David R. Liu, PhD, joined the Harvard University chemistry faculty in 1999 at the tender age of 26 years. When you excel at research through high school, undergraduate, and PhD, postdoctoral fellowships become somewhat superfluous—although Liu regrets skipping his postdoc and does not recommend others bypass such training. In 2005, he was promoted to full professor and appointed an investigator with the Howard Hughes Medical Institute (HHMI).
Although his diverse research interests include phage-directed evolution and small-molecule drug discovery, it is his laboratory’s groundbreaking work in genome editing that has truly garnered the interest of the human gene therapy community. In 2016, Liu’s postdoctoral fellow Alexis Komor, PhD, spearheaded the development of the first base editing technology, which had the ability to install specific base substitutions (C to T, or G to A) without cleaving DNA. Eighteen months later, her colleague Nicole Gaudelli, PhD, developed a complementary adenine (A to G) base editor (both studies were published in Nature). Base editing technology is being commercially developed by Beam Therapeutics.
More recently, the Liu laboratory has described prime editing, which enables the installation of all base substitutions, small insertions, and small deletions by using an RNA rather than a DNA template, as well as the first method for making precise changes to the sequence of mitochondrial DNA.
In this revealing interview, David Liu recalls his first foray into genome editing 20 years ago and discusses recent progress, including exciting preclinical results, in a range of genetic disorders applying base editing. (This interview has been lightly edited for length and clarity; a longer version can be found in GEN’s sister journal, Human Gene Therapy.)
GEN Edge: I would like to start by going back to your first interest in genome editing. I read that you even dabbled in some work in this space 20 years ago?
David Liu: You’ve done your homework, or at least you’ve heard the rumors!
I began my academic career in 1999. One of the first projects we started, when I had just a few graduate students as an assistant professor, was a project we nicknamed the Unifactor 2000, because it sounded futuristic to have 2000 in the title. The purpose of the project was to develop a universal transcription factor. Our idea, in retrospect naive in a number of ways, was to digitally address DNA using triplex formation and recruit a piece of RNA or DNA linked to a DNA-cutting enzyme, transcriptional activator, or transcriptional repressor in a sequence-programmable manner.
To support the feasibility of the Unifactor 2000, we had identified reports that researchers had successfully observed triplex formation, even in cells. And although triplex formation is a well-documented phenomenon, it requires conditions that are mostly mutually exclusive with conditions in cells. That turned out to be the Achilles heel of the project, but we were very excited about the possibility of simply recruiting different effector domains to sites in the genome in a sequence-programmable manner. It’s just that a way to do that efficiently in cells did not really exist in 1999–2000, so probably around the time the Unifactor 2000 project reached its namesake year, we killed the project.
GEN Edge: The arrival of Alexis Komor in your laboratory was hugely important in the development of base editing, but to what degree were you exploring genome editing before that?
Liu: We actually first got into genome editing through my good friend and colleague Keith Joung, who was then a young professor at Mass General Hospital at Harvard Medical School (HMS). A student who was on an MD/PhD track—Vikram Pattanayak—was an HMS student who happened to sit next to Keith at an HMS retreat dinner. They had a conversation, which resulted in Vikram becoming interested in the idea of zinc fingers, their programmability, and the extent to which they were able to bind DNA at designed targets…
We completed that initial project, “Revealing off-target cleavage specificities of zinc-finger nucleases by in vitro selection,” published in 2011. 1 That was the first article in the genome editing field we ever published. We then went on to perform similar substrate specificity profiles powered by Darwinian selection on TALENs, also in collaboration with Keith’s laboratory, and then on CRISPR-Cas9 in collaboration with Jennifer Doudna’s laboratory.
So understanding specificity was really the first foray into gene editing for our laboratory and then we quickly transitioned to how to improve some of the key features of gene editing agents as we started to identify their bottlenecks.
GEN Edge: It has been about five years since your seminal base editing report (Komor et al.) in Nature. 2 As you look back on the past five years, are you surprised just how fast this technology has taken off?
Liu: I am incredibly surprised! If you would asked me in 2016 how long would it take for people to be able to take engineered gene editing macromolecular machines, deliver them into an animal so that they could correct one base-pair causing a grievous genetic disease like progeria and rescue the symptoms in the animal, I probably would have said, “Optimistically, I hope within 10 or 20 years.”
So much has to go right: you have to develop the machine and have it function with an efficiency and a specificity that does not cause problems in the animal. You have to figure out how to deliver this machine, which has traditionally been one of the challenges in any kind of applied protein engineering. You have to hope that the biology that links the genetics to the disease is rock solid. And of course, you will be trying to forge a path for connecting the biology of the corrected allele to the rescue of the disease, which is not a guarantee either.
It still sounds like science fiction when we think about it. I think it is a testament not only to the hard work and talents of people such as Alexis Komor (Fig. 1) but also really to the entire field. There are many laboratories that are using and improving base editors. (The nonprofit) Addgene says there have been about 11,000 base editing constructs sent out just from our laboratory, and I’m sure thousands more from other laboratories as well.
GEN Edge: There was another base editing article led by Akhiko Kondo, from Kobe University, in 2016. How does their technology differ from or complement yours?
Liu: The Kondo group had been independently working on ways to mutate genomes using deaminases linked to DNA-binding proteins. Several months after our article,2 they published their study in Science3 reporting the fusion of a different deaminase to Cas9—it has slightly different properties, because the deaminase has slightly different enzymatic properties.
I think the contributions that we made that may have helped their study included nicking the nonedited strand and fusing the UGI, for which they cited our article—but they may have been thinking along those lines anyway… It established that two totally different laboratories with different deaminases and different interests—the Kondo laboratory’s focus then was agriculture and biofuels—could show that this was a robust approach to making targeted point mutations in the genome.
Second, what happens next in terms of translating the technology into some kind of societal benefit was important. After I gave one of my very first talks on base editing before the article was published, I was approached by (VC firms) F-Prime and Arch to start a company around base editing. My naive initial answer was, “I’m already a co-founder of a gene editing company, Editas, and they should have the opportunity to incorporate base editing into their company.” They said, “We think base editing should be a separate company.” Their argument was that being a different technology would require a different effort to optimize and to apply successfully, and the targets that we would pursue would also be different. Ultimately, I think it was good advice.
Another good decision we made early on was to not try to fragment the base editing field into multiple companies. If our main interest is to bring base editing technology to benefit patients, we should try to get everybody under the same tent. The group that would end up forming Beam Therapeutics reached out to Professors Kondo and Nishida, who published that Science article, and invited them to join forces and be part of the company, which they did. Their own commercialization effort sublicensed their part of base editing to Beam for human therapeutics outside of microbiome applications, while keeping their own ability to develop their base editors for microbiome use.
The important point is base editing did not become multiple companies spending resources trying to compete with each other.
GEN Edge: Your progeria study was just published in Nature4 in collaboration with NIH Director Francis Collins. Why is that study so exciting and what are the prospects for translating this study in a mouse model into patients?
Liu: The project was really interesting and had a surprising outcome for several reasons. Among all the various gene editing therapeutics efforts that we are doing in our laboratory or in collaboration with other laboratories, I would have said at the outset that progeria would be toward the challenging end. It is a systemic disease, so it must be addressed, therefore, in vivo. The proximal cause of death is often cardiovascular failure. You are not going to edit the heart ex vivo and then transplant it, presumably. And it is a grievous progressive genetic disease for which there is no treatment that has been shown to greatly extend lifespan. It also was not clear what the window of opportunity would be to correct the disease… it could be that once an animal with progeria is born, its fate is sealed. All of those features—systemic, in vivo, progressive, rapidly degenerating disease—suggested this could be a real challenge.
GEN Edge: But you did it anyway…!
Dr. Liu: Yes. I really credit the graduate student, Luke Koblan, who led the effort and decided to try our brand new A-base editor (ABE), which we had not even published at the time (the development of which was led by Nicole Gaudelli)…5
The results were encouraging enough that we started a collaboration with Jonathan Brown at Vanderbilt, who studies vascular biology and disease… We injected the single pilot mouse with AAV encoding this base editor. It is the kind of swing-for-the-fences pilot experiment that is arguably naively optimistic, because it is a stretch to go from patient-derived fibroblasts to a mouse. The split-intein dual AAV system we used to deliver the base editor into the mouse had not even been published at that time.
But they injected the single mouse. We did not have at the time the sophistication or the manpower to do extensive vascular pathology analysis or lifespan analysis. We simply took some of the basic organs such as the liver and the heart and sequenced them. We were surprised to find 20–60% editing in most of the organs that we sequenced.
And when we did an RNA analysis, we saw that the amount of the toxic progerin RNA that results from mis-splicing caused by this silent point mutation had gone down quite a bit and the amount of protein in the Western blot had also gone down in favor of the healthy normally spliced lamin A protein, which was a remarkable outcome.
We got that data shortly before I was invited to give a talk at the NIH.… Right after the talk, Francis [Collins] offered to collaborate and explained that he had this large colony of progeria mice that each has two copies of the human mutated lamin A gene. The fact that they had the human gene really makes the work more therapeutically relevant because in principle the exact same composition of matter AAV9 targeting the human progerin gene that we use to treat these mice could be used in a patient.
GEN Edge: How does this look for potentially trying this in patients?
Liu: We’ve decided to take two approaches. One approach is a recognition that although there are some treatments, including the recently FDA-approved small molecule called lonafarrnib, which works by inhibiting protein farneslyation, which offer patients some benefits in quality of life and lifespans, there has been no treatment, including lonafarnib, that seems to show the kind of lifespan extension and general animal vitality rescue that we observed from directly correcting the root cause of the mutation with a base editor.
Leslie Gordon from the Progeria Research Foundation made compelling arguments that we should take two approaches: we should advance the current treatment in its current form and do the remaining toxicity and biodistribution studies to pave the way for a potential clinical trial of pretty much what we reported in Nature.
In the meantime… we’re hoping to optimize the base editors, AAV, and the timing of the dose to get a deeper understanding of what kind of patients might benefit, and how to offer them the highest ratio of potential benefit to potential risk.
GEN Edge: The other interesting animal model study in base editing last year was from Sek Kathiresan and his team at Verve Therapeutics. They are tackling genetic forms of heart disease but looking at the potential of base editing to treat a much broader set of complex diseases. I wonder whether you are comfortable with that?
Liu: I’m not involved in Verve, although Verve and Beam are co-developing Verve’s base editing therapeutic, which Verve has announced is their lead program. From my vantage point, it is an amazing outcome. They sought to knock down PCSK9. They compared doing so with a Cas9 nuclease and with a base editor, and they found a better outcome with the base editor.
Of course, the really exciting data are that they treated nonhuman primate monkeys with a lipid that delivers a base editor into the liver; it edited quite efficiently the PCSK9 gene, knocking it down. The result is dramatically improved blood parameters, such as serum low-density lipoprotein cholesterol and triglycerides.
The second part of your question is really interesting. It was in my opinion a brilliant choice of target, because there are lots of human genetics around PCSK9 that has established a pretty strong relationship between knocking out the gene and lowering risk of cardiovascular disease. It provides an opportunity—in my opinion—to choose a patient population that is matched to both the medical need and to societal acceptance of gene editing. Of course, there are open questions that, as the field of clinical base editing matures, will help guide the best risk–benefit ratio.
In other words, you can start out by treating patients with very high-risk familial hypercholesterolemia—people who have a very high chance of having heart attacks or strokes or other cardiovascular problems much younger than normal. In that case you can say, even if this is a new experimental therapeutic modality, the potential benefits are worth the potential risks. But if the unknowns begin to be known, and the risks begin to be understood to be acceptable, and the technology is optimized with respect to minimizing off-target editing and delivering efficiently and editing on target efficiently—then as you point out, it does raise the possibility that perhaps not just the most endangered subpopulations of patients, but maybe also a broader set of patients, could benefit.
For that matter, there are other conditions in which there are known disease prevention alleles or disease risk alleles that could be corrected or installed with the base edit. Alzheimer’s disease, of course, has APOE4, which is a serious risk allele. But it also has Icelandic amyloid precursor protein (APP) variant—Ala673Thr—which confers strong protective benefit enjoyed by roughly 1 out of 1,000 or fewer Icelandic people and virtually nobody else. In other words, from the perspective of APP, the vast majority of the human population has the disease-causing allele.
It would be an interesting question: if clinical base editing begins to flourish and is viewed as safe and efficacious, whether society should pursue not just disease-correction alleles or alleles that elevate disease risk like APOE4, but also perhaps disease prevention, like the PCSK9 example or the APP example. There are others—there is a prion mutation that lowers your risk of prion disease, for example.
GEN Edge: There are some 7,000 known Mendelian genetic diseases. What proportion of those genetic diseases is base editing potentially able to help correct? How do you begin to make a bigger dent in tackling not just a few cases but hundreds or thousands?
Liu: It is a great question, really the question that I open most of my gene editing talks with: this big pie chart that says here are 75,000 known human gene variants associated with genetic disease (Fig. 2). What fraction of them is editable by which technologies? The answer to that question depends on what is the editing capability of that technology? For base editing, it is primarily converting As to Gs, Gs to As, Cs to Ts, and Ts to Cs; and second, what fraction of the mutations that fit that class of correction is accessible by those editing agents?
When Alexis Komor published our article in 2016, the answer was we could only make two kinds of changes—C-to-T and G-to-A—and we only showed that we could do this with canonical spCas9-derived base editors, we could only position the base editing activity window over ∼25% of those pathogenic mutations that were in principle correctable by making one of those two changes.
Thanks to the hard work of many laboratories including those of Keith Joung, Ben Kleinstiver, Osamu Nureki, and our own, a variety of Cas variants have been engineered or evolved that have greatly expanded the flexibility with respect to where you can park a base editor by offering tremendous PAM flexibility. If you crunch the numbers, the math works out that now say 95% of pathogenic transition mutations can be addressed with a base editor. Of course, now C base editors and ABEs, which collectively can cover something like 25–30% of all known pathogenic human gene variants, are widely used.
That still leaves a huge amount of territory, which was one of the inspirations behind our development of prime editing.6 Prime editors can make all 12 kinds of base-to-base changes, as well as small insertions and deletions. Although prime editors are much newer and have gone through fewer rounds of improvement—there have only been a couple of dozen articles or preprints published as opposed to several hundred base editor articles at this point. But the community and our laboratory are working really hard on expanding the scope of prime editing the same way that the community expanded the scope of base editing, to hopefully cover a large majority of that enormous pie chart.
GEN Edge: You presented prime editing for the first time at Cold Spring Harbor in late 2019. At the end you acknowledged the first author of the article, Andrew Anzalone, and quipped, “I’m really looking forward to seeing what Andrew can do in the second year of his postdoc!” Has he been as productive as his first year?!
Liu: [Andrew] became one of the founding scientists at a company called Prime Medicine to transform his remarkable work into therapeutics… Andrew is one of the most talented and collaborative people I have ever had the pleasure of mentoring in 22 years. In the second year, he developed some improvements of prime editing that we have not reported yet but hope to soon. He contributed to a number of projects in ways that are hard to imagine. I started to notice that when Andrew presented at group meetings over Zoom, more people than normal would come just to hear him talk. He grew quite a following.
GEN Edge: 2020 was a big year for CRISPR with the award of the Nobel Prize for Chemistry. I’m sure there were some mixed emotions at the Broad, but you put out a nice statement congratulating Jennifer and Emmanuelle. What do you think that award meant for the field of CRISPR and genome editing more broadly?
Liu: I think everybody in the gene editing field, including everybody at the Broad Institute, was really happy and excited to see the seminal work of Emmanuelle Charpentier and Jennifer Doudna recognized by the Nobel Prize in Chemistry. You cannot award prizes to everybody who has contributed to a field and in a sense, the recognition that their contributions—and by association, the contributions of everybody who have really built that field so quickly into one that is already having impact on society—is a wonderful incredibly positive development for the whole field. Their seminal article—the first to combine gene editing with CRISPR—was highly influential for many researchers, including myself. They are richly deserving of this recognition.
The full transcript of this interview can be found in the journal Human Gene Therapy (March 2021).
- Pattanayak V, Ramirez CL, Joung JK, et al. Revealing off-target cleavage specificities of zinc-finger nucleases by in vitro selection. Nat Methods 2011;8:765–770.
- Komor AC, Kim YB, Packer MS, et al. Programmable editing of a target base in genomic DNA without double-stranded DNA cleavage. Nature 2016;533:420–424.
- Nishida K, Arazoe T, Yachie N, et al. Targeted nucleotide editing using hybrid prokaryotic and vertebrate adaptive immune systems. Science 2016;353:aaf8729.
- Koblan LW, Erdos MR, Wilson C, et al. In vivo vase editing rescues Hutchinson-Gilford progeria syndrome in mice. Nature 2021;589:608–614.
- Gaudelli NM, Komor AC, Rees HA, et al. Programmable base editing of A · T to G · C in genomic DNA without DNA cleavage. Nature 2017;551:464–471.
- Anzalone AV, Randolph PB, Davis JR, et al. Search-and-replace genome editing without double-strand breaks or donor DNA. Nature 2019;576:149–157.