SID: a new carbohydrate blood group system based on a well-characterized but still mysterious antigen of great pathophysiologic interest

In 1967, two independent studies by Macvie et al.¹ and Renton et al.² described the same distinctive agglutination pattern of small agglutinates in a sea of free red blood cells (RBCs). The defined antibodies had repeatedly been causing incompatibility in crossmatch tests since the mid-1950s. There was a challenge in distinguishing the weakly positive from negative RBCs, and those individuals with positive RBCs showed a wide variation of antigen strength. The RBCs of 1 percent of the population were strongly agglutinated, while 80 percent were more moderate, and 10 percent were weakly reactive or negative. The antigen was named Sd^a after the first known donor (S.S. or Sid) with strongly reacting RBCs. The system included the Sd(a+) and Sd(a−) phenotypes.^1,2 In 1971, anti-Sd^a was associated with the rare Cad phenotype,³ which was first noted in 1962 by Ikuta and Mukarami⁴ and later defined in three individuals of a Mauritian family of Indian origin as a polyagglutinable feature.⁵ Cad RBCs are strongly agglutinated by anti-Sd^a, which explains the alternative name Sd(a++) or super-Sid.³

The International Society of Blood Transfusion (ISBT) Working Party for Red Cell Immunogenetics and Blood Group Terminology acknowledged the SID blood group system as number 038 in 2019. This designation was possible thanks to genetic studies of Sd(a−) individuals that identified alterations (Table 1) in B4GALNT2 on chromosome 17^7,8 and thereby confirmed the previously suggested genetic origin of Sd^a. The system only includes this one antigen, number 038001, previously denoted number 901012 in the 901 series of high-prevalence antigens.⁶ It is uncertain whether the related Cad phenotype (RBCs highly reactive with anti-Sd^a) only represents a quantitative difference of the antigen or if qualitative/structural aspects of the antigen are also altered, or indeed both. Therefore, it is possible that (an)other related antigen(s) could be part of this system. However, there is currently no evidence that the Cad phenotype actually depends on the same locus, B4GALNT2.

Anti-Sd^a is naturally occurring and reacts well at room temperature and to a lesser extent at 37°C. The antibodies are of immunoglobulin M (IgM) type, although IgG and a rise of this titer after transfusion has been described.^1,2,10 Although the antibodies are generally regarded as clinically insignificant, hemolytic transfusion reactions have occurred in cases where the donor has a strong expression of Sd^a.^11,12 To avoid hemolytic transfusion reactions induced by anti-Sd^a, a donor–patient crossmatch test would suffice. Unfortunately, it is difficult to detect anti-Sd^a in a laboratory setup with antibody screening test RBCs, since these RBCs are usually not high expressers of Sd^a. Furthermore, gel-based tests often used today are not optimal for detection of the unusual agglutination pattern of anti-Sd^a. Additionally, donors are generally not phenotyped for Sd^a, and therefore donors whose RBCs are highly reactive with anti-Sd^a are not usually recognized and recorded.

As mentioned earlier, approximately 90 percent of individuals of European ancestry (Table 2) carry Sd^a on at least a portion of their RBCs. Remarkably, there are only 2–4 percent of individuals who are truly negative for the antigen when measuring its presence in other tissues as well (see Biochemistry section).^13,14,20 Sd^a is lacking on RBCs of the fetus and newborn but is detected on RBCs from babies at age 7–10 months.^1,2 Furthermore, the Sd(a−) RBC phenotype is more common among pregnant women compared with normal blood donors, and possibly the number increases closer to term.^1,10,13 This finding is a feature that Sd^a shares with Le^a and Le^b in the Lewis system and may at least partially be attributed to increased blood volume during pregnancy.²¹ An additional similarity with the Lewis system is the indication that Sd^a is not of erythroid origin but rather is passively adsorbed. The gene underlying the production of the enzyme that synthesizes Sd^a, B4GALNT2 (described in more detail later), does not seem to be expressed in erythroid precursor cells.²² However, adsorption of the carrier structure onto the RBC membrane from plasma, like that seen for the antigens of the Lewis system,²³ has not been possible to prove experimentally.¹³ Interestingly, the antibodies in both systems can be neutralized by addition of body fluids. In the case of Lewis, plasma and isotonic saliva can be used; in the case of anti-Sd^a, dialyzed urine from humans, or preferably guinea pigs, is useful.²⁴

The Cad phenotype [sometimes referred to as Sd(a++) or super-Sid] is defined as RBCs highly reactive with anti-Sd^a. The name Cad was assigned in a study by Cazal et al.⁵ in 1968 and characterizes the phenotype in a Mauritian family of Indian origin. However, the phenotype was first recognized in 1962 by Japanese investigators, followed by others.^4,17,25 Cazal et al.⁵ found RBCs from a group B donor reacted with Dolichos biflorus agglutinin (DBA), a lectin otherwise used as an anti-A1 reagent. The investigation of blood samples from family members of the original donor distinguished the phenotype as polyagglutinable and hereditary. Blood from 250,000 other group O and B individuals was examined over time, but none were reactive with the DBA, classifying Cad as an extremely low-prevalence phenotype.⁵ The phenotype was differentiated from T and Tn, which are also polyagglutinable.²⁶ In 1971, Sanger et al.³ proclaimed the similarities with Sd^a. The Cad RBCs from five different donors were found to be highly reactive with anti-Sd^a, and much more so than the strongest Sd^a RBCs identified so far.³ Further studies by Cazal et al.¹⁵ distinguished Cad RBCs to vary in reactivity and detected polyagglutination only with the strongest reactors.

With the use of DBA and Gastropoda species extracts, additional individuals with the Cad phenotype were identified, and, in cross-absorption experiments, three subtypes were suggested. Cad 1 was the originally described phenotype, while the non-polyagglutinable Cad 2 and Cad 3 were found in Asians and Europeans, respectively. Reactivity with DBA absorbed by A₁ RBCs distinguished Cad 2 from Cad 3. Among individuals of Asian descent, this finding was found to be a more frequent feature.¹⁵ In the Thai population, the prevalence was estimated at 0.26 percent by phenotyping of group O and B individuals with DBA¹⁹ (Table 2). Among a Japanese population, the prevalence of Cad was 0.03 percent, assessed in donors of all ABO blood groups using DBA for group O and B donors, and anti-Cad (chicken immune serum) for group A and AB donors. Family studies settled an inheritance of autosomal-dominant character. There was no polyagglutinability with any of the Cad samples tested, but both Cad 2 and Cad 3 subtypes were represented, determined as previously described. Furthermore, the agglutination pattern of the RBCs, with both DBA and anti-Cad (chicken immune serum), displayed a mixture of agglutinates and free RBCs. Cad eluates could then be obtained from the agglutinated fraction but not the free RBCs, suggesting an uneven antigen distribution within the RBC population. The mixed agglutination pattern and high reactivity with anti-Sd^a, assessed in the individuals of one family, further settled the connection between Sd^a and Cad.¹⁷

In 1974, Sringarm et al.²⁷ studied the sub-phenotype Cad 2 in donors of Thai and Chinese origin with DBA. The dominant inheritance was confirmed, as well as the relationship to Sd^a. However, weak polyagglutinability was also detected, a feature previously only assigned to subtype Cad 1.²⁷ A large-screen study of group O and B donors in four French blood transfusion centers, using DBA, determined the mean prevalence to be 0.07 percent. Most of the Cad samples were polyagglutinable in a sensitive method with group AB sera. The agglutination strength with a selected group AB serum varied greatly between individuals, even within families.¹⁶ The French screen also provided sample material from 50 individuals with the Cad phenotype (29 families) for extended studies using DBA, Helix pomatia agglutinin (HPA), anti-Cad (chicken immune serum), and anti-Cad (group AB sera). Lopez et al.²⁵ detected polyagglutinability with all but one Cad sample. The mixed-field agglutination pattern was again described, and the portion of agglutinated RBCs varied greatly, from approximately 5 to 100 percent. The authors explained the discrepancy between studies (detecting polyagglutinability or not) by the different methods used. Furthermore, varied antigenic expression (seen even within families) made the authors question whether the subtypes really reflected the genetic background.²⁵ In Canada, a DBA screen involving blood group O and B individuals identified two individuals with Cad phenotype (0.14%). One of the two also carried the rare Wr^a antigen, belonging to the Diego blood group system and carried on the erythroid band 3 protein anion transporter (Band 3). However, further studies involving family members did not show a linkage at a significant level. Another feature suggested to coincide with Cad in the family was weak glutamic-pyruvic transaminase activity.¹⁸ The connection with Wr^a was further contradicted in a study on a weak A phenotype involving four Cad individuals. None of the individuals’ RBCs reacted with anti-Wr^a.²⁸ In 1977, Cazal et al.²⁹ reviewed the Cad phenotype and elaborated on the subtypes. They grouped Cad 1 and Cad 2 as Cad strong (Cad s). Cad weak (Cad w) comprised Cad 3 and Cad 4, with Cad 4 describing RBCs less reactive to HPA when compared with Cad 3. The subtypes describe a falling reactivity strength (Cad 1 > Cad 2 > Cad 3 > Cad 4). Within the three families studied, a consistency of antigenic strength was seen with members of the same family.²⁹ A case report from 1996 on a French donor from Bermuda with polyagglutinable RBCs describes yet another Cad subtype. The high incidence of reactive sera (50–70%) and serology suggested a strong form of Cad. However, reactivity with lectins did not mimic what had previously been described for strong Cad subtypes. The subtype therefore failed to fit into the subtype scale of falling reactivity strength; therefore, the name Cad_Ber was suggested. It is noteworthy that this donor was also Wr(a+).³⁰

Sd^a is a glycan structure, and initial studies of the Cad phenotype indicated that a terminal GalNAc moiety was involved in the antigen structure. GalNAc binding lectins such as DBA, HPA, or lectins from Helix aspersa and Wisteria sensis all agglutinate Cad RBCs.^5,15,31 The antigenic structures of Sd^a and Cad have later been assessed in different types of human tissue, with the common terminal trisaccharide being identified as GalNAcβ1-4(NeuAcα2-3)Gal-R. In Cad RBCs, the epitope has been defined on glycophorin A (GPA) on O-GalNAc core 1 glycans (there were also indications of glycophorin B as a carrier). In addition, Cad was found on glycolipids as elongated paragloboside in the neolacto-series.^32,33 In a recent study on Cad RBC glycoproteins, we also found the epitope on the N-glycan of Band 3 and O-GalNAc glycan of equilibrative nucleoside transporter 1 (ENT1).⁹ However, when investigating RBC membranes with the Sd(a+) phenotype, the epitope was not detected.^9,34 GPA, Band 3, and ENT1 are all abundant RBC proteins and known carriers of protein blood group antigens within the MNS, Diego, and Augustine systems, respectively.^24,35 In human urine from Sd(a+) individuals, the epitope is carried on N-glycans of the abundant uromodulin, also known as Tamm-Horsfall glycoprotein.^36–38 In fact, the presence of the Sd^a glycan influences uromodulin levels in serum, suggested by increased levels in Sd(a−) genotyped individuals (rs7224888T>C, see Genetics and Molecular Basis section).³⁹ Sd^a is a major structural feature in the human descending colon. The antigen carriers in the colon are core 3 O-GalNAc glycans on the highly glycosylated mucins.⁴⁰ A summary of the carrier molecules and their respective biochemically characterized glycoconjugates are shown in Figure 1.

Fig. 1

The Sd^a synthase is a 4-β-N-acetylgalactosaminyltrans-ferase (β1,4GalNAc-T). Sd^a synthase uses the donor substrate UDP-GalNAc and the acceptor group NeuAcα2-3Gal-R. The different Sd^a carrier structures provide evidence on how the synthase can work on a variation of substrates, both O-GalNAc and N-glycans, as well as sialylparagloboside. Depending on the tissue source, the synthase displays some variation in preferred substrate, although the sialylated precursor is a common requirement.^41–44 Sd^a synthase (β1,4GalNAc-T2) is a homologue of β1,4GalNAc-T1, which synthesizes the GM2 epitope from GM3. GM2 is a glycosphingolipid in the ganglioside series and is composed of an identical terminal trisaccharide as the Sd^a/Cad epitope. However, Sd^a synthase is not able to use the precursor GM3 and make GM2.⁴⁵ The β1,4GalNAc-T1 structure was evaluated in its soluble form and has cysteine residues involved in both intra- and interprotein binding, with the latter in homodimer formation.⁴⁶ The cysteines are evolutionarily conserved between the two homologues, and Sd^a synthase also forms dimers, recently shown in a study of the transiently expressed transferase.⁹ Two isoforms of Sd^a synthase have been examined, of which one carries an extended cytoplasmic tail usually not seen in glycosyltransferases. In vitro, this extended cytoplasmic tail has been shown to alter its cellular location after Golgi processing, although the significance of this is unclear.⁴⁷

A transcript coding for β1,4GalNAc-T that synthesizes the epitope was partially cloned in 1996,⁴⁸ and, subsequently, two groups cloned the whole gene B4GALNT2 on chromosome 17 in 2003.^49,50 The theory that the product β1,4GalNAc-T2 holds Sd^a synthase activity was based on a number of arguments: (1) there are sequence similarities with the orthologue and homologue, murine B4galnt2 and human B4GALNT1, respectively; (2) the human tissues where the gene is expressed coincide with Sd^a locations; and (3) the glycosyltransferase had the expected substrate specificities. However, the genetics and molecular mechanism underlying the Sd(a−) and Cad phenotypes were not explained at that time. Instead, it took another 16 years until Sd^a negativity was associated with the gene when two research groups independently identified a missense mutation that dominated the two studied cohorts of nine Sd(a−) individuals each.^7,8 One of these studies was published in full in 2019⁷ and was followed by a more detailed analysis where causality was established between the predominant single nucleotide variant (SNV) and the Sd(a−) phenotype, by transfection of a cell line and followed by flow cytometry and proteoglycomics using mass spectrometry.⁹ The identified SNV, rs7224888, causes an amino acid change of cysteine to arginine in the catalytic domain of the enzyme. This cysteine is one of the previously mentioned evolutionarily conserved residues,⁷ and the corresponding moiety in the human homologue, β1,4GalNAc-T1 (the GM2 synthase), is known to be involved in a crucial disulfide bond within the protein.⁴⁶ In the previously mentioned transfection studies, B4GALNT2 constructs were expressed in HEK293 cells, and the low-frequency variant at rs7224888 was shown to abolish Sd^a synthesis.^9,51 This finding was confirmed in a separate study on circulating uromodulin, a known carrier of Sd^a.³⁹ Furthermore, the B4GALNT2_rs7224888 expressed mutant has reduced capacity to form dimers⁹ and is retained in the endoplasmic reticulum (ER) compartment and thereby absent in Golgi, where it is normally confined.³⁹ There are individuals with the Sd(a−) phenotype who do not carry this SNV where the reason for the lack of Sd^a remains unexplained. Other SNVs have been discussed: for example, rs72835417, a splice-site mutation in intron 8, found in two Sd(a−) individuals who were compound heterozygous for rs7224888.^7,8 Additionally, compound heterozygosity for two missense mutations of low frequency, rs148441237 and rs61743617 (situated in the same region as rs7224888), was reported in one Sd(a−) individual.⁷ However, neither of these latter mutations have yet been experimentally shown to affect the Sd^a synthesis capacity of the product from the B4GALNT2 locus. In addition, another Sd(a−) individual had no genetic variation in the B4GALNT2 coding region that could be linked to the phenotype.⁷ The genetic background of the Sd^a-high-expressing Cad phenotype has not yet been defined. However, a study of a small cohort of individuals with the Cad phenotype (N = 5) could not identify common genetic differences for these five samples in the coding region of B4GALNT2.⁵¹

The gene product encoded by B4GALNT2 consists of 11 exons. However, the two isoforms that have been experimentally detected, and a third predicted, all use different exon 1 alternatives. Therefore, there are actually at least 13 different exons associated with this gene altogether.^50,52 An area around and upstream of the three exon 1 alternatives contains CpG islands, which could imply that methylation is involved in regulation of B4GALNT2, at least in cancer (see Disease Associations section).^53,54 A recent study performed on human colon cell lines showed that transcription factor ETS1 (E26-transformation specific), and to a lesser extent SP1 (stimulating protein 1), is involved in transcriptional regulation of B4GALNT2 in this tissue.⁵⁵ In addition, the same study identified a cis-regulatory region located upstream of the B4GALNT2 short transcript harboring the promoter that drives the expression of the human B4GALNT2 regardless of cell type.⁵⁵ Furthermore, the lack of Sd^a on human colon cells and RBCs at birth, and the increase of Sd^a as a function of age, suggests an ontogenetic regulation of the gene.^56,57 This feature has also been shown in other species.^58,59

As previously mentioned, RNAseq data curiously indicate that there is no expression of B4GALNT2 in erythroid cells differentiating towards RBCs.²² Also, no GATA1 or other erythroid transcription factor motif has been associated with the gene, a feature otherwise often seen in many genes coding for glycosyltransferases synthesizing blood group antigens.²² More research on the subject is required to establish the cellular origin of the RBC-carried Sd^a, but the gastrointestinal tract in general, and the colon in particular, appear to be likely sources.

The characteristic agglutination pattern of Sd(a+) blood, including the Cad phenotype, suggests that only a portion of the RBC population carries the antigen. The underlying cause remains unknown, as well as the source of RBC-carried Sd^a, since there is no suggestion of expression of the B4GALNT2 in erythroid tissue.²² Even in the first studies of Sd^a, the antigen was already found soluble in saliva.^1,2 A follow-up study in 1970 confirmed this and further reported that infants had higher concentrations of Sd^a in saliva than adults. Sd^a was also found in human urine, milk, gastric juice, and feces, and when various tissues were examined, the kidney showed strong activity. The incidence of Sd(a+) urine suggested a Sd(a−) prevalence of approximately 4 percent,¹³ while it was estimated to be 2 percent after examination of kidney and colon tissues, which are both rich sources of the epitope.¹⁴ The Sd^a synthase has been isolated and characterized from human healthy kidney, plasma, urine, and colon.^41–44 B4GALNT2 transcript and protein levels of the synthase, as well as its activity in the gastrointestinal tract, display the highest levels in the colon.^41,48,60,61 It is noteworthy that cells of fetal colon tissue originally lack Sd^a,⁶² which thus fits well with the late appearance of Sd^a on RBCs in the common Sd(a+) phenotype.

Early on, Sd^a was recognized in species such as guinea pig, mole, and hedgehog.¹³ The high concentration of the epitope in guinea pig urine makes it a useful tool in inhibitory assays by which the antibody specificity can be neutralized and thereby confirmed.⁶³ Today, research studies on B4GALNT2 orthologues are of interest in a number of species for different reasons. In most mouse strains, B4galnt2 is predominantly expressed in the intestinal endothelium, and it has been shown that deficiency of this gene expression in mice significantly alters the composition of their intestinal microbiota. This finding may suggest that patients with the null phenotype could have an altered susceptibility to gastrointestinal infections.⁶⁴ There are also mice with a B4galnt2 allele called Mvwf1, which carries a cis-acting mutation 30 kb upstream of the gene. This motif causes a regulatory switch, which makes the gene express in the vascular endothelium. A downstream consequence of this is decreased half-life and therefore lower plasma levels of the von Willebrand factor. The vascular endothelium is the primary site of von Willebrand factor synthesis, and the altered glycosylation has been hypothesized to cause clearance via the GalNAc-binding hepatic asialoglycoprotein receptor in the liver. This feature has been seen in both inbred and wild-type mouse strains.^65–67 In mice with muscular dystrophy, overexpression of B4galnt2 reduces the pathology.⁶⁸ The mechanism is poorly understood and, to translate this to humans, better understanding of the glycosylation in human muscles is needed.⁶⁹ Gene therapy resulting in overexpression of this gene in dogs with muscular dystrophy resulted in altered glycosylation, but muscle strength was not increased.⁷⁰ Despite this, clinical trials are being planned or conducted in patients with Duchenne muscular dystrophy. Furthermore, there are indications that B4galnt2 expression is regulated by progesterone and estrogen and is involved in murine embryo implantation.⁷¹ In sheep, the orthologue gene is primarily expressed in the ovaries, and it seems to have an impact on fertility. Several breeds carry a SNV in intron 7 of their B4GALNT2, called the FecL^L mutation, which leads to increased expression and glycosylation in ovarian granulosa cells. This alteration has been shown to correlate with ovulation rate and large litter size, although the mechanism of action is unknown.⁷² Another study identified two missense mutations in a different breed, also suggested to affect litter size.⁷³

Finally, the xenotransplantation research field has identified the B4GALNT2 orthologue in pigs to be of interest. This gene encodes a synthase with 76 percent amino acid sequence identity to the human β1,4GalNAc-T2. It builds similar structures as the Sd^a/Cad epitope, which can also be targeted by the DBA lectin just like human Sd^a structures. Nevertheless, this structure has been acknowledged as a xenoantigen, and most humans have both IgM and IgG antibodies against it. It has been proposed that the carrier structure differs from that in humans, and the implicated moiety might be Neu5Gc, a type of sialic acid that humans lack.⁴⁵

Many glycans, including blood group–related glycans, are known to interact with microbial pathogens. Cad RBCs have been shown to be resistant to Plasmodium falciparum invasion. GPA on the RBC surface is involved in the attachment of the parasite to the RBC, and the sialic acid residues (often terminating glycan structures) may be involved. The addition of a GalNAc to the sialic acid–carrying structures, creating the Cad epitope on GPA, may interfere with parasite–cell interaction.⁷⁴ B4GALNT2 expression has also shown an inhibitory effect on influenza A virus. Again, the virus attaches to α2,3-bound sialic acid, which is the type of precursor the β1,4GalNAc-T2 uses.⁷⁵ A recent review focuses on how the Sd^a glycosylation profile of the gastrointestinal tract can influence the host–microbe interactions there.⁷⁶

Altered glycosylation pattern is a common feature in cancer tissue.⁷⁷ In the normally Sd(a+) stomach and colon tissues, the B4GALNT2 expression is reduced in malignant cells.^41,48,60,61 Forced B4GALNT2 expression reduces the presence of the sLe^x epitope on the cell surfaces and is thought to lower risk for metastases. The sLe^x structure is built on the same precursor as Sd^a and is a ligand for E-selectin on endothelial cells, a feature important in the metastasis process.⁷⁸ However, B4GALNT2 inhibition of the stemness-associated malignant phenotype is achieved independently of altered sLe^x expression.⁷⁹ The decreased expression of B4GALNT2 may be regulated by DNA methylation. This finding is an analogy with what has been shown for some other carbohydrate histo-blood groups (e.g., ABO in hematologic disease).⁸⁰ In gastrointestinal cancer cell lines, low Sd^a levels have been associated with increased methylation status of the gene’s CpG island promoter.^53,54 However, when looking closer at patient data, this finding may not be the whole explanation.^54,81 Cancer patients who maintain B4GALNT2 expression are predicted to survive longer. This finding is seen in large data sets, as in colorectal adenocarcinoma patients,⁸¹ but also in the pathology section of B4GALNT2 expression in kidney cancer tissue, in the Human Protein Atlas.⁸² Furthermore, Sd^a has been suggested to be used as a biomarker in cancer diagnostics. By profiling the internal sialylated glycans in the plasma of gastric and pancreatic cancer patients, the Sd^a-carrying structures were found to be elevated in some case subjects compared with normal control subjects. The cause was suggested to be increased release of the epitope, but further work is needed to conclude if this explanation could be useful clinically.⁸³ For more in-depth reading on this subtopic, there are reviews focused primarily on Sd^a and its relation to cancer.^84,85

Sd^a is associated with additional medical conditions shown in animal models; these conditions were discussed in the B4GALNT2 in Other Species section.

Sd^a was described more than 50 years ago, and shortly thereafter, the related Cad phenotype was identified. The anti-Sd^a that can be found in Sd(a−) individuals is naturally occurring and may cause hemolytic transfusion reactions when the donated RBCs express Sd^a strongly, as in the case of Cad. The interest of Sd^a stretches beyond the field of transfusion medicine. An example of this is its altered expression in certain cancers and its interaction with pathogens. Over time, the biochemical background has been clarified, and recently the Sd(a−) phenotype was linked to loss-of-function variants at the B4GALNT2 locus on chromosome 17. Based on this, the ISBT in 2019 acknowledged the SID blood group system as number 038, containing one antigen. Nevertheless, the origin of the RBC-carried Sd^a is still unknown because B4GALNT2 does not appear to be transcribed in erythroid precursor cells. Finally, the molecular basis of the Cad phenotype remains an enigma waiting for resolution.