40 years of researching the Del phenotype results in a change of transfusion practice.

Anti-D cannot agglutinate red cells of any Del phenotype in routine serology. Many individuals with East Asian ancestry who type D-negative in serology harbor a Del phenotype. Almost all such individuals carry one distinct DEL variant, dubbed Asian-type DEL, known as RHD*01EL.01, RHD*DEL1, RHD:c.1227G>A, formerly known as RHD(K409K). Clinical evidence strongly suggests that Asian-type DEL individuals can safely be transfused with RhD-positive blood and do not need anti-D prophylaxis in pregnancy.

CD59 gene: 143 haplotypes of 22,718 nucleotides length by computational phasing in 113 individuals from different ethnicities.

BACKGROUND: CD59 deficiency due to rare germline variants in the CD59 gene causes disabilities, ischemic strokes, neuropathy, and hemolysis. CD59 deficiency due to common somatic variants in the PIG-A gene in hematopoietic stem cells causes paroxysmal nocturnal hemoglobinuria. The ISBT database lists one nonsense and three missense germline variants that are associated with the CD59-null phenotype. To analyze the genetic diversity of the CD59 gene, we determined long-range CD59 haplotypes among individuals from different ethnicities. METHODS: We determined a 22.7 kb genomic fragment of the CD59 gene in 113 individuals using next-generation sequencing (NGS), which covered the whole NM_203330.2 mRNA transcript of 7796 base pairs. Samples came from an FDA reference repository and our Ethiopia study cohorts. The raw genotype data were computationally phased into individual haplotype sequences. RESULTS: Nucleotide sequencing of the CD59 gene of 226 chromosomes identified 216 positions with single nucleotide variants. Only three haplotypes were observed in homozygous form, which allowed us to assign them unambiguously as experimentally verified CD59 haplotypes. They were also the most frequent haplotypes among both cohorts. An additional 140 haplotypes were imputed computationally. DISCUSSION: We provided a large set of haplotypes and proposed three verified long-range CD59 reference sequences, based on a population approach, using a generalizable rationale for our choice. Correct long-range haplotypes are useful as template sequences for allele calling in high-throughput NGS and precision medicine approaches, thus enhancing the reliability of clinical diagnostics. Long-range haplotypes can also be used to evaluate the influence of genetic variation on the risk of transfusion reactions or diseases.

Rh flow cytometry: An updated methodology for D antigen density applied to weak D types 164 and 165.

BACKGROUND: An original methodology for determining the D antigen density on red cells was published in 2000 and has been applied in many publications since. This flow cytometry-based assay remained largely unrevised utilizing monoclonal anti-Ds that are not readily available anymore. We updated the methodology to quantify erythrocyte D antigen sites using microspheres and monoclonal anti-Ds that are commercially available today. METHODS: The absolute D antigen density of a frozen standard CcDEe cell, drawn in 2003, a fresh blood donation from the same individual, drawn in 2022, and an internal control CcDEe cell, was quantified by flow cytometry using fluorescence-labeled microspheres. The internal control CcDEe cell was used in conjunction with 9 commercial anti-Ds to determine D antigen densities of 7 normal D, 4 partial D, and 11 weak D type samples, including 2 novel alleles. RESULTS: The reproducibility of the updated assay was evaluated with red cells of published D antigen densities. The current results matched the known ones closely. The new weak D types 164 and 165 carried 4500 and 1505 D antigens/red cell, respectively. The absolute D antigen density decreased from 27,231 to 26,037 in an individual over 19 years. DISCUSSION: The updated assay gave highly reproducible results for the D antigen densities of Rh phenotypes. Readily available anti-Ds allowed for the determination of the D antigen densities of 7 weak D types. The assay is suitable to evaluate the effects of distinct amino acid substitutions on the RhD phenotype.

Asian-type DEL (RHD*DEL1) with an allo-anti-D: A paradoxical observation in a healthy multiparous woman.

BACKGROUND: The DEL phenotype is the D variant expressing the least amounts of D antigen per red cell. Asian-type DEL (RHD:c:1227G > A) is the most prevalent DEL in East Asia without any anti-D alloimmunization reported before. We investigated the first observation of an anti-D in any DEL phenotype, reported in the Japanese language at a 1987 conference, only 3 years after the discovery of DEL. METHODS: We contacted the proband 35 years after the initial report. Standard hemagglutination, adsorption/elution, and flow cytometry tests were performed, as was nucleotide sequencing for the RHD, RHCE, and HLA class I and class II genes. RESULTS: The healthy multiparous Japanese woman, a regular blood donor, still had the anti-D of titer 8 representing an alloantibody by standard serologic methods. Unexpectedly, she carried an Asian-type DEL without any additional RHD gene variation. All 12 HLA alleles identified were known in the Japanese population. Interestingly, one of her HLA-DRB1 and a variant of her HLA-DQB1 alleles had previously been associated with anti-D immunization. CONCLUSION: We described an allo-anti-D, maintained for more than three decades, in an Asian-type DEL. The combination of two implicated HLA alleles were rare and could have contributed to the anti-D immunization. Continued monitoring of anti-D immunization events in patients with DEL is warranted, and we discuss possible mechanisms for further study. As only this single observation has been recognized in the last 35 years, the current recommendation is affirmed: Individuals with Asian-type DEL should be treated as Rh D-positive for transfusion and Rh immune prophylaxis purposes.

Cataloguing experimentally confirmed 80.7 kb-long ACKR1 haplotypes from the 1000 Genomes Project database.

BACKGROUND: Clinically effective and safe genotyping relies on correct reference sequences, often represented by haplotypes. The 1000 Genomes Project recorded individual genotypes across 26 different populations and, using computerized genotype phasing, reported haplotype data. In contrast, we identified long reference sequences by analyzing the homozygous genomic regions in this online database, a concept that has rarely been reported since next generation sequencing data became available. STUDY DESIGN AND METHODS: Phased genotype data for a 80.6 kb region of chromosome 1 was downloaded for all 2,504 unrelated individuals of the 1000 Genome Project Phase 3 cohort. The data was centered on the ACKR1 gene and bordered by the CADM3 and FCER1A genes. Individuals with heterozygosity at a single site or with complete homozygosity allowed unambiguous assignment of an ACKR1 haplotype. A computer algorithm was developed for extracting these haplotypes from the 1000 Genome Project in an automated fashion. A manual analysis validated the data extracted by the algorithm. RESULTS: We confirmed 902 ACKR1 haplotypes of varying lengths, the longest at 80,584 nucleotides and shortest at 1,901 nucleotides. The combined length of haplotype sequences comprised 19,895,388 nucleotides with a median of 16,014 nucleotides. Based on our approach, all haplotypes can be considered experimentally confirmed and not affected by the known errors of computerized genotype phasing. CONCLUSIONS: Tracts of homozygosity can provide definitive reference sequences for any gene. They are particularly useful when observed in unrelated individuals of large scale sequence databases. As a proof of principle, we explored the 1000 Genomes Project database for ACKR1 gene data and mined long haplotypes. These haplotypes are useful for high throughput analysis with next generation sequencing. Our approach is scalable, using automated bioinformatics tools, and can be applied to any gene.

COVID-19 antibody screening with SARS-CoV-2 red cell kodecytes using routine serologic diagnostic platforms.

BACKGROUND: The Coronavirus disease 2019 (COVID-19) pandemic is having a major global impact, and the resultant response in the development of new diagnostics is unprecedented. The detection of antibodies against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has a role in managing the pandemic. We evaluated the feasibility of using SARS-CoV-2 peptide Kode Technology-modified red cells (C19-kodecytes) to develop an assay compatible with existing routine serologic platforms. STUDY DESIGN AND METHODS: A panel of eight unique red cells modified using Kode Technology function-spacer-lipid constructs and bearing short SARS-CoV-2 peptides was developed (C19-kodecyte assay). Kodecytes were tested against undiluted expected antibody-negative and -positive plasma samples in manual tube and three column agglutination technology (CAT) platforms. Parallel analysis with the same peptides in solid phase by enzyme immunoassays was performed. Evaluation samples included >120 expected negative blood donor samples and >140 COVID-19 convalescent plasma samples, with independent serologic analysis from two centers. RESULTS: Specificity (negative reaction rate against expected negative samples) in three different CAT platforms against novel C19-kodecytes was >91%, which correlated with published literature. Sensitivity (positive reaction rate against expected positive convalescent, PCR-confirmed samples) ranged from 82% to 97% compared to 77% with the Abbott Architect SARS-CoV-2 IgG assay. Manual tube serology was less sensitive than CAT. Enzyme immunoassay results with some Kode Technology constructs also had high sensitivity. CONCLUSIONS: C19-kodecytes are viable for use as serologic reagent red cells for the detection of SARS-CoV-2 antibody with routine blood antibody screening equipment.

SCAR: The high-prevalence antigen 013.008 in the Scianna blood group system.

BACKGROUND: The Scianna (SC) blood group system comprises seven antigens. They reside on the erythroblast membrane-associated glycoprotein (ERMAP). The ERMAP and RHCE genes are juxtaposed to each other on chromosome 1. We report a novel SC antigen. STUDY DESIGN AND METHODS: Blood samples came from a patient and his two sisters in Saudi Arabia. To investigate the antibody specificity we used the column agglutination technique and soluble recombinant ERMAP protein. The significance of anti-SCAR was evaluated by the transfusion history and a monocyte monolayer assay. We determined the genomic sequence of ERMAP and RHCE genes. RESULTS: The patient's serum showed an antibody of titer 8 against a high-prevalence antigen. The soluble recombinant ERMAP protein inhibited the antibody. The propositus genotyped homozygous for an ERMAP:c.424C>G variant, for which his sisters were heterozygous. The c.424C>G variant occurred in the SC*01 allele in one haplotype with the RHCE*03 (RHCE*cE) allele. No signs of hemolysis occurred following an incompatible blood transfusion. The monocyte monolayer assay was negative. CONCLUSIONS: We characterized a high-prevalence antigen, with the proposed name "SCAR," which is the eighth antigen of the Scianna blood group system (proposed designation 013.008). Individuals homozygous for ERMAP:p.(Gln142Glu) protein variant can produce anti-SCAR. Although we did not observe any sign of hemolysis at this time, the anti-SCAR prompted a change of the treatment regimen. A review of the known reports indicated that all SC alloantibodies of sufficient titer should be considered capable of causing hemolysis.

Pharmacogenomics with red cells: a model to study protein variants of drug transporter genes.

The PharmacoScan pharmacogenomics platform screens for variation in genes that affect drug absorption, distribution, metabolism, elimination, immune adverse reactions and targets. Among the 1,191 genes tested on the platform, 12 genes are expressed in the red cell membrane: ABCC1, ABCC4, ABCC5, ABCG2, CFTR, SLC16A1, SLC19A1, SLC29A1, ATP7A, CYP4F3, EPHX1 and FLOT1. These genes represent 5 ATP-binding cassette proteins, 3 solute carrier proteins, 1 ATP transport protein and 3 genes associated with drug metabolism and adverse drug reactions. Only ABCG2 and SLC29A1 encode blood group systems, JR and AUG, respectively. We propose red cells as an ex vivo model system to study the effect of heritable variants in genes encoding the transport proteins on the pharmacokinetics of drugs. Altered pharmacodynamics in red cells could also cause adverse reactions, such as haemolysis, hitherto unexplained by other mechanisms.

A resource-conserving serologic and highthroughput molecular approach to screen for blood donors with an IN:-5 phenotype.

CONCLUSIONS: The Indian blood group system (ISBT 023) comprises one lowprevalence antigen, Ina (IN:1), and five high-prevalence antigens: Inb (IN:2), INFI (IN:3), INJA (IN:4), INRA (IN:5), and INSL (IN:6). The antigens are located on the single-pass trans-membrane glycoprotein encoded by the CD44 gene. The present study was designed to identify the prevalence of the INRA- (IN:-5) phenotype and the frequency of its associated allele (IN*02.- 05) to inform us of the probability of finding antigen-negative donors and to assess the risk of antibody formation in transfusion recipients. Buffy coats were extracted from EDTA-anticoagulated whole blood samples, collected with consent from 5261 random blood donors in Surat, Gujarat, India. Standard serologic methods were performed with a modification allowing the use of antiserum generated by recycling the antibody augmented from the test already performed. A real-time polymerase chain reaction- based assay was devised to genotype c.449G>A (p.Arg150His) single nucleotide variation in exon 5 of the CD44 gene. None of the 411 donors tested by serology or the 5261 donors tested molecularly were positive for the IN:-5 phenotype or the allele (IN*02.-05), respectively. The allele frequency estimate ranged from less than 1 in 10,522 (0.01%) to 1 in 3203 alleles (0.03%) in the study cohort (95% confidence interval, Poisson distribution). The absence of this rare allele in the present survey could be due to an ethnic difference, since the donors mostly came from the Hindu community, and the only case of the IN:-5 phenotype was found in the Muslim community. The p.150His variant may be either restricted to the index case family or only found in the Muslim community. Further studies in local subpopulations may provide more information on the frequency of p.150His and its immunogenicity in transfusion recipients if occurring among blood donors.

The phylogeny of 48 alleles, experimentally verified at 21 kb, and its application to clinical allele detection.

BACKGROUND: Sequence information generated from next generation sequencing is often computationally phased using haplotype-phasing algorithms. Utilizing experimentally derived allele or haplotype information improves this prediction, as routinely used in HLA typing. We recently established a large dataset of long ERMAP alleles, which code for protein variants in the Scianna blood group system. We propose the phylogeny of this set of 48 alleles and identify evolutionary steps to derive the observed alleles. METHODS: The nucleotide sequence of > 21 kb each was used for all physically confirmed 48 ERMAP alleles that we previously published. Full-length sequences were aligned and variant sites were extracted manually. The Bayesian coalescent algorithm implemented in BEAST v1.8.3 was used to estimate a coalescent phylogeny for these variants and the allelic ancestral states at the internal nodes of the phylogeny. RESULTS: The phylogenetic analysis allowed us to identify the evolutionary relationships among the 48 ERMAP alleles, predict 4243 potential ancestral alleles and calculate a posterior probability for each of these unobserved alleles. Some of them coincide with observed alleles that are extant in the population. CONCLUSIONS: Our proposed strategy places known alleles in a phylogenetic framework, allowing us to describe as-yet-undiscovered alleles. In this new approach, which relies heavily on the accuracy of the alleles used for the phylogenetic analysis, an expanded set of predicted alleles can be used to infer alleles when large genotype data are analyzed, as typically generated by high-throughput sequencing. The alleles identified by studies like ours may be utilized in designing of microarray technologies, imputing of genotypes and mapping of next generation sequencing data.

Scalable deep text comprehension for Cancer surveillance on high-performance computing.

BACKGROUND: Deep Learning (DL) has advanced the state-of-the-art capabilities in bioinformatics applications which has resulted in trends of increasingly sophisticated and computationally demanding models trained by larger and larger data sets. This vastly increased computational demand challenges the feasibility of conducting cutting-edge research. One solution is to distribute the vast computational workload across multiple computing cluster nodes with data parallelism algorithms. In this study, we used a High-Performance Computing environment and implemented the Downpour Stochastic Gradient Descent algorithm for data parallelism to train a Convolutional Neural Network (CNN) for the natural language processing task of information extraction from a massive dataset of cancer pathology reports. We evaluated the scalability improvements using data parallelism training and the Titan supercomputer at Oak Ridge Leadership Computing Facility. To evaluate scalability, we used different numbers of worker nodes and performed a set of experiments comparing the effects of different training batch sizes and optimizer functions. RESULTS: We found that Adadelta would consistently converge at a lower validation loss, though requiring over twice as many training epochs as the fastest converging optimizer, RMSProp. The Adam optimizer consistently achieved a close 2nd place minimum validation loss significantly faster; using a batch size of 16 and 32 allowed the network to converge in only 4.5 training epochs. CONCLUSIONS: We demonstrated that the networked training process is scalable across multiple compute nodes communicating with message passing interface while achieving higher classification accuracy compared to a traditional machine learning algorithm.

Author Correction: Two large deletions extending beyond either end of the RHD gene and their red cell phenotypes.

The authors of the above paper noticed an error in publication. In Results section, under sub-section RHD genetic variations, the deletion nomenclature for Sample 1 was incorrectly given as [NC_000001.11(NG_007494.1):c.(1-15149_1-15153)_(148+3154_148+3158)del].

Two large deletions extending beyond either end of the RHD gene and their red cell phenotypes.

Only two partial deletions longer than 655 nucleotides had been reported for the RHD gene, constrained within the gene and causing DEL phenotypes. Using a combination of quantitative PCR and long-range PCR, we examined three distinct deletions affecting parts of the RHD gene in three blood donors. Their RHD nucleotide sequences and exact boundaries of the breakpoint regions were determined. DEL phenotypes were caused by a novel 18.4 kb deletion and a previously published 5.4 kb deletion of the RHD gene; a D-negative phenotype was caused by a novel 7.6 kb deletion. Examination of the deletion-flanking regions suggested microhomology-mediated end-joining, replication slippage, and non-homologous end-joining, respectively, as the most likely mechanisms for the three distinct deletions. We described two new deletions affecting parts of the RHD gene, much longer than any previously reported partial deletion: one was the first deletion observed at the 5' end of the RHD gene extending into the intergenic region, and the other the second deletion observed at its 3' end. Large deletions present at either end are a mechanism for a much reduced RhD protein expression or its complete loss. Exact molecular characterization of such deletions is instrumental for accurate RHD genotyping.

Transfusion strategy for weak D Type 4.0 based on RHD alleles and RH haplotypes in Tunisia.

BACKGROUND: With more than 460 RHD alleles, this gene is the most complex and polymorphic among all blood group systems. The Tunisian population has the largest known prevalence of weak D Type 4.0 alleles, occurring in one of 105 RH haplotypes. We aimed to establish a rationale for the transfusion strategy of weak D Type 4.0 in Tunisia. STUDY DESIGN AND METHODS: Donors were randomly screened for the serologic weak D phenotype. The RHD coding sequence and parts of the introns were sequenced. To establish the RH haplotype, the RHCE gene was tested for characteristic single-nucleotide positions. RESULTS: We determined all RHD alleles and the RH haplotypes coding for the serologic weak D phenotype among 13,431 Tunisian donations. A serologic weak D phenotype was found in 67 individuals (0.50%). Among them, 60 carried a weak D Type 4 allele: 53 weak D Type 4.0, six weak D Type 4.2.2 (DAR), and one weak D Type 4.1. An additional four donors had one variant allele each: DVII, weak D Type 1, weak D Type 3, and weak D type 100, while three donors showed a normal RHD sequence. The weak D Type 4.0 was most often linked to RHCE*ceVS.04.01, weak D Type 4.2.2 to RHCE*ceAR, and weak D Type 4.1 to RHCE*ceVS.02, while the other RHD alleles were linked to one of the common RHCE alleles. CONCLUSIONS: Among the weak D phenotypes in Tunisia, no novel RHD allele was found and almost 90% were caused by alleles of the weak D Type 4 cluster, of which 88% represented the weak D Type 4.0 allele. Based on established RH haplotypes for variant RHD and RHCE alleles and the lack of adverse clinical reports, we recommend D+ transfusions for patients with weak D Type 4.0 in Tunisia.

BOC is a modifier gene in holoprosencephaly.

Holoprosencephaly (HPE), a common developmental defect of the forebrain and midface, has a complex etiology. Heterozygous, loss-of-function mutations in the sonic hedgehog (SHH) pathway are associated with HPE. However, mutation carriers display highly variable clinical presentation, leading to an "autosomal dominant with modifier" model, in which the penetrance and expressivity of a predisposing mutation is graded by genetic or environmental modifiers. Such modifiers have not been identified. Boc encodes a SHH coreceptor and is a silent HPE modifier gene in mice. Here, we report the identification of missense BOC variants in HPE patients. Consistent with these alleles functioning as HPE modifiers, individual variant BOC proteins had either loss- or gain-of-function properties in cell-based SHH signaling assays. Therefore, in addition to heterozygous loss-of-function mutations in specific SHH pathway genes and an ill-defined environmental component, our findings identify a third variable in HPE: low-frequency modifier genes, BOC being the first identified.

Endothelial cell-derived non-canonical Wnt ligands control vascular pruning in angiogenesis.

Multiple cell types involved in the regulation of angiogenesis express Wnt ligands. Although ß-catenin dependent and independent Wnt signaling pathways have been shown to control angiogenesis, the contribution of individual cell types to activate these downstream pathways in endothelial cells (ECs) during blood vessel formation is still elusive. To investigate the role of ECs in contributing Wnt ligands for regulation of blood vessel formation, we conditionally deleted the Wnt secretion factor Evi in mouse ECs (Evi-ECKO). Evi-ECKO mice showed decreased microvessel density during physiological and pathological angiogenesis in the postnatal retina and in tumors, respectively. The reduced microvessel density resulted from increased vessel regression accompanied by decreased EC survival and proliferation. Concomitantly, survival-related genes were downregulated and cell cycle arrest- and apoptosis-inducing genes were upregulated. EVI silencing in cultured HUVECs showed similar target gene regulation, supporting a mechanism of EC-derived Wnt ligands in controlling EC function. ECs preferentially expressed non-canonical Wnt ligands and canonical target gene expression was unaffected in Evi-ECKO mice. Furthermore, the reduced vascularization of Matrigel plugs in Evi-ECKO mice could be rescued by introduction of non-canonical Wnt5a. Treatment of mouse pups with the non-canonical Wnt inhibitor TNP470 resulted in increased vessel regression accompanied by decreased EC proliferation, thus mimicking the proliferation-dependent Evi-ECKO remodeling phenotype. Taken together, this study identified EC-derived non-canonical Wnt ligands as regulators of EC survival, proliferation and subsequent vascular pruning during developmental and pathological angiogenesis.

Full-length nucleotide sequence of ERMAP alleles encoding Scianna (SC) antigens.

BACKGROUND: Scianna (SC) blood group system comprises two antithetical antigens, Sc1 and Sc2, and five additional antigens. The antigens reside on a glycoprotein encoded by the erythroblast membrane-associated protein (ERMAP) gene. For the common ERMAP alleles, we determined the full-length nucleotide sequence that encodes the Scianna glycoprotein. STUDY DESIGN AND METHODS: Blood donor samples from five populations were analyzed including 20 African Americans, 10 Caucasians, 10 Thai, five Asians, and five Hispanics for a total of 100 chromosomes. An assay was devised to determine the genomic sequence of the ERMAP gene in one amplicon, spanning 21.4 kb and covering Exons 2 to 12 and the intervening sequence (IVS). All alleles (confirmed haplotypes) were resolved without ambiguity. RESULTS: Among 50 blood donors, we found 80 single-nucleotide polymorphisms (SNPs), including six novel SNPs, in 21,308 nucleotides covering the coding sequence of the ERMAP gene and including the introns. The noncoding sequences harbored 75 SNPs (68 in the introns and seven in the 3'-UTR). No SNP indicative of a nonfunctional allele was detected. The nucleotide sequences for 48 ERMAP alleles (confirmed haplotypes) were determined by allele-specific polymerase chain reaction and sequencing in 100 chromosomes. CONCLUSIONS: We documented 48 ERMAP alleles of 21,308 nucleotides each. The two nucleotide sequences available in GenBank for ERMAP alleles of similar length have not been found in our 100 chromosomes. Alleles determined without ambiguity can be used as templates to analyze next generation sequencing data, which will enhance the reliability in clinical diagnostics.

Genotype frequency of human neutrophil antigen-3 polymorphisms in the Yi, Han, and Tibetan populations of China.

BACKGROUND: Human neutrophil antigen-3 (HNA-3) alloantibodies can cause fatal transfusion-related acute lung injury (TRALI). Most frequencies of SLC44A2 alleles encoding the HNA-3a/b antigens have been established in Han individuals by polymerase chain reaction with sequence-specific priming (PCR-SSP). We sequenced SLC44A2 gene fragments and determined allele frequencies in three ethnicities of China. STUDY DESIGN AND METHODS: Genomic DNA was extracted from 448 samples of 100 blood donors of Yi ethnicity in Xichang, Liangshan; 248 Han in Nanjing, Jiangsu; and 100 Tibetan in Lhasa, Tibet. A PCR-SSP was applied to determine the phase of two single-nucleotide polymorphisms (SNPs); SLC44A2 haplotypes were constructed. RESULTS: In the 567 nucleotides of the SLC44A2 gene covered by our sequencing approach in Han individuals, we detected the known 331-44G>A (rs12972963) and 461G>A (rs2288904) polymorphisms. In the 243 nucleotides sequenced in Yi and Tibetan populations, we detected the known 461G>A and 503-15T>C (rs1560711) polymorphisms. A PCR-SSP for the common HNA-3a/b SNP was 100% concordant. The frequencies of the HNA-3a allele were 0.58, 0.66, and 0.69 in Yi, Han (Nanjing), and Tibetan, respectively (0.42, 0.34, and 0.31 for HNA-3b). CONCLUSIONS: The Yi population of China had the highest frequency of blood donors at risk of harboring anti-HNA-3a compared to any population studied so far. We confirmed that the underlying SLC44A2*2 allele is more common in China than in any European or African populations.