Red Cell Genotyping in Modern Blood Banking

I'm not sure if you all knew that this is also board review time. I'm going to ask you some questions, and if you have paper, you could write down your answer. Then, we're going to come back to these and answer them as we go.

Case Studies

So, while you think about all those cases, now we're going to back up and talk about why we do red cell genotyping and how it can lead to improved medical care.

Introduction to Blood Banking

First, I'll start by telling you a little bit about blood banking, in case those questions made you uncomfortable. We'll talk about what actually happens in the blood bank in its current state.

What blood groups are is a collection of erythrocyte membrane components, and they are defined by the fact that when we give a transfusion, we are giving a transient tissue transplant to someone else. Their immune system can recognize these erythrocyte membrane components and make an antibody. Over time and over history, as this has happened through pregnancy and transfusion, we've defined blood groups basically by the fact that someone made an antibody to this surface component. I show an antibody here to show that's both how they're originally defined and also generally how we're testing for them in the current state in the blood bank.

Traditional blood typing is done by serological methods. In this diagram, a red cell has these green triangles on its surface, and we're trying to see if they are there or not. We would add an antisera that would be specific for that red cell component. Because we can't see that with the naked eye—because the antisera is not big enough to cause large agglutination clumps—we add an agglutination reagent, which is really Coombs' reagent, or anti-IgG. That makes a visible agglutinate, big enough for the human eye or for a basic light reader to see. In addition, there are some reagents that are IgM or monoclonal that don't require adding the Coombs' reagent, but either way, it's mostly agglutination-based testing.

There are problems with this method, and they are illustrated here in this slide.

• If a patient's been transfused, when you try to use serology to look at the surface of the red cell, you're going to have problems because there are multiple people's red cells in that tube.
• In addition, when people have interfering antibodies coating the whole red cell surface, the antisera cannot access the surface.
• On the other side, when we have no sample available—so in the maternal-fetal setting, we don't have fresh red cells from the baby—this can be a problem for typing the baby's red blood cells if we're concerned about incompatibility.
• For many blood group systems, we actually don't have any antisera at all. If we do, we have a very limited amount that are FDA-approved. The others that are not FDA-approved are literally from blood donors that we find that have antibodies, and we use those as reagents. So, it's really important to realize that the current state of blood typing is relying on human blood products as reagents.
• In addition, I'm going to talk a lot about the lower right-hand corner bullet, which is underlying genetic differences in blood group antigens, so, antigenic epitopes and expression. When you think about the fact that the serological reagents are all based around a human, animal, or tissue culture creating antisera, they're directed at epitopes. If those epitopes or the antigen expression is actually different because of underlying genetic differences, they don't work right. And as clinical pathologists, we want our tests to work right. When they don't, they can lead us astray.

Molecular testing is now part of most reference labs. When I say reference labs, I mean Red Cell Reference Labs for transfusion. They are gaining momentum, but they've implemented it at different levels and in different amounts. The types of testing that we do in my laboratory go all the way back to RFLP; we still use this for testing red cell genes. Most of the testing is focused around SNPs, or single nucleotide polymorphisms, to look at what the DNA codes for that we know by history would predict that red cell's expression of the antigen. We do occasionally, from time to time, have to go to Sanger sequencing, and I'll tell you more about that, but we don't do it very much.

So now, that's blood typing in a nutshell and where we are today. Now, I'm going to talk to you about two blood groups and explain the underlying genetics around these blood groups, which really helps to understand how, clinically, these cases that we talked about at the beginning can happen.

The RH Blood Group System

I'm starting with the hardest one. Anybody who knows anything about blood groups knows that RH is a really difficult blood group to type, and it is really the most complex blood group. There are 417-amino acid polypeptide proteins, and here on the red cell surface, it's stylized as this 12-transmembrane polypeptide. There are actually two encoded by the RH blood group: RHD and RHCE.

People have been trying to figure out the RH blood group genetics since the '40s. This is a picture of the pub, "The Bun Shop," which is on King Street in Cambridge, where Fisher and Race proposed how they thought the blood group worked genetically. Basically, the two different ways that were being thought of in the '40s of how this blood group worked is that it was either one gene with five different antigens or it was three genes. Turns out neither of them are right, but that's where we got our notation. One of the things that baffles people when they hear blood bankers talk is we still talk in this notation, even though it was derived around these models of either a one-gene model or a three-gene model.

Over here on the right-hand side are the shorthand notations for the Wiener and the Fisher-Race nomenclature, and the reason that they stand is that we do know that RH is inherited as a haplotype. I'll show you that a little bit more.

In the RH gene, you can be D-positive or D-negative. What's interesting about this when you think about the immune system in blood groups—because we really care about them because of transfusion—is that you either have it or you don't have it, in the most simplistic sense. That means that you either have a 12-transmembrane polypeptide in your membrane or it's not there. So you can imagine that's quite immunogenic if we cross that barrier in transfusion.

Contrasting that to the CE locus is that most people, except for very, very rare people, have the CE gene, and then there are four permutations of what that allele could be—so four reference alleles, if you will. What this is showing is the D gene. If you're D-positive, in the Wiener/Fisher-Race nomenclature over here, you have the "big D" and then the "big C" or the "little e," and that correlates to the R1 haplotype, and so on down the line. In the lower one, when there's no D gene, we still keep that "little d," but little d means no gene. We keep that there, maybe to confuse people or to remind ourselves that it's not there, but that means that there's no D.

This is now what we know and understand the RH blood group genes to look like. It's two genes, and they are close to each other in the genome. They each have 10 exons. In fact, the D gene is an ancestral duplication of the CE gene. In the middle, there's another gene that doesn't have to do with the RH blood group system. And at the flanking ends of the RHD gene are these "Rhesus boxes," which are highly homologous boxes upstream and downstream from the gene.

What's also interesting about this is that these genes are incredibly similar to each other, which makes typing for someone's blood type incredibly difficult when it's important if you have D or don't have D, and then which allele of CE you have.

This is now the polypeptide—that's the consensus polypeptide for D and C/E. The circles represent the places where D and C/E differ from each other, and the boxes represent where the exons start and stop. You can see that D and C/E are actually pretty similar. What this makes difficult in the serology world is that sometimes antisera have trouble picking up D if you also have the big C gene, because they're so similar.

Then, looking just at CE: D is either there or not there, and CE is are you big C, little c, big E, or little e. These are the positions here in the amino acid chain that show where those are different. You can see that some of them, the most important two, are presented to the immune system on the extracellular side of the membrane, and the others on the intracellular or intramembrane part of the cell.

If you try to put it all together in a 3D picture, as Dr. Flegel did in Blood in 2009, you get this picture of a raft in the membrane. It includes D and C/E, and also another gene called RHAG, which we don't talk about that much in blood banking; it's not as clinically important in terms of making antibodies, but it's there nonetheless. In this picture, again, you can compare and contrast the blue spheres in both the top picture and the bottom picture. One is more simple and one is more filled in on the bottom. This is the difference between D and C/E, and the green spheres are the difference between C and E. I think it's important to see them because it tells you that when you're transfused with something that's not you—that is, foreign red cells—those are probably the positions that you're going to make an antibody to.

So that's the gene. What about when we actually transfuse? How different are we from each other? How different is a transfusion patient from the donor base? This shows—we're not going to go through this in detail—but it just shows that we're different. If you look at your blood type, you get half of your genes from your mother and half of your genes from your father, and your blood type tends to go with your ethnic background. And so you can see here that the likelihood of being an R1, or big D, little c, little e, is 42% if you're of European descent, but if you're of African descent, it's actually quite a bit lower than that.

One of the problems that happens in transfusion is that this is not a blood group that we're matching for all of these antigens right now. And so, when the blood supply in the United States is predominantly Caucasian, if you're not Caucasian and are receiving blood products, you can bet that by chance you will be getting mismatched blood at some level.

Why does that matter? Well, it matters because it can cause transfusion reactions. Transfusion reactions can be mild or can be fatal. Most of the antibodies formed when you're transfused with foreign RH are of the IgG type, and so therefore, if you're going to hemolyze incompatible blood, it will be extravascular. The transfusion reaction associated with RH can cause severe hemolytic transfusion reactions.

In addition, although we often think that we've solved the hemolytic disease of the newborn problem with RhoGAM, we have not. Antibodies to D are still one of the most common reasons for hemolytic disease of the fetus and newborn (HDFN), and the other RH antigens—antibodies to them—still cause significant HDFN. The consequences of this are actually quite disastrous in terms of the child's morbidity and mortality. In addition, there's some studies that even show when you have intrauterine anemia, your likelihood of cerebral palsy or other cardiac failure is really much, much higher.

The Challenge of Weak D and Partial D

We're going to go a little bit deeper into RH, and we're going to talk about the real problem. I hope I've already presented that RH is complicated, just if we accept that we care if you're D-positive or D-negative and if you're C-positive or C-negative. But it's far more complicated than that.

Here again is the red cell membrane, and in the membrane here is a stylized picture of the RHD polypeptide. What we have invented in blood banking is a terminology called "weak D." What that means is it's the way that a blood banker is saying, "There's less D on this red cell than what we would expect." And so, by virtue of that, our reagents don't work so well—back to those serology reagents. But in the clinical setting, we think that because the extracellular loops are still the same as what we would consider to be wild-type D, it should not make you make an antibody as long as you're D-positive.

If you contrast this to "partial D," you can see that the dots, or the mutation changes, are actually now on the extracellular side of the red cell membrane. Therefore, if you transfuse someone who is D-positive with this red cell, they could make anti-D.

What the blood industry has done because of this—which is also quite a simplistic way of looking at it—is they've made it so that if you're a patient and your blood sample is sent to a blood bank, the antisera used in that blood bank is made in such a way so that it would find a weak D person to be D-negative. And that's because we just said, "Well, let's just treat them as D-negative. It's safer. We don't know what's going to happen, so let's just treat them as D-negative."

But if that same person then becomes a blood donor and they're typed in a laboratory where blood donors are tested, those antisera have been designed to pick that weak D person up as D-positive. And that's because if you transfuse that into a D-negative person, they can make anti-D. We know that the industry asked for that and the manufacturers did that. It's kind of funny if you think about it from a pathologist's perspective that we're engineering the reagents to see different things for the same person, because it's not really the right answer in the end.

There are more than 76 classified weak D alleles right now, and I'm sure there are many, many more out there we haven't found. The reason I'm showing you this chart is just to show you that there's three very common kinds: weak D type 1, 2, and 3. They were found mostly in people of European descent, mostly in people in Germany, and they have been analyzed in Europe far more than here. We have found time and time again that these people do not make anti-D. That's important.

We actually think now that for those types—one, two, and three of weak D, represented by the line here—if they basically have these nucleotide changes, we can treat them as D-positive. They actually have enough D on their surface they should be treated as D-positive. We don't need to treat them as D-negative. Why do we even care? Well, D-negative blood is not that common, and to transfuse a D-positive person with D-negative blood is not the best use of your resources. In addition, if it's a pregnant woman, you might be giving her RhoGAM when she doesn't need it. And so you're giving her another exposure, another cost, something else that she has to worry and think about, maybe over multiple pregnancies, when in fact she's weak D type 2 and she actually could just be treated as D-positive.

For all the other weak D types, so number four through number 76, we actually don't have enough data to say that they are safe to be treated as D-positive.

Case 1: Resolution

So let me show this to a different—oh, I have my case first. So we're going to go back to this case, and then I'll tell you the guidance for what I was just talking about.

Back to our woman who was typed as Group A, RhD-negative, and her transfusion service at the hospital provided Group A, RhD-positive blood. In this case, when they called me, they did retest the blood; it was the same type again. The transfusion service retested her sample; it was still Group A, RhD-positive. So it's not "wrong blood in tube."

Okay, we ended up doing a genotype, and I'm showing you the gene again at the bottom to tell you that in my lab, when we're looking for if you're weak D or any other D variant, we need to know how many copies of D you have. It's more likely that you'll have one allele that's D-negative and one allele that has a D variant. And this is what we found with this patient.

What we do is we have a PCR that's looking for the hybrid box, which is basically the upstream and the downstream Rhesus boxes that I showed you before, brought closer together by the deletion. When we find the RHD hybrid box, we say, "Ah, one of the copies is deleted." If we find two copies of it, it means both of the copies were deleted. If we find just the upstream box—so primers designed around that region—then we know that means that there's an intact D gene there. So here, there's a control bar, and as well there's a hybrid box and an upstream box. We would say that this woman is a hemizygote; she has a D-positive allele and she has a D-negative allele, which follows with people that have D variants. We usually see this.

So then we do a very simple PCR looking for weak D types 1, 2, 3, and then some other ones here. And you can see very clearly she has that SNP for weak D type 3. And this again is someone who has the allele for which there has been no published evidence of someone with this allele, although it's relatively prevalent, forming anti-D. So we could treat her as D-positive.

What caused this situation? The answer to this one is her RhD expression is variant.

Literally going back to that case, her obstetrician is sending her blood during her prenatal visits to an outpatient lab that's using some antisera, and then she comes to the blood bank, which is using a different antisera just because they are, and it picks up this weak D differently. And so now you have a D typing discrepancy. I can tell you it makes people really upset when their blood type changes because it's one of the things people know about themselves. Often their blood type is one of the things they think about with their family members, and so people get upset when their blood type changes. Clearly, that nurse thought that there was actually a mistransfusion event, and it wasn't that at all. It was the difference in the antisera of these two different laboratories.

So going back to what I was saying, how do we treat her? The next question that follows is, is this woman D-positive or D-negative for the purposes of transfusion? And therefore, do we give her RhD-positive or negative blood, and do we give her RhoGAM?

I've been part of a consensus conference that the AABB commissioned, where we've tried to answer this question with the tools that we have now. You'll see that it's pretty pared down, but basically what the group has come up with is to say that:

• If you are D-negative, you are D-negative. You are a candidate for RhoGAM and you should be given RhD-negative blood.
• If you are like her in that middle box—discrepant or inconclusive, or the strength of reaction is quite weak—then we are saying that you should do RHD genotyping, that simple test I just showed.
• If you are, on the right-hand side, weak D type 1, 2, or 3, for the purposes of transfusion and medical care, you can be treated as D-positive.
• However, if you're not one, two, or three, you're something else, we can't comment any further. We don't have enough data to say.
• And then on the very right-hand side is that if you're D-positive, you're D-positive. You don't need RhoGAM, you don't need D-negative red cells.

This is new, it's coming out this year, and I think it's going to be something that the blood banking community is going to talk a lot about, because we're trying to push that we really need to know what the answer is. When somebody's blood type is wrong, it's a big deal. It matters to the patients, it matters to their medical care, and we should really get it right.

Hybrid Alleles and Sickle Cell Disease

So, I'm not done with D. It actually gets more confusing than that. Remember the gene picture I showed you with the D and the CE oriented in the genome? That's panel A here; it's the same picture. One thing that we found, and the theory goes, is that these two genes, because they're so close in approximation and the way that they're oriented, can actually exchange genetic material. We find these hybrid alleles. Down here on the lower right is a D gene that has exons 4 through 7 of CE in the middle. This is a relatively common finding in people of African descent.

This hybrid allele can result in what's known as a (C)ceS phenotype. An interesting study was done in France where they looked at their sickle cell patients. We know in the blood bank that this is out there, that we see this a lot. They typed them for C, and if they were C-positive serologically, they did a genotype to look for this hybrid allele. Then they watched them.

The issue with this allele is that it's a little bit of D and a little bit of C, and it makes the person express a C that's variant. They will type as C-positive but then they'll make anti-C because it's not the same C. Again, typing reagents are not calling it right. These patients were prospectively observed, and basically, 30% of them made this antibody. So even though they are big C-positive, they are making anti-big C, and it's because they carry this hybrid allele. We need to know this in the blood bank. This is one of the reasons that here in Seattle, for sickle cell disease patients, we genotype them, because we want to look for this.

Case 2: Resolution

So let's go to case two now. This is our 18-year-old male who had an initially negative antibody screen. Ten days later, his hematocrit was not responding, they did an antibody screen, and now he has evidence of an anti-big C. And you can see here, he is big C-positive. So what caused his poor hemoglobin response?

He'd probably made the antibody in the past. About 25% of antibodies actually disappear over time, and that's what his did. When he got restimulated, his immune system recalled that antibody and made the antibody come up to a higher titer.

But what's important here is that the way that in our lab we predict this right now is just by looking at pieces. There's no way I have right now to actually find that entire hybrid allele genetically. What I have to do is I have to look at pieces of information. The patient has to be C-positive serologically, predicted to be C-positive genomically, and then have these two changes that are on one of my chips that I use. Then I say they might have that allele, but that's as far as I really can go right now. It's sort of the best we can do at this point.

This is a picture of a patient that was reported in a sickle cell study out of Children's Hospital of Philadelphia with the same exact findings, showing on the left is the hemoglobin S percentage and on the right is the actual hemoglobin percentage. You can see that the arrowhead shows when his antibody screen became positive, just like our patient. And when his antibody screen became positive, you can see on the right-hand panel that the hemoglobin dropped. Indeed, they have transfusion reactions, and it's very difficult to get their typing right because they are more likely to have these hybrid genes at the RH locus. Yet, they're incredibly clinically significant, because if 30% of patients make an antibody when they have this genotype, that's really high. In the general surgical population, it's about 3% of people that make antibodies.

Just to give you a little bit more about sickle cell disease, because no talk about red cell genomics would be complete without talking about them a little more. From the same group in Philadelphia, they have a very interesting approach, and it's very much thought of as being the best approach to transfusing sickle patients. Because we know that they have hybrid genes at the RH locus and because they're heavily, heavily transfused and have a very high rate of making antibodies, they provide both RH and Kell matched product. They give more highly matched product. In addition, when their blood donors of African descent come in and donate blood, they can take a green card and put their green card with their donation, and that will direct that blood to a sickle cell pediatric patient. So we're trying to racially match for the things that we're not typing for. It's kind of a surrogate to try to get a better match.

What this study shows is that they looked at these 182 kids that were transfused over 15 years, and they wanted to see, did it work? Did it make them not make antibodies and not have transfusion reactions? What they found is that first of all, 58% of them on the left actually did still make antibodies, which is not good at all. The 15% episodic means they weren't transfused as much, but they also still made antibodies, and most of them were RH antibodies. (*Question from audience*) They were on simple transfusion every three weeks.

These patients made lots of antibodies, even though this is about the highest level of match of any big academic center in the country. In addition, what the chart is showing is that 38% of them had this problem from the case I showed you. They expressed the antigen and made an antibody. So they were big C-positive and made an anti-big C; they were little e-positive and made an anti-little e. The take-home message of that is that our antisera don't work to type their RH type because it's looking at an epitope that the antisera was designed to look at, which is different in the way that their genes have evolved over time. This means we have a lot of work to do. The authors of this paper actually concluded that we probably need to think about doing molecular-level RH matching for sickle cell disease patients because doing it at the phenotypic level is not good enough.

The Kidd Blood Group System

I'll leave RH for a while and I'll talk about the Kidd blood group system. The Kidd blood group is a urea transporter in the red cell membrane. The Kidd polymorphism that we are interested in is really, are you Jka-positive or Jkb-positive? For whatever reason, that mutation makes the immune system make antibodies if you're one or the other. Here is a chart showing, again like I showed for RH, the different Kidd types you can be: A-positive B-negative, A-negative B-positive, and so on. They differ.

I want to point out at the bottom, to be Jk(a-b-), or null, is almost not found in certain populations but found much more routinely in people of Asian descent. There are different reasons for this. For some reason, in that population—mostly Southeast Asian and South Asian—is that there have been exon deletions, premature stop codons, and exon-skipped in-frame mutations. There's probably an evolutionary derivation for this reason, but for blood typing, it can be a problem. The regular SNP assay that we do in the clinical lab is really only looking at that Jka/Jkb polymorphism; it is not looking for these other silencing mutations. Therefore, it can cause discrepancies, which I'll talk about more towards the end.

Case 3: Resolution

Let's talk about our case again. This is the 73-year-old Asian-American man who has an upper GI bleed and hepatitis, and he has an anti-Jka, an anti-Jkb, and an anti-Jk3. What that means is that the Jk3 is directed at all Kidd antigens, if you will. So how hard will it be to find compatible blood for this patient?

We'll almost definitely have to find an Asian donor. There's a few founder alleles in Europe that have been found, but mostly they would be Asian. When we have patients like this, sometimes we have to import blood from Hawaii. It's just generally you find units where those ethnic groups live.

New Approaches: Massively Parallel Sequencing

So, given the known and unknown blood group variation, are our current approaches adequate? In my lab, again, we have serology; we have lots of serological techniques. We also have genotyping, but all the genotyping really focuses on looking at positions in the genome that we know in the Caucasian population to code for changes that are common. The way that I've started to think about it, and I think it helps make more sense, is it's kind of like we're peeking under tiles. I'm looking to see, well, is the RH mutation there or no? Or is the CE here or no? Or is the other D mutation that I'm interested in there or not there? And then sometimes I come up with, "Well, I can't answer the question." And so I have to actually do sequencing. The way that we do sequencing in the clinical lab right now is Sanger, and that takes a long time, it's very labor-intensive, and it still doesn't always answer the question, which means we also scientifically have more to learn.

I've been talking a little bit about these discrepancies, and I'm going to show you some data around discrepancies. What I mean by discrepancies now is not two serology reagents being discrepant, like our case number one, but serology and genotyping being discrepant. This is again data from the Children's Hospital of Philadelphia looking at that same sickle cell patient cohort. They genotype all of their sickle patients now as well. This is basically looking at SNP genotyping versus serology. You can see the false positive in gray, the false negative in white, and not discrepant in black. And so that meant that they made a transcription error. That was 14% of their sickle cell patients.

Here in Seattle, we've been focusing on Asian blood donors and typing Asian blood donors. Back in 2005, the blood center started collecting Asian blood donors' DNA. Here, it represents all of the different self-identified ethnic heritages that the now 9,000-donor repository now holds. I'm really proud of this work from the Blood Center, that this kind of collection exists.

Similarly to what the Philadelphia group has done, we have looked at some blood group genes both serologically and genotypically. What this graph shows is, in the colored boxes, these were SNP-negative and serologically positive. And in the white and gray boxes, they're SNP-positive and serologically negative. I'm going to pause here for a minute to think about how interesting it is that it goes both ways. The unknown null or a premature stop codon in the genome makes sense, but the other way is also quite interesting, that it actually goes both ways. Overall, when you look at this cohort, we found close to 5% of Asian and Native American blood donors where their SNP typing predicted red cell type and their serologic type did not match.

I'm going to close my talk by talking about new approaches, and this is a topic that I am learning a lot about and trying to learn a lot about. And it is massively parallel genotyping. I feel really lucky that Jill Johnson and myself have been working for a while on trying to figure out how we could use next-generation technology for blood groups. We've partnered with Debbie Nickerson's lab and have, really in the past year, gotten some results, and I'm happy to show them to you today.

The advantages of doing next-generation or massively parallel genotyping is that we know what most of the blood group genes are, although we don't know all of them; there's actually a few that are still unidentified. Because of that, we can target them and get almost all of them on one test. Because of the SNP solutions that I've been talking about and I struggle with in my clinical lab, we think we can more deeply figure out what the answer is right from the beginning and not do multiple, piecemeal SNP tests trying to figure out which polymorphisms might be there or might not be there. We can also do all different types of samples, and just like in clinical work, DNA is very stable.

The shortcomings, if we were going to do all genotyping and not look at the serum anymore, is that we can't get that antibody testing. I did remind you in case two with the sickle cell patient, we do need to know if the patient's made an antibody. And then the bottom one is the one that we're all learning together: how to inform the calling. How do you make these decisions when you have your genotype done by massively parallel sequencing? How do you decide what the answer is? It's actually quite challenging.

What the group has built is a target capture Illumina platform where the DNA is made into a library, and then biotin probes are used to select the blood group genes. Then there's a cleanup step, and the captured DNA is then sequenced. This is the gene panel that's been developed. We've targeted 42 genes, and there are thousands of different possible alleles in the population. We've been focusing on these Asian blood donors that we have because we've already done serology on most of them and we've already done SNP genotyping. So it's powerful in that if we next do the next-generation sequencing, we actually have deeper knowledge to be able to formulate what our final designations will be.

The library prep consisted of two insert sizes, and I'll show you that a little bit later. Then there's a capture step, and then the HiSeq was used on all of them. In addition, 96 of the samples went through the MiSeq, and that was to get the longer base pair read. The HiSeq is doing 100 base pairs, whereas the MiSeq is doing 250.

You don't need to look at all these numbers, but this is to tell you, as pathologists, we want to validate that what we did worked. This is an initial validation. This is basically showing that these blood groups here, which are some Kell blood group genes, some MNS, Colton, and Diego—different blood group genes that we have data from our SNP chip from the clinical lab—were found in very high concordance by the next-generation assay. So it just shows it works. We're able to find the known SNPs that we know are there based on the SNP typing; we're able to also find them with next-generation sequencing. That's really good news because this is the first time we've ever done it.

So then, back to the problems, right? Coming full circle back to the issues that I was talking about is that there are SNP-serology discrepancies, there's a need for finding rare donors, and there's a need for figuring out blood typing discrepancies. So looking at the Kidd typing discrepancies—this is the same data I showed you, just a little bit more broken out into Jka and Jkb—the samples that we had that fit this description are the ones that we decided to put through the pipeline first. So we started with kind of the harder ones first.

What we found, again, you don't need to see all the details, is that we found some common mutations. This one here—these were donors that were Jkb-positive by SNP and Jkb-negative by serology. 39 of them had that, and that was found to have a splice site mutation, which is well-documented in the literature as a common Southeast Asian change that causes this issue. In addition, we found another example in a South Asian that showed another published allele. You'll see that there's some that are blank; we're still working on them, and we have more work to do. But this is really exciting that many of these known, published alleles that are out in the literature were picked up by our platform. We had known that they were there, but we hadn't seen them before because we hadn't had the technology to look for them. But now we did this, and it worked. You can see at the bottom two, there's one that we think might be novel, and we think there'll probably be more novel ones as well.

Now, thinking about the RH blood group. I think I started the whole talk saying the RH blood group is so complicated. So now if you layer on top of the complication of the locus with thinking about the approach using next-generation sequencing, the way that I would originally think about genotyping RH is that you'd want really long reads because of these conversion events, and that because of all the mutations, that would give you a stronger result. In addition, because D and CE are so homologous, how are you ever going to align them with short reads? They're all going to not align correctly.

So that's the problem that the group has been faced with, and I'm going to take some time to explain this really fascinating slide that my colleagues in the group shared with me. The approach is using map quality. What that means is taking the short reads and using only the ones that map specifically to D or CE. That would give you a map quality of one, meaning that they don't map to both genes. Then, basically doing an easy arithmetic problem by looking at those that are D-only and those that are C-only and dividing them.

• If you have the same number of D and C reads—we know that you're supposed to ancestrally have two copies of CE—and you have two copies of D, those specific reads that are only D line up, you would have a ratio of one. Here, it is showing that the highest bar of red circles are samples that had that map quality of one. We would predict those to be RHD homozygous.
• The next one down, same thinking: but if you have two copies of CE and one copy of D, then you would be at 0.5. And so the middle bars of red are basically RHD hemizygous.
• And then at the bottom, if you have no copies of D and you have two copies of CE, it would be zero. So you would be predicted to be RHD-negative.

This is really important. This is the first thing that we have to know if we're going to genotype someone: how many copies of D do you have? And so this is really, I think, pretty amazing results. The red and the black represent serological results. After this was done, we went back and looked at the serological results of these donors and labeled the D-positives as red and the D-negatives as black. And you can see that it worked. The D-negatives, the ones that we predicted to be D-negative, indeed are all on the bottom. And the ones that we predicted to be D-positive are in the middle or on the top.

But there's a couple that didn't do that. I'm going to talk about two. The ones that didn't fall on the lines have been looked at in more detail. I'll talk about two of them. This one up here had a CE deletion, and so that is why the ratios basically got flipped and it shows up way up high. In addition, the one that I think is most interesting is down here. If you see that one off to the end, it's black. So it typed as serologically D-negative, but it's mapping as if it's supposed to be a D hemizygote. The reason for that is—I'm going to show you both the 100 base pair reads and the 250—is that this here should be a UAA, and it actually is a premature stop codon in an exon. So it would be predicted that even though the donor would have a D-negative allele and a D-positive allele, in the D-positive allele, we predict a premature stop codon. So indeed, serologically, they would be D-negative. This is showing the same thing again with the 250 base pair reads.

I hope I've shown you that the blood groups are interesting. The thing that's interesting about approaching them using massively parallel sequencing is that there are a manageable number of loci to approach in a targeted capture. There are some real challenging blood groups: RH and MNS to name the top two. I've talked to you a bit about RH. There's still a lot more that our group has to do to figure out a lot of that RH polymorphism. What I've shown you is just D-positive and D-negative.

And so the question I have, and that we all have is, you know, will the future be to use genomic methods to approach blood groups? Or even flipping it around, as people do more genotyping, right now if you do a whole exome or a whole genome, we cannot impute your blood type. So this work will also be able to take that data that's already out there and be able to impute the blood type because it's not really been done. The reference alleles need to be—everyone needs to work very hard on figuring out what all the calls mean and what they align to, given the technological advantages and disadvantages of the system.

One last thing that I want to think about, as I think about this with colleagues, is what can we learn from HLA? We don't have to read all these slides, but really, HLA has done the serology-to-genotyping transition over the past 20-something years. It used to be that the gold standard was serology, and that's what it is in blood banking right now. But as science advanced, as transplantation advanced, people found out that that's not really high enough resolution, that we actually need to be looking at the allele level. Now, when you get a bone marrow transplant, they're doing high-resolution genotyping of your HLA to find a bone marrow donor.

Although blood transfusion is different in that we usually give lots of different donors to one person as they're bleeding, and in HLA you're looking for one donor in the world, so it is different, I think that as technology advances and as we gain knowledge in the polymorphisms of the blood groups, it will become more compelling that the technology we have will be useful in this setting.

Just one other thing to think about here is the HLA allele discovery rate going up to 2015. You can see how fast it grew as they did more and more clinical genotyping to find new alleles. And that is sort of where I think we are for blood groups. We already know that there are many, many of them that we haven't seen or have seen very rarely that are probably more common the more that we look.

Conclusion

In conclusion, I hope I've convinced you that blood group genotyping provides really important, critical clinical information. For weak D, we're taking sort of a baby step in terms of our consensus statement, really telling the community, "You know, if you have these D typing discrepancies, you should do a simple genotype assay and figure out if they're weak D type 1, 2, or 3, because it's important to know what their blood type is." The way that serology reagents have been designed over the past couple of decades has, in my opinion, made more confusion about what the actual blood type is at the end.

I've also tried to talk a little bit about this recombination at the RH locus. It is a significant clinical issue. It's compelling in the sickle cell disease population, and as more and more clinical studies show that blood transfusion is one of the best therapies for sickle cell disease, we're going to continue to be faced with this issue.

In addition, rare blood types are hard to find. They still are hard to find, but we need to find those in our donors to support patients when they show up with antibodies. SNP genotyping is great—I use it in my clinical lab—but it has inherent bias towards European populations. I hope I've shown you with some of the cases that genomic sequencing has both technological hurdles but a lot of power to detect rare variants and to offer clinically useful information.