Gene therapy has been heralded as the new frontier of medicine, but there are still many limitations to current technologies; among them, how to deliver therapeutic genes to specific cells, and only activate them in the right context. A team has created a new RNA-based tool called DART VADAR to bring gene editing out of the 'dark side' of those problems and into the light. Using an engineered form of a naturally occurring enzyme, their sensors can automatically sense the presence of a trigger molecule and initiate the translation of payload genes within cells. The advance broadens the scope of conditions that can be treated with RNA-based therapeutics and enables the development of highly specific treatments for a variety of diseases.
More than twelve billion doses of mRNA vaccines have been administered globally since the start of the COVID pandemic, saving millions of lives. But RNA-based therapies for other diseases have so far proven more challenging to develop. The full-body immune response caused by mRNA vaccines is fantastic for fighting off invading pathogens, but many other conditions only affect a single organ or cell type. Engineering RNA molecules to only activate their therapeutic payloads when they find themselves in the right conditions is the key to the next generation of "smart" RNA-based therapies.
A new system created by researchers at the Wyss Institute at Harvard University and MIT may help unlock that potential, as reported in Nature Communications. The team, working in the lab of Wyss Core Faculty member Jim Collins, Ph.D., developed a novel RNA sense-and-respond circuit they call Detection and Amplification of RNA Triggers via ADAR, or DART VADAR. Taking advantage of an enzyme that edits RNA in the human body, DART VADAR allows researchers to easily design circuits that trigger the translation of a delivered genetic payload in response to the presence of a specific molecular marker of disease and/or cell type. This ability broadens the scope of conditions that can be addressed with RNA-based therapeutics and enables the development of highly specific treatments for a variety of diseases.
"I am particularly excited by the fact that our DART VADAR system is a clinically relevant, compact RNA-based circuit that enables one to direct therapies in a highly programmable manner to specific cell types and cells in certain states, thereby minimizing off-target effects," said Collins, who is also the Termeer Professor of Medical Engineering & Science at MIT.
From trigger to translation to therapy
The Collins Lab has long been interested in finding ways to control the translation of RNAs in cells, and has developed several methods, including eToeholds, that allow them to initiate translation only in the presence of a specific "trigger" molecule. But the process of designing a new molecular structure for every new trigger was cumbersome and complicated. "Our technology grew from the idea that we could decouple the elements of responsive RNA sensors -- sensing, actuation, etc. -- so it's much easier to design circuits for new targets. Ideally, we wanted to be able to change the payload without modifying the sensor element every time," said co-first author Raphaël Gayet, Ph.D., a research scientist at the Wyss Institute.
The substance around which DART VADAR is built is called adenosine deaminases acting on RNA, or ADARs. ADARs are enzymes that bind to double-stranded RNA (dsRNA) molecules and make a specific base edit, converting a mismatched adenosine (A) molecule into inosine (I). This change destabilizes the dsRNA structure, and is thought to be involved in cells' responses to different viruses, many of which carry their genetic material as RNA.
The researchers reasoned that they could use the natural dsRNA-editing ability of ADARs to create a new kind of responsive RNA sensor. So they designed a single-stranded RNA circuit containing multiple modular elements:
The UAG codon sequence prevents any genes encoded after it, like the green fluorescent protein, from being translated, as UAG is the natural signal to a ribosome that it should stop the translation process and fall off an RNA strand. However, if the circuit binds to a complementary target RNA strand in the cell, it becomes a double-stranded RNA molecule. The team engineered their circuit so that the A of the UAG sequence would "mismatch" with a cytosine (C) in the target strand, rather than its rightful partner of U. This mismatch essentially frees up the A to be found and converted to I by ADAR, and the resulting UIG sequence is no longer a "stop" codon, allowing translation to occur. Thus, the green fluorescent protein is produced, signaling that the sensor circuit has found and bound to its target.
ADARs are naturally found at high concentrations in neurons, but in low amounts in other cells, and the team found that while their ADAR-based sensor worked, its activity was low. To ensure that the sensor could work in different cell types, they added the sequence of the ADAR gene to their sensors. Now, activation of the sensor by ADAR naturally present in a cell could stimulate the production of more ADAR, creating a positive feedback loop that amplifies the sensor's activity. They verified that this occurred in experiments in vitro, observing that cells that received the ADAR-containing DART VADAR sensor displayed a marked increase in fluorescent green protein levels.
"What's really exciting about this sensor is that the green protein signal sequence can be easily replaced with the sequence for any therapeutic gene that you want to express in response to the presence of a trigger RNA in the cell. So not only does this sensor detect targets, it can automatically respond to them without requiring user input, automating the delivery of a therapeutic payload at the cellular level," said co-first author Shiva Razavi, Ph.D., a Postdoctoral Fellow at MIT.
A highly sensitive sensor
While the team was thrilled that their first version of DART VADAR worked, they realized that they would have to make some modifications if it were to be useful as a therapeutic approach.
"While expressing naturally occurring ADAR via our sensor increased the A-to-I editing rate from about 3% to about 30%, that still falls short of being useful in a real-world context. Endogenous ADAR is insufficient to generate consistent results and too big to fit into clinically approved delivery vehicles, like AAV [adeno-associated virus]. In terms of translation potential, it's important to make sure whatever you're doing in the lab can realistically make it into a product one day," said co-first-author Katherine Ilia, Ph.D., a Postdoctoral Researcher at MIT.
To address those limits, the team leveraged an engineered ADAR variant that was truncated to only contain its active RNA-editing region, and added two short regions that can bind to a "hairpin" RNA structure known as MS2, which they inserted into their sensor sequence flanking the UAG codon. This chimeric protein, which they named MCP-ADAR, is much smaller than naturally occurring ADAR, and has a reduced likelihood of binding to off-target molecules. They also modified the ADAR gene in their RNA sensor so that it would be translated into their modified MCP-ADAR version. Now, when a sensor was activated by naturally occurring ADAR in a cell, the MCP-ADAR produced could bind to other sensors via the MS2 hairpins, driving the production of more MCP-ADAR and amplifying the sensors' activity.
To test the performance of their optimized DART VADAR system with its engineered enzyme, the team decided to see if it could detect a single-base mutation in the human p53 tumor
suppressor gene that is known to drive several kinds of cancer. They designed a DART VADAR sensor to detect the p53 mutant and introduced it into a line of human cells along with plasmids expressing either the normal or the mutant version of the gene. They found that cells that contained the mutant version of the gene also showed a five-fold activation of the reporter gene in the sensor, showing that the sensor did indeed register the presence of the aberrant genetic sequence with high specificity.
They next tested whether DART VADAR could distinguish between healthy cells that are in different stages of development, based on the presence of specific molecular markers of cell state. They used mouse cells called myoblasts, which are progenitor cells that can differentiate into multiple cell types including myotubes (which become muscle) and osteoblasts (which produce bone). They designed DART VADAR sensors to detect RNA markers of both cell fates, prodded the myoblasts to differentiate, and then added their sensors to the cells. They found that both sensors strongly produced their respective payload signal molecules in their corresponding cells, demonstrating that the sensors can detect molecular differences in both cell states and cell types.
The team points out that in order to apply DART VADAR to various clinical applications, additional aspects of the target site, such as the likelihood of triggering an immune response, will need to be considered during the sensor development process. They anticipate that recent computational advances in protein structure prediction could be used to further refine the selection of optimal target sites to guide the design of safe and effective DART VADAR sensors. They have applied for a grant to use DART VADAR to explore the stepwise differentiation of stem cells into other cell types, which in the future could be used to replace diseased cells in a patient with healthy cells.
"This team's ability to combine preexisting biological components into a completely new engineered technology that has the potential to make the treatment of a wide range of diseases faster and easier is a great example of how synthetic biology can change the world for the better," said Wyss Founding Director Don Ingber, M.D., Ph.D. Ingber is also the is also the Judah Folkman Professor of Vascular Biology at Harvard Medical School and Boston Children's Hospital, and Hansjörg Wyss Professor of Bioinspired Engineering at the Harvard John A. Paulson School of Engineering and Applied Sciences.
Nathaniel Tippens is another co-first-author of the paper. Additional authors include Makoto Lalwani, Kehan Zhang, and Jack Chen from the Wyss Institute and MIT, Jonathan Chen from MIT and the Broad Institute, and Jose Vargas-Asencio from the MIT Picower Institute. This work was supported by NIH grants R01EB024591 and 5RC2DK120535-03, and the Wyss Institute for Biologically Inspired Engineering.
