2023-05-24

DNA synthesis breakout session (again)

phil paik (molecular assemblies) steve r. bischoff (novohelix) telesis bio

Phil Patt: I am the CTO of Molecular Assemblies. With me is the chief science officer (Steve BIschoff of Novo... we want to talk about the future of DNA synthesis. MA is headquartered in California in San Diego. We have over 55 employees including chemistry, biochemistry, hardware engineering, software engineering, and unprecedented capabilities to write long pure DNA oligonucleotides independent of sequence complexity. We are now shipping to customers, and we have 28 issued patents in life sciences and DNA data storage. We have over $55 million faced in VC money and we have an ongoing partnership with GE Global Research through a $42m DARPA grant through which we get $9.4m.

Phil: The mission of MA has to been to deliver the next generation of synthetic DNA products. The current space is limited. The current purity you get today of long oligos has constrained abilities. The demand for DNA in expanding research markets like CRISPR and gene editing has really outpaced what phosphoramidite has provided today, and this often translates into delays while you're waiting for synthesis or processing. Sometimes you need to go back to redesign your experiments or get different purification or yield that you want.

Phil: We have a cycle efficiency of 99.9% or greater, giving us ultrapure long oligos. The DNA is experiment ready. We do de novo direct synthesis resulting in natural DNA, regardless of length and complexity. The aqueous and enzymatic nature that we do it in .... we are able to do quick delivery.

Phil: HOw do we do it? We have "fully enzymatic synthesis" or FES for short. There are two core tech. We have a proprietary TdT enzyme developed in partnership with Codexis which gives us ultra-high efficiency. We designed and engineered it in a way agnostic to sequence complexity. On the engineering side, we have a high throughput platform architected to be modular for multiple product lines. We focus on balancing thorughput, capital equipment and consumable cost and order-to-ship time. We have flexible output scale to address different applications. Maybe for high mass output, CRISPR, gene editing, or higher throughput but lower output mass for gene fragment or assembly.

Phil: Last year at Synbiobeta, Matt Miller at Codexis gave a presentation talking about our partnership and the process they use to engineer TdT. I'll give some highlights. It was a 13 month project with 40 rounds of evolution and targeted every amino acid position. 3,600 variants were characterized. The final variant had 25% coding sequence changed. The requirements were that we wanted fast kinetics at elevated temperature. We wanted performance to be unaffected by homopolymer stretches or high GC content. We wanted high cycle efficiency retained independent of length. We currently perform direct synthesis of 300mers.

Phil: The incorporation efficiency of a modified nucleotide went from <0.5% to >99%, and the reaction time went from 16 hours to less than 90 seconds. The maximum reaction temperature we wanted to be as high as possible so it went from 44c to 70c. When you think about this enzyme, it alleviates burden for the end-to-end process that allows us to produce natural DNA with no scaring.

Phil: Zooming out, let's think about FES from a process perspective. It's a two step process. We have a starting piece of DNA bound to a solid support we call an "initiator". The first step is the addition or incorporation of a modified nucleotide with a lbocking group and the blocking group allows us to not have a runaway chain reaction, in our case a 3 phosphate group using a proprietary TdT enzyme. We use a phosphotase to remove this. It's a fast process. Once it's done, we release the DNA off the solid support and it's pretty much ready to go into your experiment. There is no scarring on the final product that you get.

Phil: We have been developing and optimizing this process. The first challenge is that scientists are limited by the lenght of oligos available. We studied this and did an extensive market research survey and heard things from researchers saying the biggest challenge of length is that the longer it is the less pure it is. In CRISPR, these smaller fragments create toxicity. We hear that long oligos are a huge advantage for CRISPR to maximize delivery to cells, and they would take long DNA from synthesis companies if available.

Phil: Fundamentally the current industry is stuck at 99.5% cycle efficiency. While you're able to create smaller oligos at 50nt at 80-85% purity, it falls off exponentially beyond that. This is where the industry is today. Anything longer than that is untenable to purify or create something sustainable. Zooming in a little bit and taking a more analytical view, we want to understand what the phenotype of the problem was. We synthesized using phosphoramidite chemistry an internal control sequence that we use. It's a vanilla sequence middle of the road with 50% GC content. We went from 50nt, 100nt, and 200nt. We looked at this with capillary gel electrophoresis and we did a gel separation method where we profiled one species going all the way to the target length and we got an idea of the types of impurities we get. Here, we can see a few phenotypes. The first phenotype you see is a growth in the species that is just shy one nucleotide. We adopted a nomenclature where we call the product length n, and the product length that is 1 short is n-1 impurity. That impurity grows as you grow from a 100mer to a 200mer as a result of the natural inefficiency of the process. In addition to that, we also see a trail of impurities ogoing all the way back to the 1 nucleotide, called a lagging trail band. There's a change in the distribution and amplitude of this impunity. These two things combined result in a calculated impurity of 34% for this 200mer. You could run it on a gel or PAGE and cut it out and that would solve the lagging band trail, but it's hard to take care of the n-1 impurity. It would also cost you more money and take weeks in your turnaround time.

Phil: How does that compare to our performance? We used our FES process. Same sequence. No purification. Straight direct synthesis and release. It's a clean synthesis. No lagging band trail phenotype showing. In addition, when you look at the zoomed in picture, that n-1 peak is significantly lower. This results in an 85% purity at a 200mer. This is the type of quality that we're getting when we synthesize this length.

Phil: We didn't want to stop there. We weanted to push it out to see what reasonable product we can make pushing in length. Switching to 300nt, we still retain high cycle efficiency and we don't see any increase in laggy band trail. There is only a modest growth in the n-1 impurities. We calculate this purity to be 75% and this is really a testament to the fact that we can retain 99.5% cycle efficiency independent of length. So we're confident saying FES outperforms current industry in quality and length.

Phil: For accuracy, all of our oligos are quality controlled using Thermo-Fisher Orbitrap LC-MS. The measured mass is really close to the expected mass but by 1 dalton. By being off by a nucleotide, you expect being off by 300 daltons. So we're confident that we're producing accurate product here.

Phil: The second challenge is sequence complexity. In our market research study, our customers want oligos ordered that can't be synthesized in other companies. This restricts their experiments they can do. A lot of the human genome has segments that are not synthesis friendly. There are often repetitive elements. They need clean DNA. The crux of the problem is that you end up getting secondary structure as a result of DNA folding oer itself and you end up getting stem loop structure. The 300nt structure is shown here at its biological temperature of 37 degrees. You see the stem loop structure here. Raising the temp to 60c which is whree we synthesize at, and most of that melts away. Doing synthesis on this is just as easy as doing a very linear sequence.

Phil: We decided to test this hypothesis and we challenged ourselves with a number of different complex sequences. We had an A-T rich sequence first. This is a 141nt region of the leucine-rich repeat-containing protein which is 78% AT-rich content. We did this in 50, 100, and 141nt fragment. The synthesis is clean all the way through and there's only a modest growth in n-1 impurities, and we hypothesize this is because we're able to raise temperature to higher temperatures. Here we calculate purity to be greater than 80% at this length. With chemical synthesis, we saw an interesting pattern of degradation in the lagging band trail and the resulting impurity was 80%. ....We mass verified this too, numbers are shown on the irght.

Phil: What about 100% G-C content? This is the arguably harder thing to synthesize through. We looked at this through the lense of short tandem repeats. There is an ALS G4C2 repeat expansions... researchers believe this repeat is one of the causes for ALS and the relevant number of repeats for research starts around 30x which translates to 180mer. So you can see what this stem loop structure looks like here. It's a tight stem loop. When we took this through FES, we saw very similar results. We saw that we were able to maintain clean synthesis just a modest growth in that n-1 impurity. One would argue that this is a more challenging sequence to synthesize not just because of stem loop structures but stableplex G quad structures. We were able to synthesize cleanly all the way through 180mer. We tried to get this synthesized throug hchemical synthesis providers and they were unable to deliver at the 30x mark. Even at the 10x or 20x length mark, there was significant signs of degradation. Being able to synthesize at higher temperatures, it helps solve complex sequence synthesis.

Phil: Another test that we wanted to try was homopolymer repeats. There are guidelines that providers put in place in terms of the range of G-C content. They suggest a synthesis successful product requires limiting the length of homopolymer stretches. This is hypothesized because of impurities at the detrinilyation step causing degradation. We ran a 150mer of a polyA and we had extremely clean synthesis here. The aquoeueous pH neutral conditions is good for synthesis. This polyA mer had >99.93% cycle efficiency for us. For chemical synthesis, there was <35% yield.

Phil: What does this early data tell us? It gives us confidence that FES enables new opportunities in life sciences applications. We can make "long" high purity DNA. We have been able to demonstrate up to 300nt. We show that performance is maintained despite sequence complexity. We really are breaking down the traditional guidelines of sequence design. We want to get rid of those guidelines completely. We want to make pure long oligos despite sequence complexity like homopolymer stretches or G-C content or highly repetitive sequences. We want to deliver to customers with faster turnaround time. The green chemistry being able to produce natural nucleotides really makes it ready to use in research applications.

Phil: Since February, we have been putting our product into the market by way of our Key Customer Program which has helped drive and guide product development and get researcher feedback and also show that our oligos are relevant in today's complex applications. We also wanted to target and solve researcher's synthesis challenges. Many of our customers are focused on long donor templates for CRISPR knock-ins and aspects of gene assembly where they are unable to get certain sequences because they are too hard to synthesize traditionally. The customers in our queue in addition to those applications we also have customers have interest in synthesis for massiely parallel reporter assays, cloning, mutagenesis, and others. Steve is one of our first key customers and will give his experience on using our oligos for CRISPR applications.

Steve Bischoff: I am going to talk about an application that we used these donor oligos for recombination-mediated casette exchange (RMCE). .....

Phil: ... we're not done yet in terms of continuing to push product specifications around purity and length. Still a lot of room for us to optimize. Instead of talking about 99.9%, we need to go to 99.99%. If you want to find out more information, we do have a booth manned mainly by Ben Johnson. We're still taking orders and interest from those who want to be part of our Key Customer Program. Find out more at the booth.

Phil: I also want to acknowledge the entire Molecular Assemblies team. We have a talented group of talent working tirelessly to make automated enzymatic DNA synthesis a reality, and our investors for their continued support.

Q: Great data. Thank you for showing the data. Incredible work. Simple question. What yield are you guys offering?

Phil: It's important that we separate the notion of yield... now we really focus on purity. Just to separate these two concepts; in terms of mass output we start with a 200 picomol starting scale.

Q: Are you making the nucleotides yourself or sourcing them?

Phil: We are currently sourcing them from an external vendor.

Q: ... you optimize the reactions... the time of the reactions are optimized by... protein engineering. What time reaction?

Phil: The important thing for customers is turnaround and delivery time. The focus and the promise of enzymatic synthesis was to cut not just the synthesis cycle time, but also the total post synthesis processing time. We want to deliver a 200mer sequence in less than 1 week. That's our target.

Steve: For our application, for these n-1 products, for genome engineering we need these precision oligos and these n-1 products are in the homology arm and we can use that even if it's just dirty for n-1 right. We see challenges in indel lock regions and making it inactive.

Q: So if you become a contractor and synthesize for everything, why not put a robot in a box and send it out to customers? You said turnaround time is everything. Benchtop solves that. You also solve legal issues for sending things out or signing dozens of MTAs when you onboard anyone and nobody wants to share IP.

Phil: In terms of selling a box and provide reagents? It's a good question. We had this conversation early on in our company. Do we sell an instrument or do we provide services? Given where the current landscape is in terms of type of customers that want long oligos, I think they don't really want to mess with managing an instrument. Their ideal scenario is that they can put in an order and get a tube of DNA that they can pout directly into their experiments. With an instrument, there are more challenges: quality control. Now it is not just a set of reagents but it's all the validation processes around creating an instrument. We wanted to tightly control that. So we decided to go with a service model.

Steve: We also source oligos from IDT. I don't want to be synthesizing oligos myself as an end user.

Phil: There is room for both models. We think the industry is broad enough and wide enough to support both models. Maybe one day things will get to the point where every customer will have an instrument in their lab. Right now that's not where the industry and demand is right now.

Q: We like in-sourcing and do everything in house. We have built our own biofoundry. I like your enzyme based on your data. If we had that enzyme in-house, I think there's a market for it.

Phil: Okay.

Q: What do you think is the most major technical challenge to get a more quantitative cycle efficiency? Going from 99.9% to 100%? You're still working on that last bit, what's the most challenging aspect of that?

Phil: Things become astronomically tougher as you get closer to 100%. I think we still have headroom to push that cycle efficiency higher. Really it comes down to finer tuned optimization of our biochemistry. That's something that we feel is we can continue to do that. I can't reveal too much on what those parameters are. It's something that we believe we have headroom on.

Max: That estimate you gave on the potential upcoming turnaround time, is that for a certain length in kilobases or is that for 200mers?

Phil: The target for <1 week is for a 200mer. The synthesis time is proportional, or somewhat proportional to the length. I think it's important to note that we're synthesizing directly and providing single strand oligo. You used the term kilobase...

Max: Elegen says 7 days for 7 kb.

Phil: For the gene side of things, it would go through assembly process or something. It wouldn't be going from 200mers and then multiplying that by 100x a week. No, it would be a different process for that.

Max: Okay, cool.

Steve: ... at a gene editing company, we built constructs. Error correction-- we would take smaller oligos that were correct, rather than junky mutated oligos, and release those and do-- the error correction step of re-sequencing and so forth.