RNA is a central and universal mediator of genetic information underlying the diversity of cell types and states. The lack of tools for studying and controlling various cell types using RNAs continues to be a problem in biology and medicine. Molecular RNA sensors would enable it to investigate and treat particular cell types and states in various settings, especially in human patients and non-model organisms.
New research published by teams from Duke University Medical Center and Stanford University describes modular, programmable live RNA-sensing technologies with broad applications in biology, biotechnology, and RNA medicine. This new class of RNA-sensing-dependent protein translation technology bypasses the DNA-based transcription process. Instead, these technologies use enzymes called ADARs (adenosine deaminases acting on RNA) for the potential of RNA-based monitoring and editing of cells across organ systems and species.
The two systems, called CellREADR (Cell access through RNA sensing by Endogenous ADAR) and RADAR (RNA sensing using adenosine deaminases acting on RNA), are described in two articles: “Programmable RNA sensing for cell monitoring and manipulation” in Nature and “Modular, programmable RNA sensing using ADAR editing in living cells” in the October issue of Nature Biotechnology.
“At a fundamental level, this technology will have a very wide impact on understanding biology in general,” Z. Josh Huang, Duke School of Medicine Distinguished Professor in Neuroscience and senior author of the CellREADR article, told GEN Edge. “This is also a really useful tool to modify disease.”
A tale of two sensors
While these two stories converged on ADAR as the centerpiece for these innovations, the two groups came to the problem of RNA sensing from entirely different angles. As an undergraduate student in 2012, K. Eerik Kaseniit, lead author of the RADAR paper, has been thinking about RNA sensors for the past decade.
“I tried all the tricks you can think of, like forming weird secondary structures and disrupting them with RNA base pairing, but nothing worked out,” Kaseniit told GEN Edge.
When he ended up in grad school in the lab of Xiaojing J. Gao, who has a synthetic biology background coming from Michael B. Elowitz’s lab, Kaseniit began to think about RNA more deeply and how to make gene circuits without DNA. Kaseniit said that Gao has been thinking about the RNA problem for a while and that people have been trying to use RNA interactions to make strand displacement logic gates or anything that involves double-stranded RNA (dsRNA).
But Gao found a bug in the system: there are proteins that interact with dsRNA. “[Xiaojing] went through the literature for all these RNA-interacted proteins and saw that there was an enzyme called ADAR that binds double-stranded RNA, and that’s where it clicked,” said Kaseniit. Gao turned this so-called bug into a feature that could be used as a tool to build exciting synthetic biology circuits.
Josh Huang, the senior author from the Duke group, came at this from a different angle. Huang is a neuroscientist trying to understand brain circuits and neural cell type diversity. Along with many researchers, Huang has used germline genetic engineering—traditionally by leveraging DNA regulatory elements in the nucleus—to create genetic tools that monitor, manipulate, and edit cell types by sensing RNA in living cells.
But this gene targeting approach—the traditional method for the past two decades—is extremely labor intensive, lengthy, and not easily scalable. This gene targeting approach is cumbersome, slow, difficult to scale, generalizes, and raises ethical issues, especially in primates and human cells. All DNA-based approaches for mimicking and leveraging cell-specific RNA expression patterns are inherently indirect.
“We have been trying to see whether there is a fundamentally different strategy that will be in ways that are specific, versatile, simple, and generalizable across organ systems and species,” said Huang. “That’s how we got into using RNA sensors and then use that sensor for payload expression.”
These stories came together through a common link between the two groups that goes back more than 20 years to when Huang was at Brandeis University for graduate school. Gao’s PhD advisor was Liqun Luo, who also overlapped with and became friends with Josh Huang at Brandeis. Luo knew both groups were looking at RNA sensors and encouraged them to connect.
The RNA sensor blueprint
These programmable, modular novel RNA sensors leverage dsRNA editing by ADARs. These enzymes bind dsRNA in a largely sequence-agnostic manner and edit adenosine nucleotides to guanine-like inosine nucleotides. This conversion can be leveraged to edit away UAG stop codons gating translation of a protein payload to form RNA-responsive, protein-outputting sensors.
Although these ADAR-utilized RNAs between the two groups differ slightly, they both have two key regions. First, a ‘sensor’ domain is several hundred nucleotides complementary to a specific cellular RNA to form sequence-specific base-pairing. This sensor domain contains one or more ADAR-editable stop codons that act as a translation switch. Second, an ‘output’ domain encodes an effector protein downstream of the sensor domain. In-frame with the stop codon is a sequence encoding the self-cleaving peptide called T2A, followed in-frame by an effector RNA region that can encode all sorts of effector proteins.
So, in cells expressing the target RNA, these ADAR-utilized RNAs form dsRNA with the target RNA, which recruits ADARs to assemble a complex at the editable stop codon. Once the stop codon is converted by the ADAR complex, translation of the effector RNA is turned on. The resulting fusion protein comprises an N-terminal peptide, T2A, and a C-terminal effector, which then self-cleaves by T2A and releases the functional effector protein.
Huang is keen on using RNA sensors to understand cell type diversity in the brain, something he’s been working on for decades using genetic engineering techniques. As RNA sensors can be delivered on viral or other genetic vectors and achieve cell-type-specific expression, it removes the need for promoter identification, which has remained a major hurdle in onboarding new organisms for bioengineering or genetics-driven research.
Huang notes that single-cell sequencing technology has allowed researchers to identify all the cell types. But the technology to study each cell type in an easy, scalable way to understand the fundamental flow of information didn’t exist.
Both Huang and Kaseniit think these RNA sensors will do for cell types what CRISPR is doing for genetics. In the CellREADR paper, Huang’s group used viral delivery of the technology to confer specific cell-type access in mouse and rat brains as well as in ex vivo human brain tissues. Also, CellREADR enabled the recording and control of particular neuron types in behaving mice.
Kaseniit agrees, telling GEN Edge, “If you want to read or write genes, use CRISPR. If you want to read or write cells, you use CellREADR or RADAR technology—you use RNA sensors. We’re going from the gene to the whole cell level to understand cell types.”
Kaseniit is excited about RNA sensors because they can be the foundation of a toolkit to understand cell types using logic gates. He gives the example of using an ‘AND’ gate— when two events are necessary to trigger a response. For example, this can be used to have cells that express two markers spit out a particular protein output, whether a cytokine or any other cell signaling molecule. Kaseniit hopes to expand the logic gate grammar to include more operators, such as a ‘NOT’ gate.
By integrating multiple inputs, RNA sensing can enable high specificity and low off-target effects of the downstream interventions. It is especially suitable for increasing the specificity of RNA-based vaccination and gene therapies, the power of which was recently demonstrated during the pandemic. As RNA sensing can be delivered on viral or other genetic vectors and achieve cell-type-specific expression, it removes the need for promoter identification, which has remained a significant hurdle in onboarding new organisms for bioengineering or genetics-driven research.
Following in the footsteps of CRISPR
As with any new molecular tool, this new technology goes through multiple versions of improvement and optimization. Both Huang and Kaseniit think that the trajectory for RNA sensors will look a lot like that of CRISPR gene editing.
“If we look at the CRISPR story, people got better at predicting which [guide RNAs] are better,” said Kaseniit. “In our case, we have sensor RNAs, so we’re thinking about finding the best sequences to target, getting better at the bioinformatics there.” Just as it has for most other molecular tools, it boils down to specificity, efficiency, efficacy, and scalability.
And just like CRISPR, not only do Huang and Kaseniit see the potential for these technologies to break open the basic biology of cell types—they see an incredibly powerful therapeutic tool. Kaseniit is pondering the value of RNA sensors as therapeutics that turn on only in a particular cell subtype. “For example, you can make RNA sensors for a point mutation in cancer so that you can deliver an RNA shot that only turns on in cancer and destroys it,” said Kaseniit.
Huang agrees. “In the context of disease,” he says, “you can very specifically alter the physiology of the cell, delete harmful cells, whether cancerous or infected cells, or reprogram cells in certain developmental disorders.”
RNA sensing smart therapies
As all the RNA sensor components can be delivered via mRNA, the wide range of existing RNA synthetic biology and nanotechnology tools could be combined with RNA sensors. It is especially suitable for increasing the specificity of RNA-based vaccination and gene therapies, the power of which was recently demonstrated during the pandemic.
And there are many situations in which these tools can be combined to track cells conditionally. For example, RNA sensors can track cells as they become infected with a virus, transition from normal to pre-cancerous to cancerous to metastatic or become senescent.
On top of being tracked, RNA sensors combined with other tools can lead to smart therapies that can dynamically detect these state changes and stop or even reverse the processes driving them. Or, by integrating multiple inputs, RNA sensors can enable high specificity and low off-target effects of the downstream interventions. In addition to responding to transcript levels, RNA sensors can distinguish disease-relevant sequence alterations of transcript identities, such as point mutations and fusions.
While these back-to-back articles are proof-of-concept, it’s not a reach to say that RNA sensors will have the potential to facilitate research towards elucidating the principles of biological information flow from genotype to phenotype across cell types. Moreover, RNA sensors make possible a new generation of programmable cell-specific RNA medicine.
