Isolation of Nucleated Red Blood Cells With Intact Genomic DNA From Cord Blood by Applying G&T‐Seq

ABSTRACT

Purpose

Fetal cells in maternal blood are a pure source of fetal genomic DNA for noninvasive prenatal testing (NIPT), if successfully isolated. We assessed whether single‐cell genome and transcriptome sequencing (G&T‐seq) can be applied to efficiently isolate fetal nucleated red blood cells (fNRBCs) suitable for genetic testing.

Methods

Using umbilical cord blood as a model, we isolated 165 single NRBC candidates from four samples and 12 single lymphocytes as controls from one sample. G&T‐seq was used to estimate the maturation stage of each NRBC candidate from the transcriptome data, and genomic integrity was assessed using shallow whole‐genome sequencing (WGS) data.

Results

Multi‐dimensional scaling (MDS) of the transcriptome data revealed that five NRBC candidates clustered separately, classifying them as primitive NRBCs. Two of these cells showed high yields of WGS libraries and high mapping rates comparable to control lymphocytes, suggesting an intact nuclear genome.

Conclusions

G&T‐seq effectively identified primitive NRBCs with high‐quality DNA among candidate cells dominated by mature RBCs. Single‐cell multi‐omics technology may advance the development of fNRBC‐based NIPT.

This study presents a novel approach to identifying primitive‐stage NRBCs from umbilical cord blood using the G&T‐seq method, enabling high‐quality genomic analysis of single cells. The significance of this study lies in its ability to address the current limitations in fetal cell isolation and genomic analysis, offering a solution that could advance the field of prenatal diagnostics.

1. Introduction

Cell‐free DNA (cfDNA)‐based noninvasive prenatal testing (NIPT) has become the standard method for prenatal screening for fetal chromosomal aneuploidy [1]. However, this technology targets fragmented cfDNA primarily derived from placental cells (trophoblasts) rather than fetal cells, and the majority (80%–95%) of cfDNA in the maternal blood is of maternal origin [2]. Due to insufficient positive predictive values for detecting fetal chromosomal aneuploidy, confirmatory diagnostic tests, such as chorionic villus sampling (CVS) or amniocentesis, are recommended when cfDNA‐based NIPT detects fetal chromosomal aneuploidy [2].

Fetal nucleated red blood cells (fNRBCs) in maternal blood during pregnancy were discovered in 1959 [3] and were investigated as targets for NIPT before cfDNA. As fetal cells are the source of pure fetal genomic DNA, genetic diagnoses using genomic DNA from circulating fetal cells in the maternal blood are considered theoretically superior to those using cfDNA for detecting genomic alterations [4]. To date, four types of circulating fetal cells have been identified in the maternal blood [5]. They include fNRBCs, fetal leukocytes, fetal progenitor cells, and trophoblasts. Fetal leukocytes and progenitor cells remain in the maternal circulation for years after delivery, posing a risk of detecting fetal cells from prior pregnancies. These cell types were excluded as targets for prenatal genetic testing in cell‐based NIPT development studies [5]. However, since fNRBCs and trophoblasts are rapidly lost from the maternal circulation after delivery [5] these two cell types are being studied as targets for cell‐based NIPT. Cell‐based NIPT targeting circulating trophoblasts can detect aneuploidies [6], copy number variations with a resolution of 1–2 Mb [7], and disease‐causing variants through single‐cell targeted [8] or whole‐genome sequencing (WGS) [9]. Despite their rarity in maternal blood and the difficulty in isolating them, fNRBCs are attractive targets for cell‐based NIPT because they have a much lower risk of mosaic results than trophoblasts, which can be affected by confined placental mosaicism [5, 10, 11, 12]. The feasibility of cell‐based NIPT targeting fNRBCs has also recently been shown by detecting pathogenic variants of hereditary hearing using whole‐exome and WGS [13]. In these studies [6, 7, 8, 9, 13], isolated single trophoblasts and fNRBCs were subjected to whole‐genome amplification (WGA), followed by microarray or next‐generation sequencing analysis. Technical improvements in the efficiency of fetal cell isolation are expected to facilitate the clinical application of cell‐based NIPT. This will contribute to increasing diagnostic accuracy and may also decrease the cost and labor of the tests.

We previously reported that the combination of fluorescence‐activated cell sorting (FACS) with erythrocyte‐associated surface antigen markers and Y‐chromosome‐specific real‐time PCR efficiently isolated fNRBC candidates in male infants [14]. By conducting WGA‐based WGS analysis, we were able to detect fetal genomic DNA from isolated single cells. However, this method is applicable only to male infants, and the quality of the WGA‐WGS data from the obtained single cells was insufficient for genomic diagnosis, likely due to the loss and/or degradation of fNRBC genomic DNA via enucleation and/or apoptosis. We hypothesized that cells in the primitive stages of NRBC maturation retain intact genomic DNA (Figure 1) and screened for primitive‐stage cells based on their gene expression patterns. We introduced single‐cell genome and transcriptome sequencing (G&T‐seq), which enabled us to obtain both transcriptome and genomic sequencing data from a single cell [15, 16]. Referring to a previously reported dataset of single‐cell RNA sequencing for fetal erythroblasts at various stages of differentiation isolated from umbilical cord blood [17], we chose marker genes for primitive and mature stages of RBC maturation. In the present study, using umbilical cord blood as a model material, we verified the effectiveness of this G&T‐seq strategy for screening NRBCs with intact genomic DNA.

FIGURE 1.

Schematic representation of the G&T‐seq strategy to identify primitive‐stage NRBCs suitable for WGA‐based WGS analysis. (A) By obtaining transcriptome data for each of the NRBC candidates and assessing the expression patterns of marker genes for primitive, intermediate, and mature stages of RBC maturation, we hypothesized that primitive‐stage NRBCs, which are expected to maintain intact genomic DNA, can be suitable for subsequent WGA‐based WGS analysis. (B) The G&T‐seq workflow to first identify primitive‐stage NRBCs by assessing transcriptome data and then conduct WGA‐based WGS analysis only for selected cells.

2. Materials and Methods

2.1. Blood Sample Collection

One sample (5 mL) of peripheral blood from a non‐pregnant adult female and four samples (7.0–8.5 mL) of umbilical cord blood were collected in EDTA‐2K blood collection tubes. Cord blood samples include two female (C1 and C3) and two male (C2 and C4) samples.

2.2. Cell Enrichment, Immunofluorescent Staining, and Flow Cytometric Analysis

Enrichment of lymphocytes and NRBC candidates from blood samples was conducted as described previously [14]. FACS was performed using a BD FACSAria III Cell Sorter (BD Biosciences). The antibodies used for FACS included BD Horizon Fixable Viability Stain 450 (BD Biosciences) to distinguish dead cells from live cells, FITC anti‐human CD45 antibody (BioLegend, San Diego, CA, USA) as a leukocyte marker, and BD Pharmingen APC Mouse Anti‐Human CD71 (BD Biosciences) and BD Pharmingen PE Mouse Anti‐Human CD235a (BD Biosciences) as markers for erythroid precursor cells [14]. Lymphocytes from female peripheral and cord blood were sorted by gating cells positive for CD45 and negative for other markers (CD71, CCD235a, and Viability Stain 450), after gating the lymphocyte fraction in the forward scatter/side scatter plot. NRBC candidates were sorted from cord blood by gating cells positive for CD71 and CD235a and negative for CD45 and Viability Stain 450. Cells were sorted into the wells of a 96‐well plate (#0030129512, Eppendorf, Hamburg, Germany) preloaded with 2.5 μL of Buffer RLT Plus (Qiagen, Venlo, The Netherlands). Plates containing sorted cells were stored at −80°C.

2.3. G&T‐Seq

G&T‐seq was performed as previously described [16] with some modifications. In the reverse transcription step, ERCC RNA Spike‐In Mix was not added, and the Maxima H Minus Reverse Transcriptase (200 U/μL) (Thermo Fisher Scientific, Waltham, MA) was used. After cDNA amplification, fragmented cDNA libraries were generated using the NEBNext Ultra II FS DNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA). The supernatant fraction containing genomic DNA (Figure 2) was purified using 0.65 volume of Agencourt AMPure XP Reagent (Beckman Coulter Inc., Brea, CA). WGA‐based WGS library preparation was conducted using the SMARTer PicoPLEX Gold Single Cell DNA‐Seq Kit (Takara Bio, Shiga, Japan) according to the manufacturer's protocol.

FIGURE 2.

Confirmation of G&T‐seq performance using single‐cell lymphocytes isolated by FACS from adult female peripheral blood. RNA‐seq and WGA‐based WGS libraries were prepared for single cells (1‐cell_02 and 1‐cell_06), a five‐cell pool, a 10‐cell pool, a 1/50 volume of the cell lysate of 50‐cell pool (“1/50 vol. of 50 cells”), and 49/50 volume of the cell lysate of 50‐cell pool (“49/50 vol. of 50 cells”). (A) The correlation coefficient matrix of transcripts per million (TPM) data of six RNA‐seq libraries, whose read numbers ranging from 7.5 to 10.6 million reads (Table S1). (B) Chromosomal distribution of mapped reads observed in a control WGS library generated using 600 ng of genomic DNA from female peripheral blood [14] shown in orange, and of WGA‐based WGS libraries shown in blue (two one‐cell libraries) and in black (other four libraries).

2.4. Sequencing and Data Analyses

RNA‐seq and WGS libraries were sequenced using a MiSeq or NextSeq 550 (Illumina, San Diego, CA, USA) with paired‐end (150 or 75 bp) and dual‐index settings. RNA sequencing data were subjected to gene‐level quantification using Salmon v1.5.2 [18] and Homo_sapiens.GRCh38v38.gtf. The correlation matrix of gene expression data was generated using the R package ggcorrplot with transcript‐per‐million (TPM) data. The multi‐dimensional scaling (MDS) plot and heatmap for the TPM values of the selected genes were generated using the R packages EdgeR and gplots, respectively [19].

WGS data were subjected to the DRAGEN (Dynamic Read Analysis for GENomics) Bio‐IT Platform v3.9.5 and v4.0.3 (Illumina, San Diego, CA) to align sequencing reads to the GRCh38 reference genome with decoy sequences. Mapping, PCR duplicate rates, and mean read depths were obtained from a report file generated by DRAGEN. The number of uniquely mapped reads for each chromosome was counted from the BAM file using the view command in Samtools v1.11 (http://www.htslib.org). Correlation coefficients of the chromosomal distributions of the aligned genomic sequencing reads were calculated using the correl function in Excel.

3. Results

3.1. G&T‐Seq for Lymphocyte Cells Isolated From Adult Female Peripheral Blood

We initially tested whether we could obtain high‐quality RNA‐seq and WGS data by applying G&T‐seq to lymphocytes isolated from adult female peripheral blood. For RNA‐seq libraries, we assessed the cDNA yield and the number of genes expressed. First, the cDNA yields were assessed for the following numbers of cells (s) and wells (s): seven wells for one event (corresponding to one cell) per well, three wells for five events per well, 10 events per well, and 50 events per well (16 wells in total). The cDNA yields were approximately proportional to the number of input cells. The average (and standard deviation) values were 1.2 ng (0.74 ng) for one cell, 3.5 ng (2.34 ng) for five cells, 7.8 ng (0.52 ng) for 10 cells, and 22.4 ng (7.42 ng) for 50 cells (Table S1). We subsequently prepared RNA‐seq libraries from five cDNA samples—two for one cell and one each for five cells, 10 cells, and 50 cells—and sequenced these libraries (8–10 million reads, Table S1), and quantified the gene expression levels. The number of genes detected to be expressed (TPM > 0) was confirmed to increase proportionally with the input cell numbers: 3962 and 6968 for one cell, 10,125 for five cells, 12,392 for 10 cells, and 16,655 for 50 cells. In the correlation analysis of the TPM data from the five RNA‐seq libraries, the correlation coefficients ranged from 0.81 to 0.99 (Figure 2A). Although the number of detected genes was lower in the RNA‐seq libraries from one cell than in the libraries from 10 and 50 cells, the single‐cell libraries were expected to provide information for a sufficient number of genes to estimate the input cell types.

We prepared WGA‐based WGS libraries only for the five samples whose RNA was subjected to RNA‐seq analysis, and sequenced the libraries (1.5–32.4 million reads, Table S1). For the WGA‐based WGS libraries, we assessed library yields, PCR duplicate rates, mapping rates, and the chromosomal distribution of mapped reads. In contrast to the results of the RNA‐seq libraries, the yield of the WGA‐based WGS library was the highest when generated from one cell. The yields of the libraries generated from 5, 10, and 49 cells gradually decreased in proportion to the input cell number (Table S1). This was presumably because the WGA kit used was optimized for five cells as input. A WGS library generated from 600 ng of genomic DNA from female peripheral blood was used as a control library, showing a mapping rate of 99.2% and a PCR duplication rate of 16.3% when 836 million reads were aligned to the GRCh38 reference genome. The mapping rates of the WGA‐based WGS libraries ranged from 97.85% to 99.19%. We have previously shown that the chromosomal ratios of mapped reads were highly concordant (correlation coefficient > 0.99) among regular WGS libraries as well as WGA‐based WGS libraries from a single cell containing intact genomic DNA, and that this concordance could be confirmed by the mapped WGS data from as small as 0.13 million sequencing reads [14]. The chromosomal distribution patterns of the mapped reads of the six WGA‐based WGS libraries from lymphocytes were highly concordant with those of the control WGS library (Figure 2B). Correlation coefficients with those of the control WGS library ranged from 0.991 to 0.997 (Table S1), indicating unbiased amplification during the WGA procedure. These data demonstrate that the G&T‐seq protocol [16] we slightly modified as described in Section 2.3 enabled the satisfactory acquisition of whole transcriptome and genomic information from single cells when intact cells were subjected to library preparation.

3.2. G&T‐Seq for Single NRBC Candidates Isolated From Umbilical Cord Blood

We hypothesized that NRBCs at the primitive stages of RBC maturation contain intact genomic DNA. We conducted G&T‐seq on 165 NRBC candidates from the umbilical cord blood (C1 to C4 samples) and 12 single lymphocytes from the C1 sample. We observed different size distributions of cDNA between lymphocytes and NRBC candidates. While lymphocyte cDNA showed a smear distribution as expected, the cDNA obtained from NRBC candidates contained a sharp peak at approximately 650 bp (Figure 3A). This peak was considered most likely due to cDNAs derived from globin mRNAs known to be highly abundant in mature NRBCs. The peak size roughly corresponds to the sum of 23‐bp length adapter sequences at both ends and their known mRNA sizes: 577 bp (HBA1), 576 bp (HBA2), 628 bp (HBB), and 586 bp (HBG2), respectively, according to the NCBI GenBank accession numbers NM_000558, NM_000517, NM_000518, and NM_000184. Lymphocytes and NRBC candidates also exhibited a remarkable difference in the top‐ranked highly expressed genes: many ribosomal protein genes were top‐ranked for lymphocytes, while globin‐related genes were top‐ranked for NRBC candidates (Table S3). These results demonstrated that lymphocytes and NRBCs can be distinguished by their differences in cDNA size distribution and gene expression patterns when subjected to single‐cell transcriptome analysis (Table S1).

FIGURE 3.

Identification of NRBCs in the primitive stage of their maturation by assessing G&T‐seq data for NRBC candidates from umbilical cord blood. (A) Size distribution of complementary DNA (cDNA) generated for RNA‐seq library preparation for five‐cell pools of lymphocytes (left) and NRBC candidates (right). (B) An MDS plot for the RNA‐seq count data of 12 single lymphocytes (black) and 165 single NRBC candidates (red). Five cells estimated as NRBCs in the primitive stage of their maturation are circled. (C) A heatmap diagram of 13 marker genes for the three maturation stages of NRBC for lymphocytes and NRBC candidates. The names of five NRBC candidates estimated to be in the primitive stage of their maturation are shown in blue. (D) The metrics of RNA‐seq and WGS libraries prepared by the G&T‐seq method from NRBC candidates. The number of genes detected to be expressed by RNA‐seq analysis (top), mapping rate to the human reference genome sequence of WGS libraries (middle), and WGS library yield relative to the average yield of WGS libraries from single lymphocytes (bottom) are bar‐plotted for five primitive‐stage cells (blue) and nine mature‐stage cells (black).

In the MDS plot for the gene expression data, five NRBC candidates (circled) clustered between the lymphocyte cluster and the cluster of the other 160 NRBC candidates along the dimension‐1 axis (Figure 3B). Four out of 12 lymphocyte cells, five NRBC candidates, circled in Figure 3B, and nine cells from the other 160 NRBC candidates were visualized for the expression levels of 13 genes: three primitive‐stage, six intermediate‐stage, and four mature‐stage marker genes of NRBC maturation [17]. Five cells, C1_NRBC_1cell_13, C2_NRBC_1cell_05, C2_NRBC_1cell_25, C4_NRBC_1cell_03, and C4_NRBC_1cell_41, expressed the primitive‐stage marker GYPC and either CD36 or CD47 (Figure 3C, Table S2) [17]. The expression patterns of 13 marker genes observed in the NRBC candidates indicate that five cells represent primitive NRBCs, while the other 160 NRBC candidates are enriched with mature‐stage cells (Table S2).

The number of expressed genes, yield of the WGA‐based WGS library, and mapping rate of WGS reads to the reference genome were assessed for NRBC candidates. In the five primitive NRBCs, the number of expressed genes was higher (ranging from 875 to 2073) than in the other 160 cells (ranging from 24 to 736; average 259; standard deviation 122) (Figure 3D). We obtained sequencing data ranging from 0.19 to 1.02 million reads of the WGA‐based WGS libraries of 165 NRBC candidates, which are equivalent to x 0.0095~x 0.051 genome coverage when the mapping rate is assumed to be 100%. We aligned the obtained reads to the reference genome sequence and determined PCR duplicate rates, mapping rates, and the chromosomal distribution of mapped reads (Table S1). Regarding the C2_NRBC_05 and C2_NRBC_25 cells, their yields of the WGA‐based WGS library (3960.0 and 3720.1 fmol) and their mapping rates of WGS reads to the reference genome (98.3% and 90.2%) were comparable to those of two single lymphocyte WGA‐based WGS libraries, 1‐cell_02 and 1‐cell_06 (Figure 3D, Table S1). Their correlation coefficients of chromosomal distribution of mapped reads compared with those of a control WGS library were 0.972 and 0.969, which were slightly lower than those of two libraries, 1‐cell_02 and 1‐cell_06 (0.997 and 0.991, Table S1). These results suggest the intactness of nuclear genomic DNA in two primitive NRBC cells, C2_NRBC_05 and C2_NRBC_25.

4. Discussion

Recent studies have shown the feasibility of cell‐based NIPT targeting circulating fetal trophoblasts [6, 7, 8, 9] and fNRBCs [13] in maternal blood. These studies have also demonstrated the current technical limitations in isolating enough fetal cells with high purity at a low cost and in amplifying genomic DNA without allelic dropout [5]. Improvement in the efficiency of fetal cell isolation is expected to lower the cost of cell‐based NIPT for monogenic disorders and facilitate its clinical application.

We previously established a protocol to isolate fNRBCs from maternal blood using FACS with erythrocyte‐associated surface antigen markers [14]. One advantage of our FACS‐based approach is that it does not require specialized devices for NRBC isolation, such as cell scanning systems adopted by other groups [6, 7, 9, 13]. However, the efficiency of the WGA of isolated single cells was not consistently sufficient for subsequent WGS. The quality of genomic DNA may have deteriorated because of enucleation and/or apoptosis in most isolated fNRBCs. In our previous study [14], the mapping rates of single‐cell WGS data for NRBC candidates isolated from maternal blood ranged from 20.28% to 90.61%. It has been speculated that NRBCs are present in maternal blood at various maturation stages.

In the present study, we hypothesized that NRBCs in the primitive stage of maturation maintain intact genomic DNA, which is a suitable target for cell‐based NIPT. The data obtained for candidate NRBCs from cord blood support our hypothesis that cells in the primitive stages of NRBC maturation maintain intact genomic DNA. In this study, G&T‐seq was effective in identifying primitive‐stage cells that maintained high‐quality DNA among the single NRBC candidates isolated by FACS. Among the 165 candidate NRBCs, we identified five (5/165, 3%) in the primitive stage of maturation based on their gene expression patterns. By conducting WGA‐based WGS analysis of the five cells, we confirmed that two NRBCs (2/165, 1.2%) maintained intact genomic DNA suitable for genomic diagnosis. These two cells expressed the three primitive‐stage marker genes GYPC, CD47, and CD36. The expression levels of these three genes will serve as criteria for identifying NRBCs in the primitive stage of maturation using the G&T‐seq approach.

Because of the high rates of NRBCs present in cord blood (0.0–13.1 NRBCs/100 white blood cells) [20] and the rarity of NRBCs in adult peripheral blood [21], we assume that the NRBC candidates we isolated from cord blood are of fetal origin even though maternal microchimerism, the presence of a small number of maternal cells in the fetus, has been reported to be prevalent in cord blood [22]. However, it should be noted as a limitation of this study that we cannot determine the origin, fetal or maternal, of NRBC candidates from our shallow WGS data due to the insufficient read depth to obtain genotype information for common variants and also due to our study design in which no maternal specimen was collected. The origin of NRBC candidates, fetal or maternal, can also be estimated based on the presence or absence of Y chromosome reads when they are isolated from male cord blood. In our previous study [14], we determined the average ratios of the reads mapped to Y chromosome to the reads mapped to X chromosome (Y/X ratios) in control WGS libraries. The Y/X ratios in regular WGS libraries generated from female and male genomic DNA were 0.0031 (SD 0.0004, n = 3) and 0.109 (SD 0.001, n = 3), respectively. The Y/X ratios in WGA‐WGS libraries generated from female and male single cells were 0.0007 (SD 0.0002, n = 6) and 0.080 (SD 0.013, n = 5), respectively [14]. High sequence similarities of X and Y chromosomes at gametologous regions [23] outside the pseudoautosomal regions are likely responsible for the presence of the reads mapped to Y chromosome among the female WGS reads. To estimate the origin of the five cells isolated as those in the primitive stage of maturation in this study, we determined the Y/X ratios of five cells from their WGA‐WGS data as described previously [14]. The Y/X ratios of the cells (and the sex of cord blood from which the cell was isolated) were 0.020 (C1_NRBC_1cell_13, female), 0.075 (C2_NRBC_1cell_05, male), 0.095 (C2_NRBC_1cell_25, male), 0.040 (C4_NRBC_1cell_03, male), and 0.015 (C4_NRBC_1cell_41, male). Because of the Y/X ratios close to the average of male control WGA‐WGS libraries (0.080) and their suggested intactness of nuclear genomic DNA (Figure 3D), two C2 cells can be estimated as male cells. On the other hand, the origins of C1 and C4 cells were unclear from their Y/X ratios. Considering the low quality of their WGA‐WGS libraries (Figure 3D), their Y/X ratios may be severely biased.

We have adopted a protocol to isolate NRBC candidates using cell separation density gradient followed by FACS using cell‐surface markers, but without nuclear staining [24, 25]. The results of this and our previous studies [14] indicate that our protocol isolates a larger number of mature NRBCs than primitive NRBCs. To exclude mature and enucleated RBCs and to enrich primitive NRBCs as its consequence, nuclear staining by a fluorescent dye such as Hoechst33342 prior to FACS and selection of nuclear stain‐positive cells by FACS in addition to the selections by cell‐surface marker signals is considered effective to enrich primitive NRBCs.

As we have shown in the utility of G&T‐seq in this study, the application of single‐cell multiomics technology is expected to facilitate the development of fNRBC‐based NIPT.

Ethics Statement

All procedures were conducted in accordance with the ethical standards of the Institutional Ethical Committee on Human Experimentation (institutional and national) and the Declaration of Helsinki of 1964 and its later amendments. This study was approved by the Institutional Review Board of the National Research Institute for Child Health and Development (IRB number 699). Written informed consent was obtained from all patients.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1. The metrics of G&T‐seq libraries from lymphocytes and NRBCs.

Table S2. Transcript per million (TPM) values of 13 markers genes in single lymphocytes and NRBC candidates.

Table S3. Top 100 most significantly changed gene expression levels overall and each TPM value.

Ito N., Fujii T., Taniguchi K., et al., “Isolation of Nucleated Red Blood Cells With Intact Genomic DNA From Cord Blood by Applying G&T‐Seq,” Reproductive Medicine and Biology 24, no. 1 (2025): e12671, 10.1002/rmb2.12671.

Data Availability Statement

Raw sequencing data (WGS and RNA‐seq) are available upon request from the corresponding author (K.N.).