. Two cases showed RhD negative in the first samples but confirmed positive in the second samples.

RhD-negative frequencies show vast racial differences (Xu et al., 2003; Flegel et al., 2009; Polin et al., 2009; Cruz et al., 2012; Crottet et al., 2014; Perez-Alvarez et al., 2019) (Table 4). Thus, for diverse ethnic populations, it is necessary to adopt different RHD genotyping strategies to the spectrum of prevalent alleles. In our study, the main genotypes of RHD in D¯ pregnant women in Shaoxing were RHD deletion, RHD 1227A homozygous type, and RHD-CE(2-9)-D type. The profile of our survey was in concordance with that of another study in China (Zhang et al., 2019), indicating minor differences in the RHD genotypes of D¯ pregnant women across different regions in China. Using the PCR-SSP method, designed for eight common Chinese RHD genotypes: RHD-positive, RHD deletion, RHD-CE(2-9)-D, DVa (Hus), DVI III, weak D15, DEL RHD 1227A homozygous, and DEL RHD 1227A heterozygous, we found that about 96.1% of D¯ women can identify RHD genotype, and the rate rises to above 99% when adding the 10 RHD exons sequencing. Therefore, we established a simple molecular diagnostic strategy for Chinese D¯ pregnant women (Figure 4).

The occurrence of alloimmunity in D¯ pregnant women is related to several factors, such as maternal RHD genotype and RhD antigen epitope (Flegel, 2011); quantity of D-positive fetal red blood cells entering the body (Woodrow, 1971); and other factors such as ABO blood type, maternal HLA class, or fetal sex (Hadley and Soothill, 2002). However, the main factor is maternal RHD genotype and RhD antigen epitope. Therefore, our study guides RhIG use according to the different RHD genotypes in D¯ pregnant women (Table 5).

The RHD (c.1227G>A) allele, named the DEL type or Asian DEL type, is the most prevalent DEL allele in Asians. Asian DEL is always mistyped as D-negative by routine serological assays (Kim et al., 2020). Therefore, RHD genotyping is the gold standard for detecting DEL (Nuchnoi et al., 2014). The RHD (c.1227G>A) should be managed as D-positive for RhIG administration or selection of blood components for transfusion (Chun et al., 2020). In 2015, a study revealed that erythrocytes carrying the 1227A mutation might express very low “normal” D antigen levels (Fichou et al., 2015). This theory explains why “Asian DEL types” do not cause hemolytic disease in newborns and fetuses. Thus, at least 15% of D¯ pregnant women in China, typed as the RHD (c.1227G>A) allele, will benefit from RHD genotyping and be free from unnecessary RhIG use. At the same time, to avoid transfusion of DEL RBC units to D¯ recipients, RBC units carrying c.1227G>A mutation should be transferred into the D+ pool (Flegel et al., 2009).

RHD-CE(2-9)-D is a hybrid allele that replaces exons 2 to 9 by RHCE. Lacking D antigen expression, RBCs of RHD-CE(2-9)-D will not risk D¯ recipients (Flegel et al., 2009). At the same time, women with RHD-CE(2-9)-D should be managed as D¯ and be given RhIG for D immunoprophylaxis in pregnancy (Table 5).

DVI is the most common partial D that produces anti-D, and newborns born to DVI mothers with anti-D may develop hemolytic disease. DVI III is named a D-Ce(3-6)-D hybrid and DVI III erythrocytes carry relatively high RhD antigen densities. Therefore, DVI III recipients should be transfused with RhD negative blood, while DVI III donor should be managed as D⁺ (Wagner et al., 1998).

Our study characterized two point mutations in exons: c.340C>T and c.3G>A. According to the guideline (Flegel et al., 2020), both should be managed as D¯ and given RhIG for D immunoprophylaxis. More interesting, a recent study showed that weak-D and Asia-type DEL alleles’ coexistence would completely express the D-antigen (Chun et al., 2020), which means free from RhIG administration.

For our study’s 336-1G>A mutation, the mRNA products of the proband showed long intronic segment retention, which will induce DEL or D-negative phenotypes (Fichou et al., 2015). The serological study for family members reveals D¯ phenotypes in RHD-/336-1G>A heterozygote and D positive in RHD+/336-1G>A heterozygote (Table 2). However, lacking the detailed evidence for no D-antigen in red blood cells, we still suggested these individuals with RHD-/336-1G>A heterozygote should be managed as RhD positive donors in transfusion in an abundance of caution.

Thus, in our study, RHD genotyping for D¯ pregnant women eliminates at least 15% unnecessary ante- and postpartum injections of RhIG and gave a detailed guide of RhIG use for these D¯ pregnant women.

More than 300 variant alleles of RHD have been reported (Kawano et al., 1998; Ye et al., 2007; Fichou et al., 2015; Chen et al., 2016; Ogasawara et al., 2016; El Wafi et al., 2017; Chun et al., 2018; Raud et al., 2019), including single-nucleotide polymorphisms (SNPs), small or large fragment deletions, gene rearrangements, and complete RHD deletions. The intron mutation disrupts a constitutive splice site, resulting in improper retention of an intron or activation of a cryptic splice site in the vicinity of the mutant. Thus, this mutation will induce DEL or D-negative phenotypes by minute to no expression of the D antigen at the surface of RBCs. As many intronic mutated alleles of RHD were identified in Sequencing, several bioinformatics tools were used to predict the defect resulting from a genetic variation in silico. However, very few functional data are available to confirm these predictions. Furthermore, lacking fresh blood samples for RNA extraction, many functional studies mainly relied on recombinant plasmids. Hence, clinical transcriptome data and protein analysis are urgently needed to investigate genotype-phenotype mechanisms (Liu et al., 2010; Fichou et al., 2015; El Wafi et al., 2017).

We found a family with 336-1G>A intron mutation in our study. The mutation site of 336-1G>A is an intron 2 acceptor. The 336-1G>A mutation was first reported in Korea in 2005 (Kim et al., 2005) and first identified in Chinese blood donors in 2009 (Ye et al., 2009), named as the RHD*01N.25 type (IVS2-1G>A) by ISBT. In 2015, The first functional analysis by plasmid recombination experiments revealed that the cryptic splicing site of c.336-1G>A was activated, and mRNA product was either increased by 28 base pairs (bp) (from intron 2) or decreased by 21 bp (exon 3) (Fichou et al., 2015).

Our serological D-negative proband was an RHD-deletion/336-1G>A heterozygote in our study. SpliceAI showed a 94% decrease in the probability of acting as a splice acceptor in the 336-1G>A site. Our mRNA products of the 336-1G>A mutation under physiological conditions confirmed the prediction. A cryptic site upstream of the constitutive acceptor site was activated in the presence of c.336-1G>A, which resulted in insertion of an intronic sequence within the mature transcript. The retentional intronic segments (31 bp–197 bp) are much longer in our study than the 28 bp predicted by Fichou (Fichou et al., 2015), which suggests that the mRNA product under physiological conditions is different from the plasmid recombination experiments. This mature RHD transcript with intronic segments might be translated into a protein with an additional 10–66 amino-acid sequence, which may alter the structure of the second extracellular loop of RhD. We didn't identify the active cryptic splice site in intron 2. However, we supposed the intronic retention inducing in-frame shift because the intron 3 seems correctly splicing in our study. All reported RHD-/336-1G>A cases are completely serological D-negative, which means no antigens can be detected (Kim et al., 2005; Ye et al., 2009; Fichou et al., 2015). Our results first revealed that the 336-1G>A intron mutation induces the retention of intron 2, which will generate improper peptides.

Prenatal detection of the fetal RhD status is helpful to assess the risk of hemolytic disease in fetus. Non-invasive prenatal testing for predicting fetal RhD status is available in many countries (Finning et al., 2008; Akolekar et al., 2011; Wikman et al., 2012; Chitty et al., 2014; Clausen et al., 2014). The standard quantitative PCR method is based on the assumption of homozygous RHD deletion in D¯ pregnant women. RHD variants in D¯ women will yield high false-positive results (Yang et al., 2019). Therefore, Non-invasive fetal RHD genotyping is only feasible for D¯ women with total or partial deletion of the RHD (Clausen et al., 2019; Yang et al., 2019).

Unlike the high-throughput quantitative PCR method widely used in European countries, we use the PCR byproduct from routine NIPT for aneuploidy as DNA template, then amplified by PCR-SSP method. Because most Chinese pregnant women choose NIPT for aneuploidy after 12 gestational weeks, our NIPT + PCR-SSP protocol for D¯ pregnant women is cost-effective and suitable for the small number of D¯ samples in China. In addition, the results of routine NIPT for aneuploidy can provide good quality control data, such as fetal fraction concentration.

Many research chooses the exons 5, 7, 10 as the targeted exons to predict fetal RhD status (Finning et al., 2008; Akolekar et al., 2011; Chitty et al., 2014; Clausen et al., 2014), while Wikman (Wikman et al., 2012) choose exon 4 as target exon. However, our study found that any one of exon 1, 5, 6, 7, and 9 positive can predict fetal D+ status. Furthermore, all our predicted results were consistent with the newborn RhD status.

Our study also emphasized the importance of resampling for the first negative results. Two cases showed negative results in the first sample but were later confirmed as positive in the second sample, which was likely due to the lower fetal fraction content or PCR bias (Sabina and Leamon, 2015). Thus, we argued the second sampling after a few weeks to confirm the first negative results (Table 3).

No fetuses were predicted as RhD negative in our study. Asian D-negative pregnant women have a 96% possibility of having a D-positive fetus (Chun et al., 2020). Predicting fetal RhD status is not cost-effective in China, whatever methods are used.

To our knowledge, our study used the mRNA-seq to analyze the splicing products of the 336-1G>A under physiological conditions at the first time. However, there has a significant limitation by using an inappropriate sample type, whole peripheral blood, instead of enriched reticulocytes. Hence the reads numbers are relatively small in mRNA-seq. The active cryptic site upstream of the constitutive acceptor site is not identified in our study.