Fibrosis is the body’s attempt to heal gone wrong. When tissues are repeatedly injured — by chronic inflammation, infection, or metabolic stress — the normal repair process can spiral into excessive scarring. Functional cells are gradually replaced with stiff connective tissue rich in collagen and other extracellular matrix proteins, thickening and hardening the organ until it can no longer do its job, as seen in end-stage liver disease, kidney disease, idiopathic pulmonary fibrosis (IPF), and heart failure.
Beyond organ scarring, fibrosis also contributes to tumor invasion and metastasis, chronic graft rejection, and the progression of many muscle diseases. Although fibrogenesis is increasingly recognized as a major driver of morbidity and mortality in chronic inflammatory conditions, there are few — if any — therapies that specifically target the underlying disease process.
Now, researchers from Abalone Bio and the Icahn School of Medicine at Mount Sinai have reported the first successful demonstration of antibody drugs that selectively activate CB2 (cannabinoid receptor 2), a long-sought anti-inflammatory target for fibrosis. The findings, published in Molecular Pharmacology, suggest a new therapeutic strategy for fibrotic disease and the potential of AI-driven antibody discovery to tackle targets long considered intractable.
A target undone by specificity
At the core of fibrosis is chronic inflammation. Immune cells like macrophages normally help repair tissue, but when they stay active too long, they release cytokines that instruct fibroblasts to produce excess collagen, driving scar formation.
“Fibrosis is the common denominator in many chronic diseases, especially liver disease, which is responsible for roughly four percent of deaths globally,” Richard Yu, CEO and cofounder of Abalone Bio, told DDN. “While current 'blockbuster' drugs often target upstream drivers, like metabolism or fat buildup, they don't always stop the resulting inflammation that leads to fibrosis.”
CB2 has been an attractive target because it directly suppresses that inflammation and promotes tissue repair.
—Richard Yu, Abalone Bio
This is why CB2 is of such interest. These receptors play a key role in controlling inflammation. CB2 is a G-protein-coupled receptor (GPCR) expressed primarily on immune cells, including macrophages, T cells, B cells, and other leukocytes. When activated, CB2 dampens inflammatory signaling and reduces the release of cytokines and other pro-inflammatory signals. By limiting this overactive immune response, CB2 helps prevent the cascade of events that leads to excessive collagen deposition and organ scarring.
“CB2 has been an attractive target because it directly suppresses that inflammation and promotes tissue repair,” said Yu. However, despite decades of effort, drug developers have struggled to translate that promise into viable medicines. The receptor is highly dynamic and closely resembles CB1, its near-identical counterpart in the brain. As a result, small-molecule agonists often hit CB1 as well, leading to psychoactive side effects and, in some cases, even promoting fibrosis rather than suppressing it.
Rethinking how antibodies are discovered
Antibodies, in principle, offer a way around this specificity problem. They are large, highly selective molecules, generally excluded from the brain by the blood–brain barrier, and well-suited for chronic diseases. However, GPCR agonist antibodies are extraordinarily rare.
Traditional discovery methods prioritize binding affinity first, screening huge libraries of candidates for those that stick tightly to a receptor. But for agonists, it’s not enough for a molecule to bind — it has to bind in exactly the right way to activate the receptor and trigger the desired biological response.
Abalone Bio’s approach flips the script. Rather than training machine-learning models on binding data or structural predictions, the company’s Functional Antibody Selection Technology (FAST) platform is built around large-scale measurements of biological activity. Using engineered yeast cells expressing the receptor on their surface, each antibody variant is tested for its ability to turn the receptor “on.” High-throughput sequencing then maps which sequences produce the strongest activation signals.
FAST identifies special antibodies that recognize rare, specific conformations of the receptor that traditional methods would have simply overlooked.
—Richard Yu, Abalone Bio
“By using the FAST platform to screen for functional activity first, we found antibodies that don't just 'stick' to the receptor; they actively 'turn' it into the position that has therapeutic activities,” said Yu. “Interestingly, some of our most effective antibodies have lower binding affinity as measured by traditional methods, but higher potency. FAST identifies special antibodies that recognize rare, specific conformations of the receptor that traditional methods would have simply overlooked.”
Machine learning models take the results further, designing new antibody sequences predicted to maximize functional activity while maintaining stability and developability. With this approach, Abalone researchers identified two antibody agonists, AB120 and AB150, that activate CB2 while sparing CB1.
Targeted CB2 activation reduces scar formation and inflammation
Both of these candidates are engineered antibodies made from a small, single-domain binding unit attached to a human antibody backbone. This design helps the antibodies stay in the body longer while reducing unwanted immune reactions.
When tested in cellular assays, the antibodies showed robust suppression of inflammatory cytokines including IL-6, IL-1β, and TNF-α. Crucially, those effects were absent in CB1-expressing cells, highlighting the specificity that has eluded small-molecule approaches.
The team also tested AB120 and AB150 in precision-cut human liver slices, an ex vivo system that preserves the multicellular architecture and signaling environment of diseased tissue.
“Most importantly, we moved beyond the petri dish and demonstrated that these antibodies work in human liver tissue,” said Scott Friedman, hepatologist at the Icahn School of Medicine at Mount Sinai. “We showed they can stop the production of collagen (the building block of fibrotic scar tissue) and quiet the inflammation that drives disease. It bridges the theory and in vitro results into human tissue, one step further to a product that could help a human patient.”
“This provides the crucial evidence that our approach translates from the lab to the complex environment of human disease,” said Yu. These findings provide a strong case that AB120 and AB150 could represent a novel therapeutic strategy for treating fibrotic and inflammatory diseases driven by chronic macrophage activation. More importantly, they show that AI-driven, function-first antibody discovery could transform how challenging receptors like GPCRs are drugged, opening new possibilities to precisely modulate some of the most elusive targets in human biology.
Implications beyond CB2
To date, only a handful of GPCR agonist antibodies have been reported in the scientific literature. Many researchers have viewed GPCR activation as the domain of small molecules and peptides, not biologics. The Abalone–Mount Sinai study suggests that limitation may be more technological than biological.
As AI tools mature, function-first antibody discovery could open the door to many previously “undruggable” targets. GPCRs regulate a huge variety of physiological processes — from metabolism to immunity to cardiovascular function — yet only a small fraction have been successfully targeted with therapeutic antibodies. Abalone’s success with CB2 demonstrates that with the right combination of synthetic biology, high-throughput functional screening, and machine learning, antibodies can be designed not just to bind a receptor, but to precisely modulate its activity in therapeutically meaningful ways.
