How a pan-reactive antibody in a pregnant woman led to JAMA (ISBT 049), a blood group carried on an adhesion molecule most of us know from platelets. Bench to bedside, told as a case.
Shown
Not shown or not known
Reported gnomAD global allele frequency for rs755819660: 0.0000044. This is the frequency of the variant allele, not of JIMI-negative people.
In 2013 a blood service laboratory in Israel received a sample from a pregnant woman of Arab ethnicity. She had never been transfused. Her plasma held an antibody of moderate strength that reacted by indirect antiglobulin test with every red cell the team tried. Every cell except two sets: her own and her brother's.
That pattern tells an immunohaematologist a lot. An antibody that hits everyone except the patient and a sibling points to an inherited absence of a high-prevalence antigen. Siblings can inherit the same two copies of a rare variant from their parents. The job was simply to name the missing antigen.
It would not be named. The samples went to the International Blood Group Reference Laboratory in Bristol. The antibody reacted with cells from a large panel of rare null phenotypes, so no known "missing antigen" explained it. The siblings' cells reacted with an extensive set of rare antisera against high-prevalence antigens, so they were not negative for any antigen we already knew about.
Seven years later, in 2020, she was pregnant again. New samples, same result, still no answer. By then the tools had changed. Whole-exome sequencing had already cracked several old puzzles in red cell serology, and the Bristol team turned it on this family.
One variant survived their filters. It sat in a gene nobody had linked to a blood group before: F11R, which makes JAM-A. The acknowledgements of the paper carry a lovely detail. Dr Frances Spring had long believed that JAM-A would carry a blood group one day. She was right.
"An antibody that reacts with everyone except you and your brother is your family history written in plasma."
Teaching line · Blood🩸DoctorWhat JAM-A is. JAM-A belongs to the immunoglobulin superfamily. Picture a short arm sticking out of the cell. It has two "knuckles" outside the cell, called Ig domains. The outer one (D1) is a V-set domain at the N-terminus. The one nearer the membrane (D2) is an I-set domain. Then comes a single pass through the membrane and a short 39-amino-acid tail inside the cell.
Where it lives. JAM-A is expressed on several cell types, including epithelial and endothelial cells (where it sits in tight junctions), platelets, red cells and other blood cells. It has even been described as a marker of long-term repopulating haematopoietic stem cells.
What it does. JAM-A likes to hold hands. Two molecules on the same cell pair up side by side (a cis dimer, through the R59-V60-E61 motif on D1). Molecules on neighbouring cells grip each other across the gap (a trans interaction, through N43-N44-P45, also on D1). D2 is the docking site for the leucocyte integrin αLβ2 (LFA-1).
Why haematologists already knew it. In platelets, JAM-A keeps the fibrinogen receptor αIIbβ3 switched off. Its tail holds Csk kinase, which keeps the integrin inactive. When the platelet is activated, the tail gets phosphorylated, Csk lets go, and αIIbβ3 can switch on. JAM-A knockout mice show thrombotic and other problems.
What it does on red cells. Nobody knows yet. The authors suggest it may help hold erythroblastic islands together during red cell production, the same role proposed for the red cell adhesion molecules BCAM (Lutheran) and ICAM-4 (LW).
Codon 215 of F11R spans nucleotides 643 to 645. The variant swaps the middle letter, G to A. That turns CG· (arginine) into CA· (glutamine). Arginine carries a positive charge on a long side chain. Glutamine is similar in size but neutral. Modelling on the crystal structure shows both side chains pointing out into the plasma, and the authors describe the change as conservative, with little structural alteration.
JAM-A is a name badge every red cell wears. JIMI-positive people have one letter printed as "R". This family's badge says "Q". Same badge, same shape, one letter different. Her immune system read that letter and called it foreign.
Homozygous p.Arg215Gln removes the Arg215 epitope from the surface of D2. The protein is still made and still reaches the membrane, at lower levels on the siblings' mature red cells. She made an alloantibody against Arg215 JAM-A; the authors report that it formed during pregnancy.
Her antibody reacted with every red cell tested apart from the siblings'. Because JAM-A sits on platelets too, the antibody bound platelets as well as red cells in the laboratory. Whether that matters clinically is not yet known.
A monoclonal anti-JAM-A still labelled the siblings' red cells by flow cytometry. Their expression was lower than all eight controls, and the patient had the lowest level of all. The authors report that Arg215 and Gln215 JAM-A were expressed equally in CHO and HEK cells (data not shown in the paper). Their explanation is that the lower level on red cells arises from membrane remodelling during the last stages of red cell maturation, rather than from how the protein is made. This is a hypothesis; it was not tested directly.
Finding a rare variant is not enough. The ISBT needs proof that the antibody binds the protein, and that it binds the protein because of that one amino acid. Here is the chain the team built, in the order they needed it.
LISS tube IAT, moderate strength, all cells positive except the siblings'. An active eluate was incompatible with a large panel of rare null cells. The siblings' cells were positive with every rare antiserum tried.
Answer: no known specificityResistant to papain, trypsin, pronase and DTT. Destroyed by chymotrypsin. A useful fingerprint that matched nothing on file.
Answer: a chymotrypsin-sensitive protein epitopeNo variants in known blood group genes explained it. Filters kept shared, rare (<0.01), homozygous coding variants that change a surface-exposed part of a red cell membrane protein. Only one variant passed: F11R c.644G>A, p.Arg215Gln.
Answer: a candidate in JAM-AMouse anti-JAM-A labelled all 8 control samples and, more weakly, the siblings' cells. Patient plasma labelled controls in the same way but did not bind the siblings' cells.
Answer: yes, and the pattern fitsBoth forms were built as soluble Fc fusion proteins. Arg215 Fc inhibited the patient's antibody; Gln215 Fc and a Lutheran Fc control did not. Mouse anti-JAM-A was inhibited by both forms, so both proteins kept the region that antibody binds. The authors took this as evidence that both were biologically active. On transfected CHO cells, the patient's eluate bound only Arg215 JAM-A cells.
Answer: the epitope is Arg215Washed control platelets adsorbed the antibody from plasma. The eluate from platelets agglutinated control red cells but not the patient's. An anti-D control was not adsorbed, since platelets carry no RhD.
Answer: yes, red cells and platelets| Test | Patient antibody | Mouse anti-JAM-A | What it shows |
|---|---|---|---|
| Control red cells | Binds | Binds | Both reagents reacted with control red cells |
| Siblings' red cells | No binding | Binds (lower) | JAM-A present, JIMI absent |
| Pre-incubated with Arg215 Fc | Inhibited | Inhibited | Arg215 carries the epitope |
| Pre-incubated with Gln215 Fc | Not inhibited | Inhibited | Gln215 protein keeps the mouse epitope but lacks JIMI |
| CHO cells with Arg215 JAM-A | Binds | Binds | Binding needs Arg215 |
| CHO cells with Gln215 JAM-A | No binding | Binds | Confirms the above |
"The mouse antibody was blocked by both proteins. Hers was blocked by only one. That single contrast is the whole proof."
On step 5The antibody was never going to fit. Serology works by comparison. You match the antibody's pattern against cells and antisera that define antigens we already know. When the antigen belongs to a protein no one has linked to a blood group, every comparison comes back blank, and the case sits in the freezer.
The genetics held a second trap. Most variant pipelines rank changes by how badly they damage a protein. By that yardstick, p.Arg215Gln looks harmless. PolyPhen-2 called it benign (0.021). SIFT called it tolerated (0.76). AlphaMissense called it likely benign (0.112). A team hunting for "damaging" variants could have thrown it away. These tools estimate the likely effect of a change on protein function. They do not assess whether a surface amino-acid change can be seen by another person's immune system as an alloantigen. Nor did the study test JAM-A function directly. The Bristol filters asked a different question: is the change on the outside of a red cell membrane protein? That is the question that matters for a blood group.
Anchoring on the known list. Assuming a pan-reactive antibody must target one of the existing systems, and repeating serology rather than changing the method.
"Benign" in silico scores. Variant-effect tools estimate effects on protein function. They do not assess whether a surface change can act as an alloantigen.
Whole-exome sequencing of two affected siblings, filtered for shared homozygous, surface-exposed variants in red cell membrane proteins.
Antibodies to known high-prevalence antigens: argued against by the rare null panel and rare antisera. Autoantibody: her plasma did not react with her own cells, which supports an alloantibody, though a negative autocontrol alone does not exclude every autoantibody.
A pan-reactive alloantibody with a compatible sibling points to an inherited missing antigen. If the specificity stays unresolved after reference serology, that is the moment to think about sequencing both affected relatives, not the moment to file it as "unidentified".
| Question | What the paper shows |
|---|---|
| Was HDFN seen? | No HDFN was reported in four live births from six pregnancies. The paper does not link the two pregnancy losses to the antibody. One family cannot define the risk. |
| Can it cause a transfusion reaction? | Unknown. Neither sibling has been transfused. |
| Were the siblings unwell? | No associated pathology was reported. Both had normal Hb, MCV, MCH, RDW, platelet count and MPV. This fits with, but does not prove, normal JAM-A function. |
| Why does platelet binding matter? | The authors say any anti-JAM-A must be considered for effects on red cells and platelets, including fetal and neonatal alloimmune thrombocytopenia and platelet or granulocyte transfusion. Platelet binding was shown in vitro. FNAIT, platelet refractoriness and transfusion reactions were not shown. |
| How rare is the negative type? | The reported gnomAD global allele frequency is 0.0000044. That describes the variant allele, not the number of JIMI-negative people, and the paper does not estimate how common the negative phenotype is. |
The authors point to a warning from a neighbouring system. Csa on red cells and HNA-3a on neutrophils are defined by the same amino acid in CTL2. Yet anti-HNA-3a does not bind red cells, while anti-Csa binds both. CTL2 has a small extracellular part buried among other membrane proteins. JAM-A holds its two Ig domains well clear of the membrane, which may be why anti-JIMI binds red cells and platelets alike.
These are general principles for antibodies to high-prevalence antigens, applied to JAMA. They are my suggestions, not recommendations from the paper.
"Your blood has a very rare type. Almost everyone carries a small marker on their red cells, and yours do not have it. At some point your body made an antibody against that marker. The main practical effect we know about is that finding matching blood for you is harder, because most donors carry the marker. In the study, there was no sign of the antibody causing blood problems in four babies born alive, which is reassuring. But only one family has been studied, so we do not yet know all the possible effects. If you need a transfusion, an operation or care in a future pregnancy, tell your team early, so they can involve transfusion specialists and plan ahead."
A variant scored benign can still create a blood group, because pathogenicity tools ask whether the protein works, not whether someone else's immune system can see the change.
JIMI-negative cells still carry JAM-A, because the variant alters one surface residue rather than deleting the protein, so a monoclonal anti-JAM-A will still label them.
The proof came from the patient's own plasma: Arg215 JAM-A blocked it and Gln215 JAM-A did not, while a mouse antibody was blocked by both.
Anti-JIMI bound platelets as well as red cells in the laboratory, because JAM-A sits on both, so effects on fetal platelets and platelet support stay open questions until shown.
When an antibody reacts with everyone except a sibling, the family is the clue, and siblings are a sensible first place to look for donors, though compatibility and suitability still have to be assessed.
Tap an option to see the answer and why the others are wrong.
Before JAMA there were 398 accepted blood group antigens. 371 had a known genetic basis and sat in 48 systems. Nine of those 48 were found in the last five years alone. The authors count 12 other new systems ratified in the past decade, and a pattern is emerging: several of the newest blood groups sit on proteins we first knew from other blood cells. CD36 on platelets. Mal on leucocytes. CTL2 on granulocytes. And now JAM-A. All blood cells share a common precursor, so the overlap makes sense.
Recent relatives, from the same laboratory and its collaborators (as cited in the paper):
First, JAM-A has two extracellular Ig domains, D1 and D2. Drawings with three extracellular beads are wrong. Second, the paper does not show platelet refractoriness, FNAIT or transfusion reactions. The authors list platelet and granulocyte transfusion and FNAIT only as areas to consider, so present them as open questions.
Every clinical and laboratory statement on this page comes from reference 1 unless marked otherwise. All PMIDs and DOIs were checked against PubMed on 30 September 2026. These papers support the science on this page; none of them is a clinical management guideline. The codon arithmetic and the arginine and glutamine chemistry are standard biochemistry. The practical laboratory points and the clinic script are my own suggestions and are labelled as such. Treatment lines and clinical trials do not apply to a new antigen and are left out.
Blood🩸Doctor · contact@blooddoctor.co