©2021 by The International Society for Stem Cell Research. All rights reserved.

No part of this document may be produced in any form without written permission of The International Society for Stem Cell Research.


Appendix 5. Considerations for Genome Editing Research

Assessing tumorigenicity of genome-modified cell interventions

For gene-modified cell products early readouts of a potential tumorigenic risk could include the expansion of one or few dominant clones from a starting polyclonal graft in a (xeno)-transplanted host. The emergence of such dominant clones may highlight the occurrence within the administered cell population of some genotoxic events consequent to the genetic modification, such as integration of a gene transfer vector or editing-induced translocation nearby an oncogene. These random and presumably rare events may activate the tumorigenic potential of the oncogene and endow the affected cell with a gain-of-function mutation promoting its growth. It should be realized that cells carrying genotoxic events leading to a gain of function may not progress to the formation of a full-blown tumor in preclinical models for lack of proper supporting conditions, sufficient follow-up time or the small scale of the study. On the other hand, such could develop in humans, where more cells are administered and clinical persistence may extend far longer than in conventional preclinical models. Clonal tracking of administered cells in vivo has been primarily developed and validated as a safety readout in the field of hematopoietic stem cell gene therapy, where semirandom genome-wide insertions of vector provide a unique clonal marker of transduced cells. In studies using early generation retroviral vectors, expansion of rare clones carrying vector integration nearby certain oncogenes was often reported, both in animal models as well as in human subjects, and some of these clones eventually progressed to overt leukemia. In these cases, identification of a vector insertion next to an oncogene in the leukemic clone allows tracing the origin of the leukemia to the original genetic modification. Such clonal markers of gene-modified cells may not be available when using other engineering platforms such as genome editing or when the cell product does not undergo genetic engineering. Clonal tracking could still be attempted using surrogate readouts such as randomly acquired genomic mutations and monitoring the graft for the skewing from polyclonal to oligoclonal composition and the potential emergence and expansion of clones having a growth advantage, which may eventually progress to tumors. 

Preclinical Safety and Efficacy Involving Genome Modified Cell Interventions 

The following must be addressed and minimized through preclinical studies before initiating a first-in-human clinical trial. 

Issues Particular to Gene Replacement

Semi-random insertion of exogenous DNA may cause genotoxicity when a sporadic insertion takes place near an oncogene and causes its activation by truncation and/or transcriptional activation or disrupts tumor suppressor genes. These events may be rare but due to the very large number of insertions typically occurring in some cell therapies, they may well occur within a cell product. The few cells bearing such insertions might then expand and become dominant in vivo because of the enhanced growth potential afforded by the mutation. Genome insertions are expected with integrating vector platforms (such as retro-/lenti-viral vectors or transposons) but may also occur inadvertently and to lower extent when episomal DNA (i.e. from AAV vectors or plasmids) become incorporated by non-homologous end joining (NHEJ) at sites of DNA double-stranded breaks (DSB). For integrating vectors, a design should be used that minimizes the risk of genotoxic insertions (i.e. reducing the extent of transcriptional transactivation or readthrough from insertion site). Some knowledge should also be acquired on the genome-wide insertion pattern in the selected cell type and any existing specific biases that may increase the risk of genotoxic insertions. The genomic distribution of vector insertion should be assessed by preclinical studies in the cultured treated cells as well after in vivo administration into recipients, which should be monitored for the emergence of dominant clones with genotoxic insertions. Information available from prior studies performed with the same or similar vector backbone and target cells might alleviate the requirement for new extensive investigation.  For non-integrating platforms, the residual extent of insertion or lack thereof should be investigated or previously known. 

The potential mobilization of the vector, whether integrated or maintained as an episome, upon superinfection of the engineered cells in the recipient by wild-type virus, and the possibility of recombination of the vector genome with the wild-type viral genome should also be considered among the potential long-term risk. It is expected that recombination of vector sequence with the parental viral genome would most often result in a replication-defective virus. However, the potential risk of incorporating a new and biohazardous gene in the viral gene pool should be considered and, if present, alleviated by adopting conditions minimizing such risk. Many integrating vectors derived from retro-/lenti-viruses are commonly designed to be “self-inactivating”. This design means that upon integration the viral long terminal repeats are deleted of most transcription activating sequences. Such deletion makes the rescue of proviral expression, and its capture by the superinfecting virus, highly unlikely.

Cytoplasmic and nuclear exposure to exogenous nucleic acids, whether of viral, plasmid or other origin, and their replication intermediates might activate the innate immune sensing machinery in the treated cells. This activation may in turn trigger detrimental and inflammatory responses, potentially spreading to neighboring cells. Such responses might be minimal and only have subtle effects. However, if their activation is more robust or sustained they might impact the ability to engraft and adversely affect the clonal composition and long-term stability of an engineered cell graft. Importantly, these responses might be substantially augmented by excess impurities, such as DNA fragments and residual plasmid in the final cell product. Thus, efforts should be made to reduce impurities in the vector preparation.

Pre-existing immunity to viruses used to make gene transfer vectors may limit their application in vivo. This might be due to the presence of high-titer neutralizing antibodies in the plasma that may inactivate the vector and thus block gene transfer. Another possibility is the recognition of residual viral components in the transduced cells by T-cells, which may lead to the immune-mediated clearance of the transduced cells. The latter response might also affect ex vivo engineered cells if administered shortly after vector exposure. These immune responses may impact the in vivo survival of engineered cells and should be appropriately investigated before clinical testing.  

Issues Particular to Genome Editing 

  1. The first and best developed approach to genome editing exploits engineered endonucleases to deliver a DNA double-stranded break (DSB) to the intended target sequence. One main safety concern is the off-target activity of the nuclease. Extensive preclinical testing should be performed to interrogate the genome-wide specificity of the editing reagents using orthogonal techniques. The target sequence is first chosen to be uniquely represented in the genome and with limited or no occurrence of any highly similar homologous sequences bearing only a few mismatches. Bioinformatic prediction of potential off targets is then performed to rule out potential activity in known sensitive genomic sites (such as tumor suppressor genes). An experimental assessment of specificity is then performed on DNA in vitro or in cell lines exposed to high concentration of the nuclease by one or more techniques, thus generating a list of candidate off-target sites, which are also analyzed comparatively with the bioinformatic predictions. Finally, the top ranked off-target sites are interrogated by deep sequencing for targeting by the nuclease in the selected target cells in conditions best representative of the intended clinical protocol. These studies should be conducted with proper positive and negative controls to determine sensitivity thresholds. Standard or threshold acceptance values for off targets are hard to provide across platforms, target cells and applications, and should be determined accordingly to the intended use. 
  2. Large genomic alterations, deletions and translocations are also induced, albeit to lower extent than NHEJ and HDR-mediated repair, at the DNA DSB sites, and are all difficult to evaluate. This is particularly true for allele drop-outs due to large deletions, which can encompass large segments of DNA. These events may be of particular concern if they lead to hemizygosity or even homozygosity for a loss-of-function mutation in a tumor suppressor gene. The possible contribution to loss of heterozygosity by gene conversion in the course of repair of a DNA DSB should also be considered. Efforts should go into ruling out the occurrence of unwanted on-target genomic alterations above a threshold limit of detection and/or expectation. Moreover, the possible occurrence of genomic rearrangement involving sensitive loci should be cause for discarding the candidate reagents. When addressing the overall safety of a cell product that may comprise a small fraction of cells bearing genomic alterations below the threshold of detection one may be able to draw upon available past experience with gene- and cell-based interventions using the same or other platforms with the same target cells.
  3. Biodistribution studies of genome edited cells in suitable xenogeneic immunocompromised recipients should be performed to establish comparable behavior to untreated cells. Targeted editing by nucleases may leave a genetic scar. Such scars may be traceable, depending on the mechanism of repair of the DNA DSB.  NHEJ-mediated DSB repair usually introduces small nucleotide insertion/deletions (indels) at the target site, which can be identified by deep sequencing the locus. However, some editing events might be missed because the original sequence is reconstituted or has been lost by a large deletion or because it was involved in a translocation. If only one base has been changed, it will be difficult to distinguish it from a sequencing error. Homology directed repair (HDR) of DSB can more easily be tracked because of the templated sequence changes in the target locus. Whenever feasible, strategies should be adopted to allow reliable tracking of the edited cells, for instance by recoding part of the target sequence in the template to introduce a traceable genetic marker. These base changes might also serve to protect the template from the action of the nuclease and improve the efficiency of editing. The genetic modifications introduced during editing could be used to track the fate, survival and biodistribution of the edited cells and their progeny. These studies will help to establish safety and efficacy of the treatment and address the possible relationship of eventual adverse events with the editing process (i.e. to distinguish the possible origin of abnormal differentiation, growth or transformation of some edited cells vs. background disease or age-related events). However, some edited cells may still escape tracking. Tracking of cells edited by base editors or epigenetic editors (see below) may prove even more difficult or perhaps impossible. 
  4. DNA DSB might induce DNA damage response in a dose-dependent manner as well as other signaling and transcriptional responses in cells treated for editing. P53-mediated responses have the potential to induce cell senescence with long-term detrimental effects and selection of p53+/- or -/- variants. The occurrence, extent and specific modes of such responses to genome editing still need to be investigated in most target cells and applications. Furthermore, combination of DNA DSB with vectors used to deliver the repair template for HDR may induce cumulative activation of innate immune sensors and trigger more detrimental responses. Such responses might only have transient effects but if robust and prolonged they might impact the cell survival, extent and time to engraftment, clonal composition and long-term stability of an engineered cell or tissue graft.
  5. There are continuously emerging new technology platforms which introduce new editing modalities with broader reach and potentially improved precision and safety. For example: Break-less editors, Base editors, Prime editors (Anzalone et al., 2020). These new strategies are expected to provide improved editing precision at the target site by diminishing the spectrum of potential outcomes and to alleviate the burden imposed on the target cell by the DNA DSB. However, these new strategies also raise specific issues concerning monitoring for off-target effects. Specific tests might need to be designed to address genome wide specificity of these editors. In particular, many of these editors exploit the editing domain of an enzyme with constitutive activity independent of the binding of the fusion protein to DNA. Thus, off-target activity might be displayed semi-randomly in the genome and thus independently from the nearby occurrence of homology to the intended target sequence. Because of its semi-random occurrence such off-target activity may escape detection when investigating bulk treated cells, where semi-randomly distributed rare events would become noise. A possible strategy to address this issue is to compare SNVs among several single cell-derived clones from the treated cells.
  6. In vivo genome editing still remains challenging because it requires an effective and safe delivery of the editing machinery to sufficient numbers of the relevant cell type. Current platforms support either stable high-level expression of editors with concomitant increased risk of toxicity, off-target activity and immunogenicity (such as when using AAV vectors) or they fail to achieve satisfactory efficacy due to low expression level. Nanoparticle based delivery methods represent a promising approach for short-term expression, but are still difficult to target to tissues other than the liver. Furthermore, most genome editors comprise at least some components of bacterial origin and are thus likely to be immunogenic. The sustained expression or even the residual presence of such material in the edited cells might impact their survival in vivo and this risk should be appropriately investigated before clinical testing.