A conversation with Richard Novak, Ph.D. about the CogniXense platform and its ability to model and screen drugs for rare genetic diseases like Rett syndrome
By Jessica Leff
Five years ago, Richard Novak, Ph.D., and other Wyss members started working on the DARPA THoR project, led by Wyss Founding Director Don Ingber, M.D., Ph.D., and Mike Super, Ph.D., Lead Senior Staff Scientist, with the goal of identifying drugs that could induce human tolerance to sepsis and other forms of infection. But rather than fretting about what would happen if they failed, the team instead worried about what effect their success might have on patients’ cognitive abilities, as many therapies and drugs cause negative behavioral or cognitive effects that are not discovered until after they have been tested on people. The team decided to develop cognitive assays and behavioral screens to prevent this problem, and thus the idea for CogniXense was born. Since then, this biological modeling and drug screening platform has developed greatly and it is now being applied to rare genetic diseases, most notably Rett syndrome, a severe neurological disorder that affects mostly girls.
We sat down with Novak, who is a Senior Staff Engineer and member of the Advanced Technology Team, to discuss the current state of the CogniXense platform, and how he envisions it functioning in the future.
What exactly is CogniXense?
CogniXense is a biological modeling and drug screening platform consisting of three main parts: a drug prediction algorithm called NeMoCAD, disease models, and a screening instrument called a TadPool. The platform rapidly screens drug candidates for genetic diseases using Xenopus tadpoles.
What real-world problem does CogniXense solve?
Finding the biological pathways related to cognition, memory, and behavior is a complicated process. This means that finding drugs to treat cognitive and behavioral disorders is a big challenge, and drugs neurotypical people take can have adverse effects in these areas that aren’t discovered until they reach late-stage trials. Ideally, you’d want to screen for this earlier on in the process.
CogniXense enables us to screen compounds before they reach human trials or more advanced pre-clinical models and helps us develop treatments for cognitive and behavioral disorders. One of the goals is to find a nootropic, or a drug that will improve cognition. With CogniXense, we can work with large numbers of tadpoles, which have advanced cognitive capabilities and behavioral repertoires, and start analyzing what a drug’s function is and how it affects the brain.
How has CogniXense developed to effectively address this problem?
Each component has advanced since we had the initial idea for CogniXense. NeMoCAD, the drug prediction algorithm, has evolved over time and is even being used on other projects like the COVID-19 drug repurposing effort and Biostasis. We use that algorithm to predict drugs that might reverse the cognitive deficits, which narrows down which compounds we need to test in the screening system.
Because of the convenience of CRISPR, the number of chromosome copies in the tadpoles, and the mosaicism of engineered cells in developing Xenopus embryos, within a couple weeks we’re able to get a disease model indicative of a whole patient population, instead of just a single patient. Many diseases do not present exactly the same in every person, and we are able to see a whole range of behavior and phenotypes in our model that go from asymptomatic to severely affected. That’s really exciting because it means the model is far more representative of a genetically diverse population, which is what doctors really see.
The screening instrument, or the TadPool, has become more complex. We’re not just using a camera to observe a dish of tadpoles. We have automated mazes where we can track the animals’ decision-making skills and ability to learn. In addition, we’re tracking swimming patterns and circadian rhythms to get a more complete picture of their behavior.
How does it work?
We use NeMoCAD to evaluate known, FDA-approved drugs or other compounds to find out which has the best likelihood of being effective. Repurposed compounds can be moved to the clinic more quickly than novel compounds, so it’s ideal to use them. We microinject the CRISPR-Cas9 enzyme into the Xenopus embryos at the four- or eight-cell stage of development, which damages a target gene as the tadpole continues to grow. Since the tadpoles have four copies of each chromosome, and CRISPR is not 100% effective, each cell can have 0-4 copies of the mutated gene. This results in a spectrum of the gene mutation expression, which is actually advantageous to us. As cells continue to divide, different parts of the body may be more or less affected. This gives rise to the variability in the population.
Once we have the disease model, we use the TadPool screening instrument and video recoding of swimming to do the testing. The most effective compounds move on to a more widely-accepted animal model, such as a mouse. Additionally, once we find a compound that works, we can look into how it is achieving the desired behavioral effect, which is really cool. Then, we know which protein we should go after and researchers can work to develop drugs that are more effective at targeting that protein.
CogniXense also provides free toxicity profiles, because tadpoles have historically been used for this purpose anyway and are actually quite sensitive to toxic compounds. The computation piece also speeds things up by decreasing the number of compounds you need to test in early screens.
You’re currently using CogniXense to study Rett syndrome. What is that?
Rett syndrome is a severe neurological disorder that is linked to the X chromosome and mutations in the MEPC2 gene. It mostly shows up in young girls and its hallmark feature is retrogression. Those affected have some level of cognitive deficits, and often have seizures that are difficult to control, GI tract issues including difficulty swallowing and lack of digestion, and sleep disruption. It’s a relatively rare disorder affecting approximately 1 in 10,000 girls.
Why did you decide to apply CogniXense to Rett syndrome?
A visitor came to the Wyss Institute that had a daughter with CDKL5 Deficiency Disorder (CDD). CDD is a genetic disorder that causes seizures and developmental delays. Don Ingber mentioned the CogniXense platform, thinking it could be relevant for his foundation, which focused on CDD therapies. At the time, CDD was classified as an atypical form of Rett syndrome, though they have since been divided into two indications. The visitor was so impressed with the idea that the two of them approached me about using CogniXense for rare disorders like Rett syndrome.
I did some research and found out Rett syndrome was a single gene mutation, which makes it easier to create a disease model. I brought in Frederic Vigneault from George Church’s group, who has a background in CRISPR gene editing and synthetic biology, and we created the Rett syndrome disease model. In our first test, we saw dramatic differences in behavior between the typically developing tadpoles and those with the altered gene, confirming the efficacy of our model. That was actually the beginning of using CogniXense for rare disorders and neurogenetic disorders, and for screening hypothesis-driven ideas.
So, it was a combination of that visitor asking us to look at genetic disorders and their personal connection to CDD and Rett syndrome along with the fact that Rett syndrome was a good first target because it is a single gene mutation with some information already available but still in need of effective drugs.
Why is CogniXense particularly effective for Rett syndrome?
The way Rett syndrome presents can range. Some patients are walking and talking while others are wheelchair bound, nonverbal, and unable to feed themselves. The fact that the CogniXense platform shows a disease model with a huge amount of genetic and phenotypic diversity is immensely helpful with a disorder like Rett syndrome, because it directly mimics the patient population.
Also, MEPC2 is a critical gene; if you disrupt it, it’s going to have wide-ranging effects. Though Rett syndrome is a single gene mutation, it affects multiple organ systems, like the central nervous system, immune system, and the GI tract. This makes it really important to have a whole animal disease model, as we do in the CogniXense system.
In addition to Rett Syndrome, what are some other diseases the CogniXense approach is being used to address or could be used to address in the future?
Aside from Rett syndrome, our work with CDD is most developed. We’re also collaborating with Jenny Tam, Don Ingber, George Church, Mike Levin, and Bruce Yankner on a project to address bipolar disorder. Bipolar disorder is not totally understood, and it takes 6-8 months and a lot of money to create a mouse model, so we’re hoping to do something better. We’re also working with a foundation based at Massachusetts General Hospital focused on X-linked dystonia-parkinsonism (XDP) to create a tadpole model. This is an ultra-rare disease, affecting only about 500 people worldwide. Researchers haven’t been able to keep a whole animal model with XDP alive, and are currently relying on a Drosophila (fruit fly) eye as their disease model. So far, our tadpole model has survived and we’re waiting for phenotypes.
CogniXense can effectively look at more rare genetic diseases in the future. We’ve started looking at Fragile X syndrome, comparing how different it is computationally from Rett syndrome. We’re also looking at some rare genetic kidney diseases, which are traditionally difficult to study. Tadpole kidneys are well developed, so we’re hopeful for success there. In the future we want to primarily look at disorders that cause a cognitive or behavioral deficit, though that’s not essential. Since we can track circadian rhythms, we could also look at sleep disorders. With support from foundations and pharma companies, we can make cost-efficient advances in these areas.
How can CogniXense help both foundations and pharma companies to be more successful?
Our platform can provide a rapid way of generating more data, which in turn can generate more interest in a disease and show that there’s a body of work that would make it easier for a pharma company to jump in. Simultaneously, we can talk to patients and families and develop screening biomarkers and metrics for success to ensure that the drugs are actually addressing the patients’ needs early in the development process. We aim to be at the intersection of pharma, foundations and advocacy groups, and patients and families, working to ensure that the therapies that are developed actually fit the problem.
For example, in the case of Rett syndrome, there are many efforts focused on how to reduce seizure frequency. But, when you talk to patients’ families, they say that’s not what’s affecting their child the most. What’s really hard are the GI issues and sleep disruptions, and nobody’s working on a GI improvement or sleep enhancement drug specifically for Rett syndrome. So, we’re looking to find appropriate metrics that are relevant to the patients and bring them into the therapeutics development process earlier so that what’s developed is actually relevant to what they need. With many of these complex symptoms, you need a good model with effective metrics to screen for the behavior, not just a biological indicator like whether or not a protein is inhibited. If you can screen for the behavior earlier in the process, when you get further along and start taking bigger risks, then you’re doing it to actually meet the goals of the patients. Ultimately, that’s what we and all these groups collaborating with us care about: the patients.
This interview has been edited for length and clarity.