Fyodor D. Urnov Ph.D. Associate Director Altius Institute for Biomedical Sciences
Lasting Lessons from the “Old Testament”
Genome editing with engineered nucleases, a powerful tool for understanding biological function and revealing causality, was built in a joint effort by academia and industry in 1994–2010. Use of CRISPR/Cas9 is the most recent (2013–), and facile, implementation of the resulting editing toolbox. Principles and methods of genome editing from the pre-CRISPR era remain relevant and continue to be useful.
In the beginning…
…was the double-strand break (DSB). Induced at a specific position of a mammalian chromosome by Maria Jasin in 1994 (19 B.C. [before CRISPR*]),1 it set in motion a chain of events, including landmark studies by Dana Carroll (2000–2003; 13–10 B.C.),2–4 that brought us all directly, in 2018, to genome-edited potatoes,5 and people,6 and a Noah's ark of edited cells and organisms.
As a result, biomedical scientists now have a powerful remedy for a formerly crippling epistemological problem (Figure 1): in the absence of information, the human mind fills in the gaps with things it simply makes up. Genome editing is a tool to get away from these “just so stories” and closer to biological reality. It is also the definitive tool for the determination of causality.
The toolbox of editing was developed between 1994 and 2010 in a joint effort between academia and industry, using meganucleases for proof-of-concept7 and zinc finger nucleases (ZFNs) for editing of native loci.8 In 2010–2012, the toolbox was rapidly ported, wholesale, to a third nuclease class—one based on the TAL effector domain.9
The history of industrial progress has several examples of a “1 + 1 = 7” effect from the convergence of previously unconnected lines of effort. Aluminum, discovered in 1825 and purified at scale in 1886, found its widespread use with the invention of the airplane and the rise of aviation in the 1910s. The Corning Company developed “gorilla glass” in the 1960s for potential use in automobile and plane windshields; it became ubiquitous as the surface of the iPhone a half-century later.
In 1987–1989, an “unusual arrangement with repeated sequences” was observed at a specific locus in the Escherichia coligenome—an arrangement now known as a “clustered regularly interspaced short palindromic repeat,” or CRISPR.10,11 This launched a quarter-century of studies not in genome engineering, but in bacterial immunity centered on such CRISPR loci.12,13 As described below, this work ultimately led to the discovery, in 2012, that a key enzyme of a specific CRISPR-based system, Cas9, is an RNA-guided endonuclease.14 This discovery resonated in a special way, one unrelated to bacterial immunity, because of the decade-long precedent of genome editing with ZFNs and, more recently, TAL effector nucleases (TALENs). Thus, in 2012, Martin Jinek, Jennifer Doudna, Emmanuelle Charpentier, and colleagues wrote: “Zinc-finger nucleases and transcription-activator–like effector nucleases have attracted considerable interest as artificial enzymes engineered to manipulate genomes. We propose an alternative methodology based on RNA-programmed Cas9 that could offer considerable potential for gene-targeting and genome-editing applications.”14 A short time later, Gasiunas et al. wrote: “Taken together, these findings pave the way for the development of unique molecular tools for RNA-directed DNA surgery.”15 In January 2013, publications from four laboratories—those of George Church, Doudna, Jin-Soo Kim, and Feng Zhang—presented data reducing this suggestion to practice.16–19
The brief history of CRISPR-Cas9 gene editing has already received considerable analysis. Here, I aim to discuss the findings from the approximately 18-year-long effort on building genome editing that preceded the Age of CRISPR (Figure 2). It focuses on principles and methods that remain relevant and useful to genome editors today.
Genetic Engineering Before Editing: Of Mice and Yeast
Between the invention of recombinant DNA20 and DNA sequencing21 in the 1970s, and until the toolbox of editing was essentially complete by 2010,8 targeted genetic engineering was the province of a select few model organisms, progress in the study of which was much faster than in other, genetically intractable, systems.
34 B.C. (1979)
Gene targeting is born. Gerry Fink's and Ron Davis's laboratories showed22,23 that a yeast gene can be replaced with a selectable marker and thus knocked out. For example, the Davis lab transformed yeast cells with a plasmid in which the marker (URA3) is flanked by regions of homology to the gene of interest (HIS3). Selection for marker-positive cells shows that URA3-positive cells carry the marker at HIS3. This method became a key source of the “awesome power of yeast genetics,” and over the subsequent three decades, the yeast research community uses it, among other things, to make a collection of null and hypomorphic alleles across the entire yeast genome, widely used for reverse genetic screens.
27 B.C. (1986)
Gene targeting works in mouse embryonic stem (ES) cells. Inspired by the precedent from yeast, Mario Capecchi showed24that a gene that can be selected against in mouse ES cells (HPRT) can be knocked out by targeting to it a different gene—a marker that can be selected for (neor). In 1992, Michael Rudnicki and Rudolf Jaenisch discover,25 in the course of efforts to knock out genes that cannot be selected for, that long stretches of isogenic DNA are required for efficient targeting. Many elegant technical improvements are invented by the mouse research community, and, just as in yeast, classical gene targeting (the use of selection to transfer one gene to another) becomes the basis for comprehensive collections of knockout mouse ES cells, and mice genetically engineered in sophisticated ways. Use of classical gene targeting in settings other than mouse ES cells remains, to this day, an experimental challenge (see below).
18 B.C. (1995)
Gene targeting can work in settings other than mouse ES cells. John Sedivy, among others, uses “gene targeting vectors” to knock out cell-cycle-control related genes in primary human fibroblasts.26 The results are thought-provoking, which makes the technical challenges associated with gene targeting an unfortunate feature of the state of the field at the time.
11 B.C. (2002)
Gene targeting can be improved via the use of adeno-associated virus (AAV) vectors. David Russell makes the important discovery27 that if one uses AAV to deliver the “gene targeting construct,” its efficiencies improve. Bert Vogelstein, in 2004, expands this observation to show28 that in HCT116 cells, use of AAV-based gene targeting produces, after selection, 13 out of 244 drug-resistant clones that have one allele of a representative gene (CCR5) knocked out.
Gene targeting made yeast and mice into the dominant genetic systems used between 1980 and 2012. By 2005, however, after 11 years of effort (see below), genome engineers had found not merely a “better mousetrap.” They found a cat: a fundamentally new way to alter the genome, one that requires neither targeting vectors nor selection, one that can generate any allele in a single step, and one that works in organisms as well as in cells. That “cat,” of course, is genome editing. As will be seen below, its current utility and ubiquity derives from a thoughtful collaboration with Mother Nature, specifically, with the equally ubiquitous process of DSB repair.
Lasting Lessons from the B.C. Age
18 B.C. (1994): a DSB is editogenic in mammalian cells
This was the first key discovery of the Age of Editing. Maria Jasin found that a single DSB delivered by an enzyme to a mammalian chromosome in mitotically dividing cells is efficiently repaired either by homology-directed repair (HDR) from an investigator-provided template, or by nonhomologous end joining (NHEJ).1
To appreciate this finding, some context. In 1982, Jeffrey Strathern et al. discovered that a DSB initiates the transfer of genetic information during mating type switching in budding yeast,29 and over the subsequent decades, a wealth of knowledge about the role of DSBs in genetic exchange is obtained, notably by Jim Haber,30 from studying this system.31 In 1983, Jack Szostak, Terry Orr-Weaver, Rodney Rothstein, and Franklin Stahl proposed that a DSB is the initiating event in meiotic recombination.32 By the late 1980s, dogma in the field of DNA repair held that end joining, rather than HDR, is the dominant DSB pathway in mitotically dividing mammalian cells in culture (this was in part based on low efficiencies of classical gene targeting).
Jasin's interest in this problem began when she was a postdoctoral fellow with Paul Berg in the late 1980s. She started her laboratory at Memorial Sloan Kettering in 1990, where she reasoned (see Supplementary Data Box 1 “The Book of Jasin”; Supplementary Data are available online at www.liebertpub.com/crispr) that this problem could be resolved using a nuclease that (1) would induce a single DSB in the chromosome of a mammalian cell and (2) would be tolerated by that cell. In the first of several remarkable “leaps across species” from which the field of genome editing has benefitted, she repurposed an enzyme discovered in 1985 by Bernard Dujon: a nuclease, I-SceI, that evolved to function inside yeast mitochondria to cut an 18 base pairs (bp) target site and thereby spread its own open reading frame.33 The experiment made use of a chromosomal reporter gene in which a selectable marker was interrupted by the recognition site of I-SceI, and thus selection could be used to track down rare cells of interest. Cells transfected with a 700 bp fragment carrying the wild-type marker sequence produced no wild-type progeny. Co-delivering the DSB-inducing nuclease produced a wealth of these, however, and an analysis of their DNA sequences showed they have resulted from HDR-based replacement of the nuclease target site with wild-type DNA. Delivering the nuclease alone also produced wild-type cells, with sequence analysis showing they resulted from a microhomology-driven, end joining–based elimination of the nuclease site from the reporter gene. From one experiment with a reporter construct thus came three conclusions that resonate to this day: (1) in mitotically dividing mammalian cells in culture, a nuclease-induced DSB enhances—from undetectable to readily revealed by selection—the HDR-based transfer of genetic information into the chromosome from an ectopic fragment; (2) NHEJ is a competing pathway for repair of this DSB, which produces small insertions and deletions (indels) at the nuclease target; and (3) expression of such a nuclease is tolerated by mammalian cells.
Horace Judson's incomparable history of molecular biology, The Eighth Day of Creation, recounts the moment when Francis Crick came to the lab where Leslie Barnett had just obtained the results from the legendary “uncles and aunts” experiment that asked how many base pairs form each “letter” of the genetic code.34 Crick looked at the plates with the phage plaques, then looked at Barnett and said, “You and I are the only two people on Earth who know that the genetic code is a triplet.” Philippe Rouet, Fatima Smih, and Jasin, in examining the indels and precisely repaired chromosomes induced by I-SceI, were at that moment in 1993 or early 1994 in that rare category of scientists: unique in their knowledge of a fundamentally important truth.
13–10 B.C. (2000–2003): an enzyme can be engineered to induce a DSB and drive editing at a locus of interest in eukaryotic cells
This was the second key discovery of the Age of Editing. Aware of the implications of the Jasin results (targeted DSB = editing!) and of challenges regarding reengineering of meganucleases for genes of interest, Dana Carroll set his sights on a recently invented engineering restriction enzyme: the ZFN (a fusion between a zinc finger–based DNA binding domain and the cleavage domain from a restriction enzyme, FokI). ZFNs were originally built, at the suggestion of Hamilton Smith, by the Chandrasegaran laboratory, for use as novel restriction enzymes.35 In a telling example of how hard it is to predict the path of technology development, these engineered enzymes found their true calling in a setting completely unrelated to what they were invented for.
Dana Carroll trained with Don Brown and, in starting his own research laboratory at the University of Utah, was well aware of the unique power of the Xenopus oocyte as a model system: it allows the rapid in-cell evaluation of the action of a large number of DNA-binding proteins on any number of engineered templates.36 Further, zinc fingers were discovered by Aaron Klug in Xenopus,37 and while they represent the dominant class of DNA-binding domains across all metazoa,38 if there was one species where a zinc finger–based new enzyme had a chance, it was Xenopus. Carroll used this system for all its might, identifying the biochemical parameters that govern efficient target recognition, cutting, and recombination on a plasmid template by ZFNs in the oocyte.2 The stage was now set (see Supplementary Data Box 1: “The Book of Carroll”) for editing a gene.
The history of biomedicine has several examples of a large gap between someone proposing a method and someone actually getting it to work. Take the polymerase chain reaction (PCR), which was described as a concept, in essentially complete form, in a 1971 publication by no less than H.G. Khorana, who closed the paper stating, “Experiments based on these lines of thought are in progress.”39 Whatever happened to those experiments, PCR was not described in working form until some 17 years later by Mullis et al.40
In order not to fall into this gap between proposal and action, Carroll wisely chose a system where one could (1) deliver the nuclease efficiently and (2) score for the editing outcome by phenotype. He used Drosophila, where the transgene encoding the ZFNs could be inserted stably and randomly into the genome, and has an eye-color phenotype. As reported in Genetics in 2002,3 about 6% of the animals exposed to the ZFNs were phenotypically mutant. Sequencing of the ZFN target in the yellowgene in these mutant animals showed a range of small indels precisely at the position of the DSB. According to Carroll, his collaborator, Kent Golic, who developed a sophisticated method for gene targeting in Drosophila, looked at the data and said, “If I were you, I would be pretty excited.” In retrospect, this was an understatement (examine the nuclease-induced indels in Figure 5 in that paper and imagine flying back in time and telling Marina Bibikova and Dana Carroll what they just discovered). It is also deeply affecting that this landmark advance in molecular genetics was made using the same model organism, and the same class of phenotype, as T.H. Morgan's classic experiments a century prior.41
For the rest of this article and its references click here.
Fyodor D. Urnov, Ph.D. (firstname.lastname@example.org) is associate director at Altius Institute for Biomedical Sciences.
The CRISPR Journal, published by Mary Ann Liebert, Inc., delivers cutting-edge multidisciplinary peer-reviewed research, advances, and commentary on CRISPR, the extraordinary technology that gives scientists the power to cure disease and sculpt evolution. The above article was first published in the February, 2018 issue of The CRISPR Journal with the title “Genome Editing B.C. (Before CRISPR): Lasting Lessons from the “Old Testament". The views expressed here are those of the authors and are not necessarily those of The CRISPR Journal, Mary Ann Liebert, Inc., publishers, or their affiliates. No endorsement of any entity or technology is implied.