High-Throughput Next-Generation Sequencing of the Kidd Blood Group: Unexpected Antigen Expression Properties of Four Alleles and Detection of Novel Variants

Background The blood supply for patients with foreign ethnic backgrounds can be challenging, as they often have blood group and HPA patterns that differ from the variants prevalent in the German population. In addition, hemoglobinopathies requiring regular blood transfusion may be more common in such populations. High-throughput genotyping tests can facilitate the identification of the most compatible blood products, thereby reducing the risk of transfusion reactions. The present study reports the results of a molecular study for the Kidd (JK) blood group. Allele frequencies and antigen prevalence data are presented for >8,000 individuals of various origins. Material and Methods More than 8,000 blood donors were genotyped for 22 blood group systems and 5 HPA genes using an amplicon-based next-generation sequencing (NGS) approach. As part of the test system, we focused on the JK system in more detail. Double-ARMS PCR analysis was performed for the haplotype phasing of the JK1/JK2 and two more common synonymous polymorphisms. We performed transcript analysis to detect potential alternative splice products. For a subset of samples, a comparison between serotype and red cell genotype was conducted. Allele frequencies were determined for geographically different panels of individuals. Results We successfully genotyped the JK blood group for 99.6% of the samples. Haplotype phasing revealed 96 different alleles. For several alleles that carry one of the synonymous SNVs c.588A>G and c.810G>A, we could not confirm the reported JK phenotypes. We found a higher frequency of JK:1 alleles for all populations except Iraqis. JK*01W.01 alleles were more common in the Asian groups and sub-Saharan Africans. A variant of the allele JK*02N.01 was present exclusively in Southeast Asians. Conclusion Genotyping for JK antigens with a targeted NGS assay can easily be performed in routine. The interpretation that c.588A>G leads to a weak phenotype and c.810G>A to a null phenotype is questionable. IDs as well as the descriptions of alleles carrying these SNVs should be revised in the ISBT JK table.


Introduction
The main task in transfusion medicine is to provide suitable blood products for each patient on time. This can be a challenge for people with foreign ethnic backgrounds, as they may have rare blood group phenotypes that do not match with those of people of Western European origin. The number of patients suffering from hemoglobinopathies such as sickle cell anemia or thalassemia who rely on a regular blood supply has increased significantly since 2015, when larger numbers of people from Arab countries, particularly Syria, Iran, Iraq, and the Maghreb countries, began migrating. In addition, communities from Turkey and Eastern European countries live in Germany. In order to characterize more specifically the variability of blood group genes in these populations, we designed a targeted amplicon-based next-generation sequencing (NGS) approach for genotyping 22 different blood group systems comprising 26 different genes among the 43 blood group systems currently recognized by the International Society of Blood Transfusion (ISBT) (http://www. isbtweb.org) [1]. In addition, multiple human platelet antigens are genotyped simultaneously (online suppl. Table  S1; see www.karger.com/doi/10.1159/000525326 for all online suppl. material). In this report, we focus our attention on the Kidd blood group system.
The JK blood group (JK; ISBT 009) system consists of three antigens: JK1, JK2 (or Jk a and Jk b ), and JK3. The JK1 and JK2 antigens result from a single-nucleotide variant (SNV) at c.838G>A (p.Asp280Asn, rs1058396). In addition, the ISBT lists more than 50 alleles that encode altered or suppressed expression of JK antigens (http:// www.isbtweb.org/fileadmin/user_upload/_ISBT_009__ JK_blood_group_alleles_v6.0_01-MAR-2020.pdf). The Kidd blood group system has been known as clinically important because antibodies to JK antigens can cause immediate hemolytic transfusion reactions and hemolytic disease in neonates. They are a common cause of delayed transfusion reactions [2]. About 12% of alloimmunized patients with sickle cell disease and 4% of patients with other diseases develop antibodies to JK antigens [3][4][5].
This study aimed to develop a targeted NGS approach for the analysis of different blood groups and platelet antigens. Here, we provide and discuss the results of a molecular study for the JK blood group system, including SNVs, haplotypes, alleles, and frequencies, in >8,000 individuals of various origins. In addition, an investigation of genotype-phenotype correlation was performed in a subset of samples.

Sample Cohort
The German Red Cross Blood Service recruited 7,564 individuals for blood donation who came mainly from Syria, Turkey, Iran, and Eastern European countries. The donors presented at regular blood collections at the German Red Cross Blood Service West in North-Rhine Westphalia. We further included 499 blood samples from a random cohort of blood donors from the Institute for Transplantation Diagnostics and Cell Therapeutics at the University Hospital Düsseldorf, Germany. Donors declared their country of origin in a voluntary manner (n = 5,829). 2,217 individuals did not provide information about their mother countries. Samples were grouped in various panels according to the geographical area of their mother countries to determine the respective allele frequencies (Table 1).

DNA Extraction
The white blood cells remaining in the blood bag after the separation of the erythrocytes were used as starting material. The samples were diluted at 1:2 with PBS. Subsequent DNA extraction was performed using the DNAQiamp 96 DNA Blood Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Deviating from the instructions, the lysis time was extended to 30 min.

NGS and Workflow
The entire set of fragments was amplified in four multiplex PCR reactions. The reaction mixture for each sample consisted of 5 μL of KAPA2G Fast Multiplex Mix (Kapa Biosystems, Cape Town, South Africa), 1 μM of each primer mix, and 20-50 ng of genomic DNA in a total volume of 10 μL. The PCR protocol was performed as follows: 95°C-3 min; 30 cycles of 95°C-15 s, 63°C-2 min, 72°C-45 s; 72°C-3 min. Further processing was done as described previously except that 190 samples were pooled [7].
For verification and proof of reproducibility, 95 samples were analyzed in three different NGS runs. Depth of coverage, quality of reads, and genotyping results were compared.

Evaluation Criteria
Runs that were included in this study had to pass the following criteria: cluster density between 800 and 1,300 k/mm 2 , more than 70% of clusters pass the filter, and Q30 score must be higher than 70%.

Data Analysis
Data analysis was performed using the previously described, self-developed, and accredited analysis software Blood Group_ Analyzer (Institute for Transplantation Diagnostics and Cell Therapeutics, University Hospital Düsseldorf, Düsseldorf, Germany) [8]. After an error correction step, the final allele determination was performed by forming intersections of allele groups for all eight exons. Thresholds for minimum coverage were determined empirically by comparing genotyping and serotyping results for the entire set of validation samples. The minimum coverage for a fragment was set at 80 in the case of a homozygous allele group and 30 for each allele in a heterozygous allele group. Genotyping results were accepted if the appropriate depth of coverage was reached for the entire set of fragments.

JK-Specific Analysis
Reference Sequences Enumeration of exons is based on the reference sequences for allele JK*01 (Acc. No. NG_011775.4 and NM_015865.7 for genomic DNA and mRNA, respectively).

Haplotype Phasing by Double Amplification Refractory Mutation System (Double-ARMS)
We performed a double-ARMS test to assign the two common SNVs c.588A>G and c.810G>A to JK*01 and JK*02. Similar to Lo et al. [7], we used allele-specific primer pairs to resolve the different haplotypes. The primer combinations of the haplotype-specific PCR reactions are listed in Table 2. Amplicon sizes of fragments from c.588 to c.810 and from c.810 to c.838 were 2,773 bp and 281 bp, respectively. A 506 bp fragment encompassing exon 11 of the ABCG2 gene was co-amplified in each reaction for amplification control. The reaction mixture contained 20-50 ng of genomic DNA, 0.5 μM of each JK primer, 0.05 μM of each control primer, 0.2 mM of each dNTP, 0.25 U LongAmp Taq, and 2 μL 5× Lon-gAmp Buffer (New England Biolabs, Frankfurt/Main, Germany) in a final volume of 10 μL. The following amplification conditions were used: 95°C-3 min; 3 cycles of 95°C-15 s, 70°C-20 s or 3.5 min; 3 cycles of 95°C-15 s, 68°C-45 s, 72°C-20 s or 3.5 min; 35 cycles of 95°C-15 s, 66°C-20 s or 3.5 min; 72°C-10 min. The elongation time of 3.5 min was used for amplification of the 2,776 bp amplicon of reaction A, and an elongation time of 20 s was used for the amplification of the 281 bp amplicon of reaction B. Phasing of other SNVs concerning JK1/JK2 polymorphism was inferred from the presence of samples homozygous for JK1 or JK2, obtained from the Erythrogene.com database [9] or analyzed in double-ARMS tests (data not shown).

JK Transcript Analysis
RNA was isolated from reticulocytes using the Maxwell CSC RNA Blood Kit (Promega, Walldorf, Germany) or the Quick RNA Whole Blood Kit (Zymo Research Europe, Freiburg, Germany). JK-cDNA was amplified with primers targeting the exon 5-exon 6 and the exon 8-exon 9 boundaries according to the following protocol. In a reaction volume of 10 μL, 50 ng RNA was mixed with 50 pmol sense and reverse primer, each. The mixture was incubated for 5 min at 70°C and 5 min at 20°C. After the addition of 5 μL Scriptum standard reaction mix and 0.4 μL Scriptum standard enzyme (both; Bio&Sell, Feucht, Germany), cDNA fragments were amplified at 95°C-5 min; 40 cycles of 95°C-10 s, 63°C-20 s, 72°C-1 min; 72°C-2 min. Amplicons were subjected to Sanger sequencing analysis using standard protocols. Primers for cDNA amplification and Sanger sequence analysis are given in Table 2.

Inference of JK RBC Phenotypes
For the inference of RBC phenotypes, alleles were classified into two different categories.
Nonreferenced Alleles For alleles carrying novel SNVs, we abandoned assuming a RBC phenotype because uncharacterized mutations might alter the antigenicity of the background allele.

Serotyping of JK Red Blood Cell Antigens
The presence of JK1 and JK2 antigens was identified by RBC phenotyping for 551 randomly selected samples. In addition, the agglutination intensity was determined in 251 samples.
The first serological assay was performed on erythrocytes using an automated system (Erytra, Grifols, Leipzig, Germany) with anti-Jk a and anti-Jk b monoclonal antibodies (MS-15 and MS-8, respectively). For a second determination, serotyping was performed using the standard tube agglutination method with polyclonal antibodies (ND-Diagnostik, Sinsheim, Germany, and Bio-Rad, Neuberg, Germany, respectively). According to the manufacturer's instructions, one drop of cell suspension was mixed with one drop of reagent. The mixture was centrifuged at 800 g for 1 min and incubated at 37°C for 15-30 min. Agglutination was counted as Jk a or Jk b antigen positivity. Positive and negative control cells and Coomb's control cells were used for quality controls.

Provisional Nomenclature of Novel Alleles
For nonreferenced alleles, a provisional nomenclature was chosen that unambiguously identifies the amino acid and nucleotide changes in the coding region as well as in intron regions (shown in Fig. 1).

Statistics
Absolute allele frequencies were determined by direct counting of alleles according to the Hardy-Weinberg ratio. For homozygous genotyping results, alleles were counted twice. The percentage allele frequencies were calculated from the number of homo-and heterozygous samples. Typing failures (no calls) were not considered in the frequency calculation.

Results
We conducted a targeted NGS analysis for the determination of multiple blood groups and platelet antigens in >8,000 blood donors from different origins. Because detailed results for the entire set of genes are voluminous and complex, further analysis results are described for the JK blood group system, only.

Results for the JK Blood Group System
Validation Sample Set We performed initial genotyping and serotyping for 551 randomly selected samples. These samples were used (1) to optimize the assay in terms of data quality and metrics, (2) to adapt the BloodGroup_Analyzer software for the automatic determination of genotypes, and (3) to validate the obtained genotyping results by direct comparison with erythrocyte serotypes.

Data Quality and Metrics
The mean depth of coverage varied between various JK amplicons (300.6-873.59 and 166.44-450.69 for homozygous and heterozygous samples, respectively) (online suppl. Table S2). In addition, depending on the quality of DNA, we observed low coverage for all amplicons for some samples. For heterozygous fragments, the proportion of both alleles was near 1:1 (online suppl. Table S2; online suppl. Fig. S1).
To prove reproducibility, a series of 95 samples were analyzed in three different NGS runs. The depths of coverage were comparable, and identical genotyping results were obtained (online suppl. Table S3; online suppl. Fig.  S1). Next, we trimmed the BloodGroup_Analyzer software for automatic SNV detection and allele determination. We analyzed the depth of coverage for heterozygous and homozygous fragments for the entire set of 551 validation samples and empirically determined the minimum depth of coverage required for reliable results. The minimum coverage was set to 80 in the case of a homozygous fragment and 30 for each allele in a heterozygous allele group.

Haplotype Phasing of SNVs
The haplotype phase of the observed SNVs was then determined by applying a double-ARMS approach (shown in Fig. 1a). The common SNVs c.588A>G (rs2298718) and c.810G>A (rs17675299) were present in 209 and 32 JK1/JK2 heterozygous samples, respectively. These SNVs are thought to attenuate or suppress JK expression regardless of the background allele. Eighteen samples were tested for haplophasing SNV c.588A>G to JK1 and JK2, and thirteen samples were tested for all three SNVs. All JK2 alleles showed linkage to c.588G. Among    the JK1 alleles, c.588A is more common than c.588G, except for alleles that also carry c.130G>A. The c.810G>A change was found exclusively in haplotype phasing to c.588G and c.838A.

Transcript Analysis
Because the SNV c.810G>A affects the penultimate position of exon 7, it might disturb proper splicing. Therefore, we performed a transcript analysis of four homozygous c.810A samples. Exons 6-8 of the JK transcript were amplified and subjected to Sanger sequencing analysis. We did not observe alternative splice products suggesting that a full-length JK transcript is present (shown in Fig. 2b).
Allele Determination Haplotype phasing resulted in a total number of sixteen different alleles present in the validation sample set. Of these, six alleles were already referenced by ISBT. The remaining alleles are variants of alleles JK*01, 01W.01, 02W.03, 02N.01, and 02N.17 and include an additional missense or synonymous variant (Table 3). For nonreferenced alleles, a provisional nomenclature was chosen that uniquely identifies the amino acid and nucleotide changes (shown in Fig. 1).

Comparison of Inferred RBC Phenotypes and Serotypes
We then compared the inferred RBC phenotypes with experimentally determined serotypes (Table 3). Only samples that were either heterozygous for JK1/JK2 or homozygous for a given allele were considered. In this way, phenotypes of fourteen different alleles could be analyzed. Agglutination intensity was used as a measure of the degree of antigen expression. Inferred phenotypes of JK:1, JK:1 weak , and JK:−2 were identical to serotypes for alleles JK*01, 01W.01 (±c.588G), and 02N.01.588A_G, respectively. The two synonymous SNVs c.588A>G (rs2298718) and c.810G>A (rs17675299) occur at high frequencies of 0.6016 and 0.0517, respectively. These variants were recorded to be associated with attenuated (c.588G) and absent (c.810A) JK antigen expression regardless of whether they are present on a JK1 or JK2 allele. For JK*01W.03 and 02W.06 alleles, which both carry the c.588G variant, diminished antigen expression would have been expected, while agglutination intensities of 3-4 were observed (5 and 305 samples, respectively). Alleles JK*02N.17 with c.810A and JK*01N.20 were both expected to be JK-seronegative but were typed as JK:2 and JK:1, respectively, with high antigen expression regardless of the homozygous or heterozygous presence of JK1/JK2. Altogether, neither both synonymous SNVs nor the amino acid and nucleotide changes present in allele JK*01N.20 appear to alter the antigenicity of the JK protein.  1 Divider 1 (*) indicates a result that was obtained using molecular typing methods. 2 The most similar referenced ISBT allele as published by the ISBT working party [1]. 3 Divider 2 is always a ".". It indicates that additional amino acid changes or nucleotide changes are present. 4 Amino acid changes are listed in consecutive order. They are termed in one-letter code with the position of amino acid in the middle. 5 Divider 3 "_" is used to separate additional amino acid changes or nucleotide changes in the cDNA or changes which are present in the intron. 6 Synonymous SNVs are termed as nucleotide position in cDNA immediately followed by wild-type nucleotide and changed nucleotide separated by "_". 7 In the case of SNVs in intron regions, the nearest exon is given followed by the distance to the exon in counts of nucleotide. +, Nucleotide change is downstream of the exon. −, Nucleotide change is upstream of the exon.  (Table 3). Overall, none of these nonreferenced SNVs appears to change the JK1/JK2 antigenicity. Since allele JK*01.L148F was found in combination with JK*01, we were not able to determine the RBC serotype.    Table 5. Frequencies of ISBT-referenced alleles and alleles that were detected at least five times in various blood donor panels

Prospective Data Set
Data Quality A cohort of 7,512 additional samples derived from blood donors of a different ancestry was analyzed using the NGS assay. Of these, 7,482 (99.6%) were successfully genotyped for the JK blood group system. The remaining 30 samples did not provide a clear genotyping result. Due to poor sample quality, one or more exons did not reach the required depth of coverage (Table 1).

General Sequencing Data
Observed SNVs Besides the SNVs already described for the validation sample set, we detected a large number of SNVs scattered throughout the coding region and adjacent parts of the 5′-UTR, introns, and 3′-UTR for both JK:1 and JK:2 alleles (Table 4).

Haplotype Phasing and Allele Determination
Haplotype phasing revealed 95 different alleles, including all alleles already discovered for the validation sample set, except JK*02N.17 (Table 4). Two missense SNVs were discovered in multiple alleles, p.Met167Val and p.Val175Ile (3 and 2 alleles, respectively). Although most variants have been reported previously in the gno-mAD database [10], 26 variants are described for the first time. These variant alleles were submitted to the Gen-Bank database (https://www.ncbi.nlm.nih.gov/gene). Accession numbers are given in Table 4.

Allele Frequencies in Various Blood Donor Panels
Genotyping data from the validation and prospective data sets were combined to determine allele frequencies.
A summary of genotyping results is reported in online supplementary Table S5.
The null allele JK*02N.01.588A_G was present only in Southeast Asians (Philippines, Thailand, and Vietnam). In addition, one Filipino was homozygous for this allele. In gnomAD and ALFA databases, this allele was found in 1-2% of Asians.

Discussion
Transfusion therapy remains the mainstay of treatment for patients with hemoglobinopathy. The development of red cell alloantibodies is a serious and common complication, as it increases the risk for hemolytic transfusion reactions and occasionally life-threatening events [12]. In addition, delays in the identification of compatible red cell units may occur. While only 2-5% of the general population develop allo-RBC-antibodies after transfusion therapy, the prevalence of alloimmunization for patients with sickle cell disease ranges from 5 to 75% [13][14][15].
To ensure transfusion safety, blood donor samples are routinely tested for the major antigens of the Rh, ABO, and Kell blood group systems. However, several clinically important alloantibodies to other high-frequency antigens have also been identified. Therefore, extended RBC antigen matching was recommended for patients at risk for alloimmunization [16]. Because extended blood group typing by serology is not applicable for high sample numbers, especially when rare typing sera are not available, prospective high-throughput blood donor screening for expanded blood group panels might improve the provision of blood products for patients with rare phenotypes or patients immunized against common antigens. In the present report, we used an amplicon-based NGS approach for genotyping several blood groups and platelet antigen genes including the Kidd blood group system. Allo-JK antibodies can cause immediate hemolytic transfusion reactions, neonatal hemolytic disease, and delayed transfusion reactions [2]. It was recommended to administer JK-matched red cell units to immunized patients [16]. Here, we turn the attention to JK blood group genotyping. We provide molecular data of >8,000 blood donors of various origins. Moreover, for a subset of samples, we performed genotype-phenotype correlation.

Discordance in the Description of Alleles -Consequences for Nomenclature
For a subset of 551 samples, we inferred the RBC phenotype based on the specifications given by ISBT in the current table for the JK blood group system (http://www. isbtweb.org/fileadmin/user_upload/_ISBT_009__JK_ blood_group_alleles_v6.0_01-MAR-2020.pdf) and compared the results to experimentally obtain serotypes. We demonstrated that neither of the frequent synonymous SNVs c.588A>G and c.810G>A lead to altered JK antigen expression. Moreover, we were able to exclude disruption of proper transcript splicing caused by c.810A as previously suspected [17,18]. Similarly, all samples carrying allele JK*01N.20 showed normal JK:1 expression. In addition, c.588G was present in all but one JK:2 expressing alleles. This nucleotide change was initially described to be present in the regular JK*02 allele [19]. Because the ISBT working party focuses on variants that result in a change of antigenicity, the c.588A>G change may intentionally be omitted from some allele descriptions. However, our results strongly suggest that the IDs and phenotype descriptions of alleles JK*01W.06, 01N.20, and 02N.17, as well as JK*01N. 19, which carry the same SNV as JK*02N.17, should be revised. In addition, since the c.588G variant is part of the regular JK*02 allele, JK*02W.03 should be removed and the nucleotide description of JK*02 should be revised.

Characterization of Novel JK Alleles
In the present study, we discovered a large number of rare SNVs. In 2016, the Erythrogene.com database was established, describing the full coding regions of blood group alleles discovered during the 1000 Genomes project [9]. This simplified haplotyping of some nonreferenced alleles but no unique allele names was given. HLA typing is routinely performed by NGS, and the sequencing results obtained are compared with a reference database in which alleles are recorded with unique IDs, regardless of whether protein characteristics are affected (IPD-IMGT/HLA database, available at https://www.ebi. ac.uk/ipd/imgt/hla/). Multiple nonreferenced alleles are often found during genotyping by sequence analysis assays. Here, a streamlined protocol for naming novel alleles and a database analogous to the IPD-IMGT/HLA database in which the nucleotide sequences of these alleles are recorded would be helpful. However, the clinical significance of those novel alleles is difficult to predict without clinical correlation. Further functional studies are needed to definitively determine whether a variant results in altered expression or encodes a previously undescribed blood group antigen.

Classification of Weak and Null Alleles by Targeted JK Genotyping
In our sample cohort, a large number of donors carried at least one weakly expressed or null allele (12.5% and 0.3%, respectively). In blood group genetics, the presence of alleles with attenuated or silenced antigen expression is a well-known problem. JK typing can be performed using serological RBC phenotyping or DNA-based molecular typing methods. When using antibody-based typing methods, weak antigen expression may be missed depending on the method protocol, reagents, or antibodies. In addition, serological typing of pre-transfused patients is frequently impossible [20]. A falsely negative typed blood unit may boost antibody levels in previously sensitized patients. On the other hand, molecular typing methods are often limited to the identification of the JK1/JK2 polymorphism [21][22][23][24][25][26][27]. In donor typing, the risk of falsepositive results is negligible as they are phenotypically homozygous. Therefore, they are unlikely to harm the recipient if transfused to a heterozygous patient.

Allele Frequencies in Different Populations
Consistent with several other reports and databases [10,11,[28][29][30][31][32], we found a nearly equivalent frequency of JK1 and JK2 alleles for most of the blood donor panels and an elevated frequency of JK1 alleles in sub-Saharan Africans confirming other studies [10,11,32]. Similarly, the higher frequencies of JK*01W.01 and its variants for Southeast and Eastern Asian populations as well as sub-Saharan Africans have been described before [10,11].
The frequency of the novel allele JK*02W.03.M167V was increased in the Eastern Asian group. However, we serotyped one JK*02W.03.M167V homozygous sample and four samples carrying this allele combined with a JK1 allele for JK:2 expression. Our results suggest that this missense variant does not have an impact on JK:2 antigen expression. DOI: 10.1159/000525326 Benefits for Blood Banking Services The described assay has the potential to allow for transfusion of matched blood products to the patient's phenotype. Thus, it could prevent adverse transfusion reactions leading to increased transfusion-related mortality. In addition, it will improve transfusion safety by identifying the most compatible blood products, thereby reducing the risk of alloimmunization. The method makes it possible to test a large number of donors without concerns of expense or lack of reagents, thus ensuring a sufficient supply of typed blood donors. Currently used DNA-based methods for the JK1/JK2 polymorphism will not identify the rare JK:−1,−2 blood donors because the mutations that cause loss of antigen expression are located in other parts of the SLC14A1 gene. In contrast, our assay is based on the analysis of the entire coding sequence including parts of the adjacent introns, thereby allowing for the identification of those rare JK-negative blood donors.
In addition, besides the high accuracy and high throughput, it has the advantage of large flexibility by combining the analysis of multiple clinical important red cell antigens. The failure rate of 0.37% across the entire sample set was very low, strongly suggesting the feasibility of this novel assay for routine diagnostics. As we did not find a major variance in JK1/JK2 allele frequencies between different panels of blood donors, except for sub-Saharan Africans, the supply of JK-matched blood products can be sufficiently performed by the blood banking services in European countries.

Limitations
This study has some limitations. Serotyping data are lacking for the majority of samples. Here, haplotype phasing was performed using SNV frequencies from databases. Although this strategy is helpful and accurate in most cases, it is risky in a clinical context when phenotyping data are not available. The clinical significance of new alleles is difficult to predict without clinical correlation. Further functional studies and appropriate RBC serotype identification are of particular importance to determine the impact of novel missense variants. In addition, because of the amplicon-based approach of the NGS assay, mutations present in parts of the SLC14A1 other than the coding region and adjacent intron sequences may be missed and may lead to attenuation or suppression of JK expression.

Conclusions
Overall, we present a targeted amplicon-based NGS workflow that is suitable for high-throughput genotyping of the Kidd blood group. Rare weak and null alleles that may be missed by currently used serotyping and molecular genotyping assays were reliably detected. The com-bined strategy of NGS-based genotyping and serological RBC phenotyping revealed conflicting expression characteristics for several ISBT-assigned alleles. Particularly, we show that the interpretation that c.588A>G leads to a weak phenotype and c.810G>A to a null phenotype is questionable. We thus recommend that the IDs, as well as nucleotide and phenotype descriptions of those alleles carrying one of these synonymous SNVs, be revised. A streamlined protocol for naming unreferenced alleles in conjunction with a detailed database of allele sequences and, if available, phenotype descriptions would be helpful for further studies.