Genome editing is a set of methods used to change the DNA of a cell with single base-pair precision. It is a specific form of gene therapy, and the engineering of cells through genome editing has the potential to create a new class of medicines for the treatment of both genetic and nongenetic diseases. Genome editing has entered clinical trials: applications include the correction of variants that cause monogenic diseases, the enhancement of chimeric antigen receptor (CAR) T-cell therapy, and cell-based regenerative medicine. Here I describe the development of genome editing and discuss the ways in which efficacy, specificity, delivery, and safety are integral to this process.
Until 1994, the efficiency of genome editing in a mammalian cell was 10−6 (1 cell in 1 million would have the desired gene-targeting event).1 In 1994, Jasin and colleagues discovered that the creation of a break in a DNA double-strand in a target gene could stimulate gene targeting by a factor of more than 1000 in somatic cells when a “donor” template strand of DNA was provided at the same time that the break was created.2-5 With optimization, this system could be used to correct a reporter gene in up to 5% of cells.4 (A reporter gene delivers a signal on successful DNA editing.) In addition to showing that new sequences could be inserted at the site of the breakthrough homologous recombination, this discovery also indicated that new mutations could be created at the site of the breakthrough a process called nonhomologous end-joining (NHEJ).6-8 The discovery that a specific double-strand break in DNA could induce repair is the foundational principle of the field of genome editing.
Nuclease Platforms for Genome Editing.
A limitation of these early studies was their use of a specific homing endonuclease, an enzyme that recognizes and cuts a specific DNA sequence (a recognition site). This approach could not be applied to human cells because the recognition site does not occur in endogenous genes. The problem was solved by engineering nucleases that recognize target sites in endogenous genes and stimulate genome editing at those sites.4,7-10 The first such nucleases were zinc-finger nucleases, in which a DNA-binding protein with a specific recognition sequence was fused to a nonspecific nuclease domain.11 A wide variety of nucleases are now used in addition to zinc-finger nucleases, including homing endonucleases and transcription activator-like effector nucleases (Figure 2).7,12 Each creates a site-specific, double-strand break in the genome of the cell that activates repair through NHEJ or homology-directed repair (HDR). Nonnuclease-based systems of genome editing13-16 are in earlier stages of development.
The most commonly used Cas9 enzyme is from Streptococcus pyogenes. The gRNA molecule can be tailored to optimize hybridization with a particular DNA target site and thereby guide the Cas9–gRNA complex to the site of the desired break (Figure 2).19,21 In contrast with other genome-editing nuclease systems, the “guidance” of Cas9–gRNA to its target site is governed by Watson–Crick base-pairing, an ease-of-design feature.
DNA Editing through NHEJ
NHEJ is a form of double-stranded break repair that does not require a “repair” template.21,22 Instead, the ends of the broken DNA are held in close proximity, processed, and then joined. NHEJ-mediated editing is normally used in all cells to repair spontaneous breaks. It is generally accurate (at a rate of at ≥70%)23,24 but can create errors. NHEJ is the process naturally used by cells of the immune system to create genetic diversity in genes encoding immunoglobulins and T-cell receptors (TCRs).
These include the direct reversion of a disease-causing variant in a gene34-36; the insertion of a complementary DNA (cDNA) cassette containing a specific gene into the endogenous locus of that gene such that it is regulated by its own natural regulatory elements; the insertion of a cDNA cassette into a different locus such that it will be expressed according to the regulatory elements of the gene at that locus39; and the insertion of a transgene cassette into a “safe harbor” to avoid creating unintended insertional mutations caused by semi-random integration with viral vectors (particularly integrations with highly expressed genes) and to achieve more homogeneous expression of the transgene.
There are other aspects of delivery to consider. For example, although mRNA is better than plasmid DNA in delivering the nuclease to primary human cells, mRNA can induce an antiviral type I interferon response. Moreover, prolonged expression of a nuclease or expression of a nuclease with low specificity can result in sustained activation of the p53 pathway, thereby triggering cell-cycle arrest and apoptosis.
High frequencies of HDR-mediated editing can be achieved by delivering sufficient amounts of template DNA to cells without activating a toxic cellular response (e.g., the type I interferon response). Recombinant adeno-associated viral vectors, which have evolved to avoid cellular detection while delivering single-stranded DNA cargos to the nucleus, are efficient in delivering classic gene-targeting donor templates to cells.
Some approaches that enhance HDR-mediated editing in cells involve the use of small molecules to target specific pathways, but the effects of such interventions have been modest and inconsistent; the greatest effects have been realized when the efficiency has not been optimized.48-51 Moreover, caution is warranted: some of these interventions perturb the ways in which a cell normally repairs or responds to a double-strand break and may, therefore, compromise the repair of the 20 to 40 double-strand breaks that occur spontaneously in every cell as it progresses through its cycle.
Ex Vivo Genome Editing to Generate Cells as Drugs
Of all the approaches to genome editing, the most developed is ex vivo genome editing, in which cells are engineered outside of the body and then returned to the patient. Indeed, ex vivo engineering of cells with viral vectors (in standard gene therapy) provides commercially available products that are being used to treat a genetic immunodeficiency and cancer.
In addition, genome editing has been used to make CAR T cells “universal” through the simultaneous disruption of the genes encoding TCRα and CD52, conferring resistance to alemtuzumab, the drug used for lymphodepletion.30 These CAR T cells are being tested in the treatment of resistant leukemia. Two patients treated on an emergency basis were reported to have remission within 28 days, although graft-versus-host disease developed as a result of residual TCR-positive cells.
These trials use zinc-finger nuclease and transcription activator-like effector nuclease platforms (Figure 2). Within the next several years, however, multiple clinical trials involving ex vivo modification of HSCs and T cells that use the Cas9–the gRNA system will be initiated in the United States and Europe. The trials include the use of NHEJ-mediated knockout strategies to generate more potent CAR T cells to treat cancer.58 They also include the knockout of the erythroid-specific enhancer of BCL11A to up-regulate gamma globin within the erythroid lineage of autologous HSCs59 as a potential therapy for both sickle cell disease and β-thalassemia.
Human HDR-based genome-editing strategies are also likely to enter clinical trials in the next year. These include direct correction of the variant that causes sickle cell disease in patient-derived HSCs and the generation of more potent CAR T cells. Preclinical studies augur well for ex vivo, autologous, cell-based therapies involving genome editing for diseases such as chronic granulomatous disease, X-linked severe combined immunodeficiency,37 X-linked hyper-IgM syndromes, and HIV infection.
The overall efficiencies of genome editing of cells ex vivo are remarkably high. It is now routine to generate NHEJ-mediated indels with efficiencies exceeding 80%, large deletions with efficiencies exceeding 50%, and changes effected through HDR at frequencies between 30% and 70% in HSCs and primary human T cells.
The human immune system (adaptive and innate) has proved to be a consistent barrier to the successful use of genetic engineering in vivo. Transgene immunogenicity is a challenge to standard in vivo gene therapy and to genome-editing strategies. In the context of genome editing, the immune system may also be a barrier to the editing machinery itself. All the major nuclease platforms contain foreign proteins. Prolonged expression of the nuclease is therefore likely to invoke an adaptive immune response, which could eliminate the nuclease-expressing cell, resulting in a lack of efficacy and the generation of toxic effects. In addition, the first dose may vaccinate the patient against subsequent doses. The Cas9 nuclease used in the Cas9–gRNA system is from one of two bacterial species, S. pyogenes and Staphylococcus aureus. Since each universally infects humans, a large proportion of adults has preexisting immunity to Cas9.
Decreasing the duration of nuclease expression — for example, by delivering Cas9–gRNA as a ribonucleoprotein complex — can result in exponential improvements in specificity.23 This strategy is effective because genome editing is a “hit-and-run” process that does not require sustained nuclease expression. Changing the binding and catalytic activity of the nuclease can similarly result in improved specificity.7,12,69-71 Changing one component, however, can limit the flexibility in changing another. For example, some Cas9 variants with higher specificity had suboptimal target activity when delivered as a ribonucleoprotein complex.35 A relatively new Cas9 high-fidelity variant, when delivered as a complex with gRNA, combines high on-target activity with improved specificity.
A challenge in assessing the safety of genome editing is that there are no validated preclinical assays for this new type of medicine. There are different approaches (e.g., bioinformatics, cell capture, and in vitro) that can be used to identify sites that may harbor off-target indels, but no one approach has been established as the most effective, and each has its own intrinsic biases.
Although off-target indels can be detected to a certain level of sensitivity, there are no data to provide guidance as to what is a safe level of off-target indels for either ex vivo or in vivo uses of genome editing. It is likely that any engineered nuclease modifying a large population of cells will facilitate translocations between the on-target break and spontaneous, random breaks that continuously arise elsewhere in the genome. Current assays are not sensitive enough to measure the frequency of such events, and they have not been designed to measure the functional consequences of such events. The wisdom of growing a large population of cells from a single clone is uncertain, given that the population could become dominated by a cell with a spontaneous mutation in a tumor-suppressor gene or by an oncogene that is selected for expression during the expansion process. Further complicating the assessment of specificity is the fact that every person has a different genome, with millions of small differences at baseline, which makes it challenging to evaluate the consequence of any small potential change made by a nuclease.
The use of animal models to predict the safety of genetic engineering has not been an effective means of predicting safety in clinical trials. Although genetically engineered cells have been transplanted into immunodeficient mice to ascertain safety,41 this method cannot be relied on to identify safe or toxic genetic-engineering strategies. Developing animal models is useful, but only if they are time- and cost-efficient and can be shown to reliably predict the results in human clinical trials. Currently, the best approach to evaluating safety is in carefully controlled phase 1 human clinical trials, which not only incorporate standard measures for adjudicating adverse events but also build in analytic studies for the purpose of assessing specific toxic effects associated with genome editing, including clonality and the development of toxic immune responses.
Applications to the Treatment of Humans
For organ systems such as the hematopoietic and immune systems, the high frequency of gene correction achieved across a range of gene targets for diseases such as sickle cell disease, X-linked severe combined immunodeficiency, and X-linked chronic granulomatous disease is above the therapeutic threshold that is predicted to be curative. These systems will become the subject of clinical trials in the next several years. Hundreds of genetic diseases of the hematopoietic and immune system could, in principle, be cured with the use of this platform. Although monogenic diseases of other organ systems can also be genetically “fixed” through genome editing, challenges remain, including the isolation, expansion, and transplantation of tissue-specific stem cells (for ex vivo therapy) and the delivery of the genome-editing machinery to affected tissues.
Genome editing is being used in clinical trials for the purpose of improving CAR T-cell therapy (ClinicalTrials.gov numbers, NCT02808442 and NCT02746952; see Table 1)30 and can be used in a number of other ways. NHEJ-mediated editing, which is used to remove the alloreactivity of T cells by knocking out TCRA (which encodes TCRα), could also be used to remove immunogenicity by knocking out B2M (which encodes β2-microglobulin) and perhaps to increase the potency of cells by removing molecules that inhibit their function or accelerate their exhaustion. HDR-mediated editing can be used to ensure that genes are inserted into a specific locus.39
One generally unrealized promise of cell-based therapies is the use of cells or stem cells to replace or restore diseased, damaged, or aging tissue. Genome editing provides a method of engineering cells to increase their potency and safety. Examples of combining regenerative medicine with genome editing include engineering cells to secrete protective factors that prevent neurodegeneration and providing safety switches that readily eliminate cells if they start to cause harm.
Synthetic biology involves engineering a cell to perform a function it does not normally have. It is now possible to genetically edit cells to secrete therapeutic proteins and to use those cells to influence the physiology of an animal. Examples of combining genome editing and synthetic biology include engineering cells to secrete erythropoietin73 or wound-healing growth factors. It may become possible to engineer cells to divide, migrate, respond, signal, and secrete in ways that are therapeutically useful to the environment of diseased tissues.
The need for transparency and many of the other criteria proposed by the international study committee and the Nuffield council were violated by the unethical application of genome editing to embryos subsequently born as twins, as reported at the Second International Summit on Human Genome Editing in November 2018 in Hong Kong. No report on the case of the twins has been published, and the outcome has not been verified. In any case, even if the twins develop with no adverse events, the work was irresponsible and reckless and violated broad international norms regarding the application of genome editing to human embryos. It highlights the urgency of developing international standards that can be referred to and used to deter such unethical and irresponsible applications from occurring in the future.
A critical issue associated with the development of gene-editing therapies is the goal of making them broadly accessible. The cost of these therapies is likely to be extremely high initially, but cost-benefit analyses, including the cost of care over the lifetime of a patient, may provide justification for their use. Nonetheless, it will be important to control costs to improve equality of access, and continued attention to strategies such as prevention through genetic counseling will remain important.ail
For travel to the United States for advanced treatment - call us to book hospital bed and make travel arrangements for your loved ones!