The molecular genetic background leading to the formation of the human erythroid-specific Xg^a/CD99 blood groups

The Xg blood group system, the 12th of the 36 human blood group systems to be acknowledged, contains 2 antigens: Xg^a and CD99. Xg^a, first identified early in 1962,¹  is an X chromosome–linked blood group antigen. In 1981, Goodfellow and Tippett identified the Xg^a-related CD99 blood group antigen, and the Xg^a and CD99 antigens were found to show a unique and sex-specific phenotypic relationship on the surface of red blood cells (RBCs).²  The Xg(a+) and Xg(a−) blood groups are common polymorphic phenotypes among individuals; furthermore, CD99 expression shows erythroid-specific quantitative traits that have been classified into the CD99-high (CD99H) and CD99-low (CD99L) phenotypes. This CD99H and CD99L phenotypic variation is directly related to the Xg^a blood group.^2,3  Among females and males, the Xg(a+) phenotype is associated with the CD99H phenotype. On the other hand, Xg(a−) females show an association with the CD99L phenotype; however, Xg(a−) males, in addition to having the CD99L phenotype as 1 possibility, may also have the CD99H phenotype (Figure 1A).

A series of studies were conducted from the 1980s onwards with the aim of revealing the genetic and molecular basis for the intriguing phenotypic relationship between the Xg^a and CD99 blood groups. The Xg^a and CD99 antigens are glycoproteins⁴  and are found to be expressed separately on the membrane of RBCs.⁵  Notably, it has been found that the reticulocytes from Xg(a+)/CD99H individuals have a higher level of the XG and CD99 (MIC2) transcripts than the reticulocytes from Xg(a−)/CD99L individuals.^3,6  Based on these findings, it has been proposed that the Xg^a/CD99 blood groups and their phenotypic relationship are brought about by the transcriptional coregulation of XG and CD99 expression (reviewed in Tippett and Ellis⁶  and Daniels⁷ ). XG and CD99 are homologous genes, and each consists of 10 exons.^8,9  They are situated side by side at the tips of the X and Y chromosomes and span the pseudoautosomal boundary (Figure 2). CD99 is located in pseudoautosomal region 1 (PAR1),¹⁰  whereas XG spans the pseudoautosomal boundary with exons 1-3 being located within PAR1 and exons 4-10 being X specific.¹¹  As a result, an incomplete XG is present on the Y chromosome. Neither gene is subject to X inactivation.^12-14  After the genomic organization of XG and CD99 was recognized, a genetic element, the XG Regulator, was proposed to coregulate the expression of XG and CD99 and, thus, determine the Xg^a/CD99 blood groups.^6,15 

In the present investigation, through a pilot investigation that involved large-scale DNA sequence analysis of the relevant XG and CD99 genomic areas using targeted next-generation sequencing (NGS), together with a follow-up association study, we have demonstrated that the Xg^a/CD99 blood groups are associated with a nucleotide polymorphism that affects the single nucleotide polymorphism (SNP) rs311103. During the review process for this manuscript, a study reported that SNP rs311103 is responsible for the Xg(a+)/Xg(a−) phenotypes.¹⁶ 

Peripheral blood samples were collected from 78 healthy nonrelated Taiwanese (47 females and 31 males). The Xg^a blood group of each sample was determined by the antiglobulin test using human monoclonal anti-Xg^a reagent HIRO-123 (a gift from Tokyo Metropolitan Blood Center, Japanese Red Cross Society). Genomic DNA was purified from the peripheral blood cells of each sample using a QIAamp DNA Blood Mini Kit (QIAGEN). This human study was approved by the Institutional Review Board of National Taiwan University, and informed consent was obtained from all participants.

After blocking with 5% bovine serum albumin in phosphate-buffered saline, the RBCs were incubated with mouse monoclonal anti-CD99 antibody (clone 12E7)¹⁷  (Abcam) at a final concentration of 20 μg/mL for 1 hour at 4°C. The bound monoclonal antibodies were detected by incubation with fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgG antibody (Leinco Technologies). Finally, the cells were subjected to flow cytometry to measure their fluorescence.

The X-chromosome 2240-2810K regions of genomic DNA samples were enriched using a SureSelectXT Custom kit (Agilent Technologies). The sequences were then determined using a NextSeq 500 system (Illumina).

Five DNA segments encompassing a 4.7-kb region (X chromosome 2 663 675-2 668 362-bp region) were amplified using polymerase chain reaction (PCR), and the sequences were determined using the Sanger method. The PCR primer pairs used are listed in supplemental Table 1.

The PCR primer pair XGFa and XGRb (supplemental Table 1) was used to amplify the 2 666 284-2 666 483-bp region of the X chromosome. The PCR-amplified DNA fragments were inserted into the BamHI and SalI sites of the pGL4.23 reporter vector (Promega). The created reporter constructs, denoted 200-G and 200-C, carry a 200-bp genomic DNA segment in which the rs311103 site bears the G or C nucleotide, respectively. The primer pair XGFc and XGRf was used to amplify the X-chromosome 2 666 352-2 666 413-bp region using PCR. The amplified fragments were inserted into the KpnI and SacI sites of the pGL4.23 vector, and the created reporter constructs, denoted 62-G and 62-C, carry a 62-bp genomic DNA segment in which the rs311103 site bears the G or C nucleotide, respectively.

Transfection of the reporter vectors and the β-galactosidase–expressing plasmid pCH110 (Amersham), which served as an internal control, into the host cells and the analysis of the reporter assays were performed as described previously.¹⁸ 

The expression vectors for the GATA1-GATA6 and LEF1 transcription factors were purchased from Origen Technologies. Cotransfection of the expression vector, reporter construct, and internal control vector into the host cells in the reporter assays was performed as described previously.¹⁹ 

Oligonucleotides with sequences GGTTTGCAAAGATAAATGCCTT (sense for rs311103[G]), AAGGCATTTATCTTTGCAAACC (antisense for rs311103[G]), GGTTTGCAAACATAAATGCCTT (sense for rs311103[C]), and AAGGCATTTATGTTTGCAAACC (antisense for rs311103[C]) were synthesized. Annealing of the complementary oligonucleotides yielded 22-bp rs311103[G] and rs311103[C] DNA fragments. HEK293T cells were transfected with the expression vectors for GATA1 and GATA2 or the empty pCMV6-entry vector and cultured for 48 hr. Nuclear extracts of the transfected HEK293T cells and the native K-562 cells were prepared using a Nuclear Extract kit (Active Motif). The subsequent electrophoretic mobility shift assay (EMSA) experiments were performed as described previously.¹⁹ 

Chromatin immunoprecipitation (ChIP) analysis was performed using the Pierce Magnetic ChIP assay kit (Thermo Scientific), following the manufacturer’s protocols. A total of 4 × 10⁶ HEL cells was used in the ChIP analysis, and rabbit monoclonal anti-GATA1 antibody (clone EPR17362), rabbit polyclonal anti-GATA2 (Abcam) antibody, or normal rabbit IgG was used to immunoprecipitate the chromatin DNAs that had been cross-linked with the transcription factors. Real-time PCR was performed using a QuantiFast SYBR Green PCR Kit and the primer pair SF and SR (supplemental Table 1).

Cells were cultured as described in supplemental Methods.

Peripheral blood samples from 78 Taiwanese subjects (47 females and 31 males) were collected in the present investigation, and their Xg^a blood groups were determined (Figure 1A). Flow cytometry analysis was used to quantitatively compare the expression levels of the RBC CD99 antigen (Figure 1B) among the subjects. In total, 16 samples from the 41 Xg(a+) females, 17 samples from the 19 Xg(a+) males, together with the samples from the 6 Xg(a−) females and 12 Xg(a−) males were subjected to the flow cytometry analysis. All of the 33 Xg(a+) samples analyzed were found to have the CD99H phenotype, whereas the 6 Xg(a−) females were found to have the CD99L phenotype. Six of the 12 Xg(a−) males had the CD99L phenotype, whereas the remaining 6 Xg(a−) males were shown to have the CD99H phenotype (Figure 1A). The RBC CD99 antigen expression levels of each individual were compiled based on the different Xg^a/CD99 phenotypes present in females and males (Figure 1C).

Next, we conducted a pilot investigation that involved targeted NGS of the 2240-2810K regions of the X chromosome (Figure 2), which encompasses the region from 370 kb upstream of CD99 to 75 kb downstream of XG. This was carried out on the genomic DNAs from 16 individuals; these subjects consisted of 4 females and 4 males with the Xg(a+)/CD99H phenotype and 4 females and 4 males with the Xg(a−)/CD99L phenotype. Comparison of the 570-kb genomic sequences of these 16 individuals revealed 3536 SNPs (data not shown); among these SNPs, rs311103 and rs311104 were found to be associated with the Xg^a/CD99 phenotypic polymorphisms of the 16 individuals.

The genomic areas encompassing rs311103, rs311104, and an additional 15 contiguous SNPs from the 16 individuals who had been enrolled in the pilot study were amplified by PCR, and the sequences were determined using the Sanger method. This verified the DNA sequences of the 4.7-kb genomic regions (2 663 675-2 668 362 bp) of the 16 individuals, with the exception of an ∼270-bp region (around 2 667 140-2 667 410 bp) that contains a highly repeated region of C and T nucleotides. The Sanger resequencing confirmed the genotypes of rs311103, rs311104, and the 15 contiguous SNPs of these 16 individuals (supplemental Table 2). Comparison of the 4.7-kb DNA sequences did not identify any other common nucleotide polymorphisms among the 16 individuals. This further analysis confirmed the association between SNPs rs311103 and rs311104 and the Xg^a/CD99 phenotypic polymorphisms among the 16 individuals enrolled in the pilot study.

Subsequent to the above analysis, we completed the genotyping of the 17 SNPs (rs311103, rs311104, and the 15 contiguous SNPs) of the remaining 62 samples using Sanger sequencing. The results obtained from this larger-scale association study (summarized in supplemental Table 3) demonstrated that there was a definite association between genotypic polymorphism at rs311103 and the phenotypic polymorphisms of the Xg^a/CD99 blood groups (Figure 3A), but there was no such association for rs311104 or the other 15 contiguous SNPs.

An overview of the association status of the 17 SNPs from the 78 individuals is shown in Figure 3B, and rs311103 shows a significant association with the Xg^a/CD99 blood group phenotypes (P = 4.88 × 10^‒11). The genotypes of the 17 SNPs observed in the 78 individuals was examined for the observed vs expected heterozygosity using the Fisher exact test; the results, with P values for the 17 SNPs ranging from .31 to 1, verify that the observed genotypes of the 17 SNPs do not depart from Hardy-Weinberg equilibrium.^20,21  In addition, the haplotypes of the 17 SNPs present in the 78 individuals were reconstructed, and the most likely haplotype pair was assigned for each individual using the PHASE program.²²  In total, 19 haplotypes for the 17 SNPs were identified among the 78 individuals (supplemental Figure 1), indicating the presence of a relatively high recombination rate across this region. It is known that PAR1 has a cross-over rate much greater than the genome-wide average.^23,24  Within these highly diverse haplotype pair combinations, the genotypes at rs311103 are consistently associated with the phenotypic polymorphisms of the Xg^a/CD99 blood groups across the 78 individuals. Specifically, a comparison of the RBC CD99 antigen expression levels in the individual groups according to their rs311103 genotypes showed a significant correlation between a higher level of CD99 expression and the copy number of the rs311103[G] allele (Figure 3C). These findings further support the correlation between the polymorphism in terms of rs311103 genotypes and the Xg^a/CD99 blood group phenotypes.

To investigate the molecular connection between rs311103 nucleotide polymorphism and the differential transcription levels of allelic XG and CD99, we used a reporter assay to examine the transcriptional activity of the genomic regions spanning rs311103.

When the 200-G reporter construct, which bears a 200-bp genomic segment that encompasses rs311103[G], was introduced into K-562 erythroleukemia cells, a dramatically higher level of transcriptional activity was observed; this transcriptional activity was completely absent when a similar construct, 200-C, which had only a single G-to-C change within rs311103, was examined (Figure 4A). The reporter assays of the 62-G and 62-C constructs, which carry a 62-bp genomic region that encompassed rs311103[G] and rs311103[C], respectively, produced similar results using K-562 cells. In addition, when the erythroleukemia cell lines HEL and LAMA-84 were used as hosts for the same reporter assays, the 62-bp rs311103[G] segment also showed a high level of transcriptional activity (Figure 4B). In contrast, a similar induction of transcriptional activity by this 62-bp rs311103[G] segment was not observed in the nonerythroid lineage cell lines HEK293T and DLD-1. These findings strongly suggest that the rs311103[G] genomic region possesses a high transcription-enhancing activity with an erythroid lineage-specific characteristic.

The polymorphic rs311103 genomic regions were analyzed using TRANSFAC (geneXplain) to identify potential transcription factor binding motifs. Putative binding motifs for the GATA binding protein family (Figure 5A) and for lymphoid enhancer binding factor 1 (LEF1) (data not shown) were identified within the rs311103[G] region but not within rs311103[C] region. To explore any relevant effect, LEF1 and the 6 known members of the GATA family (GATA1-GATA6) were ectopically expressed in K-562 cells that had been transfected with the 62-G or 62-C reporter construct. The results showed that expression of each of the 6 GATA members, but not of the LEF1 factor, was able to significantly elevate the transcriptional activity of the rs311103[G] segment (Figure 5B). Specifically, the hematopoietic GATA factors, GATA1-GATA3, were found to give rise to a much greater induction of the transcription-enhancing activity than the other 3 GATA factors; this effect was most noticeable for GATA1.

The expression levels of the GATA1-GATA6 transcripts and the LEF1 transcript in the erythroleukemia cell lines and the nonerythroid lineage cell lines that were used for these experiments were quantitatively analyzed (supplemental Figure 2). GATA1 and GATA2 are known to be crucial factors for the development of erythroid-lineage cells; thus, considerable amounts of the GATA1 and GATA2 transcripts and proteins were found to be present in the erythroleukemia cell lines K-562, HEL, and LAMA-84, as would be expected. This contrasts with the situation within the nonerythroid cell lines HEK293T and DLD-1, in which the expression levels of the GATA1 and GATA2 transcripts were found to be very low. The expression levels of the other transcription factors (ie, GATA3-GATA6 and LEF1) in the erythroleukemia cell lines were also very low or virtually undetectable. These findings suggest that the high transcriptional activity of the rs311103[G] segment that was specifically found in erythroid-lineage cells is likely to be substantially due to the presence of high levels of these 2 erythroid GATA factors (GATA1 and GATA2) in these cells.

Next, EMSA was carried out to examine the binding of the GATA1 and GATA2 factors to the polymorphic rs311103 genomic segments. As shown in Figure 6A (left panel), there was specific binding of GATA1 and GATA2 to the rs311103[G] DNA fragment, but such binding was virtually absent when the rs311103[C] fragment was used. Competition experiments using unlabeled rs311103[G] and rs311103[C] fragments further demonstrated that a G-to-C nucleotide change at rs311103 position virtually abolishes the affinity of GATA1 and GATA2 for the genomic region in question (Figure 6A, right panel).

When nuclear extracts prepared from native K-562 cells were used, specific binding of the endogenous GATA1 factor in K-562 cells to the rs311103[G] fragment was found (Figure 6B). However, binding of the endogenous GATA2 factor, which is also present in K-562 cells in large amounts, to rs311103[G] was not observed.

The in vivo binding status of GATA1 and GATA2 to the rs311103 regions was assessed using ChIP analysis. The results demonstrated that there was an association of GATA1 with the rs311103 genomic region in HEL erythroleukemia cells; however, under the same circumstances, the association of GATA2 to the rs311103 region was not significant (Figure 7). The ChIP results agree well with the results obtained from the EMSA experiments and show that endogenous GATA1, but not endogenous GATA2, is able to bind to the rs311103 segment.

Examination of the ChIP sequencing data from primary human erythroblasts in the Encyclopedia of DNA Elements databank using the UCSC Genome Browser manifested significant binding of the GATA1 factor to the rs311103 region (supplemental Figure 3A). This further substantiates the proposal that there is substantial binding of GATA1 to the SNP region in erythroid-lineage cells. The ChIP sequencing data also show binding of GATA1 to the XG and CD99 promoter regions in primary human erythroblast cells (supplemental Figure 3B).

During the present study, a pilot investigation was carried out that used targeted NGS, and this was followed by an association study that was able to demonstrate a definite association between SNP rs311103 and the Xg^a/CD99 blood groups. Our results indicate that the rs311103[G] and rs311103[C] alleles on the X chromosome are responsible for the Xg(a+)/CD99H and Xg(a−)/CD99L phenotypes, respectively, and the rs311103[G] allele on the Y chromosome is responsible for the Xg(a−)/CD99H phenotype in males. The SNP rs311103 is located between XG and CD99 and is situated at a position 3.7 kb upstream and 57 kb downstream of the transcription start sites of XG and CD99, respectively (Figure 2). It is located in the PAR1 region of the X and Y chromosomes; this agrees with the previous proposition that the postulated XG Regulator element should be situated within the PAR area.^6,15  The frequencies of the XG^(a+) gene have been calculated using Xg^a phenotype frequencies in various populations.^7,25-27  These findings were compared with the rs311103[G] frequencies available from the HapMap databank; the latter results were found to parallel well with the calculated XG^(a+) frequencies (supplemental Table 4).

It has been pointed out that the Xg^a/CD99 blood groups are erythroid specific. The present investigation reveals that the rs311103[G] genomic region, but not the rs311103[C] region, brings about a high level of transcriptional enhancement that is specific to erythroid-lineage cells. In addition, our results show how GATA factors are important to the induction of the transcriptional activity of the rs311103[G] region. The high frequency of the G nucleotide at the second position of the GATA-binding consensus WGATAR (Figure 5A) points to how critical this G nucleotide is to the binding of GATA transcription factors. This situation also provides an explanation for the dramatic contrast in transcriptional activity between the rs311103[G] and rs311103[C] genomic segments in erythroid-lineage cells. Via the transcription-activation function of the erythroid GATA factors, the rs311103 G/C variation present in the human population results in significant differences in the transcription-enhancer activity of the rs311103 genomic regions in erythroid lineage between individuals, consequently leading to the different Xg^a/CD99 blood group phenotypes.

Among human blood groups, cis-regulatory SNPs have been identified that are associated with the P₁/P₂ phenotypes of the P1PK blood group system.¹⁸  Recently, the transcription factors involved in differential binding to this polymorphic SNP region have been proposed to control the differential activation of P¹-A4GALT and P²-A4GALT expression in P₁ and P₂ red cells.^19,28  The present investigation provides another example of a human blood group quantitative trait that involves a SNP cis-linked to the genes responsible for the phenotype. In addition, disruption of a GATA1 motif by a genetic variant (rs2814778) has been identified in the FY (ACKR1) gene promoter region of Africans with the Duffy-negative phenotype.²⁹  This disruption abolishes the expression of the Duffy blood group antigens on RBCs; as a consequence, this confers resistance to invasion by Plasmodium vivax and Plasmodium knowlesi malarial parasites.^30,31  The present study, the Duffy-negative phenotype findings, and various forms of congenital anemia^32-35  show the profound effect that disruption of a canonical GATA1-binding motif can have on the expression of the target genes in RBCs. In this context, it is of interest to note that other common genetic variants associated with RBC traits do frequently alter GATA1 regulatory elements; in these cases, they do not act by critically disrupting the canonical core-binding motif, but rather change the sequences flanking the core-binding motif. These changes subtly alter the binding/activity of GATA1 and various cofactors, which can result in mild changes in the expression level of the target genes.^36,37 

The expression of the Xg^a antigen appears to be restricted to a limited number of cell types, such as RBCs, and the expression levels of this antigen are relatively low. The expression of the Xg^a antigen in Ewing’s sarcoma has been shown to contribute to the tumor invasiveness and to be of prognostic value.³⁸  In contrast to Xg^a, CD99 is abundantly expressed in a wide range of tissues and has drawn considerable attention in recent years.^39-41  CD99 has been recognized to be 1 of the critical molecules involved in the leukocyte transendothelial migration signaling pathway; thus, it plays an important role in the recruitment of leukocytes into inflamed areas.^39,42,43  Particularly, subsets of CD34⁺ hematopoietic stem cells with high or low CD99 expression levels have been identified; these subsets have different levels of transendothelial migration activity and show variation in the numbers of blood cells of several downstream lineages produced after differentiation.⁴⁴  In addition, CD99 has been demonstrated to act as a surface marker for the stem and progenitor cells during acute myeloid leukemia and myelodysplastic syndrome and to be a promising therapeutic target for the treatment of both.^40,45  In view of the crucial cellular and pathological roles played by CD99, it will be of interest in the future to explore whether the rs311103 common genetic variation investigated here leads to different levels of CD99 expression in other cell types, just like in RBCs. Of particular interest would be cells in which members of the GATA family are involved in development; 1 possible outcome of this research would be the identification of different physiopathological conditions in individual groups carrying the different SNP genotypes.

Among the 6 members of the GATA family, it has been established that GATA2 is critical to the generation and functioning of hematopoietic stem and progenitor cells, including early erythroid precursor cells. Furthermore, GATA1 is essential to the development of the erythroid lineage and the regulation of RBC differentiation.^46,47  Specifically, the 2 erythroid-related GATA factors are 2 out of 6 GATA factors that are able to induce the greatest amount of transcriptional activity via the rs311103[G] segment (Figure 5B). Nonetheless, the EMSA results using K-562 nuclear extracts, as well as the ChIP analysis, suggest that GATA1, but not GATA2, is involved in the differential transcriptional activity of the rs311103[G] and rs311103[C] segments. Further studies are required to verify this. In addition, as the genetic basis for the Xg^a/CD99 blood groups, the rs311103 region must be involved in the coregulation of XG and CD99 expression in erythroid-lineage cells. Furthermore, the rs311103[G] region needs to have the ability to interact over a long distance to influence the CD99 promoter and affect the transcription of CD99. It should be noted that the recognition and binding between a specific protein factor and its DNA binding motif, followed by subsequent transcription, is a highly complex event and often involves many other protein factors. Indeed, various transcription factors and cofactors have been shown to associate with the GATA1 and GATA2 factors and to be involved in the regulation of the activation/repression functions of GATA1/GATA2 with respect to erythroid gene expression.^46-49  Some of these factors have been identified as being capable of mediating long-range interactions between the enhancers and the promoters of erythroid genes.^50-52  Further studies are needed to delineate the molecular details associated with the allelic variation in expression of the XG and CD99 genes with respect to the rs311103[G] and rs311103[C] alleles. It will be of particular interest to further explore the functions of specific erythroid GATA factors and any other factors that are found to be associated with them. This will involve investigating the details of the regulation system, especially the involvement of these factors in the long-range interaction between the rs311103[G] genomic region and the CD99 promoter and in the coregulation of XG and CD99 expression.

This work was supported in part by Grants MOST 104-2320-B-002-031 and MOST 104-2321-B-002-073 from the Ministry of Science and Technology, Taiwan.

Contribution: C.-C.Y. designed and performed the research, collected the data, analyzed the data, and interpreted the data; C.-J.C. and Y.-C.T. performed the research and interpreted the data; C.-C.C. analyzed the data; B.-S.L., J.-T.H., S.-T.H., Y.-S.C., and Y.-J.T. performed the research; S.-W.L. and M.L. contributed the analytical tools; and L.-C.Y. designed the research, collected the data, analyzed the data, interpreted the data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Lung-Chih Yu, Institute of Biochemical Sciences, College of Life Science, National Taiwan University, No. 1, Roosevelt Rd Sec. 4, Taipei 106, Taiwan; e-mail: yulc@ntu.edu.tw.