The Latest and Hottest Topics in ADCs: A Discussion With a Chemical Engineer

There is a lot of excitement among experts in the field of antibody-drug conjugates as new developments continue to come out in various disease states.

During the 16th Annual World Antibody-Drug Conjugate (ADC) Summit in San Diego, CA, top ADC experts gathered to discuss the hottest and most relevant topics in the field. Among them was Greg Thurber, PhD, a professor in chemical engineering at the University of Michigan School of Biomedical Engineering, who presented a seminar titled, “Contextualizing learnings from translational profiles of approved ADCs to unlock best approaches to preclinical dosing in mouse models”.

At the Summit, Thurber spoke with CancerNetwork® regarding his seminar, as well as other hot topics in the space of ADCs. From mouse models and patient-derived xenograft (PDX) models to new advancements, the conversation covered various facets of the space. Read here to catch the highlights:

CancerNetwork: What was the rationale for your seminar on contextualized learnings from the translational profiles of ADCs?

Thurber: We’ve seen a lot of progress and a lot of new approvals, and there’s a lot of excitement in the field of ADCs. One of the challenges the field has struggled with in previous years, because it’s taken a while to get to the current point, is why some molecules look so promising in preclinical models and then end up failing in the clinic. The take-home message from my talk was that if we dose these preclinical models at the correct level, close to the clinically tolerated doses, then the results do match what we see in the clinic. That could be a good marker moving forward to select molecules that will have success.

What takeaways from that seminar do you want to highlight?

The other aspect that I discussed was some of our more recent work looking at multiple target-independent mechanisms of action, because there are a lot of reports in the literature and in mouse models of how target-independent effects might be playing a role in the clinic. In that work, what we did was we took all these different factors, whether it’s immune effects, extracellular protease cleavage, macrophage uptake and payload release, and free payload in the blood, and put all of them into a single framework. This is another advantage of the simulations—we can take clinical data, in vitro cellular data, and preclinical animal data, put them all in the same framework, so we have an apples-to-apples comparison. That allowed us to look at the magnitude of these impacts. This helps put the clinical data into context. One of the important things that came out of that, however, was that the target-mediated uptake is the biggest driver of efficacy. All the effort that’s being put into identifying the right targets and getting local delivery into the cells is worth it in the end because that’s going to far outweigh any of these other effects. Those other effects can be important, and they can help us understand some of the clinical data more clearly.

What are the most significant disconnects between mouse model pharmacokinetics/efficacy and human clinical outcomes?

A traditional way of approaching the preclinical models is to do dose escalation until you see your response, but one of the underappreciated facts in the field is that mice tolerate extremely high doses of these molecules. Many of these payloads don't work very well on mouse cells, and so the mice tolerate very high doses. If you're using a xenograft model that's growing a human tumor inside the mouse and the tumor is very sensitive, you have this artificially high therapeutic window, and you end up getting these complete responses in mice. Once you go into humans, the tissues are much more sensitive. You have to dose much lower, and then you don't get the uptake and the distribution of the ADC in the tumor tissue, and you don't end up with responses.

How can efficacy results sometimes be counterintuitive in animals and in the clinic?

One of the things the field is learning more from these newer agents is that there’s no one-size-fits-all ADC that will work for any given target. You have to design it with purpose, for a given target. We’ve had a few different counterintuitive scenarios that we’ve shown in preclinical models where, for instance—and this also translates to the clinic—the most potent compound that kills cells easily in a petri dish when you’re just growing a monolayer of cells looks the best in vitro on cells, but that also comes with this liability of high toxicity. What might look good with some of the ultra-potent payloads, once you get to the clinic, you don’t end up with a therapeutic window. By selecting lower potency payloads—and this is one of the major reasons why the topoisomerases have been so successful—is that they have a low enough potency that you can dose high in the clinic. That allows you to penetrate the tumor tissue, reach all the cells, and drive much higher efficacy. There are other counterintuitive examples that we’ve shown in the past. Sometimes, very rapid internalization can hurt efficacy. Now, this is a unique case. This is not in all cases, but sometimes, at these lower doses, fast internalization can limit the tissue penetration, so you don’t reach all the cells. In other cases, we’ve shown that slower internalization allows the molecule to penetrate deeper and reach more cells, and you end up with higher efficacy.

How can lower toxicity be achieved in the development of ADCs?

One of the things that I've argued for, for many years, is lowering the drug-to-antibody ratio [DAR], particularly with the more potent compounds, or adding in an extra antibody to make sure that your antibody dose is high enough. With some of these more potent compounds, that is one way to improve the therapeutic window: by dosing more antibody. Now there's a lot of interest in alternative and new payloads. In those cases, I don't think you'll necessarily want to go with a lower DAR. Off the bat, you need to figure out what the potency of your compound is. If you have much lower potency, let's say lower potency than even a topoisomerase inhibitor, then you might want to go on the other end of the spectrum and try to use these very high DARs if you have a low potency compound that also might have better tolerability. Finding that right balance of the potency of the compound and the delivery will then maximize your therapeutic window and help improve its tolerability.

How do PDX and computational models help understand patient responses in settings with various approved ADCs?

The computational models are very nice [being] able to forecast what you expect to happen in the clinic. I often argue for running a parallel campaign where you’re making computational predictions of these [ADCs] at the same time as you’re doing the experimental work. What the computational work allows you to do is to forecast what you expect to happen in the clinic. Even at the early discovery stages, you might look at a new target, and you might understand something about the biology, and you can look at the internalization rate and forecast what that is ultimately going to do in terms of your efficacy in the clinic. As you’re screening for antibodies, you might find one that internalizes much faster or much slower, and then you can plug that into your computational model and predict what the impact is going to be at the clinical stage. As you move along the pipeline, once you get into your animal studies, you can update the model with any sort of differences from your predictions. It gives you more confidence to move forward, that things are behaving as you expect, and that you already have this clinical forecast. That’s one of the strong advantages of doing the computational work in parallel with the experimental work.

How does this affect the development of ADCs?

We have enough understanding now to know that we can't blindly go into these screens and pick the most potent compound. We need to understand how we're developing the animal models to pick the molecules that are going to have the greatest efficacy. Just because you can get a complete response in the mouse doesn't mean that it's going to be a good compound in the clinic. That translation during the development phase and selecting for molecules, or on the contrary, if a molecule is not giving you efficacy at a dose that you expect to be able to give in the clinic, to stop that program and move on to a more promising agent. That can help you and help companies harness the resources and focus on the most promising molecules.

What new advancements are you most excited about in the field of ADCs?

A lot of us are looking at different combination therapies and how that will impact clinical efficacy. One of the areas where our lab is interested is understanding how these molecules interact with the immune system, because these are targeted chemotherapeutics. I would argue that they have a lot of ideal properties for driving a strong immune response, and we're seeing that with some of the combinations of immune checkpoint inhibitors and ADCs. What other combination partners can we leverage to drive that immune response? Because if you can drive a strong immune response, you can get a very durable response over years, as opposed to months. That's the future of where we went ahead with cancer therapy, and ADCs are a great mechanism of doing that.

Reference

Thurber G. Contextualizing learnings from translational profiles of approved ADCs to unlock best approaches to preclinical dosing in mouse models. Presented at the 16th Annual World ADC Summit; November 3-6, 2025; San Diego, CA.