Powerful tools for genetic modification: Advances in gene editing
Abstract
Recent discoveries and technical advances in genetic engineering, methods called gene or genome editing, provide hope for repairing genes that cause diseases like cystic fibrosis (CF) or otherwise altering a gene for therapeutic benefit. There are both hopes and hurdles with these technologies, with new ideas emerging almost daily. Initial studies using intestinal organoid cultures carrying the common, F508del mutation have shown that gene editing by CRISPR/Cas9 can convert cells lacking CFTR function to cells with normal channel function, providing a precedent that this technology can be harnessed for CF. While this is an important precedent, the challenges that remain are not trivial. A logistical issue for this and many other genetic diseases is genetic heterogeneity. Approximately, 2000 mutations associated with CF have been found in CFTR, the gene responsible for CF, and thus a feasible strategy that would encompass all individuals affected by the disease is particularly difficult to envision. However, single strategies that would be applicable to all subjects affected by CF have been conceived and are being investigated. With all of these approaches, efficiency (the proportion of cells edited), accuracy (how often other sites in the genome are affected), and delivery of the gene editing components to the desired cells are perhaps the most significant, impending hurdles. Our understanding of each of these areas is increasing rapidly, and while it is impossible to predict when a successful strategy will reach the clinic, there is every reason to believe it is a question of “when” and not “if.”
1 INTRODUCTION
CFTR, the gene responsible for cystic fibrosis (CF), was identified in 1989 by virtue of its position in the genome1 and because most individuals with CF carry a common mutation in CFTR that could be tracked through pedigrees.2, 3 This mutation, originally referred to as ΔF508, is a 3-nucleotide deletion that eliminates a single amino acid, phenylalanine, at the 508th position of the CF transmembrane conductance regulator (CFTR) protein. This mutation has been found to account for about 70% of the CF chromosomes in the population and it was quickly recognized that the remaining 30% of CF alleles are comprised of hundreds of disease-causing mutations4 spread out across the nearly 200 000 base pairs that make up the CFTR gene. To date, over 2000 variants have been reported in the CFTR gene (https://cftr2.org/resources), although the relationship to disease has not been established for most.
Such extensive genetic heterogeneity warranted a therapeutic approach agnostic to the specific mutations carried by a CF patient. Early studies were carried out using a complementation approach; a shortened form of the gene, called a mini-gene, containing the protein-coding sequences of the CFTR gene (cDNA) controlled by a strong, constitutive transcriptional promoter to achieve constant, high levels of CFTR messenger RNA from this mini-gene. These trials showed that introducing a “normal” copy of the cDNA into CF cells could confer chloride transport properties that were otherwise missing from CF cells. This ion transport correction of CF cells provided hope of a DNA-based therapy for CF.5, 6 The number and distribution of mutations for this approach were of little concern, as the cDNA would provide a functional copy of the gene no matter where the mutations might lie. This approach neglected the gene's natural regulatory mechanisms, many of which are still unknown, but as this was the only gene-based strategy with the available technologies of the time, lack of endogenous gene regulation was not considered crucial.
The first generation of gene-based therapies have taught much about nucleic acid delivery, but the approach has not proven successful clinically for CF.7 However, recent successes with pharmaceuticals that augment activity of mutant CFTR have shown that therapies directed at the primary defect of this disease can be very successful at slowing and even preventing onset of clinical manifestations of CF.8, 9 While effective, these therapies are expensive and must be provided daily and thus a strategy to fix the defective gene once and for all is still quite appealing.
Recently, methods to repair chromosomal DNA have been discovered and modified toward the goal of therapeutic application. These methods, termed gene editing as a group, have evolved to describe the ability to harness DNA repair mechanisms as a means to generate precise changes in the genome. While the concept is not new, the ability to carry out such processes efficiently and quickly in mammalian cells has only recently become possible. With this technology comes a plethora of conceivable applications, both for basic research and medicine. The ability to generate genetically matched systems, such as cell lines and animal models, differing only by a single gene, should improve laboratory studies by reducing variables. The potential clinical application is staggering: an estimated 7000 single gene disorders affect humans and it is now conceivable to consider treating a genetic disease at the DNA level. If the efficiency of the editing process can be increased and methods to deliver the editing components to the intended cells can be developed, these technologies could have immense impact on genetic diseases. An overview of the technologies and potential uses for CF is outlined here.
2 GENE EDITING: HARNESSING DNA DAMAGE AND REPAIR
Our chromosomes are constantly under fire from DNA-damaging agents such as chemical carcinogens, oxidative stress, and ultraviolet light, but our cells have evolved mechanisms to repair those breaks. When those repair mechanisms fail to work effectively, diseases can occur, most commonly cancer, and as a consequence much has been learned about DNA damage and its repair processes.
One of the important discoveries in this field was that if a template were available to the broken, or damaged DNA, it could be used by the cell as a guide to the repair process, incorporating the template sequence. The fundamentals of this process led to the ability to engineer mice with specific changes in their DNA and the first to achieve that goal in the late 1980s became Nobel laureates in 2007. Mario Capecchi,10 Oliver Smithies,11 and Sir Martin Evans12 showed that it was possible to generate human genetic disease models by specifically targeting changes in the mouse genome. With this technological advance came the hope the opposite process, repairing a mutant gene, might become a reality. In the early days and until quite recently, this process was reliant on randomly occurring breaks in the DNA and thus the chance that such a break occurred at the desired location was low (on the order of one in a million), limiting how this technology could be used.
2.1 Engineered nucleases to target specific sites in the genome
The utility of harnessing this repair mechanism was clearly recognized, as well as the need for ways to make the DNA breaks more specific, and in 1996, a breakthrough was reported in which synthetic peptides called zinc-finger nucleases (ZFNs) were engineered.13 Zinc fingers are domains of many DNA binding proteins that recognize specific, but short, three base-pair DNA sequences. Mixing and matching multiple zinc finger domains provided the ability to generate very specific DNA-binding proteins and when fused to an endonuclease domain that cleaves DNA these ZFNs could make targeted breaks in the genome, limited only by the zinc finger combinations that could be generated. For more information, the reader is referred to14 for review.
Shortly after the initial reports of ZFNs, a similar technology emerged using DNA binding motifs called TALs,15 for Transcription Activator-Like proteins from bacteria. By mixing and matching various TALs and fusing to an endonuclease that cleaves DNA, Transcription activator-like effector nucleases, or TALENs were developed.16 Both ZFNs and TALENs provided immense potential for making very specific, targeted DNA breaks and were pursued extensively for many purposes, including gene repair. Multiple disorders are being approached with these methods, with clinical trials using ZFNs ongoing for HIV, hemophilia, mucopolysaccharidoses, and others (https://clinicaltrials.gov). While these examples validated the concept of gene repair as a therapeutic strategy, efficient delivery to relevant tissues along with other technical hurdles have prevented use of these systems for CF.
In 2013, a series of manuscripts were published that used a very different mechanism to target DNA breaks,17, 18 a mechanism that bacteria use to protect themselves from viral infections. This system is founded on the principal that RNA and DNA hybridize to each other, rather than relying on DNA binding motifs of proteins. The system, called CRISPR/Cas9 (CRISPR = Clustered Regularly Interspaced Short Palindromic Repeats), involves a set of RNA molecules that assemble with each other. One, called the guide RNA (gRNA), carries sequences of about 18-20 base-pairs that recognize sequences complementary to the DNA target. This gRNA is also capable of tethering a nuclease, Cas9, through a second RNA molecule. In their native state in bacteria, these RNAs along with Cas9 form a complex that is targeted to the invading DNA because of the DNA:RNA sequence complementarity between gRNA and virus. In bacteria, the gRNA recognizes viral sequences and the nuclease cleaves the viral DNA, disabling it. By modifying the components to be produced and functional in mammalian and other eurkaryotic cells, it is now possible to quickly and inexpensively target a DNA break in nearly any region of the genome.
To cleave the DNA, the Cas9 enzyme has very simple but specific requirements. Sequences of a few nucleotides are recognized that depend on the bacterial species from which the Cas9 originates, and when brought to this motif by the gRNA, the Cas9 cleaves the target. Cas9 from Streptococcocus pyogenes (SpCas9), the most widely used Cas9 currently, for example, recognizes a three base-pair motif of any nucleotide (A, C, G, or T) followed by a pair of guanosines. This motif is denoted NGG and is called the Protospacer Adjacent Motif, or PAM.
Many nuances of the CRISPR/Cas system are still being elucidated, but the spectrum of uses is already staggering. Using Cas9 as an example, one would expect a functional PAM sequence (any base followed by a “GG” pair) to occur every 16 base pairs in the 3-billion base pair human genome. However, CRISPR/Cas is not limited to S. pyogenes and it is now clear that there are many other CRISPR/Cas9 systems to be harnessed, each with different sequence specificities.19 Consequently, it would be rare to find a site in the genome that could not be targeted by some iteration of the CRISPR/Cas9 system.
3 GENE EDITING FOR CF
The gene editing methods described above have shown it is clearly possible to make very specific and accurate changes in chromosomal DNA. Among the many uses one could conceive for this technology, a clinical application is to repair small lesions in DNA that cause diseases like CF. The vast majority of CFTR mutations involve three nucleotides or less, and thus would likely be amenable to the gene editing process. The first reported successful example of this for the CFTR gene was accomplished using intestinal stem cells. These cells, carrying the ΔF508 mutation, were corrected, selected and then allowed to differentiate into 3-dimensional, spheroid structures called organoids. The cells of these organoids, if containing functional CFTR, swell when stimulated by agonists such as forskolin that activate adenylate cyclase. Schwank et al showed that without using corrected cells to generate organoids, swelling was undetectable while organoids from corrected cells were similar to non-CF organoids.20
With this successful example, gene editing for CF therapy is not a question of possibility, but rather becomes an engineering task, allowing researchers to focus on details like efficiency of correction and delivery of the editing components into cells of the body. More simply put, the steps that must be considered are how, where, when, and how often to edit the gene and which mutations should be pursued first for such strategies.
3.1 How?
There are multiple facets to how one might accomplish gene editing, most notably: 1. which system and components will be used; 2. how will the components be delivered to the cells; and 3. will the delivery be in vivo or ex vivo followed by introducing the corrected cells into the patient?
3.2 Gene editing components
The complexity of CF genetics, with nearly 2000 distinct mutations, presents a logistical problem. The gene editing process is most efficient if the desired change is located very close to the site of DNA cleavage, within a dozen or so nucleotides.21 As mentioned above, the variety of CRISPR/Cas reagents for targeting a break near a mutation will not be the limiting concern. However, these CF-causing mutations span nearly 200 000 nucleotides and warrant consideration of strategies to minimize the number of mutation-specific correction protocols that would be needed. At one end of the spectrum, 2000 different formulations could be envisioned, and at the other end a single strategy that corrects all mutations would be ideal.
An example of how any of the gene editing systems could be used to repair a CF mutation is illustrated in Figure 1. In this example, the G542X mutation, caused by a guanine to thymine substitution that creates a premature stop codon, could be repaired if a template sequence were provided to direct the repair process to incorporate the guanine residue.
Although the CFTR gene spans 190 298 base-pairs, almost all mutations are found in, or immediately flanking, the protein coding sequences of the exons. In other words, nearly all of the 2000 mutations are found within stretches of the gene that account for only about 4500 base-pairs, corresponding to about one mutation every 2.25 base-pairs. If efficient gene editing requires a cleavage site near the mutation (say, within 15 base-pairs), the number of protocols needed should be much less than 2000, but still in the hundreds.
Methods that would result in one, or a small number of strategies have been proposed. These involve insertion of DNA sequences that allow the mutant gene to be overridden. In these cases, a version of the gene, a cDNA that contains the normal CFTR coding sequence but without intronic sequences, is inserted into an exon (preferably exon 1) upstream of mutations in a manner that would allow a normal CFTR protein to be produced. In this way, the gene's endogenous regulatory elements that dictate when and where the gene is expressed are left intact, allowing CFTR production and function to be dictated by normal physiologic signals. Strategies that allow for inserting relatively large DNA sequences are being developed and show promise,22 supporting the feasibility of such a “one-size-fits-all” approach, as illustrated in Figure 2.
3.3 Introducing editing components into the cells
Delivery of nucleic acids to cells has been an obstacle for many strategies. Viral systems that transduce cells through infection can be very effective delivery vehicles, but they are limited in the size of cargo they can contain by the biophysics of the viral particles themselves. Consequently, reducing the cargo to the smallest size that will complete the job is essential. To this end, Cas9 from other bacteria have been examined for their utility. As an example, Cas9 from Staphylococcus aureus (SaCas9) has activity comparable to SpCas9, but is ∼300 amino acids smaller than SpCas9 and thus the DNA or RNA encoding it is nearly a thousand residues smaller.23
Since, the original reports that introduced DNA components, one encoding Cas9, another generating the gRNA and yet another containing a template with the desired sequence changes, different delivery schemes have been examined. gRNA molecules complexed with Cas9 protein can be transfected along with single-stranded oligonucleotide templates efficiently and effectively, demonstrating a fair amount of plasticity in the system.
3.4 In vivo versus ex vivo delivery
Early attempts at nucleic acid delivery for CF therapy were focused on in vivo administration of cDNA-based transgenes. Engineered viral vectors,24 receptor-mediated strategies,25 and liposomes26 have been proposed and used, but have not been successfully developed for therapy, partly because of the immune response mounted to the delivery vehicles but also because of the short duration of the transgenes and their expression.
While limiting the efficacy of the transgene approach, these attributes may be a benefit for the gene editing strategies. As gene editing relies on a nuclease to cause DNA breaks, one would prefer such an activity to be transient to minimize the chances of off-target DNA damage. Consequently, strategies to deliver transgenes to CF cells and tissues has renewed fervor, as much has been learned about efficiency and specificity of cell types for the various vector systems and thus may allow the editing process to move forward more quickly.
CF gene complementation as a therapeutic prospect was initially proposed after in vitro experiments in 1990 showed a cDNA encoding functional CFTR could impart chloride channel activity on CF cells.5, 6 At that time, the focus was solely on airway epithelial cells and technologies to introduce nucleic acids into them were limited. All of the early approaches to the delivery quandary did so in attempts to complement the cells in situ, but the opportunity to use ex vivo approaches has changed in recent years. The field of induced pluripotent stem (iPS) cells was born in 200627 and not available when most strategies of gene complementation were being pursued for CF. “Recipes” continue to be developed that will convert these cells into various lineages, and to do so efficiently. With these advances in stem cell technologies, additional CFTR-directed therapies may be possible.
3.5 Where?
CF is most often thought of as an epithelial disorder, but it will be a challenge to develop a single system to target the various epithelia affected. For example, a strategy to correct the epithelial cells of the conducting airways will not likely reach the intestines or pancreatic ducts.
A related issue is the role of non-epithelial cells in CF. CFTR is found in many different tissue and cell types that have been shown to contribute to disease manifestations. The first report, in 1991, suggested low levels of CFTR mRNA in neutrophils and alveolar macrophages.28 Since then, antimicrobial and inflammatory differences of function between CF and non-CF-derived cells of these types have been reported.29, 30 Other cell types have also been reported to be directly affected by CFTR's absence, including smooth muscle,31 pancreatic β-cells,32 and likely others.33, 34 Each will have its own considerations regarding “how” and “when.”
3.6 When?
A fundamental consideration is when in the disease process gene correction must occur in order to have clinical benefit. One concern has been whether correction would need to be accomplished prenatally. Fortunately, clinical experience with CFTR modulators, such as Ivacaftor, has shown that postnatal correction can have dramatic disease-suppressing and disease-correcting effects, particularly if implemented early.35 These observations imply that prenatal strategies are not necessary for considerable clinical benefit to be achieved. However, there are tissues, such as the exocrine pancreas, that are irreversibly damaged before birth, so the idea of pre-natal treatment is likely to be re-visited as the technologies advance.
3.7 How often?
A corollary to “when?” is the frequency that gene repair strategies would need to be implemented. It is difficult to answer that question currently, as it is intimately tied to “where?” If strategies were able to target stem cell populations, a single correction procedure might be sufficient. If not, repetitions will be dependent on cell turnover, the proportion of cells corrected, and the immune response to the correction components.
3.8 Which mutations?
If a single strategy for all mutations is feasible, then the issue of prioritizing mutations is moot. If not, arguably the mutations that could not be influenced by CFTR modulator drugs (stop mutations) would be of high priority. About 10% of CF-causing mutations prevent CFTR from being produced, either by changing an amino acid-encoding codon to a stop signal, or by deletion or insertion of base pairs that shift the reading frame of the gene. There are strategies being explored to alter the fidelity of the ribosome so that it incorporates an amino acid at these premature stop codons, but their efficacy has been low compared to the CFTR modulators.36
However, as the gene editing technologies are still in their infancy, one could also argue that proof of concept trials should be considered to provide evidence that this is a feasible strategy. For these, the targets most likely to succeed would be of highest priority. In this regard, a mutation that could be functionally corrected by non-homologous end joining, which is much more efficient than homology-directed repair, would be desired. One such example is a mutation in intron 22, c.3717+12191C>T, that generates a splice site and causes a non-functional CFTR to be produced. In this case, simply destroying the mutation, rather than trying to make a specific change (the “normal” sequence at that position), would prevent the mutant splicing from occurring and result in production of normal CFTR.
4 WHAT THE FUTURE MAY HOLD
The idea of gene-based therapy for CF has been pursued since the early 1990s6, 37 and its implementation has been unsuccessful in achieving clinical benefit for CF patients. However, these early studies have provided much insight into how one might deliver nucleic acids to target tissues such as the airway epithelium. As delivery methods are elucidated and efficiency of editing increased, the potential to provide therapy for genetic diseases is considerable. However, the technologies are still very new and the caveats are likely not all known yet. The ability to correct CF cells has been established, but current technologies do not provide delivery and targeting to sufficient numbers of cells for physiologic correction in vivo. Ex vivo approaches may be feasible and should be explored as well. Using induced pluripotent stem cells grown in culture combined with selection of successfully modified cells, the cDNA insertion approach has been shown to work for other genetic disorders, such as a form of muscular dystrophy.38
As with all new technologies, there is excitement about the potential, but it must be balanced with caution. There are concerns over carcinogenesis and other deleterious effects, as manipulation of the genome also carries with it the potential to alter unintended sites, or so-called “off target” effects. Strategies to assess such safety concerns are being developed and will be important as trials are proposed. Nonetheless, the potential to treat genetic diseases, such as CF, at their source has never been greater.
