Yes, MAM: how the cancer-related EMP3 protein became a regulator of erythropoiesis and the key protein underlying a new blood group system

The story of MAM began in 1992 with the investigation of an antibody to an unidentified high-prevalence antigen in a young American woman (M.A.M.) of mixed Irish-Cherokee descent in her third pregnancy.¹ The antibody was strongly reactive with all red blood cells (RBCs) tested, and the monocyte monolayer assay (MMA) predicted that the antibody was probably clinically significant. Despite the efforts of the Immunohematology Reference Laboratories both in the United States and UK, specificity could not be established. Symptoms of fetal distress at week 31 of the pregnancy prompted delivery by caesarian section. The baby’s RBCs were strongly positive by the direct antiglobulin test (DAT), but the baby had a normal bilirubin and showed no signs of clinical hemolytic disease despite a maternal antibody titer that had risen from 4 at the time of detection to 256 at delivery. However, the infant was severely thrombocytopenic and required platelet transfusion, symptoms that were at the time attributed to the presence of anti-HPA-1a also found in the proposita’s plasma.

Since then, at least 10 other cases of anti-MAM have been identified, all in women. In six of the 11 known cases, anti-MAM was detected during a current pregnancy with clinical symptoms of hemolytic disease of the fetus and newborn (HDFN) that have varied from mild to severe. In five cases, detection of anti-MAM was not associated with a current pregnancy, although the medical history suggested that pregnancy was the likely immunizing event in at least three cases.

In the second case described, a woman of Arabic origin, severe HDFN as well as fetal and neonatal alloimmune thrombocytopenia (FNAIT) prompted an unsuccessful intra-uterine transfusion attempt followed by an emergency caesarian section at week 24.5.² The hematocrit and bilirubin were stabilized to normal levels by day 13, but unfortunately the infant died at 6 months of age because of complications of premature delivery. Family studies were performed, and anti-MAM was also identified in samples from the proposita’s sister, who had had one previous pregnancy in which the newborn was unaffected.

A similarly severe course of HDFN was observed in a New Zealand woman of European descent.³ Anti-MAM was identified during her second pregnancy. Upon indications of fetal distress, an intrauterine transfusion was performed with washed, irradiated maternal RBCs. However, the patient suffered a placental abruption at 28 weeks and a major postpartum hemorrhage that necessitated the transfusion with incompatible MAM+ blood. Sadly, the fetus did not survive.

Another case study of anti-MAM followed a young Yemeni woman in her second pregnancy.⁴ Her antibody titer remained stable at 32 throughout the pregnancy, and a full-term infant was delivered at week 38. Surprisingly, the baby’s hemoglobin was slightly elevated, but the baby showed few other symptoms.

Altogether, 11 cases of anti-MAM have been described to date (all but one, summarized in Thornton et al.⁵) and have recently been used to elucidate the molecular and genetic background of the MAM– phenotype. MAM has therefore been acknowledged by the International Society of Blood Transfusion (ISBT) as the single antigen in the new MAM blood group system (ISBT 041).

Until recently, MAM was one of the few remaining high-prevalence blood group antigens to remain elusive to molecular and genetic characterization. A previous study had suggested it to be carried by N-glycans,⁶ but more recently, Thornton et al.⁵ identified the tumor-related factor EMP3 (epithelial membrane protein 3) as the gene coding for the MAM antigen using a whole exome sequencing approach. Once EMP3 was identified, targeted gene sequencing was used to characterize five genetic variants (whole-gene deletion, exon deletions, and premature stop codon) in EMP3 across 10 MAM– individuals as the cause of the phenotype (Table 1). So far, all known cases lacking MAM have been found homozygous for a genetic defect in the EMP3 gene. The study by Thornton et al.⁵ confirmed that EMP3 is necessary and sufficient for MAM antigen expression by knocking it out in BEL-A2 and overexpressing it in Daudi human immortalized cell lines. Later, yet an additional genetic variant leading to loss of EMP3 was reported.⁷

EMP3 is an 18-kDa glycoprotein consisting of 164 amino acid residues with four transmembrane domains. It is a member of the peripheral myelin protein 22 kDa (PMP22) gene family and contains two extracellular loops with two potential N-glycosylation sites, one experimentally confirmed to carry an N-glycan (N47), and also a putative intracellular phosphorylation site (T90) (Fig. 1). The EMP3 gene is located at 19q13.3.8 Based on homology with PMP22, the EMP3 protein was initially thought to be involved in regulation of membrane organization, cell-to-cell interactions, and signal transduction processes.⁹ Functional aspects of EMP3 will be further discussed in a later section.

Fig. 1

Interestingly, it was found that ex vivo culture of CD34+ cells from peripheral blood of MAM– individuals showed a marked increase in proliferation compared with cells from MAM+ individuals, resulting in a considerably higher yield of reticulocytes in the culture (Fig. 2).⁵ This finding suggests that EMP3 may act as a suppressor of erythropoiesis. Contrary to this, MAM– individuals exhibit no signs of increased hemoglobin values, erythrocytosis, or reticulocytosis, which could possibly reflect a compensatory mechanism within the erythroid niche of the bone marrow that prevents excess erythrocytes from being formed or from entering peripheral blood, or other regulatory measures not yet understood.

Fig. 2

EMP3 is expressed in many tissues throughout the human body, with particularly high expression in certain blood lineages including granulocytes, monocytes, and lymphocytes, and several epithelial and mesenchymal cell types. Given the clinical picture of both HDFN and FNAIT in some cases of anti-MAM during pregnancy, it is particularly interesting to note that EMP3 (hence, MAM antigen) expression is seen on platelets. MAM expression is independent of activation status and does not appear to affect platelet activation.⁵

The MAM blood group antigen is resistant to treatment with ficin, papain, trypsin, and α-chymotrypsin, as well as 200 mM dithiothreitol. The antigen is well developed on cord RBCs. MAM– RBCs express low levels of CD44, and consequently the antigens of the Indian blood group system are only weakly expressed. As can be seen by the case reports described earlier, anti-MAM can cause severe HDFN in some pregnancies and have little to no effect in others. The reason for this variable effect is not known. Similarly, results of the MMA performed in the different cases suggests that anti-MAM is clinically important with regard to blood transfusion, although there are too few examples to be definitive. Similarly, it is still unclear what effect anti-MAM has on platelets. Even if it has been conclusively shown that MAM is expressed on blood cells other than erythrocytes, its causal involvement in FNAIT is debatable but, because of their rarity and quite disparate outcomes in the few cases followed, clinicians should be advised to take precaution and look for both signs of anemia and thrombocytopenia if maternal anti-MAM is detected. In this context, it needs to be noted that there are extremely few MAM– blood donors available. A future hope is that, now that the genetic basis of the phenotype is known, genetic EMP3 markers could be included in commercially available platforms for blood group genotyping to help identify more donors.

The normal function of EMP3 in healthy tissue or development is yet to be elucidated and, so far, little work focusing on this function has been presented.

Disease Associations

Expression of EMP3 has been linked to a number of cancers and was initially described as a tumor suppressor in glioma and neuroblastoma, where promoter hypermethylation leads to transcriptional silencing of the EMP3 gene.^10,11 However, expression of EMP3 has since been associated with both poor and good prognoses in an increasing number of various cancers.

The most striking observation is that the effect of EMP3 appears to be highly context dependent. It has been suggested that high EMP3 expression is a marker of some cancers such as renal cell carcinoma,¹² hepatic carcinoma,¹³ gastric cancer,^14,15 upper urinary urothelial carcinoma,¹⁶ primary breast carcinoma,^17,18 and mammary carcinoma.^19,20 Conversely, loss of EMP3 expression has been linked to a poor prognosis in esophageal squamous cell carcinoma,²¹ non–small cell lung cancer,²² oral squamous cancer cells,²³ and gallbladder cancer.²⁴ EMP3 dysfunction has also been generally associated with endometrial cancer,²⁵ acute myelogenous leukemia,²⁶ and thyroid cancer.²⁷

Even among central nervous system tumors, expression of EMP3 has different impacts depending on the context. Whereas loss of EMP3 expression is observed in low-grade (oligodendroglial and astrocytic) glioma and neuroblastoma,^10,11,28,29 EMP3 appears to be a major driver of high-grade (isocitrate dehydrogenase–wild-type [IDH-wt]) glioblastoma.^30–33 Overexpression of EMP3 is frequently observed in IDH-wt glioblastoma, linking it to poor clinical outcomes, suggesting an active role of EMP3 in these tumors. Studies in this area suggest that EMP3 is part of a mesenchymal transcriptional program that imparts poorer prognosis.^34,35 Regardless of the role of EMP3 in cancers (as an oncogene or tumor-suppressor gene), it is clear that its level of expression in certain cancer types is a potentially useful prognostic marker for patient survival (Fig. 3).

Fig. 3

In the various studies of EMP3 function in cancer, EMP3 has been implicated in intracellular signaling and cancer-related pathways such as P13K/AKT,^13,14,16 MAPK/ERK,²⁴ and TNFα.^23,36 EMP3 has been shown to interact with numerous membrane proteins affecting downstream responses. For example, EMP3 directly interacts separately with TGFβ receptor II (TGFBR2)^33,37 and HER-2 (ErbB2),^16,38 both of which are part of known cancer-related pathways. EMP3 can also interact with FLOT1, which is known to cluster ErbB2 into lipid rafts.³⁹ The known roles of EMP3 in systems such as these have recently been thoroughly reviewed by Martija and Pusch.³⁴ However, it is currently not known if the lack of EMP3 in MAM– individuals has any effect on disease susceptibility or treatment response.

While the normal function of EMP3 remains unclear despite ongoing efforts to elucidate its role, the presence of EMP3 on RBCs and platelets is undisputed, and anti-MAM should be considered to be clinically important. There are few MAM– blood donors available worldwide, although the nonsense single nucleotide variant c.123C>G (p.Tyr41Ter; rs201392469) appears to be more common in people of Middle Eastern origin, which may enable a more targeted screening approach for identifying MAM– donors.