A Conversation with Chris Bond, VP at Kite Pharma: Using data to drive CAR T R&D
Q: Tell us about your background. How did you first become interested in science? And how did you decide that you wanted to get into biotech and, specifically, cell therapy?
It all started in seventh grade, learning about Mendal’s peas in science class. I found it captivating and exciting and, well, I seemed to do well at it. In college at the University of Washington, my plan was to get a PhD in physics and teach. I was working at a linear accelerator on campus when a good friend of mine at the Fred Hutchinson Cancer Research Center introduced me to her PI. This conversation led to subsequent discussions regarding the nature of protein structure and work on protein folding which I found absolutely fascinating. Ultimately I decided to study biochemistry, rather than physics, in graduate school.
I went on to do my postdoc at Genentech’s industry program and then met my now-wife, Katrina. After Genentech, Katrina and I actually ended up doing what my graduate advisor thought was career suicide: she and I took off for a year and traveled throughout Southeast Asia, India, Nepal, and China, and just took advantage of our youth. We got married at the end of that.
Once I came back to the states, I took a position with Chris Garcia at Stanford University studying determinants of MHC restriction. This was in the mid-2000s, and I started reading some of Steven Rosenberg’s seminal work on TILs (tumor-infiltrating lymphocytes) in melanoma. Reading his work was one of the “aha moments” of my life. Harnessing patients’ own immune systems to attack malignant cells was something I suddenly knew I had to pursue.
Q: In the past 10 years or so, there have been so many developments for CAR T. But we still haven’t really cracked the code. What do you think is going to lead to our real mastery over it?
I would say three things. First: Overcoming the challenges around safety. We need to continue to evolve our understanding both from the clinical management perspective, but also from the basic product design perspective to ensure that we’re developing the safest product.
Second: Driving the accessibility of CAR T therapy. The field, and Kite in particular, has seen a lot of success. But we’re still a very tailored, personalized therapy. Bringing this to as many patients as possible is going to be critical to advance the field.
Q: One of the most exciting but challenging things about CAR T therapy is the extent to which it’s personalized for each patient. How do you think about patient data with respect to preclinical efforts?
That’s a great question. It’s an area that the field has done well so far, but we need to continue to evolve. Certainly, handling patient data has to be done with a lot of caution. We need to understand the interactions of the cells with the patient’s tumors and immune system. We need to work all of that back into R&D to continue to evolve our platforms.
Q: So you’ve got to get data from the clinic, and then you need to analyze it alongside all the data from everything else that’s happening in R&D. How do you actually go about collecting all this very complex data from so many different sources?
We are collecting a lot of data including phenotypic characterization, cytokine production, cytolytic activity, as well as the transcriptional and epigenetic landscape from both the engineered cells and the tumor. It is critically important to have data warehousing and curation so those experiments can be queried in a broader sense. You need to be able to dig into the data and then correlate that to functional outcomes. Only with all of that can we start to understand, which are the most relevant attributes for an effective CAR or engineered TCR.
So how are you curating the data? How are you storing it? Are you setting up experiments in standardized ways so that you can easily compare across experiments? The front line scientists need to have access to this, analyze it, generate hypotheses, and then follow up on those leads. At Kite, we’re investing heavily to build out a system, so that our data is curated in a way that makes it accessible.
Q: Right, so there’s this huge data challenge and the need for warehousing, curation, and building out analytical models to actually crunch all of this data. But how successful have you seen companies actually be at that?
Without speaking to specific examples, I have seen it implemented in a way where you can pull predictive phenotypes out of cell products that are possibly correlative of some level of clinical response. That’s just been successfully done in a limited scale. I’m confident that with the right access to data and analytics, this type of approach is the right way to go.
But especially on the preclinical research side, there’s a lot left to do. Research groups tend to be a little more siloed, which makes it more challenging to compare apples to apples when you’re looking at different programs, experimental protocols, and procedures.
It is critically important to have data warehousing and curation so those experiments can be queried in a broader sense. You need to be able to dig into the data and then correlate that to functional outcomes. Only with all of that can we start to understand, which are the most relevant attributes for an effective CAR or engineered TCR.
Q: That data strategy can drive a lot of process iteration, but I can also see how it could help in terms of “failing fast.” Especially when there’s still the big question of whether certain candidate attributes are even predictive of downstream clinical success, how do you approach failing fast?
Some of it is having well-defined milestones that allow you to say, “All right, this may be a risky platform development endeavor, but at least can we get to this point.” That allows us some ability to say we’re on track and moving towards what we think is success.
The other part of it is comparing our work to as many of our historical controls as possible. So for a second generation CAR, how does it compare to what we’ve achieved in the past? We’ve now evolved enough to know how long things should take for some of the CAR development efforts, for example.
Q: What’s it like to make business decisions based on that data? And from the standpoint of running a business, what’s it like when so many parts of what will make you successful are, to a greater extent than perhaps some other industries, up in the air?
It’s a little frightening, because you’re asked to make decisions with uncertainty and incomplete data. Scientists are data-driven people, and we want to be able to see as much data as possible before we make a decision.
Q: When it comes to evolving a cell therapy R&D organization – both in terms of evolving the business and in terms of evolving the science – what role do you think big pharma should play? Gilead acquired Kite, and Celgene acquired Juno, and now Celgene’s being acquired by BMS. Novartis took a different approach, but largely M&A seems to be the name of the game. What are the upsides and downsides to that?
In these emerging modalities, there are a lot of things that we can only learn once we test products in the clinic. And that’s an expensive game to play. Having major players like Gilead, BMS, and Celgene enthusiastically invested in the potential of this platform really enables us to explore and ask, “What are the major drivers of response, relapse, and how do we continue to engineer our products so they’re safer, more effective, and accessible?”
How are you curating the data? How are you storing it? Are you setting up experiments in standardized ways so that you can easily compare across experiments? The front line scientists need to have access to this, analyze it, generate hypotheses, and then follow up on those leads.
Q: It’s been great to see that model working well for Juno and Kite.
Yeah, absolutely. I think people are striking that balance and really recognize the unique challenges, but you still need to test things in patients to really drive learnings. Pharma’s got a role to play there, because that’s a complicated endeavor, and getting it right takes a lot of experience and resources.
Q: To the extent that you can talk about it, what’s your transition from Juno to Kite been like?
What’s most striking is how everybody’s feeling the same constraints and pressures around, “How do you make decisions? How are we choosing between different options with sparse datasets?” I would guess that that’s the biggest single thing that’s common across the entire field. At the end of the day, there are quite a lot of things that we could engineer into cells – it’s an embarrassment of riches. But that means that you’ve got to be able to choose, and to choose with limited predictive data is a very hard thing to do.
Q: Lastly, what trends in life science R&D, in general, are you most excited about today?
I’ve been spending a lot of time thinking about how with DLBCL (diffuse large B Cell lymphoma), we’re starting to see responses with multiple types of approaches. It’s exciting for patients, and exciting for the field, and it is going to drive how we think about where cell therapy fits in and complements other therapeutic options. I’m also just really excited to improve our product in a way that allows us to provide access to more patients.