Introduction

Whole-genome data has become more accessible thanks to techniques being made easier, the availability of sequencing machines or contractors, and the release of public data. Only a small part of these entire genomes are exploited beyond the scope of their initial purposes.

Amplicon-based next-generation sequencing (NGS) assays have in many ways laid the groundwork for whole-genome analyses as they require equivalent reagents, equipment and experimental skills. Much software for amplicon-based NGS has been developed, validated and certified in clinical fields. More particularly, most immunogenetics labs are equipped with amplicon-based NGS for HLA typing and some have also developed and validated such techniques for human platelet and RBC antigens [1–3].

Many questions can be explored in various fields thanks to WGS resources and their integrative investigation; first in population genetics where such data may improve understanding of natural selection, local adaptation, demographic history and early human migration [4,5]. Then in evolutionary genetics where it can address more fundamental issues such as gene evolution and functional investigation [4,5] thanks to haplotype reconstruction or the localization of new variants. Finally, these rapidly evolving techniques have now made their entry into analysis on an individual scale, for example in forensics and for clinical purposes [3,6].

However many issues raised by WGS handling limit the implementation of these techniques both in Research and Clinical laboratories working within regulatory approved frameworks e.g. Council of Europe (CE): software availability, database accuracy and editing, coverage and read depth quality indicators [5,7]. Thus, many whole-genome experiments designed for one scientific purpose are not used for any further analyses.

The aim of this study was to overcome these limitations and to estimate the feasibility of RBC antigen typing from WGS data. We analyzed whole-genome data from 79 individuals from Central Asia [8] for the highly polymorphic HLA-DRB1 gene and for 9 blood group antigen.

The same samples had previously been typed for HLA-DRB1 by amplicon-based monoallelic sequencing and for blood group bi-allelic polymorphisms using SNaPshot analysis [9].

WGS data was analyzed with software validated for HLA typing from NGS data [10]. This software relies on an allele alignments database; whereas the HLA system has a very convenient database and consensus on allele naming [11] with monthly updates, genetic polymorphism of RBC antigens are provided in Portable Document Format (pdf) and need to be converted. Most importantly, the software used in this study allows phasing of variable positions to define alleles or haplotypes. In a second analysis, the software was set up to search WGS data for new alleles. Indeed, previous investigations for blood groups but also for specific anthropogenic analyses revealed that this cohort presented a singular genetic mosaic of components from various geographic regions of Eurasian ancestry [12].

Materials and methods

Results

Blood group analyses

Blood group genotyping analyzed by SNaPshot are given in Table 1. Most analyses focused on one SNP leading to bi-allelic results, except for KEL, DO and FY systems for which 2 or 3 SNPs were analyzed. Most antigens displayed low or no allelic diversity except for DO (p.Asn265Asp), FY (p.Gly42Asp), JK (p.Asp280Asn) and YT (p.His353Asn).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Blood group typing by SNaPshot analysis.

https://doi.org/10.1371/journal.pone.0242168.t001

Group typing based on WGS analysis was performed targeting all of the variable positions described in S1 Table. Sixty-three alleles out of 1035 described by SNaPshot could not be resolved (6.1%) by WGS analysis. For all genes analyzed, typing resolution was associated with the number of reads mapped on their genetic sequence (S3 Table).

WGS-based typing showed 100% of concordance for homozygous SNPs analyzed by SNaPshot (N = 865 SNPs) and 95.3% for heterozygous positions (N = 102/107 SNPs; Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Blood group typing by WGS analysis.

https://doi.org/10.1371/journal.pone.0242168.t002

98.6% of WGS-based typing results were concordant with SNaPshot results for KEL (p.Met193Thr), one heterozygous sample was not correctly typed (KEL*02) and 5 samples remained unresolved. The monomorphic position KEL (p.Pro597Leu) was 100% concordant, 3 samples were unresolved.

100% of concordance was observed for the monomorphic positions FY (p.Arg89Cys) and FY -67T>C, 4 and 8 samples remained unresolved respectively.

98.4% of WGS-based typing results were concordant with SNaPshot results for FY (p.Gly42Asp); among the 30 heterozygous samples, 1 was typed FY*01. Six samples were not resolved.

98.6% of WGS-based results were concordant with SNaPshot results for JK (p.Asp280Asn), with 1 incorrect typing for a heterozygous sample (JK*02). Eight samples were not resolved.

100% of concordance was observed for the monomorphic positions DO (p.Gly108Val) and DO (p.Thr117Ile); 4 and 3 samples remained unresolved respectively. 98.6% of WGS-based results were concordant with SNaPshot results for DO (p.Asn265Asp), 1 heterozygous sample was incorrectly typed (DO*02). Four samples were not resolved.

100% of concordance was observed for YT (p.His353Asn) (9 samples were unresolved), IN (p.Pro46Arg) (2 samples were not resolved), CO (p.Ala45Val) (4 samples were not resolved,) and LW (p.Gln100Arg) (1 sample not resolved).

98.7% of concordance was observed for DI (p.Pro854Leu) with 1 incorrect typing for a heterozygous sample (DI*02); 2 samples remained unresolved.

WGS-based typing targeting all of the variable positions (described in S1 Table) led to ambiguities (described in S4 Table) but also to more precise typing. WGS analysis allowed typing of JK*01W.01 allele corresponding to JK:1^WK phenotype in 28 samples [19]; 10 samples were JK*02/JK*01W.01, 6 were JK*01/JK*01W.01 and 1 sample was homozygous for JK*01W.01. FY*02 allele associated with c.298G>A (p.Ala100Thr) was found in 18 samples [20]. No SNaPshot results were available to confirm or refute these typing results.

WGS data analysis also revealed polymorphisms that were unmapped in the ISBT database. A total of 267 previously unidentified polymorphisms covered with a minimum depth of 10x and observed in a minimum of 5 samples were found (S5 Table). Among these, 5 SNPs were in exonic regions but none led to amino-acid changes. Two SNPs in the DO gene were observed in 18 and 21 samples, 2 SNPs in IN were observed in respectively 37 and 41 samples and, in the JK gene, one SNP was observed in 37 individuals (S6 Table).

HLA-DRB1 analyses

Thirty-four HLA-DRB1 alleles were defined at maximum resolution by amplicon-based monoallelic sequencing, 5 samples could not be analyzed (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. HLA-DRB1 analysis by monoallelic sequencing.

https://doi.org/10.1371/journal.pone.0242168.t003

Ninety-one percent of WGS-based HLA-DRB1 typing, i.e. 135 out of 148 alleles, showed an exact match with typing defined by monoallelic sequencing at second field resolution. Most discordances were due to insufficient coverage and low read numbers leading to differences in 3^rd and 4^th digits; two samples (counting for 4 alleles) could not be typed.

No novel polymorphism could be detected in HLA-DRB1 during the second analysis of the WGS data.

Discussion

In this study we explored diploid markers in WGS data generated for Y-chromosome analysis from 79 individuals [8]. Analyses were performed with Polypheme software validated for HLA typing from NGS data [10] and set up for RBC analysis. HLA-DRB1 gene and 9 blood group antigens were typed (KEL, ACKR1 (FY), SLC14A1 (JK), ACHE (YT), ART4 (DO), AQP1 (CO), CD44 (IN), SLC4A1 (DI) and ICAM4 (LW)) according to standard nomenclature (IMGT 3.39.0 database [11], ISBT (http://www.isbtweb.org) and RBC antigens [3]). Whereas targeted strategies, such as PCR followed by sequencing or SnaPshot, circumvent specificity issues of genes with structural changes and hybrids such as RHCE/RHD and GPA/GPB; their analysis from WGS data have requires specific bioinformatic approaches including CNV (copy number variation) analysis. Therefore, such systems were not included in this study.

Our results showed that blood group typing deduced from WGS were correct at 99.5% compared to SNaPshot analysis (967 SNP correctly identified out of 972 typed); 93% when taking into account ambiguous typing. In a clinical or research context however, ambiguous RBC results need to be reanalyzed. HLA-DRB1 typing from WGS showed 91% of concordance with those obtained by amplicon-based monoallelic sequencing. These performances on RBC antigens were similar to those presented in a former study on WGS from donor data [3] which included the typing of highly complex genes such as MNS, RHD/RHCE and ABO systems.

WGS data quality is assessed by the estimation of read depth. A former study conducted on WGS data established a minimum of 15x for RBC antigen typing in the clinical field [3,14]. Here, mean read depth of the genome was estimated at 11.8x [5.5x-18.4x] and read depth for each gene reached higher values. For each gene, typing resolution was significantly associated with the number of reads mapped on its sequence and ambiguous and incorrect typing showed low numbers of reads corresponding to the missing allele and read depth equal to or below 15x. Our study thus confirms that RBC typing from WGS should be considered reliable with read depths strictly above 15x. To reach this goal, genome sequencing of one human (3Gb) should be analyzed with at least 45 Gb of data, here mean data was 34 Gb [16–53].

In our study, WGS data analysis allowed refined typing, identification of both potential new alleles and haplotypes as PolyPheMe software used here allowed phasing of polymorphisms subject to sufficient coverage and variable positions. We were able to type the JK*01W.01 allele [19] and the FY*02 allele associated with c.298G>A [20]. The weak JK allele may present a risk of hemolytic transfusion reactions [21] as it has been shown that among samples screened as JK:-1,-2, a fraction was JK:1^WK [22]. JK*01W.01 has been reported in Caucasian, Asian and Chinese individuals [19] but there is a lack of description of this allele among different populations. Given the frequency found here, our results strongly support the need of a better description of this allele, particularly in Asia.

Serological typing is the gold standard for blood group analysis but in particular situations molecular analysis can provide valuable information. In hematology laboratories, molecular biology based on sequence analysis was superseded by ready-to-use closed systems mainly based on SNPs analysis and validated for clinical purposes. Whereas unthinkable for routine patient care, some situations would gain from WGS such as screening donors for rare blood antigens and the management of RBC disorders. In this regards, our results showing rare and potential new alleles are particularly relevant in diseases such as Sickle Cell disease for example, where allo-immunization is a major complication [23]. Research of minor alleles and their potential role in allo-immunization in these patients would be a major advance in personalized medicine.

In a second analysis, WGS data were screened for new polymorphisms. 262 new variable positions in intronic regions and 5 polymorphisms in exons were identified, none led to non-synonymous mutations. An insight of their frequencies in populations described as being related to the Afghan population would contribute to refining their origins [12].

Molecular testing for the HLA system has been integrated in immunogenetics laboratories for a long time and evolves according to new technologies. Amplicon-based NGS is suitable for donor HLA typing, with robust and certified protocols, high throughput and highly resolutive typing results. These protocols can be performed with methods requiring several days and are also suitable for patients, for whom typing results are rarely impatiently awaited.

Immunogenetics laboratories are thus quite prepared to integrate WGS in their pipeline and use it to analyze other immune markers. Patients with auto-immune diseases, solid organ and HSC transplantation, or inflammatory diseases would benefit from personalized care with specific typing of non-classical HLA, FC receptors, KIR or LILRs [24–27]. In conclusion, the implementation of WGS can serve many purposes, from anthropogenic integrative studies to handling specific diseases in clinical fields.

Supporting information