Developing Functional Antibodies That Modulate Target Activity

Abalone Bio’s co-founders CEO Richard Yu and Gustavo Pesce met while they both worked at Sydney Brenner’s non-profit Molecular Sciences Institute in Berkeley on an interdisciplinary project to measure, model, and predict the behavior of the yeast pheromone response pathway—a model system of the fundamental natural process of how cells respond to their environment. 

Ten years later, the two of them were thinking about how they could apply their learnings spanning structural biology, systems biology, synthetic biology, cell biology, genetics, and computer science. They realized they could engineer yeast into a discovery tool to find functionally active antibodies against targets such as G-protein coupled receptors, the receptors that control signal cascades that are dysregulated in many disease states. 

They founded Abalone Bio to develop antibody drugs to activate challenging targets across disease areas, including metabolism, inflammation, pain, and cancer. 

“The gap that we're trying to address for these really promising targets, like GPCRs and other membrane protein targets, is the unfulfilled potential of antibodies,” says Yu. “They're super-specific, well-behaved molecules as far as drugs go, but it's been really hard to find antibodies that can do much of anything to these targets except to bind and inhibit.” 

Yu says these antibodies are large, structurally and chemically complex molecules. There should be ones out there that can specifically bind your protein receptor target in just the right way that stabilizes a particular structure and therefore yields a particular function of interest, such as target activation (agonism). Such molecules would address the big challenge he saw with other drug modalities, like small molecules or peptides, which is finding molecules that are both specific and efficacious; you could usually get one or the other.  Unfortunately for antibody discovery, existing high-throughput measurement technologies exist for attributes like binding affinity but not the ability to affect the activity of targets. Functional activity is usually tested very late in the discovery process, and, by that point, antibodies have been pre-selected for high affinity binding on any part of the target protein, like an immunogenic epitope on an N-terminal tail, rather than enriched for activity, which likely requires binding to the multiple extracellular target epitopes in the right relative orientations to stabilize the right structure and have a functional effect.

“Our hypothesis was this was fundamental a search problem. If you directly searched far and wide enough in sequence space up front, you should be able to find antibodies that both bind specifically and can have a functional activity like agonism,” he said.

So, Abalone Bio built a platform called FAST—Functional Antibody Selection Technology—that uses engineered yeast to measure how antibodies interact and change the function of a particular target of interest. It then transforms those data with computational tools, including AI protein large language models, to understand what antibody sequences have what functions against a target. The models can then be used to predict and generate antibody sequences with improved function against a target of interest. 

“We can measure the activity of antibodies with orders of magnitude higher throughput than other technologies,” says Yu. “In a single discovery campaign experiment, we're typically measuring activities of a hundred million antibody variants at once. With this functional data, we can not only discover these rare agonists, but what’s been really exciting is that we’ve shown that, with AI, we can learn from these data and design better ones.”

Yu uses the gold mining analogy. You could go to a river and pan for gold, find some flakes, start digging around there, and if you're lucky, you can search locally in that space to find your gold nuggets. But what you really want to do is measure the landscape using modern surveying tools and combine those data with principles of geology to understand where there should be a gold vein and just go to the right place for a major gold deposit. That's the difference between discovery and design. 

“By measuring enough data, you understand the antibody function landscape so you can design these molecules rather than just depend on luck to find them,” Yu says.

Abalone Bio’s first antibody candidate targets the endocannabinoid receptor CB2 because it sits at the intersection of several interesting biological processes and, therefore, could be applicable in a lot of diseases. And several companies already had small molecules targeting CB2 in clinical trials. 

“When you activate this receptor, it has anti-inflammatory activities,” says Yu. “It's also analgesic in a non-opioid related way, and it's also directly anti-fibrotic, in addition to being indirectly antifibrotic through its anti-inflammatory actions.” 

From a target perspective, Abalone Bio sees CB2 as a really good use case for an activating antibody that binds allosterically, i.e., to parts of the target that are not where the typical chemical signal binds, and then massaging the shape of the target to activate it. 

CB1 is the other major cannabinoid receptor in humans. CB1 is primarily in the CNS, and CB2 is primarily peripheral, expressed in immune cells and peripheral neurons. Abalone Bio is looking at non-CNS diseases like peripheral neuropathy first.

“We thought that antibody agonists could address one of the major problems with this target, which is lack of specificity. CB2 small molecules- even ones that are purportedly quite specific for CB2 over CB1 in in vitro cell-based assays—once you put them in vivo, you often see some crossover reactivity with the unwanted target,” says Yu. “In certain tissues activation of CB1 is actually pro-inflammatory and pro-fibrotic. So, you really don't want to have any cross reactivity between the two because you're causing more of the disease and undercutting the effect of your drug.” 

The small molecule binding sites for CB2 and CB1 are very similar. However, the extracellular loops of the GPCRs where the antibody can bind are very divergent- those regions in CB1 and CB2 share ~25% sequence identity, making it very unlikely to find an antibody that binds both, much less activates both. 

“We thought this was a perfect target for our platform and the types of molecules that we could find,” Yu says. “And sure enough, the antibodies that we found are very specific to CB2, we've never seen a hint of CB1 cross-reactivity in in vitro or in vivo assays.”

Abalone Bio has raised $18 million to date and is in the middle of raising a new round of capital to fund the discovery and development of internal pipeline candidates as well as programs for co-development partnerships, like its CB2 program.

The company listed its first CB2 engineered candidate, ABt285, on Bioneex in early June and quickly got a couple of partnering matchups. Yu likes how the platform makes connections and is excited to start talking to partners about the interesting use cases in peripheral neuropathy. The company has a lot of data on chemotherapy-induced peripheral neuropathy and is expanding that into diabetic neuropathy. It also has some good data in liver inflammation and NASH/MASH. 

“So yeah, really exciting times for that molecule,” says Yu. The company has gotten a lot of interest in it and is ready to partner it for development, and use revenues to support development of other internal programs. The company is also interested in co-discovery partnerships to help other groups find functionally active antibodies for their targets.

“GPCRs are everywhere in biology, so we can be quite ecumenical in our disease area selection,” says Yu. “We focus on targets where you really, really need specificity, or where other drugs are poorly behaved (like some peptides) or don’t have the chemical and structural complexity to functionally interact with the target (as is needed for receptors with large ligands, like chemokine receptors).”