Allogeneic cell therapies are inherently complex. But they also hold the potential to scale to treat large segments of the population.
Martin Maiers, MS, is the Vice President of Innovation at the National Marrow Donor Program® (NMDP)/Be The Match®. He and his team develop software and analytical methods used to predict factors that can impact outcomes for patients who receive an allogeneic cell therapy.
Through Be The Match BioTherapies®, which is an extension of the NMDP/Be The Match, Maiers has worked with many allogeneic cell therapy developers in both the early stages of their therapy development and as it moves through clinical trials.
Here he discusses:
- The challenges companies have encountered when setting up their cell banks
- Strategies to reduce the complexities of HLA matching that can be used in the development of new allogeneic cell therapies
- How historical data can be used in matching algorithms and tools to make allogeneic cell therapies better
- How HLA matching impacts cell bank setups
- The keys to maximizing population coverage
- One of the biggest challenges he sees coming in the next five to 10 years (it’s probably not what you expect)
What are some of the unexpected challenges allogeneic cell therapy developers have encountered as they work to set up their cell banks?
The sheer diversity of HLA is usually the first thing that hits and is overwhelming for many people. On average, most people are different from one another. If you take two people at random, they’re most likely not the same at any HLA.
Historically at the NMDP/Be The Match, we’re looking for a match for one patient from one donor. That in itself is complex.
But with allogeneic cell therapies, their target may be the population of a whole country. Or even the whole world. How do you develop a cell bank that will cover a population that large?
Then there is the complexity of the typing itself. There has always been, and there will always be, a gap between your true phenotype and what we know about it.
I think it’s a bit of a shock when people run into it. You figure out what the HLA type could be. You’ve narrowed down the possibilities. But, there is no unambiguous HLA. That just doesn’t exist.
Fortunately, our organization has a lot of strategies for dealing with the diversity and ambiguity of HLA. Both of those realities are what we have had to deal with from the start when matching patients with donors.
We have the experience, we have the tools and we have the methods for communicating about HLA in a standard way. We’ve developed it over time as we’ve gone back and said, “How can we do this better?”
Using population genetics is one strategy for dealing with the challenges of HLA. So, using data from large groups of people to make predictions about a specific individual. Through HapLogicSM – the matching algorithm we use for finding donors or cord blood units for patients who need hematopoietic stem cell therapy (HSCT) – we make predictions based on haplotype frequencies.
The data on our donor registry and those across the world reach across spans of time where technologies have changed a lot. We’ve built an imputation method for matching that looks at the complexity, context and technologies of the typing to normalize out the data and see things consistently.
We can apply that method to look at coverage when building an allogeneic cell therapy bank. An imputation platform is something that we see as a major category of tools to be able to manage pulling data from the expanse of a donor registry across spans of time when things have changed a lot.
The NMDP/Be The Match has amassed a huge amount of data over the past 30+ years. How can that historical data be used in matching algorithms or other tools to make allogeneic cell therapies better?
When it comes to historical data and how we can apply it to future therapies, population genetics is our friend. Worldwide there are 38 million donors and over a million cord blood units listed on registries. Our experience of knowing what we know about those people helps us make predictions.
Take matching for HSCT as an example. Let’s say there is someone on the registry who has incomplete typing. There are five genes and that person is only typed at three of the genes. You can use data from the collective to predict the variance at the other two genes.
That’s one application. But we can also use the historical data as a predictor of the future at a population level.
For example, a company wants to build an allo cell bank. They say they want to treat 95% of the patient population. How do you know what the patients are going to look like? You can use this frequency data to do that.
Focusing on donor characteristics and efficient selection to improve allogeneic cell therapy outcomes
You can also ask questions like, “If the patients were in China, how would you build a bank? Would the bank be different if the patients were in Korea? How would you build a bank that would serve 10 countries equally well?”
It’s a bit of a trick and it’s getting more difficult. If you look at the world today, long-held principles of population genetics just aren’t true. Look at Vice President Kamala Harris. Her parents are from different continents from each other. Over one in five of the donors we recruit today answer categorically two different groups for race and ethnicity.
This is becoming a generation of new population genetics. So that’s a challenge. To meet the challenge, we have to continue to evolve our methods. Because if you’re trying to build a bank for the future patient, we need to have some way of predicting what that need will look like.
How does HLA matching impact the cell bank setups?
We’re moving beyond HLA matching as we’ve thought of it traditionally and moving into functional targeting, avoiding, rejecting and inducing specific immune function as part of our HLA goals.
But even so, the thing that has been generally consistent is that HLA is there and we need to try to avoid mismatching as much as possible. If there is a mismatch, it can’t mismatch in the direction of rejecting.
The interesting thing about HLA rejection is that some people have antibodies to HLA. We’re trying to identify donor-specific antibodies and characterize the patient for their antibody profile.
By capturing the patient’s antibody report, we can use that as a way to say, here is a cell that mismatches, but it’s ok because the patient doesn’t have antibodies for it. The potential to expand that is huge.
Donor Pool Size Matters for Allogeneic Cell Therapies
If you were trying to build a bank for an allogeneic CAR-T therapy, it’s not practical to have a goal of having an 8 of 8 match for everybody. But you could try to select common types and optimize it to minimize host-versus-graft direction as much as possible to avoid rejection by the patient.
In my experience, it all starts with this platform of producing something that will not get rejected by the most patients. Then from there a number of different directions of where you could go.
What do you see as the keys to maximizing population coverage?
Gene editing is the one big technology that is very exciting for maximizing population coverage. But with all the complexities that go with gene editing – the expense, the potential off-target effects – it will be another variable in the model.
For example, you can do some gene editing, which will expand how many patients you could cover with your treatment. But for certain HLA types, removing a gene solves one problem but introduces another.
We can factor that into our simulations. Maybe you’re building a bank where you come up with a set of a dozen genotypes and also recommend removing three specific HLA genes to expand your coverage. You could show that by removing HLA-A, for example, match rates go up tremendously. But not if you remove HLA-B or C because those two genes stick together.
Also understanding which side effects that editing could produce or how it could diminish the effectiveness of the therapy will be important.
This is a medical breakthrough. Unequivocally. I don’t recommend that we ignore it. But I also don’t think that just gene editing our way out of this HLA barrier is going to lead to overall improvement of patient care.
Beyond gene editing, I think the key to maximizing population coverage is to understand the genetic diversity of the world. Exploring genetic diversity will help us maximize our coverage because we’ll know what we’re trying to maximize to. Our predictions are as good as the data that goes into it. The potential is there to get a lot more precise with understanding that diversity.
Why is it important for cell therapy developers to have a diverse starting pool and how would that minimize risk as they scale towards or assume commercialization?
There are so many unknowns. That’s something that is a fairly consistent theme when we are engaging with organizations that are trying to build cell banks.
Because there are so many unknowns, having more diversity to be able to respond to the science or to the market or both is good.
What are some of the biggest challenges you foresee for allogeneic cell therapy developers over the next 5 to 10 years?
I see the challenges more so from the perspective of the physician. The number of products that are working their way through the pipeline means this is going to be a very complicated space.
Physicians will need a way to navigate that. We need to be able to say, “Here is the menu of different cellular therapies that might be suitable for your patients.”
Who do you trust to provide a more neutral navigation through that sea of opportunities and challenges? And back it up with outcomes data?
I think the registry community can mature into this space because we have been doing something similar for transplant. The CIBMTR has already expanded into this space. We collect outcomes data on every FDA-approved CAR-T therapy and other cellular therapies.
Our opportunity is to use that data to fuel clinical decisions and support systems that will help us navigate this space in the future based on evidence.
Enjoy what you read? Subscribe to our Cell Lines blog to be notified when a new story is posted.