Leveraging Directed Evolution To Optimize AAV Vector Design

By Tyler Menichiello, contributing editor

Accurate and effective delivery is one of the biggest challenges in the gene therapy space. After all, a gene therapy is pretty useless (if not dangerous) if it can’t effectively target tissues of interest. AAVs are the most popular delivery vehicle for gene therapies, and as such, have been an active area of innovation in recent years. A lot of work is being done to improve their specificity and transduction efficiency to mitigate off-target effects. The more effective an AAV is at targeting and transducing specific tissues, the more potent the therapy is, meaning less drug product is needed per dose, which ultimately lowers production costs.
Of course, this isn’t as simple as deciding what style rims should go on a concept car. Rather, it’s about understanding the complex biochemistry underlying AAV capsids and how they interact with human cells and leveraging this understanding for therapeutic benefit. One company’s approach to solving this engineering problem relies on nature’s key design mechanism — natural selection.
Building A Vector Library And Letting Nature Do The Rest
“We don’t have to be smarter than nature,” Kirn says. Mother Nature designed AAVs, so 4DMT’s philosophy is to let her improve this design by accelerating natural selection. The company’s directed evolution approach to vector design begins with a target vector profile — identifying the desired characteristics for the optimized vector. For its R100 vector (used in its ophthalmology programs), this meant a vector that could be delivered via intravitreal injection.
“We wanted to get through all the barriers that had blocked the standard AAV,” Kirn says, “and we wanted to have high-level transduction of the retina so we could treat a large number of retinal diseases.”
The next step involves building a diverse vector library (similar to a small molecule library) that can later be screened and distilled into the best candidates. “We start with three to four different AAV starting points that are present in nature, and then we use a variety of molecular biology techniques to invent a new library of a billion different versions of those,” Kirn explains.
Where directed evolution comes into play is in the screening of vectors in vivo (using monkey models) to find the “needle in the haystack” — i.e., the best-performing capsid. Kirn says it was important to deliver these AAVs to the animal by the same route intended to be used in humans (intravitreally, in the case of R100). After some weeks, the retinal tissues were examined to see which vectors made it to their intended target.
“We have a very sophisticated way of figuring that out by barcoding,” Kirn tells me. “Each AAV has the DNA sequence for its own capsid in it, which is a very difficult thing to do, but we can track the DNA.” Through deep sequencing, the development team was able to identify which AAVs made it to the retina and at what ratio. Those vectors with the highest ratios were then isolated, manufactured, and re-injected into another animal. “Each time you do that, the best vector gets a higher and higher percentage,” he explains. “So, when we did this in real life, one vector went from one in a billion to 70% of all genomes after several rounds.”
At the end of this iterative in vivo screening process, R100 outcompeted all other vectors and rose to the top. The company used this same screening technique (with different tissues and routes of administration) to optimize vectors for its pulmonary (A101 vector) and cardiology (C102 vector) programs.
“We have to be very sophisticated about how we build the libraries and how we do the selections to get rid of any bias, but after that, it’s like nature figures it out for us,” says Kirn.
Controlling For Bias
Anything that involves human decision making has biases baked in, and screening for top-performing vectors is no exception. “At every step, you can introduce bias that would push you further away from the vector you really want to find,” Kirn says. For example, some vectors may be more manufacturable than others, and manufacturability may be something you want to select for. Certain vector DNA sequences may amplify at different efficiencies than others, which can potentially influence results. “These are examples of biases that you need to be aware of and either correct for or include if you want that feature.”
Clinical Success Greases The Tracks
As the conversation turned to 4DMT’s ongoing clinical success (with one product nearing Phase 3 readiness and another expecting interim Phase 2 data in mid-2025), Kirn emphasized the importance of having a clear vision for the company and being methodical in its execution. “Our purpose statement was to boldly innovate to unlock the full potential of gene therapy,” he says. “Right up front, we decided that we are willing to take longer to get it right, even though it’s a much higher hurdle.’”
This slow-and-steady mindset was somewhat unconventional when the company was founded in 2013. “It was like a gold rush,” Kirn says with nostalgia. “Everyone was like, ‘Oh my god, AAV works! Let’s just grab whatever AAV’s in the lab, throw it in humans against a disease that we think is a rare disease and can get a big effect, and we’ll all go home!’”
Of course, the industry soon found that it was not so simple. Fast forward to today, and 4DMT’s slow-and-steady approach is paying off. As the company brings more products to the clinic with the same vector, the regulatory requirements become less stringent, “to the point where we’ve even gone into the clinic in some cases without the need for a GLP toxicology study,” Kirn says. The company has produced so much clinical data for its vectors that each new product carries less risk and is able to move through the clinic much faster.
Building Stage-Appropriate GMP Manufacturing
The almighty question most biotech companies face: To manufacture in-house or to partner with a CDMO? “There’s no right answer,” says Kirn, “but if you have the money, doing a stage-appropriate buildup of your GMP manufacturing is the optimal way to go.”
Of course, internal manufacturing carries a huge risk. Any clinical failure or setback can derail a company’s timeline and trajectory. If you build a commercial manufacturing facility too early and then encounter challenges that slow development, the product may die on the vine and the facility will sit empty. This decision around manufacturing is a constant back and forth of weighing the risk and the cost — what’s good enough to get into the clinic today versus what is optimal for Phase 3 and commercial?
Ultimately, Kirn argues that being able to raise money in a cost-effective and sustainable way enables companies to build their own GMP capabilities and balance these risks. “It allows you to control your destiny much more and take less risk down the road — less timeline risk and less quality risk — because you’re always going to care about it much more than a CDMO,” he says.
For 4DMT, the company built its own GMP in a stepwise fashion. “We started with a very small unit that would supply the eye, which is a low dose,” Kirn says. “Then, we slowly built up to a scale where we could service the lung, which is an order of magnitude higher in dose. Once we had money and proof of concept, we were able to do the IV dosing, which is a much higher dose.”
For its Phase 3 and commercial manufacturing, however, 4DMT will use a CDMO because of the huge risk in switching manufacturing sites going from Phase 3 to commercial, Kirn says. “Down the road, once we have a blockbuster on our hands or have two products approved, we can make our own commercial facility,” he says, “but not until then.”
After nearly a decade of development, 4DMT’s lead candidate, 4D-150, will enter Phase 3 in the beginning of 2025 for wet AMD. Kirn believes its product safety and efficacy profile will make it a game changer for wet AMD and DME. Its success will not only validate the company’s methodical approach to vector design, but also shine as an example of what can be achieved through directed evolution. After all, nature provided humans with AAVs, so it only makes sense to leverage nature to refine these powerful tools.