Abstract
As The CRISPR Journal kicks off its 5th year of publishing, I'm pleased to introduce a special collection of articles in this issue on the theme of “CRISPR's Path to the Clinic.”
The CRISPR-Cas system has revolutionized the field of genome editing, enabling the rapid generation of disease models and the development of therapeutic strategies for genetic and nongenetic diseases. This special issue provides an interesting collection of review and original articles on the use of CRISPR-Cas technology for disease modeling (see pages 40 and 53) and therapies (see pages 66 and 95), potential drawbacks (see pages 19 and 80) and optimization (see page 40) of current CRISPR-Cas technologies, as well as novel genome editing tools (see pages 31 and 109).
The CRISPR-Cas nuclease system has been widely used to inactivate genes or regulatory regions via nonhomologous end joining (NHEJ)–mediated repair of double-strand breaks (DSBs) or to insert/correct mutations via homology-directed repair (HDR). While the NHEJ repair mechanism is commonly used in all cells throughout the cell cycle, HDR takes place mainly in dividing cells. Furthermore, HDR-based strategies are more complex, as they require the use of a donor template (containing the desired edit) delivered as single-stranded oligo deoxynucleotides (ssODs) or using a viral vector such as adeno-associated virus (AAV).
Great effort is currently being undertaken to optimize HDR efficiency for generating disease models or correcting genetic mutations. By way of example, Simone et al. report in this issue the development of a chimeric donor ssODN template fused to the tracrRNA that is efficiently delivered to the target site, thus achieving high HDR frequencies (see page 40). Similarly, Zhao et al. describe the use of bacterial genetic elements known as retrons to produce continuously in cellula a DNA donor template fused to the guide RNA to achieve HDR-mediated genome editing (see page 31).
Of course, the use of CRISPR has raised a number of safety concerns for clinical applications (reviewed by Boutin et al. in this issue; see page 19
The generation of DSBs by CRISPR-Cas can also cause large genomic rearrangements (e.g., in the presence of concomitant DSBs at on- and off-target sites). 4 Furthermore, even the generation of a single DSB can lead to genomic aberrations, such as chromosomal loss, formation of micronuclei, and chromothripsis. 5 However, cells with chromosomal aberrations are often counterselected. 4 Finally, another concern in the field of genome editing-based therapeutic approaches is the risk associated with the use of AAVs that can integrate in oncogenic hot spots. 6 Importantly, the use of alternative donor templates such as those described in this issue (see pages 31 and 40), as well as the development of alternative vehicles for CRISPR-Cas, such as nanoparticles, 7 could avoid the potential oncogenic risk associated to the integration of viral vectors. 6
It is worth highlighting the limitations of preclinical assays to evaluate potentially genotoxic events. The sensitivity of assays currently in use is relatively low, particularly in primary cells (the target population of genome editing approaches). This might be a concern, especially when the therapeutic approach is based on the genetic modification of millions of cells. Progress has been made in this regard, as reported in this issue by Allen et al., who developed a high-throughput, flow-based imaging method to detect DNA damage in CRISPR-Cas-treated primary hematopoietic stem progenitor cells (the target population in gene therapy approaches for hematopoietic and non-hematopoietic genetic disorders; see page 80).
Another challenge is the variability among individuals that could lead to a wide diversity in off-target edits, depending on the genotype. Evaluation of off-target activity for every potential patient would be challenging and cumbersome, but novel methods, such as OligoNucleotide Enrichment and sequencing (ONE-seq), take into account the genomic sequence variation that exists within the human population, although this analysis so far is restricted to in silico predicted off-targets. 8
Similarly, preclinical models have often several limitations. For example, humanized animals carrying human target sequence are commonly used to prove the efficacy in vivo of genome editing strategies, often in long-term studies. However, they cannot be used to evaluate off-target activity and the persistence of off-target edits in vivo. On the contrary, xenotransplantation of human cells (e.g., hematopoietic stem/progenitor cells) in immunodeficient mice allow the evaluation of efficacy and safety in vivo in long-term experiments. However, whether these chimeric models are helpful in predicting clinical outcomes in patients is still a subject of debate.
Given the relatively recent introduction of CRISPR-Cas for therapeutic purposes, established guidelines for preclinical studies for each disease and patient population are still lacking. However, efforts have been made in this direction, for example the creation of the “genomic therapy work group” that, among various activities, aims to identify the necessary preclinical studies to be conducted in gene therapy approaches for sickle cell disease. Importantly, upcoming results from clinical studies will contribute to validate the current preclinical efficacy and safety studies.
Clinical trials relying on the use of genome editing tools have been initiated to treat genetic or nongenetic diseases, and have already shown early promising clinical results, notably in the CRISPR Therapeutics-sponsored trial for beta-hemoglobinopathies. 9 In one study, chromosomal translocations were observed after CRISPR-Cas multiplex editing of T lymphocytes to improve antitumor immunity, but decreased over time, suggesting that these events were counterselected. Engineered cells persisted for 9 months without the occurrence of malignant transformation. 10 A chromosomal abnormality was also observed during the Allogene CAR-T clinical trial. 11 However, that was unrelated to the TALEN genome editing and had no clinical significance.
These studies show that it is critical to define the potential genotoxic events caused by each specific genome editing strategy and to master tools and define criteria to identify genome editing–induced events and distinguish them from preexisting genomic aberrations. Notably, current or upcoming clinical trials for beta-hemoglobinopathies 9 use CRISPR-Cas strategies that have been shown to cause the formation of micronuclei and chromothripsis 5 or megabase-scale losses of heterozygosity. 12 So far, thankfully, no adverse event has been reported in patients treated with these approaches. Long-term follow-up studies are necessary to assess fully the efficacy and safety of CRISPR-Cas therapeutic approaches and evaluate the risk/benefit ratio of these treatment for each specific disease.
The frequency of genotoxic events potentially induced by CRISPR-Cas nuclease can be substantially reduced or eliminated by more recent base and prime editing systems that can insert/correct mutations without generating DSBs. However, these technologies might raise additional safety concerns such as the unpredictable (guide RNA-independent) off-target activity of base editors as well as their RNA off-target editing. Prime editing is thought to have a low off-target activity, but its efficiency in primary cells needs to be improved for potential therapeutic applications. One such potential application of prime editing is described in this issue by Laval et al. (see page 109).
Reliable preclinical assays and models should also be implemented for applications based on the use of these novel editing systems, as this field is moving fast with Beam Therapeutics, 13 which is preparing to initiate a clinical trial for patients with sickle cell disease.