Evidence that CD36 is expressed on red blood cells and constitutes a novel blood group system of clinical importance

Polymorphic molecules expressed on the surface of certain blood cells are traditionally categorized as blood groups and human platelet or neutrophil antigens. CD36 is widely considered a platelet antigen (Naka) and anti‐CD36 can cause foetal/neonatal alloimmune thrombocytopenia (FNAIT) in CD36‐negative pregnant women. CD36 is used as a marker of differentiation in early erythroid culture. During the experimental culture of CD34+ cells from random blood donors, we observed that one individual lacked CD36. We sought to investigate this observation further and determine if CD36 fulfils the International Society of Blood Transfusion criteria for becoming a blood group.


INTRODUCTION
The transmembrane glycoprotein CD36 (Figure 1) exhibits a broad distribution across various tissues.It is expressed in adipose tissue, mammary gland, epithelial cells, the membrane of placenta and many cancer cell lines [2].Additionally, it is expressed in haematopoietic cells, including monocytes, platelets, and during the maturation of erythroid progenitors to red blood cells (RBCs) [3].Individuals lacking CD36 may develop antibodies against the CD36 protein.The first case of CD36-related immunisation was reported from Japan, in a thrombocytopenic woman who became refractory to platelet transfusions [4].Her antibody, anti-Nak a , defined a platelet antigen with a prevalence of 97% among Japanese.Several family studies showed that Nak a antigen was inherited as an autosomal codominant trait [4] and Nak a was later shown to be carried on CD36 (Glycoprotein IV) [5].CD36 deficiency is divided into two groups: type I deficiency, which is characterized by lack of CD36 expression on platelets and all other cells [6,7] and type II, in which CD36 expression is lacking from platelets only [8,9].CD36 deficiency type I, while rare in Europeans, occurs with a prevalence of approximately 3% in Africans [10], 0.5%-1% among the Japanese [11,12] and 0.5% in China [13], as recently summarized by Xu et al. [14].
Type I deficiency is of clinical significance since it can be considered a null phenotype associated with the development of antibodies against CD36.These antibodies have been implicated in various clinically important conditions, including platelet transfusion refractoriness [4,15,16], post-transfusion purpura (PTP) [17], foetal-neonatal alloimmune thrombocytopenia (FNAIT) [18], and transfusion-related acute lung injury (TRALI) [19].Testing haematopoietic progenitor cell donors for CD36 type I has also been proposed due to the risk of incompatible transplantation [20].
Early work investigated the biochemistry, distribution and immunological characteristics of CD36 on different haematopoietic lineages in the adult and foetal settings [21,22].In line with this, CD36 now serves as a key cell surface marker in studies of human erythropoiesis using haematopoietic stem and progenitor cells (HSPCs).It is typically used in combination with CD34 to define the two earliest forms of erythroid progenitors, burst-forming unit-erythroid (BFU-E) and colony-forming unit-erythroid (CFU-E) [23].During the culture of HSPCs obtained from adult peripheral blood or other sources, CD36 expression increases progressively and reaches its highest level in early erythroblasts at the end of the erythroid expansion phase in vitro.Subsequently, CD36 decreases during the terminal stages of erythroid differentiation [24].
During erythropoiesis culture experiments performed in our laboratory using peripheral HSPCs from random blood donors, we observed that one donor did not express the CD36 marker as otherwise expected, making it challenging to characterize the early stages of erythroid development.We therefore wanted to investigate the genetic basis behind the absence of CD36, which lasted throughout the culture but did not appear to affect erythroblast maturation compared with other donors positive for CD36.Since anti-CD36 is known to cause platelet-related clinical effects, we asked the question if the presence of CD36 on various stages of erythroid cells may be targets for these antibodies in a similar way known for blood groups expressed on early during erythropoiesis.Foetal anaemia due to suppression of erythroid progenitors has been found to be caused by immunisation against the KEL, GE, MNS and JR blood groups [25].Indeed, reports implicating anti-CD36 in cases of hydrops fetalis and severe foetal anaemia have been described [26,27].Taken together, CD36 appears to have certain properties resembling a blood group molecule but the controversy regarding its presence or not on mature RBCs challenges the current dogma of how a blood group antigen is defined.Based on in vitro erythroid-culturing results from a blood donor that turned out to lack CD36, we performed additional experiments and scrutinized available datasets to evaluate if CD36 formally fulfils International Society of Blood Transfusion (ISBT) requirements to become a blood group system.

In silico analysis of CD36 expression
HSPC and other haematopoietic cell microarray data were obtained from the Bloodspot website (https://servers.binf.ku.dk/bloodspot) [3] using the normal human haematopoiesis (DMAP) dataset [28].RNA and protein expression data in erythroid cells from cultured HSPCs were downloaded from the supplement of Gautier et al. [24].A search for peptides derived from the seven glycoproteins that bear the 35 reported human platelets antigens (HPA) listed in the HPA gene database (https://www.versiti.org/products-services/human-plateletantigen-hpa-database/hpa-gene-database)was performed on the proteomics data from reticulocytes and RBCs, both membrane and soluble fractions [29,30].

Blood samples and ethics
RBCs, peripheral blood mononuclear cells (PBMCs) and platelets from donated whole blood units prepared in the Reveos automated blood processing system were obtained from anonymized leucocyte waste donor were cultured in a three-phase erythroid culture system and maintained at a concentration of 1-2 Â 10 6 cells/mL throughout the culture as described previously [31].FACSDiva software (v8.0-9.0,BD) was used for acquisition and FlowJo (FlowJo, LLC) for analysis of all the data.

RNA isolation and molecular analysis
Total RNA was extracted from cultured erythroblasts using the Mini-RNeasy Kit (Qiagen) following the manufacturer's protocol.Genomic DNA was removed using DNase I on-column digestion (Qiagen), then complementary DNA from the isolated RNAs was synthesized using Superscript IV VILO Master Mix (Thermo Fisher Scientific).CD36 primers encompassing exons 2-14 (CD36_cDNA22_43: 5 0 CCTGC AGAATACCATTTGATCC3 0 ; CD36_cDNA1771_1752: 5 0 TTGGCCA CCCAGAAACCAAT3 0 ) were designed using Primer BLAST based on the reference NCBI transcript NM_001001548.

Erythroid culture reveals a CD36-deficient blood donor
During routine culture of HSPCs towards erythroid differentiation (Figure 2a) from random, anonymized blood donors, we observed a donor that appeared to exhibit a complete lack of CD36 expression at all stages of erythropoiesis.This finding was confirmed by parallel culture of HSPCs from a CD36-expressing donor with the CD36deficient donor, here designated donors 1 and 2, respectively (Figure 2a).This finding presented a challenging situation as it made it difficult for us to separate the erythroid progenitor cells BFU-Es and CFU-Es based on their cell surface antigen expression [23] (Figure 2aday 7).However, donor 2 cells showed an otherwise unremarkable antigen profile as maturation progressed (Figure 2a-days 14-21).
Moreover, the CD36-negative cells throughout the erythroid culture (Figure 3a) underwent normal erythroid commitment by upregulating We also investigated the CD36 expression on platelets from both donors and showed that while the platelets from donor 1 expressed the expected, high levels of CD36, those of donor 2 lacked CD36 completely, consistent with the in vitro culture results (Figure 2b,c).

Genetic basis of CD36 deficiency in donor 2
DNA sequencing showed that donor 2 was homozygous for the SNV c.1133G>T (rs146027667) in exon 12 of the CD36 gene on chromosome 7 (Figure 4a), encoding p.Gly378Val (Figure 4b).This SNV is predicted to have a deleterious effect based on scores from both Sorting Intolerant From Tolerant (SIFT) [32] and Polymorphism Phenotyping v2 (PolyPhen-2) algorithms [33] (Figure 4c).Analysis of gnomAD frequencies showed that this SNV is common in individuals originating from Middle Eastern countries and those of Latino/Admixed American origin (Table 1) [34].The fact that CD36 is missing on both platelets and erythroblasts in donor 2 due to a germline variant previously implicated in CD36 deficiency in two other individuals [15], indicates that this variant indeed leads to CD36 deficiency but the latter study did not investigate CD36 on other cells than platelets.However, in line with the study by Toba et al.where all type I donors had CD36-negative erythroblasts [35], we conclude that the CD36

CD36 has broad tissue distribution including erythroid cells
Analysis of publicly available mRNA and protein datasets showed that CD36 expression has a diverse tissue distribution (www. proteomicsdb.org)[36] and almost all haematopoietic cells express CD36 at different levels.Notably, it is significantly higher in the myeloid, megakaryocyte and erythroid lineages (Figure S1A).When differentiating HSPCs towards the erythroid lineage, the expression of CD36 is upregulated during culture and reaches its highest level around day 14 and then is downregulated during terminal erythroid differentiation (Figure 3a).In line with these data from our erythroid culture, similar expression patterns are observed for CD36-mRNA and CD36 protein at different stages of differentiation in other erythroid culture systems [24] (Figure S1B,C).

CD36 is expressed on reticulocytes and erythrocytes in peripheral blood
Currently, there is a controversy based on the limited data available regarding whether CD36 is expressed on the RBC surface or not.To further investigate the expression of CD36 on erythroid cells in peripheral blood, we conducted a study using 20 peripheral blood samples from random blood donors.To avoid a potentially false positive result due to platelets contaminating the evaluated erythroid cell population, we stained cells from peripheral blood with a combination of erythroid-and platelet-specific cell surface markers, GPA and CD61, respectively, then gated the cells as shown in Figure 5a.Interestingly, we identified and excluded from the analysis gate a subset of cells double-positive for these two markers (Figure 5a).Their size resembles that of normal RBCs and shares characteristics with both RBCs and platelets (Figure S2), indicating that they may represent platelets adhering to RBCs.Gating on the GPA + CD61-cells and adding a dye for RNA content further enabled the separation of reticulocytes from mature RBCs (Figure 5a).By using this gating strategy, we demonstrated that CD36 is highly expressed on platelets as expected (Figure 5b,c).Moreover, we could show a distinct but significant right shift both on reticulocytes and RBCs, indicating that CD36 is indeed expressed on these cells, albeit at low levels (Figure 5b,c).
In line with this, mass spectrometric approaches have been used to assess the overall proteome of highly purified reticulocytes and mature RBCs [29,30].We analysed these two datasets for the presence of CD36-derived peptides.One study identified six unique peptides specific for CD36 and found approximately five times higher levels of CD36-derived peptides in reticulocytes compared with RBCs.Notably, CD36 peptides were only found in the membrane fractions but not in the soluble phase [29].The other study also reported six unique peptides from CD36 in mature RBCs, the vast majority in the membrane fraction of white ghosts [30].Importantly, we analysed these two datasets for the presence of peptides corresponding to the seven reported HPA-carrying proteins but were unable to detect any such peptides, suggesting the data are indeed erythroid-specific.In summary, our investigation clearly shows that CD36 is present in the reticulocyte and RBC membrane.

CD36 fulfils the ISBT criteria to become a blood group system
Based on experimental data from this study, as well as analysis of data from the literature, we asked the question if CD36 fulfils the formal ISBT requirements to form a novel blood group system.As shown in Table 2, we conclude that this is now the case.

DISCUSSION
With every challenge comes opportunity.We encountered a donor lacking CD36 expression on erythroid progenitor cells, which provided us with a unique opportunity to address a question that is both simple yet controversial in the field of transfusion medicine.Can CD36 be considered a blood group antigen?
Traditionally, blood group antigens are defined by polymorphisms found on RBC surface molecules.Some, such as RH and MNS, antigens are mainly erythroid-specific, while ABO and KEL show broader tissue distribution [40].Not unexpectedly, some blood group systems are shared between blood lineages.The Choline Transporter-Like protein 2 (CTL2; SLC44A2) is such an example that both underlies a blood group system on RBCs [41] and carries the HNA-3 antigens on neutrophils [42,43].Furthermore, several blood-group-carrying glycoproteins are expressed early during erythropoiesis [44,45].For example, antibodies targeting blood group antigens in the KEL and GE systems can cause antibody-mediated anaemia in the foetus and may even result in fatal hydrops [25].In such cases, antibodies target the erythroid progenitors, which leads to suppression of foetal erythropoiesis and decreased RBC production.Since onset of CD36 expression occurs at the CFU-E stage, a similar mechanism is at play in pregnancies complicated by maternal anti-CD36, which could lead to suppressed erythropoiesis [26].Reviewing the literature on anti-CD36, we found that it not only causes platelet-related disease including FNAIT but was also reported as implicated in several cases of foetal hydrops in pregnant women lacking CD36 [26,27].
We show in Table 2 that CD36 fulfils the current criteria for a new blood group system according to the ISBT Working Party for Red Cell Immunogenetics and Blood Group Terminology (WP-RCIBGT).As discussed above, antibodies have long been known to occur in CD36 null individuals.Whether these antibodies are defined as alloantibodies or isoantibodies is less relevant to the WP-RCIBGT because existing blood group systems display examples of both.In analogy with CD36, antibodies against the A, B and RhD antigens are produced in the absence of the corresponding antigen carrier molecules, which has not prevented them from becoming the archetypes among blood groups.Additionally, the genetics underlying the absence of CD36 are already well-defined, including the proven independence from the other three blood group loci currently residing on chromosome 7.Thus, CD36 meets all the ISBT criteria regarding inheritance and being an independent locus (Figure 4a and Table 2).
The categorisation of donor 2 as likely CD36-deficient type I relies here on the absence of CD36 from both platelets and cells of various stages of erythroid differentiation.Traditionally, lack of CD36 on monocytes has been used to define type I, partially due to their easy access in peripheral blood.Since monocytes were not available to us once we realized donor 2 lacked CD36, this can be considered a weakness of our study.It should, however, be noted that the classification of CD36 as a new blood group system does not depend on the definition of the deficiency type status of this donor.
Interestingly, the ISBT criteria have assumed but do not formally require a blood group antigen to be present on RBCs.This becomes particularly challenging in a case like CD36 where earlier erythroid cell stages express the glycoprotein at high levels while it has been difficult to detect on RBCs and reticulocytes in peripheral blood.In this study, we demonstrate that CD36 is expressed at low but significant levels on both reticulocytes and RBCs (Figure 5).The WP-RCIBGT may have to revisit its criteria regarding how much and where blood group antigens need to be expressed.Nevertheless, the WP-RCIBGT approved our proposal to make CD36 a new blood group system on 17 June 2023, substantiated not only by our experimental work [46]   that resulted in acknowledgement of CD36 as a new blood group system.
T A B L E 2 Criteria for the establishment of new blood group systems.

ISBT criteria a Yes/no Evidence
The antigen must be defined by a human alloantibody.b Yes Multiple examples known incl.antibodies made by pregnant women [37][38][39] and causing foetal anaemia [26,27] The antigen must be an inherited character.

Yes
CD36 well known to be inherited [26,27] The gene encoding it must have been identified and sequenced.

Yes
Multiple variants known to cause the CD36 null phenotype (Figure 4a) [39] Its chromosomal location must be known.

Yes
CD36 is located at 7q21.11 (Figure 4a) The gene must be different from, and not a closelylinked homologue of all other genes encoding antigens of existing blood group systems.
Yes CD36 lacks significant homology with other blood group molecules.The three other blood group genes on chromosome 7 are distant and lack homology (Figure 4a).

Potential future ISBT criterion c
The antigen should be expressed on erythroid cells normally present in the peripheral circulation Yes CD36 is expressed on maturing reticulocytes and red blood cells (Figure 5a-c) [29,30] a According to the ISBT Working Party of Red Cell Immunogenetics and Blood Group Terminology (WP-RCIBGT) (https://www.isbtweb.org/isbtworking-parties/rcibgt/blood-group-terminology.html).b The ISBT WP-RCIBGT defines this as antibodies against an antigen the individual lacks.
This additional criterion has been discussed by the Working Party on 17 June 2023, at the ISBT congress in Gothenburg and a revision will be formulated for approval at an upcoming Working Party meeting, along with revision of the antibody criterion discussed in the footnote above.
CD36 AS A NEW BLOOD GROUP
GPA and transferrin receptor (CD71) (Figure3b-day 14).Donor 2 cells also underwent apparently normal terminal erythroid differentiation later in the culture, as characterized by the expected downregulation of CD49d and upregulation of band 3, the latter being a marker specifically associated with the maturing erythroblast (Figure3b-days 18-21).These results suggest that lack of CD36 antigen expression does not affect human erythroid cell differentiation in vitro, and the absence of CD36 did not have a detectable impact on upregulating erythroid-specific cell surface markers during erythroblast maturation.

2
Lack of CD36 expression on HSPCs, erythroid cells and platelets.(a) Flow cytometry pseudocolor analysis plots showing CD36-positive cells (donor 1, top panel) and CD36-negative cells (donor 2, bottom panel) during erythroid culture (days 7-21).(BFU-E), (CFU-E), Pro-Erythroblast (Pro-E) and Early-Late-Erythroblast (E-L-Erythroblast).The timeline at the top represents the three stages of the erythropoiesis culture system used in this study.(b) Flow cytometry pseudocolor analysis plots showing the gating strategy and lack of CD36 expression on platelets from donor 2 (bottom panel) compared to those of donor 1 (top panel).(c) A histogram of platelets from donors 1 and 2 stained with anti-CD36 shows the complete absence of CD36 on donor 2 cells, comparable to the unstained control, while donor 1 cells are clearly CD36-positive.deficiency in donor 2 is most likely of type I, despite the absence of monocyte data.

F I G U R E 3 4
Lack of CD36 expression on HSPCs did not alter the erythroid lineage commitments nor the expression of erythroid-specific cell surface markers.(a) CD36 expression histogram overlay of donor 1 and donor 2 during the erythroid culture.(b) Flow cytometry pseudocolor plots showing CD36-negative donor cells (bottom panel) are differentiating towards erythroid cells like CD36-positive donor cells (top panel).CFU-E, (I) pro-erythroblast, (II) early-erythroblast, (III) late-erythroblast and (IV) orthochromatic-erythroblast.Genetic analysis of CD36 in donor 2. (a) Chromosomal location of CD36: the gene is distant from other known blood group genes on chromosome 7.(b) Sanger sequencing of CD36 transcripts from donor 2 revealed homozygosity for the SNV c.1133G>T in exon 12. (c) This SNV is predicted to have a deleterious effect based on SIFT and Polyphen-2 scores.

F
I G U R E 5 CD36 expression on reticulocytes and RBCs.(a) Pseudocolor plot showing the gating strategy used to identify platelets, reticulocytes and RBCs from peripheral blood samples using a combination of anti-GPA, anti-CD61 and thiazole orange.GPA + CD61À gated cells were further fractionated using thiazole orange, positive (reticulocytes) versus negative (RBCs).(b) Representative histograms displaying CD36 expression on platelets, reticulocytes and RBCs.(c) Bar graphs showing CD36 expression as mean fluorescence intensity (MFI) for the different cell populations.Unstained and isotype-stained cells were included as negative controls.Unpaired t-test analysis was used for statistical analysis (**** denotes p < 0.0001).
[26,27]o by a case report by Canals et al. in which anti-CD36 was reported to cause a weak panagglutination in routine RBC antibody testing.reportedpreviously,particularly from Asian countries where most cases of anti-CD36 have been identified.It appears to stand in stark contrast to the serious erythroid consequences reported by Japanese and Chinese groups[26,27].We can only speculate about the reasons behind this apparent discrepancy, but it is possible that In conclusion, we encountered a CD36-negative blood donor that prompted us to conduct a study including in vitro experiments, literature review and mining of publicly available proteomics data