Kärt Tomberg, PhD, has a vision to express every recombinant protein at the levels required to implement its intended therapeutic effect.
“There are proteins in the biotechnology field that need to be expressed in much greater abundance than they are now,” Tomberg told Inside Precision Medicine. “It might not sound revolutionary, but the outcome could be a drug—it’s enabling an outcome from nothing to something.”
These challenges to protein expression result in high development and manufacturing costs. For instance, many drugs today are still purified from human plasma because they cannot be produced recombinantly because of their size and complexity, like Factor VIII for hemophilia A—a rare inherited condition that affects 1 in 4,000 to 5,000 males worldwide at birth.
“If you go and look up the Factor VIII drug market, half of it is still purified from plasma and half of it is made recombinant, “said Tomberg. “But it is so poorly expressed in protein that its price is about a thousand times higher than an average antibody. This is how poorly this molecule is being expressed.”
To solve this problem of poor protein production, Tomberg has leveraged the utility of exon-junction complexes—protein complexes that forms on a pre-messenger RNA strand at the junction of two exons that have been joined together. Behind this concept, she launched the startup ExpressionEdits, which just announced having raised $13 million in seed funding.
Tomberg co-founded ExpressionEdits with Liliana Antunes, PhD, and Allan Bradley, PhD, best known for serving as the Wellcome Sanger Institute’s director from 2000 to 2010. The funding round was led by Octopus Ventures and redalpine, with participation from BlueYard Capital, Wilbe Capital, Acequia Capital, Amino Collective, and Hawktail.
Tomberg said, “In some ways, you could say that we are solving the production problem. It’s why certain drugs don’t exist and patients don’t get them. We call it inventing the imaginable. We’re not inventing the unimaginable. These are the proteins that the patients need. But actually getting there is quite hard.”
Genetic Syntax Engine
In March 2020, with the COVID-19 pandemic emerging, Tomberg put her academic research on hold to join the all-hands-on-deck effort to understand and combat SARS-CoV-2.
Tomberg was part of a team tasked with producing one of the proteins found on the surface of SARS-CoV-2, a spike protein (S-protein), but they had difficulty expressing the protein. So, she began to think of how to redesign the construct to maximize the production of the S-protein. Tomberg began to dig into the problem when she had an insight into how the virus works to hijack human cellular machinery.
“I started to dig into the problem, and I realized that as an RNA virus, [SARS-CoV-2] is never seen in the nucleus and lives in the cytoplasm—it doesn’t even look like a gene,” Tomberg told Inside Precision Medicine. “We were failing, whereas [the virus] was making millions of proteins in our noses with the mammalian system. So, the question became, If it doesn’t look like a gene, what would it take to make it look like a gene?”
With this in mind, Tomberg devised a solution that resulted in a 10-fold increase in S-protein expression. Tomberg’s methodology leverages the incorporation of specific introns to the gene design to make exon-junction complexes, which play essential roles in nonsense-mediated mRNA decay and mRNA splicing, transport, and translation—all of which can be part of the protein production process.
Tomberg then began to consider the universality of this solution—could designing genes with specific intronic sequences work with other proteins? So, she started developing a platform that could automate the entire process of taking a sequence and optimizing it for protein production through “intronization.”
ExpressionEdits has created a platform called the “Genetic Syntax Engine.” This platform uses AI to look at huge amounts of data and make predictions about adding introns at exact locations and with certain sequences to gene design to speed up the formation of exon-junction complexes and improve protein production. According to Tomberg, the secret sauce to the Genetic Syntax Engine is not some revolutionary AI/ML algorithm but instead lies in the data ExpressionEdits has produced, assessing different introns in different gene settings.
“We’ve created lots of libraries assessing different introns in different coding contexts,” said Tomberg. “You build a lot of different libraries with many different combinations, and the machine learning algorithms look at the features that predict good and bad performance—how well they actually do in expression and splicing. Then, the machine basically figures out what’s common between success and failure, and that can then be translated into predictive models.
No protein left behind
The Genetic Syntax Engine’s great potential is not lost on Tomberg, who said that it makes sense to find partners for certain applications. In particular, Tomberg is interested in having ExpressionEdits partner with companies pursuing gene therapy because protein expression is a major hurdle for getting the therapy safely to patients.
In these situations, ExpressionEdits will design and produce the sequence before handing it off to their partner, who will then test it out in their systems. Tomberg is limiting ExpressionEdits’ work on gene therapy to partnerships because she thinks the technology is still riddled with too many problems for the startup to undertake in their pipeline. Tomberg said that ExpressionEdits points the Genetic Syntax Engine at recombinant proteins that have eluded expression and thus prohibited their therapeutic potential.
“We decided, because we are a small team and are geared to focusing on one really hard problem, that we will focus on one pipeline of a few candidates to really solve them and take them as far as possible in the process of clinical development in-house,” said Tomberg.
When asked about pipeline details, Tomberg said she could not comment because there wasn’t really any data to show. She said ExpressionEdits is in very early stages and is combing through data to find which proteins are a good fit for its Genetic Syntax Engine.
“First and foremost, we will be looking at which proteins we are improving the expression of in really big ways,” said Tomberg. We know that proteins can fail for other reasons, like toxicity and misfolding. Just having more expression will not solve every single problem in protein production. So, we want to identify proteins where we are providing a solution.”
If ExpressionEdits’ Genetic Syntax Engine succeeds, a significant barrier to making many protein therapeutics viable and accessible will have been toppled.
