Peptone Biotechnology Spinout 2022

Peptone Biotechnology Spinout 2022

Peptone: finding opportunity in disorder

Combining experimental and computational tools, Peptone is targeting disordered regions of proteins previously thought to be undruggable.

"There are still relatively few companies with a core focus on intrinsically disordered proteins, and Kurgan is glad to see a startup embracing the risk and opportunity associated with these tricky targets."

The rise of AlphaFold may not have put structural biologists out of work, but it has undeniably transformed the field by allow- ing researchers to nearly instantaneously generate high-quality predictions of struc- tures for many proteins that would take months or years to experimentally deter- mine. But we are still far from a full solution to the ‘protein structure prediction problem’. “There’s a ridiculous amount of structural disorder in proteins,” says Kamil Tamiola, CeO and co-founder of London, england– based Peptone. The total number of such proteins remains an open question, but he estimates that as many as 18,000 human proteins—roughly 75% of the proteome—may contain stretches of 70 or more amino acids that lack a sta- ble and defined struc- tural conformation. Such ‘intrinsically disordered’ regions currently remain an insurmountable challenge for AlphaFold or any other structure-prediction program. Tamiola became fascinated with intrinsic disorder as a graduate student at the univer- sity of Groningen in the netherlands, where he learned how to combine nuclear magnetic resonance (nMR) mass spectroscopy with molecular dynamics simulations to charac- terize the dynamic behavior of such proteins. In 2016, he and chief operations officer Matt Heberling cofounded Peptone with the goal of building out this experimental and com- putational toolbox to develop new therapeu- tics based on insights derived from these still largely undruggable targets. Disordered proteins are mostly known for their involvement in conditions such as Alz- heimer’s and Parkinson’s, for which messy clumps of misfolded protein figure promi- nently in the disease pathology. But in fact, many proteins rely heavily on unstructured elements to execute their normal biological functions. Lukasz Kurgan, a bioinformatician at Virginia Commonwealth university, notes that these elements offer a flexible interface for establishing interactions with a range of DnA, RnA or protein targets. “That disordered region can essentially refold depending on the interaction, and that gives it a lot more bang for the buck,” says Kurgan, who has no ties to Peptone. “There’s still specificity there, so it’s not like you can bind to anything, but there’s diversity with that specificity.” Peptone’s strategy was initially a hard sell to investors who didn’t immediately grasp the value of going after such messy and chaotic targets, says Tamiola. But AlphaFold’s emer- gence helped crystallize the company’s pitch. “We started being perceived by the market as the guys that work on the other side of the folding problem—that is, molecules not suit- able for AlphaFold,” he says. And in June 2022, the company secured $40 million in series A funding from investors including F-Prime Capital and Bessemer Venture Partners.

Their drug-discovery process begins with identifying suitable targets; the Peptone team recently published an article describ- ing an algorithm called ADOPT (Attention Dis-Order PredicTor) that applies deep learn- ing to identify stretches of disordered amino acids. Coupling this with information from other databases reveals links to disease and pathogenic mutations that might provide additional insight into the underlying biol- ogy of the protein. Once a protein of interest has been iden- tified, it is extensively analyzed using a multi-stage drug-discovery platform that the company has dubbed ‘Oppenheimer’. First, the molecule is isolated and subjected to experimental characterization at Peptone’s recently constructed R&D facility in Bellin- zona, Switzerland. Among other techniques, the company uses a proprietary form of hydrogen–deuterium exchange mass spec- trometry (HDX-MS), a technique that analyzes the exchange of hydrogen ions in a protein with deuterium ions in its surrounding solu- tion to derive detailed insights into that pro- tein’s conformation and dynamic behavior. Tamiola says that their system “operates at a time resolution 100–1,000 times quicker than anything on the market” and can be applied to a much broader range of proteins than the nMR spectroscopy he used as a grad student. That doesn’t mean that this experimental pro- cess is necessarily easy; simply synthesizing stable and functional protein can be a hefty challenge, especially for molecules with siz- able intrinsically disordered elements. Finally, these data are analyzed in a series of molecular dynamics simulations using a software pipeline developed in collaboration with Kresten Lindorff-Larsen, a computational biologist at the university of Copenhagen in Denmark. This translation from experimental analysis to computational interpretation is the biggest challenge, according to Tamiola, but if successful, it can generate a detailed ‘map’ of the structural states that disordered regions transition through and the relative stability of those states in different environmental conditions. Armed with this knowledge, the researchers can then begin to think of ways to intervene in this structural repertoire to achieve a clinical benefit. The company’s most advanced programs currently entail finding ways to fix disordered proteins into a more therapeutically useful state. For example, cytokines are powerful signaling molecules that regulate a wide range of immunological and other biologi- cal functions, but past attempts to exploit them as drugs have fallen short because of issues including poor stability and unpredict- able side effects. The Peptone team believes this is attributable to the extensive disor- dered regions that it has identified in many cytokines. “We figured out how to engineer some of these cytokines in a part which don’t touch the receptors,” says Tamiola. “Sud- denly the protein is 20% less susceptible to proteolysis.”

These engineered cytokines can then be coupled to antibodies to produce ‘immuno- cytokine’ conjugates that are highly stable and selectively act on specific target cells. Chief strategy officer Benjamin Owens says that Peptone is currently prioritizing a derivative of interleukin-21, a cytokine that can help trig- ger a potent antitumor immune response, with the goal of completing investigational new drug (InD) application–enabling studies in late 2024 or early 2025.

Other, earlier-stage programs are using machine-learning-based models to design polypeptides or monoclonal antibodies that can strongly and efficiently engage with disor- dered regions of target proteins. “We are not yet at the stage where we can say, ‘Throw away all your antibody discovery tools, the comput- ers are here’,” says Tamiola. “But we can very rapidly advance these programs to the stage where the computer-generated models are believable enough to be expressed and bind specifically.” efforts in this arena are targeting proteins involved in inflammatory autoim- mune disorders as well as cancer indications.

Peptone has been working with several companies on individual collaborations that have allowed them to test their strategy on thorny protein targets that have left other bio- techs stumped. But the goal is not to become a contract research organization, according to Owens. “As we were going through the series A pitching process ... we made a very clear state- ment that we will develop our own therapeutic programs,” he says. That being said, the door is open to more extensive drug-development collaborations in the future.

There are still relatively few companies with a core focus on intrinsically disordered proteins, and Kurgan is glad to see a startup embracing the risk and opportunity associated with these tricky targets. “It’s a mixture of sensible experimental modeling with some sensible computational modeling,” he says. “I think they have some chops—they’ve done some modeling that is equivalent to what the academic field is doing.” However, he also expresses frustration about the opacity of Pep- tone’s proprietary computational modeling toolbox, describing it as a ‘black box’ whose capabilities are currently hard to evaluate.

Tamiola says that he aspires to achieve regular exchange of data and tools with his colleagues in the academic world, citing the example of DeepMind CeO Demis Hassabis’ decision to make his company’s AlphaFold software freely available. But even if the com- pany’s technological advances don’t trickle down right away, its success could still benefit the small but growing community of research- ers working on disordered proteins. “When you write grants or papers, it’s always good to say there’s some money in it downstream,” says Kurgan. “If we can target something new, you can get investment, new blood, new life into the whole area.”ME